tcp_subr.c revision 1.70
1 /* $NetBSD: tcp_subr.c,v 1.70 1999/07/09 22:57:22 thorpej 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 (c) 1997, 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Jason R. Thorpe and Kevin M. Lahey of the Numerical Aerospace Simulation 38 * Facility, NASA Ames Research Center. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 3. All advertising materials mentioning features or use of this software 49 * must display the following acknowledgement: 50 * This product includes software developed by the NetBSD 51 * Foundation, Inc. and its contributors. 52 * 4. Neither the name of The NetBSD Foundation nor the names of its 53 * contributors may be used to endorse or promote products derived 54 * from this software without specific prior written permission. 55 * 56 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 57 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 58 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 59 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 60 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 61 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 62 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 63 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 64 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 65 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 66 * POSSIBILITY OF SUCH DAMAGE. 67 */ 68 69 /* 70 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 71 * The Regents of the University of California. All rights reserved. 72 * 73 * Redistribution and use in source and binary forms, with or without 74 * modification, are permitted provided that the following conditions 75 * are met: 76 * 1. Redistributions of source code must retain the above copyright 77 * notice, this list of conditions and the following disclaimer. 78 * 2. Redistributions in binary form must reproduce the above copyright 79 * notice, this list of conditions and the following disclaimer in the 80 * documentation and/or other materials provided with the distribution. 81 * 3. All advertising materials mentioning features or use of this software 82 * must display the following acknowledgement: 83 * This product includes software developed by the University of 84 * California, Berkeley and its contributors. 85 * 4. Neither the name of the University nor the names of its contributors 86 * may be used to endorse or promote products derived from this software 87 * without specific prior written permission. 88 * 89 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 90 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 91 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 92 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 93 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 94 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 95 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 96 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 97 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 98 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 99 * SUCH DAMAGE. 100 * 101 * @(#)tcp_subr.c 8.2 (Berkeley) 5/24/95 102 */ 103 104 #include "opt_inet.h" 105 #include "opt_ipsec.h" 106 #include "opt_tcp_compat_42.h" 107 #include "rnd.h" 108 109 #include <sys/param.h> 110 #include <sys/proc.h> 111 #include <sys/systm.h> 112 #include <sys/malloc.h> 113 #include <sys/mbuf.h> 114 #include <sys/socket.h> 115 #include <sys/socketvar.h> 116 #include <sys/protosw.h> 117 #include <sys/errno.h> 118 #include <sys/kernel.h> 119 #include <sys/pool.h> 120 #if NRND > 0 121 #include <sys/rnd.h> 122 #endif 123 124 #include <net/route.h> 125 #include <net/if.h> 126 127 #include <netinet/in.h> 128 #include <netinet/in_systm.h> 129 #include <netinet/ip.h> 130 #include <netinet/in_pcb.h> 131 #include <netinet/ip_var.h> 132 #include <netinet/ip_icmp.h> 133 134 #ifdef INET6 135 #ifndef INET 136 #include <netinet/in.h> 137 #endif 138 #include <netinet/ip6.h> 139 #include <netinet6/in6_pcb.h> 140 #include <netinet6/ip6_var.h> 141 #endif 142 143 #include <netinet/tcp.h> 144 #include <netinet/tcp_fsm.h> 145 #include <netinet/tcp_seq.h> 146 #include <netinet/tcp_timer.h> 147 #include <netinet/tcp_var.h> 148 #include <netinet/tcpip.h> 149 150 #ifdef IPSEC 151 #include <netinet6/ipsec.h> 152 #include <netinet6/ah.h> 153 #ifdef IPSEC_ESP 154 #include <netinet6/esp.h> 155 #endif 156 #endif /*IPSEC*/ 157 158 #ifdef INET6 159 struct in6pcb tcb6; 160 #endif 161 162 /* patchable/settable parameters for tcp */ 163 int tcp_mssdflt = TCP_MSS; 164 int tcp_rttdflt = TCPTV_SRTTDFLT / PR_SLOWHZ; 165 int tcp_do_rfc1323 = 1; /* window scaling / timestamps (obsolete) */ 166 int tcp_do_sack = 1; /* selective acknowledgement */ 167 int tcp_do_win_scale = 1; /* RFC1323 window scaling */ 168 int tcp_do_timestamps = 1; /* RFC1323 timestamps */ 169 int tcp_do_newreno = 0; /* Use the New Reno algorithms */ 170 int tcp_ack_on_push = 0; /* set to enable immediate ACK-on-PUSH */ 171 int tcp_init_win = 1; 172 int tcp_mss_ifmtu = 0; 173 #ifdef TCP_COMPAT_42 174 int tcp_compat_42 = 1; 175 #else 176 int tcp_compat_42 = 0; 177 #endif 178 179 #ifndef TCBHASHSIZE 180 #define TCBHASHSIZE 128 181 #endif 182 int tcbhashsize = TCBHASHSIZE; 183 184 int tcp_freeq __P((struct tcpcb *)); 185 186 struct pool tcpcb_pool; 187 188 /* 189 * Tcp initialization 190 */ 191 void 192 tcp_init() 193 { 194 int hlen; 195 196 pool_init(&tcpcb_pool, sizeof(struct tcpcb), 0, 0, 0, "tcpcbpl", 197 0, NULL, NULL, M_PCB); 198 in_pcbinit(&tcbtable, tcbhashsize, tcbhashsize); 199 #ifdef INET6 200 tcb6.in6p_next = tcb6.in6p_prev = &tcb6; 201 #endif 202 LIST_INIT(&tcp_delacks); 203 204 hlen = sizeof(struct ip) + sizeof(struct tcphdr); 205 #ifdef INET6 206 if (sizeof(struct ip) < sizeof(struct ip6_hdr)) 207 hlen = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); 208 #endif 209 if (max_protohdr < hlen) 210 max_protohdr = hlen; 211 if (max_linkhdr + hlen > MHLEN) 212 panic("tcp_init"); 213 214 /* Initialize the compressed state engine. */ 215 syn_cache_init(); 216 } 217 218 /* 219 * Create template to be used to send tcp packets on a connection. 220 * Call after host entry created, allocates an mbuf and fills 221 * in a skeletal tcp/ip header, minimizing the amount of work 222 * necessary when the connection is used. 223 */ 224 struct mbuf * 225 tcp_template(tp) 226 struct tcpcb *tp; 227 { 228 register struct inpcb *inp = tp->t_inpcb; 229 #ifdef INET6 230 register struct in6pcb *in6p = tp->t_in6pcb; 231 #endif 232 register struct tcphdr *n; 233 register struct mbuf *m; 234 int hlen; 235 236 switch (tp->t_family) { 237 case AF_INET: 238 hlen = sizeof(struct ip); 239 if (inp) 240 break; 241 #ifdef INET6 242 if (in6p) { 243 /* mapped addr case */ 244 if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr) 245 && IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) 246 break; 247 } 248 #endif 249 return NULL; /*EINVAL*/ 250 #ifdef INET6 251 case AF_INET6: 252 hlen = sizeof(struct ip6_hdr); 253 if (in6p) { 254 /* more sainty check? */ 255 break; 256 } 257 return NULL; /*EINVAL*/ 258 #endif 259 default: 260 hlen = 0; /*pacify gcc*/ 261 return NULL; /*EAFNOSUPPORT*/ 262 } 263 if ((m = tp->t_template) == 0) { 264 MGETHDR(m, M_DONTWAIT, MT_HEADER); 265 if (m) { 266 MCLGET(m, M_DONTWAIT); 267 if ((m->m_flags & M_EXT) == 0) { 268 m_free(m); 269 m = NULL; 270 } 271 } 272 if (m == NULL) 273 return NULL; 274 m->m_len = hlen + sizeof(struct tcphdr); 275 } 276 bzero(mtod(m, caddr_t), m->m_len); 277 switch (tp->t_family) { 278 case AF_INET: 279 { 280 struct ipovly *ipov; 281 mtod(m, struct ip *)->ip_v = 4; 282 ipov = mtod(m, struct ipovly *); 283 ipov->ih_pr = IPPROTO_TCP; 284 ipov->ih_len = htons(sizeof(struct tcphdr)); 285 if (inp) { 286 ipov->ih_src = inp->inp_laddr; 287 ipov->ih_dst = inp->inp_faddr; 288 } 289 #ifdef INET6 290 else if (in6p) { 291 /* mapped addr case */ 292 bcopy(&in6p->in6p_laddr.s6_addr32[3], &ipov->ih_src, 293 sizeof(ipov->ih_src)); 294 bcopy(&in6p->in6p_faddr.s6_addr32[3], &ipov->ih_dst, 295 sizeof(ipov->ih_dst)); 296 } 297 #endif 298 break; 299 } 300 #ifdef INET6 301 case AF_INET6: 302 { 303 struct ip6_hdr *ip6; 304 mtod(m, struct ip *)->ip_v = 6; 305 ip6 = mtod(m, struct ip6_hdr *); 306 ip6->ip6_nxt = IPPROTO_TCP; 307 ip6->ip6_plen = htons(sizeof(struct tcphdr)); 308 ip6->ip6_src = in6p->in6p_laddr; 309 ip6->ip6_dst = in6p->in6p_faddr; 310 ip6->ip6_flow = in6p->in6p_flowinfo & IPV6_FLOWINFO_MASK; 311 if (ip6_auto_flowlabel) { 312 ip6->ip6_flow &= ~IPV6_FLOWLABEL_MASK; 313 ip6->ip6_flow |= 314 (htonl(ip6_flow_seq++) & IPV6_FLOWLABEL_MASK); 315 } 316 ip6->ip6_vfc = IPV6_VERSION; 317 break; 318 } 319 #endif 320 } 321 n = (struct tcphdr *)(mtod(m, caddr_t) + hlen); 322 if (inp) { 323 n->th_sport = inp->inp_lport; 324 n->th_dport = inp->inp_fport; 325 } 326 #ifdef INET6 327 else if (in6p) { 328 n->th_sport = in6p->in6p_lport; 329 n->th_dport = in6p->in6p_fport; 330 } 331 #endif 332 n->th_seq = 0; 333 n->th_ack = 0; 334 n->th_x2 = 0; 335 n->th_off = 5; 336 n->th_flags = 0; 337 n->th_win = 0; 338 n->th_sum = 0; 339 n->th_urp = 0; 340 return (m); 341 } 342 343 /* 344 * Send a single message to the TCP at address specified by 345 * the given TCP/IP header. If m == 0, then we make a copy 346 * of the tcpiphdr at ti and send directly to the addressed host. 347 * This is used to force keep alive messages out using the TCP 348 * template for a connection tp->t_template. If flags are given 349 * then we send a message back to the TCP which originated the 350 * segment ti, and discard the mbuf containing it and any other 351 * attached mbufs. 352 * 353 * In any case the ack and sequence number of the transmitted 354 * segment are as specified by the parameters. 355 */ 356 int 357 tcp_respond(tp, template, m, ack, seq, flags) 358 struct tcpcb *tp; 359 struct mbuf *template; /* XXX should be struct mbuf? */ 360 register struct mbuf *m; 361 tcp_seq ack, seq; 362 int flags; 363 { 364 #ifndef INET6 365 struct route iproute; 366 #else 367 struct route_in6 iproute; /* sizeof(route_in6) > sizeof(route) */ 368 #endif 369 struct route *ro; 370 struct rtentry *rt; 371 int error, tlen, win = 0; 372 int hlen; 373 struct ip *ip; 374 #ifdef INET6 375 struct ip6_hdr *ip6; 376 #endif 377 int family; /* family on packet, not inpcb/in6pcb! */ 378 struct tcphdr *th; 379 380 if (tp != NULL && (flags & TH_RST) == 0) { 381 if (tp->t_inpcb) 382 win = sbspace(&tp->t_inpcb->inp_socket->so_rcv); 383 #ifdef INET6 384 else if (tp->t_in6pcb) 385 win = sbspace(&tp->t_in6pcb->in6p_socket->so_rcv); 386 #endif 387 } 388 389 ip = NULL; 390 #ifdef INET6 391 ip6 = NULL; 392 #endif 393 if (m == 0) { 394 if (template 395 && template->m_len < hlen + sizeof(struct tcphdr)) { 396 if (m) 397 m_freem(m); 398 return EINVAL; 399 } 400 401 /* get family information from template */ 402 switch (mtod(template, struct ip *)->ip_v) { 403 case 4: 404 family = AF_INET; 405 hlen = sizeof(struct ip); 406 break; 407 #ifdef INET6 408 case 6: 409 family = AF_INET6; 410 hlen = sizeof(struct ip6_hdr); 411 break; 412 #endif 413 default: 414 if (m) 415 m_freem(m); 416 return EAFNOSUPPORT; 417 } 418 419 MGETHDR(m, M_DONTWAIT, MT_HEADER); 420 if (m) { 421 MCLGET(m, M_DONTWAIT); 422 if ((m->m_flags & M_EXT)) { 423 m_free(m); 424 m = NULL; 425 } 426 } 427 if (m == NULL) 428 return (ENOBUFS); 429 430 if (tcp_compat_42) 431 tlen = 1; 432 else 433 tlen = 0; 434 435 m->m_data += max_linkhdr; 436 bcopy(mtod(template, caddr_t), mtod(m, caddr_t), 437 template->m_len); 438 switch (family) { 439 case AF_INET: 440 ip = mtod(m, struct ip *); 441 th = (struct tcphdr *)(ip + 1); 442 break; 443 #ifdef INET6 444 case AF_INET6: 445 ip6 = mtod(m, struct ip6_hdr *); 446 th = (struct tcphdr *)(ip6 + 1); 447 break; 448 #endif 449 default: /*pacify gcc*/ 450 ip = NULL; 451 #ifdef INET6 452 ip6 = NULL; 453 #endif 454 th = NULL; 455 break; 456 } 457 flags = TH_ACK; 458 } else { 459 /* get family information from m */ 460 switch (mtod(m, struct ip *)->ip_v) { 461 case 4: 462 family = AF_INET; 463 hlen = sizeof(struct ip); 464 break; 465 #ifdef INET6 466 case 6: 467 family = AF_INET6; 468 hlen = sizeof(struct ip6_hdr); 469 break; 470 #endif 471 default: 472 if (m) 473 m_freem(m); 474 return EAFNOSUPPORT; 475 } 476 477 /* template pointer almost has no meaning */ 478 m_freem(m->m_next); 479 m->m_next = 0; 480 m->m_len = hlen + sizeof(struct tcphdr); 481 if ((m->m_flags & M_PKTHDR) == 0) { 482 printf("non PKTHDR to tcp_respond\n"); 483 m_freem(m); 484 return EINVAL; 485 } 486 487 tlen = 0; 488 #define xchg(a,b,type) { type t; t=a; a=b; b=t; } 489 switch (family) { 490 case AF_INET: 491 ip = mtod(m, struct ip *); 492 th = (struct tcphdr *)(ip + 1); 493 xchg(ip->ip_dst, ip->ip_src, struct in_addr); 494 break; 495 #ifdef INET6 496 case AF_INET6: 497 ip6 = mtod(m, struct ip6_hdr *); 498 th = (struct tcphdr *)(ip6 + 1); 499 xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr); 500 break; 501 #endif 502 } 503 xchg(th->th_dport, th->th_sport, u_int16_t); 504 #undef xchg 505 } 506 th->th_seq = htonl(seq); 507 th->th_ack = htonl(ack); 508 th->th_x2 = 0; 509 if ((flags & TH_SYN) == 0) { 510 if (tp) 511 th->th_win = htons((u_int16_t) (win >> tp->rcv_scale)); 512 else 513 th->th_win = htons((u_int16_t)win); 514 th->th_off = sizeof (struct tcphdr) >> 2; 515 tlen += sizeof (struct tcphdr); 516 } else 517 tlen += th->th_off << 2; 518 m->m_len = hlen + tlen; 519 m->m_pkthdr.len = hlen + tlen; 520 m->m_pkthdr.rcvif = (struct ifnet *) 0; 521 th->th_flags = flags; 522 th->th_urp = 0; 523 524 switch (family) { 525 case AF_INET: 526 { 527 struct ipovly *ipov = (struct ipovly *)ip; 528 bzero(ipov->ih_x1, sizeof ipov->ih_x1); 529 ipov->ih_len = htons((u_int16_t)tlen); 530 531 th->th_sum = 0; 532 th->th_sum = in_cksum(m, hlen + tlen); 533 ip->ip_len = hlen + tlen; /*will be flipped on output*/ 534 ip->ip_ttl = ip_defttl; 535 break; 536 } 537 #ifdef INET6 538 case AF_INET6: 539 { 540 th->th_sum = 0; 541 th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(struct ip6_hdr), 542 tlen); 543 ip6->ip6_plen = ntohs(tlen); 544 ip6->ip6_hlim = ip6_defhlim; 545 ip6->ip6_flow &= ~IPV6_FLOWINFO_MASK; 546 if (ip6_auto_flowlabel) { 547 ip6->ip6_flow |= 548 (htonl(ip6_flow_seq++) & IPV6_FLOWLABEL_MASK); 549 } 550 break; 551 } 552 #endif 553 } 554 555 #ifdef IPSEC 556 m->m_pkthdr.rcvif = NULL; 557 #endif /*IPSEC*/ 558 559 /* 560 * If we're doing Path MTU discovery, we need to set DF unless 561 * the route's MTU is locked. If we lack a route, we need to 562 * look it up now. 563 * 564 * ip_output() could do this for us, but it's convenient to just 565 * do it here unconditionally. 566 */ 567 if (tp != NULL && tp->t_inpcb != NULL) { 568 ro = &tp->t_inpcb->inp_route; 569 #ifdef IPSEC 570 m->m_pkthdr.rcvif = (struct ifnet *)tp->t_inpcb->inp_socket; 571 #endif 572 #ifdef DIAGNOSTIC 573 if (family != AF_INET) 574 panic("tcp_respond: address family mismatch"); 575 if (!in_hosteq(ip->ip_dst, tp->t_inpcb->inp_faddr)) { 576 panic("tcp_respond: ip_dst %x != inp_faddr %x", 577 ntohl(ip->ip_dst.s_addr), 578 ntohl(tp->t_inpcb->inp_faddr.s_addr)); 579 } 580 #endif 581 } 582 #ifdef INET6 583 else if (tp != NULL && tp->t_in6pcb != NULL) { 584 ro = (struct route *)&tp->t_in6pcb->in6p_route; 585 #ifdef IPSEC 586 m->m_pkthdr.rcvif = (struct ifnet *)tp->t_in6pcb->in6p_socket; 587 #endif 588 #ifdef DIAGNOSTIC 589 if (family == AF_INET) { 590 if (!IN6_IS_ADDR_V4MAPPED(&tp->t_in6pcb->in6p_faddr)) 591 panic("tcp_respond: not mapped addr"); 592 if (bcmp(&ip->ip_dst, 593 &tp->t_in6pcb->in6p_faddr.s6_addr32[3], 594 sizeof(ip->ip_dst)) != 0) { 595 panic("tcp_respond: ip_dst != in6p_faddr"); 596 } 597 } else if (family == AF_INET6) { 598 if (!IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &tp->t_in6pcb->in6p_faddr)) 599 panic("tcp_respond: ip6_dst != in6p_faddr"); 600 } else 601 panic("tcp_respond: address family mismatch"); 602 #endif 603 } 604 #endif 605 else { 606 ro = (struct route *)&iproute; 607 bzero(ro, sizeof(iproute)); 608 } 609 if ((rt = ro->ro_rt) == NULL || (rt->rt_flags & RTF_UP) == 0) { 610 if (ro->ro_rt != NULL) { 611 RTFREE(ro->ro_rt); 612 ro->ro_rt = NULL; 613 } 614 switch (family) { 615 case AF_INET: 616 { 617 struct sockaddr_in *dst; 618 dst = satosin(&ro->ro_dst); 619 dst->sin_family = AF_INET; 620 dst->sin_len = sizeof(*dst); 621 dst->sin_addr = ip->ip_dst; 622 break; 623 } 624 #ifdef INET6 625 case AF_INET6: 626 { 627 struct sockaddr_in6 *dst; 628 dst = satosin6(&ro->ro_dst); 629 bzero(dst, sizeof(*dst)); 630 dst->sin6_family = AF_INET6; 631 dst->sin6_len = sizeof(*dst); 632 dst->sin6_addr = ip6->ip6_dst; 633 break; 634 } 635 #endif 636 } 637 rtalloc(ro); 638 if ((rt = ro->ro_rt) == NULL) { 639 m_freem(m); 640 switch (family) { 641 case AF_INET: 642 ipstat.ips_noroute++; 643 break; 644 #ifdef INET6 645 case AF_INET6: 646 ip6stat.ip6s_noroute++; 647 break; 648 #endif 649 } 650 return (EHOSTUNREACH); 651 } 652 } 653 switch (family) { 654 case AF_INET: 655 if (ip_mtudisc != 0 && (rt->rt_rmx.rmx_locks & RTV_MTU) == 0) 656 ip->ip_off |= IP_DF; 657 658 error = ip_output(m, NULL, ro, 0, NULL); 659 break; 660 #ifdef INET6 661 case AF_INET6: 662 error = ip6_output(m, NULL, (struct route_in6 *)ro, 0, NULL); 663 break; 664 #endif 665 default: 666 error = EAFNOSUPPORT; 667 break; 668 } 669 670 if (ro == (struct route *)&iproute) { 671 RTFREE(ro->ro_rt); 672 ro->ro_rt = NULL; 673 } 674 675 return (error); 676 } 677 678 /* 679 * Create a new TCP control block, making an 680 * empty reassembly queue and hooking it to the argument 681 * protocol control block. 682 */ 683 struct tcpcb * 684 tcp_newtcpcb(family, aux) 685 int family; /* selects inpcb, or in6pcb */ 686 void *aux; 687 { 688 register struct tcpcb *tp; 689 690 switch (family) { 691 case PF_INET: 692 break; 693 #ifdef INET6 694 case PF_INET6: 695 break; 696 #endif 697 default: 698 return NULL; 699 } 700 701 tp = pool_get(&tcpcb_pool, PR_NOWAIT); 702 if (tp == NULL) 703 return (NULL); 704 bzero((caddr_t)tp, sizeof(struct tcpcb)); 705 LIST_INIT(&tp->segq); 706 LIST_INIT(&tp->timeq); 707 tp->t_family = family; /* may be overridden later on */ 708 tp->t_peermss = tcp_mssdflt; 709 tp->t_ourmss = tcp_mssdflt; 710 tp->t_segsz = tcp_mssdflt; 711 712 tp->t_flags = 0; 713 if (tcp_do_rfc1323 && tcp_do_win_scale) 714 tp->t_flags |= TF_REQ_SCALE; 715 if (tcp_do_rfc1323 && tcp_do_timestamps) 716 tp->t_flags |= TF_REQ_TSTMP; 717 if (tcp_do_sack == 2) 718 tp->t_flags |= TF_WILL_SACK; 719 else if (tcp_do_sack == 1) 720 tp->t_flags |= TF_WILL_SACK|TF_IGNR_RXSACK; 721 tp->t_flags |= TF_CANT_TXSACK; 722 switch (family) { 723 case PF_INET: 724 tp->t_inpcb = (struct inpcb *)aux; 725 break; 726 #ifdef INET6 727 case PF_INET6: 728 tp->t_in6pcb = (struct in6pcb *)aux; 729 break; 730 #endif 731 } 732 /* 733 * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no 734 * rtt estimate. Set rttvar so that srtt + 2 * rttvar gives 735 * reasonable initial retransmit time. 736 */ 737 tp->t_srtt = TCPTV_SRTTBASE; 738 tp->t_rttvar = tcp_rttdflt * PR_SLOWHZ << (TCP_RTTVAR_SHIFT + 2 - 1); 739 tp->t_rttmin = TCPTV_MIN; 740 TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp), 741 TCPTV_MIN, TCPTV_REXMTMAX); 742 tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; 743 tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT; 744 if (family == AF_INET) { 745 struct inpcb *inp = (struct inpcb *)aux; 746 inp->inp_ip.ip_ttl = ip_defttl; 747 inp->inp_ppcb = (caddr_t)tp; 748 } 749 #ifdef INET6 750 else if (family == AF_INET6) { 751 struct in6pcb *in6p = (struct in6pcb *)aux; 752 in6p->in6p_ip6.ip6_hlim = ip6_defhlim; 753 in6p->in6p_ppcb = (caddr_t)tp; 754 } 755 #endif 756 return (tp); 757 } 758 759 /* 760 * Drop a TCP connection, reporting 761 * the specified error. If connection is synchronized, 762 * then send a RST to peer. 763 */ 764 struct tcpcb * 765 tcp_drop(tp, errno) 766 register struct tcpcb *tp; 767 int errno; 768 { 769 struct socket *so; 770 771 if (tp->t_inpcb) 772 so = tp->t_inpcb->inp_socket; 773 #ifdef INET6 774 else if (tp->t_in6pcb) 775 so = tp->t_in6pcb->in6p_socket; 776 #endif 777 else 778 return NULL; 779 780 if (TCPS_HAVERCVDSYN(tp->t_state)) { 781 tp->t_state = TCPS_CLOSED; 782 (void) tcp_output(tp); 783 tcpstat.tcps_drops++; 784 } else 785 tcpstat.tcps_conndrops++; 786 if (errno == ETIMEDOUT && tp->t_softerror) 787 errno = tp->t_softerror; 788 so->so_error = errno; 789 return (tcp_close(tp)); 790 } 791 792 /* 793 * Close a TCP control block: 794 * discard all space held by the tcp 795 * discard internet protocol block 796 * wake up any sleepers 797 */ 798 struct tcpcb * 799 tcp_close(tp) 800 register struct tcpcb *tp; 801 { 802 struct inpcb *inp; 803 #ifdef INET6 804 struct in6pcb *in6p; 805 #endif 806 struct socket *so; 807 #ifdef RTV_RTT 808 register struct rtentry *rt; 809 #endif 810 struct route *ro; 811 812 inp = tp->t_inpcb; 813 #ifdef INET6 814 in6p = tp->t_in6pcb; 815 #endif 816 so = NULL; 817 ro = NULL; 818 if (inp) { 819 so = inp->inp_socket; 820 ro = &inp->inp_route; 821 } 822 #ifdef INET6 823 else if (in6p) { 824 so = in6p->in6p_socket; 825 ro = (struct route *)&in6p->in6p_route; 826 } 827 #endif 828 829 #ifdef RTV_RTT 830 /* 831 * If we sent enough data to get some meaningful characteristics, 832 * save them in the routing entry. 'Enough' is arbitrarily 833 * defined as the sendpipesize (default 4K) * 16. This would 834 * give us 16 rtt samples assuming we only get one sample per 835 * window (the usual case on a long haul net). 16 samples is 836 * enough for the srtt filter to converge to within 5% of the correct 837 * value; fewer samples and we could save a very bogus rtt. 838 * 839 * Don't update the default route's characteristics and don't 840 * update anything that the user "locked". 841 */ 842 if (SEQ_LT(tp->iss + so->so_snd.sb_hiwat * 16, tp->snd_max) && 843 ro && (rt = ro->ro_rt) && 844 !in_nullhost(satosin(rt_key(rt))->sin_addr)) { 845 register u_long i = 0; 846 847 if ((rt->rt_rmx.rmx_locks & RTV_RTT) == 0) { 848 i = tp->t_srtt * 849 ((RTM_RTTUNIT / PR_SLOWHZ) >> (TCP_RTT_SHIFT + 2)); 850 if (rt->rt_rmx.rmx_rtt && i) 851 /* 852 * filter this update to half the old & half 853 * the new values, converting scale. 854 * See route.h and tcp_var.h for a 855 * description of the scaling constants. 856 */ 857 rt->rt_rmx.rmx_rtt = 858 (rt->rt_rmx.rmx_rtt + i) / 2; 859 else 860 rt->rt_rmx.rmx_rtt = i; 861 } 862 if ((rt->rt_rmx.rmx_locks & RTV_RTTVAR) == 0) { 863 i = tp->t_rttvar * 864 ((RTM_RTTUNIT / PR_SLOWHZ) >> (TCP_RTTVAR_SHIFT + 2)); 865 if (rt->rt_rmx.rmx_rttvar && i) 866 rt->rt_rmx.rmx_rttvar = 867 (rt->rt_rmx.rmx_rttvar + i) / 2; 868 else 869 rt->rt_rmx.rmx_rttvar = i; 870 } 871 /* 872 * update the pipelimit (ssthresh) if it has been updated 873 * already or if a pipesize was specified & the threshhold 874 * got below half the pipesize. I.e., wait for bad news 875 * before we start updating, then update on both good 876 * and bad news. 877 */ 878 if (((rt->rt_rmx.rmx_locks & RTV_SSTHRESH) == 0 && 879 (i = tp->snd_ssthresh) && rt->rt_rmx.rmx_ssthresh) || 880 i < (rt->rt_rmx.rmx_sendpipe / 2)) { 881 /* 882 * convert the limit from user data bytes to 883 * packets then to packet data bytes. 884 */ 885 i = (i + tp->t_segsz / 2) / tp->t_segsz; 886 if (i < 2) 887 i = 2; 888 i *= (u_long)(tp->t_segsz + sizeof (struct tcpiphdr)); 889 if (rt->rt_rmx.rmx_ssthresh) 890 rt->rt_rmx.rmx_ssthresh = 891 (rt->rt_rmx.rmx_ssthresh + i) / 2; 892 else 893 rt->rt_rmx.rmx_ssthresh = i; 894 } 895 } 896 #endif /* RTV_RTT */ 897 /* free the reassembly queue, if any */ 898 TCP_REASS_LOCK(tp); 899 (void) tcp_freeq(tp); 900 TCP_REASS_UNLOCK(tp); 901 902 TCP_CLEAR_DELACK(tp); 903 904 if (tp->t_template) { 905 m_free(tp->t_template); 906 tp->t_template = NULL; 907 } 908 pool_put(&tcpcb_pool, tp); 909 if (inp) { 910 inp->inp_ppcb = 0; 911 soisdisconnected(so); 912 in_pcbdetach(inp); 913 } 914 #ifdef INET6 915 else if (in6p) { 916 in6p->in6p_ppcb = 0; 917 soisdisconnected(so); 918 in6_pcbdetach(in6p); 919 } 920 #endif 921 tcpstat.tcps_closed++; 922 return ((struct tcpcb *)0); 923 } 924 925 int 926 tcp_freeq(tp) 927 struct tcpcb *tp; 928 { 929 register struct ipqent *qe; 930 int rv = 0; 931 #ifdef TCPREASS_DEBUG 932 int i = 0; 933 #endif 934 935 TCP_REASS_LOCK_CHECK(tp); 936 937 while ((qe = tp->segq.lh_first) != NULL) { 938 #ifdef TCPREASS_DEBUG 939 printf("tcp_freeq[%p,%d]: %u:%u(%u) 0x%02x\n", 940 tp, i++, qe->ipqe_seq, qe->ipqe_seq + qe->ipqe_len, 941 qe->ipqe_len, qe->ipqe_flags & (TH_SYN|TH_FIN|TH_RST)); 942 #endif 943 LIST_REMOVE(qe, ipqe_q); 944 LIST_REMOVE(qe, ipqe_timeq); 945 m_freem(qe->ipqe_m); 946 pool_put(&ipqent_pool, qe); 947 rv = 1; 948 } 949 return (rv); 950 } 951 952 /* 953 * Protocol drain routine. Called when memory is in short supply. 954 */ 955 void 956 tcp_drain() 957 { 958 register struct inpcb *inp; 959 register struct tcpcb *tp; 960 961 /* 962 * Free the sequence queue of all TCP connections. 963 */ 964 inp = tcbtable.inpt_queue.cqh_first; 965 if (inp) /* XXX */ 966 for (; inp != (struct inpcb *)&tcbtable.inpt_queue; 967 inp = inp->inp_queue.cqe_next) { 968 if ((tp = intotcpcb(inp)) != NULL) { 969 /* 970 * We may be called from a device's interrupt 971 * context. If the tcpcb is already busy, 972 * just bail out now. 973 */ 974 if (tcp_reass_lock_try(tp) == 0) 975 continue; 976 if (tcp_freeq(tp)) 977 tcpstat.tcps_connsdrained++; 978 TCP_REASS_UNLOCK(tp); 979 } 980 } 981 } 982 983 /* 984 * Notify a tcp user of an asynchronous error; 985 * store error as soft error, but wake up user 986 * (for now, won't do anything until can select for soft error). 987 */ 988 void 989 tcp_notify(inp, error) 990 struct inpcb *inp; 991 int error; 992 { 993 register struct tcpcb *tp = (struct tcpcb *)inp->inp_ppcb; 994 register struct socket *so = inp->inp_socket; 995 996 /* 997 * Ignore some errors if we are hooked up. 998 * If connection hasn't completed, has retransmitted several times, 999 * and receives a second error, give up now. This is better 1000 * than waiting a long time to establish a connection that 1001 * can never complete. 1002 */ 1003 if (tp->t_state == TCPS_ESTABLISHED && 1004 (error == EHOSTUNREACH || error == ENETUNREACH || 1005 error == EHOSTDOWN)) { 1006 return; 1007 } else if (TCPS_HAVEESTABLISHED(tp->t_state) == 0 && 1008 tp->t_rxtshift > 3 && tp->t_softerror) 1009 so->so_error = error; 1010 else 1011 tp->t_softerror = error; 1012 wakeup((caddr_t) &so->so_timeo); 1013 sorwakeup(so); 1014 sowwakeup(so); 1015 } 1016 1017 #if defined(INET6) && !defined(TCP6) 1018 void 1019 tcp6_ctlinput(cmd, sa, ip6, m, off) 1020 int cmd; 1021 struct sockaddr *sa; 1022 register struct ip6_hdr *ip6; 1023 struct mbuf *m; 1024 int off; 1025 { 1026 (void)tcp_ctlinput(cmd, sa, (void *)ip6); 1027 } 1028 #endif 1029 1030 /* assumes that ip header and tcp header are contiguous on mbuf */ 1031 void * 1032 tcp_ctlinput(cmd, sa, v) 1033 int cmd; 1034 struct sockaddr *sa; 1035 register void *v; 1036 { 1037 register struct ip *ip = v; 1038 register struct tcphdr *th; 1039 extern int inetctlerrmap[]; 1040 void (*notify) __P((struct inpcb *, int)) = tcp_notify; 1041 int errno; 1042 int nmatch; 1043 1044 if ((unsigned)cmd >= PRC_NCMDS) 1045 return NULL; 1046 errno = inetctlerrmap[cmd]; 1047 if (cmd == PRC_QUENCH) 1048 notify = tcp_quench; 1049 else if (PRC_IS_REDIRECT(cmd)) 1050 notify = in_rtchange, ip = 0; 1051 else if (cmd == PRC_MSGSIZE && ip_mtudisc) 1052 notify = tcp_mtudisc, ip = 0; 1053 else if (cmd == PRC_HOSTDEAD) 1054 ip = 0; 1055 else if (errno == 0) 1056 return NULL; 1057 if (ip && ip->ip_v == 4 && sa->sa_family == AF_INET) { 1058 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 1059 nmatch = in_pcbnotify(&tcbtable, satosin(sa)->sin_addr, 1060 th->th_dport, ip->ip_src, th->th_sport, errno, notify); 1061 if (nmatch == 0 && syn_cache_count && 1062 (inetctlerrmap[cmd] == EHOSTUNREACH || 1063 inetctlerrmap[cmd] == ENETUNREACH || 1064 inetctlerrmap[cmd] == EHOSTDOWN)) { 1065 struct sockaddr_in sin; 1066 bzero(&sin, sizeof(sin)); 1067 sin.sin_len = sizeof(sin); 1068 sin.sin_family = AF_INET; 1069 sin.sin_port = th->th_sport; 1070 sin.sin_addr = ip->ip_src; 1071 syn_cache_unreach((struct sockaddr *)&sin, sa, th); 1072 } 1073 1074 /* XXX mapped address case */ 1075 } 1076 #ifdef INET6 1077 else if (ip && ip->ip_v == 6 && sa->sa_family == AF_INET6) { 1078 /* XXX do something for ip6 */ 1079 } 1080 #endif 1081 else { 1082 (void)in_pcbnotifyall(&tcbtable, satosin(sa)->sin_addr, errno, 1083 notify); 1084 #ifdef INET6 1085 /* XXX do something for ip6 */ 1086 #endif 1087 } 1088 return NULL; 1089 } 1090 1091 /* 1092 * When a source quence is received, we are being notifed of congestion. 1093 * Close the congestion window down to the Loss Window (one segment). 1094 * We will gradually open it again as we proceed. 1095 */ 1096 void 1097 tcp_quench(inp, errno) 1098 struct inpcb *inp; 1099 int errno; 1100 { 1101 struct tcpcb *tp = intotcpcb(inp); 1102 1103 if (tp) 1104 tp->snd_cwnd = tp->t_segsz; 1105 } 1106 1107 /* 1108 * On receipt of path MTU corrections, flush old route and replace it 1109 * with the new one. Retransmit all unacknowledged packets, to ensure 1110 * that all packets will be received. 1111 */ 1112 void 1113 tcp_mtudisc(inp, errno) 1114 struct inpcb *inp; 1115 int errno; 1116 { 1117 struct tcpcb *tp = intotcpcb(inp); 1118 struct rtentry *rt = in_pcbrtentry(inp); 1119 1120 if (tp != 0) { 1121 if (rt != 0) { 1122 /* 1123 * If this was not a host route, remove and realloc. 1124 */ 1125 if ((rt->rt_flags & RTF_HOST) == 0) { 1126 in_rtchange(inp, errno); 1127 if ((rt = in_pcbrtentry(inp)) == 0) 1128 return; 1129 } 1130 1131 /* 1132 * Slow start out of the error condition. We 1133 * use the MTU because we know it's smaller 1134 * than the previously transmitted segment. 1135 * 1136 * Note: This is more conservative than the 1137 * suggestion in draft-floyd-incr-init-win-03. 1138 */ 1139 if (rt->rt_rmx.rmx_mtu != 0) 1140 tp->snd_cwnd = 1141 TCP_INITIAL_WINDOW(tcp_init_win, 1142 rt->rt_rmx.rmx_mtu); 1143 } 1144 1145 /* 1146 * Resend unacknowledged packets. 1147 */ 1148 tp->snd_nxt = tp->snd_una; 1149 tcp_output(tp); 1150 } 1151 } 1152 1153 1154 /* 1155 * Compute the MSS to advertise to the peer. Called only during 1156 * the 3-way handshake. If we are the server (peer initiated 1157 * connection), we are called with a pointer to the interface 1158 * on which the SYN packet arrived. If we are the client (we 1159 * initiated connection), we are called with a pointer to the 1160 * interface out which this connection should go. 1161 */ 1162 u_long 1163 tcp_mss_to_advertise(ifp) 1164 const struct ifnet *ifp; 1165 { 1166 extern u_long in_maxmtu; 1167 u_long mss = 0; 1168 1169 /* 1170 * In order to avoid defeating path MTU discovery on the peer, 1171 * we advertise the max MTU of all attached networks as our MSS, 1172 * per RFC 1191, section 3.1. 1173 * 1174 * We provide the option to advertise just the MTU of 1175 * the interface on which we hope this connection will 1176 * be receiving. If we are responding to a SYN, we 1177 * will have a pretty good idea about this, but when 1178 * initiating a connection there is a bit more doubt. 1179 * 1180 * We also need to ensure that loopback has a large enough 1181 * MSS, as the loopback MTU is never included in in_maxmtu. 1182 */ 1183 1184 if (ifp != NULL) 1185 mss = ifp->if_mtu; 1186 1187 if (tcp_mss_ifmtu == 0) 1188 mss = max(in_maxmtu, mss); 1189 1190 if (mss > sizeof(struct tcpiphdr)) 1191 mss -= sizeof(struct tcpiphdr); 1192 1193 mss = max(tcp_mssdflt, mss); 1194 return (mss); 1195 } 1196 1197 /* 1198 * Set connection variables based on the peer's advertised MSS. 1199 * We are passed the TCPCB for the actual connection. If we 1200 * are the server, we are called by the compressed state engine 1201 * when the 3-way handshake is complete. If we are the client, 1202 * we are called when we recieve the SYN,ACK from the server. 1203 * 1204 * NOTE: Our advertised MSS value must be initialized in the TCPCB 1205 * before this routine is called! 1206 */ 1207 void 1208 tcp_mss_from_peer(tp, offer) 1209 struct tcpcb *tp; 1210 int offer; 1211 { 1212 struct socket *so; 1213 #if defined(RTV_SPIPE) || defined(RTV_SSTHRESH) 1214 struct rtentry *rt; 1215 #endif 1216 u_long bufsize; 1217 int mss; 1218 1219 so = NULL; 1220 rt = NULL; 1221 if (tp->t_inpcb) { 1222 so = tp->t_inpcb->inp_socket; 1223 #if defined(RTV_SPIPE) || defined(RTV_SSTHRESH) 1224 rt = in_pcbrtentry(tp->t_inpcb); 1225 #endif 1226 } 1227 #ifdef INET6 1228 else if (tp->t_in6pcb) { 1229 so = tp->t_in6pcb->in6p_socket; 1230 #if defined(RTV_SPIPE) || defined(RTV_SSTHRESH) 1231 #ifdef TCP6 1232 rt = NULL; 1233 #else 1234 rt = in6_pcbrtentry(tp->t_in6pcb); 1235 #endif 1236 #endif 1237 } 1238 #endif 1239 1240 /* 1241 * As per RFC1122, use the default MSS value, unless they 1242 * sent us an offer. Do not accept offers less than 32 bytes. 1243 */ 1244 mss = tcp_mssdflt; 1245 if (offer) 1246 mss = offer; 1247 mss = max(mss, 32); /* sanity */ 1248 tp->t_peermss = mss; 1249 mss -= tcp_optlen(tp); 1250 if (tp->t_inpcb) 1251 mss -= ip_optlen(tp->t_inpcb); 1252 #ifdef INET6 1253 else if (tp->t_in6pcb) 1254 mss -= ip6_optlen(tp->t_in6pcb); 1255 #endif 1256 1257 /* 1258 * If there's a pipesize, change the socket buffer to that size. 1259 * Make the socket buffer an integral number of MSS units. If 1260 * the MSS is larger than the socket buffer, artificially decrease 1261 * the MSS. 1262 */ 1263 #ifdef RTV_SPIPE 1264 if (rt != NULL && rt->rt_rmx.rmx_sendpipe != 0) 1265 bufsize = rt->rt_rmx.rmx_sendpipe; 1266 else 1267 #endif 1268 bufsize = so->so_snd.sb_hiwat; 1269 if (bufsize < mss) 1270 mss = bufsize; 1271 else { 1272 bufsize = roundup(bufsize, mss); 1273 if (bufsize > sb_max) 1274 bufsize = sb_max; 1275 (void) sbreserve(&so->so_snd, bufsize); 1276 } 1277 tp->t_segsz = mss; 1278 1279 #ifdef RTV_SSTHRESH 1280 if (rt != NULL && rt->rt_rmx.rmx_ssthresh) { 1281 /* 1282 * There's some sort of gateway or interface buffer 1283 * limit on the path. Use this to set the slow 1284 * start threshold, but set the threshold to no less 1285 * than 2 * MSS. 1286 */ 1287 tp->snd_ssthresh = max(2 * mss, rt->rt_rmx.rmx_ssthresh); 1288 } 1289 #endif 1290 } 1291 1292 /* 1293 * Processing necessary when a TCP connection is established. 1294 */ 1295 void 1296 tcp_established(tp) 1297 struct tcpcb *tp; 1298 { 1299 struct socket *so; 1300 #ifdef RTV_RPIPE 1301 struct rtentry *rt; 1302 #endif 1303 u_long bufsize; 1304 1305 so = NULL; 1306 rt = NULL; 1307 if (tp->t_inpcb) { 1308 so = tp->t_inpcb->inp_socket; 1309 #if defined(RTV_RPIPE) 1310 rt = in_pcbrtentry(tp->t_inpcb); 1311 #endif 1312 } 1313 #ifdef INET6 1314 else if (tp->t_in6pcb) { 1315 so = tp->t_in6pcb->in6p_socket; 1316 #if defined(RTV_RPIPE) 1317 #ifdef TCP6 1318 rt = NULL; 1319 #else 1320 rt = in6_pcbrtentry(tp->t_in6pcb); 1321 #endif 1322 #endif 1323 } 1324 #endif 1325 1326 tp->t_state = TCPS_ESTABLISHED; 1327 TCP_TIMER_ARM(tp, TCPT_KEEP, tcp_keepidle); 1328 1329 #ifdef RTV_RPIPE 1330 if (rt != NULL && rt->rt_rmx.rmx_recvpipe != 0) 1331 bufsize = rt->rt_rmx.rmx_recvpipe; 1332 else 1333 #endif 1334 bufsize = so->so_rcv.sb_hiwat; 1335 if (bufsize > tp->t_ourmss) { 1336 bufsize = roundup(bufsize, tp->t_ourmss); 1337 if (bufsize > sb_max) 1338 bufsize = sb_max; 1339 (void) sbreserve(&so->so_rcv, bufsize); 1340 } 1341 } 1342 1343 /* 1344 * Check if there's an initial rtt or rttvar. Convert from the 1345 * route-table units to scaled multiples of the slow timeout timer. 1346 * Called only during the 3-way handshake. 1347 */ 1348 void 1349 tcp_rmx_rtt(tp) 1350 struct tcpcb *tp; 1351 { 1352 #ifdef RTV_RTT 1353 struct rtentry *rt = NULL; 1354 int rtt; 1355 1356 if (tp->t_inpcb) 1357 rt = in_pcbrtentry(tp->t_inpcb); 1358 #ifdef INET6 1359 else if (tp->t_in6pcb) { 1360 #ifdef TCP6 1361 rt = NULL; 1362 #else 1363 rt = in6_pcbrtentry(tp->t_in6pcb); 1364 #endif 1365 } 1366 #endif 1367 if (rt == NULL) 1368 return; 1369 1370 if (tp->t_srtt == 0 && (rtt = rt->rt_rmx.rmx_rtt)) { 1371 /* 1372 * XXX The lock bit for MTU indicates that the value 1373 * is also a minimum value; this is subject to time. 1374 */ 1375 if (rt->rt_rmx.rmx_locks & RTV_RTT) 1376 TCPT_RANGESET(tp->t_rttmin, 1377 rtt / (RTM_RTTUNIT / PR_SLOWHZ), 1378 TCPTV_MIN, TCPTV_REXMTMAX); 1379 tp->t_srtt = rtt / 1380 ((RTM_RTTUNIT / PR_SLOWHZ) >> (TCP_RTT_SHIFT + 2)); 1381 if (rt->rt_rmx.rmx_rttvar) { 1382 tp->t_rttvar = rt->rt_rmx.rmx_rttvar / 1383 ((RTM_RTTUNIT / PR_SLOWHZ) >> 1384 (TCP_RTTVAR_SHIFT + 2)); 1385 } else { 1386 /* Default variation is +- 1 rtt */ 1387 tp->t_rttvar = 1388 tp->t_srtt >> (TCP_RTT_SHIFT - TCP_RTTVAR_SHIFT); 1389 } 1390 TCPT_RANGESET(tp->t_rxtcur, 1391 ((tp->t_srtt >> 2) + tp->t_rttvar) >> (1 + 2), 1392 tp->t_rttmin, TCPTV_REXMTMAX); 1393 } 1394 #endif 1395 } 1396 1397 tcp_seq tcp_iss_seq = 0; /* tcp initial seq # */ 1398 1399 /* 1400 * Get a new sequence value given a tcp control block 1401 */ 1402 tcp_seq 1403 tcp_new_iss(tp, len, addin) 1404 void *tp; 1405 u_long len; 1406 tcp_seq addin; 1407 { 1408 tcp_seq tcp_iss; 1409 1410 /* 1411 * Randomize. 1412 */ 1413 #if NRND > 0 1414 rnd_extract_data(&tcp_iss, sizeof(tcp_iss), RND_EXTRACT_ANY); 1415 #else 1416 tcp_iss = random(); 1417 #endif 1418 1419 /* 1420 * If we were asked to add some amount to a known value, 1421 * we will take a random value obtained above, mask off the upper 1422 * bits, and add in the known value. We also add in a constant to 1423 * ensure that we are at least a certain distance from the original 1424 * value. 1425 * 1426 * This is used when an old connection is in timed wait 1427 * and we have a new one coming in, for instance. 1428 */ 1429 if (addin != 0) { 1430 #ifdef TCPISS_DEBUG 1431 printf("Random %08x, ", tcp_iss); 1432 #endif 1433 tcp_iss &= TCP_ISS_RANDOM_MASK; 1434 tcp_iss += addin + TCP_ISSINCR; 1435 #ifdef TCPISS_DEBUG 1436 printf("Old ISS %08x, ISS %08x\n", addin, tcp_iss); 1437 #endif 1438 } else { 1439 tcp_iss &= TCP_ISS_RANDOM_MASK; 1440 tcp_iss += tcp_iss_seq; 1441 tcp_iss_seq += TCP_ISSINCR; 1442 #ifdef TCPISS_DEBUG 1443 printf("ISS %08x\n", tcp_iss); 1444 #endif 1445 } 1446 1447 if (tcp_compat_42) { 1448 /* 1449 * Limit it to the positive range for really old TCP 1450 * implementations. 1451 */ 1452 if (tcp_iss >= 0x80000000) 1453 tcp_iss &= 0x7fffffff; /* XXX */ 1454 } 1455 1456 return tcp_iss; 1457 } 1458 1459 #ifdef IPSEC 1460 /* compute ESP/AH header size for TCP, including outer IP header. */ 1461 size_t 1462 ipsec4_hdrsiz_tcp(tp) 1463 struct tcpcb *tp; 1464 { 1465 struct inpcb *inp; 1466 size_t hdrsiz; 1467 1468 /* XXX mapped addr case (tp->t_in6pcb) */ 1469 if (!tp || !tp->t_template || !(inp = tp->t_inpcb)) 1470 return 0; 1471 switch (tp->t_family) { 1472 case AF_INET: 1473 hdrsiz = ipsec4_hdrsiz(tp->t_template, inp); 1474 break; 1475 default: 1476 hdrsiz = 0; 1477 break; 1478 } 1479 1480 return hdrsiz; 1481 } 1482 1483 #if defined(INET6) && !defined(TCP6) 1484 size_t 1485 ipsec6_hdrsiz_tcp(tp) 1486 struct tcpcb *tp; 1487 { 1488 struct in6pcb *in6p; 1489 size_t hdrsiz; 1490 1491 if (!tp || !tp->t_template || !(in6p = tp->t_in6pcb)) 1492 return 0; 1493 switch (tp->t_family) { 1494 case AF_INET6: 1495 hdrsiz = ipsec6_hdrsiz(tp->t_template, in6p); 1496 break; 1497 case AF_INET: 1498 /* mapped address case - tricky */ 1499 default: 1500 hdrsiz = 0; 1501 break; 1502 } 1503 1504 return hdrsiz; 1505 } 1506 #endif 1507 #endif /*IPSEC*/ 1508 1509 /* 1510 * Determine the length of the TCP options for this connection. 1511 * 1512 * XXX: What do we do for SACK, when we add that? Just reserve 1513 * all of the space? Otherwise we can't exactly be incrementing 1514 * cwnd by an amount that varies depending on the amount we last 1515 * had to SACK! 1516 */ 1517 1518 u_int 1519 tcp_optlen(tp) 1520 struct tcpcb *tp; 1521 { 1522 if ((tp->t_flags & (TF_REQ_TSTMP|TF_RCVD_TSTMP|TF_NOOPT)) == 1523 (TF_REQ_TSTMP | TF_RCVD_TSTMP)) 1524 return TCPOLEN_TSTAMP_APPA; 1525 else 1526 return 0; 1527 } 1528
Indexes created Sun Sep 21 15:09:59 GMT 2025