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 1.38 2006/09/05 00:55:48 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/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 prison_remote_ip(td, 0, &sinp->sin_addr.s_addr); 477 478 if ((error = tcp_connect(tp, nam, td)) != 0) 479 goto out; 480 error = tcp_output(tp); 481 COMMON_END(PRU_CONNECT); 482 } 483 484 #ifdef INET6 485 static int 486 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 487 { 488 int error = 0; 489 struct inpcb *inp; 490 struct tcpcb *tp; 491 struct sockaddr_in6 *sin6p; 492 493 COMMON_START(so, inp); 494 495 /* 496 * Must disallow TCP ``connections'' to multicast addresses. 497 */ 498 sin6p = (struct sockaddr_in6 *)nam; 499 if (sin6p->sin6_family == AF_INET6 500 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { 501 error = EAFNOSUPPORT; 502 goto out; 503 } 504 505 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { 506 struct sockaddr_in sin; 507 508 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { 509 error = EINVAL; 510 goto out; 511 } 512 513 in6_sin6_2_sin(&sin, sin6p); 514 inp->inp_vflag |= INP_IPV4; 515 inp->inp_vflag &= ~INP_IPV6; 516 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0) 517 goto out; 518 error = tcp_output(tp); 519 goto out; 520 } 521 inp->inp_vflag &= ~INP_IPV4; 522 inp->inp_vflag |= INP_IPV6; 523 inp->inp_inc.inc_isipv6 = 1; 524 if ((error = tcp6_connect(tp, nam, td)) != 0) 525 goto out; 526 error = tcp_output(tp); 527 COMMON_END(PRU_CONNECT); 528 } 529 #endif /* INET6 */ 530 531 /* 532 * Initiate disconnect from peer. 533 * If connection never passed embryonic stage, just drop; 534 * else if don't need to let data drain, then can just drop anyways, 535 * else have to begin TCP shutdown process: mark socket disconnecting, 536 * drain unread data, state switch to reflect user close, and 537 * send segment (e.g. FIN) to peer. Socket will be really disconnected 538 * when peer sends FIN and acks ours. 539 * 540 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 541 */ 542 static int 543 tcp_usr_disconnect(struct socket *so) 544 { 545 int error = 0; 546 struct inpcb *inp; 547 struct tcpcb *tp; 548 549 COMMON_START(so, inp); 550 tp = tcp_disconnect(tp); 551 COMMON_END(PRU_DISCONNECT); 552 } 553 554 /* 555 * Accept a connection. Essentially all the work is 556 * done at higher levels; just return the address 557 * of the peer, storing through addr. 558 */ 559 static int 560 tcp_usr_accept(struct socket *so, struct sockaddr **nam) 561 { 562 int error = 0; 563 struct inpcb *inp; 564 struct tcpcb *tp = NULL; 565 TCPDEBUG0; 566 567 crit_enter(); 568 inp = so->so_pcb; 569 if (so->so_state & SS_ISDISCONNECTED) { 570 error = ECONNABORTED; 571 goto out; 572 } 573 if (inp == 0) { 574 crit_exit(); 575 return (EINVAL); 576 } 577 tp = intotcpcb(inp); 578 TCPDEBUG1(); 579 in_setpeeraddr(so, nam); 580 COMMON_END(PRU_ACCEPT); 581 } 582 583 #ifdef INET6 584 static int 585 tcp6_usr_accept(struct socket *so, struct sockaddr **nam) 586 { 587 int error = 0; 588 struct inpcb *inp; 589 struct tcpcb *tp = NULL; 590 TCPDEBUG0; 591 592 crit_enter(); 593 inp = so->so_pcb; 594 595 if (so->so_state & SS_ISDISCONNECTED) { 596 error = ECONNABORTED; 597 goto out; 598 } 599 if (inp == 0) { 600 crit_exit(); 601 return (EINVAL); 602 } 603 tp = intotcpcb(inp); 604 TCPDEBUG1(); 605 in6_mapped_peeraddr(so, nam); 606 COMMON_END(PRU_ACCEPT); 607 } 608 #endif /* INET6 */ 609 /* 610 * Mark the connection as being incapable of further output. 611 */ 612 static int 613 tcp_usr_shutdown(struct socket *so) 614 { 615 int error = 0; 616 struct inpcb *inp; 617 struct tcpcb *tp; 618 619 COMMON_START(so, inp); 620 socantsendmore(so); 621 tp = tcp_usrclosed(tp); 622 if (tp) 623 error = tcp_output(tp); 624 COMMON_END(PRU_SHUTDOWN); 625 } 626 627 /* 628 * After a receive, possibly send window update to peer. 629 */ 630 static int 631 tcp_usr_rcvd(struct socket *so, int flags) 632 { 633 int error = 0; 634 struct inpcb *inp; 635 struct tcpcb *tp; 636 637 COMMON_START(so, inp); 638 tcp_output(tp); 639 COMMON_END(PRU_RCVD); 640 } 641 642 /* 643 * Do a send by putting data in output queue and updating urgent 644 * marker if URG set. Possibly send more data. Unlike the other 645 * pru_*() routines, the mbuf chains are our responsibility. We 646 * must either enqueue them or free them. The other pru_* routines 647 * generally are caller-frees. 648 */ 649 static int 650 tcp_usr_send(struct socket *so, int flags, struct mbuf *m, 651 struct sockaddr *nam, struct mbuf *control, struct thread *td) 652 { 653 int error = 0; 654 struct inpcb *inp; 655 struct tcpcb *tp; 656 #ifdef INET6 657 int isipv6; 658 #endif 659 TCPDEBUG0; 660 661 crit_enter(); 662 inp = so->so_pcb; 663 664 if (inp == NULL) { 665 /* 666 * OOPS! we lost a race, the TCP session got reset after 667 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a 668 * network interrupt in the non-critical section of sosend(). 669 */ 670 if (m) 671 m_freem(m); 672 if (control) 673 m_freem(control); 674 error = ECONNRESET; /* XXX EPIPE? */ 675 tp = NULL; 676 TCPDEBUG1(); 677 goto out; 678 } 679 #ifdef INET6 680 isipv6 = nam && nam->sa_family == AF_INET6; 681 #endif /* INET6 */ 682 tp = intotcpcb(inp); 683 TCPDEBUG1(); 684 if (control) { 685 /* TCP doesn't do control messages (rights, creds, etc) */ 686 if (control->m_len) { 687 m_freem(control); 688 if (m) 689 m_freem(m); 690 error = EINVAL; 691 goto out; 692 } 693 m_freem(control); /* empty control, just free it */ 694 } 695 if(!(flags & PRUS_OOB)) { 696 sbappendstream(&so->so_snd, m); 697 if (nam && tp->t_state < TCPS_SYN_SENT) { 698 /* 699 * Do implied connect if not yet connected, 700 * initialize window to default value, and 701 * initialize maxseg/maxopd using peer's cached 702 * MSS. 703 */ 704 #ifdef INET6 705 if (isipv6) 706 error = tcp6_connect(tp, nam, td); 707 else 708 #endif /* INET6 */ 709 error = tcp_connect(tp, nam, td); 710 if (error) 711 goto out; 712 tp->snd_wnd = TTCP_CLIENT_SND_WND; 713 tcp_mss(tp, -1); 714 } 715 716 if (flags & PRUS_EOF) { 717 /* 718 * Close the send side of the connection after 719 * the data is sent. 720 */ 721 socantsendmore(so); 722 tp = tcp_usrclosed(tp); 723 } 724 if (tp != NULL) { 725 if (flags & PRUS_MORETOCOME) 726 tp->t_flags |= TF_MORETOCOME; 727 error = tcp_output(tp); 728 if (flags & PRUS_MORETOCOME) 729 tp->t_flags &= ~TF_MORETOCOME; 730 } 731 } else { 732 if (sbspace(&so->so_snd) < -512) { 733 m_freem(m); 734 error = ENOBUFS; 735 goto out; 736 } 737 /* 738 * According to RFC961 (Assigned Protocols), 739 * the urgent pointer points to the last octet 740 * of urgent data. We continue, however, 741 * to consider it to indicate the first octet 742 * of data past the urgent section. 743 * Otherwise, snd_up should be one lower. 744 */ 745 sbappendstream(&so->so_snd, m); 746 if (nam && tp->t_state < TCPS_SYN_SENT) { 747 /* 748 * Do implied connect if not yet connected, 749 * initialize window to default value, and 750 * initialize maxseg/maxopd using peer's cached 751 * MSS. 752 */ 753 #ifdef INET6 754 if (isipv6) 755 error = tcp6_connect(tp, nam, td); 756 else 757 #endif /* INET6 */ 758 error = tcp_connect(tp, nam, td); 759 if (error) 760 goto out; 761 tp->snd_wnd = TTCP_CLIENT_SND_WND; 762 tcp_mss(tp, -1); 763 } 764 tp->snd_up = tp->snd_una + so->so_snd.sb_cc; 765 tp->t_flags |= TF_FORCE; 766 error = tcp_output(tp); 767 tp->t_flags &= ~TF_FORCE; 768 } 769 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB : 770 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); 771 } 772 773 /* 774 * Abort the TCP. 775 */ 776 static int 777 tcp_usr_abort(struct socket *so) 778 { 779 int error = 0; 780 struct inpcb *inp; 781 struct tcpcb *tp; 782 783 COMMON_START(so, inp); 784 tp = tcp_drop(tp, ECONNABORTED); 785 COMMON_END(PRU_ABORT); 786 } 787 788 /* 789 * Receive out-of-band data. 790 */ 791 static int 792 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) 793 { 794 int error = 0; 795 struct inpcb *inp; 796 struct tcpcb *tp; 797 798 COMMON_START(so, inp); 799 if ((so->so_oobmark == 0 && 800 (so->so_state & SS_RCVATMARK) == 0) || 801 so->so_options & SO_OOBINLINE || 802 tp->t_oobflags & TCPOOB_HADDATA) { 803 error = EINVAL; 804 goto out; 805 } 806 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 807 error = EWOULDBLOCK; 808 goto out; 809 } 810 m->m_len = 1; 811 *mtod(m, caddr_t) = tp->t_iobc; 812 if ((flags & MSG_PEEK) == 0) 813 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 814 COMMON_END(PRU_RCVOOB); 815 } 816 817 /* xxx - should be const */ 818 struct pr_usrreqs tcp_usrreqs = { 819 tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind, 820 tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach, 821 tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd, 822 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, 823 in_setsockaddr, sosend, soreceive, sopoll 824 }; 825 826 #ifdef INET6 827 struct pr_usrreqs tcp6_usrreqs = { 828 tcp_usr_abort, tcp6_usr_accept, tcp_usr_attach, tcp6_usr_bind, 829 tcp6_usr_connect, pru_connect2_notsupp, in6_control, tcp_usr_detach, 830 tcp_usr_disconnect, tcp6_usr_listen, in6_mapped_peeraddr, tcp_usr_rcvd, 831 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, 832 in6_mapped_sockaddr, sosend, soreceive, sopoll 833 }; 834 #endif /* INET6 */ 835 836 static int 837 tcp_connect_oncpu(struct tcpcb *tp, struct sockaddr_in *sin, 838 struct sockaddr_in *if_sin) 839 { 840 struct inpcb *inp = tp->t_inpcb, *oinp; 841 struct socket *so = inp->inp_socket; 842 struct tcpcb *otp; 843 struct rmxp_tao *taop; 844 struct rmxp_tao tao_noncached; 845 846 oinp = in_pcblookup_hash(&tcbinfo[mycpu->gd_cpuid], 847 sin->sin_addr, sin->sin_port, 848 inp->inp_laddr.s_addr != INADDR_ANY ? 849 inp->inp_laddr : if_sin->sin_addr, 850 inp->inp_lport, 0, NULL); 851 if (oinp != NULL) { 852 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 853 otp->t_state == TCPS_TIME_WAIT && 854 (ticks - otp->t_starttime) < tcp_msl && 855 (otp->t_flags & TF_RCVD_CC)) 856 tcp_close(otp); 857 else 858 return (EADDRINUSE); 859 } 860 if (inp->inp_laddr.s_addr == INADDR_ANY) 861 inp->inp_laddr = if_sin->sin_addr; 862 inp->inp_faddr = sin->sin_addr; 863 inp->inp_fport = sin->sin_port; 864 inp->inp_cpcbinfo = &tcbinfo[mycpu->gd_cpuid]; 865 in_pcbinsconnhash(inp); 866 867 /* Compute window scaling to request. */ 868 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 869 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 870 tp->request_r_scale++; 871 872 soisconnecting(so); 873 tcpstat.tcps_connattempt++; 874 tp->t_state = TCPS_SYN_SENT; 875 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 876 tp->iss = tcp_new_isn(tp); 877 tcp_sendseqinit(tp); 878 879 /* 880 * Generate a CC value for this connection and 881 * check whether CC or CCnew should be used. 882 */ 883 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) { 884 taop = &tao_noncached; 885 bzero(taop, sizeof *taop); 886 } 887 888 tp->cc_send = CC_INC(tcp_ccgen); 889 if (taop->tao_ccsent != 0 && 890 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 891 taop->tao_ccsent = tp->cc_send; 892 } else { 893 taop->tao_ccsent = 0; 894 tp->t_flags |= TF_SENDCCNEW; 895 } 896 897 return (0); 898 } 899 900 #ifdef SMP 901 902 struct netmsg_tcp_connect { 903 struct lwkt_msg nm_lmsg; 904 struct tcpcb *nm_tp; 905 struct sockaddr_in *nm_sin; 906 struct sockaddr_in *nm_ifsin; 907 }; 908 909 static int 910 tcp_connect_handler(lwkt_msg_t lmsg) 911 { 912 struct netmsg_tcp_connect *msg = (void *)lmsg; 913 int error; 914 915 error = tcp_connect_oncpu(msg->nm_tp, msg->nm_sin, msg->nm_ifsin); 916 lwkt_replymsg(lmsg, error); 917 return(EASYNC); 918 } 919 920 #endif 921 922 /* 923 * Common subroutine to open a TCP connection to remote host specified 924 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local 925 * port number if needed. Call in_pcbladdr to do the routing and to choose 926 * a local host address (interface). If there is an existing incarnation 927 * of the same connection in TIME-WAIT state and if the remote host was 928 * sending CC options and if the connection duration was < MSL, then 929 * truncate the previous TIME-WAIT state and proceed. 930 * Initialize connection parameters and enter SYN-SENT state. 931 */ 932 static int 933 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td) 934 { 935 struct inpcb *inp = tp->t_inpcb; 936 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 937 struct sockaddr_in *if_sin; 938 int error; 939 #ifdef SMP 940 lwkt_port_t port; 941 #endif 942 943 if (inp->inp_lport == 0) { 944 error = in_pcbbind(inp, (struct sockaddr *)NULL, td); 945 if (error) 946 return (error); 947 } 948 949 /* 950 * Cannot simply call in_pcbconnect, because there might be an 951 * earlier incarnation of this same connection still in 952 * TIME_WAIT state, creating an ADDRINUSE error. 953 */ 954 error = in_pcbladdr(inp, nam, &if_sin); 955 if (error) 956 return (error); 957 958 #ifdef SMP 959 port = tcp_addrport(sin->sin_addr.s_addr, sin->sin_port, 960 inp->inp_laddr.s_addr ? 961 inp->inp_laddr.s_addr : if_sin->sin_addr.s_addr, 962 inp->inp_lport); 963 964 if (port->mp_td != curthread) { 965 struct netmsg_tcp_connect msg; 966 967 lwkt_initmsg(&msg.nm_lmsg, &curthread->td_msgport, 0, 968 lwkt_cmd_func(tcp_connect_handler), lwkt_cmd_op_none); 969 msg.nm_tp = tp; 970 msg.nm_sin = sin; 971 msg.nm_ifsin = if_sin; 972 error = lwkt_domsg(port, &msg.nm_lmsg); 973 } else 974 #endif 975 error = tcp_connect_oncpu(tp, sin, if_sin); 976 977 return (error); 978 } 979 980 #ifdef INET6 981 static int 982 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td) 983 { 984 struct inpcb *inp = tp->t_inpcb, *oinp; 985 struct socket *so = inp->inp_socket; 986 struct tcpcb *otp; 987 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; 988 struct in6_addr *addr6; 989 struct rmxp_tao *taop; 990 struct rmxp_tao tao_noncached; 991 int error; 992 993 if (inp->inp_lport == 0) { 994 error = in6_pcbbind(inp, (struct sockaddr *)0, td); 995 if (error) 996 return error; 997 } 998 999 /* 1000 * Cannot simply call in_pcbconnect, because there might be an 1001 * earlier incarnation of this same connection still in 1002 * TIME_WAIT state, creating an ADDRINUSE error. 1003 */ 1004 error = in6_pcbladdr(inp, nam, &addr6); 1005 if (error) 1006 return error; 1007 oinp = in6_pcblookup_hash(inp->inp_cpcbinfo, 1008 &sin6->sin6_addr, sin6->sin6_port, 1009 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ? 1010 addr6 : &inp->in6p_laddr, 1011 inp->inp_lport, 0, NULL); 1012 if (oinp) { 1013 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 1014 otp->t_state == TCPS_TIME_WAIT && 1015 (ticks - otp->t_starttime) < tcp_msl && 1016 (otp->t_flags & TF_RCVD_CC)) 1017 otp = tcp_close(otp); 1018 else 1019 return (EADDRINUSE); 1020 } 1021 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) 1022 inp->in6p_laddr = *addr6; 1023 inp->in6p_faddr = sin6->sin6_addr; 1024 inp->inp_fport = sin6->sin6_port; 1025 if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != NULL) 1026 inp->in6p_flowinfo = sin6->sin6_flowinfo; 1027 in_pcbinsconnhash(inp); 1028 1029 /* Compute window scaling to request. */ 1030 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 1031 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 1032 tp->request_r_scale++; 1033 1034 soisconnecting(so); 1035 tcpstat.tcps_connattempt++; 1036 tp->t_state = TCPS_SYN_SENT; 1037 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 1038 tp->iss = tcp_new_isn(tp); 1039 tcp_sendseqinit(tp); 1040 1041 /* 1042 * Generate a CC value for this connection and 1043 * check whether CC or CCnew should be used. 1044 */ 1045 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) { 1046 taop = &tao_noncached; 1047 bzero(taop, sizeof *taop); 1048 } 1049 1050 tp->cc_send = CC_INC(tcp_ccgen); 1051 if (taop->tao_ccsent != 0 && 1052 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 1053 taop->tao_ccsent = tp->cc_send; 1054 } else { 1055 taop->tao_ccsent = 0; 1056 tp->t_flags |= TF_SENDCCNEW; 1057 } 1058 1059 return (0); 1060 } 1061 #endif /* INET6 */ 1062 1063 /* 1064 * The new sockopt interface makes it possible for us to block in the 1065 * copyin/out step (if we take a page fault). Taking a page fault while 1066 * in a critical section is probably a Bad Thing. (Since sockets and pcbs 1067 * both now use TSM, there probably isn't any need for this function to 1068 * run in a critical section any more. This needs more examination.) 1069 */ 1070 int 1071 tcp_ctloutput(struct socket *so, struct sockopt *sopt) 1072 { 1073 int error, opt, optval; 1074 struct inpcb *inp; 1075 struct tcpcb *tp; 1076 1077 error = 0; 1078 crit_enter(); /* XXX */ 1079 inp = so->so_pcb; 1080 if (inp == NULL) { 1081 crit_exit(); 1082 return (ECONNRESET); 1083 } 1084 if (sopt->sopt_level != IPPROTO_TCP) { 1085 #ifdef INET6 1086 if (INP_CHECK_SOCKAF(so, AF_INET6)) 1087 error = ip6_ctloutput(so, sopt); 1088 else 1089 #endif /* INET6 */ 1090 error = ip_ctloutput(so, sopt); 1091 crit_exit(); 1092 return (error); 1093 } 1094 tp = intotcpcb(inp); 1095 1096 switch (sopt->sopt_dir) { 1097 case SOPT_SET: 1098 switch (sopt->sopt_name) { 1099 case TCP_NODELAY: 1100 case TCP_NOOPT: 1101 error = sooptcopyin(sopt, &optval, sizeof optval, 1102 sizeof optval); 1103 if (error) 1104 break; 1105 1106 switch (sopt->sopt_name) { 1107 case TCP_NODELAY: 1108 opt = TF_NODELAY; 1109 break; 1110 case TCP_NOOPT: 1111 opt = TF_NOOPT; 1112 break; 1113 default: 1114 opt = 0; /* dead code to fool gcc */ 1115 break; 1116 } 1117 1118 if (optval) 1119 tp->t_flags |= opt; 1120 else 1121 tp->t_flags &= ~opt; 1122 break; 1123 1124 case TCP_NOPUSH: 1125 error = sooptcopyin(sopt, &optval, sizeof optval, 1126 sizeof optval); 1127 if (error) 1128 break; 1129 1130 if (optval) 1131 tp->t_flags |= TF_NOPUSH; 1132 else { 1133 tp->t_flags &= ~TF_NOPUSH; 1134 error = tcp_output(tp); 1135 } 1136 break; 1137 1138 case TCP_MAXSEG: 1139 error = sooptcopyin(sopt, &optval, sizeof optval, 1140 sizeof optval); 1141 if (error) 1142 break; 1143 1144 if (optval > 0 && optval <= tp->t_maxseg) 1145 tp->t_maxseg = optval; 1146 else 1147 error = EINVAL; 1148 break; 1149 1150 default: 1151 error = ENOPROTOOPT; 1152 break; 1153 } 1154 break; 1155 1156 case SOPT_GET: 1157 switch (sopt->sopt_name) { 1158 case TCP_NODELAY: 1159 optval = tp->t_flags & TF_NODELAY; 1160 break; 1161 case TCP_MAXSEG: 1162 optval = tp->t_maxseg; 1163 break; 1164 case TCP_NOOPT: 1165 optval = tp->t_flags & TF_NOOPT; 1166 break; 1167 case TCP_NOPUSH: 1168 optval = tp->t_flags & TF_NOPUSH; 1169 break; 1170 default: 1171 error = ENOPROTOOPT; 1172 break; 1173 } 1174 if (error == 0) 1175 error = sooptcopyout(sopt, &optval, sizeof optval); 1176 break; 1177 } 1178 crit_exit(); 1179 return (error); 1180 } 1181 1182 /* 1183 * tcp_sendspace and tcp_recvspace are the default send and receive window 1184 * sizes, respectively. These are obsolescent (this information should 1185 * be set by the route). 1186 */ 1187 u_long tcp_sendspace = 1024*32; 1188 SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW, 1189 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size"); 1190 u_long tcp_recvspace = 57344; /* largest multiple of PAGE_SIZE < 64k */ 1191 SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW, 1192 &tcp_recvspace , 0, "Maximum incoming TCP datagram size"); 1193 1194 /* 1195 * Attach TCP protocol to socket, allocating 1196 * internet protocol control block, tcp control block, 1197 * bufer space, and entering LISTEN state if to accept connections. 1198 */ 1199 static int 1200 tcp_attach(struct socket *so, struct pru_attach_info *ai) 1201 { 1202 struct tcpcb *tp; 1203 struct inpcb *inp; 1204 int error; 1205 int cpu; 1206 #ifdef INET6 1207 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != NULL; 1208 #endif 1209 1210 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 1211 error = soreserve(so, tcp_sendspace, tcp_recvspace, 1212 ai->sb_rlimit); 1213 if (error) 1214 return (error); 1215 } 1216 cpu = mycpu->gd_cpuid; 1217 error = in_pcballoc(so, &tcbinfo[cpu]); 1218 if (error) 1219 return (error); 1220 inp = so->so_pcb; 1221 #ifdef INET6 1222 if (isipv6) { 1223 inp->inp_vflag |= INP_IPV6; 1224 inp->in6p_hops = -1; /* use kernel default */ 1225 } 1226 else 1227 #endif 1228 inp->inp_vflag |= INP_IPV4; 1229 tp = tcp_newtcpcb(inp); 1230 if (tp == 0) { 1231 int nofd = so->so_state & SS_NOFDREF; /* XXX */ 1232 1233 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ 1234 #ifdef INET6 1235 if (isipv6) 1236 in6_pcbdetach(inp); 1237 else 1238 #endif 1239 in_pcbdetach(inp); 1240 so->so_state |= nofd; 1241 return (ENOBUFS); 1242 } 1243 tp->t_state = TCPS_CLOSED; 1244 return (0); 1245 } 1246 1247 /* 1248 * Initiate (or continue) disconnect. 1249 * If embryonic state, just send reset (once). 1250 * If in ``let data drain'' option and linger null, just drop. 1251 * Otherwise (hard), mark socket disconnecting and drop 1252 * current input data; switch states based on user close, and 1253 * send segment to peer (with FIN). 1254 */ 1255 static struct tcpcb * 1256 tcp_disconnect(struct tcpcb *tp) 1257 { 1258 struct socket *so = tp->t_inpcb->inp_socket; 1259 1260 if (tp->t_state < TCPS_ESTABLISHED) 1261 tp = tcp_close(tp); 1262 else if ((so->so_options & SO_LINGER) && so->so_linger == 0) 1263 tp = tcp_drop(tp, 0); 1264 else { 1265 soisdisconnecting(so); 1266 sbflush(&so->so_rcv); 1267 tp = tcp_usrclosed(tp); 1268 if (tp) 1269 tcp_output(tp); 1270 } 1271 return (tp); 1272 } 1273 1274 /* 1275 * User issued close, and wish to trail through shutdown states: 1276 * if never received SYN, just forget it. If got a SYN from peer, 1277 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 1278 * If already got a FIN from peer, then almost done; go to LAST_ACK 1279 * state. In all other cases, have already sent FIN to peer (e.g. 1280 * after PRU_SHUTDOWN), and just have to play tedious game waiting 1281 * for peer to send FIN or not respond to keep-alives, etc. 1282 * We can let the user exit from the close as soon as the FIN is acked. 1283 */ 1284 static struct tcpcb * 1285 tcp_usrclosed(struct tcpcb *tp) 1286 { 1287 1288 switch (tp->t_state) { 1289 1290 case TCPS_CLOSED: 1291 case TCPS_LISTEN: 1292 tp->t_state = TCPS_CLOSED; 1293 tp = tcp_close(tp); 1294 break; 1295 1296 case TCPS_SYN_SENT: 1297 case TCPS_SYN_RECEIVED: 1298 tp->t_flags |= TF_NEEDFIN; 1299 break; 1300 1301 case TCPS_ESTABLISHED: 1302 tp->t_state = TCPS_FIN_WAIT_1; 1303 break; 1304 1305 case TCPS_CLOSE_WAIT: 1306 tp->t_state = TCPS_LAST_ACK; 1307 break; 1308 } 1309 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { 1310 soisdisconnected(tp->t_inpcb->inp_socket); 1311 /* To prevent the connection hanging in FIN_WAIT_2 forever. */ 1312 if (tp->t_state == TCPS_FIN_WAIT_2) 1313 callout_reset(tp->tt_2msl, tcp_maxidle, 1314 tcp_timer_2msl, tp); 1315 } 1316 return (tp); 1317 } 1318