1 /* $NetBSD: tcp_output.c,v 1.222 2024/09/08 09:36:52 rillig 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 * 93 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 94 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 95 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 96 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 97 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 98 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 99 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 100 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 101 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 102 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 103 * POSSIBILITY OF SUCH DAMAGE. 104 */ 105 106 /* 107 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 108 * The Regents of the University of California. All rights reserved. 109 * 110 * Redistribution and use in source and binary forms, with or without 111 * modification, are permitted provided that the following conditions 112 * are met: 113 * 1. Redistributions of source code must retain the above copyright 114 * notice, this list of conditions and the following disclaimer. 115 * 2. Redistributions in binary form must reproduce the above copyright 116 * notice, this list of conditions and the following disclaimer in the 117 * documentation and/or other materials provided with the distribution. 118 * 3. Neither the name of the University nor the names of its contributors 119 * may be used to endorse or promote products derived from this software 120 * without specific prior written permission. 121 * 122 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 123 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 124 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 125 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 126 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 127 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 128 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 129 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 130 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 131 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 132 * SUCH DAMAGE. 133 * 134 * @(#)tcp_output.c 8.4 (Berkeley) 5/24/95 135 */ 136 137 #include <sys/cdefs.h> 138 __KERNEL_RCSID(0, "$NetBSD: tcp_output.c,v 1.222 2024/09/08 09:36:52 rillig Exp $"); 139 140 #ifdef _KERNEL_OPT 141 #include "opt_inet.h" 142 #include "opt_ipsec.h" 143 #include "opt_tcp_debug.h" 144 #endif 145 146 #include <sys/param.h> 147 #include <sys/systm.h> 148 #include <sys/mbuf.h> 149 #include <sys/protosw.h> 150 #include <sys/socket.h> 151 #include <sys/socketvar.h> 152 #include <sys/errno.h> 153 #include <sys/domain.h> 154 #include <sys/kernel.h> 155 #ifdef TCP_SIGNATURE 156 #include <sys/md5.h> 157 #endif 158 159 #include <net/if.h> 160 #include <net/route.h> 161 162 #include <netinet/in.h> 163 #include <netinet/in_systm.h> 164 #include <netinet/ip.h> 165 #include <netinet/in_pcb.h> 166 #include <netinet/ip_var.h> 167 168 #ifdef INET6 169 #include <netinet/ip6.h> 170 #include <netinet6/in6_var.h> 171 #include <netinet6/ip6_var.h> 172 #include <netinet6/in6_pcb.h> 173 #include <netinet6/nd6.h> 174 #endif 175 176 #ifdef IPSEC 177 #include <netipsec/ipsec.h> 178 #include <netipsec/key.h> 179 #ifdef INET6 180 #include <netipsec/ipsec6.h> 181 #endif 182 #endif 183 184 #include <netinet/tcp.h> 185 #define TCPOUTFLAGS 186 #include <netinet/tcp_fsm.h> 187 #include <netinet/tcp_seq.h> 188 #include <netinet/tcp_timer.h> 189 #include <netinet/tcp_var.h> 190 #include <netinet/tcp_private.h> 191 #include <netinet/tcp_congctl.h> 192 #include <netinet/tcp_debug.h> 193 #include <netinet/in_offload.h> 194 #include <netinet6/in6_offload.h> 195 196 /* 197 * Knob to enable Congestion Window Monitoring, and control 198 * the burst size it allows. Default burst is 4 packets, per 199 * the Internet draft. 200 */ 201 int tcp_cwm = 0; 202 int tcp_cwm_burstsize = 4; 203 204 int tcp_do_autosndbuf = 1; 205 int tcp_autosndbuf_inc = 8 * 1024; 206 int tcp_autosndbuf_max = 256 * 1024; 207 208 #ifdef TCP_OUTPUT_COUNTERS 209 #include <sys/device.h> 210 211 extern struct evcnt tcp_output_bigheader; 212 extern struct evcnt tcp_output_predict_hit; 213 extern struct evcnt tcp_output_predict_miss; 214 extern struct evcnt tcp_output_copysmall; 215 extern struct evcnt tcp_output_copybig; 216 extern struct evcnt tcp_output_refbig; 217 218 #define TCP_OUTPUT_COUNTER_INCR(ev) (ev)->ev_count++ 219 #else 220 221 #define TCP_OUTPUT_COUNTER_INCR(ev) /* nothing */ 222 223 #endif /* TCP_OUTPUT_COUNTERS */ 224 225 static int 226 tcp_segsize(struct tcpcb *tp, int *txsegsizep, int *rxsegsizep, 227 bool *alwaysfragp) 228 { 229 struct inpcb *inp = tp->t_inpcb; 230 struct socket *so = NULL; 231 struct rtentry *rt; 232 struct ifnet *ifp; 233 int size; 234 int hdrlen; 235 int optlen; 236 237 *alwaysfragp = false; 238 size = tcp_mssdflt; 239 240 switch (tp->t_family) { 241 case AF_INET: 242 hdrlen = sizeof(struct ip) + sizeof(struct tcphdr); 243 break; 244 #ifdef INET6 245 case AF_INET6: 246 hdrlen = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); 247 break; 248 #endif 249 default: 250 hdrlen = 1; /* prevent zero sized segments */ 251 goto out; 252 } 253 254 rt = inpcb_rtentry(inp); 255 so = inp->inp_socket; 256 if (rt == NULL) { 257 goto out; 258 } 259 260 ifp = rt->rt_ifp; 261 262 if (tp->t_mtudisc && rt->rt_rmx.rmx_mtu != 0) { 263 #ifdef INET6 264 if (inp->inp_af == AF_INET6 && rt->rt_rmx.rmx_mtu < IPV6_MMTU) { 265 /* 266 * RFC2460 section 5, last paragraph: if path MTU is 267 * smaller than 1280, use 1280 as packet size and 268 * attach fragment header. 269 */ 270 size = IPV6_MMTU - hdrlen - sizeof(struct ip6_frag); 271 *alwaysfragp = true; 272 } else 273 size = rt->rt_rmx.rmx_mtu - hdrlen; 274 #else 275 size = rt->rt_rmx.rmx_mtu - hdrlen; 276 #endif 277 } else if (ifp->if_flags & IFF_LOOPBACK) 278 size = ifp->if_mtu - hdrlen; 279 else if (inp->inp_af == AF_INET && tp->t_mtudisc) 280 size = ifp->if_mtu - hdrlen; 281 else if (inp->inp_af == AF_INET && in_localaddr(in4p_faddr(inp))) 282 size = ifp->if_mtu - hdrlen; 283 #ifdef INET6 284 else if (inp->inp_af == AF_INET6) { 285 if (IN6_IS_ADDR_V4MAPPED(&in6p_faddr(inp))) { 286 /* mapped addr case */ 287 struct in_addr d; 288 memcpy(&d, &in6p_faddr(inp).s6_addr32[3], sizeof(d)); 289 if (tp->t_mtudisc || in_localaddr(d)) 290 size = ifp->if_mtu - hdrlen; 291 } else { 292 /* 293 * for IPv6, path MTU discovery is always turned on, 294 * or the node must use packet size <= 1280. 295 */ 296 size = tp->t_mtudisc ? ifp->if_mtu : IPV6_MMTU; 297 size -= hdrlen; 298 } 299 } 300 #endif 301 inpcb_rtentry_unref(rt, inp); 302 out: 303 /* 304 * Now we must make room for whatever extra TCP/IP options are in 305 * the packet. 306 */ 307 optlen = tcp_optlen(tp); 308 309 /* 310 * XXX tp->t_ourmss should have the right size, but without this code 311 * fragmentation will occur... need more investigation 312 */ 313 314 if (inp->inp_af == AF_INET) { 315 #if defined(IPSEC) 316 if (ipsec_used && 317 !ipsec_pcb_skip_ipsec(inp->inp_sp, IPSEC_DIR_OUTBOUND)) 318 optlen += ipsec4_hdrsiz_tcp(tp); 319 #endif 320 optlen += ip_optlen(inp); 321 } 322 323 #ifdef INET6 324 if (inp->inp_af == AF_INET6 && tp->t_family == AF_INET) { 325 #if defined(IPSEC) 326 if (ipsec_used && 327 !ipsec_pcb_skip_ipsec(inp->inp_sp, IPSEC_DIR_OUTBOUND)) 328 optlen += ipsec4_hdrsiz_tcp(tp); 329 #endif 330 /* XXX size -= ip_optlen(in6p); */ 331 } else if (inp->inp_af == AF_INET6) { 332 #if defined(IPSEC) 333 if (ipsec_used && 334 !ipsec_pcb_skip_ipsec(inp->inp_sp, IPSEC_DIR_OUTBOUND)) 335 optlen += ipsec6_hdrsiz_tcp(tp); 336 #endif 337 optlen += ip6_optlen(inp); 338 } 339 #endif 340 size -= optlen; 341 342 /* 343 * There may not be any room for data if mtu is too small. This 344 * includes zero-sized. 345 */ 346 if (size <= 0) { 347 return EMSGSIZE; 348 } 349 350 /* 351 * *rxsegsizep holds *estimated* inbound segment size (estimation 352 * assumes that path MTU is the same for both ways). this is only 353 * for silly window avoidance, do not use the value for other purposes. 354 * 355 * ipseclen is subtracted from both sides, this may not be right. 356 * I'm not quite sure about this (could someone comment). 357 */ 358 *txsegsizep = uimin(tp->t_peermss - optlen, size); 359 *rxsegsizep = uimin(tp->t_ourmss - optlen, size); 360 361 /* 362 * Never send more than half a buffer full. This insures that we can 363 * always keep 2 packets on the wire, no matter what SO_SNDBUF is, and 364 * therefore acks will never be delayed unless we run out of data to 365 * transmit. 366 */ 367 if (so) { 368 *txsegsizep = uimin(so->so_snd.sb_hiwat >> 1, *txsegsizep); 369 } 370 371 /* 372 * A segment must at least store header + options 373 */ 374 if (*txsegsizep < hdrlen + optlen) { 375 return EMSGSIZE; 376 } 377 378 if (*txsegsizep != tp->t_segsz) { 379 /* 380 * If the new segment size is larger, we don't want to 381 * mess up the congestion window, but if it is smaller 382 * we'll have to reduce the congestion window to ensure 383 * that we don't get into trouble with initial windows 384 * and the rest. In any case, if the segment size 385 * has changed, chances are the path has, too, and 386 * our congestion window will be different. 387 */ 388 if (*txsegsizep < tp->t_segsz) { 389 tp->snd_cwnd = uimax((tp->snd_cwnd / tp->t_segsz) 390 * *txsegsizep, *txsegsizep); 391 tp->snd_ssthresh = uimax((tp->snd_ssthresh / tp->t_segsz) 392 * *txsegsizep, *txsegsizep); 393 } 394 tp->t_segsz = *txsegsizep; 395 } 396 397 return 0; 398 } 399 400 static int 401 tcp_build_datapkt(struct tcpcb *tp, struct socket *so, int off, 402 long len, int hdrlen, struct mbuf **mp) 403 { 404 struct mbuf *m, *m0; 405 net_stat_ref_t tcps; 406 407 tcps = TCP_STAT_GETREF(); 408 if (tp->t_force && len == 1) 409 _NET_STATINC_REF(tcps, TCP_STAT_SNDPROBE); 410 else if (SEQ_LT(tp->snd_nxt, tp->snd_max)) { 411 tp->t_sndrexmitpack++; 412 _NET_STATINC_REF(tcps, TCP_STAT_SNDREXMITPACK); 413 _NET_STATADD_REF(tcps, TCP_STAT_SNDREXMITBYTE, len); 414 } else { 415 _NET_STATINC_REF(tcps, TCP_STAT_SNDPACK); 416 _NET_STATADD_REF(tcps, TCP_STAT_SNDBYTE, len); 417 } 418 TCP_STAT_PUTREF(); 419 420 MGETHDR(m, M_DONTWAIT, MT_HEADER); 421 if (__predict_false(m == NULL)) 422 return ENOBUFS; 423 MCLAIM(m, &tcp_tx_mowner); 424 425 /* 426 * XXX Because other code assumes headers will fit in 427 * XXX one header mbuf. 428 * 429 * (This code should almost *never* be run.) 430 */ 431 if (__predict_false((max_linkhdr + hdrlen) > MHLEN)) { 432 TCP_OUTPUT_COUNTER_INCR(&tcp_output_bigheader); 433 MCLGET(m, M_DONTWAIT); 434 if ((m->m_flags & M_EXT) == 0) { 435 m_freem(m); 436 return ENOBUFS; 437 } 438 } 439 440 m->m_data += max_linkhdr; 441 m->m_len = hdrlen; 442 443 /* 444 * To avoid traversing the whole sb_mb chain for correct 445 * data to send, remember last sent mbuf, its offset and 446 * the sent size. When called the next time, see if the 447 * data to send is directly following the previous transfer. 448 * This is important for large TCP windows. 449 */ 450 if (off == 0 || tp->t_lastm == NULL || 451 (tp->t_lastoff + tp->t_lastlen) != off) { 452 TCP_OUTPUT_COUNTER_INCR(&tcp_output_predict_miss); 453 /* 454 * Either a new packet or a retransmit. 455 * Start from the beginning. 456 */ 457 tp->t_lastm = so->so_snd.sb_mb; 458 tp->t_inoff = off; 459 } else { 460 TCP_OUTPUT_COUNTER_INCR(&tcp_output_predict_hit); 461 tp->t_inoff += tp->t_lastlen; 462 } 463 464 /* Traverse forward to next packet */ 465 while (tp->t_inoff > 0) { 466 if (tp->t_lastm == NULL) 467 panic("tp->t_lastm == NULL"); 468 if (tp->t_inoff < tp->t_lastm->m_len) 469 break; 470 tp->t_inoff -= tp->t_lastm->m_len; 471 tp->t_lastm = tp->t_lastm->m_next; 472 } 473 474 tp->t_lastoff = off; 475 tp->t_lastlen = len; 476 m0 = tp->t_lastm; 477 off = tp->t_inoff; 478 479 if (len <= M_TRAILINGSPACE(m)) { 480 m_copydata(m0, off, (int)len, mtod(m, char *) + hdrlen); 481 m->m_len += len; 482 TCP_OUTPUT_COUNTER_INCR(&tcp_output_copysmall); 483 } else { 484 m->m_next = m_copym(m0, off, (int)len, M_DONTWAIT); 485 if (m->m_next == NULL) { 486 m_freem(m); 487 return ENOBUFS; 488 } 489 #ifdef TCP_OUTPUT_COUNTERS 490 if (m->m_next->m_flags & M_EXT) 491 TCP_OUTPUT_COUNTER_INCR(&tcp_output_refbig); 492 else 493 TCP_OUTPUT_COUNTER_INCR(&tcp_output_copybig); 494 #endif 495 } 496 497 *mp = m; 498 return 0; 499 } 500 501 /* 502 * Tcp output routine: figure out what should be sent and send it. 503 */ 504 int 505 tcp_output(struct tcpcb *tp) 506 { 507 struct rtentry *rt = NULL; 508 struct socket *so; 509 struct route *ro; 510 long len, win; 511 int off, flags, error; 512 struct mbuf *m; 513 struct ip *ip; 514 #ifdef INET6 515 struct ip6_hdr *ip6; 516 #endif 517 struct tcphdr *th; 518 u_char opt[MAX_TCPOPTLEN], *optp; 519 #define OPT_FITS(more) ((optlen + (more)) <= sizeof(opt)) 520 unsigned optlen, hdrlen, packetlen; 521 unsigned int sack_numblks; 522 int idle, sendalot, txsegsize, rxsegsize; 523 int txsegsize_nosack; 524 int maxburst = TCP_MAXBURST; 525 int af; /* address family on the wire */ 526 int iphdrlen; 527 int has_tso4, has_tso6; 528 int has_tso, use_tso; 529 bool alwaysfrag; 530 int sack_rxmit; 531 int sack_bytes_rxmt; 532 int ecn_tos; 533 struct sackhole *p; 534 #ifdef TCP_SIGNATURE 535 int sigoff = 0; 536 #endif 537 net_stat_ref_t tcps; 538 539 so = tp->t_inpcb->inp_socket; 540 ro = &tp->t_inpcb->inp_route; 541 542 switch (af = tp->t_family) { 543 case AF_INET: 544 case AF_INET6: 545 if (tp->t_inpcb) 546 break; 547 return EINVAL; 548 default: 549 return EAFNOSUPPORT; 550 } 551 552 if (tcp_segsize(tp, &txsegsize, &rxsegsize, &alwaysfrag)) 553 return EMSGSIZE; 554 555 idle = (tp->snd_max == tp->snd_una); 556 557 /* 558 * Determine if we can use TCP segmentation offload: 559 * - If we're using IPv4 560 * - If there is not an IPsec policy that prevents it 561 * - If the interface can do it 562 */ 563 has_tso4 = has_tso6 = false; 564 565 has_tso4 = tp->t_inpcb->inp_af == AF_INET && 566 #if defined(IPSEC) 567 (!ipsec_used || ipsec_pcb_skip_ipsec(tp->t_inpcb->inp_sp, 568 IPSEC_DIR_OUTBOUND)) && 569 #endif 570 (rt = rtcache_validate(&tp->t_inpcb->inp_route)) != NULL && 571 (rt->rt_ifp->if_capenable & IFCAP_TSOv4) != 0; 572 if (rt != NULL) { 573 rtcache_unref(rt, &tp->t_inpcb->inp_route); 574 rt = NULL; 575 } 576 577 #if defined(INET6) 578 has_tso6 = tp->t_inpcb->inp_af == AF_INET6 && 579 #if defined(IPSEC) 580 (!ipsec_used || ipsec_pcb_skip_ipsec(tp->t_inpcb->inp_sp, 581 IPSEC_DIR_OUTBOUND)) && 582 #endif 583 (rt = rtcache_validate(&tp->t_inpcb->inp_route)) != NULL && 584 (rt->rt_ifp->if_capenable & IFCAP_TSOv6) != 0; 585 if (rt != NULL) 586 rtcache_unref(rt, &tp->t_inpcb->inp_route); 587 #endif /* defined(INET6) */ 588 has_tso = (has_tso4 || has_tso6) && !alwaysfrag; 589 590 /* 591 * Restart Window computation. From draft-floyd-incr-init-win-03: 592 * 593 * Optionally, a TCP MAY set the restart window to the 594 * minimum of the value used for the initial window and 595 * the current value of cwnd (in other words, using a 596 * larger value for the restart window should never increase 597 * the size of cwnd). 598 */ 599 if (tcp_cwm) { 600 /* 601 * Hughes/Touch/Heidemann Congestion Window Monitoring. 602 * Count the number of packets currently pending 603 * acknowledgement, and limit our congestion window 604 * to a pre-determined allowed burst size plus that count. 605 * This prevents bursting once all pending packets have 606 * been acknowledged (i.e. transmission is idle). 607 * 608 * XXX Link this to Initial Window? 609 */ 610 tp->snd_cwnd = uimin(tp->snd_cwnd, 611 (tcp_cwm_burstsize * txsegsize) + 612 (tp->snd_nxt - tp->snd_una)); 613 } else { 614 if (idle && (tcp_now - tp->t_rcvtime) >= tp->t_rxtcur) { 615 /* 616 * We have been idle for "a while" and no acks are 617 * expected to clock out any data we send -- 618 * slow start to get ack "clock" running again. 619 */ 620 int ss = tcp_init_win; 621 if (tp->t_inpcb->inp_af == AF_INET && 622 in_localaddr(in4p_faddr(tp->t_inpcb))) 623 ss = tcp_init_win_local; 624 #ifdef INET6 625 else if (tp->t_inpcb->inp_af == AF_INET6 && 626 in6_localaddr(&in6p_faddr(tp->t_inpcb))) 627 ss = tcp_init_win_local; 628 #endif 629 tp->snd_cwnd = uimin(tp->snd_cwnd, 630 TCP_INITIAL_WINDOW(ss, txsegsize)); 631 } 632 } 633 634 txsegsize_nosack = txsegsize; 635 again: 636 ecn_tos = 0; 637 use_tso = has_tso; 638 if ((tp->t_flags & (TF_ECN_SND_CWR|TF_ECN_SND_ECE)) != 0) { 639 /* don't duplicate CWR/ECE. */ 640 use_tso = 0; 641 } 642 TCP_REASS_LOCK(tp); 643 sack_numblks = tcp_sack_numblks(tp); 644 if (sack_numblks) { 645 int sackoptlen; 646 647 sackoptlen = TCP_SACK_OPTLEN(sack_numblks); 648 if (sackoptlen > txsegsize_nosack) { 649 sack_numblks = 0; /* give up SACK */ 650 txsegsize = txsegsize_nosack; 651 } else { 652 if ((tp->rcv_sack_flags & TCPSACK_HAVED) != 0) { 653 /* don't duplicate D-SACK. */ 654 use_tso = 0; 655 } 656 txsegsize = txsegsize_nosack - sackoptlen; 657 } 658 } else { 659 txsegsize = txsegsize_nosack; 660 } 661 662 /* 663 * Determine length of data that should be transmitted, and 664 * flags that should be used. If there is some data or critical 665 * controls (SYN, RST) to send, then transmit; otherwise, 666 * investigate further. 667 * 668 * Readjust SACK information to avoid resending duplicate data. 669 */ 670 if (TCP_SACK_ENABLED(tp) && SEQ_LT(tp->snd_nxt, tp->snd_max)) 671 tcp_sack_adjust(tp); 672 sendalot = 0; 673 off = tp->snd_nxt - tp->snd_una; 674 win = uimin(tp->snd_wnd, tp->snd_cwnd); 675 676 flags = tcp_outflags[tp->t_state]; 677 678 /* 679 * Send any SACK-generated retransmissions. If we're explicitly trying 680 * to send out new data (when sendalot is 1), bypass this function. 681 * If we retransmit in fast recovery mode, decrement snd_cwnd, since 682 * we're replacing a (future) new transmission with a retransmission 683 * now, and we previously incremented snd_cwnd in tcp_input(). 684 */ 685 /* 686 * Still in sack recovery, reset rxmit flag to zero. 687 */ 688 sack_rxmit = 0; 689 sack_bytes_rxmt = 0; 690 len = 0; 691 p = NULL; 692 do { 693 long cwin; 694 if (!TCP_SACK_ENABLED(tp)) 695 break; 696 if (tp->t_partialacks < 0) 697 break; 698 p = tcp_sack_output(tp, &sack_bytes_rxmt); 699 if (p == NULL) 700 break; 701 702 cwin = uimin(tp->snd_wnd, tp->snd_cwnd) - sack_bytes_rxmt; 703 if (cwin < 0) 704 cwin = 0; 705 /* Do not retransmit SACK segments beyond snd_recover */ 706 if (SEQ_GT(p->end, tp->snd_recover)) { 707 /* 708 * (At least) part of sack hole extends beyond 709 * snd_recover. Check to see if we can rexmit data 710 * for this hole. 711 */ 712 if (SEQ_GEQ(p->rxmit, tp->snd_recover)) { 713 /* 714 * Can't rexmit any more data for this hole. 715 * That data will be rexmitted in the next 716 * sack recovery episode, when snd_recover 717 * moves past p->rxmit. 718 */ 719 p = NULL; 720 break; 721 } 722 /* Can rexmit part of the current hole */ 723 len = ((long)ulmin(cwin, tp->snd_recover - p->rxmit)); 724 } else 725 len = ((long)ulmin(cwin, p->end - p->rxmit)); 726 off = p->rxmit - tp->snd_una; 727 if (off + len > so->so_snd.sb_cc) { 728 /* 1 for TH_FIN */ 729 KASSERT(off + len == so->so_snd.sb_cc + 1); 730 KASSERT(p->rxmit + len == tp->snd_max); 731 len = so->so_snd.sb_cc - off; 732 } 733 if (len > 0) { 734 sack_rxmit = 1; 735 sendalot = 1; 736 } 737 } while (/*CONSTCOND*/0); 738 739 /* 740 * If in persist timeout with window of 0, send 1 byte. 741 * Otherwise, if window is small but nonzero 742 * and timer expired, we will send what we can 743 * and go to transmit state. 744 */ 745 if (tp->t_force) { 746 if (win == 0) { 747 /* 748 * If we still have some data to send, then 749 * clear the FIN bit. Usually this would 750 * happen below when it realizes that we 751 * aren't sending all the data. However, 752 * if we have exactly 1 byte of unset data, 753 * then it won't clear the FIN bit below, 754 * and if we are in persist state, we wind 755 * up sending the packet without recording 756 * that we sent the FIN bit. 757 * 758 * We can't just blindly clear the FIN bit, 759 * because if we don't have any more data 760 * to send then the probe will be the FIN 761 * itself. 762 */ 763 if (off < so->so_snd.sb_cc) 764 flags &= ~TH_FIN; 765 win = 1; 766 } else { 767 TCP_TIMER_DISARM(tp, TCPT_PERSIST); 768 tp->t_rxtshift = 0; 769 } 770 } 771 772 if (sack_rxmit == 0) { 773 if (TCP_SACK_ENABLED(tp) && tp->t_partialacks >= 0) { 774 long cwin; 775 776 /* 777 * We are inside of a SACK recovery episode and are 778 * sending new data, having retransmitted all the 779 * data possible in the scoreboard. 780 */ 781 if (tp->snd_wnd < so->so_snd.sb_cc) { 782 len = tp->snd_wnd - off; 783 flags &= ~TH_FIN; 784 } else { 785 len = so->so_snd.sb_cc - off; 786 } 787 788 /* 789 * From FreeBSD: 790 * Don't remove this (len > 0) check ! 791 * We explicitly check for len > 0 here (although it 792 * isn't really necessary), to work around a gcc 793 * optimization issue - to force gcc to compute 794 * len above. Without this check, the computation 795 * of len is bungled by the optimizer. 796 */ 797 if (len > 0) { 798 cwin = tp->snd_cwnd - 799 (tp->snd_nxt - tp->sack_newdata) - 800 sack_bytes_rxmt; 801 if (cwin < 0) 802 cwin = 0; 803 if (cwin < len) { 804 len = cwin; 805 flags &= ~TH_FIN; 806 } 807 } 808 } else if (win < so->so_snd.sb_cc) { 809 len = win - off; 810 flags &= ~TH_FIN; 811 } else { 812 len = so->so_snd.sb_cc - off; 813 } 814 } 815 816 if (len < 0) { 817 /* 818 * If FIN has been sent but not acked, 819 * but we haven't been called to retransmit, 820 * len will be -1. Otherwise, window shrank 821 * after we sent into it. If window shrank to 0, 822 * cancel pending retransmit, pull snd_nxt back 823 * to (closed) window, and set the persist timer 824 * if it isn't already going. If the window didn't 825 * close completely, just wait for an ACK. 826 * 827 * If we have a pending FIN, either it has already been 828 * transmitted or it is outside the window, so drop it. 829 * If the FIN has been transmitted, but this is not a 830 * retransmission, then len must be -1. Therefore we also 831 * prevent here the sending of `gratuitous FINs'. This 832 * eliminates the need to check for that case below (e.g. 833 * to back up snd_nxt before the FIN so that the sequence 834 * number is correct). 835 */ 836 len = 0; 837 flags &= ~TH_FIN; 838 if (win == 0) { 839 TCP_TIMER_DISARM(tp, TCPT_REXMT); 840 tp->t_rxtshift = 0; 841 tp->snd_nxt = tp->snd_una; 842 if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) 843 tcp_setpersist(tp); 844 } 845 } 846 847 /* 848 * Automatic sizing enables the performance of large buffers 849 * and most of the efficiency of small ones by only allocating 850 * space when it is needed. 851 * 852 * The criteria to step up the send buffer one notch are: 853 * 1. receive window of remote host is larger than send buffer 854 * (with a fudge factor of 5/4th); 855 * 2. send buffer is filled to 7/8th with data (so we actually 856 * have data to make use of it); 857 * 3. send buffer fill has not hit maximal automatic size; 858 * 4. our send window (slow start and cogestion controlled) is 859 * larger than sent but unacknowledged data in send buffer. 860 * 861 * The remote host receive window scaling factor may limit the 862 * growing of the send buffer before it reaches its allowed 863 * maximum. 864 * 865 * It scales directly with slow start or congestion window 866 * and does at most one step per received ACK. This fast 867 * scaling has the drawback of growing the send buffer beyond 868 * what is strictly necessary to make full use of a given 869 * delay*bandwidth product. However testing has shown this not 870 * to be much of a problem. At worst we are trading wasting 871 * of available bandwidth (the non-use of it) for wasting some 872 * socket buffer memory. 873 * 874 * TODO: Shrink send buffer during idle periods together 875 * with congestion window. Requires another timer. 876 */ 877 if (tcp_do_autosndbuf && so->so_snd.sb_flags & SB_AUTOSIZE) { 878 if ((tp->snd_wnd / 4 * 5) >= so->so_snd.sb_hiwat && 879 so->so_snd.sb_cc >= (so->so_snd.sb_hiwat / 8 * 7) && 880 so->so_snd.sb_cc < tcp_autosndbuf_max && 881 win >= (so->so_snd.sb_cc - (tp->snd_nxt - tp->snd_una))) { 882 if (!sbreserve(&so->so_snd, 883 uimin(so->so_snd.sb_hiwat + tcp_autosndbuf_inc, 884 tcp_autosndbuf_max), so)) 885 so->so_snd.sb_flags &= ~SB_AUTOSIZE; 886 } 887 } 888 889 if (len > txsegsize) { 890 if (use_tso) { 891 /* 892 * Truncate TSO transfers to IP_MAXPACKET, and make 893 * sure that we send equal size transfers down the 894 * stack (rather than big-small-big-small-...). 895 */ 896 #ifdef INET6 897 CTASSERT(IPV6_MAXPACKET == IP_MAXPACKET); 898 #endif 899 len = (uimin(len, IP_MAXPACKET) / txsegsize) * txsegsize; 900 if (len <= txsegsize) { 901 use_tso = 0; 902 } 903 } else 904 len = txsegsize; 905 flags &= ~TH_FIN; 906 sendalot = 1; 907 } else 908 use_tso = 0; 909 if (sack_rxmit) { 910 if (SEQ_LT(p->rxmit + len, tp->snd_una + so->so_snd.sb_cc)) 911 flags &= ~TH_FIN; 912 } 913 914 win = sbspace(&so->so_rcv); 915 916 /* 917 * Sender silly window avoidance. If connection is idle 918 * and can send all data, a maximum segment, 919 * at least a maximum default-size segment do it, 920 * or are forced, do it; otherwise don't bother. 921 * If peer's buffer is tiny, then send 922 * when window is at least half open. 923 * If retransmitting (possibly after persist timer forced us 924 * to send into a small window), then must resend. 925 */ 926 if (len) { 927 if (len >= txsegsize) 928 goto send; 929 if ((so->so_state & SS_MORETOCOME) == 0 && 930 ((idle || tp->t_flags & TF_NODELAY) && 931 len + off >= so->so_snd.sb_cc)) 932 goto send; 933 if (tp->t_force) 934 goto send; 935 if (len >= tp->max_sndwnd / 2) 936 goto send; 937 if (SEQ_LT(tp->snd_nxt, tp->snd_max)) 938 goto send; 939 if (sack_rxmit) 940 goto send; 941 } 942 943 /* 944 * Compare available window to amount of window known to peer 945 * (as advertised window less next expected input). If the 946 * difference is at least twice the size of the largest segment 947 * we expect to receive (i.e. two segments) or at least 50% of 948 * the maximum possible window, then want to send a window update 949 * to peer. 950 */ 951 if (win > 0) { 952 /* 953 * "adv" is the amount we can increase the window, 954 * taking into account that we are limited by 955 * TCP_MAXWIN << tp->rcv_scale. 956 */ 957 long recwin = uimin(win, (long)TCP_MAXWIN << tp->rcv_scale); 958 long oldwin, adv; 959 960 /* 961 * rcv_nxt may overtake rcv_adv when we accept a 962 * zero-window probe. 963 */ 964 if (SEQ_GT(tp->rcv_adv, tp->rcv_nxt)) 965 oldwin = tp->rcv_adv - tp->rcv_nxt; 966 else 967 oldwin = 0; 968 969 /* 970 * If the new window size ends up being the same as or 971 * less than the old size when it is scaled, then 972 * don't force a window update. 973 */ 974 if (recwin >> tp->rcv_scale <= oldwin >> tp->rcv_scale) 975 goto dontupdate; 976 977 adv = recwin - oldwin; 978 if (adv >= (long) (2 * rxsegsize)) 979 goto send; 980 if (2 * adv >= (long) so->so_rcv.sb_hiwat) 981 goto send; 982 } 983 dontupdate: 984 985 /* 986 * Send if we owe peer an ACK. 987 */ 988 if (tp->t_flags & TF_ACKNOW) 989 goto send; 990 if (flags & (TH_SYN|TH_FIN|TH_RST)) 991 goto send; 992 if (SEQ_GT(tp->snd_up, tp->snd_una)) 993 goto send; 994 /* 995 * In SACK, it is possible for tcp_output to fail to send a segment 996 * after the retransmission timer has been turned off. Make sure 997 * that the retransmission timer is set. 998 */ 999 if (TCP_SACK_ENABLED(tp) && SEQ_GT(tp->snd_max, tp->snd_una) && 1000 !TCP_TIMER_ISARMED(tp, TCPT_REXMT) && 1001 !TCP_TIMER_ISARMED(tp, TCPT_PERSIST)) { 1002 TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur); 1003 goto just_return; 1004 } 1005 1006 /* 1007 * TCP window updates are not reliable, rather a polling protocol 1008 * using ``persist'' packets is used to insure receipt of window 1009 * updates. The three ``states'' for the output side are: 1010 * idle not doing retransmits or persists 1011 * persisting to move a small or zero window 1012 * (re)transmitting and thereby not persisting 1013 * 1014 * tp->t_timer[TCPT_PERSIST] 1015 * is set when we are in persist state. 1016 * tp->t_force 1017 * is set when we are called to send a persist packet. 1018 * tp->t_timer[TCPT_REXMT] 1019 * is set when we are retransmitting 1020 * The output side is idle when both timers are zero. 1021 * 1022 * If send window is too small, there is data to transmit, and no 1023 * retransmit or persist is pending, then go to persist state. 1024 * If nothing happens soon, send when timer expires: 1025 * if window is nonzero, transmit what we can, 1026 * otherwise force out a byte. 1027 */ 1028 if (so->so_snd.sb_cc && TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 && 1029 TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) { 1030 tp->t_rxtshift = 0; 1031 tcp_setpersist(tp); 1032 } 1033 1034 /* 1035 * No reason to send a segment, just return. 1036 */ 1037 just_return: 1038 TCP_REASS_UNLOCK(tp); 1039 return 0; 1040 1041 send: 1042 /* 1043 * Before ESTABLISHED, force sending of initial options unless TCP set 1044 * not to do any options. 1045 * 1046 * Note: we assume that the IP/TCP header plus TCP options always fit 1047 * in a single mbuf, leaving room for a maximum link header, i.e.: 1048 * max_linkhdr + IP_header + TCP_header + optlen <= MCLBYTES 1049 */ 1050 optlen = 0; 1051 optp = opt; 1052 switch (af) { 1053 case AF_INET: 1054 iphdrlen = sizeof(struct ip) + sizeof(struct tcphdr); 1055 break; 1056 #ifdef INET6 1057 case AF_INET6: 1058 iphdrlen = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); 1059 break; 1060 #endif 1061 default: /*pacify gcc*/ 1062 iphdrlen = 0; 1063 break; 1064 } 1065 hdrlen = iphdrlen; 1066 if (flags & TH_SYN) { 1067 struct rtentry *synrt; 1068 1069 synrt = inpcb_rtentry(tp->t_inpcb); 1070 tp->snd_nxt = tp->iss; 1071 tp->t_ourmss = tcp_mss_to_advertise(synrt != NULL ? 1072 synrt->rt_ifp : NULL, af); 1073 inpcb_rtentry_unref(synrt, tp->t_inpcb); 1074 if ((tp->t_flags & TF_NOOPT) == 0 && OPT_FITS(TCPOLEN_MAXSEG)) { 1075 *optp++ = TCPOPT_MAXSEG; 1076 *optp++ = TCPOLEN_MAXSEG; 1077 *optp++ = (tp->t_ourmss >> 8) & 0xff; 1078 *optp++ = tp->t_ourmss & 0xff; 1079 optlen += TCPOLEN_MAXSEG; 1080 1081 if ((tp->t_flags & TF_REQ_SCALE) && 1082 ((flags & TH_ACK) == 0 || 1083 (tp->t_flags & TF_RCVD_SCALE)) && 1084 OPT_FITS(TCPOLEN_WINDOW + TCPOLEN_NOP)) { 1085 *((uint32_t *)optp) = htonl( 1086 TCPOPT_NOP << 24 | 1087 TCPOPT_WINDOW << 16 | 1088 TCPOLEN_WINDOW << 8 | 1089 tp->request_r_scale); 1090 optp += TCPOLEN_WINDOW + TCPOLEN_NOP; 1091 optlen += TCPOLEN_WINDOW + TCPOLEN_NOP; 1092 } 1093 if (tcp_do_sack && OPT_FITS(TCPOLEN_SACK_PERMITTED)) { 1094 *optp++ = TCPOPT_SACK_PERMITTED; 1095 *optp++ = TCPOLEN_SACK_PERMITTED; 1096 optlen += TCPOLEN_SACK_PERMITTED; 1097 } 1098 } 1099 } 1100 1101 /* 1102 * Send a timestamp and echo-reply if this is a SYN and our side 1103 * wants to use timestamps (TF_REQ_TSTMP is set) or both our side 1104 * and our peer have sent timestamps in our SYN's. 1105 */ 1106 if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP && 1107 (flags & TH_RST) == 0 && 1108 ((flags & (TH_SYN|TH_ACK)) == TH_SYN || 1109 (tp->t_flags & TF_RCVD_TSTMP))) { 1110 int alen = 0; 1111 while (optlen % 4 != 2) { 1112 optlen += TCPOLEN_NOP; 1113 *optp++ = TCPOPT_NOP; 1114 alen++; 1115 } 1116 if (OPT_FITS(TCPOLEN_TIMESTAMP)) { 1117 *optp++ = TCPOPT_TIMESTAMP; 1118 *optp++ = TCPOLEN_TIMESTAMP; 1119 uint32_t *lp = (uint32_t *)optp; 1120 /* Form timestamp option (appendix A of RFC 1323) */ 1121 *lp++ = htonl(TCP_TIMESTAMP(tp)); 1122 *lp = htonl(tp->ts_recent); 1123 optp += TCPOLEN_TIMESTAMP - 2; 1124 optlen += TCPOLEN_TIMESTAMP; 1125 1126 /* Set receive buffer autosizing timestamp. */ 1127 if (tp->rfbuf_ts == 0 && 1128 (so->so_rcv.sb_flags & SB_AUTOSIZE)) 1129 tp->rfbuf_ts = TCP_TIMESTAMP(tp); 1130 } else { 1131 optp -= alen; 1132 optlen -= alen; 1133 } 1134 } 1135 1136 #ifdef TCP_SIGNATURE 1137 if (tp->t_flags & TF_SIGNATURE) { 1138 /* 1139 * Initialize TCP-MD5 option (RFC2385) 1140 */ 1141 if (!OPT_FITS(TCPOLEN_SIGNATURE)) 1142 goto reset; 1143 1144 *optp++ = TCPOPT_SIGNATURE; 1145 *optp++ = TCPOLEN_SIGNATURE; 1146 sigoff = optlen + 2; 1147 memset(optp, 0, TCP_SIGLEN); 1148 optlen += TCPOLEN_SIGNATURE; 1149 optp += TCP_SIGLEN; 1150 } 1151 #endif 1152 1153 /* 1154 * Tack on the SACK block if it is necessary. 1155 */ 1156 if (sack_numblks) { 1157 int alen = 0; 1158 int sack_len = sack_numblks * 8; 1159 while (optlen % 4 != 2) { 1160 optlen += TCPOLEN_NOP; 1161 *optp++ = TCPOPT_NOP; 1162 alen++; 1163 } 1164 if (OPT_FITS(sack_len + 2)) { 1165 struct ipqent *tiqe; 1166 *optp++ = TCPOPT_SACK; 1167 *optp++ = sack_len + 2; 1168 uint32_t *lp = (uint32_t *)optp; 1169 if ((tp->rcv_sack_flags & TCPSACK_HAVED) != 0) { 1170 sack_numblks--; 1171 *lp++ = htonl(tp->rcv_dsack_block.left); 1172 *lp++ = htonl(tp->rcv_dsack_block.right); 1173 tp->rcv_sack_flags &= ~TCPSACK_HAVED; 1174 } 1175 for (tiqe = TAILQ_FIRST(&tp->timeq); 1176 sack_numblks > 0; 1177 tiqe = TAILQ_NEXT(tiqe, ipqe_timeq)) { 1178 KASSERT(tiqe != NULL); 1179 sack_numblks--; 1180 *lp++ = htonl(tiqe->ipqe_seq); 1181 *lp++ = htonl(tiqe->ipqe_seq + tiqe->ipqe_len + 1182 ((tiqe->ipqe_flags & TH_FIN) != 0 ? 1 : 0)); 1183 } 1184 optlen += sack_len + 2; 1185 optp += sack_len; 1186 } else { 1187 optp -= alen; 1188 optlen -= alen; 1189 } 1190 } 1191 1192 /* Terminate and pad TCP options to a 4 byte boundary. */ 1193 if (optlen % 4) { 1194 if (!OPT_FITS(TCPOLEN_EOL)) { 1195 reset: TCP_REASS_UNLOCK(tp); 1196 error = ECONNABORTED; 1197 goto out; 1198 } 1199 optlen += TCPOLEN_EOL; 1200 *optp++ = TCPOPT_EOL; 1201 } 1202 /* 1203 * According to RFC 793 (STD0007): 1204 * "The content of the header beyond the End-of-Option option 1205 * must be header padding (i.e., zero)." 1206 * and later: "The padding is composed of zeros." 1207 */ 1208 while (optlen % 4) { 1209 if (!OPT_FITS(TCPOLEN_PAD)) 1210 goto reset; 1211 optlen += TCPOLEN_PAD; 1212 *optp++ = TCPOPT_PAD; 1213 } 1214 1215 TCP_REASS_UNLOCK(tp); 1216 1217 hdrlen += optlen; 1218 1219 #ifdef DIAGNOSTIC 1220 if (!use_tso && len > txsegsize) 1221 panic("tcp data to be sent is larger than segment"); 1222 else if (use_tso && len > IP_MAXPACKET) 1223 panic("tcp data to be sent is larger than max TSO size"); 1224 if (max_linkhdr + hdrlen > MCLBYTES) 1225 panic("tcphdr too big"); 1226 #endif 1227 1228 /* 1229 * Grab a header mbuf, attaching a copy of data to 1230 * be transmitted, and initialize the header from 1231 * the template for sends on this connection. 1232 */ 1233 if (len) { 1234 error = tcp_build_datapkt(tp, so, off, len, hdrlen, &m); 1235 if (error) 1236 goto out; 1237 /* 1238 * If we're sending everything we've got, set PUSH. 1239 * (This will keep happy those implementations which only 1240 * give data to the user when a buffer fills or 1241 * a PUSH comes in.) 1242 */ 1243 if (off + len == so->so_snd.sb_cc) 1244 flags |= TH_PUSH; 1245 } else { 1246 tcps = TCP_STAT_GETREF(); 1247 if (tp->t_flags & TF_ACKNOW) 1248 _NET_STATINC_REF(tcps, TCP_STAT_SNDACKS); 1249 else if (flags & (TH_SYN|TH_FIN|TH_RST)) 1250 _NET_STATINC_REF(tcps, TCP_STAT_SNDCTRL); 1251 else if (SEQ_GT(tp->snd_up, tp->snd_una)) 1252 _NET_STATINC_REF(tcps, TCP_STAT_SNDURG); 1253 else 1254 _NET_STATINC_REF(tcps, TCP_STAT_SNDWINUP); 1255 TCP_STAT_PUTREF(); 1256 1257 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1258 if (m != NULL && max_linkhdr + hdrlen > MHLEN) { 1259 MCLGET(m, M_DONTWAIT); 1260 if ((m->m_flags & M_EXT) == 0) { 1261 m_freem(m); 1262 m = NULL; 1263 } 1264 } 1265 if (m == NULL) { 1266 error = ENOBUFS; 1267 goto out; 1268 } 1269 MCLAIM(m, &tcp_tx_mowner); 1270 m->m_data += max_linkhdr; 1271 m->m_len = hdrlen; 1272 } 1273 m_reset_rcvif(m); 1274 switch (af) { 1275 case AF_INET: 1276 ip = mtod(m, struct ip *); 1277 #ifdef INET6 1278 ip6 = NULL; 1279 #endif 1280 th = (struct tcphdr *)(ip + 1); 1281 break; 1282 #ifdef INET6 1283 case AF_INET6: 1284 ip = NULL; 1285 ip6 = mtod(m, struct ip6_hdr *); 1286 th = (struct tcphdr *)(ip6 + 1); 1287 break; 1288 #endif 1289 default: /*pacify gcc*/ 1290 ip = NULL; 1291 #ifdef INET6 1292 ip6 = NULL; 1293 #endif 1294 th = NULL; 1295 break; 1296 } 1297 if (tp->t_template == NULL) 1298 panic("%s: no template", __func__); 1299 if (tp->t_template->m_len < iphdrlen) 1300 panic("%s: %d < %d", __func__, tp->t_template->m_len, iphdrlen); 1301 bcopy(mtod(tp->t_template, void *), mtod(m, void *), iphdrlen); 1302 1303 /* 1304 * If we are starting a connection, send ECN setup 1305 * SYN packet. If we are on a retransmit, we may 1306 * resend those bits a number of times as per 1307 * RFC 3168. 1308 */ 1309 if (tp->t_state == TCPS_SYN_SENT && tcp_do_ecn) { 1310 if (tp->t_flags & TF_SYN_REXMT) { 1311 if (tp->t_ecn_retries--) 1312 flags |= TH_ECE|TH_CWR; 1313 } else { 1314 flags |= TH_ECE|TH_CWR; 1315 tp->t_ecn_retries = tcp_ecn_maxretries; 1316 } 1317 } 1318 1319 if (TCP_ECN_ALLOWED(tp)) { 1320 /* 1321 * If the peer has ECN, mark data packets 1322 * ECN capable. Ignore pure ack packets, retransmissions 1323 * and window probes. 1324 */ 1325 if (len > 0 && SEQ_GEQ(tp->snd_nxt, tp->snd_max) && 1326 !(tp->t_force && len == 1)) { 1327 ecn_tos = IPTOS_ECN_ECT0; 1328 TCP_STATINC(TCP_STAT_ECN_ECT); 1329 } 1330 1331 /* 1332 * Reply with proper ECN notifications. 1333 */ 1334 if (tp->t_flags & TF_ECN_SND_CWR) { 1335 flags |= TH_CWR; 1336 tp->t_flags &= ~TF_ECN_SND_CWR; 1337 } 1338 if (tp->t_flags & TF_ECN_SND_ECE) { 1339 flags |= TH_ECE; 1340 } 1341 } 1342 1343 /* 1344 * If we are doing retransmissions, then snd_nxt will 1345 * not reflect the first unsent octet. For ACK only 1346 * packets, we do not want the sequence number of the 1347 * retransmitted packet, we want the sequence number 1348 * of the next unsent octet. So, if there is no data 1349 * (and no SYN or FIN), use snd_max instead of snd_nxt 1350 * when filling in ti_seq. But if we are in persist 1351 * state, snd_max might reflect one byte beyond the 1352 * right edge of the window, so use snd_nxt in that 1353 * case, since we know we aren't doing a retransmission. 1354 * (retransmit and persist are mutually exclusive...) 1355 */ 1356 if (TCP_SACK_ENABLED(tp) && sack_rxmit) { 1357 th->th_seq = htonl(p->rxmit); 1358 p->rxmit += len; 1359 } else { 1360 if (len || (flags & (TH_SYN|TH_FIN)) || 1361 TCP_TIMER_ISARMED(tp, TCPT_PERSIST)) 1362 th->th_seq = htonl(tp->snd_nxt); 1363 else 1364 th->th_seq = htonl(tp->snd_max); 1365 } 1366 th->th_ack = htonl(tp->rcv_nxt); 1367 if (optlen) { 1368 memcpy(th + 1, opt, optlen); 1369 th->th_off = (sizeof (struct tcphdr) + optlen) >> 2; 1370 } 1371 th->th_flags = flags; 1372 /* 1373 * Calculate receive window. Don't shrink window, 1374 * but avoid silly window syndrome. 1375 */ 1376 if (win < (long)(so->so_rcv.sb_hiwat / 4) && win < (long)rxsegsize) 1377 win = 0; 1378 if (win > (long)TCP_MAXWIN << tp->rcv_scale) 1379 win = (long)TCP_MAXWIN << tp->rcv_scale; 1380 if (win < (long)(int32_t)(tp->rcv_adv - tp->rcv_nxt)) 1381 win = (long)(int32_t)(tp->rcv_adv - tp->rcv_nxt); 1382 th->th_win = htons((u_int16_t) (win>>tp->rcv_scale)); 1383 if (th->th_win == 0) { 1384 tp->t_sndzerowin++; 1385 } 1386 if (SEQ_GT(tp->snd_up, tp->snd_nxt)) { 1387 u_int32_t urp = tp->snd_up - tp->snd_nxt; 1388 if (urp > IP_MAXPACKET) 1389 urp = IP_MAXPACKET; 1390 th->th_urp = htons((u_int16_t)urp); 1391 th->th_flags |= TH_URG; 1392 } else 1393 /* 1394 * If no urgent pointer to send, then we pull 1395 * the urgent pointer to the left edge of the send window 1396 * so that it doesn't drift into the send window on sequence 1397 * number wraparound. 1398 */ 1399 tp->snd_up = tp->snd_una; /* drag it along */ 1400 1401 #ifdef TCP_SIGNATURE 1402 if (sigoff && (tp->t_flags & TF_SIGNATURE)) { 1403 struct secasvar *sav; 1404 u_int8_t *sigp; 1405 1406 sav = tcp_signature_getsav(m); 1407 if (sav == NULL) { 1408 m_freem(m); 1409 return EPERM; 1410 } 1411 1412 m->m_pkthdr.len = hdrlen + len; 1413 sigp = (char *)th + sizeof(*th) + sigoff; 1414 tcp_signature(m, th, (char *)th - mtod(m, char *), sav, sigp); 1415 1416 key_sa_recordxfer(sav, m); 1417 KEY_SA_UNREF(&sav); 1418 } 1419 #endif 1420 1421 /* 1422 * Set ourselves up to be checksummed just before the packet 1423 * hits the wire. 1424 */ 1425 switch (af) { 1426 case AF_INET: 1427 m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); 1428 if (use_tso) { 1429 m->m_pkthdr.segsz = txsegsize; 1430 m->m_pkthdr.csum_flags = M_CSUM_TSOv4; 1431 } else { 1432 m->m_pkthdr.csum_flags = M_CSUM_TCPv4; 1433 if (len + optlen) { 1434 /* Fixup the pseudo-header checksum. */ 1435 /* XXXJRT Not IP Jumbogram safe. */ 1436 th->th_sum = in_cksum_addword(th->th_sum, 1437 htons((u_int16_t) (len + optlen))); 1438 } 1439 } 1440 break; 1441 #ifdef INET6 1442 case AF_INET6: 1443 m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); 1444 if (use_tso) { 1445 m->m_pkthdr.segsz = txsegsize; 1446 m->m_pkthdr.csum_flags = M_CSUM_TSOv6; 1447 } else { 1448 m->m_pkthdr.csum_flags = M_CSUM_TCPv6; 1449 if (len + optlen) { 1450 /* Fixup the pseudo-header checksum. */ 1451 /* XXXJRT: Not IPv6 Jumbogram safe. */ 1452 th->th_sum = in_cksum_addword(th->th_sum, 1453 htons((u_int16_t) (len + optlen))); 1454 } 1455 } 1456 break; 1457 #endif 1458 } 1459 1460 /* 1461 * In transmit state, time the transmission and arrange for 1462 * the retransmit. In persist state, just set snd_max. 1463 */ 1464 if (tp->t_force == 0 || TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) { 1465 tcp_seq startseq = tp->snd_nxt; 1466 1467 /* 1468 * Advance snd_nxt over sequence space of this segment. 1469 * There are no states in which we send both a SYN and a FIN, 1470 * so we collapse the tests for these flags. 1471 */ 1472 if (flags & (TH_SYN|TH_FIN)) 1473 tp->snd_nxt++; 1474 if (sack_rxmit) 1475 goto timer; 1476 tp->snd_nxt += len; 1477 if (SEQ_GT(tp->snd_nxt, tp->snd_max)) { 1478 tp->snd_max = tp->snd_nxt; 1479 /* 1480 * Time this transmission if not a retransmission and 1481 * not currently timing anything. 1482 */ 1483 if (tp->t_rtttime == 0) { 1484 tp->t_rtttime = tcp_now; 1485 tp->t_rtseq = startseq; 1486 TCP_STATINC(TCP_STAT_SEGSTIMED); 1487 } 1488 } 1489 1490 /* 1491 * Set retransmit timer if not currently set, 1492 * and not doing an ack or a keep-alive probe. 1493 * Initial value for retransmit timer is smoothed 1494 * round-trip time + 2 * round-trip time variance. 1495 * Initialize shift counter which is used for backoff 1496 * of retransmit time. 1497 */ 1498 timer: 1499 if (TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0) { 1500 if ((sack_rxmit && tp->snd_nxt != tp->snd_max) 1501 || tp->snd_nxt != tp->snd_una) { 1502 if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST)) { 1503 TCP_TIMER_DISARM(tp, TCPT_PERSIST); 1504 tp->t_rxtshift = 0; 1505 } 1506 TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur); 1507 } else if (len == 0 && so->so_snd.sb_cc > 0 1508 && TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) { 1509 /* 1510 * If we are sending a window probe and there's 1511 * unacked data in the socket, make sure at 1512 * least the persist timer is running. 1513 */ 1514 tp->t_rxtshift = 0; 1515 tcp_setpersist(tp); 1516 } 1517 } 1518 } else 1519 if (SEQ_GT(tp->snd_nxt + len, tp->snd_max)) 1520 tp->snd_max = tp->snd_nxt + len; 1521 1522 #ifdef TCP_DEBUG 1523 /* 1524 * Trace. 1525 */ 1526 if (so->so_options & SO_DEBUG) 1527 tcp_trace(TA_OUTPUT, tp->t_state, tp, m, 0); 1528 #endif 1529 1530 /* 1531 * Fill in IP length and desired time to live and 1532 * send to IP level. There should be a better way 1533 * to handle ttl and tos; we could keep them in 1534 * the template, but need a way to checksum without them. 1535 */ 1536 m->m_pkthdr.len = hdrlen + len; 1537 1538 switch (af) { 1539 case AF_INET: 1540 ip->ip_len = htons(m->m_pkthdr.len); 1541 packetlen = m->m_pkthdr.len; 1542 if (tp->t_inpcb->inp_af == AF_INET) { 1543 ip->ip_ttl = in4p_ip(tp->t_inpcb).ip_ttl; 1544 ip->ip_tos = in4p_ip(tp->t_inpcb).ip_tos | ecn_tos; 1545 } 1546 #ifdef INET6 1547 else if (tp->t_inpcb->inp_af == AF_INET6) { 1548 ip->ip_ttl = in6pcb_selecthlim(tp->t_inpcb, NULL); /*XXX*/ 1549 ip->ip_tos = ecn_tos; /*XXX*/ 1550 } 1551 #endif 1552 break; 1553 #ifdef INET6 1554 case AF_INET6: 1555 packetlen = m->m_pkthdr.len; 1556 ip6->ip6_nxt = IPPROTO_TCP; 1557 if (tp->t_family == AF_INET6) { 1558 /* 1559 * we separately set hoplimit for every segment, since 1560 * the user might want to change the value via 1561 * setsockopt. Also, desired default hop limit might 1562 * be changed via Neighbor Discovery. 1563 */ 1564 ip6->ip6_hlim = in6pcb_selecthlim_rt(tp->t_inpcb); 1565 } 1566 ip6->ip6_flow |= htonl(ecn_tos << 20); 1567 /* ip6->ip6_flow = ??? (from template) */ 1568 /* ip6_plen will be filled in ip6_output(). */ 1569 break; 1570 #endif 1571 default: /*pacify gcc*/ 1572 packetlen = 0; 1573 break; 1574 } 1575 1576 switch (af) { 1577 case AF_INET: 1578 { 1579 struct mbuf *opts; 1580 1581 if (tp->t_inpcb->inp_af == AF_INET) 1582 opts = tp->t_inpcb->inp_options; 1583 else 1584 opts = NULL; 1585 error = ip_output(m, opts, ro, 1586 (tp->t_mtudisc ? IP_MTUDISC : 0) | 1587 (so->so_options & SO_DONTROUTE), NULL, tp->t_inpcb); 1588 break; 1589 } 1590 #ifdef INET6 1591 case AF_INET6: 1592 { 1593 struct ip6_pktopts *opts; 1594 1595 if (tp->t_inpcb->inp_af == AF_INET6) 1596 opts = in6p_outputopts(tp->t_inpcb); 1597 else 1598 opts = NULL; 1599 error = ip6_output(m, opts, ro, so->so_options & SO_DONTROUTE, 1600 NULL, tp->t_inpcb, NULL); 1601 break; 1602 } 1603 #endif 1604 default: 1605 error = EAFNOSUPPORT; 1606 break; 1607 } 1608 if (error) { 1609 out: 1610 if (error == ENOBUFS) { 1611 TCP_STATINC(TCP_STAT_SELFQUENCH); 1612 tcp_quench(tp->t_inpcb); 1613 error = 0; 1614 } else if ((error == EHOSTUNREACH || error == ENETDOWN || 1615 error == EHOSTDOWN) && TCPS_HAVERCVDSYN(tp->t_state)) { 1616 tp->t_softerror = error; 1617 error = 0; 1618 } 1619 1620 /* Back out the sequence number advance. */ 1621 if (sack_rxmit) 1622 p->rxmit -= len; 1623 1624 /* Restart the delayed ACK timer, if necessary. */ 1625 if (tp->t_flags & TF_DELACK) 1626 TCP_RESTART_DELACK(tp); 1627 1628 return error; 1629 } 1630 1631 if (packetlen > tp->t_pmtud_mtu_sent) 1632 tp->t_pmtud_mtu_sent = packetlen; 1633 1634 tcps = TCP_STAT_GETREF(); 1635 _NET_STATINC_REF(tcps, TCP_STAT_SNDTOTAL); 1636 if (tp->t_flags & TF_DELACK) 1637 _NET_STATINC_REF(tcps, TCP_STAT_DELACK); 1638 TCP_STAT_PUTREF(); 1639 1640 /* 1641 * Data sent (as far as we can tell). 1642 * If this advertises a larger window than any other segment, 1643 * then remember the size of the advertised window. 1644 * Any pending ACK has now been sent. 1645 */ 1646 if (win > 0 && SEQ_GT(tp->rcv_nxt+win, tp->rcv_adv)) 1647 tp->rcv_adv = tp->rcv_nxt + win; 1648 tp->last_ack_sent = tp->rcv_nxt; 1649 tp->t_flags &= ~TF_ACKNOW; 1650 TCP_CLEAR_DELACK(tp); 1651 #ifdef DIAGNOSTIC 1652 if (maxburst < 0) 1653 printf("tcp_output: maxburst exceeded by %d\n", -maxburst); 1654 #endif 1655 if (sendalot && (tp->t_congctl == &tcp_reno_ctl || --maxburst)) 1656 goto again; 1657 return 0; 1658 } 1659 1660 void 1661 tcp_setpersist(struct tcpcb *tp) 1662 { 1663 int t = ((tp->t_srtt >> 2) + tp->t_rttvar) >> (1 + 2); 1664 int nticks; 1665 1666 if (TCP_TIMER_ISARMED(tp, TCPT_REXMT)) 1667 panic("tcp_output REXMT"); 1668 /* 1669 * Start/restart persistance timer. 1670 */ 1671 if (t < tp->t_rttmin) 1672 t = tp->t_rttmin; 1673 TCPT_RANGESET(nticks, t * tcp_backoff[tp->t_rxtshift], 1674 TCPTV_PERSMIN, TCPTV_PERSMAX); 1675 TCP_TIMER_ARM(tp, TCPT_PERSIST, nticks); 1676 if (tp->t_rxtshift < TCP_MAXRXTSHIFT) 1677 tp->t_rxtshift++; 1678 } 1679
Indexes created Sun Sep 21 19:09:51 GMT 2025