1/* $NetBSD$ */ 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, 2006 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 * This code is derived from software contributed to The NetBSD Foundation 42 * by Rui Paulo. 43 * 44 * Redistribution and use in source and binary forms, with or without 45 * modification, are permitted provided that the following conditions 46 * are met: 47 * 1. Redistributions of source code must retain the above copyright 48 * notice, this list of conditions and the following disclaimer. 49 * 2. Redistributions in binary form must reproduce the above copyright 50 * notice, this list of conditions and the following disclaimer in the 51 * documentation and/or other materials provided with the distribution. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 54 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 55 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 56 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 57 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 58 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 59 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 60 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 61 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 62 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 63 * POSSIBILITY OF SUCH DAMAGE. 64 */ 65 66/* 67 * Copyright (c) 1982, 1986, 1988, 1993, 1995 68 * The Regents of the University of California. All rights reserved. 69 * 70 * Redistribution and use in source and binary forms, with or without 71 * modification, are permitted provided that the following conditions 72 * are met: 73 * 1. Redistributions of source code must retain the above copyright 74 * notice, this list of conditions and the following disclaimer. 75 * 2. Redistributions in binary form must reproduce the above copyright 76 * notice, this list of conditions and the following disclaimer in the 77 * documentation and/or other materials provided with the distribution. 78 * 3. Neither the name of the University nor the names of its contributors 79 * may be used to endorse or promote products derived from this software 80 * without specific prior written permission. 81 * 82 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 83 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 84 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 85 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 86 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 87 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 88 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 89 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 90 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 91 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 92 * SUCH DAMAGE. 93 * 94 * @(#)tcp_usrreq.c 8.5 (Berkeley) 6/21/95 95 */ 96 97#include <sys/cdefs.h> 98__KERNEL_RCSID(0, "$NetBSD$"); 99 100#include "opt_inet.h" 101#include "opt_ipsec.h" 102#include "opt_tcp_debug.h" 103#include "opt_mbuftrace.h" 104 105 106#include <sys/param.h> 107#include <sys/systm.h> 108#include <sys/kernel.h> 109#include <sys/malloc.h> 110#include <sys/mbuf.h> 111#include <sys/socket.h> 112#include <sys/socketvar.h> 113#include <sys/protosw.h> 114#include <sys/errno.h> 115#include <sys/stat.h> 116#include <sys/proc.h> 117#include <sys/domain.h> 118#include <sys/sysctl.h> 119#include <sys/kauth.h> 120#include <sys/uidinfo.h> 121 122#include <net/if.h> 123#include <net/route.h> 124 125#include <netinet/in.h> 126#include <netinet/in_systm.h> 127#include <netinet/in_var.h> 128#include <netinet/ip.h> 129#include <netinet/in_pcb.h> 130#include <netinet/ip_var.h> 131#include <netinet/in_offload.h> 132 133#ifdef INET6 134#ifndef INET 135#include <netinet/in.h> 136#endif 137#include <netinet/ip6.h> 138#include <netinet6/in6_pcb.h> 139#include <netinet6/ip6_var.h> 140#include <netinet6/scope6_var.h> 141#endif 142 143#include <netinet/tcp.h> 144#include <netinet/tcp_fsm.h> 145#include <netinet/tcp_seq.h> 146#include <netinet/tcp_timer.h> 147#include <netinet/tcp_var.h> 148#include <netinet/tcp_private.h> 149#include <netinet/tcp_congctl.h> 150#include <netinet/tcpip.h> 151#include <netinet/tcp_debug.h> 152#include <netinet/tcp_vtw.h> 153 154#include "opt_tcp_space.h" 155 156#ifdef KAME_IPSEC 157#include <netinet6/ipsec.h> 158#endif /*KAME_IPSEC*/ 159 160/* 161 * TCP protocol interface to socket abstraction. 162 */ 163 164/* 165 * Process a TCP user request for TCP tb. If this is a send request 166 * then m is the mbuf chain of send data. If this is a timer expiration 167 * (called from the software clock routine), then timertype tells which timer. 168 */ 169/*ARGSUSED*/ 170int 171tcp_usrreq(struct socket *so, int req, 172 struct mbuf *m, struct mbuf *nam, struct mbuf *control, struct lwp *l) 173{ 174 struct inpcb *inp; 175#ifdef INET6 176 struct in6pcb *in6p; 177#endif 178 struct tcpcb *tp = NULL; 179 int s; 180 int error = 0; 181#ifdef TCP_DEBUG 182 int ostate = 0; 183#endif 184 int family; /* family of the socket */ 185 186 family = so->so_proto->pr_domain->dom_family; 187 188 if (req == PRU_CONTROL) { 189 switch (family) { 190#ifdef INET 191 case PF_INET: 192 return (in_control(so, (long)m, (void *)nam, 193 (struct ifnet *)control, l)); 194#endif 195#ifdef INET6 196 case PF_INET6: 197 return (in6_control(so, (long)m, (void *)nam, 198 (struct ifnet *)control, l)); 199#endif 200 default: 201 return EAFNOSUPPORT; 202 } 203 } 204 205 s = splsoftnet(); 206 207 if (req == PRU_PURGEIF) { 208 mutex_enter(softnet_lock); 209 switch (family) { 210#ifdef INET 211 case PF_INET: 212 in_pcbpurgeif0(&tcbtable, (struct ifnet *)control); 213 in_purgeif((struct ifnet *)control); 214 in_pcbpurgeif(&tcbtable, (struct ifnet *)control); 215 break; 216#endif 217#ifdef INET6 218 case PF_INET6: 219 in6_pcbpurgeif0(&tcbtable, (struct ifnet *)control); 220 in6_purgeif((struct ifnet *)control); 221 in6_pcbpurgeif(&tcbtable, (struct ifnet *)control); 222 break; 223#endif 224 default: 225 mutex_exit(softnet_lock); 226 splx(s); 227 return (EAFNOSUPPORT); 228 } 229 mutex_exit(softnet_lock); 230 splx(s); 231 return (0); 232 } 233 234 if (req == PRU_ATTACH) 235 sosetlock(so); 236 237 switch (family) { 238#ifdef INET 239 case PF_INET: 240 inp = sotoinpcb(so); 241#ifdef INET6 242 in6p = NULL; 243#endif 244 break; 245#endif 246#ifdef INET6 247 case PF_INET6: 248 inp = NULL; 249 in6p = sotoin6pcb(so); 250 break; 251#endif 252 default: 253 splx(s); 254 return EAFNOSUPPORT; 255 } 256 257#ifdef DIAGNOSTIC 258#ifdef INET6 259 if (inp && in6p) 260 panic("tcp_usrreq: both inp and in6p set to non-NULL"); 261#endif 262 if (req != PRU_SEND && req != PRU_SENDOOB && control) 263 panic("tcp_usrreq: unexpected control mbuf"); 264#endif 265 /* 266 * When a TCP is attached to a socket, then there will be 267 * a (struct inpcb) pointed at by the socket, and this 268 * structure will point at a subsidary (struct tcpcb). 269 */ 270 if ((inp == 0 271#ifdef INET6 272 && in6p == 0 273#endif 274 ) && (req != PRU_ATTACH && req != PRU_SENSE)) 275 { 276 error = EINVAL; 277 goto release; 278 } 279#ifdef INET 280 if (inp) { 281 tp = intotcpcb(inp); 282 /* WHAT IF TP IS 0? */ 283#ifdef KPROF 284 tcp_acounts[tp->t_state][req]++; 285#endif 286#ifdef TCP_DEBUG 287 ostate = tp->t_state; 288#endif 289 } 290#endif 291#ifdef INET6 292 if (in6p) { 293 tp = in6totcpcb(in6p); 294 /* WHAT IF TP IS 0? */ 295#ifdef KPROF 296 tcp_acounts[tp->t_state][req]++; 297#endif 298#ifdef TCP_DEBUG 299 ostate = tp->t_state; 300#endif 301 } 302#endif 303 304 switch (req) { 305 306 /* 307 * TCP attaches to socket via PRU_ATTACH, reserving space, 308 * and an internet control block. 309 */ 310 case PRU_ATTACH: 311#ifndef INET6 312 if (inp != 0) 313#else 314 if (inp != 0 || in6p != 0) 315#endif 316 { 317 error = EISCONN; 318 break; 319 } 320 error = tcp_attach(so); 321 if (error) 322 break; 323 if ((so->so_options & SO_LINGER) && so->so_linger == 0) 324 so->so_linger = TCP_LINGERTIME; 325 tp = sototcpcb(so); 326 break; 327 328 /* 329 * PRU_DETACH detaches the TCP protocol from the socket. 330 */ 331 case PRU_DETACH: 332 tp = tcp_disconnect(tp); 333 break; 334 335 /* 336 * Give the socket an address. 337 */ 338 case PRU_BIND: 339 switch (family) { 340#ifdef INET 341 case PF_INET: 342 error = in_pcbbind(inp, nam, l); 343 break; 344#endif 345#ifdef INET6 346 case PF_INET6: 347 error = in6_pcbbind(in6p, nam, l); 348 if (!error) { 349 /* mapped addr case */ 350 if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)) 351 tp->t_family = AF_INET; 352 else 353 tp->t_family = AF_INET6; 354 } 355 break; 356#endif 357 } 358 break; 359 360 /* 361 * Prepare to accept connections. 362 */ 363 case PRU_LISTEN: 364#ifdef INET 365 if (inp && inp->inp_lport == 0) { 366 error = in_pcbbind(inp, NULL, l); 367 if (error) 368 break; 369 } 370#endif 371#ifdef INET6 372 if (in6p && in6p->in6p_lport == 0) { 373 error = in6_pcbbind(in6p, NULL, l); 374 if (error) 375 break; 376 } 377#endif 378 tp->t_state = TCPS_LISTEN; 379 break; 380 381 /* 382 * Initiate connection to peer. 383 * Create a template for use in transmissions on this connection. 384 * Enter SYN_SENT state, and mark socket as connecting. 385 * Start keep-alive timer, and seed output sequence space. 386 * Send initial segment on connection. 387 */ 388 case PRU_CONNECT: 389#ifdef INET 390 if (inp) { 391 if (inp->inp_lport == 0) { 392 error = in_pcbbind(inp, NULL, l); 393 if (error) 394 break; 395 } 396 error = in_pcbconnect(inp, nam, l); 397 } 398#endif 399#ifdef INET6 400 if (in6p) { 401 if (in6p->in6p_lport == 0) { 402 error = in6_pcbbind(in6p, NULL, l); 403 if (error) 404 break; 405 } 406 error = in6_pcbconnect(in6p, nam, l); 407 if (!error) { 408 /* mapped addr case */ 409 if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) 410 tp->t_family = AF_INET; 411 else 412 tp->t_family = AF_INET6; 413 } 414 } 415#endif 416 if (error) 417 break; 418 tp->t_template = tcp_template(tp); 419 if (tp->t_template == 0) { 420#ifdef INET 421 if (inp) 422 in_pcbdisconnect(inp); 423#endif 424#ifdef INET6 425 if (in6p) 426 in6_pcbdisconnect(in6p); 427#endif 428 error = ENOBUFS; 429 break; 430 } 431 /* 432 * Compute window scaling to request. 433 * XXX: This should be moved to tcp_output(). 434 */ 435 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 436 (TCP_MAXWIN << tp->request_r_scale) < sb_max) 437 tp->request_r_scale++; 438 soisconnecting(so); 439 TCP_STATINC(TCP_STAT_CONNATTEMPT); 440 tp->t_state = TCPS_SYN_SENT; 441 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit); 442 tp->iss = tcp_new_iss(tp, 0); 443 tcp_sendseqinit(tp); 444 error = tcp_output(tp); 445 break; 446 447 /* 448 * Create a TCP connection between two sockets. 449 */ 450 case PRU_CONNECT2: 451 error = EOPNOTSUPP; 452 break; 453 454 /* 455 * Initiate disconnect from peer. 456 * If connection never passed embryonic stage, just drop; 457 * else if don't need to let data drain, then can just drop anyways, 458 * else have to begin TCP shutdown process: mark socket disconnecting, 459 * drain unread data, state switch to reflect user close, and 460 * send segment (e.g. FIN) to peer. Socket will be really disconnected 461 * when peer sends FIN and acks ours. 462 * 463 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 464 */ 465 case PRU_DISCONNECT: 466 tp = tcp_disconnect(tp); 467 break; 468 469 /* 470 * Accept a connection. Essentially all the work is 471 * done at higher levels; just return the address 472 * of the peer, storing through addr. 473 */ 474 case PRU_ACCEPT: 475#ifdef INET 476 if (inp) 477 in_setpeeraddr(inp, nam); 478#endif 479#ifdef INET6 480 if (in6p) 481 in6_setpeeraddr(in6p, nam); 482#endif 483 break; 484 485 /* 486 * Mark the connection as being incapable of further output. 487 */ 488 case PRU_SHUTDOWN: 489 socantsendmore(so); 490 tp = tcp_usrclosed(tp); 491 if (tp) 492 error = tcp_output(tp); 493 break; 494 495 /* 496 * After a receive, possibly send window update to peer. 497 */ 498 case PRU_RCVD: 499 /* 500 * soreceive() calls this function when a user receives 501 * ancillary data on a listening socket. We don't call 502 * tcp_output in such a case, since there is no header 503 * template for a listening socket and hence the kernel 504 * will panic. 505 */ 506 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) != 0) 507 (void) tcp_output(tp); 508 break; 509 510 /* 511 * Do a send by putting data in output queue and updating urgent 512 * marker if URG set. Possibly send more data. 513 */ 514 case PRU_SEND: 515 if (control && control->m_len) { 516 m_freem(control); 517 m_freem(m); 518 error = EINVAL; 519 break; 520 } 521 sbappendstream(&so->so_snd, m); 522 error = tcp_output(tp); 523 break; 524 525 /* 526 * Abort the TCP. 527 */ 528 case PRU_ABORT: 529 tp = tcp_drop(tp, ECONNABORTED); 530 break; 531 532 case PRU_SENSE: 533 /* 534 * stat: don't bother with a blocksize. 535 */ 536 splx(s); 537 return (0); 538 539 case PRU_RCVOOB: 540 if (control && control->m_len) { 541 m_freem(control); 542 m_freem(m); 543 error = EINVAL; 544 break; 545 } 546 if ((so->so_oobmark == 0 && 547 (so->so_state & SS_RCVATMARK) == 0) || 548 so->so_options & SO_OOBINLINE || 549 tp->t_oobflags & TCPOOB_HADDATA) { 550 error = EINVAL; 551 break; 552 } 553 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 554 error = EWOULDBLOCK; 555 break; 556 } 557 m->m_len = 1; 558 *mtod(m, char *) = tp->t_iobc; 559 if (((long)nam & MSG_PEEK) == 0) 560 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 561 break; 562 563 case PRU_SENDOOB: 564 if (sbspace(&so->so_snd) < -512) { 565 m_freem(m); 566 error = ENOBUFS; 567 break; 568 } 569 /* 570 * According to RFC961 (Assigned Protocols), 571 * the urgent pointer points to the last octet 572 * of urgent data. We continue, however, 573 * to consider it to indicate the first octet 574 * of data past the urgent section. 575 * Otherwise, snd_up should be one lower. 576 */ 577 sbappendstream(&so->so_snd, m); 578 tp->snd_up = tp->snd_una + so->so_snd.sb_cc; 579 tp->t_force = 1; 580 error = tcp_output(tp); 581 tp->t_force = 0; 582 break; 583 584 case PRU_SOCKADDR: 585#ifdef INET 586 if (inp) 587 in_setsockaddr(inp, nam); 588#endif 589#ifdef INET6 590 if (in6p) 591 in6_setsockaddr(in6p, nam); 592#endif 593 break; 594 595 case PRU_PEERADDR: 596#ifdef INET 597 if (inp) 598 in_setpeeraddr(inp, nam); 599#endif 600#ifdef INET6 601 if (in6p) 602 in6_setpeeraddr(in6p, nam); 603#endif 604 break; 605 606 default: 607 panic("tcp_usrreq"); 608 } 609#ifdef TCP_DEBUG 610 if (tp && (so->so_options & SO_DEBUG)) 611 tcp_trace(TA_USER, ostate, tp, NULL, req); 612#endif 613 614release: 615 splx(s); 616 return (error); 617} 618 619static void 620change_keepalive(struct socket *so, struct tcpcb *tp) 621{ 622 tp->t_maxidle = tp->t_keepcnt * tp->t_keepintvl; 623 TCP_TIMER_DISARM(tp, TCPT_KEEP); 624 TCP_TIMER_DISARM(tp, TCPT_2MSL); 625 626 if (tp->t_state == TCPS_SYN_RECEIVED || 627 tp->t_state == TCPS_SYN_SENT) { 628 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit); 629 } else if (so->so_options & SO_KEEPALIVE && 630 tp->t_state <= TCPS_CLOSE_WAIT) { 631 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepintvl); 632 } else { 633 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle); 634 } 635 636 if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0)) 637 TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle); 638} 639 640 641int 642tcp_ctloutput(int op, struct socket *so, struct sockopt *sopt) 643{ 644 int error = 0, s; 645 struct inpcb *inp; 646#ifdef INET6 647 struct in6pcb *in6p; 648#endif 649 struct tcpcb *tp; 650 u_int ui; 651 int family; /* family of the socket */ 652 int level, optname, optval; 653 654 level = sopt->sopt_level; 655 optname = sopt->sopt_name; 656 657 family = so->so_proto->pr_domain->dom_family; 658 659 s = splsoftnet(); 660 switch (family) { 661#ifdef INET 662 case PF_INET: 663 inp = sotoinpcb(so); 664#ifdef INET6 665 in6p = NULL; 666#endif 667 break; 668#endif 669#ifdef INET6 670 case PF_INET6: 671 inp = NULL; 672 in6p = sotoin6pcb(so); 673 break; 674#endif 675 default: 676 splx(s); 677 panic("%s: af %d", __func__, family); 678 } 679#ifndef INET6 680 if (inp == NULL) 681#else 682 if (inp == NULL && in6p == NULL) 683#endif 684 { 685 splx(s); 686 return (ECONNRESET); 687 } 688 if (level != IPPROTO_TCP) { 689 switch (family) { 690#ifdef INET 691 case PF_INET: 692 error = ip_ctloutput(op, so, sopt); 693 break; 694#endif 695#ifdef INET6 696 case PF_INET6: 697 error = ip6_ctloutput(op, so, sopt); 698 break; 699#endif 700 } 701 splx(s); 702 return (error); 703 } 704 if (inp) 705 tp = intotcpcb(inp); 706#ifdef INET6 707 else if (in6p) 708 tp = in6totcpcb(in6p); 709#endif 710 else 711 tp = NULL; 712 713 switch (op) { 714 case PRCO_SETOPT: 715 switch (optname) { 716#ifdef TCP_SIGNATURE 717 case TCP_MD5SIG: 718 error = sockopt_getint(sopt, &optval); 719 if (error) 720 break; 721 if (optval > 0) 722 tp->t_flags |= TF_SIGNATURE; 723 else 724 tp->t_flags &= ~TF_SIGNATURE; 725 break; 726#endif /* TCP_SIGNATURE */ 727 728 case TCP_NODELAY: 729 error = sockopt_getint(sopt, &optval); 730 if (error) 731 break; 732 if (optval) 733 tp->t_flags |= TF_NODELAY; 734 else 735 tp->t_flags &= ~TF_NODELAY; 736 break; 737 738 case TCP_MAXSEG: 739 error = sockopt_getint(sopt, &optval); 740 if (error) 741 break; 742 if (optval > 0 && optval <= tp->t_peermss) 743 tp->t_peermss = optval; /* limit on send size */ 744 else 745 error = EINVAL; 746 break; 747#ifdef notyet 748 case TCP_CONGCTL: 749 /* XXX string overflow XXX */ 750 error = tcp_congctl_select(tp, sopt->sopt_data); 751 break; 752#endif 753 754 case TCP_KEEPIDLE: 755 error = sockopt_get(sopt, &ui, sizeof(ui)); 756 if (error) 757 break; 758 if (ui > 0) { 759 tp->t_keepidle = ui; 760 change_keepalive(so, tp); 761 } else 762 error = EINVAL; 763 break; 764 765 case TCP_KEEPINTVL: 766 error = sockopt_get(sopt, &ui, sizeof(ui)); 767 if (error) 768 break; 769 if (ui > 0) { 770 tp->t_keepintvl = ui; 771 change_keepalive(so, tp); 772 } else 773 error = EINVAL; 774 break; 775 776 case TCP_KEEPCNT: 777 error = sockopt_get(sopt, &ui, sizeof(ui)); 778 if (error) 779 break; 780 if (ui > 0) { 781 tp->t_keepcnt = ui; 782 change_keepalive(so, tp); 783 } else 784 error = EINVAL; 785 break; 786 787 case TCP_KEEPINIT: 788 error = sockopt_get(sopt, &ui, sizeof(ui)); 789 if (error) 790 break; 791 if (ui > 0) { 792 tp->t_keepinit = ui; 793 change_keepalive(so, tp); 794 } else 795 error = EINVAL; 796 break; 797 798 default: 799 error = ENOPROTOOPT; 800 break; 801 } 802 break; 803 804 case PRCO_GETOPT: 805 switch (optname) { 806#ifdef TCP_SIGNATURE 807 case TCP_MD5SIG: 808 optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0; 809 error = sockopt_set(sopt, &optval, sizeof(optval)); 810 break; 811#endif 812 case TCP_NODELAY: 813 optval = tp->t_flags & TF_NODELAY; 814 error = sockopt_set(sopt, &optval, sizeof(optval)); 815 break; 816 case TCP_MAXSEG: 817 optval = tp->t_peermss; 818 error = sockopt_set(sopt, &optval, sizeof(optval)); 819 break; 820#ifdef notyet 821 case TCP_CONGCTL: 822 break; 823#endif 824 default: 825 error = ENOPROTOOPT; 826 break; 827 } 828 break; 829 } 830 splx(s); 831 return (error); 832} 833 834#ifndef TCP_SENDSPACE 835#define TCP_SENDSPACE 1024*32 836#endif 837int tcp_sendspace = TCP_SENDSPACE; 838#ifndef TCP_RECVSPACE 839#define TCP_RECVSPACE 1024*32 840#endif 841int tcp_recvspace = TCP_RECVSPACE; 842 843/* 844 * Attach TCP protocol to socket, allocating 845 * internet protocol control block, tcp control block, 846 * bufer space, and entering LISTEN state if to accept connections. 847 */ 848int 849tcp_attach(struct socket *so) 850{ 851 struct tcpcb *tp; 852 struct inpcb *inp; 853#ifdef INET6 854 struct in6pcb *in6p; 855#endif 856 int error; 857 int family; /* family of the socket */ 858 859 family = so->so_proto->pr_domain->dom_family; 860 861#ifdef MBUFTRACE 862 so->so_mowner = &tcp_sock_mowner; 863 so->so_rcv.sb_mowner = &tcp_sock_rx_mowner; 864 so->so_snd.sb_mowner = &tcp_sock_tx_mowner; 865#endif 866 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 867 error = soreserve(so, tcp_sendspace, tcp_recvspace); 868 if (error) 869 return (error); 870 } 871 872 so->so_rcv.sb_flags |= SB_AUTOSIZE; 873 so->so_snd.sb_flags |= SB_AUTOSIZE; 874 875 switch (family) { 876#ifdef INET 877 case PF_INET: 878 error = in_pcballoc(so, &tcbtable); 879 if (error) 880 return (error); 881 inp = sotoinpcb(so); 882#ifdef INET6 883 in6p = NULL; 884#endif 885 break; 886#endif 887#ifdef INET6 888 case PF_INET6: 889 error = in6_pcballoc(so, &tcbtable); 890 if (error) 891 return (error); 892 inp = NULL; 893 in6p = sotoin6pcb(so); 894 break; 895#endif 896 default: 897 return EAFNOSUPPORT; 898 } 899 if (inp) 900 tp = tcp_newtcpcb(family, (void *)inp); 901#ifdef INET6 902 else if (in6p) 903 tp = tcp_newtcpcb(family, (void *)in6p); 904#endif 905 else 906 tp = NULL; 907 908 if (tp == 0) { 909 int nofd = so->so_state & SS_NOFDREF; /* XXX */ 910 911 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ 912#ifdef INET 913 if (inp) 914 in_pcbdetach(inp); 915#endif 916#ifdef INET6 917 if (in6p) 918 in6_pcbdetach(in6p); 919#endif 920 so->so_state |= nofd; 921 return (ENOBUFS); 922 } 923 tp->t_state = TCPS_CLOSED; 924 return (0); 925} 926 927/* 928 * Initiate (or continue) disconnect. 929 * If embryonic state, just send reset (once). 930 * If in ``let data drain'' option and linger null, just drop. 931 * Otherwise (hard), mark socket disconnecting and drop 932 * current input data; switch states based on user close, and 933 * send segment to peer (with FIN). 934 */ 935struct tcpcb * 936tcp_disconnect(struct tcpcb *tp) 937{ 938 struct socket *so; 939 940 if (tp->t_inpcb) 941 so = tp->t_inpcb->inp_socket; 942#ifdef INET6 943 else if (tp->t_in6pcb) 944 so = tp->t_in6pcb->in6p_socket; 945#endif 946 else 947 so = NULL; 948 949 if (TCPS_HAVEESTABLISHED(tp->t_state) == 0) 950 tp = tcp_close(tp); 951 else if ((so->so_options & SO_LINGER) && so->so_linger == 0) 952 tp = tcp_drop(tp, 0); 953 else { 954 soisdisconnecting(so); 955 sbflush(&so->so_rcv); 956 tp = tcp_usrclosed(tp); 957 if (tp) 958 (void) tcp_output(tp); 959 } 960 return (tp); 961} 962 963/* 964 * User issued close, and wish to trail through shutdown states: 965 * if never received SYN, just forget it. If got a SYN from peer, 966 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 967 * If already got a FIN from peer, then almost done; go to LAST_ACK 968 * state. In all other cases, have already sent FIN to peer (e.g. 969 * after PRU_SHUTDOWN), and just have to play tedious game waiting 970 * for peer to send FIN or not respond to keep-alives, etc. 971 * We can let the user exit from the close as soon as the FIN is acked. 972 */ 973struct tcpcb * 974tcp_usrclosed(struct tcpcb *tp) 975{ 976 977 switch (tp->t_state) { 978 979 case TCPS_CLOSED: 980 case TCPS_LISTEN: 981 case TCPS_SYN_SENT: 982 tp->t_state = TCPS_CLOSED; 983 tp = tcp_close(tp); 984 break; 985 986 case TCPS_SYN_RECEIVED: 987 case TCPS_ESTABLISHED: 988 tp->t_state = TCPS_FIN_WAIT_1; 989 break; 990 991 case TCPS_CLOSE_WAIT: 992 tp->t_state = TCPS_LAST_ACK; 993 break; 994 } 995 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { 996 struct socket *so; 997 if (tp->t_inpcb) 998 so = tp->t_inpcb->inp_socket; 999#ifdef INET6 1000 else if (tp->t_in6pcb) 1001 so = tp->t_in6pcb->in6p_socket; 1002#endif 1003 else 1004 so = NULL; 1005 if (so) 1006 soisdisconnected(so); 1007 /* 1008 * If we are in FIN_WAIT_2, we arrived here because the 1009 * application did a shutdown of the send side. Like the 1010 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after 1011 * a full close, we start a timer to make sure sockets are 1012 * not left in FIN_WAIT_2 forever. 1013 */ 1014 if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0)) 1015 TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle); 1016 else if (tp->t_state == TCPS_TIME_WAIT 1017 && ((tp->t_inpcb 1018 && (tcp4_vtw_enable & 1) 1019 && vtw_add(AF_INET, tp)) 1020 || 1021 (tp->t_in6pcb 1022 && (tcp6_vtw_enable & 1) 1023 && vtw_add(AF_INET6, tp)))) { 1024 tp = 0; 1025 } 1026 } 1027 return (tp); 1028} 1029 1030/* 1031 * sysctl helper routine for net.inet.ip.mssdflt. it can't be less 1032 * than 32. 1033 */ 1034static int 1035sysctl_net_inet_tcp_mssdflt(SYSCTLFN_ARGS) 1036{ 1037 int error, mssdflt; 1038 struct sysctlnode node; 1039 1040 mssdflt = tcp_mssdflt; 1041 node = *rnode; 1042 node.sysctl_data = &mssdflt; 1043 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1044 if (error || newp == NULL) 1045 return (error); 1046 1047 if (mssdflt < 32) 1048 return (EINVAL); 1049 tcp_mssdflt = mssdflt; 1050 1051 return (0); 1052} 1053 1054/* 1055 * sysctl helper routine for setting port related values under 1056 * net.inet.ip and net.inet6.ip6. does basic range checking and does 1057 * additional checks for each type. this code has placed in 1058 * tcp_input.c since INET and INET6 both use the same tcp code. 1059 * 1060 * this helper is not static so that both inet and inet6 can use it. 1061 */ 1062int 1063sysctl_net_inet_ip_ports(SYSCTLFN_ARGS) 1064{ 1065 int error, tmp; 1066 int apmin, apmax; 1067#ifndef IPNOPRIVPORTS 1068 int lpmin, lpmax; 1069#endif /* IPNOPRIVPORTS */ 1070 struct sysctlnode node; 1071 1072 if (namelen != 0) 1073 return (EINVAL); 1074 1075 switch (name[-3]) { 1076#ifdef INET 1077 case PF_INET: 1078 apmin = anonportmin; 1079 apmax = anonportmax; 1080#ifndef IPNOPRIVPORTS 1081 lpmin = lowportmin; 1082 lpmax = lowportmax; 1083#endif /* IPNOPRIVPORTS */ 1084 break; 1085#endif /* INET */ 1086#ifdef INET6 1087 case PF_INET6: 1088 apmin = ip6_anonportmin; 1089 apmax = ip6_anonportmax; 1090#ifndef IPNOPRIVPORTS 1091 lpmin = ip6_lowportmin; 1092 lpmax = ip6_lowportmax; 1093#endif /* IPNOPRIVPORTS */ 1094 break; 1095#endif /* INET6 */ 1096 default: 1097 return (EINVAL); 1098 } 1099 1100 /* 1101 * insert temporary copy into node, perform lookup on 1102 * temporary, then restore pointer 1103 */ 1104 node = *rnode; 1105 tmp = *(int*)rnode->sysctl_data; 1106 node.sysctl_data = &tmp; 1107 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1108 if (error || newp == NULL) 1109 return (error); 1110 1111 /* 1112 * simple port range check 1113 */ 1114 if (tmp < 0 || tmp > 65535) 1115 return (EINVAL); 1116 1117 /* 1118 * per-node range checks 1119 */ 1120 switch (rnode->sysctl_num) { 1121 case IPCTL_ANONPORTMIN: 1122 case IPV6CTL_ANONPORTMIN: 1123 if (tmp >= apmax) 1124 return (EINVAL); 1125#ifndef IPNOPRIVPORTS 1126 if (tmp < IPPORT_RESERVED) 1127 return (EINVAL); 1128#endif /* IPNOPRIVPORTS */ 1129 break; 1130 1131 case IPCTL_ANONPORTMAX: 1132 case IPV6CTL_ANONPORTMAX: 1133 if (apmin >= tmp) 1134 return (EINVAL); 1135#ifndef IPNOPRIVPORTS 1136 if (tmp < IPPORT_RESERVED) 1137 return (EINVAL); 1138#endif /* IPNOPRIVPORTS */ 1139 break; 1140 1141#ifndef IPNOPRIVPORTS 1142 case IPCTL_LOWPORTMIN: 1143 case IPV6CTL_LOWPORTMIN: 1144 if (tmp >= lpmax || 1145 tmp > IPPORT_RESERVEDMAX || 1146 tmp < IPPORT_RESERVEDMIN) 1147 return (EINVAL); 1148 break; 1149 1150 case IPCTL_LOWPORTMAX: 1151 case IPV6CTL_LOWPORTMAX: 1152 if (lpmin >= tmp || 1153 tmp > IPPORT_RESERVEDMAX || 1154 tmp < IPPORT_RESERVEDMIN) 1155 return (EINVAL); 1156 break; 1157#endif /* IPNOPRIVPORTS */ 1158 1159 default: 1160 return (EINVAL); 1161 } 1162 1163 *(int*)rnode->sysctl_data = tmp; 1164 1165 return (0); 1166} 1167 1168static inline int 1169copyout_uid(struct socket *sockp, void *oldp, size_t *oldlenp) 1170{ 1171 if (oldp) { 1172 size_t sz; 1173 uid_t uid; 1174 int error; 1175 1176 if (sockp->so_cred == NULL) 1177 return EPERM; 1178 1179 uid = kauth_cred_geteuid(sockp->so_cred); 1180 sz = MIN(sizeof(uid), *oldlenp); 1181 if ((error = copyout(&uid, oldp, sz)) != 0) 1182 return error; 1183 } 1184 *oldlenp = sizeof(uid_t); 1185 return 0; 1186} 1187 1188static inline int 1189inet4_ident_core(struct in_addr raddr, u_int rport, 1190 struct in_addr laddr, u_int lport, 1191 void *oldp, size_t *oldlenp, 1192 struct lwp *l, int dodrop) 1193{ 1194 struct inpcb *inp; 1195 struct socket *sockp; 1196 1197 inp = in_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport, 0); 1198 1199 if (inp == NULL || (sockp = inp->inp_socket) == NULL) 1200 return ESRCH; 1201 1202 if (dodrop) { 1203 struct tcpcb *tp; 1204 int error; 1205 1206 if (inp == NULL || (tp = intotcpcb(inp)) == NULL || 1207 (inp->inp_socket->so_options & SO_ACCEPTCONN) != 0) 1208 return ESRCH; 1209 1210 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, 1211 KAUTH_REQ_NETWORK_SOCKET_DROP, inp->inp_socket, tp, NULL); 1212 if (error) 1213 return (error); 1214 1215 (void)tcp_drop(tp, ECONNABORTED); 1216 return 0; 1217 } 1218 else 1219 return copyout_uid(sockp, oldp, oldlenp); 1220} 1221 1222#ifdef INET6 1223static inline int 1224inet6_ident_core(struct in6_addr *raddr, u_int rport, 1225 struct in6_addr *laddr, u_int lport, 1226 void *oldp, size_t *oldlenp, 1227 struct lwp *l, int dodrop) 1228{ 1229 struct in6pcb *in6p; 1230 struct socket *sockp; 1231 1232 in6p = in6_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport, 0, 0); 1233 1234 if (in6p == NULL || (sockp = in6p->in6p_socket) == NULL) 1235 return ESRCH; 1236 1237 if (dodrop) { 1238 struct tcpcb *tp; 1239 int error; 1240 1241 if (in6p == NULL || (tp = in6totcpcb(in6p)) == NULL || 1242 (in6p->in6p_socket->so_options & SO_ACCEPTCONN) != 0) 1243 return ESRCH; 1244 1245 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, 1246 KAUTH_REQ_NETWORK_SOCKET_DROP, in6p->in6p_socket, tp, NULL); 1247 if (error) 1248 return (error); 1249 1250 (void)tcp_drop(tp, ECONNABORTED); 1251 return 0; 1252 } 1253 else 1254 return copyout_uid(sockp, oldp, oldlenp); 1255} 1256#endif 1257 1258/* 1259 * sysctl helper routine for the net.inet.tcp.drop and 1260 * net.inet6.tcp6.drop nodes. 1261 */ 1262#define sysctl_net_inet_tcp_drop sysctl_net_inet_tcp_ident 1263 1264/* 1265 * sysctl helper routine for the net.inet.tcp.ident and 1266 * net.inet6.tcp6.ident nodes. contains backwards compat code for the 1267 * old way of looking up the ident information for ipv4 which involves 1268 * stuffing the port/addr pairs into the mib lookup. 1269 */ 1270static int 1271sysctl_net_inet_tcp_ident(SYSCTLFN_ARGS) 1272{ 1273#ifdef INET 1274 struct sockaddr_in *si4[2]; 1275#endif /* INET */ 1276#ifdef INET6 1277 struct sockaddr_in6 *si6[2]; 1278#endif /* INET6 */ 1279 struct sockaddr_storage sa[2]; 1280 int error, pf, dodrop; 1281 1282 dodrop = name[-1] == TCPCTL_DROP; 1283 if (dodrop) { 1284 if (oldp != NULL || *oldlenp != 0) 1285 return EINVAL; 1286 if (newp == NULL) 1287 return EPERM; 1288 if (newlen < sizeof(sa)) 1289 return ENOMEM; 1290 } 1291 if (namelen != 4 && namelen != 0) 1292 return EINVAL; 1293 if (name[-2] != IPPROTO_TCP) 1294 return EINVAL; 1295 pf = name[-3]; 1296 1297 /* old style lookup, ipv4 only */ 1298 if (namelen == 4) { 1299#ifdef INET 1300 struct in_addr laddr, raddr; 1301 u_int lport, rport; 1302 1303 if (pf != PF_INET) 1304 return EPROTONOSUPPORT; 1305 raddr.s_addr = (uint32_t)name[0]; 1306 rport = (u_int)name[1]; 1307 laddr.s_addr = (uint32_t)name[2]; 1308 lport = (u_int)name[3]; 1309 1310 mutex_enter(softnet_lock); 1311 error = inet4_ident_core(raddr, rport, laddr, lport, 1312 oldp, oldlenp, l, dodrop); 1313 mutex_exit(softnet_lock); 1314 return error; 1315#else /* INET */ 1316 return EINVAL; 1317#endif /* INET */ 1318 } 1319 1320 if (newp == NULL || newlen != sizeof(sa)) 1321 return EINVAL; 1322 error = copyin(newp, &sa, newlen); 1323 if (error) 1324 return error; 1325 1326 /* 1327 * requested families must match 1328 */ 1329 if (pf != sa[0].ss_family || sa[0].ss_family != sa[1].ss_family) 1330 return EINVAL; 1331 1332 switch (pf) { 1333#ifdef INET6 1334 case PF_INET6: 1335 si6[0] = (struct sockaddr_in6*)&sa[0]; 1336 si6[1] = (struct sockaddr_in6*)&sa[1]; 1337 if (si6[0]->sin6_len != sizeof(*si6[0]) || 1338 si6[1]->sin6_len != sizeof(*si6[1])) 1339 return EINVAL; 1340 1341 if (!IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) && 1342 !IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr)) { 1343 error = sa6_embedscope(si6[0], ip6_use_defzone); 1344 if (error) 1345 return error; 1346 error = sa6_embedscope(si6[1], ip6_use_defzone); 1347 if (error) 1348 return error; 1349 1350 mutex_enter(softnet_lock); 1351 error = inet6_ident_core(&si6[0]->sin6_addr, 1352 si6[0]->sin6_port, &si6[1]->sin6_addr, 1353 si6[1]->sin6_port, oldp, oldlenp, l, dodrop); 1354 mutex_exit(softnet_lock); 1355 return error; 1356 } 1357 1358 if (IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) != 1359 IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr)) 1360 return EINVAL; 1361 1362 in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[0]); 1363 in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[1]); 1364 /*FALLTHROUGH*/ 1365#endif /* INET6 */ 1366#ifdef INET 1367 case PF_INET: 1368 si4[0] = (struct sockaddr_in*)&sa[0]; 1369 si4[1] = (struct sockaddr_in*)&sa[1]; 1370 if (si4[0]->sin_len != sizeof(*si4[0]) || 1371 si4[0]->sin_len != sizeof(*si4[1])) 1372 return EINVAL; 1373 1374 mutex_enter(softnet_lock); 1375 error = inet4_ident_core(si4[0]->sin_addr, si4[0]->sin_port, 1376 si4[1]->sin_addr, si4[1]->sin_port, 1377 oldp, oldlenp, l, dodrop); 1378 mutex_exit(softnet_lock); 1379 return error; 1380#endif /* INET */ 1381 default: 1382 return EPROTONOSUPPORT; 1383 } 1384} 1385 1386/* 1387 * sysctl helper for the inet and inet6 pcblists. handles tcp/udp and 1388 * inet/inet6, as well as raw pcbs for each. specifically not 1389 * declared static so that raw sockets and udp/udp6 can use it as 1390 * well. 1391 */ 1392int 1393sysctl_inpcblist(SYSCTLFN_ARGS) 1394{ 1395#ifdef INET 1396 struct sockaddr_in *in; 1397 const struct inpcb *inp; 1398#endif 1399#ifdef INET6 1400 struct sockaddr_in6 *in6; 1401 const struct in6pcb *in6p; 1402#endif 1403 /* 1404 * sysctl_data is const, but CIRCLEQ_FOREACH can't use a const 1405 * struct inpcbtable pointer, so we have to discard const. :-/ 1406 */ 1407 struct inpcbtable *pcbtbl = __UNCONST(rnode->sysctl_data); 1408 const struct inpcb_hdr *inph; 1409 struct tcpcb *tp; 1410 struct kinfo_pcb pcb; 1411 char *dp; 1412 u_int op, arg; 1413 size_t len, needed, elem_size, out_size; 1414 int error, elem_count, pf, proto, pf2; 1415 1416 if (namelen != 4) 1417 return (EINVAL); 1418 1419 if (oldp != NULL) { 1420 len = *oldlenp; 1421 elem_size = name[2]; 1422 elem_count = name[3]; 1423 if (elem_size != sizeof(pcb)) 1424 return EINVAL; 1425 } else { 1426 len = 0; 1427 elem_count = INT_MAX; 1428 elem_size = sizeof(pcb); 1429 } 1430 error = 0; 1431 dp = oldp; 1432 op = name[0]; 1433 arg = name[1]; 1434 out_size = elem_size; 1435 needed = 0; 1436 1437 if (namelen == 1 && name[0] == CTL_QUERY) 1438 return (sysctl_query(SYSCTLFN_CALL(rnode))); 1439 1440 if (name - oname != 4) 1441 return (EINVAL); 1442 1443 pf = oname[1]; 1444 proto = oname[2]; 1445 pf2 = (oldp != NULL) ? pf : 0; 1446 1447 mutex_enter(softnet_lock); 1448 1449 CIRCLEQ_FOREACH(inph, &pcbtbl->inpt_queue, inph_queue) { 1450#ifdef INET 1451 inp = (const struct inpcb *)inph; 1452#endif 1453#ifdef INET6 1454 in6p = (const struct in6pcb *)inph; 1455#endif 1456 1457 if (inph->inph_af != pf) 1458 continue; 1459 1460 if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, 1461 KAUTH_REQ_NETWORK_SOCKET_CANSEE, inph->inph_socket, NULL, 1462 NULL) != 0) 1463 continue; 1464 1465 memset(&pcb, 0, sizeof(pcb)); 1466 1467 pcb.ki_family = pf; 1468 pcb.ki_type = proto; 1469 1470 switch (pf2) { 1471 case 0: 1472 /* just probing for size */ 1473 break; 1474#ifdef INET 1475 case PF_INET: 1476 pcb.ki_family = inp->inp_socket->so_proto-> 1477 pr_domain->dom_family; 1478 pcb.ki_type = inp->inp_socket->so_proto-> 1479 pr_type; 1480 pcb.ki_protocol = inp->inp_socket->so_proto-> 1481 pr_protocol; 1482 pcb.ki_pflags = inp->inp_flags; 1483 1484 pcb.ki_sostate = inp->inp_socket->so_state; 1485 pcb.ki_prstate = inp->inp_state; 1486 if (proto == IPPROTO_TCP) { 1487 tp = intotcpcb(inp); 1488 pcb.ki_tstate = tp->t_state; 1489 pcb.ki_tflags = tp->t_flags; 1490 } 1491 1492 pcb.ki_pcbaddr = PTRTOUINT64(inp); 1493 pcb.ki_ppcbaddr = PTRTOUINT64(inp->inp_ppcb); 1494 pcb.ki_sockaddr = PTRTOUINT64(inp->inp_socket); 1495 1496 pcb.ki_rcvq = inp->inp_socket->so_rcv.sb_cc; 1497 pcb.ki_sndq = inp->inp_socket->so_snd.sb_cc; 1498 1499 in = satosin(&pcb.ki_src); 1500 in->sin_len = sizeof(*in); 1501 in->sin_family = pf; 1502 in->sin_port = inp->inp_lport; 1503 in->sin_addr = inp->inp_laddr; 1504 if (pcb.ki_prstate >= INP_CONNECTED) { 1505 in = satosin(&pcb.ki_dst); 1506 in->sin_len = sizeof(*in); 1507 in->sin_family = pf; 1508 in->sin_port = inp->inp_fport; 1509 in->sin_addr = inp->inp_faddr; 1510 } 1511 break; 1512#endif 1513#ifdef INET6 1514 case PF_INET6: 1515 pcb.ki_family = in6p->in6p_socket->so_proto-> 1516 pr_domain->dom_family; 1517 pcb.ki_type = in6p->in6p_socket->so_proto->pr_type; 1518 pcb.ki_protocol = in6p->in6p_socket->so_proto-> 1519 pr_protocol; 1520 pcb.ki_pflags = in6p->in6p_flags; 1521 1522 pcb.ki_sostate = in6p->in6p_socket->so_state; 1523 pcb.ki_prstate = in6p->in6p_state; 1524 if (proto == IPPROTO_TCP) { 1525 tp = in6totcpcb(in6p); 1526 pcb.ki_tstate = tp->t_state; 1527 pcb.ki_tflags = tp->t_flags; 1528 } 1529 1530 pcb.ki_pcbaddr = PTRTOUINT64(in6p); 1531 pcb.ki_ppcbaddr = PTRTOUINT64(in6p->in6p_ppcb); 1532 pcb.ki_sockaddr = PTRTOUINT64(in6p->in6p_socket); 1533 1534 pcb.ki_rcvq = in6p->in6p_socket->so_rcv.sb_cc; 1535 pcb.ki_sndq = in6p->in6p_socket->so_snd.sb_cc; 1536 1537 in6 = satosin6(&pcb.ki_src); 1538 in6->sin6_len = sizeof(*in6); 1539 in6->sin6_family = pf; 1540 in6->sin6_port = in6p->in6p_lport; 1541 in6->sin6_flowinfo = in6p->in6p_flowinfo; 1542 in6->sin6_addr = in6p->in6p_laddr; 1543 in6->sin6_scope_id = 0; /* XXX? */ 1544 1545 if (pcb.ki_prstate >= IN6P_CONNECTED) { 1546 in6 = satosin6(&pcb.ki_dst); 1547 in6->sin6_len = sizeof(*in6); 1548 in6->sin6_family = pf; 1549 in6->sin6_port = in6p->in6p_fport; 1550 in6->sin6_flowinfo = in6p->in6p_flowinfo; 1551 in6->sin6_addr = in6p->in6p_faddr; 1552 in6->sin6_scope_id = 0; /* XXX? */ 1553 } 1554 break; 1555#endif 1556 } 1557 1558 if (len >= elem_size && elem_count > 0) { 1559 error = copyout(&pcb, dp, out_size); 1560 if (error) { 1561 mutex_exit(softnet_lock); 1562 return (error); 1563 } 1564 dp += elem_size; 1565 len -= elem_size; 1566 } 1567 needed += elem_size; 1568 if (elem_count > 0 && elem_count != INT_MAX) 1569 elem_count--; 1570 } 1571 1572 *oldlenp = needed; 1573 if (oldp == NULL) 1574 *oldlenp += PCB_SLOP * sizeof(struct kinfo_pcb); 1575 1576 mutex_exit(softnet_lock); 1577 1578 return (error); 1579} 1580 1581static int 1582sysctl_tcp_congctl(SYSCTLFN_ARGS) 1583{ 1584 struct sysctlnode node; 1585 int error; 1586 char newname[TCPCC_MAXLEN]; 1587 1588 strlcpy(newname, tcp_congctl_global_name, sizeof(newname) - 1); 1589 1590 node = *rnode; 1591 node.sysctl_data = newname; 1592 node.sysctl_size = sizeof(newname); 1593 1594 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1595 1596 if (error || 1597 newp == NULL || 1598 strncmp(newname, tcp_congctl_global_name, sizeof(newname)) == 0) 1599 return error; 1600 1601 mutex_enter(softnet_lock); 1602 error = tcp_congctl_select(NULL, newname); 1603 mutex_exit(softnet_lock); 1604 1605 return error; 1606} 1607 1608static int 1609sysctl_tcp_keep(SYSCTLFN_ARGS) 1610{ 1611 int error; 1612 u_int tmp; 1613 struct sysctlnode node; 1614 1615 node = *rnode; 1616 tmp = *(u_int *)rnode->sysctl_data; 1617 node.sysctl_data = &tmp; 1618 1619 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1620 if (error || newp == NULL) 1621 return error; 1622 1623 mutex_enter(softnet_lock); 1624 1625 *(u_int *)rnode->sysctl_data = tmp; 1626 tcp_tcpcb_template(); /* update the template */ 1627 1628 mutex_exit(softnet_lock); 1629 return 0; 1630} 1631 1632static int 1633sysctl_net_inet_tcp_stats(SYSCTLFN_ARGS) 1634{ 1635 1636 return (NETSTAT_SYSCTL(tcpstat_percpu, TCP_NSTATS)); 1637} 1638 1639/* 1640 * this (second stage) setup routine is a replacement for tcp_sysctl() 1641 * (which is currently used for ipv4 and ipv6) 1642 */ 1643static void 1644sysctl_net_inet_tcp_setup2(struct sysctllog **clog, int pf, const char *pfname, 1645 const char *tcpname) 1646{ 1647 const struct sysctlnode *sack_node; 1648 const struct sysctlnode *abc_node; 1649 const struct sysctlnode *ecn_node; 1650 const struct sysctlnode *congctl_node; 1651 const struct sysctlnode *mslt_node; 1652 const struct sysctlnode *vtw_node; 1653#ifdef TCP_DEBUG 1654 extern struct tcp_debug tcp_debug[TCP_NDEBUG]; 1655 extern int tcp_debx; 1656#endif 1657 1658 sysctl_createv(clog, 0, NULL, NULL, 1659 CTLFLAG_PERMANENT, 1660 CTLTYPE_NODE, "net", NULL, 1661 NULL, 0, NULL, 0, 1662 CTL_NET, CTL_EOL); 1663 sysctl_createv(clog, 0, NULL, NULL, 1664 CTLFLAG_PERMANENT, 1665 CTLTYPE_NODE, pfname, NULL, 1666 NULL, 0, NULL, 0, 1667 CTL_NET, pf, CTL_EOL); 1668 sysctl_createv(clog, 0, NULL, NULL, 1669 CTLFLAG_PERMANENT, 1670 CTLTYPE_NODE, tcpname, 1671 SYSCTL_DESCR("TCP related settings"), 1672 NULL, 0, NULL, 0, 1673 CTL_NET, pf, IPPROTO_TCP, CTL_EOL); 1674 1675 sysctl_createv(clog, 0, NULL, NULL, 1676 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1677 CTLTYPE_INT, "rfc1323", 1678 SYSCTL_DESCR("Enable RFC1323 TCP extensions"), 1679 NULL, 0, &tcp_do_rfc1323, 0, 1680 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RFC1323, CTL_EOL); 1681 sysctl_createv(clog, 0, NULL, NULL, 1682 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1683 CTLTYPE_INT, "sendspace", 1684 SYSCTL_DESCR("Default TCP send buffer size"), 1685 NULL, 0, &tcp_sendspace, 0, 1686 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SENDSPACE, CTL_EOL); 1687 sysctl_createv(clog, 0, NULL, NULL, 1688 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1689 CTLTYPE_INT, "recvspace", 1690 SYSCTL_DESCR("Default TCP receive buffer size"), 1691 NULL, 0, &tcp_recvspace, 0, 1692 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RECVSPACE, CTL_EOL); 1693 sysctl_createv(clog, 0, NULL, NULL, 1694 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1695 CTLTYPE_INT, "mssdflt", 1696 SYSCTL_DESCR("Default maximum segment size"), 1697 sysctl_net_inet_tcp_mssdflt, 0, &tcp_mssdflt, 0, 1698 CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSSDFLT, CTL_EOL); 1699 sysctl_createv(clog, 0, NULL, NULL, 1700 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1701 CTLTYPE_INT, "minmss", 1702 SYSCTL_DESCR("Lower limit for TCP maximum segment size"), 1703 NULL, 0, &tcp_minmss, 0, 1704 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1705 sysctl_createv(clog, 0, NULL, NULL, 1706 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1707 CTLTYPE_INT, "msl", 1708 SYSCTL_DESCR("Maximum Segment Life"), 1709 NULL, 0, &tcp_msl, 0, 1710 CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSL, CTL_EOL); 1711 sysctl_createv(clog, 0, NULL, NULL, 1712 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1713 CTLTYPE_INT, "syn_cache_limit", 1714 SYSCTL_DESCR("Maximum number of entries in the TCP " 1715 "compressed state engine"), 1716 NULL, 0, &tcp_syn_cache_limit, 0, 1717 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_LIMIT, 1718 CTL_EOL); 1719 sysctl_createv(clog, 0, NULL, NULL, 1720 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1721 CTLTYPE_INT, "syn_bucket_limit", 1722 SYSCTL_DESCR("Maximum number of entries per hash " 1723 "bucket in the TCP compressed state " 1724 "engine"), 1725 NULL, 0, &tcp_syn_bucket_limit, 0, 1726 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_BUCKET_LIMIT, 1727 CTL_EOL); 1728#if 0 /* obsoleted */ 1729 sysctl_createv(clog, 0, NULL, NULL, 1730 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1731 CTLTYPE_INT, "syn_cache_interval", 1732 SYSCTL_DESCR("TCP compressed state engine's timer interval"), 1733 NULL, 0, &tcp_syn_cache_interval, 0, 1734 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_INTER, 1735 CTL_EOL); 1736#endif 1737 sysctl_createv(clog, 0, NULL, NULL, 1738 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1739 CTLTYPE_INT, "init_win", 1740 SYSCTL_DESCR("Initial TCP congestion window"), 1741 NULL, 0, &tcp_init_win, 0, 1742 CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN, CTL_EOL); 1743 sysctl_createv(clog, 0, NULL, NULL, 1744 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1745 CTLTYPE_INT, "mss_ifmtu", 1746 SYSCTL_DESCR("Use interface MTU for calculating MSS"), 1747 NULL, 0, &tcp_mss_ifmtu, 0, 1748 CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSS_IFMTU, CTL_EOL); 1749 sysctl_createv(clog, 0, NULL, &sack_node, 1750 CTLFLAG_PERMANENT, 1751 CTLTYPE_NODE, "sack", 1752 SYSCTL_DESCR("RFC2018 Selective ACKnowledgement tunables"), 1753 NULL, 0, NULL, 0, 1754 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_EOL); 1755 1756 /* Congctl subtree */ 1757 sysctl_createv(clog, 0, NULL, &congctl_node, 1758 CTLFLAG_PERMANENT, 1759 CTLTYPE_NODE, "congctl", 1760 SYSCTL_DESCR("TCP Congestion Control"), 1761 NULL, 0, NULL, 0, 1762 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1763 sysctl_createv(clog, 0, &congctl_node, NULL, 1764 CTLFLAG_PERMANENT, 1765 CTLTYPE_STRING, "available", 1766 SYSCTL_DESCR("Available Congestion Control Mechanisms"), 1767 NULL, 0, &tcp_congctl_avail, 0, CTL_CREATE, CTL_EOL); 1768 sysctl_createv(clog, 0, &congctl_node, NULL, 1769 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1770 CTLTYPE_STRING, "selected", 1771 SYSCTL_DESCR("Selected Congestion Control Mechanism"), 1772 sysctl_tcp_congctl, 0, NULL, TCPCC_MAXLEN, 1773 CTL_CREATE, CTL_EOL); 1774 1775 sysctl_createv(clog, 0, NULL, NULL, 1776 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1777 CTLTYPE_INT, "win_scale", 1778 SYSCTL_DESCR("Use RFC1323 window scale options"), 1779 NULL, 0, &tcp_do_win_scale, 0, 1780 CTL_NET, pf, IPPROTO_TCP, TCPCTL_WSCALE, CTL_EOL); 1781 sysctl_createv(clog, 0, NULL, NULL, 1782 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1783 CTLTYPE_INT, "timestamps", 1784 SYSCTL_DESCR("Use RFC1323 time stamp options"), 1785 NULL, 0, &tcp_do_timestamps, 0, 1786 CTL_NET, pf, IPPROTO_TCP, TCPCTL_TSTAMP, CTL_EOL); 1787 sysctl_createv(clog, 0, NULL, NULL, 1788 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1789 CTLTYPE_INT, "compat_42", 1790 SYSCTL_DESCR("Enable workarounds for 4.2BSD TCP bugs"), 1791 NULL, 0, &tcp_compat_42, 0, 1792 CTL_NET, pf, IPPROTO_TCP, TCPCTL_COMPAT_42, CTL_EOL); 1793 sysctl_createv(clog, 0, NULL, NULL, 1794 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1795 CTLTYPE_INT, "cwm", 1796 SYSCTL_DESCR("Hughes/Touch/Heidemann Congestion Window " 1797 "Monitoring"), 1798 NULL, 0, &tcp_cwm, 0, 1799 CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM, CTL_EOL); 1800 sysctl_createv(clog, 0, NULL, NULL, 1801 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1802 CTLTYPE_INT, "cwm_burstsize", 1803 SYSCTL_DESCR("Congestion Window Monitoring allowed " 1804 "burst count in packets"), 1805 NULL, 0, &tcp_cwm_burstsize, 0, 1806 CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM_BURSTSIZE, 1807 CTL_EOL); 1808 sysctl_createv(clog, 0, NULL, NULL, 1809 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1810 CTLTYPE_INT, "ack_on_push", 1811 SYSCTL_DESCR("Immediately return ACK when PSH is " 1812 "received"), 1813 NULL, 0, &tcp_ack_on_push, 0, 1814 CTL_NET, pf, IPPROTO_TCP, TCPCTL_ACK_ON_PUSH, CTL_EOL); 1815 sysctl_createv(clog, 0, NULL, NULL, 1816 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1817 CTLTYPE_INT, "keepidle", 1818 SYSCTL_DESCR("Allowed connection idle ticks before a " 1819 "keepalive probe is sent"), 1820 sysctl_tcp_keep, 0, &tcp_keepidle, 0, 1821 CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPIDLE, CTL_EOL); 1822 sysctl_createv(clog, 0, NULL, NULL, 1823 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1824 CTLTYPE_INT, "keepintvl", 1825 SYSCTL_DESCR("Ticks before next keepalive probe is sent"), 1826 sysctl_tcp_keep, 0, &tcp_keepintvl, 0, 1827 CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPINTVL, CTL_EOL); 1828 sysctl_createv(clog, 0, NULL, NULL, 1829 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1830 CTLTYPE_INT, "keepcnt", 1831 SYSCTL_DESCR("Number of keepalive probes to send"), 1832 sysctl_tcp_keep, 0, &tcp_keepcnt, 0, 1833 CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPCNT, CTL_EOL); 1834 sysctl_createv(clog, 0, NULL, NULL, 1835 CTLFLAG_PERMANENT|CTLFLAG_IMMEDIATE, 1836 CTLTYPE_INT, "slowhz", 1837 SYSCTL_DESCR("Keepalive ticks per second"), 1838 NULL, PR_SLOWHZ, NULL, 0, 1839 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SLOWHZ, CTL_EOL); 1840 sysctl_createv(clog, 0, NULL, NULL, 1841 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1842 CTLTYPE_INT, "log_refused", 1843 SYSCTL_DESCR("Log refused TCP connections"), 1844 NULL, 0, &tcp_log_refused, 0, 1845 CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOG_REFUSED, CTL_EOL); 1846#if 0 /* obsoleted */ 1847 sysctl_createv(clog, 0, NULL, NULL, 1848 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1849 CTLTYPE_INT, "rstratelimit", NULL, 1850 NULL, 0, &tcp_rst_ratelim, 0, 1851 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTRATELIMIT, CTL_EOL); 1852#endif 1853 sysctl_createv(clog, 0, NULL, NULL, 1854 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1855 CTLTYPE_INT, "rstppslimit", 1856 SYSCTL_DESCR("Maximum number of RST packets to send " 1857 "per second"), 1858 NULL, 0, &tcp_rst_ppslim, 0, 1859 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTPPSLIMIT, CTL_EOL); 1860 sysctl_createv(clog, 0, NULL, NULL, 1861 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1862 CTLTYPE_INT, "delack_ticks", 1863 SYSCTL_DESCR("Number of ticks to delay sending an ACK"), 1864 NULL, 0, &tcp_delack_ticks, 0, 1865 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DELACK_TICKS, CTL_EOL); 1866 sysctl_createv(clog, 0, NULL, NULL, 1867 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1868 CTLTYPE_INT, "init_win_local", 1869 SYSCTL_DESCR("Initial TCP window size (in segments)"), 1870 NULL, 0, &tcp_init_win_local, 0, 1871 CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN_LOCAL, 1872 CTL_EOL); 1873 sysctl_createv(clog, 0, NULL, NULL, 1874 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1875 CTLTYPE_STRUCT, "ident", 1876 SYSCTL_DESCR("RFC1413 Identification Protocol lookups"), 1877 sysctl_net_inet_tcp_ident, 0, NULL, sizeof(uid_t), 1878 CTL_NET, pf, IPPROTO_TCP, TCPCTL_IDENT, CTL_EOL); 1879 sysctl_createv(clog, 0, NULL, NULL, 1880 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1881 CTLTYPE_INT, "do_loopback_cksum", 1882 SYSCTL_DESCR("Perform TCP checksum on loopback"), 1883 NULL, 0, &tcp_do_loopback_cksum, 0, 1884 CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOOPBACKCKSUM, 1885 CTL_EOL); 1886 sysctl_createv(clog, 0, NULL, NULL, 1887 CTLFLAG_PERMANENT, 1888 CTLTYPE_STRUCT, "pcblist", 1889 SYSCTL_DESCR("TCP protocol control block list"), 1890 sysctl_inpcblist, 0, &tcbtable, 0, 1891 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, 1892 CTL_EOL); 1893 sysctl_createv(clog, 0, NULL, NULL, 1894 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1895 CTLTYPE_INT, "keepinit", 1896 SYSCTL_DESCR("Ticks before initial tcp connection times out"), 1897 sysctl_tcp_keep, 0, &tcp_keepinit, 0, 1898 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1899 1900 /* TCP socket buffers auto-sizing nodes */ 1901 sysctl_createv(clog, 0, NULL, NULL, 1902 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1903 CTLTYPE_INT, "recvbuf_auto", 1904 SYSCTL_DESCR("Enable automatic receive " 1905 "buffer sizing (experimental)"), 1906 NULL, 0, &tcp_do_autorcvbuf, 0, 1907 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1908 sysctl_createv(clog, 0, NULL, NULL, 1909 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1910 CTLTYPE_INT, "recvbuf_inc", 1911 SYSCTL_DESCR("Incrementor step size of " 1912 "automatic receive buffer"), 1913 NULL, 0, &tcp_autorcvbuf_inc, 0, 1914 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1915 sysctl_createv(clog, 0, NULL, NULL, 1916 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1917 CTLTYPE_INT, "recvbuf_max", 1918 SYSCTL_DESCR("Max size of automatic receive buffer"), 1919 NULL, 0, &tcp_autorcvbuf_max, 0, 1920 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1921 1922 sysctl_createv(clog, 0, NULL, NULL, 1923 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1924 CTLTYPE_INT, "sendbuf_auto", 1925 SYSCTL_DESCR("Enable automatic send " 1926 "buffer sizing (experimental)"), 1927 NULL, 0, &tcp_do_autosndbuf, 0, 1928 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1929 sysctl_createv(clog, 0, NULL, NULL, 1930 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1931 CTLTYPE_INT, "sendbuf_inc", 1932 SYSCTL_DESCR("Incrementor step size of " 1933 "automatic send buffer"), 1934 NULL, 0, &tcp_autosndbuf_inc, 0, 1935 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1936 sysctl_createv(clog, 0, NULL, NULL, 1937 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1938 CTLTYPE_INT, "sendbuf_max", 1939 SYSCTL_DESCR("Max size of automatic send buffer"), 1940 NULL, 0, &tcp_autosndbuf_max, 0, 1941 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1942 1943 /* ECN subtree */ 1944 sysctl_createv(clog, 0, NULL, &ecn_node, 1945 CTLFLAG_PERMANENT, 1946 CTLTYPE_NODE, "ecn", 1947 SYSCTL_DESCR("RFC3168 Explicit Congestion Notification"), 1948 NULL, 0, NULL, 0, 1949 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1950 sysctl_createv(clog, 0, &ecn_node, NULL, 1951 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1952 CTLTYPE_INT, "enable", 1953 SYSCTL_DESCR("Enable TCP Explicit Congestion " 1954 "Notification"), 1955 NULL, 0, &tcp_do_ecn, 0, CTL_CREATE, CTL_EOL); 1956 sysctl_createv(clog, 0, &ecn_node, NULL, 1957 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1958 CTLTYPE_INT, "maxretries", 1959 SYSCTL_DESCR("Number of times to retry ECN setup " 1960 "before disabling ECN on the connection"), 1961 NULL, 0, &tcp_ecn_maxretries, 0, CTL_CREATE, CTL_EOL); 1962 1963 /* SACK gets it's own little subtree. */ 1964 sysctl_createv(clog, 0, NULL, &sack_node, 1965 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1966 CTLTYPE_INT, "enable", 1967 SYSCTL_DESCR("Enable RFC2018 Selective ACKnowledgement"), 1968 NULL, 0, &tcp_do_sack, 0, 1969 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 1970 sysctl_createv(clog, 0, NULL, &sack_node, 1971 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1972 CTLTYPE_INT, "maxholes", 1973 SYSCTL_DESCR("Maximum number of TCP SACK holes allowed per connection"), 1974 NULL, 0, &tcp_sack_tp_maxholes, 0, 1975 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 1976 sysctl_createv(clog, 0, NULL, &sack_node, 1977 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1978 CTLTYPE_INT, "globalmaxholes", 1979 SYSCTL_DESCR("Global maximum number of TCP SACK holes"), 1980 NULL, 0, &tcp_sack_globalmaxholes, 0, 1981 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 1982 sysctl_createv(clog, 0, NULL, &sack_node, 1983 CTLFLAG_PERMANENT, 1984 CTLTYPE_INT, "globalholes", 1985 SYSCTL_DESCR("Global number of TCP SACK holes"), 1986 NULL, 0, &tcp_sack_globalholes, 0, 1987 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 1988 1989 sysctl_createv(clog, 0, NULL, NULL, 1990 CTLFLAG_PERMANENT, 1991 CTLTYPE_STRUCT, "stats", 1992 SYSCTL_DESCR("TCP statistics"), 1993 sysctl_net_inet_tcp_stats, 0, NULL, 0, 1994 CTL_NET, pf, IPPROTO_TCP, TCPCTL_STATS, 1995 CTL_EOL); 1996 sysctl_createv(clog, 0, NULL, NULL, 1997 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1998 CTLTYPE_INT, "local_by_rtt", 1999 SYSCTL_DESCR("Use RTT estimator to decide which hosts " 2000 "are local"), 2001 NULL, 0, &tcp_rttlocal, 0, 2002 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2003#ifdef TCP_DEBUG 2004 sysctl_createv(clog, 0, NULL, NULL, 2005 CTLFLAG_PERMANENT, 2006 CTLTYPE_STRUCT, "debug", 2007 SYSCTL_DESCR("TCP sockets debug information"), 2008 NULL, 0, &tcp_debug, sizeof(tcp_debug), 2009 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBUG, 2010 CTL_EOL); 2011 sysctl_createv(clog, 0, NULL, NULL, 2012 CTLFLAG_PERMANENT, 2013 CTLTYPE_INT, "debx", 2014 SYSCTL_DESCR("Number of TCP debug sockets messages"), 2015 NULL, 0, &tcp_debx, sizeof(tcp_debx), 2016 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBX, 2017 CTL_EOL); 2018#endif 2019 sysctl_createv(clog, 0, NULL, NULL, 2020 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2021 CTLTYPE_STRUCT, "drop", 2022 SYSCTL_DESCR("TCP drop connection"), 2023 sysctl_net_inet_tcp_drop, 0, NULL, 0, 2024 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DROP, CTL_EOL); 2025 sysctl_createv(clog, 0, NULL, NULL, 2026 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2027 CTLTYPE_INT, "iss_hash", 2028 SYSCTL_DESCR("Enable RFC 1948 ISS by cryptographic " 2029 "hash computation"), 2030 NULL, 0, &tcp_do_rfc1948, sizeof(tcp_do_rfc1948), 2031 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, 2032 CTL_EOL); 2033 2034 /* ABC subtree */ 2035 2036 sysctl_createv(clog, 0, NULL, &abc_node, 2037 CTLFLAG_PERMANENT, CTLTYPE_NODE, "abc", 2038 SYSCTL_DESCR("RFC3465 Appropriate Byte Counting (ABC)"), 2039 NULL, 0, NULL, 0, 2040 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2041 sysctl_createv(clog, 0, &abc_node, NULL, 2042 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2043 CTLTYPE_INT, "enable", 2044 SYSCTL_DESCR("Enable RFC3465 Appropriate Byte Counting"), 2045 NULL, 0, &tcp_do_abc, 0, CTL_CREATE, CTL_EOL); 2046 sysctl_createv(clog, 0, &abc_node, NULL, 2047 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2048 CTLTYPE_INT, "aggressive", 2049 SYSCTL_DESCR("1: L=2*SMSS 0: L=1*SMSS"), 2050 NULL, 0, &tcp_abc_aggressive, 0, CTL_CREATE, CTL_EOL); 2051 2052 /* MSL tuning subtree */ 2053 2054 sysctl_createv(clog, 0, NULL, &mslt_node, 2055 CTLFLAG_PERMANENT, CTLTYPE_NODE, "mslt", 2056 SYSCTL_DESCR("MSL Tuning for TIME_WAIT truncation"), 2057 NULL, 0, NULL, 0, 2058 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2059 sysctl_createv(clog, 0, &mslt_node, NULL, 2060 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2061 CTLTYPE_INT, "enable", 2062 SYSCTL_DESCR("Enable TIME_WAIT truncation"), 2063 NULL, 0, &tcp_msl_enable, 0, CTL_CREATE, CTL_EOL); 2064 sysctl_createv(clog, 0, &mslt_node, NULL, 2065 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2066 CTLTYPE_INT, "loopback", 2067 SYSCTL_DESCR("MSL value to use for loopback connections"), 2068 NULL, 0, &tcp_msl_loop, 0, CTL_CREATE, CTL_EOL); 2069 sysctl_createv(clog, 0, &mslt_node, NULL, 2070 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2071 CTLTYPE_INT, "local", 2072 SYSCTL_DESCR("MSL value to use for local connections"), 2073 NULL, 0, &tcp_msl_local, 0, CTL_CREATE, CTL_EOL); 2074 sysctl_createv(clog, 0, &mslt_node, NULL, 2075 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2076 CTLTYPE_INT, "remote", 2077 SYSCTL_DESCR("MSL value to use for remote connections"), 2078 NULL, 0, &tcp_msl_remote, 0, CTL_CREATE, CTL_EOL); 2079 sysctl_createv(clog, 0, &mslt_node, NULL, 2080 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2081 CTLTYPE_INT, "remote_threshold", 2082 SYSCTL_DESCR("RTT estimate value to promote local to remote"), 2083 NULL, 0, &tcp_msl_remote_threshold, 0, CTL_CREATE, CTL_EOL); 2084 2085 /* vestigial TIME_WAIT tuning subtree */ 2086 2087 sysctl_createv(clog, 0, NULL, &vtw_node, 2088 CTLFLAG_PERMANENT, CTLTYPE_NODE, "vtw", 2089 SYSCTL_DESCR("Tuning for Vestigial TIME_WAIT"), 2090 NULL, 0, NULL, 0, 2091 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2092 sysctl_createv(clog, 0, &vtw_node, NULL, 2093 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2094 CTLTYPE_INT, "enable", 2095 SYSCTL_DESCR("Enable Vestigial TIME_WAIT"), 2096 sysctl_tcp_vtw_enable, 0, 2097 (pf == AF_INET) ? &tcp4_vtw_enable : &tcp6_vtw_enable, 2098 0, CTL_CREATE, CTL_EOL); 2099 sysctl_createv(clog, 0, &vtw_node, NULL, 2100 CTLFLAG_PERMANENT|CTLFLAG_READONLY, 2101 CTLTYPE_INT, "entries", 2102 SYSCTL_DESCR("Maximum number of vestigial TIME_WAIT entries"), 2103 NULL, 0, &tcp_vtw_entries, 0, CTL_CREATE, CTL_EOL); 2104} 2105 2106void 2107tcp_usrreq_init(void) 2108{ 2109 2110#ifdef INET 2111 sysctl_net_inet_tcp_setup2(NULL, PF_INET, "inet", "tcp"); 2112#endif 2113#ifdef INET6 2114 sysctl_net_inet_tcp_setup2(NULL, PF_INET6, "inet6", "tcp6"); 2115#endif 2116} 2117