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 1.33 2005/02/08 22:56:19 hsu 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 = so->so_pcb; 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 = so->so_pcb; 215 struct tcpcb *tp; 216 TCPDEBUG0; 217 218 if (inp == NULL) { 219 splx(s); 220 return EINVAL; /* XXX */ 221 } 222 223 /* 224 * It's possible for the tcpcb (tp) to disconnect from the inp due 225 * to tcp_drop()->tcp_close() being called. This may occur *after* 226 * the detach message has been queued so we may find a NULL tp here. 227 */ 228 if ((tp = intotcpcb(inp)) != NULL) { 229 TCPDEBUG1(); 230 tp = tcp_disconnect(tp); 231 TCPDEBUG2(PRU_DETACH); 232 } 233 splx(s); 234 return error; 235 } 236 237 #define COMMON_START() TCPDEBUG0; \ 238 do { \ 239 if (inp == 0) { \ 240 splx(s); \ 241 return EINVAL; \ 242 } \ 243 tp = intotcpcb(inp); \ 244 TCPDEBUG1(); \ 245 } while(0) 246 247 #define COMMON_END(req) out: TCPDEBUG2(req); splx(s); return error; goto out 248 249 250 /* 251 * Give the socket an address. 252 */ 253 static int 254 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 255 { 256 int s = splnet(); 257 int error = 0; 258 struct inpcb *inp = so->so_pcb; 259 struct tcpcb *tp; 260 struct sockaddr_in *sinp; 261 262 COMMON_START(); 263 264 /* 265 * Must check for multicast addresses and disallow binding 266 * to them. 267 */ 268 sinp = (struct sockaddr_in *)nam; 269 if (sinp->sin_family == AF_INET && 270 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 271 error = EAFNOSUPPORT; 272 goto out; 273 } 274 error = in_pcbbind(inp, nam, td); 275 if (error) 276 goto out; 277 COMMON_END(PRU_BIND); 278 279 } 280 281 #ifdef INET6 282 static int 283 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 284 { 285 int s = splnet(); 286 int error = 0; 287 struct inpcb *inp = so->so_pcb; 288 struct tcpcb *tp; 289 struct sockaddr_in6 *sin6p; 290 291 COMMON_START(); 292 293 /* 294 * Must check for multicast addresses and disallow binding 295 * to them. 296 */ 297 sin6p = (struct sockaddr_in6 *)nam; 298 if (sin6p->sin6_family == AF_INET6 && 299 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { 300 error = EAFNOSUPPORT; 301 goto out; 302 } 303 inp->inp_vflag &= ~INP_IPV4; 304 inp->inp_vflag |= INP_IPV6; 305 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { 306 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr)) 307 inp->inp_vflag |= INP_IPV4; 308 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { 309 struct sockaddr_in sin; 310 311 in6_sin6_2_sin(&sin, sin6p); 312 inp->inp_vflag |= INP_IPV4; 313 inp->inp_vflag &= ~INP_IPV6; 314 error = in_pcbbind(inp, (struct sockaddr *)&sin, td); 315 goto out; 316 } 317 } 318 error = in6_pcbbind(inp, nam, td); 319 if (error) 320 goto out; 321 COMMON_END(PRU_BIND); 322 } 323 #endif /* INET6 */ 324 325 #ifdef SMP 326 struct netmsg_inswildcard { 327 struct lwkt_msg nm_lmsg; 328 struct inpcb *nm_inp; 329 struct inpcbinfo *nm_pcbinfo; 330 }; 331 332 static int 333 in_pcbinswildcardhash_handler(struct lwkt_msg *msg0) 334 { 335 struct netmsg_inswildcard *msg = (struct netmsg_inswildcard *)msg0; 336 337 in_pcbinswildcardhash_oncpu(msg->nm_inp, msg->nm_pcbinfo); 338 lwkt_replymsg(&msg->nm_lmsg, 0); 339 return (EASYNC); 340 } 341 #endif 342 343 /* 344 * Prepare to accept connections. 345 */ 346 static int 347 tcp_usr_listen(struct socket *so, struct thread *td) 348 { 349 int s = splnet(); 350 int error = 0; 351 struct inpcb *inp = so->so_pcb; 352 struct tcpcb *tp; 353 #ifdef SMP 354 int cpu; 355 #endif 356 357 COMMON_START(); 358 if (inp->inp_lport == 0) { 359 error = in_pcbbind(inp, NULL, td); 360 if (error != 0) 361 goto out; 362 } 363 364 tp->t_state = TCPS_LISTEN; 365 #ifdef SMP 366 /* 367 * We have to set the flag because we can't have other cpus 368 * messing with our inp's flags. 369 */ 370 inp->inp_flags |= INP_WILDCARD_MP; 371 for (cpu = 0; cpu < ncpus2; cpu++) { 372 struct netmsg_inswildcard *msg; 373 374 if (cpu == mycpu->gd_cpuid) { 375 in_pcbinswildcardhash_oncpu(inp, &tcbinfo[cpu]); 376 continue; 377 } 378 379 msg = malloc(sizeof(struct netmsg_inswildcard), M_LWKTMSG, 380 M_INTWAIT); 381 lwkt_initmsg(&msg->nm_lmsg, &netisr_afree_rport, 0, 382 lwkt_cmd_func(in_pcbinswildcardhash_handler), 383 lwkt_cmd_op_none); 384 msg->nm_inp = inp; 385 msg->nm_pcbinfo = &tcbinfo[cpu]; 386 lwkt_sendmsg(tcp_cport(cpu), &msg->nm_lmsg); 387 } 388 #else 389 in_pcbinswildcardhash(inp); 390 #endif 391 COMMON_END(PRU_LISTEN); 392 } 393 394 #ifdef INET6 395 static int 396 tcp6_usr_listen(struct socket *so, struct thread *td) 397 { 398 int s = splnet(); 399 int error = 0; 400 struct inpcb *inp = so->so_pcb; 401 struct tcpcb *tp; 402 #ifdef SMP 403 int cpu; 404 #endif 405 406 COMMON_START(); 407 if (inp->inp_lport == 0) { 408 if (!(inp->inp_flags & IN6P_IPV6_V6ONLY)) 409 inp->inp_vflag |= INP_IPV4; 410 else 411 inp->inp_vflag &= ~INP_IPV4; 412 error = in6_pcbbind(inp, (struct sockaddr *)0, td); 413 } 414 if (error == 0) 415 tp->t_state = TCPS_LISTEN; 416 #ifdef SMP 417 /* 418 * We have to set the flag because we can't have other cpus 419 * messing with our inp's flags. 420 */ 421 inp->inp_flags |= INP_WILDCARD_MP; 422 for (cpu = 0; cpu < ncpus2; cpu++) { 423 struct netmsg_inswildcard *msg; 424 425 if (cpu == mycpu->gd_cpuid) { 426 in_pcbinswildcardhash_oncpu(inp, &tcbinfo[cpu]); 427 continue; 428 } 429 430 msg = malloc(sizeof(struct netmsg_inswildcard), M_LWKTMSG, 431 M_INTWAIT); 432 lwkt_initmsg(&msg->nm_lmsg, &netisr_afree_rport, 0, 433 lwkt_cmd_func(in_pcbinswildcardhash_handler), 434 lwkt_cmd_op_none); 435 msg->nm_inp = inp; 436 msg->nm_pcbinfo = &tcbinfo[cpu]; 437 lwkt_sendmsg(tcp_cport(cpu), &msg->nm_lmsg); 438 } 439 #else 440 in_pcbinswildcardhash(inp); 441 #endif 442 COMMON_END(PRU_LISTEN); 443 } 444 #endif /* INET6 */ 445 446 /* 447 * Initiate connection to peer. 448 * Create a template for use in transmissions on this connection. 449 * Enter SYN_SENT state, and mark socket as connecting. 450 * Start keep-alive timer, and seed output sequence space. 451 * Send initial segment on connection. 452 */ 453 static int 454 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 455 { 456 int s = splnet(); 457 int error = 0; 458 struct inpcb *inp = so->so_pcb; 459 struct tcpcb *tp; 460 struct sockaddr_in *sinp; 461 462 COMMON_START(); 463 464 /* 465 * Must disallow TCP ``connections'' to multicast addresses. 466 */ 467 sinp = (struct sockaddr_in *)nam; 468 if (sinp->sin_family == AF_INET 469 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 470 error = EAFNOSUPPORT; 471 goto out; 472 } 473 474 prison_remote_ip(td, 0, &sinp->sin_addr.s_addr); 475 476 if ((error = tcp_connect(tp, nam, td)) != 0) 477 goto out; 478 error = tcp_output(tp); 479 COMMON_END(PRU_CONNECT); 480 } 481 482 #ifdef INET6 483 static int 484 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 485 { 486 int s = splnet(); 487 int error = 0; 488 struct inpcb *inp = so->so_pcb; 489 struct tcpcb *tp; 490 struct sockaddr_in6 *sin6p; 491 492 COMMON_START(); 493 494 /* 495 * Must disallow TCP ``connections'' to multicast addresses. 496 */ 497 sin6p = (struct sockaddr_in6 *)nam; 498 if (sin6p->sin6_family == AF_INET6 499 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) { 500 error = EAFNOSUPPORT; 501 goto out; 502 } 503 504 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { 505 struct sockaddr_in sin; 506 507 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { 508 error = EINVAL; 509 goto out; 510 } 511 512 in6_sin6_2_sin(&sin, sin6p); 513 inp->inp_vflag |= INP_IPV4; 514 inp->inp_vflag &= ~INP_IPV6; 515 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0) 516 goto out; 517 error = tcp_output(tp); 518 goto out; 519 } 520 inp->inp_vflag &= ~INP_IPV4; 521 inp->inp_vflag |= INP_IPV6; 522 inp->inp_inc.inc_isipv6 = 1; 523 if ((error = tcp6_connect(tp, nam, td)) != 0) 524 goto out; 525 error = tcp_output(tp); 526 COMMON_END(PRU_CONNECT); 527 } 528 #endif /* INET6 */ 529 530 /* 531 * Initiate disconnect from peer. 532 * If connection never passed embryonic stage, just drop; 533 * else if don't need to let data drain, then can just drop anyways, 534 * else have to begin TCP shutdown process: mark socket disconnecting, 535 * drain unread data, state switch to reflect user close, and 536 * send segment (e.g. FIN) to peer. Socket will be really disconnected 537 * when peer sends FIN and acks ours. 538 * 539 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 540 */ 541 static int 542 tcp_usr_disconnect(struct socket *so) 543 { 544 int s = splnet(); 545 int error = 0; 546 struct inpcb *inp = so->so_pcb; 547 struct tcpcb *tp; 548 549 COMMON_START(); 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 s = splnet(); 563 int error = 0; 564 struct inpcb *inp = so->so_pcb; 565 struct tcpcb *tp = NULL; 566 TCPDEBUG0; 567 568 if (so->so_state & SS_ISDISCONNECTED) { 569 error = ECONNABORTED; 570 goto out; 571 } 572 if (inp == 0) { 573 splx(s); 574 return (EINVAL); 575 } 576 tp = intotcpcb(inp); 577 TCPDEBUG1(); 578 in_setpeeraddr(so, nam); 579 COMMON_END(PRU_ACCEPT); 580 } 581 582 #ifdef INET6 583 static int 584 tcp6_usr_accept(struct socket *so, struct sockaddr **nam) 585 { 586 int s = splnet(); 587 int error = 0; 588 struct inpcb *inp = so->so_pcb; 589 struct tcpcb *tp = NULL; 590 TCPDEBUG0; 591 592 if (so->so_state & SS_ISDISCONNECTED) { 593 error = ECONNABORTED; 594 goto out; 595 } 596 if (inp == 0) { 597 splx(s); 598 return (EINVAL); 599 } 600 tp = intotcpcb(inp); 601 TCPDEBUG1(); 602 in6_mapped_peeraddr(so, nam); 603 COMMON_END(PRU_ACCEPT); 604 } 605 #endif /* INET6 */ 606 /* 607 * Mark the connection as being incapable of further output. 608 */ 609 static int 610 tcp_usr_shutdown(struct socket *so) 611 { 612 int s = splnet(); 613 int error = 0; 614 struct inpcb *inp = so->so_pcb; 615 struct tcpcb *tp; 616 617 COMMON_START(); 618 socantsendmore(so); 619 tp = tcp_usrclosed(tp); 620 if (tp) 621 error = tcp_output(tp); 622 COMMON_END(PRU_SHUTDOWN); 623 } 624 625 /* 626 * After a receive, possibly send window update to peer. 627 */ 628 static int 629 tcp_usr_rcvd(struct socket *so, int flags) 630 { 631 int s = splnet(); 632 int error = 0; 633 struct inpcb *inp = so->so_pcb; 634 struct tcpcb *tp; 635 636 COMMON_START(); 637 tcp_output(tp); 638 COMMON_END(PRU_RCVD); 639 } 640 641 /* 642 * Do a send by putting data in output queue and updating urgent 643 * marker if URG set. Possibly send more data. Unlike the other 644 * pru_*() routines, the mbuf chains are our responsibility. We 645 * must either enqueue them or free them. The other pru_* routines 646 * generally are caller-frees. 647 */ 648 static int 649 tcp_usr_send(struct socket *so, int flags, struct mbuf *m, 650 struct sockaddr *nam, struct mbuf *control, struct thread *td) 651 { 652 int s = splnet(); 653 int error = 0; 654 struct inpcb *inp = so->so_pcb; 655 struct tcpcb *tp; 656 #ifdef INET6 657 int isipv6; 658 #endif 659 TCPDEBUG0; 660 661 if (inp == NULL) { 662 /* 663 * OOPS! we lost a race, the TCP session got reset after 664 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a 665 * network interrupt in the non-splnet() section of sosend(). 666 */ 667 if (m) 668 m_freem(m); 669 if (control) 670 m_freem(control); 671 error = ECONNRESET; /* XXX EPIPE? */ 672 tp = NULL; 673 TCPDEBUG1(); 674 goto out; 675 } 676 #ifdef INET6 677 isipv6 = nam && nam->sa_family == AF_INET6; 678 #endif /* INET6 */ 679 tp = intotcpcb(inp); 680 TCPDEBUG1(); 681 if (control) { 682 /* TCP doesn't do control messages (rights, creds, etc) */ 683 if (control->m_len) { 684 m_freem(control); 685 if (m) 686 m_freem(m); 687 error = EINVAL; 688 goto out; 689 } 690 m_freem(control); /* empty control, just free it */ 691 } 692 if(!(flags & PRUS_OOB)) { 693 sbappendstream(&so->so_snd, m); 694 if (nam && tp->t_state < TCPS_SYN_SENT) { 695 /* 696 * Do implied connect if not yet connected, 697 * initialize window to default value, and 698 * initialize maxseg/maxopd using peer's cached 699 * MSS. 700 */ 701 #ifdef INET6 702 if (isipv6) 703 error = tcp6_connect(tp, nam, td); 704 else 705 #endif /* INET6 */ 706 error = tcp_connect(tp, nam, td); 707 if (error) 708 goto out; 709 tp->snd_wnd = TTCP_CLIENT_SND_WND; 710 tcp_mss(tp, -1); 711 } 712 713 if (flags & PRUS_EOF) { 714 /* 715 * Close the send side of the connection after 716 * the data is sent. 717 */ 718 socantsendmore(so); 719 tp = tcp_usrclosed(tp); 720 } 721 if (tp != NULL) { 722 if (flags & PRUS_MORETOCOME) 723 tp->t_flags |= TF_MORETOCOME; 724 error = tcp_output(tp); 725 if (flags & PRUS_MORETOCOME) 726 tp->t_flags &= ~TF_MORETOCOME; 727 } 728 } else { 729 if (sbspace(&so->so_snd) < -512) { 730 m_freem(m); 731 error = ENOBUFS; 732 goto out; 733 } 734 /* 735 * According to RFC961 (Assigned Protocols), 736 * the urgent pointer points to the last octet 737 * of urgent data. We continue, however, 738 * to consider it to indicate the first octet 739 * of data past the urgent section. 740 * Otherwise, snd_up should be one lower. 741 */ 742 sbappendstream(&so->so_snd, m); 743 if (nam && tp->t_state < TCPS_SYN_SENT) { 744 /* 745 * Do implied connect if not yet connected, 746 * initialize window to default value, and 747 * initialize maxseg/maxopd using peer's cached 748 * MSS. 749 */ 750 #ifdef INET6 751 if (isipv6) 752 error = tcp6_connect(tp, nam, td); 753 else 754 #endif /* INET6 */ 755 error = tcp_connect(tp, nam, td); 756 if (error) 757 goto out; 758 tp->snd_wnd = TTCP_CLIENT_SND_WND; 759 tcp_mss(tp, -1); 760 } 761 tp->snd_up = tp->snd_una + so->so_snd.sb_cc; 762 tp->t_flags |= TF_FORCE; 763 error = tcp_output(tp); 764 tp->t_flags &= ~TF_FORCE; 765 } 766 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB : 767 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); 768 } 769 770 /* 771 * Abort the TCP. 772 */ 773 static int 774 tcp_usr_abort(struct socket *so) 775 { 776 int s = splnet(); 777 int error = 0; 778 struct inpcb *inp = so->so_pcb; 779 struct tcpcb *tp; 780 781 COMMON_START(); 782 tp = tcp_drop(tp, ECONNABORTED); 783 COMMON_END(PRU_ABORT); 784 } 785 786 /* 787 * Receive out-of-band data. 788 */ 789 static int 790 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) 791 { 792 int s = splnet(); 793 int error = 0; 794 struct inpcb *inp = so->so_pcb; 795 struct tcpcb *tp; 796 797 COMMON_START(); 798 if ((so->so_oobmark == 0 && 799 (so->so_state & SS_RCVATMARK) == 0) || 800 so->so_options & SO_OOBINLINE || 801 tp->t_oobflags & TCPOOB_HADDATA) { 802 error = EINVAL; 803 goto out; 804 } 805 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 806 error = EWOULDBLOCK; 807 goto out; 808 } 809 m->m_len = 1; 810 *mtod(m, caddr_t) = tp->t_iobc; 811 if ((flags & MSG_PEEK) == 0) 812 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 813 COMMON_END(PRU_RCVOOB); 814 } 815 816 /* xxx - should be const */ 817 struct pr_usrreqs tcp_usrreqs = { 818 tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind, 819 tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach, 820 tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd, 821 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, 822 in_setsockaddr, sosend, soreceive, sopoll 823 }; 824 825 #ifdef INET6 826 struct pr_usrreqs tcp6_usrreqs = { 827 tcp_usr_abort, tcp6_usr_accept, tcp_usr_attach, tcp6_usr_bind, 828 tcp6_usr_connect, pru_connect2_notsupp, in6_control, tcp_usr_detach, 829 tcp_usr_disconnect, tcp6_usr_listen, in6_mapped_peeraddr, tcp_usr_rcvd, 830 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown, 831 in6_mapped_sockaddr, sosend, soreceive, sopoll 832 }; 833 #endif /* INET6 */ 834 835 static int 836 tcp_connect_oncpu(struct tcpcb *tp, struct sockaddr_in *sin, 837 struct sockaddr_in *if_sin) 838 { 839 struct inpcb *inp = tp->t_inpcb, *oinp; 840 struct socket *so = inp->inp_socket; 841 struct tcpcb *otp; 842 struct rmxp_tao *taop; 843 struct rmxp_tao tao_noncached; 844 845 oinp = in_pcblookup_hash(&tcbinfo[mycpu->gd_cpuid], 846 sin->sin_addr, sin->sin_port, 847 inp->inp_laddr.s_addr != INADDR_ANY ? 848 inp->inp_laddr : if_sin->sin_addr, 849 inp->inp_lport, 0, NULL); 850 if (oinp != NULL) { 851 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 852 otp->t_state == TCPS_TIME_WAIT && 853 (ticks - otp->t_starttime) < tcp_msl && 854 (otp->t_flags & TF_RCVD_CC)) 855 tcp_close(otp); 856 else 857 return (EADDRINUSE); 858 } 859 if (inp->inp_laddr.s_addr == INADDR_ANY) 860 inp->inp_laddr = if_sin->sin_addr; 861 inp->inp_faddr = sin->sin_addr; 862 inp->inp_fport = sin->sin_port; 863 inp->inp_cpcbinfo = &tcbinfo[mycpu->gd_cpuid]; 864 in_pcbinsconnhash(inp); 865 866 /* Compute window scaling to request. */ 867 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 868 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 869 tp->request_r_scale++; 870 871 soisconnecting(so); 872 tcpstat.tcps_connattempt++; 873 tp->t_state = TCPS_SYN_SENT; 874 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 875 tp->iss = tcp_new_isn(tp); 876 tcp_sendseqinit(tp); 877 878 /* 879 * Generate a CC value for this connection and 880 * check whether CC or CCnew should be used. 881 */ 882 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) { 883 taop = &tao_noncached; 884 bzero(taop, sizeof *taop); 885 } 886 887 tp->cc_send = CC_INC(tcp_ccgen); 888 if (taop->tao_ccsent != 0 && 889 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 890 taop->tao_ccsent = tp->cc_send; 891 } else { 892 taop->tao_ccsent = 0; 893 tp->t_flags |= TF_SENDCCNEW; 894 } 895 896 return (0); 897 } 898 899 #ifdef SMP 900 901 struct netmsg_tcp_connect { 902 struct lwkt_msg nm_lmsg; 903 struct tcpcb *nm_tp; 904 struct sockaddr_in *nm_sin; 905 struct sockaddr_in *nm_ifsin; 906 }; 907 908 static int 909 tcp_connect_handler(lwkt_msg_t lmsg) 910 { 911 struct netmsg_tcp_connect *msg = (void *)lmsg; 912 int error; 913 914 error = tcp_connect_oncpu(msg->nm_tp, msg->nm_sin, msg->nm_ifsin); 915 lwkt_replymsg(lmsg, error); 916 return(EASYNC); 917 } 918 919 #endif 920 921 /* 922 * Common subroutine to open a TCP connection to remote host specified 923 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local 924 * port number if needed. Call in_pcbladdr to do the routing and to choose 925 * a local host address (interface). If there is an existing incarnation 926 * of the same connection in TIME-WAIT state and if the remote host was 927 * sending CC options and if the connection duration was < MSL, then 928 * truncate the previous TIME-WAIT state and proceed. 929 * Initialize connection parameters and enter SYN-SENT state. 930 */ 931 static int 932 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td) 933 { 934 struct inpcb *inp = tp->t_inpcb; 935 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 936 struct sockaddr_in *if_sin; 937 int error; 938 #ifdef SMP 939 lwkt_port_t port; 940 #endif 941 942 if (inp->inp_lport == 0) { 943 error = in_pcbbind(inp, (struct sockaddr *)NULL, td); 944 if (error) 945 return (error); 946 } 947 948 /* 949 * Cannot simply call in_pcbconnect, because there might be an 950 * earlier incarnation of this same connection still in 951 * TIME_WAIT state, creating an ADDRINUSE error. 952 */ 953 error = in_pcbladdr(inp, nam, &if_sin); 954 if (error) 955 return (error); 956 957 #ifdef SMP 958 port = tcp_addrport(sin->sin_addr.s_addr, sin->sin_port, 959 inp->inp_laddr.s_addr ? 960 inp->inp_laddr.s_addr : if_sin->sin_addr.s_addr, 961 inp->inp_lport); 962 963 if (port->mp_td != curthread) { 964 struct netmsg_tcp_connect msg; 965 966 lwkt_initmsg(&msg.nm_lmsg, &curthread->td_msgport, 0, 967 lwkt_cmd_func(tcp_connect_handler), lwkt_cmd_op_none); 968 msg.nm_tp = tp; 969 msg.nm_sin = sin; 970 msg.nm_ifsin = if_sin; 971 error = lwkt_domsg(port, &msg.nm_lmsg); 972 } else 973 #endif 974 error = tcp_connect_oncpu(tp, sin, if_sin); 975 976 return (error); 977 } 978 979 #ifdef INET6 980 static int 981 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td) 982 { 983 struct inpcb *inp = tp->t_inpcb, *oinp; 984 struct socket *so = inp->inp_socket; 985 struct tcpcb *otp; 986 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; 987 struct in6_addr *addr6; 988 struct rmxp_tao *taop; 989 struct rmxp_tao tao_noncached; 990 int error; 991 992 if (inp->inp_lport == 0) { 993 error = in6_pcbbind(inp, (struct sockaddr *)0, td); 994 if (error) 995 return error; 996 } 997 998 /* 999 * Cannot simply call in_pcbconnect, because there might be an 1000 * earlier incarnation of this same connection still in 1001 * TIME_WAIT state, creating an ADDRINUSE error. 1002 */ 1003 error = in6_pcbladdr(inp, nam, &addr6); 1004 if (error) 1005 return error; 1006 oinp = in6_pcblookup_hash(inp->inp_cpcbinfo, 1007 &sin6->sin6_addr, sin6->sin6_port, 1008 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ? 1009 addr6 : &inp->in6p_laddr, 1010 inp->inp_lport, 0, NULL); 1011 if (oinp) { 1012 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL && 1013 otp->t_state == TCPS_TIME_WAIT && 1014 (ticks - otp->t_starttime) < tcp_msl && 1015 (otp->t_flags & TF_RCVD_CC)) 1016 otp = tcp_close(otp); 1017 else 1018 return (EADDRINUSE); 1019 } 1020 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) 1021 inp->in6p_laddr = *addr6; 1022 inp->in6p_faddr = sin6->sin6_addr; 1023 inp->inp_fport = sin6->sin6_port; 1024 if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != NULL) 1025 inp->in6p_flowinfo = sin6->sin6_flowinfo; 1026 in_pcbinsconnhash(inp); 1027 1028 /* Compute window scaling to request. */ 1029 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 1030 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 1031 tp->request_r_scale++; 1032 1033 soisconnecting(so); 1034 tcpstat.tcps_connattempt++; 1035 tp->t_state = TCPS_SYN_SENT; 1036 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); 1037 tp->iss = tcp_new_isn(tp); 1038 tcp_sendseqinit(tp); 1039 1040 /* 1041 * Generate a CC value for this connection and 1042 * check whether CC or CCnew should be used. 1043 */ 1044 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) { 1045 taop = &tao_noncached; 1046 bzero(taop, sizeof *taop); 1047 } 1048 1049 tp->cc_send = CC_INC(tcp_ccgen); 1050 if (taop->tao_ccsent != 0 && 1051 CC_GEQ(tp->cc_send, taop->tao_ccsent)) { 1052 taop->tao_ccsent = tp->cc_send; 1053 } else { 1054 taop->tao_ccsent = 0; 1055 tp->t_flags |= TF_SENDCCNEW; 1056 } 1057 1058 return (0); 1059 } 1060 #endif /* INET6 */ 1061 1062 /* 1063 * The new sockopt interface makes it possible for us to block in the 1064 * copyin/out step (if we take a page fault). Taking a page fault at 1065 * splnet() is probably a Bad Thing. (Since sockets and pcbs both now 1066 * use TSM, there probably isn't any need for this function to run at 1067 * splnet() any more. This needs more examination.) 1068 */ 1069 int 1070 tcp_ctloutput(so, sopt) 1071 struct socket *so; 1072 struct sockopt *sopt; 1073 { 1074 int error, opt, optval, s; 1075 struct inpcb *inp; 1076 struct tcpcb *tp; 1077 1078 error = 0; 1079 s = splnet(); /* XXX */ 1080 inp = so->so_pcb; 1081 if (inp == NULL) { 1082 splx(s); 1083 return (ECONNRESET); 1084 } 1085 if (sopt->sopt_level != IPPROTO_TCP) { 1086 #ifdef INET6 1087 if (INP_CHECK_SOCKAF(so, AF_INET6)) 1088 error = ip6_ctloutput(so, sopt); 1089 else 1090 #endif /* INET6 */ 1091 error = ip_ctloutput(so, sopt); 1092 splx(s); 1093 return (error); 1094 } 1095 tp = intotcpcb(inp); 1096 1097 switch (sopt->sopt_dir) { 1098 case SOPT_SET: 1099 switch (sopt->sopt_name) { 1100 case TCP_NODELAY: 1101 case TCP_NOOPT: 1102 error = sooptcopyin(sopt, &optval, sizeof optval, 1103 sizeof optval); 1104 if (error) 1105 break; 1106 1107 switch (sopt->sopt_name) { 1108 case TCP_NODELAY: 1109 opt = TF_NODELAY; 1110 break; 1111 case TCP_NOOPT: 1112 opt = TF_NOOPT; 1113 break; 1114 default: 1115 opt = 0; /* dead code to fool gcc */ 1116 break; 1117 } 1118 1119 if (optval) 1120 tp->t_flags |= opt; 1121 else 1122 tp->t_flags &= ~opt; 1123 break; 1124 1125 case TCP_NOPUSH: 1126 error = sooptcopyin(sopt, &optval, sizeof optval, 1127 sizeof optval); 1128 if (error) 1129 break; 1130 1131 if (optval) 1132 tp->t_flags |= TF_NOPUSH; 1133 else { 1134 tp->t_flags &= ~TF_NOPUSH; 1135 error = tcp_output(tp); 1136 } 1137 break; 1138 1139 case TCP_MAXSEG: 1140 error = sooptcopyin(sopt, &optval, sizeof optval, 1141 sizeof optval); 1142 if (error) 1143 break; 1144 1145 if (optval > 0 && optval <= tp->t_maxseg) 1146 tp->t_maxseg = optval; 1147 else 1148 error = EINVAL; 1149 break; 1150 1151 default: 1152 error = ENOPROTOOPT; 1153 break; 1154 } 1155 break; 1156 1157 case SOPT_GET: 1158 switch (sopt->sopt_name) { 1159 case TCP_NODELAY: 1160 optval = tp->t_flags & TF_NODELAY; 1161 break; 1162 case TCP_MAXSEG: 1163 optval = tp->t_maxseg; 1164 break; 1165 case TCP_NOOPT: 1166 optval = tp->t_flags & TF_NOOPT; 1167 break; 1168 case TCP_NOPUSH: 1169 optval = tp->t_flags & TF_NOPUSH; 1170 break; 1171 default: 1172 error = ENOPROTOOPT; 1173 break; 1174 } 1175 if (error == 0) 1176 error = sooptcopyout(sopt, &optval, sizeof optval); 1177 break; 1178 } 1179 splx(s); 1180 return (error); 1181 } 1182 1183 /* 1184 * tcp_sendspace and tcp_recvspace are the default send and receive window 1185 * sizes, respectively. These are obsolescent (this information should 1186 * be set by the route). 1187 */ 1188 u_long tcp_sendspace = 1024*32; 1189 SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW, 1190 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size"); 1191 u_long tcp_recvspace = 57344; /* largest multiple of PAGE_SIZE < 64k */ 1192 SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW, 1193 &tcp_recvspace , 0, "Maximum incoming TCP datagram size"); 1194 1195 /* 1196 * Attach TCP protocol to socket, allocating 1197 * internet protocol control block, tcp control block, 1198 * bufer space, and entering LISTEN state if to accept connections. 1199 */ 1200 static int 1201 tcp_attach(struct socket *so, struct pru_attach_info *ai) 1202 { 1203 struct tcpcb *tp; 1204 struct inpcb *inp; 1205 int error; 1206 int cpu; 1207 #ifdef INET6 1208 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != NULL; 1209 #endif 1210 1211 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 1212 error = soreserve(so, tcp_sendspace, tcp_recvspace, 1213 ai->sb_rlimit); 1214 if (error) 1215 return (error); 1216 } 1217 cpu = mycpu->gd_cpuid; 1218 error = in_pcballoc(so, &tcbinfo[cpu]); 1219 if (error) 1220 return (error); 1221 inp = so->so_pcb; 1222 #ifdef INET6 1223 if (isipv6) { 1224 inp->inp_vflag |= INP_IPV6; 1225 inp->in6p_hops = -1; /* use kernel default */ 1226 } 1227 else 1228 #endif 1229 inp->inp_vflag |= INP_IPV4; 1230 tp = tcp_newtcpcb(inp); 1231 if (tp == 0) { 1232 int nofd = so->so_state & SS_NOFDREF; /* XXX */ 1233 1234 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ 1235 #ifdef INET6 1236 if (isipv6) 1237 in6_pcbdetach(inp); 1238 else 1239 #endif 1240 in_pcbdetach(inp); 1241 so->so_state |= nofd; 1242 return (ENOBUFS); 1243 } 1244 tp->t_state = TCPS_CLOSED; 1245 return (0); 1246 } 1247 1248 /* 1249 * Initiate (or continue) disconnect. 1250 * If embryonic state, just send reset (once). 1251 * If in ``let data drain'' option and linger null, just drop. 1252 * Otherwise (hard), mark socket disconnecting and drop 1253 * current input data; switch states based on user close, and 1254 * send segment to peer (with FIN). 1255 */ 1256 static struct tcpcb * 1257 tcp_disconnect(tp) 1258 struct tcpcb *tp; 1259 { 1260 struct socket *so = tp->t_inpcb->inp_socket; 1261 1262 if (tp->t_state < TCPS_ESTABLISHED) 1263 tp = tcp_close(tp); 1264 else if ((so->so_options & SO_LINGER) && so->so_linger == 0) 1265 tp = tcp_drop(tp, 0); 1266 else { 1267 soisdisconnecting(so); 1268 sbflush(&so->so_rcv); 1269 tp = tcp_usrclosed(tp); 1270 if (tp) 1271 tcp_output(tp); 1272 } 1273 return (tp); 1274 } 1275 1276 /* 1277 * User issued close, and wish to trail through shutdown states: 1278 * if never received SYN, just forget it. If got a SYN from peer, 1279 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 1280 * If already got a FIN from peer, then almost done; go to LAST_ACK 1281 * state. In all other cases, have already sent FIN to peer (e.g. 1282 * after PRU_SHUTDOWN), and just have to play tedious game waiting 1283 * for peer to send FIN or not respond to keep-alives, etc. 1284 * We can let the user exit from the close as soon as the FIN is acked. 1285 */ 1286 static struct tcpcb * 1287 tcp_usrclosed(tp) 1288 struct tcpcb *tp; 1289 { 1290 1291 switch (tp->t_state) { 1292 1293 case TCPS_CLOSED: 1294 case TCPS_LISTEN: 1295 tp->t_state = TCPS_CLOSED; 1296 tp = tcp_close(tp); 1297 break; 1298 1299 case TCPS_SYN_SENT: 1300 case TCPS_SYN_RECEIVED: 1301 tp->t_flags |= TF_NEEDFIN; 1302 break; 1303 1304 case TCPS_ESTABLISHED: 1305 tp->t_state = TCPS_FIN_WAIT_1; 1306 break; 1307 1308 case TCPS_CLOSE_WAIT: 1309 tp->t_state = TCPS_LAST_ACK; 1310 break; 1311 } 1312 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { 1313 soisdisconnected(tp->t_inpcb->inp_socket); 1314 /* To prevent the connection hanging in FIN_WAIT_2 forever. */ 1315 if (tp->t_state == TCPS_FIN_WAIT_2) 1316 callout_reset(tp->tt_2msl, tcp_maxidle, 1317 tcp_timer_2msl, tp); 1318 } 1319 return (tp); 1320 } 1321