bsd/netinet/tcp_usrreq.c

   1 /*
   2  * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_LICENSE_HEADER_START@
   5  *
   6  * Copyright (c) 1999-2003 Apple Computer, Inc.  All Rights Reserved.
   7  *
   8  * This file contains Original Code and/or Modifications of Original Code
   9  * as defined in and that are subject to the Apple Public Source License
  10  * Version 2.0 (the 'License'). You may not use this file except in
  11  * compliance with the License. Please obtain a copy of the License at
  12  * http://www.opensource.apple.com/apsl/ and read it before using this
  13  * file.
  14  *
  15  * The Original Code and all software distributed under the License are
  16  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  17  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  18  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  19  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  20  * Please see the License for the specific language governing rights and
  21  * limitations under the License.
  22  *
  23  * @APPLE_LICENSE_HEADER_END@
  24  */
  25 /*
  26  * Copyright (c) 1982, 1986, 1988, 1993
  27  *      The Regents of the University of California.  All rights reserved.
  28  *
  29  * Redistribution and use in source and binary forms, with or without
  30  * modification, are permitted provided that the following conditions
  31  * are met:
  32  * 1. Redistributions of source code must retain the above copyright
  33  *    notice, this list of conditions and the following disclaimer.
  34  * 2. Redistributions in binary form must reproduce the above copyright
  35  *    notice, this list of conditions and the following disclaimer in the
  36  *    documentation and/or other materials provided with the distribution.
  37  * 3. All advertising materials mentioning features or use of this software
  38  *    must display the following acknowledgement:
  39  *      This product includes software developed by the University of
  40  *      California, Berkeley and its contributors.
  41  * 4. Neither the name of the University nor the names of its contributors
  42  *    may be used to endorse or promote products derived from this software
  43  *    without specific prior written permission.
  44  *
  45  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  46  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  47  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  48  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  49  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  50  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  51  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  52  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  53  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  54  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  55  * SUCH DAMAGE.
  56  *
  57  *      From: @(#)tcp_usrreq.c  8.2 (Berkeley) 1/3/94
  58  * $FreeBSD: src/sys/netinet/tcp_usrreq.c,v 1.51.2.9 2001/08/22 00:59:12 silby Exp $
  59  */
  60
  61
  62 #include <sys/param.h>
  63 #include <sys/systm.h>
  64 #include <sys/kernel.h>
  65 #include <sys/sysctl.h>
  66 #include <sys/mbuf.h>
  67 #if INET6
  68 #include <sys/domain.h>
  69 #endif /* INET6 */
  70 #include <sys/socket.h>
  71 #include <sys/socketvar.h>
  72 #include <sys/protosw.h>
  73
  74 #include <net/if.h>
  75 #include <net/route.h>
  76
  77 #include <netinet/in.h>
  78 #include <netinet/in_systm.h>
  79 #if INET6
  80 #include <netinet/ip6.h>
  81 #endif
  82 #include <netinet/in_pcb.h>
  83 #if INET6
  84 #include <netinet6/in6_pcb.h>
  85 #endif
  86 #include <netinet/in_var.h>
  87 #include <netinet/ip_var.h>
  88 #if INET6
  89 #include <netinet6/ip6_var.h>
  90 #endif
  91 #include <netinet/tcp.h>
  92 #include <netinet/tcp_fsm.h>
  93 #include <netinet/tcp_seq.h>
  94 #include <netinet/tcp_timer.h>
  95 #include <netinet/tcp_var.h>
  96 #include <netinet/tcpip.h>
  97 #if TCPDEBUG
  98 #include <netinet/tcp_debug.h>
  99 #endif
 100
 101 #if IPSEC
 102 #include <netinet6/ipsec.h>
 103 #endif /*IPSEC*/
 104
 105 /*
 106  * TCP protocol interface to socket abstraction.
 107  */
 108 extern  char *tcpstates[];      /* XXX ??? */
 109
 110 static int      tcp_attach __P((struct socket *, struct proc *));
 111 static int      tcp_connect __P((struct tcpcb *, struct sockaddr *,
 112                                  struct proc *));
 113 #if INET6
 114 static int      tcp6_connect __P((struct tcpcb *, struct sockaddr *,
 115                                  struct proc *));
 116 #endif /* INET6 */
 117 static struct tcpcb *
 118                 tcp_disconnect __P((struct tcpcb *));
 119 static struct tcpcb *
 120                 tcp_usrclosed __P((struct tcpcb *));
 121
 122 #if TCPDEBUG
 123 #define TCPDEBUG0       int ostate = 0
 124 #define TCPDEBUG1()     ostate = tp ? tp->t_state : 0
 125 #define TCPDEBUG2(req)  if (tp && (so->so_options & SO_DEBUG)) \
 126                                 tcp_trace(TA_USER, ostate, tp, 0, 0, req)
 127 #else
 128 #define TCPDEBUG0
 129 #define TCPDEBUG1()
 130 #define TCPDEBUG2(req)
 131 #endif
 132
 133 /*
 134  * TCP attaches to socket via pru_attach(), reserving space,
 135  * and an internet control block.
 136  */
 137 static int
 138 tcp_usr_attach(struct socket *so, int proto, struct proc *p)
 139 {
 140         int s = splnet();
 141         int error;
 142         struct inpcb *inp = sotoinpcb(so);
 143         struct tcpcb *tp = 0;
 144         TCPDEBUG0;
 145
 146         TCPDEBUG1();
 147         if (inp) {
 148                 error = EISCONN;
 149                 goto out;
 150         }
 151
 152         error = tcp_attach(so, p);
 153         if (error)
 154                 goto out;
 155
 156         if ((so->so_options & SO_LINGER) && so->so_linger == 0)
 157                 so->so_linger = TCP_LINGERTIME * hz;
 158         tp = sototcpcb(so);
 159 out:
 160         TCPDEBUG2(PRU_ATTACH);
 161         splx(s);
 162         return error;
 163 }
 164
 165 /*
 166  * pru_detach() detaches the TCP protocol from the socket.
 167  * If the protocol state is non-embryonic, then can't
 168  * do this directly: have to initiate a pru_disconnect(),
 169  * which may finish later; embryonic TCB's can just
 170  * be discarded here.
 171  */
 172 static int
 173 tcp_usr_detach(struct socket *so)
 174 {
 175         int s = splnet();
 176         int error = 0;
 177         struct inpcb *inp = sotoinpcb(so);
 178         struct tcpcb *tp;
 179         TCPDEBUG0;
 180
 181         if (inp == 0) {
 182                 splx(s);
 183                 return EINVAL;  /* XXX */
 184         }
 185         tp = intotcpcb(inp);
 186         /* In case we got disconnected from the peer */
 187         if (tp == 0)
 188             goto out;
 189         TCPDEBUG1();
 190         tp = tcp_disconnect(tp);
 191 out:
 192         TCPDEBUG2(PRU_DETACH);
 193         splx(s);
 194         return error;
 195 }
 196
 197 #define COMMON_START()  TCPDEBUG0; \
 198                         do { \
 199                                      if (inp == 0) { \
 200                                              splx(s); \
 201                                              return EINVAL; \
 202                                      } \
 203                                      tp = intotcpcb(inp); \
 204                                      TCPDEBUG1(); \
 205                      } while(0)
 206
 207 #define COMMON_END(req) out: TCPDEBUG2(req); splx(s); return error; goto out
 208
 209
 210 /*
 211  * Give the socket an address.
 212  */
 213 static int
 214 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
 215 {
 216         int s = splnet();
 217         int error = 0;
 218         struct inpcb *inp = sotoinpcb(so);
 219         struct tcpcb *tp;
 220         struct sockaddr_in *sinp;
 221
 222         COMMON_START();
 223
 224         /*
 225          * Must check for multicast addresses and disallow binding
 226          * to them.
 227          */
 228         sinp = (struct sockaddr_in *)nam;
 229         if (sinp->sin_family == AF_INET &&
 230             IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
 231                 error = EAFNOSUPPORT;
 232                 goto out;
 233         }
 234         error = in_pcbbind(inp, nam, p);
 235         if (error)
 236                 goto out;
 237         COMMON_END(PRU_BIND);
 238
 239 }
 240
 241 #if INET6
 242 static int
 243 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
 244 {
 245         int s = splnet();
 246         int error = 0;
 247         struct inpcb *inp = sotoinpcb(so);
 248         struct tcpcb *tp;
 249         struct sockaddr_in6 *sin6p;
 250
 251         COMMON_START();
 252
 253         /*
 254          * Must check for multicast addresses and disallow binding
 255          * to them.
 256          */
 257         sin6p = (struct sockaddr_in6 *)nam;
 258         if (sin6p->sin6_family == AF_INET6 &&
 259             IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) {
 260                 error = EAFNOSUPPORT;
 261                 goto out;
 262         }
 263         inp->inp_vflag &= ~INP_IPV4;
 264         inp->inp_vflag |= INP_IPV6;
 265         if (ip6_mapped_addr_on && (inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
 266                 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
 267                         inp->inp_vflag |= INP_IPV4;
 268                 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
 269                         struct sockaddr_in sin;
 270
 271                         in6_sin6_2_sin(&sin, sin6p);
 272                         inp->inp_vflag |= INP_IPV4;
 273                         inp->inp_vflag &= ~INP_IPV6;
 274                         error = in_pcbbind(inp, (struct sockaddr *)&sin, p);
 275                         goto out;
 276                 }
 277         }
 278         error = in6_pcbbind(inp, nam, p);
 279         if (error)
 280                 goto out;
 281         COMMON_END(PRU_BIND);
 282 }
 283 #endif /* INET6 */
 284
 285 /*
 286  * Prepare to accept connections.
 287  */
 288 static int
 289 tcp_usr_listen(struct socket *so, struct proc *p)
 290 {
 291         int s = splnet();
 292         int error = 0;
 293         struct inpcb *inp = sotoinpcb(so);
 294         struct tcpcb *tp;
 295
 296         COMMON_START();
 297         if (inp->inp_lport == 0)
 298                 error = in_pcbbind(inp, (struct sockaddr *)0, p);
 299         if (error == 0)
 300                 tp->t_state = TCPS_LISTEN;
 301         COMMON_END(PRU_LISTEN);
 302 }
 303
 304 #if INET6
 305 static int
 306 tcp6_usr_listen(struct socket *so, struct proc *p)
 307 {
 308         int s = splnet();
 309         int error = 0;
 310         struct inpcb *inp = sotoinpcb(so);
 311         struct tcpcb *tp;
 312
 313         COMMON_START();
 314         if (inp->inp_lport == 0) {
 315                 inp->inp_vflag &= ~INP_IPV4;
 316                 if (ip6_mapped_addr_on &&
 317                     (inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
 318                         inp->inp_vflag |= INP_IPV4;
 319                 error = in6_pcbbind(inp, (struct sockaddr *)0, p);
 320         }
 321         if (error == 0)
 322                 tp->t_state = TCPS_LISTEN;
 323         COMMON_END(PRU_LISTEN);
 324 }
 325 #endif /* INET6 */
 326
 327 /*
 328  * Initiate connection to peer.
 329  * Create a template for use in transmissions on this connection.
 330  * Enter SYN_SENT state, and mark socket as connecting.
 331  * Start keep-alive timer, and seed output sequence space.
 332  * Send initial segment on connection.
 333  */
 334 static int
 335 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
 336 {
 337         int s = splnet();
 338         int error = 0;
 339         struct inpcb *inp = sotoinpcb(so);
 340         struct tcpcb *tp;
 341         struct sockaddr_in *sinp;
 342
 343         COMMON_START();
 344
 345         /*
 346          * Must disallow TCP ``connections'' to multicast addresses.
 347          */
 348         sinp = (struct sockaddr_in *)nam;
 349         if (sinp->sin_family == AF_INET
 350             && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
 351                 error = EAFNOSUPPORT;
 352                 goto out;
 353         }
 354
 355 #ifndef __APPLE__
 356         prison_remote_ip(p, 0, &sinp->sin_addr.s_addr);
 357 #endif
 358
 359         if ((error = tcp_connect(tp, nam, p)) != 0)
 360                 goto out;
 361         error = tcp_output(tp);
 362         COMMON_END(PRU_CONNECT);
 363 }
 364
 365 #if INET6
 366 static int
 367 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
 368 {
 369         int s = splnet();
 370         int error = 0;
 371         struct inpcb *inp = sotoinpcb(so);
 372         struct tcpcb *tp;
 373         struct sockaddr_in6 *sin6p;
 374
 375         COMMON_START();
 376
 377         /*
 378          * Must disallow TCP ``connections'' to multicast addresses.
 379          */
 380         sin6p = (struct sockaddr_in6 *)nam;
 381         if (sin6p->sin6_family == AF_INET6
 382             && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) {
 383                 error = EAFNOSUPPORT;
 384                 goto out;
 385         }
 386
 387         if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
 388                 struct sockaddr_in sin;
 389
 390                 if (!ip6_mapped_addr_on ||
 391                     (inp->inp_flags & IN6P_IPV6_V6ONLY))
 392                         return(EINVAL);
 393
 394                 in6_sin6_2_sin(&sin, sin6p);
 395                 inp->inp_vflag |= INP_IPV4;
 396                 inp->inp_vflag &= ~INP_IPV6;
 397                 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, p)) != 0)
 398                         goto out;
 399                 error = tcp_output(tp);
 400                 goto out;
 401         }
 402         inp->inp_vflag &= ~INP_IPV4;
 403         inp->inp_vflag |= INP_IPV6;
 404         if ((error = tcp6_connect(tp, nam, p)) != 0)
 405                 goto out;
 406         error = tcp_output(tp);
 407         if (error)
 408                 goto out;
 409         COMMON_END(PRU_CONNECT);
 410 }
 411 #endif /* INET6 */
 412
 413 /*
 414  * Initiate disconnect from peer.
 415  * If connection never passed embryonic stage, just drop;
 416  * else if don't need to let data drain, then can just drop anyways,
 417  * else have to begin TCP shutdown process: mark socket disconnecting,
 418  * drain unread data, state switch to reflect user close, and
 419  * send segment (e.g. FIN) to peer.  Socket will be really disconnected
 420  * when peer sends FIN and acks ours.
 421  *
 422  * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
 423  */
 424 static int
 425 tcp_usr_disconnect(struct socket *so)
 426 {
 427         int s = splnet();
 428         int error = 0;
 429         struct inpcb *inp = sotoinpcb(so);
 430         struct tcpcb *tp;
 431
 432         COMMON_START();
 433         /* In case we got disconnected from the peer */
 434         if (tp == 0)
 435             goto out;
 436         tp = tcp_disconnect(tp);
 437         COMMON_END(PRU_DISCONNECT);
 438 }
 439
 440 /*
 441  * Accept a connection.  Essentially all the work is
 442  * done at higher levels; just return the address
 443  * of the peer, storing through addr.
 444  */
 445 static int
 446 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
 447 {
 448         int s = splnet();
 449         int error = 0;
 450         struct inpcb *inp = sotoinpcb(so);
 451         struct tcpcb *tp = NULL;
 452         TCPDEBUG0;
 453
 454         if (so->so_state & SS_ISDISCONNECTED) {
 455                 error = ECONNABORTED;
 456                 goto out;
 457         }
 458         if (inp == 0) {
 459                 splx(s);
 460                 return (EINVAL);
 461         }
 462         tp = intotcpcb(inp);
 463         TCPDEBUG1();
 464         in_setpeeraddr(so, nam);
 465         COMMON_END(PRU_ACCEPT);
 466 }
 467
 468 #if INET6
 469 static int
 470 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
 471 {
 472         int s = splnet();
 473         int error = 0;
 474         struct inpcb *inp = sotoinpcb(so);
 475         struct tcpcb *tp = NULL;
 476         TCPDEBUG0;
 477
 478         if (so->so_state & SS_ISDISCONNECTED) {
 479                 error = ECONNABORTED;
 480                 goto out;
 481         }
 482         if (inp == 0) {
 483                 splx(s);
 484                 return (EINVAL);
 485         }
 486         tp = intotcpcb(inp);
 487         TCPDEBUG1();
 488         in6_mapped_peeraddr(so, nam);
 489         COMMON_END(PRU_ACCEPT);
 490 }
 491 #endif /* INET6 */
 492 /*
 493  * Mark the connection as being incapable of further output.
 494  */
 495 static int
 496 tcp_usr_shutdown(struct socket *so)
 497 {
 498         int s = splnet();
 499         int error = 0;
 500         struct inpcb *inp = sotoinpcb(so);
 501         struct tcpcb *tp;
 502
 503         COMMON_START();
 504         socantsendmore(so);
 505         /* In case we got disconnected from the peer */
 506         if (tp == 0)
 507             goto out;
 508         tp = tcp_usrclosed(tp);
 509         if (tp)
 510                 error = tcp_output(tp);
 511         COMMON_END(PRU_SHUTDOWN);
 512 }
 513
 514 /*
 515  * After a receive, possibly send window update to peer.
 516  */
 517 static int
 518 tcp_usr_rcvd(struct socket *so, int flags)
 519 {
 520         int s = splnet();
 521         int error = 0;
 522         struct inpcb *inp = sotoinpcb(so);
 523         struct tcpcb *tp;
 524
 525         COMMON_START();
 526         /* In case we got disconnected from the peer */
 527         if (tp == 0)
 528             goto out;
 529         tcp_output(tp);
 530         COMMON_END(PRU_RCVD);
 531 }
 532
 533 /*
 534  * Do a send by putting data in output queue and updating urgent
 535  * marker if URG set.  Possibly send more data.  Unlike the other
 536  * pru_*() routines, the mbuf chains are our responsibility.  We
 537  * must either enqueue them or free them.  The other pru_* routines
 538  * generally are caller-frees.
 539  */
 540 static int
 541 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
 542              struct sockaddr *nam, struct mbuf *control, struct proc *p)
 543 {
 544         int s = splnet();
 545         int error = 0;
 546         struct inpcb *inp = sotoinpcb(so);
 547         struct tcpcb *tp;
 548 #if INET6
 549         int isipv6;
 550 #endif
 551         TCPDEBUG0;
 552
 553         if (inp == NULL) {
 554                 /*
 555                  * OOPS! we lost a race, the TCP session got reset after
 556                  * we checked SS_CANTSENDMORE, eg: while doing uiomove or a
 557                  * network interrupt in the non-splnet() section of sosend().
 558                  */
 559                 if (m)
 560                         m_freem(m);
 561                 if (control)
 562                         m_freem(control);
 563                 error = ECONNRESET;     /* XXX EPIPE? */
 564                 tp = NULL;
 565                 TCPDEBUG1();
 566                 goto out;
 567         }
 568 #if INET6
 569         isipv6 = nam && nam->sa_family == AF_INET6;
 570 #endif /* INET6 */
 571         tp = intotcpcb(inp);
 572         TCPDEBUG1();
 573         if (control) {
 574                 /* TCP doesn't do control messages (rights, creds, etc) */
 575                 if (control->m_len) {
 576                         m_freem(control);
 577                         if (m)
 578                                 m_freem(m);
 579                         error = EINVAL;
 580                         goto out;
 581                 }
 582                 m_freem(control);       /* empty control, just free it */
 583         }
 584         if(!(flags & PRUS_OOB)) {
 585                 sbappend(&so->so_snd, m);
 586                 if (nam && tp->t_state < TCPS_SYN_SENT) {
 587                         /*
 588                          * Do implied connect if not yet connected,
 589                          * initialize window to default value, and
 590                          * initialize maxseg/maxopd using peer's cached
 591                          * MSS.
 592                          */
 593 #if INET6
 594                         if (isipv6)
 595                                 error = tcp6_connect(tp, nam, p);
 596                         else
 597 #endif /* INET6 */
 598                         error = tcp_connect(tp, nam, p);
 599                         if (error)
 600                                 goto out;
 601                         tp->snd_wnd = TTCP_CLIENT_SND_WND;
 602                         tcp_mss(tp, -1);
 603                 }
 604
 605                 if (flags & PRUS_EOF) {
 606                         /*
 607                          * Close the send side of the connection after
 608                          * the data is sent.
 609                          */
 610                         socantsendmore(so);
 611                         tp = tcp_usrclosed(tp);
 612                 }
 613                 if (tp != NULL) {
 614                         if (flags & PRUS_MORETOCOME)
 615                                 tp->t_flags |= TF_MORETOCOME;
 616                         error = tcp_output(tp);
 617                         if (flags & PRUS_MORETOCOME)
 618                                 tp->t_flags &= ~TF_MORETOCOME;
 619                 }
 620         } else {
 621                 if (sbspace(&so->so_snd) < -512) {
 622                         m_freem(m);
 623                         error = ENOBUFS;
 624                         goto out;
 625                 }
 626                 /*
 627                  * According to RFC961 (Assigned Protocols),
 628                  * the urgent pointer points to the last octet
 629                  * of urgent data.  We continue, however,
 630                  * to consider it to indicate the first octet
 631                  * of data past the urgent section.
 632                  * Otherwise, snd_up should be one lower.
 633                  */
 634                 sbappend(&so->so_snd, m);
 635                 if (nam && tp->t_state < TCPS_SYN_SENT) {
 636                         /*
 637                          * Do implied connect if not yet connected,
 638                          * initialize window to default value, and
 639                          * initialize maxseg/maxopd using peer's cached
 640                          * MSS.
 641                          */
 642 #if INET6
 643                         if (isipv6)
 644                                 error = tcp6_connect(tp, nam, p);
 645                         else
 646 #endif /* INET6 */
 647                         error = tcp_connect(tp, nam, p);
 648                         if (error)
 649                                 goto out;
 650                         tp->snd_wnd = TTCP_CLIENT_SND_WND;
 651                         tcp_mss(tp, -1);
 652                 }
 653                 tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
 654                 tp->t_force = 1;
 655                 error = tcp_output(tp);
 656                 tp->t_force = 0;
 657         }
 658         COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
 659                    ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
 660 }
 661
 662 /*
 663  * Abort the TCP.
 664  */
 665 static int
 666 tcp_usr_abort(struct socket *so)
 667 {
 668         int s = splnet();
 669         int error = 0;
 670         struct inpcb *inp = sotoinpcb(so);
 671         struct tcpcb *tp;
 672
 673         COMMON_START();
 674         /* In case we got disconnected from the peer */
 675         if (tp == 0)
 676             goto out;
 677         tp = tcp_drop(tp, ECONNABORTED);
 678         COMMON_END(PRU_ABORT);
 679 }
 680
 681 /*
 682  * Receive out-of-band data.
 683  */
 684 static int
 685 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
 686 {
 687         int s = splnet();
 688         int error = 0;
 689         struct inpcb *inp = sotoinpcb(so);
 690         struct tcpcb *tp;
 691
 692         COMMON_START();
 693         if ((so->so_oobmark == 0 &&
 694              (so->so_state & SS_RCVATMARK) == 0) ||
 695             so->so_options & SO_OOBINLINE ||
 696             tp->t_oobflags & TCPOOB_HADDATA) {
 697                 error = EINVAL;
 698                 goto out;
 699         }
 700         if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
 701                 error = EWOULDBLOCK;
 702                 goto out;
 703         }
 704         m->m_len = 1;
 705         *mtod(m, caddr_t) = tp->t_iobc;
 706         if ((flags & MSG_PEEK) == 0)
 707                 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
 708         COMMON_END(PRU_RCVOOB);
 709 }
 710
 711 /* xxx - should be const */
 712 struct pr_usrreqs tcp_usrreqs = {
 713         tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind,
 714         tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach,
 715         tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd,
 716         tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown,
 717         in_setsockaddr, sosend, soreceive, sopoll
 718 };
 719
 720 #if INET6
 721 struct pr_usrreqs tcp6_usrreqs = {
 722         tcp_usr_abort, tcp6_usr_accept, tcp_usr_attach, tcp6_usr_bind,
 723         tcp6_usr_connect, pru_connect2_notsupp, in6_control, tcp_usr_detach,
 724         tcp_usr_disconnect, tcp6_usr_listen, in6_mapped_peeraddr, tcp_usr_rcvd,
 725         tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown,
 726         in6_mapped_sockaddr, sosend, soreceive, sopoll
 727 };
 728 #endif /* INET6 */
 729
 730 /*
 731  * Common subroutine to open a TCP connection to remote host specified
 732  * by struct sockaddr_in in mbuf *nam.  Call in_pcbbind to assign a local
 733  * port number if needed.  Call in_pcbladdr to do the routing and to choose
 734  * a local host address (interface).  If there is an existing incarnation
 735  * of the same connection in TIME-WAIT state and if the remote host was
 736  * sending CC options and if the connection duration was < MSL, then
 737  * truncate the previous TIME-WAIT state and proceed.
 738  * Initialize connection parameters and enter SYN-SENT state.
 739  */
 740 static int
 741 tcp_connect(tp, nam, p)
 742         register struct tcpcb *tp;
 743         struct sockaddr *nam;
 744         struct proc *p;
 745 {
 746         struct inpcb *inp = tp->t_inpcb, *oinp;
 747         struct socket *so = inp->inp_socket;
 748         struct tcpcb *otp;
 749         struct sockaddr_in *sin = (struct sockaddr_in *)nam;
 750         struct sockaddr_in *ifaddr;
 751         struct rmxp_tao *taop;
 752         struct rmxp_tao tao_noncached;
 753         int error;
 754
 755         if (inp->inp_lport == 0) {
 756                 error = in_pcbbind(inp, (struct sockaddr *)0, p);
 757                 if (error)
 758                         return error;
 759         }
 760
 761         /*
 762          * Cannot simply call in_pcbconnect, because there might be an
 763          * earlier incarnation of this same connection still in
 764          * TIME_WAIT state, creating an ADDRINUSE error.
 765          */
 766         error = in_pcbladdr(inp, nam, &ifaddr);
 767         if (error)
 768                 return error;
 769         oinp = in_pcblookup_hash(inp->inp_pcbinfo,
 770             sin->sin_addr, sin->sin_port,
 771             inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr
 772                                                 : ifaddr->sin_addr,
 773             inp->inp_lport,  0, NULL);
 774         if (oinp) {
 775                 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
 776                 otp->t_state == TCPS_TIME_WAIT &&
 777                     otp->t_starttime < tcp_msl &&
 778                     (otp->t_flags & TF_RCVD_CC))
 779                         otp = tcp_close(otp);
 780                 else
 781                         return EADDRINUSE;
 782         }
 783         if ((inp->inp_laddr.s_addr == INADDR_ANY ? ifaddr->sin_addr.s_addr :
 784                  inp->inp_laddr.s_addr) == sin->sin_addr.s_addr &&
 785             inp->inp_lport == sin->sin_port)
 786                         return EINVAL;
 787         if (inp->inp_laddr.s_addr == INADDR_ANY)
 788                 inp->inp_laddr = ifaddr->sin_addr;
 789         inp->inp_faddr = sin->sin_addr;
 790         inp->inp_fport = sin->sin_port;
 791         in_pcbrehash(inp);
 792
 793         /* Compute window scaling to request.  */
 794         while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
 795             (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
 796                 tp->request_r_scale++;
 797
 798         soisconnecting(so);
 799         tcpstat.tcps_connattempt++;
 800         tp->t_state = TCPS_SYN_SENT;
 801         tp->t_timer[TCPT_KEEP] = tcp_keepinit;
 802         tp->iss = tcp_new_isn(tp);
 803         tcp_sendseqinit(tp);
 804
 805         /*
 806          * Generate a CC value for this connection and
 807          * check whether CC or CCnew should be used.
 808          */
 809         if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) {
 810                 taop = &tao_noncached;
 811                 bzero(taop, sizeof(*taop));
 812         }
 813
 814         tp->cc_send = CC_INC(tcp_ccgen);
 815         if (taop->tao_ccsent != 0 &&
 816             CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
 817                 taop->tao_ccsent = tp->cc_send;
 818         } else {
 819                 taop->tao_ccsent = 0;
 820                 tp->t_flags |= TF_SENDCCNEW;
 821         }
 822
 823         return 0;
 824 }
 825
 826 #if INET6
 827 static int
 828 tcp6_connect(tp, nam, p)
 829         register struct tcpcb *tp;
 830         struct sockaddr *nam;
 831         struct proc *p;
 832 {
 833         struct inpcb *inp = tp->t_inpcb, *oinp;
 834         struct socket *so = inp->inp_socket;
 835         struct tcpcb *otp;
 836         struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
 837         struct in6_addr *addr6;
 838         struct rmxp_tao *taop;
 839         struct rmxp_tao tao_noncached;
 840         int error;
 841
 842         if (inp->inp_lport == 0) {
 843                 error = in6_pcbbind(inp, (struct sockaddr *)0, p);
 844                 if (error)
 845                         return error;
 846         }
 847
 848         /*
 849          * Cannot simply call in_pcbconnect, because there might be an
 850          * earlier incarnation of this same connection still in
 851          * TIME_WAIT state, creating an ADDRINUSE error.
 852          */
 853         error = in6_pcbladdr(inp, nam, &addr6);
 854         if (error)
 855                 return error;
 856         oinp = in6_pcblookup_hash(inp->inp_pcbinfo,
 857                                   &sin6->sin6_addr, sin6->sin6_port,
 858                                   IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
 859                                   ? addr6
 860                                   : &inp->in6p_laddr,
 861                                   inp->inp_lport,  0, NULL);
 862         if (oinp) {
 863                 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
 864                     otp->t_state == TCPS_TIME_WAIT &&
 865                     otp->t_starttime < tcp_msl &&
 866                     (otp->t_flags & TF_RCVD_CC))
 867                         otp = tcp_close(otp);
 868                 else
 869                         return EADDRINUSE;
 870         }
 871         if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
 872                 inp->in6p_laddr = *addr6;
 873         inp->in6p_faddr = sin6->sin6_addr;
 874         inp->inp_fport = sin6->sin6_port;
 875         if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != NULL)
 876                 inp->in6p_flowinfo = sin6->sin6_flowinfo;
 877         in_pcbrehash(inp);
 878
 879         /* Compute window scaling to request.  */
 880         while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
 881             (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
 882                 tp->request_r_scale++;
 883
 884         soisconnecting(so);
 885         tcpstat.tcps_connattempt++;
 886         tp->t_state = TCPS_SYN_SENT;
 887         tp->t_timer[TCPT_KEEP] = tcp_keepinit;
 888         tp->iss = tcp_new_isn(tp);
 889         tcp_sendseqinit(tp);
 890
 891         /*
 892          * Generate a CC value for this connection and
 893          * check whether CC or CCnew should be used.
 894          */
 895         if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) {
 896                 taop = &tao_noncached;
 897                 bzero(taop, sizeof(*taop));
 898         }
 899
 900         tp->cc_send = CC_INC(tcp_ccgen);
 901         if (taop->tao_ccsent != 0 &&
 902             CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
 903                 taop->tao_ccsent = tp->cc_send;
 904         } else {
 905                 taop->tao_ccsent = 0;
 906                 tp->t_flags |= TF_SENDCCNEW;
 907         }
 908
 909         return 0;
 910 }
 911 #endif /* INET6 */
 912
 913 /*
 914  * The new sockopt interface makes it possible for us to block in the
 915  * copyin/out step (if we take a page fault).  Taking a page fault at
 916  * splnet() is probably a Bad Thing.  (Since sockets and pcbs both now
 917  * use TSM, there probably isn't any need for this function to run at
 918  * splnet() any more.  This needs more examination.)
 919  */
 920 int
 921 tcp_ctloutput(so, sopt)
 922         struct socket *so;
 923         struct sockopt *sopt;
 924 {
 925         int     error, opt, optval, s;
 926         struct  inpcb *inp;
 927         struct  tcpcb *tp;
 928
 929         error = 0;
 930         s = splnet();           /* XXX */
 931         inp = sotoinpcb(so);
 932         if (inp == NULL) {
 933                 splx(s);
 934                 return (ECONNRESET);
 935         }
 936         if (sopt->sopt_level != IPPROTO_TCP) {
 937 #if INET6
 938                 if (INP_CHECK_SOCKAF(so, AF_INET6))
 939                         error = ip6_ctloutput(so, sopt);
 940                 else
 941 #endif /* INET6 */
 942                 error = ip_ctloutput(so, sopt);
 943                 splx(s);
 944                 return (error);
 945         }
 946         tp = intotcpcb(inp);
 947         if (tp == NULL) {
 948                 splx(s);
 949                 return (ECONNRESET);
 950         }
 951
 952         switch (sopt->sopt_dir) {
 953         case SOPT_SET:
 954                 switch (sopt->sopt_name) {
 955                 case TCP_NODELAY:
 956                 case TCP_NOOPT:
 957                 case TCP_NOPUSH:
 958                         error = sooptcopyin(sopt, &optval, sizeof optval,
 959                                             sizeof optval);
 960                         if (error)
 961                                 break;
 962
 963                         switch (sopt->sopt_name) {
 964                         case TCP_NODELAY:
 965                                 opt = TF_NODELAY;
 966                                 break;
 967                         case TCP_NOOPT:
 968                                 opt = TF_NOOPT;
 969                                 break;
 970                         case TCP_NOPUSH:
 971                                 opt = TF_NOPUSH;
 972                                 break;
 973                         default:
 974                                 opt = 0; /* dead code to fool gcc */
 975                                 break;
 976                         }
 977
 978                         if (optval)
 979                                 tp->t_flags |= opt;
 980                         else
 981                                 tp->t_flags &= ~opt;
 982                         break;
 983
 984                 case TCP_MAXSEG:
 985                         error = sooptcopyin(sopt, &optval, sizeof optval,
 986                                             sizeof optval);
 987                         if (error)
 988                                 break;
 989
 990                         if (optval > 0 && optval <= tp->t_maxseg)
 991                                 tp->t_maxseg = optval;
 992                         else
 993                                 error = EINVAL;
 994                         break;
 995
 996                 default:
 997                         error = ENOPROTOOPT;
 998                         break;
 999                 }
1000                 break;
1001
1002         case SOPT_GET:
1003                 switch (sopt->sopt_name) {
1004                 case TCP_NODELAY:
1005                         optval = tp->t_flags & TF_NODELAY;
1006                         break;
1007                 case TCP_MAXSEG:
1008                         optval = tp->t_maxseg;
1009                         break;
1010                 case TCP_NOOPT:
1011                         optval = tp->t_flags & TF_NOOPT;
1012                         break;
1013                 case TCP_NOPUSH:
1014                         optval = tp->t_flags & TF_NOPUSH;
1015                         break;
1016                 default:
1017                         error = ENOPROTOOPT;
1018                         break;
1019                 }
1020                 if (error == 0)
1021                         error = sooptcopyout(sopt, &optval, sizeof optval);
1022                 break;
1023         }
1024         splx(s);
1025         return (error);
1026 }
1027
1028 /*
1029  * tcp_sendspace and tcp_recvspace are the default send and receive window
1030  * sizes, respectively.  These are obsolescent (this information should
1031  * be set by the route).
1032  */
1033 u_long  tcp_sendspace = 1024*16;
1034 SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW,
1035     &tcp_sendspace , 0, "Maximum outgoing TCP datagram size");
1036 u_long  tcp_recvspace = 1024*16;
1037 SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
1038     &tcp_recvspace , 0, "Maximum incoming TCP datagram size");
1039
1040 /*
1041  * Attach TCP protocol to socket, allocating
1042  * internet protocol control block, tcp control block,
1043  * bufer space, and entering LISTEN state if to accept connections.
1044  */
1045 static int
1046 tcp_attach(so, p)
1047         struct socket *so;
1048         struct proc *p;
1049 {
1050         register struct tcpcb *tp;
1051         struct inpcb *inp;
1052         int error;
1053 #if INET6
1054         int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != NULL;
1055 #endif
1056
1057         if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1058                 error = soreserve(so, tcp_sendspace, tcp_recvspace);
1059                 if (error)
1060                         return (error);
1061         }
1062         error = in_pcballoc(so, &tcbinfo, p);
1063         if (error)
1064                 return (error);
1065         inp = sotoinpcb(so);
1066 #if INET6
1067         if (isipv6) {
1068                 inp->inp_vflag |= INP_IPV6;
1069                 inp->in6p_hops = -1;    /* use kernel default */
1070         }
1071         else
1072 #endif /* INET6 */
1073         inp->inp_vflag |= INP_IPV4;
1074         tp = tcp_newtcpcb(inp);
1075         if (tp == 0) {
1076                 int nofd = so->so_state & SS_NOFDREF;   /* XXX */
1077
1078                 so->so_state &= ~SS_NOFDREF;    /* don't free the socket yet */
1079 #if INET6
1080                 if (isipv6)
1081                         in6_pcbdetach(inp);
1082                 else
1083 #endif /* INET6 */
1084                 in_pcbdetach(inp);
1085                 so->so_state |= nofd;
1086                 return (ENOBUFS);
1087         }
1088         tp->t_state = TCPS_CLOSED;
1089         return (0);
1090 }
1091
1092 /*
1093  * Initiate (or continue) disconnect.
1094  * If embryonic state, just send reset (once).
1095  * If in ``let data drain'' option and linger null, just drop.
1096  * Otherwise (hard), mark socket disconnecting and drop
1097  * current input data; switch states based on user close, and
1098  * send segment to peer (with FIN).
1099  */
1100 static struct tcpcb *
1101 tcp_disconnect(tp)
1102         register struct tcpcb *tp;
1103 {
1104         struct socket *so = tp->t_inpcb->inp_socket;
1105
1106         if (tp->t_state < TCPS_ESTABLISHED)
1107                 tp = tcp_close(tp);
1108         else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
1109                 tp = tcp_drop(tp, 0);
1110         else {
1111                 soisdisconnecting(so);
1112                 sbflush(&so->so_rcv);
1113                 tp = tcp_usrclosed(tp);
1114                 if (tp)
1115                         (void) tcp_output(tp);
1116         }
1117         return (tp);
1118 }
1119
1120 /*
1121  * User issued close, and wish to trail through shutdown states:
1122  * if never received SYN, just forget it.  If got a SYN from peer,
1123  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1124  * If already got a FIN from peer, then almost done; go to LAST_ACK
1125  * state.  In all other cases, have already sent FIN to peer (e.g.
1126  * after PRU_SHUTDOWN), and just have to play tedious game waiting
1127  * for peer to send FIN or not respond to keep-alives, etc.
1128  * We can let the user exit from the close as soon as the FIN is acked.
1129  */
1130 static struct tcpcb *
1131 tcp_usrclosed(tp)
1132         register struct tcpcb *tp;
1133 {
1134
1135         switch (tp->t_state) {
1136
1137         case TCPS_CLOSED:
1138         case TCPS_LISTEN:
1139                 tp->t_state = TCPS_CLOSED;
1140                 tp = tcp_close(tp);
1141                 break;
1142
1143         case TCPS_SYN_SENT:
1144         case TCPS_SYN_RECEIVED:
1145                 tp->t_flags |= TF_NEEDFIN;
1146                 break;
1147
1148         case TCPS_ESTABLISHED:
1149                 tp->t_state = TCPS_FIN_WAIT_1;
1150                 break;
1151
1152         case TCPS_CLOSE_WAIT:
1153                 tp->t_state = TCPS_LAST_ACK;
1154                 break;
1155         }
1156         if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
1157                 soisdisconnected(tp->t_inpcb->inp_socket);
1158                 /* To prevent the connection hanging in FIN_WAIT_2 forever. */
1159                 if (tp->t_state == TCPS_FIN_WAIT_2)
1160                         tp->t_timer[TCPT_2MSL] = tcp_maxidle;
1161         }
1162         return (tp);
1163 }
1164