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/malloc.h>
  49 #include <sys/kernel.h>
  50 #include <sys/sysctl.h>
  51 #include <sys/mbuf.h>
  52 #ifdef INET6
  53 #include <sys/domain.h>
  54 #endif /* INET6 */
  55 #include <sys/socket.h>
  56 #include <sys/socketvar.h>
  57 #include <sys/protosw.h>
  58 #include <sys/proc.h>
  59 #include <sys/jail.h>
  60
  61 #ifdef DDB
  62 #include <ddb/ddb.h>
  63 #endif
  64
  65 #include <net/if.h>
  66 #include <net/route.h>
  67 #include <net/vnet.h>
  68
  69 #include <netinet/cc.h>
  70 #include <netinet/in.h>
  71 #include <netinet/in_pcb.h>
  72 #include <netinet/in_systm.h>
  73 #include <netinet/in_var.h>
  74 #include <netinet/ip_var.h>
  75 #ifdef INET6
  76 #include <netinet/ip6.h>
  77 #include <netinet6/in6_pcb.h>
  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 #ifdef INET
  96 static int      tcp_connect(struct tcpcb *, struct sockaddr *,
  97                     struct thread *td);
  98 #endif /* INET */
  99 #ifdef INET6
 100 static int      tcp6_connect(struct tcpcb *, struct sockaddr *,
 101                     struct thread *td);
 102 #endif /* INET6 */
 103 static void     tcp_disconnect(struct tcpcb *);
 104 static void     tcp_usrclosed(struct tcpcb *);
 105 static void     tcp_fill_info(struct tcpcb *, struct tcp_info *);
 106
 107 #ifdef TCPDEBUG
 108 #define TCPDEBUG0       int ostate = 0
 109 #define TCPDEBUG1()     ostate = tp ? tp->t_state : 0
 110 #define TCPDEBUG2(req)  if (tp && (so->so_options & SO_DEBUG)) \
 111                                 tcp_trace(TA_USER, ostate, tp, 0, 0, req)
 112 #else
 113 #define TCPDEBUG0
 114 #define TCPDEBUG1()
 115 #define TCPDEBUG2(req)
 116 #endif
 117
 118 /*
 119  * TCP attaches to socket via pru_attach(), reserving space,
 120  * and an internet control block.
 121  */
 122 static int
 123 tcp_usr_attach(struct socket *so, int proto, struct thread *td)
 124 {
 125         struct inpcb *inp;
 126         struct tcpcb *tp = NULL;
 127         int error;
 128         TCPDEBUG0;
 129
 130         inp = sotoinpcb(so);
 131         KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL"));
 132         TCPDEBUG1();
 133
 134         error = tcp_attach(so);
 135         if (error)
 136                 goto out;
 137
 138         if ((so->so_options & SO_LINGER) && so->so_linger == 0)
 139                 so->so_linger = TCP_LINGERTIME;
 140
 141         inp = sotoinpcb(so);
 142         tp = intotcpcb(inp);
 143 out:
 144         TCPDEBUG2(PRU_ATTACH);
 145         return error;
 146 }
 147
 148 /*
 149  * tcp_detach is called when the socket layer loses its final reference
 150  * to the socket, be it a file descriptor reference, a reference from TCP,
 151  * etc.  At this point, there is only one case in which we will keep around
 152  * inpcb state: time wait.
 153  *
 154  * This function can probably be re-absorbed back into tcp_usr_detach() now
 155  * that there is a single detach path.
 156  */
 157 static void
 158 tcp_detach(struct socket *so, struct inpcb *inp)
 159 {
 160         struct tcpcb *tp;
 161
 162         INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
 163         INP_WLOCK_ASSERT(inp);
 164
 165         KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp"));
 166         KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so"));
 167
 168         tp = intotcpcb(inp);
 169
 170         if (inp->inp_flags & INP_TIMEWAIT) {
 171                 /*
 172                  * There are two cases to handle: one in which the time wait
 173                  * state is being discarded (INP_DROPPED), and one in which
 174                  * this connection will remain in timewait.  In the former,
 175                  * it is time to discard all state (except tcptw, which has
 176                  * already been discarded by the timewait close code, which
 177                  * should be further up the call stack somewhere).  In the
 178                  * latter case, we detach from the socket, but leave the pcb
 179                  * present until timewait ends.
 180                  *
 181                  * XXXRW: Would it be cleaner to free the tcptw here?
 182                  */
 183                 if (inp->inp_flags & INP_DROPPED) {
 184                         KASSERT(tp == NULL, ("tcp_detach: INP_TIMEWAIT && "
 185                             "INP_DROPPED && tp != NULL"));
 186                         in_pcbdetach(inp);
 187                         in_pcbfree(inp);
 188                 } else {
 189                         in_pcbdetach(inp);
 190                         INP_WUNLOCK(inp);
 191                 }
 192         } else {
 193                 /*
 194                  * If the connection is not in timewait, we consider two
 195                  * two conditions: one in which no further processing is
 196                  * necessary (dropped || embryonic), and one in which TCP is
 197                  * not yet done, but no longer requires the socket, so the
 198                  * pcb will persist for the time being.
 199                  *
 200                  * XXXRW: Does the second case still occur?
 201                  */
 202                 if (inp->inp_flags & INP_DROPPED ||
 203                     tp->t_state < TCPS_SYN_SENT) {
 204                         tcp_discardcb(tp);
 205                         in_pcbdetach(inp);
 206                         in_pcbfree(inp);
 207                 } else
 208                         in_pcbdetach(inp);
 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 #ifdef INET
 235 /*
 236  * Give the socket an address.
 237  */
 238 static int
 239 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 240 {
 241         int error = 0;
 242         struct inpcb *inp;
 243         struct tcpcb *tp = NULL;
 244         struct sockaddr_in *sinp;
 245
 246         sinp = (struct sockaddr_in *)nam;
 247         if (nam->sa_len != sizeof (*sinp))
 248                 return (EINVAL);
 249         /*
 250          * Must check for multicast addresses and disallow binding
 251          * to them.
 252          */
 253         if (sinp->sin_family == AF_INET &&
 254             IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
 255                 return (EAFNOSUPPORT);
 256
 257         TCPDEBUG0;
 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         INP_HASH_WLOCK(&V_tcbinfo);
 268         error = in_pcbbind(inp, nam, td->td_ucred);
 269         INP_HASH_WUNLOCK(&V_tcbinfo);
 270 out:
 271         TCPDEBUG2(PRU_BIND);
 272         INP_WUNLOCK(inp);
 273
 274         return (error);
 275 }
 276 #endif /* INET */
 277
 278 #ifdef INET6
 279 static int
 280 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 281 {
 282         int error = 0;
 283         struct inpcb *inp;
 284         struct tcpcb *tp = NULL;
 285         struct sockaddr_in6 *sin6p;
 286
 287         sin6p = (struct sockaddr_in6 *)nam;
 288         if (nam->sa_len != sizeof (*sin6p))
 289                 return (EINVAL);
 290         /*
 291          * Must check for multicast addresses and disallow binding
 292          * to them.
 293          */
 294         if (sin6p->sin6_family == AF_INET6 &&
 295             IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
 296                 return (EAFNOSUPPORT);
 297
 298         TCPDEBUG0;
 299         inp = sotoinpcb(so);
 300         KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL"));
 301         INP_WLOCK(inp);
 302         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 303                 error = EINVAL;
 304                 goto out;
 305         }
 306         tp = intotcpcb(inp);
 307         TCPDEBUG1();
 308         INP_HASH_WLOCK(&V_tcbinfo);
 309         inp->inp_vflag &= ~INP_IPV4;
 310         inp->inp_vflag |= INP_IPV6;
 311 #ifdef INET
 312         if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
 313                 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
 314                         inp->inp_vflag |= INP_IPV4;
 315                 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
 316                         struct sockaddr_in sin;
 317
 318                         in6_sin6_2_sin(&sin, sin6p);
 319                         inp->inp_vflag |= INP_IPV4;
 320                         inp->inp_vflag &= ~INP_IPV6;
 321                         error = in_pcbbind(inp, (struct sockaddr *)&sin,
 322                             td->td_ucred);
 323                         INP_HASH_WUNLOCK(&V_tcbinfo);
 324                         goto out;
 325                 }
 326         }
 327 #endif
 328         error = in6_pcbbind(inp, nam, td->td_ucred);
 329         INP_HASH_WUNLOCK(&V_tcbinfo);
 330 out:
 331         TCPDEBUG2(PRU_BIND);
 332         INP_WUNLOCK(inp);
 333         return (error);
 334 }
 335 #endif /* INET6 */
 336
 337 #ifdef INET
 338 /*
 339  * Prepare to accept connections.
 340  */
 341 static int
 342 tcp_usr_listen(struct socket *so, int backlog, struct thread *td)
 343 {
 344         int error = 0;
 345         struct inpcb *inp;
 346         struct tcpcb *tp = NULL;
 347
 348         TCPDEBUG0;
 349         inp = sotoinpcb(so);
 350         KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL"));
 351         INP_WLOCK(inp);
 352         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 353                 error = EINVAL;
 354                 goto out;
 355         }
 356         tp = intotcpcb(inp);
 357         TCPDEBUG1();
 358         SOCK_LOCK(so);
 359         error = solisten_proto_check(so);
 360         INP_HASH_WLOCK(&V_tcbinfo);
 361         if (error == 0 && inp->inp_lport == 0)
 362                 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
 363         INP_HASH_WUNLOCK(&V_tcbinfo);
 364         if (error == 0) {
 365                 tp->t_state = TCPS_LISTEN;
 366                 solisten_proto(so, backlog);
 367                 tcp_offload_listen_open(tp);
 368         }
 369         SOCK_UNLOCK(so);
 370
 371 out:
 372         TCPDEBUG2(PRU_LISTEN);
 373         INP_WUNLOCK(inp);
 374         return (error);
 375 }
 376 #endif /* INET */
 377
 378 #ifdef INET6
 379 static int
 380 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td)
 381 {
 382         int error = 0;
 383         struct inpcb *inp;
 384         struct tcpcb *tp = NULL;
 385
 386         TCPDEBUG0;
 387         inp = sotoinpcb(so);
 388         KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL"));
 389         INP_WLOCK(inp);
 390         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 391                 error = EINVAL;
 392                 goto out;
 393         }
 394         tp = intotcpcb(inp);
 395         TCPDEBUG1();
 396         SOCK_LOCK(so);
 397         error = solisten_proto_check(so);
 398         INP_HASH_WLOCK(&V_tcbinfo);
 399         if (error == 0 && inp->inp_lport == 0) {
 400                 inp->inp_vflag &= ~INP_IPV4;
 401                 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
 402                         inp->inp_vflag |= INP_IPV4;
 403                 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
 404         }
 405         INP_HASH_WUNLOCK(&V_tcbinfo);
 406         if (error == 0) {
 407                 tp->t_state = TCPS_LISTEN;
 408                 solisten_proto(so, backlog);
 409         }
 410         SOCK_UNLOCK(so);
 411
 412 out:
 413         TCPDEBUG2(PRU_LISTEN);
 414         INP_WUNLOCK(inp);
 415         return (error);
 416 }
 417 #endif /* INET6 */
 418
 419 #ifdef INET
 420 /*
 421  * Initiate connection to peer.
 422  * Create a template for use in transmissions on this connection.
 423  * Enter SYN_SENT state, and mark socket as connecting.
 424  * Start keep-alive timer, and seed output sequence space.
 425  * Send initial segment on connection.
 426  */
 427 static int
 428 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 429 {
 430         int error = 0;
 431         struct inpcb *inp;
 432         struct tcpcb *tp = NULL;
 433         struct sockaddr_in *sinp;
 434
 435         sinp = (struct sockaddr_in *)nam;
 436         if (nam->sa_len != sizeof (*sinp))
 437                 return (EINVAL);
 438         /*
 439          * Must disallow TCP ``connections'' to multicast addresses.
 440          */
 441         if (sinp->sin_family == AF_INET
 442             && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
 443                 return (EAFNOSUPPORT);
 444         if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0)
 445                 return (error);
 446
 447         TCPDEBUG0;
 448         inp = sotoinpcb(so);
 449         KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL"));
 450         INP_WLOCK(inp);
 451         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 452                 error = EINVAL;
 453                 goto out;
 454         }
 455         tp = intotcpcb(inp);
 456         TCPDEBUG1();
 457         if ((error = tcp_connect(tp, nam, td)) != 0)
 458                 goto out;
 459         error = tcp_output_connect(so, nam);
 460 out:
 461         TCPDEBUG2(PRU_CONNECT);
 462         INP_WUNLOCK(inp);
 463         return (error);
 464 }
 465 #endif /* INET */
 466
 467 #ifdef INET6
 468 static int
 469 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 470 {
 471         int error = 0;
 472         struct inpcb *inp;
 473         struct tcpcb *tp = NULL;
 474         struct sockaddr_in6 *sin6p;
 475
 476         TCPDEBUG0;
 477
 478         sin6p = (struct sockaddr_in6 *)nam;
 479         if (nam->sa_len != sizeof (*sin6p))
 480                 return (EINVAL);
 481         /*
 482          * Must disallow TCP ``connections'' to multicast addresses.
 483          */
 484         if (sin6p->sin6_family == AF_INET6
 485             && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
 486                 return (EAFNOSUPPORT);
 487
 488         inp = sotoinpcb(so);
 489         KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL"));
 490         INP_WLOCK(inp);
 491         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 492                 error = EINVAL;
 493                 goto out;
 494         }
 495         tp = intotcpcb(inp);
 496         TCPDEBUG1();
 497 #ifdef INET
 498         /*
 499          * XXXRW: Some confusion: V4/V6 flags relate to binding, and
 500          * therefore probably require the hash lock, which isn't held here.
 501          * Is this a significant problem?
 502          */
 503         if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
 504                 struct sockaddr_in sin;
 505
 506                 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
 507                         error = EINVAL;
 508                         goto out;
 509                 }
 510
 511                 in6_sin6_2_sin(&sin, sin6p);
 512                 inp->inp_vflag |= INP_IPV4;
 513                 inp->inp_vflag &= ~INP_IPV6;
 514                 if ((error = prison_remote_ip4(td->td_ucred,
 515                     &sin.sin_addr)) != 0)
 516                         goto out;
 517                 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
 518                         goto out;
 519                 error = tcp_output_connect(so, nam);
 520                 goto out;
 521         }
 522 #endif
 523         inp->inp_vflag &= ~INP_IPV4;
 524         inp->inp_vflag |= INP_IPV6;
 525         inp->inp_inc.inc_flags |= INC_ISIPV6;
 526         if ((error = prison_remote_ip6(td->td_ucred, &sin6p->sin6_addr)) != 0)
 527                 goto out;
 528         if ((error = tcp6_connect(tp, nam, td)) != 0)
 529                 goto out;
 530         error = tcp_output_connect(so, nam);
 531
 532 out:
 533         TCPDEBUG2(PRU_CONNECT);
 534         INP_WUNLOCK(inp);
 535         return (error);
 536 }
 537 #endif /* INET6 */
 538
 539 /*
 540  * Initiate disconnect from peer.
 541  * If connection never passed embryonic stage, just drop;
 542  * else if don't need to let data drain, then can just drop anyways,
 543  * else have to begin TCP shutdown process: mark socket disconnecting,
 544  * drain unread data, state switch to reflect user close, and
 545  * send segment (e.g. FIN) to peer.  Socket will be really disconnected
 546  * when peer sends FIN and acks ours.
 547  *
 548  * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
 549  */
 550 static int
 551 tcp_usr_disconnect(struct socket *so)
 552 {
 553         struct inpcb *inp;
 554         struct tcpcb *tp = NULL;
 555         int error = 0;
 556
 557         TCPDEBUG0;
 558         INP_INFO_WLOCK(&V_tcbinfo);
 559         inp = sotoinpcb(so);
 560         KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
 561         INP_WLOCK(inp);
 562         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 563                 error = ECONNRESET;
 564                 goto out;
 565         }
 566         tp = intotcpcb(inp);
 567         TCPDEBUG1();
 568         tcp_disconnect(tp);
 569 out:
 570         TCPDEBUG2(PRU_DISCONNECT);
 571         INP_WUNLOCK(inp);
 572         INP_INFO_WUNLOCK(&V_tcbinfo);
 573         return (error);
 574 }
 575
 576 #ifdef INET
 577 /*
 578  * Accept a connection.  Essentially all the work is done at higher levels;
 579  * just return the address of the peer, storing through addr.
 580  *
 581  * The rationale for acquiring the tcbinfo lock here is somewhat complicated,
 582  * and is described in detail in the commit log entry for r175612.  Acquiring
 583  * it delays an accept(2) racing with sonewconn(), which inserts the socket
 584  * before the inpcb address/port fields are initialized.  A better fix would
 585  * prevent the socket from being placed in the listen queue until all fields
 586  * are fully initialized.
 587  */
 588 static int
 589 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
 590 {
 591         int error = 0;
 592         struct inpcb *inp = NULL;
 593         struct tcpcb *tp = NULL;
 594         struct in_addr addr;
 595         in_port_t port = 0;
 596         TCPDEBUG0;
 597
 598         if (so->so_state & SS_ISDISCONNECTED)
 599                 return (ECONNABORTED);
 600
 601         inp = sotoinpcb(so);
 602         KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
 603         INP_INFO_RLOCK(&V_tcbinfo);
 604         INP_WLOCK(inp);
 605         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 606                 error = ECONNABORTED;
 607                 goto out;
 608         }
 609         tp = intotcpcb(inp);
 610         TCPDEBUG1();
 611
 612         /*
 613          * We inline in_getpeeraddr and COMMON_END here, so that we can
 614          * copy the data of interest and defer the malloc until after we
 615          * release the lock.
 616          */
 617         port = inp->inp_fport;
 618         addr = inp->inp_faddr;
 619
 620 out:
 621         TCPDEBUG2(PRU_ACCEPT);
 622         INP_WUNLOCK(inp);
 623         INP_INFO_RUNLOCK(&V_tcbinfo);
 624         if (error == 0)
 625                 *nam = in_sockaddr(port, &addr);
 626         return error;
 627 }
 628 #endif /* INET */
 629
 630 #ifdef INET6
 631 static int
 632 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
 633 {
 634         struct inpcb *inp = NULL;
 635         int error = 0;
 636         struct tcpcb *tp = NULL;
 637         struct in_addr addr;
 638         struct in6_addr addr6;
 639         in_port_t port = 0;
 640         int v4 = 0;
 641         TCPDEBUG0;
 642
 643         if (so->so_state & SS_ISDISCONNECTED)
 644                 return (ECONNABORTED);
 645
 646         inp = sotoinpcb(so);
 647         KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
 648         INP_INFO_RLOCK(&V_tcbinfo);
 649         INP_WLOCK(inp);
 650         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 651                 error = ECONNABORTED;
 652                 goto out;
 653         }
 654         tp = intotcpcb(inp);
 655         TCPDEBUG1();
 656
 657         /*
 658          * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
 659          * copy the data of interest and defer the malloc until after we
 660          * release the lock.
 661          */
 662         if (inp->inp_vflag & INP_IPV4) {
 663                 v4 = 1;
 664                 port = inp->inp_fport;
 665                 addr = inp->inp_faddr;
 666         } else {
 667                 port = inp->inp_fport;
 668                 addr6 = inp->in6p_faddr;
 669         }
 670
 671 out:
 672         TCPDEBUG2(PRU_ACCEPT);
 673         INP_WUNLOCK(inp);
 674         INP_INFO_RUNLOCK(&V_tcbinfo);
 675         if (error == 0) {
 676                 if (v4)
 677                         *nam = in6_v4mapsin6_sockaddr(port, &addr);
 678                 else
 679                         *nam = in6_sockaddr(port, &addr6);
 680         }
 681         return error;
 682 }
 683 #endif /* INET6 */
 684
 685 /*
 686  * Mark the connection as being incapable of further output.
 687  */
 688 static int
 689 tcp_usr_shutdown(struct socket *so)
 690 {
 691         int error = 0;
 692         struct inpcb *inp;
 693         struct tcpcb *tp = NULL;
 694
 695         TCPDEBUG0;
 696         INP_INFO_WLOCK(&V_tcbinfo);
 697         inp = sotoinpcb(so);
 698         KASSERT(inp != NULL, ("inp == NULL"));
 699         INP_WLOCK(inp);
 700         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 701                 error = ECONNRESET;
 702                 goto out;
 703         }
 704         tp = intotcpcb(inp);
 705         TCPDEBUG1();
 706         socantsendmore(so);
 707         tcp_usrclosed(tp);
 708         if (!(inp->inp_flags & INP_DROPPED))
 709                 error = tcp_output_disconnect(tp);
 710
 711 out:
 712         TCPDEBUG2(PRU_SHUTDOWN);
 713         INP_WUNLOCK(inp);
 714         INP_INFO_WUNLOCK(&V_tcbinfo);
 715
 716         return (error);
 717 }
 718
 719 /*
 720  * After a receive, possibly send window update to peer.
 721  */
 722 static int
 723 tcp_usr_rcvd(struct socket *so, int flags)
 724 {
 725         struct inpcb *inp;
 726         struct tcpcb *tp = NULL;
 727         int error = 0;
 728
 729         TCPDEBUG0;
 730         inp = sotoinpcb(so);
 731         KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
 732         INP_WLOCK(inp);
 733         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 734                 error = ECONNRESET;
 735                 goto out;
 736         }
 737         tp = intotcpcb(inp);
 738         TCPDEBUG1();
 739         tcp_output_rcvd(tp);
 740
 741 out:
 742         TCPDEBUG2(PRU_RCVD);
 743         INP_WUNLOCK(inp);
 744         return (error);
 745 }
 746
 747 /*
 748  * Do a send by putting data in output queue and updating urgent
 749  * marker if URG set.  Possibly send more data.  Unlike the other
 750  * pru_*() routines, the mbuf chains are our responsibility.  We
 751  * must either enqueue them or free them.  The other pru_* routines
 752  * generally are caller-frees.
 753  */
 754 static int
 755 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
 756     struct sockaddr *nam, struct mbuf *control, struct thread *td)
 757 {
 758         int error = 0;
 759         struct inpcb *inp;
 760         struct tcpcb *tp = NULL;
 761 #ifdef INET6
 762         int isipv6;
 763 #endif
 764         TCPDEBUG0;
 765
 766         /*
 767          * We require the pcbinfo lock if we will close the socket as part of
 768          * this call.
 769          */
 770         if (flags & PRUS_EOF)
 771                 INP_INFO_WLOCK(&V_tcbinfo);
 772         inp = sotoinpcb(so);
 773         KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
 774         INP_WLOCK(inp);
 775         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 776                 if (control)
 777                         m_freem(control);
 778                 if (m)
 779                         m_freem(m);
 780                 error = ECONNRESET;
 781                 goto out;
 782         }
 783 #ifdef INET6
 784         isipv6 = nam && nam->sa_family == AF_INET6;
 785 #endif /* INET6 */
 786         tp = intotcpcb(inp);
 787         TCPDEBUG1();
 788         if (control) {
 789                 /* TCP doesn't do control messages (rights, creds, etc) */
 790                 if (control->m_len) {
 791                         m_freem(control);
 792                         if (m)
 793                                 m_freem(m);
 794                         error = EINVAL;
 795                         goto out;
 796                 }
 797                 m_freem(control);       /* empty control, just free it */
 798         }
 799         if (!(flags & PRUS_OOB)) {
 800                 sbappendstream(&so->so_snd, m);
 801                 if (nam && tp->t_state < TCPS_SYN_SENT) {
 802                         /*
 803                          * Do implied connect if not yet connected,
 804                          * initialize window to default value, and
 805                          * initialize maxseg/maxopd using peer's cached
 806                          * MSS.
 807                          */
 808 #ifdef INET6
 809                         if (isipv6)
 810                                 error = tcp6_connect(tp, nam, td);
 811 #endif /* INET6 */
 812 #if defined(INET6) && defined(INET)
 813                         else
 814 #endif
 815 #ifdef INET
 816                                 error = tcp_connect(tp, nam, td);
 817 #endif
 818                         if (error)
 819                                 goto out;
 820                         tp->snd_wnd = TTCP_CLIENT_SND_WND;
 821                         tcp_mss(tp, -1);
 822                 }
 823                 if (flags & PRUS_EOF) {
 824                         /*
 825                          * Close the send side of the connection after
 826                          * the data is sent.
 827                          */
 828                         INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
 829                         socantsendmore(so);
 830                         tcp_usrclosed(tp);
 831                 }
 832                 if (!(inp->inp_flags & INP_DROPPED)) {
 833                         if (flags & PRUS_MORETOCOME)
 834                                 tp->t_flags |= TF_MORETOCOME;
 835                         error = tcp_output_send(tp);
 836                         if (flags & PRUS_MORETOCOME)
 837                                 tp->t_flags &= ~TF_MORETOCOME;
 838                 }
 839         } else {
 840                 /*
 841                  * XXXRW: PRUS_EOF not implemented with PRUS_OOB?
 842                  */
 843                 SOCKBUF_LOCK(&so->so_snd);
 844                 if (sbspace(&so->so_snd) < -512) {
 845                         SOCKBUF_UNLOCK(&so->so_snd);
 846                         m_freem(m);
 847                         error = ENOBUFS;
 848                         goto out;
 849                 }
 850                 /*
 851                  * According to RFC961 (Assigned Protocols),
 852                  * the urgent pointer points to the last octet
 853                  * of urgent data.  We continue, however,
 854                  * to consider it to indicate the first octet
 855                  * of data past the urgent section.
 856                  * Otherwise, snd_up should be one lower.
 857                  */
 858                 sbappendstream_locked(&so->so_snd, m);
 859                 SOCKBUF_UNLOCK(&so->so_snd);
 860                 if (nam && tp->t_state < TCPS_SYN_SENT) {
 861                         /*
 862                          * Do implied connect if not yet connected,
 863                          * initialize window to default value, and
 864                          * initialize maxseg/maxopd using peer's cached
 865                          * MSS.
 866                          */
 867 #ifdef INET6
 868                         if (isipv6)
 869                                 error = tcp6_connect(tp, nam, td);
 870 #endif /* INET6 */
 871 #if defined(INET6) && defined(INET)
 872                         else
 873 #endif
 874 #ifdef INET
 875                                 error = tcp_connect(tp, nam, td);
 876 #endif
 877                         if (error)
 878                                 goto out;
 879                         tp->snd_wnd = TTCP_CLIENT_SND_WND;
 880                         tcp_mss(tp, -1);
 881                 }
 882                 tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
 883                 tp->t_flags |= TF_FORCEDATA;
 884                 error = tcp_output_send(tp);
 885                 tp->t_flags &= ~TF_FORCEDATA;
 886         }
 887 out:
 888         TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
 889                   ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
 890         INP_WUNLOCK(inp);
 891         if (flags & PRUS_EOF)
 892                 INP_INFO_WUNLOCK(&V_tcbinfo);
 893         return (error);
 894 }
 895
 896 /*
 897  * Abort the TCP.  Drop the connection abruptly.
 898  */
 899 static void
 900 tcp_usr_abort(struct socket *so)
 901 {
 902         struct inpcb *inp;
 903         struct tcpcb *tp = NULL;
 904         TCPDEBUG0;
 905
 906         inp = sotoinpcb(so);
 907         KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
 908
 909         INP_INFO_WLOCK(&V_tcbinfo);
 910         INP_WLOCK(inp);
 911         KASSERT(inp->inp_socket != NULL,
 912             ("tcp_usr_abort: inp_socket == NULL"));
 913
 914         /*
 915          * If we still have full TCP state, and we're not dropped, drop.
 916          */
 917         if (!(inp->inp_flags & INP_TIMEWAIT) &&
 918             !(inp->inp_flags & INP_DROPPED)) {
 919                 tp = intotcpcb(inp);
 920                 TCPDEBUG1();
 921                 tcp_drop(tp, ECONNABORTED);
 922                 TCPDEBUG2(PRU_ABORT);
 923         }
 924         if (!(inp->inp_flags & INP_DROPPED)) {
 925                 SOCK_LOCK(so);
 926                 so->so_state |= SS_PROTOREF;
 927                 SOCK_UNLOCK(so);
 928                 inp->inp_flags |= INP_SOCKREF;
 929         }
 930         INP_WUNLOCK(inp);
 931         INP_INFO_WUNLOCK(&V_tcbinfo);
 932 }
 933
 934 /*
 935  * TCP socket is closed.  Start friendly disconnect.
 936  */
 937 static void
 938 tcp_usr_close(struct socket *so)
 939 {
 940         struct inpcb *inp;
 941         struct tcpcb *tp = NULL;
 942         TCPDEBUG0;
 943
 944         inp = sotoinpcb(so);
 945         KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
 946
 947         INP_INFO_WLOCK(&V_tcbinfo);
 948         INP_WLOCK(inp);
 949         KASSERT(inp->inp_socket != NULL,
 950             ("tcp_usr_close: inp_socket == NULL"));
 951
 952         /*
 953          * If we still have full TCP state, and we're not dropped, initiate
 954          * a disconnect.
 955          */
 956         if (!(inp->inp_flags & INP_TIMEWAIT) &&
 957             !(inp->inp_flags & INP_DROPPED)) {
 958                 tp = intotcpcb(inp);
 959                 TCPDEBUG1();
 960                 tcp_disconnect(tp);
 961                 TCPDEBUG2(PRU_CLOSE);
 962         }
 963         if (!(inp->inp_flags & INP_DROPPED)) {
 964                 SOCK_LOCK(so);
 965                 so->so_state |= SS_PROTOREF;
 966                 SOCK_UNLOCK(so);
 967                 inp->inp_flags |= INP_SOCKREF;
 968         }
 969         INP_WUNLOCK(inp);
 970         INP_INFO_WUNLOCK(&V_tcbinfo);
 971 }
 972
 973 /*
 974  * Receive out-of-band data.
 975  */
 976 static int
 977 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
 978 {
 979         int error = 0;
 980         struct inpcb *inp;
 981         struct tcpcb *tp = NULL;
 982
 983         TCPDEBUG0;
 984         inp = sotoinpcb(so);
 985         KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
 986         INP_WLOCK(inp);
 987         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
 988                 error = ECONNRESET;
 989                 goto out;
 990         }
 991         tp = intotcpcb(inp);
 992         TCPDEBUG1();
 993         if ((so->so_oobmark == 0 &&
 994              (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
 995             so->so_options & SO_OOBINLINE ||
 996             tp->t_oobflags & TCPOOB_HADDATA) {
 997                 error = EINVAL;
 998                 goto out;
 999         }
1000         if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1001                 error = EWOULDBLOCK;
1002                 goto out;
1003         }
1004         m->m_len = 1;
1005         *mtod(m, caddr_t) = tp->t_iobc;
1006         if ((flags & MSG_PEEK) == 0)
1007                 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1008
1009 out:
1010         TCPDEBUG2(PRU_RCVOOB);
1011         INP_WUNLOCK(inp);
1012         return (error);
1013 }
1014
1015 #ifdef INET
1016 struct pr_usrreqs tcp_usrreqs = {
1017         .pru_abort =            tcp_usr_abort,
1018         .pru_accept =           tcp_usr_accept,
1019         .pru_attach =           tcp_usr_attach,
1020         .pru_bind =             tcp_usr_bind,
1021         .pru_connect =          tcp_usr_connect,
1022         .pru_control =          in_control,
1023         .pru_detach =           tcp_usr_detach,
1024         .pru_disconnect =       tcp_usr_disconnect,
1025         .pru_listen =           tcp_usr_listen,
1026         .pru_peeraddr =         in_getpeeraddr,
1027         .pru_rcvd =             tcp_usr_rcvd,
1028         .pru_rcvoob =           tcp_usr_rcvoob,
1029         .pru_send =             tcp_usr_send,
1030         .pru_shutdown =         tcp_usr_shutdown,
1031         .pru_sockaddr =         in_getsockaddr,
1032         .pru_sosetlabel =       in_pcbsosetlabel,
1033         .pru_close =            tcp_usr_close,
1034 };
1035 #endif /* INET */
1036
1037 #ifdef INET6
1038 struct pr_usrreqs tcp6_usrreqs = {
1039         .pru_abort =            tcp_usr_abort,
1040         .pru_accept =           tcp6_usr_accept,
1041         .pru_attach =           tcp_usr_attach,
1042         .pru_bind =             tcp6_usr_bind,
1043         .pru_connect =          tcp6_usr_connect,
1044         .pru_control =          in6_control,
1045         .pru_detach =           tcp_usr_detach,
1046         .pru_disconnect =       tcp_usr_disconnect,
1047         .pru_listen =           tcp6_usr_listen,
1048         .pru_peeraddr =         in6_mapped_peeraddr,
1049         .pru_rcvd =             tcp_usr_rcvd,
1050         .pru_rcvoob =           tcp_usr_rcvoob,
1051         .pru_send =             tcp_usr_send,
1052         .pru_shutdown =         tcp_usr_shutdown,
1053         .pru_sockaddr =         in6_mapped_sockaddr,
1054         .pru_sosetlabel =       in_pcbsosetlabel,
1055         .pru_close =            tcp_usr_close,
1056 };
1057 #endif /* INET6 */
1058
1059 #ifdef INET
1060 /*
1061  * Common subroutine to open a TCP connection to remote host specified
1062  * by struct sockaddr_in in mbuf *nam.  Call in_pcbbind to assign a local
1063  * port number if needed.  Call in_pcbconnect_setup to do the routing and
1064  * to choose a local host address (interface).  If there is an existing
1065  * incarnation of the same connection in TIME-WAIT state and if the remote
1066  * host was sending CC options and if the connection duration was < MSL, then
1067  * truncate the previous TIME-WAIT state and proceed.
1068  * Initialize connection parameters and enter SYN-SENT state.
1069  */
1070 static int
1071 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1072 {
1073         struct inpcb *inp = tp->t_inpcb, *oinp;
1074         struct socket *so = inp->inp_socket;
1075         struct in_addr laddr;
1076         u_short lport;
1077         int error;
1078
1079         INP_WLOCK_ASSERT(inp);
1080         INP_HASH_WLOCK(&V_tcbinfo);
1081
1082         if (inp->inp_lport == 0) {
1083                 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1084                 if (error)
1085                         goto out;
1086         }
1087
1088         /*
1089          * Cannot simply call in_pcbconnect, because there might be an
1090          * earlier incarnation of this same connection still in
1091          * TIME_WAIT state, creating an ADDRINUSE error.
1092          */
1093         laddr = inp->inp_laddr;
1094         lport = inp->inp_lport;
1095         error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
1096             &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
1097         if (error && oinp == NULL)
1098                 goto out;
1099         if (oinp) {
1100                 error = EADDRINUSE;
1101                 goto out;
1102         }
1103         inp->inp_laddr = laddr;
1104         in_pcbrehash(inp);
1105         INP_HASH_WUNLOCK(&V_tcbinfo);
1106
1107         /*
1108          * Compute window scaling to request:
1109          * Scale to fit into sweet spot.  See tcp_syncache.c.
1110          * XXX: This should move to tcp_output().
1111          */
1112         while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1113             (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1114                 tp->request_r_scale++;
1115
1116         soisconnecting(so);
1117         TCPSTAT_INC(tcps_connattempt);
1118         tp->t_state = TCPS_SYN_SENT;
1119         tcp_timer_activate(tp, TT_KEEP, tcp_keepinit);
1120         tp->iss = tcp_new_isn(tp);
1121         tcp_sendseqinit(tp);
1122
1123         return 0;
1124
1125 out:
1126         INP_HASH_WUNLOCK(&V_tcbinfo);
1127         return (error);
1128 }
1129 #endif /* INET */
1130
1131 #ifdef INET6
1132 static int
1133 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1134 {
1135         struct inpcb *inp = tp->t_inpcb, *oinp;
1136         struct socket *so = inp->inp_socket;
1137         struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
1138         struct in6_addr addr6;
1139         int error;
1140
1141         INP_WLOCK_ASSERT(inp);
1142         INP_HASH_WLOCK(&V_tcbinfo);
1143
1144         if (inp->inp_lport == 0) {
1145                 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1146                 if (error)
1147                         goto out;
1148         }
1149
1150         /*
1151          * Cannot simply call in_pcbconnect, because there might be an
1152          * earlier incarnation of this same connection still in
1153          * TIME_WAIT state, creating an ADDRINUSE error.
1154          * in6_pcbladdr() also handles scope zone IDs.
1155          *
1156          * XXXRW: We wouldn't need to expose in6_pcblookup_hash_locked()
1157          * outside of in6_pcb.c if there were an in6_pcbconnect_setup().
1158          */
1159         error = in6_pcbladdr(inp, nam, &addr6);
1160         if (error)
1161                 goto out;
1162         oinp = in6_pcblookup_hash_locked(inp->inp_pcbinfo,
1163                                   &sin6->sin6_addr, sin6->sin6_port,
1164                                   IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
1165                                   ? &addr6
1166                                   : &inp->in6p_laddr,
1167                                   inp->inp_lport,  0, NULL);
1168         if (oinp) {
1169                 error = EADDRINUSE;
1170                 goto out;
1171         }
1172         if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
1173                 inp->in6p_laddr = addr6;
1174         inp->in6p_faddr = sin6->sin6_addr;
1175         inp->inp_fport = sin6->sin6_port;
1176         /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
1177         inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
1178         if (inp->inp_flags & IN6P_AUTOFLOWLABEL)
1179                 inp->inp_flow |=
1180                     (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
1181         in_pcbrehash(inp);
1182         INP_HASH_WUNLOCK(&V_tcbinfo);
1183
1184         /* Compute window scaling to request.  */
1185         while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1186             (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1187                 tp->request_r_scale++;
1188
1189         soisconnecting(so);
1190         TCPSTAT_INC(tcps_connattempt);
1191         tp->t_state = TCPS_SYN_SENT;
1192         tcp_timer_activate(tp, TT_KEEP, tcp_keepinit);
1193         tp->iss = tcp_new_isn(tp);
1194         tcp_sendseqinit(tp);
1195
1196         return 0;
1197
1198 out:
1199         INP_HASH_WUNLOCK(&V_tcbinfo);
1200         return error;
1201 }
1202 #endif /* INET6 */
1203
1204 /*
1205  * Export TCP internal state information via a struct tcp_info, based on the
1206  * Linux 2.6 API.  Not ABI compatible as our constants are mapped differently
1207  * (TCP state machine, etc).  We export all information using FreeBSD-native
1208  * constants -- for example, the numeric values for tcpi_state will differ
1209  * from Linux.
1210  */
1211 static void
1212 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1213 {
1214
1215         INP_WLOCK_ASSERT(tp->t_inpcb);
1216         bzero(ti, sizeof(*ti));
1217
1218         ti->tcpi_state = tp->t_state;
1219         if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
1220                 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1221         if (tp->t_flags & TF_SACK_PERMIT)
1222                 ti->tcpi_options |= TCPI_OPT_SACK;
1223         if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
1224                 ti->tcpi_options |= TCPI_OPT_WSCALE;
1225                 ti->tcpi_snd_wscale = tp->snd_scale;
1226                 ti->tcpi_rcv_wscale = tp->rcv_scale;
1227         }
1228
1229         ti->tcpi_rto = tp->t_rxtcur * tick;
1230         ti->tcpi_last_data_recv = (long)(ticks - (int)tp->t_rcvtime) * tick;
1231         ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
1232         ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
1233
1234         ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1235         ti->tcpi_snd_cwnd = tp->snd_cwnd;
1236
1237         /*
1238          * FreeBSD-specific extension fields for tcp_info.
1239          */
1240         ti->tcpi_rcv_space = tp->rcv_wnd;
1241         ti->tcpi_rcv_nxt = tp->rcv_nxt;
1242         ti->tcpi_snd_wnd = tp->snd_wnd;
1243         ti->tcpi_snd_bwnd = 0;          /* Unused, kept for compat. */
1244         ti->tcpi_snd_nxt = tp->snd_nxt;
1245         ti->tcpi_snd_mss = tp->t_maxseg;
1246         ti->tcpi_rcv_mss = tp->t_maxseg;
1247         if (tp->t_flags & TF_TOE)
1248                 ti->tcpi_options |= TCPI_OPT_TOE;
1249         ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
1250         ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
1251         ti->tcpi_snd_zerowin = tp->t_sndzerowin;
1252 }
1253
1254 /*
1255  * tcp_ctloutput() must drop the inpcb lock before performing copyin on
1256  * socket option arguments.  When it re-acquires the lock after the copy, it
1257  * has to revalidate that the connection is still valid for the socket
1258  * option.
1259  */
1260 #define INP_WLOCK_RECHECK(inp) do {                                     \
1261         INP_WLOCK(inp);                                                 \
1262         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {            \
1263                 INP_WUNLOCK(inp);                                       \
1264                 return (ECONNRESET);                                    \
1265         }                                                               \
1266         tp = intotcpcb(inp);                                            \
1267 } while(0)
1268
1269 int
1270 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1271 {
1272         int     error, opt, optval;
1273         struct  inpcb *inp;
1274         struct  tcpcb *tp;
1275         struct  tcp_info ti;
1276         char buf[TCP_CA_NAME_MAX];
1277         struct cc_algo *algo;
1278
1279         error = 0;
1280         inp = sotoinpcb(so);
1281         KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
1282         INP_WLOCK(inp);
1283         if (sopt->sopt_level != IPPROTO_TCP) {
1284 #ifdef INET6
1285                 if (inp->inp_vflag & INP_IPV6PROTO) {
1286                         INP_WUNLOCK(inp);
1287                         error = ip6_ctloutput(so, sopt);
1288                 }
1289 #endif /* INET6 */
1290 #if defined(INET6) && defined(INET)
1291                 else
1292 #endif
1293 #ifdef INET
1294                 {
1295                         INP_WUNLOCK(inp);
1296                         error = ip_ctloutput(so, sopt);
1297                 }
1298 #endif
1299                 return (error);
1300         }
1301         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1302                 INP_WUNLOCK(inp);
1303                 return (ECONNRESET);
1304         }
1305
1306         switch (sopt->sopt_dir) {
1307         case SOPT_SET:
1308                 switch (sopt->sopt_name) {
1309 #ifdef TCP_SIGNATURE
1310                 case TCP_MD5SIG:
1311                         INP_WUNLOCK(inp);
1312                         error = sooptcopyin(sopt, &optval, sizeof optval,
1313                             sizeof optval);
1314                         if (error)
1315                                 return (error);
1316
1317                         INP_WLOCK_RECHECK(inp);
1318                         if (optval > 0)
1319                                 tp->t_flags |= TF_SIGNATURE;
1320                         else
1321                                 tp->t_flags &= ~TF_SIGNATURE;
1322                         INP_WUNLOCK(inp);
1323                         break;
1324 #endif /* TCP_SIGNATURE */
1325                 case TCP_NODELAY:
1326                 case TCP_NOOPT:
1327                         INP_WUNLOCK(inp);
1328                         error = sooptcopyin(sopt, &optval, sizeof optval,
1329                             sizeof optval);
1330                         if (error)
1331                                 return (error);
1332
1333                         INP_WLOCK_RECHECK(inp);
1334                         switch (sopt->sopt_name) {
1335                         case TCP_NODELAY:
1336                                 opt = TF_NODELAY;
1337                                 break;
1338                         case TCP_NOOPT:
1339                                 opt = TF_NOOPT;
1340                                 break;
1341                         default:
1342                                 opt = 0; /* dead code to fool gcc */
1343                                 break;
1344                         }
1345
1346                         if (optval)
1347                                 tp->t_flags |= opt;
1348                         else
1349                                 tp->t_flags &= ~opt;
1350                         INP_WUNLOCK(inp);
1351                         break;
1352
1353                 case TCP_NOPUSH:
1354                         INP_WUNLOCK(inp);
1355                         error = sooptcopyin(sopt, &optval, sizeof optval,
1356                             sizeof optval);
1357                         if (error)
1358                                 return (error);
1359
1360                         INP_WLOCK_RECHECK(inp);
1361                         if (optval)
1362                                 tp->t_flags |= TF_NOPUSH;
1363                         else if (tp->t_flags & TF_NOPUSH) {
1364                                 tp->t_flags &= ~TF_NOPUSH;
1365                                 if (TCPS_HAVEESTABLISHED(tp->t_state))
1366                                         error = tcp_output(tp);
1367                         }
1368                         INP_WUNLOCK(inp);
1369                         break;
1370
1371                 case TCP_MAXSEG:
1372                         INP_WUNLOCK(inp);
1373                         error = sooptcopyin(sopt, &optval, sizeof optval,
1374                             sizeof optval);
1375                         if (error)
1376                                 return (error);
1377
1378                         INP_WLOCK_RECHECK(inp);
1379                         if (optval > 0 && optval <= tp->t_maxseg &&
1380                             optval + 40 >= V_tcp_minmss)
1381                                 tp->t_maxseg = optval;
1382                         else
1383                                 error = EINVAL;
1384                         INP_WUNLOCK(inp);
1385                         break;
1386
1387                 case TCP_INFO:
1388                         INP_WUNLOCK(inp);
1389                         error = EINVAL;
1390                         break;
1391
1392                 case TCP_CONGESTION:
1393                         INP_WUNLOCK(inp);
1394                         bzero(buf, sizeof(buf));
1395                         error = sooptcopyin(sopt, &buf, sizeof(buf), 1);
1396                         if (error)
1397                                 break;
1398                         INP_WLOCK_RECHECK(inp);
1399                         /*
1400                          * Return EINVAL if we can't find the requested cc algo.
1401                          */
1402                         error = EINVAL;
1403                         CC_LIST_RLOCK();
1404                         STAILQ_FOREACH(algo, &cc_list, entries) {
1405                                 if (strncmp(buf, algo->name, TCP_CA_NAME_MAX)
1406                                     == 0) {
1407                                         /* We've found the requested algo. */
1408                                         error = 0;
1409                                         /*
1410                                          * We hold a write lock over the tcb
1411                                          * so it's safe to do these things
1412                                          * without ordering concerns.
1413                                          */
1414                                         if (CC_ALGO(tp)->cb_destroy != NULL)
1415                                                 CC_ALGO(tp)->cb_destroy(tp->ccv);
1416                                         CC_ALGO(tp) = algo;
1417                                         /*
1418                                          * If something goes pear shaped
1419                                          * initialising the new algo,
1420                                          * fall back to newreno (which
1421                                          * does not require initialisation).
1422                                          */
1423                                         if (algo->cb_init != NULL)
1424                                                 if (algo->cb_init(tp->ccv) > 0) {
1425                                                         CC_ALGO(tp) = &newreno_cc_algo;
1426                                                         /*
1427                                                          * The only reason init
1428                                                          * should fail is
1429                                                          * because of malloc.
1430                                                          */
1431                                                         error = ENOMEM;
1432                                                 }
1433                                         break; /* Break the STAILQ_FOREACH. */
1434                                 }
1435                         }
1436                         CC_LIST_RUNLOCK();
1437                         INP_WUNLOCK(inp);
1438                         break;
1439
1440                 default:
1441                         INP_WUNLOCK(inp);
1442                         error = ENOPROTOOPT;
1443                         break;
1444                 }
1445                 break;
1446
1447         case SOPT_GET:
1448                 tp = intotcpcb(inp);
1449                 switch (sopt->sopt_name) {
1450 #ifdef TCP_SIGNATURE
1451                 case TCP_MD5SIG:
1452                         optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0;
1453                         INP_WUNLOCK(inp);
1454                         error = sooptcopyout(sopt, &optval, sizeof optval);
1455                         break;
1456 #endif
1457
1458                 case TCP_NODELAY:
1459                         optval = tp->t_flags & TF_NODELAY;
1460                         INP_WUNLOCK(inp);
1461                         error = sooptcopyout(sopt, &optval, sizeof optval);
1462                         break;
1463                 case TCP_MAXSEG:
1464                         optval = tp->t_maxseg;
1465                         INP_WUNLOCK(inp);
1466                         error = sooptcopyout(sopt, &optval, sizeof optval);
1467                         break;
1468                 case TCP_NOOPT:
1469                         optval = tp->t_flags & TF_NOOPT;
1470                         INP_WUNLOCK(inp);
1471                         error = sooptcopyout(sopt, &optval, sizeof optval);
1472                         break;
1473                 case TCP_NOPUSH:
1474                         optval = tp->t_flags & TF_NOPUSH;
1475                         INP_WUNLOCK(inp);
1476                         error = sooptcopyout(sopt, &optval, sizeof optval);
1477                         break;
1478                 case TCP_INFO:
1479                         tcp_fill_info(tp, &ti);
1480                         INP_WUNLOCK(inp);
1481                         error = sooptcopyout(sopt, &ti, sizeof ti);
1482                         break;
1483                 case TCP_CONGESTION:
1484                         bzero(buf, sizeof(buf));
1485                         strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX);
1486                         INP_WUNLOCK(inp);
1487                         error = sooptcopyout(sopt, buf, TCP_CA_NAME_MAX);
1488                         break;
1489                 default:
1490                         INP_WUNLOCK(inp);
1491                         error = ENOPROTOOPT;
1492                         break;
1493                 }
1494                 break;
1495         }
1496         return (error);
1497 }
1498 #undef INP_WLOCK_RECHECK
1499
1500 /*
1501  * tcp_sendspace and tcp_recvspace are the default send and receive window
1502  * sizes, respectively.  These are obsolescent (this information should
1503  * be set by the route).
1504  */
1505 u_long  tcp_sendspace = 1024*32;
1506 SYSCTL_ULONG(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW,
1507     &tcp_sendspace , 0, "Maximum outgoing TCP datagram size");
1508 u_long  tcp_recvspace = 1024*64;
1509 SYSCTL_ULONG(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
1510     &tcp_recvspace , 0, "Maximum incoming TCP datagram size");
1511
1512 /*
1513  * Attach TCP protocol to socket, allocating
1514  * internet protocol control block, tcp control block,
1515  * bufer space, and entering LISTEN state if to accept connections.
1516  */
1517 static int
1518 tcp_attach(struct socket *so)
1519 {
1520         struct tcpcb *tp;
1521         struct inpcb *inp;
1522         int error;
1523
1524         if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1525                 error = soreserve(so, tcp_sendspace, tcp_recvspace);
1526                 if (error)
1527                         return (error);
1528         }
1529         so->so_rcv.sb_flags |= SB_AUTOSIZE;
1530         so->so_snd.sb_flags |= SB_AUTOSIZE;
1531         INP_INFO_WLOCK(&V_tcbinfo);
1532         error = in_pcballoc(so, &V_tcbinfo);
1533         if (error) {
1534                 INP_INFO_WUNLOCK(&V_tcbinfo);
1535                 return (error);
1536         }
1537         inp = sotoinpcb(so);
1538 #ifdef INET6
1539         if (inp->inp_vflag & INP_IPV6PROTO) {
1540                 inp->inp_vflag |= INP_IPV6;
1541                 inp->in6p_hops = -1;    /* use kernel default */
1542         }
1543         else
1544 #endif
1545         inp->inp_vflag |= INP_IPV4;
1546         tp = tcp_newtcpcb(inp);
1547         if (tp == NULL) {
1548                 in_pcbdetach(inp);
1549                 in_pcbfree(inp);
1550                 INP_INFO_WUNLOCK(&V_tcbinfo);
1551                 return (ENOBUFS);
1552         }
1553         tp->t_state = TCPS_CLOSED;
1554         INP_WUNLOCK(inp);
1555         INP_INFO_WUNLOCK(&V_tcbinfo);
1556         return (0);
1557 }
1558
1559 /*
1560  * Initiate (or continue) disconnect.
1561  * If embryonic state, just send reset (once).
1562  * If in ``let data drain'' option and linger null, just drop.
1563  * Otherwise (hard), mark socket disconnecting and drop
1564  * current input data; switch states based on user close, and
1565  * send segment to peer (with FIN).
1566  */
1567 static void
1568 tcp_disconnect(struct tcpcb *tp)
1569 {
1570         struct inpcb *inp = tp->t_inpcb;
1571         struct socket *so = inp->inp_socket;
1572
1573         INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1574         INP_WLOCK_ASSERT(inp);
1575
1576         /*
1577          * Neither tcp_close() nor tcp_drop() should return NULL, as the
1578          * socket is still open.
1579          */
1580         if (tp->t_state < TCPS_ESTABLISHED) {
1581                 tp = tcp_close(tp);
1582                 KASSERT(tp != NULL,
1583                     ("tcp_disconnect: tcp_close() returned NULL"));
1584         } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
1585                 tp = tcp_drop(tp, 0);
1586                 KASSERT(tp != NULL,
1587                     ("tcp_disconnect: tcp_drop() returned NULL"));
1588         } else {
1589                 soisdisconnecting(so);
1590                 sbflush(&so->so_rcv);
1591                 tcp_usrclosed(tp);
1592                 if (!(inp->inp_flags & INP_DROPPED))
1593                         tcp_output_disconnect(tp);
1594         }
1595 }
1596
1597 /*
1598  * User issued close, and wish to trail through shutdown states:
1599  * if never received SYN, just forget it.  If got a SYN from peer,
1600  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1601  * If already got a FIN from peer, then almost done; go to LAST_ACK
1602  * state.  In all other cases, have already sent FIN to peer (e.g.
1603  * after PRU_SHUTDOWN), and just have to play tedious game waiting
1604  * for peer to send FIN or not respond to keep-alives, etc.
1605  * We can let the user exit from the close as soon as the FIN is acked.
1606  */
1607 static void
1608 tcp_usrclosed(struct tcpcb *tp)
1609 {
1610
1611         INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1612         INP_WLOCK_ASSERT(tp->t_inpcb);
1613
1614         switch (tp->t_state) {
1615         case TCPS_LISTEN:
1616                 tcp_offload_listen_close(tp);
1617                 /* FALLTHROUGH */
1618         case TCPS_CLOSED:
1619                 tp->t_state = TCPS_CLOSED;
1620                 tp = tcp_close(tp);
1621                 /*
1622                  * tcp_close() should never return NULL here as the socket is
1623                  * still open.
1624                  */
1625                 KASSERT(tp != NULL,
1626                     ("tcp_usrclosed: tcp_close() returned NULL"));
1627                 break;
1628
1629         case TCPS_SYN_SENT:
1630         case TCPS_SYN_RECEIVED:
1631                 tp->t_flags |= TF_NEEDFIN;
1632                 break;
1633
1634         case TCPS_ESTABLISHED:
1635                 tp->t_state = TCPS_FIN_WAIT_1;
1636                 break;
1637
1638         case TCPS_CLOSE_WAIT:
1639                 tp->t_state = TCPS_LAST_ACK;
1640                 break;
1641         }
1642         if (tp->t_state >= TCPS_FIN_WAIT_2) {
1643                 soisdisconnected(tp->t_inpcb->inp_socket);
1644                 /* Prevent the connection hanging in FIN_WAIT_2 forever. */
1645                 if (tp->t_state == TCPS_FIN_WAIT_2) {
1646                         int timeout;
1647
1648                         timeout = (tcp_fast_finwait2_recycle) ?
1649                             tcp_finwait2_timeout : tcp_maxidle;
1650                         tcp_timer_activate(tp, TT_2MSL, timeout);
1651                 }
1652         }
1653 }
1654
1655 #ifdef DDB
1656 static void
1657 db_print_indent(int indent)
1658 {
1659         int i;
1660
1661         for (i = 0; i < indent; i++)
1662                 db_printf(" ");
1663 }
1664
1665 static void
1666 db_print_tstate(int t_state)
1667 {
1668
1669         switch (t_state) {
1670         case TCPS_CLOSED:
1671                 db_printf("TCPS_CLOSED");
1672                 return;
1673
1674         case TCPS_LISTEN:
1675                 db_printf("TCPS_LISTEN");
1676                 return;
1677
1678         case TCPS_SYN_SENT:
1679                 db_printf("TCPS_SYN_SENT");
1680                 return;
1681
1682         case TCPS_SYN_RECEIVED:
1683                 db_printf("TCPS_SYN_RECEIVED");
1684                 return;
1685
1686         case TCPS_ESTABLISHED:
1687                 db_printf("TCPS_ESTABLISHED");
1688                 return;
1689
1690         case TCPS_CLOSE_WAIT:
1691                 db_printf("TCPS_CLOSE_WAIT");
1692                 return;
1693
1694         case TCPS_FIN_WAIT_1:
1695                 db_printf("TCPS_FIN_WAIT_1");
1696                 return;
1697
1698         case TCPS_CLOSING:
1699                 db_printf("TCPS_CLOSING");
1700                 return;
1701
1702         case TCPS_LAST_ACK:
1703                 db_printf("TCPS_LAST_ACK");
1704                 return;
1705
1706         case TCPS_FIN_WAIT_2:
1707                 db_printf("TCPS_FIN_WAIT_2");
1708                 return;
1709
1710         case TCPS_TIME_WAIT:
1711                 db_printf("TCPS_TIME_WAIT");
1712                 return;
1713
1714         default:
1715                 db_printf("unknown");
1716                 return;
1717         }
1718 }
1719
1720 static void
1721 db_print_tflags(u_int t_flags)
1722 {
1723         int comma;
1724
1725         comma = 0;
1726         if (t_flags & TF_ACKNOW) {
1727                 db_printf("%sTF_ACKNOW", comma ? ", " : "");
1728                 comma = 1;
1729         }
1730         if (t_flags & TF_DELACK) {
1731                 db_printf("%sTF_DELACK", comma ? ", " : "");
1732                 comma = 1;
1733         }
1734         if (t_flags & TF_NODELAY) {
1735                 db_printf("%sTF_NODELAY", comma ? ", " : "");
1736                 comma = 1;
1737         }
1738         if (t_flags & TF_NOOPT) {
1739                 db_printf("%sTF_NOOPT", comma ? ", " : "");
1740                 comma = 1;
1741         }
1742         if (t_flags & TF_SENTFIN) {
1743                 db_printf("%sTF_SENTFIN", comma ? ", " : "");
1744                 comma = 1;
1745         }
1746         if (t_flags & TF_REQ_SCALE) {
1747                 db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
1748                 comma = 1;
1749         }
1750         if (t_flags & TF_RCVD_SCALE) {
1751                 db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
1752                 comma = 1;
1753         }
1754         if (t_flags & TF_REQ_TSTMP) {
1755                 db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
1756                 comma = 1;
1757         }
1758         if (t_flags & TF_RCVD_TSTMP) {
1759                 db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
1760                 comma = 1;
1761         }
1762         if (t_flags & TF_SACK_PERMIT) {
1763                 db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
1764                 comma = 1;
1765         }
1766         if (t_flags & TF_NEEDSYN) {
1767                 db_printf("%sTF_NEEDSYN", comma ? ", " : "");
1768                 comma = 1;
1769         }
1770         if (t_flags & TF_NEEDFIN) {
1771                 db_printf("%sTF_NEEDFIN", comma ? ", " : "");
1772                 comma = 1;
1773         }
1774         if (t_flags & TF_NOPUSH) {
1775                 db_printf("%sTF_NOPUSH", comma ? ", " : "");
1776                 comma = 1;
1777         }
1778         if (t_flags & TF_MORETOCOME) {
1779                 db_printf("%sTF_MORETOCOME", comma ? ", " : "");
1780                 comma = 1;
1781         }
1782         if (t_flags & TF_LQ_OVERFLOW) {
1783                 db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : "");
1784                 comma = 1;
1785         }
1786         if (t_flags & TF_LASTIDLE) {
1787                 db_printf("%sTF_LASTIDLE", comma ? ", " : "");
1788                 comma = 1;
1789         }
1790         if (t_flags & TF_RXWIN0SENT) {
1791                 db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
1792                 comma = 1;
1793         }
1794         if (t_flags & TF_FASTRECOVERY) {
1795                 db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
1796                 comma = 1;
1797         }
1798         if (t_flags & TF_CONGRECOVERY) {
1799                 db_printf("%sTF_CONGRECOVERY", comma ? ", " : "");
1800                 comma = 1;
1801         }
1802         if (t_flags & TF_WASFRECOVERY) {
1803                 db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
1804                 comma = 1;
1805         }
1806         if (t_flags & TF_SIGNATURE) {
1807                 db_printf("%sTF_SIGNATURE", comma ? ", " : "");
1808                 comma = 1;
1809         }
1810         if (t_flags & TF_FORCEDATA) {
1811                 db_printf("%sTF_FORCEDATA", comma ? ", " : "");
1812                 comma = 1;
1813         }
1814         if (t_flags & TF_TSO) {
1815                 db_printf("%sTF_TSO", comma ? ", " : "");
1816                 comma = 1;
1817         }
1818         if (t_flags & TF_ECN_PERMIT) {
1819                 db_printf("%sTF_ECN_PERMIT", comma ? ", " : "");
1820                 comma = 1;
1821         }
1822 }
1823
1824 static void
1825 db_print_toobflags(char t_oobflags)
1826 {
1827         int comma;
1828
1829         comma = 0;
1830         if (t_oobflags & TCPOOB_HAVEDATA) {
1831                 db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
1832                 comma = 1;
1833         }
1834         if (t_oobflags & TCPOOB_HADDATA) {
1835                 db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
1836                 comma = 1;
1837         }
1838 }
1839
1840 static void
1841 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
1842 {
1843
1844         db_print_indent(indent);
1845         db_printf("%s at %p\n", name, tp);
1846
1847         indent += 2;
1848
1849         db_print_indent(indent);
1850         db_printf("t_segq first: %p   t_segqlen: %d   t_dupacks: %d\n",
1851            LIST_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks);
1852
1853         db_print_indent(indent);
1854         db_printf("tt_rexmt: %p   tt_persist: %p   tt_keep: %p\n",
1855             &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep);
1856
1857         db_print_indent(indent);
1858         db_printf("tt_2msl: %p   tt_delack: %p   t_inpcb: %p\n", &tp->t_timers->tt_2msl,
1859             &tp->t_timers->tt_delack, tp->t_inpcb);
1860
1861         db_print_indent(indent);
1862         db_printf("t_state: %d (", tp->t_state);
1863         db_print_tstate(tp->t_state);
1864         db_printf(")\n");
1865
1866         db_print_indent(indent);
1867         db_printf("t_flags: 0x%x (", tp->t_flags);
1868         db_print_tflags(tp->t_flags);
1869         db_printf(")\n");
1870
1871         db_print_indent(indent);
1872         db_printf("snd_una: 0x%08x   snd_max: 0x%08x   snd_nxt: x0%08x\n",
1873             tp->snd_una, tp->snd_max, tp->snd_nxt);
1874
1875         db_print_indent(indent);
1876         db_printf("snd_up: 0x%08x   snd_wl1: 0x%08x   snd_wl2: 0x%08x\n",
1877            tp->snd_up, tp->snd_wl1, tp->snd_wl2);
1878
1879         db_print_indent(indent);
1880         db_printf("iss: 0x%08x   irs: 0x%08x   rcv_nxt: 0x%08x\n",
1881             tp->iss, tp->irs, tp->rcv_nxt);
1882
1883         db_print_indent(indent);
1884         db_printf("rcv_adv: 0x%08x   rcv_wnd: %lu   rcv_up: 0x%08x\n",
1885             tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
1886
1887         db_print_indent(indent);
1888         db_printf("snd_wnd: %lu   snd_cwnd: %lu\n",
1889            tp->snd_wnd, tp->snd_cwnd);
1890
1891         db_print_indent(indent);
1892         db_printf("snd_ssthresh: %lu   snd_recover: "
1893             "0x%08x\n", tp->snd_ssthresh, tp->snd_recover);
1894
1895         db_print_indent(indent);
1896         db_printf("t_maxopd: %u   t_rcvtime: %u   t_startime: %u\n",
1897             tp->t_maxopd, tp->t_rcvtime, tp->t_starttime);
1898
1899         db_print_indent(indent);
1900         db_printf("t_rttime: %u   t_rtsq: 0x%08x\n",
1901             tp->t_rtttime, tp->t_rtseq);
1902
1903         db_print_indent(indent);
1904         db_printf("t_rxtcur: %d   t_maxseg: %u   t_srtt: %d\n",
1905             tp->t_rxtcur, tp->t_maxseg, tp->t_srtt);
1906
1907         db_print_indent(indent);
1908         db_printf("t_rttvar: %d   t_rxtshift: %d   t_rttmin: %u   "
1909             "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin,
1910             tp->t_rttbest);
1911
1912         db_print_indent(indent);
1913         db_printf("t_rttupdated: %lu   max_sndwnd: %lu   t_softerror: %d\n",
1914             tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
1915
1916         db_print_indent(indent);
1917         db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
1918         db_print_toobflags(tp->t_oobflags);
1919         db_printf(")   t_iobc: 0x%02x\n", tp->t_iobc);
1920
1921         db_print_indent(indent);
1922         db_printf("snd_scale: %u   rcv_scale: %u   request_r_scale: %u\n",
1923             tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
1924
1925         db_print_indent(indent);
1926         db_printf("ts_recent: %u   ts_recent_age: %u\n",
1927             tp->ts_recent, tp->ts_recent_age);
1928
1929         db_print_indent(indent);
1930         db_printf("ts_offset: %u   last_ack_sent: 0x%08x   snd_cwnd_prev: "
1931             "%lu\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
1932
1933         db_print_indent(indent);
1934         db_printf("snd_ssthresh_prev: %lu   snd_recover_prev: 0x%08x   "
1935             "t_badrxtwin: %u\n", tp->snd_ssthresh_prev,
1936             tp->snd_recover_prev, tp->t_badrxtwin);
1937
1938         db_print_indent(indent);
1939         db_printf("snd_numholes: %d  snd_holes first: %p\n",
1940             tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
1941
1942         db_print_indent(indent);
1943         db_printf("snd_fack: 0x%08x   rcv_numsacks: %d   sack_newdata: "
1944             "0x%08x\n", tp->snd_fack, tp->rcv_numsacks, tp->sack_newdata);
1945
1946         /* Skip sackblks, sackhint. */
1947
1948         db_print_indent(indent);
1949         db_printf("t_rttlow: %d   rfbuf_ts: %u   rfbuf_cnt: %d\n",
1950             tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
1951 }
1952
1953 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
1954 {
1955         struct tcpcb *tp;
1956
1957         if (!have_addr) {
1958                 db_printf("usage: show tcpcb <addr>\n");
1959                 return;
1960         }
1961         tp = (struct tcpcb *)addr;
1962
1963         db_print_tcpcb(tp, "tcpcb", 0);
1964 }
1965 #endif