sys/netinet/tcp_usrreq.c

   1 /*-
   2  * Copyright (c) 1982, 1986, 1988, 1993
   3  *      The Regents of the University of California.
   4  * Copyright (c) 2006-2007 Robert N. M. Watson
   5  * Copyright (c) 2010-2011 Juniper Networks, Inc.
   6  * All rights reserved.
   7  *
   8  * Portions of this software were developed by Robert N. M. Watson under
   9  * contract to Juniper Networks, Inc.
  10  *
  11  * Redistribution and use in source and binary forms, with or without
  12  * modification, are permitted provided that the following conditions
  13  * are met:
  14  * 1. Redistributions of source code must retain the above copyright
  15  *    notice, this list of conditions and the following disclaimer.
  16  * 2. Redistributions in binary form must reproduce the above copyright
  17  *    notice, this list of conditions and the following disclaimer in the
  18  *    documentation and/or other materials provided with the distribution.
  19  * 4. Neither the name of the University nor the names of its contributors
  20  *    may be used to endorse or promote products derived from this software
  21  *    without specific prior written permission.
  22  *
  23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  33  * SUCH DAMAGE.
  34  *
  35  *      From: @(#)tcp_usrreq.c  8.2 (Berkeley) 1/3/94
  36  */
  37
  38 #include <sys/cdefs.h>
  39 __FBSDID("$FreeBSD$");
  40
  41 #include "opt_ddb.h"
  42 #include "opt_inet.h"
  43 #include "opt_inet6.h"
  44 #include "opt_tcpdebug.h"
  45
  46 #include <sys/param.h>
  47 #include <sys/systm.h>
  48 #include <sys/limits.h>
  49 #include <sys/malloc.h>
  50 #include <sys/kernel.h>
  51 #include <sys/sysctl.h>
  52 #include <sys/mbuf.h>
  53 #ifdef INET6
  54 #include <sys/domain.h>
  55 #endif /* INET6 */
  56 #include <sys/socket.h>
  57 #include <sys/socketvar.h>
  58 #include <sys/protosw.h>
  59 #include <sys/proc.h>
  60 #include <sys/jail.h>
  61
  62 #ifdef DDB
  63 #include <ddb/ddb.h>
  64 #endif
  65
  66 #include <net/if.h>
  67 #include <net/if_var.h>
  68 #include <net/route.h>
  69 #include <net/vnet.h>
  70
  71 #include <netinet/cc.h>
  72 #include <netinet/in.h>
  73 #include <netinet/in_kdtrace.h>
  74 #include <netinet/in_pcb.h>
  75 #include <netinet/in_systm.h>
  76 #include <netinet/in_var.h>
  77 #include <netinet/ip_var.h>
  78 #ifdef INET6
  79 #include <netinet/ip6.h>
  80 #include <netinet6/in6_pcb.h>
  81 #include <netinet6/ip6_var.h>
  82 #include <netinet6/scope6_var.h>
  83 #endif
  84 #include <netinet/tcp_fsm.h>
  85 #include <netinet/tcp_seq.h>
  86 #include <netinet/tcp_timer.h>
  87 #include <netinet/tcp_var.h>
  88 #include <netinet/tcpip.h>
  89 #ifdef TCPDEBUG
  90 #include <netinet/tcp_debug.h>
  91 #endif
  92 #ifdef TCP_OFFLOAD
  93 #include <netinet/tcp_offload.h>
  94 #endif
  95
  96 /*
  97  * TCP protocol interface to socket abstraction.
  98  */
  99 static int      tcp_attach(struct socket *);
 100 #ifdef INET
 101 static int      tcp_connect(struct tcpcb *, struct sockaddr *,
 102                     struct thread *td);
 103 #endif /* INET */
 104 #ifdef INET6
 105 static int      tcp6_connect(struct tcpcb *, struct sockaddr *,
 106                     struct thread *td);
 107 #endif /* INET6 */
 108 static void     tcp_disconnect(struct tcpcb *);
 109 static void     tcp_usrclosed(struct tcpcb *);
 110 static void     tcp_fill_info(struct tcpcb *, struct tcp_info *);
 111
 112 #ifdef TCPDEBUG
 113 #define TCPDEBUG0       int ostate = 0
 114 #define TCPDEBUG1()     ostate = tp ? tp->t_state : 0
 115 #define TCPDEBUG2(req)  if (tp && (so->so_options & SO_DEBUG)) \
 116                                 tcp_trace(TA_USER, ostate, tp, 0, 0, req)
 117 #else
 118 #define TCPDEBUG0
 119 #define TCPDEBUG1()
 120 #define TCPDEBUG2(req)
 121 #endif
 122
 123 /*
 124  * TCP attaches to socket via pru_attach(), reserving space,
 125  * and an internet control block.
 126  */
 127 static int
 128 tcp_usr_attach(struct socket *so, int proto, struct thread *td)
 129 {
 130         struct inpcb *inp;
 131         struct tcpcb *tp = NULL;
 132         int error;
 133         TCPDEBUG0;
 134
 135         inp = sotoinpcb(so);
 136         KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL"));
 137         TCPDEBUG1();
 138
 139         error = tcp_attach(so);
 140         if (error)
 141                 goto out;
 142
 143         if ((so->so_options & SO_LINGER) && so->so_linger == 0)
 144                 so->so_linger = TCP_LINGERTIME;
 145
 146         inp = sotoinpcb(so);
 147         tp = intotcpcb(inp);
 148 out:
 149         TCPDEBUG2(PRU_ATTACH);
 150         TCP_PROBE2(debug__user, tp, PRU_ATTACH);
 151         return error;
 152 }
 153
 154 /*
 155  * tcp_detach is called when the socket layer loses its final reference
 156  * to the socket, be it a file descriptor reference, a reference from TCP,
 157  * etc.  At this point, there is only one case in which we will keep around
 158  * inpcb state: time wait.
 159  *
 160  * This function can probably be re-absorbed back into tcp_usr_detach() now
 161  * that there is a single detach path.
 162  */
 163 static void
 164 tcp_detach(struct socket *so, struct inpcb *inp)
 165 {
 166         struct tcpcb *tp;
 167
 168         INP_INFO_LOCK_ASSERT(&V_tcbinfo);
 169         INP_WLOCK_ASSERT(inp);
 170
 171         KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp"));
 172         KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so"));
 173
 174         tp = intotcpcb(inp);
 175
 176         if (inp->inp_flags & INP_TIMEWAIT) {
 177                 /*
 178                  * There are two cases to handle: one in which the time wait
 179                  * state is being discarded (INP_DROPPED), and one in which
 180                  * this connection will remain in timewait.  In the former,
 181                  * it is time to discard all state (except tcptw, which has
 182                  * already been discarded by the timewait close code, which
 183                  * should be further up the call stack somewhere).  In the
 184                  * latter case, we detach from the socket, but leave the pcb
 185                  * present until timewait ends.
 186                  *
 187                  * XXXRW: Would it be cleaner to free the tcptw here?
 188                  *
 189                  * Astute question indeed, from twtcp perspective there are
 190                  * three cases to consider:
 191                  *
 192                  * #1 tcp_detach is called at tcptw creation time by
 193                  *  tcp_twstart, then do not discard the newly created tcptw
 194                  *  and leave inpcb present until timewait ends
 195                  * #2 tcp_detach is called at timewait end (or reuse) by
 196                  *  tcp_twclose, then the tcptw has already been discarded
 197                  *  (or reused) and inpcb is freed here
 198                  * #3 tcp_detach is called() after timewait ends (or reuse)
 199                  *  (e.g. by soclose), then tcptw has already been discarded
 200                  *  (or reused) and inpcb is freed here
 201                  *
 202                  *  In all three cases the tcptw should not be freed here.
 203                  */
 204                 if (inp->inp_flags & INP_DROPPED) {
 205                         KASSERT(tp == NULL, ("tcp_detach: INP_TIMEWAIT && "
 206                             "INP_DROPPED && tp != NULL"));
 207                         in_pcbdetach(inp);
 208                         in_pcbfree(inp);
 209                 } else {
 210                         in_pcbdetach(inp);
 211                         INP_WUNLOCK(inp);
 212                 }
 213         } else {
 214                 /*
 215                  * If the connection is not in timewait, we consider two
 216                  * two conditions: one in which no further processing is
 217                  * necessary (dropped || embryonic), and one in which TCP is
 218                  * not yet done, but no longer requires the socket, so the
 219                  * pcb will persist for the time being.
 220                  *
 221                  * XXXRW: Does the second case still occur?
 222                  */
 223                 if (inp->inp_flags & INP_DROPPED ||
 224                     tp->t_state < TCPS_SYN_SENT) {
 225                         tcp_discardcb(tp);
 226                         in_pcbdetach(inp);
 227                         in_pcbfree(inp);
 228                 } else {
 229                         in_pcbdetach(inp);
 230                         INP_WUNLOCK(inp);
 231                 }
 232         }
 233 }
 234
 235 /*
 236  * pru_detach() detaches the TCP protocol from the socket.
 237  * If the protocol state is non-embryonic, then can't
 238  * do this directly: have to initiate a pru_disconnect(),
 239  * which may finish later; embryonic TCB's can just
 240  * be discarded here.
 241  */
 242 static void
 243 tcp_usr_detach(struct socket *so)
 244 {
 245         struct inpcb *inp;
 246         int rlock = 0;
 247
 248         inp = sotoinpcb(so);
 249         KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL"));
 250         if (!INP_INFO_WLOCKED(&V_tcbinfo)) {
 251                 INP_INFO_RLOCK(&V_tcbinfo);
 252                 rlock = 1;
 253         }
 254         INP_WLOCK(inp);
 255         KASSERT(inp->inp_socket != NULL,
 256             ("tcp_usr_detach: inp_socket == NULL"));
 257         tcp_detach(so, inp);
 258         if (rlock)
 259                 INP_INFO_RUNLOCK(&V_tcbinfo);
 260 }
 261
 262 #ifdef INET
 263 /*
 264  * Give the socket an address.
 265  */
 266 static int
 267 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 268 {
 269         int error = 0;
 270         struct inpcb *inp;
 271         struct tcpcb *tp = NULL;
 272         struct sockaddr_in *sinp;
 273
 274         sinp = (struct sockaddr_in *)nam;
 275         if (nam->sa_len != sizeof (*sinp))
 276                 return (EINVAL);
 277         /*
 278          * Must check for multicast addresses and disallow binding
 279          * to them.
 280          */
 281         if (sinp->sin_family == AF_INET &&
 282             IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
 283                 return (EAFNOSUPPORT);
 284
 285         TCPDEBUG0;
 286         inp = sotoinpcb(so);
 287         KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL"));
 288         INP_WLOCK(inp);
 289         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 290                 error = EINVAL;
 291                 goto out;
 292         }
 293         tp = intotcpcb(inp);
 294         TCPDEBUG1();
 295         INP_HASH_WLOCK(&V_tcbinfo);
 296         error = in_pcbbind(inp, nam, td->td_ucred);
 297         INP_HASH_WUNLOCK(&V_tcbinfo);
 298 out:
 299         TCPDEBUG2(PRU_BIND);
 300         TCP_PROBE2(debug__user, tp, PRU_BIND);
 301         INP_WUNLOCK(inp);
 302
 303         return (error);
 304 }
 305 #endif /* INET */
 306
 307 #ifdef INET6
 308 static int
 309 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 310 {
 311         int error = 0;
 312         struct inpcb *inp;
 313         struct tcpcb *tp = NULL;
 314         struct sockaddr_in6 *sin6p;
 315
 316         sin6p = (struct sockaddr_in6 *)nam;
 317         if (nam->sa_len != sizeof (*sin6p))
 318                 return (EINVAL);
 319         /*
 320          * Must check for multicast addresses and disallow binding
 321          * to them.
 322          */
 323         if (sin6p->sin6_family == AF_INET6 &&
 324             IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
 325                 return (EAFNOSUPPORT);
 326
 327         TCPDEBUG0;
 328         inp = sotoinpcb(so);
 329         KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL"));
 330         INP_WLOCK(inp);
 331         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 332                 error = EINVAL;
 333                 goto out;
 334         }
 335         tp = intotcpcb(inp);
 336         TCPDEBUG1();
 337         INP_HASH_WLOCK(&V_tcbinfo);
 338         inp->inp_vflag &= ~INP_IPV4;
 339         inp->inp_vflag |= INP_IPV6;
 340 #ifdef INET
 341         if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
 342                 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
 343                         inp->inp_vflag |= INP_IPV4;
 344                 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
 345                         struct sockaddr_in sin;
 346
 347                         in6_sin6_2_sin(&sin, sin6p);
 348                         inp->inp_vflag |= INP_IPV4;
 349                         inp->inp_vflag &= ~INP_IPV6;
 350                         error = in_pcbbind(inp, (struct sockaddr *)&sin,
 351                             td->td_ucred);
 352                         INP_HASH_WUNLOCK(&V_tcbinfo);
 353                         goto out;
 354                 }
 355         }
 356 #endif
 357         error = in6_pcbbind(inp, nam, td->td_ucred);
 358         INP_HASH_WUNLOCK(&V_tcbinfo);
 359 out:
 360         TCPDEBUG2(PRU_BIND);
 361         TCP_PROBE2(debug__user, tp, PRU_BIND);
 362         INP_WUNLOCK(inp);
 363         return (error);
 364 }
 365 #endif /* INET6 */
 366
 367 #ifdef INET
 368 /*
 369  * Prepare to accept connections.
 370  */
 371 static int
 372 tcp_usr_listen(struct socket *so, int backlog, struct thread *td)
 373 {
 374         int error = 0;
 375         struct inpcb *inp;
 376         struct tcpcb *tp = NULL;
 377
 378         TCPDEBUG0;
 379         inp = sotoinpcb(so);
 380         KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL"));
 381         INP_WLOCK(inp);
 382         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 383                 error = EINVAL;
 384                 goto out;
 385         }
 386         tp = intotcpcb(inp);
 387         TCPDEBUG1();
 388         SOCK_LOCK(so);
 389         error = solisten_proto_check(so);
 390         INP_HASH_WLOCK(&V_tcbinfo);
 391         if (error == 0 && inp->inp_lport == 0)
 392                 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
 393         INP_HASH_WUNLOCK(&V_tcbinfo);
 394         if (error == 0) {
 395                 tcp_state_change(tp, TCPS_LISTEN);
 396                 solisten_proto(so, backlog);
 397 #ifdef TCP_OFFLOAD
 398                 if ((so->so_options & SO_NO_OFFLOAD) == 0)
 399                         tcp_offload_listen_start(tp);
 400 #endif
 401         }
 402         SOCK_UNLOCK(so);
 403
 404 out:
 405         TCPDEBUG2(PRU_LISTEN);
 406         TCP_PROBE2(debug__user, tp, PRU_LISTEN);
 407         INP_WUNLOCK(inp);
 408         return (error);
 409 }
 410 #endif /* INET */
 411
 412 #ifdef INET6
 413 static int
 414 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td)
 415 {
 416         int error = 0;
 417         struct inpcb *inp;
 418         struct tcpcb *tp = NULL;
 419
 420         TCPDEBUG0;
 421         inp = sotoinpcb(so);
 422         KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL"));
 423         INP_WLOCK(inp);
 424         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 425                 error = EINVAL;
 426                 goto out;
 427         }
 428         tp = intotcpcb(inp);
 429         TCPDEBUG1();
 430         SOCK_LOCK(so);
 431         error = solisten_proto_check(so);
 432         INP_HASH_WLOCK(&V_tcbinfo);
 433         if (error == 0 && inp->inp_lport == 0) {
 434                 inp->inp_vflag &= ~INP_IPV4;
 435                 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
 436                         inp->inp_vflag |= INP_IPV4;
 437                 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
 438         }
 439         INP_HASH_WUNLOCK(&V_tcbinfo);
 440         if (error == 0) {
 441                 tcp_state_change(tp, TCPS_LISTEN);
 442                 solisten_proto(so, backlog);
 443 #ifdef TCP_OFFLOAD
 444                 if ((so->so_options & SO_NO_OFFLOAD) == 0)
 445                         tcp_offload_listen_start(tp);
 446 #endif
 447         }
 448         SOCK_UNLOCK(so);
 449
 450 out:
 451         TCPDEBUG2(PRU_LISTEN);
 452         TCP_PROBE2(debug__user, tp, PRU_LISTEN);
 453         INP_WUNLOCK(inp);
 454         return (error);
 455 }
 456 #endif /* INET6 */
 457
 458 #ifdef INET
 459 /*
 460  * Initiate connection to peer.
 461  * Create a template for use in transmissions on this connection.
 462  * Enter SYN_SENT state, and mark socket as connecting.
 463  * Start keep-alive timer, and seed output sequence space.
 464  * Send initial segment on connection.
 465  */
 466 static int
 467 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 468 {
 469         int error = 0;
 470         struct inpcb *inp;
 471         struct tcpcb *tp = NULL;
 472         struct sockaddr_in *sinp;
 473
 474         sinp = (struct sockaddr_in *)nam;
 475         if (nam->sa_len != sizeof (*sinp))
 476                 return (EINVAL);
 477         /*
 478          * Must disallow TCP ``connections'' to multicast addresses.
 479          */
 480         if (sinp->sin_family == AF_INET
 481             && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
 482                 return (EAFNOSUPPORT);
 483         if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0)
 484                 return (error);
 485
 486         TCPDEBUG0;
 487         inp = sotoinpcb(so);
 488         KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL"));
 489         INP_WLOCK(inp);
 490         if (inp->inp_flags & INP_TIMEWAIT) {
 491                 error = EADDRINUSE;
 492                 goto out;
 493         }
 494         if (inp->inp_flags & INP_DROPPED) {
 495                 error = ECONNREFUSED;
 496                 goto out;
 497         }
 498         tp = intotcpcb(inp);
 499         TCPDEBUG1();
 500         if ((error = tcp_connect(tp, nam, td)) != 0)
 501                 goto out;
 502 #ifdef TCP_OFFLOAD
 503         if (registered_toedevs > 0 &&
 504             (so->so_options & SO_NO_OFFLOAD) == 0 &&
 505             (error = tcp_offload_connect(so, nam)) == 0)
 506                 goto out;
 507 #endif
 508         tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
 509         error = tcp_output(tp);
 510 out:
 511         TCPDEBUG2(PRU_CONNECT);
 512         INP_WUNLOCK(inp);
 513         return (error);
 514 }
 515 #endif /* INET */
 516
 517 #ifdef INET6
 518 static int
 519 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 520 {
 521         int error = 0;
 522         struct inpcb *inp;
 523         struct tcpcb *tp = NULL;
 524         struct sockaddr_in6 *sin6p;
 525
 526         TCPDEBUG0;
 527
 528         sin6p = (struct sockaddr_in6 *)nam;
 529         if (nam->sa_len != sizeof (*sin6p))
 530                 return (EINVAL);
 531         /*
 532          * Must disallow TCP ``connections'' to multicast addresses.
 533          */
 534         if (sin6p->sin6_family == AF_INET6
 535             && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
 536                 return (EAFNOSUPPORT);
 537
 538         inp = sotoinpcb(so);
 539         KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL"));
 540         INP_WLOCK(inp);
 541         if (inp->inp_flags & INP_TIMEWAIT) {
 542                 error = EADDRINUSE;
 543                 goto out;
 544         }
 545         if (inp->inp_flags & INP_DROPPED) {
 546                 error = ECONNREFUSED;
 547                 goto out;
 548         }
 549         tp = intotcpcb(inp);
 550         TCPDEBUG1();
 551 #ifdef INET
 552         /*
 553          * XXXRW: Some confusion: V4/V6 flags relate to binding, and
 554          * therefore probably require the hash lock, which isn't held here.
 555          * Is this a significant problem?
 556          */
 557         if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
 558                 struct sockaddr_in sin;
 559
 560                 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
 561                         error = EINVAL;
 562                         goto out;
 563                 }
 564
 565                 in6_sin6_2_sin(&sin, sin6p);
 566                 inp->inp_vflag |= INP_IPV4;
 567                 inp->inp_vflag &= ~INP_IPV6;
 568                 if ((error = prison_remote_ip4(td->td_ucred,
 569                     &sin.sin_addr)) != 0)
 570                         goto out;
 571                 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
 572                         goto out;
 573 #ifdef TCP_OFFLOAD
 574                 if (registered_toedevs > 0 &&
 575                     (so->so_options & SO_NO_OFFLOAD) == 0 &&
 576                     (error = tcp_offload_connect(so, nam)) == 0)
 577                         goto out;
 578 #endif
 579                 error = tcp_output(tp);
 580                 goto out;
 581         }
 582 #endif
 583         inp->inp_vflag &= ~INP_IPV4;
 584         inp->inp_vflag |= INP_IPV6;
 585         inp->inp_inc.inc_flags |= INC_ISIPV6;
 586         if ((error = prison_remote_ip6(td->td_ucred, &sin6p->sin6_addr)) != 0)
 587                 goto out;
 588         if ((error = tcp6_connect(tp, nam, td)) != 0)
 589                 goto out;
 590 #ifdef TCP_OFFLOAD
 591         if (registered_toedevs > 0 &&
 592             (so->so_options & SO_NO_OFFLOAD) == 0 &&
 593             (error = tcp_offload_connect(so, nam)) == 0)
 594                 goto out;
 595 #endif
 596         tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
 597         error = tcp_output(tp);
 598
 599 out:
 600         TCPDEBUG2(PRU_CONNECT);
 601         TCP_PROBE2(debug__user, tp, PRU_CONNECT);
 602         INP_WUNLOCK(inp);
 603         return (error);
 604 }
 605 #endif /* INET6 */
 606
 607 /*
 608  * Initiate disconnect from peer.
 609  * If connection never passed embryonic stage, just drop;
 610  * else if don't need to let data drain, then can just drop anyways,
 611  * else have to begin TCP shutdown process: mark socket disconnecting,
 612  * drain unread data, state switch to reflect user close, and
 613  * send segment (e.g. FIN) to peer.  Socket will be really disconnected
 614  * when peer sends FIN and acks ours.
 615  *
 616  * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
 617  */
 618 static int
 619 tcp_usr_disconnect(struct socket *so)
 620 {
 621         struct inpcb *inp;
 622         struct tcpcb *tp = NULL;
 623         int error = 0;
 624
 625         TCPDEBUG0;
 626         INP_INFO_RLOCK(&V_tcbinfo);
 627         inp = sotoinpcb(so);
 628         KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
 629         INP_WLOCK(inp);
 630         if (inp->inp_flags & INP_TIMEWAIT)
 631                 goto out;
 632         if (inp->inp_flags & INP_DROPPED) {
 633                 error = ECONNRESET;
 634                 goto out;
 635         }
 636         tp = intotcpcb(inp);
 637         TCPDEBUG1();
 638         tcp_disconnect(tp);
 639 out:
 640         TCPDEBUG2(PRU_DISCONNECT);
 641         TCP_PROBE2(debug__user, tp, PRU_DISCONNECT);
 642         INP_WUNLOCK(inp);
 643         INP_INFO_RUNLOCK(&V_tcbinfo);
 644         return (error);
 645 }
 646
 647 #ifdef INET
 648 /*
 649  * Accept a connection.  Essentially all the work is done at higher levels;
 650  * just return the address of the peer, storing through addr.
 651  */
 652 static int
 653 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
 654 {
 655         int error = 0;
 656         struct inpcb *inp = NULL;
 657         struct tcpcb *tp = NULL;
 658         struct in_addr addr;
 659         in_port_t port = 0;
 660         TCPDEBUG0;
 661
 662         if (so->so_state & SS_ISDISCONNECTED)
 663                 return (ECONNABORTED);
 664
 665         inp = sotoinpcb(so);
 666         KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
 667         INP_WLOCK(inp);
 668         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 669                 error = ECONNABORTED;
 670                 goto out;
 671         }
 672         tp = intotcpcb(inp);
 673         TCPDEBUG1();
 674
 675         /*
 676          * We inline in_getpeeraddr and COMMON_END here, so that we can
 677          * copy the data of interest and defer the malloc until after we
 678          * release the lock.
 679          */
 680         port = inp->inp_fport;
 681         addr = inp->inp_faddr;
 682
 683 out:
 684         TCPDEBUG2(PRU_ACCEPT);
 685         TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
 686         INP_WUNLOCK(inp);
 687         if (error == 0)
 688                 *nam = in_sockaddr(port, &addr);
 689         return error;
 690 }
 691 #endif /* INET */
 692
 693 #ifdef INET6
 694 static int
 695 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
 696 {
 697         struct inpcb *inp = NULL;
 698         int error = 0;
 699         struct tcpcb *tp = NULL;
 700         struct in_addr addr;
 701         struct in6_addr addr6;
 702         in_port_t port = 0;
 703         int v4 = 0;
 704         TCPDEBUG0;
 705
 706         if (so->so_state & SS_ISDISCONNECTED)
 707                 return (ECONNABORTED);
 708
 709         inp = sotoinpcb(so);
 710         KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
 711         INP_INFO_RLOCK(&V_tcbinfo);
 712         INP_WLOCK(inp);
 713         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 714                 error = ECONNABORTED;
 715                 goto out;
 716         }
 717         tp = intotcpcb(inp);
 718         TCPDEBUG1();
 719
 720         /*
 721          * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
 722          * copy the data of interest and defer the malloc until after we
 723          * release the lock.
 724          */
 725         if (inp->inp_vflag & INP_IPV4) {
 726                 v4 = 1;
 727                 port = inp->inp_fport;
 728                 addr = inp->inp_faddr;
 729         } else {
 730                 port = inp->inp_fport;
 731                 addr6 = inp->in6p_faddr;
 732         }
 733
 734 out:
 735         TCPDEBUG2(PRU_ACCEPT);
 736         TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
 737         INP_WUNLOCK(inp);
 738         INP_INFO_RUNLOCK(&V_tcbinfo);
 739         if (error == 0) {
 740                 if (v4)
 741                         *nam = in6_v4mapsin6_sockaddr(port, &addr);
 742                 else
 743                         *nam = in6_sockaddr(port, &addr6);
 744         }
 745         return error;
 746 }
 747 #endif /* INET6 */
 748
 749 /*
 750  * Mark the connection as being incapable of further output.
 751  */
 752 static int
 753 tcp_usr_shutdown(struct socket *so)
 754 {
 755         int error = 0;
 756         struct inpcb *inp;
 757         struct tcpcb *tp = NULL;
 758
 759         TCPDEBUG0;
 760         INP_INFO_RLOCK(&V_tcbinfo);
 761         inp = sotoinpcb(so);
 762         KASSERT(inp != NULL, ("inp == NULL"));
 763         INP_WLOCK(inp);
 764         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 765                 error = ECONNRESET;
 766                 goto out;
 767         }
 768         tp = intotcpcb(inp);
 769         TCPDEBUG1();
 770         socantsendmore(so);
 771         tcp_usrclosed(tp);
 772         if (!(inp->inp_flags & INP_DROPPED))
 773                 error = tcp_output(tp);
 774
 775 out:
 776         TCPDEBUG2(PRU_SHUTDOWN);
 777         TCP_PROBE2(debug__user, tp, PRU_SHUTDOWN);
 778         INP_WUNLOCK(inp);
 779         INP_INFO_RUNLOCK(&V_tcbinfo);
 780
 781         return (error);
 782 }
 783
 784 /*
 785  * After a receive, possibly send window update to peer.
 786  */
 787 static int
 788 tcp_usr_rcvd(struct socket *so, int flags)
 789 {
 790         struct inpcb *inp;
 791         struct tcpcb *tp = NULL;
 792         int error = 0;
 793
 794         TCPDEBUG0;
 795         inp = sotoinpcb(so);
 796         KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
 797         INP_WLOCK(inp);
 798         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 799                 error = ECONNRESET;
 800                 goto out;
 801         }
 802         tp = intotcpcb(inp);
 803         TCPDEBUG1();
 804 #ifdef TCP_OFFLOAD
 805         if (tp->t_flags & TF_TOE)
 806                 tcp_offload_rcvd(tp);
 807         else
 808 #endif
 809         tcp_output(tp);
 810
 811 out:
 812         TCPDEBUG2(PRU_RCVD);
 813         TCP_PROBE2(debug__user, tp, PRU_RCVD);
 814         INP_WUNLOCK(inp);
 815         return (error);
 816 }
 817
 818 /*
 819  * Do a send by putting data in output queue and updating urgent
 820  * marker if URG set.  Possibly send more data.  Unlike the other
 821  * pru_*() routines, the mbuf chains are our responsibility.  We
 822  * must either enqueue them or free them.  The other pru_* routines
 823  * generally are caller-frees.
 824  */
 825 static int
 826 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
 827     struct sockaddr *nam, struct mbuf *control, struct thread *td)
 828 {
 829         int error = 0;
 830         struct inpcb *inp;
 831         struct tcpcb *tp = NULL;
 832 #ifdef INET6
 833         int isipv6;
 834 #endif
 835         TCPDEBUG0;
 836
 837         /*
 838          * We require the pcbinfo lock if we will close the socket as part of
 839          * this call.
 840          */
 841         if (flags & PRUS_EOF)
 842                 INP_INFO_RLOCK(&V_tcbinfo);
 843         inp = sotoinpcb(so);
 844         KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
 845         INP_WLOCK(inp);
 846         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 847                 if (control)
 848                         m_freem(control);
 849                 /*
 850                  * In case of PRUS_NOTREADY, tcp_usr_ready() is responsible
 851                  * for freeing memory.
 852                  */
 853                 if (m && (flags & PRUS_NOTREADY) == 0)
 854                         m_freem(m);
 855                 error = ECONNRESET;
 856                 goto out;
 857         }
 858 #ifdef INET6
 859         isipv6 = nam && nam->sa_family == AF_INET6;
 860 #endif /* INET6 */
 861         tp = intotcpcb(inp);
 862         TCPDEBUG1();
 863         if (control) {
 864                 /* TCP doesn't do control messages (rights, creds, etc) */
 865                 if (control->m_len) {
 866                         m_freem(control);
 867                         if (m)
 868                                 m_freem(m);
 869                         error = EINVAL;
 870                         goto out;
 871                 }
 872                 m_freem(control);       /* empty control, just free it */
 873         }
 874         if (!(flags & PRUS_OOB)) {
 875                 sbappendstream(&so->so_snd, m, flags);
 876                 if (nam && tp->t_state < TCPS_SYN_SENT) {
 877                         /*
 878                          * Do implied connect if not yet connected,
 879                          * initialize window to default value, and
 880                          * initialize maxseg/maxopd using peer's cached
 881                          * MSS.
 882                          */
 883 #ifdef INET6
 884                         if (isipv6)
 885                                 error = tcp6_connect(tp, nam, td);
 886 #endif /* INET6 */
 887 #if defined(INET6) && defined(INET)
 888                         else
 889 #endif
 890 #ifdef INET
 891                                 error = tcp_connect(tp, nam, td);
 892 #endif
 893                         if (error)
 894                                 goto out;
 895                         tp->snd_wnd = TTCP_CLIENT_SND_WND;
 896                         tcp_mss(tp, -1);
 897                 }
 898                 if (flags & PRUS_EOF) {
 899                         /*
 900                          * Close the send side of the connection after
 901                          * the data is sent.
 902                          */
 903                         INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
 904                         socantsendmore(so);
 905                         tcp_usrclosed(tp);
 906                 }
 907                 if (!(inp->inp_flags & INP_DROPPED) &&
 908                     !(flags & PRUS_NOTREADY)) {
 909                         if (flags & PRUS_MORETOCOME)
 910                                 tp->t_flags |= TF_MORETOCOME;
 911                         error = tcp_output(tp);
 912                         if (flags & PRUS_MORETOCOME)
 913                                 tp->t_flags &= ~TF_MORETOCOME;
 914                 }
 915         } else {
 916                 /*
 917                  * XXXRW: PRUS_EOF not implemented with PRUS_OOB?
 918                  */
 919                 SOCKBUF_LOCK(&so->so_snd);
 920                 if (sbspace(&so->so_snd) < -512) {
 921                         SOCKBUF_UNLOCK(&so->so_snd);
 922                         m_freem(m);
 923                         error = ENOBUFS;
 924                         goto out;
 925                 }
 926                 /*
 927                  * According to RFC961 (Assigned Protocols),
 928                  * the urgent pointer points to the last octet
 929                  * of urgent data.  We continue, however,
 930                  * to consider it to indicate the first octet
 931                  * of data past the urgent section.
 932                  * Otherwise, snd_up should be one lower.
 933                  */
 934                 sbappendstream_locked(&so->so_snd, m, flags);
 935                 SOCKBUF_UNLOCK(&so->so_snd);
 936                 if (nam && tp->t_state < TCPS_SYN_SENT) {
 937                         /*
 938                          * Do implied connect if not yet connected,
 939                          * initialize window to default value, and
 940                          * initialize maxseg/maxopd using peer's cached
 941                          * MSS.
 942                          */
 943 #ifdef INET6
 944                         if (isipv6)
 945                                 error = tcp6_connect(tp, nam, td);
 946 #endif /* INET6 */
 947 #if defined(INET6) && defined(INET)
 948                         else
 949 #endif
 950 #ifdef INET
 951                                 error = tcp_connect(tp, nam, td);
 952 #endif
 953                         if (error)
 954                                 goto out;
 955                         tp->snd_wnd = TTCP_CLIENT_SND_WND;
 956                         tcp_mss(tp, -1);
 957                 }
 958                 tp->snd_up = tp->snd_una + sbavail(&so->so_snd);
 959                 if (!(flags & PRUS_NOTREADY)) {
 960                         tp->t_flags |= TF_FORCEDATA;
 961                         error = tcp_output(tp);
 962                         tp->t_flags &= ~TF_FORCEDATA;
 963                 }
 964         }
 965 out:
 966         TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
 967                   ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
 968         TCP_PROBE2(debug__user, tp, (flags & PRUS_OOB) ? PRU_SENDOOB :
 969                    ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
 970         INP_WUNLOCK(inp);
 971         if (flags & PRUS_EOF)
 972                 INP_INFO_RUNLOCK(&V_tcbinfo);
 973         return (error);
 974 }
 975
 976 static int
 977 tcp_usr_ready(struct socket *so, struct mbuf *m, int count)
 978 {
 979         struct inpcb *inp;
 980         struct tcpcb *tp;
 981         int error;
 982
 983         inp = sotoinpcb(so);
 984         INP_WLOCK(inp);
 985         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 986                 INP_WUNLOCK(inp);
 987                 for (int i = 0; i < count; i++)
 988                         m = m_free(m);
 989                 return (ECONNRESET);
 990         }
 991         tp = intotcpcb(inp);
 992
 993         SOCKBUF_LOCK(&so->so_snd);
 994         error = sbready(&so->so_snd, m, count);
 995         SOCKBUF_UNLOCK(&so->so_snd);
 996         if (error == 0)
 997                 error = tcp_output(tp);
 998         INP_WUNLOCK(inp);
 999
1000         return (error);
1001 }
1002
1003 /*
1004  * Abort the TCP.  Drop the connection abruptly.
1005  */
1006 static void
1007 tcp_usr_abort(struct socket *so)
1008 {
1009         struct inpcb *inp;
1010         struct tcpcb *tp = NULL;
1011         TCPDEBUG0;
1012
1013         inp = sotoinpcb(so);
1014         KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
1015
1016         INP_INFO_RLOCK(&V_tcbinfo);
1017         INP_WLOCK(inp);
1018         KASSERT(inp->inp_socket != NULL,
1019             ("tcp_usr_abort: inp_socket == NULL"));
1020
1021         /*
1022          * If we still have full TCP state, and we're not dropped, drop.
1023          */
1024         if (!(inp->inp_flags & INP_TIMEWAIT) &&
1025             !(inp->inp_flags & INP_DROPPED)) {
1026                 tp = intotcpcb(inp);
1027                 TCPDEBUG1();
1028                 tcp_drop(tp, ECONNABORTED);
1029                 TCPDEBUG2(PRU_ABORT);
1030                 TCP_PROBE2(debug__user, tp, PRU_ABORT);
1031         }
1032         if (!(inp->inp_flags & INP_DROPPED)) {
1033                 SOCK_LOCK(so);
1034                 so->so_state |= SS_PROTOREF;
1035                 SOCK_UNLOCK(so);
1036                 inp->inp_flags |= INP_SOCKREF;
1037         }
1038         INP_WUNLOCK(inp);
1039         INP_INFO_RUNLOCK(&V_tcbinfo);
1040 }
1041
1042 /*
1043  * TCP socket is closed.  Start friendly disconnect.
1044  */
1045 static void
1046 tcp_usr_close(struct socket *so)
1047 {
1048         struct inpcb *inp;
1049         struct tcpcb *tp = NULL;
1050         TCPDEBUG0;
1051
1052         inp = sotoinpcb(so);
1053         KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
1054
1055         INP_INFO_RLOCK(&V_tcbinfo);
1056         INP_WLOCK(inp);
1057         KASSERT(inp->inp_socket != NULL,
1058             ("tcp_usr_close: inp_socket == NULL"));
1059
1060         /*
1061          * If we still have full TCP state, and we're not dropped, initiate
1062          * a disconnect.
1063          */
1064         if (!(inp->inp_flags & INP_TIMEWAIT) &&
1065             !(inp->inp_flags & INP_DROPPED)) {
1066                 tp = intotcpcb(inp);
1067                 TCPDEBUG1();
1068                 tcp_disconnect(tp);
1069                 TCPDEBUG2(PRU_CLOSE);
1070                 TCP_PROBE2(debug__user, tp, PRU_CLOSE);
1071         }
1072         if (!(inp->inp_flags & INP_DROPPED)) {
1073                 SOCK_LOCK(so);
1074                 so->so_state |= SS_PROTOREF;
1075                 SOCK_UNLOCK(so);
1076                 inp->inp_flags |= INP_SOCKREF;
1077         }
1078         INP_WUNLOCK(inp);
1079         INP_INFO_RUNLOCK(&V_tcbinfo);
1080 }
1081
1082 /*
1083  * Receive out-of-band data.
1084  */
1085 static int
1086 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
1087 {
1088         int error = 0;
1089         struct inpcb *inp;
1090         struct tcpcb *tp = NULL;
1091
1092         TCPDEBUG0;
1093         inp = sotoinpcb(so);
1094         KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
1095         INP_WLOCK(inp);
1096         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1097                 error = ECONNRESET;
1098                 goto out;
1099         }
1100         tp = intotcpcb(inp);
1101         TCPDEBUG1();
1102         if ((so->so_oobmark == 0 &&
1103              (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
1104             so->so_options & SO_OOBINLINE ||
1105             tp->t_oobflags & TCPOOB_HADDATA) {
1106                 error = EINVAL;
1107                 goto out;
1108         }
1109         if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1110                 error = EWOULDBLOCK;
1111                 goto out;
1112         }
1113         m->m_len = 1;
1114         *mtod(m, caddr_t) = tp->t_iobc;
1115         if ((flags & MSG_PEEK) == 0)
1116                 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1117
1118 out:
1119         TCPDEBUG2(PRU_RCVOOB);
1120         TCP_PROBE2(debug__user, tp, PRU_RCVOOB);
1121         INP_WUNLOCK(inp);
1122         return (error);
1123 }
1124
1125 #ifdef INET
1126 struct pr_usrreqs tcp_usrreqs = {
1127         .pru_abort =            tcp_usr_abort,
1128         .pru_accept =           tcp_usr_accept,
1129         .pru_attach =           tcp_usr_attach,
1130         .pru_bind =             tcp_usr_bind,
1131         .pru_connect =          tcp_usr_connect,
1132         .pru_control =          in_control,
1133         .pru_detach =           tcp_usr_detach,
1134         .pru_disconnect =       tcp_usr_disconnect,
1135         .pru_listen =           tcp_usr_listen,
1136         .pru_peeraddr =         in_getpeeraddr,
1137         .pru_rcvd =             tcp_usr_rcvd,
1138         .pru_rcvoob =           tcp_usr_rcvoob,
1139         .pru_send =             tcp_usr_send,
1140         .pru_ready =            tcp_usr_ready,
1141         .pru_shutdown =         tcp_usr_shutdown,
1142         .pru_sockaddr =         in_getsockaddr,
1143         .pru_sosetlabel =       in_pcbsosetlabel,
1144         .pru_close =            tcp_usr_close,
1145 };
1146 #endif /* INET */
1147
1148 #ifdef INET6
1149 struct pr_usrreqs tcp6_usrreqs = {
1150         .pru_abort =            tcp_usr_abort,
1151         .pru_accept =           tcp6_usr_accept,
1152         .pru_attach =           tcp_usr_attach,
1153         .pru_bind =             tcp6_usr_bind,
1154         .pru_connect =          tcp6_usr_connect,
1155         .pru_control =          in6_control,
1156         .pru_detach =           tcp_usr_detach,
1157         .pru_disconnect =       tcp_usr_disconnect,
1158         .pru_listen =           tcp6_usr_listen,
1159         .pru_peeraddr =         in6_mapped_peeraddr,
1160         .pru_rcvd =             tcp_usr_rcvd,
1161         .pru_rcvoob =           tcp_usr_rcvoob,
1162         .pru_send =             tcp_usr_send,
1163         .pru_ready =            tcp_usr_ready,
1164         .pru_shutdown =         tcp_usr_shutdown,
1165         .pru_sockaddr =         in6_mapped_sockaddr,
1166         .pru_sosetlabel =       in_pcbsosetlabel,
1167         .pru_close =            tcp_usr_close,
1168 };
1169 #endif /* INET6 */
1170
1171 #ifdef INET
1172 /*
1173  * Common subroutine to open a TCP connection to remote host specified
1174  * by struct sockaddr_in in mbuf *nam.  Call in_pcbbind to assign a local
1175  * port number if needed.  Call in_pcbconnect_setup to do the routing and
1176  * to choose a local host address (interface).  If there is an existing
1177  * incarnation of the same connection in TIME-WAIT state and if the remote
1178  * host was sending CC options and if the connection duration was < MSL, then
1179  * truncate the previous TIME-WAIT state and proceed.
1180  * Initialize connection parameters and enter SYN-SENT state.
1181  */
1182 static int
1183 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1184 {
1185         struct inpcb *inp = tp->t_inpcb, *oinp;
1186         struct socket *so = inp->inp_socket;
1187         struct in_addr laddr;
1188         u_short lport;
1189         int error;
1190
1191         INP_WLOCK_ASSERT(inp);
1192         INP_HASH_WLOCK(&V_tcbinfo);
1193
1194         if (inp->inp_lport == 0) {
1195                 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1196                 if (error)
1197                         goto out;
1198         }
1199
1200         /*
1201          * Cannot simply call in_pcbconnect, because there might be an
1202          * earlier incarnation of this same connection still in
1203          * TIME_WAIT state, creating an ADDRINUSE error.
1204          */
1205         laddr = inp->inp_laddr;
1206         lport = inp->inp_lport;
1207         error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
1208             &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
1209         if (error && oinp == NULL)
1210                 goto out;
1211         if (oinp) {
1212                 error = EADDRINUSE;
1213                 goto out;
1214         }
1215         inp->inp_laddr = laddr;
1216         in_pcbrehash(inp);
1217         INP_HASH_WUNLOCK(&V_tcbinfo);
1218
1219         /*
1220          * Compute window scaling to request:
1221          * Scale to fit into sweet spot.  See tcp_syncache.c.
1222          * XXX: This should move to tcp_output().
1223          */
1224         while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1225             (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1226                 tp->request_r_scale++;
1227
1228         soisconnecting(so);
1229         TCPSTAT_INC(tcps_connattempt);
1230         tcp_state_change(tp, TCPS_SYN_SENT);
1231         tp->iss = tcp_new_isn(tp);
1232         tcp_sendseqinit(tp);
1233
1234         return 0;
1235
1236 out:
1237         INP_HASH_WUNLOCK(&V_tcbinfo);
1238         return (error);
1239 }
1240 #endif /* INET */
1241
1242 #ifdef INET6
1243 static int
1244 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1245 {
1246         struct inpcb *inp = tp->t_inpcb;
1247         int error;
1248
1249         INP_WLOCK_ASSERT(inp);
1250         INP_HASH_WLOCK(&V_tcbinfo);
1251
1252         if (inp->inp_lport == 0) {
1253                 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1254                 if (error)
1255                         goto out;
1256         }
1257         error = in6_pcbconnect(inp, nam, td->td_ucred);
1258         if (error != 0)
1259                 goto out;
1260         INP_HASH_WUNLOCK(&V_tcbinfo);
1261
1262         /* Compute window scaling to request.  */
1263         while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1264             (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1265                 tp->request_r_scale++;
1266
1267         soisconnecting(inp->inp_socket);
1268         TCPSTAT_INC(tcps_connattempt);
1269         tcp_state_change(tp, TCPS_SYN_SENT);
1270         tp->iss = tcp_new_isn(tp);
1271         tcp_sendseqinit(tp);
1272
1273         return 0;
1274
1275 out:
1276         INP_HASH_WUNLOCK(&V_tcbinfo);
1277         return error;
1278 }
1279 #endif /* INET6 */
1280
1281 /*
1282  * Export TCP internal state information via a struct tcp_info, based on the
1283  * Linux 2.6 API.  Not ABI compatible as our constants are mapped differently
1284  * (TCP state machine, etc).  We export all information using FreeBSD-native
1285  * constants -- for example, the numeric values for tcpi_state will differ
1286  * from Linux.
1287  */
1288 static void
1289 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1290 {
1291
1292         INP_WLOCK_ASSERT(tp->t_inpcb);
1293         bzero(ti, sizeof(*ti));
1294
1295         ti->tcpi_state = tp->t_state;
1296         if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
1297                 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1298         if (tp->t_flags & TF_SACK_PERMIT)
1299                 ti->tcpi_options |= TCPI_OPT_SACK;
1300         if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
1301                 ti->tcpi_options |= TCPI_OPT_WSCALE;
1302                 ti->tcpi_snd_wscale = tp->snd_scale;
1303                 ti->tcpi_rcv_wscale = tp->rcv_scale;
1304         }
1305
1306         ti->tcpi_rto = tp->t_rxtcur * tick;
1307         ti->tcpi_last_data_recv = (long)(ticks - (int)tp->t_rcvtime) * tick;
1308         ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
1309         ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
1310
1311         ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1312         ti->tcpi_snd_cwnd = tp->snd_cwnd;
1313
1314         /*
1315          * FreeBSD-specific extension fields for tcp_info.
1316          */
1317         ti->tcpi_rcv_space = tp->rcv_wnd;
1318         ti->tcpi_rcv_nxt = tp->rcv_nxt;
1319         ti->tcpi_snd_wnd = tp->snd_wnd;
1320         ti->tcpi_snd_bwnd = 0;          /* Unused, kept for compat. */
1321         ti->tcpi_snd_nxt = tp->snd_nxt;
1322         ti->tcpi_snd_mss = tp->t_maxseg;
1323         ti->tcpi_rcv_mss = tp->t_maxseg;
1324         if (tp->t_flags & TF_TOE)
1325                 ti->tcpi_options |= TCPI_OPT_TOE;
1326         ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
1327         ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
1328         ti->tcpi_snd_zerowin = tp->t_sndzerowin;
1329 }
1330
1331 /*
1332  * tcp_ctloutput() must drop the inpcb lock before performing copyin on
1333  * socket option arguments.  When it re-acquires the lock after the copy, it
1334  * has to revalidate that the connection is still valid for the socket
1335  * option.
1336  */
1337 #define INP_WLOCK_RECHECK(inp) do {                                     \
1338         INP_WLOCK(inp);                                                 \
1339         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {            \
1340                 INP_WUNLOCK(inp);                                       \
1341                 return (ECONNRESET);                                    \
1342         }                                                               \
1343         tp = intotcpcb(inp);                                            \
1344 } while(0)
1345
1346 int
1347 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1348 {
1349         int     error, opt, optval;
1350         u_int   ui;
1351         struct  inpcb *inp;
1352         struct  tcpcb *tp;
1353         struct  tcp_info ti;
1354         char buf[TCP_CA_NAME_MAX];
1355         struct cc_algo *algo;
1356
1357         error = 0;
1358         inp = sotoinpcb(so);
1359         KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
1360         INP_WLOCK(inp);
1361         if (sopt->sopt_level != IPPROTO_TCP) {
1362 #ifdef INET6
1363                 if (inp->inp_vflag & INP_IPV6PROTO) {
1364                         INP_WUNLOCK(inp);
1365                         error = ip6_ctloutput(so, sopt);
1366                 }
1367 #endif /* INET6 */
1368 #if defined(INET6) && defined(INET)
1369                 else
1370 #endif
1371 #ifdef INET
1372                 {
1373                         INP_WUNLOCK(inp);
1374                         error = ip_ctloutput(so, sopt);
1375                 }
1376 #endif
1377                 return (error);
1378         }
1379         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1380                 INP_WUNLOCK(inp);
1381                 return (ECONNRESET);
1382         }
1383
1384         switch (sopt->sopt_dir) {
1385         case SOPT_SET:
1386                 switch (sopt->sopt_name) {
1387 #ifdef TCP_SIGNATURE
1388                 case TCP_MD5SIG:
1389                         INP_WUNLOCK(inp);
1390                         error = sooptcopyin(sopt, &optval, sizeof optval,
1391                             sizeof optval);
1392                         if (error)
1393                                 return (error);
1394
1395                         INP_WLOCK_RECHECK(inp);
1396                         if (optval > 0)
1397                                 tp->t_flags |= TF_SIGNATURE;
1398                         else
1399                                 tp->t_flags &= ~TF_SIGNATURE;
1400                         goto unlock_and_done;
1401 #endif /* TCP_SIGNATURE */
1402
1403                 case TCP_NODELAY:
1404                 case TCP_NOOPT:
1405                         INP_WUNLOCK(inp);
1406                         error = sooptcopyin(sopt, &optval, sizeof optval,
1407                             sizeof optval);
1408                         if (error)
1409                                 return (error);
1410
1411                         INP_WLOCK_RECHECK(inp);
1412                         switch (sopt->sopt_name) {
1413                         case TCP_NODELAY:
1414                                 opt = TF_NODELAY;
1415                                 break;
1416                         case TCP_NOOPT:
1417                                 opt = TF_NOOPT;
1418                                 break;
1419                         default:
1420                                 opt = 0; /* dead code to fool gcc */
1421                                 break;
1422                         }
1423
1424                         if (optval)
1425                                 tp->t_flags |= opt;
1426                         else
1427                                 tp->t_flags &= ~opt;
1428 unlock_and_done:
1429 #ifdef TCP_OFFLOAD
1430                         if (tp->t_flags & TF_TOE) {
1431                                 tcp_offload_ctloutput(tp, sopt->sopt_dir,
1432                                     sopt->sopt_name);
1433                         }
1434 #endif
1435                         INP_WUNLOCK(inp);
1436                         break;
1437
1438                 case TCP_NOPUSH:
1439                         INP_WUNLOCK(inp);
1440                         error = sooptcopyin(sopt, &optval, sizeof optval,
1441                             sizeof optval);
1442                         if (error)
1443                                 return (error);
1444
1445                         INP_WLOCK_RECHECK(inp);
1446                         if (optval)
1447                                 tp->t_flags |= TF_NOPUSH;
1448                         else if (tp->t_flags & TF_NOPUSH) {
1449                                 tp->t_flags &= ~TF_NOPUSH;
1450                                 if (TCPS_HAVEESTABLISHED(tp->t_state))
1451                                         error = tcp_output(tp);
1452                         }
1453                         goto unlock_and_done;
1454
1455                 case TCP_MAXSEG:
1456                         INP_WUNLOCK(inp);
1457                         error = sooptcopyin(sopt, &optval, sizeof optval,
1458                             sizeof optval);
1459                         if (error)
1460                                 return (error);
1461
1462                         INP_WLOCK_RECHECK(inp);
1463                         if (optval > 0 && optval <= tp->t_maxseg &&
1464                             optval + 40 >= V_tcp_minmss)
1465                                 tp->t_maxseg = optval;
1466                         else
1467                                 error = EINVAL;
1468                         goto unlock_and_done;
1469
1470                 case TCP_INFO:
1471                         INP_WUNLOCK(inp);
1472                         error = EINVAL;
1473                         break;
1474
1475                 case TCP_CONGESTION:
1476                         INP_WUNLOCK(inp);
1477                         bzero(buf, sizeof(buf));
1478                         error = sooptcopyin(sopt, &buf, sizeof(buf), 1);
1479                         if (error)
1480                                 break;
1481                         INP_WLOCK_RECHECK(inp);
1482                         /*
1483                          * Return EINVAL if we can't find the requested cc algo.
1484                          */
1485                         error = EINVAL;
1486                         CC_LIST_RLOCK();
1487                         STAILQ_FOREACH(algo, &cc_list, entries) {
1488                                 if (strncmp(buf, algo->name, TCP_CA_NAME_MAX)
1489                                     == 0) {
1490                                         /* We've found the requested algo. */
1491                                         error = 0;
1492                                         /*
1493                                          * We hold a write lock over the tcb
1494                                          * so it's safe to do these things
1495                                          * without ordering concerns.
1496                                          */
1497                                         if (CC_ALGO(tp)->cb_destroy != NULL)
1498                                                 CC_ALGO(tp)->cb_destroy(tp->ccv);
1499                                         CC_ALGO(tp) = algo;
1500                                         /*
1501                                          * If something goes pear shaped
1502                                          * initialising the new algo,
1503                                          * fall back to newreno (which
1504                                          * does not require initialisation).
1505                                          */
1506                                         if (algo->cb_init != NULL)
1507                                                 if (algo->cb_init(tp->ccv) > 0) {
1508                                                         CC_ALGO(tp) = &newreno_cc_algo;
1509                                                         /*
1510                                                          * The only reason init
1511                                                          * should fail is
1512                                                          * because of malloc.
1513                                                          */
1514                                                         error = ENOMEM;
1515                                                 }
1516                                         break; /* Break the STAILQ_FOREACH. */
1517                                 }
1518                         }
1519                         CC_LIST_RUNLOCK();
1520                         goto unlock_and_done;
1521
1522                 case TCP_KEEPIDLE:
1523                 case TCP_KEEPINTVL:
1524                 case TCP_KEEPINIT:
1525                         INP_WUNLOCK(inp);
1526                         error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
1527                         if (error)
1528                                 return (error);
1529
1530                         if (ui > (UINT_MAX / hz)) {
1531                                 error = EINVAL;
1532                                 break;
1533                         }
1534                         ui *= hz;
1535
1536                         INP_WLOCK_RECHECK(inp);
1537                         switch (sopt->sopt_name) {
1538                         case TCP_KEEPIDLE:
1539                                 tp->t_keepidle = ui;
1540                                 /*
1541                                  * XXX: better check current remaining
1542                                  * timeout and "merge" it with new value.
1543                                  */
1544                                 if ((tp->t_state > TCPS_LISTEN) &&
1545                                     (tp->t_state <= TCPS_CLOSING))
1546                                         tcp_timer_activate(tp, TT_KEEP,
1547                                             TP_KEEPIDLE(tp));
1548                                 break;
1549                         case TCP_KEEPINTVL:
1550                                 tp->t_keepintvl = ui;
1551                                 if ((tp->t_state == TCPS_FIN_WAIT_2) &&
1552                                     (TP_MAXIDLE(tp) > 0))
1553                                         tcp_timer_activate(tp, TT_2MSL,
1554                                             TP_MAXIDLE(tp));
1555                                 break;
1556                         case TCP_KEEPINIT:
1557                                 tp->t_keepinit = ui;
1558                                 if (tp->t_state == TCPS_SYN_RECEIVED ||
1559                                     tp->t_state == TCPS_SYN_SENT)
1560                                         tcp_timer_activate(tp, TT_KEEP,
1561                                             TP_KEEPINIT(tp));
1562                                 break;
1563                         }
1564                         goto unlock_and_done;
1565
1566                 case TCP_KEEPCNT:
1567                         INP_WUNLOCK(inp);
1568                         error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
1569                         if (error)
1570                                 return (error);
1571
1572                         INP_WLOCK_RECHECK(inp);
1573                         tp->t_keepcnt = ui;
1574                         if ((tp->t_state == TCPS_FIN_WAIT_2) &&
1575                             (TP_MAXIDLE(tp) > 0))
1576                                 tcp_timer_activate(tp, TT_2MSL,
1577                                     TP_MAXIDLE(tp));
1578                         goto unlock_and_done;
1579
1580                 default:
1581                         INP_WUNLOCK(inp);
1582                         error = ENOPROTOOPT;
1583                         break;
1584                 }
1585                 break;
1586
1587         case SOPT_GET:
1588                 tp = intotcpcb(inp);
1589                 switch (sopt->sopt_name) {
1590 #ifdef TCP_SIGNATURE
1591                 case TCP_MD5SIG:
1592                         optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0;
1593                         INP_WUNLOCK(inp);
1594                         error = sooptcopyout(sopt, &optval, sizeof optval);
1595                         break;
1596 #endif
1597
1598                 case TCP_NODELAY:
1599                         optval = tp->t_flags & TF_NODELAY;
1600                         INP_WUNLOCK(inp);
1601                         error = sooptcopyout(sopt, &optval, sizeof optval);
1602                         break;
1603                 case TCP_MAXSEG:
1604                         optval = tp->t_maxseg;
1605                         INP_WUNLOCK(inp);
1606                         error = sooptcopyout(sopt, &optval, sizeof optval);
1607                         break;
1608                 case TCP_NOOPT:
1609                         optval = tp->t_flags & TF_NOOPT;
1610                         INP_WUNLOCK(inp);
1611                         error = sooptcopyout(sopt, &optval, sizeof optval);
1612                         break;
1613                 case TCP_NOPUSH:
1614                         optval = tp->t_flags & TF_NOPUSH;
1615                         INP_WUNLOCK(inp);
1616                         error = sooptcopyout(sopt, &optval, sizeof optval);
1617                         break;
1618                 case TCP_INFO:
1619                         tcp_fill_info(tp, &ti);
1620                         INP_WUNLOCK(inp);
1621                         error = sooptcopyout(sopt, &ti, sizeof ti);
1622                         break;
1623                 case TCP_CONGESTION:
1624                         bzero(buf, sizeof(buf));
1625                         strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX);
1626                         INP_WUNLOCK(inp);
1627                         error = sooptcopyout(sopt, buf, TCP_CA_NAME_MAX);
1628                         break;
1629                 case TCP_KEEPIDLE:
1630                 case TCP_KEEPINTVL:
1631                 case TCP_KEEPINIT:
1632                 case TCP_KEEPCNT:
1633                         switch (sopt->sopt_name) {
1634                         case TCP_KEEPIDLE:
1635                                 ui = tp->t_keepidle / hz;
1636                                 break;
1637                         case TCP_KEEPINTVL:
1638                                 ui = tp->t_keepintvl / hz;
1639                                 break;
1640                         case TCP_KEEPINIT:
1641                                 ui = tp->t_keepinit / hz;
1642                                 break;
1643                         case TCP_KEEPCNT:
1644                                 ui = tp->t_keepcnt;
1645                                 break;
1646                         }
1647                         INP_WUNLOCK(inp);
1648                         error = sooptcopyout(sopt, &ui, sizeof(ui));
1649                         break;
1650                 default:
1651                         INP_WUNLOCK(inp);
1652                         error = ENOPROTOOPT;
1653                         break;
1654                 }
1655                 break;
1656         }
1657         return (error);
1658 }
1659 #undef INP_WLOCK_RECHECK
1660
1661 /*
1662  * Attach TCP protocol to socket, allocating
1663  * internet protocol control block, tcp control block,
1664  * bufer space, and entering LISTEN state if to accept connections.
1665  */
1666 static int
1667 tcp_attach(struct socket *so)
1668 {
1669         struct tcpcb *tp;
1670         struct inpcb *inp;
1671         int error;
1672
1673         if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1674                 error = soreserve(so, V_tcp_sendspace, V_tcp_recvspace);
1675                 if (error)
1676                         return (error);
1677         }
1678         so->so_rcv.sb_flags |= SB_AUTOSIZE;
1679         so->so_snd.sb_flags |= SB_AUTOSIZE;
1680         INP_INFO_RLOCK(&V_tcbinfo);
1681         error = in_pcballoc(so, &V_tcbinfo);
1682         if (error) {
1683                 INP_INFO_RUNLOCK(&V_tcbinfo);
1684                 return (error);
1685         }
1686         inp = sotoinpcb(so);
1687 #ifdef INET6
1688         if (inp->inp_vflag & INP_IPV6PROTO) {
1689                 inp->inp_vflag |= INP_IPV6;
1690                 inp->in6p_hops = -1;    /* use kernel default */
1691         }
1692         else
1693 #endif
1694         inp->inp_vflag |= INP_IPV4;
1695         tp = tcp_newtcpcb(inp);
1696         if (tp == NULL) {
1697                 in_pcbdetach(inp);
1698                 in_pcbfree(inp);
1699                 INP_INFO_RUNLOCK(&V_tcbinfo);
1700                 return (ENOBUFS);
1701         }
1702         tp->t_state = TCPS_CLOSED;
1703         INP_WUNLOCK(inp);
1704         INP_INFO_RUNLOCK(&V_tcbinfo);
1705         return (0);
1706 }
1707
1708 /*
1709  * Initiate (or continue) disconnect.
1710  * If embryonic state, just send reset (once).
1711  * If in ``let data drain'' option and linger null, just drop.
1712  * Otherwise (hard), mark socket disconnecting and drop
1713  * current input data; switch states based on user close, and
1714  * send segment to peer (with FIN).
1715  */
1716 static void
1717 tcp_disconnect(struct tcpcb *tp)
1718 {
1719         struct inpcb *inp = tp->t_inpcb;
1720         struct socket *so = inp->inp_socket;
1721
1722         INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
1723         INP_WLOCK_ASSERT(inp);
1724
1725         /*
1726          * Neither tcp_close() nor tcp_drop() should return NULL, as the
1727          * socket is still open.
1728          */
1729         if (tp->t_state < TCPS_ESTABLISHED) {
1730                 tp = tcp_close(tp);
1731                 KASSERT(tp != NULL,
1732                     ("tcp_disconnect: tcp_close() returned NULL"));
1733         } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
1734                 tp = tcp_drop(tp, 0);
1735                 KASSERT(tp != NULL,
1736                     ("tcp_disconnect: tcp_drop() returned NULL"));
1737         } else {
1738                 soisdisconnecting(so);
1739                 sbflush(&so->so_rcv);
1740                 tcp_usrclosed(tp);
1741                 if (!(inp->inp_flags & INP_DROPPED))
1742                         tcp_output(tp);
1743         }
1744 }
1745
1746 /*
1747  * User issued close, and wish to trail through shutdown states:
1748  * if never received SYN, just forget it.  If got a SYN from peer,
1749  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1750  * If already got a FIN from peer, then almost done; go to LAST_ACK
1751  * state.  In all other cases, have already sent FIN to peer (e.g.
1752  * after PRU_SHUTDOWN), and just have to play tedious game waiting
1753  * for peer to send FIN or not respond to keep-alives, etc.
1754  * We can let the user exit from the close as soon as the FIN is acked.
1755  */
1756 static void
1757 tcp_usrclosed(struct tcpcb *tp)
1758 {
1759
1760         INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
1761         INP_WLOCK_ASSERT(tp->t_inpcb);
1762
1763         switch (tp->t_state) {
1764         case TCPS_LISTEN:
1765 #ifdef TCP_OFFLOAD
1766                 tcp_offload_listen_stop(tp);
1767 #endif
1768                 tcp_state_change(tp, TCPS_CLOSED);
1769                 /* FALLTHROUGH */
1770         case TCPS_CLOSED:
1771                 tp = tcp_close(tp);
1772                 /*
1773                  * tcp_close() should never return NULL here as the socket is
1774                  * still open.
1775                  */
1776                 KASSERT(tp != NULL,
1777                     ("tcp_usrclosed: tcp_close() returned NULL"));
1778                 break;
1779
1780         case TCPS_SYN_SENT:
1781         case TCPS_SYN_RECEIVED:
1782                 tp->t_flags |= TF_NEEDFIN;
1783                 break;
1784
1785         case TCPS_ESTABLISHED:
1786                 tcp_state_change(tp, TCPS_FIN_WAIT_1);
1787                 break;
1788
1789         case TCPS_CLOSE_WAIT:
1790                 tcp_state_change(tp, TCPS_LAST_ACK);
1791                 break;
1792         }
1793         if (tp->t_state >= TCPS_FIN_WAIT_2) {
1794                 soisdisconnected(tp->t_inpcb->inp_socket);
1795                 /* Prevent the connection hanging in FIN_WAIT_2 forever. */
1796                 if (tp->t_state == TCPS_FIN_WAIT_2) {
1797                         int timeout;
1798
1799                         timeout = (tcp_fast_finwait2_recycle) ?
1800                             tcp_finwait2_timeout : TP_MAXIDLE(tp);
1801                         tcp_timer_activate(tp, TT_2MSL, timeout);
1802                 }
1803         }
1804 }
1805
1806 #ifdef DDB
1807 static void
1808 db_print_indent(int indent)
1809 {
1810         int i;
1811
1812         for (i = 0; i < indent; i++)
1813                 db_printf(" ");
1814 }
1815
1816 static void
1817 db_print_tstate(int t_state)
1818 {
1819
1820         switch (t_state) {
1821         case TCPS_CLOSED:
1822                 db_printf("TCPS_CLOSED");
1823                 return;
1824
1825         case TCPS_LISTEN:
1826                 db_printf("TCPS_LISTEN");
1827                 return;
1828
1829         case TCPS_SYN_SENT:
1830                 db_printf("TCPS_SYN_SENT");
1831                 return;
1832
1833         case TCPS_SYN_RECEIVED:
1834                 db_printf("TCPS_SYN_RECEIVED");
1835                 return;
1836
1837         case TCPS_ESTABLISHED:
1838                 db_printf("TCPS_ESTABLISHED");
1839                 return;
1840
1841         case TCPS_CLOSE_WAIT:
1842                 db_printf("TCPS_CLOSE_WAIT");
1843                 return;
1844
1845         case TCPS_FIN_WAIT_1:
1846                 db_printf("TCPS_FIN_WAIT_1");
1847                 return;
1848
1849         case TCPS_CLOSING:
1850                 db_printf("TCPS_CLOSING");
1851                 return;
1852
1853         case TCPS_LAST_ACK:
1854                 db_printf("TCPS_LAST_ACK");
1855                 return;
1856
1857         case TCPS_FIN_WAIT_2:
1858                 db_printf("TCPS_FIN_WAIT_2");
1859                 return;
1860
1861         case TCPS_TIME_WAIT:
1862                 db_printf("TCPS_TIME_WAIT");
1863                 return;
1864
1865         default:
1866                 db_printf("unknown");
1867                 return;
1868         }
1869 }
1870
1871 static void
1872 db_print_tflags(u_int t_flags)
1873 {
1874         int comma;
1875
1876         comma = 0;
1877         if (t_flags & TF_ACKNOW) {
1878                 db_printf("%sTF_ACKNOW", comma ? ", " : "");
1879                 comma = 1;
1880         }
1881         if (t_flags & TF_DELACK) {
1882                 db_printf("%sTF_DELACK", comma ? ", " : "");
1883                 comma = 1;
1884         }
1885         if (t_flags & TF_NODELAY) {
1886                 db_printf("%sTF_NODELAY", comma ? ", " : "");
1887                 comma = 1;
1888         }
1889         if (t_flags & TF_NOOPT) {
1890                 db_printf("%sTF_NOOPT", comma ? ", " : "");
1891                 comma = 1;
1892         }
1893         if (t_flags & TF_SENTFIN) {
1894                 db_printf("%sTF_SENTFIN", comma ? ", " : "");
1895                 comma = 1;
1896         }
1897         if (t_flags & TF_REQ_SCALE) {
1898                 db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
1899                 comma = 1;
1900         }
1901         if (t_flags & TF_RCVD_SCALE) {
1902                 db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
1903                 comma = 1;
1904         }
1905         if (t_flags & TF_REQ_TSTMP) {
1906                 db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
1907                 comma = 1;
1908         }
1909         if (t_flags & TF_RCVD_TSTMP) {
1910                 db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
1911                 comma = 1;
1912         }
1913         if (t_flags & TF_SACK_PERMIT) {
1914                 db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
1915                 comma = 1;
1916         }
1917         if (t_flags & TF_NEEDSYN) {
1918                 db_printf("%sTF_NEEDSYN", comma ? ", " : "");
1919                 comma = 1;
1920         }
1921         if (t_flags & TF_NEEDFIN) {
1922                 db_printf("%sTF_NEEDFIN", comma ? ", " : "");
1923                 comma = 1;
1924         }
1925         if (t_flags & TF_NOPUSH) {
1926                 db_printf("%sTF_NOPUSH", comma ? ", " : "");
1927                 comma = 1;
1928         }
1929         if (t_flags & TF_MORETOCOME) {
1930                 db_printf("%sTF_MORETOCOME", comma ? ", " : "");
1931                 comma = 1;
1932         }
1933         if (t_flags & TF_LQ_OVERFLOW) {
1934                 db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : "");
1935                 comma = 1;
1936         }
1937         if (t_flags & TF_LASTIDLE) {
1938                 db_printf("%sTF_LASTIDLE", comma ? ", " : "");
1939                 comma = 1;
1940         }
1941         if (t_flags & TF_RXWIN0SENT) {
1942                 db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
1943                 comma = 1;
1944         }
1945         if (t_flags & TF_FASTRECOVERY) {
1946                 db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
1947                 comma = 1;
1948         }
1949         if (t_flags & TF_CONGRECOVERY) {
1950                 db_printf("%sTF_CONGRECOVERY", comma ? ", " : "");
1951                 comma = 1;
1952         }
1953         if (t_flags & TF_WASFRECOVERY) {
1954                 db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
1955                 comma = 1;
1956         }
1957         if (t_flags & TF_SIGNATURE) {
1958                 db_printf("%sTF_SIGNATURE", comma ? ", " : "");
1959                 comma = 1;
1960         }
1961         if (t_flags & TF_FORCEDATA) {
1962                 db_printf("%sTF_FORCEDATA", comma ? ", " : "");
1963                 comma = 1;
1964         }
1965         if (t_flags & TF_TSO) {
1966                 db_printf("%sTF_TSO", comma ? ", " : "");
1967                 comma = 1;
1968         }
1969         if (t_flags & TF_ECN_PERMIT) {
1970                 db_printf("%sTF_ECN_PERMIT", comma ? ", " : "");
1971                 comma = 1;
1972         }
1973 }
1974
1975 static void
1976 db_print_toobflags(char t_oobflags)
1977 {
1978         int comma;
1979
1980         comma = 0;
1981         if (t_oobflags & TCPOOB_HAVEDATA) {
1982                 db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
1983                 comma = 1;
1984         }
1985         if (t_oobflags & TCPOOB_HADDATA) {
1986                 db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
1987                 comma = 1;
1988         }
1989 }
1990
1991 static void
1992 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
1993 {
1994
1995         db_print_indent(indent);
1996         db_printf("%s at %p\n", name, tp);
1997
1998         indent += 2;
1999
2000         db_print_indent(indent);
2001         db_printf("t_segq first: %p   t_segqlen: %d   t_dupacks: %d\n",
2002            LIST_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks);
2003
2004         db_print_indent(indent);
2005         db_printf("tt_rexmt: %p   tt_persist: %p   tt_keep: %p\n",
2006             &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep);
2007
2008         db_print_indent(indent);
2009         db_printf("tt_2msl: %p   tt_delack: %p   t_inpcb: %p\n", &tp->t_timers->tt_2msl,
2010             &tp->t_timers->tt_delack, tp->t_inpcb);
2011
2012         db_print_indent(indent);
2013         db_printf("t_state: %d (", tp->t_state);
2014         db_print_tstate(tp->t_state);
2015         db_printf(")\n");
2016
2017         db_print_indent(indent);
2018         db_printf("t_flags: 0x%x (", tp->t_flags);
2019         db_print_tflags(tp->t_flags);
2020         db_printf(")\n");
2021
2022         db_print_indent(indent);
2023         db_printf("snd_una: 0x%08x   snd_max: 0x%08x   snd_nxt: x0%08x\n",
2024             tp->snd_una, tp->snd_max, tp->snd_nxt);
2025
2026         db_print_indent(indent);
2027         db_printf("snd_up: 0x%08x   snd_wl1: 0x%08x   snd_wl2: 0x%08x\n",
2028            tp->snd_up, tp->snd_wl1, tp->snd_wl2);
2029
2030         db_print_indent(indent);
2031         db_printf("iss: 0x%08x   irs: 0x%08x   rcv_nxt: 0x%08x\n",
2032             tp->iss, tp->irs, tp->rcv_nxt);
2033
2034         db_print_indent(indent);
2035         db_printf("rcv_adv: 0x%08x   rcv_wnd: %lu   rcv_up: 0x%08x\n",
2036             tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
2037
2038         db_print_indent(indent);
2039         db_printf("snd_wnd: %lu   snd_cwnd: %lu\n",
2040            tp->snd_wnd, tp->snd_cwnd);
2041
2042         db_print_indent(indent);
2043         db_printf("snd_ssthresh: %lu   snd_recover: "
2044             "0x%08x\n", tp->snd_ssthresh, tp->snd_recover);
2045
2046         db_print_indent(indent);
2047         db_printf("t_maxopd: %u   t_rcvtime: %u   t_startime: %u\n",
2048             tp->t_maxopd, tp->t_rcvtime, tp->t_starttime);
2049
2050         db_print_indent(indent);
2051         db_printf("t_rttime: %u   t_rtsq: 0x%08x\n",
2052             tp->t_rtttime, tp->t_rtseq);
2053
2054         db_print_indent(indent);
2055         db_printf("t_rxtcur: %d   t_maxseg: %u   t_srtt: %d\n",
2056             tp->t_rxtcur, tp->t_maxseg, tp->t_srtt);
2057
2058         db_print_indent(indent);
2059         db_printf("t_rttvar: %d   t_rxtshift: %d   t_rttmin: %u   "
2060             "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin,
2061             tp->t_rttbest);
2062
2063         db_print_indent(indent);
2064         db_printf("t_rttupdated: %lu   max_sndwnd: %lu   t_softerror: %d\n",
2065             tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
2066
2067         db_print_indent(indent);
2068         db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
2069         db_print_toobflags(tp->t_oobflags);
2070         db_printf(")   t_iobc: 0x%02x\n", tp->t_iobc);
2071
2072         db_print_indent(indent);
2073         db_printf("snd_scale: %u   rcv_scale: %u   request_r_scale: %u\n",
2074             tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
2075
2076         db_print_indent(indent);
2077         db_printf("ts_recent: %u   ts_recent_age: %u\n",
2078             tp->ts_recent, tp->ts_recent_age);
2079
2080         db_print_indent(indent);
2081         db_printf("ts_offset: %u   last_ack_sent: 0x%08x   snd_cwnd_prev: "
2082             "%lu\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
2083
2084         db_print_indent(indent);
2085         db_printf("snd_ssthresh_prev: %lu   snd_recover_prev: 0x%08x   "
2086             "t_badrxtwin: %u\n", tp->snd_ssthresh_prev,
2087             tp->snd_recover_prev, tp->t_badrxtwin);
2088
2089         db_print_indent(indent);
2090         db_printf("snd_numholes: %d  snd_holes first: %p\n",
2091             tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
2092
2093         db_print_indent(indent);
2094         db_printf("snd_fack: 0x%08x   rcv_numsacks: %d   sack_newdata: "
2095             "0x%08x\n", tp->snd_fack, tp->rcv_numsacks, tp->sack_newdata);
2096
2097         /* Skip sackblks, sackhint. */
2098
2099         db_print_indent(indent);
2100         db_printf("t_rttlow: %d   rfbuf_ts: %u   rfbuf_cnt: %d\n",
2101             tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
2102 }
2103
2104 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
2105 {
2106         struct tcpcb *tp;
2107
2108         if (!have_addr) {
2109                 db_printf("usage: show tcpcb <addr>\n");
2110                 return;
2111         }
2112         tp = (struct tcpcb *)addr;
2113
2114         db_print_tcpcb(tp, "tcpcb", 0);
2115 }
2116 #endif