tcp_subr.c revision 1.191
1 /* $NetBSD: tcp_subr.c,v 1.191 2005/05/29 21:41:23 christos 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, 2000, 2001 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. Neither the name of the University nor the names of its contributors 82 * may be used to endorse or promote products derived from this software 83 * without specific prior written permission. 84 * 85 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 86 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 87 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 88 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 89 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 90 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 91 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 92 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 93 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 94 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 95 * SUCH DAMAGE. 96 * 97 * @(#)tcp_subr.c 8.2 (Berkeley) 5/24/95 98 */ 99 100 #include <sys/cdefs.h> 101 __KERNEL_RCSID(0, "$NetBSD: tcp_subr.c,v 1.191 2005/05/29 21:41:23 christos Exp $"); 102 103 #include "opt_inet.h" 104 #include "opt_ipsec.h" 105 #include "opt_tcp_compat_42.h" 106 #include "opt_inet_csum.h" 107 #include "opt_mbuftrace.h" 108 #include "rnd.h" 109 110 #include <sys/param.h> 111 #include <sys/proc.h> 112 #include <sys/systm.h> 113 #include <sys/malloc.h> 114 #include <sys/mbuf.h> 115 #include <sys/socket.h> 116 #include <sys/socketvar.h> 117 #include <sys/protosw.h> 118 #include <sys/errno.h> 119 #include <sys/kernel.h> 120 #include <sys/pool.h> 121 #if NRND > 0 122 #include <sys/md5.h> 123 #include <sys/rnd.h> 124 #endif 125 126 #include <net/route.h> 127 #include <net/if.h> 128 129 #include <netinet/in.h> 130 #include <netinet/in_systm.h> 131 #include <netinet/ip.h> 132 #include <netinet/in_pcb.h> 133 #include <netinet/ip_var.h> 134 #include <netinet/ip_icmp.h> 135 136 #ifdef INET6 137 #ifndef INET 138 #include <netinet/in.h> 139 #endif 140 #include <netinet/ip6.h> 141 #include <netinet6/in6_pcb.h> 142 #include <netinet6/ip6_var.h> 143 #include <netinet6/in6_var.h> 144 #include <netinet6/ip6protosw.h> 145 #include <netinet/icmp6.h> 146 #include <netinet6/nd6.h> 147 #endif 148 149 #include <netinet/tcp.h> 150 #include <netinet/tcp_fsm.h> 151 #include <netinet/tcp_seq.h> 152 #include <netinet/tcp_timer.h> 153 #include <netinet/tcp_var.h> 154 #include <netinet/tcpip.h> 155 156 #ifdef IPSEC 157 #include <netinet6/ipsec.h> 158 #include <netkey/key.h> 159 #endif /*IPSEC*/ 160 161 #ifdef FAST_IPSEC 162 #include <netipsec/ipsec.h> 163 #include <netipsec/xform.h> 164 #ifdef INET6 165 #include <netipsec/ipsec6.h> 166 #endif 167 #include <netipsec/key.h> 168 #endif /* FAST_IPSEC*/ 169 170 171 struct inpcbtable tcbtable; /* head of queue of active tcpcb's */ 172 struct tcpstat tcpstat; /* tcp statistics */ 173 u_int32_t tcp_now; /* for RFC 1323 timestamps */ 174 175 /* patchable/settable parameters for tcp */ 176 int tcp_mssdflt = TCP_MSS; 177 int tcp_rttdflt = TCPTV_SRTTDFLT / PR_SLOWHZ; 178 int tcp_do_rfc1323 = 1; /* window scaling / timestamps (obsolete) */ 179 #if NRND > 0 180 int tcp_do_rfc1948 = 0; /* ISS by cryptographic hash */ 181 #endif 182 int tcp_do_sack = 1; /* selective acknowledgement */ 183 int tcp_do_win_scale = 1; /* RFC1323 window scaling */ 184 int tcp_do_timestamps = 1; /* RFC1323 timestamps */ 185 int tcp_do_newreno = 1; /* Use the New Reno algorithms */ 186 int tcp_ack_on_push = 0; /* set to enable immediate ACK-on-PUSH */ 187 #ifndef TCP_INIT_WIN 188 #define TCP_INIT_WIN 0 /* initial slow start window */ 189 #endif 190 #ifndef TCP_INIT_WIN_LOCAL 191 #define TCP_INIT_WIN_LOCAL 4 /* initial slow start window for local nets */ 192 #endif 193 int tcp_init_win = TCP_INIT_WIN; 194 int tcp_init_win_local = TCP_INIT_WIN_LOCAL; 195 int tcp_mss_ifmtu = 0; 196 #ifdef TCP_COMPAT_42 197 int tcp_compat_42 = 1; 198 #else 199 int tcp_compat_42 = 0; 200 #endif 201 int tcp_rst_ppslim = 100; /* 100pps */ 202 int tcp_ackdrop_ppslim = 100; /* 100pps */ 203 int tcp_sack_tp_maxholes = 32; 204 int tcp_sack_globalmaxholes = 1024; 205 int tcp_sack_globalholes = 0; 206 207 208 /* tcb hash */ 209 #ifndef TCBHASHSIZE 210 #define TCBHASHSIZE 128 211 #endif 212 int tcbhashsize = TCBHASHSIZE; 213 214 /* syn hash parameters */ 215 #define TCP_SYN_HASH_SIZE 293 216 #define TCP_SYN_BUCKET_SIZE 35 217 int tcp_syn_cache_size = TCP_SYN_HASH_SIZE; 218 int tcp_syn_cache_limit = TCP_SYN_HASH_SIZE*TCP_SYN_BUCKET_SIZE; 219 int tcp_syn_bucket_limit = 3*TCP_SYN_BUCKET_SIZE; 220 struct syn_cache_head tcp_syn_cache[TCP_SYN_HASH_SIZE]; 221 222 int tcp_freeq(struct tcpcb *); 223 224 #ifdef INET 225 void tcp_mtudisc_callback(struct in_addr); 226 #endif 227 #ifdef INET6 228 void tcp6_mtudisc_callback(struct in6_addr *); 229 #endif 230 231 void tcp_mtudisc(struct inpcb *, int); 232 #ifdef INET6 233 void tcp6_mtudisc(struct in6pcb *, int); 234 #endif 235 236 POOL_INIT(tcpcb_pool, sizeof(struct tcpcb), 0, 0, 0, "tcpcbpl", NULL); 237 238 #ifdef TCP_CSUM_COUNTERS 239 #include <sys/device.h> 240 241 struct evcnt tcp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 242 NULL, "tcp", "hwcsum bad"); 243 struct evcnt tcp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 244 NULL, "tcp", "hwcsum ok"); 245 struct evcnt tcp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 246 NULL, "tcp", "hwcsum data"); 247 struct evcnt tcp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 248 NULL, "tcp", "swcsum"); 249 250 EVCNT_ATTACH_STATIC(tcp_hwcsum_bad); 251 EVCNT_ATTACH_STATIC(tcp_hwcsum_ok); 252 EVCNT_ATTACH_STATIC(tcp_hwcsum_data); 253 EVCNT_ATTACH_STATIC(tcp_swcsum); 254 #endif /* TCP_CSUM_COUNTERS */ 255 256 257 #ifdef TCP_OUTPUT_COUNTERS 258 #include <sys/device.h> 259 260 struct evcnt tcp_output_bigheader = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 261 NULL, "tcp", "output big header"); 262 struct evcnt tcp_output_predict_hit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 263 NULL, "tcp", "output predict hit"); 264 struct evcnt tcp_output_predict_miss = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 265 NULL, "tcp", "output predict miss"); 266 struct evcnt tcp_output_copysmall = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 267 NULL, "tcp", "output copy small"); 268 struct evcnt tcp_output_copybig = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 269 NULL, "tcp", "output copy big"); 270 struct evcnt tcp_output_refbig = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 271 NULL, "tcp", "output reference big"); 272 273 EVCNT_ATTACH_STATIC(tcp_output_bigheader); 274 EVCNT_ATTACH_STATIC(tcp_output_predict_hit); 275 EVCNT_ATTACH_STATIC(tcp_output_predict_miss); 276 EVCNT_ATTACH_STATIC(tcp_output_copysmall); 277 EVCNT_ATTACH_STATIC(tcp_output_copybig); 278 EVCNT_ATTACH_STATIC(tcp_output_refbig); 279 280 #endif /* TCP_OUTPUT_COUNTERS */ 281 282 #ifdef TCP_REASS_COUNTERS 283 #include <sys/device.h> 284 285 struct evcnt tcp_reass_ = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 286 NULL, "tcp_reass", "calls"); 287 struct evcnt tcp_reass_empty = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 288 &tcp_reass_, "tcp_reass", "insert into empty queue"); 289 struct evcnt tcp_reass_iteration[8] = { 290 EVCNT_INITIALIZER(EVCNT_TYPE_MISC, &tcp_reass_, "tcp_reass", ">7 iterations"), 291 EVCNT_INITIALIZER(EVCNT_TYPE_MISC, &tcp_reass_, "tcp_reass", "1 iteration"), 292 EVCNT_INITIALIZER(EVCNT_TYPE_MISC, &tcp_reass_, "tcp_reass", "2 iterations"), 293 EVCNT_INITIALIZER(EVCNT_TYPE_MISC, &tcp_reass_, "tcp_reass", "3 iterations"), 294 EVCNT_INITIALIZER(EVCNT_TYPE_MISC, &tcp_reass_, "tcp_reass", "4 iterations"), 295 EVCNT_INITIALIZER(EVCNT_TYPE_MISC, &tcp_reass_, "tcp_reass", "5 iterations"), 296 EVCNT_INITIALIZER(EVCNT_TYPE_MISC, &tcp_reass_, "tcp_reass", "6 iterations"), 297 EVCNT_INITIALIZER(EVCNT_TYPE_MISC, &tcp_reass_, "tcp_reass", "7 iterations"), 298 }; 299 struct evcnt tcp_reass_prependfirst = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 300 &tcp_reass_, "tcp_reass", "prepend to first"); 301 struct evcnt tcp_reass_prepend = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 302 &tcp_reass_, "tcp_reass", "prepend"); 303 struct evcnt tcp_reass_insert = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 304 &tcp_reass_, "tcp_reass", "insert"); 305 struct evcnt tcp_reass_inserttail = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 306 &tcp_reass_, "tcp_reass", "insert at tail"); 307 struct evcnt tcp_reass_append = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 308 &tcp_reass_, "tcp_reass", "append"); 309 struct evcnt tcp_reass_appendtail = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 310 &tcp_reass_, "tcp_reass", "append to tail fragment"); 311 struct evcnt tcp_reass_overlaptail = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 312 &tcp_reass_, "tcp_reass", "overlap at end"); 313 struct evcnt tcp_reass_overlapfront = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 314 &tcp_reass_, "tcp_reass", "overlap at start"); 315 struct evcnt tcp_reass_segdup = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 316 &tcp_reass_, "tcp_reass", "duplicate segment"); 317 struct evcnt tcp_reass_fragdup = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 318 &tcp_reass_, "tcp_reass", "duplicate fragment"); 319 320 EVCNT_ATTACH_STATIC(tcp_reass_); 321 EVCNT_ATTACH_STATIC(tcp_reass_empty); 322 EVCNT_ATTACH_STATIC2(tcp_reass_iteration, 0); 323 EVCNT_ATTACH_STATIC2(tcp_reass_iteration, 1); 324 EVCNT_ATTACH_STATIC2(tcp_reass_iteration, 2); 325 EVCNT_ATTACH_STATIC2(tcp_reass_iteration, 3); 326 EVCNT_ATTACH_STATIC2(tcp_reass_iteration, 4); 327 EVCNT_ATTACH_STATIC2(tcp_reass_iteration, 5); 328 EVCNT_ATTACH_STATIC2(tcp_reass_iteration, 6); 329 EVCNT_ATTACH_STATIC2(tcp_reass_iteration, 7); 330 EVCNT_ATTACH_STATIC(tcp_reass_prependfirst); 331 EVCNT_ATTACH_STATIC(tcp_reass_prepend); 332 EVCNT_ATTACH_STATIC(tcp_reass_insert); 333 EVCNT_ATTACH_STATIC(tcp_reass_inserttail); 334 EVCNT_ATTACH_STATIC(tcp_reass_append); 335 EVCNT_ATTACH_STATIC(tcp_reass_appendtail); 336 EVCNT_ATTACH_STATIC(tcp_reass_overlaptail); 337 EVCNT_ATTACH_STATIC(tcp_reass_overlapfront); 338 EVCNT_ATTACH_STATIC(tcp_reass_segdup); 339 EVCNT_ATTACH_STATIC(tcp_reass_fragdup); 340 341 #endif /* TCP_REASS_COUNTERS */ 342 343 #ifdef MBUFTRACE 344 struct mowner tcp_mowner = { "tcp" }; 345 struct mowner tcp_rx_mowner = { "tcp", "rx" }; 346 struct mowner tcp_tx_mowner = { "tcp", "tx" }; 347 #endif 348 349 /* 350 * Tcp initialization 351 */ 352 void 353 tcp_init(void) 354 { 355 int hlen; 356 357 /* Initialize the TCPCB template. */ 358 tcp_tcpcb_template(); 359 360 in_pcbinit(&tcbtable, tcbhashsize, tcbhashsize); 361 362 hlen = sizeof(struct ip) + sizeof(struct tcphdr); 363 #ifdef INET6 364 if (sizeof(struct ip) < sizeof(struct ip6_hdr)) 365 hlen = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); 366 #endif 367 if (max_protohdr < hlen) 368 max_protohdr = hlen; 369 if (max_linkhdr + hlen > MHLEN) 370 panic("tcp_init"); 371 372 #ifdef INET 373 icmp_mtudisc_callback_register(tcp_mtudisc_callback); 374 #endif 375 #ifdef INET6 376 icmp6_mtudisc_callback_register(tcp6_mtudisc_callback); 377 #endif 378 379 /* Initialize timer state. */ 380 tcp_timer_init(); 381 382 /* Initialize the compressed state engine. */ 383 syn_cache_init(); 384 385 MOWNER_ATTACH(&tcp_tx_mowner); 386 MOWNER_ATTACH(&tcp_rx_mowner); 387 MOWNER_ATTACH(&tcp_mowner); 388 } 389 390 /* 391 * Create template to be used to send tcp packets on a connection. 392 * Call after host entry created, allocates an mbuf and fills 393 * in a skeletal tcp/ip header, minimizing the amount of work 394 * necessary when the connection is used. 395 */ 396 struct mbuf * 397 tcp_template(struct tcpcb *tp) 398 { 399 struct inpcb *inp = tp->t_inpcb; 400 #ifdef INET6 401 struct in6pcb *in6p = tp->t_in6pcb; 402 #endif 403 struct tcphdr *n; 404 struct mbuf *m; 405 int hlen; 406 407 switch (tp->t_family) { 408 case AF_INET: 409 hlen = sizeof(struct ip); 410 if (inp) 411 break; 412 #ifdef INET6 413 if (in6p) { 414 /* mapped addr case */ 415 if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr) 416 && IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) 417 break; 418 } 419 #endif 420 return NULL; /*EINVAL*/ 421 #ifdef INET6 422 case AF_INET6: 423 hlen = sizeof(struct ip6_hdr); 424 if (in6p) { 425 /* more sainty check? */ 426 break; 427 } 428 return NULL; /*EINVAL*/ 429 #endif 430 default: 431 hlen = 0; /*pacify gcc*/ 432 return NULL; /*EAFNOSUPPORT*/ 433 } 434 #ifdef DIAGNOSTIC 435 if (hlen + sizeof(struct tcphdr) > MCLBYTES) 436 panic("mclbytes too small for t_template"); 437 #endif 438 m = tp->t_template; 439 if (m && m->m_len == hlen + sizeof(struct tcphdr)) 440 ; 441 else { 442 if (m) 443 m_freem(m); 444 m = tp->t_template = NULL; 445 MGETHDR(m, M_DONTWAIT, MT_HEADER); 446 if (m && hlen + sizeof(struct tcphdr) > MHLEN) { 447 MCLGET(m, M_DONTWAIT); 448 if ((m->m_flags & M_EXT) == 0) { 449 m_free(m); 450 m = NULL; 451 } 452 } 453 if (m == NULL) 454 return NULL; 455 MCLAIM(m, &tcp_mowner); 456 m->m_pkthdr.len = m->m_len = hlen + sizeof(struct tcphdr); 457 } 458 459 bzero(mtod(m, caddr_t), m->m_len); 460 461 n = (struct tcphdr *)(mtod(m, caddr_t) + hlen); 462 463 switch (tp->t_family) { 464 case AF_INET: 465 { 466 struct ipovly *ipov; 467 mtod(m, struct ip *)->ip_v = 4; 468 mtod(m, struct ip *)->ip_hl = hlen >> 2; 469 ipov = mtod(m, struct ipovly *); 470 ipov->ih_pr = IPPROTO_TCP; 471 ipov->ih_len = htons(sizeof(struct tcphdr)); 472 if (inp) { 473 ipov->ih_src = inp->inp_laddr; 474 ipov->ih_dst = inp->inp_faddr; 475 } 476 #ifdef INET6 477 else if (in6p) { 478 /* mapped addr case */ 479 bcopy(&in6p->in6p_laddr.s6_addr32[3], &ipov->ih_src, 480 sizeof(ipov->ih_src)); 481 bcopy(&in6p->in6p_faddr.s6_addr32[3], &ipov->ih_dst, 482 sizeof(ipov->ih_dst)); 483 } 484 #endif 485 /* 486 * Compute the pseudo-header portion of the checksum 487 * now. We incrementally add in the TCP option and 488 * payload lengths later, and then compute the TCP 489 * checksum right before the packet is sent off onto 490 * the wire. 491 */ 492 n->th_sum = in_cksum_phdr(ipov->ih_src.s_addr, 493 ipov->ih_dst.s_addr, 494 htons(sizeof(struct tcphdr) + IPPROTO_TCP)); 495 break; 496 } 497 #ifdef INET6 498 case AF_INET6: 499 { 500 struct ip6_hdr *ip6; 501 mtod(m, struct ip *)->ip_v = 6; 502 ip6 = mtod(m, struct ip6_hdr *); 503 ip6->ip6_nxt = IPPROTO_TCP; 504 ip6->ip6_plen = htons(sizeof(struct tcphdr)); 505 ip6->ip6_src = in6p->in6p_laddr; 506 ip6->ip6_dst = in6p->in6p_faddr; 507 ip6->ip6_flow = in6p->in6p_flowinfo & IPV6_FLOWINFO_MASK; 508 if (ip6_auto_flowlabel) { 509 ip6->ip6_flow &= ~IPV6_FLOWLABEL_MASK; 510 ip6->ip6_flow |= 511 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK); 512 } 513 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 514 ip6->ip6_vfc |= IPV6_VERSION; 515 516 /* 517 * Compute the pseudo-header portion of the checksum 518 * now. We incrementally add in the TCP option and 519 * payload lengths later, and then compute the TCP 520 * checksum right before the packet is sent off onto 521 * the wire. 522 */ 523 n->th_sum = in6_cksum_phdr(&in6p->in6p_laddr, 524 &in6p->in6p_faddr, htonl(sizeof(struct tcphdr)), 525 htonl(IPPROTO_TCP)); 526 break; 527 } 528 #endif 529 } 530 if (inp) { 531 n->th_sport = inp->inp_lport; 532 n->th_dport = inp->inp_fport; 533 } 534 #ifdef INET6 535 else if (in6p) { 536 n->th_sport = in6p->in6p_lport; 537 n->th_dport = in6p->in6p_fport; 538 } 539 #endif 540 n->th_seq = 0; 541 n->th_ack = 0; 542 n->th_x2 = 0; 543 n->th_off = 5; 544 n->th_flags = 0; 545 n->th_win = 0; 546 n->th_urp = 0; 547 return (m); 548 } 549 550 /* 551 * Send a single message to the TCP at address specified by 552 * the given TCP/IP header. If m == 0, then we make a copy 553 * of the tcpiphdr at ti and send directly to the addressed host. 554 * This is used to force keep alive messages out using the TCP 555 * template for a connection tp->t_template. If flags are given 556 * then we send a message back to the TCP which originated the 557 * segment ti, and discard the mbuf containing it and any other 558 * attached mbufs. 559 * 560 * In any case the ack and sequence number of the transmitted 561 * segment are as specified by the parameters. 562 */ 563 int 564 tcp_respond(struct tcpcb *tp, struct mbuf *template, struct mbuf *m, 565 struct tcphdr *th0, tcp_seq ack, tcp_seq seq, int flags) 566 { 567 struct route *ro; 568 int error, tlen, win = 0; 569 int hlen; 570 struct ip *ip; 571 #ifdef INET6 572 struct ip6_hdr *ip6; 573 #endif 574 int family; /* family on packet, not inpcb/in6pcb! */ 575 struct tcphdr *th; 576 struct socket *so; 577 578 if (tp != NULL && (flags & TH_RST) == 0) { 579 #ifdef DIAGNOSTIC 580 if (tp->t_inpcb && tp->t_in6pcb) 581 panic("tcp_respond: both t_inpcb and t_in6pcb are set"); 582 #endif 583 #ifdef INET 584 if (tp->t_inpcb) 585 win = sbspace(&tp->t_inpcb->inp_socket->so_rcv); 586 #endif 587 #ifdef INET6 588 if (tp->t_in6pcb) 589 win = sbspace(&tp->t_in6pcb->in6p_socket->so_rcv); 590 #endif 591 } 592 593 th = NULL; /* Quell uninitialized warning */ 594 ip = NULL; 595 #ifdef INET6 596 ip6 = NULL; 597 #endif 598 if (m == 0) { 599 if (!template) 600 return EINVAL; 601 602 /* get family information from template */ 603 switch (mtod(template, struct ip *)->ip_v) { 604 case 4: 605 family = AF_INET; 606 hlen = sizeof(struct ip); 607 break; 608 #ifdef INET6 609 case 6: 610 family = AF_INET6; 611 hlen = sizeof(struct ip6_hdr); 612 break; 613 #endif 614 default: 615 return EAFNOSUPPORT; 616 } 617 618 MGETHDR(m, M_DONTWAIT, MT_HEADER); 619 if (m) { 620 MCLAIM(m, &tcp_tx_mowner); 621 MCLGET(m, M_DONTWAIT); 622 if ((m->m_flags & M_EXT) == 0) { 623 m_free(m); 624 m = NULL; 625 } 626 } 627 if (m == NULL) 628 return (ENOBUFS); 629 630 if (tcp_compat_42) 631 tlen = 1; 632 else 633 tlen = 0; 634 635 m->m_data += max_linkhdr; 636 bcopy(mtod(template, caddr_t), mtod(m, caddr_t), 637 template->m_len); 638 switch (family) { 639 case AF_INET: 640 ip = mtod(m, struct ip *); 641 th = (struct tcphdr *)(ip + 1); 642 break; 643 #ifdef INET6 644 case AF_INET6: 645 ip6 = mtod(m, struct ip6_hdr *); 646 th = (struct tcphdr *)(ip6 + 1); 647 break; 648 #endif 649 #if 0 650 default: 651 /* noone will visit here */ 652 m_freem(m); 653 return EAFNOSUPPORT; 654 #endif 655 } 656 flags = TH_ACK; 657 } else { 658 659 if ((m->m_flags & M_PKTHDR) == 0) { 660 #if 0 661 printf("non PKTHDR to tcp_respond\n"); 662 #endif 663 m_freem(m); 664 return EINVAL; 665 } 666 #ifdef DIAGNOSTIC 667 if (!th0) 668 panic("th0 == NULL in tcp_respond"); 669 #endif 670 671 /* get family information from m */ 672 switch (mtod(m, struct ip *)->ip_v) { 673 case 4: 674 family = AF_INET; 675 hlen = sizeof(struct ip); 676 ip = mtod(m, struct ip *); 677 break; 678 #ifdef INET6 679 case 6: 680 family = AF_INET6; 681 hlen = sizeof(struct ip6_hdr); 682 ip6 = mtod(m, struct ip6_hdr *); 683 break; 684 #endif 685 default: 686 m_freem(m); 687 return EAFNOSUPPORT; 688 } 689 /* clear h/w csum flags inherited from rx packet */ 690 m->m_pkthdr.csum_flags = 0; 691 692 if ((flags & TH_SYN) == 0 || sizeof(*th0) > (th0->th_off << 2)) 693 tlen = sizeof(*th0); 694 else 695 tlen = th0->th_off << 2; 696 697 if (m->m_len > hlen + tlen && (m->m_flags & M_EXT) == 0 && 698 mtod(m, caddr_t) + hlen == (caddr_t)th0) { 699 m->m_len = hlen + tlen; 700 m_freem(m->m_next); 701 m->m_next = NULL; 702 } else { 703 struct mbuf *n; 704 705 #ifdef DIAGNOSTIC 706 if (max_linkhdr + hlen + tlen > MCLBYTES) { 707 m_freem(m); 708 return EMSGSIZE; 709 } 710 #endif 711 MGETHDR(n, M_DONTWAIT, MT_HEADER); 712 if (n && max_linkhdr + hlen + tlen > MHLEN) { 713 MCLGET(n, M_DONTWAIT); 714 if ((n->m_flags & M_EXT) == 0) { 715 m_freem(n); 716 n = NULL; 717 } 718 } 719 if (!n) { 720 m_freem(m); 721 return ENOBUFS; 722 } 723 724 MCLAIM(n, &tcp_tx_mowner); 725 n->m_data += max_linkhdr; 726 n->m_len = hlen + tlen; 727 m_copyback(n, 0, hlen, mtod(m, caddr_t)); 728 m_copyback(n, hlen, tlen, (caddr_t)th0); 729 730 m_freem(m); 731 m = n; 732 n = NULL; 733 } 734 735 #define xchg(a,b,type) { type t; t=a; a=b; b=t; } 736 switch (family) { 737 case AF_INET: 738 ip = mtod(m, struct ip *); 739 th = (struct tcphdr *)(ip + 1); 740 ip->ip_p = IPPROTO_TCP; 741 xchg(ip->ip_dst, ip->ip_src, struct in_addr); 742 ip->ip_p = IPPROTO_TCP; 743 break; 744 #ifdef INET6 745 case AF_INET6: 746 ip6 = mtod(m, struct ip6_hdr *); 747 th = (struct tcphdr *)(ip6 + 1); 748 ip6->ip6_nxt = IPPROTO_TCP; 749 xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr); 750 ip6->ip6_nxt = IPPROTO_TCP; 751 break; 752 #endif 753 #if 0 754 default: 755 /* noone will visit here */ 756 m_freem(m); 757 return EAFNOSUPPORT; 758 #endif 759 } 760 xchg(th->th_dport, th->th_sport, u_int16_t); 761 #undef xchg 762 tlen = 0; /*be friendly with the following code*/ 763 } 764 th->th_seq = htonl(seq); 765 th->th_ack = htonl(ack); 766 th->th_x2 = 0; 767 if ((flags & TH_SYN) == 0) { 768 if (tp) 769 win >>= tp->rcv_scale; 770 if (win > TCP_MAXWIN) 771 win = TCP_MAXWIN; 772 th->th_win = htons((u_int16_t)win); 773 th->th_off = sizeof (struct tcphdr) >> 2; 774 tlen += sizeof(*th); 775 } else 776 tlen += th->th_off << 2; 777 m->m_len = hlen + tlen; 778 m->m_pkthdr.len = hlen + tlen; 779 m->m_pkthdr.rcvif = (struct ifnet *) 0; 780 th->th_flags = flags; 781 th->th_urp = 0; 782 783 switch (family) { 784 #ifdef INET 785 case AF_INET: 786 { 787 struct ipovly *ipov = (struct ipovly *)ip; 788 bzero(ipov->ih_x1, sizeof ipov->ih_x1); 789 ipov->ih_len = htons((u_int16_t)tlen); 790 791 th->th_sum = 0; 792 th->th_sum = in_cksum(m, hlen + tlen); 793 ip->ip_len = htons(hlen + tlen); 794 ip->ip_ttl = ip_defttl; 795 break; 796 } 797 #endif 798 #ifdef INET6 799 case AF_INET6: 800 { 801 th->th_sum = 0; 802 th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(struct ip6_hdr), 803 tlen); 804 ip6->ip6_plen = htons(tlen); 805 if (tp && tp->t_in6pcb) { 806 struct ifnet *oifp; 807 ro = (struct route *)&tp->t_in6pcb->in6p_route; 808 oifp = ro->ro_rt ? ro->ro_rt->rt_ifp : NULL; 809 ip6->ip6_hlim = in6_selecthlim(tp->t_in6pcb, oifp); 810 } else 811 ip6->ip6_hlim = ip6_defhlim; 812 ip6->ip6_flow &= ~IPV6_FLOWINFO_MASK; 813 if (ip6_auto_flowlabel) { 814 ip6->ip6_flow |= 815 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK); 816 } 817 break; 818 } 819 #endif 820 } 821 822 if (tp && tp->t_inpcb) 823 so = tp->t_inpcb->inp_socket; 824 #ifdef INET6 825 else if (tp && tp->t_in6pcb) 826 so = tp->t_in6pcb->in6p_socket; 827 #endif 828 else 829 so = NULL; 830 831 if (tp != NULL && tp->t_inpcb != NULL) { 832 ro = &tp->t_inpcb->inp_route; 833 #ifdef DIAGNOSTIC 834 if (family != AF_INET) 835 panic("tcp_respond: address family mismatch"); 836 if (!in_hosteq(ip->ip_dst, tp->t_inpcb->inp_faddr)) { 837 panic("tcp_respond: ip_dst %x != inp_faddr %x", 838 ntohl(ip->ip_dst.s_addr), 839 ntohl(tp->t_inpcb->inp_faddr.s_addr)); 840 } 841 #endif 842 } 843 #ifdef INET6 844 else if (tp != NULL && tp->t_in6pcb != NULL) { 845 ro = (struct route *)&tp->t_in6pcb->in6p_route; 846 #ifdef DIAGNOSTIC 847 if (family == AF_INET) { 848 if (!IN6_IS_ADDR_V4MAPPED(&tp->t_in6pcb->in6p_faddr)) 849 panic("tcp_respond: not mapped addr"); 850 if (bcmp(&ip->ip_dst, 851 &tp->t_in6pcb->in6p_faddr.s6_addr32[3], 852 sizeof(ip->ip_dst)) != 0) { 853 panic("tcp_respond: ip_dst != in6p_faddr"); 854 } 855 } else if (family == AF_INET6) { 856 if (!IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, 857 &tp->t_in6pcb->in6p_faddr)) 858 panic("tcp_respond: ip6_dst != in6p_faddr"); 859 } else 860 panic("tcp_respond: address family mismatch"); 861 #endif 862 } 863 #endif 864 else 865 ro = NULL; 866 867 switch (family) { 868 #ifdef INET 869 case AF_INET: 870 error = ip_output(m, NULL, ro, 871 (tp && tp->t_mtudisc ? IP_MTUDISC : 0), 872 (struct ip_moptions *)0, so); 873 break; 874 #endif 875 #ifdef INET6 876 case AF_INET6: 877 error = ip6_output(m, NULL, (struct route_in6 *)ro, 0, 878 (struct ip6_moptions *)0, so, NULL); 879 break; 880 #endif 881 default: 882 error = EAFNOSUPPORT; 883 break; 884 } 885 886 return (error); 887 } 888 889 /* 890 * Template TCPCB. Rather than zeroing a new TCPCB and initializing 891 * a bunch of members individually, we maintain this template for the 892 * static and mostly-static components of the TCPCB, and copy it into 893 * the new TCPCB instead. 894 */ 895 static struct tcpcb tcpcb_template = { 896 /* 897 * If TCP_NTIMERS ever changes, we'll need to update this 898 * initializer. 899 */ 900 .t_timer = { 901 CALLOUT_INITIALIZER, 902 CALLOUT_INITIALIZER, 903 CALLOUT_INITIALIZER, 904 CALLOUT_INITIALIZER, 905 }, 906 .t_delack_ch = CALLOUT_INITIALIZER, 907 908 .t_srtt = TCPTV_SRTTBASE, 909 .t_rttmin = TCPTV_MIN, 910 911 .snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT, 912 .snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT, 913 .snd_numholes = 0, 914 915 .t_partialacks = -1, 916 }; 917 918 /* 919 * Updates the TCPCB template whenever a parameter that would affect 920 * the template is changed. 921 */ 922 void 923 tcp_tcpcb_template(void) 924 { 925 struct tcpcb *tp = &tcpcb_template; 926 int flags; 927 928 tp->t_peermss = tcp_mssdflt; 929 tp->t_ourmss = tcp_mssdflt; 930 tp->t_segsz = tcp_mssdflt; 931 932 flags = 0; 933 if (tcp_do_rfc1323 && tcp_do_win_scale) 934 flags |= TF_REQ_SCALE; 935 if (tcp_do_rfc1323 && tcp_do_timestamps) 936 flags |= TF_REQ_TSTMP; 937 tp->t_flags = flags; 938 939 /* 940 * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no 941 * rtt estimate. Set rttvar so that srtt + 2 * rttvar gives 942 * reasonable initial retransmit time. 943 */ 944 tp->t_rttvar = tcp_rttdflt * PR_SLOWHZ << (TCP_RTTVAR_SHIFT + 2 - 1); 945 TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp), 946 TCPTV_MIN, TCPTV_REXMTMAX); 947 } 948 949 /* 950 * Create a new TCP control block, making an 951 * empty reassembly queue and hooking it to the argument 952 * protocol control block. 953 */ 954 /* family selects inpcb, or in6pcb */ 955 struct tcpcb * 956 tcp_newtcpcb(int family, void *aux) 957 { 958 struct tcpcb *tp; 959 int i; 960 961 /* XXX Consider using a pool_cache for speed. */ 962 tp = pool_get(&tcpcb_pool, PR_NOWAIT); 963 if (tp == NULL) 964 return (NULL); 965 memcpy(tp, &tcpcb_template, sizeof(*tp)); 966 TAILQ_INIT(&tp->segq); 967 TAILQ_INIT(&tp->timeq); 968 tp->t_family = family; /* may be overridden later on */ 969 TAILQ_INIT(&tp->snd_holes); 970 LIST_INIT(&tp->t_sc); /* XXX can template this */ 971 972 /* Don't sweat this loop; hopefully the compiler will unroll it. */ 973 for (i = 0; i < TCPT_NTIMERS; i++) 974 TCP_TIMER_INIT(tp, i); 975 976 switch (family) { 977 case AF_INET: 978 { 979 struct inpcb *inp = (struct inpcb *)aux; 980 981 inp->inp_ip.ip_ttl = ip_defttl; 982 inp->inp_ppcb = (caddr_t)tp; 983 984 tp->t_inpcb = inp; 985 tp->t_mtudisc = ip_mtudisc; 986 break; 987 } 988 #ifdef INET6 989 case AF_INET6: 990 { 991 struct in6pcb *in6p = (struct in6pcb *)aux; 992 993 in6p->in6p_ip6.ip6_hlim = in6_selecthlim(in6p, 994 in6p->in6p_route.ro_rt ? in6p->in6p_route.ro_rt->rt_ifp 995 : NULL); 996 in6p->in6p_ppcb = (caddr_t)tp; 997 998 tp->t_in6pcb = in6p; 999 /* for IPv6, always try to run path MTU discovery */ 1000 tp->t_mtudisc = 1; 1001 break; 1002 } 1003 #endif /* INET6 */ 1004 default: 1005 pool_put(&tcpcb_pool, tp); 1006 return (NULL); 1007 } 1008 1009 /* 1010 * Initialize our timebase. When we send timestamps, we take 1011 * the delta from tcp_now -- this means each connection always 1012 * gets a timebase of 0, which makes it, among other things, 1013 * more difficult to determine how long a system has been up, 1014 * and thus how many TCP sequence increments have occurred. 1015 */ 1016 tp->ts_timebase = tcp_now; 1017 1018 return (tp); 1019 } 1020 1021 /* 1022 * Drop a TCP connection, reporting 1023 * the specified error. If connection is synchronized, 1024 * then send a RST to peer. 1025 */ 1026 struct tcpcb * 1027 tcp_drop(struct tcpcb *tp, int errno) 1028 { 1029 struct socket *so = NULL; 1030 1031 #ifdef DIAGNOSTIC 1032 if (tp->t_inpcb && tp->t_in6pcb) 1033 panic("tcp_drop: both t_inpcb and t_in6pcb are set"); 1034 #endif 1035 #ifdef INET 1036 if (tp->t_inpcb) 1037 so = tp->t_inpcb->inp_socket; 1038 #endif 1039 #ifdef INET6 1040 if (tp->t_in6pcb) 1041 so = tp->t_in6pcb->in6p_socket; 1042 #endif 1043 if (!so) 1044 return NULL; 1045 1046 if (TCPS_HAVERCVDSYN(tp->t_state)) { 1047 tp->t_state = TCPS_CLOSED; 1048 (void) tcp_output(tp); 1049 tcpstat.tcps_drops++; 1050 } else 1051 tcpstat.tcps_conndrops++; 1052 if (errno == ETIMEDOUT && tp->t_softerror) 1053 errno = tp->t_softerror; 1054 so->so_error = errno; 1055 return (tcp_close(tp)); 1056 } 1057 1058 /* 1059 * Return whether this tcpcb is marked as dead, indicating 1060 * to the calling timer function that no further action should 1061 * be taken, as we are about to release this tcpcb. The release 1062 * of the storage will be done if this is the last timer running. 1063 * 1064 * This should be called from the callout handler function after 1065 * callout_ack() is done, so that the number of invoking timer 1066 * functions is 0. 1067 */ 1068 int 1069 tcp_isdead(struct tcpcb *tp) 1070 { 1071 int dead = (tp->t_flags & TF_DEAD); 1072 1073 if (__predict_false(dead)) { 1074 if (tcp_timers_invoking(tp) > 0) 1075 /* not quite there yet -- count separately? */ 1076 return dead; 1077 tcpstat.tcps_delayed_free++; 1078 pool_put(&tcpcb_pool, tp); 1079 } 1080 return dead; 1081 } 1082 1083 /* 1084 * Close a TCP control block: 1085 * discard all space held by the tcp 1086 * discard internet protocol block 1087 * wake up any sleepers 1088 */ 1089 struct tcpcb * 1090 tcp_close(struct tcpcb *tp) 1091 { 1092 struct inpcb *inp; 1093 #ifdef INET6 1094 struct in6pcb *in6p; 1095 #endif 1096 struct socket *so; 1097 #ifdef RTV_RTT 1098 struct rtentry *rt; 1099 #endif 1100 struct route *ro; 1101 1102 inp = tp->t_inpcb; 1103 #ifdef INET6 1104 in6p = tp->t_in6pcb; 1105 #endif 1106 so = NULL; 1107 ro = NULL; 1108 if (inp) { 1109 so = inp->inp_socket; 1110 ro = &inp->inp_route; 1111 } 1112 #ifdef INET6 1113 else if (in6p) { 1114 so = in6p->in6p_socket; 1115 ro = (struct route *)&in6p->in6p_route; 1116 } 1117 #endif 1118 1119 #ifdef RTV_RTT 1120 /* 1121 * If we sent enough data to get some meaningful characteristics, 1122 * save them in the routing entry. 'Enough' is arbitrarily 1123 * defined as the sendpipesize (default 4K) * 16. This would 1124 * give us 16 rtt samples assuming we only get one sample per 1125 * window (the usual case on a long haul net). 16 samples is 1126 * enough for the srtt filter to converge to within 5% of the correct 1127 * value; fewer samples and we could save a very bogus rtt. 1128 * 1129 * Don't update the default route's characteristics and don't 1130 * update anything that the user "locked". 1131 */ 1132 if (SEQ_LT(tp->iss + so->so_snd.sb_hiwat * 16, tp->snd_max) && 1133 ro && (rt = ro->ro_rt) && 1134 !in_nullhost(satosin(rt_key(rt))->sin_addr)) { 1135 u_long i = 0; 1136 1137 if ((rt->rt_rmx.rmx_locks & RTV_RTT) == 0) { 1138 i = tp->t_srtt * 1139 ((RTM_RTTUNIT / PR_SLOWHZ) >> (TCP_RTT_SHIFT + 2)); 1140 if (rt->rt_rmx.rmx_rtt && i) 1141 /* 1142 * filter this update to half the old & half 1143 * the new values, converting scale. 1144 * See route.h and tcp_var.h for a 1145 * description of the scaling constants. 1146 */ 1147 rt->rt_rmx.rmx_rtt = 1148 (rt->rt_rmx.rmx_rtt + i) / 2; 1149 else 1150 rt->rt_rmx.rmx_rtt = i; 1151 } 1152 if ((rt->rt_rmx.rmx_locks & RTV_RTTVAR) == 0) { 1153 i = tp->t_rttvar * 1154 ((RTM_RTTUNIT / PR_SLOWHZ) >> (TCP_RTTVAR_SHIFT + 2)); 1155 if (rt->rt_rmx.rmx_rttvar && i) 1156 rt->rt_rmx.rmx_rttvar = 1157 (rt->rt_rmx.rmx_rttvar + i) / 2; 1158 else 1159 rt->rt_rmx.rmx_rttvar = i; 1160 } 1161 /* 1162 * update the pipelimit (ssthresh) if it has been updated 1163 * already or if a pipesize was specified & the threshhold 1164 * got below half the pipesize. I.e., wait for bad news 1165 * before we start updating, then update on both good 1166 * and bad news. 1167 */ 1168 if (((rt->rt_rmx.rmx_locks & RTV_SSTHRESH) == 0 && 1169 (i = tp->snd_ssthresh) && rt->rt_rmx.rmx_ssthresh) || 1170 i < (rt->rt_rmx.rmx_sendpipe / 2)) { 1171 /* 1172 * convert the limit from user data bytes to 1173 * packets then to packet data bytes. 1174 */ 1175 i = (i + tp->t_segsz / 2) / tp->t_segsz; 1176 if (i < 2) 1177 i = 2; 1178 i *= (u_long)(tp->t_segsz + sizeof (struct tcpiphdr)); 1179 if (rt->rt_rmx.rmx_ssthresh) 1180 rt->rt_rmx.rmx_ssthresh = 1181 (rt->rt_rmx.rmx_ssthresh + i) / 2; 1182 else 1183 rt->rt_rmx.rmx_ssthresh = i; 1184 } 1185 } 1186 #endif /* RTV_RTT */ 1187 /* free the reassembly queue, if any */ 1188 TCP_REASS_LOCK(tp); 1189 (void) tcp_freeq(tp); 1190 TCP_REASS_UNLOCK(tp); 1191 1192 /* free the SACK holes list. */ 1193 tcp_free_sackholes(tp); 1194 1195 tcp_canceltimers(tp); 1196 TCP_CLEAR_DELACK(tp); 1197 syn_cache_cleanup(tp); 1198 1199 if (tp->t_template) { 1200 m_free(tp->t_template); 1201 tp->t_template = NULL; 1202 } 1203 if (tcp_timers_invoking(tp)) 1204 tp->t_flags |= TF_DEAD; 1205 else 1206 pool_put(&tcpcb_pool, tp); 1207 1208 if (inp) { 1209 inp->inp_ppcb = 0; 1210 soisdisconnected(so); 1211 in_pcbdetach(inp); 1212 } 1213 #ifdef INET6 1214 else if (in6p) { 1215 in6p->in6p_ppcb = 0; 1216 soisdisconnected(so); 1217 in6_pcbdetach(in6p); 1218 } 1219 #endif 1220 tcpstat.tcps_closed++; 1221 return ((struct tcpcb *)0); 1222 } 1223 1224 int 1225 tcp_freeq(tp) 1226 struct tcpcb *tp; 1227 { 1228 struct ipqent *qe; 1229 int rv = 0; 1230 #ifdef TCPREASS_DEBUG 1231 int i = 0; 1232 #endif 1233 1234 TCP_REASS_LOCK_CHECK(tp); 1235 1236 while ((qe = TAILQ_FIRST(&tp->segq)) != NULL) { 1237 #ifdef TCPREASS_DEBUG 1238 printf("tcp_freeq[%p,%d]: %u:%u(%u) 0x%02x\n", 1239 tp, i++, qe->ipqe_seq, qe->ipqe_seq + qe->ipqe_len, 1240 qe->ipqe_len, qe->ipqe_flags & (TH_SYN|TH_FIN|TH_RST)); 1241 #endif 1242 TAILQ_REMOVE(&tp->segq, qe, ipqe_q); 1243 TAILQ_REMOVE(&tp->timeq, qe, ipqe_timeq); 1244 m_freem(qe->ipqe_m); 1245 tcpipqent_free(qe); 1246 rv = 1; 1247 } 1248 tp->t_segqlen = 0; 1249 KASSERT(TAILQ_EMPTY(&tp->timeq)); 1250 return (rv); 1251 } 1252 1253 /* 1254 * Protocol drain routine. Called when memory is in short supply. 1255 */ 1256 void 1257 tcp_drain(void) 1258 { 1259 struct inpcb_hdr *inph; 1260 struct tcpcb *tp; 1261 1262 /* 1263 * Free the sequence queue of all TCP connections. 1264 */ 1265 CIRCLEQ_FOREACH(inph, &tcbtable.inpt_queue, inph_queue) { 1266 switch (inph->inph_af) { 1267 case AF_INET: 1268 tp = intotcpcb((struct inpcb *)inph); 1269 break; 1270 #ifdef INET6 1271 case AF_INET6: 1272 tp = in6totcpcb((struct in6pcb *)inph); 1273 break; 1274 #endif 1275 default: 1276 tp = NULL; 1277 break; 1278 } 1279 if (tp != NULL) { 1280 /* 1281 * We may be called from a device's interrupt 1282 * context. If the tcpcb is already busy, 1283 * just bail out now. 1284 */ 1285 if (tcp_reass_lock_try(tp) == 0) 1286 continue; 1287 if (tcp_freeq(tp)) 1288 tcpstat.tcps_connsdrained++; 1289 TCP_REASS_UNLOCK(tp); 1290 } 1291 } 1292 } 1293 1294 /* 1295 * Notify a tcp user of an asynchronous error; 1296 * store error as soft error, but wake up user 1297 * (for now, won't do anything until can select for soft error). 1298 */ 1299 void 1300 tcp_notify(struct inpcb *inp, int error) 1301 { 1302 struct tcpcb *tp = (struct tcpcb *)inp->inp_ppcb; 1303 struct socket *so = inp->inp_socket; 1304 1305 /* 1306 * Ignore some errors if we are hooked up. 1307 * If connection hasn't completed, has retransmitted several times, 1308 * and receives a second error, give up now. This is better 1309 * than waiting a long time to establish a connection that 1310 * can never complete. 1311 */ 1312 if (tp->t_state == TCPS_ESTABLISHED && 1313 (error == EHOSTUNREACH || error == ENETUNREACH || 1314 error == EHOSTDOWN)) { 1315 return; 1316 } else if (TCPS_HAVEESTABLISHED(tp->t_state) == 0 && 1317 tp->t_rxtshift > 3 && tp->t_softerror) 1318 so->so_error = error; 1319 else 1320 tp->t_softerror = error; 1321 wakeup((caddr_t) &so->so_timeo); 1322 sorwakeup(so); 1323 sowwakeup(so); 1324 } 1325 1326 #ifdef INET6 1327 void 1328 tcp6_notify(struct in6pcb *in6p, int error) 1329 { 1330 struct tcpcb *tp = (struct tcpcb *)in6p->in6p_ppcb; 1331 struct socket *so = in6p->in6p_socket; 1332 1333 /* 1334 * Ignore some errors if we are hooked up. 1335 * If connection hasn't completed, has retransmitted several times, 1336 * and receives a second error, give up now. This is better 1337 * than waiting a long time to establish a connection that 1338 * can never complete. 1339 */ 1340 if (tp->t_state == TCPS_ESTABLISHED && 1341 (error == EHOSTUNREACH || error == ENETUNREACH || 1342 error == EHOSTDOWN)) { 1343 return; 1344 } else if (TCPS_HAVEESTABLISHED(tp->t_state) == 0 && 1345 tp->t_rxtshift > 3 && tp->t_softerror) 1346 so->so_error = error; 1347 else 1348 tp->t_softerror = error; 1349 wakeup((caddr_t) &so->so_timeo); 1350 sorwakeup(so); 1351 sowwakeup(so); 1352 } 1353 #endif 1354 1355 #ifdef INET6 1356 void 1357 tcp6_ctlinput(int cmd, struct sockaddr *sa, void *d) 1358 { 1359 struct tcphdr th; 1360 void (*notify)(struct in6pcb *, int) = tcp6_notify; 1361 int nmatch; 1362 struct ip6_hdr *ip6; 1363 const struct sockaddr_in6 *sa6_src = NULL; 1364 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)sa; 1365 struct mbuf *m; 1366 int off; 1367 1368 if (sa->sa_family != AF_INET6 || 1369 sa->sa_len != sizeof(struct sockaddr_in6)) 1370 return; 1371 if ((unsigned)cmd >= PRC_NCMDS) 1372 return; 1373 else if (cmd == PRC_QUENCH) { 1374 /* XXX there's no PRC_QUENCH in IPv6 */ 1375 notify = tcp6_quench; 1376 } else if (PRC_IS_REDIRECT(cmd)) 1377 notify = in6_rtchange, d = NULL; 1378 else if (cmd == PRC_MSGSIZE) 1379 ; /* special code is present, see below */ 1380 else if (cmd == PRC_HOSTDEAD) 1381 d = NULL; 1382 else if (inet6ctlerrmap[cmd] == 0) 1383 return; 1384 1385 /* if the parameter is from icmp6, decode it. */ 1386 if (d != NULL) { 1387 struct ip6ctlparam *ip6cp = (struct ip6ctlparam *)d; 1388 m = ip6cp->ip6c_m; 1389 ip6 = ip6cp->ip6c_ip6; 1390 off = ip6cp->ip6c_off; 1391 sa6_src = ip6cp->ip6c_src; 1392 } else { 1393 m = NULL; 1394 ip6 = NULL; 1395 sa6_src = &sa6_any; 1396 off = 0; 1397 } 1398 1399 if (ip6) { 1400 /* 1401 * XXX: We assume that when ip6 is non NULL, 1402 * M and OFF are valid. 1403 */ 1404 1405 /* check if we can safely examine src and dst ports */ 1406 if (m->m_pkthdr.len < off + sizeof(th)) { 1407 if (cmd == PRC_MSGSIZE) 1408 icmp6_mtudisc_update((struct ip6ctlparam *)d, 0); 1409 return; 1410 } 1411 1412 bzero(&th, sizeof(th)); 1413 m_copydata(m, off, sizeof(th), (caddr_t)&th); 1414 1415 if (cmd == PRC_MSGSIZE) { 1416 int valid = 0; 1417 1418 /* 1419 * Check to see if we have a valid TCP connection 1420 * corresponding to the address in the ICMPv6 message 1421 * payload. 1422 */ 1423 if (in6_pcblookup_connect(&tcbtable, &sa6->sin6_addr, 1424 th.th_dport, (const struct in6_addr *)&sa6_src->sin6_addr, 1425 th.th_sport, 0)) 1426 valid++; 1427 1428 /* 1429 * Depending on the value of "valid" and routing table 1430 * size (mtudisc_{hi,lo}wat), we will: 1431 * - recalcurate the new MTU and create the 1432 * corresponding routing entry, or 1433 * - ignore the MTU change notification. 1434 */ 1435 icmp6_mtudisc_update((struct ip6ctlparam *)d, valid); 1436 1437 /* 1438 * no need to call in6_pcbnotify, it should have been 1439 * called via callback if necessary 1440 */ 1441 return; 1442 } 1443 1444 nmatch = in6_pcbnotify(&tcbtable, sa, th.th_dport, 1445 (const struct sockaddr *)sa6_src, th.th_sport, cmd, NULL, notify); 1446 if (nmatch == 0 && syn_cache_count && 1447 (inet6ctlerrmap[cmd] == EHOSTUNREACH || 1448 inet6ctlerrmap[cmd] == ENETUNREACH || 1449 inet6ctlerrmap[cmd] == EHOSTDOWN)) 1450 syn_cache_unreach((const struct sockaddr *)sa6_src, 1451 sa, &th); 1452 } else { 1453 (void) in6_pcbnotify(&tcbtable, sa, 0, 1454 (const struct sockaddr *)sa6_src, 0, cmd, NULL, notify); 1455 } 1456 } 1457 #endif 1458 1459 #ifdef INET 1460 /* assumes that ip header and tcp header are contiguous on mbuf */ 1461 void * 1462 tcp_ctlinput(int cmd, struct sockaddr *sa, void *v) 1463 { 1464 struct ip *ip = v; 1465 struct tcphdr *th; 1466 struct icmp *icp; 1467 extern const int inetctlerrmap[]; 1468 void (*notify)(struct inpcb *, int) = tcp_notify; 1469 int errno; 1470 int nmatch; 1471 #ifdef INET6 1472 struct in6_addr src6, dst6; 1473 #endif 1474 1475 if (sa->sa_family != AF_INET || 1476 sa->sa_len != sizeof(struct sockaddr_in)) 1477 return NULL; 1478 if ((unsigned)cmd >= PRC_NCMDS) 1479 return NULL; 1480 errno = inetctlerrmap[cmd]; 1481 if (cmd == PRC_QUENCH) 1482 notify = tcp_quench; 1483 else if (PRC_IS_REDIRECT(cmd)) 1484 notify = in_rtchange, ip = 0; 1485 else if (cmd == PRC_MSGSIZE && ip && ip->ip_v == 4) { 1486 /* 1487 * Check to see if we have a valid TCP connection 1488 * corresponding to the address in the ICMP message 1489 * payload. 1490 * 1491 * Boundary check is made in icmp_input(), with ICMP_ADVLENMIN. 1492 */ 1493 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 1494 #ifdef INET6 1495 memset(&src6, 0, sizeof(src6)); 1496 memset(&dst6, 0, sizeof(dst6)); 1497 src6.s6_addr16[5] = dst6.s6_addr16[5] = 0xffff; 1498 memcpy(&src6.s6_addr32[3], &ip->ip_src, sizeof(struct in_addr)); 1499 memcpy(&dst6.s6_addr32[3], &ip->ip_dst, sizeof(struct in_addr)); 1500 #endif 1501 if (in_pcblookup_connect(&tcbtable, ip->ip_dst, th->th_dport, 1502 ip->ip_src, th->th_sport) != NULL) 1503 ; 1504 #ifdef INET6 1505 else if (in6_pcblookup_connect(&tcbtable, &dst6, 1506 th->th_dport, &src6, th->th_sport, 0) != NULL) 1507 ; 1508 #endif 1509 else 1510 return NULL; 1511 1512 /* 1513 * Now that we've validated that we are actually communicating 1514 * with the host indicated in the ICMP message, locate the 1515 * ICMP header, recalculate the new MTU, and create the 1516 * corresponding routing entry. 1517 */ 1518 icp = (struct icmp *)((caddr_t)ip - 1519 offsetof(struct icmp, icmp_ip)); 1520 icmp_mtudisc(icp, ip->ip_dst); 1521 1522 return NULL; 1523 } else if (cmd == PRC_HOSTDEAD) 1524 ip = 0; 1525 else if (errno == 0) 1526 return NULL; 1527 if (ip && ip->ip_v == 4 && sa->sa_family == AF_INET) { 1528 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 1529 nmatch = in_pcbnotify(&tcbtable, satosin(sa)->sin_addr, 1530 th->th_dport, ip->ip_src, th->th_sport, errno, notify); 1531 if (nmatch == 0 && syn_cache_count && 1532 (inetctlerrmap[cmd] == EHOSTUNREACH || 1533 inetctlerrmap[cmd] == ENETUNREACH || 1534 inetctlerrmap[cmd] == EHOSTDOWN)) { 1535 struct sockaddr_in sin; 1536 bzero(&sin, sizeof(sin)); 1537 sin.sin_len = sizeof(sin); 1538 sin.sin_family = AF_INET; 1539 sin.sin_port = th->th_sport; 1540 sin.sin_addr = ip->ip_src; 1541 syn_cache_unreach((struct sockaddr *)&sin, sa, th); 1542 } 1543 1544 /* XXX mapped address case */ 1545 } else 1546 in_pcbnotifyall(&tcbtable, satosin(sa)->sin_addr, errno, 1547 notify); 1548 return NULL; 1549 } 1550 1551 /* 1552 * When a source quench is received, we are being notified of congestion. 1553 * Close the congestion window down to the Loss Window (one segment). 1554 * We will gradually open it again as we proceed. 1555 */ 1556 void 1557 tcp_quench(struct inpcb *inp, int errno) 1558 { 1559 struct tcpcb *tp = intotcpcb(inp); 1560 1561 if (tp) 1562 tp->snd_cwnd = tp->t_segsz; 1563 } 1564 #endif 1565 1566 #ifdef INET6 1567 void 1568 tcp6_quench(struct in6pcb *in6p, int errno) 1569 { 1570 struct tcpcb *tp = in6totcpcb(in6p); 1571 1572 if (tp) 1573 tp->snd_cwnd = tp->t_segsz; 1574 } 1575 #endif 1576 1577 #ifdef INET 1578 /* 1579 * Path MTU Discovery handlers. 1580 */ 1581 void 1582 tcp_mtudisc_callback(struct in_addr faddr) 1583 { 1584 #ifdef INET6 1585 struct in6_addr in6; 1586 #endif 1587 1588 in_pcbnotifyall(&tcbtable, faddr, EMSGSIZE, tcp_mtudisc); 1589 #ifdef INET6 1590 memset(&in6, 0, sizeof(in6)); 1591 in6.s6_addr16[5] = 0xffff; 1592 memcpy(&in6.s6_addr32[3], &faddr, sizeof(struct in_addr)); 1593 tcp6_mtudisc_callback(&in6); 1594 #endif 1595 } 1596 1597 /* 1598 * On receipt of path MTU corrections, flush old route and replace it 1599 * with the new one. Retransmit all unacknowledged packets, to ensure 1600 * that all packets will be received. 1601 */ 1602 void 1603 tcp_mtudisc(struct inpcb *inp, int errno) 1604 { 1605 struct tcpcb *tp = intotcpcb(inp); 1606 struct rtentry *rt = in_pcbrtentry(inp); 1607 1608 if (tp != 0) { 1609 if (rt != 0) { 1610 /* 1611 * If this was not a host route, remove and realloc. 1612 */ 1613 if ((rt->rt_flags & RTF_HOST) == 0) { 1614 in_rtchange(inp, errno); 1615 if ((rt = in_pcbrtentry(inp)) == 0) 1616 return; 1617 } 1618 1619 /* 1620 * Slow start out of the error condition. We 1621 * use the MTU because we know it's smaller 1622 * than the previously transmitted segment. 1623 * 1624 * Note: This is more conservative than the 1625 * suggestion in draft-floyd-incr-init-win-03. 1626 */ 1627 if (rt->rt_rmx.rmx_mtu != 0) 1628 tp->snd_cwnd = 1629 TCP_INITIAL_WINDOW(tcp_init_win, 1630 rt->rt_rmx.rmx_mtu); 1631 } 1632 1633 /* 1634 * Resend unacknowledged packets. 1635 */ 1636 tp->snd_nxt = tp->snd_una; 1637 tcp_output(tp); 1638 } 1639 } 1640 #endif 1641 1642 #ifdef INET6 1643 /* 1644 * Path MTU Discovery handlers. 1645 */ 1646 void 1647 tcp6_mtudisc_callback(struct in6_addr *faddr) 1648 { 1649 struct sockaddr_in6 sin6; 1650 1651 bzero(&sin6, sizeof(sin6)); 1652 sin6.sin6_family = AF_INET6; 1653 sin6.sin6_len = sizeof(struct sockaddr_in6); 1654 sin6.sin6_addr = *faddr; 1655 (void) in6_pcbnotify(&tcbtable, (struct sockaddr *)&sin6, 0, 1656 (const struct sockaddr *)&sa6_any, 0, PRC_MSGSIZE, NULL, tcp6_mtudisc); 1657 } 1658 1659 void 1660 tcp6_mtudisc(struct in6pcb *in6p, int errno) 1661 { 1662 struct tcpcb *tp = in6totcpcb(in6p); 1663 struct rtentry *rt = in6_pcbrtentry(in6p); 1664 1665 if (tp != 0) { 1666 if (rt != 0) { 1667 /* 1668 * If this was not a host route, remove and realloc. 1669 */ 1670 if ((rt->rt_flags & RTF_HOST) == 0) { 1671 in6_rtchange(in6p, errno); 1672 if ((rt = in6_pcbrtentry(in6p)) == 0) 1673 return; 1674 } 1675 1676 /* 1677 * Slow start out of the error condition. We 1678 * use the MTU because we know it's smaller 1679 * than the previously transmitted segment. 1680 * 1681 * Note: This is more conservative than the 1682 * suggestion in draft-floyd-incr-init-win-03. 1683 */ 1684 if (rt->rt_rmx.rmx_mtu != 0) 1685 tp->snd_cwnd = 1686 TCP_INITIAL_WINDOW(tcp_init_win, 1687 rt->rt_rmx.rmx_mtu); 1688 } 1689 1690 /* 1691 * Resend unacknowledged packets. 1692 */ 1693 tp->snd_nxt = tp->snd_una; 1694 tcp_output(tp); 1695 } 1696 } 1697 #endif /* INET6 */ 1698 1699 /* 1700 * Compute the MSS to advertise to the peer. Called only during 1701 * the 3-way handshake. If we are the server (peer initiated 1702 * connection), we are called with a pointer to the interface 1703 * on which the SYN packet arrived. If we are the client (we 1704 * initiated connection), we are called with a pointer to the 1705 * interface out which this connection should go. 1706 * 1707 * NOTE: Do not subtract IP option/extension header size nor IPsec 1708 * header size from MSS advertisement. MSS option must hold the maximum 1709 * segment size we can accept, so it must always be: 1710 * max(if mtu) - ip header - tcp header 1711 */ 1712 u_long 1713 tcp_mss_to_advertise(const struct ifnet *ifp, int af) 1714 { 1715 extern u_long in_maxmtu; 1716 u_long mss = 0; 1717 u_long hdrsiz; 1718 1719 /* 1720 * In order to avoid defeating path MTU discovery on the peer, 1721 * we advertise the max MTU of all attached networks as our MSS, 1722 * per RFC 1191, section 3.1. 1723 * 1724 * We provide the option to advertise just the MTU of 1725 * the interface on which we hope this connection will 1726 * be receiving. If we are responding to a SYN, we 1727 * will have a pretty good idea about this, but when 1728 * initiating a connection there is a bit more doubt. 1729 * 1730 * We also need to ensure that loopback has a large enough 1731 * MSS, as the loopback MTU is never included in in_maxmtu. 1732 */ 1733 1734 if (ifp != NULL) 1735 switch (af) { 1736 case AF_INET: 1737 mss = ifp->if_mtu; 1738 break; 1739 #ifdef INET6 1740 case AF_INET6: 1741 mss = IN6_LINKMTU(ifp); 1742 break; 1743 #endif 1744 } 1745 1746 if (tcp_mss_ifmtu == 0) 1747 switch (af) { 1748 case AF_INET: 1749 mss = max(in_maxmtu, mss); 1750 break; 1751 #ifdef INET6 1752 case AF_INET6: 1753 mss = max(in6_maxmtu, mss); 1754 break; 1755 #endif 1756 } 1757 1758 switch (af) { 1759 case AF_INET: 1760 hdrsiz = sizeof(struct ip); 1761 break; 1762 #ifdef INET6 1763 case AF_INET6: 1764 hdrsiz = sizeof(struct ip6_hdr); 1765 break; 1766 #endif 1767 default: 1768 hdrsiz = 0; 1769 break; 1770 } 1771 hdrsiz += sizeof(struct tcphdr); 1772 if (mss > hdrsiz) 1773 mss -= hdrsiz; 1774 1775 mss = max(tcp_mssdflt, mss); 1776 return (mss); 1777 } 1778 1779 /* 1780 * Set connection variables based on the peer's advertised MSS. 1781 * We are passed the TCPCB for the actual connection. If we 1782 * are the server, we are called by the compressed state engine 1783 * when the 3-way handshake is complete. If we are the client, 1784 * we are called when we receive the SYN,ACK from the server. 1785 * 1786 * NOTE: Our advertised MSS value must be initialized in the TCPCB 1787 * before this routine is called! 1788 */ 1789 void 1790 tcp_mss_from_peer(struct tcpcb *tp, int offer) 1791 { 1792 struct socket *so; 1793 #if defined(RTV_SPIPE) || defined(RTV_SSTHRESH) 1794 struct rtentry *rt; 1795 #endif 1796 u_long bufsize; 1797 int mss; 1798 1799 #ifdef DIAGNOSTIC 1800 if (tp->t_inpcb && tp->t_in6pcb) 1801 panic("tcp_mss_from_peer: both t_inpcb and t_in6pcb are set"); 1802 #endif 1803 so = NULL; 1804 rt = NULL; 1805 #ifdef INET 1806 if (tp->t_inpcb) { 1807 so = tp->t_inpcb->inp_socket; 1808 #if defined(RTV_SPIPE) || defined(RTV_SSTHRESH) 1809 rt = in_pcbrtentry(tp->t_inpcb); 1810 #endif 1811 } 1812 #endif 1813 #ifdef INET6 1814 if (tp->t_in6pcb) { 1815 so = tp->t_in6pcb->in6p_socket; 1816 #if defined(RTV_SPIPE) || defined(RTV_SSTHRESH) 1817 rt = in6_pcbrtentry(tp->t_in6pcb); 1818 #endif 1819 } 1820 #endif 1821 1822 /* 1823 * As per RFC1122, use the default MSS value, unless they 1824 * sent us an offer. Do not accept offers less than 256 bytes. 1825 */ 1826 mss = tcp_mssdflt; 1827 if (offer) 1828 mss = offer; 1829 mss = max(mss, 256); /* sanity */ 1830 tp->t_peermss = mss; 1831 mss -= tcp_optlen(tp); 1832 #ifdef INET 1833 if (tp->t_inpcb) 1834 mss -= ip_optlen(tp->t_inpcb); 1835 #endif 1836 #ifdef INET6 1837 if (tp->t_in6pcb) 1838 mss -= ip6_optlen(tp->t_in6pcb); 1839 #endif 1840 1841 /* 1842 * If there's a pipesize, change the socket buffer to that size. 1843 * Make the socket buffer an integral number of MSS units. If 1844 * the MSS is larger than the socket buffer, artificially decrease 1845 * the MSS. 1846 */ 1847 #ifdef RTV_SPIPE 1848 if (rt != NULL && rt->rt_rmx.rmx_sendpipe != 0) 1849 bufsize = rt->rt_rmx.rmx_sendpipe; 1850 else 1851 #endif 1852 bufsize = so->so_snd.sb_hiwat; 1853 if (bufsize < mss) 1854 mss = bufsize; 1855 else { 1856 bufsize = roundup(bufsize, mss); 1857 if (bufsize > sb_max) 1858 bufsize = sb_max; 1859 (void) sbreserve(&so->so_snd, bufsize, so); 1860 } 1861 tp->t_segsz = mss; 1862 1863 #ifdef RTV_SSTHRESH 1864 if (rt != NULL && rt->rt_rmx.rmx_ssthresh) { 1865 /* 1866 * There's some sort of gateway or interface buffer 1867 * limit on the path. Use this to set the slow 1868 * start threshold, but set the threshold to no less 1869 * than 2 * MSS. 1870 */ 1871 tp->snd_ssthresh = max(2 * mss, rt->rt_rmx.rmx_ssthresh); 1872 } 1873 #endif 1874 } 1875 1876 /* 1877 * Processing necessary when a TCP connection is established. 1878 */ 1879 void 1880 tcp_established(struct tcpcb *tp) 1881 { 1882 struct socket *so; 1883 #ifdef RTV_RPIPE 1884 struct rtentry *rt; 1885 #endif 1886 u_long bufsize; 1887 1888 #ifdef DIAGNOSTIC 1889 if (tp->t_inpcb && tp->t_in6pcb) 1890 panic("tcp_established: both t_inpcb and t_in6pcb are set"); 1891 #endif 1892 so = NULL; 1893 rt = NULL; 1894 #ifdef INET 1895 if (tp->t_inpcb) { 1896 so = tp->t_inpcb->inp_socket; 1897 #if defined(RTV_RPIPE) 1898 rt = in_pcbrtentry(tp->t_inpcb); 1899 #endif 1900 } 1901 #endif 1902 #ifdef INET6 1903 if (tp->t_in6pcb) { 1904 so = tp->t_in6pcb->in6p_socket; 1905 #if defined(RTV_RPIPE) 1906 rt = in6_pcbrtentry(tp->t_in6pcb); 1907 #endif 1908 } 1909 #endif 1910 1911 tp->t_state = TCPS_ESTABLISHED; 1912 TCP_TIMER_ARM(tp, TCPT_KEEP, tcp_keepidle); 1913 1914 #ifdef RTV_RPIPE 1915 if (rt != NULL && rt->rt_rmx.rmx_recvpipe != 0) 1916 bufsize = rt->rt_rmx.rmx_recvpipe; 1917 else 1918 #endif 1919 bufsize = so->so_rcv.sb_hiwat; 1920 if (bufsize > tp->t_ourmss) { 1921 bufsize = roundup(bufsize, tp->t_ourmss); 1922 if (bufsize > sb_max) 1923 bufsize = sb_max; 1924 (void) sbreserve(&so->so_rcv, bufsize, so); 1925 } 1926 } 1927 1928 /* 1929 * Check if there's an initial rtt or rttvar. Convert from the 1930 * route-table units to scaled multiples of the slow timeout timer. 1931 * Called only during the 3-way handshake. 1932 */ 1933 void 1934 tcp_rmx_rtt(struct tcpcb *tp) 1935 { 1936 #ifdef RTV_RTT 1937 struct rtentry *rt = NULL; 1938 int rtt; 1939 1940 #ifdef DIAGNOSTIC 1941 if (tp->t_inpcb && tp->t_in6pcb) 1942 panic("tcp_rmx_rtt: both t_inpcb and t_in6pcb are set"); 1943 #endif 1944 #ifdef INET 1945 if (tp->t_inpcb) 1946 rt = in_pcbrtentry(tp->t_inpcb); 1947 #endif 1948 #ifdef INET6 1949 if (tp->t_in6pcb) 1950 rt = in6_pcbrtentry(tp->t_in6pcb); 1951 #endif 1952 if (rt == NULL) 1953 return; 1954 1955 if (tp->t_srtt == 0 && (rtt = rt->rt_rmx.rmx_rtt)) { 1956 /* 1957 * XXX The lock bit for MTU indicates that the value 1958 * is also a minimum value; this is subject to time. 1959 */ 1960 if (rt->rt_rmx.rmx_locks & RTV_RTT) 1961 TCPT_RANGESET(tp->t_rttmin, 1962 rtt / (RTM_RTTUNIT / PR_SLOWHZ), 1963 TCPTV_MIN, TCPTV_REXMTMAX); 1964 tp->t_srtt = rtt / 1965 ((RTM_RTTUNIT / PR_SLOWHZ) >> (TCP_RTT_SHIFT + 2)); 1966 if (rt->rt_rmx.rmx_rttvar) { 1967 tp->t_rttvar = rt->rt_rmx.rmx_rttvar / 1968 ((RTM_RTTUNIT / PR_SLOWHZ) >> 1969 (TCP_RTTVAR_SHIFT + 2)); 1970 } else { 1971 /* Default variation is +- 1 rtt */ 1972 tp->t_rttvar = 1973 tp->t_srtt >> (TCP_RTT_SHIFT - TCP_RTTVAR_SHIFT); 1974 } 1975 TCPT_RANGESET(tp->t_rxtcur, 1976 ((tp->t_srtt >> 2) + tp->t_rttvar) >> (1 + 2), 1977 tp->t_rttmin, TCPTV_REXMTMAX); 1978 } 1979 #endif 1980 } 1981 1982 tcp_seq tcp_iss_seq = 0; /* tcp initial seq # */ 1983 #if NRND > 0 1984 u_int8_t tcp_iss_secret[16]; /* 128 bits; should be plenty */ 1985 #endif 1986 1987 /* 1988 * Get a new sequence value given a tcp control block 1989 */ 1990 tcp_seq 1991 tcp_new_iss(struct tcpcb *tp, tcp_seq addin) 1992 { 1993 1994 #ifdef INET 1995 if (tp->t_inpcb != NULL) { 1996 return (tcp_new_iss1(&tp->t_inpcb->inp_laddr, 1997 &tp->t_inpcb->inp_faddr, tp->t_inpcb->inp_lport, 1998 tp->t_inpcb->inp_fport, sizeof(tp->t_inpcb->inp_laddr), 1999 addin)); 2000 } 2001 #endif 2002 #ifdef INET6 2003 if (tp->t_in6pcb != NULL) { 2004 return (tcp_new_iss1(&tp->t_in6pcb->in6p_laddr, 2005 &tp->t_in6pcb->in6p_faddr, tp->t_in6pcb->in6p_lport, 2006 tp->t_in6pcb->in6p_fport, sizeof(tp->t_in6pcb->in6p_laddr), 2007 addin)); 2008 } 2009 #endif 2010 /* Not possible. */ 2011 panic("tcp_new_iss"); 2012 } 2013 2014 /* 2015 * This routine actually generates a new TCP initial sequence number. 2016 */ 2017 tcp_seq 2018 tcp_new_iss1(void *laddr, void *faddr, u_int16_t lport, u_int16_t fport, 2019 size_t addrsz, tcp_seq addin) 2020 { 2021 tcp_seq tcp_iss; 2022 2023 #if NRND > 0 2024 static int beenhere; 2025 2026 /* 2027 * If we haven't been here before, initialize our cryptographic 2028 * hash secret. 2029 */ 2030 if (beenhere == 0) { 2031 rnd_extract_data(tcp_iss_secret, sizeof(tcp_iss_secret), 2032 RND_EXTRACT_ANY); 2033 beenhere = 1; 2034 } 2035 2036 if (tcp_do_rfc1948) { 2037 MD5_CTX ctx; 2038 u_int8_t hash[16]; /* XXX MD5 knowledge */ 2039 2040 /* 2041 * Compute the base value of the ISS. It is a hash 2042 * of (saddr, sport, daddr, dport, secret). 2043 */ 2044 MD5Init(&ctx); 2045 2046 MD5Update(&ctx, (u_char *) laddr, addrsz); 2047 MD5Update(&ctx, (u_char *) &lport, sizeof(lport)); 2048 2049 MD5Update(&ctx, (u_char *) faddr, addrsz); 2050 MD5Update(&ctx, (u_char *) &fport, sizeof(fport)); 2051 2052 MD5Update(&ctx, tcp_iss_secret, sizeof(tcp_iss_secret)); 2053 2054 MD5Final(hash, &ctx); 2055 2056 memcpy(&tcp_iss, hash, sizeof(tcp_iss)); 2057 2058 /* 2059 * Now increment our "timer", and add it in to 2060 * the computed value. 2061 * 2062 * XXX Use `addin'? 2063 * XXX TCP_ISSINCR too large to use? 2064 */ 2065 tcp_iss_seq += TCP_ISSINCR; 2066 #ifdef TCPISS_DEBUG 2067 printf("ISS hash 0x%08x, ", tcp_iss); 2068 #endif 2069 tcp_iss += tcp_iss_seq + addin; 2070 #ifdef TCPISS_DEBUG 2071 printf("new ISS 0x%08x\n", tcp_iss); 2072 #endif 2073 } else 2074 #endif /* NRND > 0 */ 2075 { 2076 /* 2077 * Randomize. 2078 */ 2079 #if NRND > 0 2080 rnd_extract_data(&tcp_iss, sizeof(tcp_iss), RND_EXTRACT_ANY); 2081 #else 2082 tcp_iss = arc4random(); 2083 #endif 2084 2085 /* 2086 * If we were asked to add some amount to a known value, 2087 * we will take a random value obtained above, mask off 2088 * the upper bits, and add in the known value. We also 2089 * add in a constant to ensure that we are at least a 2090 * certain distance from the original value. 2091 * 2092 * This is used when an old connection is in timed wait 2093 * and we have a new one coming in, for instance. 2094 */ 2095 if (addin != 0) { 2096 #ifdef TCPISS_DEBUG 2097 printf("Random %08x, ", tcp_iss); 2098 #endif 2099 tcp_iss &= TCP_ISS_RANDOM_MASK; 2100 tcp_iss += addin + TCP_ISSINCR; 2101 #ifdef TCPISS_DEBUG 2102 printf("Old ISS %08x, ISS %08x\n", addin, tcp_iss); 2103 #endif 2104 } else { 2105 tcp_iss &= TCP_ISS_RANDOM_MASK; 2106 tcp_iss += tcp_iss_seq; 2107 tcp_iss_seq += TCP_ISSINCR; 2108 #ifdef TCPISS_DEBUG 2109 printf("ISS %08x\n", tcp_iss); 2110 #endif 2111 } 2112 } 2113 2114 if (tcp_compat_42) { 2115 /* 2116 * Limit it to the positive range for really old TCP 2117 * implementations. 2118 * Just AND off the top bit instead of checking if 2119 * is set first - saves a branch 50% of the time. 2120 */ 2121 tcp_iss &= 0x7fffffff; /* XXX */ 2122 } 2123 2124 return (tcp_iss); 2125 } 2126 2127 #if defined(IPSEC) || defined(FAST_IPSEC) 2128 /* compute ESP/AH header size for TCP, including outer IP header. */ 2129 size_t 2130 ipsec4_hdrsiz_tcp(struct tcpcb *tp) 2131 { 2132 struct inpcb *inp; 2133 size_t hdrsiz; 2134 2135 /* XXX mapped addr case (tp->t_in6pcb) */ 2136 if (!tp || !tp->t_template || !(inp = tp->t_inpcb)) 2137 return 0; 2138 switch (tp->t_family) { 2139 case AF_INET: 2140 /* XXX: should use currect direction. */ 2141 hdrsiz = ipsec4_hdrsiz(tp->t_template, IPSEC_DIR_OUTBOUND, inp); 2142 break; 2143 default: 2144 hdrsiz = 0; 2145 break; 2146 } 2147 2148 return hdrsiz; 2149 } 2150 2151 #ifdef INET6 2152 size_t 2153 ipsec6_hdrsiz_tcp(struct tcpcb *tp) 2154 { 2155 struct in6pcb *in6p; 2156 size_t hdrsiz; 2157 2158 if (!tp || !tp->t_template || !(in6p = tp->t_in6pcb)) 2159 return 0; 2160 switch (tp->t_family) { 2161 case AF_INET6: 2162 /* XXX: should use currect direction. */ 2163 hdrsiz = ipsec6_hdrsiz(tp->t_template, IPSEC_DIR_OUTBOUND, in6p); 2164 break; 2165 case AF_INET: 2166 /* mapped address case - tricky */ 2167 default: 2168 hdrsiz = 0; 2169 break; 2170 } 2171 2172 return hdrsiz; 2173 } 2174 #endif 2175 #endif /*IPSEC*/ 2176 2177 /* 2178 * Determine the length of the TCP options for this connection. 2179 * 2180 * XXX: What do we do for SACK, when we add that? Just reserve 2181 * all of the space? Otherwise we can't exactly be incrementing 2182 * cwnd by an amount that varies depending on the amount we last 2183 * had to SACK! 2184 */ 2185 2186 u_int 2187 tcp_optlen(struct tcpcb *tp) 2188 { 2189 u_int optlen; 2190 2191 optlen = 0; 2192 if ((tp->t_flags & (TF_REQ_TSTMP|TF_RCVD_TSTMP|TF_NOOPT)) == 2193 (TF_REQ_TSTMP | TF_RCVD_TSTMP)) 2194 optlen += TCPOLEN_TSTAMP_APPA; 2195 2196 #ifdef TCP_SIGNATURE 2197 #if defined(INET6) && defined(FAST_IPSEC) 2198 if (tp->t_family == AF_INET) 2199 #endif 2200 if (tp->t_flags & TF_SIGNATURE) 2201 optlen += TCPOLEN_SIGNATURE + 2; 2202 #endif /* TCP_SIGNATURE */ 2203 2204 return optlen; 2205 } 2206
Indexes created Mon Sep 22 10:09:38 GMT 2025