Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]
FreeBSD/Linux Kernel Cross Reference
sys/netinet/tcp_usrreq.c
Version:
- FREEBSD - FREEBSD-13-STABLE - FREEBSD-13-0 - FREEBSD-12-STABLE - FREEBSD-12-0 - FREEBSD-11-STABLE - FREEBSD-11-0 - FREEBSD-10-STABLE - FREEBSD-10-0 - FREEBSD-9-STABLE - FREEBSD-9-0 - FREEBSD-8-STABLE - FREEBSD-8-0 - FREEBSD-7-STABLE - FREEBSD-7-0 - FREEBSD-6-STABLE - FREEBSD-6-0 - FREEBSD-5-STABLE - FREEBSD-5-0 - FREEBSD-4-STABLE - FREEBSD-3-STABLE - FREEBSD22 - l41 - OPENBSD - linux-2.6 - MK84 - PLAN9 - xnu-8792
SearchContext: - none - 3 - 10
SearchContext: - none - 3 - 10
1 /*- 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. 4 * Copyright (c) 2006-2007 Robert N. M. Watson 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 4. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94 32 */ 33 34 #include <sys/cdefs.h> 35 __FBSDID("$FreeBSD$"); 36 37 #include "opt_ddb.h" 38 #include "opt_inet.h" 39 #include "opt_inet6.h" 40 #include "opt_tcpdebug.h" 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/malloc.h> 45 #include <sys/kernel.h> 46 #include <sys/sysctl.h> 47 #include <sys/mbuf.h> 48 #ifdef INET6 49 #include <sys/domain.h> 50 #endif /* INET6 */ 51 #include <sys/socket.h> 52 #include <sys/socketvar.h> 53 #include <sys/protosw.h> 54 #include <sys/proc.h> 55 #include <sys/jail.h> 56 57 #ifdef DDB 58 #include <ddb/ddb.h> 59 #endif 60 61 #include <net/if.h> 62 #include <net/route.h> 63 64 #include <netinet/in.h> 65 #include <netinet/in_systm.h> 66 #ifdef INET6 67 #include <netinet/ip6.h> 68 #endif 69 #include <netinet/in_pcb.h> 70 #ifdef INET6 71 #include <netinet6/in6_pcb.h> 72 #endif 73 #include <netinet/in_var.h> 74 #include <netinet/ip_var.h> 75 #ifdef INET6 76 #include <netinet6/ip6_var.h> 77 #include <netinet6/scope6_var.h> 78 #endif 79 #include <netinet/tcp.h> 80 #include <netinet/tcp_fsm.h> 81 #include <netinet/tcp_seq.h> 82 #include <netinet/tcp_timer.h> 83 #include <netinet/tcp_var.h> 84 #include <netinet/tcpip.h> 85 #ifdef TCPDEBUG 86 #include <netinet/tcp_debug.h> 87 #endif 88 #include <netinet/tcp_offload.h> 89 90 /* 91 * TCP protocol interface to socket abstraction. 92 */ 93 static int tcp_attach(struct socket *); 94 static int tcp_connect(struct tcpcb *, struct sockaddr *, 95 struct thread *td); 96 #ifdef INET6 97 static int tcp6_connect(struct tcpcb *, struct sockaddr *, 98 struct thread *td); 99 #endif /* INET6 */ 100 static void tcp_disconnect(struct tcpcb *); 101 static void tcp_usrclosed(struct tcpcb *); 102 static void tcp_fill_info(struct tcpcb *, struct tcp_info *); 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 */ 119 static int 120 tcp_usr_attach(struct socket *so, int proto, struct thread *td) 121 { 122 struct inpcb *inp; 123 struct tcpcb *tp = NULL; 124 int error; 125 TCPDEBUG0; 126 127 inp = sotoinpcb(so); 128 KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL")); 129 TCPDEBUG1(); 130 131 error = tcp_attach(so); 132 if (error) 133 goto out; 134 135 if ((so->so_options & SO_LINGER) && so->so_linger == 0) 136 so->so_linger = TCP_LINGERTIME; 137 138 inp = sotoinpcb(so); 139 tp = intotcpcb(inp); 140 out: 141 TCPDEBUG2(PRU_ATTACH); 142 return error; 143 } 144 145 /* 146 * tcp_detach is called when the socket layer loses its final reference 147 * to the socket, be it a file descriptor reference, a reference from TCP, 148 * etc. At this point, there is only one case in which we will keep around 149 * inpcb state: time wait. 150 * 151 * This function can probably be re-absorbed back into tcp_usr_detach() now 152 * that there is a single detach path. 153 */ 154 static void 155 tcp_detach(struct socket *so, struct inpcb *inp) 156 { 157 struct tcpcb *tp; 158 159 INP_INFO_WLOCK_ASSERT(&tcbinfo); 160 INP_WLOCK_ASSERT(inp); 161 162 KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp")); 163 KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so")); 164 165 tp = intotcpcb(inp); 166 167 if (inp->inp_flags & INP_TIMEWAIT) { 168 /* 169 * There are two cases to handle: one in which the time wait 170 * state is being discarded (INP_DROPPED), and one in which 171 * this connection will remain in timewait. In the former, 172 * it is time to discard all state (except tcptw, which has 173 * already been discarded by the timewait close code, which 174 * should be further up the call stack somewhere). In the 175 * latter case, we detach from the socket, but leave the pcb 176 * present until timewait ends. 177 * 178 * XXXRW: Would it be cleaner to free the tcptw here? 179 */ 180 if (inp->inp_flags & INP_DROPPED) { 181 KASSERT(tp == NULL, ("tcp_detach: INP_TIMEWAIT && " 182 "INP_DROPPED && tp != NULL")); 183 in_pcbdetach(inp); 184 in_pcbfree(inp); 185 } else { 186 in_pcbdetach(inp); 187 INP_WUNLOCK(inp); 188 } 189 } else { 190 /* 191 * If the connection is not in timewait, we consider two 192 * two conditions: one in which no further processing is 193 * necessary (dropped || embryonic), and one in which TCP is 194 * not yet done, but no longer requires the socket, so the 195 * pcb will persist for the time being. 196 * 197 * XXXRW: Does the second case still occur? 198 */ 199 if (inp->inp_flags & INP_DROPPED || 200 tp->t_state < TCPS_SYN_SENT) { 201 tcp_discardcb(tp); 202 in_pcbdetach(inp); 203 in_pcbfree(inp); 204 } else 205 in_pcbdetach(inp); 206 } 207 } 208 209 /* 210 * pru_detach() detaches the TCP protocol from the socket. 211 * If the protocol state is non-embryonic, then can't 212 * do this directly: have to initiate a pru_disconnect(), 213 * which may finish later; embryonic TCB's can just 214 * be discarded here. 215 */ 216 static void 217 tcp_usr_detach(struct socket *so) 218 { 219 struct inpcb *inp; 220 221 inp = sotoinpcb(so); 222 KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL")); 223 INP_INFO_WLOCK(&tcbinfo); 224 INP_WLOCK(inp); 225 KASSERT(inp->inp_socket != NULL, 226 ("tcp_usr_detach: inp_socket == NULL")); 227 tcp_detach(so, inp); 228 INP_INFO_WUNLOCK(&tcbinfo); 229 } 230 231 /* 232 * Give the socket an address. 233 */ 234 static int 235 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 236 { 237 int error = 0; 238 struct inpcb *inp; 239 struct tcpcb *tp = NULL; 240 struct sockaddr_in *sinp; 241 242 sinp = (struct sockaddr_in *)nam; 243 if (nam->sa_len != sizeof (*sinp)) 244 return (EINVAL); 245 /* 246 * Must check for multicast addresses and disallow binding 247 * to them. 248 */ 249 if (sinp->sin_family == AF_INET && 250 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) 251 return (EAFNOSUPPORT); 252 253 TCPDEBUG0; 254 INP_INFO_WLOCK(&tcbinfo); 255 inp = sotoinpcb(so); 256 KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL")); 257 INP_WLOCK(inp); 258 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 259 error = EINVAL; 260 goto out; 261 } 262 tp = intotcpcb(inp); 263 TCPDEBUG1(); 264 error = in_pcbbind(inp, nam, td->td_ucred); 265 out: 266 TCPDEBUG2(PRU_BIND); 267 INP_WUNLOCK(inp); 268 INP_INFO_WUNLOCK(&tcbinfo); 269 270 return (error); 271 } 272 273 #ifdef INET6 274 static int 275 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 276 { 277 int error = 0; 278 struct inpcb *inp; 279 struct tcpcb *tp = NULL; 280 struct sockaddr_in6 *sin6p; 281 282 sin6p = (struct sockaddr_in6 *)nam; 283 if (nam->sa_len != sizeof (*sin6p)) 284 return (EINVAL); 285 /* 286 * Must check for multicast addresses and disallow binding 287 * to them. 288 */ 289 if (sin6p->sin6_family == AF_INET6 && 290 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) 291 return (EAFNOSUPPORT); 292 293 TCPDEBUG0; 294 INP_INFO_WLOCK(&tcbinfo); 295 inp = sotoinpcb(so); 296 KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL")); 297 INP_WLOCK(inp); 298 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 299 error = EINVAL; 300 goto out; 301 } 302 tp = intotcpcb(inp); 303 TCPDEBUG1(); 304 inp->inp_vflag &= ~INP_IPV4; 305 inp->inp_vflag |= INP_IPV6; 306 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { 307 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr)) 308 inp->inp_vflag |= INP_IPV4; 309 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { 310 struct sockaddr_in sin; 311 312 in6_sin6_2_sin(&sin, sin6p); 313 inp->inp_vflag |= INP_IPV4; 314 inp->inp_vflag &= ~INP_IPV6; 315 error = in_pcbbind(inp, (struct sockaddr *)&sin, 316 td->td_ucred); 317 goto out; 318 } 319 } 320 error = in6_pcbbind(inp, nam, td->td_ucred); 321 out: 322 TCPDEBUG2(PRU_BIND); 323 INP_WUNLOCK(inp); 324 INP_INFO_WUNLOCK(&tcbinfo); 325 return (error); 326 } 327 #endif /* INET6 */ 328 329 /* 330 * Prepare to accept connections. 331 */ 332 static int 333 tcp_usr_listen(struct socket *so, int backlog, struct thread *td) 334 { 335 int error = 0; 336 struct inpcb *inp; 337 struct tcpcb *tp = NULL; 338 339 TCPDEBUG0; 340 INP_INFO_WLOCK(&tcbinfo); 341 inp = sotoinpcb(so); 342 KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL")); 343 INP_WLOCK(inp); 344 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 345 error = EINVAL; 346 goto out; 347 } 348 tp = intotcpcb(inp); 349 TCPDEBUG1(); 350 SOCK_LOCK(so); 351 error = solisten_proto_check(so); 352 if (error == 0 && inp->inp_lport == 0) 353 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); 354 if (error == 0) { 355 tp->t_state = TCPS_LISTEN; 356 solisten_proto(so, backlog); 357 tcp_offload_listen_open(tp); 358 } 359 SOCK_UNLOCK(so); 360 361 out: 362 TCPDEBUG2(PRU_LISTEN); 363 INP_WUNLOCK(inp); 364 INP_INFO_WUNLOCK(&tcbinfo); 365 return (error); 366 } 367 368 #ifdef INET6 369 static int 370 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td) 371 { 372 int error = 0; 373 struct inpcb *inp; 374 struct tcpcb *tp = NULL; 375 376 TCPDEBUG0; 377 INP_INFO_WLOCK(&tcbinfo); 378 inp = sotoinpcb(so); 379 KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL")); 380 INP_WLOCK(inp); 381 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 382 error = EINVAL; 383 goto out; 384 } 385 tp = intotcpcb(inp); 386 TCPDEBUG1(); 387 SOCK_LOCK(so); 388 error = solisten_proto_check(so); 389 if (error == 0 && inp->inp_lport == 0) { 390 inp->inp_vflag &= ~INP_IPV4; 391 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) 392 inp->inp_vflag |= INP_IPV4; 393 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); 394 } 395 if (error == 0) { 396 tp->t_state = TCPS_LISTEN; 397 solisten_proto(so, backlog); 398 } 399 SOCK_UNLOCK(so); 400 401 out: 402 TCPDEBUG2(PRU_LISTEN); 403 INP_WUNLOCK(inp); 404 INP_INFO_WUNLOCK(&tcbinfo); 405 return (error); 406 } 407 #endif /* INET6 */ 408 409 /* 410 * Initiate connection to peer. 411 * Create a template for use in transmissions on this connection. 412 * Enter SYN_SENT state, and mark socket as connecting. 413 * Start keep-alive timer, and seed output sequence space. 414 * Send initial segment on connection. 415 */ 416 static int 417 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 418 { 419 int error = 0; 420 struct inpcb *inp; 421 struct tcpcb *tp = NULL; 422 struct sockaddr_in *sinp; 423 424 sinp = (struct sockaddr_in *)nam; 425 if (nam->sa_len != sizeof (*sinp)) 426 return (EINVAL); 427 /* 428 * Must disallow TCP ``connections'' to multicast addresses. 429 */ 430 if (sinp->sin_family == AF_INET 431 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) 432 return (EAFNOSUPPORT); 433 if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0) 434 return (error); 435 436 TCPDEBUG0; 437 INP_INFO_WLOCK(&tcbinfo); 438 inp = sotoinpcb(so); 439 KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL")); 440 INP_WLOCK(inp); 441 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 442 error = EINVAL; 443 goto out; 444 } 445 tp = intotcpcb(inp); 446 TCPDEBUG1(); 447 if ((error = tcp_connect(tp, nam, td)) != 0) 448 goto out; 449 error = tcp_output_connect(so, nam); 450 out: 451 TCPDEBUG2(PRU_CONNECT); 452 INP_WUNLOCK(inp); 453 INP_INFO_WUNLOCK(&tcbinfo); 454 return (error); 455 } 456 457 #ifdef INET6 458 static int 459 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 460 { 461 int error = 0; 462 struct inpcb *inp; 463 struct tcpcb *tp = NULL; 464 struct sockaddr_in6 *sin6p; 465 466 TCPDEBUG0; 467 468 sin6p = (struct sockaddr_in6 *)nam; 469 if (nam->sa_len != sizeof (*sin6p)) 470 return (EINVAL); 471 /* 472 * Must disallow TCP ``connections'' to multicast addresses. 473 */ 474 if (sin6p->sin6_family == AF_INET6 475 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) 476 return (EAFNOSUPPORT); 477 478 INP_INFO_WLOCK(&tcbinfo); 479 inp = sotoinpcb(so); 480 KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL")); 481 INP_WLOCK(inp); 482 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 483 error = EINVAL; 484 goto out; 485 } 486 tp = intotcpcb(inp); 487 TCPDEBUG1(); 488 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { 489 struct sockaddr_in sin; 490 491 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { 492 error = EINVAL; 493 goto out; 494 } 495 496 in6_sin6_2_sin(&sin, sin6p); 497 inp->inp_vflag |= INP_IPV4; 498 inp->inp_vflag &= ~INP_IPV6; 499 if ((error = prison_remote_ip4(td->td_ucred, 500 &sin.sin_addr)) != 0) 501 goto out; 502 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0) 503 goto out; 504 error = tcp_output_connect(so, nam); 505 goto out; 506 } 507 inp->inp_vflag &= ~INP_IPV4; 508 inp->inp_vflag |= INP_IPV6; 509 inp->inp_inc.inc_flags |= INC_ISIPV6; 510 if ((error = prison_remote_ip6(td->td_ucred, &sin6p->sin6_addr)) != 0) 511 goto out; 512 if ((error = tcp6_connect(tp, nam, td)) != 0) 513 goto out; 514 error = tcp_output_connect(so, nam); 515 516 out: 517 TCPDEBUG2(PRU_CONNECT); 518 INP_WUNLOCK(inp); 519 INP_INFO_WUNLOCK(&tcbinfo); 520 return (error); 521 } 522 #endif /* INET6 */ 523 524 /* 525 * Initiate disconnect from peer. 526 * If connection never passed embryonic stage, just drop; 527 * else if don't need to let data drain, then can just drop anyways, 528 * else have to begin TCP shutdown process: mark socket disconnecting, 529 * drain unread data, state switch to reflect user close, and 530 * send segment (e.g. FIN) to peer. Socket will be really disconnected 531 * when peer sends FIN and acks ours. 532 * 533 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 534 */ 535 static int 536 tcp_usr_disconnect(struct socket *so) 537 { 538 struct inpcb *inp; 539 struct tcpcb *tp = NULL; 540 int error = 0; 541 542 TCPDEBUG0; 543 INP_INFO_WLOCK(&tcbinfo); 544 inp = sotoinpcb(so); 545 KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL")); 546 INP_WLOCK(inp); 547 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 548 error = ECONNRESET; 549 goto out; 550 } 551 tp = intotcpcb(inp); 552 TCPDEBUG1(); 553 tcp_disconnect(tp); 554 out: 555 TCPDEBUG2(PRU_DISCONNECT); 556 INP_WUNLOCK(inp); 557 INP_INFO_WUNLOCK(&tcbinfo); 558 return (error); 559 } 560 561 /* 562 * Accept a connection. Essentially all the work is 563 * done at higher levels; just return the address 564 * of the peer, storing through addr. 565 */ 566 static int 567 tcp_usr_accept(struct socket *so, struct sockaddr **nam) 568 { 569 int error = 0; 570 struct inpcb *inp = NULL; 571 struct tcpcb *tp = NULL; 572 struct in_addr addr; 573 in_port_t port = 0; 574 TCPDEBUG0; 575 576 if (so->so_state & SS_ISDISCONNECTED) 577 return (ECONNABORTED); 578 579 inp = sotoinpcb(so); 580 KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL")); 581 INP_INFO_RLOCK(&tcbinfo); 582 INP_WLOCK(inp); 583 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 584 error = ECONNABORTED; 585 goto out; 586 } 587 tp = intotcpcb(inp); 588 TCPDEBUG1(); 589 590 /* 591 * We inline in_getpeeraddr and COMMON_END here, so that we can 592 * copy the data of interest and defer the malloc until after we 593 * release the lock. 594 */ 595 port = inp->inp_fport; 596 addr = inp->inp_faddr; 597 598 out: 599 TCPDEBUG2(PRU_ACCEPT); 600 INP_WUNLOCK(inp); 601 INP_INFO_RUNLOCK(&tcbinfo); 602 if (error == 0) 603 *nam = in_sockaddr(port, &addr); 604 return error; 605 } 606 607 #ifdef INET6 608 static int 609 tcp6_usr_accept(struct socket *so, struct sockaddr **nam) 610 { 611 struct inpcb *inp = NULL; 612 int error = 0; 613 struct tcpcb *tp = NULL; 614 struct in_addr addr; 615 struct in6_addr addr6; 616 in_port_t port = 0; 617 int v4 = 0; 618 TCPDEBUG0; 619 620 if (so->so_state & SS_ISDISCONNECTED) 621 return (ECONNABORTED); 622 623 inp = sotoinpcb(so); 624 KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL")); 625 INP_WLOCK(inp); 626 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 627 error = ECONNABORTED; 628 goto out; 629 } 630 tp = intotcpcb(inp); 631 TCPDEBUG1(); 632 633 /* 634 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can 635 * copy the data of interest and defer the malloc until after we 636 * release the lock. 637 */ 638 if (inp->inp_vflag & INP_IPV4) { 639 v4 = 1; 640 port = inp->inp_fport; 641 addr = inp->inp_faddr; 642 } else { 643 port = inp->inp_fport; 644 addr6 = inp->in6p_faddr; 645 } 646 647 out: 648 TCPDEBUG2(PRU_ACCEPT); 649 INP_WUNLOCK(inp); 650 if (error == 0) { 651 if (v4) 652 *nam = in6_v4mapsin6_sockaddr(port, &addr); 653 else 654 *nam = in6_sockaddr(port, &addr6); 655 } 656 return error; 657 } 658 #endif /* INET6 */ 659 660 /* 661 * Mark the connection as being incapable of further output. 662 */ 663 static int 664 tcp_usr_shutdown(struct socket *so) 665 { 666 int error = 0; 667 struct inpcb *inp; 668 struct tcpcb *tp = NULL; 669 670 TCPDEBUG0; 671 INP_INFO_WLOCK(&tcbinfo); 672 inp = sotoinpcb(so); 673 KASSERT(inp != NULL, ("inp == NULL")); 674 INP_WLOCK(inp); 675 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 676 error = ECONNRESET; 677 goto out; 678 } 679 tp = intotcpcb(inp); 680 TCPDEBUG1(); 681 socantsendmore(so); 682 tcp_usrclosed(tp); 683 if (!(inp->inp_flags & INP_DROPPED)) 684 error = tcp_output_disconnect(tp); 685 686 out: 687 TCPDEBUG2(PRU_SHUTDOWN); 688 INP_WUNLOCK(inp); 689 INP_INFO_WUNLOCK(&tcbinfo); 690 691 return (error); 692 } 693 694 /* 695 * After a receive, possibly send window update to peer. 696 */ 697 static int 698 tcp_usr_rcvd(struct socket *so, int flags) 699 { 700 struct inpcb *inp; 701 struct tcpcb *tp = NULL; 702 int error = 0; 703 704 TCPDEBUG0; 705 inp = sotoinpcb(so); 706 KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL")); 707 INP_WLOCK(inp); 708 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 709 error = ECONNRESET; 710 goto out; 711 } 712 tp = intotcpcb(inp); 713 TCPDEBUG1(); 714 tcp_output_rcvd(tp); 715 716 out: 717 TCPDEBUG2(PRU_RCVD); 718 INP_WUNLOCK(inp); 719 return (error); 720 } 721 722 /* 723 * Do a send by putting data in output queue and updating urgent 724 * marker if URG set. Possibly send more data. Unlike the other 725 * pru_*() routines, the mbuf chains are our responsibility. We 726 * must either enqueue them or free them. The other pru_* routines 727 * generally are caller-frees. 728 */ 729 static int 730 tcp_usr_send(struct socket *so, int flags, struct mbuf *m, 731 struct sockaddr *nam, struct mbuf *control, struct thread *td) 732 { 733 int error = 0; 734 struct inpcb *inp; 735 struct tcpcb *tp = NULL; 736 int headlocked = 0; 737 #ifdef INET6 738 int isipv6; 739 #endif 740 TCPDEBUG0; 741 742 /* 743 * We require the pcbinfo lock in two cases: 744 * 745 * (1) An implied connect is taking place, which can result in 746 * binding IPs and ports and hence modification of the pcb hash 747 * chains. 748 * 749 * (2) PRUS_EOF is set, resulting in explicit close on the send. 750 */ 751 if ((nam != NULL) || (flags & PRUS_EOF)) { 752 INP_INFO_WLOCK(&tcbinfo); 753 headlocked = 1; 754 } 755 inp = sotoinpcb(so); 756 KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL")); 757 INP_WLOCK(inp); 758 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 759 if (control) 760 m_freem(control); 761 if (m) 762 m_freem(m); 763 error = ECONNRESET; 764 goto out; 765 } 766 #ifdef INET6 767 isipv6 = nam && nam->sa_family == AF_INET6; 768 #endif /* INET6 */ 769 tp = intotcpcb(inp); 770 TCPDEBUG1(); 771 if (control) { 772 /* TCP doesn't do control messages (rights, creds, etc) */ 773 if (control->m_len) { 774 m_freem(control); 775 if (m) 776 m_freem(m); 777 error = EINVAL; 778 goto out; 779 } 780 m_freem(control); /* empty control, just free it */ 781 } 782 if (!(flags & PRUS_OOB)) { 783 sbappendstream(&so->so_snd, m); 784 if (nam && tp->t_state < TCPS_SYN_SENT) { 785 /* 786 * Do implied connect if not yet connected, 787 * initialize window to default value, and 788 * initialize maxseg/maxopd using peer's cached 789 * MSS. 790 */ 791 INP_INFO_WLOCK_ASSERT(&tcbinfo); 792 #ifdef INET6 793 if (isipv6) 794 error = tcp6_connect(tp, nam, td); 795 else 796 #endif /* INET6 */ 797 error = tcp_connect(tp, nam, td); 798 if (error) 799 goto out; 800 tp->snd_wnd = TTCP_CLIENT_SND_WND; 801 tcp_mss(tp, -1); 802 } 803 if (flags & PRUS_EOF) { 804 /* 805 * Close the send side of the connection after 806 * the data is sent. 807 */ 808 INP_INFO_WLOCK_ASSERT(&tcbinfo); 809 socantsendmore(so); 810 tcp_usrclosed(tp); 811 } 812 if (headlocked) { 813 INP_INFO_WUNLOCK(&tcbinfo); 814 headlocked = 0; 815 } 816 if (!(inp->inp_flags & INP_DROPPED)) { 817 if (flags & PRUS_MORETOCOME) 818 tp->t_flags |= TF_MORETOCOME; 819 error = tcp_output_send(tp); 820 if (flags & PRUS_MORETOCOME) 821 tp->t_flags &= ~TF_MORETOCOME; 822 } 823 } else { 824 /* 825 * XXXRW: PRUS_EOF not implemented with PRUS_OOB? 826 */ 827 SOCKBUF_LOCK(&so->so_snd); 828 if (sbspace(&so->so_snd) < -512) { 829 SOCKBUF_UNLOCK(&so->so_snd); 830 m_freem(m); 831 error = ENOBUFS; 832 goto out; 833 } 834 /* 835 * According to RFC961 (Assigned Protocols), 836 * the urgent pointer points to the last octet 837 * of urgent data. We continue, however, 838 * to consider it to indicate the first octet 839 * of data past the urgent section. 840 * Otherwise, snd_up should be one lower. 841 */ 842 sbappendstream_locked(&so->so_snd, m); 843 SOCKBUF_UNLOCK(&so->so_snd); 844 if (nam && tp->t_state < TCPS_SYN_SENT) { 845 /* 846 * Do implied connect if not yet connected, 847 * initialize window to default value, and 848 * initialize maxseg/maxopd using peer's cached 849 * MSS. 850 */ 851 INP_INFO_WLOCK_ASSERT(&tcbinfo); 852 #ifdef INET6 853 if (isipv6) 854 error = tcp6_connect(tp, nam, td); 855 else 856 #endif /* INET6 */ 857 error = tcp_connect(tp, nam, td); 858 if (error) 859 goto out; 860 tp->snd_wnd = TTCP_CLIENT_SND_WND; 861 tcp_mss(tp, -1); 862 INP_INFO_WUNLOCK(&tcbinfo); 863 headlocked = 0; 864 } else if (nam) { 865 INP_INFO_WUNLOCK(&tcbinfo); 866 headlocked = 0; 867 } 868 tp->snd_up = tp->snd_una + so->so_snd.sb_cc; 869 tp->t_flags |= TF_FORCEDATA; 870 error = tcp_output_send(tp); 871 tp->t_flags &= ~TF_FORCEDATA; 872 } 873 out: 874 TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB : 875 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); 876 INP_WUNLOCK(inp); 877 if (headlocked) 878 INP_INFO_WUNLOCK(&tcbinfo); 879 return (error); 880 } 881 882 /* 883 * Abort the TCP. Drop the connection abruptly. 884 */ 885 static void 886 tcp_usr_abort(struct socket *so) 887 { 888 struct inpcb *inp; 889 struct tcpcb *tp = NULL; 890 TCPDEBUG0; 891 892 inp = sotoinpcb(so); 893 KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL")); 894 895 INP_INFO_WLOCK(&tcbinfo); 896 INP_WLOCK(inp); 897 KASSERT(inp->inp_socket != NULL, 898 ("tcp_usr_abort: inp_socket == NULL")); 899 900 /* 901 * If we still have full TCP state, and we're not dropped, drop. 902 */ 903 if (!(inp->inp_flags & INP_TIMEWAIT) && 904 !(inp->inp_flags & INP_DROPPED)) { 905 tp = intotcpcb(inp); 906 TCPDEBUG1(); 907 tcp_drop(tp, ECONNABORTED); 908 TCPDEBUG2(PRU_ABORT); 909 } 910 if (!(inp->inp_flags & INP_DROPPED)) { 911 SOCK_LOCK(so); 912 so->so_state |= SS_PROTOREF; 913 SOCK_UNLOCK(so); 914 inp->inp_flags |= INP_SOCKREF; 915 } 916 INP_WUNLOCK(inp); 917 INP_INFO_WUNLOCK(&tcbinfo); 918 } 919 920 /* 921 * TCP socket is closed. Start friendly disconnect. 922 */ 923 static void 924 tcp_usr_close(struct socket *so) 925 { 926 struct inpcb *inp; 927 struct tcpcb *tp = NULL; 928 TCPDEBUG0; 929 930 inp = sotoinpcb(so); 931 KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL")); 932 933 INP_INFO_WLOCK(&tcbinfo); 934 INP_WLOCK(inp); 935 KASSERT(inp->inp_socket != NULL, 936 ("tcp_usr_close: inp_socket == NULL")); 937 938 /* 939 * If we still have full TCP state, and we're not dropped, initiate 940 * a disconnect. 941 */ 942 if (!(inp->inp_flags & INP_TIMEWAIT) && 943 !(inp->inp_flags & INP_DROPPED)) { 944 tp = intotcpcb(inp); 945 TCPDEBUG1(); 946 tcp_disconnect(tp); 947 TCPDEBUG2(PRU_CLOSE); 948 } 949 if (!(inp->inp_flags & INP_DROPPED)) { 950 SOCK_LOCK(so); 951 so->so_state |= SS_PROTOREF; 952 SOCK_UNLOCK(so); 953 inp->inp_flags |= INP_SOCKREF; 954 } 955 INP_WUNLOCK(inp); 956 INP_INFO_WUNLOCK(&tcbinfo); 957 } 958 959 /* 960 * Receive out-of-band data. 961 */ 962 static int 963 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) 964 { 965 int error = 0; 966 struct inpcb *inp; 967 struct tcpcb *tp = NULL; 968 969 TCPDEBUG0; 970 inp = sotoinpcb(so); 971 KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL")); 972 INP_WLOCK(inp); 973 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 974 error = ECONNRESET; 975 goto out; 976 } 977 tp = intotcpcb(inp); 978 TCPDEBUG1(); 979 if ((so->so_oobmark == 0 && 980 (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) || 981 so->so_options & SO_OOBINLINE || 982 tp->t_oobflags & TCPOOB_HADDATA) { 983 error = EINVAL; 984 goto out; 985 } 986 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 987 error = EWOULDBLOCK; 988 goto out; 989 } 990 m->m_len = 1; 991 *mtod(m, caddr_t) = tp->t_iobc; 992 if ((flags & MSG_PEEK) == 0) 993 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 994 995 out: 996 TCPDEBUG2(PRU_RCVOOB); 997 INP_WUNLOCK(inp); 998 return (error); 999 } 1000 1001 struct pr_usrreqs tcp_usrreqs = { 1002 .pru_abort = tcp_usr_abort, 1003 .pru_accept = tcp_usr_accept, 1004 .pru_attach = tcp_usr_attach, 1005 .pru_bind = tcp_usr_bind, 1006 .pru_connect = tcp_usr_connect, 1007 .pru_control = in_control, 1008 .pru_detach = tcp_usr_detach, 1009 .pru_disconnect = tcp_usr_disconnect, 1010 .pru_listen = tcp_usr_listen, 1011 .pru_peeraddr = in_getpeeraddr, 1012 .pru_rcvd = tcp_usr_rcvd, 1013 .pru_rcvoob = tcp_usr_rcvoob, 1014 .pru_send = tcp_usr_send, 1015 .pru_shutdown = tcp_usr_shutdown, 1016 .pru_sockaddr = in_getsockaddr, 1017 .pru_sosetlabel = in_pcbsosetlabel, 1018 .pru_close = tcp_usr_close, 1019 }; 1020 1021 #ifdef INET6 1022 struct pr_usrreqs tcp6_usrreqs = { 1023 .pru_abort = tcp_usr_abort, 1024 .pru_accept = tcp6_usr_accept, 1025 .pru_attach = tcp_usr_attach, 1026 .pru_bind = tcp6_usr_bind, 1027 .pru_connect = tcp6_usr_connect, 1028 .pru_control = in6_control, 1029 .pru_detach = tcp_usr_detach, 1030 .pru_disconnect = tcp_usr_disconnect, 1031 .pru_listen = tcp6_usr_listen, 1032 .pru_peeraddr = in6_mapped_peeraddr, 1033 .pru_rcvd = tcp_usr_rcvd, 1034 .pru_rcvoob = tcp_usr_rcvoob, 1035 .pru_send = tcp_usr_send, 1036 .pru_shutdown = tcp_usr_shutdown, 1037 .pru_sockaddr = in6_mapped_sockaddr, 1038 .pru_sosetlabel = in_pcbsosetlabel, 1039 .pru_close = tcp_usr_close, 1040 }; 1041 #endif /* INET6 */ 1042 1043 /* 1044 * Common subroutine to open a TCP connection to remote host specified 1045 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local 1046 * port number if needed. Call in_pcbconnect_setup to do the routing and 1047 * to choose a local host address (interface). If there is an existing 1048 * incarnation of the same connection in TIME-WAIT state and if the remote 1049 * host was sending CC options and if the connection duration was < MSL, then 1050 * truncate the previous TIME-WAIT state and proceed. 1051 * Initialize connection parameters and enter SYN-SENT state. 1052 */ 1053 static int 1054 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td) 1055 { 1056 struct inpcb *inp = tp->t_inpcb, *oinp; 1057 struct socket *so = inp->inp_socket; 1058 struct in_addr laddr; 1059 u_short lport; 1060 int error; 1061 1062 INP_INFO_WLOCK_ASSERT(&tcbinfo); 1063 INP_WLOCK_ASSERT(inp); 1064 1065 if (inp->inp_lport == 0) { 1066 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); 1067 if (error) 1068 return error; 1069 } 1070 1071 /* 1072 * Cannot simply call in_pcbconnect, because there might be an 1073 * earlier incarnation of this same connection still in 1074 * TIME_WAIT state, creating an ADDRINUSE error. 1075 */ 1076 laddr = inp->inp_laddr; 1077 lport = inp->inp_lport; 1078 error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport, 1079 &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred); 1080 if (error && oinp == NULL) 1081 return error; 1082 if (oinp) 1083 return EADDRINUSE; 1084 inp->inp_laddr = laddr; 1085 in_pcbrehash(inp); 1086 1087 /* 1088 * Compute window scaling to request: 1089 * Scale to fit into sweet spot. See tcp_syncache.c. 1090 * XXX: This should move to tcp_output(). 1091 */ 1092 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 1093 (TCP_MAXWIN << tp->request_r_scale) < sb_max) 1094 tp->request_r_scale++; 1095 1096 soisconnecting(so); 1097 tcpstat.tcps_connattempt++; 1098 tp->t_state = TCPS_SYN_SENT; 1099 tcp_timer_activate(tp, TT_KEEP, tcp_keepinit); 1100 tp->iss = tcp_new_isn(tp); 1101 tp->t_bw_rtseq = tp->iss; 1102 tcp_sendseqinit(tp); 1103 1104 return 0; 1105 } 1106 1107 #ifdef INET6 1108 static int 1109 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td) 1110 { 1111 struct inpcb *inp = tp->t_inpcb, *oinp; 1112 struct socket *so = inp->inp_socket; 1113 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; 1114 struct in6_addr *addr6; 1115 int error; 1116 1117 INP_INFO_WLOCK_ASSERT(&tcbinfo); 1118 INP_WLOCK_ASSERT(inp); 1119 1120 if (inp->inp_lport == 0) { 1121 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); 1122 if (error) 1123 return error; 1124 } 1125 1126 /* 1127 * Cannot simply call in_pcbconnect, because there might be an 1128 * earlier incarnation of this same connection still in 1129 * TIME_WAIT state, creating an ADDRINUSE error. 1130 * in6_pcbladdr() also handles scope zone IDs. 1131 */ 1132 error = in6_pcbladdr(inp, nam, &addr6); 1133 if (error) 1134 return error; 1135 oinp = in6_pcblookup_hash(inp->inp_pcbinfo, 1136 &sin6->sin6_addr, sin6->sin6_port, 1137 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) 1138 ? addr6 1139 : &inp->in6p_laddr, 1140 inp->inp_lport, 0, NULL); 1141 if (oinp) 1142 return EADDRINUSE; 1143 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) 1144 inp->in6p_laddr = *addr6; 1145 inp->in6p_faddr = sin6->sin6_addr; 1146 inp->inp_fport = sin6->sin6_port; 1147 /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */ 1148 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK; 1149 if (inp->inp_flags & IN6P_AUTOFLOWLABEL) 1150 inp->inp_flow |= 1151 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK); 1152 in_pcbrehash(inp); 1153 1154 /* Compute window scaling to request. */ 1155 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 1156 (TCP_MAXWIN << tp->request_r_scale) < sb_max) 1157 tp->request_r_scale++; 1158 1159 soisconnecting(so); 1160 tcpstat.tcps_connattempt++; 1161 tp->t_state = TCPS_SYN_SENT; 1162 tcp_timer_activate(tp, TT_KEEP, tcp_keepinit); 1163 tp->iss = tcp_new_isn(tp); 1164 tp->t_bw_rtseq = tp->iss; 1165 tcp_sendseqinit(tp); 1166 1167 return 0; 1168 } 1169 #endif /* INET6 */ 1170 1171 /* 1172 * Export TCP internal state information via a struct tcp_info, based on the 1173 * Linux 2.6 API. Not ABI compatible as our constants are mapped differently 1174 * (TCP state machine, etc). We export all information using FreeBSD-native 1175 * constants -- for example, the numeric values for tcpi_state will differ 1176 * from Linux. 1177 */ 1178 static void 1179 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti) 1180 { 1181 1182 INP_WLOCK_ASSERT(tp->t_inpcb); 1183 bzero(ti, sizeof(*ti)); 1184 1185 ti->tcpi_state = tp->t_state; 1186 if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP)) 1187 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS; 1188 if (tp->t_flags & TF_SACK_PERMIT) 1189 ti->tcpi_options |= TCPI_OPT_SACK; 1190 if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) { 1191 ti->tcpi_options |= TCPI_OPT_WSCALE; 1192 ti->tcpi_snd_wscale = tp->snd_scale; 1193 ti->tcpi_rcv_wscale = tp->rcv_scale; 1194 } 1195 1196 ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT; 1197 ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT; 1198 1199 ti->tcpi_snd_ssthresh = tp->snd_ssthresh; 1200 ti->tcpi_snd_cwnd = tp->snd_cwnd; 1201 1202 /* 1203 * FreeBSD-specific extension fields for tcp_info. 1204 */ 1205 ti->tcpi_rcv_space = tp->rcv_wnd; 1206 ti->tcpi_rcv_nxt = tp->rcv_nxt; 1207 ti->tcpi_snd_wnd = tp->snd_wnd; 1208 ti->tcpi_snd_bwnd = tp->snd_bwnd; 1209 ti->tcpi_snd_nxt = tp->snd_nxt; 1210 ti->__tcpi_snd_mss = tp->t_maxseg; 1211 ti->__tcpi_rcv_mss = tp->t_maxseg; 1212 if (tp->t_flags & TF_TOE) 1213 ti->tcpi_options |= TCPI_OPT_TOE; 1214 } 1215 1216 /* 1217 * tcp_ctloutput() must drop the inpcb lock before performing copyin on 1218 * socket option arguments. When it re-acquires the lock after the copy, it 1219 * has to revalidate that the connection is still valid for the socket 1220 * option. 1221 */ 1222 #define INP_WLOCK_RECHECK(inp) do { \ 1223 INP_WLOCK(inp); \ 1224 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \ 1225 INP_WUNLOCK(inp); \ 1226 return (ECONNRESET); \ 1227 } \ 1228 tp = intotcpcb(inp); \ 1229 } while(0) 1230 1231 int 1232 tcp_ctloutput(struct socket *so, struct sockopt *sopt) 1233 { 1234 int error, opt, optval; 1235 struct inpcb *inp; 1236 struct tcpcb *tp; 1237 struct tcp_info ti; 1238 1239 error = 0; 1240 inp = sotoinpcb(so); 1241 KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL")); 1242 INP_WLOCK(inp); 1243 if (sopt->sopt_level != IPPROTO_TCP) { 1244 #ifdef INET6 1245 if (inp->inp_vflag & INP_IPV6PROTO) { 1246 INP_WUNLOCK(inp); 1247 error = ip6_ctloutput(so, sopt); 1248 } else { 1249 #endif /* INET6 */ 1250 INP_WUNLOCK(inp); 1251 error = ip_ctloutput(so, sopt); 1252 #ifdef INET6 1253 } 1254 #endif 1255 return (error); 1256 } 1257 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 1258 INP_WUNLOCK(inp); 1259 return (ECONNRESET); 1260 } 1261 1262 switch (sopt->sopt_dir) { 1263 case SOPT_SET: 1264 switch (sopt->sopt_name) { 1265 #ifdef TCP_SIGNATURE 1266 case TCP_MD5SIG: 1267 INP_WUNLOCK(inp); 1268 error = sooptcopyin(sopt, &optval, sizeof optval, 1269 sizeof optval); 1270 if (error) 1271 return (error); 1272 1273 INP_WLOCK_RECHECK(inp); 1274 if (optval > 0) 1275 tp->t_flags |= TF_SIGNATURE; 1276 else 1277 tp->t_flags &= ~TF_SIGNATURE; 1278 INP_WUNLOCK(inp); 1279 break; 1280 #endif /* TCP_SIGNATURE */ 1281 case TCP_NODELAY: 1282 case TCP_NOOPT: 1283 INP_WUNLOCK(inp); 1284 error = sooptcopyin(sopt, &optval, sizeof optval, 1285 sizeof optval); 1286 if (error) 1287 return (error); 1288 1289 INP_WLOCK_RECHECK(inp); 1290 switch (sopt->sopt_name) { 1291 case TCP_NODELAY: 1292 opt = TF_NODELAY; 1293 break; 1294 case TCP_NOOPT: 1295 opt = TF_NOOPT; 1296 break; 1297 default: 1298 opt = 0; /* dead code to fool gcc */ 1299 break; 1300 } 1301 1302 if (optval) 1303 tp->t_flags |= opt; 1304 else 1305 tp->t_flags &= ~opt; 1306 INP_WUNLOCK(inp); 1307 break; 1308 1309 case TCP_NOPUSH: 1310 INP_WUNLOCK(inp); 1311 error = sooptcopyin(sopt, &optval, sizeof optval, 1312 sizeof optval); 1313 if (error) 1314 return (error); 1315 1316 INP_WLOCK_RECHECK(inp); 1317 if (optval) 1318 tp->t_flags |= TF_NOPUSH; 1319 else { 1320 tp->t_flags &= ~TF_NOPUSH; 1321 error = tcp_output(tp); 1322 } 1323 INP_WUNLOCK(inp); 1324 break; 1325 1326 case TCP_MAXSEG: 1327 INP_WUNLOCK(inp); 1328 error = sooptcopyin(sopt, &optval, sizeof optval, 1329 sizeof optval); 1330 if (error) 1331 return (error); 1332 1333 INP_WLOCK_RECHECK(inp); 1334 if (optval > 0 && optval <= tp->t_maxseg && 1335 optval + 40 >= tcp_minmss) 1336 tp->t_maxseg = optval; 1337 else 1338 error = EINVAL; 1339 INP_WUNLOCK(inp); 1340 break; 1341 1342 case TCP_INFO: 1343 INP_WUNLOCK(inp); 1344 error = EINVAL; 1345 break; 1346 1347 default: 1348 INP_WUNLOCK(inp); 1349 error = ENOPROTOOPT; 1350 break; 1351 } 1352 break; 1353 1354 case SOPT_GET: 1355 tp = intotcpcb(inp); 1356 switch (sopt->sopt_name) { 1357 #ifdef TCP_SIGNATURE 1358 case TCP_MD5SIG: 1359 optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0; 1360 INP_WUNLOCK(inp); 1361 error = sooptcopyout(sopt, &optval, sizeof optval); 1362 break; 1363 #endif 1364 1365 case TCP_NODELAY: 1366 optval = tp->t_flags & TF_NODELAY; 1367 INP_WUNLOCK(inp); 1368 error = sooptcopyout(sopt, &optval, sizeof optval); 1369 break; 1370 case TCP_MAXSEG: 1371 optval = tp->t_maxseg; 1372 INP_WUNLOCK(inp); 1373 error = sooptcopyout(sopt, &optval, sizeof optval); 1374 break; 1375 case TCP_NOOPT: 1376 optval = tp->t_flags & TF_NOOPT; 1377 INP_WUNLOCK(inp); 1378 error = sooptcopyout(sopt, &optval, sizeof optval); 1379 break; 1380 case TCP_NOPUSH: 1381 optval = tp->t_flags & TF_NOPUSH; 1382 INP_WUNLOCK(inp); 1383 error = sooptcopyout(sopt, &optval, sizeof optval); 1384 break; 1385 case TCP_INFO: 1386 tcp_fill_info(tp, &ti); 1387 INP_WUNLOCK(inp); 1388 error = sooptcopyout(sopt, &ti, sizeof ti); 1389 break; 1390 default: 1391 INP_WUNLOCK(inp); 1392 error = ENOPROTOOPT; 1393 break; 1394 } 1395 break; 1396 } 1397 return (error); 1398 } 1399 #undef INP_WLOCK_RECHECK 1400 1401 /* 1402 * tcp_sendspace and tcp_recvspace are the default send and receive window 1403 * sizes, respectively. These are obsolescent (this information should 1404 * be set by the route). 1405 */ 1406 u_long tcp_sendspace = 1024*32; 1407 SYSCTL_ULONG(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW, 1408 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size"); 1409 u_long tcp_recvspace = 1024*64; 1410 SYSCTL_ULONG(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW, 1411 &tcp_recvspace , 0, "Maximum incoming TCP datagram size"); 1412 1413 /* 1414 * Attach TCP protocol to socket, allocating 1415 * internet protocol control block, tcp control block, 1416 * bufer space, and entering LISTEN state if to accept connections. 1417 */ 1418 static int 1419 tcp_attach(struct socket *so) 1420 { 1421 struct tcpcb *tp; 1422 struct inpcb *inp; 1423 int error; 1424 1425 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 1426 error = soreserve(so, tcp_sendspace, tcp_recvspace); 1427 if (error) 1428 return (error); 1429 } 1430 so->so_rcv.sb_flags |= SB_AUTOSIZE; 1431 so->so_snd.sb_flags |= SB_AUTOSIZE; 1432 INP_INFO_WLOCK(&tcbinfo); 1433 error = in_pcballoc(so, &tcbinfo); 1434 if (error) { 1435 INP_INFO_WUNLOCK(&tcbinfo); 1436 return (error); 1437 } 1438 inp = sotoinpcb(so); 1439 #ifdef INET6 1440 if (inp->inp_vflag & INP_IPV6PROTO) { 1441 inp->inp_vflag |= INP_IPV6; 1442 inp->in6p_hops = -1; /* use kernel default */ 1443 } 1444 else 1445 #endif 1446 inp->inp_vflag |= INP_IPV4; 1447 tp = tcp_newtcpcb(inp); 1448 if (tp == NULL) { 1449 in_pcbdetach(inp); 1450 in_pcbfree(inp); 1451 INP_INFO_WUNLOCK(&tcbinfo); 1452 return (ENOBUFS); 1453 } 1454 tp->t_state = TCPS_CLOSED; 1455 INP_WUNLOCK(inp); 1456 INP_INFO_WUNLOCK(&tcbinfo); 1457 return (0); 1458 } 1459 1460 /* 1461 * Initiate (or continue) disconnect. 1462 * If embryonic state, just send reset (once). 1463 * If in ``let data drain'' option and linger null, just drop. 1464 * Otherwise (hard), mark socket disconnecting and drop 1465 * current input data; switch states based on user close, and 1466 * send segment to peer (with FIN). 1467 */ 1468 static void 1469 tcp_disconnect(struct tcpcb *tp) 1470 { 1471 struct inpcb *inp = tp->t_inpcb; 1472 struct socket *so = inp->inp_socket; 1473 1474 INP_INFO_WLOCK_ASSERT(&tcbinfo); 1475 INP_WLOCK_ASSERT(inp); 1476 1477 /* 1478 * Neither tcp_close() nor tcp_drop() should return NULL, as the 1479 * socket is still open. 1480 */ 1481 if (tp->t_state < TCPS_ESTABLISHED) { 1482 tp = tcp_close(tp); 1483 KASSERT(tp != NULL, 1484 ("tcp_disconnect: tcp_close() returned NULL")); 1485 } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) { 1486 tp = tcp_drop(tp, 0); 1487 KASSERT(tp != NULL, 1488 ("tcp_disconnect: tcp_drop() returned NULL")); 1489 } else { 1490 soisdisconnecting(so); 1491 sbflush(&so->so_rcv); 1492 tcp_usrclosed(tp); 1493 if (!(inp->inp_flags & INP_DROPPED)) 1494 tcp_output_disconnect(tp); 1495 } 1496 } 1497 1498 /* 1499 * User issued close, and wish to trail through shutdown states: 1500 * if never received SYN, just forget it. If got a SYN from peer, 1501 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 1502 * If already got a FIN from peer, then almost done; go to LAST_ACK 1503 * state. In all other cases, have already sent FIN to peer (e.g. 1504 * after PRU_SHUTDOWN), and just have to play tedious game waiting 1505 * for peer to send FIN or not respond to keep-alives, etc. 1506 * We can let the user exit from the close as soon as the FIN is acked. 1507 */ 1508 static void 1509 tcp_usrclosed(struct tcpcb *tp) 1510 { 1511 1512 INP_INFO_WLOCK_ASSERT(&tcbinfo); 1513 INP_WLOCK_ASSERT(tp->t_inpcb); 1514 1515 switch (tp->t_state) { 1516 case TCPS_LISTEN: 1517 tcp_offload_listen_close(tp); 1518 /* FALLTHROUGH */ 1519 case TCPS_CLOSED: 1520 tp->t_state = TCPS_CLOSED; 1521 tp = tcp_close(tp); 1522 /* 1523 * tcp_close() should never return NULL here as the socket is 1524 * still open. 1525 */ 1526 KASSERT(tp != NULL, 1527 ("tcp_usrclosed: tcp_close() returned NULL")); 1528 break; 1529 1530 case TCPS_SYN_SENT: 1531 case TCPS_SYN_RECEIVED: 1532 tp->t_flags |= TF_NEEDFIN; 1533 break; 1534 1535 case TCPS_ESTABLISHED: 1536 tp->t_state = TCPS_FIN_WAIT_1; 1537 break; 1538 1539 case TCPS_CLOSE_WAIT: 1540 tp->t_state = TCPS_LAST_ACK; 1541 break; 1542 } 1543 if (tp->t_state >= TCPS_FIN_WAIT_2) { 1544 soisdisconnected(tp->t_inpcb->inp_socket); 1545 /* Prevent the connection hanging in FIN_WAIT_2 forever. */ 1546 if (tp->t_state == TCPS_FIN_WAIT_2) { 1547 int timeout; 1548 1549 timeout = (tcp_fast_finwait2_recycle) ? 1550 tcp_finwait2_timeout : tcp_maxidle; 1551 tcp_timer_activate(tp, TT_2MSL, timeout); 1552 } 1553 } 1554 } 1555 1556 #ifdef DDB 1557 static void 1558 db_print_indent(int indent) 1559 { 1560 int i; 1561 1562 for (i = 0; i < indent; i++) 1563 db_printf(" "); 1564 } 1565 1566 static void 1567 db_print_tstate(int t_state) 1568 { 1569 1570 switch (t_state) { 1571 case TCPS_CLOSED: 1572 db_printf("TCPS_CLOSED"); 1573 return; 1574 1575 case TCPS_LISTEN: 1576 db_printf("TCPS_LISTEN"); 1577 return; 1578 1579 case TCPS_SYN_SENT: 1580 db_printf("TCPS_SYN_SENT"); 1581 return; 1582 1583 case TCPS_SYN_RECEIVED: 1584 db_printf("TCPS_SYN_RECEIVED"); 1585 return; 1586 1587 case TCPS_ESTABLISHED: 1588 db_printf("TCPS_ESTABLISHED"); 1589 return; 1590 1591 case TCPS_CLOSE_WAIT: 1592 db_printf("TCPS_CLOSE_WAIT"); 1593 return; 1594 1595 case TCPS_FIN_WAIT_1: 1596 db_printf("TCPS_FIN_WAIT_1"); 1597 return; 1598 1599 case TCPS_CLOSING: 1600 db_printf("TCPS_CLOSING"); 1601 return; 1602 1603 case TCPS_LAST_ACK: 1604 db_printf("TCPS_LAST_ACK"); 1605 return; 1606 1607 case TCPS_FIN_WAIT_2: 1608 db_printf("TCPS_FIN_WAIT_2"); 1609 return; 1610 1611 case TCPS_TIME_WAIT: 1612 db_printf("TCPS_TIME_WAIT"); 1613 return; 1614 1615 default: 1616 db_printf("unknown"); 1617 return; 1618 } 1619 } 1620 1621 static void 1622 db_print_tflags(u_int t_flags) 1623 { 1624 int comma; 1625 1626 comma = 0; 1627 if (t_flags & TF_ACKNOW) { 1628 db_printf("%sTF_ACKNOW", comma ? ", " : ""); 1629 comma = 1; 1630 } 1631 if (t_flags & TF_DELACK) { 1632 db_printf("%sTF_DELACK", comma ? ", " : ""); 1633 comma = 1; 1634 } 1635 if (t_flags & TF_NODELAY) { 1636 db_printf("%sTF_NODELAY", comma ? ", " : ""); 1637 comma = 1; 1638 } 1639 if (t_flags & TF_NOOPT) { 1640 db_printf("%sTF_NOOPT", comma ? ", " : ""); 1641 comma = 1; 1642 } 1643 if (t_flags & TF_SENTFIN) { 1644 db_printf("%sTF_SENTFIN", comma ? ", " : ""); 1645 comma = 1; 1646 } 1647 if (t_flags & TF_REQ_SCALE) { 1648 db_printf("%sTF_REQ_SCALE", comma ? ", " : ""); 1649 comma = 1; 1650 } 1651 if (t_flags & TF_RCVD_SCALE) { 1652 db_printf("%sTF_RECVD_SCALE", comma ? ", " : ""); 1653 comma = 1; 1654 } 1655 if (t_flags & TF_REQ_TSTMP) { 1656 db_printf("%sTF_REQ_TSTMP", comma ? ", " : ""); 1657 comma = 1; 1658 } 1659 if (t_flags & TF_RCVD_TSTMP) { 1660 db_printf("%sTF_RCVD_TSTMP", comma ? ", " : ""); 1661 comma = 1; 1662 } 1663 if (t_flags & TF_SACK_PERMIT) { 1664 db_printf("%sTF_SACK_PERMIT", comma ? ", " : ""); 1665 comma = 1; 1666 } 1667 if (t_flags & TF_NEEDSYN) { 1668 db_printf("%sTF_NEEDSYN", comma ? ", " : ""); 1669 comma = 1; 1670 } 1671 if (t_flags & TF_NEEDFIN) { 1672 db_printf("%sTF_NEEDFIN", comma ? ", " : ""); 1673 comma = 1; 1674 } 1675 if (t_flags & TF_NOPUSH) { 1676 db_printf("%sTF_NOPUSH", comma ? ", " : ""); 1677 comma = 1; 1678 } 1679 if (t_flags & TF_NOPUSH) { 1680 db_printf("%sTF_NOPUSH", comma ? ", " : ""); 1681 comma = 1; 1682 } 1683 if (t_flags & TF_MORETOCOME) { 1684 db_printf("%sTF_MORETOCOME", comma ? ", " : ""); 1685 comma = 1; 1686 } 1687 if (t_flags & TF_LQ_OVERFLOW) { 1688 db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : ""); 1689 comma = 1; 1690 } 1691 if (t_flags & TF_LASTIDLE) { 1692 db_printf("%sTF_LASTIDLE", comma ? ", " : ""); 1693 comma = 1; 1694 } 1695 if (t_flags & TF_RXWIN0SENT) { 1696 db_printf("%sTF_RXWIN0SENT", comma ? ", " : ""); 1697 comma = 1; 1698 } 1699 if (t_flags & TF_FASTRECOVERY) { 1700 db_printf("%sTF_FASTRECOVERY", comma ? ", " : ""); 1701 comma = 1; 1702 } 1703 if (t_flags & TF_WASFRECOVERY) { 1704 db_printf("%sTF_WASFRECOVERY", comma ? ", " : ""); 1705 comma = 1; 1706 } 1707 if (t_flags & TF_SIGNATURE) { 1708 db_printf("%sTF_SIGNATURE", comma ? ", " : ""); 1709 comma = 1; 1710 } 1711 if (t_flags & TF_FORCEDATA) { 1712 db_printf("%sTF_FORCEDATA", comma ? ", " : ""); 1713 comma = 1; 1714 } 1715 if (t_flags & TF_TSO) { 1716 db_printf("%sTF_TSO", comma ? ", " : ""); 1717 comma = 1; 1718 } 1719 } 1720 1721 static void 1722 db_print_toobflags(char t_oobflags) 1723 { 1724 int comma; 1725 1726 comma = 0; 1727 if (t_oobflags & TCPOOB_HAVEDATA) { 1728 db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : ""); 1729 comma = 1; 1730 } 1731 if (t_oobflags & TCPOOB_HADDATA) { 1732 db_printf("%sTCPOOB_HADDATA", comma ? ", " : ""); 1733 comma = 1; 1734 } 1735 } 1736 1737 static void 1738 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent) 1739 { 1740 1741 db_print_indent(indent); 1742 db_printf("%s at %p\n", name, tp); 1743 1744 indent += 2; 1745 1746 db_print_indent(indent); 1747 db_printf("t_segq first: %p t_segqlen: %d t_dupacks: %d\n", 1748 LIST_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks); 1749 1750 db_print_indent(indent); 1751 db_printf("tt_rexmt: %p tt_persist: %p tt_keep: %p\n", 1752 &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep); 1753 1754 db_print_indent(indent); 1755 db_printf("tt_2msl: %p tt_delack: %p t_inpcb: %p\n", &tp->t_timers->tt_2msl, 1756 &tp->t_timers->tt_delack, tp->t_inpcb); 1757 1758 db_print_indent(indent); 1759 db_printf("t_state: %d (", tp->t_state); 1760 db_print_tstate(tp->t_state); 1761 db_printf(")\n"); 1762 1763 db_print_indent(indent); 1764 db_printf("t_flags: 0x%x (", tp->t_flags); 1765 db_print_tflags(tp->t_flags); 1766 db_printf(")\n"); 1767 1768 db_print_indent(indent); 1769 db_printf("snd_una: 0x%08x snd_max: 0x%08x snd_nxt: x0%08x\n", 1770 tp->snd_una, tp->snd_max, tp->snd_nxt); 1771 1772 db_print_indent(indent); 1773 db_printf("snd_up: 0x%08x snd_wl1: 0x%08x snd_wl2: 0x%08x\n", 1774 tp->snd_up, tp->snd_wl1, tp->snd_wl2); 1775 1776 db_print_indent(indent); 1777 db_printf("iss: 0x%08x irs: 0x%08x rcv_nxt: 0x%08x\n", 1778 tp->iss, tp->irs, tp->rcv_nxt); 1779 1780 db_print_indent(indent); 1781 db_printf("rcv_adv: 0x%08x rcv_wnd: %lu rcv_up: 0x%08x\n", 1782 tp->rcv_adv, tp->rcv_wnd, tp->rcv_up); 1783 1784 db_print_indent(indent); 1785 db_printf("snd_wnd: %lu snd_cwnd: %lu snd_bwnd: %lu\n", 1786 tp->snd_wnd, tp->snd_cwnd, tp->snd_bwnd); 1787 1788 db_print_indent(indent); 1789 db_printf("snd_ssthresh: %lu snd_bandwidth: %lu snd_recover: " 1790 "0x%08x\n", tp->snd_ssthresh, tp->snd_bandwidth, 1791 tp->snd_recover); 1792 1793 db_print_indent(indent); 1794 db_printf("t_maxopd: %u t_rcvtime: %lu t_startime: %lu\n", 1795 tp->t_maxopd, tp->t_rcvtime, tp->t_starttime); 1796 1797 db_print_indent(indent); 1798 db_printf("t_rttime: %d t_rtsq: 0x%08x t_bw_rtttime: %d\n", 1799 tp->t_rtttime, tp->t_rtseq, tp->t_bw_rtttime); 1800 1801 db_print_indent(indent); 1802 db_printf("t_bw_rtseq: 0x%08x t_rxtcur: %d t_maxseg: %u " 1803 "t_srtt: %d\n", tp->t_bw_rtseq, tp->t_rxtcur, tp->t_maxseg, 1804 tp->t_srtt); 1805 1806 db_print_indent(indent); 1807 db_printf("t_rttvar: %d t_rxtshift: %d t_rttmin: %u " 1808 "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin, 1809 tp->t_rttbest); 1810 1811 db_print_indent(indent); 1812 db_printf("t_rttupdated: %lu max_sndwnd: %lu t_softerror: %d\n", 1813 tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror); 1814 1815 db_print_indent(indent); 1816 db_printf("t_oobflags: 0x%x (", tp->t_oobflags); 1817 db_print_toobflags(tp->t_oobflags); 1818 db_printf(") t_iobc: 0x%02x\n", tp->t_iobc); 1819 1820 db_print_indent(indent); 1821 db_printf("snd_scale: %u rcv_scale: %u request_r_scale: %u\n", 1822 tp->snd_scale, tp->rcv_scale, tp->request_r_scale); 1823 1824 db_print_indent(indent); 1825 db_printf("ts_recent: %u ts_recent_age: %lu\n", 1826 tp->ts_recent, tp->ts_recent_age); 1827 1828 db_print_indent(indent); 1829 db_printf("ts_offset: %u last_ack_sent: 0x%08x snd_cwnd_prev: " 1830 "%lu\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev); 1831 1832 db_print_indent(indent); 1833 db_printf("snd_ssthresh_prev: %lu snd_recover_prev: 0x%08x " 1834 "t_badrxtwin: %lu\n", tp->snd_ssthresh_prev, 1835 tp->snd_recover_prev, tp->t_badrxtwin); 1836 1837 db_print_indent(indent); 1838 db_printf("snd_numholes: %d snd_holes first: %p\n", 1839 tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes)); 1840 1841 db_print_indent(indent); 1842 db_printf("snd_fack: 0x%08x rcv_numsacks: %d sack_newdata: " 1843 "0x%08x\n", tp->snd_fack, tp->rcv_numsacks, tp->sack_newdata); 1844 1845 /* Skip sackblks, sackhint. */ 1846 1847 db_print_indent(indent); 1848 db_printf("t_rttlow: %d rfbuf_ts: %u rfbuf_cnt: %d\n", 1849 tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt); 1850 } 1851 1852 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb) 1853 { 1854 struct tcpcb *tp; 1855 1856 if (!have_addr) { 1857 db_printf("usage: show tcpcb <addr>\n"); 1858 return; 1859 } 1860 tp = (struct tcpcb *)addr; 1861 1862 db_print_tcpcb(tp, "tcpcb", 0); 1863 } 1864 #endif
Cache object: 336fe656475939bc03361cdf321f4781
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]