1 /* $NetBSD: tcp_usrreq.c,v 1.158 2009/12/30 06:59:32 elad 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, 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: tcp_usrreq.c,v 1.158 2009/12/30 06:59:32 elad Exp $"); 99 100 #include "opt_inet.h" 101 #include "opt_ipsec.h" 102 #include "opt_tcp_debug.h" 103 #include "opt_mbuftrace.h" 104 #include "rnd.h" 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> 152 153 #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, (void *)nam, 192 (struct ifnet *)control, l)); 193 #endif 194 #ifdef INET6 195 case PF_INET6: 196 return (in6_control(so, (long)m, (void *)nam, 197 (struct ifnet *)control, l)); 198 #endif 199 default: 200 return EAFNOSUPPORT; 201 } 202 } 203 204 s = splsoftnet(); 205 206 if (req == PRU_PURGEIF) { 207 mutex_enter(softnet_lock); 208 switch (family) { 209 #ifdef INET 210 case PF_INET: 211 in_pcbpurgeif0(&tcbtable, (struct ifnet *)control); 212 in_purgeif((struct ifnet *)control); 213 in_pcbpurgeif(&tcbtable, (struct ifnet *)control); 214 break; 215 #endif 216 #ifdef INET6 217 case PF_INET6: 218 in6_pcbpurgeif0(&tcbtable, (struct ifnet *)control); 219 in6_purgeif((struct ifnet *)control); 220 in6_pcbpurgeif(&tcbtable, (struct ifnet *)control); 221 break; 222 #endif 223 default: 224 mutex_exit(softnet_lock); 225 splx(s); 226 return (EAFNOSUPPORT); 227 } 228 mutex_exit(softnet_lock); 229 splx(s); 230 return (0); 231 } 232 233 if (req == PRU_ATTACH) 234 sosetlock(so); 235 236 switch (family) { 237 #ifdef INET 238 case PF_INET: 239 inp = sotoinpcb(so); 240 #ifdef INET6 241 in6p = NULL; 242 #endif 243 break; 244 #endif 245 #ifdef INET6 246 case PF_INET6: 247 inp = NULL; 248 in6p = sotoin6pcb(so); 249 break; 250 #endif 251 default: 252 splx(s); 253 return EAFNOSUPPORT; 254 } 255 256 #ifdef DIAGNOSTIC 257 #ifdef INET6 258 if (inp && in6p) 259 panic("tcp_usrreq: both inp and in6p set to non-NULL"); 260 #endif 261 if (req != PRU_SEND && req != PRU_SENDOOB && control) 262 panic("tcp_usrreq: unexpected control mbuf"); 263 #endif 264 /* 265 * When a TCP is attached to a socket, then there will be 266 * a (struct inpcb) pointed at by the socket, and this 267 * structure will point at a subsidary (struct tcpcb). 268 */ 269 #ifndef INET6 270 if (inp == 0 && req != PRU_ATTACH) 271 #else 272 if ((inp == 0 && in6p == 0) && req != PRU_ATTACH) 273 #endif 274 { 275 error = EINVAL; 276 goto release; 277 } 278 #ifdef INET 279 if (inp) { 280 tp = intotcpcb(inp); 281 /* WHAT IF TP IS 0? */ 282 #ifdef KPROF 283 tcp_acounts[tp->t_state][req]++; 284 #endif 285 #ifdef TCP_DEBUG 286 ostate = tp->t_state; 287 #endif 288 } 289 #endif 290 #ifdef INET6 291 if (in6p) { 292 tp = in6totcpcb(in6p); 293 /* WHAT IF TP IS 0? */ 294 #ifdef KPROF 295 tcp_acounts[tp->t_state][req]++; 296 #endif 297 #ifdef TCP_DEBUG 298 ostate = tp->t_state; 299 #endif 300 } 301 #endif 302 303 switch (req) { 304 305 /* 306 * TCP attaches to socket via PRU_ATTACH, reserving space, 307 * and an internet control block. 308 */ 309 case PRU_ATTACH: 310 #ifndef INET6 311 if (inp != 0) 312 #else 313 if (inp != 0 || in6p != 0) 314 #endif 315 { 316 error = EISCONN; 317 break; 318 } 319 error = tcp_attach(so); 320 if (error) 321 break; 322 if ((so->so_options & SO_LINGER) && so->so_linger == 0) 323 so->so_linger = TCP_LINGERTIME; 324 tp = sototcpcb(so); 325 break; 326 327 /* 328 * PRU_DETACH detaches the TCP protocol from the socket. 329 */ 330 case PRU_DETACH: 331 tp = tcp_disconnect(tp); 332 break; 333 334 /* 335 * Give the socket an address. 336 */ 337 case PRU_BIND: 338 switch (family) { 339 #ifdef INET 340 case PF_INET: 341 error = in_pcbbind(inp, nam, l); 342 break; 343 #endif 344 #ifdef INET6 345 case PF_INET6: 346 error = in6_pcbbind(in6p, nam, l); 347 if (!error) { 348 /* mapped addr case */ 349 if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)) 350 tp->t_family = AF_INET; 351 else 352 tp->t_family = AF_INET6; 353 } 354 break; 355 #endif 356 } 357 break; 358 359 /* 360 * Prepare to accept connections. 361 */ 362 case PRU_LISTEN: 363 #ifdef INET 364 if (inp && inp->inp_lport == 0) { 365 error = in_pcbbind(inp, NULL, l); 366 if (error) 367 break; 368 } 369 #endif 370 #ifdef INET6 371 if (in6p && in6p->in6p_lport == 0) { 372 error = in6_pcbbind(in6p, NULL, l); 373 if (error) 374 break; 375 } 376 #endif 377 tp->t_state = TCPS_LISTEN; 378 break; 379 380 /* 381 * Initiate connection to peer. 382 * Create a template for use in transmissions on this connection. 383 * Enter SYN_SENT state, and mark socket as connecting. 384 * Start keep-alive timer, and seed output sequence space. 385 * Send initial segment on connection. 386 */ 387 case PRU_CONNECT: 388 #ifdef INET 389 if (inp) { 390 if (inp->inp_lport == 0) { 391 error = in_pcbbind(inp, NULL, l); 392 if (error) 393 break; 394 } 395 error = in_pcbconnect(inp, nam, l); 396 } 397 #endif 398 #ifdef INET6 399 if (in6p) { 400 if (in6p->in6p_lport == 0) { 401 error = in6_pcbbind(in6p, NULL, l); 402 if (error) 403 break; 404 } 405 error = in6_pcbconnect(in6p, nam, l); 406 if (!error) { 407 /* mapped addr case */ 408 if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) 409 tp->t_family = AF_INET; 410 else 411 tp->t_family = AF_INET6; 412 } 413 } 414 #endif 415 if (error) 416 break; 417 tp->t_template = tcp_template(tp); 418 if (tp->t_template == 0) { 419 #ifdef INET 420 if (inp) 421 in_pcbdisconnect(inp); 422 #endif 423 #ifdef INET6 424 if (in6p) 425 in6_pcbdisconnect(in6p); 426 #endif 427 error = ENOBUFS; 428 break; 429 } 430 /* 431 * Compute window scaling to request. 432 * XXX: This should be moved to tcp_output(). 433 */ 434 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 435 (TCP_MAXWIN << tp->request_r_scale) < sb_max) 436 tp->request_r_scale++; 437 soisconnecting(so); 438 TCP_STATINC(TCP_STAT_CONNATTEMPT); 439 tp->t_state = TCPS_SYN_SENT; 440 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit); 441 tp->iss = tcp_new_iss(tp, 0); 442 tcp_sendseqinit(tp); 443 error = tcp_output(tp); 444 break; 445 446 /* 447 * Create a TCP connection between two sockets. 448 */ 449 case PRU_CONNECT2: 450 error = EOPNOTSUPP; 451 break; 452 453 /* 454 * Initiate disconnect from peer. 455 * If connection never passed embryonic stage, just drop; 456 * else if don't need to let data drain, then can just drop anyways, 457 * else have to begin TCP shutdown process: mark socket disconnecting, 458 * drain unread data, state switch to reflect user close, and 459 * send segment (e.g. FIN) to peer. Socket will be really disconnected 460 * when peer sends FIN and acks ours. 461 * 462 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 463 */ 464 case PRU_DISCONNECT: 465 tp = tcp_disconnect(tp); 466 break; 467 468 /* 469 * Accept a connection. Essentially all the work is 470 * done at higher levels; just return the address 471 * of the peer, storing through addr. 472 */ 473 case PRU_ACCEPT: 474 #ifdef INET 475 if (inp) 476 in_setpeeraddr(inp, nam); 477 #endif 478 #ifdef INET6 479 if (in6p) 480 in6_setpeeraddr(in6p, nam); 481 #endif 482 break; 483 484 /* 485 * Mark the connection as being incapable of further output. 486 */ 487 case PRU_SHUTDOWN: 488 socantsendmore(so); 489 tp = tcp_usrclosed(tp); 490 if (tp) 491 error = tcp_output(tp); 492 break; 493 494 /* 495 * After a receive, possibly send window update to peer. 496 */ 497 case PRU_RCVD: 498 /* 499 * soreceive() calls this function when a user receives 500 * ancillary data on a listening socket. We don't call 501 * tcp_output in such a case, since there is no header 502 * template for a listening socket and hence the kernel 503 * will panic. 504 */ 505 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) != 0) 506 (void) tcp_output(tp); 507 break; 508 509 /* 510 * Do a send by putting data in output queue and updating urgent 511 * marker if URG set. Possibly send more data. 512 */ 513 case PRU_SEND: 514 if (control && control->m_len) { 515 m_freem(control); 516 m_freem(m); 517 error = EINVAL; 518 break; 519 } 520 sbappendstream(&so->so_snd, m); 521 error = tcp_output(tp); 522 break; 523 524 /* 525 * Abort the TCP. 526 */ 527 case PRU_ABORT: 528 tp = tcp_drop(tp, ECONNABORTED); 529 break; 530 531 case PRU_SENSE: 532 /* 533 * stat: don't bother with a blocksize. 534 */ 535 splx(s); 536 return (0); 537 538 case PRU_RCVOOB: 539 if (control && control->m_len) { 540 m_freem(control); 541 m_freem(m); 542 error = EINVAL; 543 break; 544 } 545 if ((so->so_oobmark == 0 && 546 (so->so_state & SS_RCVATMARK) == 0) || 547 so->so_options & SO_OOBINLINE || 548 tp->t_oobflags & TCPOOB_HADDATA) { 549 error = EINVAL; 550 break; 551 } 552 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 553 error = EWOULDBLOCK; 554 break; 555 } 556 m->m_len = 1; 557 *mtod(m, char *) = tp->t_iobc; 558 if (((long)nam & MSG_PEEK) == 0) 559 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 560 break; 561 562 case PRU_SENDOOB: 563 if (sbspace(&so->so_snd) < -512) { 564 m_freem(m); 565 error = ENOBUFS; 566 break; 567 } 568 /* 569 * According to RFC961 (Assigned Protocols), 570 * the urgent pointer points to the last octet 571 * of urgent data. We continue, however, 572 * to consider it to indicate the first octet 573 * of data past the urgent section. 574 * Otherwise, snd_up should be one lower. 575 */ 576 sbappendstream(&so->so_snd, m); 577 tp->snd_up = tp->snd_una + so->so_snd.sb_cc; 578 tp->t_force = 1; 579 error = tcp_output(tp); 580 tp->t_force = 0; 581 break; 582 583 case PRU_SOCKADDR: 584 #ifdef INET 585 if (inp) 586 in_setsockaddr(inp, nam); 587 #endif 588 #ifdef INET6 589 if (in6p) 590 in6_setsockaddr(in6p, nam); 591 #endif 592 break; 593 594 case PRU_PEERADDR: 595 #ifdef INET 596 if (inp) 597 in_setpeeraddr(inp, nam); 598 #endif 599 #ifdef INET6 600 if (in6p) 601 in6_setpeeraddr(in6p, nam); 602 #endif 603 break; 604 605 default: 606 panic("tcp_usrreq"); 607 } 608 #ifdef TCP_DEBUG 609 if (tp && (so->so_options & SO_DEBUG)) 610 tcp_trace(TA_USER, ostate, tp, NULL, req); 611 #endif 612 613 release: 614 splx(s); 615 return (error); 616 } 617 618 static void 619 change_keepalive(struct socket *so, struct tcpcb *tp) 620 { 621 tp->t_maxidle = tp->t_keepcnt * tp->t_keepintvl; 622 TCP_TIMER_DISARM(tp, TCPT_KEEP); 623 TCP_TIMER_DISARM(tp, TCPT_2MSL); 624 625 if (tp->t_state == TCPS_SYN_RECEIVED || 626 tp->t_state == TCPS_SYN_SENT) { 627 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit); 628 } else if (so->so_options & SO_KEEPALIVE && 629 tp->t_state <= TCPS_CLOSE_WAIT) { 630 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepintvl); 631 } else { 632 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle); 633 } 634 635 if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0)) 636 TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle); 637 } 638 639 640 int 641 tcp_ctloutput(int op, struct socket *so, struct sockopt *sopt) 642 { 643 int error = 0, s; 644 struct inpcb *inp; 645 #ifdef INET6 646 struct in6pcb *in6p; 647 #endif 648 struct tcpcb *tp; 649 u_int ui; 650 int family; /* family of the socket */ 651 int level, optname, optval; 652 653 level = sopt->sopt_level; 654 optname = sopt->sopt_name; 655 656 family = so->so_proto->pr_domain->dom_family; 657 658 s = splsoftnet(); 659 switch (family) { 660 #ifdef INET 661 case PF_INET: 662 inp = sotoinpcb(so); 663 #ifdef INET6 664 in6p = NULL; 665 #endif 666 break; 667 #endif 668 #ifdef INET6 669 case PF_INET6: 670 inp = NULL; 671 in6p = sotoin6pcb(so); 672 break; 673 #endif 674 default: 675 splx(s); 676 panic("%s: af %d", __func__, family); 677 } 678 #ifndef INET6 679 if (inp == NULL) 680 #else 681 if (inp == NULL && in6p == NULL) 682 #endif 683 { 684 splx(s); 685 return (ECONNRESET); 686 } 687 if (level != IPPROTO_TCP) { 688 switch (family) { 689 #ifdef INET 690 case PF_INET: 691 error = ip_ctloutput(op, so, sopt); 692 break; 693 #endif 694 #ifdef INET6 695 case PF_INET6: 696 error = ip6_ctloutput(op, so, sopt); 697 break; 698 #endif 699 } 700 splx(s); 701 return (error); 702 } 703 if (inp) 704 tp = intotcpcb(inp); 705 #ifdef INET6 706 else if (in6p) 707 tp = in6totcpcb(in6p); 708 #endif 709 else 710 tp = NULL; 711 712 switch (op) { 713 case PRCO_SETOPT: 714 switch (optname) { 715 #ifdef TCP_SIGNATURE 716 case TCP_MD5SIG: 717 error = sockopt_getint(sopt, &optval); 718 if (error) 719 break; 720 if (optval > 0) 721 tp->t_flags |= TF_SIGNATURE; 722 else 723 tp->t_flags &= ~TF_SIGNATURE; 724 break; 725 #endif /* TCP_SIGNATURE */ 726 727 case TCP_NODELAY: 728 error = sockopt_getint(sopt, &optval); 729 if (error) 730 break; 731 if (optval) 732 tp->t_flags |= TF_NODELAY; 733 else 734 tp->t_flags &= ~TF_NODELAY; 735 break; 736 737 case TCP_MAXSEG: 738 error = sockopt_getint(sopt, &optval); 739 if (error) 740 break; 741 if (optval > 0 && optval <= tp->t_peermss) 742 tp->t_peermss = optval; /* limit on send size */ 743 else 744 error = EINVAL; 745 break; 746 #ifdef notyet 747 case TCP_CONGCTL: 748 /* XXX string overflow XXX */ 749 error = tcp_congctl_select(tp, sopt->sopt_data); 750 break; 751 #endif 752 753 case TCP_KEEPIDLE: 754 error = sockopt_get(sopt, &ui, sizeof(ui)); 755 if (error) 756 break; 757 if (ui > 0) { 758 tp->t_keepidle = ui; 759 change_keepalive(so, tp); 760 } else 761 error = EINVAL; 762 break; 763 764 case TCP_KEEPINTVL: 765 error = sockopt_get(sopt, &ui, sizeof(ui)); 766 if (error) 767 break; 768 if (ui > 0) { 769 tp->t_keepintvl = ui; 770 change_keepalive(so, tp); 771 } else 772 error = EINVAL; 773 break; 774 775 case TCP_KEEPCNT: 776 error = sockopt_get(sopt, &ui, sizeof(ui)); 777 if (error) 778 break; 779 if (ui > 0) { 780 tp->t_keepcnt = ui; 781 change_keepalive(so, tp); 782 } else 783 error = EINVAL; 784 break; 785 786 case TCP_KEEPINIT: 787 error = sockopt_get(sopt, &ui, sizeof(ui)); 788 if (error) 789 break; 790 if (ui > 0) { 791 tp->t_keepinit = ui; 792 change_keepalive(so, tp); 793 } else 794 error = EINVAL; 795 break; 796 797 default: 798 error = ENOPROTOOPT; 799 break; 800 } 801 break; 802 803 case PRCO_GETOPT: 804 switch (optname) { 805 #ifdef TCP_SIGNATURE 806 case TCP_MD5SIG: 807 optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0; 808 error = sockopt_set(sopt, &optval, sizeof(optval)); 809 break; 810 #endif 811 case TCP_NODELAY: 812 optval = tp->t_flags & TF_NODELAY; 813 error = sockopt_set(sopt, &optval, sizeof(optval)); 814 break; 815 case TCP_MAXSEG: 816 optval = tp->t_peermss; 817 error = sockopt_set(sopt, &optval, sizeof(optval)); 818 break; 819 #ifdef notyet 820 case TCP_CONGCTL: 821 break; 822 #endif 823 default: 824 error = ENOPROTOOPT; 825 break; 826 } 827 break; 828 } 829 splx(s); 830 return (error); 831 } 832 833 #ifndef TCP_SENDSPACE 834 #define TCP_SENDSPACE 1024*32 835 #endif 836 int tcp_sendspace = TCP_SENDSPACE; 837 #ifndef TCP_RECVSPACE 838 #define TCP_RECVSPACE 1024*32 839 #endif 840 int tcp_recvspace = TCP_RECVSPACE; 841 842 /* 843 * Attach TCP protocol to socket, allocating 844 * internet protocol control block, tcp control block, 845 * bufer space, and entering LISTEN state if to accept connections. 846 */ 847 int 848 tcp_attach(struct socket *so) 849 { 850 struct tcpcb *tp; 851 struct inpcb *inp; 852 #ifdef INET6 853 struct in6pcb *in6p; 854 #endif 855 int error; 856 int family; /* family of the socket */ 857 858 family = so->so_proto->pr_domain->dom_family; 859 860 #ifdef MBUFTRACE 861 so->so_mowner = &tcp_sock_mowner; 862 so->so_rcv.sb_mowner = &tcp_sock_rx_mowner; 863 so->so_snd.sb_mowner = &tcp_sock_tx_mowner; 864 #endif 865 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 866 error = soreserve(so, tcp_sendspace, tcp_recvspace); 867 if (error) 868 return (error); 869 } 870 871 so->so_rcv.sb_flags |= SB_AUTOSIZE; 872 so->so_snd.sb_flags |= SB_AUTOSIZE; 873 874 switch (family) { 875 #ifdef INET 876 case PF_INET: 877 error = in_pcballoc(so, &tcbtable); 878 if (error) 879 return (error); 880 inp = sotoinpcb(so); 881 #ifdef INET6 882 in6p = NULL; 883 #endif 884 break; 885 #endif 886 #ifdef INET6 887 case PF_INET6: 888 error = in6_pcballoc(so, &tcbtable); 889 if (error) 890 return (error); 891 inp = NULL; 892 in6p = sotoin6pcb(so); 893 break; 894 #endif 895 default: 896 return EAFNOSUPPORT; 897 } 898 if (inp) 899 tp = tcp_newtcpcb(family, (void *)inp); 900 #ifdef INET6 901 else if (in6p) 902 tp = tcp_newtcpcb(family, (void *)in6p); 903 #endif 904 else 905 tp = NULL; 906 907 if (tp == 0) { 908 int nofd = so->so_state & SS_NOFDREF; /* XXX */ 909 910 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ 911 #ifdef INET 912 if (inp) 913 in_pcbdetach(inp); 914 #endif 915 #ifdef INET6 916 if (in6p) 917 in6_pcbdetach(in6p); 918 #endif 919 so->so_state |= nofd; 920 return (ENOBUFS); 921 } 922 tp->t_state = TCPS_CLOSED; 923 return (0); 924 } 925 926 /* 927 * Initiate (or continue) disconnect. 928 * If embryonic state, just send reset (once). 929 * If in ``let data drain'' option and linger null, just drop. 930 * Otherwise (hard), mark socket disconnecting and drop 931 * current input data; switch states based on user close, and 932 * send segment to peer (with FIN). 933 */ 934 struct tcpcb * 935 tcp_disconnect(struct tcpcb *tp) 936 { 937 struct socket *so; 938 939 if (tp->t_inpcb) 940 so = tp->t_inpcb->inp_socket; 941 #ifdef INET6 942 else if (tp->t_in6pcb) 943 so = tp->t_in6pcb->in6p_socket; 944 #endif 945 else 946 so = NULL; 947 948 if (TCPS_HAVEESTABLISHED(tp->t_state) == 0) 949 tp = tcp_close(tp); 950 else if ((so->so_options & SO_LINGER) && so->so_linger == 0) 951 tp = tcp_drop(tp, 0); 952 else { 953 soisdisconnecting(so); 954 sbflush(&so->so_rcv); 955 tp = tcp_usrclosed(tp); 956 if (tp) 957 (void) tcp_output(tp); 958 } 959 return (tp); 960 } 961 962 /* 963 * User issued close, and wish to trail through shutdown states: 964 * if never received SYN, just forget it. If got a SYN from peer, 965 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 966 * If already got a FIN from peer, then almost done; go to LAST_ACK 967 * state. In all other cases, have already sent FIN to peer (e.g. 968 * after PRU_SHUTDOWN), and just have to play tedious game waiting 969 * for peer to send FIN or not respond to keep-alives, etc. 970 * We can let the user exit from the close as soon as the FIN is acked. 971 */ 972 struct tcpcb * 973 tcp_usrclosed(struct tcpcb *tp) 974 { 975 976 switch (tp->t_state) { 977 978 case TCPS_CLOSED: 979 case TCPS_LISTEN: 980 case TCPS_SYN_SENT: 981 tp->t_state = TCPS_CLOSED; 982 tp = tcp_close(tp); 983 break; 984 985 case TCPS_SYN_RECEIVED: 986 case TCPS_ESTABLISHED: 987 tp->t_state = TCPS_FIN_WAIT_1; 988 break; 989 990 case TCPS_CLOSE_WAIT: 991 tp->t_state = TCPS_LAST_ACK; 992 break; 993 } 994 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { 995 struct socket *so; 996 if (tp->t_inpcb) 997 so = tp->t_inpcb->inp_socket; 998 #ifdef INET6 999 else if (tp->t_in6pcb) 1000 so = tp->t_in6pcb->in6p_socket; 1001 #endif 1002 else 1003 so = NULL; 1004 if (so) 1005 soisdisconnected(so); 1006 /* 1007 * If we are in FIN_WAIT_2, we arrived here because the 1008 * application did a shutdown of the send side. Like the 1009 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after 1010 * a full close, we start a timer to make sure sockets are 1011 * not left in FIN_WAIT_2 forever. 1012 */ 1013 if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0)) 1014 TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle); 1015 } 1016 return (tp); 1017 } 1018 1019 /* 1020 * sysctl helper routine for net.inet.ip.mssdflt. it can't be less 1021 * than 32. 1022 */ 1023 static int 1024 sysctl_net_inet_tcp_mssdflt(SYSCTLFN_ARGS) 1025 { 1026 int error, mssdflt; 1027 struct sysctlnode node; 1028 1029 mssdflt = tcp_mssdflt; 1030 node = *rnode; 1031 node.sysctl_data = &mssdflt; 1032 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1033 if (error || newp == NULL) 1034 return (error); 1035 1036 if (mssdflt < 32) 1037 return (EINVAL); 1038 tcp_mssdflt = mssdflt; 1039 1040 return (0); 1041 } 1042 1043 /* 1044 * sysctl helper routine for setting port related values under 1045 * net.inet.ip and net.inet6.ip6. does basic range checking and does 1046 * additional checks for each type. this code has placed in 1047 * tcp_input.c since INET and INET6 both use the same tcp code. 1048 * 1049 * this helper is not static so that both inet and inet6 can use it. 1050 */ 1051 int 1052 sysctl_net_inet_ip_ports(SYSCTLFN_ARGS) 1053 { 1054 int error, tmp; 1055 int apmin, apmax; 1056 #ifndef IPNOPRIVPORTS 1057 int lpmin, lpmax; 1058 #endif /* IPNOPRIVPORTS */ 1059 struct sysctlnode node; 1060 1061 if (namelen != 0) 1062 return (EINVAL); 1063 1064 switch (name[-3]) { 1065 #ifdef INET 1066 case PF_INET: 1067 apmin = anonportmin; 1068 apmax = anonportmax; 1069 #ifndef IPNOPRIVPORTS 1070 lpmin = lowportmin; 1071 lpmax = lowportmax; 1072 #endif /* IPNOPRIVPORTS */ 1073 break; 1074 #endif /* INET */ 1075 #ifdef INET6 1076 case PF_INET6: 1077 apmin = ip6_anonportmin; 1078 apmax = ip6_anonportmax; 1079 #ifndef IPNOPRIVPORTS 1080 lpmin = ip6_lowportmin; 1081 lpmax = ip6_lowportmax; 1082 #endif /* IPNOPRIVPORTS */ 1083 break; 1084 #endif /* INET6 */ 1085 default: 1086 return (EINVAL); 1087 } 1088 1089 /* 1090 * insert temporary copy into node, perform lookup on 1091 * temporary, then restore pointer 1092 */ 1093 node = *rnode; 1094 tmp = *(int*)rnode->sysctl_data; 1095 node.sysctl_data = &tmp; 1096 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1097 if (error || newp == NULL) 1098 return (error); 1099 1100 /* 1101 * simple port range check 1102 */ 1103 if (tmp < 0 || tmp > 65535) 1104 return (EINVAL); 1105 1106 /* 1107 * per-node range checks 1108 */ 1109 switch (rnode->sysctl_num) { 1110 case IPCTL_ANONPORTMIN: 1111 case IPV6CTL_ANONPORTMIN: 1112 if (tmp >= apmax) 1113 return (EINVAL); 1114 #ifndef IPNOPRIVPORTS 1115 if (tmp < IPPORT_RESERVED) 1116 return (EINVAL); 1117 #endif /* IPNOPRIVPORTS */ 1118 break; 1119 1120 case IPCTL_ANONPORTMAX: 1121 case IPV6CTL_ANONPORTMAX: 1122 if (apmin >= tmp) 1123 return (EINVAL); 1124 #ifndef IPNOPRIVPORTS 1125 if (tmp < IPPORT_RESERVED) 1126 return (EINVAL); 1127 #endif /* IPNOPRIVPORTS */ 1128 break; 1129 1130 #ifndef IPNOPRIVPORTS 1131 case IPCTL_LOWPORTMIN: 1132 case IPV6CTL_LOWPORTMIN: 1133 if (tmp >= lpmax || 1134 tmp > IPPORT_RESERVEDMAX || 1135 tmp < IPPORT_RESERVEDMIN) 1136 return (EINVAL); 1137 break; 1138 1139 case IPCTL_LOWPORTMAX: 1140 case IPV6CTL_LOWPORTMAX: 1141 if (lpmin >= tmp || 1142 tmp > IPPORT_RESERVEDMAX || 1143 tmp < IPPORT_RESERVEDMIN) 1144 return (EINVAL); 1145 break; 1146 #endif /* IPNOPRIVPORTS */ 1147 1148 default: 1149 return (EINVAL); 1150 } 1151 1152 *(int*)rnode->sysctl_data = tmp; 1153 1154 return (0); 1155 } 1156 1157 static inline int 1158 copyout_uid(struct socket *sockp, void *oldp, size_t *oldlenp) 1159 { 1160 size_t sz; 1161 int error; 1162 uid_t uid; 1163 1164 uid = kauth_cred_geteuid(sockp->so_cred); 1165 if (oldp) { 1166 sz = MIN(sizeof(uid), *oldlenp); 1167 error = copyout(&uid, oldp, sz); 1168 if (error) 1169 return error; 1170 } 1171 *oldlenp = sizeof(uid); 1172 return 0; 1173 } 1174 1175 static inline int 1176 inet4_ident_core(struct in_addr raddr, u_int rport, 1177 struct in_addr laddr, u_int lport, 1178 void *oldp, size_t *oldlenp, 1179 struct lwp *l, int dodrop) 1180 { 1181 struct inpcb *inp; 1182 struct socket *sockp; 1183 1184 inp = in_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport); 1185 1186 if (inp == NULL || (sockp = inp->inp_socket) == NULL) 1187 return ESRCH; 1188 1189 if (dodrop) { 1190 struct tcpcb *tp; 1191 int error; 1192 1193 if (inp == NULL || (tp = intotcpcb(inp)) == NULL || 1194 (inp->inp_socket->so_options & SO_ACCEPTCONN) != 0) 1195 return ESRCH; 1196 1197 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, 1198 KAUTH_REQ_NETWORK_SOCKET_DROP, inp->inp_socket, tp, NULL); 1199 if (error) 1200 return (error); 1201 1202 (void)tcp_drop(tp, ECONNABORTED); 1203 return 0; 1204 } 1205 else 1206 return copyout_uid(sockp, oldp, oldlenp); 1207 } 1208 1209 #ifdef INET6 1210 static inline int 1211 inet6_ident_core(struct in6_addr *raddr, u_int rport, 1212 struct in6_addr *laddr, u_int lport, 1213 void *oldp, size_t *oldlenp, 1214 struct lwp *l, int dodrop) 1215 { 1216 struct in6pcb *in6p; 1217 struct socket *sockp; 1218 1219 in6p = in6_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport, 0); 1220 1221 if (in6p == NULL || (sockp = in6p->in6p_socket) == NULL) 1222 return ESRCH; 1223 1224 if (dodrop) { 1225 struct tcpcb *tp; 1226 int error; 1227 1228 if (in6p == NULL || (tp = in6totcpcb(in6p)) == NULL || 1229 (in6p->in6p_socket->so_options & SO_ACCEPTCONN) != 0) 1230 return ESRCH; 1231 1232 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, 1233 KAUTH_REQ_NETWORK_SOCKET_DROP, in6p->in6p_socket, tp, NULL); 1234 if (error) 1235 return (error); 1236 1237 (void)tcp_drop(tp, ECONNABORTED); 1238 return 0; 1239 } 1240 else 1241 return copyout_uid(sockp, oldp, oldlenp); 1242 } 1243 #endif 1244 1245 /* 1246 * sysctl helper routine for the net.inet.tcp.drop and 1247 * net.inet6.tcp6.drop nodes. 1248 */ 1249 #define sysctl_net_inet_tcp_drop sysctl_net_inet_tcp_ident 1250 1251 /* 1252 * sysctl helper routine for the net.inet.tcp.ident and 1253 * net.inet6.tcp6.ident nodes. contains backwards compat code for the 1254 * old way of looking up the ident information for ipv4 which involves 1255 * stuffing the port/addr pairs into the mib lookup. 1256 */ 1257 static int 1258 sysctl_net_inet_tcp_ident(SYSCTLFN_ARGS) 1259 { 1260 #ifdef INET 1261 struct sockaddr_in *si4[2]; 1262 #endif /* INET */ 1263 #ifdef INET6 1264 struct sockaddr_in6 *si6[2]; 1265 #endif /* INET6 */ 1266 struct sockaddr_storage sa[2]; 1267 int error, pf, dodrop; 1268 1269 dodrop = name[-1] == TCPCTL_DROP; 1270 if (dodrop) { 1271 if (oldp != NULL || *oldlenp != 0) 1272 return EINVAL; 1273 if (newp == NULL) 1274 return EPERM; 1275 if (newlen < sizeof(sa)) 1276 return ENOMEM; 1277 } 1278 if (namelen != 4 && namelen != 0) 1279 return EINVAL; 1280 if (name[-2] != IPPROTO_TCP) 1281 return EINVAL; 1282 pf = name[-3]; 1283 1284 /* old style lookup, ipv4 only */ 1285 if (namelen == 4) { 1286 #ifdef INET 1287 struct in_addr laddr, raddr; 1288 u_int lport, rport; 1289 1290 if (pf != PF_INET) 1291 return EPROTONOSUPPORT; 1292 raddr.s_addr = (uint32_t)name[0]; 1293 rport = (u_int)name[1]; 1294 laddr.s_addr = (uint32_t)name[2]; 1295 lport = (u_int)name[3]; 1296 1297 mutex_enter(softnet_lock); 1298 error = inet4_ident_core(raddr, rport, laddr, lport, 1299 oldp, oldlenp, l, dodrop); 1300 mutex_exit(softnet_lock); 1301 return error; 1302 #else /* INET */ 1303 return EINVAL; 1304 #endif /* INET */ 1305 } 1306 1307 if (newp == NULL || newlen != sizeof(sa)) 1308 return EINVAL; 1309 error = copyin(newp, &sa, newlen); 1310 if (error) 1311 return error; 1312 1313 /* 1314 * requested families must match 1315 */ 1316 if (pf != sa[0].ss_family || sa[0].ss_family != sa[1].ss_family) 1317 return EINVAL; 1318 1319 switch (pf) { 1320 #ifdef INET6 1321 case PF_INET6: 1322 si6[0] = (struct sockaddr_in6*)&sa[0]; 1323 si6[1] = (struct sockaddr_in6*)&sa[1]; 1324 if (si6[0]->sin6_len != sizeof(*si6[0]) || 1325 si6[1]->sin6_len != sizeof(*si6[1])) 1326 return EINVAL; 1327 1328 if (!IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) && 1329 !IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr)) { 1330 error = sa6_embedscope(si6[0], ip6_use_defzone); 1331 if (error) 1332 return error; 1333 error = sa6_embedscope(si6[1], ip6_use_defzone); 1334 if (error) 1335 return error; 1336 1337 mutex_enter(softnet_lock); 1338 error = inet6_ident_core(&si6[0]->sin6_addr, 1339 si6[0]->sin6_port, &si6[1]->sin6_addr, 1340 si6[1]->sin6_port, oldp, oldlenp, l, dodrop); 1341 mutex_exit(softnet_lock); 1342 return error; 1343 } 1344 1345 if (IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) != 1346 IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr)) 1347 return EINVAL; 1348 1349 in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[0]); 1350 in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[1]); 1351 /*FALLTHROUGH*/ 1352 #endif /* INET6 */ 1353 #ifdef INET 1354 case PF_INET: 1355 si4[0] = (struct sockaddr_in*)&sa[0]; 1356 si4[1] = (struct sockaddr_in*)&sa[1]; 1357 if (si4[0]->sin_len != sizeof(*si4[0]) || 1358 si4[0]->sin_len != sizeof(*si4[1])) 1359 return EINVAL; 1360 1361 mutex_enter(softnet_lock); 1362 error = inet4_ident_core(si4[0]->sin_addr, si4[0]->sin_port, 1363 si4[1]->sin_addr, si4[1]->sin_port, 1364 oldp, oldlenp, l, dodrop); 1365 mutex_exit(softnet_lock); 1366 return error; 1367 #endif /* INET */ 1368 default: 1369 return EPROTONOSUPPORT; 1370 } 1371 } 1372 1373 /* 1374 * sysctl helper for the inet and inet6 pcblists. handles tcp/udp and 1375 * inet/inet6, as well as raw pcbs for each. specifically not 1376 * declared static so that raw sockets and udp/udp6 can use it as 1377 * well. 1378 */ 1379 int 1380 sysctl_inpcblist(SYSCTLFN_ARGS) 1381 { 1382 #ifdef INET 1383 struct sockaddr_in *in; 1384 const struct inpcb *inp; 1385 #endif 1386 #ifdef INET6 1387 struct sockaddr_in6 *in6; 1388 const struct in6pcb *in6p; 1389 #endif 1390 /* 1391 * sysctl_data is const, but CIRCLEQ_FOREACH can't use a const 1392 * struct inpcbtable pointer, so we have to discard const. :-/ 1393 */ 1394 struct inpcbtable *pcbtbl = __UNCONST(rnode->sysctl_data); 1395 const struct inpcb_hdr *inph; 1396 struct tcpcb *tp; 1397 struct kinfo_pcb pcb; 1398 char *dp; 1399 u_int op, arg; 1400 size_t len, needed, elem_size, out_size; 1401 int error, elem_count, pf, proto, pf2; 1402 1403 if (namelen != 4) 1404 return (EINVAL); 1405 1406 if (oldp != NULL) { 1407 len = *oldlenp; 1408 elem_size = name[2]; 1409 elem_count = name[3]; 1410 if (elem_size != sizeof(pcb)) 1411 return EINVAL; 1412 } else { 1413 len = 0; 1414 elem_count = INT_MAX; 1415 elem_size = sizeof(pcb); 1416 } 1417 error = 0; 1418 dp = oldp; 1419 op = name[0]; 1420 arg = name[1]; 1421 out_size = elem_size; 1422 needed = 0; 1423 1424 if (namelen == 1 && name[0] == CTL_QUERY) 1425 return (sysctl_query(SYSCTLFN_CALL(rnode))); 1426 1427 if (name - oname != 4) 1428 return (EINVAL); 1429 1430 pf = oname[1]; 1431 proto = oname[2]; 1432 pf2 = (oldp != NULL) ? pf : 0; 1433 1434 mutex_enter(softnet_lock); 1435 1436 CIRCLEQ_FOREACH(inph, &pcbtbl->inpt_queue, inph_queue) { 1437 #ifdef INET 1438 inp = (const struct inpcb *)inph; 1439 #endif 1440 #ifdef INET6 1441 in6p = (const struct in6pcb *)inph; 1442 #endif 1443 1444 if (inph->inph_af != pf) 1445 continue; 1446 1447 if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, 1448 KAUTH_REQ_NETWORK_SOCKET_CANSEE, inph->inph_socket, NULL, 1449 NULL) != 0) 1450 continue; 1451 1452 memset(&pcb, 0, sizeof(pcb)); 1453 1454 pcb.ki_family = pf; 1455 pcb.ki_type = proto; 1456 1457 switch (pf2) { 1458 case 0: 1459 /* just probing for size */ 1460 break; 1461 #ifdef INET 1462 case PF_INET: 1463 pcb.ki_family = inp->inp_socket->so_proto-> 1464 pr_domain->dom_family; 1465 pcb.ki_type = inp->inp_socket->so_proto-> 1466 pr_type; 1467 pcb.ki_protocol = inp->inp_socket->so_proto-> 1468 pr_protocol; 1469 pcb.ki_pflags = inp->inp_flags; 1470 1471 pcb.ki_sostate = inp->inp_socket->so_state; 1472 pcb.ki_prstate = inp->inp_state; 1473 if (proto == IPPROTO_TCP) { 1474 tp = intotcpcb(inp); 1475 pcb.ki_tstate = tp->t_state; 1476 pcb.ki_tflags = tp->t_flags; 1477 } 1478 1479 pcb.ki_pcbaddr = PTRTOUINT64(inp); 1480 pcb.ki_ppcbaddr = PTRTOUINT64(inp->inp_ppcb); 1481 pcb.ki_sockaddr = PTRTOUINT64(inp->inp_socket); 1482 1483 pcb.ki_rcvq = inp->inp_socket->so_rcv.sb_cc; 1484 pcb.ki_sndq = inp->inp_socket->so_snd.sb_cc; 1485 1486 in = satosin(&pcb.ki_src); 1487 in->sin_len = sizeof(*in); 1488 in->sin_family = pf; 1489 in->sin_port = inp->inp_lport; 1490 in->sin_addr = inp->inp_laddr; 1491 if (pcb.ki_prstate >= INP_CONNECTED) { 1492 in = satosin(&pcb.ki_dst); 1493 in->sin_len = sizeof(*in); 1494 in->sin_family = pf; 1495 in->sin_port = inp->inp_fport; 1496 in->sin_addr = inp->inp_faddr; 1497 } 1498 break; 1499 #endif 1500 #ifdef INET6 1501 case PF_INET6: 1502 pcb.ki_family = in6p->in6p_socket->so_proto-> 1503 pr_domain->dom_family; 1504 pcb.ki_type = in6p->in6p_socket->so_proto->pr_type; 1505 pcb.ki_protocol = in6p->in6p_socket->so_proto-> 1506 pr_protocol; 1507 pcb.ki_pflags = in6p->in6p_flags; 1508 1509 pcb.ki_sostate = in6p->in6p_socket->so_state; 1510 pcb.ki_prstate = in6p->in6p_state; 1511 if (proto == IPPROTO_TCP) { 1512 tp = in6totcpcb(in6p); 1513 pcb.ki_tstate = tp->t_state; 1514 pcb.ki_tflags = tp->t_flags; 1515 } 1516 1517 pcb.ki_pcbaddr = PTRTOUINT64(in6p); 1518 pcb.ki_ppcbaddr = PTRTOUINT64(in6p->in6p_ppcb); 1519 pcb.ki_sockaddr = PTRTOUINT64(in6p->in6p_socket); 1520 1521 pcb.ki_rcvq = in6p->in6p_socket->so_rcv.sb_cc; 1522 pcb.ki_sndq = in6p->in6p_socket->so_snd.sb_cc; 1523 1524 in6 = satosin6(&pcb.ki_src); 1525 in6->sin6_len = sizeof(*in6); 1526 in6->sin6_family = pf; 1527 in6->sin6_port = in6p->in6p_lport; 1528 in6->sin6_flowinfo = in6p->in6p_flowinfo; 1529 in6->sin6_addr = in6p->in6p_laddr; 1530 in6->sin6_scope_id = 0; /* XXX? */ 1531 1532 if (pcb.ki_prstate >= IN6P_CONNECTED) { 1533 in6 = satosin6(&pcb.ki_dst); 1534 in6->sin6_len = sizeof(*in6); 1535 in6->sin6_family = pf; 1536 in6->sin6_port = in6p->in6p_fport; 1537 in6->sin6_flowinfo = in6p->in6p_flowinfo; 1538 in6->sin6_addr = in6p->in6p_faddr; 1539 in6->sin6_scope_id = 0; /* XXX? */ 1540 } 1541 break; 1542 #endif 1543 } 1544 1545 if (len >= elem_size && elem_count > 0) { 1546 error = copyout(&pcb, dp, out_size); 1547 if (error) { 1548 mutex_exit(softnet_lock); 1549 return (error); 1550 } 1551 dp += elem_size; 1552 len -= elem_size; 1553 } 1554 needed += elem_size; 1555 if (elem_count > 0 && elem_count != INT_MAX) 1556 elem_count--; 1557 } 1558 1559 *oldlenp = needed; 1560 if (oldp == NULL) 1561 *oldlenp += PCB_SLOP * sizeof(struct kinfo_pcb); 1562 1563 mutex_exit(softnet_lock); 1564 1565 return (error); 1566 } 1567 1568 static int 1569 sysctl_tcp_congctl(SYSCTLFN_ARGS) 1570 { 1571 struct sysctlnode node; 1572 int error; 1573 char newname[TCPCC_MAXLEN]; 1574 1575 strlcpy(newname, tcp_congctl_global_name, sizeof(newname) - 1); 1576 1577 node = *rnode; 1578 node.sysctl_data = newname; 1579 node.sysctl_size = sizeof(newname); 1580 1581 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1582 1583 if (error || 1584 newp == NULL || 1585 strncmp(newname, tcp_congctl_global_name, sizeof(newname)) == 0) 1586 return error; 1587 1588 mutex_enter(softnet_lock); 1589 error = tcp_congctl_select(NULL, newname); 1590 mutex_exit(softnet_lock); 1591 1592 return error; 1593 } 1594 1595 static int 1596 sysctl_tcp_keep(SYSCTLFN_ARGS) 1597 { 1598 int error; 1599 u_int tmp; 1600 struct sysctlnode node; 1601 1602 node = *rnode; 1603 tmp = *(u_int *)rnode->sysctl_data; 1604 node.sysctl_data = &tmp; 1605 1606 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1607 if (error || newp == NULL) 1608 return error; 1609 1610 mutex_enter(softnet_lock); 1611 1612 *(u_int *)rnode->sysctl_data = tmp; 1613 tcp_tcpcb_template(); /* update the template */ 1614 1615 mutex_exit(softnet_lock); 1616 return 0; 1617 } 1618 1619 static int 1620 sysctl_net_inet_tcp_stats(SYSCTLFN_ARGS) 1621 { 1622 1623 return (NETSTAT_SYSCTL(tcpstat_percpu, TCP_NSTATS)); 1624 } 1625 1626 /* 1627 * this (second stage) setup routine is a replacement for tcp_sysctl() 1628 * (which is currently used for ipv4 and ipv6) 1629 */ 1630 static void 1631 sysctl_net_inet_tcp_setup2(struct sysctllog **clog, int pf, const char *pfname, 1632 const char *tcpname) 1633 { 1634 const struct sysctlnode *sack_node; 1635 const struct sysctlnode *abc_node; 1636 const struct sysctlnode *ecn_node; 1637 const struct sysctlnode *congctl_node; 1638 #ifdef TCP_DEBUG 1639 extern struct tcp_debug tcp_debug[TCP_NDEBUG]; 1640 extern int tcp_debx; 1641 #endif 1642 1643 sysctl_createv(clog, 0, NULL, NULL, 1644 CTLFLAG_PERMANENT, 1645 CTLTYPE_NODE, "net", NULL, 1646 NULL, 0, NULL, 0, 1647 CTL_NET, CTL_EOL); 1648 sysctl_createv(clog, 0, NULL, NULL, 1649 CTLFLAG_PERMANENT, 1650 CTLTYPE_NODE, pfname, NULL, 1651 NULL, 0, NULL, 0, 1652 CTL_NET, pf, CTL_EOL); 1653 sysctl_createv(clog, 0, NULL, NULL, 1654 CTLFLAG_PERMANENT, 1655 CTLTYPE_NODE, tcpname, 1656 SYSCTL_DESCR("TCP related settings"), 1657 NULL, 0, NULL, 0, 1658 CTL_NET, pf, IPPROTO_TCP, CTL_EOL); 1659 1660 sysctl_createv(clog, 0, NULL, NULL, 1661 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1662 CTLTYPE_INT, "rfc1323", 1663 SYSCTL_DESCR("Enable RFC1323 TCP extensions"), 1664 NULL, 0, &tcp_do_rfc1323, 0, 1665 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RFC1323, CTL_EOL); 1666 sysctl_createv(clog, 0, NULL, NULL, 1667 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1668 CTLTYPE_INT, "sendspace", 1669 SYSCTL_DESCR("Default TCP send buffer size"), 1670 NULL, 0, &tcp_sendspace, 0, 1671 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SENDSPACE, CTL_EOL); 1672 sysctl_createv(clog, 0, NULL, NULL, 1673 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1674 CTLTYPE_INT, "recvspace", 1675 SYSCTL_DESCR("Default TCP receive buffer size"), 1676 NULL, 0, &tcp_recvspace, 0, 1677 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RECVSPACE, CTL_EOL); 1678 sysctl_createv(clog, 0, NULL, NULL, 1679 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1680 CTLTYPE_INT, "mssdflt", 1681 SYSCTL_DESCR("Default maximum segment size"), 1682 sysctl_net_inet_tcp_mssdflt, 0, &tcp_mssdflt, 0, 1683 CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSSDFLT, CTL_EOL); 1684 sysctl_createv(clog, 0, NULL, NULL, 1685 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1686 CTLTYPE_INT, "minmss", 1687 SYSCTL_DESCR("Lower limit for TCP maximum segment size"), 1688 NULL, 0, &tcp_minmss, 0, 1689 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1690 sysctl_createv(clog, 0, NULL, NULL, 1691 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1692 CTLTYPE_INT, "msl", 1693 SYSCTL_DESCR("Maximum Segment Life"), 1694 NULL, 0, &tcp_msl, 0, 1695 CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSL, CTL_EOL); 1696 sysctl_createv(clog, 0, NULL, NULL, 1697 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1698 CTLTYPE_INT, "syn_cache_limit", 1699 SYSCTL_DESCR("Maximum number of entries in the TCP " 1700 "compressed state engine"), 1701 NULL, 0, &tcp_syn_cache_limit, 0, 1702 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_LIMIT, 1703 CTL_EOL); 1704 sysctl_createv(clog, 0, NULL, NULL, 1705 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1706 CTLTYPE_INT, "syn_bucket_limit", 1707 SYSCTL_DESCR("Maximum number of entries per hash " 1708 "bucket in the TCP compressed state " 1709 "engine"), 1710 NULL, 0, &tcp_syn_bucket_limit, 0, 1711 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_BUCKET_LIMIT, 1712 CTL_EOL); 1713 #if 0 /* obsoleted */ 1714 sysctl_createv(clog, 0, NULL, NULL, 1715 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1716 CTLTYPE_INT, "syn_cache_interval", 1717 SYSCTL_DESCR("TCP compressed state engine's timer interval"), 1718 NULL, 0, &tcp_syn_cache_interval, 0, 1719 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_INTER, 1720 CTL_EOL); 1721 #endif 1722 sysctl_createv(clog, 0, NULL, NULL, 1723 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1724 CTLTYPE_INT, "init_win", 1725 SYSCTL_DESCR("Initial TCP congestion window"), 1726 NULL, 0, &tcp_init_win, 0, 1727 CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN, CTL_EOL); 1728 sysctl_createv(clog, 0, NULL, NULL, 1729 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1730 CTLTYPE_INT, "mss_ifmtu", 1731 SYSCTL_DESCR("Use interface MTU for calculating MSS"), 1732 NULL, 0, &tcp_mss_ifmtu, 0, 1733 CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSS_IFMTU, CTL_EOL); 1734 sysctl_createv(clog, 0, NULL, &sack_node, 1735 CTLFLAG_PERMANENT, 1736 CTLTYPE_NODE, "sack", 1737 SYSCTL_DESCR("RFC2018 Selective ACKnowledgement tunables"), 1738 NULL, 0, NULL, 0, 1739 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_EOL); 1740 1741 /* Congctl subtree */ 1742 sysctl_createv(clog, 0, NULL, &congctl_node, 1743 CTLFLAG_PERMANENT, 1744 CTLTYPE_NODE, "congctl", 1745 SYSCTL_DESCR("TCP Congestion Control"), 1746 NULL, 0, NULL, 0, 1747 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1748 sysctl_createv(clog, 0, &congctl_node, NULL, 1749 CTLFLAG_PERMANENT, 1750 CTLTYPE_STRING, "available", 1751 SYSCTL_DESCR("Available Congestion Control Mechanisms"), 1752 NULL, 0, &tcp_congctl_avail, 0, CTL_CREATE, CTL_EOL); 1753 sysctl_createv(clog, 0, &congctl_node, NULL, 1754 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1755 CTLTYPE_STRING, "selected", 1756 SYSCTL_DESCR("Selected Congestion Control Mechanism"), 1757 sysctl_tcp_congctl, 0, NULL, TCPCC_MAXLEN, 1758 CTL_CREATE, CTL_EOL); 1759 1760 sysctl_createv(clog, 0, NULL, NULL, 1761 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1762 CTLTYPE_INT, "win_scale", 1763 SYSCTL_DESCR("Use RFC1323 window scale options"), 1764 NULL, 0, &tcp_do_win_scale, 0, 1765 CTL_NET, pf, IPPROTO_TCP, TCPCTL_WSCALE, CTL_EOL); 1766 sysctl_createv(clog, 0, NULL, NULL, 1767 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1768 CTLTYPE_INT, "timestamps", 1769 SYSCTL_DESCR("Use RFC1323 time stamp options"), 1770 NULL, 0, &tcp_do_timestamps, 0, 1771 CTL_NET, pf, IPPROTO_TCP, TCPCTL_TSTAMP, CTL_EOL); 1772 sysctl_createv(clog, 0, NULL, NULL, 1773 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1774 CTLTYPE_INT, "compat_42", 1775 SYSCTL_DESCR("Enable workarounds for 4.2BSD TCP bugs"), 1776 NULL, 0, &tcp_compat_42, 0, 1777 CTL_NET, pf, IPPROTO_TCP, TCPCTL_COMPAT_42, CTL_EOL); 1778 sysctl_createv(clog, 0, NULL, NULL, 1779 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1780 CTLTYPE_INT, "cwm", 1781 SYSCTL_DESCR("Hughes/Touch/Heidemann Congestion Window " 1782 "Monitoring"), 1783 NULL, 0, &tcp_cwm, 0, 1784 CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM, CTL_EOL); 1785 sysctl_createv(clog, 0, NULL, NULL, 1786 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1787 CTLTYPE_INT, "cwm_burstsize", 1788 SYSCTL_DESCR("Congestion Window Monitoring allowed " 1789 "burst count in packets"), 1790 NULL, 0, &tcp_cwm_burstsize, 0, 1791 CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM_BURSTSIZE, 1792 CTL_EOL); 1793 sysctl_createv(clog, 0, NULL, NULL, 1794 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1795 CTLTYPE_INT, "ack_on_push", 1796 SYSCTL_DESCR("Immediately return ACK when PSH is " 1797 "received"), 1798 NULL, 0, &tcp_ack_on_push, 0, 1799 CTL_NET, pf, IPPROTO_TCP, TCPCTL_ACK_ON_PUSH, CTL_EOL); 1800 sysctl_createv(clog, 0, NULL, NULL, 1801 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1802 CTLTYPE_INT, "keepidle", 1803 SYSCTL_DESCR("Allowed connection idle ticks before a " 1804 "keepalive probe is sent"), 1805 sysctl_tcp_keep, 0, &tcp_keepidle, 0, 1806 CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPIDLE, CTL_EOL); 1807 sysctl_createv(clog, 0, NULL, NULL, 1808 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1809 CTLTYPE_INT, "keepintvl", 1810 SYSCTL_DESCR("Ticks before next keepalive probe is sent"), 1811 sysctl_tcp_keep, 0, &tcp_keepintvl, 0, 1812 CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPINTVL, CTL_EOL); 1813 sysctl_createv(clog, 0, NULL, NULL, 1814 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1815 CTLTYPE_INT, "keepcnt", 1816 SYSCTL_DESCR("Number of keepalive probes to send"), 1817 sysctl_tcp_keep, 0, &tcp_keepcnt, 0, 1818 CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPCNT, CTL_EOL); 1819 sysctl_createv(clog, 0, NULL, NULL, 1820 CTLFLAG_PERMANENT|CTLFLAG_IMMEDIATE, 1821 CTLTYPE_INT, "slowhz", 1822 SYSCTL_DESCR("Keepalive ticks per second"), 1823 NULL, PR_SLOWHZ, NULL, 0, 1824 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SLOWHZ, CTL_EOL); 1825 sysctl_createv(clog, 0, NULL, NULL, 1826 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1827 CTLTYPE_INT, "log_refused", 1828 SYSCTL_DESCR("Log refused TCP connections"), 1829 NULL, 0, &tcp_log_refused, 0, 1830 CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOG_REFUSED, CTL_EOL); 1831 #if 0 /* obsoleted */ 1832 sysctl_createv(clog, 0, NULL, NULL, 1833 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1834 CTLTYPE_INT, "rstratelimit", NULL, 1835 NULL, 0, &tcp_rst_ratelim, 0, 1836 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTRATELIMIT, CTL_EOL); 1837 #endif 1838 sysctl_createv(clog, 0, NULL, NULL, 1839 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1840 CTLTYPE_INT, "rstppslimit", 1841 SYSCTL_DESCR("Maximum number of RST packets to send " 1842 "per second"), 1843 NULL, 0, &tcp_rst_ppslim, 0, 1844 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTPPSLIMIT, CTL_EOL); 1845 sysctl_createv(clog, 0, NULL, NULL, 1846 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1847 CTLTYPE_INT, "delack_ticks", 1848 SYSCTL_DESCR("Number of ticks to delay sending an ACK"), 1849 NULL, 0, &tcp_delack_ticks, 0, 1850 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DELACK_TICKS, CTL_EOL); 1851 sysctl_createv(clog, 0, NULL, NULL, 1852 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1853 CTLTYPE_INT, "init_win_local", 1854 SYSCTL_DESCR("Initial TCP window size (in segments)"), 1855 NULL, 0, &tcp_init_win_local, 0, 1856 CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN_LOCAL, 1857 CTL_EOL); 1858 sysctl_createv(clog, 0, NULL, NULL, 1859 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1860 CTLTYPE_STRUCT, "ident", 1861 SYSCTL_DESCR("RFC1413 Identification Protocol lookups"), 1862 sysctl_net_inet_tcp_ident, 0, NULL, sizeof(uid_t), 1863 CTL_NET, pf, IPPROTO_TCP, TCPCTL_IDENT, CTL_EOL); 1864 sysctl_createv(clog, 0, NULL, NULL, 1865 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1866 CTLTYPE_INT, "do_loopback_cksum", 1867 SYSCTL_DESCR("Perform TCP checksum on loopback"), 1868 NULL, 0, &tcp_do_loopback_cksum, 0, 1869 CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOOPBACKCKSUM, 1870 CTL_EOL); 1871 sysctl_createv(clog, 0, NULL, NULL, 1872 CTLFLAG_PERMANENT, 1873 CTLTYPE_STRUCT, "pcblist", 1874 SYSCTL_DESCR("TCP protocol control block list"), 1875 sysctl_inpcblist, 0, &tcbtable, 0, 1876 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, 1877 CTL_EOL); 1878 sysctl_createv(clog, 0, NULL, NULL, 1879 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1880 CTLTYPE_INT, "keepinit", 1881 SYSCTL_DESCR("Ticks before initial tcp connection times out"), 1882 sysctl_tcp_keep, 0, &tcp_keepinit, 0, 1883 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1884 1885 /* TCP socket buffers auto-sizing nodes */ 1886 sysctl_createv(clog, 0, NULL, NULL, 1887 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1888 CTLTYPE_INT, "recvbuf_auto", 1889 SYSCTL_DESCR("Enable automatic receive " 1890 "buffer sizing (experimental)"), 1891 NULL, 0, &tcp_do_autorcvbuf, 0, 1892 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1893 sysctl_createv(clog, 0, NULL, NULL, 1894 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1895 CTLTYPE_INT, "recvbuf_inc", 1896 SYSCTL_DESCR("Incrementor step size of " 1897 "automatic receive buffer"), 1898 NULL, 0, &tcp_autorcvbuf_inc, 0, 1899 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1900 sysctl_createv(clog, 0, NULL, NULL, 1901 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1902 CTLTYPE_INT, "recvbuf_max", 1903 SYSCTL_DESCR("Max size of automatic receive buffer"), 1904 NULL, 0, &tcp_autorcvbuf_max, 0, 1905 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1906 1907 sysctl_createv(clog, 0, NULL, NULL, 1908 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1909 CTLTYPE_INT, "sendbuf_auto", 1910 SYSCTL_DESCR("Enable automatic send " 1911 "buffer sizing (experimental)"), 1912 NULL, 0, &tcp_do_autosndbuf, 0, 1913 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1914 sysctl_createv(clog, 0, NULL, NULL, 1915 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1916 CTLTYPE_INT, "sendbuf_inc", 1917 SYSCTL_DESCR("Incrementor step size of " 1918 "automatic send buffer"), 1919 NULL, 0, &tcp_autosndbuf_inc, 0, 1920 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1921 sysctl_createv(clog, 0, NULL, NULL, 1922 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1923 CTLTYPE_INT, "sendbuf_max", 1924 SYSCTL_DESCR("Max size of automatic send buffer"), 1925 NULL, 0, &tcp_autosndbuf_max, 0, 1926 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1927 1928 /* ECN subtree */ 1929 sysctl_createv(clog, 0, NULL, &ecn_node, 1930 CTLFLAG_PERMANENT, 1931 CTLTYPE_NODE, "ecn", 1932 SYSCTL_DESCR("RFC3168 Explicit Congestion Notification"), 1933 NULL, 0, NULL, 0, 1934 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1935 sysctl_createv(clog, 0, &ecn_node, NULL, 1936 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1937 CTLTYPE_INT, "enable", 1938 SYSCTL_DESCR("Enable TCP Explicit Congestion " 1939 "Notification"), 1940 NULL, 0, &tcp_do_ecn, 0, CTL_CREATE, CTL_EOL); 1941 sysctl_createv(clog, 0, &ecn_node, NULL, 1942 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1943 CTLTYPE_INT, "maxretries", 1944 SYSCTL_DESCR("Number of times to retry ECN setup " 1945 "before disabling ECN on the connection"), 1946 NULL, 0, &tcp_ecn_maxretries, 0, CTL_CREATE, CTL_EOL); 1947 1948 /* SACK gets it's own little subtree. */ 1949 sysctl_createv(clog, 0, NULL, &sack_node, 1950 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1951 CTLTYPE_INT, "enable", 1952 SYSCTL_DESCR("Enable RFC2018 Selective ACKnowledgement"), 1953 NULL, 0, &tcp_do_sack, 0, 1954 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 1955 sysctl_createv(clog, 0, NULL, &sack_node, 1956 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1957 CTLTYPE_INT, "maxholes", 1958 SYSCTL_DESCR("Maximum number of TCP SACK holes allowed per connection"), 1959 NULL, 0, &tcp_sack_tp_maxholes, 0, 1960 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 1961 sysctl_createv(clog, 0, NULL, &sack_node, 1962 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1963 CTLTYPE_INT, "globalmaxholes", 1964 SYSCTL_DESCR("Global maximum number of TCP SACK holes"), 1965 NULL, 0, &tcp_sack_globalmaxholes, 0, 1966 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 1967 sysctl_createv(clog, 0, NULL, &sack_node, 1968 CTLFLAG_PERMANENT, 1969 CTLTYPE_INT, "globalholes", 1970 SYSCTL_DESCR("Global number of TCP SACK holes"), 1971 NULL, 0, &tcp_sack_globalholes, 0, 1972 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 1973 1974 sysctl_createv(clog, 0, NULL, NULL, 1975 CTLFLAG_PERMANENT, 1976 CTLTYPE_STRUCT, "stats", 1977 SYSCTL_DESCR("TCP statistics"), 1978 sysctl_net_inet_tcp_stats, 0, NULL, 0, 1979 CTL_NET, pf, IPPROTO_TCP, TCPCTL_STATS, 1980 CTL_EOL); 1981 #ifdef TCP_DEBUG 1982 sysctl_createv(clog, 0, NULL, NULL, 1983 CTLFLAG_PERMANENT, 1984 CTLTYPE_STRUCT, "debug", 1985 SYSCTL_DESCR("TCP sockets debug information"), 1986 NULL, 0, &tcp_debug, sizeof(tcp_debug), 1987 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBUG, 1988 CTL_EOL); 1989 sysctl_createv(clog, 0, NULL, NULL, 1990 CTLFLAG_PERMANENT, 1991 CTLTYPE_INT, "debx", 1992 SYSCTL_DESCR("Number of TCP debug sockets messages"), 1993 NULL, 0, &tcp_debx, sizeof(tcp_debx), 1994 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBX, 1995 CTL_EOL); 1996 #endif 1997 sysctl_createv(clog, 0, NULL, NULL, 1998 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1999 CTLTYPE_STRUCT, "drop", 2000 SYSCTL_DESCR("TCP drop connection"), 2001 sysctl_net_inet_tcp_drop, 0, NULL, 0, 2002 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DROP, CTL_EOL); 2003 #if NRND > 0 2004 sysctl_createv(clog, 0, NULL, NULL, 2005 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2006 CTLTYPE_INT, "iss_hash", 2007 SYSCTL_DESCR("Enable RFC 1948 ISS by cryptographic " 2008 "hash computation"), 2009 NULL, 0, &tcp_do_rfc1948, sizeof(tcp_do_rfc1948), 2010 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, 2011 CTL_EOL); 2012 #endif 2013 2014 /* ABC subtree */ 2015 2016 sysctl_createv(clog, 0, NULL, &abc_node, 2017 CTLFLAG_PERMANENT, CTLTYPE_NODE, "abc", 2018 SYSCTL_DESCR("RFC3465 Appropriate Byte Counting (ABC)"), 2019 NULL, 0, NULL, 0, 2020 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2021 sysctl_createv(clog, 0, &abc_node, NULL, 2022 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2023 CTLTYPE_INT, "enable", 2024 SYSCTL_DESCR("Enable RFC3465 Appropriate Byte Counting"), 2025 NULL, 0, &tcp_do_abc, 0, CTL_CREATE, CTL_EOL); 2026 sysctl_createv(clog, 0, &abc_node, NULL, 2027 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2028 CTLTYPE_INT, "aggressive", 2029 SYSCTL_DESCR("1: L=2*SMSS 0: L=1*SMSS"), 2030 NULL, 0, &tcp_abc_aggressive, 0, CTL_CREATE, CTL_EOL); 2031 } 2032 2033 void 2034 tcp_usrreq_init(void) 2035 { 2036 2037 #ifdef INET 2038 sysctl_net_inet_tcp_setup2(NULL, PF_INET, "inet", "tcp"); 2039 #endif 2040 #ifdef INET6 2041 sysctl_net_inet_tcp_setup2(NULL, PF_INET6, "inet6", "tcp6"); 2042 #endif 2043 } 2044