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 #include "opt_tcpdebug.h" 49 50 #include <sys/param.h> 51 #include <sys/systm.h> 52 #include <sys/arb.h> 53 #include <sys/limits.h> 54 #include <sys/malloc.h> 55 #include <sys/refcount.h> 56 #include <sys/kernel.h> 57 #include <sys/ktls.h> 58 #include <sys/qmath.h> 59 #include <sys/sysctl.h> 60 #include <sys/mbuf.h> 61 #ifdef INET6 62 #include <sys/domain.h> 63 #endif /* INET6 */ 64 #include <sys/socket.h> 65 #include <sys/socketvar.h> 66 #include <sys/protosw.h> 67 #include <sys/proc.h> 68 #include <sys/jail.h> 69 #include <sys/stats.h> 70 71 #ifdef DDB 72 #include <ddb/ddb.h> 73 #endif 74 75 #include <net/if.h> 76 #include <net/if_var.h> 77 #include <net/route.h> 78 #include <net/vnet.h> 79 80 #include <netinet/in.h> 81 #include <netinet/in_kdtrace.h> 82 #include <netinet/in_pcb.h> 83 #include <netinet/in_systm.h> 84 #include <netinet/in_var.h> 85 #include <netinet/ip.h> 86 #include <netinet/ip_var.h> 87 #ifdef INET6 88 #include <netinet/ip6.h> 89 #include <netinet6/in6_pcb.h> 90 #include <netinet6/ip6_var.h> 91 #include <netinet6/scope6_var.h> 92 #endif 93 #include <netinet/tcp.h> 94 #include <netinet/tcp_fsm.h> 95 #include <netinet/tcp_seq.h> 96 #include <netinet/tcp_timer.h> 97 #include <netinet/tcp_var.h> 98 #include <netinet/tcp_log_buf.h> 99 #include <netinet/tcpip.h> 100 #include <netinet/cc/cc.h> 101 #include <netinet/tcp_fastopen.h> 102 #include <netinet/tcp_hpts.h> 103 #ifdef TCPPCAP 104 #include <netinet/tcp_pcap.h> 105 #endif 106 #ifdef TCPDEBUG 107 #include <netinet/tcp_debug.h> 108 #endif 109 #ifdef TCP_OFFLOAD 110 #include <netinet/tcp_offload.h> 111 #endif 112 #include <netipsec/ipsec_support.h> 113 114 #include <vm/vm.h> 115 #include <vm/vm_param.h> 116 #include <vm/pmap.h> 117 #include <vm/vm_extern.h> 118 #include <vm/vm_map.h> 119 #include <vm/vm_page.h> 120 121 /* 122 * TCP protocol interface to socket abstraction. 123 */ 124 #ifdef INET 125 static int tcp_connect(struct tcpcb *, struct sockaddr *, 126 struct thread *td); 127 #endif /* INET */ 128 #ifdef INET6 129 static int tcp6_connect(struct tcpcb *, struct sockaddr *, 130 struct thread *td); 131 #endif /* INET6 */ 132 static void tcp_disconnect(struct tcpcb *); 133 static void tcp_usrclosed(struct tcpcb *); 134 static void tcp_fill_info(struct tcpcb *, struct tcp_info *); 135 136 static int tcp_pru_options_support(struct tcpcb *tp, int flags); 137 138 #ifdef TCPDEBUG 139 #define TCPDEBUG0 int ostate = 0 140 #define TCPDEBUG1() ostate = tp ? tp->t_state : 0 141 #define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \ 142 tcp_trace(TA_USER, ostate, tp, 0, 0, req) 143 #else 144 #define TCPDEBUG0 145 #define TCPDEBUG1() 146 #define TCPDEBUG2(req) 147 #endif 148 149 /* 150 * tcp_require_unique port requires a globally-unique source port for each 151 * outgoing connection. The default is to require the 4-tuple to be unique. 152 */ 153 VNET_DEFINE(int, tcp_require_unique_port) = 0; 154 SYSCTL_INT(_net_inet_tcp, OID_AUTO, require_unique_port, 155 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(tcp_require_unique_port), 0, 156 "Require globally-unique ephemeral port for outgoing connections"); 157 #define V_tcp_require_unique_port VNET(tcp_require_unique_port) 158 159 /* 160 * TCP attaches to socket via pru_attach(), reserving space, 161 * and an internet control block. 162 */ 163 static int 164 tcp_usr_attach(struct socket *so, int proto, struct thread *td) 165 { 166 struct inpcb *inp; 167 struct tcpcb *tp = NULL; 168 int error; 169 TCPDEBUG0; 170 171 inp = sotoinpcb(so); 172 KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL")); 173 TCPDEBUG1(); 174 175 error = soreserve(so, V_tcp_sendspace, V_tcp_recvspace); 176 if (error) 177 goto out; 178 179 so->so_rcv.sb_flags |= SB_AUTOSIZE; 180 so->so_snd.sb_flags |= SB_AUTOSIZE; 181 error = in_pcballoc(so, &V_tcbinfo); 182 if (error) 183 goto out; 184 inp = sotoinpcb(so); 185 tp = tcp_newtcpcb(inp); 186 if (tp == NULL) { 187 error = ENOBUFS; 188 in_pcbdetach(inp); 189 in_pcbfree(inp); 190 goto out; 191 } 192 tp->t_state = TCPS_CLOSED; 193 INP_WUNLOCK(inp); 194 TCPSTATES_INC(TCPS_CLOSED); 195 out: 196 TCPDEBUG2(PRU_ATTACH); 197 TCP_PROBE2(debug__user, tp, PRU_ATTACH); 198 return (error); 199 } 200 201 /* 202 * tcp_usr_detach is called when the socket layer loses its final reference 203 * to the socket, be it a file descriptor reference, a reference from TCP, 204 * etc. At this point, there is only one case in which we will keep around 205 * inpcb state: time wait. 206 */ 207 static void 208 tcp_usr_detach(struct socket *so) 209 { 210 struct inpcb *inp; 211 struct tcpcb *tp; 212 213 inp = sotoinpcb(so); 214 KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); 215 INP_WLOCK(inp); 216 KASSERT(so->so_pcb == inp && inp->inp_socket == so, 217 ("%s: socket %p inp %p mismatch", __func__, so, inp)); 218 219 tp = intotcpcb(inp); 220 221 if (inp->inp_flags & INP_TIMEWAIT) { 222 /* 223 * There are two cases to handle: one in which the time wait 224 * state is being discarded (INP_DROPPED), and one in which 225 * this connection will remain in timewait. In the former, 226 * it is time to discard all state (except tcptw, which has 227 * already been discarded by the timewait close code, which 228 * should be further up the call stack somewhere). In the 229 * latter case, we detach from the socket, but leave the pcb 230 * present until timewait ends. 231 * 232 * XXXRW: Would it be cleaner to free the tcptw here? 233 * 234 * Astute question indeed, from twtcp perspective there are 235 * four cases to consider: 236 * 237 * #1 tcp_usr_detach is called at tcptw creation time by 238 * tcp_twstart, then do not discard the newly created tcptw 239 * and leave inpcb present until timewait ends 240 * #2 tcp_usr_detach is called at tcptw creation time by 241 * tcp_twstart, but connection is local and tw will be 242 * discarded immediately 243 * #3 tcp_usr_detach is called at timewait end (or reuse) by 244 * tcp_twclose, then the tcptw has already been discarded 245 * (or reused) and inpcb is freed here 246 * #4 tcp_usr_detach is called() after timewait ends (or reuse) 247 * (e.g. by soclose), then tcptw has already been discarded 248 * (or reused) and inpcb is freed here 249 * 250 * In all three cases the tcptw should not be freed here. 251 */ 252 if (inp->inp_flags & INP_DROPPED) { 253 KASSERT(tp == NULL, ("tcp_detach: INP_TIMEWAIT && " 254 "INP_DROPPED && tp != NULL")); 255 in_pcbdetach(inp); 256 in_pcbfree(inp); 257 } else { 258 in_pcbdetach(inp); 259 INP_WUNLOCK(inp); 260 } 261 } else { 262 /* 263 * If the connection is not in timewait, it must be either 264 * dropped or embryonic. 265 */ 266 KASSERT(inp->inp_flags & INP_DROPPED || 267 tp->t_state < TCPS_SYN_SENT, 268 ("%s: inp %p not dropped or embryonic", __func__, inp)); 269 tcp_discardcb(tp); 270 in_pcbdetach(inp); 271 in_pcbfree(inp); 272 } 273 } 274 275 #ifdef INET 276 /* 277 * Give the socket an address. 278 */ 279 static int 280 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 281 { 282 int error = 0; 283 struct inpcb *inp; 284 #ifdef KDTRACE_HOOKS 285 struct tcpcb *tp = NULL; 286 #endif 287 struct sockaddr_in *sinp; 288 289 sinp = (struct sockaddr_in *)nam; 290 if (nam->sa_family != AF_INET) { 291 /* 292 * Preserve compatibility with old programs. 293 */ 294 if (nam->sa_family != AF_UNSPEC || 295 nam->sa_len < offsetof(struct sockaddr_in, sin_zero) || 296 sinp->sin_addr.s_addr != INADDR_ANY) 297 return (EAFNOSUPPORT); 298 nam->sa_family = AF_INET; 299 } 300 if (nam->sa_len != sizeof(*sinp)) 301 return (EINVAL); 302 303 /* 304 * Must check for multicast addresses and disallow binding 305 * to them. 306 */ 307 if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) 308 return (EAFNOSUPPORT); 309 310 TCPDEBUG0; 311 inp = sotoinpcb(so); 312 KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL")); 313 INP_WLOCK(inp); 314 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 315 error = EINVAL; 316 goto out; 317 } 318 #ifdef KDTRACE_HOOKS 319 tp = intotcpcb(inp); 320 #endif 321 TCPDEBUG1(); 322 INP_HASH_WLOCK(&V_tcbinfo); 323 error = in_pcbbind(inp, nam, td->td_ucred); 324 INP_HASH_WUNLOCK(&V_tcbinfo); 325 out: 326 TCPDEBUG2(PRU_BIND); 327 TCP_PROBE2(debug__user, tp, PRU_BIND); 328 INP_WUNLOCK(inp); 329 330 return (error); 331 } 332 #endif /* INET */ 333 334 #ifdef INET6 335 static int 336 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 337 { 338 int error = 0; 339 struct inpcb *inp; 340 #ifdef KDTRACE_HOOKS 341 struct tcpcb *tp = NULL; 342 #endif 343 struct sockaddr_in6 *sin6; 344 u_char vflagsav; 345 346 sin6 = (struct sockaddr_in6 *)nam; 347 if (nam->sa_family != AF_INET6) 348 return (EAFNOSUPPORT); 349 if (nam->sa_len != sizeof(*sin6)) 350 return (EINVAL); 351 352 /* 353 * Must check for multicast addresses and disallow binding 354 * to them. 355 */ 356 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 357 return (EAFNOSUPPORT); 358 359 TCPDEBUG0; 360 inp = sotoinpcb(so); 361 KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL")); 362 INP_WLOCK(inp); 363 vflagsav = inp->inp_vflag; 364 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 365 error = EINVAL; 366 goto out; 367 } 368 #ifdef KDTRACE_HOOKS 369 tp = intotcpcb(inp); 370 #endif 371 TCPDEBUG1(); 372 INP_HASH_WLOCK(&V_tcbinfo); 373 inp->inp_vflag &= ~INP_IPV4; 374 inp->inp_vflag |= INP_IPV6; 375 #ifdef INET 376 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { 377 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) 378 inp->inp_vflag |= INP_IPV4; 379 else if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { 380 struct sockaddr_in sin; 381 382 in6_sin6_2_sin(&sin, sin6); 383 if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) { 384 error = EAFNOSUPPORT; 385 INP_HASH_WUNLOCK(&V_tcbinfo); 386 goto out; 387 } 388 inp->inp_vflag |= INP_IPV4; 389 inp->inp_vflag &= ~INP_IPV6; 390 error = in_pcbbind(inp, (struct sockaddr *)&sin, 391 td->td_ucred); 392 INP_HASH_WUNLOCK(&V_tcbinfo); 393 goto out; 394 } 395 } 396 #endif 397 error = in6_pcbbind(inp, nam, td->td_ucred); 398 INP_HASH_WUNLOCK(&V_tcbinfo); 399 out: 400 if (error != 0) 401 inp->inp_vflag = vflagsav; 402 TCPDEBUG2(PRU_BIND); 403 TCP_PROBE2(debug__user, tp, PRU_BIND); 404 INP_WUNLOCK(inp); 405 return (error); 406 } 407 #endif /* INET6 */ 408 409 #ifdef INET 410 /* 411 * Prepare to accept connections. 412 */ 413 static int 414 tcp_usr_listen(struct socket *so, int backlog, struct thread *td) 415 { 416 int error = 0; 417 struct inpcb *inp; 418 struct tcpcb *tp = NULL; 419 420 TCPDEBUG0; 421 inp = sotoinpcb(so); 422 KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL")); 423 INP_WLOCK(inp); 424 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 425 error = EINVAL; 426 goto out; 427 } 428 tp = intotcpcb(inp); 429 TCPDEBUG1(); 430 SOCK_LOCK(so); 431 error = solisten_proto_check(so); 432 if (error != 0) { 433 SOCK_UNLOCK(so); 434 goto out; 435 } 436 if (inp->inp_lport == 0) { 437 INP_HASH_WLOCK(&V_tcbinfo); 438 error = in_pcbbind(inp, NULL, td->td_ucred); 439 INP_HASH_WUNLOCK(&V_tcbinfo); 440 } 441 if (error == 0) { 442 tcp_state_change(tp, TCPS_LISTEN); 443 solisten_proto(so, backlog); 444 #ifdef TCP_OFFLOAD 445 if ((so->so_options & SO_NO_OFFLOAD) == 0) 446 tcp_offload_listen_start(tp); 447 #endif 448 } else { 449 solisten_proto_abort(so); 450 } 451 SOCK_UNLOCK(so); 452 453 if (IS_FASTOPEN(tp->t_flags)) 454 tp->t_tfo_pending = tcp_fastopen_alloc_counter(); 455 456 out: 457 TCPDEBUG2(PRU_LISTEN); 458 TCP_PROBE2(debug__user, tp, PRU_LISTEN); 459 INP_WUNLOCK(inp); 460 return (error); 461 } 462 #endif /* INET */ 463 464 #ifdef INET6 465 static int 466 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td) 467 { 468 int error = 0; 469 struct inpcb *inp; 470 struct tcpcb *tp = NULL; 471 u_char vflagsav; 472 473 TCPDEBUG0; 474 inp = sotoinpcb(so); 475 KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL")); 476 INP_WLOCK(inp); 477 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 478 error = EINVAL; 479 goto out; 480 } 481 vflagsav = inp->inp_vflag; 482 tp = intotcpcb(inp); 483 TCPDEBUG1(); 484 SOCK_LOCK(so); 485 error = solisten_proto_check(so); 486 if (error != 0) { 487 SOCK_UNLOCK(so); 488 goto out; 489 } 490 INP_HASH_WLOCK(&V_tcbinfo); 491 if (inp->inp_lport == 0) { 492 inp->inp_vflag &= ~INP_IPV4; 493 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) 494 inp->inp_vflag |= INP_IPV4; 495 error = in6_pcbbind(inp, NULL, td->td_ucred); 496 } 497 INP_HASH_WUNLOCK(&V_tcbinfo); 498 if (error == 0) { 499 tcp_state_change(tp, TCPS_LISTEN); 500 solisten_proto(so, backlog); 501 #ifdef TCP_OFFLOAD 502 if ((so->so_options & SO_NO_OFFLOAD) == 0) 503 tcp_offload_listen_start(tp); 504 #endif 505 } else { 506 solisten_proto_abort(so); 507 } 508 SOCK_UNLOCK(so); 509 510 if (IS_FASTOPEN(tp->t_flags)) 511 tp->t_tfo_pending = tcp_fastopen_alloc_counter(); 512 513 if (error != 0) 514 inp->inp_vflag = vflagsav; 515 516 out: 517 TCPDEBUG2(PRU_LISTEN); 518 TCP_PROBE2(debug__user, tp, PRU_LISTEN); 519 INP_WUNLOCK(inp); 520 return (error); 521 } 522 #endif /* INET6 */ 523 524 #ifdef INET 525 /* 526 * Initiate connection to peer. 527 * Create a template for use in transmissions on this connection. 528 * Enter SYN_SENT state, and mark socket as connecting. 529 * Start keep-alive timer, and seed output sequence space. 530 * Send initial segment on connection. 531 */ 532 static int 533 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 534 { 535 struct epoch_tracker et; 536 int error = 0; 537 struct inpcb *inp; 538 struct tcpcb *tp = NULL; 539 struct sockaddr_in *sinp; 540 541 sinp = (struct sockaddr_in *)nam; 542 if (nam->sa_family != AF_INET) 543 return (EAFNOSUPPORT); 544 if (nam->sa_len != sizeof (*sinp)) 545 return (EINVAL); 546 547 /* 548 * Must disallow TCP ``connections'' to multicast addresses. 549 */ 550 if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) 551 return (EAFNOSUPPORT); 552 if (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST) 553 return (EACCES); 554 if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0) 555 return (error); 556 557 TCPDEBUG0; 558 inp = sotoinpcb(so); 559 KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL")); 560 INP_WLOCK(inp); 561 if (inp->inp_flags & INP_TIMEWAIT) { 562 error = EADDRINUSE; 563 goto out; 564 } 565 if (inp->inp_flags & INP_DROPPED) { 566 error = ECONNREFUSED; 567 goto out; 568 } 569 if (SOLISTENING(so)) { 570 error = EOPNOTSUPP; 571 goto out; 572 } 573 tp = intotcpcb(inp); 574 TCPDEBUG1(); 575 NET_EPOCH_ENTER(et); 576 if ((error = tcp_connect(tp, nam, td)) != 0) 577 goto out_in_epoch; 578 #ifdef TCP_OFFLOAD 579 if (registered_toedevs > 0 && 580 (so->so_options & SO_NO_OFFLOAD) == 0 && 581 (error = tcp_offload_connect(so, nam)) == 0) 582 goto out_in_epoch; 583 #endif 584 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp)); 585 error = tcp_output(tp); 586 KASSERT(error >= 0, ("TCP stack %s requested tcp_drop(%p) at connect()" 587 ", error code %d", tp->t_fb->tfb_tcp_block_name, tp, -error)); 588 out_in_epoch: 589 NET_EPOCH_EXIT(et); 590 out: 591 TCPDEBUG2(PRU_CONNECT); 592 TCP_PROBE2(debug__user, tp, PRU_CONNECT); 593 INP_WUNLOCK(inp); 594 return (error); 595 } 596 #endif /* INET */ 597 598 #ifdef INET6 599 static int 600 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 601 { 602 struct epoch_tracker et; 603 int error = 0; 604 struct inpcb *inp; 605 struct tcpcb *tp = NULL; 606 struct sockaddr_in6 *sin6; 607 u_int8_t incflagsav; 608 u_char vflagsav; 609 610 TCPDEBUG0; 611 612 sin6 = (struct sockaddr_in6 *)nam; 613 if (nam->sa_family != AF_INET6) 614 return (EAFNOSUPPORT); 615 if (nam->sa_len != sizeof (*sin6)) 616 return (EINVAL); 617 618 /* 619 * Must disallow TCP ``connections'' to multicast addresses. 620 */ 621 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 622 return (EAFNOSUPPORT); 623 624 inp = sotoinpcb(so); 625 KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL")); 626 INP_WLOCK(inp); 627 vflagsav = inp->inp_vflag; 628 incflagsav = inp->inp_inc.inc_flags; 629 if (inp->inp_flags & INP_TIMEWAIT) { 630 error = EADDRINUSE; 631 goto out; 632 } 633 if (inp->inp_flags & INP_DROPPED) { 634 error = ECONNREFUSED; 635 goto out; 636 } 637 if (SOLISTENING(so)) { 638 error = EINVAL; 639 goto out; 640 } 641 tp = intotcpcb(inp); 642 TCPDEBUG1(); 643 #ifdef INET 644 /* 645 * XXXRW: Some confusion: V4/V6 flags relate to binding, and 646 * therefore probably require the hash lock, which isn't held here. 647 * Is this a significant problem? 648 */ 649 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { 650 struct sockaddr_in sin; 651 652 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { 653 error = EINVAL; 654 goto out; 655 } 656 if ((inp->inp_vflag & INP_IPV4) == 0) { 657 error = EAFNOSUPPORT; 658 goto out; 659 } 660 661 in6_sin6_2_sin(&sin, sin6); 662 if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) { 663 error = EAFNOSUPPORT; 664 goto out; 665 } 666 if (ntohl(sin.sin_addr.s_addr) == INADDR_BROADCAST) { 667 error = EACCES; 668 goto out; 669 } 670 if ((error = prison_remote_ip4(td->td_ucred, 671 &sin.sin_addr)) != 0) 672 goto out; 673 inp->inp_vflag |= INP_IPV4; 674 inp->inp_vflag &= ~INP_IPV6; 675 NET_EPOCH_ENTER(et); 676 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0) 677 goto out_in_epoch; 678 #ifdef TCP_OFFLOAD 679 if (registered_toedevs > 0 && 680 (so->so_options & SO_NO_OFFLOAD) == 0 && 681 (error = tcp_offload_connect(so, nam)) == 0) 682 goto out_in_epoch; 683 #endif 684 error = tcp_output(tp); 685 goto out_in_epoch; 686 } else { 687 if ((inp->inp_vflag & INP_IPV6) == 0) { 688 error = EAFNOSUPPORT; 689 goto out; 690 } 691 } 692 #endif 693 if ((error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr)) != 0) 694 goto out; 695 inp->inp_vflag &= ~INP_IPV4; 696 inp->inp_vflag |= INP_IPV6; 697 inp->inp_inc.inc_flags |= INC_ISIPV6; 698 if ((error = tcp6_connect(tp, nam, td)) != 0) 699 goto out; 700 #ifdef TCP_OFFLOAD 701 if (registered_toedevs > 0 && 702 (so->so_options & SO_NO_OFFLOAD) == 0 && 703 (error = tcp_offload_connect(so, nam)) == 0) 704 goto out; 705 #endif 706 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp)); 707 NET_EPOCH_ENTER(et); 708 error = tcp_output(tp); 709 #ifdef INET 710 out_in_epoch: 711 #endif 712 NET_EPOCH_EXIT(et); 713 out: 714 KASSERT(error >= 0, ("TCP stack %s requested tcp_drop(%p) at connect()" 715 ", error code %d", tp->t_fb->tfb_tcp_block_name, tp, -error)); 716 /* 717 * If the implicit bind in the connect call fails, restore 718 * the flags we modified. 719 */ 720 if (error != 0 && inp->inp_lport == 0) { 721 inp->inp_vflag = vflagsav; 722 inp->inp_inc.inc_flags = incflagsav; 723 } 724 725 TCPDEBUG2(PRU_CONNECT); 726 TCP_PROBE2(debug__user, tp, PRU_CONNECT); 727 INP_WUNLOCK(inp); 728 return (error); 729 } 730 #endif /* INET6 */ 731 732 /* 733 * Initiate disconnect from peer. 734 * If connection never passed embryonic stage, just drop; 735 * else if don't need to let data drain, then can just drop anyways, 736 * else have to begin TCP shutdown process: mark socket disconnecting, 737 * drain unread data, state switch to reflect user close, and 738 * send segment (e.g. FIN) to peer. Socket will be really disconnected 739 * when peer sends FIN and acks ours. 740 * 741 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 742 */ 743 static int 744 tcp_usr_disconnect(struct socket *so) 745 { 746 struct inpcb *inp; 747 struct tcpcb *tp = NULL; 748 struct epoch_tracker et; 749 int error = 0; 750 751 TCPDEBUG0; 752 NET_EPOCH_ENTER(et); 753 inp = sotoinpcb(so); 754 KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL")); 755 INP_WLOCK(inp); 756 if (inp->inp_flags & INP_TIMEWAIT) 757 goto out; 758 if (inp->inp_flags & INP_DROPPED) { 759 error = ECONNRESET; 760 goto out; 761 } 762 tp = intotcpcb(inp); 763 TCPDEBUG1(); 764 tcp_disconnect(tp); 765 out: 766 TCPDEBUG2(PRU_DISCONNECT); 767 TCP_PROBE2(debug__user, tp, PRU_DISCONNECT); 768 INP_WUNLOCK(inp); 769 NET_EPOCH_EXIT(et); 770 return (error); 771 } 772 773 #ifdef INET 774 /* 775 * Accept a connection. Essentially all the work is done at higher levels; 776 * just return the address of the peer, storing through addr. 777 */ 778 static int 779 tcp_usr_accept(struct socket *so, struct sockaddr **nam) 780 { 781 int error = 0; 782 struct inpcb *inp = NULL; 783 #ifdef KDTRACE_HOOKS 784 struct tcpcb *tp = NULL; 785 #endif 786 struct in_addr addr; 787 in_port_t port = 0; 788 TCPDEBUG0; 789 790 if (so->so_state & SS_ISDISCONNECTED) 791 return (ECONNABORTED); 792 793 inp = sotoinpcb(so); 794 KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL")); 795 INP_WLOCK(inp); 796 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 797 error = ECONNABORTED; 798 goto out; 799 } 800 #ifdef KDTRACE_HOOKS 801 tp = intotcpcb(inp); 802 #endif 803 TCPDEBUG1(); 804 805 /* 806 * We inline in_getpeeraddr and COMMON_END here, so that we can 807 * copy the data of interest and defer the malloc until after we 808 * release the lock. 809 */ 810 port = inp->inp_fport; 811 addr = inp->inp_faddr; 812 813 out: 814 TCPDEBUG2(PRU_ACCEPT); 815 TCP_PROBE2(debug__user, tp, PRU_ACCEPT); 816 INP_WUNLOCK(inp); 817 if (error == 0) 818 *nam = in_sockaddr(port, &addr); 819 return error; 820 } 821 #endif /* INET */ 822 823 #ifdef INET6 824 static int 825 tcp6_usr_accept(struct socket *so, struct sockaddr **nam) 826 { 827 struct inpcb *inp = NULL; 828 int error = 0; 829 #ifdef KDTRACE_HOOKS 830 struct tcpcb *tp = NULL; 831 #endif 832 struct in_addr addr; 833 struct in6_addr addr6; 834 struct epoch_tracker et; 835 in_port_t port = 0; 836 int v4 = 0; 837 TCPDEBUG0; 838 839 if (so->so_state & SS_ISDISCONNECTED) 840 return (ECONNABORTED); 841 842 inp = sotoinpcb(so); 843 KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL")); 844 NET_EPOCH_ENTER(et); 845 INP_WLOCK(inp); 846 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 847 error = ECONNABORTED; 848 goto out; 849 } 850 #ifdef KDTRACE_HOOKS 851 tp = intotcpcb(inp); 852 #endif 853 TCPDEBUG1(); 854 855 /* 856 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can 857 * copy the data of interest and defer the malloc until after we 858 * release the lock. 859 */ 860 if (inp->inp_vflag & INP_IPV4) { 861 v4 = 1; 862 port = inp->inp_fport; 863 addr = inp->inp_faddr; 864 } else { 865 port = inp->inp_fport; 866 addr6 = inp->in6p_faddr; 867 } 868 869 out: 870 TCPDEBUG2(PRU_ACCEPT); 871 TCP_PROBE2(debug__user, tp, PRU_ACCEPT); 872 INP_WUNLOCK(inp); 873 NET_EPOCH_EXIT(et); 874 if (error == 0) { 875 if (v4) 876 *nam = in6_v4mapsin6_sockaddr(port, &addr); 877 else 878 *nam = in6_sockaddr(port, &addr6); 879 } 880 return error; 881 } 882 #endif /* INET6 */ 883 884 /* 885 * Mark the connection as being incapable of further output. 886 */ 887 static int 888 tcp_usr_shutdown(struct socket *so) 889 { 890 int error = 0; 891 struct inpcb *inp; 892 struct tcpcb *tp = NULL; 893 struct epoch_tracker et; 894 895 TCPDEBUG0; 896 inp = sotoinpcb(so); 897 KASSERT(inp != NULL, ("inp == NULL")); 898 INP_WLOCK(inp); 899 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 900 INP_WUNLOCK(inp); 901 return (ECONNRESET); 902 } 903 tp = intotcpcb(inp); 904 NET_EPOCH_ENTER(et); 905 TCPDEBUG1(); 906 socantsendmore(so); 907 tcp_usrclosed(tp); 908 if (!(inp->inp_flags & INP_DROPPED)) 909 error = tcp_output_nodrop(tp); 910 TCPDEBUG2(PRU_SHUTDOWN); 911 TCP_PROBE2(debug__user, tp, PRU_SHUTDOWN); 912 error = tcp_unlock_or_drop(tp, error); 913 NET_EPOCH_EXIT(et); 914 915 return (error); 916 } 917 918 /* 919 * After a receive, possibly send window update to peer. 920 */ 921 static int 922 tcp_usr_rcvd(struct socket *so, int flags) 923 { 924 struct epoch_tracker et; 925 struct inpcb *inp; 926 struct tcpcb *tp = NULL; 927 int outrv = 0, error = 0; 928 929 TCPDEBUG0; 930 inp = sotoinpcb(so); 931 KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL")); 932 INP_WLOCK(inp); 933 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 934 INP_WUNLOCK(inp); 935 return (ECONNRESET); 936 } 937 tp = intotcpcb(inp); 938 NET_EPOCH_ENTER(et); 939 TCPDEBUG1(); 940 /* 941 * For passively-created TFO connections, don't attempt a window 942 * update while still in SYN_RECEIVED as this may trigger an early 943 * SYN|ACK. It is preferable to have the SYN|ACK be sent along with 944 * application response data, or failing that, when the DELACK timer 945 * expires. 946 */ 947 if (IS_FASTOPEN(tp->t_flags) && 948 (tp->t_state == TCPS_SYN_RECEIVED)) 949 goto out; 950 #ifdef TCP_OFFLOAD 951 if (tp->t_flags & TF_TOE) 952 tcp_offload_rcvd(tp); 953 else 954 #endif 955 outrv = tcp_output_nodrop(tp); 956 out: 957 TCPDEBUG2(PRU_RCVD); 958 TCP_PROBE2(debug__user, tp, PRU_RCVD); 959 (void) tcp_unlock_or_drop(tp, outrv); 960 NET_EPOCH_EXIT(et); 961 return (error); 962 } 963 964 /* 965 * Do a send by putting data in output queue and updating urgent 966 * marker if URG set. Possibly send more data. Unlike the other 967 * pru_*() routines, the mbuf chains are our responsibility. We 968 * must either enqueue them or free them. The other pru_* routines 969 * generally are caller-frees. 970 */ 971 static int 972 tcp_usr_send(struct socket *so, int flags, struct mbuf *m, 973 struct sockaddr *nam, struct mbuf *control, struct thread *td) 974 { 975 struct epoch_tracker et; 976 int error = 0; 977 struct inpcb *inp; 978 struct tcpcb *tp = NULL; 979 #ifdef INET 980 #ifdef INET6 981 struct sockaddr_in sin; 982 #endif 983 struct sockaddr_in *sinp; 984 #endif 985 #ifdef INET6 986 int isipv6; 987 #endif 988 u_int8_t incflagsav; 989 u_char vflagsav; 990 bool restoreflags; 991 TCPDEBUG0; 992 993 if (control != NULL) { 994 /* TCP doesn't do control messages (rights, creds, etc) */ 995 if (control->m_len) { 996 m_freem(control); 997 return (EINVAL); 998 } 999 m_freem(control); /* empty control, just free it */ 1000 } 1001 1002 inp = sotoinpcb(so); 1003 KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL")); 1004 INP_WLOCK(inp); 1005 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 1006 if (m != NULL && (flags & PRUS_NOTREADY) == 0) 1007 m_freem(m); 1008 INP_WUNLOCK(inp); 1009 return (ECONNRESET); 1010 } 1011 1012 vflagsav = inp->inp_vflag; 1013 incflagsav = inp->inp_inc.inc_flags; 1014 restoreflags = false; 1015 tp = intotcpcb(inp); 1016 1017 NET_EPOCH_ENTER(et); 1018 if ((flags & PRUS_OOB) != 0 && 1019 (error = tcp_pru_options_support(tp, PRUS_OOB)) != 0) 1020 goto out; 1021 1022 TCPDEBUG1(); 1023 if (nam != NULL && tp->t_state < TCPS_SYN_SENT) { 1024 if (tp->t_state == TCPS_LISTEN) { 1025 error = EINVAL; 1026 goto out; 1027 } 1028 switch (nam->sa_family) { 1029 #ifdef INET 1030 case AF_INET: 1031 sinp = (struct sockaddr_in *)nam; 1032 if (sinp->sin_len != sizeof(struct sockaddr_in)) { 1033 error = EINVAL; 1034 goto out; 1035 } 1036 if ((inp->inp_vflag & INP_IPV6) != 0) { 1037 error = EAFNOSUPPORT; 1038 goto out; 1039 } 1040 if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { 1041 error = EAFNOSUPPORT; 1042 goto out; 1043 } 1044 if (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST) { 1045 error = EACCES; 1046 goto out; 1047 } 1048 if ((error = prison_remote_ip4(td->td_ucred, 1049 &sinp->sin_addr))) 1050 goto out; 1051 #ifdef INET6 1052 isipv6 = 0; 1053 #endif 1054 break; 1055 #endif /* INET */ 1056 #ifdef INET6 1057 case AF_INET6: 1058 { 1059 struct sockaddr_in6 *sin6; 1060 1061 sin6 = (struct sockaddr_in6 *)nam; 1062 if (sin6->sin6_len != sizeof(*sin6)) { 1063 error = EINVAL; 1064 goto out; 1065 } 1066 if ((inp->inp_vflag & INP_IPV6PROTO) == 0) { 1067 error = EAFNOSUPPORT; 1068 goto out; 1069 } 1070 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { 1071 error = EAFNOSUPPORT; 1072 goto out; 1073 } 1074 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { 1075 #ifdef INET 1076 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { 1077 error = EINVAL; 1078 goto out; 1079 } 1080 if ((inp->inp_vflag & INP_IPV4) == 0) { 1081 error = EAFNOSUPPORT; 1082 goto out; 1083 } 1084 restoreflags = true; 1085 inp->inp_vflag &= ~INP_IPV6; 1086 sinp = &sin; 1087 in6_sin6_2_sin(sinp, sin6); 1088 if (IN_MULTICAST( 1089 ntohl(sinp->sin_addr.s_addr))) { 1090 error = EAFNOSUPPORT; 1091 goto out; 1092 } 1093 if ((error = prison_remote_ip4(td->td_ucred, 1094 &sinp->sin_addr))) 1095 goto out; 1096 isipv6 = 0; 1097 #else /* !INET */ 1098 error = EAFNOSUPPORT; 1099 goto out; 1100 #endif /* INET */ 1101 } else { 1102 if ((inp->inp_vflag & INP_IPV6) == 0) { 1103 error = EAFNOSUPPORT; 1104 goto out; 1105 } 1106 restoreflags = true; 1107 inp->inp_vflag &= ~INP_IPV4; 1108 inp->inp_inc.inc_flags |= INC_ISIPV6; 1109 if ((error = prison_remote_ip6(td->td_ucred, 1110 &sin6->sin6_addr))) 1111 goto out; 1112 isipv6 = 1; 1113 } 1114 break; 1115 } 1116 #endif /* INET6 */ 1117 default: 1118 error = EAFNOSUPPORT; 1119 goto out; 1120 } 1121 } 1122 if (!(flags & PRUS_OOB)) { 1123 sbappendstream(&so->so_snd, m, flags); 1124 m = NULL; 1125 if (nam && tp->t_state < TCPS_SYN_SENT) { 1126 KASSERT(tp->t_state == TCPS_CLOSED, 1127 ("%s: tp %p is listening", __func__, tp)); 1128 1129 /* 1130 * Do implied connect if not yet connected, 1131 * initialize window to default value, and 1132 * initialize maxseg using peer's cached MSS. 1133 */ 1134 #ifdef INET6 1135 if (isipv6) 1136 error = tcp6_connect(tp, nam, td); 1137 #endif /* INET6 */ 1138 #if defined(INET6) && defined(INET) 1139 else 1140 #endif 1141 #ifdef INET 1142 error = tcp_connect(tp, 1143 (struct sockaddr *)sinp, td); 1144 #endif 1145 /* 1146 * The bind operation in tcp_connect succeeded. We 1147 * no longer want to restore the flags if later 1148 * operations fail. 1149 */ 1150 if (error == 0 || inp->inp_lport != 0) 1151 restoreflags = false; 1152 1153 if (error) { 1154 /* m is freed if PRUS_NOTREADY is unset. */ 1155 sbflush(&so->so_snd); 1156 goto out; 1157 } 1158 if (IS_FASTOPEN(tp->t_flags)) 1159 tcp_fastopen_connect(tp); 1160 else { 1161 tp->snd_wnd = TTCP_CLIENT_SND_WND; 1162 tcp_mss(tp, -1); 1163 } 1164 } 1165 if (flags & PRUS_EOF) { 1166 /* 1167 * Close the send side of the connection after 1168 * the data is sent. 1169 */ 1170 socantsendmore(so); 1171 tcp_usrclosed(tp); 1172 } 1173 if (TCPS_HAVEESTABLISHED(tp->t_state) && 1174 ((tp->t_flags2 & TF2_FBYTES_COMPLETE) == 0) && 1175 (tp->t_fbyte_out == 0) && 1176 (so->so_snd.sb_ccc > 0)) { 1177 tp->t_fbyte_out = ticks; 1178 if (tp->t_fbyte_out == 0) 1179 tp->t_fbyte_out = 1; 1180 if (tp->t_fbyte_out && tp->t_fbyte_in) 1181 tp->t_flags2 |= TF2_FBYTES_COMPLETE; 1182 } 1183 if (!(inp->inp_flags & INP_DROPPED) && 1184 !(flags & PRUS_NOTREADY)) { 1185 if (flags & PRUS_MORETOCOME) 1186 tp->t_flags |= TF_MORETOCOME; 1187 error = tcp_output_nodrop(tp); 1188 if (flags & PRUS_MORETOCOME) 1189 tp->t_flags &= ~TF_MORETOCOME; 1190 } 1191 } else { 1192 /* 1193 * XXXRW: PRUS_EOF not implemented with PRUS_OOB? 1194 */ 1195 SOCKBUF_LOCK(&so->so_snd); 1196 if (sbspace(&so->so_snd) < -512) { 1197 SOCKBUF_UNLOCK(&so->so_snd); 1198 error = ENOBUFS; 1199 goto out; 1200 } 1201 /* 1202 * According to RFC961 (Assigned Protocols), 1203 * the urgent pointer points to the last octet 1204 * of urgent data. We continue, however, 1205 * to consider it to indicate the first octet 1206 * of data past the urgent section. 1207 * Otherwise, snd_up should be one lower. 1208 */ 1209 sbappendstream_locked(&so->so_snd, m, flags); 1210 SOCKBUF_UNLOCK(&so->so_snd); 1211 m = NULL; 1212 if (nam && tp->t_state < TCPS_SYN_SENT) { 1213 /* 1214 * Do implied connect if not yet connected, 1215 * initialize window to default value, and 1216 * initialize maxseg using peer's cached MSS. 1217 */ 1218 1219 /* 1220 * Not going to contemplate SYN|URG 1221 */ 1222 if (IS_FASTOPEN(tp->t_flags)) 1223 tp->t_flags &= ~TF_FASTOPEN; 1224 #ifdef INET6 1225 if (isipv6) 1226 error = tcp6_connect(tp, nam, td); 1227 #endif /* INET6 */ 1228 #if defined(INET6) && defined(INET) 1229 else 1230 #endif 1231 #ifdef INET 1232 error = tcp_connect(tp, 1233 (struct sockaddr *)sinp, td); 1234 #endif 1235 /* 1236 * The bind operation in tcp_connect succeeded. We 1237 * no longer want to restore the flags if later 1238 * operations fail. 1239 */ 1240 if (error == 0 || inp->inp_lport != 0) 1241 restoreflags = false; 1242 1243 if (error != 0) { 1244 /* m is freed if PRUS_NOTREADY is unset. */ 1245 sbflush(&so->so_snd); 1246 goto out; 1247 } 1248 tp->snd_wnd = TTCP_CLIENT_SND_WND; 1249 tcp_mss(tp, -1); 1250 } 1251 tp->snd_up = tp->snd_una + sbavail(&so->so_snd); 1252 if ((flags & PRUS_NOTREADY) == 0) { 1253 tp->t_flags |= TF_FORCEDATA; 1254 error = tcp_output_nodrop(tp); 1255 tp->t_flags &= ~TF_FORCEDATA; 1256 } 1257 } 1258 TCP_LOG_EVENT(tp, NULL, 1259 &inp->inp_socket->so_rcv, 1260 &inp->inp_socket->so_snd, 1261 TCP_LOG_USERSEND, error, 1262 0, NULL, false); 1263 1264 out: 1265 /* 1266 * In case of PRUS_NOTREADY, the caller or tcp_usr_ready() is 1267 * responsible for freeing memory. 1268 */ 1269 if (m != NULL && (flags & PRUS_NOTREADY) == 0) 1270 m_freem(m); 1271 1272 /* 1273 * If the request was unsuccessful and we changed flags, 1274 * restore the original flags. 1275 */ 1276 if (error != 0 && restoreflags) { 1277 inp->inp_vflag = vflagsav; 1278 inp->inp_inc.inc_flags = incflagsav; 1279 } 1280 TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB : 1281 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); 1282 TCP_PROBE2(debug__user, tp, (flags & PRUS_OOB) ? PRU_SENDOOB : 1283 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); 1284 error = tcp_unlock_or_drop(tp, error); 1285 NET_EPOCH_EXIT(et); 1286 return (error); 1287 } 1288 1289 static int 1290 tcp_usr_ready(struct socket *so, struct mbuf *m, int count) 1291 { 1292 struct epoch_tracker et; 1293 struct inpcb *inp; 1294 struct tcpcb *tp; 1295 int error; 1296 1297 inp = sotoinpcb(so); 1298 INP_WLOCK(inp); 1299 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 1300 INP_WUNLOCK(inp); 1301 mb_free_notready(m, count); 1302 return (ECONNRESET); 1303 } 1304 tp = intotcpcb(inp); 1305 1306 SOCKBUF_LOCK(&so->so_snd); 1307 error = sbready(&so->so_snd, m, count); 1308 SOCKBUF_UNLOCK(&so->so_snd); 1309 if (error) { 1310 INP_WUNLOCK(inp); 1311 return (error); 1312 } 1313 NET_EPOCH_ENTER(et); 1314 error = tcp_output_unlock(tp); 1315 NET_EPOCH_EXIT(et); 1316 1317 return (error); 1318 } 1319 1320 /* 1321 * Abort the TCP. Drop the connection abruptly. 1322 */ 1323 static void 1324 tcp_usr_abort(struct socket *so) 1325 { 1326 struct inpcb *inp; 1327 struct tcpcb *tp = NULL; 1328 struct epoch_tracker et; 1329 TCPDEBUG0; 1330 1331 inp = sotoinpcb(so); 1332 KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL")); 1333 1334 NET_EPOCH_ENTER(et); 1335 INP_WLOCK(inp); 1336 KASSERT(inp->inp_socket != NULL, 1337 ("tcp_usr_abort: inp_socket == NULL")); 1338 1339 /* 1340 * If we still have full TCP state, and we're not dropped, drop. 1341 */ 1342 if (!(inp->inp_flags & INP_TIMEWAIT) && 1343 !(inp->inp_flags & INP_DROPPED)) { 1344 tp = intotcpcb(inp); 1345 TCPDEBUG1(); 1346 tp = tcp_drop(tp, ECONNABORTED); 1347 if (tp == NULL) 1348 goto dropped; 1349 TCPDEBUG2(PRU_ABORT); 1350 TCP_PROBE2(debug__user, tp, PRU_ABORT); 1351 } 1352 if (!(inp->inp_flags & INP_DROPPED)) { 1353 soref(so); 1354 inp->inp_flags |= INP_SOCKREF; 1355 } 1356 INP_WUNLOCK(inp); 1357 dropped: 1358 NET_EPOCH_EXIT(et); 1359 } 1360 1361 /* 1362 * TCP socket is closed. Start friendly disconnect. 1363 */ 1364 static void 1365 tcp_usr_close(struct socket *so) 1366 { 1367 struct inpcb *inp; 1368 struct tcpcb *tp = NULL; 1369 struct epoch_tracker et; 1370 TCPDEBUG0; 1371 1372 inp = sotoinpcb(so); 1373 KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL")); 1374 1375 NET_EPOCH_ENTER(et); 1376 INP_WLOCK(inp); 1377 KASSERT(inp->inp_socket != NULL, 1378 ("tcp_usr_close: inp_socket == NULL")); 1379 1380 /* 1381 * If we still have full TCP state, and we're not dropped, initiate 1382 * a disconnect. 1383 */ 1384 if (!(inp->inp_flags & INP_TIMEWAIT) && 1385 !(inp->inp_flags & INP_DROPPED)) { 1386 tp = intotcpcb(inp); 1387 tp->t_flags |= TF_CLOSED; 1388 TCPDEBUG1(); 1389 tcp_disconnect(tp); 1390 TCPDEBUG2(PRU_CLOSE); 1391 TCP_PROBE2(debug__user, tp, PRU_CLOSE); 1392 } 1393 if (!(inp->inp_flags & INP_DROPPED)) { 1394 soref(so); 1395 inp->inp_flags |= INP_SOCKREF; 1396 } 1397 INP_WUNLOCK(inp); 1398 NET_EPOCH_EXIT(et); 1399 } 1400 1401 static int 1402 tcp_pru_options_support(struct tcpcb *tp, int flags) 1403 { 1404 /* 1405 * If the specific TCP stack has a pru_options 1406 * specified then it does not always support 1407 * all the PRU_XX options and we must ask it. 1408 * If the function is not specified then all 1409 * of the PRU_XX options are supported. 1410 */ 1411 int ret = 0; 1412 1413 if (tp->t_fb->tfb_pru_options) { 1414 ret = (*tp->t_fb->tfb_pru_options)(tp, flags); 1415 } 1416 return (ret); 1417 } 1418 1419 /* 1420 * Receive out-of-band data. 1421 */ 1422 static int 1423 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) 1424 { 1425 int error = 0; 1426 struct inpcb *inp; 1427 struct tcpcb *tp = NULL; 1428 1429 TCPDEBUG0; 1430 inp = sotoinpcb(so); 1431 KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL")); 1432 INP_WLOCK(inp); 1433 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 1434 error = ECONNRESET; 1435 goto out; 1436 } 1437 tp = intotcpcb(inp); 1438 error = tcp_pru_options_support(tp, PRUS_OOB); 1439 if (error) { 1440 goto out; 1441 } 1442 TCPDEBUG1(); 1443 if ((so->so_oobmark == 0 && 1444 (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) || 1445 so->so_options & SO_OOBINLINE || 1446 tp->t_oobflags & TCPOOB_HADDATA) { 1447 error = EINVAL; 1448 goto out; 1449 } 1450 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 1451 error = EWOULDBLOCK; 1452 goto out; 1453 } 1454 m->m_len = 1; 1455 *mtod(m, caddr_t) = tp->t_iobc; 1456 if ((flags & MSG_PEEK) == 0) 1457 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 1458 1459 out: 1460 TCPDEBUG2(PRU_RCVOOB); 1461 TCP_PROBE2(debug__user, tp, PRU_RCVOOB); 1462 INP_WUNLOCK(inp); 1463 return (error); 1464 } 1465 1466 #ifdef INET 1467 struct pr_usrreqs tcp_usrreqs = { 1468 .pru_abort = tcp_usr_abort, 1469 .pru_accept = tcp_usr_accept, 1470 .pru_attach = tcp_usr_attach, 1471 .pru_bind = tcp_usr_bind, 1472 .pru_connect = tcp_usr_connect, 1473 .pru_control = in_control, 1474 .pru_detach = tcp_usr_detach, 1475 .pru_disconnect = tcp_usr_disconnect, 1476 .pru_listen = tcp_usr_listen, 1477 .pru_peeraddr = in_getpeeraddr, 1478 .pru_rcvd = tcp_usr_rcvd, 1479 .pru_rcvoob = tcp_usr_rcvoob, 1480 .pru_send = tcp_usr_send, 1481 .pru_ready = tcp_usr_ready, 1482 .pru_shutdown = tcp_usr_shutdown, 1483 .pru_sockaddr = in_getsockaddr, 1484 .pru_sosetlabel = in_pcbsosetlabel, 1485 .pru_close = tcp_usr_close, 1486 }; 1487 #endif /* INET */ 1488 1489 #ifdef INET6 1490 struct pr_usrreqs tcp6_usrreqs = { 1491 .pru_abort = tcp_usr_abort, 1492 .pru_accept = tcp6_usr_accept, 1493 .pru_attach = tcp_usr_attach, 1494 .pru_bind = tcp6_usr_bind, 1495 .pru_connect = tcp6_usr_connect, 1496 .pru_control = in6_control, 1497 .pru_detach = tcp_usr_detach, 1498 .pru_disconnect = tcp_usr_disconnect, 1499 .pru_listen = tcp6_usr_listen, 1500 .pru_peeraddr = in6_mapped_peeraddr, 1501 .pru_rcvd = tcp_usr_rcvd, 1502 .pru_rcvoob = tcp_usr_rcvoob, 1503 .pru_send = tcp_usr_send, 1504 .pru_ready = tcp_usr_ready, 1505 .pru_shutdown = tcp_usr_shutdown, 1506 .pru_sockaddr = in6_mapped_sockaddr, 1507 .pru_sosetlabel = in_pcbsosetlabel, 1508 .pru_close = tcp_usr_close, 1509 }; 1510 #endif /* INET6 */ 1511 1512 #ifdef INET 1513 /* 1514 * Common subroutine to open a TCP connection to remote host specified 1515 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local 1516 * port number if needed. Call in_pcbconnect_setup to do the routing and 1517 * to choose a local host address (interface). If there is an existing 1518 * incarnation of the same connection in TIME-WAIT state and if the remote 1519 * host was sending CC options and if the connection duration was < MSL, then 1520 * truncate the previous TIME-WAIT state and proceed. 1521 * Initialize connection parameters and enter SYN-SENT state. 1522 */ 1523 static int 1524 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td) 1525 { 1526 struct inpcb *inp = tp->t_inpcb, *oinp; 1527 struct socket *so = inp->inp_socket; 1528 struct in_addr laddr; 1529 u_short lport; 1530 int error; 1531 1532 NET_EPOCH_ASSERT(); 1533 INP_WLOCK_ASSERT(inp); 1534 INP_HASH_WLOCK(&V_tcbinfo); 1535 1536 if (V_tcp_require_unique_port && inp->inp_lport == 0) { 1537 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); 1538 if (error) 1539 goto out; 1540 } 1541 1542 /* 1543 * Cannot simply call in_pcbconnect, because there might be an 1544 * earlier incarnation of this same connection still in 1545 * TIME_WAIT state, creating an ADDRINUSE error. 1546 */ 1547 laddr = inp->inp_laddr; 1548 lport = inp->inp_lport; 1549 error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport, 1550 &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred); 1551 if (error && oinp == NULL) 1552 goto out; 1553 if (oinp) { 1554 error = EADDRINUSE; 1555 goto out; 1556 } 1557 /* Handle initial bind if it hadn't been done in advance. */ 1558 if (inp->inp_lport == 0) { 1559 inp->inp_lport = lport; 1560 if (in_pcbinshash(inp) != 0) { 1561 inp->inp_lport = 0; 1562 error = EAGAIN; 1563 goto out; 1564 } 1565 } 1566 inp->inp_laddr = laddr; 1567 in_pcbrehash(inp); 1568 INP_HASH_WUNLOCK(&V_tcbinfo); 1569 1570 /* 1571 * Compute window scaling to request: 1572 * Scale to fit into sweet spot. See tcp_syncache.c. 1573 * XXX: This should move to tcp_output(). 1574 */ 1575 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 1576 (TCP_MAXWIN << tp->request_r_scale) < sb_max) 1577 tp->request_r_scale++; 1578 1579 soisconnecting(so); 1580 TCPSTAT_INC(tcps_connattempt); 1581 tcp_state_change(tp, TCPS_SYN_SENT); 1582 tp->iss = tcp_new_isn(&inp->inp_inc); 1583 if (tp->t_flags & TF_REQ_TSTMP) 1584 tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc); 1585 tcp_sendseqinit(tp); 1586 1587 return 0; 1588 1589 out: 1590 INP_HASH_WUNLOCK(&V_tcbinfo); 1591 return (error); 1592 } 1593 #endif /* INET */ 1594 1595 #ifdef INET6 1596 static int 1597 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td) 1598 { 1599 struct inpcb *inp = tp->t_inpcb; 1600 int error; 1601 1602 INP_WLOCK_ASSERT(inp); 1603 INP_HASH_WLOCK(&V_tcbinfo); 1604 1605 if (V_tcp_require_unique_port && inp->inp_lport == 0) { 1606 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); 1607 if (error) 1608 goto out; 1609 } 1610 error = in6_pcbconnect(inp, nam, td->td_ucred); 1611 if (error != 0) 1612 goto out; 1613 INP_HASH_WUNLOCK(&V_tcbinfo); 1614 1615 /* Compute window scaling to request. */ 1616 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 1617 (TCP_MAXWIN << tp->request_r_scale) < sb_max) 1618 tp->request_r_scale++; 1619 1620 soisconnecting(inp->inp_socket); 1621 TCPSTAT_INC(tcps_connattempt); 1622 tcp_state_change(tp, TCPS_SYN_SENT); 1623 tp->iss = tcp_new_isn(&inp->inp_inc); 1624 if (tp->t_flags & TF_REQ_TSTMP) 1625 tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc); 1626 tcp_sendseqinit(tp); 1627 1628 return 0; 1629 1630 out: 1631 INP_HASH_WUNLOCK(&V_tcbinfo); 1632 return error; 1633 } 1634 #endif /* INET6 */ 1635 1636 /* 1637 * Export TCP internal state information via a struct tcp_info, based on the 1638 * Linux 2.6 API. Not ABI compatible as our constants are mapped differently 1639 * (TCP state machine, etc). We export all information using FreeBSD-native 1640 * constants -- for example, the numeric values for tcpi_state will differ 1641 * from Linux. 1642 */ 1643 static void 1644 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti) 1645 { 1646 1647 INP_WLOCK_ASSERT(tp->t_inpcb); 1648 bzero(ti, sizeof(*ti)); 1649 1650 ti->tcpi_state = tp->t_state; 1651 if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP)) 1652 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS; 1653 if (tp->t_flags & TF_SACK_PERMIT) 1654 ti->tcpi_options |= TCPI_OPT_SACK; 1655 if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) { 1656 ti->tcpi_options |= TCPI_OPT_WSCALE; 1657 ti->tcpi_snd_wscale = tp->snd_scale; 1658 ti->tcpi_rcv_wscale = tp->rcv_scale; 1659 } 1660 if (tp->t_flags2 & (TF2_ECN_PERMIT | TF2_ACE_PERMIT)) 1661 ti->tcpi_options |= TCPI_OPT_ECN; 1662 1663 ti->tcpi_rto = tp->t_rxtcur * tick; 1664 ti->tcpi_last_data_recv = ((uint32_t)ticks - tp->t_rcvtime) * tick; 1665 ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT; 1666 ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT; 1667 1668 ti->tcpi_snd_ssthresh = tp->snd_ssthresh; 1669 ti->tcpi_snd_cwnd = tp->snd_cwnd; 1670 1671 /* 1672 * FreeBSD-specific extension fields for tcp_info. 1673 */ 1674 ti->tcpi_rcv_space = tp->rcv_wnd; 1675 ti->tcpi_rcv_nxt = tp->rcv_nxt; 1676 ti->tcpi_snd_wnd = tp->snd_wnd; 1677 ti->tcpi_snd_bwnd = 0; /* Unused, kept for compat. */ 1678 ti->tcpi_snd_nxt = tp->snd_nxt; 1679 ti->tcpi_snd_mss = tp->t_maxseg; 1680 ti->tcpi_rcv_mss = tp->t_maxseg; 1681 ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack; 1682 ti->tcpi_rcv_ooopack = tp->t_rcvoopack; 1683 ti->tcpi_snd_zerowin = tp->t_sndzerowin; 1684 #ifdef TCP_OFFLOAD 1685 if (tp->t_flags & TF_TOE) { 1686 ti->tcpi_options |= TCPI_OPT_TOE; 1687 tcp_offload_tcp_info(tp, ti); 1688 } 1689 #endif 1690 } 1691 1692 /* 1693 * tcp_ctloutput() must drop the inpcb lock before performing copyin on 1694 * socket option arguments. When it re-acquires the lock after the copy, it 1695 * has to revalidate that the connection is still valid for the socket 1696 * option. 1697 */ 1698 #define INP_WLOCK_RECHECK_CLEANUP(inp, cleanup) do { \ 1699 INP_WLOCK(inp); \ 1700 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \ 1701 INP_WUNLOCK(inp); \ 1702 cleanup; \ 1703 return (ECONNRESET); \ 1704 } \ 1705 tp = intotcpcb(inp); \ 1706 } while(0) 1707 #define INP_WLOCK_RECHECK(inp) INP_WLOCK_RECHECK_CLEANUP((inp), /* noop */) 1708 1709 int 1710 tcp_ctloutput_set(struct inpcb *inp, struct sockopt *sopt) 1711 { 1712 struct socket *so = inp->inp_socket; 1713 struct tcpcb *tp = intotcpcb(inp); 1714 int error = 0; 1715 1716 MPASS(sopt->sopt_dir == SOPT_SET); 1717 INP_WLOCK_ASSERT(inp); 1718 KASSERT((inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) == 0, 1719 ("inp_flags == %x", inp->inp_flags)); 1720 KASSERT(so != NULL, ("inp_socket == NULL")); 1721 1722 if (sopt->sopt_level != IPPROTO_TCP) { 1723 INP_WUNLOCK(inp); 1724 #ifdef INET6 1725 if (inp->inp_vflag & INP_IPV6PROTO) 1726 error = ip6_ctloutput(so, sopt); 1727 #endif 1728 #if defined(INET6) && defined(INET) 1729 else 1730 #endif 1731 #ifdef INET 1732 error = ip_ctloutput(so, sopt); 1733 #endif 1734 /* 1735 * When an IP-level socket option affects TCP, pass control 1736 * down to stack tfb_tcp_ctloutput, otherwise return what 1737 * IP level returned. 1738 */ 1739 switch (sopt->sopt_level) { 1740 #ifdef INET6 1741 case IPPROTO_IPV6: 1742 if ((inp->inp_vflag & INP_IPV6PROTO) == 0) 1743 return (error); 1744 switch (sopt->sopt_name) { 1745 case IPV6_TCLASS: 1746 /* Notify tcp stacks that care (e.g. RACK). */ 1747 break; 1748 case IPV6_USE_MIN_MTU: 1749 /* Update t_maxseg accordingly. */ 1750 break; 1751 default: 1752 return (error); 1753 } 1754 break; 1755 #endif 1756 #ifdef INET 1757 case IPPROTO_IP: 1758 switch (sopt->sopt_name) { 1759 case IP_TOS: 1760 inp->inp_ip_tos &= ~IPTOS_ECN_MASK; 1761 break; 1762 case IP_TTL: 1763 /* Notify tcp stacks that care (e.g. RACK). */ 1764 break; 1765 default: 1766 return (error); 1767 } 1768 break; 1769 #endif 1770 default: 1771 return (error); 1772 } 1773 INP_WLOCK(inp); 1774 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 1775 INP_WUNLOCK(inp); 1776 return (ECONNRESET); 1777 } 1778 } else if (sopt->sopt_name == TCP_FUNCTION_BLK) { 1779 /* 1780 * Protect the TCP option TCP_FUNCTION_BLK so 1781 * that a sub-function can *never* overwrite this. 1782 */ 1783 struct tcp_function_set fsn; 1784 struct tcp_function_block *blk; 1785 1786 INP_WUNLOCK(inp); 1787 error = sooptcopyin(sopt, &fsn, sizeof fsn, sizeof fsn); 1788 if (error) 1789 return (error); 1790 1791 INP_WLOCK(inp); 1792 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 1793 INP_WUNLOCK(inp); 1794 return (ECONNRESET); 1795 } 1796 tp = intotcpcb(inp); 1797 1798 blk = find_and_ref_tcp_functions(&fsn); 1799 if (blk == NULL) { 1800 INP_WUNLOCK(inp); 1801 return (ENOENT); 1802 } 1803 if (tp->t_fb == blk) { 1804 /* You already have this */ 1805 refcount_release(&blk->tfb_refcnt); 1806 INP_WUNLOCK(inp); 1807 return (0); 1808 } 1809 if (tp->t_state != TCPS_CLOSED) { 1810 /* 1811 * The user has advanced the state 1812 * past the initial point, we may not 1813 * be able to switch. 1814 */ 1815 if (blk->tfb_tcp_handoff_ok != NULL) { 1816 /* 1817 * Does the stack provide a 1818 * query mechanism, if so it may 1819 * still be possible? 1820 */ 1821 error = (*blk->tfb_tcp_handoff_ok)(tp); 1822 } else 1823 error = EINVAL; 1824 if (error) { 1825 refcount_release(&blk->tfb_refcnt); 1826 INP_WUNLOCK(inp); 1827 return(error); 1828 } 1829 } 1830 if (blk->tfb_flags & TCP_FUNC_BEING_REMOVED) { 1831 refcount_release(&blk->tfb_refcnt); 1832 INP_WUNLOCK(inp); 1833 return (ENOENT); 1834 } 1835 /* 1836 * Release the old refcnt, the 1837 * lookup acquired a ref on the 1838 * new one already. 1839 */ 1840 if (tp->t_fb->tfb_tcp_fb_fini) { 1841 struct epoch_tracker et; 1842 /* 1843 * Tell the stack to cleanup with 0 i.e. 1844 * the tcb is not going away. 1845 */ 1846 NET_EPOCH_ENTER(et); 1847 (*tp->t_fb->tfb_tcp_fb_fini)(tp, 0); 1848 NET_EPOCH_EXIT(et); 1849 } 1850 #ifdef TCPHPTS 1851 /* Assure that we are not on any hpts */ 1852 tcp_hpts_remove(tp->t_inpcb); 1853 #endif 1854 if (blk->tfb_tcp_fb_init) { 1855 error = (*blk->tfb_tcp_fb_init)(tp); 1856 if (error) { 1857 refcount_release(&blk->tfb_refcnt); 1858 if (tp->t_fb->tfb_tcp_fb_init) { 1859 if((*tp->t_fb->tfb_tcp_fb_init)(tp) != 0) { 1860 /* Fall back failed, drop the connection */ 1861 INP_WUNLOCK(inp); 1862 soabort(so); 1863 return (error); 1864 } 1865 } 1866 goto err_out; 1867 } 1868 } 1869 refcount_release(&tp->t_fb->tfb_refcnt); 1870 tp->t_fb = blk; 1871 #ifdef TCP_OFFLOAD 1872 if (tp->t_flags & TF_TOE) { 1873 tcp_offload_ctloutput(tp, sopt->sopt_dir, 1874 sopt->sopt_name); 1875 } 1876 #endif 1877 err_out: 1878 INP_WUNLOCK(inp); 1879 return (error); 1880 } 1881 1882 /* Pass in the INP locked, callee must unlock it. */ 1883 return (tp->t_fb->tfb_tcp_ctloutput(inp, sopt)); 1884 } 1885 1886 static int 1887 tcp_ctloutput_get(struct inpcb *inp, struct sockopt *sopt) 1888 { 1889 struct socket *so = inp->inp_socket; 1890 struct tcpcb *tp = intotcpcb(inp); 1891 int error = 0; 1892 1893 MPASS(sopt->sopt_dir == SOPT_GET); 1894 INP_WLOCK_ASSERT(inp); 1895 KASSERT((inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) == 0, 1896 ("inp_flags == %x", inp->inp_flags)); 1897 KASSERT(so != NULL, ("inp_socket == NULL")); 1898 1899 if (sopt->sopt_level != IPPROTO_TCP) { 1900 INP_WUNLOCK(inp); 1901 #ifdef INET6 1902 if (inp->inp_vflag & INP_IPV6PROTO) 1903 error = ip6_ctloutput(so, sopt); 1904 #endif /* INET6 */ 1905 #if defined(INET6) && defined(INET) 1906 else 1907 #endif 1908 #ifdef INET 1909 error = ip_ctloutput(so, sopt); 1910 #endif 1911 return (error); 1912 } 1913 if (((sopt->sopt_name == TCP_FUNCTION_BLK) || 1914 (sopt->sopt_name == TCP_FUNCTION_ALIAS))) { 1915 struct tcp_function_set fsn; 1916 1917 if (sopt->sopt_name == TCP_FUNCTION_ALIAS) { 1918 memset(&fsn, 0, sizeof(fsn)); 1919 find_tcp_function_alias(tp->t_fb, &fsn); 1920 } else { 1921 strncpy(fsn.function_set_name, 1922 tp->t_fb->tfb_tcp_block_name, 1923 TCP_FUNCTION_NAME_LEN_MAX); 1924 fsn.function_set_name[TCP_FUNCTION_NAME_LEN_MAX - 1] = '\0'; 1925 } 1926 fsn.pcbcnt = tp->t_fb->tfb_refcnt; 1927 INP_WUNLOCK(inp); 1928 error = sooptcopyout(sopt, &fsn, sizeof fsn); 1929 return (error); 1930 } 1931 1932 /* Pass in the INP locked, callee must unlock it. */ 1933 return (tp->t_fb->tfb_tcp_ctloutput(inp, sopt)); 1934 } 1935 1936 int 1937 tcp_ctloutput(struct socket *so, struct sockopt *sopt) 1938 { 1939 struct inpcb *inp; 1940 1941 inp = sotoinpcb(so); 1942 KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL")); 1943 1944 INP_WLOCK(inp); 1945 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 1946 INP_WUNLOCK(inp); 1947 return (ECONNRESET); 1948 } 1949 if (sopt->sopt_dir == SOPT_SET) 1950 return (tcp_ctloutput_set(inp, sopt)); 1951 else if (sopt->sopt_dir == SOPT_GET) 1952 return (tcp_ctloutput_get(inp, sopt)); 1953 else 1954 panic("%s: sopt_dir $%d", __func__, sopt->sopt_dir); 1955 } 1956 1957 /* 1958 * If this assert becomes untrue, we need to change the size of the buf 1959 * variable in tcp_default_ctloutput(). 1960 */ 1961 #ifdef CTASSERT 1962 CTASSERT(TCP_CA_NAME_MAX <= TCP_LOG_ID_LEN); 1963 CTASSERT(TCP_LOG_REASON_LEN <= TCP_LOG_ID_LEN); 1964 #endif 1965 1966 #ifdef KERN_TLS 1967 static int 1968 copyin_tls_enable(struct sockopt *sopt, struct tls_enable *tls) 1969 { 1970 struct tls_enable_v0 tls_v0; 1971 int error; 1972 1973 if (sopt->sopt_valsize == sizeof(tls_v0)) { 1974 error = sooptcopyin(sopt, &tls_v0, sizeof(tls_v0), 1975 sizeof(tls_v0)); 1976 if (error) 1977 return (error); 1978 memset(tls, 0, sizeof(*tls)); 1979 tls->cipher_key = tls_v0.cipher_key; 1980 tls->iv = tls_v0.iv; 1981 tls->auth_key = tls_v0.auth_key; 1982 tls->cipher_algorithm = tls_v0.cipher_algorithm; 1983 tls->cipher_key_len = tls_v0.cipher_key_len; 1984 tls->iv_len = tls_v0.iv_len; 1985 tls->auth_algorithm = tls_v0.auth_algorithm; 1986 tls->auth_key_len = tls_v0.auth_key_len; 1987 tls->flags = tls_v0.flags; 1988 tls->tls_vmajor = tls_v0.tls_vmajor; 1989 tls->tls_vminor = tls_v0.tls_vminor; 1990 return (0); 1991 } 1992 1993 return (sooptcopyin(sopt, tls, sizeof(*tls), sizeof(*tls))); 1994 } 1995 #endif 1996 1997 extern struct cc_algo newreno_cc_algo; 1998 1999 static int 2000 tcp_set_cc_mod(struct inpcb *inp, struct sockopt *sopt) 2001 { 2002 struct cc_algo *algo; 2003 void *ptr = NULL; 2004 struct tcpcb *tp; 2005 struct cc_var cc_mem; 2006 char buf[TCP_CA_NAME_MAX]; 2007 size_t mem_sz; 2008 int error; 2009 2010 INP_WUNLOCK(inp); 2011 error = sooptcopyin(sopt, buf, TCP_CA_NAME_MAX - 1, 1); 2012 if (error) 2013 return(error); 2014 buf[sopt->sopt_valsize] = '\0'; 2015 CC_LIST_RLOCK(); 2016 STAILQ_FOREACH(algo, &cc_list, entries) { 2017 if (strncmp(buf, algo->name, 2018 TCP_CA_NAME_MAX) == 0) { 2019 if (algo->flags & CC_MODULE_BEING_REMOVED) { 2020 /* We can't "see" modules being unloaded */ 2021 continue; 2022 } 2023 break; 2024 } 2025 } 2026 if (algo == NULL) { 2027 CC_LIST_RUNLOCK(); 2028 return(ESRCH); 2029 } 2030 /* 2031 * With a reference the algorithm cannot be removed 2032 * so we hold a reference through the change process. 2033 */ 2034 cc_refer(algo); 2035 CC_LIST_RUNLOCK(); 2036 if (algo->cb_init != NULL) { 2037 /* We can now pre-get the memory for the CC */ 2038 mem_sz = (*algo->cc_data_sz)(); 2039 if (mem_sz == 0) { 2040 goto no_mem_needed; 2041 } 2042 ptr = malloc(mem_sz, M_CC_MEM, M_WAITOK); 2043 } else { 2044 no_mem_needed: 2045 mem_sz = 0; 2046 ptr = NULL; 2047 } 2048 /* 2049 * Make sure its all clean and zero and also get 2050 * back the inplock. 2051 */ 2052 memset(&cc_mem, 0, sizeof(cc_mem)); 2053 INP_WLOCK(inp); 2054 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { 2055 INP_WUNLOCK(inp); 2056 if (ptr) 2057 free(ptr, M_CC_MEM); 2058 /* Release our temp reference */ 2059 CC_LIST_RLOCK(); 2060 cc_release(algo); 2061 CC_LIST_RUNLOCK(); 2062 return (ECONNRESET); 2063 } 2064 tp = intotcpcb(inp); 2065 if (ptr != NULL) 2066 memset(ptr, 0, mem_sz); 2067 cc_mem.ccvc.tcp = tp; 2068 /* 2069 * We once again hold a write lock over the tcb so it's 2070 * safe to do these things without ordering concerns. 2071 * Note here we init into stack memory. 2072 */ 2073 if (algo->cb_init != NULL) 2074 error = algo->cb_init(&cc_mem, ptr); 2075 else 2076 error = 0; 2077 /* 2078 * The CC algorithms, when given their memory 2079 * should not fail we could in theory have a 2080 * KASSERT here. 2081 */ 2082 if (error == 0) { 2083 /* 2084 * Touchdown, lets go ahead and move the 2085 * connection to the new CC module by 2086 * copying in the cc_mem after we call 2087 * the old ones cleanup (if any). 2088 */ 2089 if (CC_ALGO(tp)->cb_destroy != NULL) 2090 CC_ALGO(tp)->cb_destroy(tp->ccv); 2091 /* Detach the old CC from the tcpcb */ 2092 cc_detach(tp); 2093 /* Copy in our temp memory that was inited */ 2094 memcpy(tp->ccv, &cc_mem, sizeof(struct cc_var)); 2095 /* Now attach the new, which takes a reference */ 2096 cc_attach(tp, algo); 2097 /* Ok now are we where we have gotten past any conn_init? */ 2098 if (TCPS_HAVEESTABLISHED(tp->t_state) && (CC_ALGO(tp)->conn_init != NULL)) { 2099 /* Yep run the connection init for the new CC */ 2100 CC_ALGO(tp)->conn_init(tp->ccv); 2101 } 2102 } else if (ptr) 2103 free(ptr, M_CC_MEM); 2104 INP_WUNLOCK(inp); 2105 /* Now lets release our temp reference */ 2106 CC_LIST_RLOCK(); 2107 cc_release(algo); 2108 CC_LIST_RUNLOCK(); 2109 return (error); 2110 } 2111 2112 int 2113 tcp_default_ctloutput(struct inpcb *inp, struct sockopt *sopt) 2114 { 2115 struct tcpcb *tp = intotcpcb(inp); 2116 int error, opt, optval; 2117 u_int ui; 2118 struct tcp_info ti; 2119 #ifdef KERN_TLS 2120 struct tls_enable tls; 2121 struct socket *so = inp->inp_socket; 2122 #endif 2123 char *pbuf, buf[TCP_LOG_ID_LEN]; 2124 #ifdef STATS 2125 struct statsblob *sbp; 2126 #endif 2127 size_t len; 2128 2129 INP_WLOCK_ASSERT(inp); 2130 KASSERT((inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) == 0, 2131 ("inp_flags == %x", inp->inp_flags)); 2132 KASSERT(inp->inp_socket != NULL, ("inp_socket == NULL")); 2133 2134 switch (sopt->sopt_level) { 2135 #ifdef INET6 2136 case IPPROTO_IPV6: 2137 MPASS(inp->inp_vflag & INP_IPV6PROTO); 2138 switch (sopt->sopt_name) { 2139 case IPV6_USE_MIN_MTU: 2140 tcp6_use_min_mtu(tp); 2141 /* FALLTHROUGH */ 2142 } 2143 INP_WUNLOCK(inp); 2144 return (0); 2145 #endif 2146 #ifdef INET 2147 case IPPROTO_IP: 2148 INP_WUNLOCK(inp); 2149 return (0); 2150 #endif 2151 } 2152 2153 /* 2154 * For TCP_CCALGOOPT forward the control to CC module, for both 2155 * SOPT_SET and SOPT_GET. 2156 */ 2157 switch (sopt->sopt_name) { 2158 case TCP_CCALGOOPT: 2159 INP_WUNLOCK(inp); 2160 if (sopt->sopt_valsize > CC_ALGOOPT_LIMIT) 2161 return (EINVAL); 2162 pbuf = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK | M_ZERO); 2163 error = sooptcopyin(sopt, pbuf, sopt->sopt_valsize, 2164 sopt->sopt_valsize); 2165 if (error) { 2166 free(pbuf, M_TEMP); 2167 return (error); 2168 } 2169 INP_WLOCK_RECHECK_CLEANUP(inp, free(pbuf, M_TEMP)); 2170 if (CC_ALGO(tp)->ctl_output != NULL) 2171 error = CC_ALGO(tp)->ctl_output(tp->ccv, sopt, pbuf); 2172 else 2173 error = ENOENT; 2174 INP_WUNLOCK(inp); 2175 if (error == 0 && sopt->sopt_dir == SOPT_GET) 2176 error = sooptcopyout(sopt, pbuf, sopt->sopt_valsize); 2177 free(pbuf, M_TEMP); 2178 return (error); 2179 } 2180 2181 switch (sopt->sopt_dir) { 2182 case SOPT_SET: 2183 switch (sopt->sopt_name) { 2184 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE) 2185 case TCP_MD5SIG: 2186 INP_WUNLOCK(inp); 2187 if (!TCPMD5_ENABLED()) 2188 return (ENOPROTOOPT); 2189 error = TCPMD5_PCBCTL(inp, sopt); 2190 if (error) 2191 return (error); 2192 INP_WLOCK_RECHECK(inp); 2193 goto unlock_and_done; 2194 #endif /* IPSEC */ 2195 2196 case TCP_NODELAY: 2197 case TCP_NOOPT: 2198 case TCP_LRD: 2199 INP_WUNLOCK(inp); 2200 error = sooptcopyin(sopt, &optval, sizeof optval, 2201 sizeof optval); 2202 if (error) 2203 return (error); 2204 2205 INP_WLOCK_RECHECK(inp); 2206 switch (sopt->sopt_name) { 2207 case TCP_NODELAY: 2208 opt = TF_NODELAY; 2209 break; 2210 case TCP_NOOPT: 2211 opt = TF_NOOPT; 2212 break; 2213 case TCP_LRD: 2214 opt = TF_LRD; 2215 break; 2216 default: 2217 opt = 0; /* dead code to fool gcc */ 2218 break; 2219 } 2220 2221 if (optval) 2222 tp->t_flags |= opt; 2223 else 2224 tp->t_flags &= ~opt; 2225 unlock_and_done: 2226 #ifdef TCP_OFFLOAD 2227 if (tp->t_flags & TF_TOE) { 2228 tcp_offload_ctloutput(tp, sopt->sopt_dir, 2229 sopt->sopt_name); 2230 } 2231 #endif 2232 INP_WUNLOCK(inp); 2233 break; 2234 2235 case TCP_NOPUSH: 2236 INP_WUNLOCK(inp); 2237 error = sooptcopyin(sopt, &optval, sizeof optval, 2238 sizeof optval); 2239 if (error) 2240 return (error); 2241 2242 INP_WLOCK_RECHECK(inp); 2243 if (optval) 2244 tp->t_flags |= TF_NOPUSH; 2245 else if (tp->t_flags & TF_NOPUSH) { 2246 tp->t_flags &= ~TF_NOPUSH; 2247 if (TCPS_HAVEESTABLISHED(tp->t_state)) { 2248 struct epoch_tracker et; 2249 2250 NET_EPOCH_ENTER(et); 2251 error = tcp_output_nodrop(tp); 2252 NET_EPOCH_EXIT(et); 2253 } 2254 } 2255 goto unlock_and_done; 2256 2257 case TCP_REMOTE_UDP_ENCAPS_PORT: 2258 INP_WUNLOCK(inp); 2259 error = sooptcopyin(sopt, &optval, sizeof optval, 2260 sizeof optval); 2261 if (error) 2262 return (error); 2263 if ((optval < TCP_TUNNELING_PORT_MIN) || 2264 (optval > TCP_TUNNELING_PORT_MAX)) { 2265 /* Its got to be in range */ 2266 return (EINVAL); 2267 } 2268 if ((V_tcp_udp_tunneling_port == 0) && (optval != 0)) { 2269 /* You have to have enabled a UDP tunneling port first */ 2270 return (EINVAL); 2271 } 2272 INP_WLOCK_RECHECK(inp); 2273 if (tp->t_state != TCPS_CLOSED) { 2274 /* You can't change after you are connected */ 2275 error = EINVAL; 2276 } else { 2277 /* Ok we are all good set the port */ 2278 tp->t_port = htons(optval); 2279 } 2280 goto unlock_and_done; 2281 2282 case TCP_MAXSEG: 2283 INP_WUNLOCK(inp); 2284 error = sooptcopyin(sopt, &optval, sizeof optval, 2285 sizeof optval); 2286 if (error) 2287 return (error); 2288 2289 INP_WLOCK_RECHECK(inp); 2290 if (optval > 0 && optval <= tp->t_maxseg && 2291 optval + 40 >= V_tcp_minmss) 2292 tp->t_maxseg = optval; 2293 else 2294 error = EINVAL; 2295 goto unlock_and_done; 2296 2297 case TCP_INFO: 2298 INP_WUNLOCK(inp); 2299 error = EINVAL; 2300 break; 2301 2302 case TCP_STATS: 2303 INP_WUNLOCK(inp); 2304 #ifdef STATS 2305 error = sooptcopyin(sopt, &optval, sizeof optval, 2306 sizeof optval); 2307 if (error) 2308 return (error); 2309 2310 if (optval > 0) 2311 sbp = stats_blob_alloc( 2312 V_tcp_perconn_stats_dflt_tpl, 0); 2313 else 2314 sbp = NULL; 2315 2316 INP_WLOCK_RECHECK(inp); 2317 if ((tp->t_stats != NULL && sbp == NULL) || 2318 (tp->t_stats == NULL && sbp != NULL)) { 2319 struct statsblob *t = tp->t_stats; 2320 tp->t_stats = sbp; 2321 sbp = t; 2322 } 2323 INP_WUNLOCK(inp); 2324 2325 stats_blob_destroy(sbp); 2326 #else 2327 return (EOPNOTSUPP); 2328 #endif /* !STATS */ 2329 break; 2330 2331 case TCP_CONGESTION: 2332 error = tcp_set_cc_mod(inp, sopt); 2333 break; 2334 2335 case TCP_REUSPORT_LB_NUMA: 2336 INP_WUNLOCK(inp); 2337 error = sooptcopyin(sopt, &optval, sizeof(optval), 2338 sizeof(optval)); 2339 INP_WLOCK_RECHECK(inp); 2340 if (!error) 2341 error = in_pcblbgroup_numa(inp, optval); 2342 INP_WUNLOCK(inp); 2343 break; 2344 2345 #ifdef KERN_TLS 2346 case TCP_TXTLS_ENABLE: 2347 INP_WUNLOCK(inp); 2348 error = copyin_tls_enable(sopt, &tls); 2349 if (error) 2350 break; 2351 error = ktls_enable_tx(so, &tls); 2352 break; 2353 case TCP_TXTLS_MODE: 2354 INP_WUNLOCK(inp); 2355 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui)); 2356 if (error) 2357 return (error); 2358 2359 INP_WLOCK_RECHECK(inp); 2360 error = ktls_set_tx_mode(so, ui); 2361 INP_WUNLOCK(inp); 2362 break; 2363 case TCP_RXTLS_ENABLE: 2364 INP_WUNLOCK(inp); 2365 error = sooptcopyin(sopt, &tls, sizeof(tls), 2366 sizeof(tls)); 2367 if (error) 2368 break; 2369 error = ktls_enable_rx(so, &tls); 2370 break; 2371 #endif 2372 2373 case TCP_KEEPIDLE: 2374 case TCP_KEEPINTVL: 2375 case TCP_KEEPINIT: 2376 INP_WUNLOCK(inp); 2377 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui)); 2378 if (error) 2379 return (error); 2380 2381 if (ui > (UINT_MAX / hz)) { 2382 error = EINVAL; 2383 break; 2384 } 2385 ui *= hz; 2386 2387 INP_WLOCK_RECHECK(inp); 2388 switch (sopt->sopt_name) { 2389 case TCP_KEEPIDLE: 2390 tp->t_keepidle = ui; 2391 /* 2392 * XXX: better check current remaining 2393 * timeout and "merge" it with new value. 2394 */ 2395 if ((tp->t_state > TCPS_LISTEN) && 2396 (tp->t_state <= TCPS_CLOSING)) 2397 tcp_timer_activate(tp, TT_KEEP, 2398 TP_KEEPIDLE(tp)); 2399 break; 2400 case TCP_KEEPINTVL: 2401 tp->t_keepintvl = ui; 2402 if ((tp->t_state == TCPS_FIN_WAIT_2) && 2403 (TP_MAXIDLE(tp) > 0)) 2404 tcp_timer_activate(tp, TT_2MSL, 2405 TP_MAXIDLE(tp)); 2406 break; 2407 case TCP_KEEPINIT: 2408 tp->t_keepinit = ui; 2409 if (tp->t_state == TCPS_SYN_RECEIVED || 2410 tp->t_state == TCPS_SYN_SENT) 2411 tcp_timer_activate(tp, TT_KEEP, 2412 TP_KEEPINIT(tp)); 2413 break; 2414 } 2415 goto unlock_and_done; 2416 2417 case TCP_KEEPCNT: 2418 INP_WUNLOCK(inp); 2419 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui)); 2420 if (error) 2421 return (error); 2422 2423 INP_WLOCK_RECHECK(inp); 2424 tp->t_keepcnt = ui; 2425 if ((tp->t_state == TCPS_FIN_WAIT_2) && 2426 (TP_MAXIDLE(tp) > 0)) 2427 tcp_timer_activate(tp, TT_2MSL, 2428 TP_MAXIDLE(tp)); 2429 goto unlock_and_done; 2430 2431 #ifdef TCPPCAP 2432 case TCP_PCAP_OUT: 2433 case TCP_PCAP_IN: 2434 INP_WUNLOCK(inp); 2435 error = sooptcopyin(sopt, &optval, sizeof optval, 2436 sizeof optval); 2437 if (error) 2438 return (error); 2439 2440 INP_WLOCK_RECHECK(inp); 2441 if (optval >= 0) 2442 tcp_pcap_set_sock_max(TCP_PCAP_OUT ? 2443 &(tp->t_outpkts) : &(tp->t_inpkts), 2444 optval); 2445 else 2446 error = EINVAL; 2447 goto unlock_and_done; 2448 #endif 2449 2450 case TCP_FASTOPEN: { 2451 struct tcp_fastopen tfo_optval; 2452 2453 INP_WUNLOCK(inp); 2454 if (!V_tcp_fastopen_client_enable && 2455 !V_tcp_fastopen_server_enable) 2456 return (EPERM); 2457 2458 error = sooptcopyin(sopt, &tfo_optval, 2459 sizeof(tfo_optval), sizeof(int)); 2460 if (error) 2461 return (error); 2462 2463 INP_WLOCK_RECHECK(inp); 2464 if ((tp->t_state != TCPS_CLOSED) && 2465 (tp->t_state != TCPS_LISTEN)) { 2466 error = EINVAL; 2467 goto unlock_and_done; 2468 } 2469 if (tfo_optval.enable) { 2470 if (tp->t_state == TCPS_LISTEN) { 2471 if (!V_tcp_fastopen_server_enable) { 2472 error = EPERM; 2473 goto unlock_and_done; 2474 } 2475 2476 if (tp->t_tfo_pending == NULL) 2477 tp->t_tfo_pending = 2478 tcp_fastopen_alloc_counter(); 2479 } else { 2480 /* 2481 * If a pre-shared key was provided, 2482 * stash it in the client cookie 2483 * field of the tcpcb for use during 2484 * connect. 2485 */ 2486 if (sopt->sopt_valsize == 2487 sizeof(tfo_optval)) { 2488 memcpy(tp->t_tfo_cookie.client, 2489 tfo_optval.psk, 2490 TCP_FASTOPEN_PSK_LEN); 2491 tp->t_tfo_client_cookie_len = 2492 TCP_FASTOPEN_PSK_LEN; 2493 } 2494 } 2495 tp->t_flags |= TF_FASTOPEN; 2496 } else 2497 tp->t_flags &= ~TF_FASTOPEN; 2498 goto unlock_and_done; 2499 } 2500 2501 #ifdef TCP_BLACKBOX 2502 case TCP_LOG: 2503 INP_WUNLOCK(inp); 2504 error = sooptcopyin(sopt, &optval, sizeof optval, 2505 sizeof optval); 2506 if (error) 2507 return (error); 2508 2509 INP_WLOCK_RECHECK(inp); 2510 error = tcp_log_state_change(tp, optval); 2511 goto unlock_and_done; 2512 2513 case TCP_LOGBUF: 2514 INP_WUNLOCK(inp); 2515 error = EINVAL; 2516 break; 2517 2518 case TCP_LOGID: 2519 INP_WUNLOCK(inp); 2520 error = sooptcopyin(sopt, buf, TCP_LOG_ID_LEN - 1, 0); 2521 if (error) 2522 break; 2523 buf[sopt->sopt_valsize] = '\0'; 2524 INP_WLOCK_RECHECK(inp); 2525 error = tcp_log_set_id(tp, buf); 2526 /* tcp_log_set_id() unlocks the INP. */ 2527 break; 2528 2529 case TCP_LOGDUMP: 2530 case TCP_LOGDUMPID: 2531 INP_WUNLOCK(inp); 2532 error = 2533 sooptcopyin(sopt, buf, TCP_LOG_REASON_LEN - 1, 0); 2534 if (error) 2535 break; 2536 buf[sopt->sopt_valsize] = '\0'; 2537 INP_WLOCK_RECHECK(inp); 2538 if (sopt->sopt_name == TCP_LOGDUMP) { 2539 error = tcp_log_dump_tp_logbuf(tp, buf, 2540 M_WAITOK, true); 2541 INP_WUNLOCK(inp); 2542 } else { 2543 tcp_log_dump_tp_bucket_logbufs(tp, buf); 2544 /* 2545 * tcp_log_dump_tp_bucket_logbufs() drops the 2546 * INP lock. 2547 */ 2548 } 2549 break; 2550 #endif 2551 2552 default: 2553 INP_WUNLOCK(inp); 2554 error = ENOPROTOOPT; 2555 break; 2556 } 2557 break; 2558 2559 case SOPT_GET: 2560 tp = intotcpcb(inp); 2561 switch (sopt->sopt_name) { 2562 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE) 2563 case TCP_MD5SIG: 2564 INP_WUNLOCK(inp); 2565 if (!TCPMD5_ENABLED()) 2566 return (ENOPROTOOPT); 2567 error = TCPMD5_PCBCTL(inp, sopt); 2568 break; 2569 #endif 2570 2571 case TCP_NODELAY: 2572 optval = tp->t_flags & TF_NODELAY; 2573 INP_WUNLOCK(inp); 2574 error = sooptcopyout(sopt, &optval, sizeof optval); 2575 break; 2576 case TCP_MAXSEG: 2577 optval = tp->t_maxseg; 2578 INP_WUNLOCK(inp); 2579 error = sooptcopyout(sopt, &optval, sizeof optval); 2580 break; 2581 case TCP_REMOTE_UDP_ENCAPS_PORT: 2582 optval = ntohs(tp->t_port); 2583 INP_WUNLOCK(inp); 2584 error = sooptcopyout(sopt, &optval, sizeof optval); 2585 break; 2586 case TCP_NOOPT: 2587 optval = tp->t_flags & TF_NOOPT; 2588 INP_WUNLOCK(inp); 2589 error = sooptcopyout(sopt, &optval, sizeof optval); 2590 break; 2591 case TCP_NOPUSH: 2592 optval = tp->t_flags & TF_NOPUSH; 2593 INP_WUNLOCK(inp); 2594 error = sooptcopyout(sopt, &optval, sizeof optval); 2595 break; 2596 case TCP_INFO: 2597 tcp_fill_info(tp, &ti); 2598 INP_WUNLOCK(inp); 2599 error = sooptcopyout(sopt, &ti, sizeof ti); 2600 break; 2601 case TCP_STATS: 2602 { 2603 #ifdef STATS 2604 int nheld; 2605 TYPEOF_MEMBER(struct statsblob, flags) sbflags = 0; 2606 2607 error = 0; 2608 socklen_t outsbsz = sopt->sopt_valsize; 2609 if (tp->t_stats == NULL) 2610 error = ENOENT; 2611 else if (outsbsz >= tp->t_stats->cursz) 2612 outsbsz = tp->t_stats->cursz; 2613 else if (outsbsz >= sizeof(struct statsblob)) 2614 outsbsz = sizeof(struct statsblob); 2615 else 2616 error = EINVAL; 2617 INP_WUNLOCK(inp); 2618 if (error) 2619 break; 2620 2621 sbp = sopt->sopt_val; 2622 nheld = atop(round_page(((vm_offset_t)sbp) + 2623 (vm_size_t)outsbsz) - trunc_page((vm_offset_t)sbp)); 2624 vm_page_t ma[nheld]; 2625 if (vm_fault_quick_hold_pages( 2626 &curproc->p_vmspace->vm_map, (vm_offset_t)sbp, 2627 outsbsz, VM_PROT_READ | VM_PROT_WRITE, ma, 2628 nheld) < 0) { 2629 error = EFAULT; 2630 break; 2631 } 2632 2633 if ((error = copyin_nofault(&(sbp->flags), &sbflags, 2634 SIZEOF_MEMBER(struct statsblob, flags)))) 2635 goto unhold; 2636 2637 INP_WLOCK_RECHECK(inp); 2638 error = stats_blob_snapshot(&sbp, outsbsz, tp->t_stats, 2639 sbflags | SB_CLONE_USRDSTNOFAULT); 2640 INP_WUNLOCK(inp); 2641 sopt->sopt_valsize = outsbsz; 2642 unhold: 2643 vm_page_unhold_pages(ma, nheld); 2644 #else 2645 INP_WUNLOCK(inp); 2646 error = EOPNOTSUPP; 2647 #endif /* !STATS */ 2648 break; 2649 } 2650 case TCP_CONGESTION: 2651 len = strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX); 2652 INP_WUNLOCK(inp); 2653 error = sooptcopyout(sopt, buf, len + 1); 2654 break; 2655 case TCP_KEEPIDLE: 2656 case TCP_KEEPINTVL: 2657 case TCP_KEEPINIT: 2658 case TCP_KEEPCNT: 2659 switch (sopt->sopt_name) { 2660 case TCP_KEEPIDLE: 2661 ui = TP_KEEPIDLE(tp) / hz; 2662 break; 2663 case TCP_KEEPINTVL: 2664 ui = TP_KEEPINTVL(tp) / hz; 2665 break; 2666 case TCP_KEEPINIT: 2667 ui = TP_KEEPINIT(tp) / hz; 2668 break; 2669 case TCP_KEEPCNT: 2670 ui = TP_KEEPCNT(tp); 2671 break; 2672 } 2673 INP_WUNLOCK(inp); 2674 error = sooptcopyout(sopt, &ui, sizeof(ui)); 2675 break; 2676 #ifdef TCPPCAP 2677 case TCP_PCAP_OUT: 2678 case TCP_PCAP_IN: 2679 optval = tcp_pcap_get_sock_max(TCP_PCAP_OUT ? 2680 &(tp->t_outpkts) : &(tp->t_inpkts)); 2681 INP_WUNLOCK(inp); 2682 error = sooptcopyout(sopt, &optval, sizeof optval); 2683 break; 2684 #endif 2685 case TCP_FASTOPEN: 2686 optval = tp->t_flags & TF_FASTOPEN; 2687 INP_WUNLOCK(inp); 2688 error = sooptcopyout(sopt, &optval, sizeof optval); 2689 break; 2690 #ifdef TCP_BLACKBOX 2691 case TCP_LOG: 2692 optval = tp->t_logstate; 2693 INP_WUNLOCK(inp); 2694 error = sooptcopyout(sopt, &optval, sizeof(optval)); 2695 break; 2696 case TCP_LOGBUF: 2697 /* tcp_log_getlogbuf() does INP_WUNLOCK(inp) */ 2698 error = tcp_log_getlogbuf(sopt, tp); 2699 break; 2700 case TCP_LOGID: 2701 len = tcp_log_get_id(tp, buf); 2702 INP_WUNLOCK(inp); 2703 error = sooptcopyout(sopt, buf, len + 1); 2704 break; 2705 case TCP_LOGDUMP: 2706 case TCP_LOGDUMPID: 2707 INP_WUNLOCK(inp); 2708 error = EINVAL; 2709 break; 2710 #endif 2711 #ifdef KERN_TLS 2712 case TCP_TXTLS_MODE: 2713 error = ktls_get_tx_mode(so, &optval); 2714 INP_WUNLOCK(inp); 2715 if (error == 0) 2716 error = sooptcopyout(sopt, &optval, 2717 sizeof(optval)); 2718 break; 2719 case TCP_RXTLS_MODE: 2720 error = ktls_get_rx_mode(so, &optval); 2721 INP_WUNLOCK(inp); 2722 if (error == 0) 2723 error = sooptcopyout(sopt, &optval, 2724 sizeof(optval)); 2725 break; 2726 #endif 2727 case TCP_LRD: 2728 optval = tp->t_flags & TF_LRD; 2729 INP_WUNLOCK(inp); 2730 error = sooptcopyout(sopt, &optval, sizeof optval); 2731 break; 2732 default: 2733 INP_WUNLOCK(inp); 2734 error = ENOPROTOOPT; 2735 break; 2736 } 2737 break; 2738 } 2739 return (error); 2740 } 2741 #undef INP_WLOCK_RECHECK 2742 #undef INP_WLOCK_RECHECK_CLEANUP 2743 2744 /* 2745 * Initiate (or continue) disconnect. 2746 * If embryonic state, just send reset (once). 2747 * If in ``let data drain'' option and linger null, just drop. 2748 * Otherwise (hard), mark socket disconnecting and drop 2749 * current input data; switch states based on user close, and 2750 * send segment to peer (with FIN). 2751 */ 2752 static void 2753 tcp_disconnect(struct tcpcb *tp) 2754 { 2755 struct inpcb *inp = tp->t_inpcb; 2756 struct socket *so = inp->inp_socket; 2757 2758 NET_EPOCH_ASSERT(); 2759 INP_WLOCK_ASSERT(inp); 2760 2761 /* 2762 * Neither tcp_close() nor tcp_drop() should return NULL, as the 2763 * socket is still open. 2764 */ 2765 if (tp->t_state < TCPS_ESTABLISHED && 2766 !(tp->t_state > TCPS_LISTEN && IS_FASTOPEN(tp->t_flags))) { 2767 tp = tcp_close(tp); 2768 KASSERT(tp != NULL, 2769 ("tcp_disconnect: tcp_close() returned NULL")); 2770 } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) { 2771 tp = tcp_drop(tp, 0); 2772 KASSERT(tp != NULL, 2773 ("tcp_disconnect: tcp_drop() returned NULL")); 2774 } else { 2775 soisdisconnecting(so); 2776 sbflush(&so->so_rcv); 2777 tcp_usrclosed(tp); 2778 if (!(inp->inp_flags & INP_DROPPED)) 2779 /* Ignore stack's drop request, we already at it. */ 2780 (void)tcp_output_nodrop(tp); 2781 } 2782 } 2783 2784 /* 2785 * User issued close, and wish to trail through shutdown states: 2786 * if never received SYN, just forget it. If got a SYN from peer, 2787 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 2788 * If already got a FIN from peer, then almost done; go to LAST_ACK 2789 * state. In all other cases, have already sent FIN to peer (e.g. 2790 * after PRU_SHUTDOWN), and just have to play tedious game waiting 2791 * for peer to send FIN or not respond to keep-alives, etc. 2792 * We can let the user exit from the close as soon as the FIN is acked. 2793 */ 2794 static void 2795 tcp_usrclosed(struct tcpcb *tp) 2796 { 2797 2798 NET_EPOCH_ASSERT(); 2799 INP_WLOCK_ASSERT(tp->t_inpcb); 2800 2801 switch (tp->t_state) { 2802 case TCPS_LISTEN: 2803 #ifdef TCP_OFFLOAD 2804 tcp_offload_listen_stop(tp); 2805 #endif 2806 tcp_state_change(tp, TCPS_CLOSED); 2807 /* FALLTHROUGH */ 2808 case TCPS_CLOSED: 2809 tp = tcp_close(tp); 2810 /* 2811 * tcp_close() should never return NULL here as the socket is 2812 * still open. 2813 */ 2814 KASSERT(tp != NULL, 2815 ("tcp_usrclosed: tcp_close() returned NULL")); 2816 break; 2817 2818 case TCPS_SYN_SENT: 2819 case TCPS_SYN_RECEIVED: 2820 tp->t_flags |= TF_NEEDFIN; 2821 break; 2822 2823 case TCPS_ESTABLISHED: 2824 tcp_state_change(tp, TCPS_FIN_WAIT_1); 2825 break; 2826 2827 case TCPS_CLOSE_WAIT: 2828 tcp_state_change(tp, TCPS_LAST_ACK); 2829 break; 2830 } 2831 if (tp->t_state >= TCPS_FIN_WAIT_2) { 2832 soisdisconnected(tp->t_inpcb->inp_socket); 2833 /* Prevent the connection hanging in FIN_WAIT_2 forever. */ 2834 if (tp->t_state == TCPS_FIN_WAIT_2) { 2835 int timeout; 2836 2837 timeout = (tcp_fast_finwait2_recycle) ? 2838 tcp_finwait2_timeout : TP_MAXIDLE(tp); 2839 tcp_timer_activate(tp, TT_2MSL, timeout); 2840 } 2841 } 2842 } 2843 2844 #ifdef DDB 2845 static void 2846 db_print_indent(int indent) 2847 { 2848 int i; 2849 2850 for (i = 0; i < indent; i++) 2851 db_printf(" "); 2852 } 2853 2854 static void 2855 db_print_tstate(int t_state) 2856 { 2857 2858 switch (t_state) { 2859 case TCPS_CLOSED: 2860 db_printf("TCPS_CLOSED"); 2861 return; 2862 2863 case TCPS_LISTEN: 2864 db_printf("TCPS_LISTEN"); 2865 return; 2866 2867 case TCPS_SYN_SENT: 2868 db_printf("TCPS_SYN_SENT"); 2869 return; 2870 2871 case TCPS_SYN_RECEIVED: 2872 db_printf("TCPS_SYN_RECEIVED"); 2873 return; 2874 2875 case TCPS_ESTABLISHED: 2876 db_printf("TCPS_ESTABLISHED"); 2877 return; 2878 2879 case TCPS_CLOSE_WAIT: 2880 db_printf("TCPS_CLOSE_WAIT"); 2881 return; 2882 2883 case TCPS_FIN_WAIT_1: 2884 db_printf("TCPS_FIN_WAIT_1"); 2885 return; 2886 2887 case TCPS_CLOSING: 2888 db_printf("TCPS_CLOSING"); 2889 return; 2890 2891 case TCPS_LAST_ACK: 2892 db_printf("TCPS_LAST_ACK"); 2893 return; 2894 2895 case TCPS_FIN_WAIT_2: 2896 db_printf("TCPS_FIN_WAIT_2"); 2897 return; 2898 2899 case TCPS_TIME_WAIT: 2900 db_printf("TCPS_TIME_WAIT"); 2901 return; 2902 2903 default: 2904 db_printf("unknown"); 2905 return; 2906 } 2907 } 2908 2909 static void 2910 db_print_tflags(u_int t_flags) 2911 { 2912 int comma; 2913 2914 comma = 0; 2915 if (t_flags & TF_ACKNOW) { 2916 db_printf("%sTF_ACKNOW", comma ? ", " : ""); 2917 comma = 1; 2918 } 2919 if (t_flags & TF_DELACK) { 2920 db_printf("%sTF_DELACK", comma ? ", " : ""); 2921 comma = 1; 2922 } 2923 if (t_flags & TF_NODELAY) { 2924 db_printf("%sTF_NODELAY", comma ? ", " : ""); 2925 comma = 1; 2926 } 2927 if (t_flags & TF_NOOPT) { 2928 db_printf("%sTF_NOOPT", comma ? ", " : ""); 2929 comma = 1; 2930 } 2931 if (t_flags & TF_SENTFIN) { 2932 db_printf("%sTF_SENTFIN", comma ? ", " : ""); 2933 comma = 1; 2934 } 2935 if (t_flags & TF_REQ_SCALE) { 2936 db_printf("%sTF_REQ_SCALE", comma ? ", " : ""); 2937 comma = 1; 2938 } 2939 if (t_flags & TF_RCVD_SCALE) { 2940 db_printf("%sTF_RECVD_SCALE", comma ? ", " : ""); 2941 comma = 1; 2942 } 2943 if (t_flags & TF_REQ_TSTMP) { 2944 db_printf("%sTF_REQ_TSTMP", comma ? ", " : ""); 2945 comma = 1; 2946 } 2947 if (t_flags & TF_RCVD_TSTMP) { 2948 db_printf("%sTF_RCVD_TSTMP", comma ? ", " : ""); 2949 comma = 1; 2950 } 2951 if (t_flags & TF_SACK_PERMIT) { 2952 db_printf("%sTF_SACK_PERMIT", comma ? ", " : ""); 2953 comma = 1; 2954 } 2955 if (t_flags & TF_NEEDSYN) { 2956 db_printf("%sTF_NEEDSYN", comma ? ", " : ""); 2957 comma = 1; 2958 } 2959 if (t_flags & TF_NEEDFIN) { 2960 db_printf("%sTF_NEEDFIN", comma ? ", " : ""); 2961 comma = 1; 2962 } 2963 if (t_flags & TF_NOPUSH) { 2964 db_printf("%sTF_NOPUSH", comma ? ", " : ""); 2965 comma = 1; 2966 } 2967 if (t_flags & TF_PREVVALID) { 2968 db_printf("%sTF_PREVVALID", comma ? ", " : ""); 2969 comma = 1; 2970 } 2971 if (t_flags & TF_MORETOCOME) { 2972 db_printf("%sTF_MORETOCOME", comma ? ", " : ""); 2973 comma = 1; 2974 } 2975 if (t_flags & TF_LQ_OVERFLOW) { 2976 db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : ""); 2977 comma = 1; 2978 } 2979 if (t_flags & TF_LASTIDLE) { 2980 db_printf("%sTF_LASTIDLE", comma ? ", " : ""); 2981 comma = 1; 2982 } 2983 if (t_flags & TF_RXWIN0SENT) { 2984 db_printf("%sTF_RXWIN0SENT", comma ? ", " : ""); 2985 comma = 1; 2986 } 2987 if (t_flags & TF_FASTRECOVERY) { 2988 db_printf("%sTF_FASTRECOVERY", comma ? ", " : ""); 2989 comma = 1; 2990 } 2991 if (t_flags & TF_CONGRECOVERY) { 2992 db_printf("%sTF_CONGRECOVERY", comma ? ", " : ""); 2993 comma = 1; 2994 } 2995 if (t_flags & TF_WASFRECOVERY) { 2996 db_printf("%sTF_WASFRECOVERY", comma ? ", " : ""); 2997 comma = 1; 2998 } 2999 if (t_flags & TF_WASCRECOVERY) { 3000 db_printf("%sTF_WASCRECOVERY", comma ? ", " : ""); 3001 comma = 1; 3002 } 3003 if (t_flags & TF_SIGNATURE) { 3004 db_printf("%sTF_SIGNATURE", comma ? ", " : ""); 3005 comma = 1; 3006 } 3007 if (t_flags & TF_FORCEDATA) { 3008 db_printf("%sTF_FORCEDATA", comma ? ", " : ""); 3009 comma = 1; 3010 } 3011 if (t_flags & TF_TSO) { 3012 db_printf("%sTF_TSO", comma ? ", " : ""); 3013 comma = 1; 3014 } 3015 if (t_flags & TF_FASTOPEN) { 3016 db_printf("%sTF_FASTOPEN", comma ? ", " : ""); 3017 comma = 1; 3018 } 3019 } 3020 3021 static void 3022 db_print_tflags2(u_int t_flags2) 3023 { 3024 int comma; 3025 3026 comma = 0; 3027 if (t_flags2 & TF2_PLPMTU_BLACKHOLE) { 3028 db_printf("%sTF2_PLPMTU_BLACKHOLE", comma ? ", " : ""); 3029 comma = 1; 3030 } 3031 if (t_flags2 & TF2_PLPMTU_PMTUD) { 3032 db_printf("%sTF2_PLPMTU_PMTUD", comma ? ", " : ""); 3033 comma = 1; 3034 } 3035 if (t_flags2 & TF2_PLPMTU_MAXSEGSNT) { 3036 db_printf("%sTF2_PLPMTU_MAXSEGSNT", comma ? ", " : ""); 3037 comma = 1; 3038 } 3039 if (t_flags2 & TF2_LOG_AUTO) { 3040 db_printf("%sTF2_LOG_AUTO", comma ? ", " : ""); 3041 comma = 1; 3042 } 3043 if (t_flags2 & TF2_DROP_AF_DATA) { 3044 db_printf("%sTF2_DROP_AF_DATA", comma ? ", " : ""); 3045 comma = 1; 3046 } 3047 if (t_flags2 & TF2_ECN_PERMIT) { 3048 db_printf("%sTF2_ECN_PERMIT", comma ? ", " : ""); 3049 comma = 1; 3050 } 3051 if (t_flags2 & TF2_ECN_SND_CWR) { 3052 db_printf("%sTF2_ECN_SND_CWR", comma ? ", " : ""); 3053 comma = 1; 3054 } 3055 if (t_flags2 & TF2_ECN_SND_ECE) { 3056 db_printf("%sTF2_ECN_SND_ECE", comma ? ", " : ""); 3057 comma = 1; 3058 } 3059 if (t_flags2 & TF2_ACE_PERMIT) { 3060 db_printf("%sTF2_ACE_PERMIT", comma ? ", " : ""); 3061 comma = 1; 3062 } 3063 if (t_flags2 & TF2_FBYTES_COMPLETE) { 3064 db_printf("%sTF2_FBYTES_COMPLETE", comma ? ", " : ""); 3065 comma = 1; 3066 } 3067 } 3068 3069 static void 3070 db_print_toobflags(char t_oobflags) 3071 { 3072 int comma; 3073 3074 comma = 0; 3075 if (t_oobflags & TCPOOB_HAVEDATA) { 3076 db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : ""); 3077 comma = 1; 3078 } 3079 if (t_oobflags & TCPOOB_HADDATA) { 3080 db_printf("%sTCPOOB_HADDATA", comma ? ", " : ""); 3081 comma = 1; 3082 } 3083 } 3084 3085 static void 3086 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent) 3087 { 3088 3089 db_print_indent(indent); 3090 db_printf("%s at %p\n", name, tp); 3091 3092 indent += 2; 3093 3094 db_print_indent(indent); 3095 db_printf("t_segq first: %p t_segqlen: %d t_dupacks: %d\n", 3096 TAILQ_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks); 3097 3098 db_print_indent(indent); 3099 db_printf("tt_rexmt: %p tt_persist: %p tt_keep: %p\n", 3100 &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep); 3101 3102 db_print_indent(indent); 3103 db_printf("tt_2msl: %p tt_delack: %p t_inpcb: %p\n", &tp->t_timers->tt_2msl, 3104 &tp->t_timers->tt_delack, tp->t_inpcb); 3105 3106 db_print_indent(indent); 3107 db_printf("t_state: %d (", tp->t_state); 3108 db_print_tstate(tp->t_state); 3109 db_printf(")\n"); 3110 3111 db_print_indent(indent); 3112 db_printf("t_flags: 0x%x (", tp->t_flags); 3113 db_print_tflags(tp->t_flags); 3114 db_printf(")\n"); 3115 3116 db_print_indent(indent); 3117 db_printf("t_flags2: 0x%x (", tp->t_flags2); 3118 db_print_tflags2(tp->t_flags2); 3119 db_printf(")\n"); 3120 3121 db_print_indent(indent); 3122 db_printf("snd_una: 0x%08x snd_max: 0x%08x snd_nxt: x0%08x\n", 3123 tp->snd_una, tp->snd_max, tp->snd_nxt); 3124 3125 db_print_indent(indent); 3126 db_printf("snd_up: 0x%08x snd_wl1: 0x%08x snd_wl2: 0x%08x\n", 3127 tp->snd_up, tp->snd_wl1, tp->snd_wl2); 3128 3129 db_print_indent(indent); 3130 db_printf("iss: 0x%08x irs: 0x%08x rcv_nxt: 0x%08x\n", 3131 tp->iss, tp->irs, tp->rcv_nxt); 3132 3133 db_print_indent(indent); 3134 db_printf("rcv_adv: 0x%08x rcv_wnd: %u rcv_up: 0x%08x\n", 3135 tp->rcv_adv, tp->rcv_wnd, tp->rcv_up); 3136 3137 db_print_indent(indent); 3138 db_printf("snd_wnd: %u snd_cwnd: %u\n", 3139 tp->snd_wnd, tp->snd_cwnd); 3140 3141 db_print_indent(indent); 3142 db_printf("snd_ssthresh: %u snd_recover: " 3143 "0x%08x\n", tp->snd_ssthresh, tp->snd_recover); 3144 3145 db_print_indent(indent); 3146 db_printf("t_rcvtime: %u t_startime: %u\n", 3147 tp->t_rcvtime, tp->t_starttime); 3148 3149 db_print_indent(indent); 3150 db_printf("t_rttime: %u t_rtsq: 0x%08x\n", 3151 tp->t_rtttime, tp->t_rtseq); 3152 3153 db_print_indent(indent); 3154 db_printf("t_rxtcur: %d t_maxseg: %u t_srtt: %d\n", 3155 tp->t_rxtcur, tp->t_maxseg, tp->t_srtt); 3156 3157 db_print_indent(indent); 3158 db_printf("t_rttvar: %d t_rxtshift: %d t_rttmin: %u " 3159 "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin, 3160 tp->t_rttbest); 3161 3162 db_print_indent(indent); 3163 db_printf("t_rttupdated: %lu 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 3203 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb) 3204 { 3205 struct tcpcb *tp; 3206 3207 if (!have_addr) { 3208 db_printf("usage: show tcpcb <addr>\n"); 3209 return; 3210 } 3211 tp = (struct tcpcb *)addr; 3212 3213 db_print_tcpcb(tp, "tcpcb", 0); 3214 } 3215 #endif 3216