1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1988, 1993 5 * The Regents of the University of California. 6 * Copyright (c) 2006-2007 Robert N. M. Watson 7 * Copyright (c) 2010-2011 Juniper Networks, Inc. 8 * All rights reserved. 9 * 10 * Portions of this software were developed by Robert N. M. Watson under 11 * contract to Juniper Networks, Inc. 12 * 13 * Redistribution and use in source and binary forms, with or without 14 * modification, are permitted provided that the following conditions 15 * are met: 16 * 1. Redistributions of source code must retain the above copyright 17 * notice, this list of conditions and the following disclaimer. 18 * 2. Redistributions in binary form must reproduce the above copyright 19 * notice, this list of conditions and the following disclaimer in the 20 * documentation and/or other materials provided with the distribution. 21 * 3. Neither the name of the University nor the names of its contributors 22 * may be used to endorse or promote products derived from this software 23 * without specific prior written permission. 24 * 25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 35 * SUCH DAMAGE. 36 */ 37 38 #include <sys/cdefs.h> 39 #include "opt_ddb.h" 40 #include "opt_inet.h" 41 #include "opt_inet6.h" 42 #include "opt_ipsec.h" 43 #include "opt_kern_tls.h" 44 45 #include <sys/param.h> 46 #include <sys/systm.h> 47 #include <sys/arb.h> 48 #include <sys/limits.h> 49 #include <sys/malloc.h> 50 #include <sys/refcount.h> 51 #include <sys/kernel.h> 52 #include <sys/ktls.h> 53 #include <sys/qmath.h> 54 #include <sys/sysctl.h> 55 #include <sys/mbuf.h> 56 #ifdef INET6 57 #include <sys/domain.h> 58 #endif /* INET6 */ 59 #include <sys/socket.h> 60 #include <sys/socketvar.h> 61 #include <sys/protosw.h> 62 #include <sys/proc.h> 63 #include <sys/jail.h> 64 #include <sys/stats.h> 65 66 #ifdef DDB 67 #include <ddb/ddb.h> 68 #endif 69 70 #include <net/if.h> 71 #include <net/if_var.h> 72 #include <net/route.h> 73 #include <net/vnet.h> 74 75 #include <netinet/in.h> 76 #include <netinet/in_kdtrace.h> 77 #include <netinet/in_pcb.h> 78 #include <netinet/in_systm.h> 79 #include <netinet/in_var.h> 80 #include <netinet/ip.h> 81 #include <netinet/ip_var.h> 82 #ifdef INET6 83 #include <netinet/ip6.h> 84 #include <netinet6/in6_pcb.h> 85 #include <netinet6/ip6_var.h> 86 #include <netinet6/scope6_var.h> 87 #endif 88 #include <netinet/tcp.h> 89 #include <netinet/tcp_fsm.h> 90 #include <netinet/tcp_seq.h> 91 #include <netinet/tcp_timer.h> 92 #include <netinet/tcp_var.h> 93 #include <netinet/tcp_log_buf.h> 94 #include <netinet/tcpip.h> 95 #include <netinet/cc/cc.h> 96 #include <netinet/tcp_fastopen.h> 97 #include <netinet/tcp_hpts.h> 98 #ifdef TCPPCAP 99 #include <netinet/tcp_pcap.h> 100 #endif 101 #ifdef TCP_OFFLOAD 102 #include <netinet/tcp_offload.h> 103 #endif 104 #include <netipsec/ipsec_support.h> 105 106 #include <vm/vm.h> 107 #include <vm/vm_param.h> 108 #include <vm/pmap.h> 109 #include <vm/vm_extern.h> 110 #include <vm/vm_map.h> 111 #include <vm/vm_page.h> 112 113 /* 114 * TCP protocol interface to socket abstraction. 115 */ 116 #ifdef INET 117 static int tcp_connect(struct tcpcb *, struct sockaddr_in *, 118 struct thread *td); 119 #endif /* INET */ 120 #ifdef INET6 121 static int tcp6_connect(struct tcpcb *, struct sockaddr_in6 *, 122 struct thread *td); 123 #endif /* INET6 */ 124 static void tcp_disconnect(struct tcpcb *); 125 static void tcp_usrclosed(struct tcpcb *); 126 static void tcp_fill_info(const struct tcpcb *, struct tcp_info *); 127 128 static int tcp_pru_options_support(struct tcpcb *tp, int flags); 129 130 static void 131 tcp_bblog_pru(struct tcpcb *tp, uint32_t pru, int error) 132 { 133 struct tcp_log_buffer *lgb; 134 135 KASSERT(tp != NULL, ("tcp_bblog_pru: tp == NULL")); 136 INP_WLOCK_ASSERT(tptoinpcb(tp)); 137 if (tcp_bblogging_on(tp)) { 138 lgb = tcp_log_event(tp, NULL, NULL, NULL, TCP_LOG_PRU, error, 139 0, NULL, false, NULL, NULL, 0, NULL); 140 } else { 141 lgb = NULL; 142 } 143 if (lgb != NULL) { 144 if (error >= 0) { 145 lgb->tlb_errno = (uint32_t)error; 146 } 147 lgb->tlb_flex1 = pru; 148 } 149 } 150 151 /* 152 * TCP attaches to socket via pru_attach(), reserving space, 153 * and an internet control block. 154 */ 155 static int 156 tcp_usr_attach(struct socket *so, int proto, struct thread *td) 157 { 158 struct inpcb *inp; 159 struct tcpcb *tp = NULL; 160 int error; 161 162 inp = sotoinpcb(so); 163 KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL")); 164 165 error = soreserve(so, V_tcp_sendspace, V_tcp_recvspace); 166 if (error) 167 goto out; 168 169 so->so_rcv.sb_flags |= SB_AUTOSIZE; 170 so->so_snd.sb_flags |= SB_AUTOSIZE; 171 error = in_pcballoc(so, &V_tcbinfo); 172 if (error) 173 goto out; 174 inp = sotoinpcb(so); 175 tp = tcp_newtcpcb(inp, NULL); 176 if (tp == NULL) { 177 error = ENOBUFS; 178 in_pcbfree(inp); 179 goto out; 180 } 181 tp->t_state = TCPS_CLOSED; 182 tcp_bblog_pru(tp, PRU_ATTACH, error); 183 INP_WUNLOCK(inp); 184 TCPSTATES_INC(TCPS_CLOSED); 185 out: 186 TCP_PROBE2(debug__user, tp, PRU_ATTACH); 187 return (error); 188 } 189 190 /* 191 * tcp_usr_detach is called when the socket layer loses its final reference 192 * to the socket, be it a file descriptor reference, a reference from TCP, 193 * etc. At this point, there is only one case in which we will keep around 194 * inpcb state: time wait. 195 */ 196 static void 197 tcp_usr_detach(struct socket *so) 198 { 199 struct inpcb *inp; 200 struct tcpcb *tp; 201 202 inp = sotoinpcb(so); 203 KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); 204 INP_WLOCK(inp); 205 KASSERT(so->so_pcb == inp && inp->inp_socket == so, 206 ("%s: socket %p inp %p mismatch", __func__, so, inp)); 207 208 tp = intotcpcb(inp); 209 210 KASSERT(inp->inp_flags & INP_DROPPED || 211 tp->t_state < TCPS_SYN_SENT, 212 ("%s: inp %p not dropped or embryonic", __func__, inp)); 213 214 tcp_discardcb(tp); 215 in_pcbfree(inp); 216 } 217 218 #ifdef INET 219 /* 220 * Give the socket an address. 221 */ 222 static int 223 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 224 { 225 int error = 0; 226 struct inpcb *inp; 227 struct tcpcb *tp; 228 struct sockaddr_in *sinp; 229 230 inp = sotoinpcb(so); 231 KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL")); 232 INP_WLOCK(inp); 233 if (inp->inp_flags & INP_DROPPED) { 234 INP_WUNLOCK(inp); 235 return (EINVAL); 236 } 237 tp = intotcpcb(inp); 238 239 sinp = (struct sockaddr_in *)nam; 240 if (nam->sa_family != AF_INET) { 241 /* 242 * Preserve compatibility with old programs. 243 */ 244 if (nam->sa_family != AF_UNSPEC || 245 nam->sa_len < offsetof(struct sockaddr_in, sin_zero) || 246 sinp->sin_addr.s_addr != INADDR_ANY) { 247 error = EAFNOSUPPORT; 248 goto out; 249 } 250 nam->sa_family = AF_INET; 251 } 252 if (nam->sa_len != sizeof(*sinp)) { 253 error = EINVAL; 254 goto out; 255 } 256 /* 257 * Must check for multicast addresses and disallow binding 258 * to them. 259 */ 260 if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 261 error = EAFNOSUPPORT; 262 goto out; 263 } 264 INP_HASH_WLOCK(&V_tcbinfo); 265 error = in_pcbbind(inp, sinp, td->td_ucred); 266 INP_HASH_WUNLOCK(&V_tcbinfo); 267 out: 268 tcp_bblog_pru(tp, PRU_BIND, error); 269 TCP_PROBE2(debug__user, tp, PRU_BIND); 270 INP_WUNLOCK(inp); 271 272 return (error); 273 } 274 #endif /* INET */ 275 276 #ifdef INET6 277 static int 278 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 279 { 280 int error = 0; 281 struct inpcb *inp; 282 struct tcpcb *tp; 283 struct sockaddr_in6 *sin6; 284 u_char vflagsav; 285 286 inp = sotoinpcb(so); 287 KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL")); 288 INP_WLOCK(inp); 289 if (inp->inp_flags & INP_DROPPED) { 290 INP_WUNLOCK(inp); 291 return (EINVAL); 292 } 293 tp = intotcpcb(inp); 294 295 vflagsav = inp->inp_vflag; 296 297 sin6 = (struct sockaddr_in6 *)nam; 298 if (nam->sa_family != AF_INET6) { 299 error = EAFNOSUPPORT; 300 goto out; 301 } 302 if (nam->sa_len != sizeof(*sin6)) { 303 error = EINVAL; 304 goto out; 305 } 306 /* 307 * Must check for multicast addresses and disallow binding 308 * to them. 309 */ 310 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { 311 error = EAFNOSUPPORT; 312 goto out; 313 } 314 315 INP_HASH_WLOCK(&V_tcbinfo); 316 inp->inp_vflag &= ~INP_IPV4; 317 inp->inp_vflag |= INP_IPV6; 318 #ifdef INET 319 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { 320 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) 321 inp->inp_vflag |= INP_IPV4; 322 else if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { 323 struct sockaddr_in sin; 324 325 in6_sin6_2_sin(&sin, sin6); 326 if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) { 327 error = EAFNOSUPPORT; 328 INP_HASH_WUNLOCK(&V_tcbinfo); 329 goto out; 330 } 331 inp->inp_vflag |= INP_IPV4; 332 inp->inp_vflag &= ~INP_IPV6; 333 error = in_pcbbind(inp, &sin, td->td_ucred); 334 INP_HASH_WUNLOCK(&V_tcbinfo); 335 goto out; 336 } 337 } 338 #endif 339 error = in6_pcbbind(inp, sin6, td->td_ucred); 340 INP_HASH_WUNLOCK(&V_tcbinfo); 341 out: 342 if (error != 0) 343 inp->inp_vflag = vflagsav; 344 tcp_bblog_pru(tp, PRU_BIND, error); 345 TCP_PROBE2(debug__user, tp, PRU_BIND); 346 INP_WUNLOCK(inp); 347 return (error); 348 } 349 #endif /* INET6 */ 350 351 #ifdef INET 352 /* 353 * Prepare to accept connections. 354 */ 355 static int 356 tcp_usr_listen(struct socket *so, int backlog, struct thread *td) 357 { 358 int error = 0; 359 struct inpcb *inp; 360 struct tcpcb *tp; 361 362 inp = sotoinpcb(so); 363 KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL")); 364 INP_WLOCK(inp); 365 if (inp->inp_flags & INP_DROPPED) { 366 INP_WUNLOCK(inp); 367 return (EINVAL); 368 } 369 tp = intotcpcb(inp); 370 371 SOCK_LOCK(so); 372 error = solisten_proto_check(so); 373 if (error != 0) { 374 SOCK_UNLOCK(so); 375 goto out; 376 } 377 if (inp->inp_lport == 0) { 378 INP_HASH_WLOCK(&V_tcbinfo); 379 error = in_pcbbind(inp, NULL, td->td_ucred); 380 INP_HASH_WUNLOCK(&V_tcbinfo); 381 } 382 if (error == 0) { 383 tcp_state_change(tp, TCPS_LISTEN); 384 solisten_proto(so, backlog); 385 #ifdef TCP_OFFLOAD 386 if ((so->so_options & SO_NO_OFFLOAD) == 0) 387 tcp_offload_listen_start(tp); 388 #endif 389 } else { 390 solisten_proto_abort(so); 391 } 392 SOCK_UNLOCK(so); 393 394 if (tp->t_flags & TF_FASTOPEN) 395 tp->t_tfo_pending = tcp_fastopen_alloc_counter(); 396 397 out: 398 tcp_bblog_pru(tp, PRU_LISTEN, error); 399 TCP_PROBE2(debug__user, tp, PRU_LISTEN); 400 INP_WUNLOCK(inp); 401 return (error); 402 } 403 #endif /* INET */ 404 405 #ifdef INET6 406 static int 407 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td) 408 { 409 int error = 0; 410 struct inpcb *inp; 411 struct tcpcb *tp; 412 u_char vflagsav; 413 414 inp = sotoinpcb(so); 415 KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL")); 416 INP_WLOCK(inp); 417 if (inp->inp_flags & INP_DROPPED) { 418 INP_WUNLOCK(inp); 419 return (EINVAL); 420 } 421 tp = intotcpcb(inp); 422 423 vflagsav = inp->inp_vflag; 424 425 SOCK_LOCK(so); 426 error = solisten_proto_check(so); 427 if (error != 0) { 428 SOCK_UNLOCK(so); 429 goto out; 430 } 431 INP_HASH_WLOCK(&V_tcbinfo); 432 if (inp->inp_lport == 0) { 433 inp->inp_vflag &= ~INP_IPV4; 434 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) 435 inp->inp_vflag |= INP_IPV4; 436 error = in6_pcbbind(inp, NULL, td->td_ucred); 437 } 438 INP_HASH_WUNLOCK(&V_tcbinfo); 439 if (error == 0) { 440 tcp_state_change(tp, TCPS_LISTEN); 441 solisten_proto(so, backlog); 442 #ifdef TCP_OFFLOAD 443 if ((so->so_options & SO_NO_OFFLOAD) == 0) 444 tcp_offload_listen_start(tp); 445 #endif 446 } else { 447 solisten_proto_abort(so); 448 } 449 SOCK_UNLOCK(so); 450 451 if (tp->t_flags & TF_FASTOPEN) 452 tp->t_tfo_pending = tcp_fastopen_alloc_counter(); 453 454 if (error != 0) 455 inp->inp_vflag = vflagsav; 456 457 out: 458 tcp_bblog_pru(tp, PRU_LISTEN, error); 459 TCP_PROBE2(debug__user, tp, PRU_LISTEN); 460 INP_WUNLOCK(inp); 461 return (error); 462 } 463 #endif /* INET6 */ 464 465 #ifdef INET 466 /* 467 * Initiate connection to peer. 468 * Create a template for use in transmissions on this connection. 469 * Enter SYN_SENT state, and mark socket as connecting. 470 * Start keep-alive timer, and seed output sequence space. 471 * Send initial segment on connection. 472 */ 473 static int 474 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 475 { 476 struct epoch_tracker et; 477 int error = 0; 478 struct inpcb *inp; 479 struct tcpcb *tp; 480 struct sockaddr_in *sinp; 481 482 inp = sotoinpcb(so); 483 KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL")); 484 INP_WLOCK(inp); 485 if (inp->inp_flags & INP_DROPPED) { 486 INP_WUNLOCK(inp); 487 return (ECONNREFUSED); 488 } 489 tp = intotcpcb(inp); 490 491 sinp = (struct sockaddr_in *)nam; 492 if (nam->sa_family != AF_INET) { 493 error = EAFNOSUPPORT; 494 goto out; 495 } 496 if (nam->sa_len != sizeof (*sinp)) { 497 error = EINVAL; 498 goto out; 499 } 500 /* 501 * Must disallow TCP ``connections'' to multicast addresses. 502 */ 503 if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 504 error = EAFNOSUPPORT; 505 goto out; 506 } 507 if (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST) { 508 error = EACCES; 509 goto out; 510 } 511 if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0) 512 goto out; 513 if (SOLISTENING(so)) { 514 error = EOPNOTSUPP; 515 goto out; 516 } 517 NET_EPOCH_ENTER(et); 518 if ((error = tcp_connect(tp, sinp, td)) != 0) 519 goto out_in_epoch; 520 #ifdef TCP_OFFLOAD 521 if (registered_toedevs > 0 && 522 (so->so_options & SO_NO_OFFLOAD) == 0 && 523 (error = tcp_offload_connect(so, nam)) == 0) 524 goto out_in_epoch; 525 #endif 526 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp)); 527 error = tcp_output(tp); 528 KASSERT(error >= 0, ("TCP stack %s requested tcp_drop(%p) at connect()" 529 ", error code %d", tp->t_fb->tfb_tcp_block_name, tp, -error)); 530 out_in_epoch: 531 NET_EPOCH_EXIT(et); 532 out: 533 tcp_bblog_pru(tp, PRU_CONNECT, error); 534 TCP_PROBE2(debug__user, tp, PRU_CONNECT); 535 INP_WUNLOCK(inp); 536 return (error); 537 } 538 #endif /* INET */ 539 540 #ifdef INET6 541 static int 542 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 543 { 544 struct epoch_tracker et; 545 int error = 0; 546 struct inpcb *inp; 547 struct tcpcb *tp; 548 struct sockaddr_in6 *sin6; 549 u_int8_t incflagsav; 550 u_char vflagsav; 551 552 inp = sotoinpcb(so); 553 KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL")); 554 INP_WLOCK(inp); 555 if (inp->inp_flags & INP_DROPPED) { 556 INP_WUNLOCK(inp); 557 return (ECONNREFUSED); 558 } 559 tp = intotcpcb(inp); 560 561 vflagsav = inp->inp_vflag; 562 incflagsav = inp->inp_inc.inc_flags; 563 564 sin6 = (struct sockaddr_in6 *)nam; 565 if (nam->sa_family != AF_INET6) { 566 error = EAFNOSUPPORT; 567 goto out; 568 } 569 if (nam->sa_len != sizeof (*sin6)) { 570 error = EINVAL; 571 goto out; 572 } 573 /* 574 * Must disallow TCP ``connections'' to multicast addresses. 575 */ 576 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { 577 error = EAFNOSUPPORT; 578 goto out; 579 } 580 if (SOLISTENING(so)) { 581 error = EINVAL; 582 goto out; 583 } 584 #ifdef INET 585 /* 586 * XXXRW: Some confusion: V4/V6 flags relate to binding, and 587 * therefore probably require the hash lock, which isn't held here. 588 * Is this a significant problem? 589 */ 590 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { 591 struct sockaddr_in sin; 592 593 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { 594 error = EINVAL; 595 goto out; 596 } 597 if ((inp->inp_vflag & INP_IPV4) == 0) { 598 error = EAFNOSUPPORT; 599 goto out; 600 } 601 602 in6_sin6_2_sin(&sin, sin6); 603 if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) { 604 error = EAFNOSUPPORT; 605 goto out; 606 } 607 if (ntohl(sin.sin_addr.s_addr) == INADDR_BROADCAST) { 608 error = EACCES; 609 goto out; 610 } 611 if ((error = prison_remote_ip4(td->td_ucred, 612 &sin.sin_addr)) != 0) 613 goto out; 614 inp->inp_vflag |= INP_IPV4; 615 inp->inp_vflag &= ~INP_IPV6; 616 NET_EPOCH_ENTER(et); 617 if ((error = tcp_connect(tp, &sin, td)) != 0) 618 goto out_in_epoch; 619 #ifdef TCP_OFFLOAD 620 if (registered_toedevs > 0 && 621 (so->so_options & SO_NO_OFFLOAD) == 0 && 622 (error = tcp_offload_connect(so, nam)) == 0) 623 goto out_in_epoch; 624 #endif 625 error = tcp_output(tp); 626 goto out_in_epoch; 627 } else { 628 if ((inp->inp_vflag & INP_IPV6) == 0) { 629 error = EAFNOSUPPORT; 630 goto out; 631 } 632 } 633 #endif 634 if ((error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr)) != 0) 635 goto out; 636 inp->inp_vflag &= ~INP_IPV4; 637 inp->inp_vflag |= INP_IPV6; 638 inp->inp_inc.inc_flags |= INC_ISIPV6; 639 NET_EPOCH_ENTER(et); 640 if ((error = tcp6_connect(tp, sin6, td)) != 0) 641 goto out_in_epoch; 642 #ifdef TCP_OFFLOAD 643 if (registered_toedevs > 0 && 644 (so->so_options & SO_NO_OFFLOAD) == 0 && 645 (error = tcp_offload_connect(so, nam)) == 0) 646 goto out_in_epoch; 647 #endif 648 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp)); 649 error = tcp_output(tp); 650 out_in_epoch: 651 NET_EPOCH_EXIT(et); 652 out: 653 KASSERT(error >= 0, ("TCP stack %s requested tcp_drop(%p) at connect()" 654 ", error code %d", tp->t_fb->tfb_tcp_block_name, tp, -error)); 655 /* 656 * If the implicit bind in the connect call fails, restore 657 * the flags we modified. 658 */ 659 if (error != 0 && inp->inp_lport == 0) { 660 inp->inp_vflag = vflagsav; 661 inp->inp_inc.inc_flags = incflagsav; 662 } 663 664 tcp_bblog_pru(tp, PRU_CONNECT, error); 665 TCP_PROBE2(debug__user, tp, PRU_CONNECT); 666 INP_WUNLOCK(inp); 667 return (error); 668 } 669 #endif /* INET6 */ 670 671 /* 672 * Initiate disconnect from peer. 673 * If connection never passed embryonic stage, just drop; 674 * else if don't need to let data drain, then can just drop anyways, 675 * else have to begin TCP shutdown process: mark socket disconnecting, 676 * drain unread data, state switch to reflect user close, and 677 * send segment (e.g. FIN) to peer. Socket will be really disconnected 678 * when peer sends FIN and acks ours. 679 * 680 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 681 */ 682 static int 683 tcp_usr_disconnect(struct socket *so) 684 { 685 struct inpcb *inp; 686 struct tcpcb *tp = NULL; 687 struct epoch_tracker et; 688 689 NET_EPOCH_ENTER(et); 690 inp = sotoinpcb(so); 691 KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL")); 692 INP_WLOCK(inp); 693 tp = intotcpcb(inp); 694 695 if (tp->t_state == TCPS_TIME_WAIT) 696 goto out; 697 tcp_disconnect(tp); 698 out: 699 tcp_bblog_pru(tp, PRU_DISCONNECT, 0); 700 TCP_PROBE2(debug__user, tp, PRU_DISCONNECT); 701 INP_WUNLOCK(inp); 702 NET_EPOCH_EXIT(et); 703 return (0); 704 } 705 706 #ifdef INET 707 /* 708 * Accept a connection. Essentially all the work is done at higher levels; 709 * just return the address of the peer, storing through addr. 710 */ 711 static int 712 tcp_usr_accept(struct socket *so, struct sockaddr *sa) 713 { 714 struct inpcb *inp; 715 struct tcpcb *tp; 716 int error = 0; 717 718 inp = sotoinpcb(so); 719 KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL")); 720 INP_WLOCK(inp); 721 if (inp->inp_flags & INP_DROPPED) { 722 INP_WUNLOCK(inp); 723 return (ECONNABORTED); 724 } 725 tp = intotcpcb(inp); 726 727 if (so->so_state & SS_ISDISCONNECTED) 728 error = ECONNABORTED; 729 else 730 *(struct sockaddr_in *)sa = (struct sockaddr_in ){ 731 .sin_family = AF_INET, 732 .sin_len = sizeof(struct sockaddr_in), 733 .sin_port = inp->inp_fport, 734 .sin_addr = inp->inp_faddr, 735 }; 736 tcp_bblog_pru(tp, PRU_ACCEPT, error); 737 TCP_PROBE2(debug__user, tp, PRU_ACCEPT); 738 INP_WUNLOCK(inp); 739 740 return (error); 741 } 742 #endif /* INET */ 743 744 #ifdef INET6 745 static int 746 tcp6_usr_accept(struct socket *so, struct sockaddr *sa) 747 { 748 struct inpcb *inp; 749 struct tcpcb *tp; 750 int error = 0; 751 752 inp = sotoinpcb(so); 753 KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL")); 754 INP_WLOCK(inp); 755 if (inp->inp_flags & INP_DROPPED) { 756 INP_WUNLOCK(inp); 757 return (ECONNABORTED); 758 } 759 tp = intotcpcb(inp); 760 761 if (so->so_state & SS_ISDISCONNECTED) { 762 error = ECONNABORTED; 763 } else { 764 if (inp->inp_vflag & INP_IPV4) { 765 struct sockaddr_in sin = { 766 .sin_family = AF_INET, 767 .sin_len = sizeof(struct sockaddr_in), 768 .sin_port = inp->inp_fport, 769 .sin_addr = inp->inp_faddr, 770 }; 771 in6_sin_2_v4mapsin6(&sin, (struct sockaddr_in6 *)sa); 772 } else { 773 *(struct sockaddr_in6 *)sa = (struct sockaddr_in6 ){ 774 .sin6_family = AF_INET6, 775 .sin6_len = sizeof(struct sockaddr_in6), 776 .sin6_port = inp->inp_fport, 777 .sin6_addr = inp->in6p_faddr, 778 }; 779 /* XXX: should catch errors */ 780 (void)sa6_recoverscope((struct sockaddr_in6 *)sa); 781 } 782 } 783 784 tcp_bblog_pru(tp, PRU_ACCEPT, error); 785 TCP_PROBE2(debug__user, tp, PRU_ACCEPT); 786 INP_WUNLOCK(inp); 787 788 return (error); 789 } 790 #endif /* INET6 */ 791 792 /* 793 * Mark the connection as being incapable of further output. 794 */ 795 static int 796 tcp_usr_shutdown(struct socket *so, enum shutdown_how how) 797 { 798 struct epoch_tracker et; 799 struct inpcb *inp = sotoinpcb(so); 800 struct tcpcb *tp = intotcpcb(inp); 801 int error = 0; 802 803 SOCK_LOCK(so); 804 if (SOLISTENING(so)) { 805 if (how != SHUT_WR) { 806 so->so_error = ECONNABORTED; 807 solisten_wakeup(so); /* unlocks so */ 808 } else 809 SOCK_UNLOCK(so); 810 return (ENOTCONN); 811 } else if ((so->so_state & 812 (SS_ISCONNECTED | SS_ISCONNECTING | SS_ISDISCONNECTING)) == 0) { 813 SOCK_UNLOCK(so); 814 return (ENOTCONN); 815 } 816 SOCK_UNLOCK(so); 817 818 switch (how) { 819 case SHUT_RD: 820 sorflush(so); 821 break; 822 case SHUT_RDWR: 823 sorflush(so); 824 /* FALLTHROUGH */ 825 case SHUT_WR: 826 /* 827 * XXXGL: mimicing old soshutdown() here. But shouldn't we 828 * return ECONNRESEST for SHUT_RD as well? 829 */ 830 INP_WLOCK(inp); 831 if (inp->inp_flags & INP_DROPPED) { 832 INP_WUNLOCK(inp); 833 return (ECONNRESET); 834 } 835 836 socantsendmore(so); 837 NET_EPOCH_ENTER(et); 838 tcp_usrclosed(tp); 839 error = tcp_output_nodrop(tp); 840 tcp_bblog_pru(tp, PRU_SHUTDOWN, error); 841 TCP_PROBE2(debug__user, tp, PRU_SHUTDOWN); 842 error = tcp_unlock_or_drop(tp, error); 843 NET_EPOCH_EXIT(et); 844 } 845 wakeup(&so->so_timeo); 846 847 return (error); 848 } 849 850 /* 851 * After a receive, possibly send window update to peer. 852 */ 853 static int 854 tcp_usr_rcvd(struct socket *so, int flags) 855 { 856 struct epoch_tracker et; 857 struct inpcb *inp; 858 struct tcpcb *tp; 859 int outrv = 0, error = 0; 860 861 inp = sotoinpcb(so); 862 KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL")); 863 INP_WLOCK(inp); 864 if (inp->inp_flags & INP_DROPPED) { 865 INP_WUNLOCK(inp); 866 return (ECONNRESET); 867 } 868 tp = intotcpcb(inp); 869 870 NET_EPOCH_ENTER(et); 871 /* 872 * For passively-created TFO connections, don't attempt a window 873 * update while still in SYN_RECEIVED as this may trigger an early 874 * SYN|ACK. It is preferable to have the SYN|ACK be sent along with 875 * application response data, or failing that, when the DELACK timer 876 * expires. 877 */ 878 if ((tp->t_flags & TF_FASTOPEN) && (tp->t_state == TCPS_SYN_RECEIVED)) 879 goto out; 880 #ifdef TCP_OFFLOAD 881 if (tp->t_flags & TF_TOE) 882 tcp_offload_rcvd(tp); 883 else 884 #endif 885 outrv = tcp_output_nodrop(tp); 886 out: 887 tcp_bblog_pru(tp, PRU_RCVD, error); 888 TCP_PROBE2(debug__user, tp, PRU_RCVD); 889 (void) tcp_unlock_or_drop(tp, outrv); 890 NET_EPOCH_EXIT(et); 891 return (error); 892 } 893 894 /* 895 * Do a send by putting data in output queue and updating urgent 896 * marker if URG set. Possibly send more data. Unlike the other 897 * pru_*() routines, the mbuf chains are our responsibility. We 898 * must either enqueue them or free them. The other pru_* routines 899 * generally are caller-frees. 900 */ 901 static int 902 tcp_usr_send(struct socket *so, int flags, struct mbuf *m, 903 struct sockaddr *nam, struct mbuf *control, struct thread *td) 904 { 905 struct epoch_tracker et; 906 int error = 0; 907 struct inpcb *inp; 908 struct tcpcb *tp; 909 #ifdef INET 910 #ifdef INET6 911 struct sockaddr_in sin; 912 #endif 913 struct sockaddr_in *sinp; 914 #endif 915 #ifdef INET6 916 struct sockaddr_in6 *sin6; 917 int isipv6; 918 #endif 919 u_int8_t incflagsav; 920 u_char vflagsav; 921 bool restoreflags; 922 923 inp = sotoinpcb(so); 924 KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL")); 925 INP_WLOCK(inp); 926 if (inp->inp_flags & INP_DROPPED) { 927 if (m != NULL && (flags & PRUS_NOTREADY) == 0) 928 m_freem(m); 929 INP_WUNLOCK(inp); 930 return (ECONNRESET); 931 } 932 tp = intotcpcb(inp); 933 934 vflagsav = inp->inp_vflag; 935 incflagsav = inp->inp_inc.inc_flags; 936 restoreflags = false; 937 938 NET_EPOCH_ENTER(et); 939 if (control != NULL) { 940 /* TCP doesn't do control messages (rights, creds, etc) */ 941 if (control->m_len > 0) { 942 m_freem(control); 943 error = EINVAL; 944 goto out; 945 } 946 m_freem(control); /* empty control, just free it */ 947 } 948 949 if ((flags & PRUS_OOB) != 0 && 950 (error = tcp_pru_options_support(tp, PRUS_OOB)) != 0) 951 goto out; 952 953 if (nam != NULL && tp->t_state < TCPS_SYN_SENT) { 954 if (tp->t_state == TCPS_LISTEN) { 955 error = EINVAL; 956 goto out; 957 } 958 switch (nam->sa_family) { 959 #ifdef INET 960 case AF_INET: 961 sinp = (struct sockaddr_in *)nam; 962 if (sinp->sin_len != sizeof(struct sockaddr_in)) { 963 error = EINVAL; 964 goto out; 965 } 966 if ((inp->inp_vflag & INP_IPV6) != 0) { 967 error = EAFNOSUPPORT; 968 goto out; 969 } 970 if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 971 error = EAFNOSUPPORT; 972 goto out; 973 } 974 if (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST) { 975 error = EACCES; 976 goto out; 977 } 978 if ((error = prison_remote_ip4(td->td_ucred, 979 &sinp->sin_addr))) 980 goto out; 981 #ifdef INET6 982 isipv6 = 0; 983 #endif 984 break; 985 #endif /* INET */ 986 #ifdef INET6 987 case AF_INET6: 988 sin6 = (struct sockaddr_in6 *)nam; 989 if (sin6->sin6_len != sizeof(*sin6)) { 990 error = EINVAL; 991 goto out; 992 } 993 if ((inp->inp_vflag & INP_IPV6PROTO) == 0) { 994 error = EAFNOSUPPORT; 995 goto out; 996 } 997 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { 998 error = EAFNOSUPPORT; 999 goto out; 1000 } 1001 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { 1002 #ifdef INET 1003 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { 1004 error = EINVAL; 1005 goto out; 1006 } 1007 if ((inp->inp_vflag & INP_IPV4) == 0) { 1008 error = EAFNOSUPPORT; 1009 goto out; 1010 } 1011 restoreflags = true; 1012 inp->inp_vflag &= ~INP_IPV6; 1013 sinp = &sin; 1014 in6_sin6_2_sin(sinp, sin6); 1015 if (IN_MULTICAST( 1016 ntohl(sinp->sin_addr.s_addr))) { 1017 error = EAFNOSUPPORT; 1018 goto out; 1019 } 1020 if ((error = prison_remote_ip4(td->td_ucred, 1021 &sinp->sin_addr))) 1022 goto out; 1023 isipv6 = 0; 1024 #else /* !INET */ 1025 error = EAFNOSUPPORT; 1026 goto out; 1027 #endif /* INET */ 1028 } else { 1029 if ((inp->inp_vflag & INP_IPV6) == 0) { 1030 error = EAFNOSUPPORT; 1031 goto out; 1032 } 1033 restoreflags = true; 1034 inp->inp_vflag &= ~INP_IPV4; 1035 inp->inp_inc.inc_flags |= INC_ISIPV6; 1036 if ((error = prison_remote_ip6(td->td_ucred, 1037 &sin6->sin6_addr))) 1038 goto out; 1039 isipv6 = 1; 1040 } 1041 break; 1042 #endif /* INET6 */ 1043 default: 1044 error = EAFNOSUPPORT; 1045 goto out; 1046 } 1047 } 1048 if (!(flags & PRUS_OOB)) { 1049 if (tp->t_acktime == 0) 1050 tp->t_acktime = ticks; 1051 sbappendstream(&so->so_snd, m, flags); 1052 m = NULL; 1053 if (nam && tp->t_state < TCPS_SYN_SENT) { 1054 KASSERT(tp->t_state == TCPS_CLOSED, 1055 ("%s: tp %p is listening", __func__, tp)); 1056 1057 /* 1058 * Do implied connect if not yet connected, 1059 * initialize window to default value, and 1060 * initialize maxseg using peer's cached MSS. 1061 */ 1062 #ifdef INET6 1063 if (isipv6) 1064 error = tcp6_connect(tp, sin6, td); 1065 #endif /* INET6 */ 1066 #if defined(INET6) && defined(INET) 1067 else 1068 #endif 1069 #ifdef INET 1070 error = tcp_connect(tp, sinp, td); 1071 #endif 1072 /* 1073 * The bind operation in tcp_connect succeeded. We 1074 * no longer want to restore the flags if later 1075 * operations fail. 1076 */ 1077 if (error == 0 || inp->inp_lport != 0) 1078 restoreflags = false; 1079 1080 if (error) { 1081 /* m is freed if PRUS_NOTREADY is unset. */ 1082 sbflush(&so->so_snd); 1083 goto out; 1084 } 1085 if (tp->t_flags & TF_FASTOPEN) 1086 tcp_fastopen_connect(tp); 1087 else { 1088 tp->snd_wnd = TTCP_CLIENT_SND_WND; 1089 tcp_mss(tp, -1); 1090 } 1091 } 1092 if (flags & PRUS_EOF) { 1093 /* 1094 * Close the send side of the connection after 1095 * the data is sent. 1096 */ 1097 socantsendmore(so); 1098 tcp_usrclosed(tp); 1099 } 1100 if (TCPS_HAVEESTABLISHED(tp->t_state) && 1101 ((tp->t_flags2 & TF2_FBYTES_COMPLETE) == 0) && 1102 (tp->t_fbyte_out == 0) && 1103 (so->so_snd.sb_ccc > 0)) { 1104 tp->t_fbyte_out = ticks; 1105 if (tp->t_fbyte_out == 0) 1106 tp->t_fbyte_out = 1; 1107 if (tp->t_fbyte_out && tp->t_fbyte_in) 1108 tp->t_flags2 |= TF2_FBYTES_COMPLETE; 1109 } 1110 if (!(inp->inp_flags & INP_DROPPED) && 1111 !(flags & PRUS_NOTREADY)) { 1112 if (flags & PRUS_MORETOCOME) 1113 tp->t_flags |= TF_MORETOCOME; 1114 error = tcp_output_nodrop(tp); 1115 if (flags & PRUS_MORETOCOME) 1116 tp->t_flags &= ~TF_MORETOCOME; 1117 } 1118 } else { 1119 /* 1120 * XXXRW: PRUS_EOF not implemented with PRUS_OOB? 1121 */ 1122 SOCK_SENDBUF_LOCK(so); 1123 if (sbspace(&so->so_snd) < -512) { 1124 SOCK_SENDBUF_UNLOCK(so); 1125 error = ENOBUFS; 1126 goto out; 1127 } 1128 /* 1129 * According to RFC961 (Assigned Protocols), 1130 * the urgent pointer points to the last octet 1131 * of urgent data. We continue, however, 1132 * to consider it to indicate the first octet 1133 * of data past the urgent section. 1134 * Otherwise, snd_up should be one lower. 1135 */ 1136 if (tp->t_acktime == 0) 1137 tp->t_acktime = ticks; 1138 sbappendstream_locked(&so->so_snd, m, flags); 1139 SOCK_SENDBUF_UNLOCK(so); 1140 m = NULL; 1141 if (nam && tp->t_state < TCPS_SYN_SENT) { 1142 /* 1143 * Do implied connect if not yet connected, 1144 * initialize window to default value, and 1145 * initialize maxseg using peer's cached MSS. 1146 */ 1147 1148 /* 1149 * Not going to contemplate SYN|URG 1150 */ 1151 if (tp->t_flags & TF_FASTOPEN) 1152 tp->t_flags &= ~TF_FASTOPEN; 1153 #ifdef INET6 1154 if (isipv6) 1155 error = tcp6_connect(tp, sin6, td); 1156 #endif /* INET6 */ 1157 #if defined(INET6) && defined(INET) 1158 else 1159 #endif 1160 #ifdef INET 1161 error = tcp_connect(tp, sinp, td); 1162 #endif 1163 /* 1164 * The bind operation in tcp_connect succeeded. We 1165 * no longer want to restore the flags if later 1166 * operations fail. 1167 */ 1168 if (error == 0 || inp->inp_lport != 0) 1169 restoreflags = false; 1170 1171 if (error != 0) { 1172 /* m is freed if PRUS_NOTREADY is unset. */ 1173 sbflush(&so->so_snd); 1174 goto out; 1175 } 1176 tp->snd_wnd = TTCP_CLIENT_SND_WND; 1177 tcp_mss(tp, -1); 1178 } 1179 tp->snd_up = tp->snd_una + sbavail(&so->so_snd); 1180 if ((flags & PRUS_NOTREADY) == 0) { 1181 tp->t_flags |= TF_FORCEDATA; 1182 error = tcp_output_nodrop(tp); 1183 tp->t_flags &= ~TF_FORCEDATA; 1184 } 1185 } 1186 TCP_LOG_EVENT(tp, NULL, 1187 &inp->inp_socket->so_rcv, 1188 &inp->inp_socket->so_snd, 1189 TCP_LOG_USERSEND, error, 1190 0, NULL, false); 1191 1192 out: 1193 /* 1194 * In case of PRUS_NOTREADY, the caller or tcp_usr_ready() is 1195 * responsible for freeing memory. 1196 */ 1197 if (m != NULL && (flags & PRUS_NOTREADY) == 0) 1198 m_freem(m); 1199 1200 /* 1201 * If the request was unsuccessful and we changed flags, 1202 * restore the original flags. 1203 */ 1204 if (error != 0 && restoreflags) { 1205 inp->inp_vflag = vflagsav; 1206 inp->inp_inc.inc_flags = incflagsav; 1207 } 1208 tcp_bblog_pru(tp, (flags & PRUS_OOB) ? PRU_SENDOOB : 1209 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND), error); 1210 TCP_PROBE2(debug__user, tp, (flags & PRUS_OOB) ? PRU_SENDOOB : 1211 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); 1212 error = tcp_unlock_or_drop(tp, error); 1213 NET_EPOCH_EXIT(et); 1214 return (error); 1215 } 1216 1217 static int 1218 tcp_usr_ready(struct socket *so, struct mbuf *m, int count) 1219 { 1220 struct epoch_tracker et; 1221 struct inpcb *inp; 1222 struct tcpcb *tp; 1223 int error; 1224 1225 inp = sotoinpcb(so); 1226 INP_WLOCK(inp); 1227 if (inp->inp_flags & INP_DROPPED) { 1228 INP_WUNLOCK(inp); 1229 mb_free_notready(m, count); 1230 return (ECONNRESET); 1231 } 1232 tp = intotcpcb(inp); 1233 1234 SOCK_SENDBUF_LOCK(so); 1235 error = sbready(&so->so_snd, m, count); 1236 SOCK_SENDBUF_UNLOCK(so); 1237 if (error) { 1238 INP_WUNLOCK(inp); 1239 return (error); 1240 } 1241 NET_EPOCH_ENTER(et); 1242 error = tcp_output_unlock(tp); 1243 NET_EPOCH_EXIT(et); 1244 1245 return (error); 1246 } 1247 1248 /* 1249 * Abort the TCP. Drop the connection abruptly. 1250 */ 1251 static void 1252 tcp_usr_abort(struct socket *so) 1253 { 1254 struct inpcb *inp; 1255 struct tcpcb *tp; 1256 struct epoch_tracker et; 1257 1258 inp = sotoinpcb(so); 1259 KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL")); 1260 1261 NET_EPOCH_ENTER(et); 1262 INP_WLOCK(inp); 1263 KASSERT(inp->inp_socket != NULL, 1264 ("tcp_usr_abort: inp_socket == NULL")); 1265 1266 /* 1267 * If we still have full TCP state, and we're not dropped, drop. 1268 */ 1269 if (!(inp->inp_flags & INP_DROPPED)) { 1270 tp = intotcpcb(inp); 1271 tp = tcp_drop(tp, ECONNABORTED); 1272 if (tp == NULL) 1273 goto dropped; 1274 tcp_bblog_pru(tp, PRU_ABORT, 0); 1275 TCP_PROBE2(debug__user, tp, PRU_ABORT); 1276 } 1277 if (!(inp->inp_flags & INP_DROPPED)) { 1278 soref(so); 1279 inp->inp_flags |= INP_SOCKREF; 1280 } 1281 INP_WUNLOCK(inp); 1282 dropped: 1283 NET_EPOCH_EXIT(et); 1284 } 1285 1286 /* 1287 * TCP socket is closed. Start friendly disconnect. 1288 */ 1289 static void 1290 tcp_usr_close(struct socket *so) 1291 { 1292 struct inpcb *inp; 1293 struct tcpcb *tp; 1294 struct epoch_tracker et; 1295 1296 inp = sotoinpcb(so); 1297 KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL")); 1298 1299 NET_EPOCH_ENTER(et); 1300 INP_WLOCK(inp); 1301 KASSERT(inp->inp_socket != NULL, 1302 ("tcp_usr_close: inp_socket == NULL")); 1303 1304 /* 1305 * If we are still connected and we're not dropped, initiate 1306 * a disconnect. 1307 */ 1308 if (!(inp->inp_flags & INP_DROPPED)) { 1309 tp = intotcpcb(inp); 1310 if (tp->t_state != TCPS_TIME_WAIT) { 1311 tp->t_flags |= TF_CLOSED; 1312 tcp_disconnect(tp); 1313 tcp_bblog_pru(tp, PRU_CLOSE, 0); 1314 TCP_PROBE2(debug__user, tp, PRU_CLOSE); 1315 } 1316 } 1317 if (!(inp->inp_flags & INP_DROPPED)) { 1318 soref(so); 1319 inp->inp_flags |= INP_SOCKREF; 1320 } 1321 INP_WUNLOCK(inp); 1322 NET_EPOCH_EXIT(et); 1323 } 1324 1325 static int 1326 tcp_pru_options_support(struct tcpcb *tp, int flags) 1327 { 1328 /* 1329 * If the specific TCP stack has a pru_options 1330 * specified then it does not always support 1331 * all the PRU_XX options and we must ask it. 1332 * If the function is not specified then all 1333 * of the PRU_XX options are supported. 1334 */ 1335 int ret = 0; 1336 1337 if (tp->t_fb->tfb_pru_options) { 1338 ret = (*tp->t_fb->tfb_pru_options)(tp, flags); 1339 } 1340 return (ret); 1341 } 1342 1343 /* 1344 * Receive out-of-band data. 1345 */ 1346 static int 1347 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) 1348 { 1349 int error = 0; 1350 struct inpcb *inp; 1351 struct tcpcb *tp; 1352 1353 inp = sotoinpcb(so); 1354 KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL")); 1355 INP_WLOCK(inp); 1356 if (inp->inp_flags & INP_DROPPED) { 1357 INP_WUNLOCK(inp); 1358 return (ECONNRESET); 1359 } 1360 tp = intotcpcb(inp); 1361 1362 error = tcp_pru_options_support(tp, PRUS_OOB); 1363 if (error) { 1364 goto out; 1365 } 1366 if ((so->so_oobmark == 0 && 1367 (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) || 1368 so->so_options & SO_OOBINLINE || 1369 tp->t_oobflags & TCPOOB_HADDATA) { 1370 error = EINVAL; 1371 goto out; 1372 } 1373 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 1374 error = EWOULDBLOCK; 1375 goto out; 1376 } 1377 m->m_len = 1; 1378 *mtod(m, caddr_t) = tp->t_iobc; 1379 if ((flags & MSG_PEEK) == 0) 1380 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 1381 1382 out: 1383 tcp_bblog_pru(tp, PRU_RCVOOB, error); 1384 TCP_PROBE2(debug__user, tp, PRU_RCVOOB); 1385 INP_WUNLOCK(inp); 1386 return (error); 1387 } 1388 1389 #ifdef INET 1390 struct protosw tcp_protosw = { 1391 .pr_type = SOCK_STREAM, 1392 .pr_protocol = IPPROTO_TCP, 1393 .pr_flags = PR_CONNREQUIRED | PR_IMPLOPCL | PR_WANTRCVD | 1394 PR_CAPATTACH, 1395 .pr_ctloutput = tcp_ctloutput, 1396 .pr_abort = tcp_usr_abort, 1397 .pr_accept = tcp_usr_accept, 1398 .pr_attach = tcp_usr_attach, 1399 .pr_bind = tcp_usr_bind, 1400 .pr_connect = tcp_usr_connect, 1401 .pr_control = in_control, 1402 .pr_detach = tcp_usr_detach, 1403 .pr_disconnect = tcp_usr_disconnect, 1404 .pr_listen = tcp_usr_listen, 1405 .pr_peeraddr = in_getpeeraddr, 1406 .pr_rcvd = tcp_usr_rcvd, 1407 .pr_rcvoob = tcp_usr_rcvoob, 1408 .pr_send = tcp_usr_send, 1409 .pr_ready = tcp_usr_ready, 1410 .pr_shutdown = tcp_usr_shutdown, 1411 .pr_sockaddr = in_getsockaddr, 1412 .pr_sosetlabel = in_pcbsosetlabel, 1413 .pr_close = tcp_usr_close, 1414 }; 1415 #endif /* INET */ 1416 1417 #ifdef INET6 1418 struct protosw tcp6_protosw = { 1419 .pr_type = SOCK_STREAM, 1420 .pr_protocol = IPPROTO_TCP, 1421 .pr_flags = PR_CONNREQUIRED | PR_IMPLOPCL |PR_WANTRCVD | 1422 PR_CAPATTACH, 1423 .pr_ctloutput = tcp_ctloutput, 1424 .pr_abort = tcp_usr_abort, 1425 .pr_accept = tcp6_usr_accept, 1426 .pr_attach = tcp_usr_attach, 1427 .pr_bind = tcp6_usr_bind, 1428 .pr_connect = tcp6_usr_connect, 1429 .pr_control = in6_control, 1430 .pr_detach = tcp_usr_detach, 1431 .pr_disconnect = tcp_usr_disconnect, 1432 .pr_listen = tcp6_usr_listen, 1433 .pr_peeraddr = in6_mapped_peeraddr, 1434 .pr_rcvd = tcp_usr_rcvd, 1435 .pr_rcvoob = tcp_usr_rcvoob, 1436 .pr_send = tcp_usr_send, 1437 .pr_ready = tcp_usr_ready, 1438 .pr_shutdown = tcp_usr_shutdown, 1439 .pr_sockaddr = in6_mapped_sockaddr, 1440 .pr_sosetlabel = in_pcbsosetlabel, 1441 .pr_close = tcp_usr_close, 1442 }; 1443 #endif /* INET6 */ 1444 1445 #ifdef INET 1446 /* 1447 * Common subroutine to open a TCP connection to remote host specified 1448 * by struct sockaddr_in. Call in_pcbconnect() to choose local host address 1449 * and assign a local port number and install the inpcb into the hash. 1450 * Initialize connection parameters and enter SYN-SENT state. 1451 */ 1452 static int 1453 tcp_connect(struct tcpcb *tp, struct sockaddr_in *sin, struct thread *td) 1454 { 1455 struct inpcb *inp = tptoinpcb(tp); 1456 struct socket *so = tptosocket(tp); 1457 int error; 1458 1459 NET_EPOCH_ASSERT(); 1460 INP_WLOCK_ASSERT(inp); 1461 1462 if (__predict_false((so->so_state & 1463 (SS_ISCONNECTING | SS_ISCONNECTED | SS_ISDISCONNECTING | 1464 SS_ISDISCONNECTED)) != 0)) 1465 return (EISCONN); 1466 1467 INP_HASH_WLOCK(&V_tcbinfo); 1468 error = in_pcbconnect(inp, sin, td->td_ucred); 1469 INP_HASH_WUNLOCK(&V_tcbinfo); 1470 if (error != 0) 1471 return (error); 1472 1473 /* 1474 * Compute window scaling to request: 1475 * Scale to fit into sweet spot. See tcp_syncache.c. 1476 * XXX: This should move to tcp_output(). 1477 */ 1478 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 1479 (TCP_MAXWIN << tp->request_r_scale) < sb_max) 1480 tp->request_r_scale++; 1481 1482 soisconnecting(so); 1483 TCPSTAT_INC(tcps_connattempt); 1484 tcp_state_change(tp, TCPS_SYN_SENT); 1485 tp->iss = tcp_new_isn(&inp->inp_inc); 1486 if (tp->t_flags & TF_REQ_TSTMP) 1487 tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc); 1488 tcp_sendseqinit(tp); 1489 1490 return (0); 1491 } 1492 #endif /* INET */ 1493 1494 #ifdef INET6 1495 static int 1496 tcp6_connect(struct tcpcb *tp, struct sockaddr_in6 *sin6, struct thread *td) 1497 { 1498 struct inpcb *inp = tptoinpcb(tp); 1499 struct socket *so = tptosocket(tp); 1500 int error; 1501 1502 NET_EPOCH_ASSERT(); 1503 INP_WLOCK_ASSERT(inp); 1504 1505 if (__predict_false((so->so_state & 1506 (SS_ISCONNECTING | SS_ISCONNECTED)) != 0)) 1507 return (EISCONN); 1508 1509 INP_HASH_WLOCK(&V_tcbinfo); 1510 error = in6_pcbconnect(inp, sin6, td->td_ucred, true); 1511 INP_HASH_WUNLOCK(&V_tcbinfo); 1512 if (error != 0) 1513 return (error); 1514 1515 /* Compute window scaling to request. */ 1516 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 1517 (TCP_MAXWIN << tp->request_r_scale) < sb_max) 1518 tp->request_r_scale++; 1519 1520 soisconnecting(so); 1521 TCPSTAT_INC(tcps_connattempt); 1522 tcp_state_change(tp, TCPS_SYN_SENT); 1523 tp->iss = tcp_new_isn(&inp->inp_inc); 1524 if (tp->t_flags & TF_REQ_TSTMP) 1525 tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc); 1526 tcp_sendseqinit(tp); 1527 1528 return (0); 1529 } 1530 #endif /* INET6 */ 1531 1532 /* 1533 * Export TCP internal state information via a struct tcp_info, based on the 1534 * Linux 2.6 API. Not ABI compatible as our constants are mapped differently 1535 * (TCP state machine, etc). We export all information using FreeBSD-native 1536 * constants -- for example, the numeric values for tcpi_state will differ 1537 * from Linux. 1538 */ 1539 void 1540 tcp_fill_info(const struct tcpcb *tp, struct tcp_info *ti) 1541 { 1542 1543 INP_LOCK_ASSERT(tptoinpcb(tp)); 1544 bzero(ti, sizeof(*ti)); 1545 1546 ti->tcpi_state = tp->t_state; 1547 if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP)) 1548 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS; 1549 if (tp->t_flags & TF_SACK_PERMIT) 1550 ti->tcpi_options |= TCPI_OPT_SACK; 1551 if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) { 1552 ti->tcpi_options |= TCPI_OPT_WSCALE; 1553 ti->tcpi_snd_wscale = tp->snd_scale; 1554 ti->tcpi_rcv_wscale = tp->rcv_scale; 1555 } 1556 switch (tp->t_flags2 & (TF2_ECN_PERMIT | TF2_ACE_PERMIT)) { 1557 case TF2_ECN_PERMIT: 1558 ti->tcpi_options |= TCPI_OPT_ECN; 1559 break; 1560 case TF2_ACE_PERMIT: 1561 /* FALLTHROUGH */ 1562 case TF2_ECN_PERMIT | TF2_ACE_PERMIT: 1563 ti->tcpi_options |= TCPI_OPT_ACE; 1564 break; 1565 default: 1566 break; 1567 } 1568 if (tp->t_flags & TF_FASTOPEN) 1569 ti->tcpi_options |= TCPI_OPT_TFO; 1570 1571 ti->tcpi_rto = tp->t_rxtcur * tick; 1572 ti->tcpi_last_data_recv = ((uint32_t)ticks - tp->t_rcvtime) * tick; 1573 ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT; 1574 ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT; 1575 1576 ti->tcpi_snd_ssthresh = tp->snd_ssthresh; 1577 ti->tcpi_snd_cwnd = tp->snd_cwnd; 1578 1579 /* 1580 * FreeBSD-specific extension fields for tcp_info. 1581 */ 1582 ti->tcpi_rcv_space = tp->rcv_wnd; 1583 ti->tcpi_rcv_nxt = tp->rcv_nxt; 1584 ti->tcpi_snd_wnd = tp->snd_wnd; 1585 ti->tcpi_snd_bwnd = 0; /* Unused, kept for compat. */ 1586 ti->tcpi_snd_nxt = tp->snd_nxt; 1587 ti->tcpi_snd_mss = tp->t_maxseg; 1588 ti->tcpi_rcv_mss = tp->t_maxseg; 1589 ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack; 1590 ti->tcpi_rcv_ooopack = tp->t_rcvoopack; 1591 ti->tcpi_snd_zerowin = tp->t_sndzerowin; 1592 ti->tcpi_snd_una = tp->snd_una; 1593 ti->tcpi_snd_max = tp->snd_max; 1594 ti->tcpi_rcv_numsacks = tp->rcv_numsacks; 1595 ti->tcpi_rcv_adv = tp->rcv_adv; 1596 ti->tcpi_dupacks = tp->t_dupacks; 1597 ti->tcpi_rttmin = tp->t_rttlow; 1598 #ifdef TCP_OFFLOAD 1599 if (tp->t_flags & TF_TOE) { 1600 ti->tcpi_options |= TCPI_OPT_TOE; 1601 tcp_offload_tcp_info(tp, ti); 1602 } 1603 #endif 1604 /* 1605 * AccECN related counters. 1606 */ 1607 if ((tp->t_flags2 & (TF2_ECN_PERMIT | TF2_ACE_PERMIT)) == 1608 (TF2_ECN_PERMIT | TF2_ACE_PERMIT)) 1609 /* 1610 * Internal counter starts at 5 for AccECN 1611 * but 0 for RFC3168 ECN. 1612 */ 1613 ti->tcpi_delivered_ce = tp->t_scep - 5; 1614 else 1615 ti->tcpi_delivered_ce = tp->t_scep; 1616 ti->tcpi_received_ce = tp->t_rcep; 1617 } 1618 1619 /* 1620 * tcp_ctloutput() must drop the inpcb lock before performing copyin on 1621 * socket option arguments. When it re-acquires the lock after the copy, it 1622 * has to revalidate that the connection is still valid for the socket 1623 * option. 1624 */ 1625 #define INP_WLOCK_RECHECK_CLEANUP(inp, cleanup) do { \ 1626 INP_WLOCK(inp); \ 1627 if (inp->inp_flags & INP_DROPPED) { \ 1628 INP_WUNLOCK(inp); \ 1629 cleanup; \ 1630 return (ECONNRESET); \ 1631 } \ 1632 tp = intotcpcb(inp); \ 1633 } while(0) 1634 #define INP_WLOCK_RECHECK(inp) INP_WLOCK_RECHECK_CLEANUP((inp), /* noop */) 1635 1636 int 1637 tcp_ctloutput_set(struct inpcb *inp, struct sockopt *sopt) 1638 { 1639 struct socket *so = inp->inp_socket; 1640 struct tcpcb *tp = intotcpcb(inp); 1641 int error = 0; 1642 1643 MPASS(sopt->sopt_dir == SOPT_SET); 1644 INP_WLOCK_ASSERT(inp); 1645 KASSERT((inp->inp_flags & INP_DROPPED) == 0, 1646 ("inp_flags == %x", inp->inp_flags)); 1647 KASSERT(so != NULL, ("inp_socket == NULL")); 1648 1649 if (sopt->sopt_level != IPPROTO_TCP) { 1650 INP_WUNLOCK(inp); 1651 #ifdef INET6 1652 if (inp->inp_vflag & INP_IPV6PROTO) 1653 error = ip6_ctloutput(so, sopt); 1654 #endif 1655 #if defined(INET6) && defined(INET) 1656 else 1657 #endif 1658 #ifdef INET 1659 error = ip_ctloutput(so, sopt); 1660 #endif 1661 /* 1662 * When an IP-level socket option affects TCP, pass control 1663 * down to stack tfb_tcp_ctloutput, otherwise return what 1664 * IP level returned. 1665 */ 1666 switch (sopt->sopt_level) { 1667 #ifdef INET6 1668 case IPPROTO_IPV6: 1669 if ((inp->inp_vflag & INP_IPV6PROTO) == 0) 1670 return (error); 1671 switch (sopt->sopt_name) { 1672 case IPV6_TCLASS: 1673 /* Notify tcp stacks that care (e.g. RACK). */ 1674 break; 1675 case IPV6_USE_MIN_MTU: 1676 /* Update t_maxseg accordingly. */ 1677 break; 1678 default: 1679 return (error); 1680 } 1681 break; 1682 #endif 1683 #ifdef INET 1684 case IPPROTO_IP: 1685 switch (sopt->sopt_name) { 1686 case IP_TOS: 1687 inp->inp_ip_tos &= ~IPTOS_ECN_MASK; 1688 break; 1689 case IP_TTL: 1690 /* Notify tcp stacks that care (e.g. RACK). */ 1691 break; 1692 default: 1693 return (error); 1694 } 1695 break; 1696 #endif 1697 default: 1698 return (error); 1699 } 1700 INP_WLOCK_RECHECK(inp); 1701 } else if (sopt->sopt_name == TCP_FUNCTION_BLK) { 1702 /* 1703 * Protect the TCP option TCP_FUNCTION_BLK so 1704 * that a sub-function can *never* overwrite this. 1705 */ 1706 struct tcp_function_set fsn; 1707 struct tcp_function_block *blk; 1708 void *ptr = NULL; 1709 1710 INP_WUNLOCK(inp); 1711 error = sooptcopyin(sopt, &fsn, sizeof fsn, sizeof fsn); 1712 if (error) 1713 return (error); 1714 1715 INP_WLOCK_RECHECK(inp); 1716 1717 blk = find_and_ref_tcp_functions(&fsn); 1718 if (blk == NULL) { 1719 INP_WUNLOCK(inp); 1720 return (ENOENT); 1721 } 1722 if (tp->t_fb == blk) { 1723 /* You already have this */ 1724 refcount_release(&blk->tfb_refcnt); 1725 INP_WUNLOCK(inp); 1726 return (0); 1727 } 1728 if (blk->tfb_flags & TCP_FUNC_BEING_REMOVED) { 1729 refcount_release(&blk->tfb_refcnt); 1730 INP_WUNLOCK(inp); 1731 return (ENOENT); 1732 } 1733 error = (*blk->tfb_tcp_handoff_ok)(tp); 1734 if (error) { 1735 refcount_release(&blk->tfb_refcnt); 1736 INP_WUNLOCK(inp); 1737 return (error); 1738 } 1739 /* 1740 * Ensure the new stack takes ownership with a 1741 * clean slate on peak rate threshold. 1742 */ 1743 if (tp->t_fb->tfb_tcp_timer_stop_all != NULL) 1744 tp->t_fb->tfb_tcp_timer_stop_all(tp); 1745 if (blk->tfb_tcp_fb_init) { 1746 error = (*blk->tfb_tcp_fb_init)(tp, &ptr); 1747 if (error) { 1748 /* 1749 * Release the ref count the lookup 1750 * acquired. 1751 */ 1752 refcount_release(&blk->tfb_refcnt); 1753 /* 1754 * Now there is a chance that the 1755 * init() function mucked with some 1756 * things before it failed, such as 1757 * hpts or inp_flags2 or timer granularity. 1758 * It should not of, but lets give the old 1759 * stack a chance to reset to a known good state. 1760 */ 1761 if (tp->t_fb->tfb_switch_failed) { 1762 (*tp->t_fb->tfb_switch_failed)(tp); 1763 } 1764 goto err_out; 1765 } 1766 } 1767 if (tp->t_fb->tfb_tcp_fb_fini) { 1768 struct epoch_tracker et; 1769 /* 1770 * Tell the stack to cleanup with 0 i.e. 1771 * the tcb is not going away. 1772 */ 1773 NET_EPOCH_ENTER(et); 1774 (*tp->t_fb->tfb_tcp_fb_fini)(tp, 0); 1775 NET_EPOCH_EXIT(et); 1776 } 1777 /* 1778 * Release the old refcnt, the 1779 * lookup acquired a ref on the 1780 * new one already. 1781 */ 1782 refcount_release(&tp->t_fb->tfb_refcnt); 1783 /* 1784 * Set in the new stack. 1785 */ 1786 tp->t_fb = blk; 1787 tp->t_fb_ptr = ptr; 1788 #ifdef TCP_OFFLOAD 1789 if (tp->t_flags & TF_TOE) { 1790 tcp_offload_ctloutput(tp, sopt->sopt_dir, 1791 sopt->sopt_name); 1792 } 1793 #endif 1794 err_out: 1795 INP_WUNLOCK(inp); 1796 return (error); 1797 1798 } 1799 1800 /* Pass in the INP locked, callee must unlock it. */ 1801 return (tp->t_fb->tfb_tcp_ctloutput(tp, sopt)); 1802 } 1803 1804 static int 1805 tcp_ctloutput_get(struct inpcb *inp, struct sockopt *sopt) 1806 { 1807 struct socket *so = inp->inp_socket; 1808 struct tcpcb *tp = intotcpcb(inp); 1809 int error = 0; 1810 1811 MPASS(sopt->sopt_dir == SOPT_GET); 1812 INP_WLOCK_ASSERT(inp); 1813 KASSERT((inp->inp_flags & INP_DROPPED) == 0, 1814 ("inp_flags == %x", inp->inp_flags)); 1815 KASSERT(so != NULL, ("inp_socket == NULL")); 1816 1817 if (sopt->sopt_level != IPPROTO_TCP) { 1818 INP_WUNLOCK(inp); 1819 #ifdef INET6 1820 if (inp->inp_vflag & INP_IPV6PROTO) 1821 error = ip6_ctloutput(so, sopt); 1822 #endif /* INET6 */ 1823 #if defined(INET6) && defined(INET) 1824 else 1825 #endif 1826 #ifdef INET 1827 error = ip_ctloutput(so, sopt); 1828 #endif 1829 return (error); 1830 } 1831 if (((sopt->sopt_name == TCP_FUNCTION_BLK) || 1832 (sopt->sopt_name == TCP_FUNCTION_ALIAS))) { 1833 struct tcp_function_set fsn; 1834 1835 if (sopt->sopt_name == TCP_FUNCTION_ALIAS) { 1836 memset(&fsn, 0, sizeof(fsn)); 1837 find_tcp_function_alias(tp->t_fb, &fsn); 1838 } else { 1839 strncpy(fsn.function_set_name, 1840 tp->t_fb->tfb_tcp_block_name, 1841 TCP_FUNCTION_NAME_LEN_MAX); 1842 fsn.function_set_name[TCP_FUNCTION_NAME_LEN_MAX - 1] = '\0'; 1843 } 1844 fsn.pcbcnt = tp->t_fb->tfb_refcnt; 1845 INP_WUNLOCK(inp); 1846 error = sooptcopyout(sopt, &fsn, sizeof fsn); 1847 return (error); 1848 } 1849 1850 /* Pass in the INP locked, callee must unlock it. */ 1851 return (tp->t_fb->tfb_tcp_ctloutput(tp, sopt)); 1852 } 1853 1854 int 1855 tcp_ctloutput(struct socket *so, struct sockopt *sopt) 1856 { 1857 struct inpcb *inp; 1858 1859 inp = sotoinpcb(so); 1860 KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL")); 1861 1862 INP_WLOCK(inp); 1863 if (inp->inp_flags & INP_DROPPED) { 1864 INP_WUNLOCK(inp); 1865 return (ECONNRESET); 1866 } 1867 if (sopt->sopt_dir == SOPT_SET) 1868 return (tcp_ctloutput_set(inp, sopt)); 1869 else if (sopt->sopt_dir == SOPT_GET) 1870 return (tcp_ctloutput_get(inp, sopt)); 1871 else 1872 panic("%s: sopt_dir $%d", __func__, sopt->sopt_dir); 1873 } 1874 1875 /* 1876 * If this assert becomes untrue, we need to change the size of the buf 1877 * variable in tcp_default_ctloutput(). 1878 */ 1879 #ifdef CTASSERT 1880 CTASSERT(TCP_CA_NAME_MAX <= TCP_LOG_ID_LEN); 1881 CTASSERT(TCP_LOG_REASON_LEN <= TCP_LOG_ID_LEN); 1882 #endif 1883 1884 extern struct cc_algo newreno_cc_algo; 1885 1886 static int 1887 tcp_set_cc_mod(struct inpcb *inp, struct sockopt *sopt) 1888 { 1889 struct cc_algo *algo; 1890 void *ptr = NULL; 1891 struct tcpcb *tp; 1892 struct cc_var cc_mem; 1893 char buf[TCP_CA_NAME_MAX]; 1894 size_t mem_sz; 1895 int error; 1896 1897 INP_WUNLOCK(inp); 1898 error = sooptcopyin(sopt, buf, TCP_CA_NAME_MAX - 1, 1); 1899 if (error) 1900 return(error); 1901 buf[sopt->sopt_valsize] = '\0'; 1902 CC_LIST_RLOCK(); 1903 STAILQ_FOREACH(algo, &cc_list, entries) { 1904 if (strncmp(buf, algo->name, 1905 TCP_CA_NAME_MAX) == 0) { 1906 if (algo->flags & CC_MODULE_BEING_REMOVED) { 1907 /* We can't "see" modules being unloaded */ 1908 continue; 1909 } 1910 break; 1911 } 1912 } 1913 if (algo == NULL) { 1914 CC_LIST_RUNLOCK(); 1915 return(ESRCH); 1916 } 1917 /* 1918 * With a reference the algorithm cannot be removed 1919 * so we hold a reference through the change process. 1920 */ 1921 cc_refer(algo); 1922 CC_LIST_RUNLOCK(); 1923 if (algo->cb_init != NULL) { 1924 /* We can now pre-get the memory for the CC */ 1925 mem_sz = (*algo->cc_data_sz)(); 1926 if (mem_sz == 0) { 1927 goto no_mem_needed; 1928 } 1929 ptr = malloc(mem_sz, M_CC_MEM, M_WAITOK); 1930 } else { 1931 no_mem_needed: 1932 mem_sz = 0; 1933 ptr = NULL; 1934 } 1935 /* 1936 * Make sure its all clean and zero and also get 1937 * back the inplock. 1938 */ 1939 memset(&cc_mem, 0, sizeof(cc_mem)); 1940 INP_WLOCK(inp); 1941 if (inp->inp_flags & INP_DROPPED) { 1942 INP_WUNLOCK(inp); 1943 if (ptr) 1944 free(ptr, M_CC_MEM); 1945 /* Release our temp reference */ 1946 CC_LIST_RLOCK(); 1947 cc_release(algo); 1948 CC_LIST_RUNLOCK(); 1949 return (ECONNRESET); 1950 } 1951 tp = intotcpcb(inp); 1952 if (ptr != NULL) 1953 memset(ptr, 0, mem_sz); 1954 cc_mem.tp = tp; 1955 /* 1956 * We once again hold a write lock over the tcb so it's 1957 * safe to do these things without ordering concerns. 1958 * Note here we init into stack memory. 1959 */ 1960 if (algo->cb_init != NULL) 1961 error = algo->cb_init(&cc_mem, ptr); 1962 else 1963 error = 0; 1964 /* 1965 * The CC algorithms, when given their memory 1966 * should not fail we could in theory have a 1967 * KASSERT here. 1968 */ 1969 if (error == 0) { 1970 /* 1971 * Touchdown, lets go ahead and move the 1972 * connection to the new CC module by 1973 * copying in the cc_mem after we call 1974 * the old ones cleanup (if any). 1975 */ 1976 if (CC_ALGO(tp)->cb_destroy != NULL) 1977 CC_ALGO(tp)->cb_destroy(&tp->t_ccv); 1978 /* Detach the old CC from the tcpcb */ 1979 cc_detach(tp); 1980 /* Copy in our temp memory that was inited */ 1981 memcpy(&tp->t_ccv, &cc_mem, sizeof(struct cc_var)); 1982 /* Now attach the new, which takes a reference */ 1983 cc_attach(tp, algo); 1984 /* Ok now are we where we have gotten past any conn_init? */ 1985 if (TCPS_HAVEESTABLISHED(tp->t_state) && (CC_ALGO(tp)->conn_init != NULL)) { 1986 /* Yep run the connection init for the new CC */ 1987 CC_ALGO(tp)->conn_init(&tp->t_ccv); 1988 } 1989 } else if (ptr) 1990 free(ptr, M_CC_MEM); 1991 INP_WUNLOCK(inp); 1992 /* Now lets release our temp reference */ 1993 CC_LIST_RLOCK(); 1994 cc_release(algo); 1995 CC_LIST_RUNLOCK(); 1996 return (error); 1997 } 1998 1999 int 2000 tcp_default_ctloutput(struct tcpcb *tp, struct sockopt *sopt) 2001 { 2002 struct inpcb *inp = tptoinpcb(tp); 2003 int error, opt, optval; 2004 u_int ui; 2005 struct tcp_info ti; 2006 #ifdef KERN_TLS 2007 struct tls_enable tls; 2008 struct socket *so = inp->inp_socket; 2009 #endif 2010 char *pbuf, buf[TCP_LOG_ID_LEN]; 2011 #ifdef STATS 2012 struct statsblob *sbp; 2013 #endif 2014 size_t len; 2015 2016 INP_WLOCK_ASSERT(inp); 2017 KASSERT((inp->inp_flags & INP_DROPPED) == 0, 2018 ("inp_flags == %x", inp->inp_flags)); 2019 KASSERT(inp->inp_socket != NULL, ("inp_socket == NULL")); 2020 2021 switch (sopt->sopt_level) { 2022 #ifdef INET6 2023 case IPPROTO_IPV6: 2024 MPASS(inp->inp_vflag & INP_IPV6PROTO); 2025 switch (sopt->sopt_name) { 2026 case IPV6_USE_MIN_MTU: 2027 tcp6_use_min_mtu(tp); 2028 /* FALLTHROUGH */ 2029 } 2030 INP_WUNLOCK(inp); 2031 return (0); 2032 #endif 2033 #ifdef INET 2034 case IPPROTO_IP: 2035 INP_WUNLOCK(inp); 2036 return (0); 2037 #endif 2038 } 2039 2040 /* 2041 * For TCP_CCALGOOPT forward the control to CC module, for both 2042 * SOPT_SET and SOPT_GET. 2043 */ 2044 switch (sopt->sopt_name) { 2045 case TCP_CCALGOOPT: 2046 INP_WUNLOCK(inp); 2047 if (sopt->sopt_valsize > CC_ALGOOPT_LIMIT) 2048 return (EINVAL); 2049 pbuf = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK | M_ZERO); 2050 error = sooptcopyin(sopt, pbuf, sopt->sopt_valsize, 2051 sopt->sopt_valsize); 2052 if (error) { 2053 free(pbuf, M_TEMP); 2054 return (error); 2055 } 2056 INP_WLOCK_RECHECK_CLEANUP(inp, free(pbuf, M_TEMP)); 2057 if (CC_ALGO(tp)->ctl_output != NULL) 2058 error = CC_ALGO(tp)->ctl_output(&tp->t_ccv, sopt, pbuf); 2059 else 2060 error = ENOENT; 2061 INP_WUNLOCK(inp); 2062 if (error == 0 && sopt->sopt_dir == SOPT_GET) 2063 error = sooptcopyout(sopt, pbuf, sopt->sopt_valsize); 2064 free(pbuf, M_TEMP); 2065 return (error); 2066 } 2067 2068 switch (sopt->sopt_dir) { 2069 case SOPT_SET: 2070 switch (sopt->sopt_name) { 2071 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE) 2072 case TCP_MD5SIG: 2073 INP_WUNLOCK(inp); 2074 if (!TCPMD5_ENABLED()) 2075 return (ENOPROTOOPT); 2076 error = TCPMD5_PCBCTL(inp, sopt); 2077 if (error) 2078 return (error); 2079 INP_WLOCK_RECHECK(inp); 2080 goto unlock_and_done; 2081 #endif /* IPSEC */ 2082 2083 case TCP_NODELAY: 2084 case TCP_NOOPT: 2085 INP_WUNLOCK(inp); 2086 error = sooptcopyin(sopt, &optval, sizeof optval, 2087 sizeof optval); 2088 if (error) 2089 return (error); 2090 2091 INP_WLOCK_RECHECK(inp); 2092 switch (sopt->sopt_name) { 2093 case TCP_NODELAY: 2094 opt = TF_NODELAY; 2095 break; 2096 case TCP_NOOPT: 2097 opt = TF_NOOPT; 2098 break; 2099 default: 2100 opt = 0; /* dead code to fool gcc */ 2101 break; 2102 } 2103 2104 if (optval) 2105 tp->t_flags |= opt; 2106 else 2107 tp->t_flags &= ~opt; 2108 unlock_and_done: 2109 #ifdef TCP_OFFLOAD 2110 if (tp->t_flags & TF_TOE) { 2111 tcp_offload_ctloutput(tp, sopt->sopt_dir, 2112 sopt->sopt_name); 2113 } 2114 #endif 2115 INP_WUNLOCK(inp); 2116 break; 2117 2118 case TCP_NOPUSH: 2119 INP_WUNLOCK(inp); 2120 error = sooptcopyin(sopt, &optval, sizeof optval, 2121 sizeof optval); 2122 if (error) 2123 return (error); 2124 2125 INP_WLOCK_RECHECK(inp); 2126 if (optval) 2127 tp->t_flags |= TF_NOPUSH; 2128 else if (tp->t_flags & TF_NOPUSH) { 2129 tp->t_flags &= ~TF_NOPUSH; 2130 if (TCPS_HAVEESTABLISHED(tp->t_state)) { 2131 struct epoch_tracker et; 2132 2133 NET_EPOCH_ENTER(et); 2134 error = tcp_output_nodrop(tp); 2135 NET_EPOCH_EXIT(et); 2136 } 2137 } 2138 goto unlock_and_done; 2139 2140 case TCP_REMOTE_UDP_ENCAPS_PORT: 2141 INP_WUNLOCK(inp); 2142 error = sooptcopyin(sopt, &optval, sizeof optval, 2143 sizeof optval); 2144 if (error) 2145 return (error); 2146 if ((optval < TCP_TUNNELING_PORT_MIN) || 2147 (optval > TCP_TUNNELING_PORT_MAX)) { 2148 /* Its got to be in range */ 2149 return (EINVAL); 2150 } 2151 if ((V_tcp_udp_tunneling_port == 0) && (optval != 0)) { 2152 /* You have to have enabled a UDP tunneling port first */ 2153 return (EINVAL); 2154 } 2155 INP_WLOCK_RECHECK(inp); 2156 if (tp->t_state != TCPS_CLOSED) { 2157 /* You can't change after you are connected */ 2158 error = EINVAL; 2159 } else { 2160 /* Ok we are all good set the port */ 2161 tp->t_port = htons(optval); 2162 } 2163 goto unlock_and_done; 2164 2165 case TCP_MAXSEG: 2166 INP_WUNLOCK(inp); 2167 error = sooptcopyin(sopt, &optval, sizeof optval, 2168 sizeof optval); 2169 if (error) 2170 return (error); 2171 2172 INP_WLOCK_RECHECK(inp); 2173 if (optval > 0 && optval <= tp->t_maxseg && 2174 optval + 40 >= V_tcp_minmss) { 2175 tp->t_maxseg = optval; 2176 if (tp->t_maxseg < V_tcp_mssdflt) { 2177 /* 2178 * The MSS is so small we should not process incoming 2179 * SACK's since we are subject to attack in such a 2180 * case. 2181 */ 2182 tp->t_flags2 |= TF2_PROC_SACK_PROHIBIT; 2183 } else { 2184 tp->t_flags2 &= ~TF2_PROC_SACK_PROHIBIT; 2185 } 2186 } else 2187 error = EINVAL; 2188 goto unlock_and_done; 2189 2190 case TCP_INFO: 2191 INP_WUNLOCK(inp); 2192 error = EINVAL; 2193 break; 2194 2195 case TCP_STATS: 2196 INP_WUNLOCK(inp); 2197 #ifdef STATS 2198 error = sooptcopyin(sopt, &optval, sizeof optval, 2199 sizeof optval); 2200 if (error) 2201 return (error); 2202 2203 if (optval > 0) 2204 sbp = stats_blob_alloc( 2205 V_tcp_perconn_stats_dflt_tpl, 0); 2206 else 2207 sbp = NULL; 2208 2209 INP_WLOCK_RECHECK(inp); 2210 if ((tp->t_stats != NULL && sbp == NULL) || 2211 (tp->t_stats == NULL && sbp != NULL)) { 2212 struct statsblob *t = tp->t_stats; 2213 tp->t_stats = sbp; 2214 sbp = t; 2215 } 2216 INP_WUNLOCK(inp); 2217 2218 stats_blob_destroy(sbp); 2219 #else 2220 return (EOPNOTSUPP); 2221 #endif /* !STATS */ 2222 break; 2223 2224 case TCP_CONGESTION: 2225 error = tcp_set_cc_mod(inp, sopt); 2226 break; 2227 2228 case TCP_REUSPORT_LB_NUMA: 2229 INP_WUNLOCK(inp); 2230 error = sooptcopyin(sopt, &optval, sizeof(optval), 2231 sizeof(optval)); 2232 INP_WLOCK_RECHECK(inp); 2233 if (!error) 2234 error = in_pcblbgroup_numa(inp, optval); 2235 INP_WUNLOCK(inp); 2236 break; 2237 2238 #ifdef KERN_TLS 2239 case TCP_TXTLS_ENABLE: 2240 INP_WUNLOCK(inp); 2241 error = ktls_copyin_tls_enable(sopt, &tls); 2242 if (error != 0) 2243 break; 2244 error = ktls_enable_tx(so, &tls); 2245 ktls_cleanup_tls_enable(&tls); 2246 break; 2247 case TCP_TXTLS_MODE: 2248 INP_WUNLOCK(inp); 2249 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui)); 2250 if (error != 0) 2251 return (error); 2252 2253 INP_WLOCK_RECHECK(inp); 2254 error = ktls_set_tx_mode(so, ui); 2255 INP_WUNLOCK(inp); 2256 break; 2257 case TCP_RXTLS_ENABLE: 2258 INP_WUNLOCK(inp); 2259 error = ktls_copyin_tls_enable(sopt, &tls); 2260 if (error != 0) 2261 break; 2262 error = ktls_enable_rx(so, &tls); 2263 ktls_cleanup_tls_enable(&tls); 2264 break; 2265 #endif 2266 case TCP_MAXUNACKTIME: 2267 case TCP_KEEPIDLE: 2268 case TCP_KEEPINTVL: 2269 case TCP_KEEPINIT: 2270 INP_WUNLOCK(inp); 2271 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui)); 2272 if (error) 2273 return (error); 2274 2275 if (ui > (UINT_MAX / hz)) { 2276 error = EINVAL; 2277 break; 2278 } 2279 ui *= hz; 2280 2281 INP_WLOCK_RECHECK(inp); 2282 switch (sopt->sopt_name) { 2283 case TCP_MAXUNACKTIME: 2284 tp->t_maxunacktime = ui; 2285 break; 2286 2287 case TCP_KEEPIDLE: 2288 tp->t_keepidle = ui; 2289 /* 2290 * XXX: better check current remaining 2291 * timeout and "merge" it with new value. 2292 */ 2293 if ((tp->t_state > TCPS_LISTEN) && 2294 (tp->t_state <= TCPS_CLOSING)) 2295 tcp_timer_activate(tp, TT_KEEP, 2296 TP_KEEPIDLE(tp)); 2297 break; 2298 case TCP_KEEPINTVL: 2299 tp->t_keepintvl = ui; 2300 if ((tp->t_state == TCPS_FIN_WAIT_2) && 2301 (TP_MAXIDLE(tp) > 0)) 2302 tcp_timer_activate(tp, TT_2MSL, 2303 TP_MAXIDLE(tp)); 2304 break; 2305 case TCP_KEEPINIT: 2306 tp->t_keepinit = ui; 2307 if (tp->t_state == TCPS_SYN_RECEIVED || 2308 tp->t_state == TCPS_SYN_SENT) 2309 tcp_timer_activate(tp, TT_KEEP, 2310 TP_KEEPINIT(tp)); 2311 break; 2312 } 2313 goto unlock_and_done; 2314 2315 case TCP_KEEPCNT: 2316 INP_WUNLOCK(inp); 2317 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui)); 2318 if (error) 2319 return (error); 2320 2321 INP_WLOCK_RECHECK(inp); 2322 tp->t_keepcnt = ui; 2323 if ((tp->t_state == TCPS_FIN_WAIT_2) && 2324 (TP_MAXIDLE(tp) > 0)) 2325 tcp_timer_activate(tp, TT_2MSL, 2326 TP_MAXIDLE(tp)); 2327 goto unlock_and_done; 2328 2329 #ifdef TCPPCAP 2330 case TCP_PCAP_OUT: 2331 case TCP_PCAP_IN: 2332 INP_WUNLOCK(inp); 2333 error = sooptcopyin(sopt, &optval, sizeof optval, 2334 sizeof optval); 2335 if (error) 2336 return (error); 2337 2338 INP_WLOCK_RECHECK(inp); 2339 if (optval >= 0) 2340 tcp_pcap_set_sock_max( 2341 (sopt->sopt_name == TCP_PCAP_OUT) ? 2342 &(tp->t_outpkts) : &(tp->t_inpkts), 2343 optval); 2344 else 2345 error = EINVAL; 2346 goto unlock_and_done; 2347 #endif 2348 2349 case TCP_FASTOPEN: { 2350 struct tcp_fastopen tfo_optval; 2351 2352 INP_WUNLOCK(inp); 2353 if (!V_tcp_fastopen_client_enable && 2354 !V_tcp_fastopen_server_enable) 2355 return (EPERM); 2356 2357 error = sooptcopyin(sopt, &tfo_optval, 2358 sizeof(tfo_optval), sizeof(int)); 2359 if (error) 2360 return (error); 2361 2362 INP_WLOCK_RECHECK(inp); 2363 if ((tp->t_state != TCPS_CLOSED) && 2364 (tp->t_state != TCPS_LISTEN)) { 2365 error = EINVAL; 2366 goto unlock_and_done; 2367 } 2368 if (tfo_optval.enable) { 2369 if (tp->t_state == TCPS_LISTEN) { 2370 if (!V_tcp_fastopen_server_enable) { 2371 error = EPERM; 2372 goto unlock_and_done; 2373 } 2374 2375 if (tp->t_tfo_pending == NULL) 2376 tp->t_tfo_pending = 2377 tcp_fastopen_alloc_counter(); 2378 } else { 2379 /* 2380 * If a pre-shared key was provided, 2381 * stash it in the client cookie 2382 * field of the tcpcb for use during 2383 * connect. 2384 */ 2385 if (sopt->sopt_valsize == 2386 sizeof(tfo_optval)) { 2387 memcpy(tp->t_tfo_cookie.client, 2388 tfo_optval.psk, 2389 TCP_FASTOPEN_PSK_LEN); 2390 tp->t_tfo_client_cookie_len = 2391 TCP_FASTOPEN_PSK_LEN; 2392 } 2393 } 2394 tp->t_flags |= TF_FASTOPEN; 2395 } else 2396 tp->t_flags &= ~TF_FASTOPEN; 2397 goto unlock_and_done; 2398 } 2399 2400 #ifdef TCP_BLACKBOX 2401 case TCP_LOG: 2402 INP_WUNLOCK(inp); 2403 error = sooptcopyin(sopt, &optval, sizeof optval, 2404 sizeof optval); 2405 if (error) 2406 return (error); 2407 2408 INP_WLOCK_RECHECK(inp); 2409 error = tcp_log_state_change(tp, optval); 2410 goto unlock_and_done; 2411 2412 case TCP_LOGBUF: 2413 INP_WUNLOCK(inp); 2414 error = EINVAL; 2415 break; 2416 2417 case TCP_LOGID: 2418 INP_WUNLOCK(inp); 2419 error = sooptcopyin(sopt, buf, TCP_LOG_ID_LEN - 1, 0); 2420 if (error) 2421 break; 2422 buf[sopt->sopt_valsize] = '\0'; 2423 INP_WLOCK_RECHECK(inp); 2424 error = tcp_log_set_id(tp, buf); 2425 /* tcp_log_set_id() unlocks the INP. */ 2426 break; 2427 2428 case TCP_LOGDUMP: 2429 case TCP_LOGDUMPID: 2430 INP_WUNLOCK(inp); 2431 error = 2432 sooptcopyin(sopt, buf, TCP_LOG_REASON_LEN - 1, 0); 2433 if (error) 2434 break; 2435 buf[sopt->sopt_valsize] = '\0'; 2436 INP_WLOCK_RECHECK(inp); 2437 if (sopt->sopt_name == TCP_LOGDUMP) { 2438 error = tcp_log_dump_tp_logbuf(tp, buf, 2439 M_WAITOK, true); 2440 INP_WUNLOCK(inp); 2441 } else { 2442 tcp_log_dump_tp_bucket_logbufs(tp, buf); 2443 /* 2444 * tcp_log_dump_tp_bucket_logbufs() drops the 2445 * INP lock. 2446 */ 2447 } 2448 break; 2449 #endif 2450 2451 default: 2452 INP_WUNLOCK(inp); 2453 error = ENOPROTOOPT; 2454 break; 2455 } 2456 break; 2457 2458 case SOPT_GET: 2459 tp = intotcpcb(inp); 2460 switch (sopt->sopt_name) { 2461 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE) 2462 case TCP_MD5SIG: 2463 INP_WUNLOCK(inp); 2464 if (!TCPMD5_ENABLED()) 2465 return (ENOPROTOOPT); 2466 error = TCPMD5_PCBCTL(inp, sopt); 2467 break; 2468 #endif 2469 2470 case TCP_NODELAY: 2471 optval = tp->t_flags & TF_NODELAY; 2472 INP_WUNLOCK(inp); 2473 error = sooptcopyout(sopt, &optval, sizeof optval); 2474 break; 2475 case TCP_MAXSEG: 2476 optval = tp->t_maxseg; 2477 INP_WUNLOCK(inp); 2478 error = sooptcopyout(sopt, &optval, sizeof optval); 2479 break; 2480 case TCP_REMOTE_UDP_ENCAPS_PORT: 2481 optval = ntohs(tp->t_port); 2482 INP_WUNLOCK(inp); 2483 error = sooptcopyout(sopt, &optval, sizeof optval); 2484 break; 2485 case TCP_NOOPT: 2486 optval = tp->t_flags & TF_NOOPT; 2487 INP_WUNLOCK(inp); 2488 error = sooptcopyout(sopt, &optval, sizeof optval); 2489 break; 2490 case TCP_NOPUSH: 2491 optval = tp->t_flags & TF_NOPUSH; 2492 INP_WUNLOCK(inp); 2493 error = sooptcopyout(sopt, &optval, sizeof optval); 2494 break; 2495 case TCP_INFO: 2496 tcp_fill_info(tp, &ti); 2497 INP_WUNLOCK(inp); 2498 error = sooptcopyout(sopt, &ti, sizeof ti); 2499 break; 2500 case TCP_STATS: 2501 { 2502 #ifdef STATS 2503 int nheld; 2504 TYPEOF_MEMBER(struct statsblob, flags) sbflags = 0; 2505 2506 error = 0; 2507 socklen_t outsbsz = sopt->sopt_valsize; 2508 if (tp->t_stats == NULL) 2509 error = ENOENT; 2510 else if (outsbsz >= tp->t_stats->cursz) 2511 outsbsz = tp->t_stats->cursz; 2512 else if (outsbsz >= sizeof(struct statsblob)) 2513 outsbsz = sizeof(struct statsblob); 2514 else 2515 error = EINVAL; 2516 INP_WUNLOCK(inp); 2517 if (error) 2518 break; 2519 2520 sbp = sopt->sopt_val; 2521 nheld = atop(round_page(((vm_offset_t)sbp) + 2522 (vm_size_t)outsbsz) - trunc_page((vm_offset_t)sbp)); 2523 vm_page_t ma[nheld]; 2524 if (vm_fault_quick_hold_pages( 2525 &curproc->p_vmspace->vm_map, (vm_offset_t)sbp, 2526 outsbsz, VM_PROT_READ | VM_PROT_WRITE, ma, 2527 nheld) < 0) { 2528 error = EFAULT; 2529 break; 2530 } 2531 2532 if ((error = copyin_nofault(&(sbp->flags), &sbflags, 2533 SIZEOF_MEMBER(struct statsblob, flags)))) 2534 goto unhold; 2535 2536 INP_WLOCK_RECHECK(inp); 2537 error = stats_blob_snapshot(&sbp, outsbsz, tp->t_stats, 2538 sbflags | SB_CLONE_USRDSTNOFAULT); 2539 INP_WUNLOCK(inp); 2540 sopt->sopt_valsize = outsbsz; 2541 unhold: 2542 vm_page_unhold_pages(ma, nheld); 2543 #else 2544 INP_WUNLOCK(inp); 2545 error = EOPNOTSUPP; 2546 #endif /* !STATS */ 2547 break; 2548 } 2549 case TCP_CONGESTION: 2550 len = strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX); 2551 INP_WUNLOCK(inp); 2552 error = sooptcopyout(sopt, buf, len + 1); 2553 break; 2554 case TCP_MAXUNACKTIME: 2555 case TCP_KEEPIDLE: 2556 case TCP_KEEPINTVL: 2557 case TCP_KEEPINIT: 2558 case TCP_KEEPCNT: 2559 switch (sopt->sopt_name) { 2560 case TCP_MAXUNACKTIME: 2561 ui = TP_MAXUNACKTIME(tp) / hz; 2562 break; 2563 case TCP_KEEPIDLE: 2564 ui = TP_KEEPIDLE(tp) / hz; 2565 break; 2566 case TCP_KEEPINTVL: 2567 ui = TP_KEEPINTVL(tp) / hz; 2568 break; 2569 case TCP_KEEPINIT: 2570 ui = TP_KEEPINIT(tp) / hz; 2571 break; 2572 case TCP_KEEPCNT: 2573 ui = TP_KEEPCNT(tp); 2574 break; 2575 } 2576 INP_WUNLOCK(inp); 2577 error = sooptcopyout(sopt, &ui, sizeof(ui)); 2578 break; 2579 #ifdef TCPPCAP 2580 case TCP_PCAP_OUT: 2581 case TCP_PCAP_IN: 2582 optval = tcp_pcap_get_sock_max( 2583 (sopt->sopt_name == TCP_PCAP_OUT) ? 2584 &(tp->t_outpkts) : &(tp->t_inpkts)); 2585 INP_WUNLOCK(inp); 2586 error = sooptcopyout(sopt, &optval, sizeof optval); 2587 break; 2588 #endif 2589 case TCP_FASTOPEN: 2590 optval = tp->t_flags & TF_FASTOPEN; 2591 INP_WUNLOCK(inp); 2592 error = sooptcopyout(sopt, &optval, sizeof optval); 2593 break; 2594 #ifdef TCP_BLACKBOX 2595 case TCP_LOG: 2596 optval = tcp_get_bblog_state(tp); 2597 INP_WUNLOCK(inp); 2598 error = sooptcopyout(sopt, &optval, sizeof(optval)); 2599 break; 2600 case TCP_LOGBUF: 2601 /* tcp_log_getlogbuf() does INP_WUNLOCK(inp) */ 2602 error = tcp_log_getlogbuf(sopt, tp); 2603 break; 2604 case TCP_LOGID: 2605 len = tcp_log_get_id(tp, buf); 2606 INP_WUNLOCK(inp); 2607 error = sooptcopyout(sopt, buf, len + 1); 2608 break; 2609 case TCP_LOGDUMP: 2610 case TCP_LOGDUMPID: 2611 INP_WUNLOCK(inp); 2612 error = EINVAL; 2613 break; 2614 #endif 2615 #ifdef KERN_TLS 2616 case TCP_TXTLS_MODE: 2617 error = ktls_get_tx_mode(so, &optval); 2618 INP_WUNLOCK(inp); 2619 if (error == 0) 2620 error = sooptcopyout(sopt, &optval, 2621 sizeof(optval)); 2622 break; 2623 case TCP_RXTLS_MODE: 2624 error = ktls_get_rx_mode(so, &optval); 2625 INP_WUNLOCK(inp); 2626 if (error == 0) 2627 error = sooptcopyout(sopt, &optval, 2628 sizeof(optval)); 2629 break; 2630 #endif 2631 default: 2632 INP_WUNLOCK(inp); 2633 error = ENOPROTOOPT; 2634 break; 2635 } 2636 break; 2637 } 2638 return (error); 2639 } 2640 #undef INP_WLOCK_RECHECK 2641 #undef INP_WLOCK_RECHECK_CLEANUP 2642 2643 /* 2644 * Initiate (or continue) disconnect. 2645 * If embryonic state, just send reset (once). 2646 * If in ``let data drain'' option and linger null, just drop. 2647 * Otherwise (hard), mark socket disconnecting and drop 2648 * current input data; switch states based on user close, and 2649 * send segment to peer (with FIN). 2650 */ 2651 static void 2652 tcp_disconnect(struct tcpcb *tp) 2653 { 2654 struct inpcb *inp = tptoinpcb(tp); 2655 struct socket *so = tptosocket(tp); 2656 2657 NET_EPOCH_ASSERT(); 2658 INP_WLOCK_ASSERT(inp); 2659 2660 /* 2661 * Neither tcp_close() nor tcp_drop() should return NULL, as the 2662 * socket is still open. 2663 */ 2664 if (tp->t_state < TCPS_ESTABLISHED && 2665 !(tp->t_state > TCPS_LISTEN && (tp->t_flags & TF_FASTOPEN))) { 2666 tp = tcp_close(tp); 2667 KASSERT(tp != NULL, 2668 ("tcp_disconnect: tcp_close() returned NULL")); 2669 } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) { 2670 tp = tcp_drop(tp, 0); 2671 KASSERT(tp != NULL, 2672 ("tcp_disconnect: tcp_drop() returned NULL")); 2673 } else { 2674 soisdisconnecting(so); 2675 sbflush(&so->so_rcv); 2676 tcp_usrclosed(tp); 2677 if (!(inp->inp_flags & INP_DROPPED)) 2678 /* Ignore stack's drop request, we already at it. */ 2679 (void)tcp_output_nodrop(tp); 2680 } 2681 } 2682 2683 /* 2684 * User issued close, and wish to trail through shutdown states: 2685 * if never received SYN, just forget it. If got a SYN from peer, 2686 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 2687 * If already got a FIN from peer, then almost done; go to LAST_ACK 2688 * state. In all other cases, have already sent FIN to peer (e.g. 2689 * after PRU_SHUTDOWN), and just have to play tedious game waiting 2690 * for peer to send FIN or not respond to keep-alives, etc. 2691 * We can let the user exit from the close as soon as the FIN is acked. 2692 */ 2693 static void 2694 tcp_usrclosed(struct tcpcb *tp) 2695 { 2696 2697 NET_EPOCH_ASSERT(); 2698 INP_WLOCK_ASSERT(tptoinpcb(tp)); 2699 2700 switch (tp->t_state) { 2701 case TCPS_LISTEN: 2702 #ifdef TCP_OFFLOAD 2703 tcp_offload_listen_stop(tp); 2704 #endif 2705 tcp_state_change(tp, TCPS_CLOSED); 2706 /* FALLTHROUGH */ 2707 case TCPS_CLOSED: 2708 tp = tcp_close(tp); 2709 /* 2710 * tcp_close() should never return NULL here as the socket is 2711 * still open. 2712 */ 2713 KASSERT(tp != NULL, 2714 ("tcp_usrclosed: tcp_close() returned NULL")); 2715 break; 2716 2717 case TCPS_SYN_SENT: 2718 case TCPS_SYN_RECEIVED: 2719 tp->t_flags |= TF_NEEDFIN; 2720 break; 2721 2722 case TCPS_ESTABLISHED: 2723 tcp_state_change(tp, TCPS_FIN_WAIT_1); 2724 break; 2725 2726 case TCPS_CLOSE_WAIT: 2727 tcp_state_change(tp, TCPS_LAST_ACK); 2728 break; 2729 } 2730 if (tp->t_acktime == 0) 2731 tp->t_acktime = ticks; 2732 if (tp->t_state >= TCPS_FIN_WAIT_2) { 2733 tcp_free_sackholes(tp); 2734 soisdisconnected(tptosocket(tp)); 2735 /* Prevent the connection hanging in FIN_WAIT_2 forever. */ 2736 if (tp->t_state == TCPS_FIN_WAIT_2) { 2737 int timeout; 2738 2739 timeout = (tcp_fast_finwait2_recycle) ? 2740 tcp_finwait2_timeout : TP_MAXIDLE(tp); 2741 tcp_timer_activate(tp, TT_2MSL, timeout); 2742 } 2743 } 2744 } 2745 2746 #ifdef DDB 2747 static void 2748 db_print_indent(int indent) 2749 { 2750 int i; 2751 2752 for (i = 0; i < indent; i++) 2753 db_printf(" "); 2754 } 2755 2756 static void 2757 db_print_tstate(int t_state) 2758 { 2759 2760 switch (t_state) { 2761 case TCPS_CLOSED: 2762 db_printf("TCPS_CLOSED"); 2763 return; 2764 2765 case TCPS_LISTEN: 2766 db_printf("TCPS_LISTEN"); 2767 return; 2768 2769 case TCPS_SYN_SENT: 2770 db_printf("TCPS_SYN_SENT"); 2771 return; 2772 2773 case TCPS_SYN_RECEIVED: 2774 db_printf("TCPS_SYN_RECEIVED"); 2775 return; 2776 2777 case TCPS_ESTABLISHED: 2778 db_printf("TCPS_ESTABLISHED"); 2779 return; 2780 2781 case TCPS_CLOSE_WAIT: 2782 db_printf("TCPS_CLOSE_WAIT"); 2783 return; 2784 2785 case TCPS_FIN_WAIT_1: 2786 db_printf("TCPS_FIN_WAIT_1"); 2787 return; 2788 2789 case TCPS_CLOSING: 2790 db_printf("TCPS_CLOSING"); 2791 return; 2792 2793 case TCPS_LAST_ACK: 2794 db_printf("TCPS_LAST_ACK"); 2795 return; 2796 2797 case TCPS_FIN_WAIT_2: 2798 db_printf("TCPS_FIN_WAIT_2"); 2799 return; 2800 2801 case TCPS_TIME_WAIT: 2802 db_printf("TCPS_TIME_WAIT"); 2803 return; 2804 2805 default: 2806 db_printf("unknown"); 2807 return; 2808 } 2809 } 2810 2811 static void 2812 db_print_tflags(u_int t_flags) 2813 { 2814 int comma; 2815 2816 comma = 0; 2817 if (t_flags & TF_ACKNOW) { 2818 db_printf("%sTF_ACKNOW", comma ? ", " : ""); 2819 comma = 1; 2820 } 2821 if (t_flags & TF_DELACK) { 2822 db_printf("%sTF_DELACK", comma ? ", " : ""); 2823 comma = 1; 2824 } 2825 if (t_flags & TF_NODELAY) { 2826 db_printf("%sTF_NODELAY", comma ? ", " : ""); 2827 comma = 1; 2828 } 2829 if (t_flags & TF_NOOPT) { 2830 db_printf("%sTF_NOOPT", comma ? ", " : ""); 2831 comma = 1; 2832 } 2833 if (t_flags & TF_SENTFIN) { 2834 db_printf("%sTF_SENTFIN", comma ? ", " : ""); 2835 comma = 1; 2836 } 2837 if (t_flags & TF_REQ_SCALE) { 2838 db_printf("%sTF_REQ_SCALE", comma ? ", " : ""); 2839 comma = 1; 2840 } 2841 if (t_flags & TF_RCVD_SCALE) { 2842 db_printf("%sTF_RECVD_SCALE", comma ? ", " : ""); 2843 comma = 1; 2844 } 2845 if (t_flags & TF_REQ_TSTMP) { 2846 db_printf("%sTF_REQ_TSTMP", comma ? ", " : ""); 2847 comma = 1; 2848 } 2849 if (t_flags & TF_RCVD_TSTMP) { 2850 db_printf("%sTF_RCVD_TSTMP", comma ? ", " : ""); 2851 comma = 1; 2852 } 2853 if (t_flags & TF_SACK_PERMIT) { 2854 db_printf("%sTF_SACK_PERMIT", comma ? ", " : ""); 2855 comma = 1; 2856 } 2857 if (t_flags & TF_NEEDSYN) { 2858 db_printf("%sTF_NEEDSYN", comma ? ", " : ""); 2859 comma = 1; 2860 } 2861 if (t_flags & TF_NEEDFIN) { 2862 db_printf("%sTF_NEEDFIN", comma ? ", " : ""); 2863 comma = 1; 2864 } 2865 if (t_flags & TF_NOPUSH) { 2866 db_printf("%sTF_NOPUSH", comma ? ", " : ""); 2867 comma = 1; 2868 } 2869 if (t_flags & TF_PREVVALID) { 2870 db_printf("%sTF_PREVVALID", comma ? ", " : ""); 2871 comma = 1; 2872 } 2873 if (t_flags & TF_WAKESOR) { 2874 db_printf("%sTF_WAKESOR", comma ? ", " : ""); 2875 comma = 1; 2876 } 2877 if (t_flags & TF_GPUTINPROG) { 2878 db_printf("%sTF_GPUTINPROG", comma ? ", " : ""); 2879 comma = 1; 2880 } 2881 if (t_flags & TF_MORETOCOME) { 2882 db_printf("%sTF_MORETOCOME", comma ? ", " : ""); 2883 comma = 1; 2884 } 2885 if (t_flags & TF_SONOTCONN) { 2886 db_printf("%sTF_SONOTCONN", comma ? ", " : ""); 2887 comma = 1; 2888 } 2889 if (t_flags & TF_LASTIDLE) { 2890 db_printf("%sTF_LASTIDLE", comma ? ", " : ""); 2891 comma = 1; 2892 } 2893 if (t_flags & TF_RXWIN0SENT) { 2894 db_printf("%sTF_RXWIN0SENT", comma ? ", " : ""); 2895 comma = 1; 2896 } 2897 if (t_flags & TF_FASTRECOVERY) { 2898 db_printf("%sTF_FASTRECOVERY", comma ? ", " : ""); 2899 comma = 1; 2900 } 2901 if (t_flags & TF_WASFRECOVERY) { 2902 db_printf("%sTF_WASFRECOVERY", comma ? ", " : ""); 2903 comma = 1; 2904 } 2905 if (t_flags & TF_SIGNATURE) { 2906 db_printf("%sTF_SIGNATURE", comma ? ", " : ""); 2907 comma = 1; 2908 } 2909 if (t_flags & TF_FORCEDATA) { 2910 db_printf("%sTF_FORCEDATA", comma ? ", " : ""); 2911 comma = 1; 2912 } 2913 if (t_flags & TF_TSO) { 2914 db_printf("%sTF_TSO", comma ? ", " : ""); 2915 comma = 1; 2916 } 2917 if (t_flags & TF_TOE) { 2918 db_printf("%sTF_TOE", comma ? ", " : ""); 2919 comma = 1; 2920 } 2921 if (t_flags & TF_CLOSED) { 2922 db_printf("%sTF_CLOSED", comma ? ", " : ""); 2923 comma = 1; 2924 } 2925 if (t_flags & TF_SENTSYN) { 2926 db_printf("%sTF_SENTSYN", comma ? ", " : ""); 2927 comma = 1; 2928 } 2929 if (t_flags & TF_LRD) { 2930 db_printf("%sTF_LRD", comma ? ", " : ""); 2931 comma = 1; 2932 } 2933 if (t_flags & TF_CONGRECOVERY) { 2934 db_printf("%sTF_CONGRECOVERY", comma ? ", " : ""); 2935 comma = 1; 2936 } 2937 if (t_flags & TF_WASCRECOVERY) { 2938 db_printf("%sTF_WASCRECOVERY", comma ? ", " : ""); 2939 comma = 1; 2940 } 2941 if (t_flags & TF_FASTOPEN) { 2942 db_printf("%sTF_FASTOPEN", comma ? ", " : ""); 2943 comma = 1; 2944 } 2945 } 2946 2947 static void 2948 db_print_tflags2(u_int t_flags2) 2949 { 2950 int comma; 2951 2952 comma = 0; 2953 if (t_flags2 & TF2_PLPMTU_BLACKHOLE) { 2954 db_printf("%sTF2_PLPMTU_BLACKHOLE", comma ? ", " : ""); 2955 comma = 1; 2956 } 2957 if (t_flags2 & TF2_PLPMTU_PMTUD) { 2958 db_printf("%sTF2_PLPMTU_PMTUD", comma ? ", " : ""); 2959 comma = 1; 2960 } 2961 if (t_flags2 & TF2_PLPMTU_MAXSEGSNT) { 2962 db_printf("%sTF2_PLPMTU_MAXSEGSNT", comma ? ", " : ""); 2963 comma = 1; 2964 } 2965 if (t_flags2 & TF2_LOG_AUTO) { 2966 db_printf("%sTF2_LOG_AUTO", comma ? ", " : ""); 2967 comma = 1; 2968 } 2969 if (t_flags2 & TF2_DROP_AF_DATA) { 2970 db_printf("%sTF2_DROP_AF_DATA", comma ? ", " : ""); 2971 comma = 1; 2972 } 2973 if (t_flags2 & TF2_ECN_PERMIT) { 2974 db_printf("%sTF2_ECN_PERMIT", comma ? ", " : ""); 2975 comma = 1; 2976 } 2977 if (t_flags2 & TF2_ECN_SND_CWR) { 2978 db_printf("%sTF2_ECN_SND_CWR", comma ? ", " : ""); 2979 comma = 1; 2980 } 2981 if (t_flags2 & TF2_ECN_SND_ECE) { 2982 db_printf("%sTF2_ECN_SND_ECE", comma ? ", " : ""); 2983 comma = 1; 2984 } 2985 if (t_flags2 & TF2_ACE_PERMIT) { 2986 db_printf("%sTF2_ACE_PERMIT", comma ? ", " : ""); 2987 comma = 1; 2988 } 2989 if (t_flags2 & TF2_HPTS_CPU_SET) { 2990 db_printf("%sTF2_HPTS_CPU_SET", comma ? ", " : ""); 2991 comma = 1; 2992 } 2993 if (t_flags2 & TF2_FBYTES_COMPLETE) { 2994 db_printf("%sTF2_FBYTES_COMPLETE", comma ? ", " : ""); 2995 comma = 1; 2996 } 2997 if (t_flags2 & TF2_ECN_USE_ECT1) { 2998 db_printf("%sTF2_ECN_USE_ECT1", comma ? ", " : ""); 2999 comma = 1; 3000 } 3001 if (t_flags2 & TF2_TCP_ACCOUNTING) { 3002 db_printf("%sTF2_TCP_ACCOUNTING", comma ? ", " : ""); 3003 comma = 1; 3004 } 3005 if (t_flags2 & TF2_HPTS_CALLS) { 3006 db_printf("%sTF2_HPTS_CALLS", comma ? ", " : ""); 3007 comma = 1; 3008 } 3009 if (t_flags2 & TF2_MBUF_L_ACKS) { 3010 db_printf("%sTF2_MBUF_L_ACKS", comma ? ", " : ""); 3011 comma = 1; 3012 } 3013 if (t_flags2 & TF2_MBUF_ACKCMP) { 3014 db_printf("%sTF2_MBUF_ACKCMP", comma ? ", " : ""); 3015 comma = 1; 3016 } 3017 if (t_flags2 & TF2_SUPPORTS_MBUFQ) { 3018 db_printf("%sTF2_SUPPORTS_MBUFQ", comma ? ", " : ""); 3019 comma = 1; 3020 } 3021 if (t_flags2 & TF2_MBUF_QUEUE_READY) { 3022 db_printf("%sTF2_MBUF_QUEUE_READY", comma ? ", " : ""); 3023 comma = 1; 3024 } 3025 if (t_flags2 & TF2_DONT_SACK_QUEUE) { 3026 db_printf("%sTF2_DONT_SACK_QUEUE", comma ? ", " : ""); 3027 comma = 1; 3028 } 3029 if (t_flags2 & TF2_CANNOT_DO_ECN) { 3030 db_printf("%sTF2_CANNOT_DO_ECN", comma ? ", " : ""); 3031 comma = 1; 3032 } 3033 if (t_flags2 & TF2_PROC_SACK_PROHIBIT) { 3034 db_printf("%sTF2_PROC_SACK_PROHIBIT", comma ? ", " : ""); 3035 comma = 1; 3036 } 3037 if (t_flags2 & TF2_IPSEC_TSO) { 3038 db_printf("%sTF2_IPSEC_TSO", comma ? ", " : ""); 3039 comma = 1; 3040 } 3041 if (t_flags2 & TF2_NO_ISS_CHECK) { 3042 db_printf("%sTF2_NO_ISS_CHECK", comma ? ", " : ""); 3043 comma = 1; 3044 } 3045 } 3046 3047 static void 3048 db_print_toobflags(char t_oobflags) 3049 { 3050 int comma; 3051 3052 comma = 0; 3053 if (t_oobflags & TCPOOB_HAVEDATA) { 3054 db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : ""); 3055 comma = 1; 3056 } 3057 if (t_oobflags & TCPOOB_HADDATA) { 3058 db_printf("%sTCPOOB_HADDATA", comma ? ", " : ""); 3059 comma = 1; 3060 } 3061 } 3062 3063 static void 3064 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent) 3065 { 3066 3067 db_print_indent(indent); 3068 db_printf("%s at %p\n", name, tp); 3069 3070 indent += 2; 3071 3072 db_print_indent(indent); 3073 db_printf("t_segq first: %p t_segqlen: %d t_dupacks: %d\n", 3074 TAILQ_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks); 3075 3076 db_print_indent(indent); 3077 db_printf("t_callout: %p t_timers: %p\n", 3078 &tp->t_callout, &tp->t_timers); 3079 3080 db_print_indent(indent); 3081 db_printf("t_state: %d (", tp->t_state); 3082 db_print_tstate(tp->t_state); 3083 db_printf(")\n"); 3084 3085 db_print_indent(indent); 3086 db_printf("t_flags: 0x%x (", tp->t_flags); 3087 db_print_tflags(tp->t_flags); 3088 db_printf(")\n"); 3089 3090 db_print_indent(indent); 3091 db_printf("t_flags2: 0x%x (", tp->t_flags2); 3092 db_print_tflags2(tp->t_flags2); 3093 db_printf(")\n"); 3094 3095 db_print_indent(indent); 3096 db_printf("snd_una: 0x%08x snd_max: 0x%08x snd_nxt: 0x%08x\n", 3097 tp->snd_una, tp->snd_max, tp->snd_nxt); 3098 3099 db_print_indent(indent); 3100 db_printf("snd_up: 0x%08x snd_wl1: 0x%08x snd_wl2: 0x%08x\n", 3101 tp->snd_up, tp->snd_wl1, tp->snd_wl2); 3102 3103 db_print_indent(indent); 3104 db_printf("iss: 0x%08x irs: 0x%08x rcv_nxt: 0x%08x\n", 3105 tp->iss, tp->irs, tp->rcv_nxt); 3106 3107 db_print_indent(indent); 3108 db_printf("rcv_adv: 0x%08x rcv_wnd: %u rcv_up: 0x%08x\n", 3109 tp->rcv_adv, tp->rcv_wnd, tp->rcv_up); 3110 3111 db_print_indent(indent); 3112 db_printf("snd_wnd: %u snd_cwnd: %u\n", 3113 tp->snd_wnd, tp->snd_cwnd); 3114 3115 db_print_indent(indent); 3116 db_printf("snd_ssthresh: %u snd_recover: " 3117 "0x%08x\n", tp->snd_ssthresh, tp->snd_recover); 3118 3119 db_print_indent(indent); 3120 db_printf("t_rcvtime: %u t_startime: %u\n", 3121 tp->t_rcvtime, tp->t_starttime); 3122 3123 db_print_indent(indent); 3124 db_printf("t_rttime: %u t_rtsq: 0x%08x\n", 3125 tp->t_rtttime, tp->t_rtseq); 3126 3127 db_print_indent(indent); 3128 db_printf("t_rxtcur: %d t_maxseg: %u t_srtt: %d\n", 3129 tp->t_rxtcur, tp->t_maxseg, tp->t_srtt); 3130 3131 db_print_indent(indent); 3132 db_printf("t_rttvar: %d t_rxtshift: %d t_rttmin: %u\n", 3133 tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin); 3134 3135 db_print_indent(indent); 3136 db_printf("t_rttupdated: %u max_sndwnd: %u t_softerror: %d\n", 3137 tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror); 3138 3139 db_print_indent(indent); 3140 db_printf("t_oobflags: 0x%x (", tp->t_oobflags); 3141 db_print_toobflags(tp->t_oobflags); 3142 db_printf(") t_iobc: 0x%02x\n", tp->t_iobc); 3143 3144 db_print_indent(indent); 3145 db_printf("snd_scale: %u rcv_scale: %u request_r_scale: %u\n", 3146 tp->snd_scale, tp->rcv_scale, tp->request_r_scale); 3147 3148 db_print_indent(indent); 3149 db_printf("ts_recent: %u ts_recent_age: %u\n", 3150 tp->ts_recent, tp->ts_recent_age); 3151 3152 db_print_indent(indent); 3153 db_printf("ts_offset: %u last_ack_sent: 0x%08x snd_cwnd_prev: " 3154 "%u\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev); 3155 3156 db_print_indent(indent); 3157 db_printf("snd_ssthresh_prev: %u snd_recover_prev: 0x%08x " 3158 "t_badrxtwin: %u\n", tp->snd_ssthresh_prev, 3159 tp->snd_recover_prev, tp->t_badrxtwin); 3160 3161 db_print_indent(indent); 3162 db_printf("snd_numholes: %d snd_holes first: %p\n", 3163 tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes)); 3164 3165 db_print_indent(indent); 3166 db_printf("snd_fack: 0x%08x rcv_numsacks: %d\n", 3167 tp->snd_fack, tp->rcv_numsacks); 3168 3169 /* Skip sackblks, sackhint. */ 3170 3171 db_print_indent(indent); 3172 db_printf("t_rttlow: %d rfbuf_ts: %u rfbuf_cnt: %d\n", 3173 tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt); 3174 } 3175 3176 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb) 3177 { 3178 struct tcpcb *tp; 3179 3180 if (!have_addr) { 3181 db_printf("usage: show tcpcb <addr>\n"); 3182 return; 3183 } 3184 tp = (struct tcpcb *)addr; 3185 3186 db_print_tcpcb(tp, "tcpcb", 0); 3187 } 3188 #endif 3189