tcp_usrreq.c revision 1.113.8.2
1 /* $NetBSD: tcp_usrreq.c,v 1.113.8.2 2006/08/11 15:46:33 yamt Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 /*- 33 * Copyright (c) 1997, 1998, 2005 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Jason R. Thorpe and Kevin M. Lahey of the Numerical Aerospace Simulation 38 * Facility, NASA Ames Research Center. 39 * This code is derived from software contributed to The NetBSD Foundation 40 * by Charles M. Hannum. 41 * 42 * Redistribution and use in source and binary forms, with or without 43 * modification, are permitted provided that the following conditions 44 * are met: 45 * 1. Redistributions of source code must retain the above copyright 46 * notice, this list of conditions and the following disclaimer. 47 * 2. Redistributions in binary form must reproduce the above copyright 48 * notice, this list of conditions and the following disclaimer in the 49 * documentation and/or other materials provided with the distribution. 50 * 3. All advertising materials mentioning features or use of this software 51 * must display the following acknowledgement: 52 * This product includes software developed by the NetBSD 53 * Foundation, Inc. and its contributors. 54 * 4. Neither the name of The NetBSD Foundation nor the names of its 55 * contributors may be used to endorse or promote products derived 56 * from this software without specific prior written permission. 57 * 58 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 59 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 60 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 61 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 62 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 63 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 64 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 65 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 66 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 67 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 68 * POSSIBILITY OF SUCH DAMAGE. 69 */ 70 71 /* 72 * Copyright (c) 1982, 1986, 1988, 1993, 1995 73 * The Regents of the University of California. All rights reserved. 74 * 75 * Redistribution and use in source and binary forms, with or without 76 * modification, are permitted provided that the following conditions 77 * are met: 78 * 1. Redistributions of source code must retain the above copyright 79 * notice, this list of conditions and the following disclaimer. 80 * 2. Redistributions in binary form must reproduce the above copyright 81 * notice, this list of conditions and the following disclaimer in the 82 * documentation and/or other materials provided with the distribution. 83 * 3. Neither the name of the University nor the names of its contributors 84 * may be used to endorse or promote products derived from this software 85 * without specific prior written permission. 86 * 87 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 88 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 89 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 90 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 91 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 92 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 93 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 94 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 95 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 96 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 97 * SUCH DAMAGE. 98 * 99 * @(#)tcp_usrreq.c 8.5 (Berkeley) 6/21/95 100 */ 101 102 #include <sys/cdefs.h> 103 __KERNEL_RCSID(0, "$NetBSD: tcp_usrreq.c,v 1.113.8.2 2006/08/11 15:46:33 yamt Exp $"); 104 105 #include "opt_inet.h" 106 #include "opt_ipsec.h" 107 #include "opt_tcp_debug.h" 108 #include "opt_mbuftrace.h" 109 110 #include <sys/param.h> 111 #include <sys/systm.h> 112 #include <sys/kernel.h> 113 #include <sys/malloc.h> 114 #include <sys/mbuf.h> 115 #include <sys/socket.h> 116 #include <sys/socketvar.h> 117 #include <sys/protosw.h> 118 #include <sys/errno.h> 119 #include <sys/stat.h> 120 #include <sys/proc.h> 121 #include <sys/domain.h> 122 #include <sys/sysctl.h> 123 #include <sys/kauth.h> 124 125 #include <net/if.h> 126 #include <net/route.h> 127 128 #include <netinet/in.h> 129 #include <netinet/in_systm.h> 130 #include <netinet/in_var.h> 131 #include <netinet/ip.h> 132 #include <netinet/in_pcb.h> 133 #include <netinet/ip_var.h> 134 #include <netinet/in_offload.h> 135 136 #ifdef INET6 137 #ifndef INET 138 #include <netinet/in.h> 139 #endif 140 #include <netinet/ip6.h> 141 #include <netinet6/in6_pcb.h> 142 #include <netinet6/ip6_var.h> 143 #endif 144 145 #include <netinet/tcp.h> 146 #include <netinet/tcp_fsm.h> 147 #include <netinet/tcp_seq.h> 148 #include <netinet/tcp_timer.h> 149 #include <netinet/tcp_var.h> 150 #include <netinet/tcpip.h> 151 #include <netinet/tcp_debug.h> 152 153 #include "opt_tcp_space.h" 154 155 #ifdef IPSEC 156 #include <netinet6/ipsec.h> 157 #endif /*IPSEC*/ 158 159 /* 160 * TCP protocol interface to socket abstraction. 161 */ 162 163 /* 164 * Process a TCP user request for TCP tb. If this is a send request 165 * then m is the mbuf chain of send data. If this is a timer expiration 166 * (called from the software clock routine), then timertype tells which timer. 167 */ 168 /*ARGSUSED*/ 169 int 170 tcp_usrreq(struct socket *so, int req, 171 struct mbuf *m, struct mbuf *nam, struct mbuf *control, struct lwp *l) 172 { 173 struct inpcb *inp; 174 #ifdef INET6 175 struct in6pcb *in6p; 176 #endif 177 struct tcpcb *tp = NULL; 178 int s; 179 int error = 0; 180 #ifdef TCP_DEBUG 181 int ostate = 0; 182 #endif 183 int family; /* family of the socket */ 184 185 family = so->so_proto->pr_domain->dom_family; 186 187 if (req == PRU_CONTROL) { 188 switch (family) { 189 #ifdef INET 190 case PF_INET: 191 return (in_control(so, (long)m, (caddr_t)nam, 192 (struct ifnet *)control, l)); 193 #endif 194 #ifdef INET6 195 case PF_INET6: 196 return (in6_control(so, (long)m, (caddr_t)nam, 197 (struct ifnet *)control, l)); 198 #endif 199 default: 200 return EAFNOSUPPORT; 201 } 202 } 203 204 if (req == PRU_PURGEIF) { 205 switch (family) { 206 #ifdef INET 207 case PF_INET: 208 in_pcbpurgeif0(&tcbtable, (struct ifnet *)control); 209 in_purgeif((struct ifnet *)control); 210 in_pcbpurgeif(&tcbtable, (struct ifnet *)control); 211 break; 212 #endif 213 #ifdef INET6 214 case PF_INET6: 215 in6_pcbpurgeif0(&tcbtable, (struct ifnet *)control); 216 in6_purgeif((struct ifnet *)control); 217 in6_pcbpurgeif(&tcbtable, (struct ifnet *)control); 218 break; 219 #endif 220 default: 221 return (EAFNOSUPPORT); 222 } 223 return (0); 224 } 225 226 s = splsoftnet(); 227 switch (family) { 228 #ifdef INET 229 case PF_INET: 230 inp = sotoinpcb(so); 231 #ifdef INET6 232 in6p = NULL; 233 #endif 234 break; 235 #endif 236 #ifdef INET6 237 case PF_INET6: 238 inp = NULL; 239 in6p = sotoin6pcb(so); 240 break; 241 #endif 242 default: 243 splx(s); 244 return EAFNOSUPPORT; 245 } 246 247 #ifdef DIAGNOSTIC 248 #ifdef INET6 249 if (inp && in6p) 250 panic("tcp_usrreq: both inp and in6p set to non-NULL"); 251 #endif 252 if (req != PRU_SEND && req != PRU_SENDOOB && control) 253 panic("tcp_usrreq: unexpected control mbuf"); 254 #endif 255 /* 256 * When a TCP is attached to a socket, then there will be 257 * a (struct inpcb) pointed at by the socket, and this 258 * structure will point at a subsidary (struct tcpcb). 259 */ 260 #ifndef INET6 261 if (inp == 0 && req != PRU_ATTACH) 262 #else 263 if ((inp == 0 && in6p == 0) && req != PRU_ATTACH) 264 #endif 265 { 266 error = EINVAL; 267 goto release; 268 } 269 #ifdef INET 270 if (inp) { 271 tp = intotcpcb(inp); 272 /* WHAT IF TP IS 0? */ 273 #ifdef KPROF 274 tcp_acounts[tp->t_state][req]++; 275 #endif 276 #ifdef TCP_DEBUG 277 ostate = tp->t_state; 278 #endif 279 } 280 #endif 281 #ifdef INET6 282 if (in6p) { 283 tp = in6totcpcb(in6p); 284 /* WHAT IF TP IS 0? */ 285 #ifdef KPROF 286 tcp_acounts[tp->t_state][req]++; 287 #endif 288 #ifdef TCP_DEBUG 289 ostate = tp->t_state; 290 #endif 291 } 292 #endif 293 294 switch (req) { 295 296 /* 297 * TCP attaches to socket via PRU_ATTACH, reserving space, 298 * and an internet control block. 299 */ 300 case PRU_ATTACH: 301 #ifndef INET6 302 if (inp != 0) 303 #else 304 if (inp != 0 || in6p != 0) 305 #endif 306 { 307 error = EISCONN; 308 break; 309 } 310 error = tcp_attach(so); 311 if (error) 312 break; 313 if ((so->so_options & SO_LINGER) && so->so_linger == 0) 314 so->so_linger = TCP_LINGERTIME; 315 tp = sototcpcb(so); 316 break; 317 318 /* 319 * PRU_DETACH detaches the TCP protocol from the socket. 320 */ 321 case PRU_DETACH: 322 tp = tcp_disconnect(tp); 323 break; 324 325 /* 326 * Give the socket an address. 327 */ 328 case PRU_BIND: 329 switch (family) { 330 #ifdef INET 331 case PF_INET: 332 error = in_pcbbind(inp, nam, l); 333 break; 334 #endif 335 #ifdef INET6 336 case PF_INET6: 337 error = in6_pcbbind(in6p, nam, l); 338 if (!error) { 339 /* mapped addr case */ 340 if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)) 341 tp->t_family = AF_INET; 342 else 343 tp->t_family = AF_INET6; 344 } 345 break; 346 #endif 347 } 348 break; 349 350 /* 351 * Prepare to accept connections. 352 */ 353 case PRU_LISTEN: 354 #ifdef INET 355 if (inp && inp->inp_lport == 0) { 356 error = in_pcbbind(inp, (struct mbuf *)0, 357 (struct lwp *)0); 358 if (error) 359 break; 360 } 361 #endif 362 #ifdef INET6 363 if (in6p && in6p->in6p_lport == 0) { 364 error = in6_pcbbind(in6p, (struct mbuf *)0, 365 (struct lwp *)0); 366 if (error) 367 break; 368 } 369 #endif 370 tp->t_state = TCPS_LISTEN; 371 break; 372 373 /* 374 * Initiate connection to peer. 375 * Create a template for use in transmissions on this connection. 376 * Enter SYN_SENT state, and mark socket as connecting. 377 * Start keep-alive timer, and seed output sequence space. 378 * Send initial segment on connection. 379 */ 380 case PRU_CONNECT: 381 #ifdef INET 382 if (inp) { 383 if (inp->inp_lport == 0) { 384 error = in_pcbbind(inp, (struct mbuf *)0, 385 (struct lwp *)0); 386 if (error) 387 break; 388 } 389 error = in_pcbconnect(inp, nam, l); 390 } 391 #endif 392 #ifdef INET6 393 if (in6p) { 394 if (in6p->in6p_lport == 0) { 395 error = in6_pcbbind(in6p, (struct mbuf *)0, 396 (struct lwp *)0); 397 if (error) 398 break; 399 } 400 error = in6_pcbconnect(in6p, nam, l); 401 if (!error) { 402 /* mapped addr case */ 403 if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) 404 tp->t_family = AF_INET; 405 else 406 tp->t_family = AF_INET6; 407 } 408 } 409 #endif 410 if (error) 411 break; 412 tp->t_template = tcp_template(tp); 413 if (tp->t_template == 0) { 414 #ifdef INET 415 if (inp) 416 in_pcbdisconnect(inp); 417 #endif 418 #ifdef INET6 419 if (in6p) 420 in6_pcbdisconnect(in6p); 421 #endif 422 error = ENOBUFS; 423 break; 424 } 425 /* Compute window scaling to request. */ 426 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 427 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 428 tp->request_r_scale++; 429 soisconnecting(so); 430 tcpstat.tcps_connattempt++; 431 tp->t_state = TCPS_SYN_SENT; 432 TCP_TIMER_ARM(tp, TCPT_KEEP, TCPTV_KEEP_INIT); 433 tp->iss = tcp_new_iss(tp, 0); 434 tcp_sendseqinit(tp); 435 error = tcp_output(tp); 436 break; 437 438 /* 439 * Create a TCP connection between two sockets. 440 */ 441 case PRU_CONNECT2: 442 error = EOPNOTSUPP; 443 break; 444 445 /* 446 * Initiate disconnect from peer. 447 * If connection never passed embryonic stage, just drop; 448 * else if don't need to let data drain, then can just drop anyways, 449 * else have to begin TCP shutdown process: mark socket disconnecting, 450 * drain unread data, state switch to reflect user close, and 451 * send segment (e.g. FIN) to peer. Socket will be really disconnected 452 * when peer sends FIN and acks ours. 453 * 454 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 455 */ 456 case PRU_DISCONNECT: 457 tp = tcp_disconnect(tp); 458 break; 459 460 /* 461 * Accept a connection. Essentially all the work is 462 * done at higher levels; just return the address 463 * of the peer, storing through addr. 464 */ 465 case PRU_ACCEPT: 466 #ifdef INET 467 if (inp) 468 in_setpeeraddr(inp, nam); 469 #endif 470 #ifdef INET6 471 if (in6p) 472 in6_setpeeraddr(in6p, nam); 473 #endif 474 break; 475 476 /* 477 * Mark the connection as being incapable of further output. 478 */ 479 case PRU_SHUTDOWN: 480 socantsendmore(so); 481 tp = tcp_usrclosed(tp); 482 if (tp) 483 error = tcp_output(tp); 484 break; 485 486 /* 487 * After a receive, possibly send window update to peer. 488 */ 489 case PRU_RCVD: 490 /* 491 * soreceive() calls this function when a user receives 492 * ancillary data on a listening socket. We don't call 493 * tcp_output in such a case, since there is no header 494 * template for a listening socket and hence the kernel 495 * will panic. 496 */ 497 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) != 0) 498 (void) tcp_output(tp); 499 break; 500 501 /* 502 * Do a send by putting data in output queue and updating urgent 503 * marker if URG set. Possibly send more data. 504 */ 505 case PRU_SEND: 506 if (control && control->m_len) { 507 m_freem(control); 508 m_freem(m); 509 error = EINVAL; 510 break; 511 } 512 sbappendstream(&so->so_snd, m); 513 error = tcp_output(tp); 514 break; 515 516 /* 517 * Abort the TCP. 518 */ 519 case PRU_ABORT: 520 tp = tcp_drop(tp, ECONNABORTED); 521 break; 522 523 case PRU_SENSE: 524 /* 525 * stat: don't bother with a blocksize. 526 */ 527 splx(s); 528 return (0); 529 530 case PRU_RCVOOB: 531 if (control && control->m_len) { 532 m_freem(control); 533 m_freem(m); 534 error = EINVAL; 535 break; 536 } 537 if ((so->so_oobmark == 0 && 538 (so->so_state & SS_RCVATMARK) == 0) || 539 so->so_options & SO_OOBINLINE || 540 tp->t_oobflags & TCPOOB_HADDATA) { 541 error = EINVAL; 542 break; 543 } 544 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 545 error = EWOULDBLOCK; 546 break; 547 } 548 m->m_len = 1; 549 *mtod(m, caddr_t) = tp->t_iobc; 550 if (((long)nam & MSG_PEEK) == 0) 551 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 552 break; 553 554 case PRU_SENDOOB: 555 if (sbspace(&so->so_snd) < -512) { 556 m_freem(m); 557 error = ENOBUFS; 558 break; 559 } 560 /* 561 * According to RFC961 (Assigned Protocols), 562 * the urgent pointer points to the last octet 563 * of urgent data. We continue, however, 564 * to consider it to indicate the first octet 565 * of data past the urgent section. 566 * Otherwise, snd_up should be one lower. 567 */ 568 sbappendstream(&so->so_snd, m); 569 tp->snd_up = tp->snd_una + so->so_snd.sb_cc; 570 tp->t_force = 1; 571 error = tcp_output(tp); 572 tp->t_force = 0; 573 break; 574 575 case PRU_SOCKADDR: 576 #ifdef INET 577 if (inp) 578 in_setsockaddr(inp, nam); 579 #endif 580 #ifdef INET6 581 if (in6p) 582 in6_setsockaddr(in6p, nam); 583 #endif 584 break; 585 586 case PRU_PEERADDR: 587 #ifdef INET 588 if (inp) 589 in_setpeeraddr(inp, nam); 590 #endif 591 #ifdef INET6 592 if (in6p) 593 in6_setpeeraddr(in6p, nam); 594 #endif 595 break; 596 597 default: 598 panic("tcp_usrreq"); 599 } 600 #ifdef TCP_DEBUG 601 if (tp && (so->so_options & SO_DEBUG)) 602 tcp_trace(TA_USER, ostate, tp, NULL, req); 603 #endif 604 605 release: 606 splx(s); 607 return (error); 608 } 609 610 int 611 tcp_ctloutput(int op, struct socket *so, int level, int optname, 612 struct mbuf **mp) 613 { 614 int error = 0, s; 615 struct inpcb *inp; 616 #ifdef INET6 617 struct in6pcb *in6p; 618 #endif 619 struct tcpcb *tp; 620 struct mbuf *m; 621 int i; 622 int family; /* family of the socket */ 623 624 family = so->so_proto->pr_domain->dom_family; 625 626 s = splsoftnet(); 627 switch (family) { 628 #ifdef INET 629 case PF_INET: 630 inp = sotoinpcb(so); 631 #ifdef INET6 632 in6p = NULL; 633 #endif 634 break; 635 #endif 636 #ifdef INET6 637 case PF_INET6: 638 inp = NULL; 639 in6p = sotoin6pcb(so); 640 break; 641 #endif 642 default: 643 splx(s); 644 return EAFNOSUPPORT; 645 } 646 #ifndef INET6 647 if (inp == NULL) 648 #else 649 if (inp == NULL && in6p == NULL) 650 #endif 651 { 652 splx(s); 653 if (op == PRCO_SETOPT && *mp) 654 (void) m_free(*mp); 655 return (ECONNRESET); 656 } 657 if (level != IPPROTO_TCP) { 658 switch (family) { 659 #ifdef INET 660 case PF_INET: 661 error = ip_ctloutput(op, so, level, optname, mp); 662 break; 663 #endif 664 #ifdef INET6 665 case PF_INET6: 666 error = ip6_ctloutput(op, so, level, optname, mp); 667 break; 668 #endif 669 } 670 splx(s); 671 return (error); 672 } 673 if (inp) 674 tp = intotcpcb(inp); 675 #ifdef INET6 676 else if (in6p) 677 tp = in6totcpcb(in6p); 678 #endif 679 else 680 tp = NULL; 681 682 switch (op) { 683 684 case PRCO_SETOPT: 685 m = *mp; 686 switch (optname) { 687 688 #ifdef TCP_SIGNATURE 689 case TCP_MD5SIG: 690 if (m == NULL || m->m_len < sizeof (int)) 691 error = EINVAL; 692 if (error) 693 break; 694 if (*mtod(m, int *) > 0) 695 tp->t_flags |= TF_SIGNATURE; 696 else 697 tp->t_flags &= ~TF_SIGNATURE; 698 break; 699 #endif /* TCP_SIGNATURE */ 700 701 case TCP_NODELAY: 702 if (m == NULL || m->m_len < sizeof (int)) 703 error = EINVAL; 704 else if (*mtod(m, int *)) 705 tp->t_flags |= TF_NODELAY; 706 else 707 tp->t_flags &= ~TF_NODELAY; 708 break; 709 710 case TCP_MAXSEG: 711 if (m && (i = *mtod(m, int *)) > 0 && 712 i <= tp->t_peermss) 713 tp->t_peermss = i; /* limit on send size */ 714 else 715 error = EINVAL; 716 break; 717 718 default: 719 error = ENOPROTOOPT; 720 break; 721 } 722 if (m) 723 (void) m_free(m); 724 break; 725 726 case PRCO_GETOPT: 727 *mp = m = m_get(M_WAIT, MT_SOOPTS); 728 m->m_len = sizeof(int); 729 MCLAIM(m, so->so_mowner); 730 731 switch (optname) { 732 #ifdef TCP_SIGNATURE 733 case TCP_MD5SIG: 734 *mtod(m, int *) = (tp->t_flags & TF_SIGNATURE) ? 1 : 0; 735 break; 736 #endif 737 case TCP_NODELAY: 738 *mtod(m, int *) = tp->t_flags & TF_NODELAY; 739 break; 740 case TCP_MAXSEG: 741 *mtod(m, int *) = tp->t_peermss; 742 break; 743 default: 744 error = ENOPROTOOPT; 745 break; 746 } 747 break; 748 } 749 splx(s); 750 return (error); 751 } 752 753 #ifndef TCP_SENDSPACE 754 #define TCP_SENDSPACE 1024*32 755 #endif 756 int tcp_sendspace = TCP_SENDSPACE; 757 #ifndef TCP_RECVSPACE 758 #define TCP_RECVSPACE 1024*32 759 #endif 760 int tcp_recvspace = TCP_RECVSPACE; 761 762 /* 763 * Attach TCP protocol to socket, allocating 764 * internet protocol control block, tcp control block, 765 * bufer space, and entering LISTEN state if to accept connections. 766 */ 767 int 768 tcp_attach(struct socket *so) 769 { 770 struct tcpcb *tp; 771 struct inpcb *inp; 772 #ifdef INET6 773 struct in6pcb *in6p; 774 #endif 775 int error; 776 int family; /* family of the socket */ 777 778 family = so->so_proto->pr_domain->dom_family; 779 780 #ifdef MBUFTRACE 781 so->so_mowner = &tcp_mowner; 782 so->so_rcv.sb_mowner = &tcp_rx_mowner; 783 so->so_snd.sb_mowner = &tcp_tx_mowner; 784 #endif 785 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 786 error = soreserve(so, tcp_sendspace, tcp_recvspace); 787 if (error) 788 return (error); 789 } 790 switch (family) { 791 #ifdef INET 792 case PF_INET: 793 error = in_pcballoc(so, &tcbtable); 794 if (error) 795 return (error); 796 inp = sotoinpcb(so); 797 #ifdef INET6 798 in6p = NULL; 799 #endif 800 break; 801 #endif 802 #ifdef INET6 803 case PF_INET6: 804 error = in6_pcballoc(so, &tcbtable); 805 if (error) 806 return (error); 807 inp = NULL; 808 in6p = sotoin6pcb(so); 809 break; 810 #endif 811 default: 812 return EAFNOSUPPORT; 813 } 814 if (inp) 815 tp = tcp_newtcpcb(family, (void *)inp); 816 #ifdef INET6 817 else if (in6p) 818 tp = tcp_newtcpcb(family, (void *)in6p); 819 #endif 820 else 821 tp = NULL; 822 823 if (tp == 0) { 824 int nofd = so->so_state & SS_NOFDREF; /* XXX */ 825 826 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ 827 #ifdef INET 828 if (inp) 829 in_pcbdetach(inp); 830 #endif 831 #ifdef INET6 832 if (in6p) 833 in6_pcbdetach(in6p); 834 #endif 835 so->so_state |= nofd; 836 return (ENOBUFS); 837 } 838 tp->t_state = TCPS_CLOSED; 839 return (0); 840 } 841 842 /* 843 * Initiate (or continue) disconnect. 844 * If embryonic state, just send reset (once). 845 * If in ``let data drain'' option and linger null, just drop. 846 * Otherwise (hard), mark socket disconnecting and drop 847 * current input data; switch states based on user close, and 848 * send segment to peer (with FIN). 849 */ 850 struct tcpcb * 851 tcp_disconnect(struct tcpcb *tp) 852 { 853 struct socket *so; 854 855 if (tp->t_inpcb) 856 so = tp->t_inpcb->inp_socket; 857 #ifdef INET6 858 else if (tp->t_in6pcb) 859 so = tp->t_in6pcb->in6p_socket; 860 #endif 861 else 862 so = NULL; 863 864 if (TCPS_HAVEESTABLISHED(tp->t_state) == 0) 865 tp = tcp_close(tp); 866 else if ((so->so_options & SO_LINGER) && so->so_linger == 0) 867 tp = tcp_drop(tp, 0); 868 else { 869 soisdisconnecting(so); 870 sbflush(&so->so_rcv); 871 tp = tcp_usrclosed(tp); 872 if (tp) 873 (void) tcp_output(tp); 874 } 875 return (tp); 876 } 877 878 /* 879 * User issued close, and wish to trail through shutdown states: 880 * if never received SYN, just forget it. If got a SYN from peer, 881 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 882 * If already got a FIN from peer, then almost done; go to LAST_ACK 883 * state. In all other cases, have already sent FIN to peer (e.g. 884 * after PRU_SHUTDOWN), and just have to play tedious game waiting 885 * for peer to send FIN or not respond to keep-alives, etc. 886 * We can let the user exit from the close as soon as the FIN is acked. 887 */ 888 struct tcpcb * 889 tcp_usrclosed(struct tcpcb *tp) 890 { 891 892 switch (tp->t_state) { 893 894 case TCPS_CLOSED: 895 case TCPS_LISTEN: 896 case TCPS_SYN_SENT: 897 tp->t_state = TCPS_CLOSED; 898 tp = tcp_close(tp); 899 break; 900 901 case TCPS_SYN_RECEIVED: 902 case TCPS_ESTABLISHED: 903 tp->t_state = TCPS_FIN_WAIT_1; 904 break; 905 906 case TCPS_CLOSE_WAIT: 907 tp->t_state = TCPS_LAST_ACK; 908 break; 909 } 910 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { 911 struct socket *so; 912 if (tp->t_inpcb) 913 so = tp->t_inpcb->inp_socket; 914 #ifdef INET6 915 else if (tp->t_in6pcb) 916 so = tp->t_in6pcb->in6p_socket; 917 #endif 918 else 919 so = NULL; 920 if (so) 921 soisdisconnected(so); 922 /* 923 * If we are in FIN_WAIT_2, we arrived here because the 924 * application did a shutdown of the send side. Like the 925 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after 926 * a full close, we start a timer to make sure sockets are 927 * not left in FIN_WAIT_2 forever. 928 */ 929 if ((tp->t_state == TCPS_FIN_WAIT_2) && (tcp_maxidle > 0)) 930 TCP_TIMER_ARM(tp, TCPT_2MSL, tcp_maxidle); 931 } 932 return (tp); 933 } 934 935 /* 936 * sysctl helper routine for net.inet.ip.mssdflt. it can't be less 937 * than 32. 938 */ 939 static int 940 sysctl_net_inet_tcp_mssdflt(SYSCTLFN_ARGS) 941 { 942 int error, mssdflt; 943 struct sysctlnode node; 944 945 mssdflt = tcp_mssdflt; 946 node = *rnode; 947 node.sysctl_data = &mssdflt; 948 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 949 if (error || newp == NULL) 950 return (error); 951 952 if (mssdflt < 32) 953 return (EINVAL); 954 tcp_mssdflt = mssdflt; 955 956 return (0); 957 } 958 959 /* 960 * sysctl helper routine for setting port related values under 961 * net.inet.ip and net.inet6.ip6. does basic range checking and does 962 * additional checks for each type. this code has placed in 963 * tcp_input.c since INET and INET6 both use the same tcp code. 964 * 965 * this helper is not static so that both inet and inet6 can use it. 966 */ 967 int 968 sysctl_net_inet_ip_ports(SYSCTLFN_ARGS) 969 { 970 int error, tmp; 971 int apmin, apmax; 972 #ifndef IPNOPRIVPORTS 973 int lpmin, lpmax; 974 #endif /* IPNOPRIVPORTS */ 975 struct sysctlnode node; 976 977 if (namelen != 0) 978 return (EINVAL); 979 980 switch (name[-3]) { 981 #ifdef INET 982 case PF_INET: 983 apmin = anonportmin; 984 apmax = anonportmax; 985 #ifndef IPNOPRIVPORTS 986 lpmin = lowportmin; 987 lpmax = lowportmax; 988 #endif /* IPNOPRIVPORTS */ 989 break; 990 #endif /* INET */ 991 #ifdef INET6 992 case PF_INET6: 993 apmin = ip6_anonportmin; 994 apmax = ip6_anonportmax; 995 #ifndef IPNOPRIVPORTS 996 lpmin = ip6_lowportmin; 997 lpmax = ip6_lowportmax; 998 #endif /* IPNOPRIVPORTS */ 999 break; 1000 #endif /* INET6 */ 1001 default: 1002 return (EINVAL); 1003 } 1004 1005 /* 1006 * insert temporary copy into node, perform lookup on 1007 * temporary, then restore pointer 1008 */ 1009 node = *rnode; 1010 tmp = *(int*)rnode->sysctl_data; 1011 node.sysctl_data = &tmp; 1012 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1013 if (error || newp == NULL) 1014 return (error); 1015 1016 /* 1017 * simple port range check 1018 */ 1019 if (tmp < 0 || tmp > 65535) 1020 return (EINVAL); 1021 1022 /* 1023 * per-node range checks 1024 */ 1025 switch (rnode->sysctl_num) { 1026 case IPCTL_ANONPORTMIN: 1027 if (tmp >= apmax) 1028 return (EINVAL); 1029 #ifndef IPNOPRIVPORTS 1030 if (tmp < IPPORT_RESERVED) 1031 return (EINVAL); 1032 #endif /* IPNOPRIVPORTS */ 1033 break; 1034 1035 case IPCTL_ANONPORTMAX: 1036 if (apmin >= tmp) 1037 return (EINVAL); 1038 #ifndef IPNOPRIVPORTS 1039 if (tmp < IPPORT_RESERVED) 1040 return (EINVAL); 1041 #endif /* IPNOPRIVPORTS */ 1042 break; 1043 1044 #ifndef IPNOPRIVPORTS 1045 case IPCTL_LOWPORTMIN: 1046 if (tmp >= lpmax || 1047 tmp > IPPORT_RESERVEDMAX || 1048 tmp < IPPORT_RESERVEDMIN) 1049 return (EINVAL); 1050 break; 1051 1052 case IPCTL_LOWPORTMAX: 1053 if (lpmin >= tmp || 1054 tmp > IPPORT_RESERVEDMAX || 1055 tmp < IPPORT_RESERVEDMIN) 1056 return (EINVAL); 1057 break; 1058 #endif /* IPNOPRIVPORTS */ 1059 1060 default: 1061 return (EINVAL); 1062 } 1063 1064 *(int*)rnode->sysctl_data = tmp; 1065 1066 return (0); 1067 } 1068 1069 /* 1070 * sysctl helper routine for the net.inet.tcp.ident and 1071 * net.inet6.tcp6.ident nodes. contains backwards compat code for the 1072 * old way of looking up the ident information for ipv4 which involves 1073 * stuffing the port/addr pairs into the mib lookup. 1074 */ 1075 static int 1076 sysctl_net_inet_tcp_ident(SYSCTLFN_ARGS) 1077 { 1078 #ifdef INET 1079 struct inpcb *inb; 1080 struct sockaddr_in *si4[2]; 1081 #endif /* INET */ 1082 #ifdef INET6 1083 struct in6pcb *in6b; 1084 struct sockaddr_in6 *si6[2]; 1085 #endif /* INET6 */ 1086 struct sockaddr_storage sa[2]; 1087 struct socket *sockp; 1088 size_t sz; 1089 uid_t uid; 1090 int error, pf; 1091 1092 if (namelen != 4 && namelen != 0) 1093 return (EINVAL); 1094 if (name[-2] != IPPROTO_TCP) 1095 return (EINVAL); 1096 pf = name[-3]; 1097 1098 /* old style lookup, ipv4 only */ 1099 if (namelen == 4) { 1100 #ifdef INET 1101 struct in_addr laddr, raddr; 1102 u_int lport, rport; 1103 1104 if (pf != PF_INET) 1105 return (EPROTONOSUPPORT); 1106 raddr.s_addr = (uint32_t)name[0]; 1107 rport = (u_int)name[1]; 1108 laddr.s_addr = (uint32_t)name[2]; 1109 lport = (u_int)name[3]; 1110 inb = in_pcblookup_connect(&tcbtable, raddr, rport, 1111 laddr, lport); 1112 if (inb == NULL || (sockp = inb->inp_socket) == NULL) 1113 return (ESRCH); 1114 uid = sockp->so_uidinfo->ui_uid; 1115 if (oldp) { 1116 sz = MIN(sizeof(uid), *oldlenp); 1117 error = copyout(&uid, oldp, sz); 1118 if (error) 1119 return (error); 1120 } 1121 *oldlenp = sizeof(uid); 1122 return (0); 1123 #else /* INET */ 1124 return (EINVAL); 1125 #endif /* INET */ 1126 } 1127 1128 if (newp == NULL || newlen != sizeof(sa)) 1129 return (EINVAL); 1130 error = copyin(newp, &sa, newlen); 1131 if (error) 1132 return (error); 1133 1134 /* 1135 * requested families must match 1136 */ 1137 if (pf != sa[0].ss_family || sa[0].ss_family != sa[1].ss_family) 1138 return (EINVAL); 1139 1140 switch (pf) { 1141 #ifdef INET 1142 case PF_INET: 1143 si4[0] = (struct sockaddr_in*)&sa[0]; 1144 si4[1] = (struct sockaddr_in*)&sa[1]; 1145 if (si4[0]->sin_len != sizeof(*si4[0]) || 1146 si4[0]->sin_len != si4[1]->sin_len) 1147 return (EINVAL); 1148 inb = in_pcblookup_connect(&tcbtable, 1149 si4[0]->sin_addr, si4[0]->sin_port, 1150 si4[1]->sin_addr, si4[1]->sin_port); 1151 if (inb == NULL || (sockp = inb->inp_socket) == NULL) 1152 return (ESRCH); 1153 break; 1154 #endif /* INET */ 1155 #ifdef INET6 1156 case PF_INET6: 1157 si6[0] = (struct sockaddr_in6*)&sa[0]; 1158 si6[1] = (struct sockaddr_in6*)&sa[1]; 1159 if (si6[0]->sin6_len != sizeof(*si6[0]) || 1160 si6[0]->sin6_len != si6[1]->sin6_len) 1161 return (EINVAL); 1162 in6b = in6_pcblookup_connect(&tcbtable, 1163 &si6[0]->sin6_addr, si6[0]->sin6_port, 1164 &si6[1]->sin6_addr, si6[1]->sin6_port, 0); 1165 if (in6b == NULL || (sockp = in6b->in6p_socket) == NULL) 1166 return (ESRCH); 1167 break; 1168 #endif /* INET6 */ 1169 default: 1170 return (EPROTONOSUPPORT); 1171 } 1172 *oldlenp = sizeof(uid); 1173 1174 uid = sockp->so_uidinfo->ui_uid; 1175 if (oldp) { 1176 sz = MIN(sizeof(uid), *oldlenp); 1177 error = copyout(&uid, oldp, sz); 1178 if (error) 1179 return (error); 1180 } 1181 *oldlenp = sizeof(uid); 1182 1183 return (0); 1184 } 1185 1186 /* 1187 * sysctl helper for the inet and inet6 pcblists. handles tcp/udp and 1188 * inet/inet6, as well as raw pcbs for each. specifically not 1189 * declared static so that raw sockets and udp/udp6 can use it as 1190 * well. 1191 */ 1192 int 1193 sysctl_inpcblist(SYSCTLFN_ARGS) 1194 { 1195 #ifdef INET 1196 struct sockaddr_in *in; 1197 const struct inpcb *inp; 1198 #endif 1199 #ifdef INET6 1200 struct sockaddr_in6 *in6; 1201 const struct in6pcb *in6p; 1202 #endif 1203 /* 1204 * sysctl_data is const, but CIRCLEQ_FOREACH can't use a const 1205 * struct inpcbtable pointer, so we have to discard const. :-/ 1206 */ 1207 struct inpcbtable *pcbtbl = __UNCONST(rnode->sysctl_data); 1208 const struct inpcb_hdr *inph; 1209 struct tcpcb *tp; 1210 struct kinfo_pcb pcb; 1211 char *dp; 1212 u_int op, arg; 1213 size_t len, needed, elem_size, out_size; 1214 int error, elem_count, pf, proto, pf2; 1215 1216 if (namelen != 4) 1217 return (EINVAL); 1218 1219 if (oldp != NULL) { 1220 len = *oldlenp; 1221 elem_size = name[2]; 1222 elem_count = name[3]; 1223 if (elem_size != sizeof(pcb)) 1224 return EINVAL; 1225 } else { 1226 len = 0; 1227 elem_count = INT_MAX; 1228 elem_size = sizeof(pcb); 1229 } 1230 error = 0; 1231 dp = oldp; 1232 op = name[0]; 1233 arg = name[1]; 1234 out_size = elem_size; 1235 needed = 0; 1236 1237 if (namelen == 1 && name[0] == CTL_QUERY) 1238 return (sysctl_query(SYSCTLFN_CALL(rnode))); 1239 1240 if (name - oname != 4) 1241 return (EINVAL); 1242 1243 pf = oname[1]; 1244 proto = oname[2]; 1245 pf2 = (oldp != NULL) ? pf : 0; 1246 1247 CIRCLEQ_FOREACH(inph, &pcbtbl->inpt_queue, inph_queue) { 1248 #ifdef INET 1249 inp = (const struct inpcb *)inph; 1250 #endif 1251 #ifdef INET6 1252 in6p = (const struct in6pcb *)inph; 1253 #endif 1254 1255 if (inph->inph_af != pf) 1256 continue; 1257 1258 /* XXX elad - should be done better */ 1259 if (security_curtain && 1260 (kauth_cred_geteuid(l->l_cred) != 0) && 1261 (kauth_cred_geteuid(l->l_cred) != 1262 inph->inph_socket->so_uidinfo->ui_uid)) 1263 continue; 1264 1265 memset(&pcb, 0, sizeof(pcb)); 1266 1267 pcb.ki_family = pf; 1268 pcb.ki_type = proto; 1269 1270 switch (pf2) { 1271 case 0: 1272 /* just probing for size */ 1273 break; 1274 #ifdef INET 1275 case PF_INET: 1276 pcb.ki_family = inp->inp_socket->so_proto-> 1277 pr_domain->dom_family; 1278 pcb.ki_type = inp->inp_socket->so_proto-> 1279 pr_type; 1280 pcb.ki_protocol = inp->inp_socket->so_proto-> 1281 pr_protocol; 1282 pcb.ki_pflags = inp->inp_flags; 1283 1284 pcb.ki_sostate = inp->inp_socket->so_state; 1285 pcb.ki_prstate = inp->inp_state; 1286 if (proto == IPPROTO_TCP) { 1287 tp = intotcpcb(inp); 1288 pcb.ki_tstate = tp->t_state; 1289 pcb.ki_tflags = tp->t_flags; 1290 } 1291 1292 pcb.ki_pcbaddr = PTRTOUINT64(inp); 1293 pcb.ki_ppcbaddr = PTRTOUINT64(inp->inp_ppcb); 1294 pcb.ki_sockaddr = PTRTOUINT64(inp->inp_socket); 1295 1296 pcb.ki_rcvq = inp->inp_socket->so_rcv.sb_cc; 1297 pcb.ki_sndq = inp->inp_socket->so_snd.sb_cc; 1298 1299 in = satosin(&pcb.ki_src); 1300 in->sin_len = sizeof(*in); 1301 in->sin_family = pf; 1302 in->sin_port = inp->inp_lport; 1303 in->sin_addr = inp->inp_laddr; 1304 if (pcb.ki_prstate >= INP_CONNECTED) { 1305 in = satosin(&pcb.ki_dst); 1306 in->sin_len = sizeof(*in); 1307 in->sin_family = pf; 1308 in->sin_port = inp->inp_fport; 1309 in->sin_addr = inp->inp_faddr; 1310 } 1311 break; 1312 #endif 1313 #ifdef INET6 1314 case PF_INET6: 1315 pcb.ki_family = in6p->in6p_socket->so_proto-> 1316 pr_domain->dom_family; 1317 pcb.ki_type = in6p->in6p_socket->so_proto->pr_type; 1318 pcb.ki_protocol = in6p->in6p_socket->so_proto-> 1319 pr_protocol; 1320 pcb.ki_pflags = in6p->in6p_flags; 1321 1322 pcb.ki_sostate = in6p->in6p_socket->so_state; 1323 pcb.ki_prstate = in6p->in6p_state; 1324 if (proto == IPPROTO_TCP) { 1325 tp = in6totcpcb(in6p); 1326 pcb.ki_tstate = tp->t_state; 1327 pcb.ki_tflags = tp->t_flags; 1328 } 1329 1330 pcb.ki_pcbaddr = PTRTOUINT64(in6p); 1331 pcb.ki_ppcbaddr = PTRTOUINT64(in6p->in6p_ppcb); 1332 pcb.ki_sockaddr = PTRTOUINT64(in6p->in6p_socket); 1333 1334 pcb.ki_rcvq = in6p->in6p_socket->so_rcv.sb_cc; 1335 pcb.ki_sndq = in6p->in6p_socket->so_snd.sb_cc; 1336 1337 in6 = satosin6(&pcb.ki_src); 1338 in6->sin6_len = sizeof(*in6); 1339 in6->sin6_family = pf; 1340 in6->sin6_port = in6p->in6p_lport; 1341 in6->sin6_flowinfo = in6p->in6p_flowinfo; 1342 in6->sin6_addr = in6p->in6p_laddr; 1343 in6->sin6_scope_id = 0; /* XXX? */ 1344 1345 if (pcb.ki_prstate >= IN6P_CONNECTED) { 1346 in6 = satosin6(&pcb.ki_dst); 1347 in6->sin6_len = sizeof(*in6); 1348 in6->sin6_family = pf; 1349 in6->sin6_port = in6p->in6p_fport; 1350 in6->sin6_flowinfo = in6p->in6p_flowinfo; 1351 in6->sin6_addr = in6p->in6p_faddr; 1352 in6->sin6_scope_id = 0; /* XXX? */ 1353 } 1354 break; 1355 #endif 1356 } 1357 1358 if (len >= elem_size && elem_count > 0) { 1359 error = copyout(&pcb, dp, out_size); 1360 if (error) 1361 return (error); 1362 dp += elem_size; 1363 len -= elem_size; 1364 } 1365 if (elem_count > 0) { 1366 needed += elem_size; 1367 if (elem_count != INT_MAX) 1368 elem_count--; 1369 } 1370 } 1371 1372 *oldlenp = needed; 1373 if (oldp == NULL) 1374 *oldlenp += PCB_SLOP * sizeof(struct kinfo_pcb); 1375 1376 return (error); 1377 } 1378 1379 /* 1380 * this (second stage) setup routine is a replacement for tcp_sysctl() 1381 * (which is currently used for ipv4 and ipv6) 1382 */ 1383 static void 1384 sysctl_net_inet_tcp_setup2(struct sysctllog **clog, int pf, const char *pfname, 1385 const char *tcpname) 1386 { 1387 const struct sysctlnode *sack_node; 1388 #ifdef TCP_DEBUG 1389 extern struct tcp_debug tcp_debug[TCP_NDEBUG]; 1390 extern int tcp_debx; 1391 #endif 1392 1393 sysctl_createv(clog, 0, NULL, NULL, 1394 CTLFLAG_PERMANENT, 1395 CTLTYPE_NODE, "net", NULL, 1396 NULL, 0, NULL, 0, 1397 CTL_NET, CTL_EOL); 1398 sysctl_createv(clog, 0, NULL, NULL, 1399 CTLFLAG_PERMANENT, 1400 CTLTYPE_NODE, pfname, NULL, 1401 NULL, 0, NULL, 0, 1402 CTL_NET, pf, CTL_EOL); 1403 sysctl_createv(clog, 0, NULL, NULL, 1404 CTLFLAG_PERMANENT, 1405 CTLTYPE_NODE, tcpname, 1406 SYSCTL_DESCR("TCP related settings"), 1407 NULL, 0, NULL, 0, 1408 CTL_NET, pf, IPPROTO_TCP, CTL_EOL); 1409 1410 sysctl_createv(clog, 0, NULL, NULL, 1411 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1412 CTLTYPE_INT, "rfc1323", 1413 SYSCTL_DESCR("Enable RFC1323 TCP extensions"), 1414 NULL, 0, &tcp_do_rfc1323, 0, 1415 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RFC1323, CTL_EOL); 1416 sysctl_createv(clog, 0, NULL, NULL, 1417 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1418 CTLTYPE_INT, "sendspace", 1419 SYSCTL_DESCR("Default TCP send buffer size"), 1420 NULL, 0, &tcp_sendspace, 0, 1421 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SENDSPACE, CTL_EOL); 1422 sysctl_createv(clog, 0, NULL, NULL, 1423 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1424 CTLTYPE_INT, "recvspace", 1425 SYSCTL_DESCR("Default TCP receive buffer size"), 1426 NULL, 0, &tcp_recvspace, 0, 1427 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RECVSPACE, CTL_EOL); 1428 sysctl_createv(clog, 0, NULL, NULL, 1429 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1430 CTLTYPE_INT, "mssdflt", 1431 SYSCTL_DESCR("Default maximum segment size"), 1432 sysctl_net_inet_tcp_mssdflt, 0, &tcp_mssdflt, 0, 1433 CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSSDFLT, CTL_EOL); 1434 sysctl_createv(clog, 0, NULL, NULL, 1435 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1436 CTLTYPE_INT, "syn_cache_limit", 1437 SYSCTL_DESCR("Maximum number of entries in the TCP " 1438 "compressed state engine"), 1439 NULL, 0, &tcp_syn_cache_limit, 0, 1440 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_LIMIT, 1441 CTL_EOL); 1442 sysctl_createv(clog, 0, NULL, NULL, 1443 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1444 CTLTYPE_INT, "syn_bucket_limit", 1445 SYSCTL_DESCR("Maximum number of entries per hash " 1446 "bucket in the TCP compressed state " 1447 "engine"), 1448 NULL, 0, &tcp_syn_bucket_limit, 0, 1449 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_BUCKET_LIMIT, 1450 CTL_EOL); 1451 #if 0 /* obsoleted */ 1452 sysctl_createv(clog, 0, NULL, NULL, 1453 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1454 CTLTYPE_INT, "syn_cache_interval", 1455 SYSCTL_DESCR("TCP compressed state engine's timer interval"), 1456 NULL, 0, &tcp_syn_cache_interval, 0, 1457 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_INTER, 1458 CTL_EOL); 1459 #endif 1460 sysctl_createv(clog, 0, NULL, NULL, 1461 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1462 CTLTYPE_INT, "init_win", 1463 SYSCTL_DESCR("Initial TCP congestion window"), 1464 NULL, 0, &tcp_init_win, 0, 1465 CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN, CTL_EOL); 1466 sysctl_createv(clog, 0, NULL, NULL, 1467 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1468 CTLTYPE_INT, "mss_ifmtu", 1469 SYSCTL_DESCR("Use interface MTU for calculating MSS"), 1470 NULL, 0, &tcp_mss_ifmtu, 0, 1471 CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSS_IFMTU, CTL_EOL); 1472 sysctl_createv(clog, 0, NULL, &sack_node, 1473 CTLFLAG_PERMANENT, 1474 CTLTYPE_NODE, "sack", 1475 SYSCTL_DESCR("RFC2018 Selective ACKnowledgement tunables"), 1476 NULL, 0, NULL, 0, 1477 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_EOL); 1478 sysctl_createv(clog, 0, NULL, NULL, 1479 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1480 CTLTYPE_INT, "win_scale", 1481 SYSCTL_DESCR("Use RFC1323 window scale options"), 1482 NULL, 0, &tcp_do_win_scale, 0, 1483 CTL_NET, pf, IPPROTO_TCP, TCPCTL_WSCALE, CTL_EOL); 1484 sysctl_createv(clog, 0, NULL, NULL, 1485 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1486 CTLTYPE_INT, "timestamps", 1487 SYSCTL_DESCR("Use RFC1323 time stamp options"), 1488 NULL, 0, &tcp_do_timestamps, 0, 1489 CTL_NET, pf, IPPROTO_TCP, TCPCTL_TSTAMP, CTL_EOL); 1490 sysctl_createv(clog, 0, NULL, NULL, 1491 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1492 CTLTYPE_INT, "compat_42", 1493 SYSCTL_DESCR("Enable workarounds for 4.2BSD TCP bugs"), 1494 NULL, 0, &tcp_compat_42, 0, 1495 CTL_NET, pf, IPPROTO_TCP, TCPCTL_COMPAT_42, CTL_EOL); 1496 sysctl_createv(clog, 0, NULL, NULL, 1497 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1498 CTLTYPE_INT, "cwm", 1499 SYSCTL_DESCR("Hughes/Touch/Heidemann Congestion Window " 1500 "Monitoring"), 1501 NULL, 0, &tcp_cwm, 0, 1502 CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM, CTL_EOL); 1503 sysctl_createv(clog, 0, NULL, NULL, 1504 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1505 CTLTYPE_INT, "cwm_burstsize", 1506 SYSCTL_DESCR("Congestion Window Monitoring allowed " 1507 "burst count in packets"), 1508 NULL, 0, &tcp_cwm_burstsize, 0, 1509 CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM_BURSTSIZE, 1510 CTL_EOL); 1511 sysctl_createv(clog, 0, NULL, NULL, 1512 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1513 CTLTYPE_INT, "ack_on_push", 1514 SYSCTL_DESCR("Immediately return ACK when PSH is " 1515 "received"), 1516 NULL, 0, &tcp_ack_on_push, 0, 1517 CTL_NET, pf, IPPROTO_TCP, TCPCTL_ACK_ON_PUSH, CTL_EOL); 1518 sysctl_createv(clog, 0, NULL, NULL, 1519 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1520 CTLTYPE_INT, "keepidle", 1521 SYSCTL_DESCR("Allowed connection idle ticks before a " 1522 "keepalive probe is sent"), 1523 NULL, 0, &tcp_keepidle, 0, 1524 CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPIDLE, CTL_EOL); 1525 sysctl_createv(clog, 0, NULL, NULL, 1526 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1527 CTLTYPE_INT, "keepintvl", 1528 SYSCTL_DESCR("Ticks before next keepalive probe is sent"), 1529 NULL, 0, &tcp_keepintvl, 0, 1530 CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPINTVL, CTL_EOL); 1531 sysctl_createv(clog, 0, NULL, NULL, 1532 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1533 CTLTYPE_INT, "keepcnt", 1534 SYSCTL_DESCR("Number of keepalive probes to send"), 1535 NULL, 0, &tcp_keepcnt, 0, 1536 CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPCNT, CTL_EOL); 1537 sysctl_createv(clog, 0, NULL, NULL, 1538 CTLFLAG_PERMANENT|CTLFLAG_IMMEDIATE, 1539 CTLTYPE_INT, "slowhz", 1540 SYSCTL_DESCR("Keepalive ticks per second"), 1541 NULL, PR_SLOWHZ, NULL, 0, 1542 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SLOWHZ, CTL_EOL); 1543 sysctl_createv(clog, 0, NULL, NULL, 1544 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1545 CTLTYPE_INT, "newreno", 1546 SYSCTL_DESCR("NewReno congestion control algorithm"), 1547 NULL, 0, &tcp_do_newreno, 0, 1548 CTL_NET, pf, IPPROTO_TCP, TCPCTL_NEWRENO, CTL_EOL); 1549 sysctl_createv(clog, 0, NULL, NULL, 1550 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1551 CTLTYPE_INT, "log_refused", 1552 SYSCTL_DESCR("Log refused TCP connections"), 1553 NULL, 0, &tcp_log_refused, 0, 1554 CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOG_REFUSED, CTL_EOL); 1555 #if 0 /* obsoleted */ 1556 sysctl_createv(clog, 0, NULL, NULL, 1557 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1558 CTLTYPE_INT, "rstratelimit", NULL, 1559 NULL, 0, &tcp_rst_ratelim, 0, 1560 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTRATELIMIT, CTL_EOL); 1561 #endif 1562 sysctl_createv(clog, 0, NULL, NULL, 1563 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1564 CTLTYPE_INT, "rstppslimit", 1565 SYSCTL_DESCR("Maximum number of RST packets to send " 1566 "per second"), 1567 NULL, 0, &tcp_rst_ppslim, 0, 1568 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTPPSLIMIT, CTL_EOL); 1569 sysctl_createv(clog, 0, NULL, NULL, 1570 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1571 CTLTYPE_INT, "delack_ticks", 1572 SYSCTL_DESCR("Number of ticks to delay sending an ACK"), 1573 NULL, 0, &tcp_delack_ticks, 0, 1574 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DELACK_TICKS, CTL_EOL); 1575 sysctl_createv(clog, 0, NULL, NULL, 1576 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1577 CTLTYPE_INT, "init_win_local", 1578 SYSCTL_DESCR("Initial TCP window size (in segments)"), 1579 NULL, 0, &tcp_init_win_local, 0, 1580 CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN_LOCAL, 1581 CTL_EOL); 1582 sysctl_createv(clog, 0, NULL, NULL, 1583 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1584 CTLTYPE_STRUCT, "ident", 1585 SYSCTL_DESCR("RFC1413 Identification Protocol lookups"), 1586 sysctl_net_inet_tcp_ident, 0, NULL, sizeof(uid_t), 1587 CTL_NET, pf, IPPROTO_TCP, TCPCTL_IDENT, CTL_EOL); 1588 sysctl_createv(clog, 0, NULL, NULL, 1589 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1590 CTLTYPE_INT, "do_loopback_cksum", 1591 SYSCTL_DESCR("Perform TCP checksum on loopback"), 1592 NULL, 0, &tcp_do_loopback_cksum, 0, 1593 CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOOPBACKCKSUM, 1594 CTL_EOL); 1595 sysctl_createv(clog, 0, NULL, NULL, 1596 CTLFLAG_PERMANENT, 1597 CTLTYPE_STRUCT, "pcblist", 1598 SYSCTL_DESCR("TCP protocol control block list"), 1599 sysctl_inpcblist, 0, &tcbtable, 0, 1600 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, 1601 CTL_EOL); 1602 1603 /* SACK gets it's own little subtree. */ 1604 sysctl_createv(clog, 0, NULL, &sack_node, 1605 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1606 CTLTYPE_INT, "enable", 1607 SYSCTL_DESCR("Enable RFC2018 Selective ACKnowledgement"), 1608 NULL, 0, &tcp_do_sack, 0, 1609 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 1610 sysctl_createv(clog, 0, NULL, &sack_node, 1611 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1612 CTLTYPE_INT, "maxholes", 1613 SYSCTL_DESCR("Maximum number of TCP SACK holes allowed per connection"), 1614 NULL, 0, &tcp_sack_tp_maxholes, 0, 1615 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 1616 sysctl_createv(clog, 0, NULL, &sack_node, 1617 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1618 CTLTYPE_INT, "globalmaxholes", 1619 SYSCTL_DESCR("Global maximum number of TCP SACK holes"), 1620 NULL, 0, &tcp_sack_globalmaxholes, 0, 1621 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 1622 sysctl_createv(clog, 0, NULL, &sack_node, 1623 CTLFLAG_PERMANENT, 1624 CTLTYPE_INT, "globalholes", 1625 SYSCTL_DESCR("Global number of TCP SACK holes"), 1626 NULL, 0, &tcp_sack_globalholes, 0, 1627 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 1628 1629 sysctl_createv(clog, 0, NULL, NULL, 1630 CTLFLAG_PERMANENT, 1631 CTLTYPE_STRUCT, "stats", 1632 SYSCTL_DESCR("TCP statistics"), 1633 NULL, 0, &tcpstat, sizeof(tcpstat), 1634 CTL_NET, pf, IPPROTO_TCP, TCPCTL_STATS, 1635 CTL_EOL); 1636 #ifdef TCP_DEBUG 1637 sysctl_createv(clog, 0, NULL, NULL, 1638 CTLFLAG_PERMANENT, 1639 CTLTYPE_STRUCT, "debug", 1640 SYSCTL_DESCR("TCP sockets debug information"), 1641 NULL, 0, &tcp_debug, sizeof(tcp_debug), 1642 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBUG, 1643 CTL_EOL); 1644 sysctl_createv(clog, 0, NULL, NULL, 1645 CTLFLAG_PERMANENT, 1646 CTLTYPE_INT, "debx", 1647 SYSCTL_DESCR("Number of TCP debug sockets messages"), 1648 NULL, 0, &tcp_debx, sizeof(tcp_debx), 1649 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBX, 1650 CTL_EOL); 1651 #endif 1652 1653 } 1654 1655 /* 1656 * Sysctl for tcp variables. 1657 */ 1658 #ifdef INET 1659 SYSCTL_SETUP(sysctl_net_inet_tcp_setup, "sysctl net.inet.tcp subtree setup") 1660 { 1661 1662 sysctl_net_inet_tcp_setup2(clog, PF_INET, "inet", "tcp"); 1663 } 1664 #endif /* INET */ 1665 1666 #ifdef INET6 1667 SYSCTL_SETUP(sysctl_net_inet6_tcp6_setup, "sysctl net.inet6.tcp6 subtree setup") 1668 { 1669 1670 sysctl_net_inet_tcp_setup2(clog, PF_INET6, "inet6", "tcp6"); 1671 } 1672 #endif /* INET6 */ 1673
Indexes created Sun Sep 21 16:09:51 GMT 2025