1 /* $OpenBSD: tcp_subr.c,v 1.155 2016/09/15 02:00:18 dlg 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/mbuf.h> 74 #include <sys/socket.h> 75 #include <sys/socketvar.h> 76 #include <sys/timeout.h> 77 #include <sys/protosw.h> 78 #include <sys/kernel.h> 79 #include <sys/pool.h> 80 81 #include <net/route.h> 82 83 #include <netinet/in.h> 84 #include <netinet/ip.h> 85 #include <netinet/in_pcb.h> 86 #include <netinet/ip_var.h> 87 #include <netinet/ip_icmp.h> 88 #include <netinet/tcp.h> 89 #include <netinet/tcp_fsm.h> 90 #include <netinet/tcp_seq.h> 91 #include <netinet/tcp_timer.h> 92 #include <netinet/tcp_var.h> 93 #include <netinet/tcpip.h> 94 95 #ifdef INET6 96 #include <netinet6/ip6protosw.h> 97 #endif /* INET6 */ 98 99 #include <crypto/md5.h> 100 #include <crypto/sha2.h> 101 102 /* patchable/settable parameters for tcp */ 103 int tcp_mssdflt = TCP_MSS; 104 int tcp_rttdflt = TCPTV_SRTTDFLT / PR_SLOWHZ; 105 106 /* values controllable via sysctl */ 107 int tcp_do_rfc1323 = 1; 108 #ifdef TCP_SACK 109 int tcp_do_sack = 1; /* RFC 2018 selective ACKs */ 110 #endif 111 int tcp_ack_on_push = 0; /* set to enable immediate ACK-on-PUSH */ 112 #ifdef TCP_ECN 113 int tcp_do_ecn = 0; /* RFC3168 ECN enabled/disabled? */ 114 #endif 115 int tcp_do_rfc3390 = 2; /* Increase TCP's Initial Window to 10*mss */ 116 117 u_int32_t tcp_now = 1; 118 119 #ifndef TCB_INITIAL_HASH_SIZE 120 #define TCB_INITIAL_HASH_SIZE 128 121 #endif 122 123 int tcp_reass_limit = NMBCLUSTERS / 8; /* hardlimit for tcpqe_pool */ 124 #ifdef TCP_SACK 125 int tcp_sackhole_limit = 32*1024; /* hardlimit for sackhl_pool */ 126 #endif 127 128 struct pool tcpcb_pool; 129 struct pool tcpqe_pool; 130 #ifdef TCP_SACK 131 struct pool sackhl_pool; 132 #endif 133 134 struct tcpstat tcpstat; /* tcp statistics */ 135 tcp_seq tcp_iss; 136 137 /* 138 * Tcp initialization 139 */ 140 void 141 tcp_init(void) 142 { 143 tcp_iss = 1; /* wrong */ 144 pool_init(&tcpcb_pool, sizeof(struct tcpcb), 0, IPL_SOFTNET, 0, 145 "tcpcb", NULL); 146 pool_init(&tcpqe_pool, sizeof(struct tcpqent), 0, IPL_SOFTNET, 0, 147 "tcpqe", NULL); 148 pool_sethardlimit(&tcpqe_pool, tcp_reass_limit, NULL, 0); 149 #ifdef TCP_SACK 150 pool_init(&sackhl_pool, sizeof(struct sackhole), 0, IPL_SOFTNET, 0, 151 "sackhl", NULL); 152 pool_sethardlimit(&sackhl_pool, tcp_sackhole_limit, NULL, 0); 153 #endif /* TCP_SACK */ 154 in_pcbinit(&tcbtable, TCB_INITIAL_HASH_SIZE); 155 156 #ifdef INET6 157 /* 158 * Since sizeof(struct ip6_hdr) > sizeof(struct ip), we 159 * do max length checks/computations only on the former. 160 */ 161 if (max_protohdr < (sizeof(struct ip6_hdr) + sizeof(struct tcphdr))) 162 max_protohdr = (sizeof(struct ip6_hdr) + sizeof(struct tcphdr)); 163 if ((max_linkhdr + sizeof(struct ip6_hdr) + sizeof(struct tcphdr)) > 164 MHLEN) 165 panic("tcp_init"); 166 167 icmp6_mtudisc_callback_register(tcp6_mtudisc_callback); 168 #endif /* INET6 */ 169 170 /* Initialize the compressed state engine. */ 171 syn_cache_init(); 172 173 /* Initialize timer state. */ 174 tcp_timer_init(); 175 } 176 177 /* 178 * Create template to be used to send tcp packets on a connection. 179 * Call after host entry created, allocates an mbuf and fills 180 * in a skeletal tcp/ip header, minimizing the amount of work 181 * necessary when the connection is used. 182 * 183 * To support IPv6 in addition to IPv4 and considering that the sizes of 184 * the IPv4 and IPv6 headers are not the same, we now use a separate pointer 185 * for the TCP header. Also, we made the former tcpiphdr header pointer 186 * into just an IP overlay pointer, with casting as appropriate for v6. rja 187 */ 188 struct mbuf * 189 tcp_template(tp) 190 struct tcpcb *tp; 191 { 192 struct inpcb *inp = tp->t_inpcb; 193 struct mbuf *m; 194 struct tcphdr *th; 195 196 if ((m = tp->t_template) == 0) { 197 m = m_get(M_DONTWAIT, MT_HEADER); 198 if (m == NULL) 199 return (0); 200 201 switch (tp->pf) { 202 case 0: /*default to PF_INET*/ 203 case AF_INET: 204 m->m_len = sizeof(struct ip); 205 break; 206 #ifdef INET6 207 case AF_INET6: 208 m->m_len = sizeof(struct ip6_hdr); 209 break; 210 #endif /* INET6 */ 211 } 212 m->m_len += sizeof (struct tcphdr); 213 214 /* 215 * The link header, network header, TCP header, and TCP options 216 * all must fit in this mbuf. For now, assume the worst case of 217 * TCP options size. Eventually, compute this from tp flags. 218 */ 219 if (m->m_len + MAX_TCPOPTLEN + max_linkhdr >= MHLEN) { 220 MCLGET(m, M_DONTWAIT); 221 if ((m->m_flags & M_EXT) == 0) { 222 m_free(m); 223 return (0); 224 } 225 } 226 } 227 228 switch(tp->pf) { 229 case AF_INET: 230 { 231 struct ipovly *ipovly; 232 233 ipovly = mtod(m, struct ipovly *); 234 235 bzero(ipovly->ih_x1, sizeof ipovly->ih_x1); 236 ipovly->ih_pr = IPPROTO_TCP; 237 ipovly->ih_len = htons(sizeof (struct tcphdr)); 238 ipovly->ih_src = inp->inp_laddr; 239 ipovly->ih_dst = inp->inp_faddr; 240 241 th = (struct tcphdr *)(mtod(m, caddr_t) + 242 sizeof(struct ip)); 243 } 244 break; 245 #ifdef INET6 246 case AF_INET6: 247 { 248 struct ip6_hdr *ip6; 249 250 ip6 = mtod(m, struct ip6_hdr *); 251 252 ip6->ip6_src = inp->inp_laddr6; 253 ip6->ip6_dst = inp->inp_faddr6; 254 ip6->ip6_flow = htonl(0x60000000) | 255 (inp->inp_flowinfo & IPV6_FLOWLABEL_MASK); 256 257 ip6->ip6_nxt = IPPROTO_TCP; 258 ip6->ip6_plen = htons(sizeof(struct tcphdr)); /*XXX*/ 259 ip6->ip6_hlim = in6_selecthlim(inp); /*XXX*/ 260 261 th = (struct tcphdr *)(mtod(m, caddr_t) + 262 sizeof(struct ip6_hdr)); 263 } 264 break; 265 #endif /* INET6 */ 266 } 267 268 th->th_sport = inp->inp_lport; 269 th->th_dport = inp->inp_fport; 270 th->th_seq = 0; 271 th->th_ack = 0; 272 th->th_x2 = 0; 273 th->th_off = 5; 274 th->th_flags = 0; 275 th->th_win = 0; 276 th->th_urp = 0; 277 th->th_sum = 0; 278 return (m); 279 } 280 281 /* 282 * Send a single message to the TCP at address specified by 283 * the given TCP/IP header. If m == 0, then we make a copy 284 * of the tcpiphdr at ti and send directly to the addressed host. 285 * This is used to force keep alive messages out using the TCP 286 * template for a connection tp->t_template. If flags are given 287 * then we send a message back to the TCP which originated the 288 * segment ti, and discard the mbuf containing it and any other 289 * attached mbufs. 290 * 291 * In any case the ack and sequence number of the transmitted 292 * segment are as specified by the parameters. 293 */ 294 void 295 tcp_respond(struct tcpcb *tp, caddr_t template, struct tcphdr *th0, 296 tcp_seq ack, tcp_seq seq, int flags, u_int rtableid) 297 { 298 int tlen; 299 int win = 0; 300 struct mbuf *m = NULL; 301 struct tcphdr *th; 302 struct ip *ip; 303 #ifdef INET6 304 struct ip6_hdr *ip6; 305 #endif 306 int af; /* af on wire */ 307 308 if (tp) { 309 win = sbspace(&tp->t_inpcb->inp_socket->so_rcv); 310 /* 311 * If this is called with an unconnected 312 * socket/tp/pcb (tp->pf is 0), we lose. 313 */ 314 af = tp->pf; 315 } else 316 af = (((struct ip *)template)->ip_v == 6) ? AF_INET6 : AF_INET; 317 318 m = m_gethdr(M_DONTWAIT, MT_HEADER); 319 if (m == NULL) 320 return; 321 m->m_data += max_linkhdr; 322 tlen = 0; 323 324 #define xchg(a,b,type) do { type t; t=a; a=b; b=t; } while (0) 325 switch (af) { 326 #ifdef INET6 327 case AF_INET6: 328 ip6 = mtod(m, struct ip6_hdr *); 329 th = (struct tcphdr *)(ip6 + 1); 330 tlen = sizeof(*ip6) + sizeof(*th); 331 if (th0) { 332 bcopy(template, ip6, sizeof(*ip6)); 333 bcopy(th0, th, sizeof(*th)); 334 xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr); 335 } else { 336 bcopy(template, ip6, tlen); 337 } 338 break; 339 #endif /* INET6 */ 340 case AF_INET: 341 ip = mtod(m, struct ip *); 342 th = (struct tcphdr *)(ip + 1); 343 tlen = sizeof(*ip) + sizeof(*th); 344 if (th0) { 345 bcopy(template, ip, sizeof(*ip)); 346 bcopy(th0, th, sizeof(*th)); 347 xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, u_int32_t); 348 } else { 349 bcopy(template, ip, tlen); 350 } 351 break; 352 } 353 if (th0) 354 xchg(th->th_dport, th->th_sport, u_int16_t); 355 else 356 flags = TH_ACK; 357 #undef xchg 358 359 th->th_seq = htonl(seq); 360 th->th_ack = htonl(ack); 361 th->th_x2 = 0; 362 th->th_off = sizeof (struct tcphdr) >> 2; 363 th->th_flags = flags; 364 if (tp) 365 win >>= tp->rcv_scale; 366 if (win > TCP_MAXWIN) 367 win = TCP_MAXWIN; 368 th->th_win = htons((u_int16_t)win); 369 th->th_urp = 0; 370 371 if (tp && (tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP && 372 (flags & TH_RST) == 0 && (tp->t_flags & TF_RCVD_TSTMP)) { 373 u_int32_t *lp = (u_int32_t *)(th + 1); 374 /* Form timestamp option as shown in appendix A of RFC 1323. */ 375 *lp++ = htonl(TCPOPT_TSTAMP_HDR); 376 *lp++ = htonl(tcp_now + tp->ts_modulate); 377 *lp = htonl(tp->ts_recent); 378 tlen += TCPOLEN_TSTAMP_APPA; 379 th->th_off = (sizeof(struct tcphdr) + TCPOLEN_TSTAMP_APPA) >> 2; 380 } 381 382 m->m_len = tlen; 383 m->m_pkthdr.len = tlen; 384 m->m_pkthdr.ph_ifidx = 0; 385 m->m_pkthdr.csum_flags |= M_TCP_CSUM_OUT; 386 387 /* force routing table */ 388 if (tp) 389 m->m_pkthdr.ph_rtableid = tp->t_inpcb->inp_rtableid; 390 else 391 m->m_pkthdr.ph_rtableid = rtableid; 392 393 switch (af) { 394 #ifdef INET6 395 case AF_INET6: 396 ip6->ip6_flow = htonl(0x60000000); 397 ip6->ip6_nxt = IPPROTO_TCP; 398 ip6->ip6_hlim = in6_selecthlim(tp ? tp->t_inpcb : NULL); /*XXX*/ 399 ip6->ip6_plen = tlen - sizeof(struct ip6_hdr); 400 ip6->ip6_plen = htons(ip6->ip6_plen); 401 ip6_output(m, tp ? tp->t_inpcb->inp_outputopts6 : NULL, 402 tp ? &tp->t_inpcb->inp_route6 : NULL, 403 0, NULL, 404 tp ? tp->t_inpcb : NULL); 405 break; 406 #endif /* INET6 */ 407 case AF_INET: 408 ip->ip_len = htons(tlen); 409 ip->ip_ttl = ip_defttl; 410 ip->ip_tos = 0; 411 ip_output(m, NULL, 412 tp ? &tp->t_inpcb->inp_route : NULL, 413 ip_mtudisc ? IP_MTUDISC : 0, NULL, 414 tp ? tp->t_inpcb : NULL, 0); 415 break; 416 } 417 } 418 419 /* 420 * Create a new TCP control block, making an 421 * empty reassembly queue and hooking it to the argument 422 * protocol control block. 423 */ 424 struct tcpcb * 425 tcp_newtcpcb(struct inpcb *inp) 426 { 427 struct tcpcb *tp; 428 int i; 429 430 tp = pool_get(&tcpcb_pool, PR_NOWAIT|PR_ZERO); 431 if (tp == NULL) 432 return (NULL); 433 TAILQ_INIT(&tp->t_segq); 434 tp->t_maxseg = tcp_mssdflt; 435 tp->t_maxopd = 0; 436 437 TCP_INIT_DELACK(tp); 438 for (i = 0; i < TCPT_NTIMERS; i++) 439 TCP_TIMER_INIT(tp, i); 440 timeout_set(&tp->t_reap_to, tcp_reaper, tp); 441 442 #ifdef TCP_SACK 443 tp->sack_enable = tcp_do_sack; 444 #endif 445 tp->t_flags = tcp_do_rfc1323 ? (TF_REQ_SCALE|TF_REQ_TSTMP) : 0; 446 tp->t_inpcb = inp; 447 /* 448 * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no 449 * rtt estimate. Set rttvar so that srtt + 2 * rttvar gives 450 * reasonable initial retransmit time. 451 */ 452 tp->t_srtt = TCPTV_SRTTBASE; 453 tp->t_rttvar = tcp_rttdflt * PR_SLOWHZ << 454 (TCP_RTTVAR_SHIFT + TCP_RTT_BASE_SHIFT - 1); 455 tp->t_rttmin = TCPTV_MIN; 456 TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp), 457 TCPTV_MIN, TCPTV_REXMTMAX); 458 tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; 459 tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT; 460 461 tp->t_pmtud_mtu_sent = 0; 462 tp->t_pmtud_mss_acked = 0; 463 464 #ifdef INET6 465 /* we disallow IPv4 mapped address completely. */ 466 if ((inp->inp_flags & INP_IPV6) == 0) 467 tp->pf = PF_INET; 468 else 469 tp->pf = PF_INET6; 470 #else 471 tp->pf = PF_INET; 472 #endif 473 474 #ifdef INET6 475 if (inp->inp_flags & INP_IPV6) 476 inp->inp_ipv6.ip6_hlim = ip6_defhlim; 477 else 478 #endif /* INET6 */ 479 inp->inp_ip.ip_ttl = ip_defttl; 480 481 inp->inp_ppcb = (caddr_t)tp; 482 return (tp); 483 } 484 485 /* 486 * Drop a TCP connection, reporting 487 * the specified error. If connection is synchronized, 488 * then send a RST to peer. 489 */ 490 struct tcpcb * 491 tcp_drop(tp, errno) 492 struct tcpcb *tp; 493 int errno; 494 { 495 struct socket *so = tp->t_inpcb->inp_socket; 496 497 if (TCPS_HAVERCVDSYN(tp->t_state)) { 498 tp->t_state = TCPS_CLOSED; 499 (void) tcp_output(tp); 500 tcpstat.tcps_drops++; 501 } else 502 tcpstat.tcps_conndrops++; 503 if (errno == ETIMEDOUT && tp->t_softerror) 504 errno = tp->t_softerror; 505 so->so_error = errno; 506 return (tcp_close(tp)); 507 } 508 509 /* 510 * Close a TCP control block: 511 * discard all space held by the tcp 512 * discard internet protocol block 513 * wake up any sleepers 514 */ 515 struct tcpcb * 516 tcp_close(struct tcpcb *tp) 517 { 518 struct inpcb *inp = tp->t_inpcb; 519 struct socket *so = inp->inp_socket; 520 #ifdef TCP_SACK 521 struct sackhole *p, *q; 522 #endif 523 524 /* free the reassembly queue, if any */ 525 tcp_freeq(tp); 526 527 tcp_canceltimers(tp); 528 TCP_CLEAR_DELACK(tp); 529 syn_cache_cleanup(tp); 530 531 #ifdef TCP_SACK 532 /* Free SACK holes. */ 533 q = p = tp->snd_holes; 534 while (p != 0) { 535 q = p->next; 536 pool_put(&sackhl_pool, p); 537 p = q; 538 } 539 #endif 540 if (tp->t_template) 541 (void) m_free(tp->t_template); 542 543 tp->t_flags |= TF_DEAD; 544 timeout_add(&tp->t_reap_to, 0); 545 546 inp->inp_ppcb = 0; 547 soisdisconnected(so); 548 in_pcbdetach(inp); 549 return (NULL); 550 } 551 552 void 553 tcp_reaper(void *arg) 554 { 555 struct tcpcb *tp = arg; 556 int s; 557 558 s = splsoftnet(); 559 pool_put(&tcpcb_pool, tp); 560 splx(s); 561 tcpstat.tcps_closed++; 562 } 563 564 int 565 tcp_freeq(struct tcpcb *tp) 566 { 567 struct tcpqent *qe; 568 int rv = 0; 569 570 while ((qe = TAILQ_FIRST(&tp->t_segq)) != NULL) { 571 TAILQ_REMOVE(&tp->t_segq, qe, tcpqe_q); 572 m_freem(qe->tcpqe_m); 573 pool_put(&tcpqe_pool, qe); 574 rv = 1; 575 } 576 return (rv); 577 } 578 579 /* 580 * Compute proper scaling value for receiver window from buffer space 581 */ 582 583 void 584 tcp_rscale(struct tcpcb *tp, u_long hiwat) 585 { 586 tp->request_r_scale = 0; 587 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 588 TCP_MAXWIN << tp->request_r_scale < hiwat) 589 tp->request_r_scale++; 590 } 591 592 /* 593 * Notify a tcp user of an asynchronous error; 594 * store error as soft error, but wake up user 595 * (for now, won't do anything until can select for soft error). 596 */ 597 void 598 tcp_notify(inp, error) 599 struct inpcb *inp; 600 int error; 601 { 602 struct tcpcb *tp = intotcpcb(inp); 603 struct socket *so = inp->inp_socket; 604 605 /* 606 * Ignore some errors if we are hooked up. 607 * If connection hasn't completed, has retransmitted several times, 608 * and receives a second error, give up now. This is better 609 * than waiting a long time to establish a connection that 610 * can never complete. 611 */ 612 if (tp->t_state == TCPS_ESTABLISHED && 613 (error == EHOSTUNREACH || error == ENETUNREACH || 614 error == EHOSTDOWN)) { 615 return; 616 } else if (TCPS_HAVEESTABLISHED(tp->t_state) == 0 && 617 tp->t_rxtshift > 3 && tp->t_softerror) 618 so->so_error = error; 619 else 620 tp->t_softerror = error; 621 wakeup((caddr_t) &so->so_timeo); 622 sorwakeup(so); 623 sowwakeup(so); 624 } 625 626 #ifdef INET6 627 void 628 tcp6_ctlinput(int cmd, struct sockaddr *sa, u_int rdomain, void *d) 629 { 630 struct tcphdr th; 631 struct tcpcb *tp; 632 void (*notify)(struct inpcb *, int) = tcp_notify; 633 struct ip6_hdr *ip6; 634 const struct sockaddr_in6 *sa6_src = NULL; 635 struct sockaddr_in6 *sa6 = satosin6(sa); 636 struct inpcb *inp; 637 struct mbuf *m; 638 tcp_seq seq; 639 int off; 640 struct { 641 u_int16_t th_sport; 642 u_int16_t th_dport; 643 u_int32_t th_seq; 644 } *thp; 645 646 if (sa->sa_family != AF_INET6 || 647 sa->sa_len != sizeof(struct sockaddr_in6) || 648 IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 649 IN6_IS_ADDR_V4MAPPED(&sa6->sin6_addr)) 650 return; 651 if ((unsigned)cmd >= PRC_NCMDS) 652 return; 653 else if (cmd == PRC_QUENCH) { 654 /* 655 * Don't honor ICMP Source Quench messages meant for 656 * TCP connections. 657 */ 658 /* XXX there's no PRC_QUENCH in IPv6 */ 659 return; 660 } else if (PRC_IS_REDIRECT(cmd)) 661 notify = in_rtchange, d = NULL; 662 else if (cmd == PRC_MSGSIZE) 663 ; /* special code is present, see below */ 664 else if (cmd == PRC_HOSTDEAD) 665 d = NULL; 666 else if (inet6ctlerrmap[cmd] == 0) 667 return; 668 669 /* if the parameter is from icmp6, decode it. */ 670 if (d != NULL) { 671 struct ip6ctlparam *ip6cp = (struct ip6ctlparam *)d; 672 m = ip6cp->ip6c_m; 673 ip6 = ip6cp->ip6c_ip6; 674 off = ip6cp->ip6c_off; 675 sa6_src = ip6cp->ip6c_src; 676 } else { 677 m = NULL; 678 ip6 = NULL; 679 sa6_src = &sa6_any; 680 } 681 682 if (ip6) { 683 /* 684 * XXX: We assume that when ip6 is non NULL, 685 * M and OFF are valid. 686 */ 687 688 /* check if we can safely examine src and dst ports */ 689 if (m->m_pkthdr.len < off + sizeof(*thp)) 690 return; 691 692 bzero(&th, sizeof(th)); 693 #ifdef DIAGNOSTIC 694 if (sizeof(*thp) > sizeof(th)) 695 panic("assumption failed in tcp6_ctlinput"); 696 #endif 697 m_copydata(m, off, sizeof(*thp), (caddr_t)&th); 698 699 /* 700 * Check to see if we have a valid TCP connection 701 * corresponding to the address in the ICMPv6 message 702 * payload. 703 */ 704 inp = in6_pcbhashlookup(&tcbtable, &sa6->sin6_addr, 705 th.th_dport, (struct in6_addr *)&sa6_src->sin6_addr, 706 th.th_sport, rdomain); 707 if (cmd == PRC_MSGSIZE) { 708 /* 709 * Depending on the value of "valid" and routing table 710 * size (mtudisc_{hi,lo}wat), we will: 711 * - recalcurate the new MTU and create the 712 * corresponding routing entry, or 713 * - ignore the MTU change notification. 714 */ 715 icmp6_mtudisc_update((struct ip6ctlparam *)d, inp != NULL); 716 return; 717 } 718 if (inp) { 719 seq = ntohl(th.th_seq); 720 if (inp->inp_socket && 721 (tp = intotcpcb(inp)) && 722 SEQ_GEQ(seq, tp->snd_una) && 723 SEQ_LT(seq, tp->snd_max)) 724 notify(inp, inet6ctlerrmap[cmd]); 725 } else if (inet6ctlerrmap[cmd] == EHOSTUNREACH || 726 inet6ctlerrmap[cmd] == ENETUNREACH || 727 inet6ctlerrmap[cmd] == EHOSTDOWN) 728 syn_cache_unreach((struct sockaddr *)sa6_src, 729 sa, &th, rdomain); 730 } else { 731 (void) in6_pcbnotify(&tcbtable, sa6, 0, 732 sa6_src, 0, rdomain, cmd, NULL, notify); 733 } 734 } 735 #endif 736 737 void * 738 tcp_ctlinput(int cmd, struct sockaddr *sa, u_int rdomain, void *v) 739 { 740 struct ip *ip = v; 741 struct tcphdr *th; 742 struct tcpcb *tp; 743 struct inpcb *inp; 744 struct in_addr faddr; 745 tcp_seq seq; 746 u_int mtu; 747 void (*notify)(struct inpcb *, int) = tcp_notify; 748 int errno; 749 750 if (sa->sa_family != AF_INET) 751 return NULL; 752 faddr = satosin(sa)->sin_addr; 753 if (faddr.s_addr == INADDR_ANY) 754 return NULL; 755 756 if ((unsigned)cmd >= PRC_NCMDS) 757 return NULL; 758 errno = inetctlerrmap[cmd]; 759 if (cmd == PRC_QUENCH) 760 /* 761 * Don't honor ICMP Source Quench messages meant for 762 * TCP connections. 763 */ 764 return NULL; 765 else if (PRC_IS_REDIRECT(cmd)) 766 notify = in_rtchange, ip = 0; 767 else if (cmd == PRC_MSGSIZE && ip_mtudisc && ip) { 768 /* 769 * Verify that the packet in the icmp payload refers 770 * to an existing TCP connection. 771 */ 772 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 773 seq = ntohl(th->th_seq); 774 inp = in_pcbhashlookup(&tcbtable, 775 ip->ip_dst, th->th_dport, ip->ip_src, th->th_sport, 776 rdomain); 777 if (inp && (tp = intotcpcb(inp)) && 778 SEQ_GEQ(seq, tp->snd_una) && 779 SEQ_LT(seq, tp->snd_max)) { 780 struct icmp *icp; 781 icp = (struct icmp *)((caddr_t)ip - 782 offsetof(struct icmp, icmp_ip)); 783 784 /* 785 * If the ICMP message advertises a Next-Hop MTU 786 * equal or larger than the maximum packet size we have 787 * ever sent, drop the message. 788 */ 789 mtu = (u_int)ntohs(icp->icmp_nextmtu); 790 if (mtu >= tp->t_pmtud_mtu_sent) 791 return NULL; 792 if (mtu >= tcp_hdrsz(tp) + tp->t_pmtud_mss_acked) { 793 /* 794 * Calculate new MTU, and create corresponding 795 * route (traditional PMTUD). 796 */ 797 tp->t_flags &= ~TF_PMTUD_PEND; 798 icmp_mtudisc(icp, inp->inp_rtableid); 799 } else { 800 /* 801 * Record the information got in the ICMP 802 * message; act on it later. 803 * If we had already recorded an ICMP message, 804 * replace the old one only if the new message 805 * refers to an older TCP segment 806 */ 807 if (tp->t_flags & TF_PMTUD_PEND) { 808 if (SEQ_LT(tp->t_pmtud_th_seq, seq)) 809 return NULL; 810 } else 811 tp->t_flags |= TF_PMTUD_PEND; 812 tp->t_pmtud_th_seq = seq; 813 tp->t_pmtud_nextmtu = icp->icmp_nextmtu; 814 tp->t_pmtud_ip_len = icp->icmp_ip.ip_len; 815 tp->t_pmtud_ip_hl = icp->icmp_ip.ip_hl; 816 return NULL; 817 } 818 } else { 819 /* ignore if we don't have a matching connection */ 820 return NULL; 821 } 822 notify = tcp_mtudisc, ip = 0; 823 } else if (cmd == PRC_MTUINC) 824 notify = tcp_mtudisc_increase, ip = 0; 825 else if (cmd == PRC_HOSTDEAD) 826 ip = 0; 827 else if (errno == 0) 828 return NULL; 829 830 if (ip) { 831 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 832 inp = in_pcbhashlookup(&tcbtable, 833 ip->ip_dst, th->th_dport, ip->ip_src, th->th_sport, 834 rdomain); 835 if (inp) { 836 seq = ntohl(th->th_seq); 837 if (inp->inp_socket && 838 (tp = intotcpcb(inp)) && 839 SEQ_GEQ(seq, tp->snd_una) && 840 SEQ_LT(seq, tp->snd_max)) 841 notify(inp, errno); 842 } else if (inetctlerrmap[cmd] == EHOSTUNREACH || 843 inetctlerrmap[cmd] == ENETUNREACH || 844 inetctlerrmap[cmd] == EHOSTDOWN) { 845 struct sockaddr_in sin; 846 847 bzero(&sin, sizeof(sin)); 848 sin.sin_len = sizeof(sin); 849 sin.sin_family = AF_INET; 850 sin.sin_port = th->th_sport; 851 sin.sin_addr = ip->ip_src; 852 syn_cache_unreach(sintosa(&sin), sa, th, rdomain); 853 } 854 } else 855 in_pcbnotifyall(&tcbtable, sa, rdomain, errno, notify); 856 857 return NULL; 858 } 859 860 861 #ifdef INET6 862 /* 863 * Path MTU Discovery handlers. 864 */ 865 void 866 tcp6_mtudisc_callback(sin6, rdomain) 867 struct sockaddr_in6 *sin6; 868 u_int rdomain; 869 { 870 (void) in6_pcbnotify(&tcbtable, sin6, 0, 871 &sa6_any, 0, rdomain, PRC_MSGSIZE, NULL, tcp_mtudisc); 872 } 873 #endif /* INET6 */ 874 875 /* 876 * On receipt of path MTU corrections, flush old route and replace it 877 * with the new one. Retransmit all unacknowledged packets, to ensure 878 * that all packets will be received. 879 */ 880 void 881 tcp_mtudisc(inp, errno) 882 struct inpcb *inp; 883 int errno; 884 { 885 struct tcpcb *tp = intotcpcb(inp); 886 struct rtentry *rt = in_pcbrtentry(inp); 887 int change = 0; 888 889 if (tp != 0) { 890 int orig_maxseg = tp->t_maxseg; 891 if (rt != 0) { 892 /* 893 * If this was not a host route, remove and realloc. 894 */ 895 if ((rt->rt_flags & RTF_HOST) == 0) { 896 in_rtchange(inp, errno); 897 if ((rt = in_pcbrtentry(inp)) == 0) 898 return; 899 } 900 if (orig_maxseg != tp->t_maxseg || 901 (rt->rt_rmx.rmx_locks & RTV_MTU)) 902 change = 1; 903 } 904 tcp_mss(tp, -1); 905 906 /* 907 * Resend unacknowledged packets 908 */ 909 tp->snd_nxt = tp->snd_una; 910 if (change || errno > 0) 911 tcp_output(tp); 912 } 913 } 914 915 void 916 tcp_mtudisc_increase(inp, errno) 917 struct inpcb *inp; 918 int errno; 919 { 920 struct tcpcb *tp = intotcpcb(inp); 921 struct rtentry *rt = in_pcbrtentry(inp); 922 923 if (tp != 0 && rt != 0) { 924 /* 925 * If this was a host route, remove and realloc. 926 */ 927 if (rt->rt_flags & RTF_HOST) 928 in_rtchange(inp, errno); 929 930 /* also takes care of congestion window */ 931 tcp_mss(tp, -1); 932 } 933 } 934 935 /* 936 * Generate new ISNs with a method based on RFC1948 937 */ 938 #define TCP_ISS_CONN_INC 4096 939 int tcp_secret_init; 940 u_char tcp_secret[16]; 941 SHA2_CTX tcp_secret_ctx; 942 943 void 944 tcp_set_iss_tsm(struct tcpcb *tp) 945 { 946 SHA2_CTX ctx; 947 union { 948 uint8_t bytes[SHA512_DIGEST_LENGTH]; 949 uint32_t words[2]; 950 } digest; 951 u_int rdomain = rtable_l2(tp->t_inpcb->inp_rtableid); 952 953 if (tcp_secret_init == 0) { 954 arc4random_buf(tcp_secret, sizeof(tcp_secret)); 955 SHA512Init(&tcp_secret_ctx); 956 SHA512Update(&tcp_secret_ctx, tcp_secret, sizeof(tcp_secret)); 957 tcp_secret_init = 1; 958 } 959 ctx = tcp_secret_ctx; 960 SHA512Update(&ctx, &rdomain, sizeof(rdomain)); 961 SHA512Update(&ctx, &tp->t_inpcb->inp_lport, sizeof(u_short)); 962 SHA512Update(&ctx, &tp->t_inpcb->inp_fport, sizeof(u_short)); 963 if (tp->pf == AF_INET6) { 964 SHA512Update(&ctx, &tp->t_inpcb->inp_laddr6, 965 sizeof(struct in6_addr)); 966 SHA512Update(&ctx, &tp->t_inpcb->inp_faddr6, 967 sizeof(struct in6_addr)); 968 } else { 969 SHA512Update(&ctx, &tp->t_inpcb->inp_laddr, 970 sizeof(struct in_addr)); 971 SHA512Update(&ctx, &tp->t_inpcb->inp_faddr, 972 sizeof(struct in_addr)); 973 } 974 SHA512Final(digest.bytes, &ctx); 975 tcp_iss += TCP_ISS_CONN_INC; 976 tp->iss = digest.words[0] + tcp_iss; 977 tp->ts_modulate = digest.words[1]; 978 } 979 980 #ifdef TCP_SIGNATURE 981 int 982 tcp_signature_tdb_attach(void) 983 { 984 return (0); 985 } 986 987 int 988 tcp_signature_tdb_init(tdbp, xsp, ii) 989 struct tdb *tdbp; 990 struct xformsw *xsp; 991 struct ipsecinit *ii; 992 { 993 if ((ii->ii_authkeylen < 1) || (ii->ii_authkeylen > 80)) 994 return (EINVAL); 995 996 tdbp->tdb_amxkey = malloc(ii->ii_authkeylen, M_XDATA, M_NOWAIT); 997 if (tdbp->tdb_amxkey == NULL) 998 return (ENOMEM); 999 bcopy(ii->ii_authkey, tdbp->tdb_amxkey, ii->ii_authkeylen); 1000 tdbp->tdb_amxkeylen = ii->ii_authkeylen; 1001 1002 return (0); 1003 } 1004 1005 int 1006 tcp_signature_tdb_zeroize(tdbp) 1007 struct tdb *tdbp; 1008 { 1009 if (tdbp->tdb_amxkey) { 1010 explicit_bzero(tdbp->tdb_amxkey, tdbp->tdb_amxkeylen); 1011 free(tdbp->tdb_amxkey, M_XDATA, 0); 1012 tdbp->tdb_amxkey = NULL; 1013 } 1014 1015 return (0); 1016 } 1017 1018 int 1019 tcp_signature_tdb_input(m, tdbp, skip, protoff) 1020 struct mbuf *m; 1021 struct tdb *tdbp; 1022 int skip, protoff; 1023 { 1024 return (0); 1025 } 1026 1027 int 1028 tcp_signature_tdb_output(m, tdbp, mp, skip, protoff) 1029 struct mbuf *m; 1030 struct tdb *tdbp; 1031 struct mbuf **mp; 1032 int skip, protoff; 1033 { 1034 return (EINVAL); 1035 } 1036 1037 int 1038 tcp_signature_apply(fstate, data, len) 1039 caddr_t fstate; 1040 caddr_t data; 1041 unsigned int len; 1042 { 1043 MD5Update((MD5_CTX *)fstate, (char *)data, len); 1044 return 0; 1045 } 1046 1047 int 1048 tcp_signature(struct tdb *tdb, int af, struct mbuf *m, struct tcphdr *th, 1049 int iphlen, int doswap, char *sig) 1050 { 1051 MD5_CTX ctx; 1052 int len; 1053 struct tcphdr th0; 1054 1055 MD5Init(&ctx); 1056 1057 switch(af) { 1058 case 0: 1059 case AF_INET: { 1060 struct ippseudo ippseudo; 1061 struct ip *ip; 1062 1063 ip = mtod(m, struct ip *); 1064 1065 ippseudo.ippseudo_src = ip->ip_src; 1066 ippseudo.ippseudo_dst = ip->ip_dst; 1067 ippseudo.ippseudo_pad = 0; 1068 ippseudo.ippseudo_p = IPPROTO_TCP; 1069 ippseudo.ippseudo_len = htons(m->m_pkthdr.len - iphlen); 1070 1071 MD5Update(&ctx, (char *)&ippseudo, 1072 sizeof(struct ippseudo)); 1073 break; 1074 } 1075 #ifdef INET6 1076 case AF_INET6: { 1077 struct ip6_hdr_pseudo ip6pseudo; 1078 struct ip6_hdr *ip6; 1079 1080 ip6 = mtod(m, struct ip6_hdr *); 1081 bzero(&ip6pseudo, sizeof(ip6pseudo)); 1082 ip6pseudo.ip6ph_src = ip6->ip6_src; 1083 ip6pseudo.ip6ph_dst = ip6->ip6_dst; 1084 in6_clearscope(&ip6pseudo.ip6ph_src); 1085 in6_clearscope(&ip6pseudo.ip6ph_dst); 1086 ip6pseudo.ip6ph_nxt = IPPROTO_TCP; 1087 ip6pseudo.ip6ph_len = htonl(m->m_pkthdr.len - iphlen); 1088 1089 MD5Update(&ctx, (char *)&ip6pseudo, 1090 sizeof(ip6pseudo)); 1091 break; 1092 } 1093 #endif 1094 } 1095 1096 th0 = *th; 1097 th0.th_sum = 0; 1098 1099 if (doswap) { 1100 th0.th_seq = htonl(th0.th_seq); 1101 th0.th_ack = htonl(th0.th_ack); 1102 th0.th_win = htons(th0.th_win); 1103 th0.th_urp = htons(th0.th_urp); 1104 } 1105 MD5Update(&ctx, (char *)&th0, sizeof(th0)); 1106 1107 len = m->m_pkthdr.len - iphlen - th->th_off * sizeof(uint32_t); 1108 1109 if (len > 0 && 1110 m_apply(m, iphlen + th->th_off * sizeof(uint32_t), len, 1111 tcp_signature_apply, (caddr_t)&ctx)) 1112 return (-1); 1113 1114 MD5Update(&ctx, tdb->tdb_amxkey, tdb->tdb_amxkeylen); 1115 MD5Final(sig, &ctx); 1116 1117 return (0); 1118 } 1119 #endif /* TCP_SIGNATURE */ 1120