tcp_usrreq.c revision 1.185
1 /* $NetBSD: tcp_usrreq.c,v 1.185 2014/07/09 04:54:04 rtr 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 1.185 2014/07/09 04:54:04 rtr 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 static int 160 tcp_debug_capture(struct tcpcb *tp, int req) 161 { 162 #ifdef KPROF 163 tcp_acounts[tp->t_state][req]++; 164 #endif 165 #ifdef TCP_DEBUG 166 return tp->t_state; 167 #endif 168 return 0; 169 } 170 171 static inline void 172 tcp_debug_trace(struct socket *so, struct tcpcb *tp, int ostate, int req) 173 { 174 #ifdef TCP_DEBUG 175 if (tp && (so->so_options & SO_DEBUG)) 176 tcp_trace(TA_USER, ostate, tp, NULL, req); 177 #endif 178 } 179 180 /* 181 * Process a TCP user request for TCP tb. If this is a send request 182 * then m is the mbuf chain of send data. If this is a timer expiration 183 * (called from the software clock routine), then timertype tells which timer. 184 */ 185 static int 186 tcp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam, 187 struct mbuf *control, struct lwp *l) 188 { 189 struct inpcb *inp; 190 #ifdef INET6 191 struct in6pcb *in6p; 192 #endif 193 struct tcpcb *tp = NULL; 194 int s; 195 int error = 0; 196 int ostate = 0; 197 int family; /* family of the socket */ 198 199 KASSERT(req != PRU_ATTACH); 200 KASSERT(req != PRU_DETACH); 201 KASSERT(req != PRU_CONTROL); 202 KASSERT(req != PRU_SENSE); 203 KASSERT(req != PRU_PEERADDR); 204 KASSERT(req != PRU_SOCKADDR); 205 206 family = so->so_proto->pr_domain->dom_family; 207 208 s = splsoftnet(); 209 210 if (req == PRU_PURGEIF) { 211 mutex_enter(softnet_lock); 212 switch (family) { 213 #ifdef INET 214 case PF_INET: 215 in_pcbpurgeif0(&tcbtable, (struct ifnet *)control); 216 in_purgeif((struct ifnet *)control); 217 in_pcbpurgeif(&tcbtable, (struct ifnet *)control); 218 break; 219 #endif 220 #ifdef INET6 221 case PF_INET6: 222 in6_pcbpurgeif0(&tcbtable, (struct ifnet *)control); 223 in6_purgeif((struct ifnet *)control); 224 in6_pcbpurgeif(&tcbtable, (struct ifnet *)control); 225 break; 226 #endif 227 default: 228 mutex_exit(softnet_lock); 229 splx(s); 230 return (EAFNOSUPPORT); 231 } 232 mutex_exit(softnet_lock); 233 splx(s); 234 return (0); 235 } 236 237 KASSERT(solocked(so)); 238 239 switch (family) { 240 #ifdef INET 241 case PF_INET: 242 inp = sotoinpcb(so); 243 #ifdef INET6 244 in6p = NULL; 245 #endif 246 break; 247 #endif 248 #ifdef INET6 249 case PF_INET6: 250 inp = NULL; 251 in6p = sotoin6pcb(so); 252 break; 253 #endif 254 default: 255 splx(s); 256 return EAFNOSUPPORT; 257 } 258 KASSERT(!control || (req == PRU_SEND || req == PRU_SENDOOB)); 259 #ifdef INET6 260 /* XXX: KASSERT((inp != NULL) ^ (in6p != NULL)); */ 261 #endif 262 /* 263 * When a TCP is attached to a socket, then there will be 264 * a (struct inpcb) pointed at by the socket, and this 265 * structure will point at a subsidary (struct tcpcb). 266 */ 267 if (inp == NULL 268 #ifdef INET6 269 && in6p == NULL 270 #endif 271 ) 272 { 273 error = EINVAL; 274 goto release; 275 } 276 #ifdef INET 277 if (inp) { 278 tp = intotcpcb(inp); 279 /* WHAT IF TP IS 0? */ 280 ostate = tcp_debug_capture(tp, req); 281 } 282 #endif 283 #ifdef INET6 284 if (in6p) { 285 tp = in6totcpcb(in6p); 286 /* WHAT IF TP IS 0? */ 287 ostate = tcp_debug_capture(tp, req); 288 } 289 #endif 290 291 switch (req) { 292 293 /* 294 * Give the socket an address. 295 */ 296 case PRU_BIND: 297 switch (family) { 298 #ifdef INET 299 case PF_INET: 300 error = in_pcbbind(inp, nam, l); 301 break; 302 #endif 303 #ifdef INET6 304 case PF_INET6: 305 error = in6_pcbbind(in6p, nam, l); 306 if (!error) { 307 /* mapped addr case */ 308 if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)) 309 tp->t_family = AF_INET; 310 else 311 tp->t_family = AF_INET6; 312 } 313 break; 314 #endif 315 } 316 break; 317 318 /* 319 * Prepare to accept connections. 320 */ 321 case PRU_LISTEN: 322 #ifdef INET 323 if (inp && inp->inp_lport == 0) { 324 error = in_pcbbind(inp, NULL, l); 325 if (error) 326 break; 327 } 328 #endif 329 #ifdef INET6 330 if (in6p && in6p->in6p_lport == 0) { 331 error = in6_pcbbind(in6p, NULL, l); 332 if (error) 333 break; 334 } 335 #endif 336 tp->t_state = TCPS_LISTEN; 337 break; 338 339 /* 340 * Initiate connection to peer. 341 * Create a template for use in transmissions on this connection. 342 * Enter SYN_SENT state, and mark socket as connecting. 343 * Start keep-alive timer, and seed output sequence space. 344 * Send initial segment on connection. 345 */ 346 case PRU_CONNECT: 347 #ifdef INET 348 if (inp) { 349 if (inp->inp_lport == 0) { 350 error = in_pcbbind(inp, NULL, l); 351 if (error) 352 break; 353 } 354 error = in_pcbconnect(inp, nam, l); 355 } 356 #endif 357 #ifdef INET6 358 if (in6p) { 359 if (in6p->in6p_lport == 0) { 360 error = in6_pcbbind(in6p, NULL, l); 361 if (error) 362 break; 363 } 364 error = in6_pcbconnect(in6p, nam, l); 365 if (!error) { 366 /* mapped addr case */ 367 if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) 368 tp->t_family = AF_INET; 369 else 370 tp->t_family = AF_INET6; 371 } 372 } 373 #endif 374 if (error) 375 break; 376 tp->t_template = tcp_template(tp); 377 if (tp->t_template == 0) { 378 #ifdef INET 379 if (inp) 380 in_pcbdisconnect(inp); 381 #endif 382 #ifdef INET6 383 if (in6p) 384 in6_pcbdisconnect(in6p); 385 #endif 386 error = ENOBUFS; 387 break; 388 } 389 /* 390 * Compute window scaling to request. 391 * XXX: This should be moved to tcp_output(). 392 */ 393 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 394 (TCP_MAXWIN << tp->request_r_scale) < sb_max) 395 tp->request_r_scale++; 396 soisconnecting(so); 397 TCP_STATINC(TCP_STAT_CONNATTEMPT); 398 tp->t_state = TCPS_SYN_SENT; 399 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit); 400 tp->iss = tcp_new_iss(tp, 0); 401 tcp_sendseqinit(tp); 402 error = tcp_output(tp); 403 break; 404 405 /* 406 * Create a TCP connection between two sockets. 407 */ 408 case PRU_CONNECT2: 409 error = EOPNOTSUPP; 410 break; 411 412 /* 413 * Initiate disconnect from peer. 414 * If connection never passed embryonic stage, just drop; 415 * else if don't need to let data drain, then can just drop anyways, 416 * else have to begin TCP shutdown process: mark socket disconnecting, 417 * drain unread data, state switch to reflect user close, and 418 * send segment (e.g. FIN) to peer. Socket will be really disconnected 419 * when peer sends FIN and acks ours. 420 * 421 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 422 */ 423 case PRU_DISCONNECT: 424 tp = tcp_disconnect(tp); 425 break; 426 427 /* 428 * Accept a connection. Essentially all the work is 429 * done at higher levels; just return the address 430 * of the peer, storing through addr. 431 */ 432 case PRU_ACCEPT: 433 #ifdef INET 434 if (inp) 435 in_setpeeraddr(inp, nam); 436 #endif 437 #ifdef INET6 438 if (in6p) 439 in6_setpeeraddr(in6p, nam); 440 #endif 441 break; 442 443 /* 444 * Mark the connection as being incapable of further output. 445 */ 446 case PRU_SHUTDOWN: 447 socantsendmore(so); 448 tp = tcp_usrclosed(tp); 449 if (tp) 450 error = tcp_output(tp); 451 break; 452 453 /* 454 * After a receive, possibly send window update to peer. 455 */ 456 case PRU_RCVD: 457 /* 458 * soreceive() calls this function when a user receives 459 * ancillary data on a listening socket. We don't call 460 * tcp_output in such a case, since there is no header 461 * template for a listening socket and hence the kernel 462 * will panic. 463 */ 464 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) != 0) 465 (void) tcp_output(tp); 466 break; 467 468 /* 469 * Do a send by putting data in output queue and updating urgent 470 * marker if URG set. Possibly send more data. 471 */ 472 case PRU_SEND: 473 if (control && control->m_len) { 474 m_freem(control); 475 m_freem(m); 476 error = EINVAL; 477 break; 478 } 479 sbappendstream(&so->so_snd, m); 480 error = tcp_output(tp); 481 break; 482 483 /* 484 * Abort the TCP. 485 */ 486 case PRU_ABORT: 487 tp = tcp_drop(tp, ECONNABORTED); 488 break; 489 490 case PRU_RCVOOB: 491 if (control && control->m_len) { 492 m_freem(control); 493 m_freem(m); 494 error = EINVAL; 495 break; 496 } 497 if ((so->so_oobmark == 0 && 498 (so->so_state & SS_RCVATMARK) == 0) || 499 so->so_options & SO_OOBINLINE || 500 tp->t_oobflags & TCPOOB_HADDATA) { 501 error = EINVAL; 502 break; 503 } 504 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 505 error = EWOULDBLOCK; 506 break; 507 } 508 m->m_len = 1; 509 *mtod(m, char *) = tp->t_iobc; 510 if (((long)nam & MSG_PEEK) == 0) 511 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 512 break; 513 514 case PRU_SENDOOB: 515 if (sbspace(&so->so_snd) < -512) { 516 m_freem(m); 517 error = ENOBUFS; 518 break; 519 } 520 /* 521 * According to RFC961 (Assigned Protocols), 522 * the urgent pointer points to the last octet 523 * of urgent data. We continue, however, 524 * to consider it to indicate the first octet 525 * of data past the urgent section. 526 * Otherwise, snd_up should be one lower. 527 */ 528 sbappendstream(&so->so_snd, m); 529 tp->snd_up = tp->snd_una + so->so_snd.sb_cc; 530 tp->t_force = 1; 531 error = tcp_output(tp); 532 tp->t_force = 0; 533 break; 534 535 default: 536 panic("tcp_usrreq"); 537 } 538 539 tcp_debug_trace(so, tp, ostate, req); 540 541 release: 542 splx(s); 543 return (error); 544 } 545 546 static void 547 change_keepalive(struct socket *so, struct tcpcb *tp) 548 { 549 tp->t_maxidle = tp->t_keepcnt * tp->t_keepintvl; 550 TCP_TIMER_DISARM(tp, TCPT_KEEP); 551 TCP_TIMER_DISARM(tp, TCPT_2MSL); 552 553 if (tp->t_state == TCPS_SYN_RECEIVED || 554 tp->t_state == TCPS_SYN_SENT) { 555 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit); 556 } else if (so->so_options & SO_KEEPALIVE && 557 tp->t_state <= TCPS_CLOSE_WAIT) { 558 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepintvl); 559 } else { 560 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle); 561 } 562 563 if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0)) 564 TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle); 565 } 566 567 568 int 569 tcp_ctloutput(int op, struct socket *so, struct sockopt *sopt) 570 { 571 int error = 0, s; 572 struct inpcb *inp; 573 #ifdef INET6 574 struct in6pcb *in6p; 575 #endif 576 struct tcpcb *tp; 577 u_int ui; 578 int family; /* family of the socket */ 579 int level, optname, optval; 580 581 level = sopt->sopt_level; 582 optname = sopt->sopt_name; 583 584 family = so->so_proto->pr_domain->dom_family; 585 586 s = splsoftnet(); 587 switch (family) { 588 #ifdef INET 589 case PF_INET: 590 inp = sotoinpcb(so); 591 #ifdef INET6 592 in6p = NULL; 593 #endif 594 break; 595 #endif 596 #ifdef INET6 597 case PF_INET6: 598 inp = NULL; 599 in6p = sotoin6pcb(so); 600 break; 601 #endif 602 default: 603 splx(s); 604 panic("%s: af %d", __func__, family); 605 } 606 #ifndef INET6 607 if (inp == NULL) 608 #else 609 if (inp == NULL && in6p == NULL) 610 #endif 611 { 612 splx(s); 613 return (ECONNRESET); 614 } 615 if (level != IPPROTO_TCP) { 616 switch (family) { 617 #ifdef INET 618 case PF_INET: 619 error = ip_ctloutput(op, so, sopt); 620 break; 621 #endif 622 #ifdef INET6 623 case PF_INET6: 624 error = ip6_ctloutput(op, so, sopt); 625 break; 626 #endif 627 } 628 splx(s); 629 return (error); 630 } 631 if (inp) 632 tp = intotcpcb(inp); 633 #ifdef INET6 634 else if (in6p) 635 tp = in6totcpcb(in6p); 636 #endif 637 else 638 tp = NULL; 639 640 switch (op) { 641 case PRCO_SETOPT: 642 switch (optname) { 643 #ifdef TCP_SIGNATURE 644 case TCP_MD5SIG: 645 error = sockopt_getint(sopt, &optval); 646 if (error) 647 break; 648 if (optval > 0) 649 tp->t_flags |= TF_SIGNATURE; 650 else 651 tp->t_flags &= ~TF_SIGNATURE; 652 break; 653 #endif /* TCP_SIGNATURE */ 654 655 case TCP_NODELAY: 656 error = sockopt_getint(sopt, &optval); 657 if (error) 658 break; 659 if (optval) 660 tp->t_flags |= TF_NODELAY; 661 else 662 tp->t_flags &= ~TF_NODELAY; 663 break; 664 665 case TCP_MAXSEG: 666 error = sockopt_getint(sopt, &optval); 667 if (error) 668 break; 669 if (optval > 0 && optval <= tp->t_peermss) 670 tp->t_peermss = optval; /* limit on send size */ 671 else 672 error = EINVAL; 673 break; 674 #ifdef notyet 675 case TCP_CONGCTL: 676 /* XXX string overflow XXX */ 677 error = tcp_congctl_select(tp, sopt->sopt_data); 678 break; 679 #endif 680 681 case TCP_KEEPIDLE: 682 error = sockopt_get(sopt, &ui, sizeof(ui)); 683 if (error) 684 break; 685 if (ui > 0) { 686 tp->t_keepidle = ui; 687 change_keepalive(so, tp); 688 } else 689 error = EINVAL; 690 break; 691 692 case TCP_KEEPINTVL: 693 error = sockopt_get(sopt, &ui, sizeof(ui)); 694 if (error) 695 break; 696 if (ui > 0) { 697 tp->t_keepintvl = ui; 698 change_keepalive(so, tp); 699 } else 700 error = EINVAL; 701 break; 702 703 case TCP_KEEPCNT: 704 error = sockopt_get(sopt, &ui, sizeof(ui)); 705 if (error) 706 break; 707 if (ui > 0) { 708 tp->t_keepcnt = ui; 709 change_keepalive(so, tp); 710 } else 711 error = EINVAL; 712 break; 713 714 case TCP_KEEPINIT: 715 error = sockopt_get(sopt, &ui, sizeof(ui)); 716 if (error) 717 break; 718 if (ui > 0) { 719 tp->t_keepinit = ui; 720 change_keepalive(so, tp); 721 } else 722 error = EINVAL; 723 break; 724 725 default: 726 error = ENOPROTOOPT; 727 break; 728 } 729 break; 730 731 case PRCO_GETOPT: 732 switch (optname) { 733 #ifdef TCP_SIGNATURE 734 case TCP_MD5SIG: 735 optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0; 736 error = sockopt_set(sopt, &optval, sizeof(optval)); 737 break; 738 #endif 739 case TCP_NODELAY: 740 optval = tp->t_flags & TF_NODELAY; 741 error = sockopt_set(sopt, &optval, sizeof(optval)); 742 break; 743 case TCP_MAXSEG: 744 optval = tp->t_peermss; 745 error = sockopt_set(sopt, &optval, sizeof(optval)); 746 break; 747 #ifdef notyet 748 case TCP_CONGCTL: 749 break; 750 #endif 751 default: 752 error = ENOPROTOOPT; 753 break; 754 } 755 break; 756 } 757 splx(s); 758 return (error); 759 } 760 761 #ifndef TCP_SENDSPACE 762 #define TCP_SENDSPACE 1024*32 763 #endif 764 int tcp_sendspace = TCP_SENDSPACE; 765 #ifndef TCP_RECVSPACE 766 #define TCP_RECVSPACE 1024*32 767 #endif 768 int tcp_recvspace = TCP_RECVSPACE; 769 770 /* 771 * tcp_attach: attach TCP protocol to socket, allocating internet protocol 772 * control block, TCP control block, buffer space and entering LISTEN state 773 * if to accept connections. 774 */ 775 static int 776 tcp_attach(struct socket *so, int proto) 777 { 778 struct tcpcb *tp; 779 struct inpcb *inp; 780 #ifdef INET6 781 struct in6pcb *in6p; 782 #endif 783 int s, error, family; 784 785 /* Assign the lock (must happen even if we will error out). */ 786 s = splsoftnet(); 787 sosetlock(so); 788 KASSERT(solocked(so)); 789 790 family = so->so_proto->pr_domain->dom_family; 791 switch (family) { 792 #ifdef INET 793 case PF_INET: 794 inp = sotoinpcb(so); 795 #ifdef INET6 796 in6p = NULL; 797 #endif 798 break; 799 #endif 800 #ifdef INET6 801 case PF_INET6: 802 inp = NULL; 803 in6p = sotoin6pcb(so); 804 break; 805 #endif 806 default: 807 error = EAFNOSUPPORT; 808 goto out; 809 } 810 811 KASSERT(inp == NULL); 812 #ifdef INET6 813 KASSERT(in6p == NULL); 814 #endif 815 816 #ifdef MBUFTRACE 817 so->so_mowner = &tcp_sock_mowner; 818 so->so_rcv.sb_mowner = &tcp_sock_rx_mowner; 819 so->so_snd.sb_mowner = &tcp_sock_tx_mowner; 820 #endif 821 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 822 error = soreserve(so, tcp_sendspace, tcp_recvspace); 823 if (error) 824 goto out; 825 } 826 827 so->so_rcv.sb_flags |= SB_AUTOSIZE; 828 so->so_snd.sb_flags |= SB_AUTOSIZE; 829 830 switch (family) { 831 #ifdef INET 832 case PF_INET: 833 error = in_pcballoc(so, &tcbtable); 834 if (error) 835 goto out; 836 inp = sotoinpcb(so); 837 #ifdef INET6 838 in6p = NULL; 839 #endif 840 break; 841 #endif 842 #ifdef INET6 843 case PF_INET6: 844 error = in6_pcballoc(so, &tcbtable); 845 if (error) 846 goto out; 847 inp = NULL; 848 in6p = sotoin6pcb(so); 849 break; 850 #endif 851 default: 852 error = EAFNOSUPPORT; 853 goto out; 854 } 855 if (inp) 856 tp = tcp_newtcpcb(family, (void *)inp); 857 #ifdef INET6 858 else if (in6p) 859 tp = tcp_newtcpcb(family, (void *)in6p); 860 #endif 861 else 862 tp = NULL; 863 864 if (tp == NULL) { 865 int nofd = so->so_state & SS_NOFDREF; /* XXX */ 866 867 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ 868 #ifdef INET 869 if (inp) 870 in_pcbdetach(inp); 871 #endif 872 #ifdef INET6 873 if (in6p) 874 in6_pcbdetach(in6p); 875 #endif 876 so->so_state |= nofd; 877 error = ENOBUFS; 878 goto out; 879 } 880 tp->t_state = TCPS_CLOSED; 881 if ((so->so_options & SO_LINGER) && so->so_linger == 0) { 882 so->so_linger = TCP_LINGERTIME; 883 } 884 out: 885 KASSERT(solocked(so)); 886 splx(s); 887 return error; 888 } 889 890 static void 891 tcp_detach(struct socket *so) 892 { 893 struct inpcb *inp; 894 #ifdef INET6 895 struct in6pcb *in6p; 896 #endif 897 struct tcpcb *tp = NULL; 898 int s, family; 899 900 KASSERT(solocked(so)); 901 902 s = splsoftnet(); 903 family = so->so_proto->pr_domain->dom_family; 904 switch (family) { 905 #ifdef INET 906 case PF_INET: 907 inp = sotoinpcb(so); 908 tp = intotcpcb(inp); 909 break; 910 #endif 911 #ifdef INET6 912 case PF_INET6: 913 in6p = sotoin6pcb(so); 914 tp = in6totcpcb(in6p); 915 break; 916 #endif 917 default: 918 splx(s); 919 return; 920 } 921 KASSERT(tp != NULL); 922 (void)tcp_disconnect(tp); 923 splx(s); 924 } 925 926 static int 927 tcp_ioctl(struct socket *so, u_long cmd, void *nam, struct ifnet *ifp) 928 { 929 switch (so->so_proto->pr_domain->dom_family) { 930 #ifdef INET 931 case PF_INET: 932 return in_control(so, cmd, nam, ifp); 933 #endif 934 #ifdef INET6 935 case PF_INET6: 936 return in6_control(so, cmd, nam, ifp); 937 #endif 938 default: 939 return EAFNOSUPPORT; 940 } 941 } 942 943 static int 944 tcp_stat(struct socket *so, struct stat *ub) 945 { 946 KASSERT(solocked(so)); 947 948 /* stat: don't bother with a blocksize. */ 949 return 0; 950 } 951 952 static int 953 tcp_peeraddr(struct socket *so, struct mbuf *nam) 954 { 955 struct inpcb *inp = NULL; 956 #ifdef INET6 957 struct in6pcb *in6p = NULL; 958 #endif 959 struct tcpcb *tp = NULL; 960 int ostate = 0; 961 962 switch (so->so_proto->pr_domain->dom_family) { 963 #ifdef INET 964 case PF_INET: 965 inp = sotoinpcb(so); 966 break; 967 #endif 968 #ifdef INET6 969 case PF_INET6: 970 in6p = sotoin6pcb(so); 971 break; 972 #endif 973 default: 974 return EAFNOSUPPORT; 975 } 976 977 if (inp == NULL 978 #ifdef INET6 979 && in6p == NULL 980 #endif 981 ) 982 return EINVAL; 983 984 #ifdef INET 985 if (inp) { 986 tp = intotcpcb(inp); 987 ostate = tcp_debug_capture(tp, PRU_PEERADDR); 988 in_setpeeraddr(inp, nam); 989 } 990 #endif 991 #ifdef INET6 992 if (in6p) { 993 tp = in6totcpcb(in6p); 994 ostate = tcp_debug_capture(tp, PRU_PEERADDR); 995 in6_setpeeraddr(in6p, nam); 996 } 997 #endif 998 999 tcp_debug_trace(so, tp, ostate, PRU_PEERADDR); 1000 1001 return 0; 1002 } 1003 1004 static int 1005 tcp_sockaddr(struct socket *so, struct mbuf *nam) 1006 { 1007 struct inpcb *inp = NULL; 1008 #ifdef INET6 1009 struct in6pcb *in6p = NULL; 1010 #endif 1011 struct tcpcb *tp = NULL; 1012 int ostate = 0; 1013 1014 switch (so->so_proto->pr_domain->dom_family) { 1015 #ifdef INET 1016 case PF_INET: 1017 inp = sotoinpcb(so); 1018 break; 1019 #endif 1020 #ifdef INET6 1021 case PF_INET6: 1022 in6p = sotoin6pcb(so); 1023 break; 1024 #endif 1025 default: 1026 return EAFNOSUPPORT; 1027 } 1028 1029 if (inp == NULL 1030 #ifdef INET6 1031 && in6p == NULL 1032 #endif 1033 ) 1034 return EINVAL; 1035 1036 #ifdef INET 1037 if (inp) { 1038 tp = intotcpcb(inp); 1039 ostate = tcp_debug_capture(tp, PRU_SOCKADDR); 1040 in_setsockaddr(inp, nam); 1041 } 1042 #endif 1043 #ifdef INET6 1044 if (in6p) { 1045 tp = in6totcpcb(in6p); 1046 ostate = tcp_debug_capture(tp, PRU_SOCKADDR); 1047 in6_setsockaddr(in6p, nam); 1048 } 1049 #endif 1050 1051 tcp_debug_trace(so, tp, ostate, PRU_SOCKADDR); 1052 1053 return 0; 1054 } 1055 1056 /* 1057 * Initiate (or continue) disconnect. 1058 * If embryonic state, just send reset (once). 1059 * If in ``let data drain'' option and linger null, just drop. 1060 * Otherwise (hard), mark socket disconnecting and drop 1061 * current input data; switch states based on user close, and 1062 * send segment to peer (with FIN). 1063 */ 1064 struct tcpcb * 1065 tcp_disconnect(struct tcpcb *tp) 1066 { 1067 struct socket *so; 1068 1069 if (tp->t_inpcb) 1070 so = tp->t_inpcb->inp_socket; 1071 #ifdef INET6 1072 else if (tp->t_in6pcb) 1073 so = tp->t_in6pcb->in6p_socket; 1074 #endif 1075 else 1076 so = NULL; 1077 1078 if (TCPS_HAVEESTABLISHED(tp->t_state) == 0) 1079 tp = tcp_close(tp); 1080 else if ((so->so_options & SO_LINGER) && so->so_linger == 0) 1081 tp = tcp_drop(tp, 0); 1082 else { 1083 soisdisconnecting(so); 1084 sbflush(&so->so_rcv); 1085 tp = tcp_usrclosed(tp); 1086 if (tp) 1087 (void) tcp_output(tp); 1088 } 1089 return (tp); 1090 } 1091 1092 /* 1093 * User issued close, and wish to trail through shutdown states: 1094 * if never received SYN, just forget it. If got a SYN from peer, 1095 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 1096 * If already got a FIN from peer, then almost done; go to LAST_ACK 1097 * state. In all other cases, have already sent FIN to peer (e.g. 1098 * after PRU_SHUTDOWN), and just have to play tedious game waiting 1099 * for peer to send FIN or not respond to keep-alives, etc. 1100 * We can let the user exit from the close as soon as the FIN is acked. 1101 */ 1102 struct tcpcb * 1103 tcp_usrclosed(struct tcpcb *tp) 1104 { 1105 1106 switch (tp->t_state) { 1107 1108 case TCPS_CLOSED: 1109 case TCPS_LISTEN: 1110 case TCPS_SYN_SENT: 1111 tp->t_state = TCPS_CLOSED; 1112 tp = tcp_close(tp); 1113 break; 1114 1115 case TCPS_SYN_RECEIVED: 1116 case TCPS_ESTABLISHED: 1117 tp->t_state = TCPS_FIN_WAIT_1; 1118 break; 1119 1120 case TCPS_CLOSE_WAIT: 1121 tp->t_state = TCPS_LAST_ACK; 1122 break; 1123 } 1124 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { 1125 struct socket *so; 1126 if (tp->t_inpcb) 1127 so = tp->t_inpcb->inp_socket; 1128 #ifdef INET6 1129 else if (tp->t_in6pcb) 1130 so = tp->t_in6pcb->in6p_socket; 1131 #endif 1132 else 1133 so = NULL; 1134 if (so) 1135 soisdisconnected(so); 1136 /* 1137 * If we are in FIN_WAIT_2, we arrived here because the 1138 * application did a shutdown of the send side. Like the 1139 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after 1140 * a full close, we start a timer to make sure sockets are 1141 * not left in FIN_WAIT_2 forever. 1142 */ 1143 if ((tp->t_state == TCPS_FIN_WAIT_2) && (tp->t_maxidle > 0)) 1144 TCP_TIMER_ARM(tp, TCPT_2MSL, tp->t_maxidle); 1145 else if (tp->t_state == TCPS_TIME_WAIT 1146 && ((tp->t_inpcb 1147 && (tcp4_vtw_enable & 1) 1148 && vtw_add(AF_INET, tp)) 1149 || 1150 (tp->t_in6pcb 1151 && (tcp6_vtw_enable & 1) 1152 && vtw_add(AF_INET6, tp)))) { 1153 tp = 0; 1154 } 1155 } 1156 return (tp); 1157 } 1158 1159 /* 1160 * sysctl helper routine for net.inet.ip.mssdflt. it can't be less 1161 * than 32. 1162 */ 1163 static int 1164 sysctl_net_inet_tcp_mssdflt(SYSCTLFN_ARGS) 1165 { 1166 int error, mssdflt; 1167 struct sysctlnode node; 1168 1169 mssdflt = tcp_mssdflt; 1170 node = *rnode; 1171 node.sysctl_data = &mssdflt; 1172 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1173 if (error || newp == NULL) 1174 return (error); 1175 1176 if (mssdflt < 32) 1177 return (EINVAL); 1178 tcp_mssdflt = mssdflt; 1179 1180 mutex_enter(softnet_lock); 1181 tcp_tcpcb_template(); 1182 mutex_exit(softnet_lock); 1183 1184 return (0); 1185 } 1186 1187 /* 1188 * sysctl helper for TCP CB template update 1189 */ 1190 static int 1191 sysctl_update_tcpcb_template(SYSCTLFN_ARGS) 1192 { 1193 int t, error; 1194 struct sysctlnode node; 1195 1196 /* follow procedures in sysctl(9) manpage */ 1197 t = *(int *)rnode->sysctl_data; 1198 node = *rnode; 1199 node.sysctl_data = &t; 1200 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1201 if (error || newp == NULL) 1202 return error; 1203 1204 if (t < 0) 1205 return EINVAL; 1206 1207 *(int *)rnode->sysctl_data = t; 1208 1209 mutex_enter(softnet_lock); 1210 tcp_tcpcb_template(); 1211 mutex_exit(softnet_lock); 1212 1213 return 0; 1214 } 1215 1216 /* 1217 * sysctl helper routine for setting port related values under 1218 * net.inet.ip and net.inet6.ip6. does basic range checking and does 1219 * additional checks for each type. this code has placed in 1220 * tcp_input.c since INET and INET6 both use the same tcp code. 1221 * 1222 * this helper is not static so that both inet and inet6 can use it. 1223 */ 1224 int 1225 sysctl_net_inet_ip_ports(SYSCTLFN_ARGS) 1226 { 1227 int error, tmp; 1228 int apmin, apmax; 1229 #ifndef IPNOPRIVPORTS 1230 int lpmin, lpmax; 1231 #endif /* IPNOPRIVPORTS */ 1232 struct sysctlnode node; 1233 1234 if (namelen != 0) 1235 return (EINVAL); 1236 1237 switch (name[-3]) { 1238 #ifdef INET 1239 case PF_INET: 1240 apmin = anonportmin; 1241 apmax = anonportmax; 1242 #ifndef IPNOPRIVPORTS 1243 lpmin = lowportmin; 1244 lpmax = lowportmax; 1245 #endif /* IPNOPRIVPORTS */ 1246 break; 1247 #endif /* INET */ 1248 #ifdef INET6 1249 case PF_INET6: 1250 apmin = ip6_anonportmin; 1251 apmax = ip6_anonportmax; 1252 #ifndef IPNOPRIVPORTS 1253 lpmin = ip6_lowportmin; 1254 lpmax = ip6_lowportmax; 1255 #endif /* IPNOPRIVPORTS */ 1256 break; 1257 #endif /* INET6 */ 1258 default: 1259 return (EINVAL); 1260 } 1261 1262 /* 1263 * insert temporary copy into node, perform lookup on 1264 * temporary, then restore pointer 1265 */ 1266 node = *rnode; 1267 tmp = *(int*)rnode->sysctl_data; 1268 node.sysctl_data = &tmp; 1269 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1270 if (error || newp == NULL) 1271 return (error); 1272 1273 /* 1274 * simple port range check 1275 */ 1276 if (tmp < 0 || tmp > 65535) 1277 return (EINVAL); 1278 1279 /* 1280 * per-node range checks 1281 */ 1282 switch (rnode->sysctl_num) { 1283 case IPCTL_ANONPORTMIN: 1284 case IPV6CTL_ANONPORTMIN: 1285 if (tmp >= apmax) 1286 return (EINVAL); 1287 #ifndef IPNOPRIVPORTS 1288 if (tmp < IPPORT_RESERVED) 1289 return (EINVAL); 1290 #endif /* IPNOPRIVPORTS */ 1291 break; 1292 1293 case IPCTL_ANONPORTMAX: 1294 case IPV6CTL_ANONPORTMAX: 1295 if (apmin >= tmp) 1296 return (EINVAL); 1297 #ifndef IPNOPRIVPORTS 1298 if (tmp < IPPORT_RESERVED) 1299 return (EINVAL); 1300 #endif /* IPNOPRIVPORTS */ 1301 break; 1302 1303 #ifndef IPNOPRIVPORTS 1304 case IPCTL_LOWPORTMIN: 1305 case IPV6CTL_LOWPORTMIN: 1306 if (tmp >= lpmax || 1307 tmp > IPPORT_RESERVEDMAX || 1308 tmp < IPPORT_RESERVEDMIN) 1309 return (EINVAL); 1310 break; 1311 1312 case IPCTL_LOWPORTMAX: 1313 case IPV6CTL_LOWPORTMAX: 1314 if (lpmin >= tmp || 1315 tmp > IPPORT_RESERVEDMAX || 1316 tmp < IPPORT_RESERVEDMIN) 1317 return (EINVAL); 1318 break; 1319 #endif /* IPNOPRIVPORTS */ 1320 1321 default: 1322 return (EINVAL); 1323 } 1324 1325 *(int*)rnode->sysctl_data = tmp; 1326 1327 return (0); 1328 } 1329 1330 static inline int 1331 copyout_uid(struct socket *sockp, void *oldp, size_t *oldlenp) 1332 { 1333 if (oldp) { 1334 size_t sz; 1335 uid_t uid; 1336 int error; 1337 1338 if (sockp->so_cred == NULL) 1339 return EPERM; 1340 1341 uid = kauth_cred_geteuid(sockp->so_cred); 1342 sz = MIN(sizeof(uid), *oldlenp); 1343 if ((error = copyout(&uid, oldp, sz)) != 0) 1344 return error; 1345 } 1346 *oldlenp = sizeof(uid_t); 1347 return 0; 1348 } 1349 1350 static inline int 1351 inet4_ident_core(struct in_addr raddr, u_int rport, 1352 struct in_addr laddr, u_int lport, 1353 void *oldp, size_t *oldlenp, 1354 struct lwp *l, int dodrop) 1355 { 1356 struct inpcb *inp; 1357 struct socket *sockp; 1358 1359 inp = in_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport, 0); 1360 1361 if (inp == NULL || (sockp = inp->inp_socket) == NULL) 1362 return ESRCH; 1363 1364 if (dodrop) { 1365 struct tcpcb *tp; 1366 int error; 1367 1368 if (inp == NULL || (tp = intotcpcb(inp)) == NULL || 1369 (inp->inp_socket->so_options & SO_ACCEPTCONN) != 0) 1370 return ESRCH; 1371 1372 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, 1373 KAUTH_REQ_NETWORK_SOCKET_DROP, inp->inp_socket, tp, NULL); 1374 if (error) 1375 return (error); 1376 1377 (void)tcp_drop(tp, ECONNABORTED); 1378 return 0; 1379 } 1380 else 1381 return copyout_uid(sockp, oldp, oldlenp); 1382 } 1383 1384 #ifdef INET6 1385 static inline int 1386 inet6_ident_core(struct in6_addr *raddr, u_int rport, 1387 struct in6_addr *laddr, u_int lport, 1388 void *oldp, size_t *oldlenp, 1389 struct lwp *l, int dodrop) 1390 { 1391 struct in6pcb *in6p; 1392 struct socket *sockp; 1393 1394 in6p = in6_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport, 0, 0); 1395 1396 if (in6p == NULL || (sockp = in6p->in6p_socket) == NULL) 1397 return ESRCH; 1398 1399 if (dodrop) { 1400 struct tcpcb *tp; 1401 int error; 1402 1403 if (in6p == NULL || (tp = in6totcpcb(in6p)) == NULL || 1404 (in6p->in6p_socket->so_options & SO_ACCEPTCONN) != 0) 1405 return ESRCH; 1406 1407 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, 1408 KAUTH_REQ_NETWORK_SOCKET_DROP, in6p->in6p_socket, tp, NULL); 1409 if (error) 1410 return (error); 1411 1412 (void)tcp_drop(tp, ECONNABORTED); 1413 return 0; 1414 } 1415 else 1416 return copyout_uid(sockp, oldp, oldlenp); 1417 } 1418 #endif 1419 1420 /* 1421 * sysctl helper routine for the net.inet.tcp.drop and 1422 * net.inet6.tcp6.drop nodes. 1423 */ 1424 #define sysctl_net_inet_tcp_drop sysctl_net_inet_tcp_ident 1425 1426 /* 1427 * sysctl helper routine for the net.inet.tcp.ident and 1428 * net.inet6.tcp6.ident nodes. contains backwards compat code for the 1429 * old way of looking up the ident information for ipv4 which involves 1430 * stuffing the port/addr pairs into the mib lookup. 1431 */ 1432 static int 1433 sysctl_net_inet_tcp_ident(SYSCTLFN_ARGS) 1434 { 1435 #ifdef INET 1436 struct sockaddr_in *si4[2]; 1437 #endif /* INET */ 1438 #ifdef INET6 1439 struct sockaddr_in6 *si6[2]; 1440 #endif /* INET6 */ 1441 struct sockaddr_storage sa[2]; 1442 int error, pf, dodrop; 1443 1444 dodrop = name[-1] == TCPCTL_DROP; 1445 if (dodrop) { 1446 if (oldp != NULL || *oldlenp != 0) 1447 return EINVAL; 1448 if (newp == NULL) 1449 return EPERM; 1450 if (newlen < sizeof(sa)) 1451 return ENOMEM; 1452 } 1453 if (namelen != 4 && namelen != 0) 1454 return EINVAL; 1455 if (name[-2] != IPPROTO_TCP) 1456 return EINVAL; 1457 pf = name[-3]; 1458 1459 /* old style lookup, ipv4 only */ 1460 if (namelen == 4) { 1461 #ifdef INET 1462 struct in_addr laddr, raddr; 1463 u_int lport, rport; 1464 1465 if (pf != PF_INET) 1466 return EPROTONOSUPPORT; 1467 raddr.s_addr = (uint32_t)name[0]; 1468 rport = (u_int)name[1]; 1469 laddr.s_addr = (uint32_t)name[2]; 1470 lport = (u_int)name[3]; 1471 1472 mutex_enter(softnet_lock); 1473 error = inet4_ident_core(raddr, rport, laddr, lport, 1474 oldp, oldlenp, l, dodrop); 1475 mutex_exit(softnet_lock); 1476 return error; 1477 #else /* INET */ 1478 return EINVAL; 1479 #endif /* INET */ 1480 } 1481 1482 if (newp == NULL || newlen != sizeof(sa)) 1483 return EINVAL; 1484 error = copyin(newp, &sa, newlen); 1485 if (error) 1486 return error; 1487 1488 /* 1489 * requested families must match 1490 */ 1491 if (pf != sa[0].ss_family || sa[0].ss_family != sa[1].ss_family) 1492 return EINVAL; 1493 1494 switch (pf) { 1495 #ifdef INET6 1496 case PF_INET6: 1497 si6[0] = (struct sockaddr_in6*)&sa[0]; 1498 si6[1] = (struct sockaddr_in6*)&sa[1]; 1499 if (si6[0]->sin6_len != sizeof(*si6[0]) || 1500 si6[1]->sin6_len != sizeof(*si6[1])) 1501 return EINVAL; 1502 1503 if (!IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) && 1504 !IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr)) { 1505 error = sa6_embedscope(si6[0], ip6_use_defzone); 1506 if (error) 1507 return error; 1508 error = sa6_embedscope(si6[1], ip6_use_defzone); 1509 if (error) 1510 return error; 1511 1512 mutex_enter(softnet_lock); 1513 error = inet6_ident_core(&si6[0]->sin6_addr, 1514 si6[0]->sin6_port, &si6[1]->sin6_addr, 1515 si6[1]->sin6_port, oldp, oldlenp, l, dodrop); 1516 mutex_exit(softnet_lock); 1517 return error; 1518 } 1519 1520 if (IN6_IS_ADDR_V4MAPPED(&si6[0]->sin6_addr) != 1521 IN6_IS_ADDR_V4MAPPED(&si6[1]->sin6_addr)) 1522 return EINVAL; 1523 1524 in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[0]); 1525 in6_sin6_2_sin_in_sock((struct sockaddr *)&sa[1]); 1526 /*FALLTHROUGH*/ 1527 #endif /* INET6 */ 1528 #ifdef INET 1529 case PF_INET: 1530 si4[0] = (struct sockaddr_in*)&sa[0]; 1531 si4[1] = (struct sockaddr_in*)&sa[1]; 1532 if (si4[0]->sin_len != sizeof(*si4[0]) || 1533 si4[0]->sin_len != sizeof(*si4[1])) 1534 return EINVAL; 1535 1536 mutex_enter(softnet_lock); 1537 error = inet4_ident_core(si4[0]->sin_addr, si4[0]->sin_port, 1538 si4[1]->sin_addr, si4[1]->sin_port, 1539 oldp, oldlenp, l, dodrop); 1540 mutex_exit(softnet_lock); 1541 return error; 1542 #endif /* INET */ 1543 default: 1544 return EPROTONOSUPPORT; 1545 } 1546 } 1547 1548 /* 1549 * sysctl helper for the inet and inet6 pcblists. handles tcp/udp and 1550 * inet/inet6, as well as raw pcbs for each. specifically not 1551 * declared static so that raw sockets and udp/udp6 can use it as 1552 * well. 1553 */ 1554 int 1555 sysctl_inpcblist(SYSCTLFN_ARGS) 1556 { 1557 #ifdef INET 1558 struct sockaddr_in *in; 1559 const struct inpcb *inp; 1560 #endif 1561 #ifdef INET6 1562 struct sockaddr_in6 *in6; 1563 const struct in6pcb *in6p; 1564 #endif 1565 struct inpcbtable *pcbtbl = __UNCONST(rnode->sysctl_data); 1566 const struct inpcb_hdr *inph; 1567 struct tcpcb *tp; 1568 struct kinfo_pcb pcb; 1569 char *dp; 1570 size_t len, needed, elem_size, out_size; 1571 int error, elem_count, pf, proto, pf2; 1572 1573 if (namelen != 4) 1574 return (EINVAL); 1575 1576 if (oldp != NULL) { 1577 len = *oldlenp; 1578 elem_size = name[2]; 1579 elem_count = name[3]; 1580 if (elem_size != sizeof(pcb)) 1581 return EINVAL; 1582 } else { 1583 len = 0; 1584 elem_count = INT_MAX; 1585 elem_size = sizeof(pcb); 1586 } 1587 error = 0; 1588 dp = oldp; 1589 out_size = elem_size; 1590 needed = 0; 1591 1592 if (namelen == 1 && name[0] == CTL_QUERY) 1593 return (sysctl_query(SYSCTLFN_CALL(rnode))); 1594 1595 if (name - oname != 4) 1596 return (EINVAL); 1597 1598 pf = oname[1]; 1599 proto = oname[2]; 1600 pf2 = (oldp != NULL) ? pf : 0; 1601 1602 mutex_enter(softnet_lock); 1603 1604 TAILQ_FOREACH(inph, &pcbtbl->inpt_queue, inph_queue) { 1605 #ifdef INET 1606 inp = (const struct inpcb *)inph; 1607 #endif 1608 #ifdef INET6 1609 in6p = (const struct in6pcb *)inph; 1610 #endif 1611 1612 if (inph->inph_af != pf) 1613 continue; 1614 1615 if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, 1616 KAUTH_REQ_NETWORK_SOCKET_CANSEE, inph->inph_socket, NULL, 1617 NULL) != 0) 1618 continue; 1619 1620 memset(&pcb, 0, sizeof(pcb)); 1621 1622 pcb.ki_family = pf; 1623 pcb.ki_type = proto; 1624 1625 switch (pf2) { 1626 case 0: 1627 /* just probing for size */ 1628 break; 1629 #ifdef INET 1630 case PF_INET: 1631 pcb.ki_family = inp->inp_socket->so_proto-> 1632 pr_domain->dom_family; 1633 pcb.ki_type = inp->inp_socket->so_proto-> 1634 pr_type; 1635 pcb.ki_protocol = inp->inp_socket->so_proto-> 1636 pr_protocol; 1637 pcb.ki_pflags = inp->inp_flags; 1638 1639 pcb.ki_sostate = inp->inp_socket->so_state; 1640 pcb.ki_prstate = inp->inp_state; 1641 if (proto == IPPROTO_TCP) { 1642 tp = intotcpcb(inp); 1643 pcb.ki_tstate = tp->t_state; 1644 pcb.ki_tflags = tp->t_flags; 1645 } 1646 1647 pcb.ki_pcbaddr = PTRTOUINT64(inp); 1648 pcb.ki_ppcbaddr = PTRTOUINT64(inp->inp_ppcb); 1649 pcb.ki_sockaddr = PTRTOUINT64(inp->inp_socket); 1650 1651 pcb.ki_rcvq = inp->inp_socket->so_rcv.sb_cc; 1652 pcb.ki_sndq = inp->inp_socket->so_snd.sb_cc; 1653 1654 in = satosin(&pcb.ki_src); 1655 in->sin_len = sizeof(*in); 1656 in->sin_family = pf; 1657 in->sin_port = inp->inp_lport; 1658 in->sin_addr = inp->inp_laddr; 1659 if (pcb.ki_prstate >= INP_CONNECTED) { 1660 in = satosin(&pcb.ki_dst); 1661 in->sin_len = sizeof(*in); 1662 in->sin_family = pf; 1663 in->sin_port = inp->inp_fport; 1664 in->sin_addr = inp->inp_faddr; 1665 } 1666 break; 1667 #endif 1668 #ifdef INET6 1669 case PF_INET6: 1670 pcb.ki_family = in6p->in6p_socket->so_proto-> 1671 pr_domain->dom_family; 1672 pcb.ki_type = in6p->in6p_socket->so_proto->pr_type; 1673 pcb.ki_protocol = in6p->in6p_socket->so_proto-> 1674 pr_protocol; 1675 pcb.ki_pflags = in6p->in6p_flags; 1676 1677 pcb.ki_sostate = in6p->in6p_socket->so_state; 1678 pcb.ki_prstate = in6p->in6p_state; 1679 if (proto == IPPROTO_TCP) { 1680 tp = in6totcpcb(in6p); 1681 pcb.ki_tstate = tp->t_state; 1682 pcb.ki_tflags = tp->t_flags; 1683 } 1684 1685 pcb.ki_pcbaddr = PTRTOUINT64(in6p); 1686 pcb.ki_ppcbaddr = PTRTOUINT64(in6p->in6p_ppcb); 1687 pcb.ki_sockaddr = PTRTOUINT64(in6p->in6p_socket); 1688 1689 pcb.ki_rcvq = in6p->in6p_socket->so_rcv.sb_cc; 1690 pcb.ki_sndq = in6p->in6p_socket->so_snd.sb_cc; 1691 1692 in6 = satosin6(&pcb.ki_src); 1693 in6->sin6_len = sizeof(*in6); 1694 in6->sin6_family = pf; 1695 in6->sin6_port = in6p->in6p_lport; 1696 in6->sin6_flowinfo = in6p->in6p_flowinfo; 1697 in6->sin6_addr = in6p->in6p_laddr; 1698 in6->sin6_scope_id = 0; /* XXX? */ 1699 1700 if (pcb.ki_prstate >= IN6P_CONNECTED) { 1701 in6 = satosin6(&pcb.ki_dst); 1702 in6->sin6_len = sizeof(*in6); 1703 in6->sin6_family = pf; 1704 in6->sin6_port = in6p->in6p_fport; 1705 in6->sin6_flowinfo = in6p->in6p_flowinfo; 1706 in6->sin6_addr = in6p->in6p_faddr; 1707 in6->sin6_scope_id = 0; /* XXX? */ 1708 } 1709 break; 1710 #endif 1711 } 1712 1713 if (len >= elem_size && elem_count > 0) { 1714 error = copyout(&pcb, dp, out_size); 1715 if (error) { 1716 mutex_exit(softnet_lock); 1717 return (error); 1718 } 1719 dp += elem_size; 1720 len -= elem_size; 1721 } 1722 needed += elem_size; 1723 if (elem_count > 0 && elem_count != INT_MAX) 1724 elem_count--; 1725 } 1726 1727 *oldlenp = needed; 1728 if (oldp == NULL) 1729 *oldlenp += PCB_SLOP * sizeof(struct kinfo_pcb); 1730 1731 mutex_exit(softnet_lock); 1732 1733 return (error); 1734 } 1735 1736 static int 1737 sysctl_tcp_congctl(SYSCTLFN_ARGS) 1738 { 1739 struct sysctlnode node; 1740 int error; 1741 char newname[TCPCC_MAXLEN]; 1742 1743 strlcpy(newname, tcp_congctl_global_name, sizeof(newname) - 1); 1744 1745 node = *rnode; 1746 node.sysctl_data = newname; 1747 node.sysctl_size = sizeof(newname); 1748 1749 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1750 1751 if (error || 1752 newp == NULL || 1753 strncmp(newname, tcp_congctl_global_name, sizeof(newname)) == 0) 1754 return error; 1755 1756 mutex_enter(softnet_lock); 1757 error = tcp_congctl_select(NULL, newname); 1758 mutex_exit(softnet_lock); 1759 1760 return error; 1761 } 1762 1763 static int 1764 sysctl_tcp_init_win(SYSCTLFN_ARGS) 1765 { 1766 int error; 1767 u_int iw; 1768 struct sysctlnode node; 1769 1770 iw = *(u_int *)rnode->sysctl_data; 1771 node = *rnode; 1772 node.sysctl_data = &iw; 1773 node.sysctl_size = sizeof(iw); 1774 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1775 if (error || newp == NULL) 1776 return error; 1777 1778 if (iw >= __arraycount(tcp_init_win_max)) 1779 return EINVAL; 1780 *(u_int *)rnode->sysctl_data = iw; 1781 return 0; 1782 } 1783 1784 static int 1785 sysctl_tcp_keep(SYSCTLFN_ARGS) 1786 { 1787 int error; 1788 u_int tmp; 1789 struct sysctlnode node; 1790 1791 node = *rnode; 1792 tmp = *(u_int *)rnode->sysctl_data; 1793 node.sysctl_data = &tmp; 1794 1795 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1796 if (error || newp == NULL) 1797 return error; 1798 1799 mutex_enter(softnet_lock); 1800 1801 *(u_int *)rnode->sysctl_data = tmp; 1802 tcp_tcpcb_template(); /* update the template */ 1803 1804 mutex_exit(softnet_lock); 1805 return 0; 1806 } 1807 1808 static int 1809 sysctl_net_inet_tcp_stats(SYSCTLFN_ARGS) 1810 { 1811 1812 return (NETSTAT_SYSCTL(tcpstat_percpu, TCP_NSTATS)); 1813 } 1814 1815 /* 1816 * this (second stage) setup routine is a replacement for tcp_sysctl() 1817 * (which is currently used for ipv4 and ipv6) 1818 */ 1819 static void 1820 sysctl_net_inet_tcp_setup2(struct sysctllog **clog, int pf, const char *pfname, 1821 const char *tcpname) 1822 { 1823 const struct sysctlnode *sack_node; 1824 const struct sysctlnode *abc_node; 1825 const struct sysctlnode *ecn_node; 1826 const struct sysctlnode *congctl_node; 1827 const struct sysctlnode *mslt_node; 1828 const struct sysctlnode *vtw_node; 1829 #ifdef TCP_DEBUG 1830 extern struct tcp_debug tcp_debug[TCP_NDEBUG]; 1831 extern int tcp_debx; 1832 #endif 1833 1834 sysctl_createv(clog, 0, NULL, NULL, 1835 CTLFLAG_PERMANENT, 1836 CTLTYPE_NODE, pfname, NULL, 1837 NULL, 0, NULL, 0, 1838 CTL_NET, pf, CTL_EOL); 1839 sysctl_createv(clog, 0, NULL, NULL, 1840 CTLFLAG_PERMANENT, 1841 CTLTYPE_NODE, tcpname, 1842 SYSCTL_DESCR("TCP related settings"), 1843 NULL, 0, NULL, 0, 1844 CTL_NET, pf, IPPROTO_TCP, CTL_EOL); 1845 1846 sysctl_createv(clog, 0, NULL, NULL, 1847 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1848 CTLTYPE_INT, "rfc1323", 1849 SYSCTL_DESCR("Enable RFC1323 TCP extensions"), 1850 sysctl_update_tcpcb_template, 0, &tcp_do_rfc1323, 0, 1851 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RFC1323, CTL_EOL); 1852 sysctl_createv(clog, 0, NULL, NULL, 1853 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1854 CTLTYPE_INT, "sendspace", 1855 SYSCTL_DESCR("Default TCP send buffer size"), 1856 NULL, 0, &tcp_sendspace, 0, 1857 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SENDSPACE, CTL_EOL); 1858 sysctl_createv(clog, 0, NULL, NULL, 1859 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1860 CTLTYPE_INT, "recvspace", 1861 SYSCTL_DESCR("Default TCP receive buffer size"), 1862 NULL, 0, &tcp_recvspace, 0, 1863 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RECVSPACE, CTL_EOL); 1864 sysctl_createv(clog, 0, NULL, NULL, 1865 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1866 CTLTYPE_INT, "mssdflt", 1867 SYSCTL_DESCR("Default maximum segment size"), 1868 sysctl_net_inet_tcp_mssdflt, 0, &tcp_mssdflt, 0, 1869 CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSSDFLT, CTL_EOL); 1870 sysctl_createv(clog, 0, NULL, NULL, 1871 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1872 CTLTYPE_INT, "minmss", 1873 SYSCTL_DESCR("Lower limit for TCP maximum segment size"), 1874 NULL, 0, &tcp_minmss, 0, 1875 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1876 sysctl_createv(clog, 0, NULL, NULL, 1877 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1878 CTLTYPE_INT, "msl", 1879 SYSCTL_DESCR("Maximum Segment Life"), 1880 NULL, 0, &tcp_msl, 0, 1881 CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSL, CTL_EOL); 1882 sysctl_createv(clog, 0, NULL, NULL, 1883 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1884 CTLTYPE_INT, "syn_cache_limit", 1885 SYSCTL_DESCR("Maximum number of entries in the TCP " 1886 "compressed state engine"), 1887 NULL, 0, &tcp_syn_cache_limit, 0, 1888 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_LIMIT, 1889 CTL_EOL); 1890 sysctl_createv(clog, 0, NULL, NULL, 1891 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1892 CTLTYPE_INT, "syn_bucket_limit", 1893 SYSCTL_DESCR("Maximum number of entries per hash " 1894 "bucket in the TCP compressed state " 1895 "engine"), 1896 NULL, 0, &tcp_syn_bucket_limit, 0, 1897 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_BUCKET_LIMIT, 1898 CTL_EOL); 1899 #if 0 /* obsoleted */ 1900 sysctl_createv(clog, 0, NULL, NULL, 1901 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1902 CTLTYPE_INT, "syn_cache_interval", 1903 SYSCTL_DESCR("TCP compressed state engine's timer interval"), 1904 NULL, 0, &tcp_syn_cache_interval, 0, 1905 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SYN_CACHE_INTER, 1906 CTL_EOL); 1907 #endif 1908 sysctl_createv(clog, 0, NULL, NULL, 1909 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1910 CTLTYPE_INT, "init_win", 1911 SYSCTL_DESCR("Initial TCP congestion window"), 1912 sysctl_tcp_init_win, 0, &tcp_init_win, 0, 1913 CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN, CTL_EOL); 1914 sysctl_createv(clog, 0, NULL, NULL, 1915 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1916 CTLTYPE_INT, "mss_ifmtu", 1917 SYSCTL_DESCR("Use interface MTU for calculating MSS"), 1918 NULL, 0, &tcp_mss_ifmtu, 0, 1919 CTL_NET, pf, IPPROTO_TCP, TCPCTL_MSS_IFMTU, CTL_EOL); 1920 sysctl_createv(clog, 0, NULL, &sack_node, 1921 CTLFLAG_PERMANENT, 1922 CTLTYPE_NODE, "sack", 1923 SYSCTL_DESCR("RFC2018 Selective ACKnowledgement tunables"), 1924 NULL, 0, NULL, 0, 1925 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_EOL); 1926 1927 /* Congctl subtree */ 1928 sysctl_createv(clog, 0, NULL, &congctl_node, 1929 CTLFLAG_PERMANENT, 1930 CTLTYPE_NODE, "congctl", 1931 SYSCTL_DESCR("TCP Congestion Control"), 1932 NULL, 0, NULL, 0, 1933 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 1934 sysctl_createv(clog, 0, &congctl_node, NULL, 1935 CTLFLAG_PERMANENT, 1936 CTLTYPE_STRING, "available", 1937 SYSCTL_DESCR("Available Congestion Control Mechanisms"), 1938 NULL, 0, tcp_congctl_avail, 0, CTL_CREATE, CTL_EOL); 1939 sysctl_createv(clog, 0, &congctl_node, NULL, 1940 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1941 CTLTYPE_STRING, "selected", 1942 SYSCTL_DESCR("Selected Congestion Control Mechanism"), 1943 sysctl_tcp_congctl, 0, NULL, TCPCC_MAXLEN, 1944 CTL_CREATE, CTL_EOL); 1945 1946 sysctl_createv(clog, 0, NULL, NULL, 1947 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1948 CTLTYPE_INT, "win_scale", 1949 SYSCTL_DESCR("Use RFC1323 window scale options"), 1950 sysctl_update_tcpcb_template, 0, &tcp_do_win_scale, 0, 1951 CTL_NET, pf, IPPROTO_TCP, TCPCTL_WSCALE, CTL_EOL); 1952 sysctl_createv(clog, 0, NULL, NULL, 1953 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1954 CTLTYPE_INT, "timestamps", 1955 SYSCTL_DESCR("Use RFC1323 time stamp options"), 1956 sysctl_update_tcpcb_template, 0, &tcp_do_timestamps, 0, 1957 CTL_NET, pf, IPPROTO_TCP, TCPCTL_TSTAMP, CTL_EOL); 1958 sysctl_createv(clog, 0, NULL, NULL, 1959 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1960 CTLTYPE_INT, "compat_42", 1961 SYSCTL_DESCR("Enable workarounds for 4.2BSD TCP bugs"), 1962 NULL, 0, &tcp_compat_42, 0, 1963 CTL_NET, pf, IPPROTO_TCP, TCPCTL_COMPAT_42, CTL_EOL); 1964 sysctl_createv(clog, 0, NULL, NULL, 1965 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1966 CTLTYPE_INT, "cwm", 1967 SYSCTL_DESCR("Hughes/Touch/Heidemann Congestion Window " 1968 "Monitoring"), 1969 NULL, 0, &tcp_cwm, 0, 1970 CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM, CTL_EOL); 1971 sysctl_createv(clog, 0, NULL, NULL, 1972 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1973 CTLTYPE_INT, "cwm_burstsize", 1974 SYSCTL_DESCR("Congestion Window Monitoring allowed " 1975 "burst count in packets"), 1976 NULL, 0, &tcp_cwm_burstsize, 0, 1977 CTL_NET, pf, IPPROTO_TCP, TCPCTL_CWM_BURSTSIZE, 1978 CTL_EOL); 1979 sysctl_createv(clog, 0, NULL, NULL, 1980 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1981 CTLTYPE_INT, "ack_on_push", 1982 SYSCTL_DESCR("Immediately return ACK when PSH is " 1983 "received"), 1984 NULL, 0, &tcp_ack_on_push, 0, 1985 CTL_NET, pf, IPPROTO_TCP, TCPCTL_ACK_ON_PUSH, CTL_EOL); 1986 sysctl_createv(clog, 0, NULL, NULL, 1987 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1988 CTLTYPE_INT, "keepidle", 1989 SYSCTL_DESCR("Allowed connection idle ticks before a " 1990 "keepalive probe is sent"), 1991 sysctl_tcp_keep, 0, &tcp_keepidle, 0, 1992 CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPIDLE, CTL_EOL); 1993 sysctl_createv(clog, 0, NULL, NULL, 1994 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1995 CTLTYPE_INT, "keepintvl", 1996 SYSCTL_DESCR("Ticks before next keepalive probe is sent"), 1997 sysctl_tcp_keep, 0, &tcp_keepintvl, 0, 1998 CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPINTVL, CTL_EOL); 1999 sysctl_createv(clog, 0, NULL, NULL, 2000 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2001 CTLTYPE_INT, "keepcnt", 2002 SYSCTL_DESCR("Number of keepalive probes to send"), 2003 sysctl_tcp_keep, 0, &tcp_keepcnt, 0, 2004 CTL_NET, pf, IPPROTO_TCP, TCPCTL_KEEPCNT, CTL_EOL); 2005 sysctl_createv(clog, 0, NULL, NULL, 2006 CTLFLAG_PERMANENT|CTLFLAG_IMMEDIATE, 2007 CTLTYPE_INT, "slowhz", 2008 SYSCTL_DESCR("Keepalive ticks per second"), 2009 NULL, PR_SLOWHZ, NULL, 0, 2010 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SLOWHZ, CTL_EOL); 2011 sysctl_createv(clog, 0, NULL, NULL, 2012 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2013 CTLTYPE_INT, "log_refused", 2014 SYSCTL_DESCR("Log refused TCP connections"), 2015 NULL, 0, &tcp_log_refused, 0, 2016 CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOG_REFUSED, CTL_EOL); 2017 #if 0 /* obsoleted */ 2018 sysctl_createv(clog, 0, NULL, NULL, 2019 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2020 CTLTYPE_INT, "rstratelimit", NULL, 2021 NULL, 0, &tcp_rst_ratelim, 0, 2022 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTRATELIMIT, CTL_EOL); 2023 #endif 2024 sysctl_createv(clog, 0, NULL, NULL, 2025 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2026 CTLTYPE_INT, "rstppslimit", 2027 SYSCTL_DESCR("Maximum number of RST packets to send " 2028 "per second"), 2029 NULL, 0, &tcp_rst_ppslim, 0, 2030 CTL_NET, pf, IPPROTO_TCP, TCPCTL_RSTPPSLIMIT, CTL_EOL); 2031 sysctl_createv(clog, 0, NULL, NULL, 2032 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2033 CTLTYPE_INT, "delack_ticks", 2034 SYSCTL_DESCR("Number of ticks to delay sending an ACK"), 2035 NULL, 0, &tcp_delack_ticks, 0, 2036 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DELACK_TICKS, CTL_EOL); 2037 sysctl_createv(clog, 0, NULL, NULL, 2038 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2039 CTLTYPE_INT, "init_win_local", 2040 SYSCTL_DESCR("Initial TCP window size (in segments)"), 2041 sysctl_tcp_init_win, 0, &tcp_init_win_local, 0, 2042 CTL_NET, pf, IPPROTO_TCP, TCPCTL_INIT_WIN_LOCAL, 2043 CTL_EOL); 2044 sysctl_createv(clog, 0, NULL, NULL, 2045 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2046 CTLTYPE_STRUCT, "ident", 2047 SYSCTL_DESCR("RFC1413 Identification Protocol lookups"), 2048 sysctl_net_inet_tcp_ident, 0, NULL, sizeof(uid_t), 2049 CTL_NET, pf, IPPROTO_TCP, TCPCTL_IDENT, CTL_EOL); 2050 sysctl_createv(clog, 0, NULL, NULL, 2051 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2052 CTLTYPE_INT, "do_loopback_cksum", 2053 SYSCTL_DESCR("Perform TCP checksum on loopback"), 2054 NULL, 0, &tcp_do_loopback_cksum, 0, 2055 CTL_NET, pf, IPPROTO_TCP, TCPCTL_LOOPBACKCKSUM, 2056 CTL_EOL); 2057 sysctl_createv(clog, 0, NULL, NULL, 2058 CTLFLAG_PERMANENT, 2059 CTLTYPE_STRUCT, "pcblist", 2060 SYSCTL_DESCR("TCP protocol control block list"), 2061 sysctl_inpcblist, 0, &tcbtable, 0, 2062 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, 2063 CTL_EOL); 2064 sysctl_createv(clog, 0, NULL, NULL, 2065 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2066 CTLTYPE_INT, "keepinit", 2067 SYSCTL_DESCR("Ticks before initial tcp connection times out"), 2068 sysctl_tcp_keep, 0, &tcp_keepinit, 0, 2069 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2070 2071 /* TCP socket buffers auto-sizing nodes */ 2072 sysctl_createv(clog, 0, NULL, NULL, 2073 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2074 CTLTYPE_INT, "recvbuf_auto", 2075 SYSCTL_DESCR("Enable automatic receive " 2076 "buffer sizing (experimental)"), 2077 NULL, 0, &tcp_do_autorcvbuf, 0, 2078 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2079 sysctl_createv(clog, 0, NULL, NULL, 2080 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2081 CTLTYPE_INT, "recvbuf_inc", 2082 SYSCTL_DESCR("Incrementor step size of " 2083 "automatic receive buffer"), 2084 NULL, 0, &tcp_autorcvbuf_inc, 0, 2085 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2086 sysctl_createv(clog, 0, NULL, NULL, 2087 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2088 CTLTYPE_INT, "recvbuf_max", 2089 SYSCTL_DESCR("Max size of automatic receive buffer"), 2090 NULL, 0, &tcp_autorcvbuf_max, 0, 2091 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2092 2093 sysctl_createv(clog, 0, NULL, NULL, 2094 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2095 CTLTYPE_INT, "sendbuf_auto", 2096 SYSCTL_DESCR("Enable automatic send " 2097 "buffer sizing (experimental)"), 2098 NULL, 0, &tcp_do_autosndbuf, 0, 2099 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2100 sysctl_createv(clog, 0, NULL, NULL, 2101 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2102 CTLTYPE_INT, "sendbuf_inc", 2103 SYSCTL_DESCR("Incrementor step size of " 2104 "automatic send buffer"), 2105 NULL, 0, &tcp_autosndbuf_inc, 0, 2106 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2107 sysctl_createv(clog, 0, NULL, NULL, 2108 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2109 CTLTYPE_INT, "sendbuf_max", 2110 SYSCTL_DESCR("Max size of automatic send buffer"), 2111 NULL, 0, &tcp_autosndbuf_max, 0, 2112 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2113 2114 /* ECN subtree */ 2115 sysctl_createv(clog, 0, NULL, &ecn_node, 2116 CTLFLAG_PERMANENT, 2117 CTLTYPE_NODE, "ecn", 2118 SYSCTL_DESCR("RFC3168 Explicit Congestion Notification"), 2119 NULL, 0, NULL, 0, 2120 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2121 sysctl_createv(clog, 0, &ecn_node, NULL, 2122 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2123 CTLTYPE_INT, "enable", 2124 SYSCTL_DESCR("Enable TCP Explicit Congestion " 2125 "Notification"), 2126 NULL, 0, &tcp_do_ecn, 0, CTL_CREATE, CTL_EOL); 2127 sysctl_createv(clog, 0, &ecn_node, NULL, 2128 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2129 CTLTYPE_INT, "maxretries", 2130 SYSCTL_DESCR("Number of times to retry ECN setup " 2131 "before disabling ECN on the connection"), 2132 NULL, 0, &tcp_ecn_maxretries, 0, CTL_CREATE, CTL_EOL); 2133 2134 /* SACK gets it's own little subtree. */ 2135 sysctl_createv(clog, 0, NULL, &sack_node, 2136 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2137 CTLTYPE_INT, "enable", 2138 SYSCTL_DESCR("Enable RFC2018 Selective ACKnowledgement"), 2139 NULL, 0, &tcp_do_sack, 0, 2140 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 2141 sysctl_createv(clog, 0, NULL, &sack_node, 2142 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2143 CTLTYPE_INT, "maxholes", 2144 SYSCTL_DESCR("Maximum number of TCP SACK holes allowed per connection"), 2145 NULL, 0, &tcp_sack_tp_maxholes, 0, 2146 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 2147 sysctl_createv(clog, 0, NULL, &sack_node, 2148 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2149 CTLTYPE_INT, "globalmaxholes", 2150 SYSCTL_DESCR("Global maximum number of TCP SACK holes"), 2151 NULL, 0, &tcp_sack_globalmaxholes, 0, 2152 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 2153 sysctl_createv(clog, 0, NULL, &sack_node, 2154 CTLFLAG_PERMANENT, 2155 CTLTYPE_INT, "globalholes", 2156 SYSCTL_DESCR("Global number of TCP SACK holes"), 2157 NULL, 0, &tcp_sack_globalholes, 0, 2158 CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_CREATE, CTL_EOL); 2159 2160 sysctl_createv(clog, 0, NULL, NULL, 2161 CTLFLAG_PERMANENT, 2162 CTLTYPE_STRUCT, "stats", 2163 SYSCTL_DESCR("TCP statistics"), 2164 sysctl_net_inet_tcp_stats, 0, NULL, 0, 2165 CTL_NET, pf, IPPROTO_TCP, TCPCTL_STATS, 2166 CTL_EOL); 2167 sysctl_createv(clog, 0, NULL, NULL, 2168 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2169 CTLTYPE_INT, "local_by_rtt", 2170 SYSCTL_DESCR("Use RTT estimator to decide which hosts " 2171 "are local"), 2172 NULL, 0, &tcp_rttlocal, 0, 2173 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2174 #ifdef TCP_DEBUG 2175 sysctl_createv(clog, 0, NULL, NULL, 2176 CTLFLAG_PERMANENT, 2177 CTLTYPE_STRUCT, "debug", 2178 SYSCTL_DESCR("TCP sockets debug information"), 2179 NULL, 0, &tcp_debug, sizeof(tcp_debug), 2180 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBUG, 2181 CTL_EOL); 2182 sysctl_createv(clog, 0, NULL, NULL, 2183 CTLFLAG_PERMANENT, 2184 CTLTYPE_INT, "debx", 2185 SYSCTL_DESCR("Number of TCP debug sockets messages"), 2186 NULL, 0, &tcp_debx, sizeof(tcp_debx), 2187 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DEBX, 2188 CTL_EOL); 2189 #endif 2190 sysctl_createv(clog, 0, NULL, NULL, 2191 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2192 CTLTYPE_STRUCT, "drop", 2193 SYSCTL_DESCR("TCP drop connection"), 2194 sysctl_net_inet_tcp_drop, 0, NULL, 0, 2195 CTL_NET, pf, IPPROTO_TCP, TCPCTL_DROP, CTL_EOL); 2196 sysctl_createv(clog, 0, NULL, NULL, 2197 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2198 CTLTYPE_INT, "iss_hash", 2199 SYSCTL_DESCR("Enable RFC 1948 ISS by cryptographic " 2200 "hash computation"), 2201 NULL, 0, &tcp_do_rfc1948, sizeof(tcp_do_rfc1948), 2202 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, 2203 CTL_EOL); 2204 2205 /* ABC subtree */ 2206 2207 sysctl_createv(clog, 0, NULL, &abc_node, 2208 CTLFLAG_PERMANENT, CTLTYPE_NODE, "abc", 2209 SYSCTL_DESCR("RFC3465 Appropriate Byte Counting (ABC)"), 2210 NULL, 0, NULL, 0, 2211 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2212 sysctl_createv(clog, 0, &abc_node, NULL, 2213 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2214 CTLTYPE_INT, "enable", 2215 SYSCTL_DESCR("Enable RFC3465 Appropriate Byte Counting"), 2216 NULL, 0, &tcp_do_abc, 0, CTL_CREATE, CTL_EOL); 2217 sysctl_createv(clog, 0, &abc_node, NULL, 2218 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2219 CTLTYPE_INT, "aggressive", 2220 SYSCTL_DESCR("1: L=2*SMSS 0: L=1*SMSS"), 2221 NULL, 0, &tcp_abc_aggressive, 0, CTL_CREATE, CTL_EOL); 2222 2223 /* MSL tuning subtree */ 2224 2225 sysctl_createv(clog, 0, NULL, &mslt_node, 2226 CTLFLAG_PERMANENT, CTLTYPE_NODE, "mslt", 2227 SYSCTL_DESCR("MSL Tuning for TIME_WAIT truncation"), 2228 NULL, 0, NULL, 0, 2229 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2230 sysctl_createv(clog, 0, &mslt_node, NULL, 2231 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2232 CTLTYPE_INT, "enable", 2233 SYSCTL_DESCR("Enable TIME_WAIT truncation"), 2234 NULL, 0, &tcp_msl_enable, 0, CTL_CREATE, CTL_EOL); 2235 sysctl_createv(clog, 0, &mslt_node, NULL, 2236 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2237 CTLTYPE_INT, "loopback", 2238 SYSCTL_DESCR("MSL value to use for loopback connections"), 2239 NULL, 0, &tcp_msl_loop, 0, CTL_CREATE, CTL_EOL); 2240 sysctl_createv(clog, 0, &mslt_node, NULL, 2241 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2242 CTLTYPE_INT, "local", 2243 SYSCTL_DESCR("MSL value to use for local connections"), 2244 NULL, 0, &tcp_msl_local, 0, CTL_CREATE, CTL_EOL); 2245 sysctl_createv(clog, 0, &mslt_node, NULL, 2246 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2247 CTLTYPE_INT, "remote", 2248 SYSCTL_DESCR("MSL value to use for remote connections"), 2249 NULL, 0, &tcp_msl_remote, 0, CTL_CREATE, CTL_EOL); 2250 sysctl_createv(clog, 0, &mslt_node, NULL, 2251 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2252 CTLTYPE_INT, "remote_threshold", 2253 SYSCTL_DESCR("RTT estimate value to promote local to remote"), 2254 NULL, 0, &tcp_msl_remote_threshold, 0, CTL_CREATE, CTL_EOL); 2255 2256 /* vestigial TIME_WAIT tuning subtree */ 2257 2258 sysctl_createv(clog, 0, NULL, &vtw_node, 2259 CTLFLAG_PERMANENT, CTLTYPE_NODE, "vtw", 2260 SYSCTL_DESCR("Tuning for Vestigial TIME_WAIT"), 2261 NULL, 0, NULL, 0, 2262 CTL_NET, pf, IPPROTO_TCP, CTL_CREATE, CTL_EOL); 2263 sysctl_createv(clog, 0, &vtw_node, NULL, 2264 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2265 CTLTYPE_INT, "enable", 2266 SYSCTL_DESCR("Enable Vestigial TIME_WAIT"), 2267 sysctl_tcp_vtw_enable, 0, 2268 (pf == AF_INET) ? &tcp4_vtw_enable : &tcp6_vtw_enable, 2269 0, CTL_CREATE, CTL_EOL); 2270 sysctl_createv(clog, 0, &vtw_node, NULL, 2271 CTLFLAG_PERMANENT|CTLFLAG_READONLY, 2272 CTLTYPE_INT, "entries", 2273 SYSCTL_DESCR("Maximum number of vestigial TIME_WAIT entries"), 2274 NULL, 0, &tcp_vtw_entries, 0, CTL_CREATE, CTL_EOL); 2275 } 2276 2277 void 2278 tcp_usrreq_init(void) 2279 { 2280 2281 #ifdef INET 2282 sysctl_net_inet_tcp_setup2(NULL, PF_INET, "inet", "tcp"); 2283 #endif 2284 #ifdef INET6 2285 sysctl_net_inet_tcp_setup2(NULL, PF_INET6, "inet6", "tcp6"); 2286 #endif 2287 } 2288 2289 PR_WRAP_USRREQS(tcp) 2290 #define tcp_attach tcp_attach_wrapper 2291 #define tcp_detach tcp_detach_wrapper 2292 #define tcp_ioctl tcp_ioctl_wrapper 2293 #define tcp_stat tcp_stat_wrapper 2294 #define tcp_peeraddr tcp_peeraddr_wrapper 2295 #define tcp_sockaddr tcp_sockaddr_wrapper 2296 #define tcp_usrreq tcp_usrreq_wrapper 2297 2298 const struct pr_usrreqs tcp_usrreqs = { 2299 .pr_attach = tcp_attach, 2300 .pr_detach = tcp_detach, 2301 .pr_ioctl = tcp_ioctl, 2302 .pr_stat = tcp_stat, 2303 .pr_peeraddr = tcp_peeraddr, 2304 .pr_sockaddr = tcp_sockaddr, 2305 .pr_generic = tcp_usrreq, 2306 }; 2307
Indexes created Sun Sep 21 16:09:51 GMT 2025