tcp_usrreq.c revision 1817
1/* 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94 34 * $Id$ 35 */ 36 37#include <sys/param.h> 38#include <sys/systm.h> 39#include <sys/malloc.h> 40#include <sys/mbuf.h> 41#include <sys/socket.h> 42#include <sys/socketvar.h> 43#include <sys/protosw.h> 44#include <sys/errno.h> 45#include <sys/stat.h> 46 47#include <net/if.h> 48#include <net/route.h> 49 50#include <netinet/in.h> 51#include <netinet/in_systm.h> 52#include <netinet/ip.h> 53#include <netinet/in_pcb.h> 54#include <netinet/ip_var.h> 55#include <netinet/tcp.h> 56#include <netinet/tcp_fsm.h> 57#include <netinet/tcp_seq.h> 58#include <netinet/tcp_timer.h> 59#include <netinet/tcp_var.h> 60#include <netinet/tcpip.h> 61#include <netinet/tcp_debug.h> 62 63/* 64 * TCP protocol interface to socket abstraction. 65 */ 66extern char *tcpstates[]; 67 68/* 69 * Process a TCP user request for TCP tb. If this is a send request 70 * then m is the mbuf chain of send data. If this is a timer expiration 71 * (called from the software clock routine), then timertype tells which timer. 72 */ 73/*ARGSUSED*/ 74int 75tcp_usrreq(so, req, m, nam, control) 76 struct socket *so; 77 int req; 78 struct mbuf *m, *nam, *control; 79{ 80 register struct inpcb *inp; 81 register struct tcpcb *tp = 0; 82 int s; 83 int error = 0; 84 int ostate; 85 86 if (req == PRU_CONTROL) 87 return (in_control(so, (int)m, (caddr_t)nam, 88 (struct ifnet *)control)); 89 if (control && control->m_len) { 90 m_freem(control); 91 if (m) 92 m_freem(m); 93 return (EINVAL); 94 } 95 96 s = splnet(); 97 inp = sotoinpcb(so); 98 /* 99 * When a TCP is attached to a socket, then there will be 100 * a (struct inpcb) pointed at by the socket, and this 101 * structure will point at a subsidary (struct tcpcb). 102 */ 103 if (inp == 0 && req != PRU_ATTACH) { 104 splx(s); 105 return (EINVAL); /* XXX */ 106 } 107 if (inp) { 108 tp = intotcpcb(inp); 109 /* WHAT IF TP IS 0? */ 110#ifdef KPROF 111 tcp_acounts[tp->t_state][req]++; 112#endif 113 ostate = tp->t_state; 114 } else 115 ostate = 0; 116 switch (req) { 117 118 /* 119 * TCP attaches to socket via PRU_ATTACH, reserving space, 120 * and an internet control block. 121 */ 122 case PRU_ATTACH: 123 if (inp) { 124 error = EISCONN; 125 break; 126 } 127 error = tcp_attach(so); 128 if (error) 129 break; 130 if ((so->so_options & SO_LINGER) && so->so_linger == 0) 131 so->so_linger = TCP_LINGERTIME; 132 tp = sototcpcb(so); 133 break; 134 135 /* 136 * PRU_DETACH detaches the TCP protocol from the socket. 137 * If the protocol state is non-embryonic, then can't 138 * do this directly: have to initiate a PRU_DISCONNECT, 139 * which may finish later; embryonic TCB's can just 140 * be discarded here. 141 */ 142 case PRU_DETACH: 143 if (tp->t_state > TCPS_LISTEN) 144 tp = tcp_disconnect(tp); 145 else 146 tp = tcp_close(tp); 147 break; 148 149 /* 150 * Give the socket an address. 151 */ 152 case PRU_BIND: 153 error = in_pcbbind(inp, nam); 154 if (error) 155 break; 156 break; 157 158 /* 159 * Prepare to accept connections. 160 */ 161 case PRU_LISTEN: 162 if (inp->inp_lport == 0) 163 error = in_pcbbind(inp, (struct mbuf *)0); 164 if (error == 0) 165 tp->t_state = TCPS_LISTEN; 166 break; 167 168 /* 169 * Initiate connection to peer. 170 * Create a template for use in transmissions on this connection. 171 * Enter SYN_SENT state, and mark socket as connecting. 172 * Start keep-alive timer, and seed output sequence space. 173 * Send initial segment on connection. 174 */ 175 case PRU_CONNECT: 176 if (inp->inp_lport == 0) { 177 error = in_pcbbind(inp, (struct mbuf *)0); 178 if (error) 179 break; 180 } 181 error = in_pcbconnect(inp, nam); 182 if (error) 183 break; 184 tp->t_template = tcp_template(tp); 185 if (tp->t_template == 0) { 186 in_pcbdisconnect(inp); 187 error = ENOBUFS; 188 break; 189 } 190 /* Compute window scaling to request. */ 191 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 192 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) 193 tp->request_r_scale++; 194 soisconnecting(so); 195 tcpstat.tcps_connattempt++; 196 tp->t_state = TCPS_SYN_SENT; 197 tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT; 198 tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2; 199 tcp_sendseqinit(tp); 200 error = tcp_output(tp); 201 break; 202 203 /* 204 * Create a TCP connection between two sockets. 205 */ 206 case PRU_CONNECT2: 207 error = EOPNOTSUPP; 208 break; 209 210 /* 211 * Initiate disconnect from peer. 212 * If connection never passed embryonic stage, just drop; 213 * else if don't need to let data drain, then can just drop anyways, 214 * else have to begin TCP shutdown process: mark socket disconnecting, 215 * drain unread data, state switch to reflect user close, and 216 * send segment (e.g. FIN) to peer. Socket will be really disconnected 217 * when peer sends FIN and acks ours. 218 * 219 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. 220 */ 221 case PRU_DISCONNECT: 222 tp = tcp_disconnect(tp); 223 break; 224 225 /* 226 * Accept a connection. Essentially all the work is 227 * done at higher levels; just return the address 228 * of the peer, storing through addr. 229 */ 230 case PRU_ACCEPT: 231 in_setpeeraddr(inp, nam); 232 break; 233 234 /* 235 * Mark the connection as being incapable of further output. 236 */ 237 case PRU_SHUTDOWN: 238 socantsendmore(so); 239 tp = tcp_usrclosed(tp); 240 if (tp) 241 error = tcp_output(tp); 242 break; 243 244 /* 245 * After a receive, possibly send window update to peer. 246 */ 247 case PRU_RCVD: 248 (void) tcp_output(tp); 249 break; 250 251 /* 252 * Do a send by putting data in output queue and updating urgent 253 * marker if URG set. Possibly send more data. 254 */ 255 case PRU_SEND: 256 sbappend(&so->so_snd, m); 257 error = tcp_output(tp); 258 break; 259 260 /* 261 * Abort the TCP. 262 */ 263 case PRU_ABORT: 264 tp = tcp_drop(tp, ECONNABORTED); 265 break; 266 267 case PRU_SENSE: 268 ((struct stat *) m)->st_blksize = so->so_snd.sb_hiwat; 269 (void) splx(s); 270 return (0); 271 272 case PRU_RCVOOB: 273 if ((so->so_oobmark == 0 && 274 (so->so_state & SS_RCVATMARK) == 0) || 275 so->so_options & SO_OOBINLINE || 276 tp->t_oobflags & TCPOOB_HADDATA) { 277 error = EINVAL; 278 break; 279 } 280 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { 281 error = EWOULDBLOCK; 282 break; 283 } 284 m->m_len = 1; 285 *mtod(m, caddr_t) = tp->t_iobc; 286 if (((int)nam & MSG_PEEK) == 0) 287 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); 288 break; 289 290 case PRU_SENDOOB: 291 if (sbspace(&so->so_snd) < -512) { 292 m_freem(m); 293 error = ENOBUFS; 294 break; 295 } 296 /* 297 * According to RFC961 (Assigned Protocols), 298 * the urgent pointer points to the last octet 299 * of urgent data. We continue, however, 300 * to consider it to indicate the first octet 301 * of data past the urgent section. 302 * Otherwise, snd_up should be one lower. 303 */ 304 sbappend(&so->so_snd, m); 305 tp->snd_up = tp->snd_una + so->so_snd.sb_cc; 306 tp->t_force = 1; 307 error = tcp_output(tp); 308 tp->t_force = 0; 309 break; 310 311 case PRU_SOCKADDR: 312 in_setsockaddr(inp, nam); 313 break; 314 315 case PRU_PEERADDR: 316 in_setpeeraddr(inp, nam); 317 break; 318 319 /* 320 * TCP slow timer went off; going through this 321 * routine for tracing's sake. 322 */ 323 case PRU_SLOWTIMO: 324 tp = tcp_timers(tp, (int)nam); 325 req |= (int)nam << 8; /* for debug's sake */ 326 break; 327 328 default: 329 panic("tcp_usrreq"); 330 } 331 if (tp && (so->so_options & SO_DEBUG)) 332 tcp_trace(TA_USER, ostate, tp, (struct tcpiphdr *)0, req); 333 splx(s); 334 return (error); 335} 336 337int 338tcp_ctloutput(op, so, level, optname, mp) 339 int op; 340 struct socket *so; 341 int level, optname; 342 struct mbuf **mp; 343{ 344 int error = 0, s; 345 struct inpcb *inp; 346 register struct tcpcb *tp; 347 register struct mbuf *m; 348 register int i; 349 350 s = splnet(); 351 inp = sotoinpcb(so); 352 if (inp == NULL) { 353 splx(s); 354 if (op == PRCO_SETOPT && *mp) 355 (void) m_free(*mp); 356 return (ECONNRESET); 357 } 358 if (level != IPPROTO_TCP) { 359 error = ip_ctloutput(op, so, level, optname, mp); 360 splx(s); 361 return (error); 362 } 363 tp = intotcpcb(inp); 364 365 switch (op) { 366 367 case PRCO_SETOPT: 368 m = *mp; 369 switch (optname) { 370 371 case TCP_NODELAY: 372 if (m == NULL || m->m_len < sizeof (int)) 373 error = EINVAL; 374 else if (*mtod(m, int *)) 375 tp->t_flags |= TF_NODELAY; 376 else 377 tp->t_flags &= ~TF_NODELAY; 378 break; 379 380 case TCP_MAXSEG: 381 if (m && (i = *mtod(m, int *)) > 0 && i <= tp->t_maxseg) 382 tp->t_maxseg = i; 383 else 384 error = EINVAL; 385 break; 386 387 default: 388 error = ENOPROTOOPT; 389 break; 390 } 391 if (m) 392 (void) m_free(m); 393 break; 394 395 case PRCO_GETOPT: 396 *mp = m = m_get(M_WAIT, MT_SOOPTS); 397 m->m_len = sizeof(int); 398 399 switch (optname) { 400 case TCP_NODELAY: 401 *mtod(m, int *) = tp->t_flags & TF_NODELAY; 402 break; 403 case TCP_MAXSEG: 404 *mtod(m, int *) = tp->t_maxseg; 405 break; 406 default: 407 error = ENOPROTOOPT; 408 break; 409 } 410 break; 411 } 412 splx(s); 413 return (error); 414} 415 416/* 417 * tcp_sendspace and tcp_recvspace are the default send and receive window 418 * sizes, respectively. These are obsolescent (this information should 419 * be set by the route). 420 */ 421#ifdef TCP_SMALLSPACE 422u_long tcp_sendspace = 1024*4; 423u_long tcp_recvspace = 1024*4; 424#else 425u_long tcp_sendspace = 1024*16; 426u_long tcp_recvspace = 1024*16; 427#endif 428 429/* 430 * Attach TCP protocol to socket, allocating 431 * internet protocol control block, tcp control block, 432 * bufer space, and entering LISTEN state if to accept connections. 433 */ 434int 435tcp_attach(so) 436 struct socket *so; 437{ 438 register struct tcpcb *tp; 439 struct inpcb *inp; 440 int error; 441 442 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { 443 error = soreserve(so, tcp_sendspace, tcp_recvspace); 444 if (error) 445 return (error); 446 } 447 error = in_pcballoc(so, &tcb); 448 if (error) 449 return (error); 450 inp = sotoinpcb(so); 451 tp = tcp_newtcpcb(inp); 452 if (tp == 0) { 453 int nofd = so->so_state & SS_NOFDREF; /* XXX */ 454 455 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */ 456 in_pcbdetach(inp); 457 so->so_state |= nofd; 458 return (ENOBUFS); 459 } 460 tp->t_state = TCPS_CLOSED; 461 return (0); 462} 463 464/* 465 * Initiate (or continue) disconnect. 466 * If embryonic state, just send reset (once). 467 * If in ``let data drain'' option and linger null, just drop. 468 * Otherwise (hard), mark socket disconnecting and drop 469 * current input data; switch states based on user close, and 470 * send segment to peer (with FIN). 471 */ 472struct tcpcb * 473tcp_disconnect(tp) 474 register struct tcpcb *tp; 475{ 476 struct socket *so = tp->t_inpcb->inp_socket; 477 478 if (tp->t_state < TCPS_ESTABLISHED) 479 tp = tcp_close(tp); 480 else if ((so->so_options & SO_LINGER) && so->so_linger == 0) 481 tp = tcp_drop(tp, 0); 482 else { 483 soisdisconnecting(so); 484 sbflush(&so->so_rcv); 485 tp = tcp_usrclosed(tp); 486 if (tp) 487 (void) tcp_output(tp); 488 } 489 return (tp); 490} 491 492/* 493 * User issued close, and wish to trail through shutdown states: 494 * if never received SYN, just forget it. If got a SYN from peer, 495 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. 496 * If already got a FIN from peer, then almost done; go to LAST_ACK 497 * state. In all other cases, have already sent FIN to peer (e.g. 498 * after PRU_SHUTDOWN), and just have to play tedious game waiting 499 * for peer to send FIN or not respond to keep-alives, etc. 500 * We can let the user exit from the close as soon as the FIN is acked. 501 */ 502struct tcpcb * 503tcp_usrclosed(tp) 504 register struct tcpcb *tp; 505{ 506 507 switch (tp->t_state) { 508 509 case TCPS_CLOSED: 510 case TCPS_LISTEN: 511 case TCPS_SYN_SENT: 512 tp->t_state = TCPS_CLOSED; 513 tp = tcp_close(tp); 514 break; 515 516 case TCPS_SYN_RECEIVED: 517 case TCPS_ESTABLISHED: 518 tp->t_state = TCPS_FIN_WAIT_1; 519 break; 520 521 case TCPS_CLOSE_WAIT: 522 tp->t_state = TCPS_LAST_ACK; 523 break; 524 } 525 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) 526 soisdisconnected(tp->t_inpcb->inp_socket); 527 return (tp); 528} 529