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 1.39 2006/12/29 18:02:56 victor 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/thread2.h> 109 #include <sys/msgport2.h> 110 111 #include <net/if.h> 112 #include <net/netisr.h> 113 #include <net/route.h> 114 115 #include <netinet/in.h> 116 #include <netinet/in_systm.h> 117 #ifdef INET6 118 #include <netinet/ip6.h> 119 #endif 120 #include <netinet/in_pcb.h> 121 #ifdef INET6 122 #include <netinet6/in6_pcb.h> 123 #endif 124 #include <netinet/in_var.h> 125 #include <netinet/ip_var.h> 126 #ifdef INET6 127 #include <netinet6/ip6_var.h> 128 #endif 129 #include <netinet/tcp.h> 130 #include <netinet/tcp_fsm.h> 131 #include <netinet/tcp_seq.h> 132 #include <netinet/tcp_timer.h> 133 #include <netinet/tcp_var.h> 134 #include <netinet/tcpip.h> 135 #ifdef TCPDEBUG 136 #include <netinet/tcp_debug.h> 137 #endif 138 139 #ifdef IPSEC 140 #include <netinet6/ipsec.h> 141 #endif /*IPSEC*/ 142 143 /* 144 * TCP protocol interface to socket abstraction. 145 */ 146 extern char *tcpstates[]; /* XXX ??? */ 147 148 static int tcp_attach (struct socket *, struct pru_attach_info *); 149 static int tcp_connect (struct tcpcb *, struct sockaddr *, 150 struct thread *); 151 #ifdef INET6 152 static int tcp6_connect (struct tcpcb *, struct sockaddr *, 153 struct thread *); 154 #endif /* INET6 */ 155 static struct tcpcb * 156 tcp_disconnect (struct tcpcb *); 157 static struct tcpcb * 158 tcp_usrclosed (struct tcpcb *); 159 160 #ifdef TCPDEBUG 161 #define TCPDEBUG0 int ostate = 0 162 #define TCPDEBUG1() ostate = tp ? tp->t_state : 0 163 #define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \ 164 tcp_trace(TA_USER, ostate, tp, 0, 0, req) 165 #else 166 #define TCPDEBUG0 167 #define TCPDEBUG1() 168 #define TCPDEBUG2(req) 169 #endif 170 171 /* 172 * TCP attaches to socket via pru_attach(), reserving space, 173 * and an internet control block. 174 */ 175 static int 176 tcp_usr_attach(struct socket *so, int proto, struct pru_attach_info *ai) 177 { 178 int error; 179 struct inpcb *inp; 180 struct tcpcb *tp = 0; 181 TCPDEBUG0; 182 183 crit_enter(); 184 inp = so->so_pcb; 185 TCPDEBUG1(); 186 if (inp) { 187 error = EISCONN; 188 goto out; 189 } 190 191 error = tcp_attach(so, ai); 192 if (error) 193 goto out; 194 195 if ((so->so_options & SO_LINGER) && so->so_linger == 0) 196 so->so_linger = TCP_LINGERTIME; 197 tp = sototcpcb(so); 198 out: 199 TCPDEBUG2(PRU_ATTACH); 200 crit_exit(); 201 return error; 202 } 203 204 /* 205 * pru_detach() detaches the TCP protocol from the socket. 206 * If the protocol state is non-embryonic, then can't 207 * do this directly: have to initiate a pru_disconnect(), 208 * which may finish later; embryonic TCB's can just 209 * be discarded here. 210 */ 211 static int 212 tcp_usr_detach(struct socket *so) 213 { 214 int error = 0; 215 struct inpcb *inp; 216 struct tcpcb *tp; 217 TCPDEBUG0; 218 219 crit_enter(); 220 inp = so->so_pcb; 221 if (inp == NULL) { 222 crit_exit(); 223 return EINVAL; /* XXX */ 224 } 225 226 /* 227 * It's possible for the tcpcb (tp) to disconnect from the inp due 228 * to tcp_drop()->tcp_close() being called. This may occur *after* 229 * the detach message has been queued so we may find a NULL tp here. 230 */ 231 if ((tp = intotcpcb(inp)) != NULL) { 232 TCPDEBUG1(); 233 tp = tcp_disconnect(tp); 234 TCPDEBUG2(PRU_DETACH); 235 } 236 crit_exit(); 237 return error; 238 } 239 240 #define COMMON_START(so, inp) \ 241 TCPDEBUG0; \ 242 \ 243 crit_enter(); \ 244 inp = so->so_pcb; \ 245 do { \ 246 if (inp == 0) { \ 247 crit_exit(); \ 248 return EINVAL; \ 249 } \ 250 tp = intotcpcb(inp); \ 251 TCPDEBUG1(); \ 252 } while(0) 253 254 #define COMMON_END(req) out: TCPDEBUG2(req); crit_exit(); return error; goto out 255 256 257 /* 258 * Give the socket an address. 259 */ 260 static int 261 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 262 { 263 int error = 0; 264 struct inpcb *inp; 265 struct tcpcb *tp; 266 struct sockaddr_in *sinp; 267 268 COMMON_START(so, inp); 269 270 /* 271 * Must check for multicast addresses and disallow binding 272 * to them. 273 */ 274 sinp = (struct sockaddr_in *)nam; 275 if (sinp->sin_family == AF_INET && 276 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 277 error = EAFNOSUPPORT; 278 goto out; 279 } 280 error = in_pcbbind(inp, nam, td); 281 if (error) 282 goto out; 283 COMMON_END(PRU_BIND); 284 285 } 286 287 #ifdef INET6 288 static int 289 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 290 { 291 int error = 0; 292 struct inpcb *inp; 293 struct tcpcb *tp; 294 struct sockaddr_in6 *sin6p; 295 296 COMMON_START(so, inp); 297 298 /* 299 * Must check for multicast addresses and disallow binding 300 * to them. 301 */ 302 sin6p = (struct sockaddr_in6 *)nam; 303 if (sin6p->sin6_family == AF_INET6 && 304 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { 305 error = EAFNOSUPPORT; 306 goto out; 307 } 308 inp->inp_vflag &= ~INP_IPV4; 309 inp->inp_vflag |= INP_IPV6; 310 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { 311 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr)) 312 inp->inp_vflag |= INP_IPV4; 313 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { 314 struct sockaddr_in sin; 315 316 in6_sin6_2_sin(&sin, sin6p); 317 inp->inp_vflag |= INP_IPV4; 318 inp->inp_vflag &= ~INP_IPV6; 319 error = in_pcbbind(inp, (struct sockaddr *)&sin, td); 320 goto out; 321 } 322 } 323 error = in6_pcbbind(inp, nam, td); 324 if (error) 325 goto out; 326 COMMON_END(PRU_BIND); 327 } 328 #endif /* INET6 */ 329 330 #ifdef SMP 331 struct netmsg_inswildcard { 332 struct lwkt_msg nm_lmsg; 333 struct inpcb *nm_inp; 334 struct inpcbinfo *nm_pcbinfo; 335 }; 336 337 static int 338 in_pcbinswildcardhash_handler(struct lwkt_msg *msg0) 339 { 340 struct netmsg_inswildcard *msg = (struct netmsg_inswildcard *)msg0; 341 342 in_pcbinswildcardhash_oncpu(msg->nm_inp, msg->nm_pcbinfo); 343 lwkt_replymsg(&msg->nm_lmsg, 0); 344 return (EASYNC); 345 } 346 #endif 347 348 /* 349 * Prepare to accept connections. 350 */ 351 static int 352 tcp_usr_listen(struct socket *so, struct thread *td) 353 { 354 int error = 0; 355 struct inpcb *inp; 356 struct tcpcb *tp; 357 #ifdef SMP 358 int cpu; 359 #endif 360 361 COMMON_START(so, inp); 362 if (inp->inp_lport == 0) { 363 error = in_pcbbind(inp, NULL, td); 364 if (error != 0) 365 goto out; 366 } 367 368 tp->t_state = TCPS_LISTEN; 369 #ifdef SMP 370 /* 371 * We have to set the flag because we can't have other cpus 372 * messing with our inp's flags. 373 */ 374 inp->inp_flags |= INP_WILDCARD_MP; 375 for (cpu = 0; cpu < ncpus2; cpu++) { 376 struct netmsg_inswildcard *msg; 377 378 if (cpu == mycpu->gd_cpuid) { 379 in_pcbinswildcardhash(inp); 380 continue; 381 } 382 383 msg = kmalloc(sizeof(struct netmsg_inswildcard), M_LWKTMSG, 384 M_INTWAIT); 385 lwkt_initmsg(&msg->nm_lmsg, &netisr_afree_rport, 0, 386 lwkt_cmd_func(in_pcbinswildcardhash_handler), 387 lwkt_cmd_op_none); 388 msg->nm_inp = inp; 389 msg->nm_pcbinfo = &tcbinfo[cpu]; 390 lwkt_sendmsg(tcp_cport(cpu), &msg->nm_lmsg); 391 } 392 #else 393 in_pcbinswildcardhash(inp); 394 #endif 395 COMMON_END(PRU_LISTEN); 396 } 397 398 #ifdef INET6 399 static int 400 tcp6_usr_listen(struct socket *so, struct thread *td) 401 { 402 int error = 0; 403 struct inpcb *inp; 404 struct tcpcb *tp; 405 #ifdef SMP 406 int cpu; 407 #endif 408 409 COMMON_START(so, inp); 410 if (inp->inp_lport == 0) { 411 if (!(inp->inp_flags & IN6P_IPV6_V6ONLY)) 412 inp->inp_vflag |= INP_IPV4; 413 else 414 inp->inp_vflag &= ~INP_IPV4; 415 error = in6_pcbbind(inp, (struct sockaddr *)0, td); 416 } 417 if (error == 0) 418 tp->t_state = TCPS_LISTEN; 419 #ifdef SMP 420 /* 421 * We have to set the flag because we can't have other cpus 422 * messing with our inp's flags. 423 */ 424 inp->inp_flags |= INP_WILDCARD_MP; 425 for (cpu = 0; cpu < ncpus2; cpu++) { 426 struct netmsg_inswildcard *msg; 427 428 if (cpu == mycpu->gd_cpuid) { 429 in_pcbinswildcardhash(inp); 430 continue; 431 } 432 433 msg = kmalloc(sizeof(struct netmsg_inswildcard), M_LWKTMSG, 434 M_INTWAIT); 435 lwkt_initmsg(&msg->nm_lmsg, &netisr_afree_rport, 0, 436 lwkt_cmd_func(in_pcbinswildcardhash_handler), 437 lwkt_cmd_op_none); 438 msg->nm_inp = inp; 439 msg->nm_pcbinfo = &tcbinfo[cpu]; 440 lwkt_sendmsg(tcp_cport(cpu), &msg->nm_lmsg); 441 } 442 #else 443 in_pcbinswildcardhash(inp); 444 #endif 445 COMMON_END(PRU_LISTEN); 446 } 447 #endif /* INET6 */ 448 449 /* 450 * Initiate connection to peer. 451 * Create a template for use in transmissions on this connection. 452 * Enter SYN_SENT state, and mark socket as connecting. 453 * Start keep-alive timer, and seed output sequence space. 454 * Send initial segment on connection. 455 */ 456 static int 457 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 458 { 459 int error = 0; 460 struct inpcb *inp; 461 struct tcpcb *tp; 462 struct sockaddr_in *sinp; 463 464 COMMON_START(so, inp); 465 466 /* 467 * Must disallow TCP ``connections'' to multicast addresses. 468 */ 469 sinp = (struct sockaddr_in *)nam; 470 if (sinp->sin_family == AF_INET 471 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 472 error = EAFNOSUPPORT; 473 goto out; 474 } 475 476 if (!prison_remote_ip(td, (struct sockaddr*)sinp)) { 477 error = EAFNOSUPPORT; /* IPv6 only jail */ 478 goto out; 479 } 480 481 if ((error = tcp_connect(tp, nam, td)) != 0) 482 goto out; 483 error = tcp_output(tp); 484 COMMON_END(PRU_CONNECT); 485 } 486 487 #ifdef INET6 488 static int 489 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 490 { 491 int error = 0; 492 struct inpcb *inp; 493 struct tcpcb *tp; 494 struct sockaddr_in6 *sin6p; 495 496 COMMON_START(so, inp); 497 498 /* 499 * Must disallow TCP ``connections'' to multicast addresses. 500 */ 501 sin6p = (struct sockaddr_in6 *)nam; 502 if (sin6p->sin6_family == AF_INET6 503 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { 504 error = EAFNOSUPPORT; 505 goto out; 506 } 507 508 if (!prison_remote_ip(td, nam)) { 509 error = EAFNOSUPPORT; /* IPv4 only jail */ 510 goto out; 511 } 512 513 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { 514 struct sockaddr_in sin; 515 516 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { 517 error = EINVAL; 518 goto out; 519 } 520 521 in6_sin6_2_sin(&sin, sin6p); 522 inp->inp_vflag |= INP_IPV4; 523 inp->inp_vflag &= ~INP_IPV6; 524 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0) 525 goto out; 526 error = tcp_output(tp); 527 goto out; 528 } 529 inp->inp_vflag &= ~INP_IPV4; 530 inp->inp_vflag |= INP_IPV6; 531 inp->inp_inc.inc_isipv6 = 1; 532 if ((error = tcp6_connect(tp, nam, td)) != 0) 533 goto out; 534 error = tcp_output(tp); 535 COMMON_END(PRU_CONNECT); 536 } 537 #endif /* INET6 */ 538 539 /* 540 * Initiate disconnect from peer. 541 * If connection never passed embryonic stage, just drop; 542 * else if don't need to let data drain, then can just drop anyways, 543 * else have to begin TCP shutdown process: mark socket disconnecting, 544 * drain unread data, state switch to reflect user close, and 545 * send segment (e.g. FIN) to peer. Socket will be really disconnected 546 * when peer sends FIN and acks ours. 547 * 548 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 549 */ 550 static int 551 tcp_usr_disconnect(struct socket *so) 552 { 553 int error = 0; 554 struct inpcb *inp; 555 struct tcpcb *tp; 556 557 COMMON_START(so, inp); 558 tp = tcp_disconnect(tp); 559 COMMON_END(PRU_DISCONNECT); 560 } 561 562 /* 563 * Accept a connection. Essentially all the work is 564 * done at higher levels; just return the address 565 * of the peer, storing through addr. 566 */ 567 static int 568 tcp_usr_accept(struct socket *so, struct sockaddr **nam) 569 { 570 int error = 0; 571 struct inpcb *inp; 572 struct tcpcb *tp = NULL; 573 TCPDEBUG0; 574 575 crit_enter(); 576 inp = so->so_pcb; 577 if (so->so_state & SS_ISDISCONNECTED) { 578 error = ECONNABORTED; 579 goto out; 580 } 581 if (inp == 0) { 582 crit_exit(); 583 return (EINVAL); 584 } 585 tp = intotcpcb(inp); 586 TCPDEBUG1(); 587 in_setpeeraddr(so, nam); 588 COMMON_END(PRU_ACCEPT); 589 } 590 591 #ifdef INET6 592 static int 593 tcp6_usr_accept(struct socket *so, struct sockaddr **nam) 594 { 595 int error = 0; 596 struct inpcb *inp; 597 struct tcpcb *tp = NULL; 598 TCPDEBUG0; 599 600 crit_enter(); 601 inp = so->so_pcb; 602 603 if (so->so_state & SS_ISDISCONNECTED) { 604 error = ECONNABORTED; 605 goto out; 606 } 607 if (inp == 0) { 608 crit_exit(); 609 return (EINVAL); 610 } 611 tp = intotcpcb(inp); 612 TCPDEBUG1(); 613 in6_mapped_peeraddr(so, nam); 614 COMMON_END(PRU_ACCEPT); 615 } 616 #endif /* INET6 */ 617 /* 618 * Mark the connection as being incapable of further output. 619 */ 620 static int 621 tcp_usr_shutdown(struct socket *so) 622 { 623 int error = 0; 624 struct inpcb *inp; 625 struct tcpcb *tp; 626 627 COMMON_START(so, inp); 628 socantsendmore(so); 629 tp = tcp_usrclosed(tp); 630 if (tp) 631 error = tcp_output(tp); 632 COMMON_END(PRU_SHUTDOWN); 633 } 634 635 /* 636 * After a receive, possibly send window update to peer. 637 */ 638 static int 639 tcp_usr_rcvd(struct socket *so, int flags) 640 { 641 int error = 0; 642 struct inpcb *inp; 643 struct tcpcb *tp; 644 645 COMMON_START(so, inp); 646 tcp_output(tp); 647 COMMON_END(PRU_RCVD); 648 } 649 650 /* 651 * Do a send by putting data in output queue and updating urgent 652 * marker if URG set. Possibly send more data. Unlike the other 653 * pru_*() routines, the mbuf chains are our responsibility. We 654 * must either enqueue them or free them. The other pru_* routines 655 * generally are caller-frees. 656 */ 657 static int 658 tcp_usr_send(struct socket *so, int flags, struct mbuf *m, 659 struct sockaddr *nam, struct mbuf *control, struct thread *td) 660 { 661 int error = 0; 662 struct inpcb *inp; 663 struct tcpcb *tp; 664 #ifdef INET6 665 int isipv6; 666 #endif 667 TCPDEBUG0; 668 669 crit_enter(); 670 inp = so->so_pcb; 671 672 if (inp == NULL) { 673 /* 674 * OOPS! we lost a race, the TCP session got reset after 675 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a 676 * network interrupt in the non-critical section of sosend(). 677 */ 678 if (m) 679 m_freem(m); 680 if (control) 681 m_freem(control); 682 error = ECONNRESET; /* XXX EPIPE? */ 683 tp = NULL; 684 TCPDEBUG1(); 685 goto out; 686 } 687 #ifdef INET6 688 isipv6 = nam && nam->sa_family == AF_INET6; 689 #endif /* INET6 */ 690 tp = intotcpcb(inp); 691 TCPDEBUG1(); 692 if (control) { 693 /* TCP doesn't do control messages (rights, creds, etc) */ 694 if (control->m_len) { 695 m_freem(control); 696 if (m) 697 m_freem(m); 698 error = EINVAL; 699 goto out; 700 } 701 m_freem(control); /* empty control, just free it */ 702 } 703 if(!(flags & PRUS_OOB)) { 704 sbappendstream(&so->so_snd, m); 705 if (nam && tp->t_state < TCPS_SYN_SENT) { 706 /* 707 * Do implied connect if not yet connected, 708 * initialize window to default value, and 709 * initialize maxseg/maxopd using peer's cached 710 * MSS. 711 */ 712 #ifdef INET6 713 if (isipv6) 714 error = tcp6_connect(tp, nam, td); 715 else 716 #endif /* INET6 */ 717 error = tcp_connect(tp, nam, td); 718 if (error) 719 goto out; 720 tp->snd_wnd = TTCP_CLIENT_SND_WND; 721 tcp_mss(tp, -1); 722 } 723 724 if (flags & PRUS_EOF) { 725 /* 726 * Close the send side of the connection after 727 * the data is sent. 728 */ 729 socantsendmore(so); 730 tp = tcp_usrclosed(tp); 731 } 732 if (tp != NULL) { 733 if (flags & PRUS_MORETOCOME) 734 tp->t_flags |= TF_MORETOCOME; 735 error = tcp_output(tp); 736 if (flags & PRUS_MORETOCOME) 737 tp->t_flags &= ~TF_MORETOCOME; 738 } 739 } else { 740 if (sbspace(&so->so_snd) < -512) { 741 m_freem(m); 742 error = ENOBUFS; 743 goto out; 744 } 745 /* 746 * According to RFC961 (Assigned Protocols), 747 * the urgent pointer points to the last octet 748 * of urgent data. We continue, however, 749 * to consider it to indicate the first octet 750 * of data past the urgent section. 751 * Otherwise, snd_up should be one lower. 752 */ 753 sbappendstream(&so->so_snd, m); 754 if (nam && tp->t_state < TCPS_SYN_SENT) { 755 /* 756 * Do implied connect if not yet connected, 757 * initialize window to default value, and 758 * initialize maxseg/maxopd using peer's cached 759 * MSS. 760 */ 761 #ifdef INET6 762 if (isipv6) 763 error = tcp6_connect(tp, nam, td); 764 else 765 #endif /* INET6 */ 766 error = tcp_connect(tp, nam, td); 767 if (error) 768 goto out; 769 tp->snd_wnd = TTCP_CLIENT_SND_WND; 770 tcp_mss(tp, -1); 771 } 772 tp->snd_up = tp->snd_una + so->so_snd.sb_cc; 773 tp->t_flags |= TF_FORCE; 774 error = tcp_output(tp); 775 tp->t_flags &= ~TF_FORCE; 776 } 777 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB : 778 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); 779 } 780 781 /* 782 * Abort the TCP. 783 */ 784 static int 785 tcp_usr_abort(struct socket *so) 786 { 787 int error = 0; 788 struct inpcb *inp; 789 struct tcpcb *tp; 790 791 COMMON_START(so, inp); 792 tp = tcp_drop(tp, ECONNABORTED); 793 COMMON_END(PRU_ABORT); 794 } 795 796 /* 797 * Receive out-of-band data. 798 */ 799 static int 800 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) 801 { 802 int error = 0; 803 struct inpcb *inp; 804 struct tcpcb *tp; 805 806 COMMON_START(so, inp); 807 if ((so->so_oobmark == 0 && 808 (so->so_state & SS_RCVATMARK) == 0) || 809 so->so_options & SO_OOBINLINE || 810 tp->t_oobflags & TCPOOB_HADDATA) { 811 error = EINVAL; 812 goto out; 813 } 814 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 815 error = EWOULDBLOCK; 816 goto out; 817 } 818 m->m_len = 1; 819 *mtod(m, caddr_t) = tp->t_iobc; 820 if ((flags & MSG_PEEK) == 0) 821 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 822 COMMON_END(PRU_RCVOOB); 823 } 824 825 /* xxx - should be const */ 826 struct pr_usrreqs tcp_usrreqs = { 827 tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind, 828 tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach, 829 tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd, 830 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, 831 in_setsockaddr, sosend, soreceive, sopoll 832 }; 833 834 #ifdef INET6 835 struct pr_usrreqs tcp6_usrreqs = { 836 tcp_usr_abort, tcp6_usr_accept, tcp_usr_attach, tcp6_usr_bind, 837 tcp6_usr_connect, pru_connect2_notsupp, in6_control, tcp_usr_detach, 838 tcp_usr_disconnect, tcp6_usr_listen, in6_mapped_peeraddr, tcp_usr_rcvd, 839 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, 840 in6_mapped_sockaddr, sosend, soreceive, sopoll 841 }; 842 #endif /* INET6 */ 843 844 static int 845 tcp_connect_oncpu(struct tcpcb *tp, struct sockaddr_in *sin, 846 struct sockaddr_in *if_sin) 847 { 848 struct inpcb *inp = tp->t_inpcb, *oinp; 849 struct socket *so = inp->inp_socket; 850 struct tcpcb *otp; 851 struct rmxp_tao *taop; 852 struct rmxp_tao tao_noncached; 853 854 oinp = in_pcblookup_hash(&tcbinfo[mycpu->gd_cpuid], 855 sin->sin_addr, sin->sin_port, 856 inp->inp_laddr.s_addr != INADDR_ANY ? 857 inp->inp_laddr : if_sin->sin_addr, 858 inp->inp_lport, 0, NULL); 859 if (oinp != NULL) { 860 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 861 otp->t_state == TCPS_TIME_WAIT && 862 (ticks - otp->t_starttime) < tcp_msl && 863 (otp->t_flags & TF_RCVD_CC)) 864 tcp_close(otp); 865 else 866 return (EADDRINUSE); 867 } 868 if (inp->inp_laddr.s_addr == INADDR_ANY) 869 inp->inp_laddr = if_sin->sin_addr; 870 inp->inp_faddr = sin->sin_addr; 871 inp->inp_fport = sin->sin_port; 872 inp->inp_cpcbinfo = &tcbinfo[mycpu->gd_cpuid]; 873 in_pcbinsconnhash(inp); 874 875 /* Compute window scaling to request. */ 876 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 877 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 878 tp->request_r_scale++; 879 880 soisconnecting(so); 881 tcpstat.tcps_connattempt++; 882 tp->t_state = TCPS_SYN_SENT; 883 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 884 tp->iss = tcp_new_isn(tp); 885 tcp_sendseqinit(tp); 886 887 /* 888 * Generate a CC value for this connection and 889 * check whether CC or CCnew should be used. 890 */ 891 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) { 892 taop = &tao_noncached; 893 bzero(taop, sizeof *taop); 894 } 895 896 tp->cc_send = CC_INC(tcp_ccgen); 897 if (taop->tao_ccsent != 0 && 898 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 899 taop->tao_ccsent = tp->cc_send; 900 } else { 901 taop->tao_ccsent = 0; 902 tp->t_flags |= TF_SENDCCNEW; 903 } 904 905 return (0); 906 } 907 908 #ifdef SMP 909 910 struct netmsg_tcp_connect { 911 struct lwkt_msg nm_lmsg; 912 struct tcpcb *nm_tp; 913 struct sockaddr_in *nm_sin; 914 struct sockaddr_in *nm_ifsin; 915 }; 916 917 static int 918 tcp_connect_handler(lwkt_msg_t lmsg) 919 { 920 struct netmsg_tcp_connect *msg = (void *)lmsg; 921 int error; 922 923 error = tcp_connect_oncpu(msg->nm_tp, msg->nm_sin, msg->nm_ifsin); 924 lwkt_replymsg(lmsg, error); 925 return(EASYNC); 926 } 927 928 #endif 929 930 /* 931 * Common subroutine to open a TCP connection to remote host specified 932 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local 933 * port number if needed. Call in_pcbladdr to do the routing and to choose 934 * a local host address (interface). If there is an existing incarnation 935 * of the same connection in TIME-WAIT state and if the remote host was 936 * sending CC options and if the connection duration was < MSL, then 937 * truncate the previous TIME-WAIT state and proceed. 938 * Initialize connection parameters and enter SYN-SENT state. 939 */ 940 static int 941 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td) 942 { 943 struct inpcb *inp = tp->t_inpcb; 944 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 945 struct sockaddr_in *if_sin; 946 int error; 947 #ifdef SMP 948 lwkt_port_t port; 949 #endif 950 951 if (inp->inp_lport == 0) { 952 error = in_pcbbind(inp, (struct sockaddr *)NULL, td); 953 if (error) 954 return (error); 955 } 956 957 /* 958 * Cannot simply call in_pcbconnect, because there might be an 959 * earlier incarnation of this same connection still in 960 * TIME_WAIT state, creating an ADDRINUSE error. 961 */ 962 error = in_pcbladdr(inp, nam, &if_sin, td); 963 if (error) 964 return (error); 965 966 #ifdef SMP 967 port = tcp_addrport(sin->sin_addr.s_addr, sin->sin_port, 968 inp->inp_laddr.s_addr ? 969 inp->inp_laddr.s_addr : if_sin->sin_addr.s_addr, 970 inp->inp_lport); 971 972 if (port->mp_td != curthread) { 973 struct netmsg_tcp_connect msg; 974 975 lwkt_initmsg(&msg.nm_lmsg, &curthread->td_msgport, 0, 976 lwkt_cmd_func(tcp_connect_handler), lwkt_cmd_op_none); 977 msg.nm_tp = tp; 978 msg.nm_sin = sin; 979 msg.nm_ifsin = if_sin; 980 error = lwkt_domsg(port, &msg.nm_lmsg); 981 } else 982 #endif 983 error = tcp_connect_oncpu(tp, sin, if_sin); 984 985 return (error); 986 } 987 988 #ifdef INET6 989 static int 990 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td) 991 { 992 struct inpcb *inp = tp->t_inpcb, *oinp; 993 struct socket *so = inp->inp_socket; 994 struct tcpcb *otp; 995 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; 996 struct in6_addr *addr6; 997 struct rmxp_tao *taop; 998 struct rmxp_tao tao_noncached; 999 int error; 1000 1001 if (inp->inp_lport == 0) { 1002 error = in6_pcbbind(inp, (struct sockaddr *)0, td); 1003 if (error) 1004 return error; 1005 } 1006 1007 /* 1008 * Cannot simply call in_pcbconnect, because there might be an 1009 * earlier incarnation of this same connection still in 1010 * TIME_WAIT state, creating an ADDRINUSE error. 1011 */ 1012 error = in6_pcbladdr(inp, nam, &addr6, td); 1013 if (error) 1014 return error; 1015 oinp = in6_pcblookup_hash(inp->inp_cpcbinfo, 1016 &sin6->sin6_addr, sin6->sin6_port, 1017 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ? 1018 addr6 : &inp->in6p_laddr, 1019 inp->inp_lport, 0, NULL); 1020 if (oinp) { 1021 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 1022 otp->t_state == TCPS_TIME_WAIT && 1023 (ticks - otp->t_starttime) < tcp_msl && 1024 (otp->t_flags & TF_RCVD_CC)) 1025 otp = tcp_close(otp); 1026 else 1027 return (EADDRINUSE); 1028 } 1029 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) 1030 inp->in6p_laddr = *addr6; 1031 inp->in6p_faddr = sin6->sin6_addr; 1032 inp->inp_fport = sin6->sin6_port; 1033 if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != NULL) 1034 inp->in6p_flowinfo = sin6->sin6_flowinfo; 1035 in_pcbinsconnhash(inp); 1036 1037 /* Compute window scaling to request. */ 1038 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 1039 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 1040 tp->request_r_scale++; 1041 1042 soisconnecting(so); 1043 tcpstat.tcps_connattempt++; 1044 tp->t_state = TCPS_SYN_SENT; 1045 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 1046 tp->iss = tcp_new_isn(tp); 1047 tcp_sendseqinit(tp); 1048 1049 /* 1050 * Generate a CC value for this connection and 1051 * check whether CC or CCnew should be used. 1052 */ 1053 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) { 1054 taop = &tao_noncached; 1055 bzero(taop, sizeof *taop); 1056 } 1057 1058 tp->cc_send = CC_INC(tcp_ccgen); 1059 if (taop->tao_ccsent != 0 && 1060 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 1061 taop->tao_ccsent = tp->cc_send; 1062 } else { 1063 taop->tao_ccsent = 0; 1064 tp->t_flags |= TF_SENDCCNEW; 1065 } 1066 1067 return (0); 1068 } 1069 #endif /* INET6 */ 1070 1071 /* 1072 * The new sockopt interface makes it possible for us to block in the 1073 * copyin/out step (if we take a page fault). Taking a page fault while 1074 * in a critical section is probably a Bad Thing. (Since sockets and pcbs 1075 * both now use TSM, there probably isn't any need for this function to 1076 * run in a critical section any more. This needs more examination.) 1077 */ 1078 int 1079 tcp_ctloutput(struct socket *so, struct sockopt *sopt) 1080 { 1081 int error, opt, optval; 1082 struct inpcb *inp; 1083 struct tcpcb *tp; 1084 1085 error = 0; 1086 crit_enter(); /* XXX */ 1087 inp = so->so_pcb; 1088 if (inp == NULL) { 1089 crit_exit(); 1090 return (ECONNRESET); 1091 } 1092 if (sopt->sopt_level != IPPROTO_TCP) { 1093 #ifdef INET6 1094 if (INP_CHECK_SOCKAF(so, AF_INET6)) 1095 error = ip6_ctloutput(so, sopt); 1096 else 1097 #endif /* INET6 */ 1098 error = ip_ctloutput(so, sopt); 1099 crit_exit(); 1100 return (error); 1101 } 1102 tp = intotcpcb(inp); 1103 1104 switch (sopt->sopt_dir) { 1105 case SOPT_SET: 1106 switch (sopt->sopt_name) { 1107 case TCP_NODELAY: 1108 case TCP_NOOPT: 1109 error = sooptcopyin(sopt, &optval, sizeof optval, 1110 sizeof optval); 1111 if (error) 1112 break; 1113 1114 switch (sopt->sopt_name) { 1115 case TCP_NODELAY: 1116 opt = TF_NODELAY; 1117 break; 1118 case TCP_NOOPT: 1119 opt = TF_NOOPT; 1120 break; 1121 default: 1122 opt = 0; /* dead code to fool gcc */ 1123 break; 1124 } 1125 1126 if (optval) 1127 tp->t_flags |= opt; 1128 else 1129 tp->t_flags &= ~opt; 1130 break; 1131 1132 case TCP_NOPUSH: 1133 error = sooptcopyin(sopt, &optval, sizeof optval, 1134 sizeof optval); 1135 if (error) 1136 break; 1137 1138 if (optval) 1139 tp->t_flags |= TF_NOPUSH; 1140 else { 1141 tp->t_flags &= ~TF_NOPUSH; 1142 error = tcp_output(tp); 1143 } 1144 break; 1145 1146 case TCP_MAXSEG: 1147 error = sooptcopyin(sopt, &optval, sizeof optval, 1148 sizeof optval); 1149 if (error) 1150 break; 1151 1152 if (optval > 0 && optval <= tp->t_maxseg) 1153 tp->t_maxseg = optval; 1154 else 1155 error = EINVAL; 1156 break; 1157 1158 default: 1159 error = ENOPROTOOPT; 1160 break; 1161 } 1162 break; 1163 1164 case SOPT_GET: 1165 switch (sopt->sopt_name) { 1166 case TCP_NODELAY: 1167 optval = tp->t_flags & TF_NODELAY; 1168 break; 1169 case TCP_MAXSEG: 1170 optval = tp->t_maxseg; 1171 break; 1172 case TCP_NOOPT: 1173 optval = tp->t_flags & TF_NOOPT; 1174 break; 1175 case TCP_NOPUSH: 1176 optval = tp->t_flags & TF_NOPUSH; 1177 break; 1178 default: 1179 error = ENOPROTOOPT; 1180 break; 1181 } 1182 if (error == 0) 1183 error = sooptcopyout(sopt, &optval, sizeof optval); 1184 break; 1185 } 1186 crit_exit(); 1187 return (error); 1188 } 1189 1190 /* 1191 * tcp_sendspace and tcp_recvspace are the default send and receive window 1192 * sizes, respectively. These are obsolescent (this information should 1193 * be set by the route). 1194 */ 1195 u_long tcp_sendspace = 1024*32; 1196 SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW, 1197 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size"); 1198 u_long tcp_recvspace = 57344; /* largest multiple of PAGE_SIZE < 64k */ 1199 SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW, 1200 &tcp_recvspace , 0, "Maximum incoming TCP datagram size"); 1201 1202 /* 1203 * Attach TCP protocol to socket, allocating 1204 * internet protocol control block, tcp control block, 1205 * bufer space, and entering LISTEN state if to accept connections. 1206 */ 1207 static int 1208 tcp_attach(struct socket *so, struct pru_attach_info *ai) 1209 { 1210 struct tcpcb *tp; 1211 struct inpcb *inp; 1212 int error; 1213 int cpu; 1214 #ifdef INET6 1215 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != NULL; 1216 #endif 1217 1218 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 1219 error = soreserve(so, tcp_sendspace, tcp_recvspace, 1220 ai->sb_rlimit); 1221 if (error) 1222 return (error); 1223 } 1224 cpu = mycpu->gd_cpuid; 1225 error = in_pcballoc(so, &tcbinfo[cpu]); 1226 if (error) 1227 return (error); 1228 inp = so->so_pcb; 1229 #ifdef INET6 1230 if (isipv6) { 1231 inp->inp_vflag |= INP_IPV6; 1232 inp->in6p_hops = -1; /* use kernel default */ 1233 } 1234 else 1235 #endif 1236 inp->inp_vflag |= INP_IPV4; 1237 tp = tcp_newtcpcb(inp); 1238 if (tp == 0) { 1239 int nofd = so->so_state & SS_NOFDREF; /* XXX */ 1240 1241 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ 1242 #ifdef INET6 1243 if (isipv6) 1244 in6_pcbdetach(inp); 1245 else 1246 #endif 1247 in_pcbdetach(inp); 1248 so->so_state |= nofd; 1249 return (ENOBUFS); 1250 } 1251 tp->t_state = TCPS_CLOSED; 1252 return (0); 1253 } 1254 1255 /* 1256 * Initiate (or continue) disconnect. 1257 * If embryonic state, just send reset (once). 1258 * If in ``let data drain'' option and linger null, just drop. 1259 * Otherwise (hard), mark socket disconnecting and drop 1260 * current input data; switch states based on user close, and 1261 * send segment to peer (with FIN). 1262 */ 1263 static struct tcpcb * 1264 tcp_disconnect(struct tcpcb *tp) 1265 { 1266 struct socket *so = tp->t_inpcb->inp_socket; 1267 1268 if (tp->t_state < TCPS_ESTABLISHED) 1269 tp = tcp_close(tp); 1270 else if ((so->so_options & SO_LINGER) && so->so_linger == 0) 1271 tp = tcp_drop(tp, 0); 1272 else { 1273 soisdisconnecting(so); 1274 sbflush(&so->so_rcv); 1275 tp = tcp_usrclosed(tp); 1276 if (tp) 1277 tcp_output(tp); 1278 } 1279 return (tp); 1280 } 1281 1282 /* 1283 * User issued close, and wish to trail through shutdown states: 1284 * if never received SYN, just forget it. If got a SYN from peer, 1285 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 1286 * If already got a FIN from peer, then almost done; go to LAST_ACK 1287 * state. In all other cases, have already sent FIN to peer (e.g. 1288 * after PRU_SHUTDOWN), and just have to play tedious game waiting 1289 * for peer to send FIN or not respond to keep-alives, etc. 1290 * We can let the user exit from the close as soon as the FIN is acked. 1291 */ 1292 static struct tcpcb * 1293 tcp_usrclosed(struct tcpcb *tp) 1294 { 1295 1296 switch (tp->t_state) { 1297 1298 case TCPS_CLOSED: 1299 case TCPS_LISTEN: 1300 tp->t_state = TCPS_CLOSED; 1301 tp = tcp_close(tp); 1302 break; 1303 1304 case TCPS_SYN_SENT: 1305 case TCPS_SYN_RECEIVED: 1306 tp->t_flags |= TF_NEEDFIN; 1307 break; 1308 1309 case TCPS_ESTABLISHED: 1310 tp->t_state = TCPS_FIN_WAIT_1; 1311 break; 1312 1313 case TCPS_CLOSE_WAIT: 1314 tp->t_state = TCPS_LAST_ACK; 1315 break; 1316 } 1317 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { 1318 soisdisconnected(tp->t_inpcb->inp_socket); 1319 /* To prevent the connection hanging in FIN_WAIT_2 forever. */ 1320 if (tp->t_state == TCPS_FIN_WAIT_2) 1321 callout_reset(tp->tt_2msl, tcp_maxidle, 1322 tcp_timer_2msl, tp); 1323 } 1324 return (tp); 1325 } 1326