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) 2003, 2004 Jeffrey M. Hsu. All rights reserved. 36 * 37 * License terms: all terms for the DragonFly license above plus the following: 38 * 39 * 4. All advertising materials mentioning features or use of this software 40 * must display the following acknowledgement: 41 * 42 * This product includes software developed by Jeffrey M. Hsu 43 * for the DragonFly Project. 44 * 45 * This requirement may be waived with permission from Jeffrey Hsu. 46 * This requirement will sunset and may be removed on July 8 2005, 47 * after which the standard DragonFly license (as shown above) will 48 * apply. 49 */ 50 51 /* 52 * Copyright (c) 1982, 1986, 1988, 1993 53 * The Regents of the University of California. All rights reserved. 54 * 55 * Redistribution and use in source and binary forms, with or without 56 * modification, are permitted provided that the following conditions 57 * are met: 58 * 1. Redistributions of source code must retain the above copyright 59 * notice, this list of conditions and the following disclaimer. 60 * 2. Redistributions in binary form must reproduce the above copyright 61 * notice, this list of conditions and the following disclaimer in the 62 * documentation and/or other materials provided with the distribution. 63 * 3. All advertising materials mentioning features or use of this software 64 * must display the following acknowledgement: 65 * This product includes software developed by the University of 66 * California, Berkeley and its contributors. 67 * 4. Neither the name of the University nor the names of its contributors 68 * may be used to endorse or promote products derived from this software 69 * without specific prior written permission. 70 * 71 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 72 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 73 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 74 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 75 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 76 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 77 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 78 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 79 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 80 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 81 * SUCH DAMAGE. 82 * 83 * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94 84 * $FreeBSD: src/sys/netinet/tcp_usrreq.c,v 1.51.2.17 2002/10/11 11:46:44 ume Exp $ 85 * $DragonFly: src/sys/netinet/tcp_usrreq.c,v 1.27 2004/10/27 03:43:47 dillon Exp $ 86 */ 87 88 #include "opt_ipsec.h" 89 #include "opt_inet6.h" 90 #include "opt_tcpdebug.h" 91 92 #include <sys/param.h> 93 #include <sys/systm.h> 94 #include <sys/kernel.h> 95 #include <sys/malloc.h> 96 #include <sys/sysctl.h> 97 #include <sys/globaldata.h> 98 #include <sys/thread.h> 99 100 #include <sys/mbuf.h> 101 #ifdef INET6 102 #include <sys/domain.h> 103 #endif /* INET6 */ 104 #include <sys/socket.h> 105 #include <sys/socketvar.h> 106 #include <sys/protosw.h> 107 108 #include <sys/msgport2.h> 109 110 #include <net/if.h> 111 #include <net/netisr.h> 112 #include <net/route.h> 113 114 #include <netinet/in.h> 115 #include <netinet/in_systm.h> 116 #ifdef INET6 117 #include <netinet/ip6.h> 118 #endif 119 #include <netinet/in_pcb.h> 120 #ifdef INET6 121 #include <netinet6/in6_pcb.h> 122 #endif 123 #include <netinet/in_var.h> 124 #include <netinet/ip_var.h> 125 #ifdef INET6 126 #include <netinet6/ip6_var.h> 127 #endif 128 #include <netinet/tcp.h> 129 #include <netinet/tcp_fsm.h> 130 #include <netinet/tcp_seq.h> 131 #include <netinet/tcp_timer.h> 132 #include <netinet/tcp_var.h> 133 #include <netinet/tcpip.h> 134 #ifdef TCPDEBUG 135 #include <netinet/tcp_debug.h> 136 #endif 137 138 #ifdef IPSEC 139 #include <netinet6/ipsec.h> 140 #endif /*IPSEC*/ 141 142 /* 143 * TCP protocol interface to socket abstraction. 144 */ 145 extern char *tcpstates[]; /* XXX ??? */ 146 147 static int tcp_attach (struct socket *, struct pru_attach_info *); 148 static int tcp_connect (struct tcpcb *, struct sockaddr *, 149 struct thread *); 150 #ifdef INET6 151 static int tcp6_connect (struct tcpcb *, struct sockaddr *, 152 struct thread *); 153 #endif /* INET6 */ 154 static struct tcpcb * 155 tcp_disconnect (struct tcpcb *); 156 static struct tcpcb * 157 tcp_usrclosed (struct tcpcb *); 158 159 #ifdef TCPDEBUG 160 #define TCPDEBUG0 int ostate = 0 161 #define TCPDEBUG1() ostate = tp ? tp->t_state : 0 162 #define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \ 163 tcp_trace(TA_USER, ostate, tp, 0, 0, req) 164 #else 165 #define TCPDEBUG0 166 #define TCPDEBUG1() 167 #define TCPDEBUG2(req) 168 #endif 169 170 /* 171 * TCP attaches to socket via pru_attach(), reserving space, 172 * and an internet control block. 173 */ 174 static int 175 tcp_usr_attach(struct socket *so, int proto, struct pru_attach_info *ai) 176 { 177 int s = splnet(); 178 int error; 179 struct inpcb *inp = sotoinpcb(so); 180 struct tcpcb *tp = 0; 181 TCPDEBUG0; 182 183 TCPDEBUG1(); 184 if (inp) { 185 error = EISCONN; 186 goto out; 187 } 188 189 error = tcp_attach(so, ai); 190 if (error) 191 goto out; 192 193 if ((so->so_options & SO_LINGER) && so->so_linger == 0) 194 so->so_linger = TCP_LINGERTIME; 195 tp = sototcpcb(so); 196 out: 197 TCPDEBUG2(PRU_ATTACH); 198 splx(s); 199 return error; 200 } 201 202 /* 203 * pru_detach() detaches the TCP protocol from the socket. 204 * If the protocol state is non-embryonic, then can't 205 * do this directly: have to initiate a pru_disconnect(), 206 * which may finish later; embryonic TCB's can just 207 * be discarded here. 208 */ 209 static int 210 tcp_usr_detach(struct socket *so) 211 { 212 int s = splnet(); 213 int error = 0; 214 struct inpcb *inp = sotoinpcb(so); 215 struct tcpcb *tp; 216 TCPDEBUG0; 217 218 if (inp == NULL) { 219 splx(s); 220 return EINVAL; /* XXX */ 221 } 222 223 /* 224 * It's possible for the tcpcb (tp) to disconnect from the inp due 225 * to tcp_drop()->tcp_close() being called. This may occur *after* 226 * the detach message has been queued so we may find a NULL tp here. 227 */ 228 if ((tp = intotcpcb(inp)) != NULL) { 229 TCPDEBUG1(); 230 tp = tcp_disconnect(tp); 231 TCPDEBUG2(PRU_DETACH); 232 } 233 splx(s); 234 return error; 235 } 236 237 #define COMMON_START() TCPDEBUG0; \ 238 do { \ 239 if (inp == 0) { \ 240 splx(s); \ 241 return EINVAL; \ 242 } \ 243 tp = intotcpcb(inp); \ 244 TCPDEBUG1(); \ 245 } while(0) 246 247 #define COMMON_END(req) out: TCPDEBUG2(req); splx(s); return error; goto out 248 249 250 /* 251 * Give the socket an address. 252 */ 253 static int 254 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 255 { 256 int s = splnet(); 257 int error = 0; 258 struct inpcb *inp = sotoinpcb(so); 259 struct tcpcb *tp; 260 struct sockaddr_in *sinp; 261 262 COMMON_START(); 263 264 /* 265 * Must check for multicast addresses and disallow binding 266 * to them. 267 */ 268 sinp = (struct sockaddr_in *)nam; 269 if (sinp->sin_family == AF_INET && 270 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 271 error = EAFNOSUPPORT; 272 goto out; 273 } 274 error = in_pcbbind(inp, nam, td); 275 if (error) 276 goto out; 277 COMMON_END(PRU_BIND); 278 279 } 280 281 #ifdef INET6 282 static int 283 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 284 { 285 int s = splnet(); 286 int error = 0; 287 struct inpcb *inp = sotoinpcb(so); 288 struct tcpcb *tp; 289 struct sockaddr_in6 *sin6p; 290 291 COMMON_START(); 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 lwkt_msg nm_lmsg; 328 struct inpcb *nm_inp; 329 struct inpcbinfo *nm_pcbinfo; 330 }; 331 332 static int 333 in_pcbinswildcardhash_handler(struct lwkt_msg *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_lmsg, 0); 339 return (EASYNC); 340 } 341 #endif 342 343 /* 344 * Prepare to accept connections. 345 */ 346 static int 347 tcp_usr_listen(struct socket *so, struct thread *td) 348 { 349 int s = splnet(); 350 int error = 0; 351 struct inpcb *inp = sotoinpcb(so); 352 struct tcpcb *tp; 353 #ifdef SMP 354 int cpu; 355 #endif 356 357 COMMON_START(); 358 if (inp->inp_lport == 0) { 359 error = in_pcbbind(inp, NULL, td); 360 if (error != 0) 361 goto out; 362 } 363 364 tp->t_state = TCPS_LISTEN; 365 #ifdef SMP 366 /* 367 * We have to set the flag because we can't have other cpus messing 368 * with our inp's flags. 369 */ 370 inp->inp_flags |= INP_WILDCARD_MP; 371 for (cpu = 0; cpu < ncpus2; cpu++) { 372 struct netmsg_inswildcard *msg; 373 374 if (cpu == mycpu->gd_cpuid) { 375 in_pcbinswildcardhash_oncpu(inp, &tcbinfo[cpu]); 376 continue; 377 } 378 379 msg = malloc(sizeof(struct netmsg_inswildcard), M_LWKTMSG, 380 M_INTWAIT); 381 lwkt_initmsg(&msg->nm_lmsg, &netisr_afree_rport, 0, 382 lwkt_cmd_func(in_pcbinswildcardhash_handler), 383 lwkt_cmd_op_none); 384 msg->nm_inp = inp; 385 msg->nm_pcbinfo = &tcbinfo[cpu]; 386 lwkt_sendmsg(tcp_cport(cpu), &msg->nm_lmsg); 387 } 388 #else 389 in_pcbinswildcardhash(inp); 390 #endif 391 COMMON_END(PRU_LISTEN); 392 } 393 394 #ifdef INET6 395 static int 396 tcp6_usr_listen(struct socket *so, struct thread *td) 397 { 398 int s = splnet(); 399 int error = 0; 400 struct inpcb *inp = sotoinpcb(so); 401 struct tcpcb *tp; 402 403 COMMON_START(); 404 if (inp->inp_lport == 0) { 405 inp->inp_vflag &= ~INP_IPV4; 406 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) 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 in_pcbinswildcardhash(inp); 413 COMMON_END(PRU_LISTEN); 414 } 415 #endif /* INET6 */ 416 417 /* 418 * Initiate connection to peer. 419 * Create a template for use in transmissions on this connection. 420 * Enter SYN_SENT state, and mark socket as connecting. 421 * Start keep-alive timer, and seed output sequence space. 422 * Send initial segment on connection. 423 */ 424 static int 425 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 426 { 427 int s = splnet(); 428 int error = 0; 429 struct inpcb *inp = sotoinpcb(so); 430 struct tcpcb *tp; 431 struct sockaddr_in *sinp; 432 433 COMMON_START(); 434 435 /* 436 * Must disallow TCP ``connections'' to multicast addresses. 437 */ 438 sinp = (struct sockaddr_in *)nam; 439 if (sinp->sin_family == AF_INET 440 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 441 error = EAFNOSUPPORT; 442 goto out; 443 } 444 445 prison_remote_ip(td, 0, &sinp->sin_addr.s_addr); 446 447 if ((error = tcp_connect(tp, nam, td)) != 0) 448 goto out; 449 error = tcp_output(tp); 450 COMMON_END(PRU_CONNECT); 451 } 452 453 #ifdef INET6 454 static int 455 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 456 { 457 int s = splnet(); 458 int error = 0; 459 struct inpcb *inp = sotoinpcb(so); 460 struct tcpcb *tp; 461 struct sockaddr_in6 *sin6p; 462 463 COMMON_START(); 464 465 /* 466 * Must disallow TCP ``connections'' to multicast addresses. 467 */ 468 sin6p = (struct sockaddr_in6 *)nam; 469 if (sin6p->sin6_family == AF_INET6 470 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { 471 error = EAFNOSUPPORT; 472 goto out; 473 } 474 475 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { 476 struct sockaddr_in sin; 477 478 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { 479 error = EINVAL; 480 goto out; 481 } 482 483 in6_sin6_2_sin(&sin, sin6p); 484 inp->inp_vflag |= INP_IPV4; 485 inp->inp_vflag &= ~INP_IPV6; 486 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0) 487 goto out; 488 error = tcp_output(tp); 489 goto out; 490 } 491 inp->inp_vflag &= ~INP_IPV4; 492 inp->inp_vflag |= INP_IPV6; 493 inp->inp_inc.inc_isipv6 = 1; 494 if ((error = tcp6_connect(tp, nam, td)) != 0) 495 goto out; 496 error = tcp_output(tp); 497 COMMON_END(PRU_CONNECT); 498 } 499 #endif /* INET6 */ 500 501 /* 502 * Initiate disconnect from peer. 503 * If connection never passed embryonic stage, just drop; 504 * else if don't need to let data drain, then can just drop anyways, 505 * else have to begin TCP shutdown process: mark socket disconnecting, 506 * drain unread data, state switch to reflect user close, and 507 * send segment (e.g. FIN) to peer. Socket will be really disconnected 508 * when peer sends FIN and acks ours. 509 * 510 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 511 */ 512 static int 513 tcp_usr_disconnect(struct socket *so) 514 { 515 int s = splnet(); 516 int error = 0; 517 struct inpcb *inp = sotoinpcb(so); 518 struct tcpcb *tp; 519 520 COMMON_START(); 521 tp = tcp_disconnect(tp); 522 COMMON_END(PRU_DISCONNECT); 523 } 524 525 /* 526 * Accept a connection. Essentially all the work is 527 * done at higher levels; just return the address 528 * of the peer, storing through addr. 529 */ 530 static int 531 tcp_usr_accept(struct socket *so, struct sockaddr **nam) 532 { 533 int s = splnet(); 534 int error = 0; 535 struct inpcb *inp = sotoinpcb(so); 536 struct tcpcb *tp = NULL; 537 TCPDEBUG0; 538 539 if (so->so_state & SS_ISDISCONNECTED) { 540 error = ECONNABORTED; 541 goto out; 542 } 543 if (inp == 0) { 544 splx(s); 545 return (EINVAL); 546 } 547 tp = intotcpcb(inp); 548 TCPDEBUG1(); 549 in_setpeeraddr(so, nam); 550 COMMON_END(PRU_ACCEPT); 551 } 552 553 #ifdef INET6 554 static int 555 tcp6_usr_accept(struct socket *so, struct sockaddr **nam) 556 { 557 int s = splnet(); 558 int error = 0; 559 struct inpcb *inp = sotoinpcb(so); 560 struct tcpcb *tp = NULL; 561 TCPDEBUG0; 562 563 if (so->so_state & SS_ISDISCONNECTED) { 564 error = ECONNABORTED; 565 goto out; 566 } 567 if (inp == 0) { 568 splx(s); 569 return (EINVAL); 570 } 571 tp = intotcpcb(inp); 572 TCPDEBUG1(); 573 in6_mapped_peeraddr(so, nam); 574 COMMON_END(PRU_ACCEPT); 575 } 576 #endif /* INET6 */ 577 /* 578 * Mark the connection as being incapable of further output. 579 */ 580 static int 581 tcp_usr_shutdown(struct socket *so) 582 { 583 int s = splnet(); 584 int error = 0; 585 struct inpcb *inp = sotoinpcb(so); 586 struct tcpcb *tp; 587 588 COMMON_START(); 589 socantsendmore(so); 590 tp = tcp_usrclosed(tp); 591 if (tp) 592 error = tcp_output(tp); 593 COMMON_END(PRU_SHUTDOWN); 594 } 595 596 /* 597 * After a receive, possibly send window update to peer. 598 */ 599 static int 600 tcp_usr_rcvd(struct socket *so, int flags) 601 { 602 int s = splnet(); 603 int error = 0; 604 struct inpcb *inp = sotoinpcb(so); 605 struct tcpcb *tp; 606 607 COMMON_START(); 608 tcp_output(tp); 609 COMMON_END(PRU_RCVD); 610 } 611 612 /* 613 * Do a send by putting data in output queue and updating urgent 614 * marker if URG set. Possibly send more data. Unlike the other 615 * pru_*() routines, the mbuf chains are our responsibility. We 616 * must either enqueue them or free them. The other pru_* routines 617 * generally are caller-frees. 618 */ 619 static int 620 tcp_usr_send(struct socket *so, int flags, struct mbuf *m, 621 struct sockaddr *nam, struct mbuf *control, struct thread *td) 622 { 623 int s = splnet(); 624 int error = 0; 625 struct inpcb *inp = sotoinpcb(so); 626 struct tcpcb *tp; 627 #ifdef INET6 628 int isipv6; 629 #endif 630 TCPDEBUG0; 631 632 if (inp == NULL) { 633 /* 634 * OOPS! we lost a race, the TCP session got reset after 635 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a 636 * network interrupt in the non-splnet() section of sosend(). 637 */ 638 if (m) 639 m_freem(m); 640 if (control) 641 m_freem(control); 642 error = ECONNRESET; /* XXX EPIPE? */ 643 tp = NULL; 644 TCPDEBUG1(); 645 goto out; 646 } 647 #ifdef INET6 648 isipv6 = nam && nam->sa_family == AF_INET6; 649 #endif /* INET6 */ 650 tp = intotcpcb(inp); 651 TCPDEBUG1(); 652 if (control) { 653 /* TCP doesn't do control messages (rights, creds, etc) */ 654 if (control->m_len) { 655 m_freem(control); 656 if (m) 657 m_freem(m); 658 error = EINVAL; 659 goto out; 660 } 661 m_freem(control); /* empty control, just free it */ 662 } 663 if(!(flags & PRUS_OOB)) { 664 sbappend(&so->so_snd, m); 665 if (nam && tp->t_state < TCPS_SYN_SENT) { 666 /* 667 * Do implied connect if not yet connected, 668 * initialize window to default value, and 669 * initialize maxseg/maxopd using peer's cached 670 * MSS. 671 */ 672 #ifdef INET6 673 if (isipv6) 674 error = tcp6_connect(tp, nam, td); 675 else 676 #endif /* INET6 */ 677 error = tcp_connect(tp, nam, td); 678 if (error) 679 goto out; 680 tp->snd_wnd = TTCP_CLIENT_SND_WND; 681 tcp_mss(tp, -1); 682 } 683 684 if (flags & PRUS_EOF) { 685 /* 686 * Close the send side of the connection after 687 * the data is sent. 688 */ 689 socantsendmore(so); 690 tp = tcp_usrclosed(tp); 691 } 692 if (tp != NULL) { 693 if (flags & PRUS_MORETOCOME) 694 tp->t_flags |= TF_MORETOCOME; 695 error = tcp_output(tp); 696 if (flags & PRUS_MORETOCOME) 697 tp->t_flags &= ~TF_MORETOCOME; 698 } 699 } else { 700 if (sbspace(&so->so_snd) < -512) { 701 m_freem(m); 702 error = ENOBUFS; 703 goto out; 704 } 705 /* 706 * According to RFC961 (Assigned Protocols), 707 * the urgent pointer points to the last octet 708 * of urgent data. We continue, however, 709 * to consider it to indicate the first octet 710 * of data past the urgent section. 711 * Otherwise, snd_up should be one lower. 712 */ 713 sbappend(&so->so_snd, m); 714 if (nam && tp->t_state < TCPS_SYN_SENT) { 715 /* 716 * Do implied connect if not yet connected, 717 * initialize window to default value, and 718 * initialize maxseg/maxopd using peer's cached 719 * MSS. 720 */ 721 #ifdef INET6 722 if (isipv6) 723 error = tcp6_connect(tp, nam, td); 724 else 725 #endif /* INET6 */ 726 error = tcp_connect(tp, nam, td); 727 if (error) 728 goto out; 729 tp->snd_wnd = TTCP_CLIENT_SND_WND; 730 tcp_mss(tp, -1); 731 } 732 tp->snd_up = tp->snd_una + so->so_snd.sb_cc; 733 tp->t_flags |= TF_FORCE; 734 error = tcp_output(tp); 735 tp->t_flags &= ~TF_FORCE; 736 } 737 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB : 738 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); 739 } 740 741 /* 742 * Abort the TCP. 743 */ 744 static int 745 tcp_usr_abort(struct socket *so) 746 { 747 int s = splnet(); 748 int error = 0; 749 struct inpcb *inp = sotoinpcb(so); 750 struct tcpcb *tp; 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 */ 760 static int 761 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) 762 { 763 int s = splnet(); 764 int error = 0; 765 struct inpcb *inp = sotoinpcb(so); 766 struct tcpcb *tp; 767 768 COMMON_START(); 769 if ((so->so_oobmark == 0 && 770 (so->so_state & SS_RCVATMARK) == 0) || 771 so->so_options & SO_OOBINLINE || 772 tp->t_oobflags & TCPOOB_HADDATA) { 773 error = EINVAL; 774 goto out; 775 } 776 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 777 error = EWOULDBLOCK; 778 goto out; 779 } 780 m->m_len = 1; 781 *mtod(m, caddr_t) = tp->t_iobc; 782 if ((flags & MSG_PEEK) == 0) 783 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 784 COMMON_END(PRU_RCVOOB); 785 } 786 787 /* xxx - should be const */ 788 struct pr_usrreqs tcp_usrreqs = { 789 tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind, 790 tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach, 791 tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd, 792 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, 793 in_setsockaddr, sosend, soreceive, sopoll 794 }; 795 796 #ifdef INET6 797 struct pr_usrreqs tcp6_usrreqs = { 798 tcp_usr_abort, tcp6_usr_accept, tcp_usr_attach, tcp6_usr_bind, 799 tcp6_usr_connect, pru_connect2_notsupp, in6_control, tcp_usr_detach, 800 tcp_usr_disconnect, tcp6_usr_listen, in6_mapped_peeraddr, tcp_usr_rcvd, 801 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, 802 in6_mapped_sockaddr, sosend, soreceive, sopoll 803 }; 804 #endif /* INET6 */ 805 806 static int 807 tcp_connect_oncpu(struct tcpcb *tp, struct sockaddr_in *sin, 808 struct sockaddr_in *if_sin) 809 { 810 struct inpcb *inp = tp->t_inpcb, *oinp; 811 struct socket *so = inp->inp_socket; 812 struct tcpcb *otp; 813 struct rmxp_tao *taop; 814 struct rmxp_tao tao_noncached; 815 816 oinp = in_pcblookup_hash(&tcbinfo[mycpu->gd_cpuid], 817 sin->sin_addr, sin->sin_port, 818 inp->inp_laddr.s_addr != INADDR_ANY ? 819 inp->inp_laddr : if_sin->sin_addr, 820 inp->inp_lport, 0, NULL); 821 if (oinp != NULL) { 822 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 823 otp->t_state == TCPS_TIME_WAIT && 824 (ticks - otp->t_starttime) < tcp_msl && 825 (otp->t_flags & TF_RCVD_CC)) 826 (void) tcp_close(otp); 827 else 828 return (EADDRINUSE); 829 } 830 if (inp->inp_laddr.s_addr == INADDR_ANY) 831 inp->inp_laddr = if_sin->sin_addr; 832 inp->inp_faddr = sin->sin_addr; 833 inp->inp_fport = sin->sin_port; 834 inp->inp_cpcbinfo = &tcbinfo[mycpu->gd_cpuid]; 835 in_pcbinsconnhash(inp); 836 837 /* Compute window scaling to request. */ 838 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 839 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 840 tp->request_r_scale++; 841 842 soisconnecting(so); 843 tcpstat.tcps_connattempt++; 844 tp->t_state = TCPS_SYN_SENT; 845 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 846 tp->iss = tcp_new_isn(tp); 847 tp->t_bw_rtseq = tp->iss; 848 tcp_sendseqinit(tp); 849 850 /* 851 * Generate a CC value for this connection and 852 * check whether CC or CCnew should be used. 853 */ 854 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) { 855 taop = &tao_noncached; 856 bzero(taop, sizeof(*taop)); 857 } 858 859 tp->cc_send = CC_INC(tcp_ccgen); 860 if (taop->tao_ccsent != 0 && 861 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 862 taop->tao_ccsent = tp->cc_send; 863 } else { 864 taop->tao_ccsent = 0; 865 tp->t_flags |= TF_SENDCCNEW; 866 } 867 868 return (0); 869 } 870 871 #ifdef SMP 872 873 struct netmsg_tcp_connect { 874 struct lwkt_msg nm_lmsg; 875 struct tcpcb *nm_tp; 876 struct sockaddr_in *nm_sin; 877 struct sockaddr_in *nm_ifsin; 878 }; 879 880 static int 881 tcp_connect_handler(lwkt_msg_t lmsg) 882 { 883 struct netmsg_tcp_connect *msg = (void *)lmsg; 884 int error; 885 886 error = tcp_connect_oncpu(msg->nm_tp, msg->nm_sin, msg->nm_ifsin); 887 lwkt_replymsg(lmsg, error); 888 return(EASYNC); 889 } 890 891 #endif 892 893 /* 894 * Common subroutine to open a TCP connection to remote host specified 895 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local 896 * port number if needed. Call in_pcbladdr to do the routing and to choose 897 * a local host address (interface). If there is an existing incarnation 898 * of the same connection in TIME-WAIT state and if the remote host was 899 * sending CC options and if the connection duration was < MSL, then 900 * truncate the previous TIME-WAIT state and proceed. 901 * Initialize connection parameters and enter SYN-SENT state. 902 */ 903 static int 904 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td) 905 { 906 struct inpcb *inp = tp->t_inpcb; 907 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 908 struct sockaddr_in *if_sin; 909 int error; 910 #ifdef SMP 911 lwkt_port_t port; 912 #endif 913 914 if (inp->inp_lport == 0) { 915 error = in_pcbbind(inp, (struct sockaddr *)NULL, td); 916 if (error) 917 return (error); 918 } 919 920 /* 921 * Cannot simply call in_pcbconnect, because there might be an 922 * earlier incarnation of this same connection still in 923 * TIME_WAIT state, creating an ADDRINUSE error. 924 */ 925 error = in_pcbladdr(inp, nam, &if_sin); 926 if (error) 927 return (error); 928 929 #ifdef SMP 930 port = tcp_addrport(sin->sin_addr.s_addr, sin->sin_port, 931 inp->inp_laddr.s_addr ? 932 inp->inp_laddr.s_addr : if_sin->sin_addr.s_addr, 933 inp->inp_lport); 934 935 if (port->mp_td != curthread) { 936 struct netmsg_tcp_connect msg; 937 938 lwkt_initmsg(&msg.nm_lmsg, &curthread->td_msgport, 0, 939 lwkt_cmd_func(tcp_connect_handler), lwkt_cmd_op_none); 940 msg.nm_tp = tp; 941 msg.nm_sin = sin; 942 msg.nm_ifsin = if_sin; 943 error = lwkt_domsg(port, &msg.nm_lmsg); 944 } else 945 #endif 946 error = tcp_connect_oncpu(tp, sin, if_sin); 947 948 return (error); 949 } 950 951 #ifdef INET6 952 static int 953 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td) 954 { 955 struct inpcb *inp = tp->t_inpcb, *oinp; 956 struct socket *so = inp->inp_socket; 957 struct tcpcb *otp; 958 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; 959 struct in6_addr *addr6; 960 struct rmxp_tao *taop; 961 struct rmxp_tao tao_noncached; 962 int error; 963 964 if (inp->inp_lport == 0) { 965 error = in6_pcbbind(inp, (struct sockaddr *)0, td); 966 if (error) 967 return error; 968 } 969 970 /* 971 * Cannot simply call in_pcbconnect, because there might be an 972 * earlier incarnation of this same connection still in 973 * TIME_WAIT state, creating an ADDRINUSE error. 974 */ 975 error = in6_pcbladdr(inp, nam, &addr6); 976 if (error) 977 return error; 978 oinp = in6_pcblookup_hash(inp->inp_cpcbinfo, 979 &sin6->sin6_addr, sin6->sin6_port, 980 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ? 981 addr6 : &inp->in6p_laddr, 982 inp->inp_lport, 0, NULL); 983 if (oinp) { 984 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 985 otp->t_state == TCPS_TIME_WAIT && 986 (ticks - otp->t_starttime) < tcp_msl && 987 (otp->t_flags & TF_RCVD_CC)) 988 otp = tcp_close(otp); 989 else 990 return (EADDRINUSE); 991 } 992 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) 993 inp->in6p_laddr = *addr6; 994 inp->in6p_faddr = sin6->sin6_addr; 995 inp->inp_fport = sin6->sin6_port; 996 if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != NULL) 997 inp->in6p_flowinfo = sin6->sin6_flowinfo; 998 in_pcbinsconnhash(inp); 999 1000 /* Compute window scaling to request. */ 1001 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 1002 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 1003 tp->request_r_scale++; 1004 1005 soisconnecting(so); 1006 tcpstat.tcps_connattempt++; 1007 tp->t_state = TCPS_SYN_SENT; 1008 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 1009 tp->iss = tcp_new_isn(tp); 1010 tp->t_bw_rtseq = tp->iss; 1011 tcp_sendseqinit(tp); 1012 1013 /* 1014 * Generate a CC value for this connection and 1015 * check whether CC or CCnew should be used. 1016 */ 1017 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) { 1018 taop = &tao_noncached; 1019 bzero(taop, sizeof(*taop)); 1020 } 1021 1022 tp->cc_send = CC_INC(tcp_ccgen); 1023 if (taop->tao_ccsent != 0 && 1024 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 1025 taop->tao_ccsent = tp->cc_send; 1026 } else { 1027 taop->tao_ccsent = 0; 1028 tp->t_flags |= TF_SENDCCNEW; 1029 } 1030 1031 return (0); 1032 } 1033 #endif /* INET6 */ 1034 1035 /* 1036 * The new sockopt interface makes it possible for us to block in the 1037 * copyin/out step (if we take a page fault). Taking a page fault at 1038 * splnet() is probably a Bad Thing. (Since sockets and pcbs both now 1039 * use TSM, there probably isn't any need for this function to run at 1040 * splnet() any more. This needs more examination.) 1041 */ 1042 int 1043 tcp_ctloutput(so, sopt) 1044 struct socket *so; 1045 struct sockopt *sopt; 1046 { 1047 int error, opt, optval, s; 1048 struct inpcb *inp; 1049 struct tcpcb *tp; 1050 1051 error = 0; 1052 s = splnet(); /* XXX */ 1053 inp = sotoinpcb(so); 1054 if (inp == NULL) { 1055 splx(s); 1056 return (ECONNRESET); 1057 } 1058 if (sopt->sopt_level != IPPROTO_TCP) { 1059 #ifdef INET6 1060 if (INP_CHECK_SOCKAF(so, AF_INET6)) 1061 error = ip6_ctloutput(so, sopt); 1062 else 1063 #endif /* INET6 */ 1064 error = ip_ctloutput(so, sopt); 1065 splx(s); 1066 return (error); 1067 } 1068 tp = intotcpcb(inp); 1069 1070 switch (sopt->sopt_dir) { 1071 case SOPT_SET: 1072 switch (sopt->sopt_name) { 1073 case TCP_NODELAY: 1074 case TCP_NOOPT: 1075 error = sooptcopyin(sopt, &optval, sizeof optval, 1076 sizeof optval); 1077 if (error) 1078 break; 1079 1080 switch (sopt->sopt_name) { 1081 case TCP_NODELAY: 1082 opt = TF_NODELAY; 1083 break; 1084 case TCP_NOOPT: 1085 opt = TF_NOOPT; 1086 break; 1087 default: 1088 opt = 0; /* dead code to fool gcc */ 1089 break; 1090 } 1091 1092 if (optval) 1093 tp->t_flags |= opt; 1094 else 1095 tp->t_flags &= ~opt; 1096 break; 1097 1098 case TCP_NOPUSH: 1099 error = sooptcopyin(sopt, &optval, sizeof optval, 1100 sizeof optval); 1101 if (error) 1102 break; 1103 1104 if (optval) 1105 tp->t_flags |= TF_NOPUSH; 1106 else { 1107 tp->t_flags &= ~TF_NOPUSH; 1108 error = tcp_output(tp); 1109 } 1110 break; 1111 1112 case TCP_MAXSEG: 1113 error = sooptcopyin(sopt, &optval, sizeof optval, 1114 sizeof optval); 1115 if (error) 1116 break; 1117 1118 if (optval > 0 && optval <= tp->t_maxseg) 1119 tp->t_maxseg = optval; 1120 else 1121 error = EINVAL; 1122 break; 1123 1124 default: 1125 error = ENOPROTOOPT; 1126 break; 1127 } 1128 break; 1129 1130 case SOPT_GET: 1131 switch (sopt->sopt_name) { 1132 case TCP_NODELAY: 1133 optval = tp->t_flags & TF_NODELAY; 1134 break; 1135 case TCP_MAXSEG: 1136 optval = tp->t_maxseg; 1137 break; 1138 case TCP_NOOPT: 1139 optval = tp->t_flags & TF_NOOPT; 1140 break; 1141 case TCP_NOPUSH: 1142 optval = tp->t_flags & TF_NOPUSH; 1143 break; 1144 default: 1145 error = ENOPROTOOPT; 1146 break; 1147 } 1148 if (error == 0) 1149 error = sooptcopyout(sopt, &optval, sizeof optval); 1150 break; 1151 } 1152 splx(s); 1153 return (error); 1154 } 1155 1156 /* 1157 * tcp_sendspace and tcp_recvspace are the default send and receive window 1158 * sizes, respectively. These are obsolescent (this information should 1159 * be set by the route). 1160 */ 1161 u_long tcp_sendspace = 1024*32; 1162 SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW, 1163 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size"); 1164 u_long tcp_recvspace = 57344; /* largest multiple of PAGE_SIZE < 64k */ 1165 SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW, 1166 &tcp_recvspace , 0, "Maximum incoming TCP datagram size"); 1167 1168 /* 1169 * Attach TCP protocol to socket, allocating 1170 * internet protocol control block, tcp control block, 1171 * bufer space, and entering LISTEN state if to accept connections. 1172 */ 1173 static int 1174 tcp_attach(struct socket *so, struct pru_attach_info *ai) 1175 { 1176 struct tcpcb *tp; 1177 struct inpcb *inp; 1178 int error; 1179 int cpu; 1180 #ifdef INET6 1181 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != NULL; 1182 #endif 1183 1184 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 1185 error = soreserve(so, tcp_sendspace, tcp_recvspace, 1186 ai->sb_rlimit); 1187 if (error) 1188 return (error); 1189 } 1190 cpu = mycpu->gd_cpuid; 1191 error = in_pcballoc(so, &tcbinfo[cpu]); 1192 if (error) 1193 return (error); 1194 inp = sotoinpcb(so); 1195 #ifdef INET6 1196 if (isipv6) { 1197 inp->inp_vflag |= INP_IPV6; 1198 inp->in6p_hops = -1; /* use kernel default */ 1199 } 1200 else 1201 #endif 1202 inp->inp_vflag |= INP_IPV4; 1203 tp = tcp_newtcpcb(inp); 1204 if (tp == 0) { 1205 int nofd = so->so_state & SS_NOFDREF; /* XXX */ 1206 1207 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ 1208 #ifdef INET6 1209 if (isipv6) 1210 in6_pcbdetach(inp); 1211 else 1212 #endif 1213 in_pcbdetach(inp); 1214 so->so_state |= nofd; 1215 return (ENOBUFS); 1216 } 1217 tp->t_state = TCPS_CLOSED; 1218 return (0); 1219 } 1220 1221 /* 1222 * Initiate (or continue) disconnect. 1223 * If embryonic state, just send reset (once). 1224 * If in ``let data drain'' option and linger null, just drop. 1225 * Otherwise (hard), mark socket disconnecting and drop 1226 * current input data; switch states based on user close, and 1227 * send segment to peer (with FIN). 1228 */ 1229 static struct tcpcb * 1230 tcp_disconnect(tp) 1231 struct tcpcb *tp; 1232 { 1233 struct socket *so = tp->t_inpcb->inp_socket; 1234 1235 if (tp->t_state < TCPS_ESTABLISHED) 1236 tp = tcp_close(tp); 1237 else if ((so->so_options & SO_LINGER) && so->so_linger == 0) 1238 tp = tcp_drop(tp, 0); 1239 else { 1240 soisdisconnecting(so); 1241 sbflush(&so->so_rcv); 1242 tp = tcp_usrclosed(tp); 1243 if (tp) 1244 (void) tcp_output(tp); 1245 } 1246 return (tp); 1247 } 1248 1249 /* 1250 * User issued close, and wish to trail through shutdown states: 1251 * if never received SYN, just forget it. If got a SYN from peer, 1252 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 1253 * If already got a FIN from peer, then almost done; go to LAST_ACK 1254 * state. In all other cases, have already sent FIN to peer (e.g. 1255 * after PRU_SHUTDOWN), and just have to play tedious game waiting 1256 * for peer to send FIN or not respond to keep-alives, etc. 1257 * We can let the user exit from the close as soon as the FIN is acked. 1258 */ 1259 static struct tcpcb * 1260 tcp_usrclosed(tp) 1261 struct tcpcb *tp; 1262 { 1263 1264 switch (tp->t_state) { 1265 1266 case TCPS_CLOSED: 1267 case TCPS_LISTEN: 1268 tp->t_state = TCPS_CLOSED; 1269 tp = tcp_close(tp); 1270 break; 1271 1272 case TCPS_SYN_SENT: 1273 case TCPS_SYN_RECEIVED: 1274 tp->t_flags |= TF_NEEDFIN; 1275 break; 1276 1277 case TCPS_ESTABLISHED: 1278 tp->t_state = TCPS_FIN_WAIT_1; 1279 break; 1280 1281 case TCPS_CLOSE_WAIT: 1282 tp->t_state = TCPS_LAST_ACK; 1283 break; 1284 } 1285 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { 1286 soisdisconnected(tp->t_inpcb->inp_socket); 1287 /* To prevent the connection hanging in FIN_WAIT_2 forever. */ 1288 if (tp->t_state == TCPS_FIN_WAIT_2) 1289 callout_reset(tp->tt_2msl, tcp_maxidle, 1290 tcp_timer_2msl, tp); 1291 } 1292 return (tp); 1293 } 1294