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 * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94 38 */ 39 40 #include <sys/cdefs.h> 41 __FBSDID("$FreeBSD$"); 42 43 #include "opt_ddb.h" 44 #include "opt_inet.h" 45 #include "opt_inet6.h" 46 #include "opt_ipsec.h" 47 #include "opt_kern_tls.h" 48 49 #include <sys/param.h> 50 #include <sys/systm.h> 51 #include <sys/arb.h> 52 #include <sys/limits.h> 53 #include <sys/malloc.h> 54 #include <sys/refcount.h> 55 #include <sys/kernel.h> 56 #include <sys/ktls.h> 57 #include <sys/qmath.h> 58 #include <sys/sysctl.h> 59 #include <sys/mbuf.h> 60 #ifdef INET6 61 #include <sys/domain.h> 62 #endif /* INET6 */ 63 #include <sys/socket.h> 64 #include <sys/socketvar.h> 65 #include <sys/protosw.h> 66 #include <sys/proc.h> 67 #include <sys/jail.h> 68 #include <sys/stats.h> 69 70 #ifdef DDB 71 #include <ddb/ddb.h> 72 #endif 73 74 #include <net/if.h> 75 #include <net/if_var.h> 76 #include <net/route.h> 77 #include <net/vnet.h> 78 79 #include <netinet/in.h> 80 #include <netinet/in_kdtrace.h> 81 #include <netinet/in_pcb.h> 82 #include <netinet/in_systm.h> 83 #include <netinet/in_var.h> 84 #include <netinet/ip.h> 85 #include <netinet/ip_var.h> 86 #ifdef INET6 87 #include <netinet/ip6.h> 88 #include <netinet6/in6_pcb.h> 89 #include <netinet6/ip6_var.h> 90 #include <netinet6/scope6_var.h> 91 #endif 92 #include <netinet/tcp.h> 93 #include <netinet/tcp_fsm.h> 94 #include <netinet/tcp_seq.h> 95 #include <netinet/tcp_timer.h> 96 #include <netinet/tcp_var.h> 97 #include <netinet/tcp_log_buf.h> 98 #include <netinet/tcpip.h> 99 #include <netinet/cc/cc.h> 100 #include <netinet/tcp_fastopen.h> 101 #include <netinet/tcp_hpts.h> 102 #ifdef TCPPCAP 103 #include <netinet/tcp_pcap.h> 104 #endif 105 #ifdef TCP_OFFLOAD 106 #include <netinet/tcp_offload.h> 107 #endif 108 #include <netipsec/ipsec_support.h> 109 110 #include <vm/vm.h> 111 #include <vm/vm_param.h> 112 #include <vm/pmap.h> 113 #include <vm/vm_extern.h> 114 #include <vm/vm_map.h> 115 #include <vm/vm_page.h> 116 117 /* 118 * TCP protocol interface to socket abstraction. 119 */ 120 #ifdef INET 121 static int tcp_connect(struct tcpcb *, struct sockaddr_in *, 122 struct thread *td); 123 #endif /* INET */ 124 #ifdef INET6 125 static int tcp6_connect(struct tcpcb *, struct sockaddr_in6 *, 126 struct thread *td); 127 #endif /* INET6 */ 128 static void tcp_disconnect(struct tcpcb *); 129 static void tcp_usrclosed(struct tcpcb *); 130 static void tcp_fill_info(struct tcpcb *, struct tcp_info *); 131 132 static int tcp_pru_options_support(struct tcpcb *tp, int flags); 133 134 /* 135 * TCP attaches to socket via pru_attach(), reserving space, 136 * and an internet control block. 137 */ 138 static int 139 tcp_usr_attach(struct socket *so, int proto, struct thread *td) 140 { 141 struct inpcb *inp; 142 struct tcpcb *tp = NULL; 143 int error; 144 145 inp = sotoinpcb(so); 146 KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL")); 147 148 error = soreserve(so, V_tcp_sendspace, V_tcp_recvspace); 149 if (error) 150 goto out; 151 152 so->so_rcv.sb_flags |= SB_AUTOSIZE; 153 so->so_snd.sb_flags |= SB_AUTOSIZE; 154 error = in_pcballoc(so, &V_tcbinfo); 155 if (error) 156 goto out; 157 inp = sotoinpcb(so); 158 tp = tcp_newtcpcb(inp); 159 if (tp == NULL) { 160 error = ENOBUFS; 161 in_pcbdetach(inp); 162 in_pcbfree(inp); 163 goto out; 164 } 165 tp->t_state = TCPS_CLOSED; 166 INP_WUNLOCK(inp); 167 TCPSTATES_INC(TCPS_CLOSED); 168 out: 169 TCP_PROBE2(debug__user, tp, PRU_ATTACH); 170 return (error); 171 } 172 173 /* 174 * tcp_usr_detach is called when the socket layer loses its final reference 175 * to the socket, be it a file descriptor reference, a reference from TCP, 176 * etc. At this point, there is only one case in which we will keep around 177 * inpcb state: time wait. 178 */ 179 static void 180 tcp_usr_detach(struct socket *so) 181 { 182 struct inpcb *inp; 183 struct tcpcb *tp; 184 185 inp = sotoinpcb(so); 186 KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); 187 INP_WLOCK(inp); 188 KASSERT(so->so_pcb == inp && inp->inp_socket == so, 189 ("%s: socket %p inp %p mismatch", __func__, so, inp)); 190 191 tp = intotcpcb(inp); 192 193 KASSERT(inp->inp_flags & INP_DROPPED || 194 tp->t_state < TCPS_SYN_SENT, 195 ("%s: inp %p not dropped or embryonic", __func__, inp)); 196 197 tcp_discardcb(tp); 198 in_pcbdetach(inp); 199 in_pcbfree(inp); 200 } 201 202 #ifdef INET 203 /* 204 * Give the socket an address. 205 */ 206 static int 207 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 208 { 209 int error = 0; 210 struct inpcb *inp; 211 #ifdef KDTRACE_HOOKS 212 struct tcpcb *tp = NULL; 213 #endif 214 struct sockaddr_in *sinp; 215 216 sinp = (struct sockaddr_in *)nam; 217 if (nam->sa_family != AF_INET) { 218 /* 219 * Preserve compatibility with old programs. 220 */ 221 if (nam->sa_family != AF_UNSPEC || 222 nam->sa_len < offsetof(struct sockaddr_in, sin_zero) || 223 sinp->sin_addr.s_addr != INADDR_ANY) 224 return (EAFNOSUPPORT); 225 nam->sa_family = AF_INET; 226 } 227 if (nam->sa_len != sizeof(*sinp)) 228 return (EINVAL); 229 230 /* 231 * Must check for multicast addresses and disallow binding 232 * to them. 233 */ 234 if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) 235 return (EAFNOSUPPORT); 236 237 inp = sotoinpcb(so); 238 KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL")); 239 INP_WLOCK(inp); 240 if (inp->inp_flags & INP_DROPPED) { 241 error = EINVAL; 242 goto out; 243 } 244 #ifdef KDTRACE_HOOKS 245 tp = intotcpcb(inp); 246 #endif 247 INP_HASH_WLOCK(&V_tcbinfo); 248 error = in_pcbbind(inp, nam, td->td_ucred); 249 INP_HASH_WUNLOCK(&V_tcbinfo); 250 out: 251 TCP_PROBE2(debug__user, tp, PRU_BIND); 252 INP_WUNLOCK(inp); 253 254 return (error); 255 } 256 #endif /* INET */ 257 258 #ifdef INET6 259 static int 260 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 261 { 262 int error = 0; 263 struct inpcb *inp; 264 #ifdef KDTRACE_HOOKS 265 struct tcpcb *tp = NULL; 266 #endif 267 struct sockaddr_in6 *sin6; 268 u_char vflagsav; 269 270 sin6 = (struct sockaddr_in6 *)nam; 271 if (nam->sa_family != AF_INET6) 272 return (EAFNOSUPPORT); 273 if (nam->sa_len != sizeof(*sin6)) 274 return (EINVAL); 275 276 /* 277 * Must check for multicast addresses and disallow binding 278 * to them. 279 */ 280 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 281 return (EAFNOSUPPORT); 282 283 inp = sotoinpcb(so); 284 KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL")); 285 INP_WLOCK(inp); 286 vflagsav = inp->inp_vflag; 287 if (inp->inp_flags & INP_DROPPED) { 288 error = EINVAL; 289 goto out; 290 } 291 #ifdef KDTRACE_HOOKS 292 tp = intotcpcb(inp); 293 #endif 294 INP_HASH_WLOCK(&V_tcbinfo); 295 inp->inp_vflag &= ~INP_IPV4; 296 inp->inp_vflag |= INP_IPV6; 297 #ifdef INET 298 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { 299 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) 300 inp->inp_vflag |= INP_IPV4; 301 else if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { 302 struct sockaddr_in sin; 303 304 in6_sin6_2_sin(&sin, sin6); 305 if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) { 306 error = EAFNOSUPPORT; 307 INP_HASH_WUNLOCK(&V_tcbinfo); 308 goto out; 309 } 310 inp->inp_vflag |= INP_IPV4; 311 inp->inp_vflag &= ~INP_IPV6; 312 error = in_pcbbind(inp, (struct sockaddr *)&sin, 313 td->td_ucred); 314 INP_HASH_WUNLOCK(&V_tcbinfo); 315 goto out; 316 } 317 } 318 #endif 319 error = in6_pcbbind(inp, nam, td->td_ucred); 320 INP_HASH_WUNLOCK(&V_tcbinfo); 321 out: 322 if (error != 0) 323 inp->inp_vflag = vflagsav; 324 TCP_PROBE2(debug__user, tp, PRU_BIND); 325 INP_WUNLOCK(inp); 326 return (error); 327 } 328 #endif /* INET6 */ 329 330 #ifdef INET 331 /* 332 * Prepare to accept connections. 333 */ 334 static int 335 tcp_usr_listen(struct socket *so, int backlog, struct thread *td) 336 { 337 int error = 0; 338 struct inpcb *inp; 339 struct tcpcb *tp = NULL; 340 341 inp = sotoinpcb(so); 342 KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL")); 343 INP_WLOCK(inp); 344 if (inp->inp_flags & INP_DROPPED) { 345 error = EINVAL; 346 goto out; 347 } 348 tp = intotcpcb(inp); 349 SOCK_LOCK(so); 350 error = solisten_proto_check(so); 351 if (error != 0) { 352 SOCK_UNLOCK(so); 353 goto out; 354 } 355 if (inp->inp_lport == 0) { 356 INP_HASH_WLOCK(&V_tcbinfo); 357 error = in_pcbbind(inp, NULL, td->td_ucred); 358 INP_HASH_WUNLOCK(&V_tcbinfo); 359 } 360 if (error == 0) { 361 tcp_state_change(tp, TCPS_LISTEN); 362 solisten_proto(so, backlog); 363 #ifdef TCP_OFFLOAD 364 if ((so->so_options & SO_NO_OFFLOAD) == 0) 365 tcp_offload_listen_start(tp); 366 #endif 367 } else { 368 solisten_proto_abort(so); 369 } 370 SOCK_UNLOCK(so); 371 372 if (IS_FASTOPEN(tp->t_flags)) 373 tp->t_tfo_pending = tcp_fastopen_alloc_counter(); 374 375 out: 376 TCP_PROBE2(debug__user, tp, PRU_LISTEN); 377 INP_WUNLOCK(inp); 378 return (error); 379 } 380 #endif /* INET */ 381 382 #ifdef INET6 383 static int 384 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td) 385 { 386 int error = 0; 387 struct inpcb *inp; 388 struct tcpcb *tp = NULL; 389 u_char vflagsav; 390 391 inp = sotoinpcb(so); 392 KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL")); 393 INP_WLOCK(inp); 394 if (inp->inp_flags & INP_DROPPED) { 395 error = EINVAL; 396 goto out; 397 } 398 vflagsav = inp->inp_vflag; 399 tp = intotcpcb(inp); 400 SOCK_LOCK(so); 401 error = solisten_proto_check(so); 402 if (error != 0) { 403 SOCK_UNLOCK(so); 404 goto out; 405 } 406 INP_HASH_WLOCK(&V_tcbinfo); 407 if (inp->inp_lport == 0) { 408 inp->inp_vflag &= ~INP_IPV4; 409 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) 410 inp->inp_vflag |= INP_IPV4; 411 error = in6_pcbbind(inp, NULL, td->td_ucred); 412 } 413 INP_HASH_WUNLOCK(&V_tcbinfo); 414 if (error == 0) { 415 tcp_state_change(tp, TCPS_LISTEN); 416 solisten_proto(so, backlog); 417 #ifdef TCP_OFFLOAD 418 if ((so->so_options & SO_NO_OFFLOAD) == 0) 419 tcp_offload_listen_start(tp); 420 #endif 421 } else { 422 solisten_proto_abort(so); 423 } 424 SOCK_UNLOCK(so); 425 426 if (IS_FASTOPEN(tp->t_flags)) 427 tp->t_tfo_pending = tcp_fastopen_alloc_counter(); 428 429 if (error != 0) 430 inp->inp_vflag = vflagsav; 431 432 out: 433 TCP_PROBE2(debug__user, tp, PRU_LISTEN); 434 INP_WUNLOCK(inp); 435 return (error); 436 } 437 #endif /* INET6 */ 438 439 #ifdef INET 440 /* 441 * Initiate connection to peer. 442 * Create a template for use in transmissions on this connection. 443 * Enter SYN_SENT state, and mark socket as connecting. 444 * Start keep-alive timer, and seed output sequence space. 445 * Send initial segment on connection. 446 */ 447 static int 448 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 449 { 450 struct epoch_tracker et; 451 int error = 0; 452 struct inpcb *inp; 453 struct tcpcb *tp = NULL; 454 struct sockaddr_in *sinp; 455 456 sinp = (struct sockaddr_in *)nam; 457 if (nam->sa_family != AF_INET) 458 return (EAFNOSUPPORT); 459 if (nam->sa_len != sizeof (*sinp)) 460 return (EINVAL); 461 462 /* 463 * Must disallow TCP ``connections'' to multicast addresses. 464 */ 465 if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) 466 return (EAFNOSUPPORT); 467 if (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST) 468 return (EACCES); 469 if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0) 470 return (error); 471 472 inp = sotoinpcb(so); 473 KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL")); 474 INP_WLOCK(inp); 475 if (inp->inp_flags & INP_DROPPED) { 476 error = ECONNREFUSED; 477 goto out; 478 } 479 if (SOLISTENING(so)) { 480 error = EOPNOTSUPP; 481 goto out; 482 } 483 tp = intotcpcb(inp); 484 NET_EPOCH_ENTER(et); 485 if ((error = tcp_connect(tp, sinp, td)) != 0) 486 goto out_in_epoch; 487 #ifdef TCP_OFFLOAD 488 if (registered_toedevs > 0 && 489 (so->so_options & SO_NO_OFFLOAD) == 0 && 490 (error = tcp_offload_connect(so, nam)) == 0) 491 goto out_in_epoch; 492 #endif 493 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp)); 494 error = tcp_output(tp); 495 KASSERT(error >= 0, ("TCP stack %s requested tcp_drop(%p) at connect()" 496 ", error code %d", tp->t_fb->tfb_tcp_block_name, tp, -error)); 497 out_in_epoch: 498 NET_EPOCH_EXIT(et); 499 out: 500 TCP_PROBE2(debug__user, tp, PRU_CONNECT); 501 INP_WUNLOCK(inp); 502 return (error); 503 } 504 #endif /* INET */ 505 506 #ifdef INET6 507 static int 508 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 509 { 510 struct epoch_tracker et; 511 int error = 0; 512 struct inpcb *inp; 513 struct tcpcb *tp = NULL; 514 struct sockaddr_in6 *sin6; 515 u_int8_t incflagsav; 516 u_char vflagsav; 517 518 sin6 = (struct sockaddr_in6 *)nam; 519 if (nam->sa_family != AF_INET6) 520 return (EAFNOSUPPORT); 521 if (nam->sa_len != sizeof (*sin6)) 522 return (EINVAL); 523 524 /* 525 * Must disallow TCP ``connections'' to multicast addresses. 526 */ 527 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 528 return (EAFNOSUPPORT); 529 530 inp = sotoinpcb(so); 531 KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL")); 532 INP_WLOCK(inp); 533 vflagsav = inp->inp_vflag; 534 incflagsav = inp->inp_inc.inc_flags; 535 if (inp->inp_flags & INP_DROPPED) { 536 error = ECONNREFUSED; 537 goto out; 538 } 539 if (SOLISTENING(so)) { 540 error = EINVAL; 541 goto out; 542 } 543 tp = intotcpcb(inp); 544 #ifdef INET 545 /* 546 * XXXRW: Some confusion: V4/V6 flags relate to binding, and 547 * therefore probably require the hash lock, which isn't held here. 548 * Is this a significant problem? 549 */ 550 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { 551 struct sockaddr_in sin; 552 553 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { 554 error = EINVAL; 555 goto out; 556 } 557 if ((inp->inp_vflag & INP_IPV4) == 0) { 558 error = EAFNOSUPPORT; 559 goto out; 560 } 561 562 in6_sin6_2_sin(&sin, sin6); 563 if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) { 564 error = EAFNOSUPPORT; 565 goto out; 566 } 567 if (ntohl(sin.sin_addr.s_addr) == INADDR_BROADCAST) { 568 error = EACCES; 569 goto out; 570 } 571 if ((error = prison_remote_ip4(td->td_ucred, 572 &sin.sin_addr)) != 0) 573 goto out; 574 inp->inp_vflag |= INP_IPV4; 575 inp->inp_vflag &= ~INP_IPV6; 576 NET_EPOCH_ENTER(et); 577 if ((error = tcp_connect(tp, &sin, td)) != 0) 578 goto out_in_epoch; 579 #ifdef TCP_OFFLOAD 580 if (registered_toedevs > 0 && 581 (so->so_options & SO_NO_OFFLOAD) == 0 && 582 (error = tcp_offload_connect(so, nam)) == 0) 583 goto out_in_epoch; 584 #endif 585 error = tcp_output(tp); 586 goto out_in_epoch; 587 } else { 588 if ((inp->inp_vflag & INP_IPV6) == 0) { 589 error = EAFNOSUPPORT; 590 goto out; 591 } 592 } 593 #endif 594 if ((error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr)) != 0) 595 goto out; 596 inp->inp_vflag &= ~INP_IPV4; 597 inp->inp_vflag |= INP_IPV6; 598 inp->inp_inc.inc_flags |= INC_ISIPV6; 599 NET_EPOCH_ENTER(et); 600 if ((error = tcp6_connect(tp, sin6, td)) != 0) 601 goto out_in_epoch; 602 #ifdef TCP_OFFLOAD 603 if (registered_toedevs > 0 && 604 (so->so_options & SO_NO_OFFLOAD) == 0 && 605 (error = tcp_offload_connect(so, nam)) == 0) 606 goto out_in_epoch; 607 #endif 608 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp)); 609 error = tcp_output(tp); 610 out_in_epoch: 611 NET_EPOCH_EXIT(et); 612 out: 613 KASSERT(error >= 0, ("TCP stack %s requested tcp_drop(%p) at connect()" 614 ", error code %d", tp->t_fb->tfb_tcp_block_name, tp, -error)); 615 /* 616 * If the implicit bind in the connect call fails, restore 617 * the flags we modified. 618 */ 619 if (error != 0 && inp->inp_lport == 0) { 620 inp->inp_vflag = vflagsav; 621 inp->inp_inc.inc_flags = incflagsav; 622 } 623 624 TCP_PROBE2(debug__user, tp, PRU_CONNECT); 625 INP_WUNLOCK(inp); 626 return (error); 627 } 628 #endif /* INET6 */ 629 630 /* 631 * Initiate disconnect from peer. 632 * If connection never passed embryonic stage, just drop; 633 * else if don't need to let data drain, then can just drop anyways, 634 * else have to begin TCP shutdown process: mark socket disconnecting, 635 * drain unread data, state switch to reflect user close, and 636 * send segment (e.g. FIN) to peer. Socket will be really disconnected 637 * when peer sends FIN and acks ours. 638 * 639 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 640 */ 641 static int 642 tcp_usr_disconnect(struct socket *so) 643 { 644 struct inpcb *inp; 645 struct tcpcb *tp = NULL; 646 struct epoch_tracker et; 647 int error = 0; 648 649 NET_EPOCH_ENTER(et); 650 inp = sotoinpcb(so); 651 KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL")); 652 INP_WLOCK(inp); 653 if (inp->inp_flags & INP_DROPPED) { 654 error = ECONNRESET; 655 goto out; 656 } 657 tp = intotcpcb(inp); 658 tcp_disconnect(tp); 659 out: 660 TCP_PROBE2(debug__user, tp, PRU_DISCONNECT); 661 INP_WUNLOCK(inp); 662 NET_EPOCH_EXIT(et); 663 return (error); 664 } 665 666 #ifdef INET 667 /* 668 * Accept a connection. Essentially all the work is done at higher levels; 669 * just return the address of the peer, storing through addr. 670 */ 671 static int 672 tcp_usr_accept(struct socket *so, struct sockaddr **nam) 673 { 674 int error = 0; 675 struct inpcb *inp = NULL; 676 #ifdef KDTRACE_HOOKS 677 struct tcpcb *tp = NULL; 678 #endif 679 struct in_addr addr; 680 in_port_t port = 0; 681 682 if (so->so_state & SS_ISDISCONNECTED) 683 return (ECONNABORTED); 684 685 inp = sotoinpcb(so); 686 KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL")); 687 INP_WLOCK(inp); 688 if (inp->inp_flags & INP_DROPPED) { 689 error = ECONNABORTED; 690 goto out; 691 } 692 #ifdef KDTRACE_HOOKS 693 tp = intotcpcb(inp); 694 #endif 695 696 /* 697 * We inline in_getpeeraddr and COMMON_END here, so that we can 698 * copy the data of interest and defer the malloc until after we 699 * release the lock. 700 */ 701 port = inp->inp_fport; 702 addr = inp->inp_faddr; 703 704 out: 705 TCP_PROBE2(debug__user, tp, PRU_ACCEPT); 706 INP_WUNLOCK(inp); 707 if (error == 0) 708 *nam = in_sockaddr(port, &addr); 709 return error; 710 } 711 #endif /* INET */ 712 713 #ifdef INET6 714 static int 715 tcp6_usr_accept(struct socket *so, struct sockaddr **nam) 716 { 717 struct inpcb *inp = NULL; 718 int error = 0; 719 #ifdef KDTRACE_HOOKS 720 struct tcpcb *tp = NULL; 721 #endif 722 struct in_addr addr; 723 struct in6_addr addr6; 724 struct epoch_tracker et; 725 in_port_t port = 0; 726 int v4 = 0; 727 728 if (so->so_state & SS_ISDISCONNECTED) 729 return (ECONNABORTED); 730 731 inp = sotoinpcb(so); 732 KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL")); 733 NET_EPOCH_ENTER(et); 734 INP_WLOCK(inp); 735 if (inp->inp_flags & INP_DROPPED) { 736 error = ECONNABORTED; 737 goto out; 738 } 739 #ifdef KDTRACE_HOOKS 740 tp = intotcpcb(inp); 741 #endif 742 743 /* 744 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can 745 * copy the data of interest and defer the malloc until after we 746 * release the lock. 747 */ 748 if (inp->inp_vflag & INP_IPV4) { 749 v4 = 1; 750 port = inp->inp_fport; 751 addr = inp->inp_faddr; 752 } else { 753 port = inp->inp_fport; 754 addr6 = inp->in6p_faddr; 755 } 756 757 out: 758 TCP_PROBE2(debug__user, tp, PRU_ACCEPT); 759 INP_WUNLOCK(inp); 760 NET_EPOCH_EXIT(et); 761 if (error == 0) { 762 if (v4) 763 *nam = in6_v4mapsin6_sockaddr(port, &addr); 764 else 765 *nam = in6_sockaddr(port, &addr6); 766 } 767 return error; 768 } 769 #endif /* INET6 */ 770 771 /* 772 * Mark the connection as being incapable of further output. 773 */ 774 static int 775 tcp_usr_shutdown(struct socket *so) 776 { 777 int error = 0; 778 struct inpcb *inp; 779 struct tcpcb *tp = NULL; 780 struct epoch_tracker et; 781 782 inp = sotoinpcb(so); 783 KASSERT(inp != NULL, ("inp == NULL")); 784 INP_WLOCK(inp); 785 if (inp->inp_flags & INP_DROPPED) { 786 INP_WUNLOCK(inp); 787 return (ECONNRESET); 788 } 789 tp = intotcpcb(inp); 790 NET_EPOCH_ENTER(et); 791 socantsendmore(so); 792 tcp_usrclosed(tp); 793 if (!(inp->inp_flags & INP_DROPPED)) 794 error = tcp_output_nodrop(tp); 795 TCP_PROBE2(debug__user, tp, PRU_SHUTDOWN); 796 error = tcp_unlock_or_drop(tp, error); 797 NET_EPOCH_EXIT(et); 798 799 return (error); 800 } 801 802 /* 803 * After a receive, possibly send window update to peer. 804 */ 805 static int 806 tcp_usr_rcvd(struct socket *so, int flags) 807 { 808 struct epoch_tracker et; 809 struct inpcb *inp; 810 struct tcpcb *tp = NULL; 811 int outrv = 0, error = 0; 812 813 inp = sotoinpcb(so); 814 KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL")); 815 INP_WLOCK(inp); 816 if (inp->inp_flags & INP_DROPPED) { 817 INP_WUNLOCK(inp); 818 return (ECONNRESET); 819 } 820 tp = intotcpcb(inp); 821 NET_EPOCH_ENTER(et); 822 /* 823 * For passively-created TFO connections, don't attempt a window 824 * update while still in SYN_RECEIVED as this may trigger an early 825 * SYN|ACK. It is preferable to have the SYN|ACK be sent along with 826 * application response data, or failing that, when the DELACK timer 827 * expires. 828 */ 829 if (IS_FASTOPEN(tp->t_flags) && 830 (tp->t_state == TCPS_SYN_RECEIVED)) 831 goto out; 832 #ifdef TCP_OFFLOAD 833 if (tp->t_flags & TF_TOE) 834 tcp_offload_rcvd(tp); 835 else 836 #endif 837 outrv = tcp_output_nodrop(tp); 838 out: 839 TCP_PROBE2(debug__user, tp, PRU_RCVD); 840 (void) tcp_unlock_or_drop(tp, outrv); 841 NET_EPOCH_EXIT(et); 842 return (error); 843 } 844 845 /* 846 * Do a send by putting data in output queue and updating urgent 847 * marker if URG set. Possibly send more data. Unlike the other 848 * pru_*() routines, the mbuf chains are our responsibility. We 849 * must either enqueue them or free them. The other pru_* routines 850 * generally are caller-frees. 851 */ 852 static int 853 tcp_usr_send(struct socket *so, int flags, struct mbuf *m, 854 struct sockaddr *nam, struct mbuf *control, struct thread *td) 855 { 856 struct epoch_tracker et; 857 int error = 0; 858 struct inpcb *inp; 859 struct tcpcb *tp = NULL; 860 #ifdef INET 861 #ifdef INET6 862 struct sockaddr_in sin; 863 #endif 864 struct sockaddr_in *sinp; 865 #endif 866 #ifdef INET6 867 struct sockaddr_in6 *sin6; 868 int isipv6; 869 #endif 870 u_int8_t incflagsav; 871 u_char vflagsav; 872 bool restoreflags; 873 874 if (control != NULL) { 875 /* TCP doesn't do control messages (rights, creds, etc) */ 876 if (control->m_len) { 877 m_freem(control); 878 return (EINVAL); 879 } 880 m_freem(control); /* empty control, just free it */ 881 } 882 883 inp = sotoinpcb(so); 884 KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL")); 885 INP_WLOCK(inp); 886 if (inp->inp_flags & INP_DROPPED) { 887 if (m != NULL && (flags & PRUS_NOTREADY) == 0) 888 m_freem(m); 889 INP_WUNLOCK(inp); 890 return (ECONNRESET); 891 } 892 893 vflagsav = inp->inp_vflag; 894 incflagsav = inp->inp_inc.inc_flags; 895 restoreflags = false; 896 tp = intotcpcb(inp); 897 898 NET_EPOCH_ENTER(et); 899 if ((flags & PRUS_OOB) != 0 && 900 (error = tcp_pru_options_support(tp, PRUS_OOB)) != 0) 901 goto out; 902 903 if (nam != NULL && tp->t_state < TCPS_SYN_SENT) { 904 if (tp->t_state == TCPS_LISTEN) { 905 error = EINVAL; 906 goto out; 907 } 908 switch (nam->sa_family) { 909 #ifdef INET 910 case AF_INET: 911 sinp = (struct sockaddr_in *)nam; 912 if (sinp->sin_len != sizeof(struct sockaddr_in)) { 913 error = EINVAL; 914 goto out; 915 } 916 if ((inp->inp_vflag & INP_IPV6) != 0) { 917 error = EAFNOSUPPORT; 918 goto out; 919 } 920 if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 921 error = EAFNOSUPPORT; 922 goto out; 923 } 924 if (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST) { 925 error = EACCES; 926 goto out; 927 } 928 if ((error = prison_remote_ip4(td->td_ucred, 929 &sinp->sin_addr))) 930 goto out; 931 #ifdef INET6 932 isipv6 = 0; 933 #endif 934 break; 935 #endif /* INET */ 936 #ifdef INET6 937 case AF_INET6: 938 sin6 = (struct sockaddr_in6 *)nam; 939 if (sin6->sin6_len != sizeof(*sin6)) { 940 error = EINVAL; 941 goto out; 942 } 943 if ((inp->inp_vflag & INP_IPV6PROTO) == 0) { 944 error = EAFNOSUPPORT; 945 goto out; 946 } 947 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { 948 error = EAFNOSUPPORT; 949 goto out; 950 } 951 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { 952 #ifdef INET 953 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { 954 error = EINVAL; 955 goto out; 956 } 957 if ((inp->inp_vflag & INP_IPV4) == 0) { 958 error = EAFNOSUPPORT; 959 goto out; 960 } 961 restoreflags = true; 962 inp->inp_vflag &= ~INP_IPV6; 963 sinp = &sin; 964 in6_sin6_2_sin(sinp, sin6); 965 if (IN_MULTICAST( 966 ntohl(sinp->sin_addr.s_addr))) { 967 error = EAFNOSUPPORT; 968 goto out; 969 } 970 if ((error = prison_remote_ip4(td->td_ucred, 971 &sinp->sin_addr))) 972 goto out; 973 isipv6 = 0; 974 #else /* !INET */ 975 error = EAFNOSUPPORT; 976 goto out; 977 #endif /* INET */ 978 } else { 979 if ((inp->inp_vflag & INP_IPV6) == 0) { 980 error = EAFNOSUPPORT; 981 goto out; 982 } 983 restoreflags = true; 984 inp->inp_vflag &= ~INP_IPV4; 985 inp->inp_inc.inc_flags |= INC_ISIPV6; 986 if ((error = prison_remote_ip6(td->td_ucred, 987 &sin6->sin6_addr))) 988 goto out; 989 isipv6 = 1; 990 } 991 break; 992 #endif /* INET6 */ 993 default: 994 error = EAFNOSUPPORT; 995 goto out; 996 } 997 } 998 if (!(flags & PRUS_OOB)) { 999 if (tp->t_acktime == 0) 1000 tp->t_acktime = ticks; 1001 sbappendstream(&so->so_snd, m, flags); 1002 m = NULL; 1003 if (nam && tp->t_state < TCPS_SYN_SENT) { 1004 KASSERT(tp->t_state == TCPS_CLOSED, 1005 ("%s: tp %p is listening", __func__, tp)); 1006 1007 /* 1008 * Do implied connect if not yet connected, 1009 * initialize window to default value, and 1010 * initialize maxseg using peer's cached MSS. 1011 */ 1012 #ifdef INET6 1013 if (isipv6) 1014 error = tcp6_connect(tp, sin6, td); 1015 #endif /* INET6 */ 1016 #if defined(INET6) && defined(INET) 1017 else 1018 #endif 1019 #ifdef INET 1020 error = tcp_connect(tp, sinp, td); 1021 #endif 1022 /* 1023 * The bind operation in tcp_connect succeeded. We 1024 * no longer want to restore the flags if later 1025 * operations fail. 1026 */ 1027 if (error == 0 || inp->inp_lport != 0) 1028 restoreflags = false; 1029 1030 if (error) { 1031 /* m is freed if PRUS_NOTREADY is unset. */ 1032 sbflush(&so->so_snd); 1033 goto out; 1034 } 1035 if (IS_FASTOPEN(tp->t_flags)) 1036 tcp_fastopen_connect(tp); 1037 else { 1038 tp->snd_wnd = TTCP_CLIENT_SND_WND; 1039 tcp_mss(tp, -1); 1040 } 1041 } 1042 if (flags & PRUS_EOF) { 1043 /* 1044 * Close the send side of the connection after 1045 * the data is sent. 1046 */ 1047 socantsendmore(so); 1048 tcp_usrclosed(tp); 1049 } 1050 if (TCPS_HAVEESTABLISHED(tp->t_state) && 1051 ((tp->t_flags2 & TF2_FBYTES_COMPLETE) == 0) && 1052 (tp->t_fbyte_out == 0) && 1053 (so->so_snd.sb_ccc > 0)) { 1054 tp->t_fbyte_out = ticks; 1055 if (tp->t_fbyte_out == 0) 1056 tp->t_fbyte_out = 1; 1057 if (tp->t_fbyte_out && tp->t_fbyte_in) 1058 tp->t_flags2 |= TF2_FBYTES_COMPLETE; 1059 } 1060 if (!(inp->inp_flags & INP_DROPPED) && 1061 !(flags & PRUS_NOTREADY)) { 1062 if (flags & PRUS_MORETOCOME) 1063 tp->t_flags |= TF_MORETOCOME; 1064 error = tcp_output_nodrop(tp); 1065 if (flags & PRUS_MORETOCOME) 1066 tp->t_flags &= ~TF_MORETOCOME; 1067 } 1068 } else { 1069 /* 1070 * XXXRW: PRUS_EOF not implemented with PRUS_OOB? 1071 */ 1072 SOCKBUF_LOCK(&so->so_snd); 1073 if (sbspace(&so->so_snd) < -512) { 1074 SOCKBUF_UNLOCK(&so->so_snd); 1075 error = ENOBUFS; 1076 goto out; 1077 } 1078 /* 1079 * According to RFC961 (Assigned Protocols), 1080 * the urgent pointer points to the last octet 1081 * of urgent data. We continue, however, 1082 * to consider it to indicate the first octet 1083 * of data past the urgent section. 1084 * Otherwise, snd_up should be one lower. 1085 */ 1086 if (tp->t_acktime == 0) 1087 tp->t_acktime = ticks; 1088 sbappendstream_locked(&so->so_snd, m, flags); 1089 SOCKBUF_UNLOCK(&so->so_snd); 1090 m = NULL; 1091 if (nam && tp->t_state < TCPS_SYN_SENT) { 1092 /* 1093 * Do implied connect if not yet connected, 1094 * initialize window to default value, and 1095 * initialize maxseg using peer's cached MSS. 1096 */ 1097 1098 /* 1099 * Not going to contemplate SYN|URG 1100 */ 1101 if (IS_FASTOPEN(tp->t_flags)) 1102 tp->t_flags &= ~TF_FASTOPEN; 1103 #ifdef INET6 1104 if (isipv6) 1105 error = tcp6_connect(tp, sin6, td); 1106 #endif /* INET6 */ 1107 #if defined(INET6) && defined(INET) 1108 else 1109 #endif 1110 #ifdef INET 1111 error = tcp_connect(tp, sinp, td); 1112 #endif 1113 /* 1114 * The bind operation in tcp_connect succeeded. We 1115 * no longer want to restore the flags if later 1116 * operations fail. 1117 */ 1118 if (error == 0 || inp->inp_lport != 0) 1119 restoreflags = false; 1120 1121 if (error != 0) { 1122 /* m is freed if PRUS_NOTREADY is unset. */ 1123 sbflush(&so->so_snd); 1124 goto out; 1125 } 1126 tp->snd_wnd = TTCP_CLIENT_SND_WND; 1127 tcp_mss(tp, -1); 1128 } 1129 tp->snd_up = tp->snd_una + sbavail(&so->so_snd); 1130 if ((flags & PRUS_NOTREADY) == 0) { 1131 tp->t_flags |= TF_FORCEDATA; 1132 error = tcp_output_nodrop(tp); 1133 tp->t_flags &= ~TF_FORCEDATA; 1134 } 1135 } 1136 TCP_LOG_EVENT(tp, NULL, 1137 &inp->inp_socket->so_rcv, 1138 &inp->inp_socket->so_snd, 1139 TCP_LOG_USERSEND, error, 1140 0, NULL, false); 1141 1142 out: 1143 /* 1144 * In case of PRUS_NOTREADY, the caller or tcp_usr_ready() is 1145 * responsible for freeing memory. 1146 */ 1147 if (m != NULL && (flags & PRUS_NOTREADY) == 0) 1148 m_freem(m); 1149 1150 /* 1151 * If the request was unsuccessful and we changed flags, 1152 * restore the original flags. 1153 */ 1154 if (error != 0 && restoreflags) { 1155 inp->inp_vflag = vflagsav; 1156 inp->inp_inc.inc_flags = incflagsav; 1157 } 1158 TCP_PROBE2(debug__user, tp, (flags & PRUS_OOB) ? PRU_SENDOOB : 1159 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); 1160 error = tcp_unlock_or_drop(tp, error); 1161 NET_EPOCH_EXIT(et); 1162 return (error); 1163 } 1164 1165 static int 1166 tcp_usr_ready(struct socket *so, struct mbuf *m, int count) 1167 { 1168 struct epoch_tracker et; 1169 struct inpcb *inp; 1170 struct tcpcb *tp; 1171 int error; 1172 1173 inp = sotoinpcb(so); 1174 INP_WLOCK(inp); 1175 if (inp->inp_flags & INP_DROPPED) { 1176 INP_WUNLOCK(inp); 1177 mb_free_notready(m, count); 1178 return (ECONNRESET); 1179 } 1180 tp = intotcpcb(inp); 1181 1182 SOCKBUF_LOCK(&so->so_snd); 1183 error = sbready(&so->so_snd, m, count); 1184 SOCKBUF_UNLOCK(&so->so_snd); 1185 if (error) { 1186 INP_WUNLOCK(inp); 1187 return (error); 1188 } 1189 NET_EPOCH_ENTER(et); 1190 error = tcp_output_unlock(tp); 1191 NET_EPOCH_EXIT(et); 1192 1193 return (error); 1194 } 1195 1196 /* 1197 * Abort the TCP. Drop the connection abruptly. 1198 */ 1199 static void 1200 tcp_usr_abort(struct socket *so) 1201 { 1202 struct inpcb *inp; 1203 struct tcpcb *tp = NULL; 1204 struct epoch_tracker et; 1205 1206 inp = sotoinpcb(so); 1207 KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL")); 1208 1209 NET_EPOCH_ENTER(et); 1210 INP_WLOCK(inp); 1211 KASSERT(inp->inp_socket != NULL, 1212 ("tcp_usr_abort: inp_socket == NULL")); 1213 1214 /* 1215 * If we still have full TCP state, and we're not dropped, drop. 1216 */ 1217 if (!(inp->inp_flags & INP_DROPPED)) { 1218 tp = intotcpcb(inp); 1219 tp = tcp_drop(tp, ECONNABORTED); 1220 if (tp == NULL) 1221 goto dropped; 1222 TCP_PROBE2(debug__user, tp, PRU_ABORT); 1223 } 1224 if (!(inp->inp_flags & INP_DROPPED)) { 1225 soref(so); 1226 inp->inp_flags |= INP_SOCKREF; 1227 } 1228 INP_WUNLOCK(inp); 1229 dropped: 1230 NET_EPOCH_EXIT(et); 1231 } 1232 1233 /* 1234 * TCP socket is closed. Start friendly disconnect. 1235 */ 1236 static void 1237 tcp_usr_close(struct socket *so) 1238 { 1239 struct inpcb *inp; 1240 struct tcpcb *tp = NULL; 1241 struct epoch_tracker et; 1242 1243 inp = sotoinpcb(so); 1244 KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL")); 1245 1246 NET_EPOCH_ENTER(et); 1247 INP_WLOCK(inp); 1248 KASSERT(inp->inp_socket != NULL, 1249 ("tcp_usr_close: inp_socket == NULL")); 1250 1251 /* 1252 * If we still have full TCP state, and we're not dropped, initiate 1253 * a disconnect. 1254 */ 1255 if (!(inp->inp_flags & INP_DROPPED)) { 1256 tp = intotcpcb(inp); 1257 tp->t_flags |= TF_CLOSED; 1258 tcp_disconnect(tp); 1259 TCP_PROBE2(debug__user, tp, PRU_CLOSE); 1260 } 1261 if (!(inp->inp_flags & INP_DROPPED)) { 1262 soref(so); 1263 inp->inp_flags |= INP_SOCKREF; 1264 } 1265 INP_WUNLOCK(inp); 1266 NET_EPOCH_EXIT(et); 1267 } 1268 1269 static int 1270 tcp_pru_options_support(struct tcpcb *tp, int flags) 1271 { 1272 /* 1273 * If the specific TCP stack has a pru_options 1274 * specified then it does not always support 1275 * all the PRU_XX options and we must ask it. 1276 * If the function is not specified then all 1277 * of the PRU_XX options are supported. 1278 */ 1279 int ret = 0; 1280 1281 if (tp->t_fb->tfb_pru_options) { 1282 ret = (*tp->t_fb->tfb_pru_options)(tp, flags); 1283 } 1284 return (ret); 1285 } 1286 1287 /* 1288 * Receive out-of-band data. 1289 */ 1290 static int 1291 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) 1292 { 1293 int error = 0; 1294 struct inpcb *inp; 1295 struct tcpcb *tp = NULL; 1296 1297 inp = sotoinpcb(so); 1298 KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL")); 1299 INP_WLOCK(inp); 1300 if (inp->inp_flags & INP_DROPPED) { 1301 error = ECONNRESET; 1302 goto out; 1303 } 1304 tp = intotcpcb(inp); 1305 error = tcp_pru_options_support(tp, PRUS_OOB); 1306 if (error) { 1307 goto out; 1308 } 1309 if ((so->so_oobmark == 0 && 1310 (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) || 1311 so->so_options & SO_OOBINLINE || 1312 tp->t_oobflags & TCPOOB_HADDATA) { 1313 error = EINVAL; 1314 goto out; 1315 } 1316 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 1317 error = EWOULDBLOCK; 1318 goto out; 1319 } 1320 m->m_len = 1; 1321 *mtod(m, caddr_t) = tp->t_iobc; 1322 if ((flags & MSG_PEEK) == 0) 1323 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 1324 1325 out: 1326 TCP_PROBE2(debug__user, tp, PRU_RCVOOB); 1327 INP_WUNLOCK(inp); 1328 return (error); 1329 } 1330 1331 #ifdef INET 1332 struct protosw tcp_protosw = { 1333 .pr_type = SOCK_STREAM, 1334 .pr_protocol = IPPROTO_TCP, 1335 .pr_flags = PR_CONNREQUIRED | PR_IMPLOPCL | PR_WANTRCVD | 1336 PR_CAPATTACH, 1337 .pr_ctloutput = tcp_ctloutput, 1338 .pr_abort = tcp_usr_abort, 1339 .pr_accept = tcp_usr_accept, 1340 .pr_attach = tcp_usr_attach, 1341 .pr_bind = tcp_usr_bind, 1342 .pr_connect = tcp_usr_connect, 1343 .pr_control = in_control, 1344 .pr_detach = tcp_usr_detach, 1345 .pr_disconnect = tcp_usr_disconnect, 1346 .pr_listen = tcp_usr_listen, 1347 .pr_peeraddr = in_getpeeraddr, 1348 .pr_rcvd = tcp_usr_rcvd, 1349 .pr_rcvoob = tcp_usr_rcvoob, 1350 .pr_send = tcp_usr_send, 1351 .pr_ready = tcp_usr_ready, 1352 .pr_shutdown = tcp_usr_shutdown, 1353 .pr_sockaddr = in_getsockaddr, 1354 .pr_sosetlabel = in_pcbsosetlabel, 1355 .pr_close = tcp_usr_close, 1356 }; 1357 #endif /* INET */ 1358 1359 #ifdef INET6 1360 struct protosw tcp6_protosw = { 1361 .pr_type = SOCK_STREAM, 1362 .pr_protocol = IPPROTO_TCP, 1363 .pr_flags = PR_CONNREQUIRED | PR_IMPLOPCL |PR_WANTRCVD | 1364 PR_CAPATTACH, 1365 .pr_ctloutput = tcp_ctloutput, 1366 .pr_abort = tcp_usr_abort, 1367 .pr_accept = tcp6_usr_accept, 1368 .pr_attach = tcp_usr_attach, 1369 .pr_bind = tcp6_usr_bind, 1370 .pr_connect = tcp6_usr_connect, 1371 .pr_control = in6_control, 1372 .pr_detach = tcp_usr_detach, 1373 .pr_disconnect = tcp_usr_disconnect, 1374 .pr_listen = tcp6_usr_listen, 1375 .pr_peeraddr = in6_mapped_peeraddr, 1376 .pr_rcvd = tcp_usr_rcvd, 1377 .pr_rcvoob = tcp_usr_rcvoob, 1378 .pr_send = tcp_usr_send, 1379 .pr_ready = tcp_usr_ready, 1380 .pr_shutdown = tcp_usr_shutdown, 1381 .pr_sockaddr = in6_mapped_sockaddr, 1382 .pr_sosetlabel = in_pcbsosetlabel, 1383 .pr_close = tcp_usr_close, 1384 }; 1385 #endif /* INET6 */ 1386 1387 #ifdef INET 1388 /* 1389 * Common subroutine to open a TCP connection to remote host specified 1390 * by struct sockaddr_in. Call in_pcbconnect() to choose local host address 1391 * and assign a local port number and install the inpcb into the hash. 1392 * Initialize connection parameters and enter SYN-SENT state. 1393 */ 1394 static int 1395 tcp_connect(struct tcpcb *tp, struct sockaddr_in *sin, struct thread *td) 1396 { 1397 struct inpcb *inp = tptoinpcb(tp); 1398 struct socket *so = tptosocket(tp); 1399 int error; 1400 1401 NET_EPOCH_ASSERT(); 1402 INP_WLOCK_ASSERT(inp); 1403 1404 INP_HASH_WLOCK(&V_tcbinfo); 1405 error = in_pcbconnect(inp, sin, td->td_ucred, true); 1406 INP_HASH_WUNLOCK(&V_tcbinfo); 1407 if (error != 0) 1408 return (error); 1409 1410 /* 1411 * Compute window scaling to request: 1412 * Scale to fit into sweet spot. See tcp_syncache.c. 1413 * XXX: This should move to tcp_output(). 1414 */ 1415 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 1416 (TCP_MAXWIN << tp->request_r_scale) < sb_max) 1417 tp->request_r_scale++; 1418 1419 soisconnecting(so); 1420 TCPSTAT_INC(tcps_connattempt); 1421 tcp_state_change(tp, TCPS_SYN_SENT); 1422 tp->iss = tcp_new_isn(&inp->inp_inc); 1423 if (tp->t_flags & TF_REQ_TSTMP) 1424 tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc); 1425 tcp_sendseqinit(tp); 1426 1427 return (0); 1428 } 1429 #endif /* INET */ 1430 1431 #ifdef INET6 1432 static int 1433 tcp6_connect(struct tcpcb *tp, struct sockaddr_in6 *sin6, struct thread *td) 1434 { 1435 struct inpcb *inp = tptoinpcb(tp); 1436 struct epoch_tracker et; 1437 int error; 1438 1439 INP_WLOCK_ASSERT(inp); 1440 1441 NET_EPOCH_ENTER(et); 1442 INP_HASH_WLOCK(&V_tcbinfo); 1443 error = in6_pcbconnect(inp, sin6, td->td_ucred, true); 1444 INP_HASH_WUNLOCK(&V_tcbinfo); 1445 NET_EPOCH_EXIT(et); 1446 if (error != 0) 1447 return (error); 1448 1449 /* Compute window scaling to request. */ 1450 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 1451 (TCP_MAXWIN << tp->request_r_scale) < sb_max) 1452 tp->request_r_scale++; 1453 1454 soisconnecting(inp->inp_socket); 1455 TCPSTAT_INC(tcps_connattempt); 1456 tcp_state_change(tp, TCPS_SYN_SENT); 1457 tp->iss = tcp_new_isn(&inp->inp_inc); 1458 if (tp->t_flags & TF_REQ_TSTMP) 1459 tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc); 1460 tcp_sendseqinit(tp); 1461 1462 return (0); 1463 } 1464 #endif /* INET6 */ 1465 1466 /* 1467 * Export TCP internal state information via a struct tcp_info, based on the 1468 * Linux 2.6 API. Not ABI compatible as our constants are mapped differently 1469 * (TCP state machine, etc). We export all information using FreeBSD-native 1470 * constants -- for example, the numeric values for tcpi_state will differ 1471 * from Linux. 1472 */ 1473 static void 1474 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti) 1475 { 1476 1477 INP_WLOCK_ASSERT(tptoinpcb(tp)); 1478 bzero(ti, sizeof(*ti)); 1479 1480 ti->tcpi_state = tp->t_state; 1481 if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP)) 1482 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS; 1483 if (tp->t_flags & TF_SACK_PERMIT) 1484 ti->tcpi_options |= TCPI_OPT_SACK; 1485 if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) { 1486 ti->tcpi_options |= TCPI_OPT_WSCALE; 1487 ti->tcpi_snd_wscale = tp->snd_scale; 1488 ti->tcpi_rcv_wscale = tp->rcv_scale; 1489 } 1490 if (tp->t_flags2 & (TF2_ECN_PERMIT | TF2_ACE_PERMIT)) 1491 ti->tcpi_options |= TCPI_OPT_ECN; 1492 1493 ti->tcpi_rto = tp->t_rxtcur * tick; 1494 ti->tcpi_last_data_recv = ((uint32_t)ticks - tp->t_rcvtime) * tick; 1495 ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT; 1496 ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT; 1497 1498 ti->tcpi_snd_ssthresh = tp->snd_ssthresh; 1499 ti->tcpi_snd_cwnd = tp->snd_cwnd; 1500 1501 /* 1502 * FreeBSD-specific extension fields for tcp_info. 1503 */ 1504 ti->tcpi_rcv_space = tp->rcv_wnd; 1505 ti->tcpi_rcv_nxt = tp->rcv_nxt; 1506 ti->tcpi_snd_wnd = tp->snd_wnd; 1507 ti->tcpi_snd_bwnd = 0; /* Unused, kept for compat. */ 1508 ti->tcpi_snd_nxt = tp->snd_nxt; 1509 ti->tcpi_snd_mss = tp->t_maxseg; 1510 ti->tcpi_rcv_mss = tp->t_maxseg; 1511 ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack; 1512 ti->tcpi_rcv_ooopack = tp->t_rcvoopack; 1513 ti->tcpi_snd_zerowin = tp->t_sndzerowin; 1514 #ifdef TCP_OFFLOAD 1515 if (tp->t_flags & TF_TOE) { 1516 ti->tcpi_options |= TCPI_OPT_TOE; 1517 tcp_offload_tcp_info(tp, ti); 1518 } 1519 #endif 1520 /* 1521 * AccECN related counters. 1522 */ 1523 if ((tp->t_flags2 & (TF2_ECN_PERMIT | TF2_ACE_PERMIT)) == 1524 (TF2_ECN_PERMIT | TF2_ACE_PERMIT)) 1525 /* 1526 * Internal counter starts at 5 for AccECN 1527 * but 0 for RFC3168 ECN. 1528 */ 1529 ti->tcpi_delivered_ce = tp->t_scep - 5; 1530 else 1531 ti->tcpi_delivered_ce = tp->t_scep; 1532 ti->tcpi_received_ce = tp->t_rcep; 1533 } 1534 1535 /* 1536 * tcp_ctloutput() must drop the inpcb lock before performing copyin on 1537 * socket option arguments. When it re-acquires the lock after the copy, it 1538 * has to revalidate that the connection is still valid for the socket 1539 * option. 1540 */ 1541 #define INP_WLOCK_RECHECK_CLEANUP(inp, cleanup) do { \ 1542 INP_WLOCK(inp); \ 1543 if (inp->inp_flags & INP_DROPPED) { \ 1544 INP_WUNLOCK(inp); \ 1545 cleanup; \ 1546 return (ECONNRESET); \ 1547 } \ 1548 tp = intotcpcb(inp); \ 1549 } while(0) 1550 #define INP_WLOCK_RECHECK(inp) INP_WLOCK_RECHECK_CLEANUP((inp), /* noop */) 1551 1552 int 1553 tcp_ctloutput_set(struct inpcb *inp, struct sockopt *sopt) 1554 { 1555 struct socket *so = inp->inp_socket; 1556 struct tcpcb *tp = intotcpcb(inp); 1557 int error = 0; 1558 1559 MPASS(sopt->sopt_dir == SOPT_SET); 1560 INP_WLOCK_ASSERT(inp); 1561 KASSERT((inp->inp_flags & INP_DROPPED) == 0, 1562 ("inp_flags == %x", inp->inp_flags)); 1563 KASSERT(so != NULL, ("inp_socket == NULL")); 1564 1565 if (sopt->sopt_level != IPPROTO_TCP) { 1566 INP_WUNLOCK(inp); 1567 #ifdef INET6 1568 if (inp->inp_vflag & INP_IPV6PROTO) 1569 error = ip6_ctloutput(so, sopt); 1570 #endif 1571 #if defined(INET6) && defined(INET) 1572 else 1573 #endif 1574 #ifdef INET 1575 error = ip_ctloutput(so, sopt); 1576 #endif 1577 /* 1578 * When an IP-level socket option affects TCP, pass control 1579 * down to stack tfb_tcp_ctloutput, otherwise return what 1580 * IP level returned. 1581 */ 1582 switch (sopt->sopt_level) { 1583 #ifdef INET6 1584 case IPPROTO_IPV6: 1585 if ((inp->inp_vflag & INP_IPV6PROTO) == 0) 1586 return (error); 1587 switch (sopt->sopt_name) { 1588 case IPV6_TCLASS: 1589 /* Notify tcp stacks that care (e.g. RACK). */ 1590 break; 1591 case IPV6_USE_MIN_MTU: 1592 /* Update t_maxseg accordingly. */ 1593 break; 1594 default: 1595 return (error); 1596 } 1597 break; 1598 #endif 1599 #ifdef INET 1600 case IPPROTO_IP: 1601 switch (sopt->sopt_name) { 1602 case IP_TOS: 1603 inp->inp_ip_tos &= ~IPTOS_ECN_MASK; 1604 break; 1605 case IP_TTL: 1606 /* Notify tcp stacks that care (e.g. RACK). */ 1607 break; 1608 default: 1609 return (error); 1610 } 1611 break; 1612 #endif 1613 default: 1614 return (error); 1615 } 1616 INP_WLOCK(inp); 1617 if (inp->inp_flags & INP_DROPPED) { 1618 INP_WUNLOCK(inp); 1619 return (ECONNRESET); 1620 } 1621 } else if (sopt->sopt_name == TCP_FUNCTION_BLK) { 1622 /* 1623 * Protect the TCP option TCP_FUNCTION_BLK so 1624 * that a sub-function can *never* overwrite this. 1625 */ 1626 struct tcp_function_set fsn; 1627 struct tcp_function_block *blk; 1628 1629 INP_WUNLOCK(inp); 1630 error = sooptcopyin(sopt, &fsn, sizeof fsn, sizeof fsn); 1631 if (error) 1632 return (error); 1633 1634 INP_WLOCK(inp); 1635 if (inp->inp_flags & INP_DROPPED) { 1636 INP_WUNLOCK(inp); 1637 return (ECONNRESET); 1638 } 1639 tp = intotcpcb(inp); 1640 1641 blk = find_and_ref_tcp_functions(&fsn); 1642 if (blk == NULL) { 1643 INP_WUNLOCK(inp); 1644 return (ENOENT); 1645 } 1646 if (tp->t_fb == blk) { 1647 /* You already have this */ 1648 refcount_release(&blk->tfb_refcnt); 1649 INP_WUNLOCK(inp); 1650 return (0); 1651 } 1652 if (tp->t_state != TCPS_CLOSED) { 1653 /* 1654 * The user has advanced the state 1655 * past the initial point, we may not 1656 * be able to switch. 1657 */ 1658 if (blk->tfb_tcp_handoff_ok != NULL) { 1659 /* 1660 * Does the stack provide a 1661 * query mechanism, if so it may 1662 * still be possible? 1663 */ 1664 error = (*blk->tfb_tcp_handoff_ok)(tp); 1665 } else 1666 error = EINVAL; 1667 if (error) { 1668 refcount_release(&blk->tfb_refcnt); 1669 INP_WUNLOCK(inp); 1670 return(error); 1671 } 1672 } 1673 if (blk->tfb_flags & TCP_FUNC_BEING_REMOVED) { 1674 refcount_release(&blk->tfb_refcnt); 1675 INP_WUNLOCK(inp); 1676 return (ENOENT); 1677 } 1678 /* 1679 * Release the old refcnt, the 1680 * lookup acquired a ref on the 1681 * new one already. 1682 */ 1683 if (tp->t_fb->tfb_tcp_fb_fini) { 1684 struct epoch_tracker et; 1685 /* 1686 * Tell the stack to cleanup with 0 i.e. 1687 * the tcb is not going away. 1688 */ 1689 NET_EPOCH_ENTER(et); 1690 (*tp->t_fb->tfb_tcp_fb_fini)(tp, 0); 1691 NET_EPOCH_EXIT(et); 1692 } 1693 #ifdef TCPHPTS 1694 /* Assure that we are not on any hpts */ 1695 tcp_hpts_remove(tptoinpcb(tp)); 1696 #endif 1697 if (blk->tfb_tcp_fb_init) { 1698 error = (*blk->tfb_tcp_fb_init)(tp); 1699 if (error) { 1700 refcount_release(&blk->tfb_refcnt); 1701 if (tp->t_fb->tfb_tcp_fb_init) { 1702 if((*tp->t_fb->tfb_tcp_fb_init)(tp) != 0) { 1703 /* Fall back failed, drop the connection */ 1704 INP_WUNLOCK(inp); 1705 soabort(so); 1706 return (error); 1707 } 1708 } 1709 goto err_out; 1710 } 1711 } 1712 refcount_release(&tp->t_fb->tfb_refcnt); 1713 tp->t_fb = blk; 1714 #ifdef TCP_OFFLOAD 1715 if (tp->t_flags & TF_TOE) { 1716 tcp_offload_ctloutput(tp, sopt->sopt_dir, 1717 sopt->sopt_name); 1718 } 1719 #endif 1720 err_out: 1721 INP_WUNLOCK(inp); 1722 return (error); 1723 } 1724 1725 /* Pass in the INP locked, callee must unlock it. */ 1726 return (tp->t_fb->tfb_tcp_ctloutput(inp, sopt)); 1727 } 1728 1729 static int 1730 tcp_ctloutput_get(struct inpcb *inp, struct sockopt *sopt) 1731 { 1732 struct socket *so = inp->inp_socket; 1733 struct tcpcb *tp = intotcpcb(inp); 1734 int error = 0; 1735 1736 MPASS(sopt->sopt_dir == SOPT_GET); 1737 INP_WLOCK_ASSERT(inp); 1738 KASSERT((inp->inp_flags & INP_DROPPED) == 0, 1739 ("inp_flags == %x", inp->inp_flags)); 1740 KASSERT(so != NULL, ("inp_socket == NULL")); 1741 1742 if (sopt->sopt_level != IPPROTO_TCP) { 1743 INP_WUNLOCK(inp); 1744 #ifdef INET6 1745 if (inp->inp_vflag & INP_IPV6PROTO) 1746 error = ip6_ctloutput(so, sopt); 1747 #endif /* INET6 */ 1748 #if defined(INET6) && defined(INET) 1749 else 1750 #endif 1751 #ifdef INET 1752 error = ip_ctloutput(so, sopt); 1753 #endif 1754 return (error); 1755 } 1756 if (((sopt->sopt_name == TCP_FUNCTION_BLK) || 1757 (sopt->sopt_name == TCP_FUNCTION_ALIAS))) { 1758 struct tcp_function_set fsn; 1759 1760 if (sopt->sopt_name == TCP_FUNCTION_ALIAS) { 1761 memset(&fsn, 0, sizeof(fsn)); 1762 find_tcp_function_alias(tp->t_fb, &fsn); 1763 } else { 1764 strncpy(fsn.function_set_name, 1765 tp->t_fb->tfb_tcp_block_name, 1766 TCP_FUNCTION_NAME_LEN_MAX); 1767 fsn.function_set_name[TCP_FUNCTION_NAME_LEN_MAX - 1] = '\0'; 1768 } 1769 fsn.pcbcnt = tp->t_fb->tfb_refcnt; 1770 INP_WUNLOCK(inp); 1771 error = sooptcopyout(sopt, &fsn, sizeof fsn); 1772 return (error); 1773 } 1774 1775 /* Pass in the INP locked, callee must unlock it. */ 1776 return (tp->t_fb->tfb_tcp_ctloutput(inp, sopt)); 1777 } 1778 1779 int 1780 tcp_ctloutput(struct socket *so, struct sockopt *sopt) 1781 { 1782 struct inpcb *inp; 1783 1784 inp = sotoinpcb(so); 1785 KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL")); 1786 1787 INP_WLOCK(inp); 1788 if (inp->inp_flags & INP_DROPPED) { 1789 INP_WUNLOCK(inp); 1790 return (ECONNRESET); 1791 } 1792 if (sopt->sopt_dir == SOPT_SET) 1793 return (tcp_ctloutput_set(inp, sopt)); 1794 else if (sopt->sopt_dir == SOPT_GET) 1795 return (tcp_ctloutput_get(inp, sopt)); 1796 else 1797 panic("%s: sopt_dir $%d", __func__, sopt->sopt_dir); 1798 } 1799 1800 /* 1801 * If this assert becomes untrue, we need to change the size of the buf 1802 * variable in tcp_default_ctloutput(). 1803 */ 1804 #ifdef CTASSERT 1805 CTASSERT(TCP_CA_NAME_MAX <= TCP_LOG_ID_LEN); 1806 CTASSERT(TCP_LOG_REASON_LEN <= TCP_LOG_ID_LEN); 1807 #endif 1808 1809 #ifdef KERN_TLS 1810 static int 1811 copyin_tls_enable(struct sockopt *sopt, struct tls_enable *tls) 1812 { 1813 struct tls_enable_v0 tls_v0; 1814 int error; 1815 1816 if (sopt->sopt_valsize == sizeof(tls_v0)) { 1817 error = sooptcopyin(sopt, &tls_v0, sizeof(tls_v0), 1818 sizeof(tls_v0)); 1819 if (error) 1820 return (error); 1821 memset(tls, 0, sizeof(*tls)); 1822 tls->cipher_key = tls_v0.cipher_key; 1823 tls->iv = tls_v0.iv; 1824 tls->auth_key = tls_v0.auth_key; 1825 tls->cipher_algorithm = tls_v0.cipher_algorithm; 1826 tls->cipher_key_len = tls_v0.cipher_key_len; 1827 tls->iv_len = tls_v0.iv_len; 1828 tls->auth_algorithm = tls_v0.auth_algorithm; 1829 tls->auth_key_len = tls_v0.auth_key_len; 1830 tls->flags = tls_v0.flags; 1831 tls->tls_vmajor = tls_v0.tls_vmajor; 1832 tls->tls_vminor = tls_v0.tls_vminor; 1833 return (0); 1834 } 1835 1836 return (sooptcopyin(sopt, tls, sizeof(*tls), sizeof(*tls))); 1837 } 1838 #endif 1839 1840 extern struct cc_algo newreno_cc_algo; 1841 1842 static int 1843 tcp_set_cc_mod(struct inpcb *inp, struct sockopt *sopt) 1844 { 1845 struct cc_algo *algo; 1846 void *ptr = NULL; 1847 struct tcpcb *tp; 1848 struct cc_var cc_mem; 1849 char buf[TCP_CA_NAME_MAX]; 1850 size_t mem_sz; 1851 int error; 1852 1853 INP_WUNLOCK(inp); 1854 error = sooptcopyin(sopt, buf, TCP_CA_NAME_MAX - 1, 1); 1855 if (error) 1856 return(error); 1857 buf[sopt->sopt_valsize] = '\0'; 1858 CC_LIST_RLOCK(); 1859 STAILQ_FOREACH(algo, &cc_list, entries) { 1860 if (strncmp(buf, algo->name, 1861 TCP_CA_NAME_MAX) == 0) { 1862 if (algo->flags & CC_MODULE_BEING_REMOVED) { 1863 /* We can't "see" modules being unloaded */ 1864 continue; 1865 } 1866 break; 1867 } 1868 } 1869 if (algo == NULL) { 1870 CC_LIST_RUNLOCK(); 1871 return(ESRCH); 1872 } 1873 /* 1874 * With a reference the algorithm cannot be removed 1875 * so we hold a reference through the change process. 1876 */ 1877 cc_refer(algo); 1878 CC_LIST_RUNLOCK(); 1879 if (algo->cb_init != NULL) { 1880 /* We can now pre-get the memory for the CC */ 1881 mem_sz = (*algo->cc_data_sz)(); 1882 if (mem_sz == 0) { 1883 goto no_mem_needed; 1884 } 1885 ptr = malloc(mem_sz, M_CC_MEM, M_WAITOK); 1886 } else { 1887 no_mem_needed: 1888 mem_sz = 0; 1889 ptr = NULL; 1890 } 1891 /* 1892 * Make sure its all clean and zero and also get 1893 * back the inplock. 1894 */ 1895 memset(&cc_mem, 0, sizeof(cc_mem)); 1896 INP_WLOCK(inp); 1897 if (inp->inp_flags & INP_DROPPED) { 1898 INP_WUNLOCK(inp); 1899 if (ptr) 1900 free(ptr, M_CC_MEM); 1901 /* Release our temp reference */ 1902 CC_LIST_RLOCK(); 1903 cc_release(algo); 1904 CC_LIST_RUNLOCK(); 1905 return (ECONNRESET); 1906 } 1907 tp = intotcpcb(inp); 1908 if (ptr != NULL) 1909 memset(ptr, 0, mem_sz); 1910 cc_mem.ccvc.tcp = tp; 1911 /* 1912 * We once again hold a write lock over the tcb so it's 1913 * safe to do these things without ordering concerns. 1914 * Note here we init into stack memory. 1915 */ 1916 if (algo->cb_init != NULL) 1917 error = algo->cb_init(&cc_mem, ptr); 1918 else 1919 error = 0; 1920 /* 1921 * The CC algorithms, when given their memory 1922 * should not fail we could in theory have a 1923 * KASSERT here. 1924 */ 1925 if (error == 0) { 1926 /* 1927 * Touchdown, lets go ahead and move the 1928 * connection to the new CC module by 1929 * copying in the cc_mem after we call 1930 * the old ones cleanup (if any). 1931 */ 1932 if (CC_ALGO(tp)->cb_destroy != NULL) 1933 CC_ALGO(tp)->cb_destroy(&tp->t_ccv); 1934 /* Detach the old CC from the tcpcb */ 1935 cc_detach(tp); 1936 /* Copy in our temp memory that was inited */ 1937 memcpy(&tp->t_ccv, &cc_mem, sizeof(struct cc_var)); 1938 /* Now attach the new, which takes a reference */ 1939 cc_attach(tp, algo); 1940 /* Ok now are we where we have gotten past any conn_init? */ 1941 if (TCPS_HAVEESTABLISHED(tp->t_state) && (CC_ALGO(tp)->conn_init != NULL)) { 1942 /* Yep run the connection init for the new CC */ 1943 CC_ALGO(tp)->conn_init(&tp->t_ccv); 1944 } 1945 } else if (ptr) 1946 free(ptr, M_CC_MEM); 1947 INP_WUNLOCK(inp); 1948 /* Now lets release our temp reference */ 1949 CC_LIST_RLOCK(); 1950 cc_release(algo); 1951 CC_LIST_RUNLOCK(); 1952 return (error); 1953 } 1954 1955 int 1956 tcp_default_ctloutput(struct inpcb *inp, struct sockopt *sopt) 1957 { 1958 struct tcpcb *tp = intotcpcb(inp); 1959 int error, opt, optval; 1960 u_int ui; 1961 struct tcp_info ti; 1962 #ifdef KERN_TLS 1963 struct tls_enable tls; 1964 struct socket *so = inp->inp_socket; 1965 #endif 1966 char *pbuf, buf[TCP_LOG_ID_LEN]; 1967 #ifdef STATS 1968 struct statsblob *sbp; 1969 #endif 1970 size_t len; 1971 1972 INP_WLOCK_ASSERT(inp); 1973 KASSERT((inp->inp_flags & INP_DROPPED) == 0, 1974 ("inp_flags == %x", inp->inp_flags)); 1975 KASSERT(inp->inp_socket != NULL, ("inp_socket == NULL")); 1976 1977 switch (sopt->sopt_level) { 1978 #ifdef INET6 1979 case IPPROTO_IPV6: 1980 MPASS(inp->inp_vflag & INP_IPV6PROTO); 1981 switch (sopt->sopt_name) { 1982 case IPV6_USE_MIN_MTU: 1983 tcp6_use_min_mtu(tp); 1984 /* FALLTHROUGH */ 1985 } 1986 INP_WUNLOCK(inp); 1987 return (0); 1988 #endif 1989 #ifdef INET 1990 case IPPROTO_IP: 1991 INP_WUNLOCK(inp); 1992 return (0); 1993 #endif 1994 } 1995 1996 /* 1997 * For TCP_CCALGOOPT forward the control to CC module, for both 1998 * SOPT_SET and SOPT_GET. 1999 */ 2000 switch (sopt->sopt_name) { 2001 case TCP_CCALGOOPT: 2002 INP_WUNLOCK(inp); 2003 if (sopt->sopt_valsize > CC_ALGOOPT_LIMIT) 2004 return (EINVAL); 2005 pbuf = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK | M_ZERO); 2006 error = sooptcopyin(sopt, pbuf, sopt->sopt_valsize, 2007 sopt->sopt_valsize); 2008 if (error) { 2009 free(pbuf, M_TEMP); 2010 return (error); 2011 } 2012 INP_WLOCK_RECHECK_CLEANUP(inp, free(pbuf, M_TEMP)); 2013 if (CC_ALGO(tp)->ctl_output != NULL) 2014 error = CC_ALGO(tp)->ctl_output(&tp->t_ccv, sopt, pbuf); 2015 else 2016 error = ENOENT; 2017 INP_WUNLOCK(inp); 2018 if (error == 0 && sopt->sopt_dir == SOPT_GET) 2019 error = sooptcopyout(sopt, pbuf, sopt->sopt_valsize); 2020 free(pbuf, M_TEMP); 2021 return (error); 2022 } 2023 2024 switch (sopt->sopt_dir) { 2025 case SOPT_SET: 2026 switch (sopt->sopt_name) { 2027 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE) 2028 case TCP_MD5SIG: 2029 INP_WUNLOCK(inp); 2030 if (!TCPMD5_ENABLED()) 2031 return (ENOPROTOOPT); 2032 error = TCPMD5_PCBCTL(inp, sopt); 2033 if (error) 2034 return (error); 2035 INP_WLOCK_RECHECK(inp); 2036 goto unlock_and_done; 2037 #endif /* IPSEC */ 2038 2039 case TCP_NODELAY: 2040 case TCP_NOOPT: 2041 case TCP_LRD: 2042 INP_WUNLOCK(inp); 2043 error = sooptcopyin(sopt, &optval, sizeof optval, 2044 sizeof optval); 2045 if (error) 2046 return (error); 2047 2048 INP_WLOCK_RECHECK(inp); 2049 switch (sopt->sopt_name) { 2050 case TCP_NODELAY: 2051 opt = TF_NODELAY; 2052 break; 2053 case TCP_NOOPT: 2054 opt = TF_NOOPT; 2055 break; 2056 case TCP_LRD: 2057 opt = TF_LRD; 2058 break; 2059 default: 2060 opt = 0; /* dead code to fool gcc */ 2061 break; 2062 } 2063 2064 if (optval) 2065 tp->t_flags |= opt; 2066 else 2067 tp->t_flags &= ~opt; 2068 unlock_and_done: 2069 #ifdef TCP_OFFLOAD 2070 if (tp->t_flags & TF_TOE) { 2071 tcp_offload_ctloutput(tp, sopt->sopt_dir, 2072 sopt->sopt_name); 2073 } 2074 #endif 2075 INP_WUNLOCK(inp); 2076 break; 2077 2078 case TCP_NOPUSH: 2079 INP_WUNLOCK(inp); 2080 error = sooptcopyin(sopt, &optval, sizeof optval, 2081 sizeof optval); 2082 if (error) 2083 return (error); 2084 2085 INP_WLOCK_RECHECK(inp); 2086 if (optval) 2087 tp->t_flags |= TF_NOPUSH; 2088 else if (tp->t_flags & TF_NOPUSH) { 2089 tp->t_flags &= ~TF_NOPUSH; 2090 if (TCPS_HAVEESTABLISHED(tp->t_state)) { 2091 struct epoch_tracker et; 2092 2093 NET_EPOCH_ENTER(et); 2094 error = tcp_output_nodrop(tp); 2095 NET_EPOCH_EXIT(et); 2096 } 2097 } 2098 goto unlock_and_done; 2099 2100 case TCP_REMOTE_UDP_ENCAPS_PORT: 2101 INP_WUNLOCK(inp); 2102 error = sooptcopyin(sopt, &optval, sizeof optval, 2103 sizeof optval); 2104 if (error) 2105 return (error); 2106 if ((optval < TCP_TUNNELING_PORT_MIN) || 2107 (optval > TCP_TUNNELING_PORT_MAX)) { 2108 /* Its got to be in range */ 2109 return (EINVAL); 2110 } 2111 if ((V_tcp_udp_tunneling_port == 0) && (optval != 0)) { 2112 /* You have to have enabled a UDP tunneling port first */ 2113 return (EINVAL); 2114 } 2115 INP_WLOCK_RECHECK(inp); 2116 if (tp->t_state != TCPS_CLOSED) { 2117 /* You can't change after you are connected */ 2118 error = EINVAL; 2119 } else { 2120 /* Ok we are all good set the port */ 2121 tp->t_port = htons(optval); 2122 } 2123 goto unlock_and_done; 2124 2125 case TCP_MAXSEG: 2126 INP_WUNLOCK(inp); 2127 error = sooptcopyin(sopt, &optval, sizeof optval, 2128 sizeof optval); 2129 if (error) 2130 return (error); 2131 2132 INP_WLOCK_RECHECK(inp); 2133 if (optval > 0 && optval <= tp->t_maxseg && 2134 optval + 40 >= V_tcp_minmss) 2135 tp->t_maxseg = optval; 2136 else 2137 error = EINVAL; 2138 goto unlock_and_done; 2139 2140 case TCP_INFO: 2141 INP_WUNLOCK(inp); 2142 error = EINVAL; 2143 break; 2144 2145 case TCP_STATS: 2146 INP_WUNLOCK(inp); 2147 #ifdef STATS 2148 error = sooptcopyin(sopt, &optval, sizeof optval, 2149 sizeof optval); 2150 if (error) 2151 return (error); 2152 2153 if (optval > 0) 2154 sbp = stats_blob_alloc( 2155 V_tcp_perconn_stats_dflt_tpl, 0); 2156 else 2157 sbp = NULL; 2158 2159 INP_WLOCK_RECHECK(inp); 2160 if ((tp->t_stats != NULL && sbp == NULL) || 2161 (tp->t_stats == NULL && sbp != NULL)) { 2162 struct statsblob *t = tp->t_stats; 2163 tp->t_stats = sbp; 2164 sbp = t; 2165 } 2166 INP_WUNLOCK(inp); 2167 2168 stats_blob_destroy(sbp); 2169 #else 2170 return (EOPNOTSUPP); 2171 #endif /* !STATS */ 2172 break; 2173 2174 case TCP_CONGESTION: 2175 error = tcp_set_cc_mod(inp, sopt); 2176 break; 2177 2178 case TCP_REUSPORT_LB_NUMA: 2179 INP_WUNLOCK(inp); 2180 error = sooptcopyin(sopt, &optval, sizeof(optval), 2181 sizeof(optval)); 2182 INP_WLOCK_RECHECK(inp); 2183 if (!error) 2184 error = in_pcblbgroup_numa(inp, optval); 2185 INP_WUNLOCK(inp); 2186 break; 2187 2188 #ifdef KERN_TLS 2189 case TCP_TXTLS_ENABLE: 2190 INP_WUNLOCK(inp); 2191 error = copyin_tls_enable(sopt, &tls); 2192 if (error) 2193 break; 2194 error = ktls_enable_tx(so, &tls); 2195 break; 2196 case TCP_TXTLS_MODE: 2197 INP_WUNLOCK(inp); 2198 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui)); 2199 if (error) 2200 return (error); 2201 2202 INP_WLOCK_RECHECK(inp); 2203 error = ktls_set_tx_mode(so, ui); 2204 INP_WUNLOCK(inp); 2205 break; 2206 case TCP_RXTLS_ENABLE: 2207 INP_WUNLOCK(inp); 2208 error = sooptcopyin(sopt, &tls, sizeof(tls), 2209 sizeof(tls)); 2210 if (error) 2211 break; 2212 error = ktls_enable_rx(so, &tls); 2213 break; 2214 #endif 2215 case TCP_MAXUNACKTIME: 2216 case TCP_KEEPIDLE: 2217 case TCP_KEEPINTVL: 2218 case TCP_KEEPINIT: 2219 INP_WUNLOCK(inp); 2220 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui)); 2221 if (error) 2222 return (error); 2223 2224 if (ui > (UINT_MAX / hz)) { 2225 error = EINVAL; 2226 break; 2227 } 2228 ui *= hz; 2229 2230 INP_WLOCK_RECHECK(inp); 2231 switch (sopt->sopt_name) { 2232 case TCP_MAXUNACKTIME: 2233 tp->t_maxunacktime = ui; 2234 break; 2235 2236 case TCP_KEEPIDLE: 2237 tp->t_keepidle = ui; 2238 /* 2239 * XXX: better check current remaining 2240 * timeout and "merge" it with new value. 2241 */ 2242 if ((tp->t_state > TCPS_LISTEN) && 2243 (tp->t_state <= TCPS_CLOSING)) 2244 tcp_timer_activate(tp, TT_KEEP, 2245 TP_KEEPIDLE(tp)); 2246 break; 2247 case TCP_KEEPINTVL: 2248 tp->t_keepintvl = ui; 2249 if ((tp->t_state == TCPS_FIN_WAIT_2) && 2250 (TP_MAXIDLE(tp) > 0)) 2251 tcp_timer_activate(tp, TT_2MSL, 2252 TP_MAXIDLE(tp)); 2253 break; 2254 case TCP_KEEPINIT: 2255 tp->t_keepinit = ui; 2256 if (tp->t_state == TCPS_SYN_RECEIVED || 2257 tp->t_state == TCPS_SYN_SENT) 2258 tcp_timer_activate(tp, TT_KEEP, 2259 TP_KEEPINIT(tp)); 2260 break; 2261 } 2262 goto unlock_and_done; 2263 2264 case TCP_KEEPCNT: 2265 INP_WUNLOCK(inp); 2266 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui)); 2267 if (error) 2268 return (error); 2269 2270 INP_WLOCK_RECHECK(inp); 2271 tp->t_keepcnt = ui; 2272 if ((tp->t_state == TCPS_FIN_WAIT_2) && 2273 (TP_MAXIDLE(tp) > 0)) 2274 tcp_timer_activate(tp, TT_2MSL, 2275 TP_MAXIDLE(tp)); 2276 goto unlock_and_done; 2277 2278 #ifdef TCPPCAP 2279 case TCP_PCAP_OUT: 2280 case TCP_PCAP_IN: 2281 INP_WUNLOCK(inp); 2282 error = sooptcopyin(sopt, &optval, sizeof optval, 2283 sizeof optval); 2284 if (error) 2285 return (error); 2286 2287 INP_WLOCK_RECHECK(inp); 2288 if (optval >= 0) 2289 tcp_pcap_set_sock_max(TCP_PCAP_OUT ? 2290 &(tp->t_outpkts) : &(tp->t_inpkts), 2291 optval); 2292 else 2293 error = EINVAL; 2294 goto unlock_and_done; 2295 #endif 2296 2297 case TCP_FASTOPEN: { 2298 struct tcp_fastopen tfo_optval; 2299 2300 INP_WUNLOCK(inp); 2301 if (!V_tcp_fastopen_client_enable && 2302 !V_tcp_fastopen_server_enable) 2303 return (EPERM); 2304 2305 error = sooptcopyin(sopt, &tfo_optval, 2306 sizeof(tfo_optval), sizeof(int)); 2307 if (error) 2308 return (error); 2309 2310 INP_WLOCK_RECHECK(inp); 2311 if ((tp->t_state != TCPS_CLOSED) && 2312 (tp->t_state != TCPS_LISTEN)) { 2313 error = EINVAL; 2314 goto unlock_and_done; 2315 } 2316 if (tfo_optval.enable) { 2317 if (tp->t_state == TCPS_LISTEN) { 2318 if (!V_tcp_fastopen_server_enable) { 2319 error = EPERM; 2320 goto unlock_and_done; 2321 } 2322 2323 if (tp->t_tfo_pending == NULL) 2324 tp->t_tfo_pending = 2325 tcp_fastopen_alloc_counter(); 2326 } else { 2327 /* 2328 * If a pre-shared key was provided, 2329 * stash it in the client cookie 2330 * field of the tcpcb for use during 2331 * connect. 2332 */ 2333 if (sopt->sopt_valsize == 2334 sizeof(tfo_optval)) { 2335 memcpy(tp->t_tfo_cookie.client, 2336 tfo_optval.psk, 2337 TCP_FASTOPEN_PSK_LEN); 2338 tp->t_tfo_client_cookie_len = 2339 TCP_FASTOPEN_PSK_LEN; 2340 } 2341 } 2342 tp->t_flags |= TF_FASTOPEN; 2343 } else 2344 tp->t_flags &= ~TF_FASTOPEN; 2345 goto unlock_and_done; 2346 } 2347 2348 #ifdef TCP_BLACKBOX 2349 case TCP_LOG: 2350 INP_WUNLOCK(inp); 2351 error = sooptcopyin(sopt, &optval, sizeof optval, 2352 sizeof optval); 2353 if (error) 2354 return (error); 2355 2356 INP_WLOCK_RECHECK(inp); 2357 error = tcp_log_state_change(tp, optval); 2358 goto unlock_and_done; 2359 2360 case TCP_LOGBUF: 2361 INP_WUNLOCK(inp); 2362 error = EINVAL; 2363 break; 2364 2365 case TCP_LOGID: 2366 INP_WUNLOCK(inp); 2367 error = sooptcopyin(sopt, buf, TCP_LOG_ID_LEN - 1, 0); 2368 if (error) 2369 break; 2370 buf[sopt->sopt_valsize] = '\0'; 2371 INP_WLOCK_RECHECK(inp); 2372 error = tcp_log_set_id(tp, buf); 2373 /* tcp_log_set_id() unlocks the INP. */ 2374 break; 2375 2376 case TCP_LOGDUMP: 2377 case TCP_LOGDUMPID: 2378 INP_WUNLOCK(inp); 2379 error = 2380 sooptcopyin(sopt, buf, TCP_LOG_REASON_LEN - 1, 0); 2381 if (error) 2382 break; 2383 buf[sopt->sopt_valsize] = '\0'; 2384 INP_WLOCK_RECHECK(inp); 2385 if (sopt->sopt_name == TCP_LOGDUMP) { 2386 error = tcp_log_dump_tp_logbuf(tp, buf, 2387 M_WAITOK, true); 2388 INP_WUNLOCK(inp); 2389 } else { 2390 tcp_log_dump_tp_bucket_logbufs(tp, buf); 2391 /* 2392 * tcp_log_dump_tp_bucket_logbufs() drops the 2393 * INP lock. 2394 */ 2395 } 2396 break; 2397 #endif 2398 2399 default: 2400 INP_WUNLOCK(inp); 2401 error = ENOPROTOOPT; 2402 break; 2403 } 2404 break; 2405 2406 case SOPT_GET: 2407 tp = intotcpcb(inp); 2408 switch (sopt->sopt_name) { 2409 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE) 2410 case TCP_MD5SIG: 2411 INP_WUNLOCK(inp); 2412 if (!TCPMD5_ENABLED()) 2413 return (ENOPROTOOPT); 2414 error = TCPMD5_PCBCTL(inp, sopt); 2415 break; 2416 #endif 2417 2418 case TCP_NODELAY: 2419 optval = tp->t_flags & TF_NODELAY; 2420 INP_WUNLOCK(inp); 2421 error = sooptcopyout(sopt, &optval, sizeof optval); 2422 break; 2423 case TCP_MAXSEG: 2424 optval = tp->t_maxseg; 2425 INP_WUNLOCK(inp); 2426 error = sooptcopyout(sopt, &optval, sizeof optval); 2427 break; 2428 case TCP_REMOTE_UDP_ENCAPS_PORT: 2429 optval = ntohs(tp->t_port); 2430 INP_WUNLOCK(inp); 2431 error = sooptcopyout(sopt, &optval, sizeof optval); 2432 break; 2433 case TCP_NOOPT: 2434 optval = tp->t_flags & TF_NOOPT; 2435 INP_WUNLOCK(inp); 2436 error = sooptcopyout(sopt, &optval, sizeof optval); 2437 break; 2438 case TCP_NOPUSH: 2439 optval = tp->t_flags & TF_NOPUSH; 2440 INP_WUNLOCK(inp); 2441 error = sooptcopyout(sopt, &optval, sizeof optval); 2442 break; 2443 case TCP_INFO: 2444 tcp_fill_info(tp, &ti); 2445 INP_WUNLOCK(inp); 2446 error = sooptcopyout(sopt, &ti, sizeof ti); 2447 break; 2448 case TCP_STATS: 2449 { 2450 #ifdef STATS 2451 int nheld; 2452 TYPEOF_MEMBER(struct statsblob, flags) sbflags = 0; 2453 2454 error = 0; 2455 socklen_t outsbsz = sopt->sopt_valsize; 2456 if (tp->t_stats == NULL) 2457 error = ENOENT; 2458 else if (outsbsz >= tp->t_stats->cursz) 2459 outsbsz = tp->t_stats->cursz; 2460 else if (outsbsz >= sizeof(struct statsblob)) 2461 outsbsz = sizeof(struct statsblob); 2462 else 2463 error = EINVAL; 2464 INP_WUNLOCK(inp); 2465 if (error) 2466 break; 2467 2468 sbp = sopt->sopt_val; 2469 nheld = atop(round_page(((vm_offset_t)sbp) + 2470 (vm_size_t)outsbsz) - trunc_page((vm_offset_t)sbp)); 2471 vm_page_t ma[nheld]; 2472 if (vm_fault_quick_hold_pages( 2473 &curproc->p_vmspace->vm_map, (vm_offset_t)sbp, 2474 outsbsz, VM_PROT_READ | VM_PROT_WRITE, ma, 2475 nheld) < 0) { 2476 error = EFAULT; 2477 break; 2478 } 2479 2480 if ((error = copyin_nofault(&(sbp->flags), &sbflags, 2481 SIZEOF_MEMBER(struct statsblob, flags)))) 2482 goto unhold; 2483 2484 INP_WLOCK_RECHECK(inp); 2485 error = stats_blob_snapshot(&sbp, outsbsz, tp->t_stats, 2486 sbflags | SB_CLONE_USRDSTNOFAULT); 2487 INP_WUNLOCK(inp); 2488 sopt->sopt_valsize = outsbsz; 2489 unhold: 2490 vm_page_unhold_pages(ma, nheld); 2491 #else 2492 INP_WUNLOCK(inp); 2493 error = EOPNOTSUPP; 2494 #endif /* !STATS */ 2495 break; 2496 } 2497 case TCP_CONGESTION: 2498 len = strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX); 2499 INP_WUNLOCK(inp); 2500 error = sooptcopyout(sopt, buf, len + 1); 2501 break; 2502 case TCP_MAXUNACKTIME: 2503 case TCP_KEEPIDLE: 2504 case TCP_KEEPINTVL: 2505 case TCP_KEEPINIT: 2506 case TCP_KEEPCNT: 2507 switch (sopt->sopt_name) { 2508 case TCP_MAXUNACKTIME: 2509 ui = TP_MAXUNACKTIME(tp) / hz; 2510 break; 2511 case TCP_KEEPIDLE: 2512 ui = TP_KEEPIDLE(tp) / hz; 2513 break; 2514 case TCP_KEEPINTVL: 2515 ui = TP_KEEPINTVL(tp) / hz; 2516 break; 2517 case TCP_KEEPINIT: 2518 ui = TP_KEEPINIT(tp) / hz; 2519 break; 2520 case TCP_KEEPCNT: 2521 ui = TP_KEEPCNT(tp); 2522 break; 2523 } 2524 INP_WUNLOCK(inp); 2525 error = sooptcopyout(sopt, &ui, sizeof(ui)); 2526 break; 2527 #ifdef TCPPCAP 2528 case TCP_PCAP_OUT: 2529 case TCP_PCAP_IN: 2530 optval = tcp_pcap_get_sock_max(TCP_PCAP_OUT ? 2531 &(tp->t_outpkts) : &(tp->t_inpkts)); 2532 INP_WUNLOCK(inp); 2533 error = sooptcopyout(sopt, &optval, sizeof optval); 2534 break; 2535 #endif 2536 case TCP_FASTOPEN: 2537 optval = tp->t_flags & TF_FASTOPEN; 2538 INP_WUNLOCK(inp); 2539 error = sooptcopyout(sopt, &optval, sizeof optval); 2540 break; 2541 #ifdef TCP_BLACKBOX 2542 case TCP_LOG: 2543 optval = tp->t_logstate; 2544 INP_WUNLOCK(inp); 2545 error = sooptcopyout(sopt, &optval, sizeof(optval)); 2546 break; 2547 case TCP_LOGBUF: 2548 /* tcp_log_getlogbuf() does INP_WUNLOCK(inp) */ 2549 error = tcp_log_getlogbuf(sopt, tp); 2550 break; 2551 case TCP_LOGID: 2552 len = tcp_log_get_id(tp, buf); 2553 INP_WUNLOCK(inp); 2554 error = sooptcopyout(sopt, buf, len + 1); 2555 break; 2556 case TCP_LOGDUMP: 2557 case TCP_LOGDUMPID: 2558 INP_WUNLOCK(inp); 2559 error = EINVAL; 2560 break; 2561 #endif 2562 #ifdef KERN_TLS 2563 case TCP_TXTLS_MODE: 2564 error = ktls_get_tx_mode(so, &optval); 2565 INP_WUNLOCK(inp); 2566 if (error == 0) 2567 error = sooptcopyout(sopt, &optval, 2568 sizeof(optval)); 2569 break; 2570 case TCP_RXTLS_MODE: 2571 error = ktls_get_rx_mode(so, &optval); 2572 INP_WUNLOCK(inp); 2573 if (error == 0) 2574 error = sooptcopyout(sopt, &optval, 2575 sizeof(optval)); 2576 break; 2577 #endif 2578 case TCP_LRD: 2579 optval = tp->t_flags & TF_LRD; 2580 INP_WUNLOCK(inp); 2581 error = sooptcopyout(sopt, &optval, sizeof optval); 2582 break; 2583 default: 2584 INP_WUNLOCK(inp); 2585 error = ENOPROTOOPT; 2586 break; 2587 } 2588 break; 2589 } 2590 return (error); 2591 } 2592 #undef INP_WLOCK_RECHECK 2593 #undef INP_WLOCK_RECHECK_CLEANUP 2594 2595 /* 2596 * Initiate (or continue) disconnect. 2597 * If embryonic state, just send reset (once). 2598 * If in ``let data drain'' option and linger null, just drop. 2599 * Otherwise (hard), mark socket disconnecting and drop 2600 * current input data; switch states based on user close, and 2601 * send segment to peer (with FIN). 2602 */ 2603 static void 2604 tcp_disconnect(struct tcpcb *tp) 2605 { 2606 struct inpcb *inp = tptoinpcb(tp); 2607 struct socket *so = tptosocket(tp); 2608 2609 NET_EPOCH_ASSERT(); 2610 INP_WLOCK_ASSERT(inp); 2611 2612 /* 2613 * Neither tcp_close() nor tcp_drop() should return NULL, as the 2614 * socket is still open. 2615 */ 2616 if (tp->t_state < TCPS_ESTABLISHED && 2617 !(tp->t_state > TCPS_LISTEN && IS_FASTOPEN(tp->t_flags))) { 2618 tp = tcp_close(tp); 2619 KASSERT(tp != NULL, 2620 ("tcp_disconnect: tcp_close() returned NULL")); 2621 } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) { 2622 tp = tcp_drop(tp, 0); 2623 KASSERT(tp != NULL, 2624 ("tcp_disconnect: tcp_drop() returned NULL")); 2625 } else { 2626 soisdisconnecting(so); 2627 sbflush(&so->so_rcv); 2628 tcp_usrclosed(tp); 2629 if (!(inp->inp_flags & INP_DROPPED)) 2630 /* Ignore stack's drop request, we already at it. */ 2631 (void)tcp_output_nodrop(tp); 2632 } 2633 } 2634 2635 /* 2636 * User issued close, and wish to trail through shutdown states: 2637 * if never received SYN, just forget it. If got a SYN from peer, 2638 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 2639 * If already got a FIN from peer, then almost done; go to LAST_ACK 2640 * state. In all other cases, have already sent FIN to peer (e.g. 2641 * after PRU_SHUTDOWN), and just have to play tedious game waiting 2642 * for peer to send FIN or not respond to keep-alives, etc. 2643 * We can let the user exit from the close as soon as the FIN is acked. 2644 */ 2645 static void 2646 tcp_usrclosed(struct tcpcb *tp) 2647 { 2648 2649 NET_EPOCH_ASSERT(); 2650 INP_WLOCK_ASSERT(tptoinpcb(tp)); 2651 2652 switch (tp->t_state) { 2653 case TCPS_LISTEN: 2654 #ifdef TCP_OFFLOAD 2655 tcp_offload_listen_stop(tp); 2656 #endif 2657 tcp_state_change(tp, TCPS_CLOSED); 2658 /* FALLTHROUGH */ 2659 case TCPS_CLOSED: 2660 tp = tcp_close(tp); 2661 /* 2662 * tcp_close() should never return NULL here as the socket is 2663 * still open. 2664 */ 2665 KASSERT(tp != NULL, 2666 ("tcp_usrclosed: tcp_close() returned NULL")); 2667 break; 2668 2669 case TCPS_SYN_SENT: 2670 case TCPS_SYN_RECEIVED: 2671 tp->t_flags |= TF_NEEDFIN; 2672 break; 2673 2674 case TCPS_ESTABLISHED: 2675 tcp_state_change(tp, TCPS_FIN_WAIT_1); 2676 break; 2677 2678 case TCPS_CLOSE_WAIT: 2679 tcp_state_change(tp, TCPS_LAST_ACK); 2680 break; 2681 } 2682 if (tp->t_acktime == 0) 2683 tp->t_acktime = ticks; 2684 if (tp->t_state >= TCPS_FIN_WAIT_2) { 2685 soisdisconnected(tptosocket(tp)); 2686 /* Prevent the connection hanging in FIN_WAIT_2 forever. */ 2687 if (tp->t_state == TCPS_FIN_WAIT_2) { 2688 int timeout; 2689 2690 timeout = (tcp_fast_finwait2_recycle) ? 2691 tcp_finwait2_timeout : TP_MAXIDLE(tp); 2692 tcp_timer_activate(tp, TT_2MSL, timeout); 2693 } 2694 } 2695 } 2696 2697 #ifdef DDB 2698 static void 2699 db_print_indent(int indent) 2700 { 2701 int i; 2702 2703 for (i = 0; i < indent; i++) 2704 db_printf(" "); 2705 } 2706 2707 static void 2708 db_print_tstate(int t_state) 2709 { 2710 2711 switch (t_state) { 2712 case TCPS_CLOSED: 2713 db_printf("TCPS_CLOSED"); 2714 return; 2715 2716 case TCPS_LISTEN: 2717 db_printf("TCPS_LISTEN"); 2718 return; 2719 2720 case TCPS_SYN_SENT: 2721 db_printf("TCPS_SYN_SENT"); 2722 return; 2723 2724 case TCPS_SYN_RECEIVED: 2725 db_printf("TCPS_SYN_RECEIVED"); 2726 return; 2727 2728 case TCPS_ESTABLISHED: 2729 db_printf("TCPS_ESTABLISHED"); 2730 return; 2731 2732 case TCPS_CLOSE_WAIT: 2733 db_printf("TCPS_CLOSE_WAIT"); 2734 return; 2735 2736 case TCPS_FIN_WAIT_1: 2737 db_printf("TCPS_FIN_WAIT_1"); 2738 return; 2739 2740 case TCPS_CLOSING: 2741 db_printf("TCPS_CLOSING"); 2742 return; 2743 2744 case TCPS_LAST_ACK: 2745 db_printf("TCPS_LAST_ACK"); 2746 return; 2747 2748 case TCPS_FIN_WAIT_2: 2749 db_printf("TCPS_FIN_WAIT_2"); 2750 return; 2751 2752 case TCPS_TIME_WAIT: 2753 db_printf("TCPS_TIME_WAIT"); 2754 return; 2755 2756 default: 2757 db_printf("unknown"); 2758 return; 2759 } 2760 } 2761 2762 static void 2763 db_print_tflags(u_int t_flags) 2764 { 2765 int comma; 2766 2767 comma = 0; 2768 if (t_flags & TF_ACKNOW) { 2769 db_printf("%sTF_ACKNOW", comma ? ", " : ""); 2770 comma = 1; 2771 } 2772 if (t_flags & TF_DELACK) { 2773 db_printf("%sTF_DELACK", comma ? ", " : ""); 2774 comma = 1; 2775 } 2776 if (t_flags & TF_NODELAY) { 2777 db_printf("%sTF_NODELAY", comma ? ", " : ""); 2778 comma = 1; 2779 } 2780 if (t_flags & TF_NOOPT) { 2781 db_printf("%sTF_NOOPT", comma ? ", " : ""); 2782 comma = 1; 2783 } 2784 if (t_flags & TF_SENTFIN) { 2785 db_printf("%sTF_SENTFIN", comma ? ", " : ""); 2786 comma = 1; 2787 } 2788 if (t_flags & TF_REQ_SCALE) { 2789 db_printf("%sTF_REQ_SCALE", comma ? ", " : ""); 2790 comma = 1; 2791 } 2792 if (t_flags & TF_RCVD_SCALE) { 2793 db_printf("%sTF_RECVD_SCALE", comma ? ", " : ""); 2794 comma = 1; 2795 } 2796 if (t_flags & TF_REQ_TSTMP) { 2797 db_printf("%sTF_REQ_TSTMP", comma ? ", " : ""); 2798 comma = 1; 2799 } 2800 if (t_flags & TF_RCVD_TSTMP) { 2801 db_printf("%sTF_RCVD_TSTMP", comma ? ", " : ""); 2802 comma = 1; 2803 } 2804 if (t_flags & TF_SACK_PERMIT) { 2805 db_printf("%sTF_SACK_PERMIT", comma ? ", " : ""); 2806 comma = 1; 2807 } 2808 if (t_flags & TF_NEEDSYN) { 2809 db_printf("%sTF_NEEDSYN", comma ? ", " : ""); 2810 comma = 1; 2811 } 2812 if (t_flags & TF_NEEDFIN) { 2813 db_printf("%sTF_NEEDFIN", comma ? ", " : ""); 2814 comma = 1; 2815 } 2816 if (t_flags & TF_NOPUSH) { 2817 db_printf("%sTF_NOPUSH", comma ? ", " : ""); 2818 comma = 1; 2819 } 2820 if (t_flags & TF_PREVVALID) { 2821 db_printf("%sTF_PREVVALID", comma ? ", " : ""); 2822 comma = 1; 2823 } 2824 if (t_flags & TF_MORETOCOME) { 2825 db_printf("%sTF_MORETOCOME", comma ? ", " : ""); 2826 comma = 1; 2827 } 2828 if (t_flags & TF_SONOTCONN) { 2829 db_printf("%sTF_SONOTCONN", comma ? ", " : ""); 2830 comma = 1; 2831 } 2832 if (t_flags & TF_LASTIDLE) { 2833 db_printf("%sTF_LASTIDLE", comma ? ", " : ""); 2834 comma = 1; 2835 } 2836 if (t_flags & TF_RXWIN0SENT) { 2837 db_printf("%sTF_RXWIN0SENT", comma ? ", " : ""); 2838 comma = 1; 2839 } 2840 if (t_flags & TF_FASTRECOVERY) { 2841 db_printf("%sTF_FASTRECOVERY", comma ? ", " : ""); 2842 comma = 1; 2843 } 2844 if (t_flags & TF_CONGRECOVERY) { 2845 db_printf("%sTF_CONGRECOVERY", comma ? ", " : ""); 2846 comma = 1; 2847 } 2848 if (t_flags & TF_WASFRECOVERY) { 2849 db_printf("%sTF_WASFRECOVERY", comma ? ", " : ""); 2850 comma = 1; 2851 } 2852 if (t_flags & TF_WASCRECOVERY) { 2853 db_printf("%sTF_WASCRECOVERY", comma ? ", " : ""); 2854 comma = 1; 2855 } 2856 if (t_flags & TF_SIGNATURE) { 2857 db_printf("%sTF_SIGNATURE", comma ? ", " : ""); 2858 comma = 1; 2859 } 2860 if (t_flags & TF_FORCEDATA) { 2861 db_printf("%sTF_FORCEDATA", comma ? ", " : ""); 2862 comma = 1; 2863 } 2864 if (t_flags & TF_TSO) { 2865 db_printf("%sTF_TSO", comma ? ", " : ""); 2866 comma = 1; 2867 } 2868 if (t_flags & TF_FASTOPEN) { 2869 db_printf("%sTF_FASTOPEN", comma ? ", " : ""); 2870 comma = 1; 2871 } 2872 } 2873 2874 static void 2875 db_print_tflags2(u_int t_flags2) 2876 { 2877 int comma; 2878 2879 comma = 0; 2880 if (t_flags2 & TF2_PLPMTU_BLACKHOLE) { 2881 db_printf("%sTF2_PLPMTU_BLACKHOLE", comma ? ", " : ""); 2882 comma = 1; 2883 } 2884 if (t_flags2 & TF2_PLPMTU_PMTUD) { 2885 db_printf("%sTF2_PLPMTU_PMTUD", comma ? ", " : ""); 2886 comma = 1; 2887 } 2888 if (t_flags2 & TF2_PLPMTU_MAXSEGSNT) { 2889 db_printf("%sTF2_PLPMTU_MAXSEGSNT", comma ? ", " : ""); 2890 comma = 1; 2891 } 2892 if (t_flags2 & TF2_LOG_AUTO) { 2893 db_printf("%sTF2_LOG_AUTO", comma ? ", " : ""); 2894 comma = 1; 2895 } 2896 if (t_flags2 & TF2_DROP_AF_DATA) { 2897 db_printf("%sTF2_DROP_AF_DATA", comma ? ", " : ""); 2898 comma = 1; 2899 } 2900 if (t_flags2 & TF2_ECN_PERMIT) { 2901 db_printf("%sTF2_ECN_PERMIT", comma ? ", " : ""); 2902 comma = 1; 2903 } 2904 if (t_flags2 & TF2_ECN_SND_CWR) { 2905 db_printf("%sTF2_ECN_SND_CWR", comma ? ", " : ""); 2906 comma = 1; 2907 } 2908 if (t_flags2 & TF2_ECN_SND_ECE) { 2909 db_printf("%sTF2_ECN_SND_ECE", comma ? ", " : ""); 2910 comma = 1; 2911 } 2912 if (t_flags2 & TF2_ACE_PERMIT) { 2913 db_printf("%sTF2_ACE_PERMIT", comma ? ", " : ""); 2914 comma = 1; 2915 } 2916 if (t_flags2 & TF2_FBYTES_COMPLETE) { 2917 db_printf("%sTF2_FBYTES_COMPLETE", comma ? ", " : ""); 2918 comma = 1; 2919 } 2920 } 2921 2922 static void 2923 db_print_toobflags(char t_oobflags) 2924 { 2925 int comma; 2926 2927 comma = 0; 2928 if (t_oobflags & TCPOOB_HAVEDATA) { 2929 db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : ""); 2930 comma = 1; 2931 } 2932 if (t_oobflags & TCPOOB_HADDATA) { 2933 db_printf("%sTCPOOB_HADDATA", comma ? ", " : ""); 2934 comma = 1; 2935 } 2936 } 2937 2938 static void 2939 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent) 2940 { 2941 2942 db_print_indent(indent); 2943 db_printf("%s at %p\n", name, tp); 2944 2945 indent += 2; 2946 2947 db_print_indent(indent); 2948 db_printf("t_segq first: %p t_segqlen: %d t_dupacks: %d\n", 2949 TAILQ_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks); 2950 2951 db_print_indent(indent); 2952 db_printf("t_callout: %p t_timers: %p\n", 2953 &tp->t_callout, &tp->t_timers); 2954 2955 db_print_indent(indent); 2956 db_printf("t_state: %d (", tp->t_state); 2957 db_print_tstate(tp->t_state); 2958 db_printf(")\n"); 2959 2960 db_print_indent(indent); 2961 db_printf("t_flags: 0x%x (", tp->t_flags); 2962 db_print_tflags(tp->t_flags); 2963 db_printf(")\n"); 2964 2965 db_print_indent(indent); 2966 db_printf("t_flags2: 0x%x (", tp->t_flags2); 2967 db_print_tflags2(tp->t_flags2); 2968 db_printf(")\n"); 2969 2970 db_print_indent(indent); 2971 db_printf("snd_una: 0x%08x snd_max: 0x%08x snd_nxt: 0x%08x\n", 2972 tp->snd_una, tp->snd_max, tp->snd_nxt); 2973 2974 db_print_indent(indent); 2975 db_printf("snd_up: 0x%08x snd_wl1: 0x%08x snd_wl2: 0x%08x\n", 2976 tp->snd_up, tp->snd_wl1, tp->snd_wl2); 2977 2978 db_print_indent(indent); 2979 db_printf("iss: 0x%08x irs: 0x%08x rcv_nxt: 0x%08x\n", 2980 tp->iss, tp->irs, tp->rcv_nxt); 2981 2982 db_print_indent(indent); 2983 db_printf("rcv_adv: 0x%08x rcv_wnd: %u rcv_up: 0x%08x\n", 2984 tp->rcv_adv, tp->rcv_wnd, tp->rcv_up); 2985 2986 db_print_indent(indent); 2987 db_printf("snd_wnd: %u snd_cwnd: %u\n", 2988 tp->snd_wnd, tp->snd_cwnd); 2989 2990 db_print_indent(indent); 2991 db_printf("snd_ssthresh: %u snd_recover: " 2992 "0x%08x\n", tp->snd_ssthresh, tp->snd_recover); 2993 2994 db_print_indent(indent); 2995 db_printf("t_rcvtime: %u t_startime: %u\n", 2996 tp->t_rcvtime, tp->t_starttime); 2997 2998 db_print_indent(indent); 2999 db_printf("t_rttime: %u t_rtsq: 0x%08x\n", 3000 tp->t_rtttime, tp->t_rtseq); 3001 3002 db_print_indent(indent); 3003 db_printf("t_rxtcur: %d t_maxseg: %u t_srtt: %d\n", 3004 tp->t_rxtcur, tp->t_maxseg, tp->t_srtt); 3005 3006 db_print_indent(indent); 3007 db_printf("t_rttvar: %d t_rxtshift: %d t_rttmin: %u\n", 3008 tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin); 3009 3010 db_print_indent(indent); 3011 db_printf("t_rttupdated: %u max_sndwnd: %u t_softerror: %d\n", 3012 tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror); 3013 3014 db_print_indent(indent); 3015 db_printf("t_oobflags: 0x%x (", tp->t_oobflags); 3016 db_print_toobflags(tp->t_oobflags); 3017 db_printf(") t_iobc: 0x%02x\n", tp->t_iobc); 3018 3019 db_print_indent(indent); 3020 db_printf("snd_scale: %u rcv_scale: %u request_r_scale: %u\n", 3021 tp->snd_scale, tp->rcv_scale, tp->request_r_scale); 3022 3023 db_print_indent(indent); 3024 db_printf("ts_recent: %u ts_recent_age: %u\n", 3025 tp->ts_recent, tp->ts_recent_age); 3026 3027 db_print_indent(indent); 3028 db_printf("ts_offset: %u last_ack_sent: 0x%08x snd_cwnd_prev: " 3029 "%u\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev); 3030 3031 db_print_indent(indent); 3032 db_printf("snd_ssthresh_prev: %u snd_recover_prev: 0x%08x " 3033 "t_badrxtwin: %u\n", tp->snd_ssthresh_prev, 3034 tp->snd_recover_prev, tp->t_badrxtwin); 3035 3036 db_print_indent(indent); 3037 db_printf("snd_numholes: %d snd_holes first: %p\n", 3038 tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes)); 3039 3040 db_print_indent(indent); 3041 db_printf("snd_fack: 0x%08x rcv_numsacks: %d\n", 3042 tp->snd_fack, tp->rcv_numsacks); 3043 3044 /* Skip sackblks, sackhint. */ 3045 3046 db_print_indent(indent); 3047 db_printf("t_rttlow: %d rfbuf_ts: %u rfbuf_cnt: %d\n", 3048 tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt); 3049 } 3050 3051 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb) 3052 { 3053 struct tcpcb *tp; 3054 3055 if (!have_addr) { 3056 db_printf("usage: show tcpcb <addr>\n"); 3057 return; 3058 } 3059 tp = (struct tcpcb *)addr; 3060 3061 db_print_tcpcb(tp, "tcpcb", 0); 3062 } 3063 #endif 3064