1 /* 2 * Copyright (c) 1982 Regents of the University of California. 3 * All rights reserved. The Berkeley software License Agreement 4 * specifies the terms and conditions for redistribution. 5 * 6 * @(#)tcp_input.c 6.16 (Berkeley) 10/23/85 7 */ 8 9 #include "param.h" 10 #include "systm.h" 11 #include "mbuf.h" 12 #include "protosw.h" 13 #include "socket.h" 14 #include "socketvar.h" 15 #include "errno.h" 16 17 #include "../net/if.h" 18 #include "../net/route.h" 19 20 #include "in.h" 21 #include "in_pcb.h" 22 #include "in_systm.h" 23 #include "ip.h" 24 #include "ip_var.h" 25 #include "tcp.h" 26 #include "tcp_fsm.h" 27 #include "tcp_seq.h" 28 #include "tcp_timer.h" 29 #include "tcp_var.h" 30 #include "tcpip.h" 31 #include "tcp_debug.h" 32 33 int tcpprintfs = 0; 34 int tcpcksum = 1; 35 struct tcpiphdr tcp_saveti; 36 extern tcpnodelack; 37 38 struct tcpcb *tcp_newtcpcb(); 39 40 /* 41 * Insert segment ti into reassembly queue of tcp with 42 * control block tp. Return TH_FIN if reassembly now includes 43 * a segment with FIN. The macro form does the common case inline 44 * (segment is the next to be received on an established connection, 45 * and the queue is empty), avoiding linkage into and removal 46 * from the queue and repetition of various conversions. 47 */ 48 #define TCP_REASS(tp, ti, m, so, flags) { \ 49 if ((ti)->ti_seq == (tp)->rcv_nxt && \ 50 (tp)->seg_next == (struct tcpiphdr *)(tp) && \ 51 (tp)->t_state == TCPS_ESTABLISHED) { \ 52 (tp)->rcv_nxt += (ti)->ti_len; \ 53 flags = (ti)->ti_flags & TH_FIN; \ 54 sbappend(&(so)->so_rcv, (m)); \ 55 sorwakeup(so); \ 56 } else \ 57 (flags) = tcp_reass((tp), (ti)); \ 58 } 59 60 tcp_reass(tp, ti) 61 register struct tcpcb *tp; 62 register struct tcpiphdr *ti; 63 { 64 register struct tcpiphdr *q; 65 struct socket *so = tp->t_inpcb->inp_socket; 66 struct mbuf *m; 67 int flags; 68 69 /* 70 * Call with ti==0 after become established to 71 * force pre-ESTABLISHED data up to user socket. 72 */ 73 if (ti == 0) 74 goto present; 75 76 /* 77 * Find a segment which begins after this one does. 78 */ 79 for (q = tp->seg_next; q != (struct tcpiphdr *)tp; 80 q = (struct tcpiphdr *)q->ti_next) 81 if (SEQ_GT(q->ti_seq, ti->ti_seq)) 82 break; 83 84 /* 85 * If there is a preceding segment, it may provide some of 86 * our data already. If so, drop the data from the incoming 87 * segment. If it provides all of our data, drop us. 88 */ 89 if ((struct tcpiphdr *)q->ti_prev != (struct tcpiphdr *)tp) { 90 register int i; 91 q = (struct tcpiphdr *)q->ti_prev; 92 /* conversion to int (in i) handles seq wraparound */ 93 i = q->ti_seq + q->ti_len - ti->ti_seq; 94 if (i > 0) { 95 if (i >= ti->ti_len) 96 goto drop; 97 m_adj(dtom(ti), i); 98 ti->ti_len -= i; 99 ti->ti_seq += i; 100 } 101 q = (struct tcpiphdr *)(q->ti_next); 102 } 103 104 /* 105 * While we overlap succeeding segments trim them or, 106 * if they are completely covered, dequeue them. 107 */ 108 while (q != (struct tcpiphdr *)tp) { 109 register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq; 110 if (i <= 0) 111 break; 112 if (i < q->ti_len) { 113 q->ti_seq += i; 114 q->ti_len -= i; 115 m_adj(dtom(q), i); 116 break; 117 } 118 q = (struct tcpiphdr *)q->ti_next; 119 m = dtom(q->ti_prev); 120 remque(q->ti_prev); 121 m_freem(m); 122 } 123 124 /* 125 * Stick new segment in its place. 126 */ 127 insque(ti, q->ti_prev); 128 129 present: 130 /* 131 * Present data to user, advancing rcv_nxt through 132 * completed sequence space. 133 */ 134 if (TCPS_HAVERCVDSYN(tp->t_state) == 0) 135 return (0); 136 ti = tp->seg_next; 137 if (ti == (struct tcpiphdr *)tp || ti->ti_seq != tp->rcv_nxt) 138 return (0); 139 if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len) 140 return (0); 141 do { 142 tp->rcv_nxt += ti->ti_len; 143 flags = ti->ti_flags & TH_FIN; 144 remque(ti); 145 m = dtom(ti); 146 ti = (struct tcpiphdr *)ti->ti_next; 147 if (so->so_state & SS_CANTRCVMORE) 148 m_freem(m); 149 else 150 sbappend(&so->so_rcv, m); 151 } while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt); 152 sorwakeup(so); 153 return (flags); 154 drop: 155 m_freem(dtom(ti)); 156 return (0); 157 } 158 159 /* 160 * TCP input routine, follows pages 65-76 of the 161 * protocol specification dated September, 1981 very closely. 162 */ 163 tcp_input(m0) 164 struct mbuf *m0; 165 { 166 register struct tcpiphdr *ti; 167 struct inpcb *inp; 168 register struct mbuf *m; 169 struct mbuf *om = 0; 170 int len, tlen, off; 171 register struct tcpcb *tp = 0; 172 register int tiflags; 173 struct socket *so; 174 int todrop, acked; 175 short ostate; 176 struct in_addr laddr; 177 int dropsocket = 0; 178 179 /* 180 * Get IP and TCP header together in first mbuf. 181 * Note: IP leaves IP header in first mbuf. 182 */ 183 m = m0; 184 ti = mtod(m, struct tcpiphdr *); 185 if (((struct ip *)ti)->ip_hl > (sizeof (struct ip) >> 2)) 186 ip_stripoptions((struct ip *)ti, (struct mbuf *)0); 187 if (m->m_off > MMAXOFF || m->m_len < sizeof (struct tcpiphdr)) { 188 if ((m = m_pullup(m, sizeof (struct tcpiphdr))) == 0) { 189 tcpstat.tcps_hdrops++; 190 return; 191 } 192 ti = mtod(m, struct tcpiphdr *); 193 } 194 195 /* 196 * Checksum extended TCP header and data. 197 */ 198 tlen = ((struct ip *)ti)->ip_len; 199 len = sizeof (struct ip) + tlen; 200 if (tcpcksum) { 201 ti->ti_next = ti->ti_prev = 0; 202 ti->ti_x1 = 0; 203 ti->ti_len = (u_short)tlen; 204 ti->ti_len = htons((u_short)ti->ti_len); 205 if (ti->ti_sum = in_cksum(m, len)) { 206 if (tcpprintfs) 207 printf("tcp sum: src %x\n", ti->ti_src); 208 tcpstat.tcps_badsum++; 209 goto drop; 210 } 211 } 212 213 /* 214 * Check that TCP offset makes sense, 215 * pull out TCP options and adjust length. 216 */ 217 off = ti->ti_off << 2; 218 if (off < sizeof (struct tcphdr) || off > tlen) { 219 if (tcpprintfs) 220 printf("tcp off: src %x off %d\n", ti->ti_src, off); 221 tcpstat.tcps_badoff++; 222 goto drop; 223 } 224 tlen -= off; 225 ti->ti_len = tlen; 226 if (off > sizeof (struct tcphdr)) { 227 if (m->m_len < sizeof(struct ip) + off) { 228 if ((m = m_pullup(m, sizeof (struct ip) + off)) == 0) { 229 tcpstat.tcps_hdrops++; 230 return; 231 } 232 ti = mtod(m, struct tcpiphdr *); 233 } 234 om = m_get(M_DONTWAIT, MT_DATA); 235 if (om == 0) 236 goto drop; 237 om->m_len = off - sizeof (struct tcphdr); 238 { caddr_t op = mtod(m, caddr_t) + sizeof (struct tcpiphdr); 239 bcopy(op, mtod(om, caddr_t), (unsigned)om->m_len); 240 m->m_len -= om->m_len; 241 bcopy(op+om->m_len, op, 242 (unsigned)(m->m_len-sizeof (struct tcpiphdr))); 243 } 244 } 245 tiflags = ti->ti_flags; 246 247 /* 248 * Drop TCP and IP headers. 249 */ 250 off += sizeof (struct ip); 251 m->m_off += off; 252 m->m_len -= off; 253 254 /* 255 * Convert TCP protocol specific fields to host format. 256 */ 257 ti->ti_seq = ntohl(ti->ti_seq); 258 ti->ti_ack = ntohl(ti->ti_ack); 259 ti->ti_win = ntohs(ti->ti_win); 260 ti->ti_urp = ntohs(ti->ti_urp); 261 262 /* 263 * Locate pcb for segment. 264 */ 265 inp = in_pcblookup 266 (&tcb, ti->ti_src, ti->ti_sport, ti->ti_dst, ti->ti_dport, 267 INPLOOKUP_WILDCARD); 268 269 /* 270 * If the state is CLOSED (i.e., TCB does not exist) then 271 * all data in the incoming segment is discarded. 272 */ 273 if (inp == 0) 274 goto dropwithreset; 275 tp = intotcpcb(inp); 276 if (tp == 0) 277 goto dropwithreset; 278 so = inp->inp_socket; 279 if (so->so_options & SO_DEBUG) { 280 ostate = tp->t_state; 281 tcp_saveti = *ti; 282 } 283 if (so->so_options & SO_ACCEPTCONN) { 284 so = sonewconn(so); 285 if (so == 0) 286 goto drop; 287 /* 288 * This is ugly, but .... 289 * 290 * Mark socket as temporary until we're 291 * committed to keeping it. The code at 292 * ``drop'' and ``dropwithreset'' check the 293 * flag dropsocket to see if the temporary 294 * socket created here should be discarded. 295 * We mark the socket as discardable until 296 * we're committed to it below in TCPS_LISTEN. 297 */ 298 dropsocket++; 299 inp = (struct inpcb *)so->so_pcb; 300 inp->inp_laddr = ti->ti_dst; 301 inp->inp_lport = ti->ti_dport; 302 inp->inp_options = ip_srcroute(); 303 tp = intotcpcb(inp); 304 tp->t_state = TCPS_LISTEN; 305 } 306 307 /* 308 * Segment received on connection. 309 * Reset idle time and keep-alive timer. 310 */ 311 tp->t_idle = 0; 312 tp->t_timer[TCPT_KEEP] = TCPTV_KEEP; 313 314 /* 315 * Process options if not in LISTEN state, 316 * else do it below (after getting remote address). 317 */ 318 if (om && tp->t_state != TCPS_LISTEN) { 319 tcp_dooptions(tp, om, ti); 320 om = 0; 321 } 322 323 /* 324 * Calculate amount of space in receive window, 325 * and then do TCP input processing. 326 * Receive window is amount of space in rcv queue, 327 * but not less than advertised window. 328 */ 329 tp->rcv_wnd = sbspace(&so->so_rcv); 330 if (tp->rcv_wnd < 0) 331 tp->rcv_wnd = 0; 332 tp->rcv_wnd = MAX(tp->rcv_wnd, (short)(tp->rcv_adv - tp->rcv_nxt)); 333 334 switch (tp->t_state) { 335 336 /* 337 * If the state is LISTEN then ignore segment if it contains an RST. 338 * If the segment contains an ACK then it is bad and send a RST. 339 * If it does not contain a SYN then it is not interesting; drop it. 340 * Don't bother responding if the destination was a broadcast. 341 * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial 342 * tp->iss, and send a segment: 343 * <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK> 344 * Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss. 345 * Fill in remote peer address fields if not previously specified. 346 * Enter SYN_RECEIVED state, and process any other fields of this 347 * segment in this state. 348 */ 349 case TCPS_LISTEN: { 350 struct mbuf *am; 351 register struct sockaddr_in *sin; 352 353 if (tiflags & TH_RST) 354 goto drop; 355 if (tiflags & TH_ACK) 356 goto dropwithreset; 357 if ((tiflags & TH_SYN) == 0) 358 goto drop; 359 if (in_broadcast(ti->ti_dst)) 360 goto drop; 361 am = m_get(M_DONTWAIT, MT_SONAME); 362 if (am == NULL) 363 goto drop; 364 am->m_len = sizeof (struct sockaddr_in); 365 sin = mtod(am, struct sockaddr_in *); 366 sin->sin_family = AF_INET; 367 sin->sin_addr = ti->ti_src; 368 sin->sin_port = ti->ti_sport; 369 laddr = inp->inp_laddr; 370 if (inp->inp_laddr.s_addr == INADDR_ANY) 371 inp->inp_laddr = ti->ti_dst; 372 if (in_pcbconnect(inp, am)) { 373 inp->inp_laddr = laddr; 374 (void) m_free(am); 375 goto drop; 376 } 377 (void) m_free(am); 378 tp->t_template = tcp_template(tp); 379 if (tp->t_template == 0) { 380 in_pcbdisconnect(inp); 381 dropsocket = 0; /* socket is already gone */ 382 tp = 0; 383 goto drop; 384 } 385 if (om) { 386 tcp_dooptions(tp, om, ti); 387 om = 0; 388 } 389 tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2; 390 tp->irs = ti->ti_seq; 391 tcp_sendseqinit(tp); 392 tcp_rcvseqinit(tp); 393 tp->t_state = TCPS_SYN_RECEIVED; 394 tp->t_timer[TCPT_KEEP] = TCPTV_KEEP; 395 dropsocket = 0; /* committed to socket */ 396 goto trimthenstep6; 397 } 398 399 /* 400 * If the state is SYN_SENT: 401 * if seg contains an ACK, but not for our SYN, drop the input. 402 * if seg contains a RST, then drop the connection. 403 * if seg does not contain SYN, then drop it. 404 * Otherwise this is an acceptable SYN segment 405 * initialize tp->rcv_nxt and tp->irs 406 * if seg contains ack then advance tp->snd_una 407 * if SYN has been acked change to ESTABLISHED else SYN_RCVD state 408 * arrange for segment to be acked (eventually) 409 * continue processing rest of data/controls, beginning with URG 410 */ 411 case TCPS_SYN_SENT: 412 if ((tiflags & TH_ACK) && 413 (SEQ_LEQ(ti->ti_ack, tp->iss) || 414 SEQ_GT(ti->ti_ack, tp->snd_max))) 415 goto dropwithreset; 416 if (tiflags & TH_RST) { 417 if (tiflags & TH_ACK) 418 tp = tcp_drop(tp, ECONNREFUSED); 419 goto drop; 420 } 421 if ((tiflags & TH_SYN) == 0) 422 goto drop; 423 tp->snd_una = ti->ti_ack; 424 if (SEQ_LT(tp->snd_nxt, tp->snd_una)) 425 tp->snd_nxt = tp->snd_una; 426 tp->t_timer[TCPT_REXMT] = 0; 427 tp->irs = ti->ti_seq; 428 tcp_rcvseqinit(tp); 429 tp->t_flags |= TF_ACKNOW; 430 if (SEQ_GT(tp->snd_una, tp->iss)) { 431 soisconnected(so); 432 tp->t_state = TCPS_ESTABLISHED; 433 tp->t_maxseg = MIN(tp->t_maxseg, tcp_mss(tp)); 434 (void) tcp_reass(tp, (struct tcpiphdr *)0); 435 } else 436 tp->t_state = TCPS_SYN_RECEIVED; 437 goto trimthenstep6; 438 439 trimthenstep6: 440 /* 441 * Advance ti->ti_seq to correspond to first data byte. 442 * If data, trim to stay within window, 443 * dropping FIN if necessary. 444 */ 445 ti->ti_seq++; 446 if (ti->ti_len > tp->rcv_wnd) { 447 todrop = ti->ti_len - tp->rcv_wnd; 448 m_adj(m, -todrop); 449 ti->ti_len = tp->rcv_wnd; 450 ti->ti_flags &= ~TH_FIN; 451 } 452 tp->snd_wl1 = ti->ti_seq - 1; 453 goto step6; 454 } 455 456 /* 457 * If data is received on a connection after the 458 * user processes are gone, then RST the other end. 459 */ 460 if ((so->so_state & SS_NOFDREF) && tp->t_state > TCPS_CLOSE_WAIT && 461 ti->ti_len) { 462 tp = tcp_close(tp); 463 goto dropwithreset; 464 } 465 466 /* 467 * States other than LISTEN or SYN_SENT. 468 * First check that at least some bytes of segment are within 469 * receive window. 470 */ 471 if (tp->rcv_wnd == 0) { 472 /* 473 * If window is closed can only take segments at 474 * window edge, and have to drop data and PUSH from 475 * incoming segments. 476 */ 477 if (tp->rcv_nxt != ti->ti_seq) 478 goto dropafterack; 479 if (ti->ti_len > 0) { 480 m_adj(m, ti->ti_len); 481 ti->ti_len = 0; 482 ti->ti_flags &= ~(TH_PUSH|TH_FIN); 483 } 484 } else { 485 /* 486 * If segment begins before rcv_nxt, drop leading 487 * data (and SYN); if nothing left, just ack. 488 */ 489 todrop = tp->rcv_nxt - ti->ti_seq; 490 if (todrop > 0) { 491 if (tiflags & TH_SYN) { 492 tiflags &= ~TH_SYN; 493 ti->ti_flags &= ~TH_SYN; 494 ti->ti_seq++; 495 if (ti->ti_urp > 1) 496 ti->ti_urp--; 497 else 498 tiflags &= ~TH_URG; 499 todrop--; 500 } 501 if (todrop > ti->ti_len || 502 todrop == ti->ti_len && (tiflags&TH_FIN) == 0) 503 goto dropafterack; 504 m_adj(m, todrop); 505 ti->ti_seq += todrop; 506 ti->ti_len -= todrop; 507 if (ti->ti_urp > todrop) 508 ti->ti_urp -= todrop; 509 else { 510 tiflags &= ~TH_URG; 511 ti->ti_flags &= ~TH_URG; 512 ti->ti_urp = 0; 513 } 514 } 515 /* 516 * If segment ends after window, drop trailing data 517 * (and PUSH and FIN); if nothing left, just ACK. 518 */ 519 todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd); 520 if (todrop > 0) { 521 if (todrop >= ti->ti_len) 522 goto dropafterack; 523 m_adj(m, -todrop); 524 ti->ti_len -= todrop; 525 ti->ti_flags &= ~(TH_PUSH|TH_FIN); 526 } 527 } 528 529 /* 530 * If the RST bit is set examine the state: 531 * SYN_RECEIVED STATE: 532 * If passive open, return to LISTEN state. 533 * If active open, inform user that connection was refused. 534 * ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES: 535 * Inform user that connection was reset, and close tcb. 536 * CLOSING, LAST_ACK, TIME_WAIT STATES 537 * Close the tcb. 538 */ 539 if (tiflags&TH_RST) switch (tp->t_state) { 540 541 case TCPS_SYN_RECEIVED: 542 tp = tcp_drop(tp, ECONNREFUSED); 543 goto drop; 544 545 case TCPS_ESTABLISHED: 546 case TCPS_FIN_WAIT_1: 547 case TCPS_FIN_WAIT_2: 548 case TCPS_CLOSE_WAIT: 549 tp = tcp_drop(tp, ECONNRESET); 550 goto drop; 551 552 case TCPS_CLOSING: 553 case TCPS_LAST_ACK: 554 case TCPS_TIME_WAIT: 555 tp = tcp_close(tp); 556 goto drop; 557 } 558 559 /* 560 * If a SYN is in the window, then this is an 561 * error and we send an RST and drop the connection. 562 */ 563 if (tiflags & TH_SYN) { 564 tp = tcp_drop(tp, ECONNRESET); 565 goto dropwithreset; 566 } 567 568 /* 569 * If the ACK bit is off we drop the segment and return. 570 */ 571 if ((tiflags & TH_ACK) == 0) 572 goto drop; 573 574 /* 575 * Ack processing. 576 */ 577 switch (tp->t_state) { 578 579 /* 580 * In SYN_RECEIVED state if the ack ACKs our SYN then enter 581 * ESTABLISHED state and continue processing, othewise 582 * send an RST. 583 */ 584 case TCPS_SYN_RECEIVED: 585 if (SEQ_GT(tp->snd_una, ti->ti_ack) || 586 SEQ_GT(ti->ti_ack, tp->snd_max)) 587 goto dropwithreset; 588 tp->snd_una++; /* SYN acked */ 589 if (SEQ_LT(tp->snd_nxt, tp->snd_una)) 590 tp->snd_nxt = tp->snd_una; 591 tp->t_timer[TCPT_REXMT] = 0; 592 soisconnected(so); 593 tp->t_state = TCPS_ESTABLISHED; 594 tp->t_maxseg = MIN(tp->t_maxseg, tcp_mss(tp)); 595 (void) tcp_reass(tp, (struct tcpiphdr *)0); 596 tp->snd_wl1 = ti->ti_seq - 1; 597 /* fall into ... */ 598 599 /* 600 * In ESTABLISHED state: drop duplicate ACKs; ACK out of range 601 * ACKs. If the ack is in the range 602 * tp->snd_una < ti->ti_ack <= tp->snd_max 603 * then advance tp->snd_una to ti->ti_ack and drop 604 * data from the retransmission queue. If this ACK reflects 605 * more up to date window information we update our window information. 606 */ 607 case TCPS_ESTABLISHED: 608 case TCPS_FIN_WAIT_1: 609 case TCPS_FIN_WAIT_2: 610 case TCPS_CLOSE_WAIT: 611 case TCPS_CLOSING: 612 case TCPS_LAST_ACK: 613 case TCPS_TIME_WAIT: 614 #define ourfinisacked (acked > 0) 615 616 if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) 617 break; 618 if (SEQ_GT(ti->ti_ack, tp->snd_max)) 619 goto dropafterack; 620 acked = ti->ti_ack - tp->snd_una; 621 622 /* 623 * If transmit timer is running and timed sequence 624 * number was acked, update smoothed round trip time. 625 */ 626 if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq)) { 627 if (tp->t_srtt == 0) 628 tp->t_srtt = tp->t_rtt; 629 else 630 tp->t_srtt = 631 tcp_alpha * tp->t_srtt + 632 (1 - tcp_alpha) * tp->t_rtt; 633 tp->t_rtt = 0; 634 } 635 636 if (ti->ti_ack == tp->snd_max) 637 tp->t_timer[TCPT_REXMT] = 0; 638 else { 639 TCPT_RANGESET(tp->t_timer[TCPT_REXMT], 640 tcp_beta * tp->t_srtt, TCPTV_MIN, TCPTV_MAX); 641 tp->t_rxtshift = 0; 642 } 643 /* 644 * When new data is acked, open the congestion window a bit. 645 */ 646 if (acked > 0) 647 tp->snd_cwnd = MIN(11 * tp->snd_cwnd / 10, 65535); 648 if (acked > so->so_snd.sb_cc) { 649 tp->snd_wnd -= so->so_snd.sb_cc; 650 sbdrop(&so->so_snd, so->so_snd.sb_cc); 651 } else { 652 sbdrop(&so->so_snd, acked); 653 tp->snd_wnd -= acked; 654 acked = 0; 655 } 656 if ((so->so_snd.sb_flags & SB_WAIT) || so->so_snd.sb_sel) 657 sowwakeup(so); 658 tp->snd_una = ti->ti_ack; 659 if (SEQ_LT(tp->snd_nxt, tp->snd_una)) 660 tp->snd_nxt = tp->snd_una; 661 662 switch (tp->t_state) { 663 664 /* 665 * In FIN_WAIT_1 STATE in addition to the processing 666 * for the ESTABLISHED state if our FIN is now acknowledged 667 * then enter FIN_WAIT_2. 668 */ 669 case TCPS_FIN_WAIT_1: 670 if (ourfinisacked) { 671 /* 672 * If we can't receive any more 673 * data, then closing user can proceed. 674 * Starting the timer is contrary to the 675 * specification, but if we don't get a FIN 676 * we'll hang forever. 677 */ 678 if (so->so_state & SS_CANTRCVMORE) { 679 soisdisconnected(so); 680 tp->t_timer[TCPT_2MSL] = TCPTV_MAXIDLE; 681 } 682 tp->t_state = TCPS_FIN_WAIT_2; 683 } 684 break; 685 686 /* 687 * In CLOSING STATE in addition to the processing for 688 * the ESTABLISHED state if the ACK acknowledges our FIN 689 * then enter the TIME-WAIT state, otherwise ignore 690 * the segment. 691 */ 692 case TCPS_CLOSING: 693 if (ourfinisacked) { 694 tp->t_state = TCPS_TIME_WAIT; 695 tcp_canceltimers(tp); 696 tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL; 697 soisdisconnected(so); 698 } 699 break; 700 701 /* 702 * The only thing that can arrive in LAST_ACK state 703 * is an acknowledgment of our FIN. If our FIN is now 704 * acknowledged, delete the TCB, enter the closed state 705 * and return. 706 */ 707 case TCPS_LAST_ACK: 708 if (ourfinisacked) 709 tp = tcp_close(tp); 710 goto drop; 711 712 /* 713 * In TIME_WAIT state the only thing that should arrive 714 * is a retransmission of the remote FIN. Acknowledge 715 * it and restart the finack timer. 716 */ 717 case TCPS_TIME_WAIT: 718 tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL; 719 goto dropafterack; 720 } 721 #undef ourfinisacked 722 } 723 724 step6: 725 /* 726 * Update window information. 727 */ 728 if (SEQ_LT(tp->snd_wl1, ti->ti_seq) || tp->snd_wl1 == ti->ti_seq && 729 (SEQ_LT(tp->snd_wl2, ti->ti_ack) || 730 tp->snd_wl2 == ti->ti_ack && ti->ti_win > tp->snd_wnd)) { 731 tp->snd_wnd = ti->ti_win; 732 tp->snd_wl1 = ti->ti_seq; 733 tp->snd_wl2 = ti->ti_ack; 734 if (tp->snd_wnd > tp->max_sndwnd) 735 tp->max_sndwnd = tp->snd_wnd; 736 } 737 738 /* 739 * Process segments with URG. 740 */ 741 if ((tiflags & TH_URG) && ti->ti_urp && 742 TCPS_HAVERCVDFIN(tp->t_state) == 0) { 743 /* 744 * This is a kludge, but if we receive accept 745 * random urgent pointers, we'll crash in 746 * soreceive. It's hard to imagine someone 747 * actually wanting to send this much urgent data. 748 */ 749 if (ti->ti_urp + (unsigned) so->so_rcv.sb_cc > 32767) { 750 ti->ti_urp = 0; /* XXX */ 751 tiflags &= ~TH_URG; /* XXX */ 752 ti->ti_flags &= ~TH_URG; /* XXX */ 753 goto badurp; /* XXX */ 754 } 755 /* 756 * If this segment advances the known urgent pointer, 757 * then mark the data stream. This should not happen 758 * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since 759 * a FIN has been received from the remote side. 760 * In these states we ignore the URG. 761 */ 762 if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) { 763 tp->rcv_up = ti->ti_seq + ti->ti_urp; 764 so->so_oobmark = so->so_rcv.sb_cc + 765 (tp->rcv_up - tp->rcv_nxt) - 1; 766 if (so->so_oobmark == 0) 767 so->so_state |= SS_RCVATMARK; 768 sohasoutofband(so); 769 tp->t_oobflags &= ~(TCPOOB_HAVEDATA | TCPOOB_HADDATA); 770 } 771 /* 772 * Remove out of band data so doesn't get presented to user. 773 * This can happen independent of advancing the URG pointer, 774 * but if two URG's are pending at once, some out-of-band 775 * data may creep in... ick. 776 */ 777 if (ti->ti_urp <= ti->ti_len) 778 tcp_pulloutofband(so, ti); 779 } 780 badurp: /* XXX */ 781 782 /* 783 * Process the segment text, merging it into the TCP sequencing queue, 784 * and arranging for acknowledgment of receipt if necessary. 785 * This process logically involves adjusting tp->rcv_wnd as data 786 * is presented to the user (this happens in tcp_usrreq.c, 787 * case PRU_RCVD). If a FIN has already been received on this 788 * connection then we just ignore the text. 789 */ 790 if ((ti->ti_len || (tiflags&TH_FIN)) && 791 TCPS_HAVERCVDFIN(tp->t_state) == 0) { 792 TCP_REASS(tp, ti, m, so, tiflags); 793 if (tcpnodelack == 0) 794 tp->t_flags |= TF_DELACK; 795 else 796 tp->t_flags |= TF_ACKNOW; 797 /* 798 * Note the amount of data that peer has sent into 799 * our window, in order to estimate the sender's 800 * buffer size. 801 */ 802 len = so->so_rcv.sb_hiwat - (tp->rcv_nxt - tp->rcv_adv); 803 if (len > tp->max_rcvd) 804 tp->max_rcvd = len; 805 } else { 806 m_freem(m); 807 tiflags &= ~TH_FIN; 808 } 809 810 /* 811 * If FIN is received ACK the FIN and let the user know 812 * that the connection is closing. 813 */ 814 if (tiflags & TH_FIN) { 815 if (TCPS_HAVERCVDFIN(tp->t_state) == 0) { 816 socantrcvmore(so); 817 tp->t_flags |= TF_ACKNOW; 818 tp->rcv_nxt++; 819 } 820 switch (tp->t_state) { 821 822 /* 823 * In SYN_RECEIVED and ESTABLISHED STATES 824 * enter the CLOSE_WAIT state. 825 */ 826 case TCPS_SYN_RECEIVED: 827 case TCPS_ESTABLISHED: 828 tp->t_state = TCPS_CLOSE_WAIT; 829 break; 830 831 /* 832 * If still in FIN_WAIT_1 STATE FIN has not been acked so 833 * enter the CLOSING state. 834 */ 835 case TCPS_FIN_WAIT_1: 836 tp->t_state = TCPS_CLOSING; 837 break; 838 839 /* 840 * In FIN_WAIT_2 state enter the TIME_WAIT state, 841 * starting the time-wait timer, turning off the other 842 * standard timers. 843 */ 844 case TCPS_FIN_WAIT_2: 845 tp->t_state = TCPS_TIME_WAIT; 846 tcp_canceltimers(tp); 847 tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL; 848 soisdisconnected(so); 849 break; 850 851 /* 852 * In TIME_WAIT state restart the 2 MSL time_wait timer. 853 */ 854 case TCPS_TIME_WAIT: 855 tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL; 856 break; 857 } 858 } 859 if (so->so_options & SO_DEBUG) 860 tcp_trace(TA_INPUT, ostate, tp, &tcp_saveti, 0); 861 862 /* 863 * Return any desired output. 864 */ 865 (void) tcp_output(tp); 866 return; 867 868 dropafterack: 869 /* 870 * Generate an ACK dropping incoming segment if it occupies 871 * sequence space, where the ACK reflects our state. 872 */ 873 if ((tiflags&TH_RST) || 874 tlen == 0 && (tiflags&(TH_SYN|TH_FIN)) == 0) 875 goto drop; 876 if (tp->t_inpcb->inp_socket->so_options & SO_DEBUG) 877 tcp_trace(TA_RESPOND, ostate, tp, &tcp_saveti, 0); 878 tcp_respond(tp, ti, tp->rcv_nxt, tp->snd_nxt, TH_ACK); 879 return; 880 881 dropwithreset: 882 if (om) { 883 (void) m_free(om); 884 om = 0; 885 } 886 /* 887 * Generate a RST, dropping incoming segment. 888 * Make ACK acceptable to originator of segment. 889 * Don't bother to respond if destination was broadcast. 890 */ 891 if ((tiflags & TH_RST) || in_broadcast(ti->ti_dst)) 892 goto drop; 893 if (tiflags & TH_ACK) 894 tcp_respond(tp, ti, (tcp_seq)0, ti->ti_ack, TH_RST); 895 else { 896 if (tiflags & TH_SYN) 897 ti->ti_len++; 898 tcp_respond(tp, ti, ti->ti_seq+ti->ti_len, (tcp_seq)0, 899 TH_RST|TH_ACK); 900 } 901 /* destroy temporarily created socket */ 902 if (dropsocket) 903 (void) soabort(so); 904 return; 905 906 drop: 907 if (om) 908 (void) m_free(om); 909 /* 910 * Drop space held by incoming segment and return. 911 */ 912 if (tp && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) 913 tcp_trace(TA_DROP, ostate, tp, &tcp_saveti, 0); 914 m_freem(m); 915 /* destroy temporarily created socket */ 916 if (dropsocket) 917 (void) soabort(so); 918 return; 919 } 920 921 tcp_dooptions(tp, om, ti) 922 struct tcpcb *tp; 923 struct mbuf *om; 924 struct tcpiphdr *ti; 925 { 926 register u_char *cp; 927 int opt, optlen, cnt; 928 929 cp = mtod(om, u_char *); 930 cnt = om->m_len; 931 for (; cnt > 0; cnt -= optlen, cp += optlen) { 932 opt = cp[0]; 933 if (opt == TCPOPT_EOL) 934 break; 935 if (opt == TCPOPT_NOP) 936 optlen = 1; 937 else { 938 optlen = cp[1]; 939 if (optlen <= 0) 940 break; 941 } 942 switch (opt) { 943 944 default: 945 break; 946 947 case TCPOPT_MAXSEG: 948 if (optlen != 4) 949 continue; 950 if (!(ti->ti_flags & TH_SYN)) 951 continue; 952 tp->t_maxseg = *(u_short *)(cp + 2); 953 tp->t_maxseg = ntohs((u_short)tp->t_maxseg); 954 tp->t_maxseg = MIN(tp->t_maxseg, tcp_mss(tp)); 955 break; 956 } 957 } 958 (void) m_free(om); 959 } 960 961 /* 962 * Pull out of band byte out of a segment so 963 * it doesn't appear in the user's data queue. 964 * It is still reflected in the segment length for 965 * sequencing purposes. 966 */ 967 tcp_pulloutofband(so, ti) 968 struct socket *so; 969 struct tcpiphdr *ti; 970 { 971 register struct mbuf *m; 972 int cnt = ti->ti_urp - 1; 973 974 m = dtom(ti); 975 while (cnt >= 0) { 976 if (m->m_len > cnt) { 977 char *cp = mtod(m, caddr_t) + cnt; 978 struct tcpcb *tp = sototcpcb(so); 979 980 tp->t_iobc = *cp; 981 tp->t_oobflags |= TCPOOB_HAVEDATA; 982 bcopy(cp+1, cp, (unsigned)(m->m_len - cnt - 1)); 983 m->m_len--; 984 return; 985 } 986 cnt -= m->m_len; 987 m = m->m_next; 988 if (m == 0) 989 break; 990 } 991 panic("tcp_pulloutofband"); 992 } 993 994 /* 995 * Determine a reasonable value for maxseg size. 996 * If the route is known, use one that can be handled 997 * on the given interface without forcing IP to fragment. 998 * If bigger than a page (CLBYTES), round down to nearest pagesize 999 * to utilize pagesize mbufs. 1000 * If interface pointer is unavailable, or the destination isn't local, 1001 * use a conservative size (512 or the default IP max size, but no more 1002 * than the mtu of the interface through which we route), 1003 * as we can't discover anything about intervening gateways or networks. 1004 * 1005 * This is ugly, and doesn't belong at this level, but has to happen somehow. 1006 */ 1007 tcp_mss(tp) 1008 register struct tcpcb *tp; 1009 { 1010 struct route *ro; 1011 struct ifnet *ifp; 1012 int mss; 1013 struct inpcb *inp; 1014 1015 inp = tp->t_inpcb; 1016 ro = &inp->inp_route; 1017 if ((ro->ro_rt == (struct rtentry *)0) || 1018 (ifp = ro->ro_rt->rt_ifp) == (struct ifnet *)0) { 1019 /* No route yet, so try to acquire one */ 1020 if (inp->inp_faddr.s_addr != INADDR_ANY) { 1021 ro->ro_dst.sa_family = AF_INET; 1022 ((struct sockaddr_in *) &ro->ro_dst)->sin_addr = 1023 inp->inp_faddr; 1024 rtalloc(ro); 1025 } 1026 if ((ro->ro_rt == 0) || (ifp = ro->ro_rt->rt_ifp) == 0) 1027 return (TCP_MSS); 1028 } 1029 1030 mss = ifp->if_mtu - sizeof(struct tcpiphdr); 1031 #if (CLBYTES & (CLBYTES - 1)) == 0 1032 if (mss > CLBYTES) 1033 mss &= ~(CLBYTES-1); 1034 #else 1035 if (mss > CLBYTES) 1036 mss = mss / CLBYTES * CLBYTES; 1037 #endif 1038 if (in_localaddr(inp->inp_faddr)) 1039 return (mss); 1040 return (MIN(mss, TCP_MSS)); 1041 } 1042