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 1.26 2004/08/11 02:36:22 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 == 0) { 219 splx(s); 220 return EINVAL; /* XXX */ 221 } 222 tp = intotcpcb(inp); 223 TCPDEBUG1(); 224 tp = tcp_disconnect(tp); 225 226 TCPDEBUG2(PRU_DETACH); 227 splx(s); 228 return error; 229 } 230 231 #define COMMON_START() TCPDEBUG0; \ 232 do { \ 233 if (inp == 0) { \ 234 splx(s); \ 235 return EINVAL; \ 236 } \ 237 tp = intotcpcb(inp); \ 238 TCPDEBUG1(); \ 239 } while(0) 240 241 #define COMMON_END(req) out: TCPDEBUG2(req); splx(s); return error; goto out 242 243 244 /* 245 * Give the socket an address. 246 */ 247 static int 248 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 249 { 250 int s = splnet(); 251 int error = 0; 252 struct inpcb *inp = sotoinpcb(so); 253 struct tcpcb *tp; 254 struct sockaddr_in *sinp; 255 256 COMMON_START(); 257 258 /* 259 * Must check for multicast addresses and disallow binding 260 * to them. 261 */ 262 sinp = (struct sockaddr_in *)nam; 263 if (sinp->sin_family == AF_INET && 264 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 265 error = EAFNOSUPPORT; 266 goto out; 267 } 268 error = in_pcbbind(inp, nam, td); 269 if (error) 270 goto out; 271 COMMON_END(PRU_BIND); 272 273 } 274 275 #ifdef INET6 276 static int 277 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 278 { 279 int s = splnet(); 280 int error = 0; 281 struct inpcb *inp = sotoinpcb(so); 282 struct tcpcb *tp; 283 struct sockaddr_in6 *sin6p; 284 285 COMMON_START(); 286 287 /* 288 * Must check for multicast addresses and disallow binding 289 * to them. 290 */ 291 sin6p = (struct sockaddr_in6 *)nam; 292 if (sin6p->sin6_family == AF_INET6 && 293 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { 294 error = EAFNOSUPPORT; 295 goto out; 296 } 297 inp->inp_vflag &= ~INP_IPV4; 298 inp->inp_vflag |= INP_IPV6; 299 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { 300 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr)) 301 inp->inp_vflag |= INP_IPV4; 302 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { 303 struct sockaddr_in sin; 304 305 in6_sin6_2_sin(&sin, sin6p); 306 inp->inp_vflag |= INP_IPV4; 307 inp->inp_vflag &= ~INP_IPV6; 308 error = in_pcbbind(inp, (struct sockaddr *)&sin, td); 309 goto out; 310 } 311 } 312 error = in6_pcbbind(inp, nam, td); 313 if (error) 314 goto out; 315 COMMON_END(PRU_BIND); 316 } 317 #endif /* INET6 */ 318 319 #ifdef SMP 320 struct netmsg_inswildcard { 321 struct lwkt_msg nm_lmsg; 322 struct inpcb *nm_inp; 323 struct inpcbinfo *nm_pcbinfo; 324 }; 325 326 static int 327 in_pcbinswildcardhash_handler(struct lwkt_msg *msg0) 328 { 329 struct netmsg_inswildcard *msg = (struct netmsg_inswildcard *)msg0; 330 331 in_pcbinswildcardhash_oncpu(msg->nm_inp, msg->nm_pcbinfo); 332 lwkt_replymsg(&msg->nm_lmsg, 0); 333 return (EASYNC); 334 } 335 #endif 336 337 /* 338 * Prepare to accept connections. 339 */ 340 static int 341 tcp_usr_listen(struct socket *so, struct thread *td) 342 { 343 int s = splnet(); 344 int error = 0; 345 struct inpcb *inp = sotoinpcb(so); 346 struct tcpcb *tp; 347 #ifdef SMP 348 int cpu; 349 #endif 350 351 COMMON_START(); 352 if (inp->inp_lport == 0) { 353 error = in_pcbbind(inp, NULL, td); 354 if (error != 0) 355 goto out; 356 } 357 358 tp->t_state = TCPS_LISTEN; 359 #ifdef SMP 360 /* 361 * We have to set the flag because we can't have other cpus messing 362 * with our inp's flags. 363 */ 364 inp->inp_flags |= INP_WILDCARD_MP; 365 for (cpu = 0; cpu < ncpus2; cpu++) { 366 struct netmsg_inswildcard *msg; 367 368 if (cpu == mycpu->gd_cpuid) { 369 in_pcbinswildcardhash_oncpu(inp, &tcbinfo[cpu]); 370 continue; 371 } 372 373 msg = malloc(sizeof(struct netmsg_inswildcard), M_LWKTMSG, 374 M_INTWAIT); 375 lwkt_initmsg(&msg->nm_lmsg, &netisr_afree_rport, 0, 376 lwkt_cmd_func(in_pcbinswildcardhash_handler), 377 lwkt_cmd_op_none); 378 msg->nm_inp = inp; 379 msg->nm_pcbinfo = &tcbinfo[cpu]; 380 lwkt_sendmsg(tcp_cport(cpu), &msg->nm_lmsg); 381 } 382 #else 383 in_pcbinswildcardhash(inp); 384 #endif 385 COMMON_END(PRU_LISTEN); 386 } 387 388 #ifdef INET6 389 static int 390 tcp6_usr_listen(struct socket *so, struct thread *td) 391 { 392 int s = splnet(); 393 int error = 0; 394 struct inpcb *inp = sotoinpcb(so); 395 struct tcpcb *tp; 396 397 COMMON_START(); 398 if (inp->inp_lport == 0) { 399 inp->inp_vflag &= ~INP_IPV4; 400 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) 401 inp->inp_vflag |= INP_IPV4; 402 error = in6_pcbbind(inp, (struct sockaddr *)0, td); 403 } 404 if (error == 0) 405 tp->t_state = TCPS_LISTEN; 406 in_pcbinswildcardhash(inp); 407 COMMON_END(PRU_LISTEN); 408 } 409 #endif /* INET6 */ 410 411 /* 412 * Initiate connection to peer. 413 * Create a template for use in transmissions on this connection. 414 * Enter SYN_SENT state, and mark socket as connecting. 415 * Start keep-alive timer, and seed output sequence space. 416 * Send initial segment on connection. 417 */ 418 static int 419 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 420 { 421 int s = splnet(); 422 int error = 0; 423 struct inpcb *inp = sotoinpcb(so); 424 struct tcpcb *tp; 425 struct sockaddr_in *sinp; 426 427 COMMON_START(); 428 429 /* 430 * Must disallow TCP ``connections'' to multicast addresses. 431 */ 432 sinp = (struct sockaddr_in *)nam; 433 if (sinp->sin_family == AF_INET 434 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 435 error = EAFNOSUPPORT; 436 goto out; 437 } 438 439 prison_remote_ip(td, 0, &sinp->sin_addr.s_addr); 440 441 if ((error = tcp_connect(tp, nam, td)) != 0) 442 goto out; 443 error = tcp_output(tp); 444 COMMON_END(PRU_CONNECT); 445 } 446 447 #ifdef INET6 448 static int 449 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 450 { 451 int s = splnet(); 452 int error = 0; 453 struct inpcb *inp = sotoinpcb(so); 454 struct tcpcb *tp; 455 struct sockaddr_in6 *sin6p; 456 457 COMMON_START(); 458 459 /* 460 * Must disallow TCP ``connections'' to multicast addresses. 461 */ 462 sin6p = (struct sockaddr_in6 *)nam; 463 if (sin6p->sin6_family == AF_INET6 464 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { 465 error = EAFNOSUPPORT; 466 goto out; 467 } 468 469 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { 470 struct sockaddr_in sin; 471 472 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { 473 error = EINVAL; 474 goto out; 475 } 476 477 in6_sin6_2_sin(&sin, sin6p); 478 inp->inp_vflag |= INP_IPV4; 479 inp->inp_vflag &= ~INP_IPV6; 480 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0) 481 goto out; 482 error = tcp_output(tp); 483 goto out; 484 } 485 inp->inp_vflag &= ~INP_IPV4; 486 inp->inp_vflag |= INP_IPV6; 487 inp->inp_inc.inc_isipv6 = 1; 488 if ((error = tcp6_connect(tp, nam, td)) != 0) 489 goto out; 490 error = tcp_output(tp); 491 COMMON_END(PRU_CONNECT); 492 } 493 #endif /* INET6 */ 494 495 /* 496 * Initiate disconnect from peer. 497 * If connection never passed embryonic stage, just drop; 498 * else if don't need to let data drain, then can just drop anyways, 499 * else have to begin TCP shutdown process: mark socket disconnecting, 500 * drain unread data, state switch to reflect user close, and 501 * send segment (e.g. FIN) to peer. Socket will be really disconnected 502 * when peer sends FIN and acks ours. 503 * 504 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 505 */ 506 static int 507 tcp_usr_disconnect(struct socket *so) 508 { 509 int s = splnet(); 510 int error = 0; 511 struct inpcb *inp = sotoinpcb(so); 512 struct tcpcb *tp; 513 514 COMMON_START(); 515 tp = tcp_disconnect(tp); 516 COMMON_END(PRU_DISCONNECT); 517 } 518 519 /* 520 * Accept a connection. Essentially all the work is 521 * done at higher levels; just return the address 522 * of the peer, storing through addr. 523 */ 524 static int 525 tcp_usr_accept(struct socket *so, struct sockaddr **nam) 526 { 527 int s = splnet(); 528 int error = 0; 529 struct inpcb *inp = sotoinpcb(so); 530 struct tcpcb *tp = NULL; 531 TCPDEBUG0; 532 533 if (so->so_state & SS_ISDISCONNECTED) { 534 error = ECONNABORTED; 535 goto out; 536 } 537 if (inp == 0) { 538 splx(s); 539 return (EINVAL); 540 } 541 tp = intotcpcb(inp); 542 TCPDEBUG1(); 543 in_setpeeraddr(so, nam); 544 COMMON_END(PRU_ACCEPT); 545 } 546 547 #ifdef INET6 548 static int 549 tcp6_usr_accept(struct socket *so, struct sockaddr **nam) 550 { 551 int s = splnet(); 552 int error = 0; 553 struct inpcb *inp = sotoinpcb(so); 554 struct tcpcb *tp = NULL; 555 TCPDEBUG0; 556 557 if (so->so_state & SS_ISDISCONNECTED) { 558 error = ECONNABORTED; 559 goto out; 560 } 561 if (inp == 0) { 562 splx(s); 563 return (EINVAL); 564 } 565 tp = intotcpcb(inp); 566 TCPDEBUG1(); 567 in6_mapped_peeraddr(so, nam); 568 COMMON_END(PRU_ACCEPT); 569 } 570 #endif /* INET6 */ 571 /* 572 * Mark the connection as being incapable of further output. 573 */ 574 static int 575 tcp_usr_shutdown(struct socket *so) 576 { 577 int s = splnet(); 578 int error = 0; 579 struct inpcb *inp = sotoinpcb(so); 580 struct tcpcb *tp; 581 582 COMMON_START(); 583 socantsendmore(so); 584 tp = tcp_usrclosed(tp); 585 if (tp) 586 error = tcp_output(tp); 587 COMMON_END(PRU_SHUTDOWN); 588 } 589 590 /* 591 * After a receive, possibly send window update to peer. 592 */ 593 static int 594 tcp_usr_rcvd(struct socket *so, int flags) 595 { 596 int s = splnet(); 597 int error = 0; 598 struct inpcb *inp = sotoinpcb(so); 599 struct tcpcb *tp; 600 601 COMMON_START(); 602 tcp_output(tp); 603 COMMON_END(PRU_RCVD); 604 } 605 606 /* 607 * Do a send by putting data in output queue and updating urgent 608 * marker if URG set. Possibly send more data. Unlike the other 609 * pru_*() routines, the mbuf chains are our responsibility. We 610 * must either enqueue them or free them. The other pru_* routines 611 * generally are caller-frees. 612 */ 613 static int 614 tcp_usr_send(struct socket *so, int flags, struct mbuf *m, 615 struct sockaddr *nam, struct mbuf *control, struct thread *td) 616 { 617 int s = splnet(); 618 int error = 0; 619 struct inpcb *inp = sotoinpcb(so); 620 struct tcpcb *tp; 621 #ifdef INET6 622 int isipv6; 623 #endif 624 TCPDEBUG0; 625 626 if (inp == NULL) { 627 /* 628 * OOPS! we lost a race, the TCP session got reset after 629 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a 630 * network interrupt in the non-splnet() section of sosend(). 631 */ 632 if (m) 633 m_freem(m); 634 if (control) 635 m_freem(control); 636 error = ECONNRESET; /* XXX EPIPE? */ 637 tp = NULL; 638 TCPDEBUG1(); 639 goto out; 640 } 641 #ifdef INET6 642 isipv6 = nam && nam->sa_family == AF_INET6; 643 #endif /* INET6 */ 644 tp = intotcpcb(inp); 645 TCPDEBUG1(); 646 if (control) { 647 /* TCP doesn't do control messages (rights, creds, etc) */ 648 if (control->m_len) { 649 m_freem(control); 650 if (m) 651 m_freem(m); 652 error = EINVAL; 653 goto out; 654 } 655 m_freem(control); /* empty control, just free it */ 656 } 657 if(!(flags & PRUS_OOB)) { 658 sbappend(&so->so_snd, m); 659 if (nam && tp->t_state < TCPS_SYN_SENT) { 660 /* 661 * Do implied connect if not yet connected, 662 * initialize window to default value, and 663 * initialize maxseg/maxopd using peer's cached 664 * MSS. 665 */ 666 #ifdef INET6 667 if (isipv6) 668 error = tcp6_connect(tp, nam, td); 669 else 670 #endif /* INET6 */ 671 error = tcp_connect(tp, nam, td); 672 if (error) 673 goto out; 674 tp->snd_wnd = TTCP_CLIENT_SND_WND; 675 tcp_mss(tp, -1); 676 } 677 678 if (flags & PRUS_EOF) { 679 /* 680 * Close the send side of the connection after 681 * the data is sent. 682 */ 683 socantsendmore(so); 684 tp = tcp_usrclosed(tp); 685 } 686 if (tp != NULL) { 687 if (flags & PRUS_MORETOCOME) 688 tp->t_flags |= TF_MORETOCOME; 689 error = tcp_output(tp); 690 if (flags & PRUS_MORETOCOME) 691 tp->t_flags &= ~TF_MORETOCOME; 692 } 693 } else { 694 if (sbspace(&so->so_snd) < -512) { 695 m_freem(m); 696 error = ENOBUFS; 697 goto out; 698 } 699 /* 700 * According to RFC961 (Assigned Protocols), 701 * the urgent pointer points to the last octet 702 * of urgent data. We continue, however, 703 * to consider it to indicate the first octet 704 * of data past the urgent section. 705 * Otherwise, snd_up should be one lower. 706 */ 707 sbappend(&so->so_snd, m); 708 if (nam && tp->t_state < TCPS_SYN_SENT) { 709 /* 710 * Do implied connect if not yet connected, 711 * initialize window to default value, and 712 * initialize maxseg/maxopd using peer's cached 713 * MSS. 714 */ 715 #ifdef INET6 716 if (isipv6) 717 error = tcp6_connect(tp, nam, td); 718 else 719 #endif /* INET6 */ 720 error = tcp_connect(tp, nam, td); 721 if (error) 722 goto out; 723 tp->snd_wnd = TTCP_CLIENT_SND_WND; 724 tcp_mss(tp, -1); 725 } 726 tp->snd_up = tp->snd_una + so->so_snd.sb_cc; 727 tp->t_flags |= TF_FORCE; 728 error = tcp_output(tp); 729 tp->t_flags &= ~TF_FORCE; 730 } 731 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB : 732 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); 733 } 734 735 /* 736 * Abort the TCP. 737 */ 738 static int 739 tcp_usr_abort(struct socket *so) 740 { 741 int s = splnet(); 742 int error = 0; 743 struct inpcb *inp = sotoinpcb(so); 744 struct tcpcb *tp; 745 746 COMMON_START(); 747 tp = tcp_drop(tp, ECONNABORTED); 748 COMMON_END(PRU_ABORT); 749 } 750 751 /* 752 * Receive out-of-band data. 753 */ 754 static int 755 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) 756 { 757 int s = splnet(); 758 int error = 0; 759 struct inpcb *inp = sotoinpcb(so); 760 struct tcpcb *tp; 761 762 COMMON_START(); 763 if ((so->so_oobmark == 0 && 764 (so->so_state & SS_RCVATMARK) == 0) || 765 so->so_options & SO_OOBINLINE || 766 tp->t_oobflags & TCPOOB_HADDATA) { 767 error = EINVAL; 768 goto out; 769 } 770 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 771 error = EWOULDBLOCK; 772 goto out; 773 } 774 m->m_len = 1; 775 *mtod(m, caddr_t) = tp->t_iobc; 776 if ((flags & MSG_PEEK) == 0) 777 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 778 COMMON_END(PRU_RCVOOB); 779 } 780 781 /* xxx - should be const */ 782 struct pr_usrreqs tcp_usrreqs = { 783 tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind, 784 tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach, 785 tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd, 786 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, 787 in_setsockaddr, sosend, soreceive, sopoll 788 }; 789 790 #ifdef INET6 791 struct pr_usrreqs tcp6_usrreqs = { 792 tcp_usr_abort, tcp6_usr_accept, tcp_usr_attach, tcp6_usr_bind, 793 tcp6_usr_connect, pru_connect2_notsupp, in6_control, tcp_usr_detach, 794 tcp_usr_disconnect, tcp6_usr_listen, in6_mapped_peeraddr, tcp_usr_rcvd, 795 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, 796 in6_mapped_sockaddr, sosend, soreceive, sopoll 797 }; 798 #endif /* INET6 */ 799 800 static int 801 tcp_connect_oncpu(struct tcpcb *tp, struct sockaddr_in *sin, 802 struct sockaddr_in *if_sin) 803 { 804 struct inpcb *inp = tp->t_inpcb, *oinp; 805 struct socket *so = inp->inp_socket; 806 struct tcpcb *otp; 807 struct rmxp_tao *taop; 808 struct rmxp_tao tao_noncached; 809 810 oinp = in_pcblookup_hash(&tcbinfo[mycpu->gd_cpuid], 811 sin->sin_addr, sin->sin_port, 812 inp->inp_laddr.s_addr != INADDR_ANY ? 813 inp->inp_laddr : if_sin->sin_addr, 814 inp->inp_lport, 0, NULL); 815 if (oinp != NULL) { 816 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 817 otp->t_state == TCPS_TIME_WAIT && 818 (ticks - otp->t_starttime) < tcp_msl && 819 (otp->t_flags & TF_RCVD_CC)) 820 (void) tcp_close(otp); 821 else 822 return (EADDRINUSE); 823 } 824 if (inp->inp_laddr.s_addr == INADDR_ANY) 825 inp->inp_laddr = if_sin->sin_addr; 826 inp->inp_faddr = sin->sin_addr; 827 inp->inp_fport = sin->sin_port; 828 inp->inp_cpcbinfo = &tcbinfo[mycpu->gd_cpuid]; 829 in_pcbinsconnhash(inp); 830 831 /* Compute window scaling to request. */ 832 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 833 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 834 tp->request_r_scale++; 835 836 soisconnecting(so); 837 tcpstat.tcps_connattempt++; 838 tp->t_state = TCPS_SYN_SENT; 839 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 840 tp->iss = tcp_new_isn(tp); 841 tp->t_bw_rtseq = tp->iss; 842 tcp_sendseqinit(tp); 843 844 /* 845 * Generate a CC value for this connection and 846 * check whether CC or CCnew should be used. 847 */ 848 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) { 849 taop = &tao_noncached; 850 bzero(taop, sizeof(*taop)); 851 } 852 853 tp->cc_send = CC_INC(tcp_ccgen); 854 if (taop->tao_ccsent != 0 && 855 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 856 taop->tao_ccsent = tp->cc_send; 857 } else { 858 taop->tao_ccsent = 0; 859 tp->t_flags |= TF_SENDCCNEW; 860 } 861 862 return (0); 863 } 864 865 #ifdef SMP 866 867 struct netmsg_tcp_connect { 868 struct lwkt_msg nm_lmsg; 869 struct tcpcb *nm_tp; 870 struct sockaddr_in *nm_sin; 871 struct sockaddr_in *nm_ifsin; 872 }; 873 874 static int 875 tcp_connect_handler(lwkt_msg_t lmsg) 876 { 877 struct netmsg_tcp_connect *msg = (void *)lmsg; 878 int error; 879 880 error = tcp_connect_oncpu(msg->nm_tp, msg->nm_sin, msg->nm_ifsin); 881 lwkt_replymsg(lmsg, error); 882 return(EASYNC); 883 } 884 885 #endif 886 887 /* 888 * Common subroutine to open a TCP connection to remote host specified 889 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local 890 * port number if needed. Call in_pcbladdr to do the routing and to choose 891 * a local host address (interface). If there is an existing incarnation 892 * of the same connection in TIME-WAIT state and if the remote host was 893 * sending CC options and if the connection duration was < MSL, then 894 * truncate the previous TIME-WAIT state and proceed. 895 * Initialize connection parameters and enter SYN-SENT state. 896 */ 897 static int 898 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td) 899 { 900 struct inpcb *inp = tp->t_inpcb; 901 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 902 struct sockaddr_in *if_sin; 903 int error; 904 #ifdef SMP 905 lwkt_port_t port; 906 #endif 907 908 if (inp->inp_lport == 0) { 909 error = in_pcbbind(inp, (struct sockaddr *)NULL, td); 910 if (error) 911 return (error); 912 } 913 914 /* 915 * Cannot simply call in_pcbconnect, because there might be an 916 * earlier incarnation of this same connection still in 917 * TIME_WAIT state, creating an ADDRINUSE error. 918 */ 919 error = in_pcbladdr(inp, nam, &if_sin); 920 if (error) 921 return (error); 922 923 #ifdef SMP 924 port = tcp_addrport(sin->sin_addr.s_addr, sin->sin_port, 925 inp->inp_laddr.s_addr ? 926 inp->inp_laddr.s_addr : if_sin->sin_addr.s_addr, 927 inp->inp_lport); 928 929 if (port->mp_td != curthread) { 930 struct netmsg_tcp_connect msg; 931 932 lwkt_initmsg(&msg.nm_lmsg, &curthread->td_msgport, 0, 933 lwkt_cmd_func(tcp_connect_handler), lwkt_cmd_op_none); 934 msg.nm_tp = tp; 935 msg.nm_sin = sin; 936 msg.nm_ifsin = if_sin; 937 error = lwkt_domsg(port, &msg.nm_lmsg); 938 } else 939 #endif 940 error = tcp_connect_oncpu(tp, sin, if_sin); 941 942 return (error); 943 } 944 945 #ifdef INET6 946 static int 947 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td) 948 { 949 struct inpcb *inp = tp->t_inpcb, *oinp; 950 struct socket *so = inp->inp_socket; 951 struct tcpcb *otp; 952 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; 953 struct in6_addr *addr6; 954 struct rmxp_tao *taop; 955 struct rmxp_tao tao_noncached; 956 int error; 957 958 if (inp->inp_lport == 0) { 959 error = in6_pcbbind(inp, (struct sockaddr *)0, td); 960 if (error) 961 return error; 962 } 963 964 /* 965 * Cannot simply call in_pcbconnect, because there might be an 966 * earlier incarnation of this same connection still in 967 * TIME_WAIT state, creating an ADDRINUSE error. 968 */ 969 error = in6_pcbladdr(inp, nam, &addr6); 970 if (error) 971 return error; 972 oinp = in6_pcblookup_hash(inp->inp_cpcbinfo, 973 &sin6->sin6_addr, sin6->sin6_port, 974 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ? 975 addr6 : &inp->in6p_laddr, 976 inp->inp_lport, 0, NULL); 977 if (oinp) { 978 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 979 otp->t_state == TCPS_TIME_WAIT && 980 (ticks - otp->t_starttime) < tcp_msl && 981 (otp->t_flags & TF_RCVD_CC)) 982 otp = tcp_close(otp); 983 else 984 return (EADDRINUSE); 985 } 986 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) 987 inp->in6p_laddr = *addr6; 988 inp->in6p_faddr = sin6->sin6_addr; 989 inp->inp_fport = sin6->sin6_port; 990 if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != NULL) 991 inp->in6p_flowinfo = sin6->sin6_flowinfo; 992 in_pcbinsconnhash(inp); 993 994 /* Compute window scaling to request. */ 995 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 996 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 997 tp->request_r_scale++; 998 999 soisconnecting(so); 1000 tcpstat.tcps_connattempt++; 1001 tp->t_state = TCPS_SYN_SENT; 1002 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 1003 tp->iss = tcp_new_isn(tp); 1004 tp->t_bw_rtseq = tp->iss; 1005 tcp_sendseqinit(tp); 1006 1007 /* 1008 * Generate a CC value for this connection and 1009 * check whether CC or CCnew should be used. 1010 */ 1011 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) { 1012 taop = &tao_noncached; 1013 bzero(taop, sizeof(*taop)); 1014 } 1015 1016 tp->cc_send = CC_INC(tcp_ccgen); 1017 if (taop->tao_ccsent != 0 && 1018 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 1019 taop->tao_ccsent = tp->cc_send; 1020 } else { 1021 taop->tao_ccsent = 0; 1022 tp->t_flags |= TF_SENDCCNEW; 1023 } 1024 1025 return (0); 1026 } 1027 #endif /* INET6 */ 1028 1029 /* 1030 * The new sockopt interface makes it possible for us to block in the 1031 * copyin/out step (if we take a page fault). Taking a page fault at 1032 * splnet() is probably a Bad Thing. (Since sockets and pcbs both now 1033 * use TSM, there probably isn't any need for this function to run at 1034 * splnet() any more. This needs more examination.) 1035 */ 1036 int 1037 tcp_ctloutput(so, sopt) 1038 struct socket *so; 1039 struct sockopt *sopt; 1040 { 1041 int error, opt, optval, s; 1042 struct inpcb *inp; 1043 struct tcpcb *tp; 1044 1045 error = 0; 1046 s = splnet(); /* XXX */ 1047 inp = sotoinpcb(so); 1048 if (inp == NULL) { 1049 splx(s); 1050 return (ECONNRESET); 1051 } 1052 if (sopt->sopt_level != IPPROTO_TCP) { 1053 #ifdef INET6 1054 if (INP_CHECK_SOCKAF(so, AF_INET6)) 1055 error = ip6_ctloutput(so, sopt); 1056 else 1057 #endif /* INET6 */ 1058 error = ip_ctloutput(so, sopt); 1059 splx(s); 1060 return (error); 1061 } 1062 tp = intotcpcb(inp); 1063 1064 switch (sopt->sopt_dir) { 1065 case SOPT_SET: 1066 switch (sopt->sopt_name) { 1067 case TCP_NODELAY: 1068 case TCP_NOOPT: 1069 error = sooptcopyin(sopt, &optval, sizeof optval, 1070 sizeof optval); 1071 if (error) 1072 break; 1073 1074 switch (sopt->sopt_name) { 1075 case TCP_NODELAY: 1076 opt = TF_NODELAY; 1077 break; 1078 case TCP_NOOPT: 1079 opt = TF_NOOPT; 1080 break; 1081 default: 1082 opt = 0; /* dead code to fool gcc */ 1083 break; 1084 } 1085 1086 if (optval) 1087 tp->t_flags |= opt; 1088 else 1089 tp->t_flags &= ~opt; 1090 break; 1091 1092 case TCP_NOPUSH: 1093 error = sooptcopyin(sopt, &optval, sizeof optval, 1094 sizeof optval); 1095 if (error) 1096 break; 1097 1098 if (optval) 1099 tp->t_flags |= TF_NOPUSH; 1100 else { 1101 tp->t_flags &= ~TF_NOPUSH; 1102 error = tcp_output(tp); 1103 } 1104 break; 1105 1106 case TCP_MAXSEG: 1107 error = sooptcopyin(sopt, &optval, sizeof optval, 1108 sizeof optval); 1109 if (error) 1110 break; 1111 1112 if (optval > 0 && optval <= tp->t_maxseg) 1113 tp->t_maxseg = optval; 1114 else 1115 error = EINVAL; 1116 break; 1117 1118 default: 1119 error = ENOPROTOOPT; 1120 break; 1121 } 1122 break; 1123 1124 case SOPT_GET: 1125 switch (sopt->sopt_name) { 1126 case TCP_NODELAY: 1127 optval = tp->t_flags & TF_NODELAY; 1128 break; 1129 case TCP_MAXSEG: 1130 optval = tp->t_maxseg; 1131 break; 1132 case TCP_NOOPT: 1133 optval = tp->t_flags & TF_NOOPT; 1134 break; 1135 case TCP_NOPUSH: 1136 optval = tp->t_flags & TF_NOPUSH; 1137 break; 1138 default: 1139 error = ENOPROTOOPT; 1140 break; 1141 } 1142 if (error == 0) 1143 error = sooptcopyout(sopt, &optval, sizeof optval); 1144 break; 1145 } 1146 splx(s); 1147 return (error); 1148 } 1149 1150 /* 1151 * tcp_sendspace and tcp_recvspace are the default send and receive window 1152 * sizes, respectively. These are obsolescent (this information should 1153 * be set by the route). 1154 */ 1155 u_long tcp_sendspace = 1024*32; 1156 SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW, 1157 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size"); 1158 u_long tcp_recvspace = 57344; /* largest multiple of PAGE_SIZE < 64k */ 1159 SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW, 1160 &tcp_recvspace , 0, "Maximum incoming TCP datagram size"); 1161 1162 /* 1163 * Attach TCP protocol to socket, allocating 1164 * internet protocol control block, tcp control block, 1165 * bufer space, and entering LISTEN state if to accept connections. 1166 */ 1167 static int 1168 tcp_attach(struct socket *so, struct pru_attach_info *ai) 1169 { 1170 struct tcpcb *tp; 1171 struct inpcb *inp; 1172 int error; 1173 int cpu; 1174 #ifdef INET6 1175 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != NULL; 1176 #endif 1177 1178 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 1179 error = soreserve(so, tcp_sendspace, tcp_recvspace, 1180 ai->sb_rlimit); 1181 if (error) 1182 return (error); 1183 } 1184 cpu = mycpu->gd_cpuid; 1185 error = in_pcballoc(so, &tcbinfo[cpu]); 1186 if (error) 1187 return (error); 1188 inp = sotoinpcb(so); 1189 #ifdef INET6 1190 if (isipv6) { 1191 inp->inp_vflag |= INP_IPV6; 1192 inp->in6p_hops = -1; /* use kernel default */ 1193 } 1194 else 1195 #endif 1196 inp->inp_vflag |= INP_IPV4; 1197 tp = tcp_newtcpcb(inp); 1198 if (tp == 0) { 1199 int nofd = so->so_state & SS_NOFDREF; /* XXX */ 1200 1201 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ 1202 #ifdef INET6 1203 if (isipv6) 1204 in6_pcbdetach(inp); 1205 else 1206 #endif 1207 in_pcbdetach(inp); 1208 so->so_state |= nofd; 1209 return (ENOBUFS); 1210 } 1211 tp->t_state = TCPS_CLOSED; 1212 return (0); 1213 } 1214 1215 /* 1216 * Initiate (or continue) disconnect. 1217 * If embryonic state, just send reset (once). 1218 * If in ``let data drain'' option and linger null, just drop. 1219 * Otherwise (hard), mark socket disconnecting and drop 1220 * current input data; switch states based on user close, and 1221 * send segment to peer (with FIN). 1222 */ 1223 static struct tcpcb * 1224 tcp_disconnect(tp) 1225 struct tcpcb *tp; 1226 { 1227 struct socket *so = tp->t_inpcb->inp_socket; 1228 1229 if (tp->t_state < TCPS_ESTABLISHED) 1230 tp = tcp_close(tp); 1231 else if ((so->so_options & SO_LINGER) && so->so_linger == 0) 1232 tp = tcp_drop(tp, 0); 1233 else { 1234 soisdisconnecting(so); 1235 sbflush(&so->so_rcv); 1236 tp = tcp_usrclosed(tp); 1237 if (tp) 1238 (void) tcp_output(tp); 1239 } 1240 return (tp); 1241 } 1242 1243 /* 1244 * User issued close, and wish to trail through shutdown states: 1245 * if never received SYN, just forget it. If got a SYN from peer, 1246 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 1247 * If already got a FIN from peer, then almost done; go to LAST_ACK 1248 * state. In all other cases, have already sent FIN to peer (e.g. 1249 * after PRU_SHUTDOWN), and just have to play tedious game waiting 1250 * for peer to send FIN or not respond to keep-alives, etc. 1251 * We can let the user exit from the close as soon as the FIN is acked. 1252 */ 1253 static struct tcpcb * 1254 tcp_usrclosed(tp) 1255 struct tcpcb *tp; 1256 { 1257 1258 switch (tp->t_state) { 1259 1260 case TCPS_CLOSED: 1261 case TCPS_LISTEN: 1262 tp->t_state = TCPS_CLOSED; 1263 tp = tcp_close(tp); 1264 break; 1265 1266 case TCPS_SYN_SENT: 1267 case TCPS_SYN_RECEIVED: 1268 tp->t_flags |= TF_NEEDFIN; 1269 break; 1270 1271 case TCPS_ESTABLISHED: 1272 tp->t_state = TCPS_FIN_WAIT_1; 1273 break; 1274 1275 case TCPS_CLOSE_WAIT: 1276 tp->t_state = TCPS_LAST_ACK; 1277 break; 1278 } 1279 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { 1280 soisdisconnected(tp->t_inpcb->inp_socket); 1281 /* To prevent the connection hanging in FIN_WAIT_2 forever. */ 1282 if (tp->t_state == TCPS_FIN_WAIT_2) 1283 callout_reset(tp->tt_2msl, tcp_maxidle, 1284 tcp_timer_2msl, tp); 1285 } 1286 return (tp); 1287 } 1288