1 /* $OpenBSD: tcp_subr.c,v 1.173 2018/09/20 18:59:10 bluhm 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 94 #ifdef INET6 95 #include <netinet6/ip6protosw.h> 96 #endif /* INET6 */ 97 98 #include <crypto/md5.h> 99 #include <crypto/sha2.h> 100 101 /* patchable/settable parameters for tcp */ 102 int tcp_mssdflt = TCP_MSS; 103 int tcp_rttdflt = TCPTV_SRTTDFLT / PR_SLOWHZ; 104 105 /* values controllable via sysctl */ 106 int tcp_do_rfc1323 = 1; 107 int tcp_do_sack = 1; /* RFC 2018 selective ACKs */ 108 int tcp_ack_on_push = 0; /* set to enable immediate ACK-on-PUSH */ 109 #ifdef TCP_ECN 110 int tcp_do_ecn = 0; /* RFC3168 ECN enabled/disabled? */ 111 #endif 112 int tcp_do_rfc3390 = 2; /* Increase TCP's Initial Window to 10*mss */ 113 114 u_int32_t tcp_now = 1; 115 116 #ifndef TCB_INITIAL_HASH_SIZE 117 #define TCB_INITIAL_HASH_SIZE 128 118 #endif 119 120 int tcp_reass_limit = NMBCLUSTERS / 8; /* hardlimit for tcpqe_pool */ 121 int tcp_sackhole_limit = 32*1024; /* hardlimit for sackhl_pool */ 122 123 struct pool tcpcb_pool; 124 struct pool tcpqe_pool; 125 struct pool sackhl_pool; 126 127 struct cpumem *tcpcounters; /* tcp statistics */ 128 129 u_char tcp_secret[16]; 130 SHA2_CTX tcp_secret_ctx; 131 tcp_seq tcp_iss; 132 133 /* 134 * Tcp initialization 135 */ 136 void 137 tcp_init(void) 138 { 139 tcp_iss = 1; /* wrong */ 140 pool_init(&tcpcb_pool, sizeof(struct tcpcb), 0, IPL_SOFTNET, 0, 141 "tcpcb", NULL); 142 pool_init(&tcpqe_pool, sizeof(struct tcpqent), 0, IPL_SOFTNET, 0, 143 "tcpqe", NULL); 144 pool_sethardlimit(&tcpqe_pool, tcp_reass_limit, NULL, 0); 145 pool_init(&sackhl_pool, sizeof(struct sackhole), 0, IPL_SOFTNET, 0, 146 "sackhl", NULL); 147 pool_sethardlimit(&sackhl_pool, tcp_sackhole_limit, NULL, 0); 148 in_pcbinit(&tcbtable, TCB_INITIAL_HASH_SIZE); 149 tcpcounters = counters_alloc(tcps_ncounters); 150 151 arc4random_buf(tcp_secret, sizeof(tcp_secret)); 152 SHA512Init(&tcp_secret_ctx); 153 SHA512Update(&tcp_secret_ctx, tcp_secret, sizeof(tcp_secret)); 154 155 #ifdef INET6 156 /* 157 * Since sizeof(struct ip6_hdr) > sizeof(struct ip), we 158 * do max length checks/computations only on the former. 159 */ 160 if (max_protohdr < (sizeof(struct ip6_hdr) + sizeof(struct tcphdr))) 161 max_protohdr = (sizeof(struct ip6_hdr) + sizeof(struct tcphdr)); 162 if ((max_linkhdr + sizeof(struct ip6_hdr) + sizeof(struct tcphdr)) > 163 MHLEN) 164 panic("tcp_init"); 165 166 icmp6_mtudisc_callback_register(tcp6_mtudisc_callback); 167 #endif /* INET6 */ 168 169 /* Initialize the compressed state engine. */ 170 syn_cache_init(); 171 172 /* Initialize timer state. */ 173 tcp_timer_init(); 174 } 175 176 /* 177 * Create template to be used to send tcp packets on a connection. 178 * Call after host entry created, allocates an mbuf and fills 179 * in a skeletal tcp/ip header, minimizing the amount of work 180 * necessary when the connection is used. 181 * 182 * To support IPv6 in addition to IPv4 and considering that the sizes of 183 * the IPv4 and IPv6 headers are not the same, we now use a separate pointer 184 * for the TCP header. Also, we made the former tcpiphdr header pointer 185 * into just an IP overlay pointer, with casting as appropriate for v6. rja 186 */ 187 struct mbuf * 188 tcp_template(struct tcpcb *tp) 189 { 190 struct inpcb *inp = tp->t_inpcb; 191 struct mbuf *m; 192 struct tcphdr *th; 193 194 CTASSERT(sizeof(struct ip) + sizeof(struct tcphdr) <= MHLEN); 195 CTASSERT(sizeof(struct ip6_hdr) + sizeof(struct tcphdr) <= MHLEN); 196 197 if ((m = tp->t_template) == 0) { 198 m = m_get(M_DONTWAIT, MT_HEADER); 199 if (m == NULL) 200 return (0); 201 202 switch (tp->pf) { 203 case 0: /*default to PF_INET*/ 204 case AF_INET: 205 m->m_len = sizeof(struct ip); 206 break; 207 #ifdef INET6 208 case AF_INET6: 209 m->m_len = sizeof(struct ip6_hdr); 210 break; 211 #endif /* INET6 */ 212 } 213 m->m_len += sizeof (struct tcphdr); 214 } 215 216 switch(tp->pf) { 217 case AF_INET: 218 { 219 struct ipovly *ipovly; 220 221 ipovly = mtod(m, struct ipovly *); 222 223 bzero(ipovly->ih_x1, sizeof ipovly->ih_x1); 224 ipovly->ih_pr = IPPROTO_TCP; 225 ipovly->ih_len = htons(sizeof (struct tcphdr)); 226 ipovly->ih_src = inp->inp_laddr; 227 ipovly->ih_dst = inp->inp_faddr; 228 229 th = (struct tcphdr *)(mtod(m, caddr_t) + 230 sizeof(struct ip)); 231 } 232 break; 233 #ifdef INET6 234 case AF_INET6: 235 { 236 struct ip6_hdr *ip6; 237 238 ip6 = mtod(m, struct ip6_hdr *); 239 240 ip6->ip6_src = inp->inp_laddr6; 241 ip6->ip6_dst = inp->inp_faddr6; 242 ip6->ip6_flow = htonl(0x60000000) | 243 (inp->inp_flowinfo & IPV6_FLOWLABEL_MASK); 244 245 ip6->ip6_nxt = IPPROTO_TCP; 246 ip6->ip6_plen = htons(sizeof(struct tcphdr)); /*XXX*/ 247 ip6->ip6_hlim = in6_selecthlim(inp); /*XXX*/ 248 249 th = (struct tcphdr *)(mtod(m, caddr_t) + 250 sizeof(struct ip6_hdr)); 251 } 252 break; 253 #endif /* INET6 */ 254 } 255 256 th->th_sport = inp->inp_lport; 257 th->th_dport = inp->inp_fport; 258 th->th_seq = 0; 259 th->th_ack = 0; 260 th->th_x2 = 0; 261 th->th_off = 5; 262 th->th_flags = 0; 263 th->th_win = 0; 264 th->th_urp = 0; 265 th->th_sum = 0; 266 return (m); 267 } 268 269 /* 270 * Send a single message to the TCP at address specified by 271 * the given TCP/IP header. If m == 0, then we make a copy 272 * of the tcpiphdr at ti and send directly to the addressed host. 273 * This is used to force keep alive messages out using the TCP 274 * template for a connection tp->t_template. If flags are given 275 * then we send a message back to the TCP which originated the 276 * segment ti, and discard the mbuf containing it and any other 277 * attached mbufs. 278 * 279 * In any case the ack and sequence number of the transmitted 280 * segment are as specified by the parameters. 281 */ 282 void 283 tcp_respond(struct tcpcb *tp, caddr_t template, struct tcphdr *th0, 284 tcp_seq ack, tcp_seq seq, int flags, u_int rtableid) 285 { 286 int tlen; 287 int win = 0; 288 struct mbuf *m = NULL; 289 struct tcphdr *th; 290 struct ip *ip; 291 #ifdef INET6 292 struct ip6_hdr *ip6; 293 #endif 294 int af; /* af on wire */ 295 296 if (tp) { 297 struct socket *so = tp->t_inpcb->inp_socket; 298 win = sbspace(so, &so->so_rcv); 299 /* 300 * If this is called with an unconnected 301 * socket/tp/pcb (tp->pf is 0), we lose. 302 */ 303 af = tp->pf; 304 } else 305 af = (((struct ip *)template)->ip_v == 6) ? AF_INET6 : AF_INET; 306 307 m = m_gethdr(M_DONTWAIT, MT_HEADER); 308 if (m == NULL) 309 return; 310 m->m_data += max_linkhdr; 311 tlen = 0; 312 313 #define xchg(a,b,type) do { type t; t=a; a=b; b=t; } while (0) 314 switch (af) { 315 #ifdef INET6 316 case AF_INET6: 317 ip6 = mtod(m, struct ip6_hdr *); 318 th = (struct tcphdr *)(ip6 + 1); 319 tlen = sizeof(*ip6) + sizeof(*th); 320 if (th0) { 321 bcopy(template, ip6, sizeof(*ip6)); 322 bcopy(th0, th, sizeof(*th)); 323 xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr); 324 } else { 325 bcopy(template, ip6, tlen); 326 } 327 break; 328 #endif /* INET6 */ 329 case AF_INET: 330 ip = mtod(m, struct ip *); 331 th = (struct tcphdr *)(ip + 1); 332 tlen = sizeof(*ip) + sizeof(*th); 333 if (th0) { 334 bcopy(template, ip, sizeof(*ip)); 335 bcopy(th0, th, sizeof(*th)); 336 xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, u_int32_t); 337 } else { 338 bcopy(template, ip, tlen); 339 } 340 break; 341 } 342 if (th0) 343 xchg(th->th_dport, th->th_sport, u_int16_t); 344 else 345 flags = TH_ACK; 346 #undef xchg 347 348 th->th_seq = htonl(seq); 349 th->th_ack = htonl(ack); 350 th->th_x2 = 0; 351 th->th_off = sizeof (struct tcphdr) >> 2; 352 th->th_flags = flags; 353 if (tp) 354 win >>= tp->rcv_scale; 355 if (win > TCP_MAXWIN) 356 win = TCP_MAXWIN; 357 th->th_win = htons((u_int16_t)win); 358 th->th_urp = 0; 359 360 if (tp && (tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP && 361 (flags & TH_RST) == 0 && (tp->t_flags & TF_RCVD_TSTMP)) { 362 u_int32_t *lp = (u_int32_t *)(th + 1); 363 /* Form timestamp option as shown in appendix A of RFC 1323. */ 364 *lp++ = htonl(TCPOPT_TSTAMP_HDR); 365 *lp++ = htonl(tcp_now + tp->ts_modulate); 366 *lp = htonl(tp->ts_recent); 367 tlen += TCPOLEN_TSTAMP_APPA; 368 th->th_off = (sizeof(struct tcphdr) + TCPOLEN_TSTAMP_APPA) >> 2; 369 } 370 371 m->m_len = tlen; 372 m->m_pkthdr.len = tlen; 373 m->m_pkthdr.ph_ifidx = 0; 374 m->m_pkthdr.csum_flags |= M_TCP_CSUM_OUT; 375 376 /* force routing table */ 377 if (tp) 378 m->m_pkthdr.ph_rtableid = tp->t_inpcb->inp_rtableid; 379 else 380 m->m_pkthdr.ph_rtableid = rtableid; 381 382 switch (af) { 383 #ifdef INET6 384 case AF_INET6: 385 ip6->ip6_flow = htonl(0x60000000); 386 ip6->ip6_nxt = IPPROTO_TCP; 387 ip6->ip6_hlim = in6_selecthlim(tp ? tp->t_inpcb : NULL); /*XXX*/ 388 ip6->ip6_plen = tlen - sizeof(struct ip6_hdr); 389 ip6->ip6_plen = htons(ip6->ip6_plen); 390 ip6_output(m, tp ? tp->t_inpcb->inp_outputopts6 : NULL, 391 tp ? &tp->t_inpcb->inp_route6 : NULL, 392 0, NULL, 393 tp ? tp->t_inpcb : NULL); 394 break; 395 #endif /* INET6 */ 396 case AF_INET: 397 ip->ip_len = htons(tlen); 398 ip->ip_ttl = ip_defttl; 399 ip->ip_tos = 0; 400 ip_output(m, NULL, 401 tp ? &tp->t_inpcb->inp_route : NULL, 402 ip_mtudisc ? IP_MTUDISC : 0, NULL, 403 tp ? tp->t_inpcb : NULL, 0); 404 break; 405 } 406 } 407 408 /* 409 * Create a new TCP control block, making an 410 * empty reassembly queue and hooking it to the argument 411 * protocol control block. 412 */ 413 struct tcpcb * 414 tcp_newtcpcb(struct inpcb *inp) 415 { 416 struct tcpcb *tp; 417 int i; 418 419 tp = pool_get(&tcpcb_pool, PR_NOWAIT|PR_ZERO); 420 if (tp == NULL) 421 return (NULL); 422 TAILQ_INIT(&tp->t_segq); 423 tp->t_maxseg = tcp_mssdflt; 424 tp->t_maxopd = 0; 425 426 for (i = 0; i < TCPT_NTIMERS; i++) 427 TCP_TIMER_INIT(tp, i); 428 429 tp->sack_enable = tcp_do_sack; 430 tp->t_flags = tcp_do_rfc1323 ? (TF_REQ_SCALE|TF_REQ_TSTMP) : 0; 431 tp->t_inpcb = inp; 432 /* 433 * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no 434 * rtt estimate. Set rttvar so that srtt + 2 * rttvar gives 435 * reasonable initial retransmit time. 436 */ 437 tp->t_srtt = TCPTV_SRTTBASE; 438 tp->t_rttvar = tcp_rttdflt * PR_SLOWHZ << 439 (TCP_RTTVAR_SHIFT + TCP_RTT_BASE_SHIFT - 1); 440 tp->t_rttmin = TCPTV_MIN; 441 TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp), 442 TCPTV_MIN, TCPTV_REXMTMAX); 443 tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; 444 tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT; 445 446 tp->t_pmtud_mtu_sent = 0; 447 tp->t_pmtud_mss_acked = 0; 448 449 #ifdef INET6 450 /* we disallow IPv4 mapped address completely. */ 451 if ((inp->inp_flags & INP_IPV6) == 0) 452 tp->pf = PF_INET; 453 else 454 tp->pf = PF_INET6; 455 #else 456 tp->pf = PF_INET; 457 #endif 458 459 #ifdef INET6 460 if (inp->inp_flags & INP_IPV6) 461 inp->inp_ipv6.ip6_hlim = ip6_defhlim; 462 else 463 #endif /* INET6 */ 464 inp->inp_ip.ip_ttl = ip_defttl; 465 466 inp->inp_ppcb = (caddr_t)tp; 467 return (tp); 468 } 469 470 /* 471 * Drop a TCP connection, reporting 472 * the specified error. If connection is synchronized, 473 * then send a RST to peer. 474 */ 475 struct tcpcb * 476 tcp_drop(struct tcpcb *tp, int errno) 477 { 478 struct socket *so = tp->t_inpcb->inp_socket; 479 480 if (TCPS_HAVERCVDSYN(tp->t_state)) { 481 tp->t_state = TCPS_CLOSED; 482 (void) tcp_output(tp); 483 tcpstat_inc(tcps_drops); 484 } else 485 tcpstat_inc(tcps_conndrops); 486 if (errno == ETIMEDOUT && tp->t_softerror) 487 errno = tp->t_softerror; 488 so->so_error = errno; 489 return (tcp_close(tp)); 490 } 491 492 /* 493 * Close a TCP control block: 494 * discard all space held by the tcp 495 * discard internet protocol block 496 * wake up any sleepers 497 */ 498 struct tcpcb * 499 tcp_close(struct tcpcb *tp) 500 { 501 struct inpcb *inp = tp->t_inpcb; 502 struct socket *so = inp->inp_socket; 503 struct sackhole *p, *q; 504 505 /* free the reassembly queue, if any */ 506 tcp_freeq(tp); 507 508 tcp_canceltimers(tp); 509 syn_cache_cleanup(tp); 510 511 /* Free SACK holes. */ 512 q = p = tp->snd_holes; 513 while (p != 0) { 514 q = p->next; 515 pool_put(&sackhl_pool, p); 516 p = q; 517 } 518 519 m_free(tp->t_template); 520 /* Free tcpcb after all pending timers have been run. */ 521 TCP_TIMER_ARM(tp, TCPT_REAPER, 0); 522 523 mtx_enter(&inp->inp_mtx); 524 inp->inp_ppcb = NULL; 525 mtx_leave(&inp->inp_mtx); 526 soisdisconnected(so); 527 in_pcbdetach(inp); 528 return (NULL); 529 } 530 531 int 532 tcp_freeq(struct tcpcb *tp) 533 { 534 struct tcpqent *qe; 535 int rv = 0; 536 537 while ((qe = TAILQ_FIRST(&tp->t_segq)) != NULL) { 538 TAILQ_REMOVE(&tp->t_segq, qe, tcpqe_q); 539 m_freem(qe->tcpqe_m); 540 pool_put(&tcpqe_pool, qe); 541 rv = 1; 542 } 543 return (rv); 544 } 545 546 /* 547 * Compute proper scaling value for receiver window from buffer space 548 */ 549 550 void 551 tcp_rscale(struct tcpcb *tp, u_long hiwat) 552 { 553 tp->request_r_scale = 0; 554 while (tp->request_r_scale < TCP_MAX_WINSHIFT && 555 TCP_MAXWIN << tp->request_r_scale < hiwat) 556 tp->request_r_scale++; 557 } 558 559 /* 560 * Notify a tcp user of an asynchronous error; 561 * store error as soft error, but wake up user 562 * (for now, won't do anything until can select for soft error). 563 */ 564 void 565 tcp_notify(struct inpcb *inp, int error) 566 { 567 struct tcpcb *tp = intotcpcb(inp); 568 struct socket *so = inp->inp_socket; 569 570 /* 571 * Ignore some errors if we are hooked up. 572 * If connection hasn't completed, has retransmitted several times, 573 * and receives a second error, give up now. This is better 574 * than waiting a long time to establish a connection that 575 * can never complete. 576 */ 577 if (tp->t_state == TCPS_ESTABLISHED && 578 (error == EHOSTUNREACH || error == ENETUNREACH || 579 error == EHOSTDOWN)) { 580 return; 581 } else if (TCPS_HAVEESTABLISHED(tp->t_state) == 0 && 582 tp->t_rxtshift > 3 && tp->t_softerror) 583 so->so_error = error; 584 else 585 tp->t_softerror = error; 586 wakeup((caddr_t) &so->so_timeo); 587 sorwakeup(so); 588 sowwakeup(so); 589 } 590 591 #ifdef INET6 592 void 593 tcp6_ctlinput(int cmd, struct sockaddr *sa, u_int rdomain, void *d) 594 { 595 struct tcphdr th; 596 struct tcpcb *tp; 597 void (*notify)(struct inpcb *, int) = tcp_notify; 598 struct ip6_hdr *ip6; 599 const struct sockaddr_in6 *sa6_src = NULL; 600 struct sockaddr_in6 *sa6 = satosin6(sa); 601 struct inpcb *inp; 602 struct mbuf *m; 603 tcp_seq seq; 604 int off; 605 struct { 606 u_int16_t th_sport; 607 u_int16_t th_dport; 608 u_int32_t th_seq; 609 } *thp; 610 611 CTASSERT(sizeof(*thp) <= sizeof(th)); 612 if (sa->sa_family != AF_INET6 || 613 sa->sa_len != sizeof(struct sockaddr_in6) || 614 IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || 615 IN6_IS_ADDR_V4MAPPED(&sa6->sin6_addr)) 616 return; 617 if ((unsigned)cmd >= PRC_NCMDS) 618 return; 619 else if (cmd == PRC_QUENCH) { 620 /* 621 * Don't honor ICMP Source Quench messages meant for 622 * TCP connections. 623 */ 624 /* XXX there's no PRC_QUENCH in IPv6 */ 625 return; 626 } else if (PRC_IS_REDIRECT(cmd)) 627 notify = in_rtchange, d = NULL; 628 else if (cmd == PRC_MSGSIZE) 629 ; /* special code is present, see below */ 630 else if (cmd == PRC_HOSTDEAD) 631 d = NULL; 632 else if (inet6ctlerrmap[cmd] == 0) 633 return; 634 635 /* if the parameter is from icmp6, decode it. */ 636 if (d != NULL) { 637 struct ip6ctlparam *ip6cp = (struct ip6ctlparam *)d; 638 m = ip6cp->ip6c_m; 639 ip6 = ip6cp->ip6c_ip6; 640 off = ip6cp->ip6c_off; 641 sa6_src = ip6cp->ip6c_src; 642 } else { 643 m = NULL; 644 ip6 = NULL; 645 sa6_src = &sa6_any; 646 } 647 648 if (ip6) { 649 /* 650 * XXX: We assume that when ip6 is non NULL, 651 * M and OFF are valid. 652 */ 653 654 /* check if we can safely examine src and dst ports */ 655 if (m->m_pkthdr.len < off + sizeof(*thp)) 656 return; 657 658 bzero(&th, sizeof(th)); 659 m_copydata(m, off, sizeof(*thp), (caddr_t)&th); 660 661 /* 662 * Check to see if we have a valid TCP connection 663 * corresponding to the address in the ICMPv6 message 664 * payload. 665 */ 666 inp = in6_pcbhashlookup(&tcbtable, &sa6->sin6_addr, 667 th.th_dport, &sa6_src->sin6_addr, th.th_sport, rdomain); 668 if (cmd == PRC_MSGSIZE) { 669 /* 670 * Depending on the value of "valid" and routing table 671 * size (mtudisc_{hi,lo}wat), we will: 672 * - recalcurate the new MTU and create the 673 * corresponding routing entry, or 674 * - ignore the MTU change notification. 675 */ 676 icmp6_mtudisc_update((struct ip6ctlparam *)d, inp != NULL); 677 return; 678 } 679 if (inp) { 680 seq = ntohl(th.th_seq); 681 if (inp->inp_socket && 682 (tp = intotcpcb(inp)) && 683 SEQ_GEQ(seq, tp->snd_una) && 684 SEQ_LT(seq, tp->snd_max)) 685 notify(inp, inet6ctlerrmap[cmd]); 686 } else if (inet6ctlerrmap[cmd] == EHOSTUNREACH || 687 inet6ctlerrmap[cmd] == ENETUNREACH || 688 inet6ctlerrmap[cmd] == EHOSTDOWN) 689 syn_cache_unreach((struct sockaddr *)sa6_src, 690 sa, &th, rdomain); 691 } else { 692 (void) in6_pcbnotify(&tcbtable, sa6, 0, 693 sa6_src, 0, rdomain, cmd, NULL, notify); 694 } 695 } 696 #endif 697 698 void 699 tcp_ctlinput(int cmd, struct sockaddr *sa, u_int rdomain, void *v) 700 { 701 struct ip *ip = v; 702 struct tcphdr *th; 703 struct tcpcb *tp; 704 struct inpcb *inp; 705 struct in_addr faddr; 706 tcp_seq seq; 707 u_int mtu; 708 void (*notify)(struct inpcb *, int) = tcp_notify; 709 int errno; 710 711 if (sa->sa_family != AF_INET) 712 return; 713 faddr = satosin(sa)->sin_addr; 714 if (faddr.s_addr == INADDR_ANY) 715 return; 716 717 if ((unsigned)cmd >= PRC_NCMDS) 718 return; 719 errno = inetctlerrmap[cmd]; 720 if (cmd == PRC_QUENCH) 721 /* 722 * Don't honor ICMP Source Quench messages meant for 723 * TCP connections. 724 */ 725 return; 726 else if (PRC_IS_REDIRECT(cmd)) 727 notify = in_rtchange, ip = 0; 728 else if (cmd == PRC_MSGSIZE && ip_mtudisc && ip) { 729 /* 730 * Verify that the packet in the icmp payload refers 731 * to an existing TCP connection. 732 */ 733 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 734 seq = ntohl(th->th_seq); 735 inp = in_pcbhashlookup(&tcbtable, 736 ip->ip_dst, th->th_dport, ip->ip_src, th->th_sport, 737 rdomain); 738 if (inp && (tp = intotcpcb(inp)) && 739 SEQ_GEQ(seq, tp->snd_una) && 740 SEQ_LT(seq, tp->snd_max)) { 741 struct icmp *icp; 742 icp = (struct icmp *)((caddr_t)ip - 743 offsetof(struct icmp, icmp_ip)); 744 745 /* 746 * If the ICMP message advertises a Next-Hop MTU 747 * equal or larger than the maximum packet size we have 748 * ever sent, drop the message. 749 */ 750 mtu = (u_int)ntohs(icp->icmp_nextmtu); 751 if (mtu >= tp->t_pmtud_mtu_sent) 752 return; 753 if (mtu >= tcp_hdrsz(tp) + tp->t_pmtud_mss_acked) { 754 /* 755 * Calculate new MTU, and create corresponding 756 * route (traditional PMTUD). 757 */ 758 tp->t_flags &= ~TF_PMTUD_PEND; 759 icmp_mtudisc(icp, inp->inp_rtableid); 760 } else { 761 /* 762 * Record the information got in the ICMP 763 * message; act on it later. 764 * If we had already recorded an ICMP message, 765 * replace the old one only if the new message 766 * refers to an older TCP segment 767 */ 768 if (tp->t_flags & TF_PMTUD_PEND) { 769 if (SEQ_LT(tp->t_pmtud_th_seq, seq)) 770 return; 771 } else 772 tp->t_flags |= TF_PMTUD_PEND; 773 tp->t_pmtud_th_seq = seq; 774 tp->t_pmtud_nextmtu = icp->icmp_nextmtu; 775 tp->t_pmtud_ip_len = icp->icmp_ip.ip_len; 776 tp->t_pmtud_ip_hl = icp->icmp_ip.ip_hl; 777 return; 778 } 779 } else { 780 /* ignore if we don't have a matching connection */ 781 return; 782 } 783 notify = tcp_mtudisc, ip = 0; 784 } else if (cmd == PRC_MTUINC) 785 notify = tcp_mtudisc_increase, ip = 0; 786 else if (cmd == PRC_HOSTDEAD) 787 ip = 0; 788 else if (errno == 0) 789 return; 790 791 if (ip) { 792 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 793 inp = in_pcbhashlookup(&tcbtable, 794 ip->ip_dst, th->th_dport, ip->ip_src, th->th_sport, 795 rdomain); 796 if (inp) { 797 seq = ntohl(th->th_seq); 798 if (inp->inp_socket && 799 (tp = intotcpcb(inp)) && 800 SEQ_GEQ(seq, tp->snd_una) && 801 SEQ_LT(seq, tp->snd_max)) 802 notify(inp, errno); 803 } else if (inetctlerrmap[cmd] == EHOSTUNREACH || 804 inetctlerrmap[cmd] == ENETUNREACH || 805 inetctlerrmap[cmd] == EHOSTDOWN) { 806 struct sockaddr_in sin; 807 808 bzero(&sin, sizeof(sin)); 809 sin.sin_len = sizeof(sin); 810 sin.sin_family = AF_INET; 811 sin.sin_port = th->th_sport; 812 sin.sin_addr = ip->ip_src; 813 syn_cache_unreach(sintosa(&sin), sa, th, rdomain); 814 } 815 } else 816 in_pcbnotifyall(&tcbtable, sa, rdomain, errno, notify); 817 } 818 819 820 #ifdef INET6 821 /* 822 * Path MTU Discovery handlers. 823 */ 824 void 825 tcp6_mtudisc_callback(struct sockaddr_in6 *sin6, u_int rdomain) 826 { 827 (void) in6_pcbnotify(&tcbtable, sin6, 0, 828 &sa6_any, 0, rdomain, PRC_MSGSIZE, NULL, tcp_mtudisc); 829 } 830 #endif /* INET6 */ 831 832 /* 833 * On receipt of path MTU corrections, flush old route and replace it 834 * with the new one. Retransmit all unacknowledged packets, to ensure 835 * that all packets will be received. 836 */ 837 void 838 tcp_mtudisc(struct inpcb *inp, int errno) 839 { 840 struct tcpcb *tp = intotcpcb(inp); 841 struct rtentry *rt; 842 int change = 0; 843 844 if (tp == NULL) 845 return; 846 847 rt = in_pcbrtentry(inp); 848 if (rt != NULL) { 849 int orig_maxseg = tp->t_maxseg; 850 851 /* 852 * If this was not a host route, remove and realloc. 853 */ 854 if ((rt->rt_flags & RTF_HOST) == 0) { 855 in_rtchange(inp, errno); 856 if ((rt = in_pcbrtentry(inp)) == NULL) 857 return; 858 } 859 if (orig_maxseg != tp->t_maxseg || 860 (rt->rt_locks & RTV_MTU)) 861 change = 1; 862 } 863 tcp_mss(tp, -1); 864 865 /* 866 * Resend unacknowledged packets 867 */ 868 tp->snd_nxt = tp->snd_una; 869 if (change || errno > 0) 870 tcp_output(tp); 871 } 872 873 void 874 tcp_mtudisc_increase(struct inpcb *inp, int errno) 875 { 876 struct tcpcb *tp = intotcpcb(inp); 877 struct rtentry *rt = in_pcbrtentry(inp); 878 879 if (tp != 0 && rt != 0) { 880 /* 881 * If this was a host route, remove and realloc. 882 */ 883 if (rt->rt_flags & RTF_HOST) 884 in_rtchange(inp, errno); 885 886 /* also takes care of congestion window */ 887 tcp_mss(tp, -1); 888 } 889 } 890 891 /* 892 * Generate new ISNs with a method based on RFC1948 893 */ 894 #define TCP_ISS_CONN_INC 4096 895 896 void 897 tcp_set_iss_tsm(struct tcpcb *tp) 898 { 899 SHA2_CTX ctx; 900 union { 901 uint8_t bytes[SHA512_DIGEST_LENGTH]; 902 uint32_t words[2]; 903 } digest; 904 u_int rdomain = rtable_l2(tp->t_inpcb->inp_rtableid); 905 906 ctx = tcp_secret_ctx; 907 SHA512Update(&ctx, &rdomain, sizeof(rdomain)); 908 SHA512Update(&ctx, &tp->t_inpcb->inp_lport, sizeof(u_short)); 909 SHA512Update(&ctx, &tp->t_inpcb->inp_fport, sizeof(u_short)); 910 if (tp->pf == AF_INET6) { 911 SHA512Update(&ctx, &tp->t_inpcb->inp_laddr6, 912 sizeof(struct in6_addr)); 913 SHA512Update(&ctx, &tp->t_inpcb->inp_faddr6, 914 sizeof(struct in6_addr)); 915 } else { 916 SHA512Update(&ctx, &tp->t_inpcb->inp_laddr, 917 sizeof(struct in_addr)); 918 SHA512Update(&ctx, &tp->t_inpcb->inp_faddr, 919 sizeof(struct in_addr)); 920 } 921 SHA512Final(digest.bytes, &ctx); 922 tcp_iss += TCP_ISS_CONN_INC; 923 tp->iss = digest.words[0] + tcp_iss; 924 tp->ts_modulate = digest.words[1]; 925 } 926 927 #ifdef TCP_SIGNATURE 928 int 929 tcp_signature_tdb_attach(void) 930 { 931 return (0); 932 } 933 934 int 935 tcp_signature_tdb_init(struct tdb *tdbp, struct xformsw *xsp, 936 struct ipsecinit *ii) 937 { 938 if ((ii->ii_authkeylen < 1) || (ii->ii_authkeylen > 80)) 939 return (EINVAL); 940 941 tdbp->tdb_amxkey = malloc(ii->ii_authkeylen, M_XDATA, M_NOWAIT); 942 if (tdbp->tdb_amxkey == NULL) 943 return (ENOMEM); 944 memcpy(tdbp->tdb_amxkey, ii->ii_authkey, ii->ii_authkeylen); 945 tdbp->tdb_amxkeylen = ii->ii_authkeylen; 946 947 return (0); 948 } 949 950 int 951 tcp_signature_tdb_zeroize(struct tdb *tdbp) 952 { 953 if (tdbp->tdb_amxkey) { 954 explicit_bzero(tdbp->tdb_amxkey, tdbp->tdb_amxkeylen); 955 free(tdbp->tdb_amxkey, M_XDATA, tdbp->tdb_amxkeylen); 956 tdbp->tdb_amxkey = NULL; 957 } 958 959 return (0); 960 } 961 962 int 963 tcp_signature_tdb_input(struct mbuf *m, struct tdb *tdbp, int skip, int protoff) 964 { 965 return (0); 966 } 967 968 int 969 tcp_signature_tdb_output(struct mbuf *m, struct tdb *tdbp, struct mbuf **mp, 970 int skip, int protoff) 971 { 972 return (EINVAL); 973 } 974 975 int 976 tcp_signature_apply(caddr_t fstate, caddr_t data, unsigned int len) 977 { 978 MD5Update((MD5_CTX *)fstate, (char *)data, len); 979 return 0; 980 } 981 982 int 983 tcp_signature(struct tdb *tdb, int af, struct mbuf *m, struct tcphdr *th, 984 int iphlen, int doswap, char *sig) 985 { 986 MD5_CTX ctx; 987 int len; 988 struct tcphdr th0; 989 990 MD5Init(&ctx); 991 992 switch(af) { 993 case 0: 994 case AF_INET: { 995 struct ippseudo ippseudo; 996 struct ip *ip; 997 998 ip = mtod(m, struct ip *); 999 1000 ippseudo.ippseudo_src = ip->ip_src; 1001 ippseudo.ippseudo_dst = ip->ip_dst; 1002 ippseudo.ippseudo_pad = 0; 1003 ippseudo.ippseudo_p = IPPROTO_TCP; 1004 ippseudo.ippseudo_len = htons(m->m_pkthdr.len - iphlen); 1005 1006 MD5Update(&ctx, (char *)&ippseudo, 1007 sizeof(struct ippseudo)); 1008 break; 1009 } 1010 #ifdef INET6 1011 case AF_INET6: { 1012 struct ip6_hdr_pseudo ip6pseudo; 1013 struct ip6_hdr *ip6; 1014 1015 ip6 = mtod(m, struct ip6_hdr *); 1016 bzero(&ip6pseudo, sizeof(ip6pseudo)); 1017 ip6pseudo.ip6ph_src = ip6->ip6_src; 1018 ip6pseudo.ip6ph_dst = ip6->ip6_dst; 1019 in6_clearscope(&ip6pseudo.ip6ph_src); 1020 in6_clearscope(&ip6pseudo.ip6ph_dst); 1021 ip6pseudo.ip6ph_nxt = IPPROTO_TCP; 1022 ip6pseudo.ip6ph_len = htonl(m->m_pkthdr.len - iphlen); 1023 1024 MD5Update(&ctx, (char *)&ip6pseudo, 1025 sizeof(ip6pseudo)); 1026 break; 1027 } 1028 #endif 1029 } 1030 1031 th0 = *th; 1032 th0.th_sum = 0; 1033 1034 if (doswap) { 1035 th0.th_seq = htonl(th0.th_seq); 1036 th0.th_ack = htonl(th0.th_ack); 1037 th0.th_win = htons(th0.th_win); 1038 th0.th_urp = htons(th0.th_urp); 1039 } 1040 MD5Update(&ctx, (char *)&th0, sizeof(th0)); 1041 1042 len = m->m_pkthdr.len - iphlen - th->th_off * sizeof(uint32_t); 1043 1044 if (len > 0 && 1045 m_apply(m, iphlen + th->th_off * sizeof(uint32_t), len, 1046 tcp_signature_apply, (caddr_t)&ctx)) 1047 return (-1); 1048 1049 MD5Update(&ctx, tdb->tdb_amxkey, tdb->tdb_amxkeylen); 1050 MD5Final(sig, &ctx); 1051 1052 return (0); 1053 } 1054 #endif /* TCP_SIGNATURE */ 1055