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  * All rights reserved.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  * 1. Redistributions of source code must retain the above copyright
  11  *    notice, this list of conditions and the following disclaimer.
  12  * 2. Redistributions in binary form must reproduce the above copyright
  13  *    notice, this list of conditions and the following disclaimer in the
  14  *    documentation and/or other materials provided with the distribution.
  15  * 4. Neither the name of the University nor the names of its contributors
  16  *    may be used to endorse or promote products derived from this software
  17  *    without specific prior written permission.
  18  *
  19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  29  * SUCH DAMAGE.
  30  *
  31  *      From: @(#)tcp_usrreq.c  8.2 (Berkeley) 1/3/94
  32  */
  33
  34 #include <sys/cdefs.h>
  35 __FBSDID("$FreeBSD$");
  36
  37 #include "opt_ddb.h"
  38 #include "opt_inet.h"
  39 #include "opt_inet6.h"
  40 #include "opt_tcpdebug.h"
  41
  42 #include <sys/param.h>
  43 #include <sys/systm.h>
  44 #include <sys/malloc.h>
  45 #include <sys/kernel.h>
  46 #include <sys/sysctl.h>
  47 #include <sys/mbuf.h>
  48 #ifdef INET6
  49 #include <sys/domain.h>
  50 #endif /* INET6 */
  51 #include <sys/socket.h>
  52 #include <sys/socketvar.h>
  53 #include <sys/protosw.h>
  54 #include <sys/proc.h>
  55 #include <sys/jail.h>
  56
  57 #ifdef DDB
  58 #include <ddb/ddb.h>
  59 #endif
  60
  61 #include <net/if.h>
  62 #include <net/route.h>
  63 #include <net/vnet.h>
  64
  65 #include <netinet/cc.h>
  66 #include <netinet/in.h>
  67 #include <netinet/in_systm.h>
  68 #ifdef INET6
  69 #include <netinet/ip6.h>
  70 #endif
  71 #include <netinet/in_pcb.h>
  72 #ifdef INET6
  73 #include <netinet6/in6_pcb.h>
  74 #endif
  75 #include <netinet/in_var.h>
  76 #include <netinet/ip_var.h>
  77 #ifdef INET6
  78 #include <netinet6/ip6_var.h>
  79 #include <netinet6/scope6_var.h>
  80 #endif
  81 #include <netinet/tcp_fsm.h>
  82 #include <netinet/tcp_seq.h>
  83 #include <netinet/tcp_timer.h>
  84 #include <netinet/tcp_var.h>
  85 #include <netinet/tcpip.h>
  86 #ifdef TCPDEBUG
  87 #include <netinet/tcp_debug.h>
  88 #endif
  89 #include <netinet/tcp_offload.h>
  90
  91 /*
  92  * TCP protocol interface to socket abstraction.
  93  */
  94 static int      tcp_attach(struct socket *);
  95 static int      tcp_connect(struct tcpcb *, struct sockaddr *,
  96                     struct thread *td);
  97 #ifdef INET6
  98 static int      tcp6_connect(struct tcpcb *, struct sockaddr *,
  99                     struct thread *td);
 100 #endif /* INET6 */
 101 static void     tcp_disconnect(struct tcpcb *);
 102 static void     tcp_usrclosed(struct tcpcb *);
 103 static void     tcp_fill_info(struct tcpcb *, struct tcp_info *);
 104
 105 #ifdef TCPDEBUG
 106 #define TCPDEBUG0       int ostate = 0
 107 #define TCPDEBUG1()     ostate = tp ? tp->t_state : 0
 108 #define TCPDEBUG2(req)  if (tp && (so->so_options & SO_DEBUG)) \
 109                                 tcp_trace(TA_USER, ostate, tp, 0, 0, req)
 110 #else
 111 #define TCPDEBUG0
 112 #define TCPDEBUG1()
 113 #define TCPDEBUG2(req)
 114 #endif
 115
 116 /*
 117  * TCP attaches to socket via pru_attach(), reserving space,
 118  * and an internet control block.
 119  */
 120 static int
 121 tcp_usr_attach(struct socket *so, int proto, struct thread *td)
 122 {
 123         struct inpcb *inp;
 124         struct tcpcb *tp = NULL;
 125         int error;
 126         TCPDEBUG0;
 127
 128         inp = sotoinpcb(so);
 129         KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL"));
 130         TCPDEBUG1();
 131
 132         error = tcp_attach(so);
 133         if (error)
 134                 goto out;
 135
 136         if ((so->so_options & SO_LINGER) && so->so_linger == 0)
 137                 so->so_linger = TCP_LINGERTIME;
 138
 139         inp = sotoinpcb(so);
 140         tp = intotcpcb(inp);
 141 out:
 142         TCPDEBUG2(PRU_ATTACH);
 143         return error;
 144 }
 145
 146 /*
 147  * tcp_detach is called when the socket layer loses its final reference
 148  * to the socket, be it a file descriptor reference, a reference from TCP,
 149  * etc.  At this point, there is only one case in which we will keep around
 150  * inpcb state: time wait.
 151  *
 152  * This function can probably be re-absorbed back into tcp_usr_detach() now
 153  * that there is a single detach path.
 154  */
 155 static void
 156 tcp_detach(struct socket *so, struct inpcb *inp)
 157 {
 158         struct tcpcb *tp;
 159
 160         INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
 161         INP_WLOCK_ASSERT(inp);
 162
 163         KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp"));
 164         KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so"));
 165
 166         tp = intotcpcb(inp);
 167
 168         if (inp->inp_flags & INP_TIMEWAIT) {
 169                 /*
 170                  * There are two cases to handle: one in which the time wait
 171                  * state is being discarded (INP_DROPPED), and one in which
 172                  * this connection will remain in timewait.  In the former,
 173                  * it is time to discard all state (except tcptw, which has
 174                  * already been discarded by the timewait close code, which
 175                  * should be further up the call stack somewhere).  In the
 176                  * latter case, we detach from the socket, but leave the pcb
 177                  * present until timewait ends.
 178                  *
 179                  * XXXRW: Would it be cleaner to free the tcptw here?
 180                  */
 181                 if (inp->inp_flags & INP_DROPPED) {
 182                         KASSERT(tp == NULL, ("tcp_detach: INP_TIMEWAIT && "
 183                             "INP_DROPPED && tp != NULL"));
 184                         in_pcbdetach(inp);
 185                         in_pcbfree(inp);
 186                 } else {
 187                         in_pcbdetach(inp);
 188                         INP_WUNLOCK(inp);
 189                 }
 190         } else {
 191                 /*
 192                  * If the connection is not in timewait, we consider two
 193                  * two conditions: one in which no further processing is
 194                  * necessary (dropped || embryonic), and one in which TCP is
 195                  * not yet done, but no longer requires the socket, so the
 196                  * pcb will persist for the time being.
 197                  *
 198                  * XXXRW: Does the second case still occur?
 199                  */
 200                 if (inp->inp_flags & INP_DROPPED ||
 201                     tp->t_state < TCPS_SYN_SENT) {
 202                         tcp_discardcb(tp);
 203                         in_pcbdetach(inp);
 204                         in_pcbfree(inp);
 205                 } else {
 206                         in_pcbdetach(inp);
 207                         INP_WUNLOCK(inp);
 208                 }
 209         }
 210 }
 211
 212 /*
 213  * pru_detach() detaches the TCP protocol from the socket.
 214  * If the protocol state is non-embryonic, then can't
 215  * do this directly: have to initiate a pru_disconnect(),
 216  * which may finish later; embryonic TCB's can just
 217  * be discarded here.
 218  */
 219 static void
 220 tcp_usr_detach(struct socket *so)
 221 {
 222         struct inpcb *inp;
 223
 224         inp = sotoinpcb(so);
 225         KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL"));
 226         INP_INFO_WLOCK(&V_tcbinfo);
 227         INP_WLOCK(inp);
 228         KASSERT(inp->inp_socket != NULL,
 229             ("tcp_usr_detach: inp_socket == NULL"));
 230         tcp_detach(so, inp);
 231         INP_INFO_WUNLOCK(&V_tcbinfo);
 232 }
 233
 234 /*
 235  * Give the socket an address.
 236  */
 237 static int
 238 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 239 {
 240         int error = 0;
 241         struct inpcb *inp;
 242         struct tcpcb *tp = NULL;
 243         struct sockaddr_in *sinp;
 244
 245         sinp = (struct sockaddr_in *)nam;
 246         if (nam->sa_len != sizeof (*sinp))
 247                 return (EINVAL);
 248         /*
 249          * Must check for multicast addresses and disallow binding
 250          * to them.
 251          */
 252         if (sinp->sin_family == AF_INET &&
 253             IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
 254                 return (EAFNOSUPPORT);
 255
 256         TCPDEBUG0;
 257         INP_INFO_WLOCK(&V_tcbinfo);
 258         inp = sotoinpcb(so);
 259         KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL"));
 260         INP_WLOCK(inp);
 261         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 262                 error = EINVAL;
 263                 goto out;
 264         }
 265         tp = intotcpcb(inp);
 266         TCPDEBUG1();
 267         error = in_pcbbind(inp, nam, td->td_ucred);
 268 out:
 269         TCPDEBUG2(PRU_BIND);
 270         INP_WUNLOCK(inp);
 271         INP_INFO_WUNLOCK(&V_tcbinfo);
 272
 273         return (error);
 274 }
 275
 276 #ifdef INET6
 277 static int
 278 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 279 {
 280         int error = 0;
 281         struct inpcb *inp;
 282         struct tcpcb *tp = NULL;
 283         struct sockaddr_in6 *sin6p;
 284
 285         sin6p = (struct sockaddr_in6 *)nam;
 286         if (nam->sa_len != sizeof (*sin6p))
 287                 return (EINVAL);
 288         /*
 289          * Must check for multicast addresses and disallow binding
 290          * to them.
 291          */
 292         if (sin6p->sin6_family == AF_INET6 &&
 293             IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
 294                 return (EAFNOSUPPORT);
 295
 296         TCPDEBUG0;
 297         INP_INFO_WLOCK(&V_tcbinfo);
 298         inp = sotoinpcb(so);
 299         KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL"));
 300         INP_WLOCK(inp);
 301         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 302                 error = EINVAL;
 303                 goto out;
 304         }
 305         tp = intotcpcb(inp);
 306         TCPDEBUG1();
 307         inp->inp_vflag &= ~INP_IPV4;
 308         inp->inp_vflag |= INP_IPV6;
 309         if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
 310                 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
 311                         inp->inp_vflag |= INP_IPV4;
 312                 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
 313                         struct sockaddr_in sin;
 314
 315                         in6_sin6_2_sin(&sin, sin6p);
 316                         inp->inp_vflag |= INP_IPV4;
 317                         inp->inp_vflag &= ~INP_IPV6;
 318                         error = in_pcbbind(inp, (struct sockaddr *)&sin,
 319                             td->td_ucred);
 320                         goto out;
 321                 }
 322         }
 323         error = in6_pcbbind(inp, nam, td->td_ucred);
 324 out:
 325         TCPDEBUG2(PRU_BIND);
 326         INP_WUNLOCK(inp);
 327         INP_INFO_WUNLOCK(&V_tcbinfo);
 328         return (error);
 329 }
 330 #endif /* INET6 */
 331
 332 /*
 333  * Prepare to accept connections.
 334  */
 335 static int
 336 tcp_usr_listen(struct socket *so, int backlog, struct thread *td)
 337 {
 338         int error = 0;
 339         struct inpcb *inp;
 340         struct tcpcb *tp = NULL;
 341
 342         TCPDEBUG0;
 343         INP_INFO_WLOCK(&V_tcbinfo);
 344         inp = sotoinpcb(so);
 345         KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL"));
 346         INP_WLOCK(inp);
 347         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 348                 error = EINVAL;
 349                 goto out;
 350         }
 351         tp = intotcpcb(inp);
 352         TCPDEBUG1();
 353         SOCK_LOCK(so);
 354         error = solisten_proto_check(so);
 355         if (error == 0 && inp->inp_lport == 0)
 356                 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
 357         if (error == 0) {
 358                 tp->t_state = TCPS_LISTEN;
 359                 solisten_proto(so, backlog);
 360                 tcp_offload_listen_open(tp);
 361         }
 362         SOCK_UNLOCK(so);
 363
 364 out:
 365         TCPDEBUG2(PRU_LISTEN);
 366         INP_WUNLOCK(inp);
 367         INP_INFO_WUNLOCK(&V_tcbinfo);
 368         return (error);
 369 }
 370
 371 #ifdef INET6
 372 static int
 373 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td)
 374 {
 375         int error = 0;
 376         struct inpcb *inp;
 377         struct tcpcb *tp = NULL;
 378
 379         TCPDEBUG0;
 380         INP_INFO_WLOCK(&V_tcbinfo);
 381         inp = sotoinpcb(so);
 382         KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL"));
 383         INP_WLOCK(inp);
 384         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 385                 error = EINVAL;
 386                 goto out;
 387         }
 388         tp = intotcpcb(inp);
 389         TCPDEBUG1();
 390         SOCK_LOCK(so);
 391         error = solisten_proto_check(so);
 392         if (error == 0 && inp->inp_lport == 0) {
 393                 inp->inp_vflag &= ~INP_IPV4;
 394                 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
 395                         inp->inp_vflag |= INP_IPV4;
 396                 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
 397         }
 398         if (error == 0) {
 399                 tp->t_state = TCPS_LISTEN;
 400                 solisten_proto(so, backlog);
 401         }
 402         SOCK_UNLOCK(so);
 403
 404 out:
 405         TCPDEBUG2(PRU_LISTEN);
 406         INP_WUNLOCK(inp);
 407         INP_INFO_WUNLOCK(&V_tcbinfo);
 408         return (error);
 409 }
 410 #endif /* INET6 */
 411
 412 /*
 413  * Initiate connection to peer.
 414  * Create a template for use in transmissions on this connection.
 415  * Enter SYN_SENT state, and mark socket as connecting.
 416  * Start keep-alive timer, and seed output sequence space.
 417  * Send initial segment on connection.
 418  */
 419 static int
 420 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 421 {
 422         int error = 0;
 423         struct inpcb *inp;
 424         struct tcpcb *tp = NULL;
 425         struct sockaddr_in *sinp;
 426
 427         sinp = (struct sockaddr_in *)nam;
 428         if (nam->sa_len != sizeof (*sinp))
 429                 return (EINVAL);
 430         /*
 431          * Must disallow TCP ``connections'' to multicast addresses.
 432          */
 433         if (sinp->sin_family == AF_INET
 434             && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
 435                 return (EAFNOSUPPORT);
 436         if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0)
 437                 return (error);
 438
 439         TCPDEBUG0;
 440         INP_INFO_WLOCK(&V_tcbinfo);
 441         inp = sotoinpcb(so);
 442         KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL"));
 443         INP_WLOCK(inp);
 444         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 445                 error = EINVAL;
 446                 goto out;
 447         }
 448         tp = intotcpcb(inp);
 449         TCPDEBUG1();
 450         if ((error = tcp_connect(tp, nam, td)) != 0)
 451                 goto out;
 452         error = tcp_output_connect(so, nam);
 453 out:
 454         TCPDEBUG2(PRU_CONNECT);
 455         INP_WUNLOCK(inp);
 456         INP_INFO_WUNLOCK(&V_tcbinfo);
 457         return (error);
 458 }
 459
 460 #ifdef INET6
 461 static int
 462 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 463 {
 464         int error = 0;
 465         struct inpcb *inp;
 466         struct tcpcb *tp = NULL;
 467         struct sockaddr_in6 *sin6p;
 468
 469         TCPDEBUG0;
 470
 471         sin6p = (struct sockaddr_in6 *)nam;
 472         if (nam->sa_len != sizeof (*sin6p))
 473                 return (EINVAL);
 474         /*
 475          * Must disallow TCP ``connections'' to multicast addresses.
 476          */
 477         if (sin6p->sin6_family == AF_INET6
 478             && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
 479                 return (EAFNOSUPPORT);
 480
 481         INP_INFO_WLOCK(&V_tcbinfo);
 482         inp = sotoinpcb(so);
 483         KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL"));
 484         INP_WLOCK(inp);
 485         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 486                 error = EINVAL;
 487                 goto out;
 488         }
 489         tp = intotcpcb(inp);
 490         TCPDEBUG1();
 491         if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
 492                 struct sockaddr_in sin;
 493
 494                 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
 495                         error = EINVAL;
 496                         goto out;
 497                 }
 498
 499                 in6_sin6_2_sin(&sin, sin6p);
 500                 inp->inp_vflag |= INP_IPV4;
 501                 inp->inp_vflag &= ~INP_IPV6;
 502                 if ((error = prison_remote_ip4(td->td_ucred,
 503                     &sin.sin_addr)) != 0)
 504                         goto out;
 505                 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
 506                         goto out;
 507                 error = tcp_output_connect(so, nam);
 508                 goto out;
 509         }
 510         inp->inp_vflag &= ~INP_IPV4;
 511         inp->inp_vflag |= INP_IPV6;
 512         inp->inp_inc.inc_flags |= INC_ISIPV6;
 513         if ((error = prison_remote_ip6(td->td_ucred, &sin6p->sin6_addr)) != 0)
 514                 goto out;
 515         if ((error = tcp6_connect(tp, nam, td)) != 0)
 516                 goto out;
 517         error = tcp_output_connect(so, nam);
 518
 519 out:
 520         TCPDEBUG2(PRU_CONNECT);
 521         INP_WUNLOCK(inp);
 522         INP_INFO_WUNLOCK(&V_tcbinfo);
 523         return (error);
 524 }
 525 #endif /* INET6 */
 526
 527 /*
 528  * Initiate disconnect from peer.
 529  * If connection never passed embryonic stage, just drop;
 530  * else if don't need to let data drain, then can just drop anyways,
 531  * else have to begin TCP shutdown process: mark socket disconnecting,
 532  * drain unread data, state switch to reflect user close, and
 533  * send segment (e.g. FIN) to peer.  Socket will be really disconnected
 534  * when peer sends FIN and acks ours.
 535  *
 536  * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
 537  */
 538 static int
 539 tcp_usr_disconnect(struct socket *so)
 540 {
 541         struct inpcb *inp;
 542         struct tcpcb *tp = NULL;
 543         int error = 0;
 544
 545         TCPDEBUG0;
 546         INP_INFO_WLOCK(&V_tcbinfo);
 547         inp = sotoinpcb(so);
 548         KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
 549         INP_WLOCK(inp);
 550         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 551                 error = ECONNRESET;
 552                 goto out;
 553         }
 554         tp = intotcpcb(inp);
 555         TCPDEBUG1();
 556         tcp_disconnect(tp);
 557 out:
 558         TCPDEBUG2(PRU_DISCONNECT);
 559         INP_WUNLOCK(inp);
 560         INP_INFO_WUNLOCK(&V_tcbinfo);
 561         return (error);
 562 }
 563
 564 /*
 565  * Accept a connection.  Essentially all the work is done at higher levels;
 566  * just return the address of the peer, storing through addr.
 567  *
 568  * The rationale for acquiring the tcbinfo lock here is somewhat complicated,
 569  * and is described in detail in the commit log entry for r175612.  Acquiring
 570  * it delays an accept(2) racing with sonewconn(), which inserts the socket
 571  * before the inpcb address/port fields are initialized.  A better fix would
 572  * prevent the socket from being placed in the listen queue until all fields
 573  * are fully initialized.
 574  */
 575 static int
 576 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
 577 {
 578         int error = 0;
 579         struct inpcb *inp = NULL;
 580         struct tcpcb *tp = NULL;
 581         struct in_addr addr;
 582         in_port_t port = 0;
 583         TCPDEBUG0;
 584
 585         if (so->so_state & SS_ISDISCONNECTED)
 586                 return (ECONNABORTED);
 587
 588         inp = sotoinpcb(so);
 589         KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
 590         INP_INFO_RLOCK(&V_tcbinfo);
 591         INP_WLOCK(inp);
 592         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 593                 error = ECONNABORTED;
 594                 goto out;
 595         }
 596         tp = intotcpcb(inp);
 597         TCPDEBUG1();
 598
 599         /*
 600          * We inline in_getpeeraddr and COMMON_END here, so that we can
 601          * copy the data of interest and defer the malloc until after we
 602          * release the lock.
 603          */
 604         port = inp->inp_fport;
 605         addr = inp->inp_faddr;
 606
 607 out:
 608         TCPDEBUG2(PRU_ACCEPT);
 609         INP_WUNLOCK(inp);
 610         INP_INFO_RUNLOCK(&V_tcbinfo);
 611         if (error == 0)
 612                 *nam = in_sockaddr(port, &addr);
 613         return error;
 614 }
 615
 616 #ifdef INET6
 617 static int
 618 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
 619 {
 620         struct inpcb *inp = NULL;
 621         int error = 0;
 622         struct tcpcb *tp = NULL;
 623         struct in_addr addr;
 624         struct in6_addr addr6;
 625         in_port_t port = 0;
 626         int v4 = 0;
 627         TCPDEBUG0;
 628
 629         if (so->so_state & SS_ISDISCONNECTED)
 630                 return (ECONNABORTED);
 631
 632         inp = sotoinpcb(so);
 633         KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
 634         INP_WLOCK(inp);
 635         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 636                 error = ECONNABORTED;
 637                 goto out;
 638         }
 639         tp = intotcpcb(inp);
 640         TCPDEBUG1();
 641
 642         /*
 643          * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
 644          * copy the data of interest and defer the malloc until after we
 645          * release the lock.
 646          */
 647         if (inp->inp_vflag & INP_IPV4) {
 648                 v4 = 1;
 649                 port = inp->inp_fport;
 650                 addr = inp->inp_faddr;
 651         } else {
 652                 port = inp->inp_fport;
 653                 addr6 = inp->in6p_faddr;
 654         }
 655
 656 out:
 657         TCPDEBUG2(PRU_ACCEPT);
 658         INP_WUNLOCK(inp);
 659         if (error == 0) {
 660                 if (v4)
 661                         *nam = in6_v4mapsin6_sockaddr(port, &addr);
 662                 else
 663                         *nam = in6_sockaddr(port, &addr6);
 664         }
 665         return error;
 666 }
 667 #endif /* INET6 */
 668
 669 /*
 670  * Mark the connection as being incapable of further output.
 671  */
 672 static int
 673 tcp_usr_shutdown(struct socket *so)
 674 {
 675         int error = 0;
 676         struct inpcb *inp;
 677         struct tcpcb *tp = NULL;
 678
 679         TCPDEBUG0;
 680         INP_INFO_WLOCK(&V_tcbinfo);
 681         inp = sotoinpcb(so);
 682         KASSERT(inp != NULL, ("inp == NULL"));
 683         INP_WLOCK(inp);
 684         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 685                 error = ECONNRESET;
 686                 goto out;
 687         }
 688         tp = intotcpcb(inp);
 689         TCPDEBUG1();
 690         socantsendmore(so);
 691         tcp_usrclosed(tp);
 692         if (!(inp->inp_flags & INP_DROPPED))
 693                 error = tcp_output_disconnect(tp);
 694
 695 out:
 696         TCPDEBUG2(PRU_SHUTDOWN);
 697         INP_WUNLOCK(inp);
 698         INP_INFO_WUNLOCK(&V_tcbinfo);
 699
 700         return (error);
 701 }
 702
 703 /*
 704  * After a receive, possibly send window update to peer.
 705  */
 706 static int
 707 tcp_usr_rcvd(struct socket *so, int flags)
 708 {
 709         struct inpcb *inp;
 710         struct tcpcb *tp = NULL;
 711         int error = 0;
 712
 713         TCPDEBUG0;
 714         inp = sotoinpcb(so);
 715         KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
 716         INP_WLOCK(inp);
 717         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 718                 error = ECONNRESET;
 719                 goto out;
 720         }
 721         tp = intotcpcb(inp);
 722         TCPDEBUG1();
 723         tcp_output_rcvd(tp);
 724
 725 out:
 726         TCPDEBUG2(PRU_RCVD);
 727         INP_WUNLOCK(inp);
 728         return (error);
 729 }
 730
 731 /*
 732  * Do a send by putting data in output queue and updating urgent
 733  * marker if URG set.  Possibly send more data.  Unlike the other
 734  * pru_*() routines, the mbuf chains are our responsibility.  We
 735  * must either enqueue them or free them.  The other pru_* routines
 736  * generally are caller-frees.
 737  */
 738 static int
 739 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
 740     struct sockaddr *nam, struct mbuf *control, struct thread *td)
 741 {
 742         int error = 0;
 743         struct inpcb *inp;
 744         struct tcpcb *tp = NULL;
 745         int headlocked = 0;
 746 #ifdef INET6
 747         int isipv6;
 748 #endif
 749         TCPDEBUG0;
 750
 751         /*
 752          * We require the pcbinfo lock in two cases:
 753          *
 754          * (1) An implied connect is taking place, which can result in
 755          *     binding IPs and ports and hence modification of the pcb hash
 756          *     chains.
 757          *
 758          * (2) PRUS_EOF is set, resulting in explicit close on the send.
 759          */
 760         if ((nam != NULL) || (flags & PRUS_EOF)) {
 761                 INP_INFO_WLOCK(&V_tcbinfo);
 762                 headlocked = 1;
 763         }
 764         inp = sotoinpcb(so);
 765         KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
 766         INP_WLOCK(inp);
 767         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 768                 if (control)
 769                         m_freem(control);
 770                 if (m)
 771                         m_freem(m);
 772                 error = ECONNRESET;
 773                 goto out;
 774         }
 775 #ifdef INET6
 776         isipv6 = nam && nam->sa_family == AF_INET6;
 777 #endif /* INET6 */
 778         tp = intotcpcb(inp);
 779         TCPDEBUG1();
 780         if (control) {
 781                 /* TCP doesn't do control messages (rights, creds, etc) */
 782                 if (control->m_len) {
 783                         m_freem(control);
 784                         if (m)
 785                                 m_freem(m);
 786                         error = EINVAL;
 787                         goto out;
 788                 }
 789                 m_freem(control);       /* empty control, just free it */
 790         }
 791         if (!(flags & PRUS_OOB)) {
 792                 sbappendstream(&so->so_snd, m);
 793                 if (nam && tp->t_state < TCPS_SYN_SENT) {
 794                         /*
 795                          * Do implied connect if not yet connected,
 796                          * initialize window to default value, and
 797                          * initialize maxseg/maxopd using peer's cached
 798                          * MSS.
 799                          */
 800                         INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
 801 #ifdef INET6
 802                         if (isipv6)
 803                                 error = tcp6_connect(tp, nam, td);
 804                         else
 805 #endif /* INET6 */
 806                         error = tcp_connect(tp, nam, td);
 807                         if (error)
 808                                 goto out;
 809                         tp->snd_wnd = TTCP_CLIENT_SND_WND;
 810                         tcp_mss(tp, -1);
 811                 }
 812                 if (flags & PRUS_EOF) {
 813                         /*
 814                          * Close the send side of the connection after
 815                          * the data is sent.
 816                          */
 817                         INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
 818                         socantsendmore(so);
 819                         tcp_usrclosed(tp);
 820                 }
 821                 if (headlocked) {
 822                         INP_INFO_WUNLOCK(&V_tcbinfo);
 823                         headlocked = 0;
 824                 }
 825                 if (!(inp->inp_flags & INP_DROPPED)) {
 826                         if (flags & PRUS_MORETOCOME)
 827                                 tp->t_flags |= TF_MORETOCOME;
 828                         error = tcp_output_send(tp);
 829                         if (flags & PRUS_MORETOCOME)
 830                                 tp->t_flags &= ~TF_MORETOCOME;
 831                 }
 832         } else {
 833                 /*
 834                  * XXXRW: PRUS_EOF not implemented with PRUS_OOB?
 835                  */
 836                 SOCKBUF_LOCK(&so->so_snd);
 837                 if (sbspace(&so->so_snd) < -512) {
 838                         SOCKBUF_UNLOCK(&so->so_snd);
 839                         m_freem(m);
 840                         error = ENOBUFS;
 841                         goto out;
 842                 }
 843                 /*
 844                  * According to RFC961 (Assigned Protocols),
 845                  * the urgent pointer points to the last octet
 846                  * of urgent data.  We continue, however,
 847                  * to consider it to indicate the first octet
 848                  * of data past the urgent section.
 849                  * Otherwise, snd_up should be one lower.
 850                  */
 851                 sbappendstream_locked(&so->so_snd, m);
 852                 SOCKBUF_UNLOCK(&so->so_snd);
 853                 if (nam && tp->t_state < TCPS_SYN_SENT) {
 854                         /*
 855                          * Do implied connect if not yet connected,
 856                          * initialize window to default value, and
 857                          * initialize maxseg/maxopd using peer's cached
 858                          * MSS.
 859                          */
 860                         INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
 861 #ifdef INET6
 862                         if (isipv6)
 863                                 error = tcp6_connect(tp, nam, td);
 864                         else
 865 #endif /* INET6 */
 866                         error = tcp_connect(tp, nam, td);
 867                         if (error)
 868                                 goto out;
 869                         tp->snd_wnd = TTCP_CLIENT_SND_WND;
 870                         tcp_mss(tp, -1);
 871                         INP_INFO_WUNLOCK(&V_tcbinfo);
 872                         headlocked = 0;
 873                 } else if (nam) {
 874                         INP_INFO_WUNLOCK(&V_tcbinfo);
 875                         headlocked = 0;
 876                 }
 877                 tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
 878                 tp->t_flags |= TF_FORCEDATA;
 879                 error = tcp_output_send(tp);
 880                 tp->t_flags &= ~TF_FORCEDATA;
 881         }
 882 out:
 883         TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
 884                   ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
 885         INP_WUNLOCK(inp);
 886         if (headlocked)
 887                 INP_INFO_WUNLOCK(&V_tcbinfo);
 888         return (error);
 889 }
 890
 891 /*
 892  * Abort the TCP.  Drop the connection abruptly.
 893  */
 894 static void
 895 tcp_usr_abort(struct socket *so)
 896 {
 897         struct inpcb *inp;
 898         struct tcpcb *tp = NULL;
 899         TCPDEBUG0;
 900
 901         inp = sotoinpcb(so);
 902         KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
 903
 904         INP_INFO_WLOCK(&V_tcbinfo);
 905         INP_WLOCK(inp);
 906         KASSERT(inp->inp_socket != NULL,
 907             ("tcp_usr_abort: inp_socket == NULL"));
 908
 909         /*
 910          * If we still have full TCP state, and we're not dropped, drop.
 911          */
 912         if (!(inp->inp_flags & INP_TIMEWAIT) &&
 913             !(inp->inp_flags & INP_DROPPED)) {
 914                 tp = intotcpcb(inp);
 915                 TCPDEBUG1();
 916                 tcp_drop(tp, ECONNABORTED);
 917                 TCPDEBUG2(PRU_ABORT);
 918         }
 919         if (!(inp->inp_flags & INP_DROPPED)) {
 920                 SOCK_LOCK(so);
 921                 so->so_state |= SS_PROTOREF;
 922                 SOCK_UNLOCK(so);
 923                 inp->inp_flags |= INP_SOCKREF;
 924         }
 925         INP_WUNLOCK(inp);
 926         INP_INFO_WUNLOCK(&V_tcbinfo);
 927 }
 928
 929 /*
 930  * TCP socket is closed.  Start friendly disconnect.
 931  */
 932 static void
 933 tcp_usr_close(struct socket *so)
 934 {
 935         struct inpcb *inp;
 936         struct tcpcb *tp = NULL;
 937         TCPDEBUG0;
 938
 939         inp = sotoinpcb(so);
 940         KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
 941
 942         INP_INFO_WLOCK(&V_tcbinfo);
 943         INP_WLOCK(inp);
 944         KASSERT(inp->inp_socket != NULL,
 945             ("tcp_usr_close: inp_socket == NULL"));
 946
 947         /*
 948          * If we still have full TCP state, and we're not dropped, initiate
 949          * a disconnect.
 950          */
 951         if (!(inp->inp_flags & INP_TIMEWAIT) &&
 952             !(inp->inp_flags & INP_DROPPED)) {
 953                 tp = intotcpcb(inp);
 954                 TCPDEBUG1();
 955                 tcp_disconnect(tp);
 956                 TCPDEBUG2(PRU_CLOSE);
 957         }
 958         if (!(inp->inp_flags & INP_DROPPED)) {
 959                 SOCK_LOCK(so);
 960                 so->so_state |= SS_PROTOREF;
 961                 SOCK_UNLOCK(so);
 962                 inp->inp_flags |= INP_SOCKREF;
 963         }
 964         INP_WUNLOCK(inp);
 965         INP_INFO_WUNLOCK(&V_tcbinfo);
 966 }
 967
 968 /*
 969  * Receive out-of-band data.
 970  */
 971 static int
 972 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
 973 {
 974         int error = 0;
 975         struct inpcb *inp;
 976         struct tcpcb *tp = NULL;
 977
 978         TCPDEBUG0;
 979         inp = sotoinpcb(so);
 980         KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
 981         INP_WLOCK(inp);
 982         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 983                 error = ECONNRESET;
 984                 goto out;
 985         }
 986         tp = intotcpcb(inp);
 987         TCPDEBUG1();
 988         if ((so->so_oobmark == 0 &&
 989              (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
 990             so->so_options & SO_OOBINLINE ||
 991             tp->t_oobflags & TCPOOB_HADDATA) {
 992                 error = EINVAL;
 993                 goto out;
 994         }
 995         if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
 996                 error = EWOULDBLOCK;
 997                 goto out;
 998         }
 999         m->m_len = 1;
1000         *mtod(m, caddr_t) = tp->t_iobc;
1001         if ((flags & MSG_PEEK) == 0)
1002                 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1003
1004 out:
1005         TCPDEBUG2(PRU_RCVOOB);
1006         INP_WUNLOCK(inp);
1007         return (error);
1008 }
1009
1010 struct pr_usrreqs tcp_usrreqs = {
1011         .pru_abort =            tcp_usr_abort,
1012         .pru_accept =           tcp_usr_accept,
1013         .pru_attach =           tcp_usr_attach,
1014         .pru_bind =             tcp_usr_bind,
1015         .pru_connect =          tcp_usr_connect,
1016         .pru_control =          in_control,
1017         .pru_detach =           tcp_usr_detach,
1018         .pru_disconnect =       tcp_usr_disconnect,
1019         .pru_listen =           tcp_usr_listen,
1020         .pru_peeraddr =         in_getpeeraddr,
1021         .pru_rcvd =             tcp_usr_rcvd,
1022         .pru_rcvoob =           tcp_usr_rcvoob,
1023         .pru_send =             tcp_usr_send,
1024         .pru_shutdown =         tcp_usr_shutdown,
1025         .pru_sockaddr =         in_getsockaddr,
1026 #if 0
1027         .pru_soreceive =        soreceive_stream,
1028 #endif
1029         .pru_sosetlabel =       in_pcbsosetlabel,
1030         .pru_close =            tcp_usr_close,
1031 };
1032
1033 #ifdef INET6
1034 struct pr_usrreqs tcp6_usrreqs = {
1035         .pru_abort =            tcp_usr_abort,
1036         .pru_accept =           tcp6_usr_accept,
1037         .pru_attach =           tcp_usr_attach,
1038         .pru_bind =             tcp6_usr_bind,
1039         .pru_connect =          tcp6_usr_connect,
1040         .pru_control =          in6_control,
1041         .pru_detach =           tcp_usr_detach,
1042         .pru_disconnect =       tcp_usr_disconnect,
1043         .pru_listen =           tcp6_usr_listen,
1044         .pru_peeraddr =         in6_mapped_peeraddr,
1045         .pru_rcvd =             tcp_usr_rcvd,
1046         .pru_rcvoob =           tcp_usr_rcvoob,
1047         .pru_send =             tcp_usr_send,
1048         .pru_shutdown =         tcp_usr_shutdown,
1049         .pru_sockaddr =         in6_mapped_sockaddr,
1050 #if 0
1051         .pru_soreceive =        soreceive_stream,
1052 #endif
1053         .pru_sosetlabel =       in_pcbsosetlabel,
1054         .pru_close =            tcp_usr_close,
1055 };
1056 #endif /* INET6 */
1057
1058 /*
1059  * Common subroutine to open a TCP connection to remote host specified
1060  * by struct sockaddr_in in mbuf *nam.  Call in_pcbbind to assign a local
1061  * port number if needed.  Call in_pcbconnect_setup to do the routing and
1062  * to choose a local host address (interface).  If there is an existing
1063  * incarnation of the same connection in TIME-WAIT state and if the remote
1064  * host was sending CC options and if the connection duration was < MSL, then
1065  * truncate the previous TIME-WAIT state and proceed.
1066  * Initialize connection parameters and enter SYN-SENT state.
1067  */
1068 static int
1069 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1070 {
1071         struct inpcb *inp = tp->t_inpcb, *oinp;
1072         struct socket *so = inp->inp_socket;
1073         struct in_addr laddr;
1074         u_short lport;
1075         int error;
1076
1077         INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1078         INP_WLOCK_ASSERT(inp);
1079
1080         if (inp->inp_lport == 0) {
1081                 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1082                 if (error)
1083                         return error;
1084         }
1085
1086         /*
1087          * Cannot simply call in_pcbconnect, because there might be an
1088          * earlier incarnation of this same connection still in
1089          * TIME_WAIT state, creating an ADDRINUSE error.
1090          */
1091         laddr = inp->inp_laddr;
1092         lport = inp->inp_lport;
1093         error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
1094             &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
1095         if (error && oinp == NULL)
1096                 return error;
1097         if (oinp)
1098                 return EADDRINUSE;
1099         inp->inp_laddr = laddr;
1100         in_pcbrehash(inp);
1101
1102         /*
1103          * Compute window scaling to request:
1104          * Scale to fit into sweet spot.  See tcp_syncache.c.
1105          * XXX: This should move to tcp_output().
1106          */
1107         while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1108             (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1109                 tp->request_r_scale++;
1110
1111         soisconnecting(so);
1112         TCPSTAT_INC(tcps_connattempt);
1113         tp->t_state = TCPS_SYN_SENT;
1114         tcp_timer_activate(tp, TT_KEEP, tcp_keepinit);
1115         tp->iss = tcp_new_isn(tp);
1116         tp->t_bw_rtseq = tp->iss;
1117         tcp_sendseqinit(tp);
1118
1119         return 0;
1120 }
1121
1122 #ifdef INET6
1123 static int
1124 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1125 {
1126         struct inpcb *inp = tp->t_inpcb, *oinp;
1127         struct socket *so = inp->inp_socket;
1128         struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
1129         struct in6_addr addr6;
1130         int error;
1131
1132         INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1133         INP_WLOCK_ASSERT(inp);
1134
1135         if (inp->inp_lport == 0) {
1136                 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1137                 if (error)
1138                         return error;
1139         }
1140
1141         /*
1142          * Cannot simply call in_pcbconnect, because there might be an
1143          * earlier incarnation of this same connection still in
1144          * TIME_WAIT state, creating an ADDRINUSE error.
1145          * in6_pcbladdr() also handles scope zone IDs.
1146          */
1147         error = in6_pcbladdr(inp, nam, &addr6);
1148         if (error)
1149                 return error;
1150         oinp = in6_pcblookup_hash(inp->inp_pcbinfo,
1151                                   &sin6->sin6_addr, sin6->sin6_port,
1152                                   IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
1153                                   ? &addr6
1154                                   : &inp->in6p_laddr,
1155                                   inp->inp_lport,  0, NULL);
1156         if (oinp)
1157                 return EADDRINUSE;
1158         if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
1159                 inp->in6p_laddr = addr6;
1160         inp->in6p_faddr = sin6->sin6_addr;
1161         inp->inp_fport = sin6->sin6_port;
1162         /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
1163         inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
1164         if (inp->inp_flags & IN6P_AUTOFLOWLABEL)
1165                 inp->inp_flow |=
1166                     (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
1167         in_pcbrehash(inp);
1168
1169         /* Compute window scaling to request.  */
1170         while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1171             (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1172                 tp->request_r_scale++;
1173
1174         soisconnecting(so);
1175         TCPSTAT_INC(tcps_connattempt);
1176         tp->t_state = TCPS_SYN_SENT;
1177         tcp_timer_activate(tp, TT_KEEP, tcp_keepinit);
1178         tp->iss = tcp_new_isn(tp);
1179         tp->t_bw_rtseq = tp->iss;
1180         tcp_sendseqinit(tp);
1181
1182         return 0;
1183 }
1184 #endif /* INET6 */
1185
1186 /*
1187  * Export TCP internal state information via a struct tcp_info, based on the
1188  * Linux 2.6 API.  Not ABI compatible as our constants are mapped differently
1189  * (TCP state machine, etc).  We export all information using FreeBSD-native
1190  * constants -- for example, the numeric values for tcpi_state will differ
1191  * from Linux.
1192  */
1193 static void
1194 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1195 {
1196
1197         INP_WLOCK_ASSERT(tp->t_inpcb);
1198         bzero(ti, sizeof(*ti));
1199
1200         ti->tcpi_state = tp->t_state;
1201         if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
1202                 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1203         if (tp->t_flags & TF_SACK_PERMIT)
1204                 ti->tcpi_options |= TCPI_OPT_SACK;
1205         if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
1206                 ti->tcpi_options |= TCPI_OPT_WSCALE;
1207                 ti->tcpi_snd_wscale = tp->snd_scale;
1208                 ti->tcpi_rcv_wscale = tp->rcv_scale;
1209         }
1210
1211         ti->tcpi_rto = tp->t_rxtcur * tick;
1212         ti->tcpi_last_data_recv = (long)(ticks - (int)tp->t_rcvtime) * tick;
1213         ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
1214         ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
1215
1216         ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1217         ti->tcpi_snd_cwnd = tp->snd_cwnd;
1218
1219         /*
1220          * FreeBSD-specific extension fields for tcp_info.
1221          */
1222         ti->tcpi_rcv_space = tp->rcv_wnd;
1223         ti->tcpi_rcv_nxt = tp->rcv_nxt;
1224         ti->tcpi_snd_wnd = tp->snd_wnd;
1225         ti->tcpi_snd_bwnd = tp->snd_bwnd;
1226         ti->tcpi_snd_nxt = tp->snd_nxt;
1227         ti->tcpi_snd_mss = tp->t_maxseg;
1228         ti->tcpi_rcv_mss = tp->t_maxseg;
1229         if (tp->t_flags & TF_TOE)
1230                 ti->tcpi_options |= TCPI_OPT_TOE;
1231         ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
1232         ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
1233         ti->tcpi_snd_zerowin = tp->t_sndzerowin;
1234 }
1235
1236 /*
1237  * tcp_ctloutput() must drop the inpcb lock before performing copyin on
1238  * socket option arguments.  When it re-acquires the lock after the copy, it
1239  * has to revalidate that the connection is still valid for the socket
1240  * option.
1241  */
1242 #define INP_WLOCK_RECHECK(inp) do {                                     \
1243         INP_WLOCK(inp);                                                 \
1244         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {            \
1245                 INP_WUNLOCK(inp);                                       \
1246                 return (ECONNRESET);                                    \
1247         }                                                               \
1248         tp = intotcpcb(inp);                                            \
1249 } while(0)
1250
1251 int
1252 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1253 {
1254         int     error, opt, optval;
1255         struct  inpcb *inp;
1256         struct  tcpcb *tp;
1257         struct  tcp_info ti;
1258         char buf[TCP_CA_NAME_MAX];
1259         struct cc_algo *algo;
1260
1261         error = 0;
1262         inp = sotoinpcb(so);
1263         KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
1264         INP_WLOCK(inp);
1265         if (sopt->sopt_level != IPPROTO_TCP) {
1266 #ifdef INET6
1267                 if (inp->inp_vflag & INP_IPV6PROTO) {
1268                         INP_WUNLOCK(inp);
1269                         error = ip6_ctloutput(so, sopt);
1270                 } else {
1271 #endif /* INET6 */
1272                         INP_WUNLOCK(inp);
1273                         error = ip_ctloutput(so, sopt);
1274 #ifdef INET6
1275                 }
1276 #endif
1277                 return (error);
1278         }
1279         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1280                 INP_WUNLOCK(inp);
1281                 return (ECONNRESET);
1282         }
1283
1284         switch (sopt->sopt_dir) {
1285         case SOPT_SET:
1286                 switch (sopt->sopt_name) {
1287 #ifdef TCP_SIGNATURE
1288                 case TCP_MD5SIG:
1289                         INP_WUNLOCK(inp);
1290                         error = sooptcopyin(sopt, &optval, sizeof optval,
1291                             sizeof optval);
1292                         if (error)
1293                                 return (error);
1294
1295                         INP_WLOCK_RECHECK(inp);
1296                         if (optval > 0)
1297                                 tp->t_flags |= TF_SIGNATURE;
1298                         else
1299                                 tp->t_flags &= ~TF_SIGNATURE;
1300                         INP_WUNLOCK(inp);
1301                         break;
1302 #endif /* TCP_SIGNATURE */
1303                 case TCP_NODELAY:
1304                 case TCP_NOOPT:
1305                         INP_WUNLOCK(inp);
1306                         error = sooptcopyin(sopt, &optval, sizeof optval,
1307                             sizeof optval);
1308                         if (error)
1309                                 return (error);
1310
1311                         INP_WLOCK_RECHECK(inp);
1312                         switch (sopt->sopt_name) {
1313                         case TCP_NODELAY:
1314                                 opt = TF_NODELAY;
1315                                 break;
1316                         case TCP_NOOPT:
1317                                 opt = TF_NOOPT;
1318                                 break;
1319                         default:
1320                                 opt = 0; /* dead code to fool gcc */
1321                                 break;
1322                         }
1323
1324                         if (optval)
1325                                 tp->t_flags |= opt;
1326                         else
1327                                 tp->t_flags &= ~opt;
1328                         INP_WUNLOCK(inp);
1329                         break;
1330
1331                 case TCP_NOPUSH:
1332                         INP_WUNLOCK(inp);
1333                         error = sooptcopyin(sopt, &optval, sizeof optval,
1334                             sizeof optval);
1335                         if (error)
1336                                 return (error);
1337
1338                         INP_WLOCK_RECHECK(inp);
1339                         if (optval)
1340                                 tp->t_flags |= TF_NOPUSH;
1341                         else if (tp->t_flags & TF_NOPUSH) {
1342                                 tp->t_flags &= ~TF_NOPUSH;
1343                                 if (TCPS_HAVEESTABLISHED(tp->t_state))
1344                                         error = tcp_output(tp);
1345                         }
1346                         INP_WUNLOCK(inp);
1347                         break;
1348
1349                 case TCP_MAXSEG:
1350                         INP_WUNLOCK(inp);
1351                         error = sooptcopyin(sopt, &optval, sizeof optval,
1352                             sizeof optval);
1353                         if (error)
1354                                 return (error);
1355
1356                         INP_WLOCK_RECHECK(inp);
1357                         if (optval > 0 && optval <= tp->t_maxseg &&
1358                             optval + 40 >= V_tcp_minmss)
1359                                 tp->t_maxseg = optval;
1360                         else
1361                                 error = EINVAL;
1362                         INP_WUNLOCK(inp);
1363                         break;
1364
1365                 case TCP_INFO:
1366                         INP_WUNLOCK(inp);
1367                         error = EINVAL;
1368                         break;
1369
1370                 case TCP_CONGESTION:
1371                         INP_WUNLOCK(inp);
1372                         bzero(buf, sizeof(buf));
1373                         error = sooptcopyin(sopt, &buf, sizeof(buf), 1);
1374                         if (error)
1375                                 break;
1376                         INP_WLOCK_RECHECK(inp);
1377                         /*
1378                          * Return EINVAL if we can't find the requested cc algo.
1379                          */
1380                         error = EINVAL;
1381                         CC_LIST_RLOCK();
1382                         STAILQ_FOREACH(algo, &cc_list, entries) {
1383                                 if (strncmp(buf, algo->name, TCP_CA_NAME_MAX)
1384                                     == 0) {
1385                                         /* We've found the requested algo. */
1386                                         error = 0;
1387                                         /*
1388                                          * We hold a write lock over the tcb
1389                                          * so it's safe to do these things
1390                                          * without ordering concerns.
1391                                          */
1392                                         if (CC_ALGO(tp)->cb_destroy != NULL)
1393                                                 CC_ALGO(tp)->cb_destroy(tp->ccv);
1394                                         CC_ALGO(tp) = algo;
1395                                         /*
1396                                          * If something goes pear shaped
1397                                          * initialising the new algo,
1398                                          * fall back to newreno (which
1399                                          * does not require initialisation).
1400                                          */
1401                                         if (algo->cb_init != NULL)
1402                                                 if (algo->cb_init(tp->ccv) > 0) {
1403                                                         CC_ALGO(tp) = &newreno_cc_algo;
1404                                                         /*
1405                                                          * The only reason init
1406                                                          * should fail is
1407                                                          * because of malloc.
1408                                                          */
1409                                                         error = ENOMEM;
1410                                                 }
1411                                         break; /* Break the STAILQ_FOREACH. */
1412                                 }
1413                         }
1414                         CC_LIST_RUNLOCK();
1415                         INP_WUNLOCK(inp);
1416                         break;
1417
1418                 default:
1419                         INP_WUNLOCK(inp);
1420                         error = ENOPROTOOPT;
1421                         break;
1422                 }
1423                 break;
1424
1425         case SOPT_GET:
1426                 tp = intotcpcb(inp);
1427                 switch (sopt->sopt_name) {
1428 #ifdef TCP_SIGNATURE
1429                 case TCP_MD5SIG:
1430                         optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0;
1431                         INP_WUNLOCK(inp);
1432                         error = sooptcopyout(sopt, &optval, sizeof optval);
1433                         break;
1434 #endif
1435
1436                 case TCP_NODELAY:
1437                         optval = tp->t_flags & TF_NODELAY;
1438                         INP_WUNLOCK(inp);
1439                         error = sooptcopyout(sopt, &optval, sizeof optval);
1440                         break;
1441                 case TCP_MAXSEG:
1442                         optval = tp->t_maxseg;
1443                         INP_WUNLOCK(inp);
1444                         error = sooptcopyout(sopt, &optval, sizeof optval);
1445                         break;
1446                 case TCP_NOOPT:
1447                         optval = tp->t_flags & TF_NOOPT;
1448                         INP_WUNLOCK(inp);
1449                         error = sooptcopyout(sopt, &optval, sizeof optval);
1450                         break;
1451                 case TCP_NOPUSH:
1452                         optval = tp->t_flags & TF_NOPUSH;
1453                         INP_WUNLOCK(inp);
1454                         error = sooptcopyout(sopt, &optval, sizeof optval);
1455                         break;
1456                 case TCP_INFO:
1457                         tcp_fill_info(tp, &ti);
1458                         INP_WUNLOCK(inp);
1459                         error = sooptcopyout(sopt, &ti, sizeof ti);
1460                         break;
1461                 case TCP_CONGESTION:
1462                         bzero(buf, sizeof(buf));
1463                         strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX);
1464                         INP_WUNLOCK(inp);
1465                         error = sooptcopyout(sopt, buf, TCP_CA_NAME_MAX);
1466                         break;
1467                 default:
1468                         INP_WUNLOCK(inp);
1469                         error = ENOPROTOOPT;
1470                         break;
1471                 }
1472                 break;
1473         }
1474         return (error);
1475 }
1476 #undef INP_WLOCK_RECHECK
1477
1478 /*
1479  * tcp_sendspace and tcp_recvspace are the default send and receive window
1480  * sizes, respectively.  These are obsolescent (this information should
1481  * be set by the route).
1482  */
1483 u_long  tcp_sendspace = 1024*32;
1484 SYSCTL_ULONG(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW,
1485     &tcp_sendspace , 0, "Maximum outgoing TCP datagram size");
1486 u_long  tcp_recvspace = 1024*64;
1487 SYSCTL_ULONG(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
1488     &tcp_recvspace , 0, "Maximum incoming TCP datagram size");
1489
1490 /*
1491  * Attach TCP protocol to socket, allocating
1492  * internet protocol control block, tcp control block,
1493  * bufer space, and entering LISTEN state if to accept connections.
1494  */
1495 static int
1496 tcp_attach(struct socket *so)
1497 {
1498         struct tcpcb *tp;
1499         struct inpcb *inp;
1500         int error;
1501
1502         if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1503                 error = soreserve(so, tcp_sendspace, tcp_recvspace);
1504                 if (error)
1505                         return (error);
1506         }
1507         so->so_rcv.sb_flags |= SB_AUTOSIZE;
1508         so->so_snd.sb_flags |= SB_AUTOSIZE;
1509         INP_INFO_WLOCK(&V_tcbinfo);
1510         error = in_pcballoc(so, &V_tcbinfo);
1511         if (error) {
1512                 INP_INFO_WUNLOCK(&V_tcbinfo);
1513                 return (error);
1514         }
1515         inp = sotoinpcb(so);
1516 #ifdef INET6
1517         if (inp->inp_vflag & INP_IPV6PROTO) {
1518                 inp->inp_vflag |= INP_IPV6;
1519                 inp->in6p_hops = -1;    /* use kernel default */
1520         }
1521         else
1522 #endif
1523         inp->inp_vflag |= INP_IPV4;
1524         tp = tcp_newtcpcb(inp);
1525         if (tp == NULL) {
1526                 in_pcbdetach(inp);
1527                 in_pcbfree(inp);
1528                 INP_INFO_WUNLOCK(&V_tcbinfo);
1529                 return (ENOBUFS);
1530         }
1531         tp->t_state = TCPS_CLOSED;
1532         INP_WUNLOCK(inp);
1533         INP_INFO_WUNLOCK(&V_tcbinfo);
1534         return (0);
1535 }
1536
1537 /*
1538  * Initiate (or continue) disconnect.
1539  * If embryonic state, just send reset (once).
1540  * If in ``let data drain'' option and linger null, just drop.
1541  * Otherwise (hard), mark socket disconnecting and drop
1542  * current input data; switch states based on user close, and
1543  * send segment to peer (with FIN).
1544  */
1545 static void
1546 tcp_disconnect(struct tcpcb *tp)
1547 {
1548         struct inpcb *inp = tp->t_inpcb;
1549         struct socket *so = inp->inp_socket;
1550
1551         INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1552         INP_WLOCK_ASSERT(inp);
1553
1554         /*
1555          * Neither tcp_close() nor tcp_drop() should return NULL, as the
1556          * socket is still open.
1557          */
1558         if (tp->t_state < TCPS_ESTABLISHED) {
1559                 tp = tcp_close(tp);
1560                 KASSERT(tp != NULL,
1561                     ("tcp_disconnect: tcp_close() returned NULL"));
1562         } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
1563                 tp = tcp_drop(tp, 0);
1564                 KASSERT(tp != NULL,
1565                     ("tcp_disconnect: tcp_drop() returned NULL"));
1566         } else {
1567                 soisdisconnecting(so);
1568                 sbflush(&so->so_rcv);
1569                 tcp_usrclosed(tp);
1570                 if (!(inp->inp_flags & INP_DROPPED))
1571                         tcp_output_disconnect(tp);
1572         }
1573 }
1574
1575 /*
1576  * User issued close, and wish to trail through shutdown states:
1577  * if never received SYN, just forget it.  If got a SYN from peer,
1578  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1579  * If already got a FIN from peer, then almost done; go to LAST_ACK
1580  * state.  In all other cases, have already sent FIN to peer (e.g.
1581  * after PRU_SHUTDOWN), and just have to play tedious game waiting
1582  * for peer to send FIN or not respond to keep-alives, etc.
1583  * We can let the user exit from the close as soon as the FIN is acked.
1584  */
1585 static void
1586 tcp_usrclosed(struct tcpcb *tp)
1587 {
1588
1589         INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1590         INP_WLOCK_ASSERT(tp->t_inpcb);
1591
1592         switch (tp->t_state) {
1593         case TCPS_LISTEN:
1594                 tcp_offload_listen_close(tp);
1595                 /* FALLTHROUGH */
1596         case TCPS_CLOSED:
1597                 tp->t_state = TCPS_CLOSED;
1598                 tp = tcp_close(tp);
1599                 /*
1600                  * tcp_close() should never return NULL here as the socket is
1601                  * still open.
1602                  */
1603                 KASSERT(tp != NULL,
1604                     ("tcp_usrclosed: tcp_close() returned NULL"));
1605                 break;
1606
1607         case TCPS_SYN_SENT:
1608         case TCPS_SYN_RECEIVED:
1609                 tp->t_flags |= TF_NEEDFIN;
1610                 break;
1611
1612         case TCPS_ESTABLISHED:
1613                 tp->t_state = TCPS_FIN_WAIT_1;
1614                 break;
1615
1616         case TCPS_CLOSE_WAIT:
1617                 tp->t_state = TCPS_LAST_ACK;
1618                 break;
1619         }
1620         if (tp->t_state >= TCPS_FIN_WAIT_2) {
1621                 soisdisconnected(tp->t_inpcb->inp_socket);
1622                 /* Prevent the connection hanging in FIN_WAIT_2 forever. */
1623                 if (tp->t_state == TCPS_FIN_WAIT_2) {
1624                         int timeout;
1625
1626                         timeout = (tcp_fast_finwait2_recycle) ?
1627                             tcp_finwait2_timeout : tcp_maxidle;
1628                         tcp_timer_activate(tp, TT_2MSL, timeout);
1629                 }
1630         }
1631 }
1632
1633 #ifdef DDB
1634 static void
1635 db_print_indent(int indent)
1636 {
1637         int i;
1638
1639         for (i = 0; i < indent; i++)
1640                 db_printf(" ");
1641 }
1642
1643 static void
1644 db_print_tstate(int t_state)
1645 {
1646
1647         switch (t_state) {
1648         case TCPS_CLOSED:
1649                 db_printf("TCPS_CLOSED");
1650                 return;
1651
1652         case TCPS_LISTEN:
1653                 db_printf("TCPS_LISTEN");
1654                 return;
1655
1656         case TCPS_SYN_SENT:
1657                 db_printf("TCPS_SYN_SENT");
1658                 return;
1659
1660         case TCPS_SYN_RECEIVED:
1661                 db_printf("TCPS_SYN_RECEIVED");
1662                 return;
1663
1664         case TCPS_ESTABLISHED:
1665                 db_printf("TCPS_ESTABLISHED");
1666                 return;
1667
1668         case TCPS_CLOSE_WAIT:
1669                 db_printf("TCPS_CLOSE_WAIT");
1670                 return;
1671
1672         case TCPS_FIN_WAIT_1:
1673                 db_printf("TCPS_FIN_WAIT_1");
1674                 return;
1675
1676         case TCPS_CLOSING:
1677                 db_printf("TCPS_CLOSING");
1678                 return;
1679
1680         case TCPS_LAST_ACK:
1681                 db_printf("TCPS_LAST_ACK");
1682                 return;
1683
1684         case TCPS_FIN_WAIT_2:
1685                 db_printf("TCPS_FIN_WAIT_2");
1686                 return;
1687
1688         case TCPS_TIME_WAIT:
1689                 db_printf("TCPS_TIME_WAIT");
1690                 return;
1691
1692         default:
1693                 db_printf("unknown");
1694                 return;
1695         }
1696 }
1697
1698 static void
1699 db_print_tflags(u_int t_flags)
1700 {
1701         int comma;
1702
1703         comma = 0;
1704         if (t_flags & TF_ACKNOW) {
1705                 db_printf("%sTF_ACKNOW", comma ? ", " : "");
1706                 comma = 1;
1707         }
1708         if (t_flags & TF_DELACK) {
1709                 db_printf("%sTF_DELACK", comma ? ", " : "");
1710                 comma = 1;
1711         }
1712         if (t_flags & TF_NODELAY) {
1713                 db_printf("%sTF_NODELAY", comma ? ", " : "");
1714                 comma = 1;
1715         }
1716         if (t_flags & TF_NOOPT) {
1717                 db_printf("%sTF_NOOPT", comma ? ", " : "");
1718                 comma = 1;
1719         }
1720         if (t_flags & TF_SENTFIN) {
1721                 db_printf("%sTF_SENTFIN", comma ? ", " : "");
1722                 comma = 1;
1723         }
1724         if (t_flags & TF_REQ_SCALE) {
1725                 db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
1726                 comma = 1;
1727         }
1728         if (t_flags & TF_RCVD_SCALE) {
1729                 db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
1730                 comma = 1;
1731         }
1732         if (t_flags & TF_REQ_TSTMP) {
1733                 db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
1734                 comma = 1;
1735         }
1736         if (t_flags & TF_RCVD_TSTMP) {
1737                 db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
1738                 comma = 1;
1739         }
1740         if (t_flags & TF_SACK_PERMIT) {
1741                 db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
1742                 comma = 1;
1743         }
1744         if (t_flags & TF_NEEDSYN) {
1745                 db_printf("%sTF_NEEDSYN", comma ? ", " : "");
1746                 comma = 1;
1747         }
1748         if (t_flags & TF_NEEDFIN) {
1749                 db_printf("%sTF_NEEDFIN", comma ? ", " : "");
1750                 comma = 1;
1751         }
1752         if (t_flags & TF_NOPUSH) {
1753                 db_printf("%sTF_NOPUSH", comma ? ", " : "");
1754                 comma = 1;
1755         }
1756         if (t_flags & TF_MORETOCOME) {
1757                 db_printf("%sTF_MORETOCOME", comma ? ", " : "");
1758                 comma = 1;
1759         }
1760         if (t_flags & TF_LQ_OVERFLOW) {
1761                 db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : "");
1762                 comma = 1;
1763         }
1764         if (t_flags & TF_LASTIDLE) {
1765                 db_printf("%sTF_LASTIDLE", comma ? ", " : "");
1766                 comma = 1;
1767         }
1768         if (t_flags & TF_RXWIN0SENT) {
1769                 db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
1770                 comma = 1;
1771         }
1772         if (t_flags & TF_FASTRECOVERY) {
1773                 db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
1774                 comma = 1;
1775         }
1776         if (t_flags & TF_CONGRECOVERY) {
1777                 db_printf("%sTF_CONGRECOVERY", comma ? ", " : "");
1778                 comma = 1;
1779         }
1780         if (t_flags & TF_WASFRECOVERY) {
1781                 db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
1782                 comma = 1;
1783         }
1784         if (t_flags & TF_SIGNATURE) {
1785                 db_printf("%sTF_SIGNATURE", comma ? ", " : "");
1786                 comma = 1;
1787         }
1788         if (t_flags & TF_FORCEDATA) {
1789                 db_printf("%sTF_FORCEDATA", comma ? ", " : "");
1790                 comma = 1;
1791         }
1792         if (t_flags & TF_TSO) {
1793                 db_printf("%sTF_TSO", comma ? ", " : "");
1794                 comma = 1;
1795         }
1796         if (t_flags & TF_ECN_PERMIT) {
1797                 db_printf("%sTF_ECN_PERMIT", comma ? ", " : "");
1798                 comma = 1;
1799         }
1800 }
1801
1802 static void
1803 db_print_toobflags(char t_oobflags)
1804 {
1805         int comma;
1806
1807         comma = 0;
1808         if (t_oobflags & TCPOOB_HAVEDATA) {
1809                 db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
1810                 comma = 1;
1811         }
1812         if (t_oobflags & TCPOOB_HADDATA) {
1813                 db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
1814                 comma = 1;
1815         }
1816 }
1817
1818 static void
1819 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
1820 {
1821
1822         db_print_indent(indent);
1823         db_printf("%s at %p\n", name, tp);
1824
1825         indent += 2;
1826
1827         db_print_indent(indent);
1828         db_printf("t_segq first: %p   t_segqlen: %d   t_dupacks: %d\n",
1829            LIST_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks);
1830
1831         db_print_indent(indent);
1832         db_printf("tt_rexmt: %p   tt_persist: %p   tt_keep: %p\n",
1833             &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep);
1834
1835         db_print_indent(indent);
1836         db_printf("tt_2msl: %p   tt_delack: %p   t_inpcb: %p\n", &tp->t_timers->tt_2msl,
1837             &tp->t_timers->tt_delack, tp->t_inpcb);
1838
1839         db_print_indent(indent);
1840         db_printf("t_state: %d (", tp->t_state);
1841         db_print_tstate(tp->t_state);
1842         db_printf(")\n");
1843
1844         db_print_indent(indent);
1845         db_printf("t_flags: 0x%x (", tp->t_flags);
1846         db_print_tflags(tp->t_flags);
1847         db_printf(")\n");
1848
1849         db_print_indent(indent);
1850         db_printf("snd_una: 0x%08x   snd_max: 0x%08x   snd_nxt: x0%08x\n",
1851             tp->snd_una, tp->snd_max, tp->snd_nxt);
1852
1853         db_print_indent(indent);
1854         db_printf("snd_up: 0x%08x   snd_wl1: 0x%08x   snd_wl2: 0x%08x\n",
1855            tp->snd_up, tp->snd_wl1, tp->snd_wl2);
1856
1857         db_print_indent(indent);
1858         db_printf("iss: 0x%08x   irs: 0x%08x   rcv_nxt: 0x%08x\n",
1859             tp->iss, tp->irs, tp->rcv_nxt);
1860
1861         db_print_indent(indent);
1862         db_printf("rcv_adv: 0x%08x   rcv_wnd: %lu   rcv_up: 0x%08x\n",
1863             tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
1864
1865         db_print_indent(indent);
1866         db_printf("snd_wnd: %lu   snd_cwnd: %lu   snd_bwnd: %lu\n",
1867            tp->snd_wnd, tp->snd_cwnd, tp->snd_bwnd);
1868
1869         db_print_indent(indent);
1870         db_printf("snd_ssthresh: %lu   snd_bandwidth: %lu   snd_recover: "
1871             "0x%08x\n", tp->snd_ssthresh, tp->snd_bandwidth,
1872             tp->snd_recover);
1873
1874         db_print_indent(indent);
1875         db_printf("t_maxopd: %u   t_rcvtime: %u   t_startime: %u\n",
1876             tp->t_maxopd, tp->t_rcvtime, tp->t_starttime);
1877
1878         db_print_indent(indent);
1879         db_printf("t_rttime: %u   t_rtsq: 0x%08x   t_bw_rtttime: %u\n",
1880             tp->t_rtttime, tp->t_rtseq, tp->t_bw_rtttime);
1881
1882         db_print_indent(indent);
1883         db_printf("t_bw_rtseq: 0x%08x   t_rxtcur: %d   t_maxseg: %u   "
1884             "t_srtt: %d\n", tp->t_bw_rtseq, tp->t_rxtcur, tp->t_maxseg,
1885             tp->t_srtt);
1886
1887         db_print_indent(indent);
1888         db_printf("t_rttvar: %d   t_rxtshift: %d   t_rttmin: %u   "
1889             "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin,
1890             tp->t_rttbest);
1891
1892         db_print_indent(indent);
1893         db_printf("t_rttupdated: %lu   max_sndwnd: %lu   t_softerror: %d\n",
1894             tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
1895
1896         db_print_indent(indent);
1897         db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
1898         db_print_toobflags(tp->t_oobflags);
1899         db_printf(")   t_iobc: 0x%02x\n", tp->t_iobc);
1900
1901         db_print_indent(indent);
1902         db_printf("snd_scale: %u   rcv_scale: %u   request_r_scale: %u\n",
1903             tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
1904
1905         db_print_indent(indent);
1906         db_printf("ts_recent: %u   ts_recent_age: %u\n",
1907             tp->ts_recent, tp->ts_recent_age);
1908
1909         db_print_indent(indent);
1910         db_printf("ts_offset: %u   last_ack_sent: 0x%08x   snd_cwnd_prev: "
1911             "%lu\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
1912
1913         db_print_indent(indent);
1914         db_printf("snd_ssthresh_prev: %lu   snd_recover_prev: 0x%08x   "
1915             "t_badrxtwin: %u\n", tp->snd_ssthresh_prev,
1916             tp->snd_recover_prev, tp->t_badrxtwin);
1917
1918         db_print_indent(indent);
1919         db_printf("snd_numholes: %d  snd_holes first: %p\n",
1920             tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
1921
1922         db_print_indent(indent);
1923         db_printf("snd_fack: 0x%08x   rcv_numsacks: %d   sack_newdata: "
1924             "0x%08x\n", tp->snd_fack, tp->rcv_numsacks, tp->sack_newdata);
1925
1926         /* Skip sackblks, sackhint. */
1927
1928         db_print_indent(indent);
1929         db_printf("t_rttlow: %d   rfbuf_ts: %u   rfbuf_cnt: %d\n",
1930             tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
1931 }
1932
1933 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
1934 {
1935         struct tcpcb *tp;
1936
1937         if (!have_addr) {
1938                 db_printf("usage: show tcpcb <addr>\n");
1939                 return;
1940         }
1941         tp = (struct tcpcb *)addr;
1942
1943         db_print_tcpcb(tp, "tcpcb", 0);
1944 }
1945 #endif