tcp_usrreq.c revision 1.174
1 /* $NetBSD: tcp_usrreq.c,v 1.174 2014/05/20 19:04:00 rmind 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 /* 98 * TCP protocol interface to socket abstraction. 99 */ 100 101 #include <sys/cdefs.h> 102 __KERNEL_RCSID(0, "$NetBSD: tcp_usrreq.c,v 1.174 2014/05/20 19:04:00 rmind Exp $"); 103 104 #include "opt_inet.h" 105 #include "opt_ipsec.h" 106 #include "opt_tcp_debug.h" 107 #include "opt_mbuftrace.h" 108 109 #include <sys/param.h> 110 #include <sys/systm.h> 111 #include <sys/kernel.h> 112 #include <sys/malloc.h> 113 #include <sys/mbuf.h> 114 #include <sys/socket.h> 115 #include <sys/socketvar.h> 116 #include <sys/protosw.h> 117 #include <sys/errno.h> 118 #include <sys/stat.h> 119 #include <sys/proc.h> 120 #include <sys/domain.h> 121 #include <sys/sysctl.h> 122 #include <sys/kauth.h> 123 #include <sys/uidinfo.h> 124 125 #include <net/if.h> 126 #include <net/route.h> 127 128 #include <netinet/in.h> 129 #include <netinet/in_systm.h> 130 #include <netinet/in_var.h> 131 #include <netinet/ip.h> 132 #include <netinet/in_pcb.h> 133 #include <netinet/ip_var.h> 134 #include <netinet/in_offload.h> 135 136 #ifdef INET6 137 #ifndef INET 138 #include <netinet/in.h> 139 #endif 140 #include <netinet/ip6.h> 141 #include <netinet6/in6_pcb.h> 142 #include <netinet6/ip6_var.h> 143 #include <netinet6/scope6_var.h> 144 #endif 145 146 #include <netinet/tcp.h> 147 #include <netinet/tcp_fsm.h> 148 #include <netinet/tcp_seq.h> 149 #include <netinet/tcp_timer.h> 150 #include <netinet/tcp_var.h> 151 #include <netinet/tcp_private.h> 152 #include <netinet/tcp_congctl.h> 153 #include <netinet/tcpip.h> 154 #include <netinet/tcp_debug.h> 155 #include <netinet/tcp_vtw.h> 156 157 #include "opt_tcp_space.h" 158 159 /* 160 * Process a TCP user request for TCP tb. If this is a send request 161 * then m is the mbuf chain of send data. If this is a timer expiration 162 * (called from the software clock routine), then timertype tells which timer. 163 */ 164 static int 165 tcp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam, 166 struct mbuf *control, struct lwp *l) 167 { 168 struct inpcb *inp; 169 #ifdef INET6 170 struct in6pcb *in6p; 171 #endif 172 struct tcpcb *tp = NULL; 173 int s; 174 int error = 0; 175 #ifdef TCP_DEBUG 176 int ostate = 0; 177 #endif 178 int family; /* family of the socket */ 179 180 KASSERT(req != PRU_ATTACH); 181 KASSERT(req != PRU_DETACH); 182 183 family = so->so_proto->pr_domain->dom_family; 184 185 if (req == PRU_CONTROL) { 186 switch (family) { 187 #ifdef INET 188 case PF_INET: 189 return (in_control(so, (long)m, (void *)nam, 190 (struct ifnet *)control, l)); 191 #endif 192 #ifdef INET6 193 case PF_INET6: 194 return (in6_control(so, (long)m, (void *)nam, 195 (struct ifnet *)control, l)); 196 #endif 197 default: 198 return EAFNOSUPPORT; 199 } 200 } 201 202 s = splsoftnet(); 203 204 if (req == PRU_PURGEIF) { 205 mutex_enter(softnet_lock); 206 switch (family) { 207 #ifdef INET 208 case PF_INET: 209 in_pcbpurgeif0(&tcbtable, (struct ifnet *)control); 210 in_purgeif((struct ifnet *)control); 211 in_pcbpurgeif(&tcbtable, (struct ifnet *)control); 212 break; 213 #endif 214 #ifdef INET6 215 case PF_INET6: 216 in6_pcbpurgeif0(&tcbtable, (struct ifnet *)control); 217 in6_purgeif((struct ifnet *)control); 218 in6_pcbpurgeif(&tcbtable, (struct ifnet *)control); 219 break; 220 #endif 221 default: 222 mutex_exit(softnet_lock); 223 splx(s); 224 return (EAFNOSUPPORT); 225 } 226 mutex_exit(softnet_lock); 227 splx(s); 228 return (0); 229 } 230 231 KASSERT(solocked(so)); 232 233 switch (family) { 234 #ifdef INET 235 case PF_INET: 236 inp = sotoinpcb(so); 237 #ifdef INET6 238 in6p = NULL; 239 #endif 240 break; 241 #endif 242 #ifdef INET6 243 case PF_INET6: 244 inp = NULL; 245 in6p = sotoin6pcb(so); 246 break; 247 #endif 248 default: 249 splx(s); 250 return EAFNOSUPPORT; 251 } 252 253 #ifdef INET6 254 KASSERT((inp != NULL) ^ (in6p != NULL)); 255 #endif 256 KASSERT(!control || (req == PRU_SEND || req == PRU_SENDOOB)); 257 258 /* 259 * When a TCP is attached to a socket, then there will be 260 * a (struct inpcb) pointed at by the socket, and this 261 * structure will point at a subsidary (struct tcpcb). 262 */ 263 if ((inp == NULL 264 #ifdef INET6 265 && in6p == NULL 266 #endif 267 ) && req != PRU_SENSE) 268 { 269 error = EINVAL; 270 goto release; 271 } 272 #ifdef INET 273 if (inp) { 274 tp = intotcpcb(inp); 275 /* WHAT IF TP IS 0? */ 276 #ifdef KPROF 277 tcp_acounts[tp->t_state][req]++; 278 #endif 279 #ifdef TCP_DEBUG 280 ostate = tp->t_state; 281 #endif 282 } 283 #endif 284 #ifdef INET6 285 if (in6p) { 286 tp = in6totcpcb(in6p); 287 /* WHAT IF TP IS 0? */ 288 #ifdef KPROF 289 tcp_acounts[tp->t_state][req]++; 290 #endif 291 #ifdef TCP_DEBUG 292 ostate = tp->t_state; 293 #endif 294 } 295 #endif 296 297 switch (req) { 298 299 /* 300 * Give the socket an address. 301 */ 302 case PRU_BIND: 303 switch (family) { 304 #ifdef INET 305 case PF_INET: 306 error = in_pcbbind(inp, nam, l); 307 break; 308 #endif 309 #ifdef INET6 310 case PF_INET6: 311 error = in6_pcbbind(in6p, nam, l); 312 if (!error) { 313 /* mapped addr case */ 314 if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)) 315 tp->t_family = AF_INET; 316 else 317 tp->t_family = AF_INET6; 318 } 319 break; 320 #endif 321 } 322 break; 323 324 /* 325 * Prepare to accept connections. 326 */ 327 case PRU_LISTEN: 328 #ifdef INET 329 if (inp && inp->inp_lport == 0) { 330 error = in_pcbbind(inp, NULL, l); 331 if (error) 332 break; 333 } 334 #endif 335 #ifdef INET6 336 if (in6p && in6p->in6p_lport == 0) { 337 error = in6_pcbbind(in6p, NULL, l); 338 if (error) 339 break; 340 } 341 #endif 342 tp->t_state = TCPS_LISTEN; 343 break; 344 345 /* 346 * Initiate connection to peer. 347 * Create a template for use in transmissions on this connection. 348 * Enter SYN_SENT state, and mark socket as connecting. 349 * Start keep-alive timer, and seed output sequence space. 350 * Send initial segment on connection. 351 */ 352 case PRU_CONNECT: 353 #ifdef INET 354 if (inp) { 355 if (inp->inp_lport == 0) { 356 error = in_pcbbind(inp, NULL, l); 357 if (error) 358 break; 359 } 360 error = in_pcbconnect(inp, nam, l); 361 } 362 #endif 363 #ifdef INET6 364 if (in6p) { 365 if (in6p->in6p_lport == 0) { 366 error = in6_pcbbind(in6p, NULL, l); 367 if (error) 368 break; 369 } 370 error = in6_pcbconnect(in6p, nam, l); 371 if (!error) { 372 /* mapped addr case */ 373 if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) 374 tp->t_family = AF_INET; 375 else 376 tp->t_family = AF_INET6; 377 } 378 } 379 #endif 380 if (error) 381 break; 382 tp->t_template = tcp_template(tp); 383 if (tp->t_template == 0) { 384 #ifdef INET 385 if (inp) 386 in_pcbdisconnect(inp); 387 #endif 388 #ifdef INET6 389 if (in6p) 390 in6_pcbdisconnect(in6p); 391 #endif 392 error = ENOBUFS; 393 break; 394 } 395 /* 396 * Compute window scaling to request. 397 * XXX: This should be moved to tcp_output(). 398 */ 399 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 400 (TCP_MAXWIN << tp->request_r_scale) < sb_max) 401 tp->request_r_scale++; 402 soisconnecting(so); 403 TCP_STATINC(TCP_STAT_CONNATTEMPT); 404 tp->t_state = TCPS_SYN_SENT; 405 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit); 406 tp->iss = tcp_new_iss(tp, 0); 407 tcp_sendseqinit(tp); 408 error = tcp_output(tp); 409 break; 410 411 /* 412 * Create a TCP connection between two sockets. 413 */ 414 case PRU_CONNECT2: 415 error = EOPNOTSUPP; 416 break; 417 418 /* 419 * Initiate disconnect from peer. 420 * If connection never passed embryonic stage, just drop; 421 * else if don't need to let data drain, then can just drop anyways, 422 * else have to begin TCP shutdown process: mark socket disconnecting, 423 * drain unread data, state switch to reflect user close, and 424 * send segment (e.g. FIN) to peer. Socket will be really disconnected 425 * when peer sends FIN and acks ours. 426 * 427 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 428 */ 429 case PRU_DISCONNECT: 430 tp = tcp_disconnect(tp); 431 break; 432 433 /* 434 * Accept a connection. Essentially all the work is 435 * done at higher levels; just return the address 436 * of the peer, storing through addr. 437 */ 438 case PRU_ACCEPT: 439 #ifdef INET 440 if (inp) 441 in_setpeeraddr(inp, nam); 442 #endif 443 #ifdef INET6 444 if (in6p) 445 in6_setpeeraddr(in6p, nam); 446 #endif 447 break; 448 449 /* 450 * Mark the connection as being incapable of further output. 451 */ 452 case PRU_SHUTDOWN: 453 socantsendmore(so); 454 tp = tcp_usrclosed(tp); 455 if (tp) 456 error = tcp_output(tp); 457 break; 458 459 /* 460 * After a receive, possibly send window update to peer. 461 */ 462 case PRU_RCVD: 463 /* 464 * soreceive() calls this function when a user receives 465 * ancillary data on a listening socket. We don't call 466 * tcp_output in such a case, since there is no header 467 * template for a listening socket and hence the kernel 468 * will panic. 469 */ 470 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) != 0) 471 (void) tcp_output(tp); 472 break; 473 474 /* 475 * Do a send by putting data in output queue and updating urgent 476 * marker if URG set. Possibly send more data. 477 */ 478 case PRU_SEND: 479 if (control && control->m_len) { 480 m_freem(control); 481 m_freem(m); 482 error = EINVAL; 483 break; 484 } 485 sbappendstream(&so->so_snd, m); 486 error = tcp_output(tp); 487 break; 488 489 /* 490 * Abort the TCP. 491 */ 492 case PRU_ABORT: 493 tp = tcp_drop(tp, ECONNABORTED); 494 break; 495 496 case PRU_SENSE: 497 /* 498 * stat: don't bother with a blocksize. 499 */ 500 splx(s); 501 return (0); 502 503 case PRU_RCVOOB: 504 if (control && control->m_len) { 505 m_freem(control); 506 m_freem(m); 507 error = EINVAL; 508 break; 509 } 510 if ((so->so_oobmark == 0 && 511 (so->so_state & SS_RCVATMARK) == 0) || 512 so->so_options & SO_OOBINLINE || 513 tp->t_oobflags & TCPOOB_HADDATA) { 514 error = EINVAL; 515 break; 516 } 517 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 518 error = EWOULDBLOCK; 519 break; 520 } 521 m->m_len = 1; 522 *mtod(m, char *) = tp->t_iobc; 523 if (((long)nam & MSG_PEEK) == 0) 524 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 525 break; 526 527 case PRU_SENDOOB: 528 if (sbspace(&so->so_snd) < -512) { 529 m_freem(m); 530 error = ENOBUFS; 531 break; 532 } 533 /* 534 * According to RFC961 (Assigned Protocols), 535 * the urgent pointer points to the last octet 536 * of urgent data. We continue, however, 537 * to consider it to indicate the first octet 538 * of data past the urgent section. 539 * Otherwise, snd_up should be one lower. 540 */ 541 sbappendstream(&so->so_snd, m); 542 tp->snd_up = tp->snd_una + so->so_snd.sb_cc; 543 tp->t_force = 1; 544 error = tcp_output(tp); 545 tp->t_force = 0; 546 break; 547 548 case PRU_SOCKADDR: 549 #ifdef INET 550 if (inp) 551 in_setsockaddr(inp, nam); 552 #endif 553 #ifdef INET6 554 if (in6p) 555 in6_setsockaddr(in6p, nam); 556 #endif 557 break; 558 559 case PRU_PEERADDR: 560 #ifdef INET 561 if (inp) 562 in_setpeeraddr(inp, nam); 563 #endif 564 #ifdef INET6 565 if (in6p) 566 in6_setpeeraddr(in6p, nam); 567 #endif 568 break; 569 570 default: 571 panic("tcp_usrreq"); 572 } 573 #ifdef TCP_DEBUG 574 if (tp && (so->so_options & SO_DEBUG)) 575 tcp_trace(TA_USER, ostate, tp, NULL, req); 576 #endif 577 578 release: 579 splx(s); 580 return (error); 581 } 582 583 static void 584 change_keepalive(struct socket *so, struct tcpcb *tp) 585 { 586 tp->t_maxidle = tp->t_keepcnt * tp->t_keepintvl; 587 TCP_TIMER_DISARM(tp, TCPT_KEEP); 588 TCP_TIMER_DISARM(tp, TCPT_2MSL); 589 590 if (tp->t_state == TCPS_SYN_RECEIVED || 591 tp->t_state == TCPS_SYN_SENT) { 592 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit); 593 } else if (so->so_options & SO_KEEPALIVE && 594 tp->t_state <= TCPS_CLOSE_WAIT) { 595 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepintvl); 596 } else { 597 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle); 598 } 599 600 if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0)) 601 TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle); 602 } 603 604 605 int 606 tcp_ctloutput(int op, struct socket *so, struct sockopt *sopt) 607 { 608 int error = 0, s; 609 struct inpcb *inp; 610 #ifdef INET6 611 struct in6pcb *in6p; 612 #endif 613 struct tcpcb *tp; 614 u_int ui; 615 int family; /* family of the socket */ 616 int level, optname, optval; 617 618 level = sopt->sopt_level; 619 optname = sopt->sopt_name; 620 621 family = so->so_proto->pr_domain->dom_family; 622 623 s = splsoftnet(); 624 switch (family) { 625 #ifdef INET 626 case PF_INET: 627 inp = sotoinpcb(so); 628 #ifdef INET6 629 in6p = NULL; 630 #endif 631 break; 632 #endif 633 #ifdef INET6 634 case PF_INET6: 635 inp = NULL; 636 in6p = sotoin6pcb(so); 637 break; 638 #endif 639 default: 640 splx(s); 641 panic("%s: af %d", __func__, family); 642 } 643 #ifndef INET6 644 if (inp == NULL) 645 #else 646 if (inp == NULL && in6p == NULL) 647 #endif 648 { 649 splx(s); 650 return (ECONNRESET); 651 } 652 if (level != IPPROTO_TCP) { 653 switch (family) { 654 #ifdef INET 655 case PF_INET: 656 error = ip_ctloutput(op, so, sopt); 657 break; 658 #endif 659 #ifdef INET6 660 case PF_INET6: 661 error = ip6_ctloutput(op, so, sopt); 662 break; 663 #endif 664 } 665 splx(s); 666 return (error); 667 } 668 if (inp) 669 tp = intotcpcb(inp); 670 #ifdef INET6 671 else if (in6p) 672 tp = in6totcpcb(in6p); 673 #endif 674 else 675 tp = NULL; 676 677 switch (op) { 678 case PRCO_SETOPT: 679 switch (optname) { 680 #ifdef TCP_SIGNATURE 681 case TCP_MD5SIG: 682 error = sockopt_getint(sopt, &optval); 683 if (error) 684 break; 685 if (optval > 0) 686 tp->t_flags |= TF_SIGNATURE; 687 else 688 tp->t_flags &= ~TF_SIGNATURE; 689 break; 690 #endif /* TCP_SIGNATURE */ 691 692 case TCP_NODELAY: 693 error = sockopt_getint(sopt, &optval); 694 if (error) 695 break; 696 if (optval) 697 tp->t_flags |= TF_NODELAY; 698 else 699 tp->t_flags &= ~TF_NODELAY; 700 break; 701 702 case TCP_MAXSEG: 703 error = sockopt_getint(sopt, &optval); 704 if (error) 705 break; 706 if (optval > 0 && optval <= tp->t_peermss) 707 tp->t_peermss = optval; /* limit on send size */ 708 else 709 error = EINVAL; 710 break; 711 #ifdef notyet 712 case TCP_CONGCTL: 713 /* XXX string overflow XXX */ 714 error = tcp_congctl_select(tp, sopt->sopt_data); 715 break; 716 #endif 717 718 case TCP_KEEPIDLE: 719 error = sockopt_get(sopt, &ui, sizeof(ui)); 720 if (error) 721 break; 722 if (ui > 0) { 723 tp->t_keepidle = ui; 724 change_keepalive(so, tp); 725 } else 726 error = EINVAL; 727 break; 728 729 case TCP_KEEPINTVL: 730 error = sockopt_get(sopt, &ui, sizeof(ui)); 731 if (error) 732 break; 733 if (ui > 0) { 734 tp->t_keepintvl = ui; 735 change_keepalive(so, tp); 736 } else 737 error = EINVAL; 738 break; 739 740 case TCP_KEEPCNT: 741 error = sockopt_get(sopt, &ui, sizeof(ui)); 742 if (error) 743 break; 744 if (ui > 0) { 745 tp->t_keepcnt = ui; 746 change_keepalive(so, tp); 747 } else 748 error = EINVAL; 749 break; 750 751 case TCP_KEEPINIT: 752 error = sockopt_get(sopt, &ui, sizeof(ui)); 753 if (error) 754 break; 755 if (ui > 0) { 756 tp->t_keepinit = ui; 757 change_keepalive(so, tp); 758 } else 759 error = EINVAL; 760 break; 761 762 default: 763 error = ENOPROTOOPT; 764 break; 765 } 766 break; 767 768 case PRCO_GETOPT: 769 switch (optname) { 770 #ifdef TCP_SIGNATURE 771 case TCP_MD5SIG: 772 optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0; 773 error = sockopt_set(sopt, &optval, sizeof(optval)); 774 break; 775 #endif 776 case TCP_NODELAY: 777 optval = tp->t_flags & TF_NODELAY; 778 error = sockopt_set(sopt, &optval, sizeof(optval)); 779 break; 780 case TCP_MAXSEG: 781 optval = tp->t_peermss; 782 error = sockopt_set(sopt, &optval, sizeof(optval)); 783 break; 784 #ifdef notyet 785 case TCP_CONGCTL: 786 break; 787 #endif 788 default: 789 error = ENOPROTOOPT; 790 break; 791 } 792 break; 793 } 794 splx(s); 795 return (error); 796 } 797 798 #ifndef TCP_SENDSPACE 799 #define TCP_SENDSPACE 1024*32 800 #endif 801 int tcp_sendspace = TCP_SENDSPACE; 802 #ifndef TCP_RECVSPACE 803 #define TCP_RECVSPACE 1024*32 804 #endif 805 int tcp_recvspace = TCP_RECVSPACE; 806 807 /* 808 * tcp_attach: attach TCP protocol to socket, allocating internet protocol 809 * control block, TCP control block, buffer space and entering LISTEN state 810 * if to accept connections. 811 */ 812 static int 813 tcp_attach(struct socket *so, int proto) 814 { 815 struct tcpcb *tp; 816 struct inpcb *inp; 817 #ifdef INET6 818 struct in6pcb *in6p; 819 #endif 820 int s, error, family; 821 822 /* Assign the lock (must happen even if we will error out). */ 823 s = splsoftnet(); 824 sosetlock(so); 825 KASSERT(solocked(so)); 826 827 family = so->so_proto->pr_domain->dom_family; 828 switch (family) { 829 #ifdef INET 830 case PF_INET: 831 inp = sotoinpcb(so); 832 #ifdef INET6 833 in6p = NULL; 834 #endif 835 break; 836 #endif 837 #ifdef INET6 838 case PF_INET6: 839 inp = NULL; 840 in6p = sotoin6pcb(so); 841 break; 842 #endif 843 default: 844 error = EAFNOSUPPORT; 845 goto out; 846 } 847 848 KASSERT(inp == NULL); 849 #ifdef INET6 850 KASSERT(in6p == NULL); 851 #endif 852 853 #ifdef MBUFTRACE 854 so->so_mowner = &tcp_sock_mowner; 855 so->so_rcv.sb_mowner = &tcp_sock_rx_mowner; 856 so->so_snd.sb_mowner = &tcp_sock_tx_mowner; 857 #endif 858 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 859 error = soreserve(so, tcp_sendspace, tcp_recvspace); 860 if (error) 861 goto out; 862 } 863 864 so->so_rcv.sb_flags |= SB_AUTOSIZE; 865 so->so_snd.sb_flags |= SB_AUTOSIZE; 866 867 switch (family) { 868 #ifdef INET 869 case PF_INET: 870 error = in_pcballoc(so, &tcbtable); 871 if (error) 872 goto out; 873 inp = sotoinpcb(so); 874 #ifdef INET6 875 in6p = NULL; 876 #endif 877 break; 878 #endif 879 #ifdef INET6 880 case PF_INET6: 881 error = in6_pcballoc(so, &tcbtable); 882 if (error) 883 goto out; 884 inp = NULL; 885 in6p = sotoin6pcb(so); 886 break; 887 #endif 888 default: 889 error = EAFNOSUPPORT; 890 goto out; 891 } 892 if (inp) 893 tp = tcp_newtcpcb(family, (void *)inp); 894 #ifdef INET6 895 else if (in6p) 896 tp = tcp_newtcpcb(family, (void *)in6p); 897 #endif 898 else 899 tp = NULL; 900 901 if (tp == NULL) { 902 int nofd = so->so_state & SS_NOFDREF; /* XXX */ 903 904 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ 905 #ifdef INET 906 if (inp) 907 in_pcbdetach(inp); 908 #endif 909 #ifdef INET6 910 if (in6p) 911 in6_pcbdetach(in6p); 912 #endif 913 so->so_state |= nofd; 914 error = ENOBUFS; 915 goto out; 916 } 917 tp->t_state = TCPS_CLOSED; 918 if ((so->so_options & SO_LINGER) && so->so_linger == 0) { 919 so->so_linger = TCP_LINGERTIME; 920 } 921 out: 922 KASSERT(solocked(so)); 923 splx(s); 924 return error; 925 } 926 927 static void 928 tcp_detach(struct socket *so) 929 { 930 struct inpcb *inp; 931 #ifdef INET6 932 struct in6pcb *in6p; 933 #endif 934 struct tcpcb *tp = NULL; 935 int s, family; 936 937 KASSERT(solocked(so)); 938 939 s = splsoftnet(); 940 family = so->so_proto->pr_domain->dom_family; 941 switch (family) { 942 #ifdef INET 943 case PF_INET: 944 inp = sotoinpcb(so); 945 tp = intotcpcb(inp); 946 break; 947 #endif 948 #ifdef INET6 949 case PF_INET6: 950 in6p = sotoin6pcb(so); 951 tp = in6totcpcb(in6p); 952 break; 953 #endif 954 default: 955 splx(s); 956 return; 957 } 958 KASSERT(tp != NULL); 959 (void)tcp_disconnect(tp); 960 splx(s); 961 } 962 963 /* 964 * Initiate (or continue) disconnect. 965 * If embryonic state, just send reset (once). 966 * If in ``let data drain'' option and linger null, just drop. 967 * Otherwise (hard), mark socket disconnecting and drop 968 * current input data; switch states based on user close, and 969 * send segment to peer (with FIN). 970 */ 971 struct tcpcb * 972 tcp_disconnect(struct tcpcb *tp) 973 { 974 struct socket *so; 975 976 if (tp->t_inpcb) 977 so = tp->t_inpcb->inp_socket; 978 #ifdef INET6 979 else if (tp->t_in6pcb) 980 so = tp->t_in6pcb->in6p_socket; 981 #endif 982 else 983 so = NULL; 984 985 if (TCPS_HAVEESTABLISHED(tp->t_state) == 0) 986 tp = tcp_close(tp); 987 else if ((so->so_options & SO_LINGER) && so->so_linger == 0) 988 tp = tcp_drop(tp, 0); 989 else { 990 soisdisconnecting(so); 991 sbflush(&so->so_rcv); 992 tp = tcp_usrclosed(tp); 993 if (tp) 994 (void) tcp_output(tp); 995 } 996 return (tp); 997 } 998 999 /* 1000 * User issued close, and wish to trail through shutdown states: 1001 * if never received SYN, just forget it. If got a SYN from peer, 1002 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 1003 * If already got a FIN from peer, then almost done; go to LAST_ACK 1004 * state. In all other cases, have already sent FIN to peer (e.g. 1005 * after PRU_SHUTDOWN), and just have to play tedious game waiting 1006 * for peer to send FIN or not respond to keep-alives, etc. 1007 * We can let the user exit from the close as soon as the FIN is acked. 1008 */ 1009 struct tcpcb * 1010 tcp_usrclosed(struct tcpcb *tp) 1011 { 1012 1013 switch (tp->t_state) { 1014 1015 case TCPS_CLOSED: 1016 case TCPS_LISTEN: 1017 case TCPS_SYN_SENT: 1018 tp->t_state = TCPS_CLOSED; 1019 tp = tcp_close(tp); 1020 break; 1021 1022 case TCPS_SYN_RECEIVED: 1023 case TCPS_ESTABLISHED: 1024 tp->t_state = TCPS_FIN_WAIT_1; 1025 break; 1026 1027 case TCPS_CLOSE_WAIT: 1028 tp->t_state = TCPS_LAST_ACK; 1029 break; 1030 } 1031 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { 1032 struct socket *so; 1033 if (tp->t_inpcb) 1034 so = tp->t_inpcb->inp_socket; 1035 #ifdef INET6 1036 else if (tp->t_in6pcb) 1037 so = tp->t_in6pcb->in6p_socket; 1038 #endif 1039 else 1040 so = NULL; 1041 if (so) 1042 soisdisconnected(so); 1043 /* 1044 * If we are in FIN_WAIT_2, we arrived here because the 1045 * application did a shutdown of the send side. Like the 1046 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after 1047 * a full close, we start a timer to make sure sockets are 1048 * not left in FIN_WAIT_2 forever. 1049 */ 1050 if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0)) 1051 TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle); 1052 else if (tp->t_state == TCPS_TIME_WAIT 1053 && ((tp->t_inpcb 1054 && (tcp4_vtw_enable & 1) 1055 && vtw_add(AF_INET, tp)) 1056 || 1057 (tp->t_in6pcb 1058 && (tcp6_vtw_enable & 1) 1059 && vtw_add(AF_INET6, tp)))) { 1060 tp = 0; 1061 } 1062 } 1063 return (tp); 1064 } 1065 1066 /* 1067 * sysctl helper routine for net.inet.ip.mssdflt. it can't be less 1068 * than 32. 1069 */ 1070 static int 1071 sysctl_net_inet_tcp_mssdflt(SYSCTLFN_ARGS) 1072 { 1073 int error, mssdflt; 1074 struct sysctlnode node; 1075 1076 mssdflt = tcp_mssdflt; 1077 node = *rnode; 1078 node.sysctl_data = &mssdflt; 1079 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1080 if (error || newp == NULL) 1081 return (error); 1082 1083 if (mssdflt < 32) 1084 return (EINVAL); 1085 tcp_mssdflt = mssdflt; 1086 1087 mutex_enter(softnet_lock); 1088 tcp_tcpcb_template(); 1089 mutex_exit(softnet_lock); 1090 1091 return (0); 1092 } 1093 1094 /* 1095 * sysctl helper for TCP CB template update 1096 */ 1097 static int 1098 sysctl_update_tcpcb_template(SYSCTLFN_ARGS) 1099 { 1100 int t, error; 1101 struct sysctlnode node; 1102 1103 /* follow procedures in sysctl(9) manpage */ 1104 t = *(int *)rnode->sysctl_data; 1105 node = *rnode; 1106 node.sysctl_data = &t; 1107 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1108 if (error || newp == NULL) 1109 return error; 1110 1111 if (t < 0) 1112 return EINVAL; 1113 1114 *(int *)rnode->sysctl_data = t; 1115 1116 mutex_enter(softnet_lock); 1117 tcp_tcpcb_template(); 1118 mutex_exit(softnet_lock); 1119 1120 return 0; 1121 } 1122 1123 /* 1124 * sysctl helper routine for setting port related values under 1125 * net.inet.ip and net.inet6.ip6. does basic range checking and does 1126 * additional checks for each type. this code has placed in 1127 * tcp_input.c since INET and INET6 both use the same tcp code. 1128 * 1129 * this helper is not static so that both inet and inet6 can use it. 1130 */ 1131 int 1132 sysctl_net_inet_ip_ports(SYSCTLFN_ARGS) 1133 { 1134 int error, tmp; 1135 int apmin, apmax; 1136 #ifndef IPNOPRIVPORTS 1137 int lpmin, lpmax; 1138 #endif /* IPNOPRIVPORTS */ 1139 struct sysctlnode node; 1140 1141 if (namelen != 0) 1142 return (EINVAL); 1143 1144 switch (name[-3]) { 1145 #ifdef INET 1146 case PF_INET: 1147 apmin = anonportmin; 1148 apmax = anonportmax; 1149 #ifndef IPNOPRIVPORTS 1150 lpmin = lowportmin; 1151 lpmax = lowportmax; 1152 #endif /* IPNOPRIVPORTS */ 1153 break; 1154 #endif /* INET */ 1155 #ifdef INET6 1156 case PF_INET6: 1157 apmin = ip6_anonportmin; 1158 apmax = ip6_anonportmax; 1159 #ifndef IPNOPRIVPORTS 1160 lpmin = ip6_lowportmin; 1161 lpmax = ip6_lowportmax; 1162 #endif /* IPNOPRIVPORTS */ 1163 break; 1164 #endif /* INET6 */ 1165 default: 1166 return (EINVAL); 1167 } 1168 1169 /* 1170 * insert temporary copy into node, perform lookup on 1171 * temporary, then restore pointer 1172 */ 1173 node = *rnode; 1174 tmp = *(int*)rnode->sysctl_data; 1175 node.sysctl_data = &tmp; 1176 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1177 if (error || newp == NULL) 1178 return (error); 1179 1180 /* 1181 * simple port range check 1182 */ 1183 if (tmp < 0 || tmp > 65535) 1184 return (EINVAL); 1185 1186 /* 1187 * per-node range checks 1188 */ 1189 switch (rnode->sysctl_num) { 1190 case IPCTL_ANONPORTMIN: 1191 case IPV6CTL_ANONPORTMIN: 1192 if (tmp >= apmax) 1193 return (EINVAL); 1194 #ifndef IPNOPRIVPORTS 1195 if (tmp < IPPORT_RESERVED) 1196 return (EINVAL); 1197 #endif /* IPNOPRIVPORTS */ 1198 break; 1199 1200 case IPCTL_ANONPORTMAX: 1201 case IPV6CTL_ANONPORTMAX: 1202 if (apmin >= tmp) 1203 return (EINVAL); 1204 #ifndef IPNOPRIVPORTS 1205 if (tmp < IPPORT_RESERVED) 1206 return (EINVAL); 1207 #endif /* IPNOPRIVPORTS */ 1208 break; 1209 1210 #ifndef IPNOPRIVPORTS 1211 case IPCTL_LOWPORTMIN: 1212 case IPV6CTL_LOWPORTMIN: 1213 if (tmp >= lpmax || 1214 tmp > IPPORT_RESERVEDMAX || 1215 tmp < IPPORT_RESERVEDMIN) 1216 return (EINVAL); 1217 break; 1218 1219 case IPCTL_LOWPORTMAX: 1220 case IPV6CTL_LOWPORTMAX: 1221 if (lpmin >= tmp || 1222 tmp > IPPORT_RESERVEDMAX || 1223 tmp < IPPORT_RESERVEDMIN) 1224 return (EINVAL); 1225 break; 1226 #endif /* IPNOPRIVPORTS */ 1227 1228 default: 1229 return (EINVAL); 1230 } 1231 1232 *(int*)rnode->sysctl_data = tmp; 1233 1234 return (0); 1235 } 1236 1237 static inline int 1238 copyout_uid(struct socket *sockp, void *oldp, size_t *oldlenp) 1239 { 1240 if (oldp) { 1241 size_t sz; 1242 uid_t uid; 1243 int error; 1244 1245 if (sockp->so_cred == NULL) 1246 return EPERM; 1247 1248 uid = kauth_cred_geteuid(sockp->so_cred); 1249 sz = MIN(sizeof(uid), *oldlenp); 1250 if ((error = copyout(&uid, oldp, sz)) != 0) 1251 return error; 1252 } 1253 *oldlenp = sizeof(uid_t); 1254 return 0; 1255 } 1256 1257 static inline int 1258 inet4_ident_core(struct in_addr raddr, u_int rport, 1259 struct in_addr laddr, u_int lport, 1260 void *oldp, size_t *oldlenp, 1261 struct lwp *l, int dodrop) 1262 { 1263 struct inpcb *inp; 1264 struct socket *sockp; 1265 1266 inp = in_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport, 0); 1267 1268 if (inp == NULL || (sockp = inp->inp_socket) == NULL) 1269 return ESRCH; 1270 1271 if (dodrop) { 1272 struct tcpcb *tp; 1273 int error; 1274 1275 if (inp == NULL || (tp = intotcpcb(inp)) == NULL || 1276 (inp->inp_socket->so_options & SO_ACCEPTCONN) != 0) 1277 return ESRCH; 1278 1279 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, 1280 KAUTH_REQ_NETWORK_SOCKET_DROP, inp->inp_socket, tp, NULL); 1281 if (error) 1282 return (error); 1283 1284 (void)tcp_drop(tp, ECONNABORTED); 1285 return 0; 1286 } 1287 else 1288 return copyout_uid(sockp, oldp, oldlenp); 1289 } 1290 1291 #ifdef INET6 1292 static inline int 1293 inet6_ident_core(struct in6_addr *raddr, u_int rport, 1294 struct in6_addr *laddr, u_int lport, 1295 void *oldp, size_t *oldlenp, 1296 struct lwp *l, int dodrop) 1297 { 1298 struct in6pcb *in6p; 1299 struct socket *sockp; 1300 1301 in6p = in6_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport, 0, 0); 1302 1303 if (in6p == NULL || (sockp = in6p->in6p_socket) == NULL) 1304 return ESRCH; 1305 1306 if (dodrop) { 1307 struct tcpcb *tp; 1308 int error; 1309 1310 if (in6p == NULL || (tp = in6totcpcb(in6p)) == NULL || 1311 (in6p->in6p_socket->so_options & SO_ACCEPTCONN) != 0) 1312 return ESRCH; 1313 1314 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, 1315 KAUTH_REQ_NETWORK_SOCKET_DROP, in6p->in6p_socket, tp, NULL); 1316 if (error) 1317 return (error); 1318 1319 (void)tcp_drop(tp, ECONNABORTED); 1320 return 0; 1321 } 1322 else 1323 return copyout_uid(sockp, oldp, oldlenp); 1324 } 1325 #endif 1326 1327 /* 1328 * sysctl helper routine for the net.inet.tcp.drop and 1329 * net.inet6.tcp6.drop nodes. 1330 */ 1331 #define sysctl_net_inet_tcp_drop sysctl_net_inet_tcp_ident 1332 1333 /* 1334 * sysctl helper routine for the net.inet.tcp.ident and 1335 * net.inet6.tcp6.ident nodes. contains backwards compat code for the 1336 * old way of looking up the ident information for ipv4 which involves 1337 * stuffing the port/addr pairs into the mib lookup. 1338 */ 1339 static int 1340 sysctl_net_inet_tcp_ident(SYSCTLFN_ARGS) 1341 { 1342 #ifdef INET 1343 struct sockaddr_in *si4[2]; 1344 #endif /* INET */ 1345 #ifdef INET6 1346 struct sockaddr_in6 *si6[2]; 1347 #endif /* INET6 */ 1348 struct sockaddr_storage sa[2]; 1349 int error, pf, dodrop; 1350 1351 dodrop = name[-1] == TCPCTL_DROP; 1352 if (dodrop) { 1353 if (oldp != NULL || *oldlenp != 0) 1354 return EINVAL; 1355 if (newp == NULL) 1356 return EPERM; 1357 if (newlen < sizeof(sa)) 1358 return ENOMEM; 1359 } 1360 if (namelen != 4 && namelen != 0) 1361 return EINVAL; 1362 if (name[-2] != IPPROTO_TCP) 1363 return EINVAL; 1364 pf = name[-3]; 1365 1366 /* old style lookup, ipv4 only */ 1367 if (namelen == 4) { 1368 #ifdef INET 1369 struct in_addr laddr, raddr; 1370 u_int lport, rport; 1371 1372 if (pf != PF_INET) 1373 return EPROTONOSUPPORT; 1374 raddr.s_addr = (uint32_t)name[0]; 1375 rport = (u_int)name[1]; 1376 laddr.s_addr = (uint32_t)name[2]; 1377 lport = (u_int)name[3]; 1378 1379 mutex_enter(softnet_lock); 1380 error = inet4_ident_core(raddr, rport, laddr, lport, 1381 oldp, oldlenp, l, dodrop); 1382 mutex_exit(softnet_lock); 1383 return error; 1384 #else /* INET */ 1385 return EINVAL; 1386 #endif /* INET */ 1387 } 1388 1389 if (newp == NULL || newlen != sizeof(sa)) 1390 return EINVAL; 1391 error = copyin(newp, &sa, newlen); 1392 if (error) 1393 return error; 1394 1395 /* 1396 * requested families must match 1397 */ 1398 if (pf != sa[0].ss_family || sa[0].ss_family != sa[1].ss_family) 1399 return EINVAL; 1400 1401 switch (pf) { 1402 #ifdef INET6 1403 case PF_INET6: 1404 si6[0] = (struct sockaddr_in6*)&sa[0]; 1405 si6[1] = (struct sockaddr_in6*)&sa[1]; 1406 if (si6[0]->sin6_len != sizeof(*si6[0]) || 1407 si6[1]->sin6_len != sizeof(*si6[1])) 1408 return EINVAL; 1409 1410 if (!IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) && 1411 !IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr)) { 1412 error = sa6_embedscope(si6[0], ip6_use_defzone); 1413 if (error) 1414 return error; 1415 error = sa6_embedscope(si6[1], ip6_use_defzone); 1416 if (error) 1417 return error; 1418 1419 mutex_enter(softnet_lock); 1420 error = inet6_ident_core(&si6[0]->sin6_addr, 1421 si6[0]->sin6_port, &si6[1]->sin6_addr, 1422 si6[1]->sin6_port, oldp, oldlenp, l, dodrop); 1423 mutex_exit(softnet_lock); 1424 return error; 1425 } 1426 1427 if (IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) != 1428 IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr)) 1429 return EINVAL; 1430 1431 in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[0]); 1432 in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[1]); 1433 /*FALLTHROUGH*/ 1434 #endif /* INET6 */ 1435 #ifdef INET 1436 case PF_INET: 1437 si4[0] = (struct sockaddr_in*)&sa[0]; 1438 si4[1] = (struct sockaddr_in*)&sa[1]; 1439 if (si4[0]->sin_len != sizeof(*si4[0]) || 1440 si4[0]->sin_len != sizeof(*si4[1])) 1441 return EINVAL; 1442 1443 mutex_enter(softnet_lock); 1444 error = inet4_ident_core(si4[0]->sin_addr, si4[0]->sin_port, 1445 si4[1]->sin_addr, si4[1]->sin_port, 1446 oldp, oldlenp, l, dodrop); 1447 mutex_exit(softnet_lock); 1448 return error; 1449 #endif /* INET */ 1450 default: 1451 return EPROTONOSUPPORT; 1452 } 1453 } 1454 1455 /* 1456 * sysctl helper for the inet and inet6 pcblists. handles tcp/udp and 1457 * inet/inet6, as well as raw pcbs for each. specifically not 1458 * declared static so that raw sockets and udp/udp6 can use it as 1459 * well. 1460 */ 1461 int 1462 sysctl_inpcblist(SYSCTLFN_ARGS) 1463 { 1464 #ifdef INET 1465 struct sockaddr_in *in; 1466 const struct inpcb *inp; 1467 #endif 1468 #ifdef INET6 1469 struct sockaddr_in6 *in6; 1470 const struct in6pcb *in6p; 1471 #endif 1472 struct inpcbtable *pcbtbl = __UNCONST(rnode->sysctl_data); 1473 const struct inpcb_hdr *inph; 1474 struct tcpcb *tp; 1475 struct kinfo_pcb pcb; 1476 char *dp; 1477 size_t len, needed, elem_size, out_size; 1478 int error, elem_count, pf, proto, pf2; 1479 1480 if (namelen != 4) 1481 return (EINVAL); 1482 1483 if (oldp != NULL) { 1484 len = *oldlenp; 1485 elem_size = name[2]; 1486 elem_count = name[3]; 1487 if (elem_size != sizeof(pcb)) 1488 return EINVAL; 1489 } else { 1490 len = 0; 1491 elem_count = INT_MAX; 1492 elem_size = sizeof(pcb); 1493 } 1494 error = 0; 1495 dp = oldp; 1496 out_size = elem_size; 1497 needed = 0; 1498 1499 if (namelen == 1 && name[0] == CTL_QUERY) 1500 return (sysctl_query(SYSCTLFN_CALL(rnode))); 1501 1502 if (name - oname != 4) 1503 return (EINVAL); 1504 1505 pf = oname[1]; 1506 proto = oname[2]; 1507 pf2 = (oldp != NULL) ? pf : 0; 1508 1509 mutex_enter(softnet_lock); 1510 1511 TAILQ_FOREACH(inph, &pcbtbl->inpt_queue, inph_queue) { 1512 #ifdef INET 1513 inp = (const struct inpcb *)inph; 1514 #endif 1515 #ifdef INET6 1516 in6p = (const struct in6pcb *)inph; 1517 #endif 1518 1519 if (inph->inph_af != pf) 1520 continue; 1521 1522 if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, 1523 KAUTH_REQ_NETWORK_SOCKET_CANSEE, inph->inph_socket, NULL, 1524 NULL) != 0) 1525 continue; 1526 1527 memset(&pcb, 0, sizeof(pcb)); 1528 1529 pcb.ki_family = pf; 1530 pcb.ki_type = proto; 1531 1532 switch (pf2) { 1533 case 0: 1534 /* just probing for size */ 1535 break; 1536 #ifdef INET 1537 case PF_INET: 1538 pcb.ki_family = inp->inp_socket->so_proto-> 1539 pr_domain->dom_family; 1540 pcb.ki_type = inp->inp_socket->so_proto-> 1541 pr_type; 1542 pcb.ki_protocol = inp->inp_socket->so_proto-> 1543 pr_protocol; 1544 pcb.ki_pflags = inp->inp_flags; 1545 1546 pcb.ki_sostate = inp->inp_socket->so_state; 1547 pcb.ki_prstate = inp->inp_state; 1548 if (proto == IPPROTO_TCP) { 1549 tp = intotcpcb(inp); 1550 pcb.ki_tstate = tp->t_state; 1551 pcb.ki_tflags = tp->t_flags; 1552 } 1553 1554 pcb.ki_pcbaddr = PTRTOUINT64(inp); 1555 pcb.ki_ppcbaddr = PTRTOUINT64(inp->inp_ppcb); 1556 pcb.ki_sockaddr = PTRTOUINT64(inp->inp_socket); 1557 1558 pcb.ki_rcvq = inp->inp_socket->so_rcv.sb_cc; 1559 pcb.ki_sndq = inp->inp_socket->so_snd.sb_cc; 1560 1561 in = satosin(&pcb.ki_src); 1562 in->sin_len = sizeof(*in); 1563 in->sin_family = pf; 1564 in->sin_port = inp->inp_lport; 1565 in->sin_addr = inp->inp_laddr; 1566 if (pcb.ki_prstate >= INP_CONNECTED) { 1567 in = satosin(&pcb.ki_dst); 1568 in->sin_len = sizeof(*in); 1569 in->sin_family = pf; 1570 in->sin_port = inp->inp_fport; 1571 in->sin_addr = inp->inp_faddr; 1572 } 1573 break; 1574 #endif 1575 #ifdef INET6 1576 case PF_INET6: 1577 pcb.ki_family = in6p->in6p_socket->so_proto-> 1578 pr_domain->dom_family; 1579 pcb.ki_type = in6p->in6p_socket->so_proto->pr_type; 1580 pcb.ki_protocol = in6p->in6p_socket->so_proto-> 1581 pr_protocol; 1582 pcb.ki_pflags = in6p->in6p_flags; 1583 1584 pcb.ki_sostate = in6p->in6p_socket->so_state; 1585 pcb.ki_prstate = in6p->in6p_state; 1586 if (proto == IPPROTO_TCP) { 1587 tp = in6totcpcb(in6p); 1588 pcb.ki_tstate = tp->t_state; 1589 pcb.ki_tflags = tp->t_flags; 1590 } 1591 1592 pcb.ki_pcbaddr = PTRTOUINT64(in6p); 1593 pcb.ki_ppcbaddr = PTRTOUINT64(in6p->in6p_ppcb); 1594 pcb.ki_sockaddr = PTRTOUINT64(in6p->in6p_socket); 1595 1596 pcb.ki_rcvq = in6p->in6p_socket->so_rcv.sb_cc; 1597 pcb.ki_sndq = in6p->in6p_socket->so_snd.sb_cc; 1598 1599 in6 = satosin6(&pcb.ki_src); 1600 in6->sin6_len = sizeof(*in6); 1601 in6->sin6_family = pf; 1602 in6->sin6_port = in6p->in6p_lport; 1603 in6->sin6_flowinfo = in6p->in6p_flowinfo; 1604 in6->sin6_addr = in6p->in6p_laddr; 1605 in6->sin6_scope_id = 0; /* XXX? */ 1606 1607 if (pcb.ki_prstate >= IN6P_CONNECTED) { 1608 in6 = satosin6(&pcb.ki_dst); 1609 in6->sin6_len = sizeof(*in6); 1610 in6->sin6_family = pf; 1611 in6->sin6_port = in6p->in6p_fport; 1612 in6->sin6_flowinfo = in6p->in6p_flowinfo; 1613 in6->sin6_addr = in6p->in6p_faddr; 1614 in6->sin6_scope_id = 0; /* XXX? */ 1615 } 1616 break; 1617 #endif 1618 } 1619 1620 if (len >= elem_size && elem_count > 0) { 1621 error = copyout(&pcb, dp, out_size); 1622 if (error) { 1623 mutex_exit(softnet_lock); 1624 return (error); 1625 } 1626 dp += elem_size; 1627 len -= elem_size; 1628 } 1629 needed += elem_size; 1630 if (elem_count > 0 && elem_count != INT_MAX) 1631 elem_count--; 1632 } 1633 1634 *oldlenp = needed; 1635 if (oldp == NULL) 1636 *oldlenp += PCB_SLOP * sizeof(struct kinfo_pcb); 1637 1638 mutex_exit(softnet_lock); 1639 1640 return (error); 1641 } 1642 1643 static int 1644 sysctl_tcp_congctl(SYSCTLFN_ARGS) 1645 { 1646 struct sysctlnode node; 1647 int error; 1648 char newname[TCPCC_MAXLEN]; 1649 1650 strlcpy(newname, tcp_congctl_global_name, sizeof(newname) - 1); 1651 1652 node = *rnode; 1653 node.sysctl_data = newname; 1654 node.sysctl_size = sizeof(newname); 1655 1656 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1657 1658 if (error || 1659 newp == NULL || 1660 strncmp(newname, tcp_congctl_global_name, sizeof(newname)) == 0) 1661 return error; 1662 1663 mutex_enter(softnet_lock); 1664 error = tcp_congctl_select(NULL, newname); 1665 mutex_exit(softnet_lock); 1666 1667 return error; 1668 } 1669 1670 static int 1671 sysctl_tcp_init_win(SYSCTLFN_ARGS) 1672 { 1673 int error; 1674 u_int iw; 1675 struct sysctlnode node; 1676 1677 iw = *(u_int *)rnode->sysctl_data; 1678 node = *rnode; 1679 node.sysctl_data = &iw; 1680 node.sysctl_size = sizeof(iw); 1681 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1682 if (error || newp == NULL) 1683 return error; 1684 1685 if (iw >= __arraycount(tcp_init_win_max)) 1686 return EINVAL; 1687 *(u_int *)rnode->sysctl_data = iw; 1688 return 0; 1689 } 1690 1691 static int 1692 sysctl_tcp_keep(SYSCTLFN_ARGS) 1693 { 1694 int error; 1695 u_int tmp; 1696 struct sysctlnode node; 1697 1698 node = *rnode; 1699 tmp = *(u_int *)rnode->sysctl_data; 1700 node.sysctl_data = &tmp; 1701 1702 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1703 if (error || newp == NULL) 1704 return error; 1705 1706 mutex_enter(softnet_lock); 1707 1708 *(u_int *)rnode->sysctl_data = tmp; 1709 tcp_tcpcb_template(); /* update the template */ 1710 1711 mutex_exit(softnet_lock); 1712 return 0; 1713 } 1714 1715 static int 1716 sysctl_net_inet_tcp_stats(SYSCTLFN_ARGS) 1717 { 1718 1719 return (NETSTAT_SYSCTL(tcpstat_percpu, TCP_NSTATS)); 1720 } 1721 1722 /* 1723 * this (second stage) setup routine is a replacement for tcp_sysctl() 1724 * (which is currently used for ipv4 and ipv6) 1725 */ 1726 static void 1727 sysctl_net_inet_tcp_setup2(struct sysctllog **clog, int pf, const char *pfname, 1728 const char *tcpname) 1729 { 1730 const struct sysctlnode *sack_node; 1731 const struct sysctlnode *abc_node; 1732 const struct sysctlnode *ecn_node; 1733 const struct sysctlnode *congctl_node; 1734 const struct sysctlnode *mslt_node; 1735 const struct sysctlnode *vtw_node; 1736 #ifdef TCP_DEBUG 1737 extern struct tcp_debug tcp_debug[TCP_NDEBUG]; 1738 extern int tcp_debx; 1739 #endif 1740 1741 sysctl_createv(clog, 0, NULL, NULL, 1742 CTLFLAG_PERMANENT, 1743 CTLTYPE_NODE, pfname, NULL, 1744 NULL, 0, NULL, 0, 1745 CTL_NET, pf, CTL_EOL); 1746 sysctl_createv(clog, 0, NULL, NULL, 1747 CTLFLAG_PERMANENT, 1748 CTLTYPE_NODE, tcpname, 1749 SYSCTL_DESCR("TCP related settings"), 1750 NULL, 0, NULL, 0, 1751 CTL_NET, pf, IPPROTO_TCP, CTL_EOL); 1752 1753 sysctl_createv(clog, 0, NULL, NULL, 1754 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1755 CTLTYPE_INT, "rfc1323", 1756 SYSCTL_DESCR("Enable RFC1323 TCP extensions"), 1757 sysctl_update_tcpcb_template, 0, &tcp_do_rfc1323, 0, 1758 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RFC1323, CTL_EOL); 1759 sysctl_createv(clog, 0, NULL, NULL, 1760 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1761 CTLTYPE_INT, "sendspace", 1762 SYSCTL_DESCR("Default TCP send buffer size"), 1763 NULL, 0, &tcp_sendspace, 0, 1764 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SENDSPACE, CTL_EOL); 1765 sysctl_createv(clog, 0, NULL, NULL, 1766 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1767 CTLTYPE_INT, "recvspace", 1768 SYSCTL_DESCR("Default TCP receive buffer size"), 1769 NULL, 0, &tcp_recvspace, 0, 1770 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RECVSPACE, CTL_EOL); 1771 sysctl_createv(clog, 0, NULL, NULL, 1772 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1773 CTLTYPE_INT, "mssdflt", 1774 SYSCTL_DESCR("Default maximum segment size"), 1775 sysctl_net_inet_tcp_mssdflt, 0, &tcp_mssdflt, 0, 1776 CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSSDFLT, CTL_EOL); 1777 sysctl_createv(clog, 0, NULL, NULL, 1778 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1779 CTLTYPE_INT, "minmss", 1780 SYSCTL_DESCR("Lower limit for TCP maximum segment size"), 1781 NULL, 0, &tcp_minmss, 0, 1782 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1783 sysctl_createv(clog, 0, NULL, NULL, 1784 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1785 CTLTYPE_INT, "msl", 1786 SYSCTL_DESCR("Maximum Segment Life"), 1787 NULL, 0, &tcp_msl, 0, 1788 CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSL, CTL_EOL); 1789 sysctl_createv(clog, 0, NULL, NULL, 1790 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1791 CTLTYPE_INT, "syn_cache_limit", 1792 SYSCTL_DESCR("Maximum number of entries in the TCP " 1793 "compressed state engine"), 1794 NULL, 0, &tcp_syn_cache_limit, 0, 1795 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_LIMIT, 1796 CTL_EOL); 1797 sysctl_createv(clog, 0, NULL, NULL, 1798 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1799 CTLTYPE_INT, "syn_bucket_limit", 1800 SYSCTL_DESCR("Maximum number of entries per hash " 1801 "bucket in the TCP compressed state " 1802 "engine"), 1803 NULL, 0, &tcp_syn_bucket_limit, 0, 1804 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_BUCKET_LIMIT, 1805 CTL_EOL); 1806 #if 0 /* obsoleted */ 1807 sysctl_createv(clog, 0, NULL, NULL, 1808 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1809 CTLTYPE_INT, "syn_cache_interval", 1810 SYSCTL_DESCR("TCP compressed state engine's timer interval"), 1811 NULL, 0, &tcp_syn_cache_interval, 0, 1812 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_INTER, 1813 CTL_EOL); 1814 #endif 1815 sysctl_createv(clog, 0, NULL, NULL, 1816 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1817 CTLTYPE_INT, "init_win", 1818 SYSCTL_DESCR("Initial TCP congestion window"), 1819 sysctl_tcp_init_win, 0, &tcp_init_win, 0, 1820 CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN, CTL_EOL); 1821 sysctl_createv(clog, 0, NULL, NULL, 1822 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1823 CTLTYPE_INT, "mss_ifmtu", 1824 SYSCTL_DESCR("Use interface MTU for calculating MSS"), 1825 NULL, 0, &tcp_mss_ifmtu, 0, 1826 CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSS_IFMTU, CTL_EOL); 1827 sysctl_createv(clog, 0, NULL, &sack_node, 1828 CTLFLAG_PERMANENT, 1829 CTLTYPE_NODE, "sack", 1830 SYSCTL_DESCR("RFC2018 Selective ACKnowledgement tunables"), 1831 NULL, 0, NULL, 0, 1832 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_EOL); 1833 1834 /* Congctl subtree */ 1835 sysctl_createv(clog, 0, NULL, &congctl_node, 1836 CTLFLAG_PERMANENT, 1837 CTLTYPE_NODE, "congctl", 1838 SYSCTL_DESCR("TCP Congestion Control"), 1839 NULL, 0, NULL, 0, 1840 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1841 sysctl_createv(clog, 0, &congctl_node, NULL, 1842 CTLFLAG_PERMANENT, 1843 CTLTYPE_STRING, "available", 1844 SYSCTL_DESCR("Available Congestion Control Mechanisms"), 1845 NULL, 0, tcp_congctl_avail, 0, CTL_CREATE, CTL_EOL); 1846 sysctl_createv(clog, 0, &congctl_node, NULL, 1847 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1848 CTLTYPE_STRING, "selected", 1849 SYSCTL_DESCR("Selected Congestion Control Mechanism"), 1850 sysctl_tcp_congctl, 0, NULL, TCPCC_MAXLEN, 1851 CTL_CREATE, CTL_EOL); 1852 1853 sysctl_createv(clog, 0, NULL, NULL, 1854 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1855 CTLTYPE_INT, "win_scale", 1856 SYSCTL_DESCR("Use RFC1323 window scale options"), 1857 sysctl_update_tcpcb_template, 0, &tcp_do_win_scale, 0, 1858 CTL_NET, pf, IPPROTO_TCP, TCPCTL_WSCALE, CTL_EOL); 1859 sysctl_createv(clog, 0, NULL, NULL, 1860 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1861 CTLTYPE_INT, "timestamps", 1862 SYSCTL_DESCR("Use RFC1323 time stamp options"), 1863 sysctl_update_tcpcb_template, 0, &tcp_do_timestamps, 0, 1864 CTL_NET, pf, IPPROTO_TCP, TCPCTL_TSTAMP, CTL_EOL); 1865 sysctl_createv(clog, 0, NULL, NULL, 1866 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1867 CTLTYPE_INT, "compat_42", 1868 SYSCTL_DESCR("Enable workarounds for 4.2BSD TCP bugs"), 1869 NULL, 0, &tcp_compat_42, 0, 1870 CTL_NET, pf, IPPROTO_TCP, TCPCTL_COMPAT_42, CTL_EOL); 1871 sysctl_createv(clog, 0, NULL, NULL, 1872 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1873 CTLTYPE_INT, "cwm", 1874 SYSCTL_DESCR("Hughes/Touch/Heidemann Congestion Window " 1875 "Monitoring"), 1876 NULL, 0, &tcp_cwm, 0, 1877 CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM, CTL_EOL); 1878 sysctl_createv(clog, 0, NULL, NULL, 1879 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1880 CTLTYPE_INT, "cwm_burstsize", 1881 SYSCTL_DESCR("Congestion Window Monitoring allowed " 1882 "burst count in packets"), 1883 NULL, 0, &tcp_cwm_burstsize, 0, 1884 CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM_BURSTSIZE, 1885 CTL_EOL); 1886 sysctl_createv(clog, 0, NULL, NULL, 1887 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1888 CTLTYPE_INT, "ack_on_push", 1889 SYSCTL_DESCR("Immediately return ACK when PSH is " 1890 "received"), 1891 NULL, 0, &tcp_ack_on_push, 0, 1892 CTL_NET, pf, IPPROTO_TCP, TCPCTL_ACK_ON_PUSH, CTL_EOL); 1893 sysctl_createv(clog, 0, NULL, NULL, 1894 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1895 CTLTYPE_INT, "keepidle", 1896 SYSCTL_DESCR("Allowed connection idle ticks before a " 1897 "keepalive probe is sent"), 1898 sysctl_tcp_keep, 0, &tcp_keepidle, 0, 1899 CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPIDLE, CTL_EOL); 1900 sysctl_createv(clog, 0, NULL, NULL, 1901 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1902 CTLTYPE_INT, "keepintvl", 1903 SYSCTL_DESCR("Ticks before next keepalive probe is sent"), 1904 sysctl_tcp_keep, 0, &tcp_keepintvl, 0, 1905 CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPINTVL, CTL_EOL); 1906 sysctl_createv(clog, 0, NULL, NULL, 1907 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1908 CTLTYPE_INT, "keepcnt", 1909 SYSCTL_DESCR("Number of keepalive probes to send"), 1910 sysctl_tcp_keep, 0, &tcp_keepcnt, 0, 1911 CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPCNT, CTL_EOL); 1912 sysctl_createv(clog, 0, NULL, NULL, 1913 CTLFLAG_PERMANENT|CTLFLAG_IMMEDIATE, 1914 CTLTYPE_INT, "slowhz", 1915 SYSCTL_DESCR("Keepalive ticks per second"), 1916 NULL, PR_SLOWHZ, NULL, 0, 1917 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SLOWHZ, CTL_EOL); 1918 sysctl_createv(clog, 0, NULL, NULL, 1919 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1920 CTLTYPE_INT, "log_refused", 1921 SYSCTL_DESCR("Log refused TCP connections"), 1922 NULL, 0, &tcp_log_refused, 0, 1923 CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOG_REFUSED, CTL_EOL); 1924 #if 0 /* obsoleted */ 1925 sysctl_createv(clog, 0, NULL, NULL, 1926 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1927 CTLTYPE_INT, "rstratelimit", NULL, 1928 NULL, 0, &tcp_rst_ratelim, 0, 1929 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTRATELIMIT, CTL_EOL); 1930 #endif 1931 sysctl_createv(clog, 0, NULL, NULL, 1932 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1933 CTLTYPE_INT, "rstppslimit", 1934 SYSCTL_DESCR("Maximum number of RST packets to send " 1935 "per second"), 1936 NULL, 0, &tcp_rst_ppslim, 0, 1937 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTPPSLIMIT, CTL_EOL); 1938 sysctl_createv(clog, 0, NULL, NULL, 1939 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1940 CTLTYPE_INT, "delack_ticks", 1941 SYSCTL_DESCR("Number of ticks to delay sending an ACK"), 1942 NULL, 0, &tcp_delack_ticks, 0, 1943 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DELACK_TICKS, CTL_EOL); 1944 sysctl_createv(clog, 0, NULL, NULL, 1945 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1946 CTLTYPE_INT, "init_win_local", 1947 SYSCTL_DESCR("Initial TCP window size (in segments)"), 1948 sysctl_tcp_init_win, 0, &tcp_init_win_local, 0, 1949 CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN_LOCAL, 1950 CTL_EOL); 1951 sysctl_createv(clog, 0, NULL, NULL, 1952 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1953 CTLTYPE_STRUCT, "ident", 1954 SYSCTL_DESCR("RFC1413 Identification Protocol lookups"), 1955 sysctl_net_inet_tcp_ident, 0, NULL, sizeof(uid_t), 1956 CTL_NET, pf, IPPROTO_TCP, TCPCTL_IDENT, CTL_EOL); 1957 sysctl_createv(clog, 0, NULL, NULL, 1958 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1959 CTLTYPE_INT, "do_loopback_cksum", 1960 SYSCTL_DESCR("Perform TCP checksum on loopback"), 1961 NULL, 0, &tcp_do_loopback_cksum, 0, 1962 CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOOPBACKCKSUM, 1963 CTL_EOL); 1964 sysctl_createv(clog, 0, NULL, NULL, 1965 CTLFLAG_PERMANENT, 1966 CTLTYPE_STRUCT, "pcblist", 1967 SYSCTL_DESCR("TCP protocol control block list"), 1968 sysctl_inpcblist, 0, &tcbtable, 0, 1969 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, 1970 CTL_EOL); 1971 sysctl_createv(clog, 0, NULL, NULL, 1972 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1973 CTLTYPE_INT, "keepinit", 1974 SYSCTL_DESCR("Ticks before initial tcp connection times out"), 1975 sysctl_tcp_keep, 0, &tcp_keepinit, 0, 1976 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1977 1978 /* TCP socket buffers auto-sizing nodes */ 1979 sysctl_createv(clog, 0, NULL, NULL, 1980 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1981 CTLTYPE_INT, "recvbuf_auto", 1982 SYSCTL_DESCR("Enable automatic receive " 1983 "buffer sizing (experimental)"), 1984 NULL, 0, &tcp_do_autorcvbuf, 0, 1985 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1986 sysctl_createv(clog, 0, NULL, NULL, 1987 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1988 CTLTYPE_INT, "recvbuf_inc", 1989 SYSCTL_DESCR("Incrementor step size of " 1990 "automatic receive buffer"), 1991 NULL, 0, &tcp_autorcvbuf_inc, 0, 1992 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1993 sysctl_createv(clog, 0, NULL, NULL, 1994 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1995 CTLTYPE_INT, "recvbuf_max", 1996 SYSCTL_DESCR("Max size of automatic receive buffer"), 1997 NULL, 0, &tcp_autorcvbuf_max, 0, 1998 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1999 2000 sysctl_createv(clog, 0, NULL, NULL, 2001 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2002 CTLTYPE_INT, "sendbuf_auto", 2003 SYSCTL_DESCR("Enable automatic send " 2004 "buffer sizing (experimental)"), 2005 NULL, 0, &tcp_do_autosndbuf, 0, 2006 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2007 sysctl_createv(clog, 0, NULL, NULL, 2008 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2009 CTLTYPE_INT, "sendbuf_inc", 2010 SYSCTL_DESCR("Incrementor step size of " 2011 "automatic send buffer"), 2012 NULL, 0, &tcp_autosndbuf_inc, 0, 2013 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2014 sysctl_createv(clog, 0, NULL, NULL, 2015 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2016 CTLTYPE_INT, "sendbuf_max", 2017 SYSCTL_DESCR("Max size of automatic send buffer"), 2018 NULL, 0, &tcp_autosndbuf_max, 0, 2019 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2020 2021 /* ECN subtree */ 2022 sysctl_createv(clog, 0, NULL, &ecn_node, 2023 CTLFLAG_PERMANENT, 2024 CTLTYPE_NODE, "ecn", 2025 SYSCTL_DESCR("RFC3168 Explicit Congestion Notification"), 2026 NULL, 0, NULL, 0, 2027 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2028 sysctl_createv(clog, 0, &ecn_node, NULL, 2029 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2030 CTLTYPE_INT, "enable", 2031 SYSCTL_DESCR("Enable TCP Explicit Congestion " 2032 "Notification"), 2033 NULL, 0, &tcp_do_ecn, 0, CTL_CREATE, CTL_EOL); 2034 sysctl_createv(clog, 0, &ecn_node, NULL, 2035 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2036 CTLTYPE_INT, "maxretries", 2037 SYSCTL_DESCR("Number of times to retry ECN setup " 2038 "before disabling ECN on the connection"), 2039 NULL, 0, &tcp_ecn_maxretries, 0, CTL_CREATE, CTL_EOL); 2040 2041 /* SACK gets it's own little subtree. */ 2042 sysctl_createv(clog, 0, NULL, &sack_node, 2043 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2044 CTLTYPE_INT, "enable", 2045 SYSCTL_DESCR("Enable RFC2018 Selective ACKnowledgement"), 2046 NULL, 0, &tcp_do_sack, 0, 2047 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 2048 sysctl_createv(clog, 0, NULL, &sack_node, 2049 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2050 CTLTYPE_INT, "maxholes", 2051 SYSCTL_DESCR("Maximum number of TCP SACK holes allowed per connection"), 2052 NULL, 0, &tcp_sack_tp_maxholes, 0, 2053 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 2054 sysctl_createv(clog, 0, NULL, &sack_node, 2055 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2056 CTLTYPE_INT, "globalmaxholes", 2057 SYSCTL_DESCR("Global maximum number of TCP SACK holes"), 2058 NULL, 0, &tcp_sack_globalmaxholes, 0, 2059 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 2060 sysctl_createv(clog, 0, NULL, &sack_node, 2061 CTLFLAG_PERMANENT, 2062 CTLTYPE_INT, "globalholes", 2063 SYSCTL_DESCR("Global number of TCP SACK holes"), 2064 NULL, 0, &tcp_sack_globalholes, 0, 2065 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 2066 2067 sysctl_createv(clog, 0, NULL, NULL, 2068 CTLFLAG_PERMANENT, 2069 CTLTYPE_STRUCT, "stats", 2070 SYSCTL_DESCR("TCP statistics"), 2071 sysctl_net_inet_tcp_stats, 0, NULL, 0, 2072 CTL_NET, pf, IPPROTO_TCP, TCPCTL_STATS, 2073 CTL_EOL); 2074 sysctl_createv(clog, 0, NULL, NULL, 2075 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2076 CTLTYPE_INT, "local_by_rtt", 2077 SYSCTL_DESCR("Use RTT estimator to decide which hosts " 2078 "are local"), 2079 NULL, 0, &tcp_rttlocal, 0, 2080 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2081 #ifdef TCP_DEBUG 2082 sysctl_createv(clog, 0, NULL, NULL, 2083 CTLFLAG_PERMANENT, 2084 CTLTYPE_STRUCT, "debug", 2085 SYSCTL_DESCR("TCP sockets debug information"), 2086 NULL, 0, &tcp_debug, sizeof(tcp_debug), 2087 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBUG, 2088 CTL_EOL); 2089 sysctl_createv(clog, 0, NULL, NULL, 2090 CTLFLAG_PERMANENT, 2091 CTLTYPE_INT, "debx", 2092 SYSCTL_DESCR("Number of TCP debug sockets messages"), 2093 NULL, 0, &tcp_debx, sizeof(tcp_debx), 2094 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBX, 2095 CTL_EOL); 2096 #endif 2097 sysctl_createv(clog, 0, NULL, NULL, 2098 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2099 CTLTYPE_STRUCT, "drop", 2100 SYSCTL_DESCR("TCP drop connection"), 2101 sysctl_net_inet_tcp_drop, 0, NULL, 0, 2102 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DROP, CTL_EOL); 2103 sysctl_createv(clog, 0, NULL, NULL, 2104 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2105 CTLTYPE_INT, "iss_hash", 2106 SYSCTL_DESCR("Enable RFC 1948 ISS by cryptographic " 2107 "hash computation"), 2108 NULL, 0, &tcp_do_rfc1948, sizeof(tcp_do_rfc1948), 2109 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, 2110 CTL_EOL); 2111 2112 /* ABC subtree */ 2113 2114 sysctl_createv(clog, 0, NULL, &abc_node, 2115 CTLFLAG_PERMANENT, CTLTYPE_NODE, "abc", 2116 SYSCTL_DESCR("RFC3465 Appropriate Byte Counting (ABC)"), 2117 NULL, 0, NULL, 0, 2118 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2119 sysctl_createv(clog, 0, &abc_node, NULL, 2120 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2121 CTLTYPE_INT, "enable", 2122 SYSCTL_DESCR("Enable RFC3465 Appropriate Byte Counting"), 2123 NULL, 0, &tcp_do_abc, 0, CTL_CREATE, CTL_EOL); 2124 sysctl_createv(clog, 0, &abc_node, NULL, 2125 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2126 CTLTYPE_INT, "aggressive", 2127 SYSCTL_DESCR("1: L=2*SMSS 0: L=1*SMSS"), 2128 NULL, 0, &tcp_abc_aggressive, 0, CTL_CREATE, CTL_EOL); 2129 2130 /* MSL tuning subtree */ 2131 2132 sysctl_createv(clog, 0, NULL, &mslt_node, 2133 CTLFLAG_PERMANENT, CTLTYPE_NODE, "mslt", 2134 SYSCTL_DESCR("MSL Tuning for TIME_WAIT truncation"), 2135 NULL, 0, NULL, 0, 2136 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2137 sysctl_createv(clog, 0, &mslt_node, NULL, 2138 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2139 CTLTYPE_INT, "enable", 2140 SYSCTL_DESCR("Enable TIME_WAIT truncation"), 2141 NULL, 0, &tcp_msl_enable, 0, CTL_CREATE, CTL_EOL); 2142 sysctl_createv(clog, 0, &mslt_node, NULL, 2143 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2144 CTLTYPE_INT, "loopback", 2145 SYSCTL_DESCR("MSL value to use for loopback connections"), 2146 NULL, 0, &tcp_msl_loop, 0, CTL_CREATE, CTL_EOL); 2147 sysctl_createv(clog, 0, &mslt_node, NULL, 2148 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2149 CTLTYPE_INT, "local", 2150 SYSCTL_DESCR("MSL value to use for local connections"), 2151 NULL, 0, &tcp_msl_local, 0, CTL_CREATE, CTL_EOL); 2152 sysctl_createv(clog, 0, &mslt_node, NULL, 2153 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2154 CTLTYPE_INT, "remote", 2155 SYSCTL_DESCR("MSL value to use for remote connections"), 2156 NULL, 0, &tcp_msl_remote, 0, CTL_CREATE, CTL_EOL); 2157 sysctl_createv(clog, 0, &mslt_node, NULL, 2158 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2159 CTLTYPE_INT, "remote_threshold", 2160 SYSCTL_DESCR("RTT estimate value to promote local to remote"), 2161 NULL, 0, &tcp_msl_remote_threshold, 0, CTL_CREATE, CTL_EOL); 2162 2163 /* vestigial TIME_WAIT tuning subtree */ 2164 2165 sysctl_createv(clog, 0, NULL, &vtw_node, 2166 CTLFLAG_PERMANENT, CTLTYPE_NODE, "vtw", 2167 SYSCTL_DESCR("Tuning for Vestigial TIME_WAIT"), 2168 NULL, 0, NULL, 0, 2169 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2170 sysctl_createv(clog, 0, &vtw_node, NULL, 2171 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2172 CTLTYPE_INT, "enable", 2173 SYSCTL_DESCR("Enable Vestigial TIME_WAIT"), 2174 sysctl_tcp_vtw_enable, 0, 2175 (pf == AF_INET) ? &tcp4_vtw_enable : &tcp6_vtw_enable, 2176 0, CTL_CREATE, CTL_EOL); 2177 sysctl_createv(clog, 0, &vtw_node, NULL, 2178 CTLFLAG_PERMANENT|CTLFLAG_READONLY, 2179 CTLTYPE_INT, "entries", 2180 SYSCTL_DESCR("Maximum number of vestigial TIME_WAIT entries"), 2181 NULL, 0, &tcp_vtw_entries, 0, CTL_CREATE, CTL_EOL); 2182 } 2183 2184 void 2185 tcp_usrreq_init(void) 2186 { 2187 2188 #ifdef INET 2189 sysctl_net_inet_tcp_setup2(NULL, PF_INET, "inet", "tcp"); 2190 #endif 2191 #ifdef INET6 2192 sysctl_net_inet_tcp_setup2(NULL, PF_INET6, "inet6", "tcp6"); 2193 #endif 2194 } 2195 2196 PR_WRAP_USRREQS(tcp) 2197 #define tcp_attach tcp_attach_wrapper 2198 #define tcp_detach tcp_detach_wrapper 2199 #define tcp_usrreq tcp_usrreq_wrapper 2200 2201 const struct pr_usrreqs tcp_usrreqs = { 2202 .pr_attach = tcp_attach, 2203 .pr_detach = tcp_detach, 2204 .pr_generic = tcp_usrreq, 2205 }; 2206
Indexes created Sun Sep 21 20:09:37 GMT 2025