tcp_output.c revision 1.63.2.2
1 /* $NetBSD: tcp_output.c,v 1.63.2.2 2001/06/21 20:08:43 nathanw Exp $ */ 2 3 /* 4 %%% portions-copyright-nrl-95 5 Portions of this software are Copyright 1995-1998 by Randall Atkinson, 6 Ronald Lee, Daniel McDonald, Bao Phan, and Chris Winters. All Rights 7 Reserved. All rights under this copyright have been assigned to the US 8 Naval Research Laboratory (NRL). The NRL Copyright Notice and License 9 Agreement Version 1.1 (January 17, 1995) applies to these portions of the 10 software. 11 You should have received a copy of the license with this software. If you 12 didn't get a copy, you may request one from <license (at) ipv6.nrl.navy.mil>. 13 14 */ 15 16 /* 17 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 18 * All rights reserved. 19 * 20 * Redistribution and use in source and binary forms, with or without 21 * modification, are permitted provided that the following conditions 22 * are met: 23 * 1. Redistributions of source code must retain the above copyright 24 * notice, this list of conditions and the following disclaimer. 25 * 2. Redistributions in binary form must reproduce the above copyright 26 * notice, this list of conditions and the following disclaimer in the 27 * documentation and/or other materials provided with the distribution. 28 * 3. Neither the name of the project nor the names of its contributors 29 * may be used to endorse or promote products derived from this software 30 * without specific prior written permission. 31 * 32 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 33 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 34 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 35 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 36 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 37 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 38 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 39 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 40 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 41 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 42 * SUCH DAMAGE. 43 */ 44 45 /*- 46 * Copyright (c) 1997, 1998, 2001 The NetBSD Foundation, Inc. 47 * All rights reserved. 48 * 49 * This code is derived from software contributed to The NetBSD Foundation 50 * by Jason R. Thorpe and Kevin M. Lahey of the Numerical Aerospace Simulation 51 * Facility, NASA Ames Research Center. 52 * 53 * Redistribution and use in source and binary forms, with or without 54 * modification, are permitted provided that the following conditions 55 * are met: 56 * 1. Redistributions of source code must retain the above copyright 57 * notice, this list of conditions and the following disclaimer. 58 * 2. Redistributions in binary form must reproduce the above copyright 59 * notice, this list of conditions and the following disclaimer in the 60 * documentation and/or other materials provided with the distribution. 61 * 3. All advertising materials mentioning features or use of this software 62 * must display the following acknowledgement: 63 * This product includes software developed by the NetBSD 64 * Foundation, Inc. and its contributors. 65 * 4. Neither the name of The NetBSD Foundation nor the names of its 66 * contributors may be used to endorse or promote products derived 67 * from this software without specific prior written permission. 68 * 69 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 70 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 71 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 72 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 73 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 74 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 75 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 76 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 77 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 78 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 79 * POSSIBILITY OF SUCH DAMAGE. 80 */ 81 82 /* 83 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 84 * The Regents of the University of California. All rights reserved. 85 * 86 * Redistribution and use in source and binary forms, with or without 87 * modification, are permitted provided that the following conditions 88 * are met: 89 * 1. Redistributions of source code must retain the above copyright 90 * notice, this list of conditions and the following disclaimer. 91 * 2. Redistributions in binary form must reproduce the above copyright 92 * notice, this list of conditions and the following disclaimer in the 93 * documentation and/or other materials provided with the distribution. 94 * 3. All advertising materials mentioning features or use of this software 95 * must display the following acknowledgement: 96 * This product includes software developed by the University of 97 * California, Berkeley and its contributors. 98 * 4. Neither the name of the University nor the names of its contributors 99 * may be used to endorse or promote products derived from this software 100 * without specific prior written permission. 101 * 102 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 103 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 104 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 105 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 106 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 107 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 108 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 109 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 110 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 111 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 112 * SUCH DAMAGE. 113 * 114 * @(#)tcp_output.c 8.4 (Berkeley) 5/24/95 115 */ 116 117 #include "opt_inet.h" 118 #include "opt_ipsec.h" 119 120 #include <sys/param.h> 121 #include <sys/systm.h> 122 #include <sys/malloc.h> 123 #include <sys/mbuf.h> 124 #include <sys/protosw.h> 125 #include <sys/socket.h> 126 #include <sys/socketvar.h> 127 #include <sys/errno.h> 128 #include <sys/domain.h> 129 130 #include <net/if.h> 131 #include <net/route.h> 132 133 #include <netinet/in.h> 134 #include <netinet/in_systm.h> 135 #include <netinet/ip.h> 136 #include <netinet/in_pcb.h> 137 #include <netinet/ip_var.h> 138 139 #ifdef INET6 140 #ifndef INET 141 #include <netinet/in.h> 142 #endif 143 #include <netinet/ip6.h> 144 #include <netinet6/in6_pcb.h> 145 #include <netinet6/ip6_var.h> 146 #endif 147 148 #include <netinet/tcp.h> 149 #define TCPOUTFLAGS 150 #include <netinet/tcp_fsm.h> 151 #include <netinet/tcp_seq.h> 152 #include <netinet/tcp_timer.h> 153 #include <netinet/tcp_var.h> 154 #include <netinet/tcpip.h> 155 #include <netinet/tcp_debug.h> 156 157 #ifdef notyet 158 extern struct mbuf *m_copypack(); 159 #endif 160 161 #define MAX_TCPOPTLEN 32 /* max # bytes that go in options */ 162 163 /* 164 * Knob to enable Congestion Window Monitoring, and control the 165 * the burst size it allows. Default burst is 4 packets, per 166 * the Internet draft. 167 */ 168 int tcp_cwm = 0; 169 int tcp_cwm_burstsize = 4; 170 171 static __inline void tcp_segsize __P((struct tcpcb *, int *, int *)); 172 static __inline void 173 tcp_segsize(tp, txsegsizep, rxsegsizep) 174 struct tcpcb *tp; 175 int *txsegsizep, *rxsegsizep; 176 { 177 #ifdef INET 178 struct inpcb *inp = tp->t_inpcb; 179 #endif 180 #ifdef INET6 181 struct in6pcb *in6p = tp->t_in6pcb; 182 #endif 183 struct rtentry *rt; 184 struct ifnet *ifp; 185 int size; 186 int iphlen; 187 188 #ifdef DIAGNOSTIC 189 if (tp->t_inpcb && tp->t_in6pcb) 190 panic("tcp_segsize: both t_inpcb and t_in6pcb are set"); 191 #endif 192 switch (tp->t_family) { 193 #ifdef INET 194 case AF_INET: 195 iphlen = sizeof(struct ip); 196 break; 197 #endif 198 #ifdef INET6 199 case AF_INET6: 200 iphlen = sizeof(struct ip6_hdr); 201 break; 202 #endif 203 default: 204 size = tcp_mssdflt; 205 goto out; 206 } 207 208 rt = NULL; 209 #ifdef INET 210 if (inp) 211 rt = in_pcbrtentry(inp); 212 #endif 213 #ifdef INET6 214 if (in6p) 215 rt = in6_pcbrtentry(in6p); 216 #endif 217 if (rt == NULL) { 218 size = tcp_mssdflt; 219 goto out; 220 } 221 222 ifp = rt->rt_ifp; 223 224 size = tcp_mssdflt; 225 if (rt->rt_rmx.rmx_mtu != 0) 226 size = rt->rt_rmx.rmx_mtu - iphlen - sizeof(struct tcphdr); 227 else if (ifp->if_flags & IFF_LOOPBACK) 228 size = ifp->if_mtu - iphlen - sizeof(struct tcphdr); 229 #ifdef INET 230 else if (inp && ip_mtudisc) 231 size = ifp->if_mtu - iphlen - sizeof(struct tcphdr); 232 else if (inp && in_localaddr(inp->inp_faddr)) 233 size = ifp->if_mtu - iphlen - sizeof(struct tcphdr); 234 #endif 235 #ifdef INET6 236 else if (in6p) { 237 #ifdef INET 238 if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) { 239 /* mapped addr case */ 240 struct in_addr d; 241 bcopy(&in6p->in6p_faddr.s6_addr32[3], &d, sizeof(d)); 242 if (ip_mtudisc || in_localaddr(d)) 243 size = ifp->if_mtu - iphlen - sizeof(struct tcphdr); 244 } else 245 #endif 246 { 247 /* 248 * for IPv6, path MTU discovery is always turned on, 249 * or the node must use packet size <= 1280. 250 */ 251 size = ifp->if_mtu - iphlen - sizeof(struct tcphdr); 252 } 253 } 254 #endif 255 size -= tcp_optlen(tp); 256 /* 257 * XXX tp->t_ourmss should have the right size, but without this code 258 * fragmentation will occur... need more investigation 259 */ 260 #ifdef INET 261 if (inp) { 262 #ifdef IPSEC 263 size -= ipsec4_hdrsiz_tcp(tp); 264 #endif 265 size -= ip_optlen(inp); 266 } 267 #endif 268 #ifdef INET6 269 #ifdef INET 270 if (in6p && tp->t_family == AF_INET) { 271 #ifdef IPSEC 272 size -= ipsec4_hdrsiz_tcp(tp); 273 #endif 274 /* XXX size -= ip_optlen(in6p); */ 275 } else 276 #endif 277 if (in6p && tp->t_family == AF_INET6) { 278 #ifdef IPSEC 279 size -= ipsec6_hdrsiz_tcp(tp); 280 #endif 281 size -= ip6_optlen(in6p); 282 } 283 #endif 284 285 out: 286 /* 287 * *rxsegsizep holds *estimated* inbound segment size (estimation 288 * assumes that path MTU is the same for both ways). this is only 289 * for silly window avoidance, do not use the value for other purposes. 290 * 291 * ipseclen is subtracted from both sides, this may not be right. 292 * I'm not quite sure about this (could someone comment). 293 */ 294 *txsegsizep = min(tp->t_peermss, size); 295 *rxsegsizep = min(tp->t_ourmss, size); 296 297 if (*txsegsizep != tp->t_segsz) { 298 /* 299 * If the new segment size is larger, we don't want to 300 * mess up the congestion window, but if it is smaller 301 * we'll have to reduce the congestion window to ensure 302 * that we don't get into trouble with initial windows 303 * and the rest. In any case, if the segment size 304 * has changed, chances are the path has, too, and 305 * our congestion window will be different. 306 */ 307 if (*txsegsizep < tp->t_segsz) { 308 tp->snd_cwnd = max((tp->snd_cwnd / tp->t_segsz) 309 * *txsegsizep, *txsegsizep); 310 tp->snd_ssthresh = max((tp->snd_ssthresh / tp->t_segsz) 311 * *txsegsizep, *txsegsizep); 312 } 313 tp->t_segsz = *txsegsizep; 314 } 315 } 316 317 /* 318 * Tcp output routine: figure out what should be sent and send it. 319 */ 320 int 321 tcp_output(tp) 322 struct tcpcb *tp; 323 { 324 struct socket *so; 325 struct route *ro; 326 long len, win; 327 int off, flags, error; 328 struct mbuf *m; 329 struct ip *ip; 330 #ifdef INET6 331 struct ip6_hdr *ip6; 332 #endif 333 struct tcphdr *th; 334 u_char opt[MAX_TCPOPTLEN]; 335 unsigned optlen, hdrlen; 336 int idle, sendalot, txsegsize, rxsegsize; 337 int maxburst = TCP_MAXBURST; 338 int af; /* address family on the wire */ 339 int iphdrlen; 340 341 #ifdef DIAGNOSTIC 342 if (tp->t_inpcb && tp->t_in6pcb) 343 panic("tcp_output: both t_inpcb and t_in6pcb are set"); 344 #endif 345 so = NULL; 346 ro = NULL; 347 if (tp->t_inpcb) { 348 so = tp->t_inpcb->inp_socket; 349 ro = &tp->t_inpcb->inp_route; 350 } 351 #ifdef INET6 352 else if (tp->t_in6pcb) { 353 so = tp->t_in6pcb->in6p_socket; 354 ro = (struct route *)&tp->t_in6pcb->in6p_route; 355 } 356 #endif 357 358 switch (af = tp->t_family) { 359 #ifdef INET 360 case AF_INET: 361 if (tp->t_inpcb) 362 break; 363 #ifdef INET6 364 /* mapped addr case */ 365 if (tp->t_in6pcb) 366 break; 367 #endif 368 return EINVAL; 369 #endif 370 #ifdef INET6 371 case AF_INET6: 372 if (tp->t_in6pcb) 373 break; 374 return EINVAL; 375 #endif 376 default: 377 return EAFNOSUPPORT; 378 } 379 380 tcp_segsize(tp, &txsegsize, &rxsegsize); 381 382 idle = (tp->snd_max == tp->snd_una); 383 384 /* 385 * Restart Window computation. From draft-floyd-incr-init-win-03: 386 * 387 * Optionally, a TCP MAY set the restart window to the 388 * minimum of the value used for the initial window and 389 * the current value of cwnd (in other words, using a 390 * larger value for the restart window should never increase 391 * the size of cwnd). 392 */ 393 if (tcp_cwm) { 394 /* 395 * Hughes/Touch/Heidemann Congestion Window Monitoring. 396 * Count the number of packets currently pending 397 * acknowledgement, and limit our congestion window 398 * to a pre-determined allowed burst size plus that count. 399 * This prevents bursting once all pending packets have 400 * been acknowledged (i.e. transmission is idle). 401 * 402 * XXX Link this to Initial Window? 403 */ 404 tp->snd_cwnd = min(tp->snd_cwnd, 405 (tcp_cwm_burstsize * txsegsize) + 406 (tp->snd_nxt - tp->snd_una)); 407 } else { 408 if (idle && tp->t_idle >= tp->t_rxtcur) { 409 /* 410 * We have been idle for "a while" and no acks are 411 * expected to clock out any data we send -- 412 * slow start to get ack "clock" running again. 413 */ 414 tp->snd_cwnd = min(tp->snd_cwnd, 415 TCP_INITIAL_WINDOW(tcp_init_win, txsegsize)); 416 } 417 } 418 419 again: 420 /* 421 * Determine length of data that should be transmitted, and 422 * flags that should be used. If there is some data or critical 423 * controls (SYN, RST) to send, then transmit; otherwise, 424 * investigate further. 425 */ 426 sendalot = 0; 427 off = tp->snd_nxt - tp->snd_una; 428 win = min(tp->snd_wnd, tp->snd_cwnd); 429 430 flags = tcp_outflags[tp->t_state]; 431 /* 432 * If in persist timeout with window of 0, send 1 byte. 433 * Otherwise, if window is small but nonzero 434 * and timer expired, we will send what we can 435 * and go to transmit state. 436 */ 437 if (tp->t_force) { 438 if (win == 0) { 439 /* 440 * If we still have some data to send, then 441 * clear the FIN bit. Usually this would 442 * happen below when it realizes that we 443 * aren't sending all the data. However, 444 * if we have exactly 1 byte of unset data, 445 * then it won't clear the FIN bit below, 446 * and if we are in persist state, we wind 447 * up sending the packet without recording 448 * that we sent the FIN bit. 449 * 450 * We can't just blindly clear the FIN bit, 451 * because if we don't have any more data 452 * to send then the probe will be the FIN 453 * itself. 454 */ 455 if (off < so->so_snd.sb_cc) 456 flags &= ~TH_FIN; 457 win = 1; 458 } else { 459 TCP_TIMER_DISARM(tp, TCPT_PERSIST); 460 tp->t_rxtshift = 0; 461 } 462 } 463 464 if (win < so->so_snd.sb_cc) { 465 len = win - off; 466 flags &= ~TH_FIN; 467 } else 468 len = so->so_snd.sb_cc - off; 469 470 if (len < 0) { 471 /* 472 * If FIN has been sent but not acked, 473 * but we haven't been called to retransmit, 474 * len will be -1. Otherwise, window shrank 475 * after we sent into it. If window shrank to 0, 476 * cancel pending retransmit, pull snd_nxt back 477 * to (closed) window, and set the persist timer 478 * if it isn't already going. If the window didn't 479 * close completely, just wait for an ACK. 480 * 481 * If we have a pending FIN, either it has already been 482 * transmitted or it is outside the window, so drop it. 483 * If the FIN has been transmitted, but this is not a 484 * retransmission, then len must be -1. Therefore we also 485 * prevent here the sending of `gratuitous FINs'. This 486 * eliminates the need to check for that case below (e.g. 487 * to back up snd_nxt before the FIN so that the sequence 488 * number is correct). 489 */ 490 len = 0; 491 flags &= ~TH_FIN; 492 if (win == 0) { 493 TCP_TIMER_DISARM(tp, TCPT_REXMT); 494 tp->t_rxtshift = 0; 495 tp->snd_nxt = tp->snd_una; 496 if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) 497 tcp_setpersist(tp); 498 } 499 } 500 if (len > txsegsize) { 501 len = txsegsize; 502 flags &= ~TH_FIN; 503 sendalot = 1; 504 } 505 506 win = sbspace(&so->so_rcv); 507 508 /* 509 * Sender silly window avoidance. If connection is idle 510 * and can send all data, a maximum segment, 511 * at least a maximum default-size segment do it, 512 * or are forced, do it; otherwise don't bother. 513 * If peer's buffer is tiny, then send 514 * when window is at least half open. 515 * If retransmitting (possibly after persist timer forced us 516 * to send into a small window), then must resend. 517 */ 518 if (len) { 519 if (len == txsegsize) 520 goto send; 521 if ((so->so_state & SS_MORETOCOME) == 0 && 522 ((idle || tp->t_flags & TF_NODELAY) && 523 len + off >= so->so_snd.sb_cc)) 524 goto send; 525 if (tp->t_force) 526 goto send; 527 if (len >= tp->max_sndwnd / 2) 528 goto send; 529 if (SEQ_LT(tp->snd_nxt, tp->snd_max)) 530 goto send; 531 } 532 533 /* 534 * Compare available window to amount of window known to peer 535 * (as advertised window less next expected input). If the 536 * difference is at least twice the size of the largest segment 537 * we expect to receive (i.e. two segments) or at least 50% of 538 * the maximum possible window, then want to send a window update 539 * to peer. 540 */ 541 if (win > 0) { 542 /* 543 * "adv" is the amount we can increase the window, 544 * taking into account that we are limited by 545 * TCP_MAXWIN << tp->rcv_scale. 546 */ 547 long adv = min(win, (long)TCP_MAXWIN << tp->rcv_scale) - 548 (tp->rcv_adv - tp->rcv_nxt); 549 550 if (adv >= (long) (2 * rxsegsize)) 551 goto send; 552 if (2 * adv >= (long) so->so_rcv.sb_hiwat) 553 goto send; 554 } 555 556 /* 557 * Send if we owe peer an ACK. 558 */ 559 if (tp->t_flags & TF_ACKNOW) 560 goto send; 561 if (flags & (TH_SYN|TH_FIN|TH_RST)) 562 goto send; 563 if (SEQ_GT(tp->snd_up, tp->snd_una)) 564 goto send; 565 566 /* 567 * TCP window updates are not reliable, rather a polling protocol 568 * using ``persist'' packets is used to insure receipt of window 569 * updates. The three ``states'' for the output side are: 570 * idle not doing retransmits or persists 571 * persisting to move a small or zero window 572 * (re)transmitting and thereby not persisting 573 * 574 * tp->t_timer[TCPT_PERSIST] 575 * is set when we are in persist state. 576 * tp->t_force 577 * is set when we are called to send a persist packet. 578 * tp->t_timer[TCPT_REXMT] 579 * is set when we are retransmitting 580 * The output side is idle when both timers are zero. 581 * 582 * If send window is too small, there is data to transmit, and no 583 * retransmit or persist is pending, then go to persist state. 584 * If nothing happens soon, send when timer expires: 585 * if window is nonzero, transmit what we can, 586 * otherwise force out a byte. 587 */ 588 if (so->so_snd.sb_cc && TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 && 589 TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) { 590 tp->t_rxtshift = 0; 591 tcp_setpersist(tp); 592 } 593 594 /* 595 * No reason to send a segment, just return. 596 */ 597 return (0); 598 599 send: 600 /* 601 * Before ESTABLISHED, force sending of initial options 602 * unless TCP set not to do any options. 603 * NOTE: we assume that the IP/TCP header plus TCP options 604 * always fit in a single mbuf, leaving room for a maximum 605 * link header, i.e. 606 * max_linkhdr + sizeof (struct tcpiphdr) + optlen <= MCLBYTES 607 */ 608 optlen = 0; 609 switch (af) { 610 #ifdef INET 611 case AF_INET: 612 iphdrlen = sizeof(struct ip) + sizeof(struct tcphdr); 613 break; 614 #endif 615 #ifdef INET6 616 case AF_INET6: 617 iphdrlen = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); 618 break; 619 #endif 620 default: /*pacify gcc*/ 621 iphdrlen = 0; 622 break; 623 } 624 hdrlen = iphdrlen; 625 if (flags & TH_SYN) { 626 struct rtentry *rt; 627 628 rt = NULL; 629 #ifdef INET 630 if (tp->t_inpcb) 631 rt = in_pcbrtentry(tp->t_inpcb); 632 #endif 633 #ifdef INET6 634 if (tp->t_in6pcb) 635 rt = in6_pcbrtentry(tp->t_in6pcb); 636 #endif 637 638 tp->snd_nxt = tp->iss; 639 tp->t_ourmss = tcp_mss_to_advertise(rt != NULL ? 640 rt->rt_ifp : NULL, af); 641 if ((tp->t_flags & TF_NOOPT) == 0) { 642 opt[0] = TCPOPT_MAXSEG; 643 opt[1] = 4; 644 opt[2] = (tp->t_ourmss >> 8) & 0xff; 645 opt[3] = tp->t_ourmss & 0xff; 646 optlen = 4; 647 648 if ((tp->t_flags & TF_REQ_SCALE) && 649 ((flags & TH_ACK) == 0 || 650 (tp->t_flags & TF_RCVD_SCALE))) { 651 *((u_int32_t *) (opt + optlen)) = htonl( 652 TCPOPT_NOP << 24 | 653 TCPOPT_WINDOW << 16 | 654 TCPOLEN_WINDOW << 8 | 655 tp->request_r_scale); 656 optlen += 4; 657 } 658 } 659 } 660 661 /* 662 * Send a timestamp and echo-reply if this is a SYN and our side 663 * wants to use timestamps (TF_REQ_TSTMP is set) or both our side 664 * and our peer have sent timestamps in our SYN's. 665 */ 666 if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP && 667 (flags & TH_RST) == 0 && 668 ((flags & (TH_SYN|TH_ACK)) == TH_SYN || 669 (tp->t_flags & TF_RCVD_TSTMP))) { 670 u_int32_t *lp = (u_int32_t *)(opt + optlen); 671 672 /* Form timestamp option as shown in appendix A of RFC 1323. */ 673 *lp++ = htonl(TCPOPT_TSTAMP_HDR); 674 *lp++ = htonl(TCP_TIMESTAMP(tp)); 675 *lp = htonl(tp->ts_recent); 676 optlen += TCPOLEN_TSTAMP_APPA; 677 } 678 679 hdrlen += optlen; 680 681 #ifdef DIAGNOSTIC 682 if (len > txsegsize) 683 panic("tcp data to be sent is larger than segment"); 684 if (max_linkhdr + hdrlen > MCLBYTES) 685 panic("tcphdr too big"); 686 #endif 687 688 /* 689 * Grab a header mbuf, attaching a copy of data to 690 * be transmitted, and initialize the header from 691 * the template for sends on this connection. 692 */ 693 if (len) { 694 if (tp->t_force && len == 1) 695 tcpstat.tcps_sndprobe++; 696 else if (SEQ_LT(tp->snd_nxt, tp->snd_max)) { 697 tcpstat.tcps_sndrexmitpack++; 698 tcpstat.tcps_sndrexmitbyte += len; 699 } else { 700 tcpstat.tcps_sndpack++; 701 tcpstat.tcps_sndbyte += len; 702 } 703 #ifdef notyet 704 if ((m = m_copypack(so->so_snd.sb_mb, off, 705 (int)len, max_linkhdr + hdrlen)) == 0) { 706 error = ENOBUFS; 707 goto out; 708 } 709 /* 710 * m_copypack left space for our hdr; use it. 711 */ 712 m->m_len += hdrlen; 713 m->m_data -= hdrlen; 714 #else 715 MGETHDR(m, M_DONTWAIT, MT_HEADER); 716 if (m != NULL && 717 (max_linkhdr + hdrlen > MHLEN || 718 max_linkhdr + hdrlen + len <= MCLBYTES)) { 719 MCLGET(m, M_DONTWAIT); 720 if ((m->m_flags & M_EXT) == 0) { 721 m_freem(m); 722 m = NULL; 723 } 724 } 725 if (m == NULL) { 726 error = ENOBUFS; 727 goto out; 728 } 729 m->m_data += max_linkhdr; 730 m->m_len = hdrlen; 731 if (len <= M_TRAILINGSPACE(m)) { 732 m_copydata(so->so_snd.sb_mb, off, (int) len, 733 mtod(m, caddr_t) + hdrlen); 734 m->m_len += len; 735 } else { 736 m->m_next = m_copy(so->so_snd.sb_mb, off, (int) len); 737 if (m->m_next == NULL) { 738 m_freem(m); 739 error = ENOBUFS; 740 goto out; 741 } 742 } 743 #endif 744 /* 745 * If we're sending everything we've got, set PUSH. 746 * (This will keep happy those implementations which only 747 * give data to the user when a buffer fills or 748 * a PUSH comes in.) 749 */ 750 if (off + len == so->so_snd.sb_cc) 751 flags |= TH_PUSH; 752 } else { 753 if (tp->t_flags & TF_ACKNOW) 754 tcpstat.tcps_sndacks++; 755 else if (flags & (TH_SYN|TH_FIN|TH_RST)) 756 tcpstat.tcps_sndctrl++; 757 else if (SEQ_GT(tp->snd_up, tp->snd_una)) 758 tcpstat.tcps_sndurg++; 759 else 760 tcpstat.tcps_sndwinup++; 761 762 MGETHDR(m, M_DONTWAIT, MT_HEADER); 763 if (m != NULL && max_linkhdr + hdrlen > MHLEN) { 764 MCLGET(m, M_DONTWAIT); 765 if ((m->m_flags & M_EXT) == 0) { 766 m_freem(m); 767 m = NULL; 768 } 769 } 770 if (m == NULL) { 771 error = ENOBUFS; 772 goto out; 773 } 774 m->m_data += max_linkhdr; 775 m->m_len = hdrlen; 776 } 777 m->m_pkthdr.rcvif = (struct ifnet *)0; 778 switch (af) { 779 #ifdef INET 780 case AF_INET: 781 ip = mtod(m, struct ip *); 782 #ifdef INET6 783 ip6 = NULL; 784 #endif 785 th = (struct tcphdr *)(ip + 1); 786 break; 787 #endif 788 #ifdef INET6 789 case AF_INET6: 790 ip = NULL; 791 ip6 = mtod(m, struct ip6_hdr *); 792 th = (struct tcphdr *)(ip6 + 1); 793 break; 794 #endif 795 default: /*pacify gcc*/ 796 ip = NULL; 797 #ifdef INET6 798 ip6 = NULL; 799 #endif 800 th = NULL; 801 break; 802 } 803 if (tp->t_template == 0) 804 panic("tcp_output"); 805 if (tp->t_template->m_len < iphdrlen) 806 panic("tcp_output"); 807 bcopy(mtod(tp->t_template, caddr_t), mtod(m, caddr_t), iphdrlen); 808 809 /* 810 * If we are doing retransmissions, then snd_nxt will 811 * not reflect the first unsent octet. For ACK only 812 * packets, we do not want the sequence number of the 813 * retransmitted packet, we want the sequence number 814 * of the next unsent octet. So, if there is no data 815 * (and no SYN or FIN), use snd_max instead of snd_nxt 816 * when filling in ti_seq. But if we are in persist 817 * state, snd_max might reflect one byte beyond the 818 * right edge of the window, so use snd_nxt in that 819 * case, since we know we aren't doing a retransmission. 820 * (retransmit and persist are mutually exclusive...) 821 */ 822 if (len || (flags & (TH_SYN|TH_FIN)) || 823 TCP_TIMER_ISARMED(tp, TCPT_PERSIST)) 824 th->th_seq = htonl(tp->snd_nxt); 825 else 826 th->th_seq = htonl(tp->snd_max); 827 th->th_ack = htonl(tp->rcv_nxt); 828 if (optlen) { 829 bcopy((caddr_t)opt, (caddr_t)(th + 1), optlen); 830 th->th_off = (sizeof (struct tcphdr) + optlen) >> 2; 831 } 832 th->th_flags = flags; 833 /* 834 * Calculate receive window. Don't shrink window, 835 * but avoid silly window syndrome. 836 */ 837 if (win < (long)(so->so_rcv.sb_hiwat / 4) && win < (long)rxsegsize) 838 win = 0; 839 if (win > (long)TCP_MAXWIN << tp->rcv_scale) 840 win = (long)TCP_MAXWIN << tp->rcv_scale; 841 if (win < (long)(tp->rcv_adv - tp->rcv_nxt)) 842 win = (long)(tp->rcv_adv - tp->rcv_nxt); 843 th->th_win = htons((u_int16_t) (win>>tp->rcv_scale)); 844 if (SEQ_GT(tp->snd_up, tp->snd_nxt)) { 845 u_int32_t urp = tp->snd_up - tp->snd_nxt; 846 if (urp > IP_MAXPACKET) 847 urp = IP_MAXPACKET; 848 th->th_urp = htons((u_int16_t)urp); 849 th->th_flags |= TH_URG; 850 } else 851 /* 852 * If no urgent pointer to send, then we pull 853 * the urgent pointer to the left edge of the send window 854 * so that it doesn't drift into the send window on sequence 855 * number wraparound. 856 */ 857 tp->snd_up = tp->snd_una; /* drag it along */ 858 859 /* 860 * Set ourselves up to be checksummed just before the packet 861 * hits the wire. 862 */ 863 switch (af) { 864 #ifdef INET 865 case AF_INET: 866 m->m_pkthdr.csum_flags = M_CSUM_TCPv4; 867 m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); 868 if (len + optlen) { 869 /* Fixup the pseudo-header checksum. */ 870 /* XXXJRT Not IP Jumbogram safe. */ 871 th->th_sum = in_cksum_addword(th->th_sum, 872 htons((u_int16_t) (len + optlen))); 873 } 874 break; 875 #endif 876 #ifdef INET6 877 case AF_INET6: 878 /* 879 * XXX Actually delaying the checksum is Hard 880 * XXX (well, maybe not for Itojun, but it is 881 * XXX for me), but we can still take advantage 882 * XXX of the cached pseudo-header checksum. 883 */ 884 /* equals to hdrlen + len */ 885 m->m_pkthdr.len = sizeof(struct ip6_hdr) 886 + sizeof(struct tcphdr) + optlen + len; 887 #ifdef notyet 888 m->m_pkthdr.csum_flags = M_CSUM_TCPv6; 889 m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); 890 #endif 891 if (len + optlen) { 892 /* Fixup the pseudo-header checksum. */ 893 /* XXXJRT: Not IPv6 Jumbogram safe. */ 894 th->th_sum = in_cksum_addword(th->th_sum, 895 htons((u_int16_t) (len + optlen))); 896 } 897 #ifndef notyet 898 th->th_sum = in6_cksum(m, 0, sizeof(struct ip6_hdr), 899 sizeof(struct tcphdr) + optlen + len); 900 #endif 901 break; 902 #endif 903 } 904 905 /* 906 * In transmit state, time the transmission and arrange for 907 * the retransmit. In persist state, just set snd_max. 908 */ 909 if (tp->t_force == 0 || TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) { 910 tcp_seq startseq = tp->snd_nxt; 911 912 /* 913 * Advance snd_nxt over sequence space of this segment. 914 * There are no states in which we send both a SYN and a FIN, 915 * so we collapse the tests for these flags. 916 */ 917 if (flags & (TH_SYN|TH_FIN)) 918 tp->snd_nxt++; 919 tp->snd_nxt += len; 920 if (SEQ_GT(tp->snd_nxt, tp->snd_max)) { 921 tp->snd_max = tp->snd_nxt; 922 /* 923 * Time this transmission if not a retransmission and 924 * not currently timing anything. 925 */ 926 if (tp->t_rtt == 0) { 927 tp->t_rtt = 1; 928 tp->t_rtseq = startseq; 929 tcpstat.tcps_segstimed++; 930 } 931 } 932 933 /* 934 * Set retransmit timer if not currently set, 935 * and not doing an ack or a keep-alive probe. 936 * Initial value for retransmit timer is smoothed 937 * round-trip time + 2 * round-trip time variance. 938 * Initialize shift counter which is used for backoff 939 * of retransmit time. 940 */ 941 if (TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 && 942 tp->snd_nxt != tp->snd_una) { 943 TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur); 944 if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST)) { 945 TCP_TIMER_DISARM(tp, TCPT_PERSIST); 946 tp->t_rxtshift = 0; 947 } 948 } 949 } else 950 if (SEQ_GT(tp->snd_nxt + len, tp->snd_max)) 951 tp->snd_max = tp->snd_nxt + len; 952 953 /* 954 * Trace. 955 */ 956 if (so->so_options & SO_DEBUG) { 957 /* 958 * need to recover version # field, which was overwritten 959 * on ip_cksum computation. 960 */ 961 struct ip *sip; 962 sip = mtod(m, struct ip *); 963 switch (af) { 964 #ifdef INET 965 case AF_INET: 966 sip->ip_v = 4; 967 break; 968 #endif 969 #ifdef INET6 970 case AF_INET6: 971 sip->ip_v = 6; 972 break; 973 #endif 974 } 975 tcp_trace(TA_OUTPUT, tp->t_state, tp, m, 0); 976 } 977 978 /* 979 * Fill in IP length and desired time to live and 980 * send to IP level. There should be a better way 981 * to handle ttl and tos; we could keep them in 982 * the template, but need a way to checksum without them. 983 */ 984 m->m_pkthdr.len = hdrlen + len; 985 986 switch (af) { 987 #ifdef INET 988 case AF_INET: 989 ip->ip_len = m->m_pkthdr.len; 990 if (tp->t_inpcb) { 991 ip->ip_ttl = tp->t_inpcb->inp_ip.ip_ttl; 992 ip->ip_tos = tp->t_inpcb->inp_ip.ip_tos; 993 } 994 #ifdef INET6 995 else if (tp->t_in6pcb) { 996 ip->ip_ttl = in6_selecthlim(tp->t_in6pcb, NULL); /*XXX*/ 997 ip->ip_tos = 0; /*XXX*/ 998 } 999 #endif 1000 break; 1001 #endif 1002 #ifdef INET6 1003 case AF_INET6: 1004 ip6->ip6_nxt = IPPROTO_TCP; 1005 if (tp->t_in6pcb) { 1006 /* 1007 * we separately set hoplimit for every segment, since 1008 * the user might want to change the value via 1009 * setsockopt. Also, desired default hop limit might 1010 * be changed via Neighbor Discovery. 1011 */ 1012 ip6->ip6_hlim = in6_selecthlim(tp->t_in6pcb, 1013 ro->ro_rt ? ro->ro_rt->rt_ifp : NULL); 1014 } 1015 /* ip6->ip6_flow = ??? */ 1016 /* ip6_plen will be filled in ip6_output(). */ 1017 break; 1018 #endif 1019 } 1020 1021 #ifdef IPSEC 1022 if (ipsec_setsocket(m, so) != 0) { 1023 m_freem(m); 1024 error = ENOBUFS; 1025 goto out; 1026 } 1027 #endif /*IPSEC*/ 1028 1029 switch (af) { 1030 #ifdef INET 1031 case AF_INET: 1032 { 1033 struct mbuf *opts; 1034 1035 if (tp->t_inpcb) 1036 opts = tp->t_inpcb->inp_options; 1037 else 1038 opts = NULL; 1039 error = ip_output(m, opts, ro, 1040 (ip_mtudisc ? IP_MTUDISC : 0) | 1041 (so->so_options & SO_DONTROUTE), 1042 0); 1043 break; 1044 } 1045 #endif 1046 #ifdef INET6 1047 case AF_INET6: 1048 { 1049 struct ip6_pktopts *opts; 1050 1051 if (tp->t_in6pcb) 1052 opts = tp->t_in6pcb->in6p_outputopts; 1053 else 1054 opts = NULL; 1055 error = ip6_output(m, opts, (struct route_in6 *)ro, 1056 so->so_options & SO_DONTROUTE, 0, NULL); 1057 break; 1058 } 1059 #endif 1060 default: 1061 error = EAFNOSUPPORT; 1062 break; 1063 } 1064 if (error) { 1065 out: 1066 if (error == ENOBUFS) { 1067 #ifdef INET 1068 if (tp->t_inpcb) 1069 tcp_quench(tp->t_inpcb, 0); 1070 #endif 1071 #ifdef INET6 1072 if (tp->t_in6pcb) 1073 tcp6_quench(tp->t_in6pcb, 0); 1074 #endif 1075 return (0); 1076 } 1077 if ((error == EHOSTUNREACH || error == ENETDOWN) 1078 && TCPS_HAVERCVDSYN(tp->t_state)) { 1079 tp->t_softerror = error; 1080 return (0); 1081 } 1082 return (error); 1083 } 1084 tcpstat.tcps_sndtotal++; 1085 if (tp->t_flags & TF_DELACK) 1086 tcpstat.tcps_delack++; 1087 1088 /* 1089 * Data sent (as far as we can tell). 1090 * If this advertises a larger window than any other segment, 1091 * then remember the size of the advertised window. 1092 * Any pending ACK has now been sent. 1093 */ 1094 if (win > 0 && SEQ_GT(tp->rcv_nxt+win, tp->rcv_adv)) 1095 tp->rcv_adv = tp->rcv_nxt + win; 1096 tp->last_ack_sent = tp->rcv_nxt; 1097 tp->t_flags &= ~TF_ACKNOW; 1098 TCP_CLEAR_DELACK(tp); 1099 #ifdef DIAGNOSTIC 1100 if (maxburst < 0) 1101 printf("tcp_output: maxburst exceeded by %d\n", -maxburst); 1102 #endif 1103 if (sendalot && (!tcp_do_newreno || --maxburst)) 1104 goto again; 1105 return (0); 1106 } 1107 1108 void 1109 tcp_setpersist(tp) 1110 struct tcpcb *tp; 1111 { 1112 int t = ((tp->t_srtt >> 2) + tp->t_rttvar) >> (1 + 2); 1113 int nticks; 1114 1115 if (TCP_TIMER_ISARMED(tp, TCPT_REXMT)) 1116 panic("tcp_output REXMT"); 1117 /* 1118 * Start/restart persistance timer. 1119 */ 1120 if (t < tp->t_rttmin) 1121 t = tp->t_rttmin; 1122 TCPT_RANGESET(nticks, t * tcp_backoff[tp->t_rxtshift], 1123 TCPTV_PERSMIN, TCPTV_PERSMAX); 1124 TCP_TIMER_ARM(tp, TCPT_PERSIST, nticks); 1125 if (tp->t_rxtshift < TCP_MAXRXTSHIFT) 1126 tp->t_rxtshift++; 1127 } 1128
Indexes created Tue Sep 23 14:10:03 GMT 2025