sys/netinet/tcp_usrreq.c

   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_tcpdebug.h"
  48
  49 #include <sys/param.h>
  50 #include <sys/systm.h>
  51 #include <sys/limits.h>
  52 #include <sys/malloc.h>
  53 #include <sys/refcount.h>
  54 #include <sys/kernel.h>
  55 #include <sys/sysctl.h>
  56 #include <sys/mbuf.h>
  57 #ifdef INET6
  58 #include <sys/domain.h>
  59 #endif /* INET6 */
  60 #include <sys/socket.h>
  61 #include <sys/socketvar.h>
  62 #include <sys/protosw.h>
  63 #include <sys/proc.h>
  64 #include <sys/jail.h>
  65 #include <sys/syslog.h>
  66
  67 #ifdef DDB
  68 #include <ddb/ddb.h>
  69 #endif
  70
  71 #include <net/if.h>
  72 #include <net/if_var.h>
  73 #include <net/route.h>
  74 #include <net/vnet.h>
  75
  76 #include <netinet/in.h>
  77 #include <netinet/in_kdtrace.h>
  78 #include <netinet/in_pcb.h>
  79 #include <netinet/in_systm.h>
  80 #include <netinet/in_var.h>
  81 #include <netinet/ip_var.h>
  82 #ifdef INET6
  83 #include <netinet/ip6.h>
  84 #include <netinet6/in6_pcb.h>
  85 #include <netinet6/ip6_var.h>
  86 #include <netinet6/scope6_var.h>
  87 #endif
  88 #include <netinet/tcp.h>
  89 #include <netinet/tcp_fsm.h>
  90 #include <netinet/tcp_seq.h>
  91 #include <netinet/tcp_timer.h>
  92 #include <netinet/tcp_var.h>
  93 #include <netinet/tcp_log_buf.h>
  94 #include <netinet/tcpip.h>
  95 #include <netinet/cc/cc.h>
  96 #include <netinet/tcp_fastopen.h>
  97 #ifdef TCPPCAP
  98 #include <netinet/tcp_pcap.h>
  99 #endif
 100 #ifdef TCPDEBUG
 101 #include <netinet/tcp_debug.h>
 102 #endif
 103 #ifdef TCP_OFFLOAD
 104 #include <netinet/tcp_offload.h>
 105 #endif
 106 #include <netipsec/ipsec_support.h>
 107
 108 /*
 109  * TCP protocol interface to socket abstraction.
 110  */
 111 static int      tcp_attach(struct socket *);
 112 #ifdef INET
 113 static int      tcp_connect(struct tcpcb *, struct sockaddr *,
 114                     struct thread *td);
 115 #endif /* INET */
 116 #ifdef INET6
 117 static int      tcp6_connect(struct tcpcb *, struct sockaddr *,
 118                     struct thread *td);
 119 #endif /* INET6 */
 120 static void     tcp_disconnect(struct tcpcb *);
 121 static void     tcp_usrclosed(struct tcpcb *);
 122 static void     tcp_fill_info(struct tcpcb *, struct tcp_info *);
 123
 124 #ifdef TCPDEBUG
 125 #define TCPDEBUG0       int ostate = 0
 126 #define TCPDEBUG1()     ostate = tp ? tp->t_state : 0
 127 #define TCPDEBUG2(req)  if (tp && (so->so_options & SO_DEBUG)) \
 128                                 tcp_trace(TA_USER, ostate, tp, 0, 0, req)
 129 #else
 130 #define TCPDEBUG0
 131 #define TCPDEBUG1()
 132 #define TCPDEBUG2(req)
 133 #endif
 134
 135 /*
 136  * TCP attaches to socket via pru_attach(), reserving space,
 137  * and an internet control block.
 138  */
 139 static int
 140 tcp_usr_attach(struct socket *so, int proto, struct thread *td)
 141 {
 142         struct inpcb *inp;
 143         struct tcpcb *tp = NULL;
 144         int error;
 145         TCPDEBUG0;
 146
 147         inp = sotoinpcb(so);
 148         KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL"));
 149         TCPDEBUG1();
 150
 151         error = tcp_attach(so);
 152         if (error)
 153                 goto out;
 154
 155         if ((so->so_options & SO_LINGER) && so->so_linger == 0)
 156                 so->so_linger = TCP_LINGERTIME;
 157
 158         inp = sotoinpcb(so);
 159         tp = intotcpcb(inp);
 160 out:
 161         TCPDEBUG2(PRU_ATTACH);
 162         TCP_PROBE2(debug__user, tp, PRU_ATTACH);
 163         return error;
 164 }
 165
 166 /*
 167  * tcp_detach is called when the socket layer loses its final reference
 168  * to the socket, be it a file descriptor reference, a reference from TCP,
 169  * etc.  At this point, there is only one case in which we will keep around
 170  * inpcb state: time wait.
 171  *
 172  * This function can probably be re-absorbed back into tcp_usr_detach() now
 173  * that there is a single detach path.
 174  */
 175 static void
 176 tcp_detach(struct socket *so, struct inpcb *inp)
 177 {
 178         struct tcpcb *tp;
 179
 180         INP_INFO_LOCK_ASSERT(&V_tcbinfo);
 181         INP_WLOCK_ASSERT(inp);
 182
 183         KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp"));
 184         KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so"));
 185
 186         tp = intotcpcb(inp);
 187
 188         if (inp->inp_flags & INP_TIMEWAIT) {
 189                 /*
 190                  * There are two cases to handle: one in which the time wait
 191                  * state is being discarded (INP_DROPPED), and one in which
 192                  * this connection will remain in timewait.  In the former,
 193                  * it is time to discard all state (except tcptw, which has
 194                  * already been discarded by the timewait close code, which
 195                  * should be further up the call stack somewhere).  In the
 196                  * latter case, we detach from the socket, but leave the pcb
 197                  * present until timewait ends.
 198                  *
 199                  * XXXRW: Would it be cleaner to free the tcptw here?
 200                  *
 201                  * Astute question indeed, from twtcp perspective there are
 202                  * four cases to consider:
 203                  *
 204                  * #1 tcp_detach is called at tcptw creation time by
 205                  *  tcp_twstart, then do not discard the newly created tcptw
 206                  *  and leave inpcb present until timewait ends
 207                  * #2 tcp_detach is called at tcptw creation time by
 208                  *  tcp_twstart, but connection is local and tw will be
 209                  *  discarded immediately
 210                  * #3 tcp_detach is called at timewait end (or reuse) by
 211                  *  tcp_twclose, then the tcptw has already been discarded
 212                  *  (or reused) and inpcb is freed here
 213                  * #4 tcp_detach is called() after timewait ends (or reuse)
 214                  *  (e.g. by soclose), then tcptw has already been discarded
 215                  *  (or reused) and inpcb is freed here
 216                  *
 217                  *  In all three cases the tcptw should not be freed here.
 218                  */
 219                 if (inp->inp_flags & INP_DROPPED) {
 220                         in_pcbdetach(inp);
 221                         if (__predict_true(tp == NULL)) {
 222                                 in_pcbfree(inp);
 223                         } else {
 224                                 /*
 225                                  * This case should not happen as in TIMEWAIT
 226                                  * state the inp should not be destroyed before
 227                                  * its tcptw.  If INVARIANTS is defined, panic.
 228                                  */
 229 #ifdef INVARIANTS
 230                                 panic("%s: Panic before an inp double-free: "
 231                                     "INP_TIMEWAIT && INP_DROPPED && tp != NULL"
 232                                     , __func__);
 233 #else
 234                                 log(LOG_ERR, "%s: Avoid an inp double-free: "
 235                                     "INP_TIMEWAIT && INP_DROPPED && tp != NULL"
 236                                     , __func__);
 237 #endif
 238                                 INP_WUNLOCK(inp);
 239                         }
 240                 } else {
 241                         in_pcbdetach(inp);
 242                         INP_WUNLOCK(inp);
 243                 }
 244         } else {
 245                 /*
 246                  * If the connection is not in timewait, we consider two
 247                  * two conditions: one in which no further processing is
 248                  * necessary (dropped || embryonic), and one in which TCP is
 249                  * not yet done, but no longer requires the socket, so the
 250                  * pcb will persist for the time being.
 251                  *
 252                  * XXXRW: Does the second case still occur?
 253                  */
 254                 if (inp->inp_flags & INP_DROPPED ||
 255                     tp->t_state < TCPS_SYN_SENT) {
 256                         tcp_discardcb(tp);
 257                         in_pcbdetach(inp);
 258                         in_pcbfree(inp);
 259                 } else {
 260                         in_pcbdetach(inp);
 261                         INP_WUNLOCK(inp);
 262                 }
 263         }
 264 }
 265
 266 /*
 267  * pru_detach() detaches the TCP protocol from the socket.
 268  * If the protocol state is non-embryonic, then can't
 269  * do this directly: have to initiate a pru_disconnect(),
 270  * which may finish later; embryonic TCB's can just
 271  * be discarded here.
 272  */
 273 static void
 274 tcp_usr_detach(struct socket *so)
 275 {
 276         struct inpcb *inp;
 277         int rlock = 0;
 278
 279         inp = sotoinpcb(so);
 280         KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL"));
 281         if (!INP_INFO_WLOCKED(&V_tcbinfo)) {
 282                 INP_INFO_RLOCK(&V_tcbinfo);
 283                 rlock = 1;
 284         }
 285         INP_WLOCK(inp);
 286         KASSERT(inp->inp_socket != NULL,
 287             ("tcp_usr_detach: inp_socket == NULL"));
 288         tcp_detach(so, inp);
 289         if (rlock)
 290                 INP_INFO_RUNLOCK(&V_tcbinfo);
 291 }
 292
 293 #ifdef INET
 294 /*
 295  * Give the socket an address.
 296  */
 297 static int
 298 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 299 {
 300         int error = 0;
 301         struct inpcb *inp;
 302         struct tcpcb *tp = NULL;
 303         struct sockaddr_in *sinp;
 304
 305         sinp = (struct sockaddr_in *)nam;
 306         if (nam->sa_len != sizeof (*sinp))
 307                 return (EINVAL);
 308         /*
 309          * Must check for multicast addresses and disallow binding
 310          * to them.
 311          */
 312         if (sinp->sin_family == AF_INET &&
 313             IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
 314                 return (EAFNOSUPPORT);
 315
 316         TCPDEBUG0;
 317         inp = sotoinpcb(so);
 318         KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL"));
 319         INP_WLOCK(inp);
 320         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 321                 error = EINVAL;
 322                 goto out;
 323         }
 324         tp = intotcpcb(inp);
 325         TCPDEBUG1();
 326         INP_HASH_WLOCK(&V_tcbinfo);
 327         error = in_pcbbind(inp, nam, td->td_ucred);
 328         INP_HASH_WUNLOCK(&V_tcbinfo);
 329 out:
 330         TCPDEBUG2(PRU_BIND);
 331         TCP_PROBE2(debug__user, tp, PRU_BIND);
 332         INP_WUNLOCK(inp);
 333
 334         return (error);
 335 }
 336 #endif /* INET */
 337
 338 #ifdef INET6
 339 static int
 340 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 341 {
 342         int error = 0;
 343         struct inpcb *inp;
 344         struct tcpcb *tp = NULL;
 345         struct sockaddr_in6 *sin6p;
 346
 347         sin6p = (struct sockaddr_in6 *)nam;
 348         if (nam->sa_len != sizeof (*sin6p))
 349                 return (EINVAL);
 350         /*
 351          * Must check for multicast addresses and disallow binding
 352          * to them.
 353          */
 354         if (sin6p->sin6_family == AF_INET6 &&
 355             IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
 356                 return (EAFNOSUPPORT);
 357
 358         TCPDEBUG0;
 359         inp = sotoinpcb(so);
 360         KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL"));
 361         INP_WLOCK(inp);
 362         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 363                 error = EINVAL;
 364                 goto out;
 365         }
 366         tp = intotcpcb(inp);
 367         TCPDEBUG1();
 368         INP_HASH_WLOCK(&V_tcbinfo);
 369         inp->inp_vflag &= ~INP_IPV4;
 370         inp->inp_vflag |= INP_IPV6;
 371 #ifdef INET
 372         if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
 373                 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
 374                         inp->inp_vflag |= INP_IPV4;
 375                 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
 376                         struct sockaddr_in sin;
 377
 378                         in6_sin6_2_sin(&sin, sin6p);
 379                         inp->inp_vflag |= INP_IPV4;
 380                         inp->inp_vflag &= ~INP_IPV6;
 381                         error = in_pcbbind(inp, (struct sockaddr *)&sin,
 382                             td->td_ucred);
 383                         INP_HASH_WUNLOCK(&V_tcbinfo);
 384                         goto out;
 385                 }
 386         }
 387 #endif
 388         error = in6_pcbbind(inp, nam, td->td_ucred);
 389         INP_HASH_WUNLOCK(&V_tcbinfo);
 390 out:
 391         TCPDEBUG2(PRU_BIND);
 392         TCP_PROBE2(debug__user, tp, PRU_BIND);
 393         INP_WUNLOCK(inp);
 394         return (error);
 395 }
 396 #endif /* INET6 */
 397
 398 #ifdef INET
 399 /*
 400  * Prepare to accept connections.
 401  */
 402 static int
 403 tcp_usr_listen(struct socket *so, int backlog, struct thread *td)
 404 {
 405         int error = 0;
 406         struct inpcb *inp;
 407         struct tcpcb *tp = NULL;
 408
 409         TCPDEBUG0;
 410         inp = sotoinpcb(so);
 411         KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL"));
 412         INP_WLOCK(inp);
 413         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 414                 error = EINVAL;
 415                 goto out;
 416         }
 417         tp = intotcpcb(inp);
 418         TCPDEBUG1();
 419         SOCK_LOCK(so);
 420         error = solisten_proto_check(so);
 421         INP_HASH_WLOCK(&V_tcbinfo);
 422         if (error == 0 && inp->inp_lport == 0)
 423                 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
 424         INP_HASH_WUNLOCK(&V_tcbinfo);
 425         if (error == 0) {
 426                 tcp_state_change(tp, TCPS_LISTEN);
 427                 solisten_proto(so, backlog);
 428 #ifdef TCP_OFFLOAD
 429                 if ((so->so_options & SO_NO_OFFLOAD) == 0)
 430                         tcp_offload_listen_start(tp);
 431 #endif
 432         }
 433         SOCK_UNLOCK(so);
 434
 435         if (IS_FASTOPEN(tp->t_flags))
 436                 tp->t_tfo_pending = tcp_fastopen_alloc_counter();
 437
 438 out:
 439         TCPDEBUG2(PRU_LISTEN);
 440         TCP_PROBE2(debug__user, tp, PRU_LISTEN);
 441         INP_WUNLOCK(inp);
 442         return (error);
 443 }
 444 #endif /* INET */
 445
 446 #ifdef INET6
 447 static int
 448 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td)
 449 {
 450         int error = 0;
 451         struct inpcb *inp;
 452         struct tcpcb *tp = NULL;
 453
 454         TCPDEBUG0;
 455         inp = sotoinpcb(so);
 456         KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL"));
 457         INP_WLOCK(inp);
 458         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 459                 error = EINVAL;
 460                 goto out;
 461         }
 462         tp = intotcpcb(inp);
 463         TCPDEBUG1();
 464         SOCK_LOCK(so);
 465         error = solisten_proto_check(so);
 466         INP_HASH_WLOCK(&V_tcbinfo);
 467         if (error == 0 && inp->inp_lport == 0) {
 468                 inp->inp_vflag &= ~INP_IPV4;
 469                 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
 470                         inp->inp_vflag |= INP_IPV4;
 471                 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
 472         }
 473         INP_HASH_WUNLOCK(&V_tcbinfo);
 474         if (error == 0) {
 475                 tcp_state_change(tp, TCPS_LISTEN);
 476                 solisten_proto(so, backlog);
 477 #ifdef TCP_OFFLOAD
 478                 if ((so->so_options & SO_NO_OFFLOAD) == 0)
 479                         tcp_offload_listen_start(tp);
 480 #endif
 481         }
 482         SOCK_UNLOCK(so);
 483
 484         if (IS_FASTOPEN(tp->t_flags))
 485                 tp->t_tfo_pending = tcp_fastopen_alloc_counter();
 486
 487 out:
 488         TCPDEBUG2(PRU_LISTEN);
 489         TCP_PROBE2(debug__user, tp, PRU_LISTEN);
 490         INP_WUNLOCK(inp);
 491         return (error);
 492 }
 493 #endif /* INET6 */
 494
 495 #ifdef INET
 496 /*
 497  * Initiate connection to peer.
 498  * Create a template for use in transmissions on this connection.
 499  * Enter SYN_SENT state, and mark socket as connecting.
 500  * Start keep-alive timer, and seed output sequence space.
 501  * Send initial segment on connection.
 502  */
 503 static int
 504 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 505 {
 506         int error = 0;
 507         struct inpcb *inp;
 508         struct tcpcb *tp = NULL;
 509         struct sockaddr_in *sinp;
 510
 511         sinp = (struct sockaddr_in *)nam;
 512         if (nam->sa_len != sizeof (*sinp))
 513                 return (EINVAL);
 514         /*
 515          * Must disallow TCP ``connections'' to multicast addresses.
 516          */
 517         if (sinp->sin_family == AF_INET
 518             && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
 519                 return (EAFNOSUPPORT);
 520         if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0)
 521                 return (error);
 522
 523         TCPDEBUG0;
 524         inp = sotoinpcb(so);
 525         KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL"));
 526         INP_WLOCK(inp);
 527         if (inp->inp_flags & INP_TIMEWAIT) {
 528                 error = EADDRINUSE;
 529                 goto out;
 530         }
 531         if (inp->inp_flags & INP_DROPPED) {
 532                 error = ECONNREFUSED;
 533                 goto out;
 534         }
 535         tp = intotcpcb(inp);
 536         TCPDEBUG1();
 537         if ((error = tcp_connect(tp, nam, td)) != 0)
 538                 goto out;
 539 #ifdef TCP_OFFLOAD
 540         if (registered_toedevs > 0 &&
 541             (so->so_options & SO_NO_OFFLOAD) == 0 &&
 542             (error = tcp_offload_connect(so, nam)) == 0)
 543                 goto out;
 544 #endif
 545         tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
 546         error = tp->t_fb->tfb_tcp_output(tp);
 547 out:
 548         TCPDEBUG2(PRU_CONNECT);
 549         TCP_PROBE2(debug__user, tp, PRU_CONNECT);
 550         INP_WUNLOCK(inp);
 551         return (error);
 552 }
 553 #endif /* INET */
 554
 555 #ifdef INET6
 556 static int
 557 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 558 {
 559         int error = 0;
 560         struct inpcb *inp;
 561         struct tcpcb *tp = NULL;
 562         struct sockaddr_in6 *sin6p;
 563
 564         TCPDEBUG0;
 565
 566         sin6p = (struct sockaddr_in6 *)nam;
 567         if (nam->sa_len != sizeof (*sin6p))
 568                 return (EINVAL);
 569         /*
 570          * Must disallow TCP ``connections'' to multicast addresses.
 571          */
 572         if (sin6p->sin6_family == AF_INET6
 573             && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
 574                 return (EAFNOSUPPORT);
 575
 576         inp = sotoinpcb(so);
 577         KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL"));
 578         INP_WLOCK(inp);
 579         if (inp->inp_flags & INP_TIMEWAIT) {
 580                 error = EADDRINUSE;
 581                 goto out;
 582         }
 583         if (inp->inp_flags & INP_DROPPED) {
 584                 error = ECONNREFUSED;
 585                 goto out;
 586         }
 587         tp = intotcpcb(inp);
 588         TCPDEBUG1();
 589 #ifdef INET
 590         /*
 591          * XXXRW: Some confusion: V4/V6 flags relate to binding, and
 592          * therefore probably require the hash lock, which isn't held here.
 593          * Is this a significant problem?
 594          */
 595         if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
 596                 struct sockaddr_in sin;
 597
 598                 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
 599                         error = EINVAL;
 600                         goto out;
 601                 }
 602                 if ((inp->inp_vflag & INP_IPV4) == 0) {
 603                         error = EAFNOSUPPORT;
 604                         goto out;
 605                 }
 606
 607                 in6_sin6_2_sin(&sin, sin6p);
 608                 inp->inp_vflag |= INP_IPV4;
 609                 inp->inp_vflag &= ~INP_IPV6;
 610                 if ((error = prison_remote_ip4(td->td_ucred,
 611                     &sin.sin_addr)) != 0)
 612                         goto out;
 613                 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
 614                         goto out;
 615 #ifdef TCP_OFFLOAD
 616                 if (registered_toedevs > 0 &&
 617                     (so->so_options & SO_NO_OFFLOAD) == 0 &&
 618                     (error = tcp_offload_connect(so, nam)) == 0)
 619                         goto out;
 620 #endif
 621                 error = tp->t_fb->tfb_tcp_output(tp);
 622                 goto out;
 623         } else {
 624                 if ((inp->inp_vflag & INP_IPV6) == 0) {
 625                         error = EAFNOSUPPORT;
 626                         goto out;
 627                 }
 628         }
 629 #endif
 630         inp->inp_vflag &= ~INP_IPV4;
 631         inp->inp_vflag |= INP_IPV6;
 632         inp->inp_inc.inc_flags |= INC_ISIPV6;
 633         if ((error = prison_remote_ip6(td->td_ucred, &sin6p->sin6_addr)) != 0)
 634                 goto out;
 635         if ((error = tcp6_connect(tp, nam, td)) != 0)
 636                 goto out;
 637 #ifdef TCP_OFFLOAD
 638         if (registered_toedevs > 0 &&
 639             (so->so_options & SO_NO_OFFLOAD) == 0 &&
 640             (error = tcp_offload_connect(so, nam)) == 0)
 641                 goto out;
 642 #endif
 643         tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
 644         error = tp->t_fb->tfb_tcp_output(tp);
 645
 646 out:
 647         TCPDEBUG2(PRU_CONNECT);
 648         TCP_PROBE2(debug__user, tp, PRU_CONNECT);
 649         INP_WUNLOCK(inp);
 650         return (error);
 651 }
 652 #endif /* INET6 */
 653
 654 /*
 655  * Initiate disconnect from peer.
 656  * If connection never passed embryonic stage, just drop;
 657  * else if don't need to let data drain, then can just drop anyways,
 658  * else have to begin TCP shutdown process: mark socket disconnecting,
 659  * drain unread data, state switch to reflect user close, and
 660  * send segment (e.g. FIN) to peer.  Socket will be really disconnected
 661  * when peer sends FIN and acks ours.
 662  *
 663  * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
 664  */
 665 static int
 666 tcp_usr_disconnect(struct socket *so)
 667 {
 668         struct inpcb *inp;
 669         struct tcpcb *tp = NULL;
 670         int error = 0;
 671
 672         TCPDEBUG0;
 673         INP_INFO_RLOCK(&V_tcbinfo);
 674         inp = sotoinpcb(so);
 675         KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
 676         INP_WLOCK(inp);
 677         if (inp->inp_flags & INP_TIMEWAIT)
 678                 goto out;
 679         if (inp->inp_flags & INP_DROPPED) {
 680                 error = ECONNRESET;
 681                 goto out;
 682         }
 683         tp = intotcpcb(inp);
 684         TCPDEBUG1();
 685         tcp_disconnect(tp);
 686 out:
 687         TCPDEBUG2(PRU_DISCONNECT);
 688         TCP_PROBE2(debug__user, tp, PRU_DISCONNECT);
 689         INP_WUNLOCK(inp);
 690         INP_INFO_RUNLOCK(&V_tcbinfo);
 691         return (error);
 692 }
 693
 694 #ifdef INET
 695 /*
 696  * Accept a connection.  Essentially all the work is done at higher levels;
 697  * just return the address of the peer, storing through addr.
 698  */
 699 static int
 700 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
 701 {
 702         int error = 0;
 703         struct inpcb *inp = NULL;
 704         struct tcpcb *tp = NULL;
 705         struct in_addr addr;
 706         in_port_t port = 0;
 707         TCPDEBUG0;
 708
 709         if (so->so_state & SS_ISDISCONNECTED)
 710                 return (ECONNABORTED);
 711
 712         inp = sotoinpcb(so);
 713         KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
 714         INP_WLOCK(inp);
 715         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 716                 error = ECONNABORTED;
 717                 goto out;
 718         }
 719         tp = intotcpcb(inp);
 720         TCPDEBUG1();
 721
 722         /*
 723          * We inline in_getpeeraddr and COMMON_END here, so that we can
 724          * copy the data of interest and defer the malloc until after we
 725          * release the lock.
 726          */
 727         port = inp->inp_fport;
 728         addr = inp->inp_faddr;
 729
 730 out:
 731         TCPDEBUG2(PRU_ACCEPT);
 732         TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
 733         INP_WUNLOCK(inp);
 734         if (error == 0)
 735                 *nam = in_sockaddr(port, &addr);
 736         return error;
 737 }
 738 #endif /* INET */
 739
 740 #ifdef INET6
 741 static int
 742 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
 743 {
 744         struct inpcb *inp = NULL;
 745         int error = 0;
 746         struct tcpcb *tp = NULL;
 747         struct in_addr addr;
 748         struct in6_addr addr6;
 749         in_port_t port = 0;
 750         int v4 = 0;
 751         TCPDEBUG0;
 752
 753         if (so->so_state & SS_ISDISCONNECTED)
 754                 return (ECONNABORTED);
 755
 756         inp = sotoinpcb(so);
 757         KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
 758         INP_INFO_RLOCK(&V_tcbinfo);
 759         INP_WLOCK(inp);
 760         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 761                 error = ECONNABORTED;
 762                 goto out;
 763         }
 764         tp = intotcpcb(inp);
 765         TCPDEBUG1();
 766
 767         /*
 768          * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
 769          * copy the data of interest and defer the malloc until after we
 770          * release the lock.
 771          */
 772         if (inp->inp_vflag & INP_IPV4) {
 773                 v4 = 1;
 774                 port = inp->inp_fport;
 775                 addr = inp->inp_faddr;
 776         } else {
 777                 port = inp->inp_fport;
 778                 addr6 = inp->in6p_faddr;
 779         }
 780
 781 out:
 782         TCPDEBUG2(PRU_ACCEPT);
 783         TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
 784         INP_WUNLOCK(inp);
 785         INP_INFO_RUNLOCK(&V_tcbinfo);
 786         if (error == 0) {
 787                 if (v4)
 788                         *nam = in6_v4mapsin6_sockaddr(port, &addr);
 789                 else
 790                         *nam = in6_sockaddr(port, &addr6);
 791         }
 792         return error;
 793 }
 794 #endif /* INET6 */
 795
 796 /*
 797  * Mark the connection as being incapable of further output.
 798  */
 799 static int
 800 tcp_usr_shutdown(struct socket *so)
 801 {
 802         int error = 0;
 803         struct inpcb *inp;
 804         struct tcpcb *tp = NULL;
 805
 806         TCPDEBUG0;
 807         INP_INFO_RLOCK(&V_tcbinfo);
 808         inp = sotoinpcb(so);
 809         KASSERT(inp != NULL, ("inp == NULL"));
 810         INP_WLOCK(inp);
 811         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 812                 error = ECONNRESET;
 813                 goto out;
 814         }
 815         tp = intotcpcb(inp);
 816         TCPDEBUG1();
 817         socantsendmore(so);
 818         tcp_usrclosed(tp);
 819         if (!(inp->inp_flags & INP_DROPPED))
 820                 error = tp->t_fb->tfb_tcp_output(tp);
 821
 822 out:
 823         TCPDEBUG2(PRU_SHUTDOWN);
 824         TCP_PROBE2(debug__user, tp, PRU_SHUTDOWN);
 825         INP_WUNLOCK(inp);
 826         INP_INFO_RUNLOCK(&V_tcbinfo);
 827
 828         return (error);
 829 }
 830
 831 /*
 832  * After a receive, possibly send window update to peer.
 833  */
 834 static int
 835 tcp_usr_rcvd(struct socket *so, int flags)
 836 {
 837         struct inpcb *inp;
 838         struct tcpcb *tp = NULL;
 839         int error = 0;
 840
 841         TCPDEBUG0;
 842         inp = sotoinpcb(so);
 843         KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
 844         INP_WLOCK(inp);
 845         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 846                 error = ECONNRESET;
 847                 goto out;
 848         }
 849         tp = intotcpcb(inp);
 850         TCPDEBUG1();
 851         /*
 852          * For passively-created TFO connections, don't attempt a window
 853          * update while still in SYN_RECEIVED as this may trigger an early
 854          * SYN|ACK.  It is preferable to have the SYN|ACK be sent along with
 855          * application response data, or failing that, when the DELACK timer
 856          * expires.
 857          */
 858         if (IS_FASTOPEN(tp->t_flags) &&
 859             (tp->t_state == TCPS_SYN_RECEIVED))
 860                 goto out;
 861 #ifdef TCP_OFFLOAD
 862         if (tp->t_flags & TF_TOE)
 863                 tcp_offload_rcvd(tp);
 864         else
 865 #endif
 866         tp->t_fb->tfb_tcp_output(tp);
 867
 868 out:
 869         TCPDEBUG2(PRU_RCVD);
 870         TCP_PROBE2(debug__user, tp, PRU_RCVD);
 871         INP_WUNLOCK(inp);
 872         return (error);
 873 }
 874
 875 /*
 876  * Do a send by putting data in output queue and updating urgent
 877  * marker if URG set.  Possibly send more data.  Unlike the other
 878  * pru_*() routines, the mbuf chains are our responsibility.  We
 879  * must either enqueue them or free them.  The other pru_* routines
 880  * generally are caller-frees.
 881  */
 882 static int
 883 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
 884     struct sockaddr *nam, struct mbuf *control, struct thread *td)
 885 {
 886         int error = 0;
 887         struct inpcb *inp;
 888         struct tcpcb *tp = NULL;
 889 #ifdef INET6
 890         int isipv6;
 891 #endif
 892         TCPDEBUG0;
 893
 894         /*
 895          * We require the pcbinfo lock if we will close the socket as part of
 896          * this call.
 897          */
 898         if (flags & PRUS_EOF)
 899                 INP_INFO_RLOCK(&V_tcbinfo);
 900         inp = sotoinpcb(so);
 901         KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
 902         INP_WLOCK(inp);
 903         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 904                 if (control)
 905                         m_freem(control);
 906                 /*
 907                  * In case of PRUS_NOTREADY, tcp_usr_ready() is responsible
 908                  * for freeing memory.
 909                  */
 910                 if (m && (flags & PRUS_NOTREADY) == 0)
 911                         m_freem(m);
 912                 error = ECONNRESET;
 913                 goto out;
 914         }
 915 #ifdef INET6
 916         isipv6 = nam && nam->sa_family == AF_INET6;
 917 #endif /* INET6 */
 918         tp = intotcpcb(inp);
 919         TCPDEBUG1();
 920         if (control) {
 921                 /* TCP doesn't do control messages (rights, creds, etc) */
 922                 if (control->m_len) {
 923                         m_freem(control);
 924                         if (m)
 925                                 m_freem(m);
 926                         error = EINVAL;
 927                         goto out;
 928                 }
 929                 m_freem(control);       /* empty control, just free it */
 930         }
 931         if (!(flags & PRUS_OOB)) {
 932                 sbappendstream(&so->so_snd, m, flags);
 933                 if (nam && tp->t_state < TCPS_SYN_SENT) {
 934                         /*
 935                          * Do implied connect if not yet connected,
 936                          * initialize window to default value, and
 937                          * initialize maxseg using peer's cached MSS.
 938                          */
 939 #ifdef INET6
 940                         if (isipv6)
 941                                 error = tcp6_connect(tp, nam, td);
 942 #endif /* INET6 */
 943 #if defined(INET6) && defined(INET)
 944                         else
 945 #endif
 946 #ifdef INET
 947                                 error = tcp_connect(tp, nam, td);
 948 #endif
 949                         if (error)
 950                                 goto out;
 951                         if (IS_FASTOPEN(tp->t_flags))
 952                                 tcp_fastopen_connect(tp);
 953                         else {
 954                                 tp->snd_wnd = TTCP_CLIENT_SND_WND;
 955                                 tcp_mss(tp, -1);
 956                         }
 957                 }
 958                 if (flags & PRUS_EOF) {
 959                         /*
 960                          * Close the send side of the connection after
 961                          * the data is sent.
 962                          */
 963                         INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
 964                         socantsendmore(so);
 965                         tcp_usrclosed(tp);
 966                 }
 967                 if (!(inp->inp_flags & INP_DROPPED) &&
 968                     !(flags & PRUS_NOTREADY)) {
 969                         if (flags & PRUS_MORETOCOME)
 970                                 tp->t_flags |= TF_MORETOCOME;
 971                         error = tp->t_fb->tfb_tcp_output(tp);
 972                         if (flags & PRUS_MORETOCOME)
 973                                 tp->t_flags &= ~TF_MORETOCOME;
 974                 }
 975         } else {
 976                 /*
 977                  * XXXRW: PRUS_EOF not implemented with PRUS_OOB?
 978                  */
 979                 SOCKBUF_LOCK(&so->so_snd);
 980                 if (sbspace(&so->so_snd) < -512) {
 981                         SOCKBUF_UNLOCK(&so->so_snd);
 982                         m_freem(m);
 983                         error = ENOBUFS;
 984                         goto out;
 985                 }
 986                 /*
 987                  * According to RFC961 (Assigned Protocols),
 988                  * the urgent pointer points to the last octet
 989                  * of urgent data.  We continue, however,
 990                  * to consider it to indicate the first octet
 991                  * of data past the urgent section.
 992                  * Otherwise, snd_up should be one lower.
 993                  */
 994                 sbappendstream_locked(&so->so_snd, m, flags);
 995                 SOCKBUF_UNLOCK(&so->so_snd);
 996                 if (nam && tp->t_state < TCPS_SYN_SENT) {
 997                         /*
 998                          * Do implied connect if not yet connected,
 999                          * initialize window to default value, and
1000                          * initialize maxseg using peer's cached MSS.
1001                          */
1002
1003                         /*
1004                          * Not going to contemplate SYN|URG
1005                          */
1006                         if (IS_FASTOPEN(tp->t_flags))
1007                                 tp->t_flags &= ~TF_FASTOPEN;
1008 #ifdef INET6
1009                         if (isipv6)
1010                                 error = tcp6_connect(tp, nam, td);
1011 #endif /* INET6 */
1012 #if defined(INET6) && defined(INET)
1013                         else
1014 #endif
1015 #ifdef INET
1016                                 error = tcp_connect(tp, nam, td);
1017 #endif
1018                         if (error)
1019                                 goto out;
1020                         tp->snd_wnd = TTCP_CLIENT_SND_WND;
1021                         tcp_mss(tp, -1);
1022                 }
1023                 tp->snd_up = tp->snd_una + sbavail(&so->so_snd);
1024                 if (!(flags & PRUS_NOTREADY)) {
1025                         tp->t_flags |= TF_FORCEDATA;
1026                         error = tp->t_fb->tfb_tcp_output(tp);
1027                         tp->t_flags &= ~TF_FORCEDATA;
1028                 }
1029         }
1030         TCP_LOG_EVENT(tp, NULL,
1031             &inp->inp_socket->so_rcv,
1032             &inp->inp_socket->so_snd,
1033             TCP_LOG_USERSEND, error,
1034             0, NULL, false);
1035 out:
1036         TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
1037                   ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1038         TCP_PROBE2(debug__user, tp, (flags & PRUS_OOB) ? PRU_SENDOOB :
1039                    ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1040         INP_WUNLOCK(inp);
1041         if (flags & PRUS_EOF)
1042                 INP_INFO_RUNLOCK(&V_tcbinfo);
1043         return (error);
1044 }
1045
1046 static int
1047 tcp_usr_ready(struct socket *so, struct mbuf *m, int count)
1048 {
1049         struct inpcb *inp;
1050         struct tcpcb *tp;
1051         int error;
1052
1053         inp = sotoinpcb(so);
1054         INP_WLOCK(inp);
1055         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1056                 INP_WUNLOCK(inp);
1057                 for (int i = 0; i < count; i++)
1058                         m = m_free(m);
1059                 return (ECONNRESET);
1060         }
1061         tp = intotcpcb(inp);
1062
1063         SOCKBUF_LOCK(&so->so_snd);
1064         error = sbready(&so->so_snd, m, count);
1065         SOCKBUF_UNLOCK(&so->so_snd);
1066         if (error == 0)
1067                 error = tp->t_fb->tfb_tcp_output(tp);
1068         INP_WUNLOCK(inp);
1069
1070         return (error);
1071 }
1072
1073 /*
1074  * Abort the TCP.  Drop the connection abruptly.
1075  */
1076 static void
1077 tcp_usr_abort(struct socket *so)
1078 {
1079         struct inpcb *inp;
1080         struct tcpcb *tp = NULL;
1081         TCPDEBUG0;
1082
1083         inp = sotoinpcb(so);
1084         KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
1085
1086         INP_INFO_RLOCK(&V_tcbinfo);
1087         INP_WLOCK(inp);
1088         KASSERT(inp->inp_socket != NULL,
1089             ("tcp_usr_abort: inp_socket == NULL"));
1090
1091         /*
1092          * If we still have full TCP state, and we're not dropped, drop.
1093          */
1094         if (!(inp->inp_flags & INP_TIMEWAIT) &&
1095             !(inp->inp_flags & INP_DROPPED)) {
1096                 tp = intotcpcb(inp);
1097                 TCPDEBUG1();
1098                 tp = tcp_drop(tp, ECONNABORTED);
1099                 if (tp == NULL)
1100                         goto dropped;
1101                 TCPDEBUG2(PRU_ABORT);
1102                 TCP_PROBE2(debug__user, tp, PRU_ABORT);
1103         }
1104         if (!(inp->inp_flags & INP_DROPPED)) {
1105                 SOCK_LOCK(so);
1106                 so->so_state |= SS_PROTOREF;
1107                 SOCK_UNLOCK(so);
1108                 inp->inp_flags |= INP_SOCKREF;
1109         }
1110         INP_WUNLOCK(inp);
1111 dropped:
1112         INP_INFO_RUNLOCK(&V_tcbinfo);
1113 }
1114
1115 /*
1116  * TCP socket is closed.  Start friendly disconnect.
1117  */
1118 static void
1119 tcp_usr_close(struct socket *so)
1120 {
1121         struct inpcb *inp;
1122         struct tcpcb *tp = NULL;
1123         TCPDEBUG0;
1124
1125         inp = sotoinpcb(so);
1126         KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
1127
1128         INP_INFO_RLOCK(&V_tcbinfo);
1129         INP_WLOCK(inp);
1130         KASSERT(inp->inp_socket != NULL,
1131             ("tcp_usr_close: inp_socket == NULL"));
1132
1133         /*
1134          * If we still have full TCP state, and we're not dropped, initiate
1135          * a disconnect.
1136          */
1137         if (!(inp->inp_flags & INP_TIMEWAIT) &&
1138             !(inp->inp_flags & INP_DROPPED)) {
1139                 tp = intotcpcb(inp);
1140                 TCPDEBUG1();
1141                 tcp_disconnect(tp);
1142                 TCPDEBUG2(PRU_CLOSE);
1143                 TCP_PROBE2(debug__user, tp, PRU_CLOSE);
1144         }
1145         if (!(inp->inp_flags & INP_DROPPED)) {
1146                 SOCK_LOCK(so);
1147                 so->so_state |= SS_PROTOREF;
1148                 SOCK_UNLOCK(so);
1149                 inp->inp_flags |= INP_SOCKREF;
1150         }
1151         INP_WUNLOCK(inp);
1152         INP_INFO_RUNLOCK(&V_tcbinfo);
1153 }
1154
1155 /*
1156  * Receive out-of-band data.
1157  */
1158 static int
1159 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
1160 {
1161         int error = 0;
1162         struct inpcb *inp;
1163         struct tcpcb *tp = NULL;
1164
1165         TCPDEBUG0;
1166         inp = sotoinpcb(so);
1167         KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
1168         INP_WLOCK(inp);
1169         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1170                 error = ECONNRESET;
1171                 goto out;
1172         }
1173         tp = intotcpcb(inp);
1174         TCPDEBUG1();
1175         if ((so->so_oobmark == 0 &&
1176              (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
1177             so->so_options & SO_OOBINLINE ||
1178             tp->t_oobflags & TCPOOB_HADDATA) {
1179                 error = EINVAL;
1180                 goto out;
1181         }
1182         if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1183                 error = EWOULDBLOCK;
1184                 goto out;
1185         }
1186         m->m_len = 1;
1187         *mtod(m, caddr_t) = tp->t_iobc;
1188         if ((flags & MSG_PEEK) == 0)
1189                 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1190
1191 out:
1192         TCPDEBUG2(PRU_RCVOOB);
1193         TCP_PROBE2(debug__user, tp, PRU_RCVOOB);
1194         INP_WUNLOCK(inp);
1195         return (error);
1196 }
1197
1198 #ifdef INET
1199 struct pr_usrreqs tcp_usrreqs = {
1200         .pru_abort =            tcp_usr_abort,
1201         .pru_accept =           tcp_usr_accept,
1202         .pru_attach =           tcp_usr_attach,
1203         .pru_bind =             tcp_usr_bind,
1204         .pru_connect =          tcp_usr_connect,
1205         .pru_control =          in_control,
1206         .pru_detach =           tcp_usr_detach,
1207         .pru_disconnect =       tcp_usr_disconnect,
1208         .pru_listen =           tcp_usr_listen,
1209         .pru_peeraddr =         in_getpeeraddr,
1210         .pru_rcvd =             tcp_usr_rcvd,
1211         .pru_rcvoob =           tcp_usr_rcvoob,
1212         .pru_send =             tcp_usr_send,
1213         .pru_ready =            tcp_usr_ready,
1214         .pru_shutdown =         tcp_usr_shutdown,
1215         .pru_sockaddr =         in_getsockaddr,
1216         .pru_sosetlabel =       in_pcbsosetlabel,
1217         .pru_close =            tcp_usr_close,
1218 };
1219 #endif /* INET */
1220
1221 #ifdef INET6
1222 struct pr_usrreqs tcp6_usrreqs = {
1223         .pru_abort =            tcp_usr_abort,
1224         .pru_accept =           tcp6_usr_accept,
1225         .pru_attach =           tcp_usr_attach,
1226         .pru_bind =             tcp6_usr_bind,
1227         .pru_connect =          tcp6_usr_connect,
1228         .pru_control =          in6_control,
1229         .pru_detach =           tcp_usr_detach,
1230         .pru_disconnect =       tcp_usr_disconnect,
1231         .pru_listen =           tcp6_usr_listen,
1232         .pru_peeraddr =         in6_mapped_peeraddr,
1233         .pru_rcvd =             tcp_usr_rcvd,
1234         .pru_rcvoob =           tcp_usr_rcvoob,
1235         .pru_send =             tcp_usr_send,
1236         .pru_ready =            tcp_usr_ready,
1237         .pru_shutdown =         tcp_usr_shutdown,
1238         .pru_sockaddr =         in6_mapped_sockaddr,
1239         .pru_sosetlabel =       in_pcbsosetlabel,
1240         .pru_close =            tcp_usr_close,
1241 };
1242 #endif /* INET6 */
1243
1244 #ifdef INET
1245 /*
1246  * Common subroutine to open a TCP connection to remote host specified
1247  * by struct sockaddr_in in mbuf *nam.  Call in_pcbbind to assign a local
1248  * port number if needed.  Call in_pcbconnect_setup to do the routing and
1249  * to choose a local host address (interface).  If there is an existing
1250  * incarnation of the same connection in TIME-WAIT state and if the remote
1251  * host was sending CC options and if the connection duration was < MSL, then
1252  * truncate the previous TIME-WAIT state and proceed.
1253  * Initialize connection parameters and enter SYN-SENT state.
1254  */
1255 static int
1256 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1257 {
1258         struct inpcb *inp = tp->t_inpcb, *oinp;
1259         struct socket *so = inp->inp_socket;
1260         struct in_addr laddr;
1261         u_short lport;
1262         int error;
1263
1264         INP_WLOCK_ASSERT(inp);
1265         INP_HASH_WLOCK(&V_tcbinfo);
1266
1267         if (inp->inp_lport == 0) {
1268                 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1269                 if (error)
1270                         goto out;
1271         }
1272
1273         /*
1274          * Cannot simply call in_pcbconnect, because there might be an
1275          * earlier incarnation of this same connection still in
1276          * TIME_WAIT state, creating an ADDRINUSE error.
1277          */
1278         laddr = inp->inp_laddr;
1279         lport = inp->inp_lport;
1280         error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
1281             &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
1282         if (error && oinp == NULL)
1283                 goto out;
1284         if (oinp) {
1285                 error = EADDRINUSE;
1286                 goto out;
1287         }
1288         inp->inp_laddr = laddr;
1289         in_pcbrehash(inp);
1290         INP_HASH_WUNLOCK(&V_tcbinfo);
1291
1292         /*
1293          * Compute window scaling to request:
1294          * Scale to fit into sweet spot.  See tcp_syncache.c.
1295          * XXX: This should move to tcp_output().
1296          */
1297         while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1298             (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1299                 tp->request_r_scale++;
1300
1301         soisconnecting(so);
1302         TCPSTAT_INC(tcps_connattempt);
1303         tcp_state_change(tp, TCPS_SYN_SENT);
1304         tp->iss = tcp_new_isn(tp);
1305         tcp_sendseqinit(tp);
1306
1307         return 0;
1308
1309 out:
1310         INP_HASH_WUNLOCK(&V_tcbinfo);
1311         return (error);
1312 }
1313 #endif /* INET */
1314
1315 #ifdef INET6
1316 static int
1317 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1318 {
1319         struct inpcb *inp = tp->t_inpcb;
1320         int error;
1321
1322         INP_WLOCK_ASSERT(inp);
1323         INP_HASH_WLOCK(&V_tcbinfo);
1324
1325         if (inp->inp_lport == 0) {
1326                 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1327                 if (error)
1328                         goto out;
1329         }
1330         error = in6_pcbconnect(inp, nam, td->td_ucred);
1331         if (error != 0)
1332                 goto out;
1333         INP_HASH_WUNLOCK(&V_tcbinfo);
1334
1335         /* Compute window scaling to request.  */
1336         while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1337             (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1338                 tp->request_r_scale++;
1339
1340         soisconnecting(inp->inp_socket);
1341         TCPSTAT_INC(tcps_connattempt);
1342         tcp_state_change(tp, TCPS_SYN_SENT);
1343         tp->iss = tcp_new_isn(tp);
1344         tcp_sendseqinit(tp);
1345
1346         return 0;
1347
1348 out:
1349         INP_HASH_WUNLOCK(&V_tcbinfo);
1350         return error;
1351 }
1352 #endif /* INET6 */
1353
1354 /*
1355  * Export TCP internal state information via a struct tcp_info, based on the
1356  * Linux 2.6 API.  Not ABI compatible as our constants are mapped differently
1357  * (TCP state machine, etc).  We export all information using FreeBSD-native
1358  * constants -- for example, the numeric values for tcpi_state will differ
1359  * from Linux.
1360  */
1361 static void
1362 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1363 {
1364
1365         INP_WLOCK_ASSERT(tp->t_inpcb);
1366         bzero(ti, sizeof(*ti));
1367
1368         ti->tcpi_state = tp->t_state;
1369         if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
1370                 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1371         if (tp->t_flags & TF_SACK_PERMIT)
1372                 ti->tcpi_options |= TCPI_OPT_SACK;
1373         if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
1374                 ti->tcpi_options |= TCPI_OPT_WSCALE;
1375                 ti->tcpi_snd_wscale = tp->snd_scale;
1376                 ti->tcpi_rcv_wscale = tp->rcv_scale;
1377         }
1378         if (tp->t_flags & TF_ECN_PERMIT)
1379                 ti->tcpi_options |= TCPI_OPT_ECN;
1380
1381         ti->tcpi_rto = tp->t_rxtcur * tick;
1382         ti->tcpi_last_data_recv = ((uint32_t)ticks - tp->t_rcvtime) * tick;
1383         ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
1384         ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
1385
1386         ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1387         ti->tcpi_snd_cwnd = tp->snd_cwnd;
1388
1389         /*
1390          * FreeBSD-specific extension fields for tcp_info.
1391          */
1392         ti->tcpi_rcv_space = tp->rcv_wnd;
1393         ti->tcpi_rcv_nxt = tp->rcv_nxt;
1394         ti->tcpi_snd_wnd = tp->snd_wnd;
1395         ti->tcpi_snd_bwnd = 0;          /* Unused, kept for compat. */
1396         ti->tcpi_snd_nxt = tp->snd_nxt;
1397         ti->tcpi_snd_mss = tp->t_maxseg;
1398         ti->tcpi_rcv_mss = tp->t_maxseg;
1399         ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
1400         ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
1401         ti->tcpi_snd_zerowin = tp->t_sndzerowin;
1402 #ifdef TCP_OFFLOAD
1403         if (tp->t_flags & TF_TOE) {
1404                 ti->tcpi_options |= TCPI_OPT_TOE;
1405                 tcp_offload_tcp_info(tp, ti);
1406         }
1407 #endif
1408 }
1409
1410 /*
1411  * tcp_ctloutput() must drop the inpcb lock before performing copyin on
1412  * socket option arguments.  When it re-acquires the lock after the copy, it
1413  * has to revalidate that the connection is still valid for the socket
1414  * option.
1415  */
1416 #define INP_WLOCK_RECHECK_CLEANUP(inp, cleanup) do {                    \
1417         INP_WLOCK(inp);                                                 \
1418         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {            \
1419                 INP_WUNLOCK(inp);                                       \
1420                 cleanup;                                                \
1421                 return (ECONNRESET);                                    \
1422         }                                                               \
1423         tp = intotcpcb(inp);                                            \
1424 } while(0)
1425 #define INP_WLOCK_RECHECK(inp) INP_WLOCK_RECHECK_CLEANUP((inp), /* noop */)
1426
1427 int
1428 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1429 {
1430         int     error;
1431         struct  inpcb *inp;
1432         struct  tcpcb *tp;
1433         struct tcp_function_block *blk;
1434         struct tcp_function_set fsn;
1435
1436         error = 0;
1437         inp = sotoinpcb(so);
1438         KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
1439         INP_WLOCK(inp);
1440         if (sopt->sopt_level != IPPROTO_TCP) {
1441 #ifdef INET6
1442                 if (inp->inp_vflag & INP_IPV6PROTO) {
1443                         INP_WUNLOCK(inp);
1444                         error = ip6_ctloutput(so, sopt);
1445                 }
1446 #endif /* INET6 */
1447 #if defined(INET6) && defined(INET)
1448                 else
1449 #endif
1450 #ifdef INET
1451                 {
1452                         INP_WUNLOCK(inp);
1453                         error = ip_ctloutput(so, sopt);
1454                 }
1455 #endif
1456                 return (error);
1457         }
1458         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1459                 INP_WUNLOCK(inp);
1460                 return (ECONNRESET);
1461         }
1462         tp = intotcpcb(inp);
1463         /*
1464          * Protect the TCP option TCP_FUNCTION_BLK so
1465          * that a sub-function can *never* overwrite this.
1466          */
1467         if ((sopt->sopt_dir == SOPT_SET) &&
1468             (sopt->sopt_name == TCP_FUNCTION_BLK)) {
1469                 INP_WUNLOCK(inp);
1470                 error = sooptcopyin(sopt, &fsn, sizeof fsn,
1471                     sizeof fsn);
1472                 if (error)
1473                         return (error);
1474                 INP_WLOCK_RECHECK(inp);
1475                 blk = find_and_ref_tcp_functions(&fsn);
1476                 if (blk == NULL) {
1477                         INP_WUNLOCK(inp);
1478                         return (ENOENT);
1479                 }
1480                 if (tp->t_fb == blk) {
1481                         /* You already have this */
1482                         refcount_release(&blk->tfb_refcnt);
1483                         INP_WUNLOCK(inp);
1484                         return (0);
1485                 }
1486                 if (tp->t_state != TCPS_CLOSED) {
1487                         int error=EINVAL;
1488                         /*
1489                          * The user has advanced the state
1490                          * past the initial point, we may not
1491                          * be able to switch.
1492                          */
1493                         if (blk->tfb_tcp_handoff_ok != NULL) {
1494                                 /*
1495                                  * Does the stack provide a
1496                                  * query mechanism, if so it may
1497                                  * still be possible?
1498                                  */
1499                                 error = (*blk->tfb_tcp_handoff_ok)(tp);
1500                         }
1501                         if (error) {
1502                                 refcount_release(&blk->tfb_refcnt);
1503                                 INP_WUNLOCK(inp);
1504                                 return(error);
1505                         }
1506                 }
1507                 if (blk->tfb_flags & TCP_FUNC_BEING_REMOVED) {
1508                         refcount_release(&blk->tfb_refcnt);
1509                         INP_WUNLOCK(inp);
1510                         return (ENOENT);
1511                 }
1512                 /*
1513                  * Release the old refcnt, the
1514                  * lookup acquired a ref on the
1515                  * new one already.
1516                  */
1517                 if (tp->t_fb->tfb_tcp_fb_fini) {
1518                         /*
1519                          * Tell the stack to cleanup with 0 i.e.
1520                          * the tcb is not going away.
1521                          */
1522                         (*tp->t_fb->tfb_tcp_fb_fini)(tp, 0);
1523                 }
1524 #ifdef TCPHPTS
1525                 /* Assure that we are not on any hpts */
1526                 tcp_hpts_remove(tp->t_inpcb, HPTS_REMOVE_ALL);
1527 #endif
1528                 if (blk->tfb_tcp_fb_init) {
1529                         error = (*blk->tfb_tcp_fb_init)(tp);
1530                         if (error) {
1531                                 refcount_release(&blk->tfb_refcnt);
1532                                 if (tp->t_fb->tfb_tcp_fb_init) {
1533                                         if((*tp->t_fb->tfb_tcp_fb_init)(tp) != 0)  {
1534                                                 /* Fall back failed, drop the connection */
1535                                                 INP_WUNLOCK(inp);
1536                                                 soabort(so);
1537                                                 return(error);
1538                                         }
1539                                 }
1540                                 goto err_out;
1541                         }
1542                 }
1543                 refcount_release(&tp->t_fb->tfb_refcnt);
1544                 tp->t_fb = blk;
1545 #ifdef TCP_OFFLOAD
1546                 if (tp->t_flags & TF_TOE) {
1547                         tcp_offload_ctloutput(tp, sopt->sopt_dir,
1548                              sopt->sopt_name);
1549                 }
1550 #endif
1551 err_out:
1552                 INP_WUNLOCK(inp);
1553                 return (error);
1554         } else if ((sopt->sopt_dir == SOPT_GET) &&
1555             (sopt->sopt_name == TCP_FUNCTION_BLK)) {
1556                 strncpy(fsn.function_set_name, tp->t_fb->tfb_tcp_block_name,
1557                     TCP_FUNCTION_NAME_LEN_MAX);
1558                 fsn.function_set_name[TCP_FUNCTION_NAME_LEN_MAX - 1] = '\0';
1559                 fsn.pcbcnt = tp->t_fb->tfb_refcnt;
1560                 INP_WUNLOCK(inp);
1561                 error = sooptcopyout(sopt, &fsn, sizeof fsn);
1562                 return (error);
1563         }
1564         /* Pass in the INP locked, called must unlock it */
1565         return (tp->t_fb->tfb_tcp_ctloutput(so, sopt, inp, tp));
1566 }
1567
1568 /*
1569  * If this assert becomes untrue, we need to change the size of the buf
1570  * variable in tcp_default_ctloutput().
1571  */
1572 #ifdef CTASSERT
1573 CTASSERT(TCP_CA_NAME_MAX <= TCP_LOG_ID_LEN);
1574 CTASSERT(TCP_LOG_REASON_LEN <= TCP_LOG_ID_LEN);
1575 #endif
1576
1577 int
1578 tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp)
1579 {
1580         int     error, opt, optval;
1581         u_int   ui;
1582         struct  tcp_info ti;
1583         struct cc_algo *algo;
1584         char    *pbuf, buf[TCP_LOG_ID_LEN];
1585         size_t  len;
1586
1587         /*
1588          * For TCP_CCALGOOPT forward the control to CC module, for both
1589          * SOPT_SET and SOPT_GET.
1590          */
1591         switch (sopt->sopt_name) {
1592         case TCP_CCALGOOPT:
1593                 INP_WUNLOCK(inp);
1594                 pbuf = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK | M_ZERO);
1595                 error = sooptcopyin(sopt, pbuf, sopt->sopt_valsize,
1596                     sopt->sopt_valsize);
1597                 if (error) {
1598                         free(pbuf, M_TEMP);
1599                         return (error);
1600                 }
1601                 INP_WLOCK_RECHECK_CLEANUP(inp, free(pbuf, M_TEMP));
1602                 if (CC_ALGO(tp)->ctl_output != NULL)
1603                         error = CC_ALGO(tp)->ctl_output(tp->ccv, sopt, pbuf);
1604                 else
1605                         error = ENOENT;
1606                 INP_WUNLOCK(inp);
1607                 if (error == 0 && sopt->sopt_dir == SOPT_GET)
1608                         error = sooptcopyout(sopt, pbuf, sopt->sopt_valsize);
1609                 free(pbuf, M_TEMP);
1610                 return (error);
1611         }
1612
1613         switch (sopt->sopt_dir) {
1614         case SOPT_SET:
1615                 switch (sopt->sopt_name) {
1616 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE)
1617                 case TCP_MD5SIG:
1618                         if (!TCPMD5_ENABLED()) {
1619                                 INP_WUNLOCK(inp);
1620                                 return (ENOPROTOOPT);
1621                         }
1622                         error = TCPMD5_PCBCTL(inp, sopt);
1623                         if (error)
1624                                 return (error);
1625                         goto unlock_and_done;
1626 #endif /* IPSEC */
1627
1628                 case TCP_NODELAY:
1629                 case TCP_NOOPT:
1630                         INP_WUNLOCK(inp);
1631                         error = sooptcopyin(sopt, &optval, sizeof optval,
1632                             sizeof optval);
1633                         if (error)
1634                                 return (error);
1635
1636                         INP_WLOCK_RECHECK(inp);
1637                         switch (sopt->sopt_name) {
1638                         case TCP_NODELAY:
1639                                 opt = TF_NODELAY;
1640                                 break;
1641                         case TCP_NOOPT:
1642                                 opt = TF_NOOPT;
1643                                 break;
1644                         default:
1645                                 opt = 0; /* dead code to fool gcc */
1646                                 break;
1647                         }
1648
1649                         if (optval)
1650                                 tp->t_flags |= opt;
1651                         else
1652                                 tp->t_flags &= ~opt;
1653 unlock_and_done:
1654 #ifdef TCP_OFFLOAD
1655                         if (tp->t_flags & TF_TOE) {
1656                                 tcp_offload_ctloutput(tp, sopt->sopt_dir,
1657                                     sopt->sopt_name);
1658                         }
1659 #endif
1660                         INP_WUNLOCK(inp);
1661                         break;
1662
1663                 case TCP_NOPUSH:
1664                         INP_WUNLOCK(inp);
1665                         error = sooptcopyin(sopt, &optval, sizeof optval,
1666                             sizeof optval);
1667                         if (error)
1668                                 return (error);
1669
1670                         INP_WLOCK_RECHECK(inp);
1671                         if (optval)
1672                                 tp->t_flags |= TF_NOPUSH;
1673                         else if (tp->t_flags & TF_NOPUSH) {
1674                                 tp->t_flags &= ~TF_NOPUSH;
1675                                 if (TCPS_HAVEESTABLISHED(tp->t_state))
1676                                         error = tp->t_fb->tfb_tcp_output(tp);
1677                         }
1678                         goto unlock_and_done;
1679
1680                 case TCP_MAXSEG:
1681                         INP_WUNLOCK(inp);
1682                         error = sooptcopyin(sopt, &optval, sizeof optval,
1683                             sizeof optval);
1684                         if (error)
1685                                 return (error);
1686
1687                         INP_WLOCK_RECHECK(inp);
1688                         if (optval > 0 && optval <= tp->t_maxseg &&
1689                             optval + 40 >= V_tcp_minmss)
1690                                 tp->t_maxseg = optval;
1691                         else
1692                                 error = EINVAL;
1693                         goto unlock_and_done;
1694
1695                 case TCP_INFO:
1696                         INP_WUNLOCK(inp);
1697                         error = EINVAL;
1698                         break;
1699
1700                 case TCP_CONGESTION:
1701                         INP_WUNLOCK(inp);
1702                         error = sooptcopyin(sopt, buf, TCP_CA_NAME_MAX - 1, 1);
1703                         if (error)
1704                                 break;
1705                         buf[sopt->sopt_valsize] = '\0';
1706                         INP_WLOCK_RECHECK(inp);
1707                         CC_LIST_RLOCK();
1708                         STAILQ_FOREACH(algo, &cc_list, entries)
1709                                 if (strncmp(buf, algo->name,
1710                                     TCP_CA_NAME_MAX) == 0)
1711                                         break;
1712                         CC_LIST_RUNLOCK();
1713                         if (algo == NULL) {
1714                                 INP_WUNLOCK(inp);
1715                                 error = EINVAL;
1716                                 break;
1717                         }
1718                         /*
1719                          * We hold a write lock over the tcb so it's safe to
1720                          * do these things without ordering concerns.
1721                          */
1722                         if (CC_ALGO(tp)->cb_destroy != NULL)
1723                                 CC_ALGO(tp)->cb_destroy(tp->ccv);
1724                         CC_ALGO(tp) = algo;
1725                         /*
1726                          * If something goes pear shaped initialising the new
1727                          * algo, fall back to newreno (which does not
1728                          * require initialisation).
1729                          */
1730                         if (algo->cb_init != NULL &&
1731                             algo->cb_init(tp->ccv) != 0) {
1732                                 CC_ALGO(tp) = &newreno_cc_algo;
1733                                 /*
1734                                  * The only reason init should fail is
1735                                  * because of malloc.
1736                                  */
1737                                 error = ENOMEM;
1738                         }
1739                         INP_WUNLOCK(inp);
1740                         break;
1741
1742                 case TCP_KEEPIDLE:
1743                 case TCP_KEEPINTVL:
1744                 case TCP_KEEPINIT:
1745                         INP_WUNLOCK(inp);
1746                         error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
1747                         if (error)
1748                                 return (error);
1749
1750                         if (ui > (UINT_MAX / hz)) {
1751                                 error = EINVAL;
1752                                 break;
1753                         }
1754                         ui *= hz;
1755
1756                         INP_WLOCK_RECHECK(inp);
1757                         switch (sopt->sopt_name) {
1758                         case TCP_KEEPIDLE:
1759                                 tp->t_keepidle = ui;
1760                                 /*
1761                                  * XXX: better check current remaining
1762                                  * timeout and "merge" it with new value.
1763                                  */
1764                                 if ((tp->t_state > TCPS_LISTEN) &&
1765                                     (tp->t_state <= TCPS_CLOSING))
1766                                         tcp_timer_activate(tp, TT_KEEP,
1767                                             TP_KEEPIDLE(tp));
1768                                 break;
1769                         case TCP_KEEPINTVL:
1770                                 tp->t_keepintvl = ui;
1771                                 if ((tp->t_state == TCPS_FIN_WAIT_2) &&
1772                                     (TP_MAXIDLE(tp) > 0))
1773                                         tcp_timer_activate(tp, TT_2MSL,
1774                                             TP_MAXIDLE(tp));
1775                                 break;
1776                         case TCP_KEEPINIT:
1777                                 tp->t_keepinit = ui;
1778                                 if (tp->t_state == TCPS_SYN_RECEIVED ||
1779                                     tp->t_state == TCPS_SYN_SENT)
1780                                         tcp_timer_activate(tp, TT_KEEP,
1781                                             TP_KEEPINIT(tp));
1782                                 break;
1783                         }
1784                         goto unlock_and_done;
1785
1786                 case TCP_KEEPCNT:
1787                         INP_WUNLOCK(inp);
1788                         error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
1789                         if (error)
1790                                 return (error);
1791
1792                         INP_WLOCK_RECHECK(inp);
1793                         tp->t_keepcnt = ui;
1794                         if ((tp->t_state == TCPS_FIN_WAIT_2) &&
1795                             (TP_MAXIDLE(tp) > 0))
1796                                 tcp_timer_activate(tp, TT_2MSL,
1797                                     TP_MAXIDLE(tp));
1798                         goto unlock_and_done;
1799
1800 #ifdef TCPPCAP
1801                 case TCP_PCAP_OUT:
1802                 case TCP_PCAP_IN:
1803                         INP_WUNLOCK(inp);
1804                         error = sooptcopyin(sopt, &optval, sizeof optval,
1805                             sizeof optval);
1806                         if (error)
1807                                 return (error);
1808
1809                         INP_WLOCK_RECHECK(inp);
1810                         if (optval >= 0)
1811                                 tcp_pcap_set_sock_max(TCP_PCAP_OUT ?
1812                                         &(tp->t_outpkts) : &(tp->t_inpkts),
1813                                         optval);
1814                         else
1815                                 error = EINVAL;
1816                         goto unlock_and_done;
1817 #endif
1818
1819                 case TCP_FASTOPEN: {
1820                         struct tcp_fastopen tfo_optval;
1821
1822                         INP_WUNLOCK(inp);
1823                         if (!V_tcp_fastopen_client_enable &&
1824                             !V_tcp_fastopen_server_enable)
1825                                 return (EPERM);
1826
1827                         error = sooptcopyin(sopt, &tfo_optval,
1828                                     sizeof(tfo_optval), sizeof(int));
1829                         if (error)
1830                                 return (error);
1831
1832                         INP_WLOCK_RECHECK(inp);
1833                         if (tfo_optval.enable) {
1834                                 if (tp->t_state == TCPS_LISTEN) {
1835                                         if (!V_tcp_fastopen_server_enable) {
1836                                                 error = EPERM;
1837                                                 goto unlock_and_done;
1838                                         }
1839
1840                                         tp->t_flags |= TF_FASTOPEN;
1841                                         if (tp->t_tfo_pending == NULL)
1842                                                 tp->t_tfo_pending =
1843                                                     tcp_fastopen_alloc_counter();
1844                                 } else {
1845                                         /*
1846                                          * If a pre-shared key was provided,
1847                                          * stash it in the client cookie
1848                                          * field of the tcpcb for use during
1849                                          * connect.
1850                                          */
1851                                         if (sopt->sopt_valsize ==
1852                                             sizeof(tfo_optval)) {
1853                                                 memcpy(tp->t_tfo_cookie.client,
1854                                                        tfo_optval.psk,
1855                                                        TCP_FASTOPEN_PSK_LEN);
1856                                                 tp->t_tfo_client_cookie_len =
1857                                                     TCP_FASTOPEN_PSK_LEN;
1858                                         }
1859                                         tp->t_flags |= TF_FASTOPEN;
1860                                 }
1861                         } else
1862                                 tp->t_flags &= ~TF_FASTOPEN;
1863                         goto unlock_and_done;
1864                 }
1865
1866 #ifdef TCP_BLACKBOX
1867                 case TCP_LOG:
1868                         INP_WUNLOCK(inp);
1869                         error = sooptcopyin(sopt, &optval, sizeof optval,
1870                             sizeof optval);
1871                         if (error)
1872                                 return (error);
1873
1874                         INP_WLOCK_RECHECK(inp);
1875                         error = tcp_log_state_change(tp, optval);
1876                         goto unlock_and_done;
1877
1878                 case TCP_LOGBUF:
1879                         INP_WUNLOCK(inp);
1880                         error = EINVAL;
1881                         break;
1882
1883                 case TCP_LOGID:
1884                         INP_WUNLOCK(inp);
1885                         error = sooptcopyin(sopt, buf, TCP_LOG_ID_LEN - 1, 0);
1886                         if (error)
1887                                 break;
1888                         buf[sopt->sopt_valsize] = '\0';
1889                         INP_WLOCK_RECHECK(inp);
1890                         error = tcp_log_set_id(tp, buf);
1891                         /* tcp_log_set_id() unlocks the INP. */
1892                         break;
1893
1894                 case TCP_LOGDUMP:
1895                 case TCP_LOGDUMPID:
1896                         INP_WUNLOCK(inp);
1897                         error =
1898                             sooptcopyin(sopt, buf, TCP_LOG_REASON_LEN - 1, 0);
1899                         if (error)
1900                                 break;
1901                         buf[sopt->sopt_valsize] = '\0';
1902                         INP_WLOCK_RECHECK(inp);
1903                         if (sopt->sopt_name == TCP_LOGDUMP) {
1904                                 error = tcp_log_dump_tp_logbuf(tp, buf,
1905                                     M_WAITOK, true);
1906                                 INP_WUNLOCK(inp);
1907                         } else {
1908                                 tcp_log_dump_tp_bucket_logbufs(tp, buf);
1909                                 /*
1910                                  * tcp_log_dump_tp_bucket_logbufs() drops the
1911                                  * INP lock.
1912                                  */
1913                         }
1914                         break;
1915 #endif
1916
1917                 default:
1918                         INP_WUNLOCK(inp);
1919                         error = ENOPROTOOPT;
1920                         break;
1921                 }
1922                 break;
1923
1924         case SOPT_GET:
1925                 tp = intotcpcb(inp);
1926                 switch (sopt->sopt_name) {
1927 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE)
1928                 case TCP_MD5SIG:
1929                         if (!TCPMD5_ENABLED()) {
1930                                 INP_WUNLOCK(inp);
1931                                 return (ENOPROTOOPT);
1932                         }
1933                         error = TCPMD5_PCBCTL(inp, sopt);
1934                         break;
1935 #endif
1936
1937                 case TCP_NODELAY:
1938                         optval = tp->t_flags & TF_NODELAY;
1939                         INP_WUNLOCK(inp);
1940                         error = sooptcopyout(sopt, &optval, sizeof optval);
1941                         break;
1942                 case TCP_MAXSEG:
1943                         optval = tp->t_maxseg;
1944                         INP_WUNLOCK(inp);
1945                         error = sooptcopyout(sopt, &optval, sizeof optval);
1946                         break;
1947                 case TCP_NOOPT:
1948                         optval = tp->t_flags & TF_NOOPT;
1949                         INP_WUNLOCK(inp);
1950                         error = sooptcopyout(sopt, &optval, sizeof optval);
1951                         break;
1952                 case TCP_NOPUSH:
1953                         optval = tp->t_flags & TF_NOPUSH;
1954                         INP_WUNLOCK(inp);
1955                         error = sooptcopyout(sopt, &optval, sizeof optval);
1956                         break;
1957                 case TCP_INFO:
1958                         tcp_fill_info(tp, &ti);
1959                         INP_WUNLOCK(inp);
1960                         error = sooptcopyout(sopt, &ti, sizeof ti);
1961                         break;
1962                 case TCP_CONGESTION:
1963                         len = strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX);
1964                         INP_WUNLOCK(inp);
1965                         error = sooptcopyout(sopt, buf, len + 1);
1966                         break;
1967                 case TCP_KEEPIDLE:
1968                 case TCP_KEEPINTVL:
1969                 case TCP_KEEPINIT:
1970                 case TCP_KEEPCNT:
1971                         switch (sopt->sopt_name) {
1972                         case TCP_KEEPIDLE:
1973                                 ui = TP_KEEPIDLE(tp) / hz;
1974                                 break;
1975                         case TCP_KEEPINTVL:
1976                                 ui = TP_KEEPINTVL(tp) / hz;
1977                                 break;
1978                         case TCP_KEEPINIT:
1979                                 ui = TP_KEEPINIT(tp) / hz;
1980                                 break;
1981                         case TCP_KEEPCNT:
1982                                 ui = TP_KEEPCNT(tp);
1983                                 break;
1984                         }
1985                         INP_WUNLOCK(inp);
1986                         error = sooptcopyout(sopt, &ui, sizeof(ui));
1987                         break;
1988 #ifdef TCPPCAP
1989                 case TCP_PCAP_OUT:
1990                 case TCP_PCAP_IN:
1991                         optval = tcp_pcap_get_sock_max(TCP_PCAP_OUT ?
1992                                         &(tp->t_outpkts) : &(tp->t_inpkts));
1993                         INP_WUNLOCK(inp);
1994                         error = sooptcopyout(sopt, &optval, sizeof optval);
1995                         break;
1996 #endif
1997                 case TCP_FASTOPEN:
1998                         optval = tp->t_flags & TF_FASTOPEN;
1999                         INP_WUNLOCK(inp);
2000                         error = sooptcopyout(sopt, &optval, sizeof optval);
2001                         break;
2002 #ifdef TCP_BLACKBOX
2003                 case TCP_LOG:
2004                         optval = tp->t_logstate;
2005                         INP_WUNLOCK(inp);
2006                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2007                         break;
2008                 case TCP_LOGBUF:
2009                         /* tcp_log_getlogbuf() does INP_WUNLOCK(inp) */
2010                         error = tcp_log_getlogbuf(sopt, tp);
2011                         break;
2012                 case TCP_LOGID:
2013                         len = tcp_log_get_id(tp, buf);
2014                         INP_WUNLOCK(inp);
2015                         error = sooptcopyout(sopt, buf, len + 1);
2016                         break;
2017                 case TCP_LOGDUMP:
2018                 case TCP_LOGDUMPID:
2019                         INP_WUNLOCK(inp);
2020                         error = EINVAL;
2021                         break;
2022 #endif
2023                 default:
2024                         INP_WUNLOCK(inp);
2025                         error = ENOPROTOOPT;
2026                         break;
2027                 }
2028                 break;
2029         }
2030         return (error);
2031 }
2032 #undef INP_WLOCK_RECHECK
2033 #undef INP_WLOCK_RECHECK_CLEANUP
2034
2035 /*
2036  * Attach TCP protocol to socket, allocating
2037  * internet protocol control block, tcp control block,
2038  * bufer space, and entering LISTEN state if to accept connections.
2039  */
2040 static int
2041 tcp_attach(struct socket *so)
2042 {
2043         struct tcpcb *tp;
2044         struct inpcb *inp;
2045         int error;
2046
2047         if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
2048                 error = soreserve(so, V_tcp_sendspace, V_tcp_recvspace);
2049                 if (error)
2050                         return (error);
2051         }
2052         so->so_rcv.sb_flags |= SB_AUTOSIZE;
2053         so->so_snd.sb_flags |= SB_AUTOSIZE;
2054         INP_INFO_RLOCK(&V_tcbinfo);
2055         error = in_pcballoc(so, &V_tcbinfo);
2056         if (error) {
2057                 INP_INFO_RUNLOCK(&V_tcbinfo);
2058                 return (error);
2059         }
2060         inp = sotoinpcb(so);
2061 #ifdef INET6
2062         if (inp->inp_vflag & INP_IPV6PROTO) {
2063                 inp->inp_vflag |= INP_IPV6;
2064                 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
2065                         inp->inp_vflag |= INP_IPV4;
2066                 inp->in6p_hops = -1;    /* use kernel default */
2067         }
2068         else
2069 #endif
2070         inp->inp_vflag |= INP_IPV4;
2071         tp = tcp_newtcpcb(inp);
2072         if (tp == NULL) {
2073                 in_pcbdetach(inp);
2074                 in_pcbfree(inp);
2075                 INP_INFO_RUNLOCK(&V_tcbinfo);
2076                 return (ENOBUFS);
2077         }
2078         tp->t_state = TCPS_CLOSED;
2079         INP_WUNLOCK(inp);
2080         INP_INFO_RUNLOCK(&V_tcbinfo);
2081         TCPSTATES_INC(TCPS_CLOSED);
2082         return (0);
2083 }
2084
2085 /*
2086  * Initiate (or continue) disconnect.
2087  * If embryonic state, just send reset (once).
2088  * If in ``let data drain'' option and linger null, just drop.
2089  * Otherwise (hard), mark socket disconnecting and drop
2090  * current input data; switch states based on user close, and
2091  * send segment to peer (with FIN).
2092  */
2093 static void
2094 tcp_disconnect(struct tcpcb *tp)
2095 {
2096         struct inpcb *inp = tp->t_inpcb;
2097         struct socket *so = inp->inp_socket;
2098
2099         INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
2100         INP_WLOCK_ASSERT(inp);
2101
2102         /*
2103          * Neither tcp_close() nor tcp_drop() should return NULL, as the
2104          * socket is still open.
2105          */
2106         if (tp->t_state < TCPS_ESTABLISHED) {
2107                 tp = tcp_close(tp);
2108                 KASSERT(tp != NULL,
2109                     ("tcp_disconnect: tcp_close() returned NULL"));
2110         } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
2111                 tp = tcp_drop(tp, 0);
2112                 KASSERT(tp != NULL,
2113                     ("tcp_disconnect: tcp_drop() returned NULL"));
2114         } else {
2115                 soisdisconnecting(so);
2116                 sbflush(&so->so_rcv);
2117                 tcp_usrclosed(tp);
2118                 if (!(inp->inp_flags & INP_DROPPED))
2119                         tp->t_fb->tfb_tcp_output(tp);
2120         }
2121 }
2122
2123 /*
2124  * User issued close, and wish to trail through shutdown states:
2125  * if never received SYN, just forget it.  If got a SYN from peer,
2126  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
2127  * If already got a FIN from peer, then almost done; go to LAST_ACK
2128  * state.  In all other cases, have already sent FIN to peer (e.g.
2129  * after PRU_SHUTDOWN), and just have to play tedious game waiting
2130  * for peer to send FIN or not respond to keep-alives, etc.
2131  * We can let the user exit from the close as soon as the FIN is acked.
2132  */
2133 static void
2134 tcp_usrclosed(struct tcpcb *tp)
2135 {
2136
2137         INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
2138         INP_WLOCK_ASSERT(tp->t_inpcb);
2139
2140         switch (tp->t_state) {
2141         case TCPS_LISTEN:
2142 #ifdef TCP_OFFLOAD
2143                 tcp_offload_listen_stop(tp);
2144 #endif
2145                 tcp_state_change(tp, TCPS_CLOSED);
2146                 /* FALLTHROUGH */
2147         case TCPS_CLOSED:
2148                 tp = tcp_close(tp);
2149                 /*
2150                  * tcp_close() should never return NULL here as the socket is
2151                  * still open.
2152                  */
2153                 KASSERT(tp != NULL,
2154                     ("tcp_usrclosed: tcp_close() returned NULL"));
2155                 break;
2156
2157         case TCPS_SYN_SENT:
2158         case TCPS_SYN_RECEIVED:
2159                 tp->t_flags |= TF_NEEDFIN;
2160                 break;
2161
2162         case TCPS_ESTABLISHED:
2163                 tcp_state_change(tp, TCPS_FIN_WAIT_1);
2164                 break;
2165
2166         case TCPS_CLOSE_WAIT:
2167                 tcp_state_change(tp, TCPS_LAST_ACK);
2168                 break;
2169         }
2170         if (tp->t_state >= TCPS_FIN_WAIT_2) {
2171                 soisdisconnected(tp->t_inpcb->inp_socket);
2172                 /* Prevent the connection hanging in FIN_WAIT_2 forever. */
2173                 if (tp->t_state == TCPS_FIN_WAIT_2) {
2174                         int timeout;
2175
2176                         timeout = (tcp_fast_finwait2_recycle) ?
2177                             tcp_finwait2_timeout : TP_MAXIDLE(tp);
2178                         tcp_timer_activate(tp, TT_2MSL, timeout);
2179                 }
2180         }
2181 }
2182
2183 #ifdef DDB
2184 static void
2185 db_print_indent(int indent)
2186 {
2187         int i;
2188
2189         for (i = 0; i < indent; i++)
2190                 db_printf(" ");
2191 }
2192
2193 static void
2194 db_print_tstate(int t_state)
2195 {
2196
2197         switch (t_state) {
2198         case TCPS_CLOSED:
2199                 db_printf("TCPS_CLOSED");
2200                 return;
2201
2202         case TCPS_LISTEN:
2203                 db_printf("TCPS_LISTEN");
2204                 return;
2205
2206         case TCPS_SYN_SENT:
2207                 db_printf("TCPS_SYN_SENT");
2208                 return;
2209
2210         case TCPS_SYN_RECEIVED:
2211                 db_printf("TCPS_SYN_RECEIVED");
2212                 return;
2213
2214         case TCPS_ESTABLISHED:
2215                 db_printf("TCPS_ESTABLISHED");
2216                 return;
2217
2218         case TCPS_CLOSE_WAIT:
2219                 db_printf("TCPS_CLOSE_WAIT");
2220                 return;
2221
2222         case TCPS_FIN_WAIT_1:
2223                 db_printf("TCPS_FIN_WAIT_1");
2224                 return;
2225
2226         case TCPS_CLOSING:
2227                 db_printf("TCPS_CLOSING");
2228                 return;
2229
2230         case TCPS_LAST_ACK:
2231                 db_printf("TCPS_LAST_ACK");
2232                 return;
2233
2234         case TCPS_FIN_WAIT_2:
2235                 db_printf("TCPS_FIN_WAIT_2");
2236                 return;
2237
2238         case TCPS_TIME_WAIT:
2239                 db_printf("TCPS_TIME_WAIT");
2240                 return;
2241
2242         default:
2243                 db_printf("unknown");
2244                 return;
2245         }
2246 }
2247
2248 static void
2249 db_print_tflags(u_int t_flags)
2250 {
2251         int comma;
2252
2253         comma = 0;
2254         if (t_flags & TF_ACKNOW) {
2255                 db_printf("%sTF_ACKNOW", comma ? ", " : "");
2256                 comma = 1;
2257         }
2258         if (t_flags & TF_DELACK) {
2259                 db_printf("%sTF_DELACK", comma ? ", " : "");
2260                 comma = 1;
2261         }
2262         if (t_flags & TF_NODELAY) {
2263                 db_printf("%sTF_NODELAY", comma ? ", " : "");
2264                 comma = 1;
2265         }
2266         if (t_flags & TF_NOOPT) {
2267                 db_printf("%sTF_NOOPT", comma ? ", " : "");
2268                 comma = 1;
2269         }
2270         if (t_flags & TF_SENTFIN) {
2271                 db_printf("%sTF_SENTFIN", comma ? ", " : "");
2272                 comma = 1;
2273         }
2274         if (t_flags & TF_REQ_SCALE) {
2275                 db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
2276                 comma = 1;
2277         }
2278         if (t_flags & TF_RCVD_SCALE) {
2279                 db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
2280                 comma = 1;
2281         }
2282         if (t_flags & TF_REQ_TSTMP) {
2283                 db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
2284                 comma = 1;
2285         }
2286         if (t_flags & TF_RCVD_TSTMP) {
2287                 db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
2288                 comma = 1;
2289         }
2290         if (t_flags & TF_SACK_PERMIT) {
2291                 db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
2292                 comma = 1;
2293         }
2294         if (t_flags & TF_NEEDSYN) {
2295                 db_printf("%sTF_NEEDSYN", comma ? ", " : "");
2296                 comma = 1;
2297         }
2298         if (t_flags & TF_NEEDFIN) {
2299                 db_printf("%sTF_NEEDFIN", comma ? ", " : "");
2300                 comma = 1;
2301         }
2302         if (t_flags & TF_NOPUSH) {
2303                 db_printf("%sTF_NOPUSH", comma ? ", " : "");
2304                 comma = 1;
2305         }
2306         if (t_flags & TF_MORETOCOME) {
2307                 db_printf("%sTF_MORETOCOME", comma ? ", " : "");
2308                 comma = 1;
2309         }
2310         if (t_flags & TF_LQ_OVERFLOW) {
2311                 db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : "");
2312                 comma = 1;
2313         }
2314         if (t_flags & TF_LASTIDLE) {
2315                 db_printf("%sTF_LASTIDLE", comma ? ", " : "");
2316                 comma = 1;
2317         }
2318         if (t_flags & TF_RXWIN0SENT) {
2319                 db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
2320                 comma = 1;
2321         }
2322         if (t_flags & TF_FASTRECOVERY) {
2323                 db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
2324                 comma = 1;
2325         }
2326         if (t_flags & TF_CONGRECOVERY) {
2327                 db_printf("%sTF_CONGRECOVERY", comma ? ", " : "");
2328                 comma = 1;
2329         }
2330         if (t_flags & TF_WASFRECOVERY) {
2331                 db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
2332                 comma = 1;
2333         }
2334         if (t_flags & TF_SIGNATURE) {
2335                 db_printf("%sTF_SIGNATURE", comma ? ", " : "");
2336                 comma = 1;
2337         }
2338         if (t_flags & TF_FORCEDATA) {
2339                 db_printf("%sTF_FORCEDATA", comma ? ", " : "");
2340                 comma = 1;
2341         }
2342         if (t_flags & TF_TSO) {
2343                 db_printf("%sTF_TSO", comma ? ", " : "");
2344                 comma = 1;
2345         }
2346         if (t_flags & TF_ECN_PERMIT) {
2347                 db_printf("%sTF_ECN_PERMIT", comma ? ", " : "");
2348                 comma = 1;
2349         }
2350         if (t_flags & TF_FASTOPEN) {
2351                 db_printf("%sTF_FASTOPEN", comma ? ", " : "");
2352                 comma = 1;
2353         }
2354 }
2355
2356 static void
2357 db_print_toobflags(char t_oobflags)
2358 {
2359         int comma;
2360
2361         comma = 0;
2362         if (t_oobflags & TCPOOB_HAVEDATA) {
2363                 db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
2364                 comma = 1;
2365         }
2366         if (t_oobflags & TCPOOB_HADDATA) {
2367                 db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
2368                 comma = 1;
2369         }
2370 }
2371
2372 static void
2373 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
2374 {
2375
2376         db_print_indent(indent);
2377         db_printf("%s at %p\n", name, tp);
2378
2379         indent += 2;
2380
2381         db_print_indent(indent);
2382         db_printf("t_segq first: %p   t_segqlen: %d   t_dupacks: %d\n",
2383            LIST_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks);
2384
2385         db_print_indent(indent);
2386         db_printf("tt_rexmt: %p   tt_persist: %p   tt_keep: %p\n",
2387             &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep);
2388
2389         db_print_indent(indent);
2390         db_printf("tt_2msl: %p   tt_delack: %p   t_inpcb: %p\n", &tp->t_timers->tt_2msl,
2391             &tp->t_timers->tt_delack, tp->t_inpcb);
2392
2393         db_print_indent(indent);
2394         db_printf("t_state: %d (", tp->t_state);
2395         db_print_tstate(tp->t_state);
2396         db_printf(")\n");
2397
2398         db_print_indent(indent);
2399         db_printf("t_flags: 0x%x (", tp->t_flags);
2400         db_print_tflags(tp->t_flags);
2401         db_printf(")\n");
2402
2403         db_print_indent(indent);
2404         db_printf("snd_una: 0x%08x   snd_max: 0x%08x   snd_nxt: x0%08x\n",
2405             tp->snd_una, tp->snd_max, tp->snd_nxt);
2406
2407         db_print_indent(indent);
2408         db_printf("snd_up: 0x%08x   snd_wl1: 0x%08x   snd_wl2: 0x%08x\n",
2409            tp->snd_up, tp->snd_wl1, tp->snd_wl2);
2410
2411         db_print_indent(indent);
2412         db_printf("iss: 0x%08x   irs: 0x%08x   rcv_nxt: 0x%08x\n",
2413             tp->iss, tp->irs, tp->rcv_nxt);
2414
2415         db_print_indent(indent);
2416         db_printf("rcv_adv: 0x%08x   rcv_wnd: %u   rcv_up: 0x%08x\n",
2417             tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
2418
2419         db_print_indent(indent);
2420         db_printf("snd_wnd: %u   snd_cwnd: %u\n",
2421            tp->snd_wnd, tp->snd_cwnd);
2422
2423         db_print_indent(indent);
2424         db_printf("snd_ssthresh: %u   snd_recover: "
2425             "0x%08x\n", tp->snd_ssthresh, tp->snd_recover);
2426
2427         db_print_indent(indent);
2428         db_printf("t_rcvtime: %u   t_startime: %u\n",
2429             tp->t_rcvtime, tp->t_starttime);
2430
2431         db_print_indent(indent);
2432         db_printf("t_rttime: %u   t_rtsq: 0x%08x\n",
2433             tp->t_rtttime, tp->t_rtseq);
2434
2435         db_print_indent(indent);
2436         db_printf("t_rxtcur: %d   t_maxseg: %u   t_srtt: %d\n",
2437             tp->t_rxtcur, tp->t_maxseg, tp->t_srtt);
2438
2439         db_print_indent(indent);
2440         db_printf("t_rttvar: %d   t_rxtshift: %d   t_rttmin: %u   "
2441             "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin,
2442             tp->t_rttbest);
2443
2444         db_print_indent(indent);
2445         db_printf("t_rttupdated: %lu   max_sndwnd: %u   t_softerror: %d\n",
2446             tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
2447
2448         db_print_indent(indent);
2449         db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
2450         db_print_toobflags(tp->t_oobflags);
2451         db_printf(")   t_iobc: 0x%02x\n", tp->t_iobc);
2452
2453         db_print_indent(indent);
2454         db_printf("snd_scale: %u   rcv_scale: %u   request_r_scale: %u\n",
2455             tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
2456
2457         db_print_indent(indent);
2458         db_printf("ts_recent: %u   ts_recent_age: %u\n",
2459             tp->ts_recent, tp->ts_recent_age);
2460
2461         db_print_indent(indent);
2462         db_printf("ts_offset: %u   last_ack_sent: 0x%08x   snd_cwnd_prev: "
2463             "%u\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
2464
2465         db_print_indent(indent);
2466         db_printf("snd_ssthresh_prev: %u   snd_recover_prev: 0x%08x   "
2467             "t_badrxtwin: %u\n", tp->snd_ssthresh_prev,
2468             tp->snd_recover_prev, tp->t_badrxtwin);
2469
2470         db_print_indent(indent);
2471         db_printf("snd_numholes: %d  snd_holes first: %p\n",
2472             tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
2473
2474         db_print_indent(indent);
2475         db_printf("snd_fack: 0x%08x   rcv_numsacks: %d   sack_newdata: "
2476             "0x%08x\n", tp->snd_fack, tp->rcv_numsacks, tp->sack_newdata);
2477
2478         /* Skip sackblks, sackhint. */
2479
2480         db_print_indent(indent);
2481         db_printf("t_rttlow: %d   rfbuf_ts: %u   rfbuf_cnt: %d\n",
2482             tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
2483 }
2484
2485 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
2486 {
2487         struct tcpcb *tp;
2488
2489         if (!have_addr) {
2490                 db_printf("usage: show tcpcb <addr>\n");
2491                 return;
2492         }
2493         tp = (struct tcpcb *)addr;
2494
2495         db_print_tcpcb(tp, "tcpcb", 0);
2496 }
2497 #endif