tcp_output.c revision 1.149
1 /* $NetBSD: tcp_output.c,v 1.149 2006/10/09 16:27:07 rpaulo Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 /* 33 * @(#)COPYRIGHT 1.1 (NRL) 17 January 1995 34 * 35 * NRL grants permission for redistribution and use in source and binary 36 * forms, with or without modification, of the software and documentation 37 * created at NRL provided that the following conditions are met: 38 * 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 3. All advertising materials mentioning features or use of this software 45 * must display the following acknowledgements: 46 * This product includes software developed by the University of 47 * California, Berkeley and its contributors. 48 * This product includes software developed at the Information 49 * Technology Division, US Naval Research Laboratory. 50 * 4. Neither the name of the NRL nor the names of its contributors 51 * may be used to endorse or promote products derived from this software 52 * without specific prior written permission. 53 * 54 * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS 55 * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 56 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A 57 * PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NRL OR 58 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 59 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 60 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 61 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 62 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 63 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 64 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 65 * 66 * The views and conclusions contained in the software and documentation 67 * are those of the authors and should not be interpreted as representing 68 * official policies, either expressed or implied, of the US Naval 69 * Research Laboratory (NRL). 70 */ 71 72 /*- 73 * Copyright (c) 1997, 1998, 2001, 2005, 2006 The NetBSD Foundation, Inc. 74 * All rights reserved. 75 * 76 * This code is derived from software contributed to The NetBSD Foundation 77 * by Jason R. Thorpe and Kevin M. Lahey of the Numerical Aerospace Simulation 78 * Facility, NASA Ames Research Center. 79 * This code is derived from software contributed to The NetBSD Foundation 80 * by Charles M. Hannum. 81 * This code is derived from software contributed to The NetBSD Foundation 82 * by Rui Paulo. 83 * 84 * Redistribution and use in source and binary forms, with or without 85 * modification, are permitted provided that the following conditions 86 * are met: 87 * 1. Redistributions of source code must retain the above copyright 88 * notice, this list of conditions and the following disclaimer. 89 * 2. Redistributions in binary form must reproduce the above copyright 90 * notice, this list of conditions and the following disclaimer in the 91 * documentation and/or other materials provided with the distribution. 92 * 3. All advertising materials mentioning features or use of this software 93 * must display the following acknowledgement: 94 * This product includes software developed by the NetBSD 95 * Foundation, Inc. and its contributors. 96 * 4. Neither the name of The NetBSD Foundation nor the names of its 97 * contributors may be used to endorse or promote products derived 98 * from this software without specific prior written permission. 99 * 100 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 101 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 102 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 103 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 104 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 105 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 106 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 107 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 108 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 109 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 110 * POSSIBILITY OF SUCH DAMAGE. 111 */ 112 113 /* 114 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 115 * The Regents of the University of California. All rights reserved. 116 * 117 * Redistribution and use in source and binary forms, with or without 118 * modification, are permitted provided that the following conditions 119 * are met: 120 * 1. Redistributions of source code must retain the above copyright 121 * notice, this list of conditions and the following disclaimer. 122 * 2. Redistributions in binary form must reproduce the above copyright 123 * notice, this list of conditions and the following disclaimer in the 124 * documentation and/or other materials provided with the distribution. 125 * 3. Neither the name of the University nor the names of its contributors 126 * may be used to endorse or promote products derived from this software 127 * without specific prior written permission. 128 * 129 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 130 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 131 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 132 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 133 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 134 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 135 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 136 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 137 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 138 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 139 * SUCH DAMAGE. 140 * 141 * @(#)tcp_output.c 8.4 (Berkeley) 5/24/95 142 */ 143 144 #include <sys/cdefs.h> 145 __KERNEL_RCSID(0, "$NetBSD: tcp_output.c,v 1.149 2006/10/09 16:27:07 rpaulo Exp $"); 146 147 #include "opt_inet.h" 148 #include "opt_ipsec.h" 149 #include "opt_tcp_debug.h" 150 151 #include <sys/param.h> 152 #include <sys/systm.h> 153 #include <sys/malloc.h> 154 #include <sys/mbuf.h> 155 #include <sys/protosw.h> 156 #include <sys/socket.h> 157 #include <sys/socketvar.h> 158 #include <sys/errno.h> 159 #include <sys/domain.h> 160 #include <sys/kernel.h> 161 #ifdef TCP_SIGNATURE 162 #include <sys/md5.h> 163 #endif 164 165 #include <net/if.h> 166 #include <net/route.h> 167 168 #include <netinet/in.h> 169 #include <netinet/in_systm.h> 170 #include <netinet/ip.h> 171 #include <netinet/in_pcb.h> 172 #include <netinet/ip_var.h> 173 174 #ifdef INET6 175 #ifndef INET 176 #include <netinet/in.h> 177 #endif 178 #include <netinet/ip6.h> 179 #include <netinet6/in6_var.h> 180 #include <netinet6/ip6_var.h> 181 #include <netinet6/in6_pcb.h> 182 #include <netinet6/nd6.h> 183 #endif 184 185 #ifdef FAST_IPSEC 186 #include <netipsec/ipsec.h> 187 #include <netipsec/key.h> 188 #endif /* FAST_IPSEC*/ 189 #ifdef IPSEC 190 #include <netinet6/ipsec.h> 191 #endif 192 193 #include <netinet/tcp.h> 194 #define TCPOUTFLAGS 195 #include <netinet/tcp_fsm.h> 196 #include <netinet/tcp_seq.h> 197 #include <netinet/tcp_timer.h> 198 #include <netinet/tcp_var.h> 199 #include <netinet/tcp_congctl.h> 200 #include <netinet/tcpip.h> 201 #include <netinet/tcp_debug.h> 202 #include <netinet/in_offload.h> 203 204 #ifdef IPSEC 205 #include <netkey/key.h> 206 #endif 207 208 #ifdef notyet 209 extern struct mbuf *m_copypack(); 210 #endif 211 212 /* 213 * Knob to enable Congestion Window Monitoring, and control 214 * the burst size it allows. Default burst is 4 packets, per 215 * the Internet draft. 216 */ 217 int tcp_cwm = 0; 218 int tcp_cwm_burstsize = 4; 219 220 #ifdef TCP_OUTPUT_COUNTERS 221 #include <sys/device.h> 222 223 extern struct evcnt tcp_output_bigheader; 224 extern struct evcnt tcp_output_predict_hit; 225 extern struct evcnt tcp_output_predict_miss; 226 extern struct evcnt tcp_output_copysmall; 227 extern struct evcnt tcp_output_copybig; 228 extern struct evcnt tcp_output_refbig; 229 230 #define TCP_OUTPUT_COUNTER_INCR(ev) (ev)->ev_count++ 231 #else 232 233 #define TCP_OUTPUT_COUNTER_INCR(ev) /* nothing */ 234 235 #endif /* TCP_OUTPUT_COUNTERS */ 236 237 static 238 #ifndef GPROF 239 inline 240 #endif 241 int 242 tcp_segsize(struct tcpcb *tp, int *txsegsizep, int *rxsegsizep) 243 { 244 #ifdef INET 245 struct inpcb *inp = tp->t_inpcb; 246 #endif 247 #ifdef INET6 248 struct in6pcb *in6p = tp->t_in6pcb; 249 #endif 250 struct socket *so = NULL; 251 struct rtentry *rt; 252 struct ifnet *ifp; 253 int size; 254 int hdrlen; 255 int optlen; 256 257 #ifdef DIAGNOSTIC 258 if (tp->t_inpcb && tp->t_in6pcb) 259 panic("tcp_segsize: both t_inpcb and t_in6pcb are set"); 260 #endif 261 switch (tp->t_family) { 262 #ifdef INET 263 case AF_INET: 264 hdrlen = sizeof(struct ip) + sizeof(struct tcphdr); 265 break; 266 #endif 267 #ifdef INET6 268 case AF_INET6: 269 hdrlen = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); 270 break; 271 #endif 272 default: 273 size = tcp_mssdflt; 274 goto out; 275 } 276 277 rt = NULL; 278 #ifdef INET 279 if (inp) { 280 rt = in_pcbrtentry(inp); 281 so = inp->inp_socket; 282 } 283 #endif 284 #ifdef INET6 285 if (in6p) { 286 rt = in6_pcbrtentry(in6p); 287 so = in6p->in6p_socket; 288 } 289 #endif 290 if (rt == NULL) { 291 size = tcp_mssdflt; 292 goto out; 293 } 294 295 ifp = rt->rt_ifp; 296 297 size = tcp_mssdflt; 298 if (tp->t_mtudisc && rt->rt_rmx.rmx_mtu != 0) { 299 #ifdef INET6 300 if (in6p && rt->rt_rmx.rmx_mtu < IPV6_MMTU) { 301 /* 302 * RFC2460 section 5, last paragraph: if path MTU is 303 * smaller than 1280, use 1280 as packet size and 304 * attach fragment header. 305 */ 306 size = IPV6_MMTU - hdrlen - sizeof(struct ip6_frag); 307 } else 308 size = rt->rt_rmx.rmx_mtu - hdrlen; 309 #else 310 size = rt->rt_rmx.rmx_mtu - hdrlen; 311 #endif 312 } else if (ifp->if_flags & IFF_LOOPBACK) 313 size = ifp->if_mtu - hdrlen; 314 #ifdef INET 315 else if (inp && tp->t_mtudisc) 316 size = ifp->if_mtu - hdrlen; 317 else if (inp && in_localaddr(inp->inp_faddr)) 318 size = ifp->if_mtu - hdrlen; 319 #endif 320 #ifdef INET6 321 else if (in6p) { 322 #ifdef INET 323 if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) { 324 /* mapped addr case */ 325 struct in_addr d; 326 bcopy(&in6p->in6p_faddr.s6_addr32[3], &d, sizeof(d)); 327 if (tp->t_mtudisc || in_localaddr(d)) 328 size = ifp->if_mtu - hdrlen; 329 } else 330 #endif 331 { 332 /* 333 * for IPv6, path MTU discovery is always turned on, 334 * or the node must use packet size <= 1280. 335 */ 336 size = tp->t_mtudisc ? IN6_LINKMTU(ifp) : IPV6_MMTU; 337 size -= hdrlen; 338 } 339 } 340 #endif 341 out: 342 /* 343 * Now we must make room for whatever extra TCP/IP options are in 344 * the packet. 345 */ 346 optlen = tcp_optlen(tp); 347 348 /* 349 * XXX tp->t_ourmss should have the right size, but without this code 350 * fragmentation will occur... need more investigation 351 */ 352 #ifdef INET 353 if (inp) { 354 #if defined(IPSEC) || defined(FAST_IPSEC) 355 if (! IPSEC_PCB_SKIP_IPSEC(inp->inp_sp, IPSEC_DIR_OUTBOUND)) 356 optlen += ipsec4_hdrsiz_tcp(tp); 357 #endif 358 optlen += ip_optlen(inp); 359 } 360 #endif 361 #ifdef INET6 362 #ifdef INET 363 if (in6p && tp->t_family == AF_INET) { 364 #if defined(IPSEC) || defined(FAST_IPSEC) 365 if (! IPSEC_PCB_SKIP_IPSEC(in6p->in6p_sp, IPSEC_DIR_OUTBOUND)) 366 optlen += ipsec4_hdrsiz_tcp(tp); 367 #endif 368 /* XXX size -= ip_optlen(in6p); */ 369 } else 370 #endif 371 if (in6p && tp->t_family == AF_INET6) { 372 #ifdef IPSEC 373 if (! IPSEC_PCB_SKIP_IPSEC(in6p->in6p_sp, IPSEC_DIR_OUTBOUND)) 374 optlen += ipsec6_hdrsiz_tcp(tp); 375 #endif 376 optlen += ip6_optlen(in6p); 377 } 378 #endif 379 size -= optlen; 380 381 /* there may not be any room for data if mtu is too small */ 382 if (size < 0) 383 return (EMSGSIZE); 384 385 /* 386 * *rxsegsizep holds *estimated* inbound segment size (estimation 387 * assumes that path MTU is the same for both ways). this is only 388 * for silly window avoidance, do not use the value for other purposes. 389 * 390 * ipseclen is subtracted from both sides, this may not be right. 391 * I'm not quite sure about this (could someone comment). 392 */ 393 *txsegsizep = min(tp->t_peermss - optlen, size); 394 /* 395 * Never send more than half a buffer full. This insures that we can 396 * always keep 2 packets on the wire, no matter what SO_SNDBUF is, and 397 * therefore acks will never be delayed unless we run out of data to 398 * transmit. 399 */ 400 if (so) 401 *txsegsizep = min(so->so_snd.sb_hiwat >> 1, *txsegsizep); 402 *rxsegsizep = min(tp->t_ourmss - optlen, size); 403 404 if (*txsegsizep != tp->t_segsz) { 405 /* 406 * If the new segment size is larger, we don't want to 407 * mess up the congestion window, but if it is smaller 408 * we'll have to reduce the congestion window to ensure 409 * that we don't get into trouble with initial windows 410 * and the rest. In any case, if the segment size 411 * has changed, chances are the path has, too, and 412 * our congestion window will be different. 413 */ 414 if (*txsegsizep < tp->t_segsz) { 415 tp->snd_cwnd = max((tp->snd_cwnd / tp->t_segsz) 416 * *txsegsizep, *txsegsizep); 417 tp->snd_ssthresh = max((tp->snd_ssthresh / tp->t_segsz) 418 * *txsegsizep, *txsegsizep); 419 } 420 tp->t_segsz = *txsegsizep; 421 } 422 423 return (0); 424 } 425 426 static 427 #ifndef GPROF 428 inline 429 #endif 430 int 431 tcp_build_datapkt(struct tcpcb *tp, struct socket *so, int off, 432 long len, int hdrlen, struct mbuf **mp) 433 { 434 struct mbuf *m, *m0; 435 436 if (tp->t_force && len == 1) 437 tcpstat.tcps_sndprobe++; 438 else if (SEQ_LT(tp->snd_nxt, tp->snd_max)) { 439 tcpstat.tcps_sndrexmitpack++; 440 tcpstat.tcps_sndrexmitbyte += len; 441 } else { 442 tcpstat.tcps_sndpack++; 443 tcpstat.tcps_sndbyte += len; 444 } 445 #ifdef notyet 446 if ((m = m_copypack(so->so_snd.sb_mb, off, 447 (int)len, max_linkhdr + hdrlen)) == 0) 448 return (ENOBUFS); 449 /* 450 * m_copypack left space for our hdr; use it. 451 */ 452 m->m_len += hdrlen; 453 m->m_data -= hdrlen; 454 #else 455 MGETHDR(m, M_DONTWAIT, MT_HEADER); 456 if (__predict_false(m == NULL)) 457 return (ENOBUFS); 458 MCLAIM(m, &tcp_tx_mowner); 459 460 /* 461 * XXX Because other code assumes headers will fit in 462 * XXX one header mbuf. 463 * 464 * (This code should almost *never* be run.) 465 */ 466 if (__predict_false((max_linkhdr + hdrlen) > MHLEN)) { 467 TCP_OUTPUT_COUNTER_INCR(&tcp_output_bigheader); 468 MCLGET(m, M_DONTWAIT); 469 if ((m->m_flags & M_EXT) == 0) { 470 m_freem(m); 471 return (ENOBUFS); 472 } 473 } 474 475 m->m_data += max_linkhdr; 476 m->m_len = hdrlen; 477 478 /* 479 * To avoid traversing the whole sb_mb chain for correct 480 * data to send, remember last sent mbuf, its offset and 481 * the sent size. When called the next time, see if the 482 * data to send is directly following the previous transfer. 483 * This is important for large TCP windows. 484 */ 485 if (off == 0 || tp->t_lastm == NULL || 486 (tp->t_lastoff + tp->t_lastlen) != off) { 487 TCP_OUTPUT_COUNTER_INCR(&tcp_output_predict_miss); 488 /* 489 * Either a new packet or a retransmit. 490 * Start from the beginning. 491 */ 492 tp->t_lastm = so->so_snd.sb_mb; 493 tp->t_inoff = off; 494 } else { 495 TCP_OUTPUT_COUNTER_INCR(&tcp_output_predict_hit); 496 tp->t_inoff += tp->t_lastlen; 497 } 498 499 /* Traverse forward to next packet */ 500 while (tp->t_inoff > 0) { 501 if (tp->t_lastm == NULL) 502 panic("tp->t_lastm == NULL"); 503 if (tp->t_inoff < tp->t_lastm->m_len) 504 break; 505 tp->t_inoff -= tp->t_lastm->m_len; 506 tp->t_lastm = tp->t_lastm->m_next; 507 } 508 509 tp->t_lastoff = off; 510 tp->t_lastlen = len; 511 m0 = tp->t_lastm; 512 off = tp->t_inoff; 513 514 if (len <= M_TRAILINGSPACE(m)) { 515 m_copydata(m0, off, (int) len, mtod(m, caddr_t) + hdrlen); 516 m->m_len += len; 517 TCP_OUTPUT_COUNTER_INCR(&tcp_output_copysmall); 518 } else { 519 m->m_next = m_copy(m0, off, (int) len); 520 if (m->m_next == NULL) { 521 m_freem(m); 522 return (ENOBUFS); 523 } 524 #ifdef TCP_OUTPUT_COUNTERS 525 if (m->m_next->m_flags & M_EXT) 526 TCP_OUTPUT_COUNTER_INCR(&tcp_output_refbig); 527 else 528 TCP_OUTPUT_COUNTER_INCR(&tcp_output_copybig); 529 #endif /* TCP_OUTPUT_COUNTERS */ 530 } 531 #endif 532 533 *mp = m; 534 return (0); 535 } 536 537 /* 538 * Tcp output routine: figure out what should be sent and send it. 539 */ 540 int 541 tcp_output(struct tcpcb *tp) 542 { 543 struct socket *so; 544 struct route *ro; 545 long len, win; 546 int off, flags, error; 547 struct mbuf *m; 548 struct ip *ip; 549 #ifdef INET6 550 struct ip6_hdr *ip6; 551 #endif 552 struct tcphdr *th; 553 u_char opt[MAX_TCPOPTLEN]; 554 unsigned optlen, hdrlen, packetlen; 555 unsigned int sack_numblks; 556 int idle, sendalot, txsegsize, rxsegsize; 557 int txsegsize_nosack; 558 int maxburst = TCP_MAXBURST; 559 int af; /* address family on the wire */ 560 int iphdrlen; 561 int has_tso, use_tso; 562 int sack_rxmit; 563 int sack_bytes_rxmt; 564 struct sackhole *p; 565 #ifdef TCP_SIGNATURE 566 int sigoff = 0; 567 #endif 568 569 #ifdef DIAGNOSTIC 570 if (tp->t_inpcb && tp->t_in6pcb) 571 panic("tcp_output: both t_inpcb and t_in6pcb are set"); 572 #endif 573 so = NULL; 574 ro = NULL; 575 if (tp->t_inpcb) { 576 so = tp->t_inpcb->inp_socket; 577 ro = &tp->t_inpcb->inp_route; 578 } 579 #ifdef INET6 580 else if (tp->t_in6pcb) { 581 so = tp->t_in6pcb->in6p_socket; 582 ro = (struct route *)&tp->t_in6pcb->in6p_route; 583 } 584 #endif 585 586 switch (af = tp->t_family) { 587 #ifdef INET 588 case AF_INET: 589 if (tp->t_inpcb) 590 break; 591 #ifdef INET6 592 /* mapped addr case */ 593 if (tp->t_in6pcb) 594 break; 595 #endif 596 return (EINVAL); 597 #endif 598 #ifdef INET6 599 case AF_INET6: 600 if (tp->t_in6pcb) 601 break; 602 return (EINVAL); 603 #endif 604 default: 605 return (EAFNOSUPPORT); 606 } 607 608 if (tcp_segsize(tp, &txsegsize, &rxsegsize)) 609 return (EMSGSIZE); 610 611 idle = (tp->snd_max == tp->snd_una); 612 613 /* 614 * Determine if we can use TCP segmentation offload: 615 * - If we're using IPv4 616 * - If there is not an IPsec policy that prevents it 617 * - If the interface can do it 618 */ 619 has_tso = tp->t_inpcb != NULL && 620 #if defined(IPSEC) || defined(FAST_IPSEC) 621 IPSEC_PCB_SKIP_IPSEC(tp->t_inpcb->inp_sp, 622 IPSEC_DIR_OUTBOUND) && 623 #endif 624 tp->t_inpcb->inp_route.ro_rt != NULL && 625 (tp->t_inpcb->inp_route.ro_rt->rt_ifp->if_capenable & 626 IFCAP_TSOv4) != 0; 627 628 /* 629 * Restart Window computation. From draft-floyd-incr-init-win-03: 630 * 631 * Optionally, a TCP MAY set the restart window to the 632 * minimum of the value used for the initial window and 633 * the current value of cwnd (in other words, using a 634 * larger value for the restart window should never increase 635 * the size of cwnd). 636 */ 637 if (tcp_cwm) { 638 /* 639 * Hughes/Touch/Heidemann Congestion Window Monitoring. 640 * Count the number of packets currently pending 641 * acknowledgement, and limit our congestion window 642 * to a pre-determined allowed burst size plus that count. 643 * This prevents bursting once all pending packets have 644 * been acknowledged (i.e. transmission is idle). 645 * 646 * XXX Link this to Initial Window? 647 */ 648 tp->snd_cwnd = min(tp->snd_cwnd, 649 (tcp_cwm_burstsize * txsegsize) + 650 (tp->snd_nxt - tp->snd_una)); 651 } else { 652 if (idle && (tcp_now - tp->t_rcvtime) >= tp->t_rxtcur) { 653 /* 654 * We have been idle for "a while" and no acks are 655 * expected to clock out any data we send -- 656 * slow start to get ack "clock" running again. 657 */ 658 int ss = tcp_init_win; 659 #ifdef INET 660 if (tp->t_inpcb && 661 in_localaddr(tp->t_inpcb->inp_faddr)) 662 ss = tcp_init_win_local; 663 #endif 664 #ifdef INET6 665 if (tp->t_in6pcb && 666 in6_localaddr(&tp->t_in6pcb->in6p_faddr)) 667 ss = tcp_init_win_local; 668 #endif 669 tp->snd_cwnd = min(tp->snd_cwnd, 670 TCP_INITIAL_WINDOW(ss, txsegsize)); 671 } 672 } 673 674 txsegsize_nosack = txsegsize; 675 again: 676 use_tso = has_tso; 677 if ((tp->t_flags & (TF_ECN_SND_CWR|TF_ECN_SND_ECE)) != 0) { 678 /* don't duplicate CWR/ECE. */ 679 use_tso = 0; 680 } 681 TCP_REASS_LOCK(tp); 682 sack_numblks = tcp_sack_numblks(tp); 683 if (sack_numblks) { 684 int sackoptlen; 685 686 sackoptlen = TCP_SACK_OPTLEN(sack_numblks); 687 if (sackoptlen > txsegsize_nosack) { 688 sack_numblks = 0; /* give up SACK */ 689 txsegsize = txsegsize_nosack; 690 } else { 691 if ((tp->rcv_sack_flags & TCPSACK_HAVED) != 0) { 692 /* don't duplicate D-SACK. */ 693 use_tso = 0; 694 } 695 txsegsize = txsegsize_nosack - sackoptlen; 696 } 697 } else { 698 txsegsize = txsegsize_nosack; 699 } 700 701 /* 702 * Determine length of data that should be transmitted, and 703 * flags that should be used. If there is some data or critical 704 * controls (SYN, RST) to send, then transmit; otherwise, 705 * investigate further. 706 * 707 * Readjust SACK information to avoid resending duplicate data. 708 */ 709 if (TCP_SACK_ENABLED(tp) && SEQ_LT(tp->snd_nxt, tp->snd_max)) 710 tcp_sack_adjust(tp); 711 sendalot = 0; 712 off = tp->snd_nxt - tp->snd_una; 713 win = min(tp->snd_wnd, tp->snd_cwnd); 714 715 flags = tcp_outflags[tp->t_state]; 716 717 /* 718 * Send any SACK-generated retransmissions. If we're explicitly trying 719 * to send out new data (when sendalot is 1), bypass this function. 720 * If we retransmit in fast recovery mode, decrement snd_cwnd, since 721 * we're replacing a (future) new transmission with a retransmission 722 * now, and we previously incremented snd_cwnd in tcp_input(). 723 */ 724 /* 725 * Still in sack recovery , reset rxmit flag to zero. 726 */ 727 sack_rxmit = 0; 728 sack_bytes_rxmt = 0; 729 len = 0; 730 p = NULL; 731 do { 732 long cwin; 733 if (!TCP_SACK_ENABLED(tp)) 734 break; 735 if (tp->t_partialacks < 0) 736 break; 737 p = tcp_sack_output(tp, &sack_bytes_rxmt); 738 if (p == NULL) 739 break; 740 741 cwin = min(tp->snd_wnd, tp->snd_cwnd) - sack_bytes_rxmt; 742 if (cwin < 0) 743 cwin = 0; 744 /* Do not retransmit SACK segments beyond snd_recover */ 745 if (SEQ_GT(p->end, tp->snd_recover)) { 746 /* 747 * (At least) part of sack hole extends beyond 748 * snd_recover. Check to see if we can rexmit data 749 * for this hole. 750 */ 751 if (SEQ_GEQ(p->rxmit, tp->snd_recover)) { 752 /* 753 * Can't rexmit any more data for this hole. 754 * That data will be rexmitted in the next 755 * sack recovery episode, when snd_recover 756 * moves past p->rxmit. 757 */ 758 p = NULL; 759 break; 760 } 761 /* Can rexmit part of the current hole */ 762 len = ((long)ulmin(cwin, tp->snd_recover - p->rxmit)); 763 } else 764 len = ((long)ulmin(cwin, p->end - p->rxmit)); 765 off = p->rxmit - tp->snd_una; 766 if (off + len > so->so_snd.sb_cc) { 767 /* 1 for TH_FIN */ 768 KASSERT(off + len == so->so_snd.sb_cc + 1); 769 KASSERT(p->rxmit + len == tp->snd_max); 770 len = so->so_snd.sb_cc - off; 771 } 772 if (len > 0) { 773 sack_rxmit = 1; 774 sendalot = 1; 775 } 776 } while (/*CONSTCOND*/0); 777 778 /* 779 * If in persist timeout with window of 0, send 1 byte. 780 * Otherwise, if window is small but nonzero 781 * and timer expired, we will send what we can 782 * and go to transmit state. 783 */ 784 if (tp->t_force) { 785 if (win == 0) { 786 /* 787 * If we still have some data to send, then 788 * clear the FIN bit. Usually this would 789 * happen below when it realizes that we 790 * aren't sending all the data. However, 791 * if we have exactly 1 byte of unset data, 792 * then it won't clear the FIN bit below, 793 * and if we are in persist state, we wind 794 * up sending the packet without recording 795 * that we sent the FIN bit. 796 * 797 * We can't just blindly clear the FIN bit, 798 * because if we don't have any more data 799 * to send then the probe will be the FIN 800 * itself. 801 */ 802 if (off < so->so_snd.sb_cc) 803 flags &= ~TH_FIN; 804 win = 1; 805 } else { 806 TCP_TIMER_DISARM(tp, TCPT_PERSIST); 807 tp->t_rxtshift = 0; 808 } 809 } 810 811 if (sack_rxmit == 0) { 812 if (sack_bytes_rxmt != 0) { 813 long cwin; 814 815 /* 816 * We are inside of a SACK recovery episode and are 817 * sending new data, having retransmitted all the 818 * data possible in the scoreboard. 819 */ 820 if (tp->snd_wnd < so->so_snd.sb_cc) { 821 len = tp->snd_wnd - off; 822 flags &= ~TH_FIN; 823 } else { 824 len = so->so_snd.sb_cc - off; 825 } 826 827 /* 828 * From FreeBSD: 829 * Don't remove this (len > 0) check ! 830 * We explicitly check for len > 0 here (although it 831 * isn't really necessary), to work around a gcc 832 * optimization issue - to force gcc to compute 833 * len above. Without this check, the computation 834 * of len is bungled by the optimizer. 835 */ 836 if (len > 0) { 837 cwin = tp->snd_cwnd - 838 (tp->snd_nxt - tp->sack_newdata) - 839 sack_bytes_rxmt; 840 if (cwin < 0) 841 cwin = 0; 842 if (cwin < len) { 843 len = cwin; 844 flags &= ~TH_FIN; 845 } 846 } 847 } else if (win < so->so_snd.sb_cc) { 848 len = win - off; 849 flags &= ~TH_FIN; 850 } else { 851 len = so->so_snd.sb_cc - off; 852 } 853 } 854 855 if (len < 0) { 856 /* 857 * If FIN has been sent but not acked, 858 * but we haven't been called to retransmit, 859 * len will be -1. Otherwise, window shrank 860 * after we sent into it. If window shrank to 0, 861 * cancel pending retransmit, pull snd_nxt back 862 * to (closed) window, and set the persist timer 863 * if it isn't already going. If the window didn't 864 * close completely, just wait for an ACK. 865 * 866 * If we have a pending FIN, either it has already been 867 * transmitted or it is outside the window, so drop it. 868 * If the FIN has been transmitted, but this is not a 869 * retransmission, then len must be -1. Therefore we also 870 * prevent here the sending of `gratuitous FINs'. This 871 * eliminates the need to check for that case below (e.g. 872 * to back up snd_nxt before the FIN so that the sequence 873 * number is correct). 874 */ 875 len = 0; 876 flags &= ~TH_FIN; 877 if (win == 0) { 878 TCP_TIMER_DISARM(tp, TCPT_REXMT); 879 tp->t_rxtshift = 0; 880 tp->snd_nxt = tp->snd_una; 881 if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) 882 tcp_setpersist(tp); 883 } 884 } 885 if (len > txsegsize) { 886 if (use_tso) { 887 /* 888 * Truncate TSO transfers to IP_MAXPACKET, and make 889 * sure that we send equal size transfers down the 890 * stack (rather than big-small-big-small-...). 891 */ 892 len = (min(len, IP_MAXPACKET) / txsegsize) * txsegsize; 893 if (len <= txsegsize) { 894 use_tso = 0; 895 } 896 } else 897 len = txsegsize; 898 flags &= ~TH_FIN; 899 sendalot = 1; 900 } else 901 use_tso = 0; 902 if (sack_rxmit) { 903 if (SEQ_LT(p->rxmit + len, tp->snd_una + so->so_snd.sb_cc)) 904 flags &= ~TH_FIN; 905 } 906 907 win = sbspace(&so->so_rcv); 908 909 /* 910 * Sender silly window avoidance. If connection is idle 911 * and can send all data, a maximum segment, 912 * at least a maximum default-size segment do it, 913 * or are forced, do it; otherwise don't bother. 914 * If peer's buffer is tiny, then send 915 * when window is at least half open. 916 * If retransmitting (possibly after persist timer forced us 917 * to send into a small window), then must resend. 918 */ 919 if (len) { 920 if (len >= txsegsize) 921 goto send; 922 if ((so->so_state & SS_MORETOCOME) == 0 && 923 ((idle || tp->t_flags & TF_NODELAY) && 924 len + off >= so->so_snd.sb_cc)) 925 goto send; 926 if (tp->t_force) 927 goto send; 928 if (len >= tp->max_sndwnd / 2) 929 goto send; 930 if (SEQ_LT(tp->snd_nxt, tp->snd_max)) 931 goto send; 932 if (sack_rxmit) 933 goto send; 934 } 935 936 /* 937 * Compare available window to amount of window known to peer 938 * (as advertised window less next expected input). If the 939 * difference is at least twice the size of the largest segment 940 * we expect to receive (i.e. two segments) or at least 50% of 941 * the maximum possible window, then want to send a window update 942 * to peer. 943 */ 944 if (win > 0) { 945 /* 946 * "adv" is the amount we can increase the window, 947 * taking into account that we are limited by 948 * TCP_MAXWIN << tp->rcv_scale. 949 */ 950 long adv = min(win, (long)TCP_MAXWIN << tp->rcv_scale) - 951 (tp->rcv_adv - tp->rcv_nxt); 952 953 if (adv >= (long) (2 * rxsegsize)) 954 goto send; 955 if (2 * adv >= (long) so->so_rcv.sb_hiwat) 956 goto send; 957 } 958 959 /* 960 * Send if we owe peer an ACK. 961 */ 962 if (tp->t_flags & TF_ACKNOW) 963 goto send; 964 if (flags & (TH_SYN|TH_FIN|TH_RST)) 965 goto send; 966 if (SEQ_GT(tp->snd_up, tp->snd_una)) 967 goto send; 968 /* 969 * In SACK, it is possible for tcp_output to fail to send a segment 970 * after the retransmission timer has been turned off. Make sure 971 * that the retransmission timer is set. 972 */ 973 if (TCP_SACK_ENABLED(tp) && SEQ_GT(tp->snd_max, tp->snd_una) && 974 !TCP_TIMER_ISARMED(tp, TCPT_REXMT) && 975 !TCP_TIMER_ISARMED(tp, TCPT_PERSIST)) { 976 TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur); 977 goto just_return; 978 } 979 980 /* 981 * TCP window updates are not reliable, rather a polling protocol 982 * using ``persist'' packets is used to insure receipt of window 983 * updates. The three ``states'' for the output side are: 984 * idle not doing retransmits or persists 985 * persisting to move a small or zero window 986 * (re)transmitting and thereby not persisting 987 * 988 * tp->t_timer[TCPT_PERSIST] 989 * is set when we are in persist state. 990 * tp->t_force 991 * is set when we are called to send a persist packet. 992 * tp->t_timer[TCPT_REXMT] 993 * is set when we are retransmitting 994 * The output side is idle when both timers are zero. 995 * 996 * If send window is too small, there is data to transmit, and no 997 * retransmit or persist is pending, then go to persist state. 998 * If nothing happens soon, send when timer expires: 999 * if window is nonzero, transmit what we can, 1000 * otherwise force out a byte. 1001 */ 1002 if (so->so_snd.sb_cc && TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 && 1003 TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) { 1004 tp->t_rxtshift = 0; 1005 tcp_setpersist(tp); 1006 } 1007 1008 /* 1009 * No reason to send a segment, just return. 1010 */ 1011 just_return: 1012 TCP_REASS_UNLOCK(tp); 1013 return (0); 1014 1015 send: 1016 /* 1017 * Before ESTABLISHED, force sending of initial options 1018 * unless TCP set not to do any options. 1019 * NOTE: we assume that the IP/TCP header plus TCP options 1020 * always fit in a single mbuf, leaving room for a maximum 1021 * link header, i.e. 1022 * max_linkhdr + sizeof (struct tcpiphdr) + optlen <= MCLBYTES 1023 */ 1024 optlen = 0; 1025 switch (af) { 1026 #ifdef INET 1027 case AF_INET: 1028 iphdrlen = sizeof(struct ip) + sizeof(struct tcphdr); 1029 break; 1030 #endif 1031 #ifdef INET6 1032 case AF_INET6: 1033 iphdrlen = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); 1034 break; 1035 #endif 1036 default: /*pacify gcc*/ 1037 iphdrlen = 0; 1038 break; 1039 } 1040 hdrlen = iphdrlen; 1041 if (flags & TH_SYN) { 1042 struct rtentry *rt; 1043 1044 rt = NULL; 1045 #ifdef INET 1046 if (tp->t_inpcb) 1047 rt = in_pcbrtentry(tp->t_inpcb); 1048 #endif 1049 #ifdef INET6 1050 if (tp->t_in6pcb) 1051 rt = in6_pcbrtentry(tp->t_in6pcb); 1052 #endif 1053 1054 tp->snd_nxt = tp->iss; 1055 tp->t_ourmss = tcp_mss_to_advertise(rt != NULL ? 1056 rt->rt_ifp : NULL, af); 1057 if ((tp->t_flags & TF_NOOPT) == 0) { 1058 opt[0] = TCPOPT_MAXSEG; 1059 opt[1] = 4; 1060 opt[2] = (tp->t_ourmss >> 8) & 0xff; 1061 opt[3] = tp->t_ourmss & 0xff; 1062 optlen = 4; 1063 1064 if ((tp->t_flags & TF_REQ_SCALE) && 1065 ((flags & TH_ACK) == 0 || 1066 (tp->t_flags & TF_RCVD_SCALE))) { 1067 *((u_int32_t *) (opt + optlen)) = htonl( 1068 TCPOPT_NOP << 24 | 1069 TCPOPT_WINDOW << 16 | 1070 TCPOLEN_WINDOW << 8 | 1071 tp->request_r_scale); 1072 optlen += 4; 1073 } 1074 if (tcp_do_sack) { 1075 u_int8_t *cp = (u_int8_t *)(opt + optlen); 1076 1077 cp[0] = TCPOPT_SACK_PERMITTED; 1078 cp[1] = 2; 1079 cp[2] = TCPOPT_NOP; 1080 cp[3] = TCPOPT_NOP; 1081 optlen += 4; 1082 } 1083 } 1084 } 1085 1086 /* 1087 * Send a timestamp and echo-reply if this is a SYN and our side 1088 * wants to use timestamps (TF_REQ_TSTMP is set) or both our side 1089 * and our peer have sent timestamps in our SYN's. 1090 */ 1091 if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP && 1092 (flags & TH_RST) == 0 && 1093 ((flags & (TH_SYN|TH_ACK)) == TH_SYN || 1094 (tp->t_flags & TF_RCVD_TSTMP))) { 1095 u_int32_t *lp = (u_int32_t *)(opt + optlen); 1096 1097 /* Form timestamp option as shown in appendix A of RFC 1323. */ 1098 *lp++ = htonl(TCPOPT_TSTAMP_HDR); 1099 *lp++ = htonl(TCP_TIMESTAMP(tp)); 1100 *lp = htonl(tp->ts_recent); 1101 optlen += TCPOLEN_TSTAMP_APPA; 1102 } 1103 1104 /* 1105 * Tack on the SACK block if it is necessary. 1106 */ 1107 if (sack_numblks) { 1108 int sack_len; 1109 u_char *bp = (u_char *)(opt + optlen); 1110 u_int32_t *lp = (u_int32_t *)(bp + 4); 1111 struct ipqent *tiqe; 1112 1113 sack_len = sack_numblks * 8 + 2; 1114 bp[0] = TCPOPT_NOP; 1115 bp[1] = TCPOPT_NOP; 1116 bp[2] = TCPOPT_SACK; 1117 bp[3] = sack_len; 1118 if ((tp->rcv_sack_flags & TCPSACK_HAVED) != 0) { 1119 sack_numblks--; 1120 *lp++ = htonl(tp->rcv_dsack_block.left); 1121 *lp++ = htonl(tp->rcv_dsack_block.right); 1122 tp->rcv_sack_flags &= ~TCPSACK_HAVED; 1123 } 1124 for (tiqe = TAILQ_FIRST(&tp->timeq); 1125 sack_numblks > 0; tiqe = TAILQ_NEXT(tiqe, ipqe_timeq)) { 1126 KASSERT(tiqe != NULL); 1127 sack_numblks--; 1128 *lp++ = htonl(tiqe->ipqe_seq); 1129 *lp++ = htonl(tiqe->ipqe_seq + tiqe->ipqe_len + 1130 ((tiqe->ipqe_flags & TH_FIN) != 0 ? 1 : 0)); 1131 } 1132 optlen += sack_len + 2; 1133 } 1134 TCP_REASS_UNLOCK(tp); 1135 1136 #ifdef TCP_SIGNATURE 1137 #if defined(INET6) && defined(FAST_IPSEC) 1138 if (tp->t_family == AF_INET) 1139 #endif 1140 if (tp->t_flags & TF_SIGNATURE) { 1141 u_char *bp; 1142 /* 1143 * Initialize TCP-MD5 option (RFC2385) 1144 */ 1145 bp = (u_char *)opt + optlen; 1146 *bp++ = TCPOPT_SIGNATURE; 1147 *bp++ = TCPOLEN_SIGNATURE; 1148 sigoff = optlen + 2; 1149 bzero(bp, TCP_SIGLEN); 1150 bp += TCP_SIGLEN; 1151 optlen += TCPOLEN_SIGNATURE; 1152 /* 1153 * Terminate options list and maintain 32-bit alignment. 1154 */ 1155 *bp++ = TCPOPT_NOP; 1156 *bp++ = TCPOPT_EOL; 1157 optlen += 2; 1158 } 1159 #endif /* TCP_SIGNATURE */ 1160 1161 hdrlen += optlen; 1162 1163 #ifdef DIAGNOSTIC 1164 if (!use_tso && len > txsegsize) 1165 panic("tcp data to be sent is larger than segment"); 1166 else if (use_tso && len > IP_MAXPACKET) 1167 panic("tcp data to be sent is larger than max TSO size"); 1168 if (max_linkhdr + hdrlen > MCLBYTES) 1169 panic("tcphdr too big"); 1170 #endif 1171 1172 /* 1173 * Grab a header mbuf, attaching a copy of data to 1174 * be transmitted, and initialize the header from 1175 * the template for sends on this connection. 1176 */ 1177 if (len) { 1178 error = tcp_build_datapkt(tp, so, off, len, hdrlen, &m); 1179 if (error) 1180 goto out; 1181 /* 1182 * If we're sending everything we've got, set PUSH. 1183 * (This will keep happy those implementations which only 1184 * give data to the user when a buffer fills or 1185 * a PUSH comes in.) 1186 */ 1187 if (off + len == so->so_snd.sb_cc) 1188 flags |= TH_PUSH; 1189 } else { 1190 if (tp->t_flags & TF_ACKNOW) 1191 tcpstat.tcps_sndacks++; 1192 else if (flags & (TH_SYN|TH_FIN|TH_RST)) 1193 tcpstat.tcps_sndctrl++; 1194 else if (SEQ_GT(tp->snd_up, tp->snd_una)) 1195 tcpstat.tcps_sndurg++; 1196 else 1197 tcpstat.tcps_sndwinup++; 1198 1199 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1200 if (m != NULL && max_linkhdr + hdrlen > MHLEN) { 1201 MCLGET(m, M_DONTWAIT); 1202 if ((m->m_flags & M_EXT) == 0) { 1203 m_freem(m); 1204 m = NULL; 1205 } 1206 } 1207 if (m == NULL) { 1208 error = ENOBUFS; 1209 goto out; 1210 } 1211 MCLAIM(m, &tcp_tx_mowner); 1212 m->m_data += max_linkhdr; 1213 m->m_len = hdrlen; 1214 } 1215 m->m_pkthdr.rcvif = (struct ifnet *)0; 1216 switch (af) { 1217 #ifdef INET 1218 case AF_INET: 1219 ip = mtod(m, struct ip *); 1220 #ifdef INET6 1221 ip6 = NULL; 1222 #endif 1223 th = (struct tcphdr *)(ip + 1); 1224 break; 1225 #endif 1226 #ifdef INET6 1227 case AF_INET6: 1228 ip = NULL; 1229 ip6 = mtod(m, struct ip6_hdr *); 1230 th = (struct tcphdr *)(ip6 + 1); 1231 break; 1232 #endif 1233 default: /*pacify gcc*/ 1234 ip = NULL; 1235 #ifdef INET6 1236 ip6 = NULL; 1237 #endif 1238 th = NULL; 1239 break; 1240 } 1241 if (tp->t_template == 0) 1242 panic("tcp_output"); 1243 if (tp->t_template->m_len < iphdrlen) 1244 panic("tcp_output"); 1245 bcopy(mtod(tp->t_template, caddr_t), mtod(m, caddr_t), iphdrlen); 1246 1247 /* 1248 * If we are starting a connection, send ECN setup 1249 * SYN packet. If we are on a retransmit, we may 1250 * resend those bits a number of times as per 1251 * RFC 3168. 1252 */ 1253 if (tp->t_state == TCPS_SYN_SENT && tcp_do_ecn) { 1254 if (tp->t_flags & TF_SYN_REXMT) { 1255 if (tp->t_ecn_retries--) 1256 flags |= TH_ECE|TH_CWR; 1257 } else { 1258 flags |= TH_ECE|TH_CWR; 1259 tp->t_ecn_retries = tcp_ecn_maxretries; 1260 } 1261 } 1262 1263 if (TCP_ECN_ALLOWED(tp)) { 1264 /* 1265 * If the peer has ECN, mark data packets 1266 * ECN capable. Ignore pure ack packets, retransmissions 1267 * and window probes. 1268 */ 1269 if (len > 0 && SEQ_GEQ(tp->snd_nxt, tp->snd_max) && 1270 !(tp->t_force && len == 1)) { 1271 switch (af) { 1272 #ifdef INET 1273 case AF_INET: 1274 tp->t_inpcb->inp_ip.ip_tos |= IPTOS_ECN_ECT0; 1275 break; 1276 #endif 1277 #ifdef INET6 1278 case AF_INET6: 1279 ip6->ip6_flow |= htonl(IPTOS_ECN_ECT0 << 20); 1280 break; 1281 #endif 1282 } 1283 tcpstat.tcps_ecn_ect++; 1284 } 1285 1286 /* 1287 * Reply with proper ECN notifications. 1288 */ 1289 if (tp->t_flags & TF_ECN_SND_CWR) { 1290 flags |= TH_CWR; 1291 tp->t_flags &= ~TF_ECN_SND_CWR; 1292 } 1293 if (tp->t_flags & TF_ECN_SND_ECE) { 1294 flags |= TH_ECE; 1295 } 1296 } 1297 1298 1299 /* 1300 * If we are doing retransmissions, then snd_nxt will 1301 * not reflect the first unsent octet. For ACK only 1302 * packets, we do not want the sequence number of the 1303 * retransmitted packet, we want the sequence number 1304 * of the next unsent octet. So, if there is no data 1305 * (and no SYN or FIN), use snd_max instead of snd_nxt 1306 * when filling in ti_seq. But if we are in persist 1307 * state, snd_max might reflect one byte beyond the 1308 * right edge of the window, so use snd_nxt in that 1309 * case, since we know we aren't doing a retransmission. 1310 * (retransmit and persist are mutually exclusive...) 1311 */ 1312 if (TCP_SACK_ENABLED(tp) && sack_rxmit) { 1313 th->th_seq = htonl(p->rxmit); 1314 p->rxmit += len; 1315 } else { 1316 if (len || (flags & (TH_SYN|TH_FIN)) || 1317 TCP_TIMER_ISARMED(tp, TCPT_PERSIST)) 1318 th->th_seq = htonl(tp->snd_nxt); 1319 else 1320 th->th_seq = htonl(tp->snd_max); 1321 } 1322 th->th_ack = htonl(tp->rcv_nxt); 1323 if (optlen) { 1324 bcopy((caddr_t)opt, (caddr_t)(th + 1), optlen); 1325 th->th_off = (sizeof (struct tcphdr) + optlen) >> 2; 1326 } 1327 th->th_flags = flags; 1328 /* 1329 * Calculate receive window. Don't shrink window, 1330 * but avoid silly window syndrome. 1331 */ 1332 if (win < (long)(so->so_rcv.sb_hiwat / 4) && win < (long)rxsegsize) 1333 win = 0; 1334 if (win > (long)TCP_MAXWIN << tp->rcv_scale) 1335 win = (long)TCP_MAXWIN << tp->rcv_scale; 1336 if (win < (long)(int32_t)(tp->rcv_adv - tp->rcv_nxt)) 1337 win = (long)(int32_t)(tp->rcv_adv - tp->rcv_nxt); 1338 th->th_win = htons((u_int16_t) (win>>tp->rcv_scale)); 1339 if (SEQ_GT(tp->snd_up, tp->snd_nxt)) { 1340 u_int32_t urp = tp->snd_up - tp->snd_nxt; 1341 if (urp > IP_MAXPACKET) 1342 urp = IP_MAXPACKET; 1343 th->th_urp = htons((u_int16_t)urp); 1344 th->th_flags |= TH_URG; 1345 } else 1346 /* 1347 * If no urgent pointer to send, then we pull 1348 * the urgent pointer to the left edge of the send window 1349 * so that it doesn't drift into the send window on sequence 1350 * number wraparound. 1351 */ 1352 tp->snd_up = tp->snd_una; /* drag it along */ 1353 1354 #ifdef TCP_SIGNATURE 1355 #if defined(INET6) && defined(FAST_IPSEC) 1356 if (tp->t_family == AF_INET) /* XXX */ 1357 #endif 1358 if (sigoff && (tp->t_flags & TF_SIGNATURE)) { 1359 struct secasvar *sav; 1360 u_int8_t *sigp; 1361 1362 sav = tcp_signature_getsav(m, th); 1363 1364 if (sav == NULL) { 1365 if (m) 1366 m_freem(m); 1367 return (EPERM); 1368 } 1369 1370 m->m_pkthdr.len = hdrlen + len; 1371 sigp = (caddr_t)th + sizeof(*th) + sigoff; 1372 tcp_signature(m, th, (caddr_t)th - mtod(m, caddr_t), sav, sigp); 1373 1374 key_sa_recordxfer(sav, m); 1375 #ifdef FAST_IPSEC 1376 KEY_FREESAV(&sav); 1377 #else 1378 key_freesav(sav); 1379 #endif 1380 } 1381 #endif 1382 1383 /* 1384 * Set ourselves up to be checksummed just before the packet 1385 * hits the wire. 1386 */ 1387 switch (af) { 1388 #ifdef INET 1389 case AF_INET: 1390 m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); 1391 if (use_tso) { 1392 m->m_pkthdr.segsz = txsegsize; 1393 m->m_pkthdr.csum_flags = M_CSUM_TSOv4; 1394 } else { 1395 m->m_pkthdr.csum_flags = M_CSUM_TCPv4; 1396 if (len + optlen) { 1397 /* Fixup the pseudo-header checksum. */ 1398 /* XXXJRT Not IP Jumbogram safe. */ 1399 th->th_sum = in_cksum_addword(th->th_sum, 1400 htons((u_int16_t) (len + optlen))); 1401 } 1402 } 1403 break; 1404 #endif 1405 #ifdef INET6 1406 case AF_INET6: 1407 m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); 1408 m->m_pkthdr.csum_flags = M_CSUM_TCPv6; 1409 if (len + optlen) { 1410 /* Fixup the pseudo-header checksum. */ 1411 /* XXXJRT: Not IPv6 Jumbogram safe. */ 1412 th->th_sum = in_cksum_addword(th->th_sum, 1413 htons((u_int16_t) (len + optlen))); 1414 } 1415 break; 1416 #endif 1417 } 1418 1419 /* 1420 * In transmit state, time the transmission and arrange for 1421 * the retransmit. In persist state, just set snd_max. 1422 */ 1423 if (tp->t_force == 0 || TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) { 1424 tcp_seq startseq = tp->snd_nxt; 1425 1426 /* 1427 * Advance snd_nxt over sequence space of this segment. 1428 * There are no states in which we send both a SYN and a FIN, 1429 * so we collapse the tests for these flags. 1430 */ 1431 if (flags & (TH_SYN|TH_FIN)) 1432 tp->snd_nxt++; 1433 if (sack_rxmit) 1434 goto timer; 1435 tp->snd_nxt += len; 1436 if (SEQ_GT(tp->snd_nxt, tp->snd_max)) { 1437 tp->snd_max = tp->snd_nxt; 1438 /* 1439 * Time this transmission if not a retransmission and 1440 * not currently timing anything. 1441 */ 1442 if (tp->t_rtttime == 0) { 1443 tp->t_rtttime = tcp_now; 1444 tp->t_rtseq = startseq; 1445 tcpstat.tcps_segstimed++; 1446 } 1447 } 1448 1449 /* 1450 * Set retransmit timer if not currently set, 1451 * and not doing an ack or a keep-alive probe. 1452 * Initial value for retransmit timer is smoothed 1453 * round-trip time + 2 * round-trip time variance. 1454 * Initialize shift counter which is used for backoff 1455 * of retransmit time. 1456 */ 1457 timer: 1458 if (TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 && 1459 ((sack_rxmit && tp->snd_nxt != tp->snd_max) || 1460 tp->snd_nxt != tp->snd_una)) { 1461 if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST)) { 1462 TCP_TIMER_DISARM(tp, TCPT_PERSIST); 1463 tp->t_rxtshift = 0; 1464 } 1465 TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur); 1466 } 1467 } else 1468 if (SEQ_GT(tp->snd_nxt + len, tp->snd_max)) 1469 tp->snd_max = tp->snd_nxt + len; 1470 1471 #ifdef TCP_DEBUG 1472 /* 1473 * Trace. 1474 */ 1475 if (so->so_options & SO_DEBUG) 1476 tcp_trace(TA_OUTPUT, tp->t_state, tp, m, 0); 1477 #endif 1478 1479 /* 1480 * Fill in IP length and desired time to live and 1481 * send to IP level. There should be a better way 1482 * to handle ttl and tos; we could keep them in 1483 * the template, but need a way to checksum without them. 1484 */ 1485 m->m_pkthdr.len = hdrlen + len; 1486 1487 switch (af) { 1488 #ifdef INET 1489 case AF_INET: 1490 ip->ip_len = htons(m->m_pkthdr.len); 1491 packetlen = m->m_pkthdr.len; 1492 if (tp->t_inpcb) { 1493 ip->ip_ttl = tp->t_inpcb->inp_ip.ip_ttl; 1494 ip->ip_tos = tp->t_inpcb->inp_ip.ip_tos; 1495 } 1496 #ifdef INET6 1497 else if (tp->t_in6pcb) { 1498 ip->ip_ttl = in6_selecthlim(tp->t_in6pcb, NULL); /*XXX*/ 1499 ip->ip_tos = 0; /*XXX*/ 1500 } 1501 #endif 1502 break; 1503 #endif 1504 #ifdef INET6 1505 case AF_INET6: 1506 packetlen = m->m_pkthdr.len; 1507 ip6->ip6_nxt = IPPROTO_TCP; 1508 if (tp->t_in6pcb) { 1509 /* 1510 * we separately set hoplimit for every segment, since 1511 * the user might want to change the value via 1512 * setsockopt. Also, desired default hop limit might 1513 * be changed via Neighbor Discovery. 1514 */ 1515 ip6->ip6_hlim = in6_selecthlim(tp->t_in6pcb, 1516 ro->ro_rt ? ro->ro_rt->rt_ifp : NULL); 1517 } 1518 /* ip6->ip6_flow = ??? */ 1519 /* ip6_plen will be filled in ip6_output(). */ 1520 break; 1521 #endif 1522 default: /*pacify gcc*/ 1523 packetlen = 0; 1524 break; 1525 } 1526 1527 switch (af) { 1528 #ifdef INET 1529 case AF_INET: 1530 { 1531 struct mbuf *opts; 1532 1533 if (tp->t_inpcb) 1534 opts = tp->t_inpcb->inp_options; 1535 else 1536 opts = NULL; 1537 error = ip_output(m, opts, ro, 1538 (tp->t_mtudisc ? IP_MTUDISC : 0) | 1539 (so->so_options & SO_DONTROUTE), 1540 (struct ip_moptions *)0, so); 1541 break; 1542 } 1543 #endif 1544 #ifdef INET6 1545 case AF_INET6: 1546 { 1547 struct ip6_pktopts *opts; 1548 1549 if (tp->t_in6pcb) 1550 opts = tp->t_in6pcb->in6p_outputopts; 1551 else 1552 opts = NULL; 1553 error = ip6_output(m, opts, (struct route_in6 *)ro, 1554 so->so_options & SO_DONTROUTE, 1555 (struct ip6_moptions *)0, so, NULL); 1556 break; 1557 } 1558 #endif 1559 default: 1560 error = EAFNOSUPPORT; 1561 break; 1562 } 1563 if (error) { 1564 out: 1565 if (error == ENOBUFS) { 1566 tcpstat.tcps_selfquench++; 1567 #ifdef INET 1568 if (tp->t_inpcb) 1569 tcp_quench(tp->t_inpcb, 0); 1570 #endif 1571 #ifdef INET6 1572 if (tp->t_in6pcb) 1573 tcp6_quench(tp->t_in6pcb, 0); 1574 #endif 1575 error = 0; 1576 } else if ((error == EHOSTUNREACH || error == ENETDOWN) && 1577 TCPS_HAVERCVDSYN(tp->t_state)) { 1578 tp->t_softerror = error; 1579 error = 0; 1580 } 1581 1582 /* Back out the seqence number advance. */ 1583 if (sack_rxmit) 1584 p->rxmit -= len; 1585 1586 /* Restart the delayed ACK timer, if necessary. */ 1587 if (tp->t_flags & TF_DELACK) 1588 TCP_RESTART_DELACK(tp); 1589 1590 return (error); 1591 } 1592 1593 if (packetlen > tp->t_pmtud_mtu_sent) 1594 tp->t_pmtud_mtu_sent = packetlen; 1595 1596 tcpstat.tcps_sndtotal++; 1597 if (tp->t_flags & TF_DELACK) 1598 tcpstat.tcps_delack++; 1599 1600 /* 1601 * Data sent (as far as we can tell). 1602 * If this advertises a larger window than any other segment, 1603 * then remember the size of the advertised window. 1604 * Any pending ACK has now been sent. 1605 */ 1606 if (win > 0 && SEQ_GT(tp->rcv_nxt+win, tp->rcv_adv)) 1607 tp->rcv_adv = tp->rcv_nxt + win; 1608 tp->last_ack_sent = tp->rcv_nxt; 1609 tp->t_flags &= ~TF_ACKNOW; 1610 TCP_CLEAR_DELACK(tp); 1611 #ifdef DIAGNOSTIC 1612 if (maxburst < 0) 1613 printf("tcp_output: maxburst exceeded by %d\n", -maxburst); 1614 #endif 1615 if (sendalot && (tp->t_congctl == &tcp_reno_ctl || --maxburst)) 1616 goto again; 1617 return (0); 1618 } 1619 1620 void 1621 tcp_setpersist(struct tcpcb *tp) 1622 { 1623 int t = ((tp->t_srtt >> 2) + tp->t_rttvar) >> (1 + 2); 1624 int nticks; 1625 1626 if (TCP_TIMER_ISARMED(tp, TCPT_REXMT)) 1627 panic("tcp_output REXMT"); 1628 /* 1629 * Start/restart persistance timer. 1630 */ 1631 if (t < tp->t_rttmin) 1632 t = tp->t_rttmin; 1633 TCPT_RANGESET(nticks, t * tcp_backoff[tp->t_rxtshift], 1634 TCPTV_PERSMIN, TCPTV_PERSMAX); 1635 TCP_TIMER_ARM(tp, TCPT_PERSIST, nticks); 1636 if (tp->t_rxtshift < TCP_MAXRXTSHIFT) 1637 tp->t_rxtshift++; 1638 } 1639 1640 #if defined(INET) 1641 /* 1642 * tcp4_segment: handle M_CSUM_TSOv4 by software. 1643 * 1644 * => always consume m. 1645 * => call output_func with output_arg for each segments. 1646 */ 1647 1648 int 1649 tcp4_segment(struct mbuf *m, int (*output_func)(void *, struct mbuf *), 1650 void *output_arg) 1651 { 1652 int mss; 1653 int iphlen; 1654 int thlen; 1655 int hlen; 1656 int len; 1657 struct ip *iph; 1658 struct tcphdr *th; 1659 uint16_t ipid; 1660 uint32_t tcpseq; 1661 struct mbuf *hdr = NULL; 1662 struct mbuf *t; 1663 int error = 0; 1664 1665 KASSERT((m->m_flags & M_PKTHDR) != 0); 1666 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0); 1667 1668 m->m_pkthdr.csum_flags = 0; 1669 1670 len = m->m_pkthdr.len; 1671 KASSERT(len >= sizeof(*iph) + sizeof(*th)); 1672 1673 if (m->m_len < sizeof(*iph)) { 1674 m = m_pullup(m, sizeof(*iph)); 1675 if (m == NULL) { 1676 error = ENOMEM; 1677 goto quit; 1678 } 1679 } 1680 iph = mtod(m, struct ip *); 1681 iphlen = iph->ip_hl * 4; 1682 KASSERT(iph->ip_v == IPVERSION); 1683 KASSERT(iphlen >= sizeof(*iph)); 1684 KASSERT(iph->ip_p == IPPROTO_TCP); 1685 ipid = ntohs(iph->ip_id); 1686 1687 hlen = iphlen + sizeof(*th); 1688 if (m->m_len < hlen) { 1689 m = m_pullup(m, hlen); 1690 if (m == NULL) { 1691 error = ENOMEM; 1692 goto quit; 1693 } 1694 } 1695 th = (void *)(mtod(m, char *) + iphlen); 1696 tcpseq = ntohl(th->th_seq); 1697 thlen = th->th_off * 4; 1698 hlen = iphlen + thlen; 1699 1700 mss = m->m_pkthdr.segsz; 1701 KASSERT(mss != 0); 1702 KASSERT(len > hlen); 1703 1704 t = m_split(m, hlen, M_NOWAIT); 1705 if (t == NULL) { 1706 error = ENOMEM; 1707 goto quit; 1708 } 1709 hdr = m; 1710 m = t; 1711 len -= hlen; 1712 KASSERT(len % mss == 0); 1713 while (len > 0) { 1714 struct mbuf *n; 1715 1716 n = m_dup(hdr, 0, hlen, M_NOWAIT); 1717 if (n == NULL) { 1718 error = ENOMEM; 1719 goto quit; 1720 } 1721 KASSERT(n->m_len == hlen); /* XXX */ 1722 1723 t = m_split(m, mss, M_NOWAIT); 1724 if (t == NULL) { 1725 m_freem(n); 1726 error = ENOMEM; 1727 goto quit; 1728 } 1729 m_cat(n, m); 1730 m = t; 1731 1732 KASSERT(n->m_len >= hlen); /* XXX */ 1733 1734 n->m_pkthdr.len = hlen + mss; 1735 iph = mtod(n, struct ip *); 1736 KASSERT(iph->ip_v == IPVERSION); 1737 iph->ip_len = htons(n->m_pkthdr.len); 1738 iph->ip_id = htons(ipid); 1739 th = (void *)(mtod(n, char *) + iphlen); 1740 th->th_seq = htonl(tcpseq); 1741 iph->ip_sum = 0; 1742 iph->ip_sum = in_cksum(n, iphlen); 1743 th->th_sum = 0; 1744 th->th_sum = in4_cksum(n, IPPROTO_TCP, iphlen, thlen + mss); 1745 1746 error = (*output_func)(output_arg, n); 1747 if (error) { 1748 goto quit; 1749 } 1750 1751 tcpseq += mss; 1752 ipid++; 1753 len -= mss; 1754 } 1755 1756 quit: 1757 if (hdr != NULL) { 1758 m_freem(hdr); 1759 } 1760 if (m != NULL) { 1761 m_freem(m); 1762 } 1763 1764 return error; 1765 } 1766 #endif /* defined(INET) */ 1767
Indexes created Tue Sep 23 07:09:52 GMT 2025