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 TCP_OFFLOAD 99 #include <netinet/tcp_offload.h> 100 #endif 101 #include <netipsec/ipsec_support.h> 102 103 #include <vm/vm.h> 104 #include <vm/vm_param.h> 105 #include <vm/pmap.h> 106 #include <vm/vm_extern.h> 107 #include <vm/vm_map.h> 108 #include <vm/vm_page.h> 109 110 /* 111 * TCP protocol interface to socket abstraction. 112 */ 113 #ifdef INET 114 static int tcp_connect(struct tcpcb *, struct sockaddr_in *, 115 struct thread *td); 116 #endif /* INET */ 117 #ifdef INET6 118 static int tcp6_connect(struct tcpcb *, struct sockaddr_in6 *, 119 struct thread *td); 120 #endif /* INET6 */ 121 static void tcp_disconnect(struct tcpcb *); 122 static void tcp_usrclosed(struct tcpcb *); 123 static void tcp_fill_info(const struct tcpcb *, struct tcp_info *); 124 125 static int tcp_pru_options_support(struct tcpcb *tp, int flags); 126 127 static void 128 tcp_bblog_pru(struct tcpcb *tp, uint32_t pru, int error) 129 { 130 struct tcp_log_buffer *lgb; 131 132 KASSERT(tp != NULL, ("tcp_bblog_pru: tp == NULL")); 133 INP_WLOCK_ASSERT(tptoinpcb(tp)); 134 if (tcp_bblogging_on(tp)) { 135 lgb = tcp_log_event(tp, NULL, NULL, NULL, TCP_LOG_PRU, error, 136 0, NULL, false, NULL, NULL, 0, NULL); 137 } else { 138 lgb = NULL; 139 } 140 if (lgb != NULL) { 141 if (error >= 0) { 142 lgb->tlb_errno = (uint32_t)error; 143 } 144 lgb->tlb_flex1 = pru; 145 } 146 } 147 148 /* 149 * TCP attaches to socket via pr_attach(), reserving space, 150 * and an internet control block. 151 */ 152 static int 153 tcp_usr_attach(struct socket *so, int proto, struct thread *td) 154 { 155 struct inpcb *inp; 156 struct tcpcb *tp = NULL; 157 int error; 158 159 inp = sotoinpcb(so); 160 KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL")); 161 162 error = soreserve(so, V_tcp_sendspace, V_tcp_recvspace); 163 if (error) 164 goto out; 165 166 so->so_rcv.sb_flags |= SB_AUTOSIZE; 167 so->so_snd.sb_flags |= (SB_AUTOLOWAT | SB_AUTOSIZE); 168 error = in_pcballoc(so, &V_tcbinfo); 169 if (error) 170 goto out; 171 inp = sotoinpcb(so); 172 tp = tcp_newtcpcb(inp, NULL); 173 if (tp == NULL) { 174 error = ENOBUFS; 175 in_pcbfree(inp); 176 goto out; 177 } 178 tp->t_state = TCPS_CLOSED; 179 tcp_bblog_pru(tp, PRU_ATTACH, error); 180 INP_WUNLOCK(inp); 181 TCPSTATES_INC(TCPS_CLOSED); 182 out: 183 TCP_PROBE2(debug__user, tp, PRU_ATTACH); 184 return (error); 185 } 186 187 /* 188 * tcp_usr_detach is called when the socket layer loses its final reference 189 * to the socket, be it a file descriptor reference, a reference from TCP, 190 * etc. At this point, there is only one case in which we will keep around 191 * inpcb state: time wait. 192 */ 193 static void 194 tcp_usr_detach(struct socket *so) 195 { 196 struct inpcb *inp; 197 struct tcpcb *tp; 198 199 inp = sotoinpcb(so); 200 KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); 201 INP_WLOCK(inp); 202 KASSERT(so->so_pcb == inp && inp->inp_socket == so, 203 ("%s: socket %p inp %p mismatch", __func__, so, inp)); 204 205 tp = intotcpcb(inp); 206 207 KASSERT(inp->inp_flags & INP_DROPPED || 208 tp->t_state < TCPS_SYN_SENT, 209 ("%s: inp %p not dropped or embryonic", __func__, inp)); 210 211 tcp_discardcb(tp); 212 in_pcbfree(inp); 213 } 214 215 #ifdef INET 216 /* 217 * Give the socket an address. 218 */ 219 static int 220 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 221 { 222 int error = 0; 223 struct inpcb *inp; 224 struct tcpcb *tp; 225 struct sockaddr_in *sinp; 226 227 inp = sotoinpcb(so); 228 KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL")); 229 INP_WLOCK(inp); 230 if (inp->inp_flags & INP_DROPPED) { 231 INP_WUNLOCK(inp); 232 return (EINVAL); 233 } 234 tp = intotcpcb(inp); 235 236 sinp = (struct sockaddr_in *)nam; 237 if (nam->sa_family != AF_INET) { 238 /* 239 * Preserve compatibility with old programs. 240 */ 241 if (nam->sa_family != AF_UNSPEC || 242 nam->sa_len < offsetof(struct sockaddr_in, sin_zero) || 243 sinp->sin_addr.s_addr != INADDR_ANY) { 244 error = EAFNOSUPPORT; 245 goto out; 246 } 247 nam->sa_family = AF_INET; 248 } 249 if (nam->sa_len != sizeof(*sinp)) { 250 error = EINVAL; 251 goto out; 252 } 253 /* 254 * Must check for multicast addresses and disallow binding 255 * to them. 256 */ 257 if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 258 error = EAFNOSUPPORT; 259 goto out; 260 } 261 INP_HASH_WLOCK(&V_tcbinfo); 262 error = in_pcbbind(inp, sinp, V_tcp_bind_all_fibs ? 0 : INPBIND_FIB, 263 td->td_ucred); 264 INP_HASH_WUNLOCK(&V_tcbinfo); 265 out: 266 tcp_bblog_pru(tp, PRU_BIND, error); 267 TCP_PROBE2(debug__user, tp, PRU_BIND); 268 INP_WUNLOCK(inp); 269 270 return (error); 271 } 272 #endif /* INET */ 273 274 #ifdef INET6 275 static int 276 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 277 { 278 int error = 0; 279 struct inpcb *inp; 280 struct tcpcb *tp; 281 struct sockaddr_in6 *sin6; 282 u_char vflagsav; 283 284 inp = sotoinpcb(so); 285 KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL")); 286 INP_WLOCK(inp); 287 if (inp->inp_flags & INP_DROPPED) { 288 INP_WUNLOCK(inp); 289 return (EINVAL); 290 } 291 tp = intotcpcb(inp); 292 293 vflagsav = inp->inp_vflag; 294 295 sin6 = (struct sockaddr_in6 *)nam; 296 if (nam->sa_family != AF_INET6) { 297 error = EAFNOSUPPORT; 298 goto out; 299 } 300 if (nam->sa_len != sizeof(*sin6)) { 301 error = EINVAL; 302 goto out; 303 } 304 /* 305 * Must check for multicast addresses and disallow binding 306 * to them. 307 */ 308 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { 309 error = EAFNOSUPPORT; 310 goto out; 311 } 312 313 INP_HASH_WLOCK(&V_tcbinfo); 314 inp->inp_vflag &= ~INP_IPV4; 315 inp->inp_vflag |= INP_IPV6; 316 #ifdef INET 317 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { 318 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) 319 inp->inp_vflag |= INP_IPV4; 320 else if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { 321 struct sockaddr_in sin; 322 323 in6_sin6_2_sin(&sin, sin6); 324 if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) { 325 error = EAFNOSUPPORT; 326 INP_HASH_WUNLOCK(&V_tcbinfo); 327 goto out; 328 } 329 inp->inp_vflag |= INP_IPV4; 330 inp->inp_vflag &= ~INP_IPV6; 331 error = in_pcbbind(inp, &sin, 0, td->td_ucred); 332 INP_HASH_WUNLOCK(&V_tcbinfo); 333 goto out; 334 } 335 } 336 #endif 337 error = in6_pcbbind(inp, sin6, V_tcp_bind_all_fibs ? 0 : INPBIND_FIB, 338 td->td_ucred); 339 INP_HASH_WUNLOCK(&V_tcbinfo); 340 out: 341 if (error != 0) 342 inp->inp_vflag = vflagsav; 343 tcp_bblog_pru(tp, PRU_BIND, error); 344 TCP_PROBE2(debug__user, tp, PRU_BIND); 345 INP_WUNLOCK(inp); 346 return (error); 347 } 348 #endif /* INET6 */ 349 350 #ifdef INET 351 /* 352 * Prepare to accept connections. 353 */ 354 static int 355 tcp_usr_listen(struct socket *so, int backlog, struct thread *td) 356 { 357 struct inpcb *inp; 358 struct tcpcb *tp; 359 int error = 0; 360 bool already_listening; 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 already_listening = SOLISTENING(so); 373 error = solisten_proto_check(so); 374 if (error != 0) { 375 SOCK_UNLOCK(so); 376 goto out; 377 } 378 if (inp->inp_lport == 0) { 379 INP_HASH_WLOCK(&V_tcbinfo); 380 error = in_pcbbind(inp, NULL, 381 V_tcp_bind_all_fibs ? 0 : INPBIND_FIB, td->td_ucred); 382 INP_HASH_WUNLOCK(&V_tcbinfo); 383 } 384 if (error == 0) { 385 tcp_state_change(tp, TCPS_LISTEN); 386 solisten_proto(so, backlog); 387 #ifdef TCP_OFFLOAD 388 if ((so->so_options & SO_NO_OFFLOAD) == 0) 389 tcp_offload_listen_start(tp); 390 #endif 391 } else { 392 solisten_proto_abort(so); 393 } 394 SOCK_UNLOCK(so); 395 if (already_listening) 396 goto out; 397 398 if (error == 0) 399 in_pcblisten(inp); 400 if (tp->t_flags & TF_FASTOPEN) 401 tp->t_tfo_pending = tcp_fastopen_alloc_counter(); 402 403 out: 404 tcp_bblog_pru(tp, PRU_LISTEN, error); 405 TCP_PROBE2(debug__user, tp, PRU_LISTEN); 406 INP_WUNLOCK(inp); 407 return (error); 408 } 409 #endif /* INET */ 410 411 #ifdef INET6 412 static int 413 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td) 414 { 415 struct inpcb *inp; 416 struct tcpcb *tp; 417 u_char vflagsav; 418 int error = 0; 419 bool already_listening; 420 421 inp = sotoinpcb(so); 422 KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL")); 423 INP_WLOCK(inp); 424 if (inp->inp_flags & INP_DROPPED) { 425 INP_WUNLOCK(inp); 426 return (EINVAL); 427 } 428 tp = intotcpcb(inp); 429 430 vflagsav = inp->inp_vflag; 431 432 SOCK_LOCK(so); 433 already_listening = SOLISTENING(so); 434 error = solisten_proto_check(so); 435 if (error != 0) { 436 SOCK_UNLOCK(so); 437 goto out; 438 } 439 INP_HASH_WLOCK(&V_tcbinfo); 440 if (inp->inp_lport == 0) { 441 inp->inp_vflag &= ~INP_IPV4; 442 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) 443 inp->inp_vflag |= INP_IPV4; 444 error = in6_pcbbind(inp, NULL, 445 V_tcp_bind_all_fibs ? 0 : INPBIND_FIB, td->td_ucred); 446 } 447 INP_HASH_WUNLOCK(&V_tcbinfo); 448 if (error == 0) { 449 tcp_state_change(tp, TCPS_LISTEN); 450 solisten_proto(so, backlog); 451 #ifdef TCP_OFFLOAD 452 if ((so->so_options & SO_NO_OFFLOAD) == 0) 453 tcp_offload_listen_start(tp); 454 #endif 455 } else { 456 solisten_proto_abort(so); 457 } 458 SOCK_UNLOCK(so); 459 if (already_listening) 460 goto out; 461 462 if (error == 0) 463 in_pcblisten(inp); 464 if (tp->t_flags & TF_FASTOPEN) 465 tp->t_tfo_pending = tcp_fastopen_alloc_counter(); 466 467 if (error != 0) 468 inp->inp_vflag = vflagsav; 469 470 out: 471 tcp_bblog_pru(tp, PRU_LISTEN, error); 472 TCP_PROBE2(debug__user, tp, PRU_LISTEN); 473 INP_WUNLOCK(inp); 474 return (error); 475 } 476 #endif /* INET6 */ 477 478 #ifdef INET 479 /* 480 * Initiate connection to peer. 481 * Create a template for use in transmissions on this connection. 482 * Enter SYN_SENT state, and mark socket as connecting. 483 * Start keep-alive timer, and seed output sequence space. 484 * Send initial segment on connection. 485 */ 486 static int 487 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 488 { 489 struct epoch_tracker et; 490 int error = 0; 491 struct inpcb *inp; 492 struct tcpcb *tp; 493 struct sockaddr_in *sinp; 494 495 inp = sotoinpcb(so); 496 KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL")); 497 INP_WLOCK(inp); 498 if (inp->inp_flags & INP_DROPPED) { 499 INP_WUNLOCK(inp); 500 return (ECONNREFUSED); 501 } 502 tp = intotcpcb(inp); 503 504 sinp = (struct sockaddr_in *)nam; 505 if (nam->sa_family != AF_INET) { 506 error = EAFNOSUPPORT; 507 goto out; 508 } 509 if (nam->sa_len != sizeof (*sinp)) { 510 error = EINVAL; 511 goto out; 512 } 513 /* 514 * Must disallow TCP ``connections'' to multicast addresses. 515 */ 516 if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 517 error = EAFNOSUPPORT; 518 goto out; 519 } 520 if (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST) { 521 error = EACCES; 522 goto out; 523 } 524 if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0) 525 goto out; 526 if (SOLISTENING(so)) { 527 error = EOPNOTSUPP; 528 goto out; 529 } 530 NET_EPOCH_ENTER(et); 531 if ((error = tcp_connect(tp, sinp, td)) != 0) 532 goto out_in_epoch; 533 #ifdef TCP_OFFLOAD 534 if (registered_toedevs > 0 && 535 (so->so_options & SO_NO_OFFLOAD) == 0 && 536 (error = tcp_offload_connect(so, nam)) == 0) 537 goto out_in_epoch; 538 #endif 539 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp)); 540 error = tcp_output(tp); 541 KASSERT(error >= 0, ("TCP stack %s requested tcp_drop(%p) at connect()" 542 ", error code %d", tp->t_fb->tfb_tcp_block_name, tp, -error)); 543 out_in_epoch: 544 NET_EPOCH_EXIT(et); 545 out: 546 tcp_bblog_pru(tp, PRU_CONNECT, error); 547 TCP_PROBE2(debug__user, tp, PRU_CONNECT); 548 INP_WUNLOCK(inp); 549 return (error); 550 } 551 #endif /* INET */ 552 553 #ifdef INET6 554 static int 555 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 556 { 557 struct epoch_tracker et; 558 int error = 0; 559 struct inpcb *inp; 560 struct tcpcb *tp; 561 struct sockaddr_in6 *sin6; 562 u_int8_t incflagsav; 563 u_char vflagsav; 564 565 inp = sotoinpcb(so); 566 KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL")); 567 INP_WLOCK(inp); 568 if (inp->inp_flags & INP_DROPPED) { 569 INP_WUNLOCK(inp); 570 return (ECONNREFUSED); 571 } 572 tp = intotcpcb(inp); 573 574 vflagsav = inp->inp_vflag; 575 incflagsav = inp->inp_inc.inc_flags; 576 577 sin6 = (struct sockaddr_in6 *)nam; 578 if (nam->sa_family != AF_INET6) { 579 error = EAFNOSUPPORT; 580 goto out; 581 } 582 if (nam->sa_len != sizeof (*sin6)) { 583 error = EINVAL; 584 goto out; 585 } 586 /* 587 * Must disallow TCP ``connections'' to multicast addresses. 588 */ 589 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { 590 error = EAFNOSUPPORT; 591 goto out; 592 } 593 if (SOLISTENING(so)) { 594 error = EOPNOTSUPP; 595 goto out; 596 } 597 #ifdef INET 598 /* 599 * XXXRW: Some confusion: V4/V6 flags relate to binding, and 600 * therefore probably require the hash lock, which isn't held here. 601 * Is this a significant problem? 602 */ 603 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { 604 struct sockaddr_in sin; 605 606 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { 607 error = EINVAL; 608 goto out; 609 } 610 if ((inp->inp_vflag & INP_IPV4) == 0) { 611 error = EAFNOSUPPORT; 612 goto out; 613 } 614 615 in6_sin6_2_sin(&sin, sin6); 616 if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) { 617 error = EAFNOSUPPORT; 618 goto out; 619 } 620 if (ntohl(sin.sin_addr.s_addr) == INADDR_BROADCAST) { 621 error = EACCES; 622 goto out; 623 } 624 if ((error = prison_remote_ip4(td->td_ucred, 625 &sin.sin_addr)) != 0) 626 goto out; 627 inp->inp_vflag |= INP_IPV4; 628 inp->inp_vflag &= ~INP_IPV6; 629 NET_EPOCH_ENTER(et); 630 if ((error = tcp_connect(tp, &sin, td)) != 0) 631 goto out_in_epoch; 632 #ifdef TCP_OFFLOAD 633 if (registered_toedevs > 0 && 634 (so->so_options & SO_NO_OFFLOAD) == 0 && 635 (error = tcp_offload_connect(so, nam)) == 0) 636 goto out_in_epoch; 637 #endif 638 error = tcp_output(tp); 639 goto out_in_epoch; 640 } else { 641 if ((inp->inp_vflag & INP_IPV6) == 0) { 642 error = EAFNOSUPPORT; 643 goto out; 644 } 645 } 646 #endif 647 if ((error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr)) != 0) 648 goto out; 649 inp->inp_vflag &= ~INP_IPV4; 650 inp->inp_vflag |= INP_IPV6; 651 inp->inp_inc.inc_flags |= INC_ISIPV6; 652 NET_EPOCH_ENTER(et); 653 if ((error = tcp6_connect(tp, sin6, td)) != 0) 654 goto out_in_epoch; 655 #ifdef TCP_OFFLOAD 656 if (registered_toedevs > 0 && 657 (so->so_options & SO_NO_OFFLOAD) == 0 && 658 (error = tcp_offload_connect(so, nam)) == 0) 659 goto out_in_epoch; 660 #endif 661 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp)); 662 error = tcp_output(tp); 663 out_in_epoch: 664 NET_EPOCH_EXIT(et); 665 out: 666 KASSERT(error >= 0, ("TCP stack %s requested tcp_drop(%p) at connect()" 667 ", error code %d", tp->t_fb->tfb_tcp_block_name, tp, -error)); 668 /* 669 * If the implicit bind in the connect call fails, restore 670 * the flags we modified. 671 */ 672 if (error != 0 && inp->inp_lport == 0) { 673 inp->inp_vflag = vflagsav; 674 inp->inp_inc.inc_flags = incflagsav; 675 } 676 677 tcp_bblog_pru(tp, PRU_CONNECT, error); 678 TCP_PROBE2(debug__user, tp, PRU_CONNECT); 679 INP_WUNLOCK(inp); 680 return (error); 681 } 682 #endif /* INET6 */ 683 684 /* 685 * Initiate disconnect from peer. 686 * If connection never passed embryonic stage, just drop; 687 * else if don't need to let data drain, then can just drop anyways, 688 * else have to begin TCP shutdown process: mark socket disconnecting, 689 * drain unread data, state switch to reflect user close, and 690 * send segment (e.g. FIN) to peer. Socket will be really disconnected 691 * when peer sends FIN and acks ours. 692 * 693 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 694 */ 695 static int 696 tcp_usr_disconnect(struct socket *so) 697 { 698 struct inpcb *inp; 699 struct tcpcb *tp = NULL; 700 struct epoch_tracker et; 701 702 NET_EPOCH_ENTER(et); 703 inp = sotoinpcb(so); 704 KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL")); 705 INP_WLOCK(inp); 706 tp = intotcpcb(inp); 707 708 if (tp->t_state == TCPS_TIME_WAIT) 709 goto out; 710 tcp_disconnect(tp); 711 out: 712 tcp_bblog_pru(tp, PRU_DISCONNECT, 0); 713 TCP_PROBE2(debug__user, tp, PRU_DISCONNECT); 714 INP_WUNLOCK(inp); 715 NET_EPOCH_EXIT(et); 716 return (0); 717 } 718 719 #ifdef INET 720 /* 721 * Accept a connection. Essentially all the work is done at higher levels; 722 * just return the address of the peer, storing through addr. 723 */ 724 static int 725 tcp_usr_accept(struct socket *so, struct sockaddr *sa) 726 { 727 struct inpcb *inp; 728 struct tcpcb *tp; 729 int error = 0; 730 731 inp = sotoinpcb(so); 732 KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL")); 733 INP_WLOCK(inp); 734 if (inp->inp_flags & INP_DROPPED) { 735 INP_WUNLOCK(inp); 736 return (ECONNABORTED); 737 } 738 tp = intotcpcb(inp); 739 740 if (so->so_state & SS_ISDISCONNECTED) 741 error = ECONNABORTED; 742 else 743 *(struct sockaddr_in *)sa = (struct sockaddr_in ){ 744 .sin_family = AF_INET, 745 .sin_len = sizeof(struct sockaddr_in), 746 .sin_port = inp->inp_fport, 747 .sin_addr = inp->inp_faddr, 748 }; 749 tcp_bblog_pru(tp, PRU_ACCEPT, error); 750 TCP_PROBE2(debug__user, tp, PRU_ACCEPT); 751 INP_WUNLOCK(inp); 752 753 return (error); 754 } 755 #endif /* INET */ 756 757 #ifdef INET6 758 static int 759 tcp6_usr_accept(struct socket *so, struct sockaddr *sa) 760 { 761 struct inpcb *inp; 762 struct tcpcb *tp; 763 int error = 0; 764 765 inp = sotoinpcb(so); 766 KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL")); 767 INP_WLOCK(inp); 768 if (inp->inp_flags & INP_DROPPED) { 769 INP_WUNLOCK(inp); 770 return (ECONNABORTED); 771 } 772 tp = intotcpcb(inp); 773 774 if (so->so_state & SS_ISDISCONNECTED) { 775 error = ECONNABORTED; 776 } else { 777 if (inp->inp_vflag & INP_IPV4) { 778 struct sockaddr_in sin = { 779 .sin_family = AF_INET, 780 .sin_len = sizeof(struct sockaddr_in), 781 .sin_port = inp->inp_fport, 782 .sin_addr = inp->inp_faddr, 783 }; 784 in6_sin_2_v4mapsin6(&sin, (struct sockaddr_in6 *)sa); 785 } else { 786 *(struct sockaddr_in6 *)sa = (struct sockaddr_in6 ){ 787 .sin6_family = AF_INET6, 788 .sin6_len = sizeof(struct sockaddr_in6), 789 .sin6_port = inp->inp_fport, 790 .sin6_addr = inp->in6p_faddr, 791 }; 792 /* XXX: should catch errors */ 793 (void)sa6_recoverscope((struct sockaddr_in6 *)sa); 794 } 795 } 796 797 tcp_bblog_pru(tp, PRU_ACCEPT, error); 798 TCP_PROBE2(debug__user, tp, PRU_ACCEPT); 799 INP_WUNLOCK(inp); 800 801 return (error); 802 } 803 #endif /* INET6 */ 804 805 /* 806 * Mark the connection as being incapable of further output. 807 */ 808 static int 809 tcp_usr_shutdown(struct socket *so, enum shutdown_how how) 810 { 811 struct epoch_tracker et; 812 struct inpcb *inp = sotoinpcb(so); 813 struct tcpcb *tp = intotcpcb(inp); 814 int error = 0; 815 816 SOCK_LOCK(so); 817 if (SOLISTENING(so)) { 818 if (how != SHUT_WR) { 819 so->so_error = ECONNABORTED; 820 solisten_wakeup(so); /* unlocks so */ 821 } else 822 SOCK_UNLOCK(so); 823 return (ENOTCONN); 824 } else if ((so->so_state & 825 (SS_ISCONNECTED | SS_ISCONNECTING | SS_ISDISCONNECTING)) == 0) { 826 SOCK_UNLOCK(so); 827 return (ENOTCONN); 828 } 829 SOCK_UNLOCK(so); 830 831 switch (how) { 832 case SHUT_RD: 833 sorflush(so); 834 break; 835 case SHUT_RDWR: 836 sorflush(so); 837 /* FALLTHROUGH */ 838 case SHUT_WR: 839 /* 840 * XXXGL: mimicing old soshutdown() here. But shouldn't we 841 * return ECONNRESEST for SHUT_RD as well? 842 */ 843 INP_WLOCK(inp); 844 if (inp->inp_flags & INP_DROPPED) { 845 INP_WUNLOCK(inp); 846 return (ECONNRESET); 847 } 848 849 socantsendmore(so); 850 NET_EPOCH_ENTER(et); 851 tcp_usrclosed(tp); 852 error = tcp_output_nodrop(tp); 853 tcp_bblog_pru(tp, PRU_SHUTDOWN, error); 854 TCP_PROBE2(debug__user, tp, PRU_SHUTDOWN); 855 error = tcp_unlock_or_drop(tp, error); 856 NET_EPOCH_EXIT(et); 857 } 858 wakeup(&so->so_timeo); 859 860 return (error); 861 } 862 863 /* 864 * After a receive, possibly send window update to peer. 865 */ 866 static int 867 tcp_usr_rcvd(struct socket *so, int flags) 868 { 869 struct epoch_tracker et; 870 struct inpcb *inp; 871 struct tcpcb *tp; 872 int outrv = 0, error = 0; 873 874 inp = sotoinpcb(so); 875 KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL")); 876 INP_WLOCK(inp); 877 if (inp->inp_flags & INP_DROPPED) { 878 INP_WUNLOCK(inp); 879 return (ECONNRESET); 880 } 881 tp = intotcpcb(inp); 882 883 NET_EPOCH_ENTER(et); 884 /* 885 * For passively-created TFO connections, don't attempt a window 886 * update while still in SYN_RECEIVED as this may trigger an early 887 * SYN|ACK. It is preferable to have the SYN|ACK be sent along with 888 * application response data, or failing that, when the DELACK timer 889 * expires. 890 */ 891 if ((tp->t_flags & TF_FASTOPEN) && (tp->t_state == TCPS_SYN_RECEIVED)) 892 goto out; 893 #ifdef TCP_OFFLOAD 894 if (tp->t_flags & TF_TOE) 895 tcp_offload_rcvd(tp); 896 else 897 #endif 898 outrv = tcp_output_nodrop(tp); 899 out: 900 tcp_bblog_pru(tp, PRU_RCVD, error); 901 TCP_PROBE2(debug__user, tp, PRU_RCVD); 902 (void) tcp_unlock_or_drop(tp, outrv); 903 NET_EPOCH_EXIT(et); 904 return (error); 905 } 906 907 /* 908 * Do a send by putting data in output queue and updating urgent 909 * marker if URG set. Possibly send more data. Unlike the other 910 * pr_*() routines, the mbuf chains are our responsibility. We 911 * must either enqueue them or free them. The other pr_*() routines 912 * generally are caller-frees. 913 */ 914 static int 915 tcp_usr_send(struct socket *so, int flags, struct mbuf *m, 916 struct sockaddr *nam, struct mbuf *control, struct thread *td) 917 { 918 struct epoch_tracker et; 919 int error = 0; 920 struct inpcb *inp; 921 struct tcpcb *tp; 922 #ifdef INET 923 #ifdef INET6 924 struct sockaddr_in sin; 925 #endif 926 struct sockaddr_in *sinp; 927 #endif 928 #ifdef INET6 929 struct sockaddr_in6 *sin6; 930 int isipv6; 931 #endif 932 u_int8_t incflagsav; 933 u_char vflagsav; 934 bool restoreflags; 935 936 inp = sotoinpcb(so); 937 KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL")); 938 INP_WLOCK(inp); 939 if (inp->inp_flags & INP_DROPPED) { 940 if (m != NULL && (flags & PRUS_NOTREADY) == 0) 941 m_freem(m); 942 INP_WUNLOCK(inp); 943 return (ECONNRESET); 944 } 945 tp = intotcpcb(inp); 946 947 vflagsav = inp->inp_vflag; 948 incflagsav = inp->inp_inc.inc_flags; 949 restoreflags = false; 950 951 NET_EPOCH_ENTER(et); 952 if (control != NULL) { 953 /* TCP doesn't do control messages (rights, creds, etc) */ 954 if (control->m_len > 0) { 955 m_freem(control); 956 error = EINVAL; 957 goto out; 958 } 959 m_freem(control); /* empty control, just free it */ 960 } 961 962 if ((flags & PRUS_OOB) != 0 && 963 (error = tcp_pru_options_support(tp, PRUS_OOB)) != 0) 964 goto out; 965 966 if (nam != NULL && tp->t_state < TCPS_SYN_SENT) { 967 if (tp->t_state == TCPS_LISTEN) { 968 error = EINVAL; 969 goto out; 970 } 971 switch (nam->sa_family) { 972 #ifdef INET 973 case AF_INET: 974 sinp = (struct sockaddr_in *)nam; 975 if (sinp->sin_len != sizeof(struct sockaddr_in)) { 976 error = EINVAL; 977 goto out; 978 } 979 if ((inp->inp_vflag & INP_IPV6) != 0) { 980 error = EAFNOSUPPORT; 981 goto out; 982 } 983 if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 984 error = EAFNOSUPPORT; 985 goto out; 986 } 987 if (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST) { 988 error = EACCES; 989 goto out; 990 } 991 if ((error = prison_remote_ip4(td->td_ucred, 992 &sinp->sin_addr))) 993 goto out; 994 #ifdef INET6 995 isipv6 = 0; 996 #endif 997 break; 998 #endif /* INET */ 999 #ifdef INET6 1000 case AF_INET6: 1001 sin6 = (struct sockaddr_in6 *)nam; 1002 if (sin6->sin6_len != sizeof(*sin6)) { 1003 error = EINVAL; 1004 goto out; 1005 } 1006 if ((inp->inp_vflag & INP_IPV6PROTO) == 0) { 1007 error = EAFNOSUPPORT; 1008 goto out; 1009 } 1010 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { 1011 error = EAFNOSUPPORT; 1012 goto out; 1013 } 1014 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { 1015 #ifdef INET 1016 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { 1017 error = EINVAL; 1018 goto out; 1019 } 1020 if ((inp->inp_vflag & INP_IPV4) == 0) { 1021 error = EAFNOSUPPORT; 1022 goto out; 1023 } 1024 restoreflags = true; 1025 inp->inp_vflag &= ~INP_IPV6; 1026 sinp = &sin; 1027 in6_sin6_2_sin(sinp, sin6); 1028 if (IN_MULTICAST( 1029 ntohl(sinp->sin_addr.s_addr))) { 1030 error = EAFNOSUPPORT; 1031 goto out; 1032 } 1033 if ((error = prison_remote_ip4(td->td_ucred, 1034 &sinp->sin_addr))) 1035 goto out; 1036 isipv6 = 0; 1037 #else /* !INET */ 1038 error = EAFNOSUPPORT; 1039 goto out; 1040 #endif /* INET */ 1041 } else { 1042 if ((inp->inp_vflag & INP_IPV6) == 0) { 1043 error = EAFNOSUPPORT; 1044 goto out; 1045 } 1046 restoreflags = true; 1047 inp->inp_vflag &= ~INP_IPV4; 1048 inp->inp_inc.inc_flags |= INC_ISIPV6; 1049 if ((error = prison_remote_ip6(td->td_ucred, 1050 &sin6->sin6_addr))) 1051 goto out; 1052 isipv6 = 1; 1053 } 1054 break; 1055 #endif /* INET6 */ 1056 default: 1057 error = EAFNOSUPPORT; 1058 goto out; 1059 } 1060 } 1061 if (!(flags & PRUS_OOB)) { 1062 if (tp->t_acktime == 0) 1063 tp->t_acktime = ticks; 1064 sbappendstream(&so->so_snd, m, flags); 1065 m = NULL; 1066 if (nam && tp->t_state < TCPS_SYN_SENT) { 1067 KASSERT(tp->t_state == TCPS_CLOSED, 1068 ("%s: tp %p is listening", __func__, tp)); 1069 1070 /* 1071 * Do implied connect if not yet connected, 1072 * initialize window to default value, and 1073 * initialize maxseg using peer's cached MSS. 1074 */ 1075 #ifdef INET6 1076 if (isipv6) 1077 error = tcp6_connect(tp, sin6, td); 1078 #endif /* INET6 */ 1079 #if defined(INET6) && defined(INET) 1080 else 1081 #endif 1082 #ifdef INET 1083 error = tcp_connect(tp, sinp, td); 1084 #endif 1085 /* 1086 * The bind operation in tcp_connect succeeded. We 1087 * no longer want to restore the flags if later 1088 * operations fail. 1089 */ 1090 if (error == 0 || inp->inp_lport != 0) 1091 restoreflags = false; 1092 1093 if (error) { 1094 /* m is freed if PRUS_NOTREADY is unset. */ 1095 sbflush(&so->so_snd); 1096 goto out; 1097 } 1098 if (tp->t_flags & TF_FASTOPEN) 1099 tcp_fastopen_connect(tp); 1100 else { 1101 tp->snd_wnd = TTCP_CLIENT_SND_WND; 1102 tcp_mss(tp, -1); 1103 } 1104 } 1105 if (flags & PRUS_EOF) { 1106 /* 1107 * Close the send side of the connection after 1108 * the data is sent. 1109 */ 1110 socantsendmore(so); 1111 tcp_usrclosed(tp); 1112 } 1113 if (TCPS_HAVEESTABLISHED(tp->t_state) && 1114 ((tp->t_flags2 & TF2_FBYTES_COMPLETE) == 0) && 1115 (tp->t_fbyte_out == 0) && 1116 (so->so_snd.sb_ccc > 0)) { 1117 tp->t_fbyte_out = ticks; 1118 if (tp->t_fbyte_out == 0) 1119 tp->t_fbyte_out = 1; 1120 if (tp->t_fbyte_out && tp->t_fbyte_in) 1121 tp->t_flags2 |= TF2_FBYTES_COMPLETE; 1122 } 1123 if (!(inp->inp_flags & INP_DROPPED) && 1124 !(flags & PRUS_NOTREADY)) { 1125 if (flags & PRUS_MORETOCOME) 1126 tp->t_flags |= TF_MORETOCOME; 1127 error = tcp_output_nodrop(tp); 1128 if (flags & PRUS_MORETOCOME) 1129 tp->t_flags &= ~TF_MORETOCOME; 1130 } 1131 } else { 1132 /* 1133 * XXXRW: PRUS_EOF not implemented with PRUS_OOB? 1134 */ 1135 SOCK_SENDBUF_LOCK(so); 1136 if (sbspace(&so->so_snd) < -512) { 1137 SOCK_SENDBUF_UNLOCK(so); 1138 error = ENOBUFS; 1139 goto out; 1140 } 1141 /* 1142 * According to RFC961 (Assigned Protocols), 1143 * the urgent pointer points to the last octet 1144 * of urgent data. We continue, however, 1145 * to consider it to indicate the first octet 1146 * of data past the urgent section. 1147 * Otherwise, snd_up should be one lower. 1148 */ 1149 if (tp->t_acktime == 0) 1150 tp->t_acktime = ticks; 1151 sbappendstream_locked(&so->so_snd, m, flags); 1152 SOCK_SENDBUF_UNLOCK(so); 1153 m = NULL; 1154 if (nam && tp->t_state < TCPS_SYN_SENT) { 1155 /* 1156 * Do implied connect if not yet connected, 1157 * initialize window to default value, and 1158 * initialize maxseg using peer's cached MSS. 1159 */ 1160 1161 /* 1162 * Not going to contemplate SYN|URG 1163 */ 1164 if (tp->t_flags & TF_FASTOPEN) 1165 tp->t_flags &= ~TF_FASTOPEN; 1166 #ifdef INET6 1167 if (isipv6) 1168 error = tcp6_connect(tp, sin6, td); 1169 #endif /* INET6 */ 1170 #if defined(INET6) && defined(INET) 1171 else 1172 #endif 1173 #ifdef INET 1174 error = tcp_connect(tp, sinp, td); 1175 #endif 1176 /* 1177 * The bind operation in tcp_connect succeeded. We 1178 * no longer want to restore the flags if later 1179 * operations fail. 1180 */ 1181 if (error == 0 || inp->inp_lport != 0) 1182 restoreflags = false; 1183 1184 if (error != 0) { 1185 /* m is freed if PRUS_NOTREADY is unset. */ 1186 sbflush(&so->so_snd); 1187 goto out; 1188 } 1189 tp->snd_wnd = TTCP_CLIENT_SND_WND; 1190 tcp_mss(tp, -1); 1191 } 1192 tp->snd_up = tp->snd_una + sbavail(&so->so_snd); 1193 if ((flags & PRUS_NOTREADY) == 0) { 1194 tp->t_flags |= TF_FORCEDATA; 1195 error = tcp_output_nodrop(tp); 1196 tp->t_flags &= ~TF_FORCEDATA; 1197 } 1198 } 1199 TCP_LOG_EVENT(tp, NULL, 1200 &inp->inp_socket->so_rcv, 1201 &inp->inp_socket->so_snd, 1202 TCP_LOG_USERSEND, error, 1203 0, NULL, false); 1204 1205 out: 1206 /* 1207 * In case of PRUS_NOTREADY, the caller or tcp_usr_ready() is 1208 * responsible for freeing memory. 1209 */ 1210 if (m != NULL && (flags & PRUS_NOTREADY) == 0) 1211 m_freem(m); 1212 1213 /* 1214 * If the request was unsuccessful and we changed flags, 1215 * restore the original flags. 1216 */ 1217 if (error != 0 && restoreflags) { 1218 inp->inp_vflag = vflagsav; 1219 inp->inp_inc.inc_flags = incflagsav; 1220 } 1221 tcp_bblog_pru(tp, (flags & PRUS_OOB) ? PRU_SENDOOB : 1222 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND), error); 1223 TCP_PROBE2(debug__user, tp, (flags & PRUS_OOB) ? PRU_SENDOOB : 1224 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); 1225 error = tcp_unlock_or_drop(tp, error); 1226 NET_EPOCH_EXIT(et); 1227 return (error); 1228 } 1229 1230 static int 1231 tcp_usr_ready(struct socket *so, struct mbuf *m, int count) 1232 { 1233 struct epoch_tracker et; 1234 struct inpcb *inp; 1235 struct tcpcb *tp; 1236 int error; 1237 1238 inp = sotoinpcb(so); 1239 INP_WLOCK(inp); 1240 if (inp->inp_flags & INP_DROPPED) { 1241 INP_WUNLOCK(inp); 1242 mb_free_notready(m, count); 1243 return (ECONNRESET); 1244 } 1245 tp = intotcpcb(inp); 1246 1247 SOCK_SENDBUF_LOCK(so); 1248 error = sbready(&so->so_snd, m, count); 1249 SOCK_SENDBUF_UNLOCK(so); 1250 if (error) { 1251 INP_WUNLOCK(inp); 1252 return (error); 1253 } 1254 NET_EPOCH_ENTER(et); 1255 error = tcp_output_unlock(tp); 1256 NET_EPOCH_EXIT(et); 1257 1258 return (error); 1259 } 1260 1261 /* 1262 * Abort the TCP. Drop the connection abruptly. 1263 */ 1264 static void 1265 tcp_usr_abort(struct socket *so) 1266 { 1267 struct inpcb *inp; 1268 struct tcpcb *tp; 1269 struct epoch_tracker et; 1270 1271 inp = sotoinpcb(so); 1272 KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL")); 1273 1274 NET_EPOCH_ENTER(et); 1275 INP_WLOCK(inp); 1276 KASSERT(inp->inp_socket != NULL, 1277 ("tcp_usr_abort: inp_socket == NULL")); 1278 1279 /* 1280 * If we still have full TCP state, and we're not dropped, drop. 1281 */ 1282 if (!(inp->inp_flags & INP_DROPPED)) { 1283 tp = intotcpcb(inp); 1284 tp = tcp_drop(tp, ECONNABORTED); 1285 if (tp == NULL) 1286 goto dropped; 1287 tcp_bblog_pru(tp, PRU_ABORT, 0); 1288 TCP_PROBE2(debug__user, tp, PRU_ABORT); 1289 } 1290 if (!(inp->inp_flags & INP_DROPPED)) { 1291 soref(so); 1292 inp->inp_flags |= INP_SOCKREF; 1293 } 1294 INP_WUNLOCK(inp); 1295 dropped: 1296 NET_EPOCH_EXIT(et); 1297 } 1298 1299 /* 1300 * TCP socket is closed. Start friendly disconnect. 1301 */ 1302 static void 1303 tcp_usr_close(struct socket *so) 1304 { 1305 struct inpcb *inp; 1306 struct tcpcb *tp; 1307 struct epoch_tracker et; 1308 1309 inp = sotoinpcb(so); 1310 KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL")); 1311 1312 NET_EPOCH_ENTER(et); 1313 INP_WLOCK(inp); 1314 KASSERT(inp->inp_socket != NULL, 1315 ("tcp_usr_close: inp_socket == NULL")); 1316 1317 /* 1318 * If we are still connected and we're not dropped, initiate 1319 * a disconnect. 1320 */ 1321 if (!(inp->inp_flags & INP_DROPPED)) { 1322 tp = intotcpcb(inp); 1323 if (tp->t_state != TCPS_TIME_WAIT) { 1324 tp->t_flags |= TF_CLOSED; 1325 tcp_disconnect(tp); 1326 tcp_bblog_pru(tp, PRU_CLOSE, 0); 1327 TCP_PROBE2(debug__user, tp, PRU_CLOSE); 1328 } 1329 } 1330 if (!(inp->inp_flags & INP_DROPPED)) { 1331 soref(so); 1332 inp->inp_flags |= INP_SOCKREF; 1333 } 1334 INP_WUNLOCK(inp); 1335 NET_EPOCH_EXIT(et); 1336 } 1337 1338 static int 1339 tcp_pru_options_support(struct tcpcb *tp, int flags) 1340 { 1341 /* 1342 * If the specific TCP stack has a pru_options 1343 * specified then it does not always support 1344 * all the PRU_XX options and we must ask it. 1345 * If the function is not specified then all 1346 * of the PRU_XX options are supported. 1347 */ 1348 int ret = 0; 1349 1350 if (tp->t_fb->tfb_pru_options) { 1351 ret = (*tp->t_fb->tfb_pru_options)(tp, flags); 1352 } 1353 return (ret); 1354 } 1355 1356 /* 1357 * Receive out-of-band data. 1358 */ 1359 static int 1360 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) 1361 { 1362 int error = 0; 1363 struct inpcb *inp; 1364 struct tcpcb *tp; 1365 1366 inp = sotoinpcb(so); 1367 KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL")); 1368 INP_WLOCK(inp); 1369 if (inp->inp_flags & INP_DROPPED) { 1370 INP_WUNLOCK(inp); 1371 return (ECONNRESET); 1372 } 1373 tp = intotcpcb(inp); 1374 1375 error = tcp_pru_options_support(tp, PRUS_OOB); 1376 if (error) { 1377 goto out; 1378 } 1379 if ((so->so_oobmark == 0 && 1380 (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) || 1381 so->so_options & SO_OOBINLINE || 1382 tp->t_oobflags & TCPOOB_HADDATA) { 1383 error = EINVAL; 1384 goto out; 1385 } 1386 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 1387 error = EWOULDBLOCK; 1388 goto out; 1389 } 1390 m->m_len = 1; 1391 *mtod(m, caddr_t) = tp->t_iobc; 1392 if ((flags & MSG_PEEK) == 0) 1393 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 1394 1395 out: 1396 tcp_bblog_pru(tp, PRU_RCVOOB, error); 1397 TCP_PROBE2(debug__user, tp, PRU_RCVOOB); 1398 INP_WUNLOCK(inp); 1399 return (error); 1400 } 1401 1402 #ifdef INET 1403 struct protosw tcp_protosw = { 1404 .pr_type = SOCK_STREAM, 1405 .pr_protocol = IPPROTO_TCP, 1406 .pr_flags = PR_CONNREQUIRED | PR_IMPLOPCL | PR_WANTRCVD | 1407 PR_CAPATTACH, 1408 .pr_ctloutput = tcp_ctloutput, 1409 .pr_abort = tcp_usr_abort, 1410 .pr_accept = tcp_usr_accept, 1411 .pr_attach = tcp_usr_attach, 1412 .pr_bind = tcp_usr_bind, 1413 .pr_connect = tcp_usr_connect, 1414 .pr_control = in_control, 1415 .pr_detach = tcp_usr_detach, 1416 .pr_disconnect = tcp_usr_disconnect, 1417 .pr_listen = tcp_usr_listen, 1418 .pr_peeraddr = in_getpeeraddr, 1419 .pr_rcvd = tcp_usr_rcvd, 1420 .pr_rcvoob = tcp_usr_rcvoob, 1421 .pr_send = tcp_usr_send, 1422 .pr_sendfile_wait = sendfile_wait_generic, 1423 .pr_ready = tcp_usr_ready, 1424 .pr_shutdown = tcp_usr_shutdown, 1425 .pr_sockaddr = in_getsockaddr, 1426 .pr_sosetlabel = in_pcbsosetlabel, 1427 .pr_close = tcp_usr_close, 1428 }; 1429 #endif /* INET */ 1430 1431 #ifdef INET6 1432 struct protosw tcp6_protosw = { 1433 .pr_type = SOCK_STREAM, 1434 .pr_protocol = IPPROTO_TCP, 1435 .pr_flags = PR_CONNREQUIRED | PR_IMPLOPCL |PR_WANTRCVD | 1436 PR_CAPATTACH, 1437 .pr_ctloutput = tcp_ctloutput, 1438 .pr_abort = tcp_usr_abort, 1439 .pr_accept = tcp6_usr_accept, 1440 .pr_attach = tcp_usr_attach, 1441 .pr_bind = tcp6_usr_bind, 1442 .pr_connect = tcp6_usr_connect, 1443 .pr_control = in6_control, 1444 .pr_detach = tcp_usr_detach, 1445 .pr_disconnect = tcp_usr_disconnect, 1446 .pr_listen = tcp6_usr_listen, 1447 .pr_peeraddr = in6_mapped_peeraddr, 1448 .pr_rcvd = tcp_usr_rcvd, 1449 .pr_rcvoob = tcp_usr_rcvoob, 1450 .pr_send = tcp_usr_send, 1451 .pr_sendfile_wait = sendfile_wait_generic, 1452 .pr_ready = tcp_usr_ready, 1453 .pr_shutdown = tcp_usr_shutdown, 1454 .pr_sockaddr = in6_mapped_sockaddr, 1455 .pr_sosetlabel = in_pcbsosetlabel, 1456 .pr_close = tcp_usr_close, 1457 }; 1458 #endif /* INET6 */ 1459 1460 #ifdef INET 1461 /* 1462 * Common subroutine to open a TCP connection to remote host specified 1463 * by struct sockaddr_in. Call in_pcbconnect() to choose local host address 1464 * and assign a local port number and install the inpcb into the hash. 1465 * Initialize connection parameters and enter SYN-SENT state. 1466 */ 1467 static int 1468 tcp_connect(struct tcpcb *tp, struct sockaddr_in *sin, struct thread *td) 1469 { 1470 struct inpcb *inp = tptoinpcb(tp); 1471 struct socket *so = tptosocket(tp); 1472 int error; 1473 1474 NET_EPOCH_ASSERT(); 1475 INP_WLOCK_ASSERT(inp); 1476 1477 if (__predict_false((so->so_state & 1478 (SS_ISCONNECTING | SS_ISCONNECTED | SS_ISDISCONNECTING | 1479 SS_ISDISCONNECTED)) != 0)) 1480 return (EISCONN); 1481 if (__predict_false((so->so_options & SO_REUSEPORT_LB) != 0)) 1482 return (EOPNOTSUPP); 1483 1484 INP_HASH_WLOCK(&V_tcbinfo); 1485 error = in_pcbconnect(inp, sin, td->td_ucred); 1486 INP_HASH_WUNLOCK(&V_tcbinfo); 1487 if (error != 0) 1488 return (error); 1489 1490 /* 1491 * Compute window scaling to request: 1492 * Scale to fit into sweet spot. See tcp_syncache.c. 1493 * XXX: This should move to tcp_output(). 1494 */ 1495 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 1496 (TCP_MAXWIN << tp->request_r_scale) < sb_max) 1497 tp->request_r_scale++; 1498 1499 soisconnecting(so); 1500 TCPSTAT_INC(tcps_connattempt); 1501 tcp_state_change(tp, TCPS_SYN_SENT); 1502 tp->iss = tcp_new_isn(&inp->inp_inc); 1503 if (tp->t_flags & TF_REQ_TSTMP) 1504 tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc); 1505 tcp_sendseqinit(tp); 1506 1507 return (0); 1508 } 1509 #endif /* INET */ 1510 1511 #ifdef INET6 1512 static int 1513 tcp6_connect(struct tcpcb *tp, struct sockaddr_in6 *sin6, struct thread *td) 1514 { 1515 struct inpcb *inp = tptoinpcb(tp); 1516 struct socket *so = tptosocket(tp); 1517 int error; 1518 1519 NET_EPOCH_ASSERT(); 1520 INP_WLOCK_ASSERT(inp); 1521 1522 if (__predict_false((so->so_state & 1523 (SS_ISCONNECTING | SS_ISCONNECTED | SS_ISDISCONNECTING | 1524 SS_ISDISCONNECTED)) != 0)) 1525 return (EISCONN); 1526 if (__predict_false((so->so_options & SO_REUSEPORT_LB) != 0)) 1527 return (EOPNOTSUPP); 1528 1529 INP_HASH_WLOCK(&V_tcbinfo); 1530 error = in6_pcbconnect(inp, sin6, td->td_ucred, true); 1531 INP_HASH_WUNLOCK(&V_tcbinfo); 1532 if (error != 0) 1533 return (error); 1534 1535 /* Compute window scaling to request. */ 1536 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 1537 (TCP_MAXWIN << tp->request_r_scale) < sb_max) 1538 tp->request_r_scale++; 1539 1540 soisconnecting(so); 1541 TCPSTAT_INC(tcps_connattempt); 1542 tcp_state_change(tp, TCPS_SYN_SENT); 1543 tp->iss = tcp_new_isn(&inp->inp_inc); 1544 if (tp->t_flags & TF_REQ_TSTMP) 1545 tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc); 1546 tcp_sendseqinit(tp); 1547 1548 return (0); 1549 } 1550 #endif /* INET6 */ 1551 1552 /* 1553 * Export TCP internal state information via a struct tcp_info, based on the 1554 * Linux 2.6 API. Not ABI compatible as our constants are mapped differently 1555 * (TCP state machine, etc). We export all information using FreeBSD-native 1556 * constants -- for example, the numeric values for tcpi_state will differ 1557 * from Linux. 1558 */ 1559 void 1560 tcp_fill_info(const struct tcpcb *tp, struct tcp_info *ti) 1561 { 1562 1563 INP_LOCK_ASSERT(tptoinpcb(tp)); 1564 bzero(ti, sizeof(*ti)); 1565 1566 ti->tcpi_state = tp->t_state; 1567 if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP)) 1568 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS; 1569 if (tp->t_flags & TF_SACK_PERMIT) 1570 ti->tcpi_options |= TCPI_OPT_SACK; 1571 if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) { 1572 ti->tcpi_options |= TCPI_OPT_WSCALE; 1573 ti->tcpi_snd_wscale = tp->snd_scale; 1574 ti->tcpi_rcv_wscale = tp->rcv_scale; 1575 } 1576 switch (tp->t_flags2 & (TF2_ECN_PERMIT | TF2_ACE_PERMIT)) { 1577 case TF2_ECN_PERMIT: 1578 ti->tcpi_options |= TCPI_OPT_ECN; 1579 break; 1580 case TF2_ACE_PERMIT: 1581 /* FALLTHROUGH */ 1582 case TF2_ECN_PERMIT | TF2_ACE_PERMIT: 1583 ti->tcpi_options |= TCPI_OPT_ACE; 1584 break; 1585 default: 1586 break; 1587 } 1588 if (tp->t_flags & TF_FASTOPEN) 1589 ti->tcpi_options |= TCPI_OPT_TFO; 1590 1591 ti->tcpi_rto = tp->t_rxtcur * tick; 1592 ti->tcpi_last_data_recv = ((uint32_t)ticks - tp->t_rcvtime) * tick; 1593 ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT; 1594 ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT; 1595 1596 ti->tcpi_snd_ssthresh = tp->snd_ssthresh; 1597 ti->tcpi_snd_cwnd = tp->snd_cwnd; 1598 1599 /* 1600 * FreeBSD-specific extension fields for tcp_info. 1601 */ 1602 ti->tcpi_rcv_space = tp->rcv_wnd; 1603 ti->tcpi_rcv_nxt = tp->rcv_nxt; 1604 ti->tcpi_snd_wnd = tp->snd_wnd; 1605 ti->tcpi_snd_bwnd = 0; /* Unused, kept for compat. */ 1606 ti->tcpi_snd_nxt = tp->snd_nxt; 1607 ti->tcpi_snd_mss = tp->t_maxseg; 1608 ti->tcpi_rcv_mss = tp->t_maxseg; 1609 ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack; 1610 ti->tcpi_rcv_ooopack = tp->t_rcvoopack; 1611 ti->tcpi_snd_zerowin = tp->t_sndzerowin; 1612 ti->tcpi_snd_una = tp->snd_una; 1613 ti->tcpi_snd_max = tp->snd_max; 1614 ti->tcpi_rcv_numsacks = tp->rcv_numsacks; 1615 ti->tcpi_rcv_adv = tp->rcv_adv; 1616 ti->tcpi_dupacks = tp->t_dupacks; 1617 ti->tcpi_rttmin = tp->t_rttlow; 1618 #ifdef TCP_OFFLOAD 1619 if (tp->t_flags & TF_TOE) { 1620 ti->tcpi_options |= TCPI_OPT_TOE; 1621 tcp_offload_tcp_info(tp, ti); 1622 } 1623 #endif 1624 /* 1625 * AccECN related counters. 1626 */ 1627 if ((tp->t_flags2 & (TF2_ECN_PERMIT | TF2_ACE_PERMIT)) == 1628 (TF2_ECN_PERMIT | TF2_ACE_PERMIT)) 1629 /* 1630 * Internal counter starts at 5 for AccECN 1631 * but 0 for RFC3168 ECN. 1632 */ 1633 ti->tcpi_delivered_ce = tp->t_scep - 5; 1634 else 1635 ti->tcpi_delivered_ce = tp->t_scep; 1636 ti->tcpi_received_ce = tp->t_rcep; 1637 } 1638 1639 /* 1640 * tcp_ctloutput() must drop the inpcb lock before performing copyin on 1641 * socket option arguments. When it re-acquires the lock after the copy, it 1642 * has to revalidate that the connection is still valid for the socket 1643 * option. 1644 */ 1645 #define INP_WLOCK_RECHECK_CLEANUP(inp, cleanup) do { \ 1646 INP_WLOCK(inp); \ 1647 if (inp->inp_flags & INP_DROPPED) { \ 1648 INP_WUNLOCK(inp); \ 1649 cleanup; \ 1650 return (ECONNRESET); \ 1651 } \ 1652 tp = intotcpcb(inp); \ 1653 } while(0) 1654 #define INP_WLOCK_RECHECK(inp) INP_WLOCK_RECHECK_CLEANUP((inp), /* noop */) 1655 1656 int 1657 tcp_ctloutput_set(struct inpcb *inp, struct sockopt *sopt) 1658 { 1659 struct socket *so = inp->inp_socket; 1660 struct tcpcb *tp = intotcpcb(inp); 1661 int error = 0; 1662 1663 MPASS(sopt->sopt_dir == SOPT_SET); 1664 INP_WLOCK_ASSERT(inp); 1665 KASSERT((inp->inp_flags & INP_DROPPED) == 0, 1666 ("inp_flags == %x", inp->inp_flags)); 1667 KASSERT(so != NULL, ("inp_socket == NULL")); 1668 1669 if (sopt->sopt_level != IPPROTO_TCP) { 1670 INP_WUNLOCK(inp); 1671 #ifdef INET6 1672 if (inp->inp_vflag & INP_IPV6PROTO) 1673 error = ip6_ctloutput(so, sopt); 1674 #endif 1675 #if defined(INET6) && defined(INET) 1676 else 1677 #endif 1678 #ifdef INET 1679 error = ip_ctloutput(so, sopt); 1680 #endif 1681 /* 1682 * When an IP-level socket option affects TCP, pass control 1683 * down to stack tfb_tcp_ctloutput, otherwise return what 1684 * IP level returned. 1685 */ 1686 switch (sopt->sopt_level) { 1687 #ifdef INET6 1688 case IPPROTO_IPV6: 1689 if ((inp->inp_vflag & INP_IPV6PROTO) == 0) 1690 return (error); 1691 switch (sopt->sopt_name) { 1692 case IPV6_TCLASS: 1693 /* Notify tcp stacks that care (e.g. RACK). */ 1694 break; 1695 case IPV6_USE_MIN_MTU: 1696 /* Update t_maxseg accordingly. */ 1697 break; 1698 default: 1699 return (error); 1700 } 1701 break; 1702 #endif 1703 #ifdef INET 1704 case IPPROTO_IP: 1705 switch (sopt->sopt_name) { 1706 case IP_TOS: 1707 inp->inp_ip_tos &= ~IPTOS_ECN_MASK; 1708 break; 1709 case IP_TTL: 1710 /* Notify tcp stacks that care (e.g. RACK). */ 1711 break; 1712 default: 1713 return (error); 1714 } 1715 break; 1716 #endif 1717 default: 1718 return (error); 1719 } 1720 INP_WLOCK_RECHECK(inp); 1721 } else if (sopt->sopt_name == TCP_FUNCTION_BLK) { 1722 /* 1723 * Protect the TCP option TCP_FUNCTION_BLK so 1724 * that a sub-function can *never* overwrite this. 1725 */ 1726 struct tcp_function_set fsn; 1727 struct tcp_function_block *blk; 1728 void *ptr = NULL; 1729 1730 INP_WUNLOCK(inp); 1731 error = sooptcopyin(sopt, &fsn, sizeof fsn, sizeof fsn); 1732 if (error) 1733 return (error); 1734 1735 INP_WLOCK_RECHECK(inp); 1736 1737 blk = find_and_ref_tcp_functions(&fsn); 1738 if (blk == NULL) { 1739 INP_WUNLOCK(inp); 1740 return (ENOENT); 1741 } 1742 if (tp->t_fb == blk) { 1743 /* You already have this */ 1744 refcount_release(&blk->tfb_refcnt); 1745 INP_WUNLOCK(inp); 1746 return (0); 1747 } 1748 if (blk->tfb_flags & TCP_FUNC_BEING_REMOVED) { 1749 refcount_release(&blk->tfb_refcnt); 1750 INP_WUNLOCK(inp); 1751 return (ENOENT); 1752 } 1753 error = (*blk->tfb_tcp_handoff_ok)(tp); 1754 if (error) { 1755 refcount_release(&blk->tfb_refcnt); 1756 INP_WUNLOCK(inp); 1757 return (error); 1758 } 1759 /* 1760 * Ensure the new stack takes ownership with a 1761 * clean slate on peak rate threshold. 1762 */ 1763 if (tp->t_fb->tfb_tcp_timer_stop_all != NULL) 1764 tp->t_fb->tfb_tcp_timer_stop_all(tp); 1765 if (blk->tfb_tcp_fb_init) { 1766 error = (*blk->tfb_tcp_fb_init)(tp, &ptr); 1767 if (error) { 1768 /* 1769 * Release the ref count the lookup 1770 * acquired. 1771 */ 1772 refcount_release(&blk->tfb_refcnt); 1773 /* 1774 * Now there is a chance that the 1775 * init() function mucked with some 1776 * things before it failed, such as 1777 * hpts or inp_flags2 or timer granularity. 1778 * It should not of, but lets give the old 1779 * stack a chance to reset to a known good state. 1780 */ 1781 if (tp->t_fb->tfb_switch_failed) { 1782 (*tp->t_fb->tfb_switch_failed)(tp); 1783 } 1784 goto err_out; 1785 } 1786 } 1787 if (tp->t_fb->tfb_tcp_fb_fini) { 1788 struct epoch_tracker et; 1789 /* 1790 * Tell the stack to cleanup with 0 i.e. 1791 * the tcb is not going away. 1792 */ 1793 NET_EPOCH_ENTER(et); 1794 (*tp->t_fb->tfb_tcp_fb_fini)(tp, 0); 1795 NET_EPOCH_EXIT(et); 1796 } 1797 /* 1798 * Release the old refcnt, the 1799 * lookup acquired a ref on the 1800 * new one already. 1801 */ 1802 refcount_release(&tp->t_fb->tfb_refcnt); 1803 /* 1804 * Set in the new stack. 1805 */ 1806 tp->t_fb = blk; 1807 tp->t_fb_ptr = ptr; 1808 #ifdef TCP_OFFLOAD 1809 if (tp->t_flags & TF_TOE) { 1810 tcp_offload_ctloutput(tp, sopt->sopt_dir, 1811 sopt->sopt_name); 1812 } 1813 #endif 1814 err_out: 1815 INP_WUNLOCK(inp); 1816 return (error); 1817 1818 } 1819 1820 /* Pass in the INP locked, callee must unlock it. */ 1821 return (tp->t_fb->tfb_tcp_ctloutput(tp, sopt)); 1822 } 1823 1824 static int 1825 tcp_ctloutput_get(struct inpcb *inp, struct sockopt *sopt) 1826 { 1827 struct socket *so = inp->inp_socket; 1828 struct tcpcb *tp = intotcpcb(inp); 1829 int error = 0; 1830 1831 MPASS(sopt->sopt_dir == SOPT_GET); 1832 INP_WLOCK_ASSERT(inp); 1833 KASSERT((inp->inp_flags & INP_DROPPED) == 0, 1834 ("inp_flags == %x", inp->inp_flags)); 1835 KASSERT(so != NULL, ("inp_socket == NULL")); 1836 1837 if (sopt->sopt_level != IPPROTO_TCP) { 1838 INP_WUNLOCK(inp); 1839 #ifdef INET6 1840 if (inp->inp_vflag & INP_IPV6PROTO) 1841 error = ip6_ctloutput(so, sopt); 1842 #endif /* INET6 */ 1843 #if defined(INET6) && defined(INET) 1844 else 1845 #endif 1846 #ifdef INET 1847 error = ip_ctloutput(so, sopt); 1848 #endif 1849 return (error); 1850 } 1851 if (((sopt->sopt_name == TCP_FUNCTION_BLK) || 1852 (sopt->sopt_name == TCP_FUNCTION_ALIAS))) { 1853 struct tcp_function_set fsn; 1854 1855 if (sopt->sopt_name == TCP_FUNCTION_ALIAS) { 1856 memset(&fsn, 0, sizeof(fsn)); 1857 find_tcp_function_alias(tp->t_fb, &fsn); 1858 } else { 1859 strncpy(fsn.function_set_name, 1860 tp->t_fb->tfb_tcp_block_name, 1861 TCP_FUNCTION_NAME_LEN_MAX); 1862 fsn.function_set_name[TCP_FUNCTION_NAME_LEN_MAX - 1] = '\0'; 1863 } 1864 fsn.pcbcnt = tp->t_fb->tfb_refcnt; 1865 INP_WUNLOCK(inp); 1866 error = sooptcopyout(sopt, &fsn, sizeof fsn); 1867 return (error); 1868 } 1869 1870 /* Pass in the INP locked, callee must unlock it. */ 1871 return (tp->t_fb->tfb_tcp_ctloutput(tp, sopt)); 1872 } 1873 1874 int 1875 tcp_ctloutput(struct socket *so, struct sockopt *sopt) 1876 { 1877 struct inpcb *inp; 1878 1879 inp = sotoinpcb(so); 1880 KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL")); 1881 1882 INP_WLOCK(inp); 1883 if (inp->inp_flags & INP_DROPPED) { 1884 INP_WUNLOCK(inp); 1885 return (ECONNRESET); 1886 } 1887 if (sopt->sopt_dir == SOPT_SET) 1888 return (tcp_ctloutput_set(inp, sopt)); 1889 else if (sopt->sopt_dir == SOPT_GET) 1890 return (tcp_ctloutput_get(inp, sopt)); 1891 else 1892 panic("%s: sopt_dir $%d", __func__, sopt->sopt_dir); 1893 } 1894 1895 /* 1896 * If this assert becomes untrue, we need to change the size of the buf 1897 * variable in tcp_default_ctloutput(). 1898 */ 1899 #ifdef CTASSERT 1900 CTASSERT(TCP_CA_NAME_MAX <= TCP_LOG_ID_LEN); 1901 CTASSERT(TCP_LOG_REASON_LEN <= TCP_LOG_ID_LEN); 1902 #endif 1903 1904 extern struct cc_algo newreno_cc_algo; 1905 1906 static int 1907 tcp_set_cc_mod(struct inpcb *inp, struct sockopt *sopt) 1908 { 1909 struct cc_algo *algo; 1910 void *ptr = NULL; 1911 struct tcpcb *tp; 1912 struct cc_var cc_mem; 1913 char buf[TCP_CA_NAME_MAX]; 1914 size_t mem_sz; 1915 int error; 1916 1917 INP_WUNLOCK(inp); 1918 error = sooptcopyin(sopt, buf, TCP_CA_NAME_MAX - 1, 1); 1919 if (error) 1920 return(error); 1921 buf[sopt->sopt_valsize] = '\0'; 1922 CC_LIST_RLOCK(); 1923 STAILQ_FOREACH(algo, &cc_list, entries) { 1924 if (strncmp(buf, algo->name, 1925 TCP_CA_NAME_MAX) == 0) { 1926 if (algo->flags & CC_MODULE_BEING_REMOVED) { 1927 /* We can't "see" modules being unloaded */ 1928 continue; 1929 } 1930 break; 1931 } 1932 } 1933 if (algo == NULL) { 1934 CC_LIST_RUNLOCK(); 1935 return(ESRCH); 1936 } 1937 /* 1938 * With a reference the algorithm cannot be removed 1939 * so we hold a reference through the change process. 1940 */ 1941 cc_refer(algo); 1942 CC_LIST_RUNLOCK(); 1943 if (algo->cb_init != NULL) { 1944 /* We can now pre-get the memory for the CC */ 1945 mem_sz = (*algo->cc_data_sz)(); 1946 if (mem_sz == 0) { 1947 goto no_mem_needed; 1948 } 1949 ptr = malloc(mem_sz, M_CC_MEM, M_WAITOK); 1950 } else { 1951 no_mem_needed: 1952 mem_sz = 0; 1953 ptr = NULL; 1954 } 1955 /* 1956 * Make sure its all clean and zero and also get 1957 * back the inplock. 1958 */ 1959 memset(&cc_mem, 0, sizeof(cc_mem)); 1960 INP_WLOCK(inp); 1961 if (inp->inp_flags & INP_DROPPED) { 1962 INP_WUNLOCK(inp); 1963 if (ptr) 1964 free(ptr, M_CC_MEM); 1965 /* Release our temp reference */ 1966 CC_LIST_RLOCK(); 1967 cc_release(algo); 1968 CC_LIST_RUNLOCK(); 1969 return (ECONNRESET); 1970 } 1971 tp = intotcpcb(inp); 1972 if (ptr != NULL) 1973 memset(ptr, 0, mem_sz); 1974 cc_mem.tp = tp; 1975 /* 1976 * We once again hold a write lock over the tcb so it's 1977 * safe to do these things without ordering concerns. 1978 * Note here we init into stack memory. 1979 */ 1980 if (algo->cb_init != NULL) 1981 error = algo->cb_init(&cc_mem, ptr); 1982 else 1983 error = 0; 1984 /* 1985 * The CC algorithms, when given their memory 1986 * should not fail we could in theory have a 1987 * KASSERT here. 1988 */ 1989 if (error == 0) { 1990 /* 1991 * Touchdown, lets go ahead and move the 1992 * connection to the new CC module by 1993 * copying in the cc_mem after we call 1994 * the old ones cleanup (if any). 1995 */ 1996 if (CC_ALGO(tp)->cb_destroy != NULL) 1997 CC_ALGO(tp)->cb_destroy(&tp->t_ccv); 1998 /* Detach the old CC from the tcpcb */ 1999 cc_detach(tp); 2000 /* Copy in our temp memory that was inited */ 2001 memcpy(&tp->t_ccv, &cc_mem, sizeof(struct cc_var)); 2002 /* Now attach the new, which takes a reference */ 2003 cc_attach(tp, algo); 2004 /* Ok now are we where we have gotten past any conn_init? */ 2005 if (TCPS_HAVEESTABLISHED(tp->t_state) && (CC_ALGO(tp)->conn_init != NULL)) { 2006 /* Yep run the connection init for the new CC */ 2007 CC_ALGO(tp)->conn_init(&tp->t_ccv); 2008 } 2009 } else if (ptr) 2010 free(ptr, M_CC_MEM); 2011 INP_WUNLOCK(inp); 2012 /* Now lets release our temp reference */ 2013 CC_LIST_RLOCK(); 2014 cc_release(algo); 2015 CC_LIST_RUNLOCK(); 2016 return (error); 2017 } 2018 2019 int 2020 tcp_default_ctloutput(struct tcpcb *tp, struct sockopt *sopt) 2021 { 2022 struct inpcb *inp = tptoinpcb(tp); 2023 int error, opt, optval; 2024 u_int ui; 2025 struct tcp_info ti; 2026 #ifdef KERN_TLS 2027 struct tls_enable tls; 2028 struct socket *so = inp->inp_socket; 2029 #endif 2030 char *pbuf, buf[TCP_LOG_ID_LEN]; 2031 #ifdef STATS 2032 struct statsblob *sbp; 2033 #endif 2034 size_t len; 2035 2036 INP_WLOCK_ASSERT(inp); 2037 KASSERT((inp->inp_flags & INP_DROPPED) == 0, 2038 ("inp_flags == %x", inp->inp_flags)); 2039 KASSERT(inp->inp_socket != NULL, ("inp_socket == NULL")); 2040 2041 switch (sopt->sopt_level) { 2042 #ifdef INET6 2043 case IPPROTO_IPV6: 2044 MPASS(inp->inp_vflag & INP_IPV6PROTO); 2045 switch (sopt->sopt_name) { 2046 case IPV6_USE_MIN_MTU: 2047 tcp6_use_min_mtu(tp); 2048 /* FALLTHROUGH */ 2049 } 2050 INP_WUNLOCK(inp); 2051 return (0); 2052 #endif 2053 #ifdef INET 2054 case IPPROTO_IP: 2055 INP_WUNLOCK(inp); 2056 return (0); 2057 #endif 2058 } 2059 2060 /* 2061 * For TCP_CCALGOOPT forward the control to CC module, for both 2062 * SOPT_SET and SOPT_GET. 2063 */ 2064 switch (sopt->sopt_name) { 2065 case TCP_CCALGOOPT: 2066 INP_WUNLOCK(inp); 2067 if (sopt->sopt_valsize > CC_ALGOOPT_LIMIT) 2068 return (EINVAL); 2069 pbuf = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK | M_ZERO); 2070 error = sooptcopyin(sopt, pbuf, sopt->sopt_valsize, 2071 sopt->sopt_valsize); 2072 if (error) { 2073 free(pbuf, M_TEMP); 2074 return (error); 2075 } 2076 INP_WLOCK_RECHECK_CLEANUP(inp, free(pbuf, M_TEMP)); 2077 if (CC_ALGO(tp)->ctl_output != NULL) 2078 error = CC_ALGO(tp)->ctl_output(&tp->t_ccv, sopt, pbuf); 2079 else 2080 error = ENOENT; 2081 INP_WUNLOCK(inp); 2082 if (error == 0 && sopt->sopt_dir == SOPT_GET) 2083 error = sooptcopyout(sopt, pbuf, sopt->sopt_valsize); 2084 free(pbuf, M_TEMP); 2085 return (error); 2086 } 2087 2088 switch (sopt->sopt_dir) { 2089 case SOPT_SET: 2090 switch (sopt->sopt_name) { 2091 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE) 2092 case TCP_MD5SIG: 2093 INP_WUNLOCK(inp); 2094 if (!TCPMD5_ENABLED()) 2095 return (ENOPROTOOPT); 2096 error = TCPMD5_PCBCTL(inp, sopt); 2097 if (error) 2098 return (error); 2099 INP_WLOCK_RECHECK(inp); 2100 goto unlock_and_done; 2101 #endif /* IPSEC */ 2102 2103 case TCP_NODELAY: 2104 case TCP_NOOPT: 2105 INP_WUNLOCK(inp); 2106 error = sooptcopyin(sopt, &optval, sizeof optval, 2107 sizeof optval); 2108 if (error) 2109 return (error); 2110 2111 INP_WLOCK_RECHECK(inp); 2112 switch (sopt->sopt_name) { 2113 case TCP_NODELAY: 2114 opt = TF_NODELAY; 2115 break; 2116 case TCP_NOOPT: 2117 opt = TF_NOOPT; 2118 break; 2119 default: 2120 opt = 0; /* dead code to fool gcc */ 2121 break; 2122 } 2123 2124 if (optval) 2125 tp->t_flags |= opt; 2126 else 2127 tp->t_flags &= ~opt; 2128 unlock_and_done: 2129 #ifdef TCP_OFFLOAD 2130 if (tp->t_flags & TF_TOE) { 2131 tcp_offload_ctloutput(tp, sopt->sopt_dir, 2132 sopt->sopt_name); 2133 } 2134 #endif 2135 INP_WUNLOCK(inp); 2136 break; 2137 2138 case TCP_NOPUSH: 2139 INP_WUNLOCK(inp); 2140 error = sooptcopyin(sopt, &optval, sizeof optval, 2141 sizeof optval); 2142 if (error) 2143 return (error); 2144 2145 INP_WLOCK_RECHECK(inp); 2146 if (optval) 2147 tp->t_flags |= TF_NOPUSH; 2148 else if (tp->t_flags & TF_NOPUSH) { 2149 tp->t_flags &= ~TF_NOPUSH; 2150 if (TCPS_HAVEESTABLISHED(tp->t_state)) { 2151 struct epoch_tracker et; 2152 2153 NET_EPOCH_ENTER(et); 2154 error = tcp_output_nodrop(tp); 2155 NET_EPOCH_EXIT(et); 2156 } 2157 } 2158 goto unlock_and_done; 2159 2160 case TCP_REMOTE_UDP_ENCAPS_PORT: 2161 INP_WUNLOCK(inp); 2162 error = sooptcopyin(sopt, &optval, sizeof optval, 2163 sizeof optval); 2164 if (error) 2165 return (error); 2166 if ((optval < TCP_TUNNELING_PORT_MIN) || 2167 (optval > TCP_TUNNELING_PORT_MAX)) { 2168 /* Its got to be in range */ 2169 return (EINVAL); 2170 } 2171 if ((V_tcp_udp_tunneling_port == 0) && (optval != 0)) { 2172 /* You have to have enabled a UDP tunneling port first */ 2173 return (EINVAL); 2174 } 2175 INP_WLOCK_RECHECK(inp); 2176 if (tp->t_state != TCPS_CLOSED) { 2177 /* You can't change after you are connected */ 2178 error = EINVAL; 2179 } else { 2180 /* Ok we are all good set the port */ 2181 tp->t_port = htons(optval); 2182 } 2183 goto unlock_and_done; 2184 2185 case TCP_MAXSEG: 2186 INP_WUNLOCK(inp); 2187 error = sooptcopyin(sopt, &optval, sizeof optval, 2188 sizeof optval); 2189 if (error) 2190 return (error); 2191 2192 INP_WLOCK_RECHECK(inp); 2193 if (optval > 0 && optval <= tp->t_maxseg && 2194 optval + 40 >= V_tcp_minmss) { 2195 tp->t_maxseg = optval; 2196 if (tp->t_maxseg < V_tcp_mssdflt) { 2197 /* 2198 * The MSS is so small we should not process incoming 2199 * SACK's since we are subject to attack in such a 2200 * case. 2201 */ 2202 tp->t_flags2 |= TF2_PROC_SACK_PROHIBIT; 2203 } else { 2204 tp->t_flags2 &= ~TF2_PROC_SACK_PROHIBIT; 2205 } 2206 } else 2207 error = EINVAL; 2208 goto unlock_and_done; 2209 2210 case TCP_INFO: 2211 INP_WUNLOCK(inp); 2212 error = EINVAL; 2213 break; 2214 2215 case TCP_STATS: 2216 INP_WUNLOCK(inp); 2217 #ifdef STATS 2218 error = sooptcopyin(sopt, &optval, sizeof optval, 2219 sizeof optval); 2220 if (error) 2221 return (error); 2222 2223 if (optval > 0) 2224 sbp = stats_blob_alloc( 2225 V_tcp_perconn_stats_dflt_tpl, 0); 2226 else 2227 sbp = NULL; 2228 2229 INP_WLOCK_RECHECK(inp); 2230 if ((tp->t_stats != NULL && sbp == NULL) || 2231 (tp->t_stats == NULL && sbp != NULL)) { 2232 struct statsblob *t = tp->t_stats; 2233 tp->t_stats = sbp; 2234 sbp = t; 2235 } 2236 INP_WUNLOCK(inp); 2237 2238 stats_blob_destroy(sbp); 2239 #else 2240 return (EOPNOTSUPP); 2241 #endif /* !STATS */ 2242 break; 2243 2244 case TCP_CONGESTION: 2245 error = tcp_set_cc_mod(inp, sopt); 2246 break; 2247 2248 case TCP_REUSPORT_LB_NUMA: 2249 INP_WUNLOCK(inp); 2250 error = sooptcopyin(sopt, &optval, sizeof(optval), 2251 sizeof(optval)); 2252 INP_WLOCK_RECHECK(inp); 2253 if (!error) 2254 error = in_pcblbgroup_numa(inp, optval); 2255 INP_WUNLOCK(inp); 2256 break; 2257 2258 #ifdef KERN_TLS 2259 case TCP_TXTLS_ENABLE: 2260 INP_WUNLOCK(inp); 2261 error = ktls_copyin_tls_enable(sopt, &tls); 2262 if (error != 0) 2263 break; 2264 error = ktls_enable_tx(so, &tls); 2265 ktls_cleanup_tls_enable(&tls); 2266 break; 2267 case TCP_TXTLS_MODE: 2268 INP_WUNLOCK(inp); 2269 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui)); 2270 if (error != 0) 2271 return (error); 2272 2273 INP_WLOCK_RECHECK(inp); 2274 error = ktls_set_tx_mode(so, ui); 2275 INP_WUNLOCK(inp); 2276 break; 2277 case TCP_RXTLS_ENABLE: 2278 INP_WUNLOCK(inp); 2279 error = ktls_copyin_tls_enable(sopt, &tls); 2280 if (error != 0) 2281 break; 2282 error = ktls_enable_rx(so, &tls); 2283 ktls_cleanup_tls_enable(&tls); 2284 break; 2285 #endif 2286 case TCP_MAXUNACKTIME: 2287 case TCP_KEEPIDLE: 2288 case TCP_KEEPINTVL: 2289 case TCP_KEEPINIT: 2290 INP_WUNLOCK(inp); 2291 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui)); 2292 if (error) 2293 return (error); 2294 2295 if (ui > (UINT_MAX / hz)) { 2296 error = EINVAL; 2297 break; 2298 } 2299 ui *= hz; 2300 2301 INP_WLOCK_RECHECK(inp); 2302 switch (sopt->sopt_name) { 2303 case TCP_MAXUNACKTIME: 2304 tp->t_maxunacktime = ui; 2305 break; 2306 2307 case TCP_KEEPIDLE: 2308 tp->t_keepidle = ui; 2309 /* 2310 * XXX: better check current remaining 2311 * timeout and "merge" it with new value. 2312 */ 2313 if ((tp->t_state > TCPS_LISTEN) && 2314 (tp->t_state <= TCPS_CLOSING)) 2315 tcp_timer_activate(tp, TT_KEEP, 2316 TP_KEEPIDLE(tp)); 2317 break; 2318 case TCP_KEEPINTVL: 2319 tp->t_keepintvl = ui; 2320 if ((tp->t_state == TCPS_FIN_WAIT_2) && 2321 (TP_MAXIDLE(tp) > 0)) 2322 tcp_timer_activate(tp, TT_2MSL, 2323 TP_MAXIDLE(tp)); 2324 break; 2325 case TCP_KEEPINIT: 2326 tp->t_keepinit = ui; 2327 if (tp->t_state == TCPS_SYN_RECEIVED || 2328 tp->t_state == TCPS_SYN_SENT) 2329 tcp_timer_activate(tp, TT_KEEP, 2330 TP_KEEPINIT(tp)); 2331 break; 2332 } 2333 goto unlock_and_done; 2334 2335 case TCP_KEEPCNT: 2336 INP_WUNLOCK(inp); 2337 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui)); 2338 if (error) 2339 return (error); 2340 2341 INP_WLOCK_RECHECK(inp); 2342 tp->t_keepcnt = ui; 2343 if ((tp->t_state == TCPS_FIN_WAIT_2) && 2344 (TP_MAXIDLE(tp) > 0)) 2345 tcp_timer_activate(tp, TT_2MSL, 2346 TP_MAXIDLE(tp)); 2347 goto unlock_and_done; 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 case TCP_FASTOPEN: 2580 optval = tp->t_flags & TF_FASTOPEN; 2581 INP_WUNLOCK(inp); 2582 error = sooptcopyout(sopt, &optval, sizeof optval); 2583 break; 2584 #ifdef TCP_BLACKBOX 2585 case TCP_LOG: 2586 optval = tcp_get_bblog_state(tp); 2587 INP_WUNLOCK(inp); 2588 error = sooptcopyout(sopt, &optval, sizeof(optval)); 2589 break; 2590 case TCP_LOGBUF: 2591 /* tcp_log_getlogbuf() does INP_WUNLOCK(inp) */ 2592 error = tcp_log_getlogbuf(sopt, tp); 2593 break; 2594 case TCP_LOGID: 2595 len = tcp_log_get_id(tp, buf); 2596 INP_WUNLOCK(inp); 2597 error = sooptcopyout(sopt, buf, len + 1); 2598 break; 2599 case TCP_LOGDUMP: 2600 case TCP_LOGDUMPID: 2601 INP_WUNLOCK(inp); 2602 error = EINVAL; 2603 break; 2604 #endif 2605 #ifdef KERN_TLS 2606 case TCP_TXTLS_MODE: 2607 error = ktls_get_tx_mode(so, &optval); 2608 INP_WUNLOCK(inp); 2609 if (error == 0) 2610 error = sooptcopyout(sopt, &optval, 2611 sizeof(optval)); 2612 break; 2613 case TCP_RXTLS_MODE: 2614 error = ktls_get_rx_mode(so, &optval); 2615 INP_WUNLOCK(inp); 2616 if (error == 0) 2617 error = sooptcopyout(sopt, &optval, 2618 sizeof(optval)); 2619 break; 2620 #endif 2621 default: 2622 INP_WUNLOCK(inp); 2623 error = ENOPROTOOPT; 2624 break; 2625 } 2626 break; 2627 } 2628 return (error); 2629 } 2630 #undef INP_WLOCK_RECHECK 2631 #undef INP_WLOCK_RECHECK_CLEANUP 2632 2633 /* 2634 * Initiate (or continue) disconnect. 2635 * If embryonic state, just send reset (once). 2636 * If in ``let data drain'' option and linger null, just drop. 2637 * Otherwise (hard), mark socket disconnecting and drop 2638 * current input data; switch states based on user close, and 2639 * send segment to peer (with FIN). 2640 */ 2641 static void 2642 tcp_disconnect(struct tcpcb *tp) 2643 { 2644 struct inpcb *inp = tptoinpcb(tp); 2645 struct socket *so = tptosocket(tp); 2646 2647 NET_EPOCH_ASSERT(); 2648 INP_WLOCK_ASSERT(inp); 2649 2650 /* 2651 * Neither tcp_close() nor tcp_drop() should return NULL, as the 2652 * socket is still open. 2653 */ 2654 if (tp->t_state < TCPS_ESTABLISHED && 2655 !(tp->t_state > TCPS_LISTEN && (tp->t_flags & TF_FASTOPEN))) { 2656 tp = tcp_close(tp); 2657 KASSERT(tp != NULL, 2658 ("tcp_disconnect: tcp_close() returned NULL")); 2659 } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) { 2660 tp = tcp_drop(tp, 0); 2661 KASSERT(tp != NULL, 2662 ("tcp_disconnect: tcp_drop() returned NULL")); 2663 } else { 2664 soisdisconnecting(so); 2665 sbflush(&so->so_rcv); 2666 tcp_usrclosed(tp); 2667 if (!(inp->inp_flags & INP_DROPPED)) 2668 /* Ignore stack's drop request, we already at it. */ 2669 (void)tcp_output_nodrop(tp); 2670 } 2671 } 2672 2673 /* 2674 * User issued close, and wish to trail through shutdown states: 2675 * if never received SYN, just forget it. If got a SYN from peer, 2676 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 2677 * If already got a FIN from peer, then almost done; go to LAST_ACK 2678 * state. In all other cases, have already sent FIN to peer (e.g. 2679 * after PRU_SHUTDOWN), and just have to play tedious game waiting 2680 * for peer to send FIN or not respond to keep-alives, etc. 2681 * We can let the user exit from the close as soon as the FIN is acked. 2682 */ 2683 static void 2684 tcp_usrclosed(struct tcpcb *tp) 2685 { 2686 2687 NET_EPOCH_ASSERT(); 2688 INP_WLOCK_ASSERT(tptoinpcb(tp)); 2689 2690 switch (tp->t_state) { 2691 case TCPS_LISTEN: 2692 #ifdef TCP_OFFLOAD 2693 tcp_offload_listen_stop(tp); 2694 #endif 2695 tcp_state_change(tp, TCPS_CLOSED); 2696 /* FALLTHROUGH */ 2697 case TCPS_CLOSED: 2698 tp = tcp_close(tp); 2699 /* 2700 * tcp_close() should never return NULL here as the socket is 2701 * still open. 2702 */ 2703 KASSERT(tp != NULL, 2704 ("tcp_usrclosed: tcp_close() returned NULL")); 2705 break; 2706 2707 case TCPS_SYN_SENT: 2708 case TCPS_SYN_RECEIVED: 2709 tp->t_flags |= TF_NEEDFIN; 2710 break; 2711 2712 case TCPS_ESTABLISHED: 2713 tcp_state_change(tp, TCPS_FIN_WAIT_1); 2714 break; 2715 2716 case TCPS_CLOSE_WAIT: 2717 tcp_state_change(tp, TCPS_LAST_ACK); 2718 break; 2719 } 2720 if (tp->t_acktime == 0) 2721 tp->t_acktime = ticks; 2722 if (tp->t_state >= TCPS_FIN_WAIT_2) { 2723 tcp_free_sackholes(tp); 2724 soisdisconnected(tptosocket(tp)); 2725 /* Prevent the connection hanging in FIN_WAIT_2 forever. */ 2726 if (tp->t_state == TCPS_FIN_WAIT_2) { 2727 int timeout; 2728 2729 timeout = (tcp_fast_finwait2_recycle) ? 2730 tcp_finwait2_timeout : TP_MAXIDLE(tp); 2731 tcp_timer_activate(tp, TT_2MSL, timeout); 2732 } 2733 } 2734 } 2735 2736 #ifdef DDB 2737 static void 2738 db_print_indent(int indent) 2739 { 2740 int i; 2741 2742 for (i = 0; i < indent; i++) 2743 db_printf(" "); 2744 } 2745 2746 static void 2747 db_print_tstate(int t_state) 2748 { 2749 2750 switch (t_state) { 2751 case TCPS_CLOSED: 2752 db_printf("TCPS_CLOSED"); 2753 return; 2754 2755 case TCPS_LISTEN: 2756 db_printf("TCPS_LISTEN"); 2757 return; 2758 2759 case TCPS_SYN_SENT: 2760 db_printf("TCPS_SYN_SENT"); 2761 return; 2762 2763 case TCPS_SYN_RECEIVED: 2764 db_printf("TCPS_SYN_RECEIVED"); 2765 return; 2766 2767 case TCPS_ESTABLISHED: 2768 db_printf("TCPS_ESTABLISHED"); 2769 return; 2770 2771 case TCPS_CLOSE_WAIT: 2772 db_printf("TCPS_CLOSE_WAIT"); 2773 return; 2774 2775 case TCPS_FIN_WAIT_1: 2776 db_printf("TCPS_FIN_WAIT_1"); 2777 return; 2778 2779 case TCPS_CLOSING: 2780 db_printf("TCPS_CLOSING"); 2781 return; 2782 2783 case TCPS_LAST_ACK: 2784 db_printf("TCPS_LAST_ACK"); 2785 return; 2786 2787 case TCPS_FIN_WAIT_2: 2788 db_printf("TCPS_FIN_WAIT_2"); 2789 return; 2790 2791 case TCPS_TIME_WAIT: 2792 db_printf("TCPS_TIME_WAIT"); 2793 return; 2794 2795 default: 2796 db_printf("unknown"); 2797 return; 2798 } 2799 } 2800 2801 static void 2802 db_print_tflags(u_int t_flags) 2803 { 2804 int comma; 2805 2806 comma = 0; 2807 if (t_flags & TF_ACKNOW) { 2808 db_printf("%sTF_ACKNOW", comma ? ", " : ""); 2809 comma = 1; 2810 } 2811 if (t_flags & TF_DELACK) { 2812 db_printf("%sTF_DELACK", comma ? ", " : ""); 2813 comma = 1; 2814 } 2815 if (t_flags & TF_NODELAY) { 2816 db_printf("%sTF_NODELAY", comma ? ", " : ""); 2817 comma = 1; 2818 } 2819 if (t_flags & TF_NOOPT) { 2820 db_printf("%sTF_NOOPT", comma ? ", " : ""); 2821 comma = 1; 2822 } 2823 if (t_flags & TF_SENTFIN) { 2824 db_printf("%sTF_SENTFIN", comma ? ", " : ""); 2825 comma = 1; 2826 } 2827 if (t_flags & TF_REQ_SCALE) { 2828 db_printf("%sTF_REQ_SCALE", comma ? ", " : ""); 2829 comma = 1; 2830 } 2831 if (t_flags & TF_RCVD_SCALE) { 2832 db_printf("%sTF_RECVD_SCALE", comma ? ", " : ""); 2833 comma = 1; 2834 } 2835 if (t_flags & TF_REQ_TSTMP) { 2836 db_printf("%sTF_REQ_TSTMP", comma ? ", " : ""); 2837 comma = 1; 2838 } 2839 if (t_flags & TF_RCVD_TSTMP) { 2840 db_printf("%sTF_RCVD_TSTMP", comma ? ", " : ""); 2841 comma = 1; 2842 } 2843 if (t_flags & TF_SACK_PERMIT) { 2844 db_printf("%sTF_SACK_PERMIT", comma ? ", " : ""); 2845 comma = 1; 2846 } 2847 if (t_flags & TF_NEEDSYN) { 2848 db_printf("%sTF_NEEDSYN", comma ? ", " : ""); 2849 comma = 1; 2850 } 2851 if (t_flags & TF_NEEDFIN) { 2852 db_printf("%sTF_NEEDFIN", comma ? ", " : ""); 2853 comma = 1; 2854 } 2855 if (t_flags & TF_NOPUSH) { 2856 db_printf("%sTF_NOPUSH", comma ? ", " : ""); 2857 comma = 1; 2858 } 2859 if (t_flags & TF_PREVVALID) { 2860 db_printf("%sTF_PREVVALID", comma ? ", " : ""); 2861 comma = 1; 2862 } 2863 if (t_flags & TF_WAKESOR) { 2864 db_printf("%sTF_WAKESOR", comma ? ", " : ""); 2865 comma = 1; 2866 } 2867 if (t_flags & TF_GPUTINPROG) { 2868 db_printf("%sTF_GPUTINPROG", comma ? ", " : ""); 2869 comma = 1; 2870 } 2871 if (t_flags & TF_MORETOCOME) { 2872 db_printf("%sTF_MORETOCOME", comma ? ", " : ""); 2873 comma = 1; 2874 } 2875 if (t_flags & TF_SONOTCONN) { 2876 db_printf("%sTF_SONOTCONN", comma ? ", " : ""); 2877 comma = 1; 2878 } 2879 if (t_flags & TF_LASTIDLE) { 2880 db_printf("%sTF_LASTIDLE", comma ? ", " : ""); 2881 comma = 1; 2882 } 2883 if (t_flags & TF_RXWIN0SENT) { 2884 db_printf("%sTF_RXWIN0SENT", comma ? ", " : ""); 2885 comma = 1; 2886 } 2887 if (t_flags & TF_FASTRECOVERY) { 2888 db_printf("%sTF_FASTRECOVERY", comma ? ", " : ""); 2889 comma = 1; 2890 } 2891 if (t_flags & TF_WASFRECOVERY) { 2892 db_printf("%sTF_WASFRECOVERY", comma ? ", " : ""); 2893 comma = 1; 2894 } 2895 if (t_flags & TF_SIGNATURE) { 2896 db_printf("%sTF_SIGNATURE", comma ? ", " : ""); 2897 comma = 1; 2898 } 2899 if (t_flags & TF_FORCEDATA) { 2900 db_printf("%sTF_FORCEDATA", comma ? ", " : ""); 2901 comma = 1; 2902 } 2903 if (t_flags & TF_TSO) { 2904 db_printf("%sTF_TSO", comma ? ", " : ""); 2905 comma = 1; 2906 } 2907 if (t_flags & TF_TOE) { 2908 db_printf("%sTF_TOE", comma ? ", " : ""); 2909 comma = 1; 2910 } 2911 if (t_flags & TF_CLOSED) { 2912 db_printf("%sTF_CLOSED", comma ? ", " : ""); 2913 comma = 1; 2914 } 2915 if (t_flags & TF_SENTSYN) { 2916 db_printf("%sTF_SENTSYN", comma ? ", " : ""); 2917 comma = 1; 2918 } 2919 if (t_flags & TF_LRD) { 2920 db_printf("%sTF_LRD", comma ? ", " : ""); 2921 comma = 1; 2922 } 2923 if (t_flags & TF_CONGRECOVERY) { 2924 db_printf("%sTF_CONGRECOVERY", comma ? ", " : ""); 2925 comma = 1; 2926 } 2927 if (t_flags & TF_WASCRECOVERY) { 2928 db_printf("%sTF_WASCRECOVERY", comma ? ", " : ""); 2929 comma = 1; 2930 } 2931 if (t_flags & TF_FASTOPEN) { 2932 db_printf("%sTF_FASTOPEN", comma ? ", " : ""); 2933 comma = 1; 2934 } 2935 } 2936 2937 static void 2938 db_print_tflags2(u_int t_flags2) 2939 { 2940 int comma; 2941 2942 comma = 0; 2943 if (t_flags2 & TF2_PLPMTU_BLACKHOLE) { 2944 db_printf("%sTF2_PLPMTU_BLACKHOLE", comma ? ", " : ""); 2945 comma = 1; 2946 } 2947 if (t_flags2 & TF2_PLPMTU_PMTUD) { 2948 db_printf("%sTF2_PLPMTU_PMTUD", comma ? ", " : ""); 2949 comma = 1; 2950 } 2951 if (t_flags2 & TF2_PLPMTU_MAXSEGSNT) { 2952 db_printf("%sTF2_PLPMTU_MAXSEGSNT", comma ? ", " : ""); 2953 comma = 1; 2954 } 2955 if (t_flags2 & TF2_LOG_AUTO) { 2956 db_printf("%sTF2_LOG_AUTO", comma ? ", " : ""); 2957 comma = 1; 2958 } 2959 if (t_flags2 & TF2_DROP_AF_DATA) { 2960 db_printf("%sTF2_DROP_AF_DATA", comma ? ", " : ""); 2961 comma = 1; 2962 } 2963 if (t_flags2 & TF2_ECN_PERMIT) { 2964 db_printf("%sTF2_ECN_PERMIT", comma ? ", " : ""); 2965 comma = 1; 2966 } 2967 if (t_flags2 & TF2_ECN_SND_CWR) { 2968 db_printf("%sTF2_ECN_SND_CWR", comma ? ", " : ""); 2969 comma = 1; 2970 } 2971 if (t_flags2 & TF2_ECN_SND_ECE) { 2972 db_printf("%sTF2_ECN_SND_ECE", comma ? ", " : ""); 2973 comma = 1; 2974 } 2975 if (t_flags2 & TF2_ACE_PERMIT) { 2976 db_printf("%sTF2_ACE_PERMIT", comma ? ", " : ""); 2977 comma = 1; 2978 } 2979 if (t_flags2 & TF2_HPTS_CPU_SET) { 2980 db_printf("%sTF2_HPTS_CPU_SET", comma ? ", " : ""); 2981 comma = 1; 2982 } 2983 if (t_flags2 & TF2_FBYTES_COMPLETE) { 2984 db_printf("%sTF2_FBYTES_COMPLETE", comma ? ", " : ""); 2985 comma = 1; 2986 } 2987 if (t_flags2 & TF2_ECN_USE_ECT1) { 2988 db_printf("%sTF2_ECN_USE_ECT1", comma ? ", " : ""); 2989 comma = 1; 2990 } 2991 if (t_flags2 & TF2_TCP_ACCOUNTING) { 2992 db_printf("%sTF2_TCP_ACCOUNTING", comma ? ", " : ""); 2993 comma = 1; 2994 } 2995 if (t_flags2 & TF2_HPTS_CALLS) { 2996 db_printf("%sTF2_HPTS_CALLS", comma ? ", " : ""); 2997 comma = 1; 2998 } 2999 if (t_flags2 & TF2_MBUF_L_ACKS) { 3000 db_printf("%sTF2_MBUF_L_ACKS", comma ? ", " : ""); 3001 comma = 1; 3002 } 3003 if (t_flags2 & TF2_MBUF_ACKCMP) { 3004 db_printf("%sTF2_MBUF_ACKCMP", comma ? ", " : ""); 3005 comma = 1; 3006 } 3007 if (t_flags2 & TF2_SUPPORTS_MBUFQ) { 3008 db_printf("%sTF2_SUPPORTS_MBUFQ", comma ? ", " : ""); 3009 comma = 1; 3010 } 3011 if (t_flags2 & TF2_MBUF_QUEUE_READY) { 3012 db_printf("%sTF2_MBUF_QUEUE_READY", comma ? ", " : ""); 3013 comma = 1; 3014 } 3015 if (t_flags2 & TF2_DONT_SACK_QUEUE) { 3016 db_printf("%sTF2_DONT_SACK_QUEUE", comma ? ", " : ""); 3017 comma = 1; 3018 } 3019 if (t_flags2 & TF2_CANNOT_DO_ECN) { 3020 db_printf("%sTF2_CANNOT_DO_ECN", comma ? ", " : ""); 3021 comma = 1; 3022 } 3023 if (t_flags2 & TF2_PROC_SACK_PROHIBIT) { 3024 db_printf("%sTF2_PROC_SACK_PROHIBIT", comma ? ", " : ""); 3025 comma = 1; 3026 } 3027 if (t_flags2 & TF2_IPSEC_TSO) { 3028 db_printf("%sTF2_IPSEC_TSO", comma ? ", " : ""); 3029 comma = 1; 3030 } 3031 if (t_flags2 & TF2_NO_ISS_CHECK) { 3032 db_printf("%sTF2_NO_ISS_CHECK", comma ? ", " : ""); 3033 comma = 1; 3034 } 3035 } 3036 3037 static void 3038 db_print_toobflags(char t_oobflags) 3039 { 3040 int comma; 3041 3042 comma = 0; 3043 if (t_oobflags & TCPOOB_HAVEDATA) { 3044 db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : ""); 3045 comma = 1; 3046 } 3047 if (t_oobflags & TCPOOB_HADDATA) { 3048 db_printf("%sTCPOOB_HADDATA", comma ? ", " : ""); 3049 comma = 1; 3050 } 3051 } 3052 3053 static void 3054 db_print_bblog_state(int state) 3055 { 3056 switch (state) { 3057 case TCP_LOG_STATE_RATIO_OFF: 3058 db_printf("TCP_LOG_STATE_RATIO_OFF"); 3059 break; 3060 case TCP_LOG_STATE_CLEAR: 3061 db_printf("TCP_LOG_STATE_CLEAR"); 3062 break; 3063 case TCP_LOG_STATE_OFF: 3064 db_printf("TCP_LOG_STATE_OFF"); 3065 break; 3066 case TCP_LOG_STATE_TAIL: 3067 db_printf("TCP_LOG_STATE_TAIL"); 3068 break; 3069 case TCP_LOG_STATE_HEAD: 3070 db_printf("TCP_LOG_STATE_HEAD"); 3071 break; 3072 case TCP_LOG_STATE_HEAD_AUTO: 3073 db_printf("TCP_LOG_STATE_HEAD_AUTO"); 3074 break; 3075 case TCP_LOG_STATE_CONTINUAL: 3076 db_printf("TCP_LOG_STATE_CONTINUAL"); 3077 break; 3078 case TCP_LOG_STATE_TAIL_AUTO: 3079 db_printf("TCP_LOG_STATE_TAIL_AUTO"); 3080 break; 3081 case TCP_LOG_VIA_BBPOINTS: 3082 db_printf("TCP_LOG_STATE_BBPOINTS"); 3083 break; 3084 default: 3085 db_printf("UNKNOWN(%d)", state); 3086 break; 3087 } 3088 } 3089 3090 static void 3091 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent, bool show_bblog) 3092 { 3093 3094 db_print_indent(indent); 3095 db_printf("%s at %p\n", name, tp); 3096 3097 indent += 2; 3098 3099 db_print_indent(indent); 3100 db_printf("t_segq first: %p t_segqlen: %d t_dupacks: %d\n", 3101 TAILQ_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks); 3102 3103 db_print_indent(indent); 3104 db_printf("t_callout: %p t_timers: %p\n", 3105 &tp->t_callout, &tp->t_timers); 3106 3107 db_print_indent(indent); 3108 db_printf("t_state: %d (", tp->t_state); 3109 db_print_tstate(tp->t_state); 3110 db_printf(")\n"); 3111 3112 db_print_indent(indent); 3113 db_printf("t_flags: 0x%x (", tp->t_flags); 3114 db_print_tflags(tp->t_flags); 3115 db_printf(")\n"); 3116 3117 db_print_indent(indent); 3118 db_printf("t_flags2: 0x%x (", tp->t_flags2); 3119 db_print_tflags2(tp->t_flags2); 3120 db_printf(")\n"); 3121 3122 db_print_indent(indent); 3123 db_printf("snd_una: 0x%08x snd_max: 0x%08x snd_nxt: 0x%08x\n", 3124 tp->snd_una, tp->snd_max, tp->snd_nxt); 3125 3126 db_print_indent(indent); 3127 db_printf("snd_up: 0x%08x snd_wl1: 0x%08x snd_wl2: 0x%08x\n", 3128 tp->snd_up, tp->snd_wl1, tp->snd_wl2); 3129 3130 db_print_indent(indent); 3131 db_printf("iss: 0x%08x irs: 0x%08x rcv_nxt: 0x%08x\n", 3132 tp->iss, tp->irs, tp->rcv_nxt); 3133 3134 db_print_indent(indent); 3135 db_printf("rcv_adv: 0x%08x rcv_wnd: %u rcv_up: 0x%08x\n", 3136 tp->rcv_adv, tp->rcv_wnd, tp->rcv_up); 3137 3138 db_print_indent(indent); 3139 db_printf("snd_wnd: %u snd_cwnd: %u\n", 3140 tp->snd_wnd, tp->snd_cwnd); 3141 3142 db_print_indent(indent); 3143 db_printf("snd_ssthresh: %u snd_recover: " 3144 "0x%08x\n", tp->snd_ssthresh, tp->snd_recover); 3145 3146 db_print_indent(indent); 3147 db_printf("t_rcvtime: %u t_startime: %u\n", 3148 tp->t_rcvtime, tp->t_starttime); 3149 3150 db_print_indent(indent); 3151 db_printf("t_rttime: %u t_rtsq: 0x%08x\n", 3152 tp->t_rtttime, tp->t_rtseq); 3153 3154 db_print_indent(indent); 3155 db_printf("t_rxtcur: %d t_maxseg: %u t_srtt: %d\n", 3156 tp->t_rxtcur, tp->t_maxseg, tp->t_srtt); 3157 3158 db_print_indent(indent); 3159 db_printf("t_rttvar: %d t_rxtshift: %d t_rttmin: %u\n", 3160 tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin); 3161 3162 db_print_indent(indent); 3163 db_printf("t_rttupdated: %u max_sndwnd: %u t_softerror: %d\n", 3164 tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror); 3165 3166 db_print_indent(indent); 3167 db_printf("t_oobflags: 0x%x (", tp->t_oobflags); 3168 db_print_toobflags(tp->t_oobflags); 3169 db_printf(") t_iobc: 0x%02x\n", tp->t_iobc); 3170 3171 db_print_indent(indent); 3172 db_printf("snd_scale: %u rcv_scale: %u request_r_scale: %u\n", 3173 tp->snd_scale, tp->rcv_scale, tp->request_r_scale); 3174 3175 db_print_indent(indent); 3176 db_printf("ts_recent: %u ts_recent_age: %u\n", 3177 tp->ts_recent, tp->ts_recent_age); 3178 3179 db_print_indent(indent); 3180 db_printf("ts_offset: %u last_ack_sent: 0x%08x snd_cwnd_prev: " 3181 "%u\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev); 3182 3183 db_print_indent(indent); 3184 db_printf("snd_ssthresh_prev: %u snd_recover_prev: 0x%08x " 3185 "t_badrxtwin: %u\n", tp->snd_ssthresh_prev, 3186 tp->snd_recover_prev, tp->t_badrxtwin); 3187 3188 db_print_indent(indent); 3189 db_printf("snd_numholes: %d snd_holes first: %p\n", 3190 tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes)); 3191 3192 db_print_indent(indent); 3193 db_printf("snd_fack: 0x%08x rcv_numsacks: %d\n", 3194 tp->snd_fack, tp->rcv_numsacks); 3195 3196 /* Skip sackblks, sackhint. */ 3197 3198 db_print_indent(indent); 3199 db_printf("t_rttlow: %d rfbuf_ts: %u rfbuf_cnt: %d\n", 3200 tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt); 3201 3202 db_print_indent(indent); 3203 db_printf("t_fb.tfb_tcp_block_name: %s\n", tp->t_fb->tfb_tcp_block_name); 3204 3205 db_print_indent(indent); 3206 db_printf("t_cc.name: %s\n", tp->t_cc->name); 3207 3208 db_print_indent(indent); 3209 db_printf("_t_logstate: %d (", tp->_t_logstate); 3210 db_print_bblog_state(tp->_t_logstate); 3211 db_printf(")\n"); 3212 3213 db_print_indent(indent); 3214 db_printf("t_lognum: %d t_loglimit: %d t_logsn: %u\n", 3215 tp->t_lognum, tp->t_loglimit, tp->t_logsn); 3216 3217 if (show_bblog) { 3218 #ifdef TCP_BLACKBOX 3219 db_print_bblog_entries(&tp->t_logs, indent); 3220 #else 3221 db_print_indent(indent); 3222 db_printf("BBLog not supported\n"); 3223 #endif 3224 } 3225 } 3226 3227 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb) 3228 { 3229 struct tcpcb *tp; 3230 bool show_bblog; 3231 3232 if (!have_addr) { 3233 db_printf("usage: show tcpcb <addr>\n"); 3234 return; 3235 } 3236 show_bblog = strchr(modif, 'b') != NULL; 3237 tp = (struct tcpcb *)addr; 3238 3239 db_print_tcpcb(tp, "tcpcb", 0, show_bblog); 3240 } 3241 3242 DB_SHOW_ALL_COMMAND(tcpcbs, db_show_all_tcpcbs) 3243 { 3244 VNET_ITERATOR_DECL(vnet_iter); 3245 struct inpcb *inp; 3246 bool only_locked, show_bblog; 3247 3248 only_locked = strchr(modif, 'l') != NULL; 3249 show_bblog = strchr(modif, 'b') != NULL; 3250 VNET_FOREACH(vnet_iter) { 3251 CURVNET_SET(vnet_iter); 3252 CK_LIST_FOREACH(inp, &V_tcbinfo.ipi_listhead, inp_list) { 3253 if (only_locked && 3254 inp->inp_lock.rw_lock == RW_UNLOCKED) 3255 continue; 3256 db_print_tcpcb(intotcpcb(inp), "tcpcb", 0, show_bblog); 3257 if (db_pager_quit) 3258 break; 3259 } 3260 CURVNET_RESTORE(); 3261 if (db_pager_quit) 3262 break; 3263 } 3264 } 3265 #endif 3266