1 /* $OpenBSD: tcp_subr.c,v 1.107 2009/08/10 10:13:43 claudio Exp $ */ 2 /* $NetBSD: tcp_subr.c,v 1.22 1996/02/13 23:44:00 christos Exp $ */ 3 4 /* 5 * Copyright (c) 1982, 1986, 1988, 1990, 1993 6 * The Regents of the University of California. All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 * 32 * @(#)COPYRIGHT 1.1 (NRL) 17 January 1995 33 * 34 * NRL grants permission for redistribution and use in source and binary 35 * forms, with or without modification, of the software and documentation 36 * created at NRL provided that the following conditions are met: 37 * 38 * 1. Redistributions of source code must retain the above copyright 39 * notice, this list of conditions and the following disclaimer. 40 * 2. Redistributions in binary form must reproduce the above copyright 41 * notice, this list of conditions and the following disclaimer in the 42 * documentation and/or other materials provided with the distribution. 43 * 3. All advertising materials mentioning features or use of this software 44 * must display the following acknowledgements: 45 * This product includes software developed by the University of 46 * California, Berkeley and its contributors. 47 * This product includes software developed at the Information 48 * Technology Division, US Naval Research Laboratory. 49 * 4. Neither the name of the NRL nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS 54 * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 55 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A 56 * PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NRL OR 57 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 58 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 59 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 60 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 61 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 62 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 63 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 64 * 65 * The views and conclusions contained in the software and documentation 66 * are those of the authors and should not be interpreted as representing 67 * official policies, either expressed or implied, of the US Naval 68 * Research Laboratory (NRL). 69 */ 70 71 #include <sys/param.h> 72 #include <sys/systm.h> 73 #include <sys/proc.h> 74 #include <sys/mbuf.h> 75 #include <sys/socket.h> 76 #include <sys/socketvar.h> 77 #include <sys/protosw.h> 78 #include <sys/kernel.h> 79 #include <sys/pool.h> 80 81 #include <net/route.h> 82 #include <net/if.h> 83 84 #include <netinet/in.h> 85 #include <netinet/in_systm.h> 86 #include <netinet/ip.h> 87 #include <netinet/in_pcb.h> 88 #include <netinet/ip_var.h> 89 #include <netinet/ip_icmp.h> 90 #include <netinet/tcp.h> 91 #include <netinet/tcp_fsm.h> 92 #include <netinet/tcp_seq.h> 93 #include <netinet/tcp_timer.h> 94 #include <netinet/tcp_var.h> 95 #include <netinet/tcpip.h> 96 #include <dev/rndvar.h> 97 98 #ifdef INET6 99 #include <netinet6/in6_var.h> 100 #include <netinet6/ip6protosw.h> 101 #endif /* INET6 */ 102 103 #include <crypto/md5.h> 104 105 /* patchable/settable parameters for tcp */ 106 int tcp_mssdflt = TCP_MSS; 107 int tcp_rttdflt = TCPTV_SRTTDFLT / PR_SLOWHZ; 108 109 /* values controllable via sysctl */ 110 int tcp_do_rfc1323 = 1; 111 #ifdef TCP_SACK 112 int tcp_do_sack = 1; /* RFC 2018 selective ACKs */ 113 #endif 114 int tcp_ack_on_push = 0; /* set to enable immediate ACK-on-PUSH */ 115 #ifdef TCP_ECN 116 int tcp_do_ecn = 0; /* RFC3168 ECN enabled/disabled? */ 117 #endif 118 int tcp_do_rfc3390 = 1; /* RFC3390 Increasing TCP's Initial Window */ 119 120 u_int32_t tcp_now = 1; 121 122 #ifndef TCBHASHSIZE 123 #define TCBHASHSIZE 128 124 #endif 125 int tcbhashsize = TCBHASHSIZE; 126 127 /* syn hash parameters */ 128 #define TCP_SYN_HASH_SIZE 293 129 #define TCP_SYN_BUCKET_SIZE 35 130 int tcp_syn_cache_size = TCP_SYN_HASH_SIZE; 131 int tcp_syn_cache_limit = TCP_SYN_HASH_SIZE*TCP_SYN_BUCKET_SIZE; 132 int tcp_syn_bucket_limit = 3*TCP_SYN_BUCKET_SIZE; 133 struct syn_cache_head tcp_syn_cache[TCP_SYN_HASH_SIZE]; 134 135 int tcp_reass_limit = NMBCLUSTERS / 2; /* hardlimit for tcpqe_pool */ 136 #ifdef TCP_SACK 137 int tcp_sackhole_limit = 32*1024; /* hardlimit for sackhl_pool */ 138 #endif 139 140 #ifdef INET6 141 extern int ip6_defhlim; 142 #endif /* INET6 */ 143 144 struct pool tcpcb_pool; 145 struct pool tcpqe_pool; 146 #ifdef TCP_SACK 147 struct pool sackhl_pool; 148 #endif 149 150 struct tcpstat tcpstat; /* tcp statistics */ 151 tcp_seq tcp_iss; 152 153 /* 154 * Tcp initialization 155 */ 156 void 157 tcp_init() 158 { 159 tcp_iss = 1; /* wrong */ 160 pool_init(&tcpcb_pool, sizeof(struct tcpcb), 0, 0, 0, "tcpcbpl", 161 NULL); 162 pool_init(&tcpqe_pool, sizeof(struct tcpqent), 0, 0, 0, "tcpqepl", 163 NULL); 164 pool_sethardlimit(&tcpqe_pool, tcp_reass_limit, NULL, 0); 165 #ifdef TCP_SACK 166 pool_init(&sackhl_pool, sizeof(struct sackhole), 0, 0, 0, "sackhlpl", 167 NULL); 168 pool_sethardlimit(&sackhl_pool, tcp_sackhole_limit, NULL, 0); 169 #endif /* TCP_SACK */ 170 in_pcbinit(&tcbtable, tcbhashsize); 171 172 #ifdef INET6 173 /* 174 * Since sizeof(struct ip6_hdr) > sizeof(struct ip), we 175 * do max length checks/computations only on the former. 176 */ 177 if (max_protohdr < (sizeof(struct ip6_hdr) + sizeof(struct tcphdr))) 178 max_protohdr = (sizeof(struct ip6_hdr) + sizeof(struct tcphdr)); 179 if ((max_linkhdr + sizeof(struct ip6_hdr) + sizeof(struct tcphdr)) > 180 MHLEN) 181 panic("tcp_init"); 182 183 icmp6_mtudisc_callback_register(tcp6_mtudisc_callback); 184 #endif /* INET6 */ 185 186 /* Initialize the compressed state engine. */ 187 syn_cache_init(); 188 189 /* Initialize timer state. */ 190 tcp_timer_init(); 191 } 192 193 /* 194 * Create template to be used to send tcp packets on a connection. 195 * Call after host entry created, allocates an mbuf and fills 196 * in a skeletal tcp/ip header, minimizing the amount of work 197 * necessary when the connection is used. 198 * 199 * To support IPv6 in addition to IPv4 and considering that the sizes of 200 * the IPv4 and IPv6 headers are not the same, we now use a separate pointer 201 * for the TCP header. Also, we made the former tcpiphdr header pointer 202 * into just an IP overlay pointer, with casting as appropriate for v6. rja 203 */ 204 struct mbuf * 205 tcp_template(tp) 206 struct tcpcb *tp; 207 { 208 struct inpcb *inp = tp->t_inpcb; 209 struct mbuf *m; 210 struct tcphdr *th; 211 212 if ((m = tp->t_template) == 0) { 213 m = m_get(M_DONTWAIT, MT_HEADER); 214 if (m == NULL) 215 return (0); 216 217 switch (tp->pf) { 218 case 0: /*default to PF_INET*/ 219 #ifdef INET 220 case AF_INET: 221 m->m_len = sizeof(struct ip); 222 break; 223 #endif /* INET */ 224 #ifdef INET6 225 case AF_INET6: 226 m->m_len = sizeof(struct ip6_hdr); 227 break; 228 #endif /* INET6 */ 229 } 230 m->m_len += sizeof (struct tcphdr); 231 232 /* 233 * The link header, network header, TCP header, and TCP options 234 * all must fit in this mbuf. For now, assume the worst case of 235 * TCP options size. Eventually, compute this from tp flags. 236 */ 237 if (m->m_len + MAX_TCPOPTLEN + max_linkhdr >= MHLEN) { 238 MCLGET(m, M_DONTWAIT); 239 if ((m->m_flags & M_EXT) == 0) { 240 m_free(m); 241 return (0); 242 } 243 } 244 } 245 246 switch(tp->pf) { 247 #ifdef INET 248 case AF_INET: 249 { 250 struct ipovly *ipovly; 251 252 ipovly = mtod(m, struct ipovly *); 253 254 bzero(ipovly->ih_x1, sizeof ipovly->ih_x1); 255 ipovly->ih_pr = IPPROTO_TCP; 256 ipovly->ih_len = htons(sizeof (struct tcphdr)); 257 ipovly->ih_src = inp->inp_laddr; 258 ipovly->ih_dst = inp->inp_faddr; 259 260 th = (struct tcphdr *)(mtod(m, caddr_t) + 261 sizeof(struct ip)); 262 th->th_sum = in_cksum_phdr(ipovly->ih_src.s_addr, 263 ipovly->ih_dst.s_addr, 264 htons(sizeof (struct tcphdr) + IPPROTO_TCP)); 265 } 266 break; 267 #endif /* INET */ 268 #ifdef INET6 269 case AF_INET6: 270 { 271 struct ip6_hdr *ip6; 272 273 ip6 = mtod(m, struct ip6_hdr *); 274 275 ip6->ip6_src = inp->inp_laddr6; 276 ip6->ip6_dst = inp->inp_faddr6; 277 ip6->ip6_flow = htonl(0x60000000) | 278 (inp->inp_flowinfo & IPV6_FLOWLABEL_MASK); 279 280 ip6->ip6_nxt = IPPROTO_TCP; 281 ip6->ip6_plen = htons(sizeof(struct tcphdr)); /*XXX*/ 282 ip6->ip6_hlim = in6_selecthlim(inp, NULL); /*XXX*/ 283 284 th = (struct tcphdr *)(mtod(m, caddr_t) + 285 sizeof(struct ip6_hdr)); 286 th->th_sum = 0; 287 } 288 break; 289 #endif /* INET6 */ 290 } 291 292 th->th_sport = inp->inp_lport; 293 th->th_dport = inp->inp_fport; 294 th->th_seq = 0; 295 th->th_ack = 0; 296 th->th_x2 = 0; 297 th->th_off = 5; 298 th->th_flags = 0; 299 th->th_win = 0; 300 th->th_urp = 0; 301 return (m); 302 } 303 304 /* 305 * Send a single message to the TCP at address specified by 306 * the given TCP/IP header. If m == 0, then we make a copy 307 * of the tcpiphdr at ti and send directly to the addressed host. 308 * This is used to force keep alive messages out using the TCP 309 * template for a connection tp->t_template. If flags are given 310 * then we send a message back to the TCP which originated the 311 * segment ti, and discard the mbuf containing it and any other 312 * attached mbufs. 313 * 314 * In any case the ack and sequence number of the transmitted 315 * segment are as specified by the parameters. 316 */ 317 #ifdef INET6 318 /* This function looks hairy, because it was so IPv4-dependent. */ 319 #endif /* INET6 */ 320 void 321 tcp_respond(tp, template, th0, ack, seq, flags) 322 struct tcpcb *tp; 323 caddr_t template; 324 struct tcphdr *th0; 325 tcp_seq ack, seq; 326 int flags; 327 { 328 int tlen; 329 int win = 0; 330 struct mbuf *m = 0; 331 struct route *ro = 0; 332 struct tcphdr *th; 333 struct ip *ip; 334 struct ipovly *ih; 335 #ifdef INET6 336 struct ip6_hdr *ip6; 337 #endif 338 int af; /* af on wire */ 339 340 if (tp) { 341 win = sbspace(&tp->t_inpcb->inp_socket->so_rcv); 342 /* 343 * If this is called with an unconnected 344 * socket/tp/pcb (tp->pf is 0), we lose. 345 */ 346 af = tp->pf; 347 348 /* 349 * The route/route6 distinction is meaningless 350 * unless you're allocating space or passing parameters. 351 */ 352 ro = &tp->t_inpcb->inp_route; 353 } else 354 af = (((struct ip *)template)->ip_v == 6) ? AF_INET6 : AF_INET; 355 356 m = m_gethdr(M_DONTWAIT, MT_HEADER); 357 if (m == NULL) 358 return; 359 m->m_data += max_linkhdr; 360 tlen = 0; 361 362 #define xchg(a,b,type) do { type t; t=a; a=b; b=t; } while (0) 363 switch (af) { 364 #ifdef INET6 365 case AF_INET6: 366 ip6 = mtod(m, struct ip6_hdr *); 367 th = (struct tcphdr *)(ip6 + 1); 368 tlen = sizeof(*ip6) + sizeof(*th); 369 if (th0) { 370 bcopy(template, ip6, sizeof(*ip6)); 371 bcopy(th0, th, sizeof(*th)); 372 xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr); 373 } else { 374 bcopy(template, ip6, tlen); 375 } 376 break; 377 #endif /* INET6 */ 378 case AF_INET: 379 ip = mtod(m, struct ip *); 380 th = (struct tcphdr *)(ip + 1); 381 tlen = sizeof(*ip) + sizeof(*th); 382 if (th0) { 383 bcopy(template, ip, sizeof(*ip)); 384 bcopy(th0, th, sizeof(*th)); 385 xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, u_int32_t); 386 } else { 387 bcopy(template, ip, tlen); 388 } 389 break; 390 } 391 if (th0) 392 xchg(th->th_dport, th->th_sport, u_int16_t); 393 else 394 flags = TH_ACK; 395 #undef xchg 396 397 m->m_len = tlen; 398 m->m_pkthdr.len = tlen; 399 m->m_pkthdr.rcvif = (struct ifnet *) 0; 400 th->th_seq = htonl(seq); 401 th->th_ack = htonl(ack); 402 th->th_x2 = 0; 403 th->th_off = sizeof (struct tcphdr) >> 2; 404 th->th_flags = flags; 405 if (tp) 406 win >>= tp->rcv_scale; 407 if (win > TCP_MAXWIN) 408 win = TCP_MAXWIN; 409 th->th_win = htons((u_int16_t)win); 410 th->th_urp = 0; 411 412 switch (af) { 413 #ifdef INET6 414 case AF_INET6: 415 ip6->ip6_flow = htonl(0x60000000); 416 ip6->ip6_nxt = IPPROTO_TCP; 417 ip6->ip6_hlim = in6_selecthlim(tp ? tp->t_inpcb : NULL, NULL); /*XXX*/ 418 ip6->ip6_plen = tlen - sizeof(struct ip6_hdr); 419 th->th_sum = 0; 420 th->th_sum = in6_cksum(m, IPPROTO_TCP, 421 sizeof(struct ip6_hdr), ip6->ip6_plen); 422 HTONS(ip6->ip6_plen); 423 ip6_output(m, tp ? tp->t_inpcb->inp_outputopts6 : NULL, 424 (struct route_in6 *)ro, 0, NULL, NULL, 425 tp ? tp->t_inpcb : NULL); 426 break; 427 #endif /* INET6 */ 428 case AF_INET: 429 ih = (struct ipovly *)ip; 430 bzero(ih->ih_x1, sizeof ih->ih_x1); 431 ih->ih_len = htons((u_short)tlen - sizeof(struct ip)); 432 433 /* 434 * There's no point deferring to hardware checksum processing 435 * here, as we only send a minimal TCP packet whose checksum 436 * we need to compute in any case. 437 */ 438 th->th_sum = 0; 439 th->th_sum = in_cksum(m, tlen); 440 ip->ip_len = htons(tlen); 441 ip->ip_ttl = ip_defttl; 442 ip_output(m, (void *)NULL, ro, ip_mtudisc ? IP_MTUDISC : 0, 443 (void *)NULL, tp ? tp->t_inpcb : (void *)NULL); 444 } 445 } 446 447 /* 448 * Create a new TCP control block, making an 449 * empty reassembly queue and hooking it to the argument 450 * protocol control block. 451 */ 452 struct tcpcb * 453 tcp_newtcpcb(struct inpcb *inp) 454 { 455 struct tcpcb *tp; 456 int i; 457 458 tp = pool_get(&tcpcb_pool, PR_NOWAIT); 459 if (tp == NULL) 460 return ((struct tcpcb *)0); 461 bzero((char *) tp, sizeof(struct tcpcb)); 462 TAILQ_INIT(&tp->t_segq); 463 tp->t_maxseg = tcp_mssdflt; 464 tp->t_maxopd = 0; 465 466 TCP_INIT_DELACK(tp); 467 for (i = 0; i < TCPT_NTIMERS; i++) 468 TCP_TIMER_INIT(tp, i); 469 timeout_set(&tp->t_reap_to, tcp_reaper, tp); 470 471 #ifdef TCP_SACK 472 tp->sack_enable = tcp_do_sack; 473 #endif 474 tp->t_flags = tcp_do_rfc1323 ? (TF_REQ_SCALE|TF_REQ_TSTMP) : 0; 475 tp->t_inpcb = inp; 476 /* 477 * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no 478 * rtt estimate. Set rttvar so that srtt + 2 * rttvar gives 479 * reasonable initial retransmit time. 480 */ 481 tp->t_srtt = TCPTV_SRTTBASE; 482 tp->t_rttvar = tcp_rttdflt * PR_SLOWHZ << 483 (TCP_RTTVAR_SHIFT + TCP_RTT_BASE_SHIFT - 1); 484 tp->t_rttmin = TCPTV_MIN; 485 TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp), 486 TCPTV_MIN, TCPTV_REXMTMAX); 487 tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; 488 tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT; 489 490 tp->t_pmtud_mtu_sent = 0; 491 tp->t_pmtud_mss_acked = 0; 492 493 #ifdef INET6 494 /* we disallow IPv4 mapped address completely. */ 495 if ((inp->inp_flags & INP_IPV6) == 0) 496 tp->pf = PF_INET; 497 else 498 tp->pf = PF_INET6; 499 #else 500 tp->pf = PF_INET; 501 #endif 502 503 #ifdef INET6 504 if (inp->inp_flags & INP_IPV6) 505 inp->inp_ipv6.ip6_hlim = ip6_defhlim; 506 else 507 #endif /* INET6 */ 508 inp->inp_ip.ip_ttl = ip_defttl; 509 510 inp->inp_ppcb = (caddr_t)tp; 511 return (tp); 512 } 513 514 /* 515 * Drop a TCP connection, reporting 516 * the specified error. If connection is synchronized, 517 * then send a RST to peer. 518 */ 519 struct tcpcb * 520 tcp_drop(tp, errno) 521 struct tcpcb *tp; 522 int errno; 523 { 524 struct socket *so = tp->t_inpcb->inp_socket; 525 526 if (TCPS_HAVERCVDSYN(tp->t_state)) { 527 tp->t_state = TCPS_CLOSED; 528 (void) tcp_output(tp); 529 tcpstat.tcps_drops++; 530 } else 531 tcpstat.tcps_conndrops++; 532 if (errno == ETIMEDOUT && tp->t_softerror) 533 errno = tp->t_softerror; 534 so->so_error = errno; 535 return (tcp_close(tp)); 536 } 537 538 /* 539 * Close a TCP control block: 540 * discard all space held by the tcp 541 * discard internet protocol block 542 * wake up any sleepers 543 */ 544 struct tcpcb * 545 tcp_close(struct tcpcb *tp) 546 { 547 struct inpcb *inp = tp->t_inpcb; 548 struct socket *so = inp->inp_socket; 549 #ifdef TCP_SACK 550 struct sackhole *p, *q; 551 #endif 552 553 /* free the reassembly queue, if any */ 554 tcp_freeq(tp); 555 556 tcp_canceltimers(tp); 557 TCP_CLEAR_DELACK(tp); 558 syn_cache_cleanup(tp); 559 560 #ifdef TCP_SACK 561 /* Free SACK holes. */ 562 q = p = tp->snd_holes; 563 while (p != 0) { 564 q = p->next; 565 pool_put(&sackhl_pool, p); 566 p = q; 567 } 568 #endif 569 if (tp->t_template) 570 (void) m_free(tp->t_template); 571 572 tp->t_flags |= TF_DEAD; 573 timeout_add(&tp->t_reap_to, 0); 574 575 inp->inp_ppcb = 0; 576 soisdisconnected(so); 577 in_pcbdetach(inp); 578 return ((struct tcpcb *)0); 579 } 580 581 void 582 tcp_reaper(void *arg) 583 { 584 struct tcpcb *tp = arg; 585 int s; 586 587 s = splsoftnet(); 588 pool_put(&tcpcb_pool, tp); 589 splx(s); 590 tcpstat.tcps_closed++; 591 } 592 593 int 594 tcp_freeq(struct tcpcb *tp) 595 { 596 struct tcpqent *qe; 597 int rv = 0; 598 599 while ((qe = TAILQ_FIRST(&tp->t_segq)) != NULL) { 600 TAILQ_REMOVE(&tp->t_segq, qe, tcpqe_q); 601 m_freem(qe->tcpqe_m); 602 pool_put(&tcpqe_pool, qe); 603 rv = 1; 604 } 605 return (rv); 606 } 607 608 /* 609 * Compute proper scaling value for receiver window from buffer space 610 */ 611 612 void 613 tcp_rscale(struct tcpcb *tp, u_long hiwat) 614 { 615 tp->request_r_scale = 0; 616 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 617 TCP_MAXWIN << tp->request_r_scale < hiwat) 618 tp->request_r_scale++; 619 } 620 621 /* 622 * Notify a tcp user of an asynchronous error; 623 * store error as soft error, but wake up user 624 * (for now, won't do anything until can select for soft error). 625 */ 626 void 627 tcp_notify(inp, error) 628 struct inpcb *inp; 629 int error; 630 { 631 struct tcpcb *tp = (struct tcpcb *)inp->inp_ppcb; 632 struct socket *so = inp->inp_socket; 633 634 /* 635 * Ignore some errors if we are hooked up. 636 * If connection hasn't completed, has retransmitted several times, 637 * and receives a second error, give up now. This is better 638 * than waiting a long time to establish a connection that 639 * can never complete. 640 */ 641 if (tp->t_state == TCPS_ESTABLISHED && 642 (error == EHOSTUNREACH || error == ENETUNREACH || 643 error == EHOSTDOWN)) { 644 return; 645 } else if (TCPS_HAVEESTABLISHED(tp->t_state) == 0 && 646 tp->t_rxtshift > 3 && tp->t_softerror) 647 so->so_error = error; 648 else 649 tp->t_softerror = error; 650 wakeup((caddr_t) &so->so_timeo); 651 sorwakeup(so); 652 sowwakeup(so); 653 } 654 655 #ifdef INET6 656 void 657 tcp6_ctlinput(cmd, sa, d) 658 int cmd; 659 struct sockaddr *sa; 660 void *d; 661 { 662 struct tcphdr th; 663 struct tcpcb *tp; 664 void (*notify)(struct inpcb *, int) = tcp_notify; 665 struct ip6_hdr *ip6; 666 const struct sockaddr_in6 *sa6_src = NULL; 667 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)sa; 668 struct inpcb *inp; 669 struct mbuf *m; 670 tcp_seq seq; 671 int off; 672 struct { 673 u_int16_t th_sport; 674 u_int16_t th_dport; 675 u_int32_t th_seq; 676 } *thp; 677 678 if (sa->sa_family != AF_INET6 || 679 sa->sa_len != sizeof(struct sockaddr_in6) || 680 IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 681 IN6_IS_ADDR_V4MAPPED(&sa6->sin6_addr)) 682 return; 683 if ((unsigned)cmd >= PRC_NCMDS) 684 return; 685 else if (cmd == PRC_QUENCH) { 686 /* 687 * Don't honor ICMP Source Quench messages meant for 688 * TCP connections. 689 */ 690 /* XXX there's no PRC_QUENCH in IPv6 */ 691 return; 692 } else if (PRC_IS_REDIRECT(cmd)) 693 notify = in_rtchange, d = NULL; 694 else if (cmd == PRC_MSGSIZE) 695 ; /* special code is present, see below */ 696 else if (cmd == PRC_HOSTDEAD) 697 d = NULL; 698 else if (inet6ctlerrmap[cmd] == 0) 699 return; 700 701 /* if the parameter is from icmp6, decode it. */ 702 if (d != NULL) { 703 struct ip6ctlparam *ip6cp = (struct ip6ctlparam *)d; 704 m = ip6cp->ip6c_m; 705 ip6 = ip6cp->ip6c_ip6; 706 off = ip6cp->ip6c_off; 707 sa6_src = ip6cp->ip6c_src; 708 } else { 709 m = NULL; 710 ip6 = NULL; 711 sa6_src = &sa6_any; 712 } 713 714 if (ip6) { 715 /* 716 * XXX: We assume that when ip6 is non NULL, 717 * M and OFF are valid. 718 */ 719 720 /* check if we can safely examine src and dst ports */ 721 if (m->m_pkthdr.len < off + sizeof(*thp)) 722 return; 723 724 bzero(&th, sizeof(th)); 725 #ifdef DIAGNOSTIC 726 if (sizeof(*thp) > sizeof(th)) 727 panic("assumption failed in tcp6_ctlinput"); 728 #endif 729 m_copydata(m, off, sizeof(*thp), (caddr_t)&th); 730 731 /* 732 * Check to see if we have a valid TCP connection 733 * corresponding to the address in the ICMPv6 message 734 * payload. 735 */ 736 inp = in6_pcbhashlookup(&tcbtable, &sa6->sin6_addr, 737 th.th_dport, (struct in6_addr *)&sa6_src->sin6_addr, 738 th.th_sport); 739 if (cmd == PRC_MSGSIZE) { 740 /* 741 * Depending on the value of "valid" and routing table 742 * size (mtudisc_{hi,lo}wat), we will: 743 * - recalcurate the new MTU and create the 744 * corresponding routing entry, or 745 * - ignore the MTU change notification. 746 */ 747 icmp6_mtudisc_update((struct ip6ctlparam *)d, inp != NULL); 748 return; 749 } 750 if (inp) { 751 seq = ntohl(th.th_seq); 752 if (inp->inp_socket && 753 (tp = intotcpcb(inp)) && 754 SEQ_GEQ(seq, tp->snd_una) && 755 SEQ_LT(seq, tp->snd_max)) 756 notify(inp, inet6ctlerrmap[cmd]); 757 } else if (syn_cache_count && 758 (inet6ctlerrmap[cmd] == EHOSTUNREACH || 759 inet6ctlerrmap[cmd] == ENETUNREACH || 760 inet6ctlerrmap[cmd] == EHOSTDOWN)) 761 syn_cache_unreach((struct sockaddr *)sa6_src, 762 sa, &th, /* XXX */ 0); 763 } else { 764 (void) in6_pcbnotify(&tcbtable, sa, 0, 765 (struct sockaddr *)sa6_src, 0, cmd, NULL, notify); 766 } 767 } 768 #endif 769 770 void * 771 tcp_ctlinput(cmd, sa, v) 772 int cmd; 773 struct sockaddr *sa; 774 void *v; 775 { 776 struct ip *ip = v; 777 struct tcphdr *th; 778 struct tcpcb *tp; 779 struct inpcb *inp; 780 struct in_addr faddr; 781 tcp_seq seq; 782 u_int mtu; 783 extern int inetctlerrmap[]; 784 void (*notify)(struct inpcb *, int) = tcp_notify; 785 int errno; 786 787 if (sa->sa_family != AF_INET) 788 return NULL; 789 faddr = satosin(sa)->sin_addr; 790 if (faddr.s_addr == INADDR_ANY) 791 return NULL; 792 793 if ((unsigned)cmd >= PRC_NCMDS) 794 return NULL; 795 errno = inetctlerrmap[cmd]; 796 if (cmd == PRC_QUENCH) 797 /* 798 * Don't honor ICMP Source Quench messages meant for 799 * TCP connections. 800 */ 801 return NULL; 802 else if (PRC_IS_REDIRECT(cmd)) 803 notify = in_rtchange, ip = 0; 804 else if (cmd == PRC_MSGSIZE && ip_mtudisc && ip) { 805 /* 806 * Verify that the packet in the icmp payload refers 807 * to an existing TCP connection. 808 */ 809 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 810 seq = ntohl(th->th_seq); 811 inp = in_pcbhashlookup(&tcbtable, 812 ip->ip_dst, th->th_dport, ip->ip_src, th->th_sport, 813 /* XXX */ 0); 814 if (inp && (tp = intotcpcb(inp)) && 815 SEQ_GEQ(seq, tp->snd_una) && 816 SEQ_LT(seq, tp->snd_max)) { 817 struct icmp *icp; 818 icp = (struct icmp *)((caddr_t)ip - 819 offsetof(struct icmp, icmp_ip)); 820 821 /* 822 * If the ICMP message advertises a Next-Hop MTU 823 * equal or larger than the maximum packet size we have 824 * ever sent, drop the message. 825 */ 826 mtu = (u_int)ntohs(icp->icmp_nextmtu); 827 if (mtu >= tp->t_pmtud_mtu_sent) 828 return NULL; 829 if (mtu >= tcp_hdrsz(tp) + tp->t_pmtud_mss_acked) { 830 /* 831 * Calculate new MTU, and create corresponding 832 * route (traditional PMTUD). 833 */ 834 tp->t_flags &= ~TF_PMTUD_PEND; 835 /* XXX inherit rdomain from PCB */ 836 icmp_mtudisc(icp, 0); 837 } else { 838 /* 839 * Record the information got in the ICMP 840 * message; act on it later. 841 * If we had already recorded an ICMP message, 842 * replace the old one only if the new message 843 * refers to an older TCP segment 844 */ 845 if (tp->t_flags & TF_PMTUD_PEND) { 846 if (SEQ_LT(tp->t_pmtud_th_seq, seq)) 847 return NULL; 848 } else 849 tp->t_flags |= TF_PMTUD_PEND; 850 tp->t_pmtud_th_seq = seq; 851 tp->t_pmtud_nextmtu = icp->icmp_nextmtu; 852 tp->t_pmtud_ip_len = icp->icmp_ip.ip_len; 853 tp->t_pmtud_ip_hl = icp->icmp_ip.ip_hl; 854 return NULL; 855 } 856 } else { 857 /* ignore if we don't have a matching connection */ 858 return NULL; 859 } 860 notify = tcp_mtudisc, ip = 0; 861 } else if (cmd == PRC_MTUINC) 862 notify = tcp_mtudisc_increase, ip = 0; 863 else if (cmd == PRC_HOSTDEAD) 864 ip = 0; 865 else if (errno == 0) 866 return NULL; 867 868 if (ip) { 869 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 870 inp = in_pcbhashlookup(&tcbtable, 871 ip->ip_dst, th->th_dport, ip->ip_src, th->th_sport, 872 /* XXX */ 0); 873 if (inp) { 874 seq = ntohl(th->th_seq); 875 if (inp->inp_socket && 876 (tp = intotcpcb(inp)) && 877 SEQ_GEQ(seq, tp->snd_una) && 878 SEQ_LT(seq, tp->snd_max)) 879 notify(inp, errno); 880 } else if (syn_cache_count && 881 (inetctlerrmap[cmd] == EHOSTUNREACH || 882 inetctlerrmap[cmd] == ENETUNREACH || 883 inetctlerrmap[cmd] == EHOSTDOWN)) { 884 struct sockaddr_in sin; 885 886 bzero(&sin, sizeof(sin)); 887 sin.sin_len = sizeof(sin); 888 sin.sin_family = AF_INET; 889 sin.sin_port = th->th_sport; 890 sin.sin_addr = ip->ip_src; 891 syn_cache_unreach((struct sockaddr *)&sin, 892 sa, th, /* XXX */ 0); 893 } 894 } else 895 in_pcbnotifyall(&tcbtable, sa, errno, notify); 896 897 return NULL; 898 } 899 900 901 #ifdef INET6 902 /* 903 * Path MTU Discovery handlers. 904 */ 905 void 906 tcp6_mtudisc_callback(faddr) 907 struct in6_addr *faddr; 908 { 909 struct sockaddr_in6 sin6; 910 911 bzero(&sin6, sizeof(sin6)); 912 sin6.sin6_family = AF_INET6; 913 sin6.sin6_len = sizeof(struct sockaddr_in6); 914 sin6.sin6_addr = *faddr; 915 (void) in6_pcbnotify(&tcbtable, (struct sockaddr *)&sin6, 0, 916 (struct sockaddr *)&sa6_any, 0, PRC_MSGSIZE, NULL, tcp_mtudisc); 917 } 918 #endif /* INET6 */ 919 920 /* 921 * On receipt of path MTU corrections, flush old route and replace it 922 * with the new one. Retransmit all unacknowledged packets, to ensure 923 * that all packets will be received. 924 */ 925 void 926 tcp_mtudisc(inp, errno) 927 struct inpcb *inp; 928 int errno; 929 { 930 struct tcpcb *tp = intotcpcb(inp); 931 struct rtentry *rt = in_pcbrtentry(inp); 932 int change = 0; 933 934 if (tp != 0) { 935 int orig_maxseg = tp->t_maxseg; 936 if (rt != 0) { 937 /* 938 * If this was not a host route, remove and realloc. 939 */ 940 if ((rt->rt_flags & RTF_HOST) == 0) { 941 in_rtchange(inp, errno); 942 if ((rt = in_pcbrtentry(inp)) == 0) 943 return; 944 } 945 if (orig_maxseg != tp->t_maxseg || 946 (rt->rt_rmx.rmx_locks & RTV_MTU)) 947 change = 1; 948 } 949 tcp_mss(tp, -1); 950 951 /* 952 * Resend unacknowledged packets 953 */ 954 tp->snd_nxt = tp->snd_una; 955 if (change || errno > 0) 956 tcp_output(tp); 957 } 958 } 959 960 void 961 tcp_mtudisc_increase(inp, errno) 962 struct inpcb *inp; 963 int errno; 964 { 965 struct tcpcb *tp = intotcpcb(inp); 966 struct rtentry *rt = in_pcbrtentry(inp); 967 968 if (tp != 0 && rt != 0) { 969 /* 970 * If this was a host route, remove and realloc. 971 */ 972 if (rt->rt_flags & RTF_HOST) 973 in_rtchange(inp, errno); 974 975 /* also takes care of congestion window */ 976 tcp_mss(tp, -1); 977 } 978 } 979 980 #define TCP_ISS_CONN_INC 4096 981 int tcp_secret_init; 982 u_char tcp_secret[16]; 983 MD5_CTX tcp_secret_ctx; 984 985 void 986 tcp_set_iss_tsm(struct tcpcb *tp) 987 { 988 MD5_CTX ctx; 989 u_int32_t digest[4]; 990 991 if (tcp_secret_init == 0) { 992 arc4random_buf(tcp_secret, sizeof(tcp_secret)); 993 MD5Init(&tcp_secret_ctx); 994 MD5Update(&tcp_secret_ctx, tcp_secret, sizeof(tcp_secret)); 995 tcp_secret_init = 1; 996 } 997 ctx = tcp_secret_ctx; 998 MD5Update(&ctx, (char *)&tp->t_inpcb->inp_lport, sizeof(u_short)); 999 MD5Update(&ctx, (char *)&tp->t_inpcb->inp_fport, sizeof(u_short)); 1000 if (tp->pf == AF_INET6) { 1001 MD5Update(&ctx, (char *)&tp->t_inpcb->inp_laddr6, 1002 sizeof(struct in6_addr)); 1003 MD5Update(&ctx, (char *)&tp->t_inpcb->inp_faddr6, 1004 sizeof(struct in6_addr)); 1005 } else { 1006 MD5Update(&ctx, (char *)&tp->t_inpcb->inp_laddr, 1007 sizeof(struct in_addr)); 1008 MD5Update(&ctx, (char *)&tp->t_inpcb->inp_faddr, 1009 sizeof(struct in_addr)); 1010 } 1011 MD5Final((u_char *)digest, &ctx); 1012 tcp_iss += TCP_ISS_CONN_INC; 1013 tp->iss = digest[0] + tcp_iss; 1014 tp->ts_modulate = digest[1]; 1015 } 1016 1017 #ifdef TCP_SIGNATURE 1018 int 1019 tcp_signature_tdb_attach() 1020 { 1021 return (0); 1022 } 1023 1024 int 1025 tcp_signature_tdb_init(tdbp, xsp, ii) 1026 struct tdb *tdbp; 1027 struct xformsw *xsp; 1028 struct ipsecinit *ii; 1029 { 1030 if ((ii->ii_authkeylen < 1) || (ii->ii_authkeylen > 80)) 1031 return (EINVAL); 1032 1033 tdbp->tdb_amxkey = malloc(ii->ii_authkeylen, M_XDATA, M_DONTWAIT); 1034 if (tdbp->tdb_amxkey == NULL) 1035 return (ENOMEM); 1036 bcopy(ii->ii_authkey, tdbp->tdb_amxkey, ii->ii_authkeylen); 1037 tdbp->tdb_amxkeylen = ii->ii_authkeylen; 1038 1039 return (0); 1040 } 1041 1042 int 1043 tcp_signature_tdb_zeroize(tdbp) 1044 struct tdb *tdbp; 1045 { 1046 if (tdbp->tdb_amxkey) { 1047 bzero(tdbp->tdb_amxkey, tdbp->tdb_amxkeylen); 1048 free(tdbp->tdb_amxkey, M_XDATA); 1049 tdbp->tdb_amxkey = NULL; 1050 } 1051 1052 return (0); 1053 } 1054 1055 int 1056 tcp_signature_tdb_input(m, tdbp, skip, protoff) 1057 struct mbuf *m; 1058 struct tdb *tdbp; 1059 int skip, protoff; 1060 { 1061 return (0); 1062 } 1063 1064 int 1065 tcp_signature_tdb_output(m, tdbp, mp, skip, protoff) 1066 struct mbuf *m; 1067 struct tdb *tdbp; 1068 struct mbuf **mp; 1069 int skip, protoff; 1070 { 1071 return (EINVAL); 1072 } 1073 1074 int 1075 tcp_signature_apply(fstate, data, len) 1076 caddr_t fstate; 1077 caddr_t data; 1078 unsigned int len; 1079 { 1080 MD5Update((MD5_CTX *)fstate, (char *)data, len); 1081 return 0; 1082 } 1083 1084 int 1085 tcp_signature(struct tdb *tdb, int af, struct mbuf *m, struct tcphdr *th, 1086 int iphlen, int doswap, char *sig) 1087 { 1088 MD5_CTX ctx; 1089 int len; 1090 struct tcphdr th0; 1091 1092 MD5Init(&ctx); 1093 1094 switch(af) { 1095 case 0: 1096 #ifdef INET 1097 case AF_INET: { 1098 struct ippseudo ippseudo; 1099 struct ip *ip; 1100 1101 ip = mtod(m, struct ip *); 1102 1103 ippseudo.ippseudo_src = ip->ip_src; 1104 ippseudo.ippseudo_dst = ip->ip_dst; 1105 ippseudo.ippseudo_pad = 0; 1106 ippseudo.ippseudo_p = IPPROTO_TCP; 1107 ippseudo.ippseudo_len = htons(m->m_pkthdr.len - iphlen); 1108 1109 MD5Update(&ctx, (char *)&ippseudo, 1110 sizeof(struct ippseudo)); 1111 break; 1112 } 1113 #endif 1114 #ifdef INET6 1115 case AF_INET6: { 1116 struct ip6_hdr_pseudo ip6pseudo; 1117 struct ip6_hdr *ip6; 1118 1119 ip6 = mtod(m, struct ip6_hdr *); 1120 bzero(&ip6pseudo, sizeof(ip6pseudo)); 1121 ip6pseudo.ip6ph_src = ip6->ip6_src; 1122 ip6pseudo.ip6ph_dst = ip6->ip6_dst; 1123 in6_clearscope(&ip6pseudo.ip6ph_src); 1124 in6_clearscope(&ip6pseudo.ip6ph_dst); 1125 ip6pseudo.ip6ph_nxt = IPPROTO_TCP; 1126 ip6pseudo.ip6ph_len = htonl(m->m_pkthdr.len - iphlen); 1127 1128 MD5Update(&ctx, (char *)&ip6pseudo, 1129 sizeof(ip6pseudo)); 1130 break; 1131 } 1132 #endif 1133 } 1134 1135 th0 = *th; 1136 th0.th_sum = 0; 1137 1138 if (doswap) { 1139 HTONL(th0.th_seq); 1140 HTONL(th0.th_ack); 1141 HTONS(th0.th_win); 1142 HTONS(th0.th_urp); 1143 } 1144 MD5Update(&ctx, (char *)&th0, sizeof(th0)); 1145 1146 len = m->m_pkthdr.len - iphlen - th->th_off * sizeof(uint32_t); 1147 1148 if (len > 0 && 1149 m_apply(m, iphlen + th->th_off * sizeof(uint32_t), len, 1150 tcp_signature_apply, (caddr_t)&ctx)) 1151 return (-1); 1152 1153 MD5Update(&ctx, tdb->tdb_amxkey, tdb->tdb_amxkeylen); 1154 MD5Final(sig, &ctx); 1155 1156 return (0); 1157 } 1158 #endif /* TCP_SIGNATURE */ 1159