tcp_timewait.c revision 70103
1/* 2 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)tcp_subr.c 8.2 (Berkeley) 5/24/95 34 * $FreeBSD: head/sys/netinet/tcp_timewait.c 70103 2000-12-16 19:42:06Z phk $ 35 */ 36 37#include "opt_compat.h" 38#include "opt_inet6.h" 39#include "opt_ipsec.h" 40#include "opt_tcpdebug.h" 41 42#include <sys/param.h> 43#include <sys/systm.h> 44#include <sys/callout.h> 45#include <sys/kernel.h> 46#include <sys/sysctl.h> 47#include <sys/malloc.h> 48#include <sys/mbuf.h> 49#ifdef INET6 50#include <sys/domain.h> 51#endif 52#include <sys/proc.h> 53#include <sys/socket.h> 54#include <sys/socketvar.h> 55#include <sys/protosw.h> 56 57#include <vm/vm_zone.h> 58 59#include <net/route.h> 60#include <net/if.h> 61 62#define _IP_VHL 63#include <netinet/in.h> 64#include <netinet/in_systm.h> 65#include <netinet/ip.h> 66#ifdef INET6 67#include <netinet/ip6.h> 68#endif 69#include <netinet/in_pcb.h> 70#ifdef INET6 71#include <netinet6/in6_pcb.h> 72#endif 73#include <netinet/in_var.h> 74#include <netinet/ip_var.h> 75#ifdef INET6 76#include <netinet6/ip6_var.h> 77#endif 78#include <netinet/tcp.h> 79#include <netinet/tcp_fsm.h> 80#include <netinet/tcp_seq.h> 81#include <netinet/tcp_timer.h> 82#include <netinet/tcp_var.h> 83#ifdef INET6 84#include <netinet6/tcp6_var.h> 85#endif 86#include <netinet/tcpip.h> 87#ifdef TCPDEBUG 88#include <netinet/tcp_debug.h> 89#endif 90#include <netinet6/ip6protosw.h> 91 92#ifdef IPSEC 93#include <netinet6/ipsec.h> 94#ifdef INET6 95#include <netinet6/ipsec6.h> 96#endif 97#endif /*IPSEC*/ 98 99#include <machine/in_cksum.h> 100 101int tcp_mssdflt = TCP_MSS; 102SYSCTL_INT(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt, CTLFLAG_RW, 103 &tcp_mssdflt , 0, "Default TCP Maximum Segment Size"); 104 105#ifdef INET6 106int tcp_v6mssdflt = TCP6_MSS; 107SYSCTL_INT(_net_inet_tcp, TCPCTL_V6MSSDFLT, v6mssdflt, 108 CTLFLAG_RW, &tcp_v6mssdflt , 0, 109 "Default TCP Maximum Segment Size for IPv6"); 110#endif 111 112#if 0 113static int tcp_rttdflt = TCPTV_SRTTDFLT / PR_SLOWHZ; 114SYSCTL_INT(_net_inet_tcp, TCPCTL_RTTDFLT, rttdflt, CTLFLAG_RW, 115 &tcp_rttdflt , 0, "Default maximum TCP Round Trip Time"); 116#endif 117 118static int tcp_do_rfc1323 = 1; 119SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323, CTLFLAG_RW, 120 &tcp_do_rfc1323 , 0, "Enable rfc1323 (high performance TCP) extensions"); 121 122static int tcp_do_rfc1644 = 0; 123SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1644, rfc1644, CTLFLAG_RW, 124 &tcp_do_rfc1644 , 0, "Enable rfc1644 (TTCP) extensions"); 125 126static int tcp_tcbhashsize = 0; 127SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcbhashsize, CTLFLAG_RD, 128 &tcp_tcbhashsize, 0, "Size of TCP control-block hashtable"); 129 130static int do_tcpdrain = 1; 131SYSCTL_INT(_net_inet_tcp, OID_AUTO, do_tcpdrain, CTLFLAG_RW, &do_tcpdrain, 0, 132 "Enable tcp_drain routine for extra help when low on mbufs"); 133 134SYSCTL_INT(_net_inet_tcp, OID_AUTO, pcbcount, CTLFLAG_RD, 135 &tcbinfo.ipi_count, 0, "Number of active PCBs"); 136 137/* 138 * Treat ICMP administratively prohibited like a TCP RST 139 * as required by rfc1122 section 3.2.2.1 140 */ 141 142static int icmp_admin_prohib_like_rst = 0; 143SYSCTL_INT(_net_inet_tcp, OID_AUTO, icmp_admin_prohib_like_rst, CTLFLAG_RW, 144 &icmp_admin_prohib_like_rst, 0, "Treat ICMP administratively prohibited messages like TCP RST, rfc1122 section 3.2.2.1"); 145 146static void tcp_cleartaocache __P((void)); 147static void tcp_notify __P((struct inpcb *, int)); 148 149/* 150 * Target size of TCP PCB hash tables. Must be a power of two. 151 * 152 * Note that this can be overridden by the kernel environment 153 * variable net.inet.tcp.tcbhashsize 154 */ 155#ifndef TCBHASHSIZE 156#define TCBHASHSIZE 512 157#endif 158 159/* 160 * This is the actual shape of what we allocate using the zone 161 * allocator. Doing it this way allows us to protect both structures 162 * using the same generation count, and also eliminates the overhead 163 * of allocating tcpcbs separately. By hiding the structure here, 164 * we avoid changing most of the rest of the code (although it needs 165 * to be changed, eventually, for greater efficiency). 166 */ 167#define ALIGNMENT 32 168#define ALIGNM1 (ALIGNMENT - 1) 169struct inp_tp { 170 union { 171 struct inpcb inp; 172 char align[(sizeof(struct inpcb) + ALIGNM1) & ~ALIGNM1]; 173 } inp_tp_u; 174 struct tcpcb tcb; 175 struct callout inp_tp_rexmt, inp_tp_persist, inp_tp_keep, inp_tp_2msl; 176 struct callout inp_tp_delack; 177}; 178#undef ALIGNMENT 179#undef ALIGNM1 180 181/* 182 * Tcp initialization 183 */ 184void 185tcp_init() 186{ 187 int hashsize; 188 189 tcp_iss = arc4random(); /* wrong, but better than a constant */ 190 tcp_ccgen = 1; 191 tcp_cleartaocache(); 192 193 tcp_delacktime = TCPTV_DELACK; 194 tcp_keepinit = TCPTV_KEEP_INIT; 195 tcp_keepidle = TCPTV_KEEP_IDLE; 196 tcp_keepintvl = TCPTV_KEEPINTVL; 197 tcp_maxpersistidle = TCPTV_KEEP_IDLE; 198 tcp_msl = TCPTV_MSL; 199 200 LIST_INIT(&tcb); 201 tcbinfo.listhead = &tcb; 202 TUNABLE_INT_FETCH("net.inet.tcp.tcbhashsize", TCBHASHSIZE, hashsize); 203 if (!powerof2(hashsize)) { 204 printf("WARNING: TCB hash size not a power of 2\n"); 205 hashsize = 512; /* safe default */ 206 } 207 tcp_tcbhashsize = hashsize; 208 tcbinfo.hashbase = hashinit(hashsize, M_PCB, &tcbinfo.hashmask); 209 tcbinfo.porthashbase = hashinit(hashsize, M_PCB, 210 &tcbinfo.porthashmask); 211 tcbinfo.ipi_zone = zinit("tcpcb", sizeof(struct inp_tp), maxsockets, 212 ZONE_INTERRUPT, 0); 213#ifdef INET6 214#define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr)) 215#else /* INET6 */ 216#define TCP_MINPROTOHDR (sizeof(struct tcpiphdr)) 217#endif /* INET6 */ 218 if (max_protohdr < TCP_MINPROTOHDR) 219 max_protohdr = TCP_MINPROTOHDR; 220 if (max_linkhdr + TCP_MINPROTOHDR > MHLEN) 221 panic("tcp_init"); 222#undef TCP_MINPROTOHDR 223} 224 225/* 226 * Create template to be used to send tcp packets on a connection. 227 * Call after host entry created, allocates an mbuf and fills 228 * in a skeletal tcp/ip header, minimizing the amount of work 229 * necessary when the connection is used. 230 */ 231struct tcptemp * 232tcp_template(tp) 233 struct tcpcb *tp; 234{ 235 register struct inpcb *inp = tp->t_inpcb; 236 register struct mbuf *m; 237 register struct tcptemp *n; 238 239 if ((n = tp->t_template) == 0) { 240 m = m_get(M_DONTWAIT, MT_HEADER); 241 if (m == NULL) 242 return (0); 243 m->m_len = sizeof (struct tcptemp); 244 n = mtod(m, struct tcptemp *); 245 } 246#ifdef INET6 247 if ((inp->inp_vflag & INP_IPV6) != 0) { 248 register struct ip6_hdr *ip6; 249 250 ip6 = (struct ip6_hdr *)n->tt_ipgen; 251 ip6->ip6_flow = (ip6->ip6_flow & ~IPV6_FLOWINFO_MASK) | 252 (inp->in6p_flowinfo & IPV6_FLOWINFO_MASK); 253 ip6->ip6_vfc = (ip6->ip6_vfc & ~IPV6_VERSION_MASK) | 254 (IPV6_VERSION & IPV6_VERSION_MASK); 255 ip6->ip6_nxt = IPPROTO_TCP; 256 ip6->ip6_plen = sizeof(struct tcphdr); 257 ip6->ip6_src = inp->in6p_laddr; 258 ip6->ip6_dst = inp->in6p_faddr; 259 n->tt_t.th_sum = 0; 260 } else 261#endif 262 { 263 struct ip *ip = (struct ip *)n->tt_ipgen; 264 265 bzero(ip, sizeof(struct ip)); /* XXX overkill? */ 266 ip->ip_vhl = IP_VHL_BORING; 267 ip->ip_p = IPPROTO_TCP; 268 ip->ip_src = inp->inp_laddr; 269 ip->ip_dst = inp->inp_faddr; 270 n->tt_t.th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, 271 htons(sizeof(struct tcphdr) + IPPROTO_TCP)); 272 } 273 n->tt_t.th_sport = inp->inp_lport; 274 n->tt_t.th_dport = inp->inp_fport; 275 n->tt_t.th_seq = 0; 276 n->tt_t.th_ack = 0; 277 n->tt_t.th_x2 = 0; 278 n->tt_t.th_off = 5; 279 n->tt_t.th_flags = 0; 280 n->tt_t.th_win = 0; 281 n->tt_t.th_urp = 0; 282 return (n); 283} 284 285/* 286 * Send a single message to the TCP at address specified by 287 * the given TCP/IP header. If m == 0, then we make a copy 288 * of the tcpiphdr at ti and send directly to the addressed host. 289 * This is used to force keep alive messages out using the TCP 290 * template for a connection tp->t_template. If flags are given 291 * then we send a message back to the TCP which originated the 292 * segment ti, and discard the mbuf containing it and any other 293 * attached mbufs. 294 * 295 * In any case the ack and sequence number of the transmitted 296 * segment are as specified by the parameters. 297 * 298 * NOTE: If m != NULL, then ti must point to *inside* the mbuf. 299 */ 300void 301tcp_respond(tp, ipgen, th, m, ack, seq, flags) 302 struct tcpcb *tp; 303 void *ipgen; 304 register struct tcphdr *th; 305 register struct mbuf *m; 306 tcp_seq ack, seq; 307 int flags; 308{ 309 register int tlen; 310 int win = 0; 311 struct route *ro = 0; 312 struct route sro; 313 struct ip *ip; 314 struct tcphdr *nth; 315#ifdef INET6 316 struct route_in6 *ro6 = 0; 317 struct route_in6 sro6; 318 struct ip6_hdr *ip6; 319 int isipv6; 320#endif /* INET6 */ 321 int ipflags = 0; 322 323#ifdef INET6 324 isipv6 = IP_VHL_V(((struct ip *)ipgen)->ip_vhl) == 6; 325 ip6 = ipgen; 326#endif /* INET6 */ 327 ip = ipgen; 328 329 if (tp) { 330 if (!(flags & TH_RST)) { 331 win = sbspace(&tp->t_inpcb->inp_socket->so_rcv); 332 if (win > (long)TCP_MAXWIN << tp->rcv_scale) 333 win = (long)TCP_MAXWIN << tp->rcv_scale; 334 } 335#ifdef INET6 336 if (isipv6) 337 ro6 = &tp->t_inpcb->in6p_route; 338 else 339#endif /* INET6 */ 340 ro = &tp->t_inpcb->inp_route; 341 } else { 342#ifdef INET6 343 if (isipv6) { 344 ro6 = &sro6; 345 bzero(ro6, sizeof *ro6); 346 } else 347#endif /* INET6 */ 348 { 349 ro = &sro; 350 bzero(ro, sizeof *ro); 351 } 352 } 353 if (m == 0) { 354 m = m_gethdr(M_DONTWAIT, MT_HEADER); 355 if (m == NULL) 356 return; 357#ifdef TCP_COMPAT_42 358 tlen = 1; 359#else 360 tlen = 0; 361#endif 362 m->m_data += max_linkhdr; 363#ifdef INET6 364 if (isipv6) { 365 bcopy((caddr_t)ip6, mtod(m, caddr_t), 366 sizeof(struct ip6_hdr)); 367 ip6 = mtod(m, struct ip6_hdr *); 368 nth = (struct tcphdr *)(ip6 + 1); 369 } else 370#endif /* INET6 */ 371 { 372 bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 373 ip = mtod(m, struct ip *); 374 nth = (struct tcphdr *)(ip + 1); 375 } 376 bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr)); 377 flags = TH_ACK; 378 } else { 379 m_freem(m->m_next); 380 m->m_next = 0; 381 m->m_data = (caddr_t)ipgen; 382 /* m_len is set later */ 383 tlen = 0; 384#define xchg(a,b,type) { type t; t=a; a=b; b=t; } 385#ifdef INET6 386 if (isipv6) { 387 xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr); 388 nth = (struct tcphdr *)(ip6 + 1); 389 } else 390#endif /* INET6 */ 391 { 392 xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, n_long); 393 nth = (struct tcphdr *)(ip + 1); 394 } 395 if (th != nth) { 396 /* 397 * this is usually a case when an extension header 398 * exists between the IPv6 header and the 399 * TCP header. 400 */ 401 nth->th_sport = th->th_sport; 402 nth->th_dport = th->th_dport; 403 } 404 xchg(nth->th_dport, nth->th_sport, n_short); 405#undef xchg 406 } 407#ifdef INET6 408 if (isipv6) { 409 ip6->ip6_plen = htons((u_short)(sizeof (struct tcphdr) + 410 tlen)); 411 tlen += sizeof (struct ip6_hdr) + sizeof (struct tcphdr); 412 } else 413#endif 414 { 415 tlen += sizeof (struct tcpiphdr); 416 ip->ip_len = tlen; 417 ip->ip_ttl = ip_defttl; 418 } 419 m->m_len = tlen; 420 m->m_pkthdr.len = tlen; 421 m->m_pkthdr.rcvif = (struct ifnet *) 0; 422 nth->th_seq = htonl(seq); 423 nth->th_ack = htonl(ack); 424 nth->th_x2 = 0; 425 nth->th_off = sizeof (struct tcphdr) >> 2; 426 nth->th_flags = flags; 427 if (tp) 428 nth->th_win = htons((u_short) (win >> tp->rcv_scale)); 429 else 430 nth->th_win = htons((u_short)win); 431 nth->th_urp = 0; 432#ifdef INET6 433 if (isipv6) { 434 nth->th_sum = 0; 435 nth->th_sum = in6_cksum(m, IPPROTO_TCP, 436 sizeof(struct ip6_hdr), 437 tlen - sizeof(struct ip6_hdr)); 438 ip6->ip6_hlim = in6_selecthlim(tp ? tp->t_inpcb : NULL, 439 ro6 && ro6->ro_rt ? 440 ro6->ro_rt->rt_ifp : 441 NULL); 442 } else 443#endif /* INET6 */ 444 { 445 nth->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, 446 htons((u_short)(tlen - sizeof(struct ip) + ip->ip_p))); 447 m->m_pkthdr.csum_flags = CSUM_TCP; 448 m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); 449 } 450#ifdef TCPDEBUG 451 if (tp == NULL || (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) 452 tcp_trace(TA_OUTPUT, 0, tp, mtod(m, void *), th, 0); 453#endif 454#ifdef IPSEC 455 ipsec_setsocket(m, tp ? tp->t_inpcb->inp_socket : NULL); 456#endif 457#ifdef INET6 458 if (isipv6) { 459 (void)ip6_output(m, NULL, ro6, ipflags, NULL, NULL); 460 if (ro6 == &sro6 && ro6->ro_rt) { 461 RTFREE(ro6->ro_rt); 462 ro6->ro_rt = NULL; 463 } 464 } else 465#endif /* INET6 */ 466 { 467 (void) ip_output(m, NULL, ro, ipflags, NULL); 468 if (ro == &sro && ro->ro_rt) { 469 RTFREE(ro->ro_rt); 470 ro->ro_rt = NULL; 471 } 472 } 473} 474 475/* 476 * Create a new TCP control block, making an 477 * empty reassembly queue and hooking it to the argument 478 * protocol control block. The `inp' parameter must have 479 * come from the zone allocator set up in tcp_init(). 480 */ 481struct tcpcb * 482tcp_newtcpcb(inp) 483 struct inpcb *inp; 484{ 485 struct inp_tp *it; 486 register struct tcpcb *tp; 487#ifdef INET6 488 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; 489#endif /* INET6 */ 490 491 it = (struct inp_tp *)inp; 492 tp = &it->tcb; 493 bzero((char *) tp, sizeof(struct tcpcb)); 494 LIST_INIT(&tp->t_segq); 495 tp->t_maxseg = tp->t_maxopd = 496#ifdef INET6 497 isipv6 ? tcp_v6mssdflt : 498#endif /* INET6 */ 499 tcp_mssdflt; 500 501 /* Set up our timeouts. */ 502 callout_init(tp->tt_rexmt = &it->inp_tp_rexmt, 0); 503 callout_init(tp->tt_persist = &it->inp_tp_persist, 0); 504 callout_init(tp->tt_keep = &it->inp_tp_keep, 0); 505 callout_init(tp->tt_2msl = &it->inp_tp_2msl, 0); 506 callout_init(tp->tt_delack = &it->inp_tp_delack, 0); 507 508 if (tcp_do_rfc1323) 509 tp->t_flags = (TF_REQ_SCALE|TF_REQ_TSTMP); 510 if (tcp_do_rfc1644) 511 tp->t_flags |= TF_REQ_CC; 512 tp->t_inpcb = inp; /* XXX */ 513 /* 514 * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no 515 * rtt estimate. Set rttvar so that srtt + 4 * rttvar gives 516 * reasonable initial retransmit time. 517 */ 518 tp->t_srtt = TCPTV_SRTTBASE; 519 tp->t_rttvar = ((TCPTV_RTOBASE - TCPTV_SRTTBASE) << TCP_RTTVAR_SHIFT) / 4; 520 tp->t_rttmin = TCPTV_MIN; 521 tp->t_rxtcur = TCPTV_RTOBASE; 522 tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; 523 tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT; 524 tp->t_rcvtime = ticks; 525 /* 526 * IPv4 TTL initialization is necessary for an IPv6 socket as well, 527 * because the socket may be bound to an IPv6 wildcard address, 528 * which may match an IPv4-mapped IPv6 address. 529 */ 530 inp->inp_ip_ttl = ip_defttl; 531 inp->inp_ppcb = (caddr_t)tp; 532 return (tp); /* XXX */ 533} 534 535/* 536 * Drop a TCP connection, reporting 537 * the specified error. If connection is synchronized, 538 * then send a RST to peer. 539 */ 540struct tcpcb * 541tcp_drop(tp, errno) 542 register struct tcpcb *tp; 543 int errno; 544{ 545 struct socket *so = tp->t_inpcb->inp_socket; 546 547 if (TCPS_HAVERCVDSYN(tp->t_state)) { 548 tp->t_state = TCPS_CLOSED; 549 (void) tcp_output(tp); 550 tcpstat.tcps_drops++; 551 } else 552 tcpstat.tcps_conndrops++; 553 if (errno == ETIMEDOUT && tp->t_softerror) 554 errno = tp->t_softerror; 555 so->so_error = errno; 556 return (tcp_close(tp)); 557} 558 559/* 560 * Close a TCP control block: 561 * discard all space held by the tcp 562 * discard internet protocol block 563 * wake up any sleepers 564 */ 565struct tcpcb * 566tcp_close(tp) 567 register struct tcpcb *tp; 568{ 569 register struct tseg_qent *q; 570 struct inpcb *inp = tp->t_inpcb; 571 struct socket *so = inp->inp_socket; 572#ifdef INET6 573 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; 574#endif /* INET6 */ 575 register struct rtentry *rt; 576 int dosavessthresh; 577 578 /* 579 * Make sure that all of our timers are stopped before we 580 * delete the PCB. 581 */ 582 callout_stop(tp->tt_rexmt); 583 callout_stop(tp->tt_persist); 584 callout_stop(tp->tt_keep); 585 callout_stop(tp->tt_2msl); 586 callout_stop(tp->tt_delack); 587 588 /* 589 * If we got enough samples through the srtt filter, 590 * save the rtt and rttvar in the routing entry. 591 * 'Enough' is arbitrarily defined as the 16 samples. 592 * 16 samples is enough for the srtt filter to converge 593 * to within 5% of the correct value; fewer samples and 594 * we could save a very bogus rtt. 595 * 596 * Don't update the default route's characteristics and don't 597 * update anything that the user "locked". 598 */ 599 if (tp->t_rttupdated >= 16) { 600 register u_long i = 0; 601#ifdef INET6 602 if (isipv6) { 603 struct sockaddr_in6 *sin6; 604 605 if ((rt = inp->in6p_route.ro_rt) == NULL) 606 goto no_valid_rt; 607 sin6 = (struct sockaddr_in6 *)rt_key(rt); 608 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) 609 goto no_valid_rt; 610 } 611 else 612#endif /* INET6 */ 613 if ((rt = inp->inp_route.ro_rt) == NULL || 614 ((struct sockaddr_in *)rt_key(rt))->sin_addr.s_addr 615 == INADDR_ANY) 616 goto no_valid_rt; 617 618 if ((rt->rt_rmx.rmx_locks & RTV_RTT) == 0) { 619 i = tp->t_srtt * 620 (RTM_RTTUNIT / (hz * TCP_RTT_SCALE)); 621 if (rt->rt_rmx.rmx_rtt && i) 622 /* 623 * filter this update to half the old & half 624 * the new values, converting scale. 625 * See route.h and tcp_var.h for a 626 * description of the scaling constants. 627 */ 628 rt->rt_rmx.rmx_rtt = 629 (rt->rt_rmx.rmx_rtt + i) / 2; 630 else 631 rt->rt_rmx.rmx_rtt = i; 632 tcpstat.tcps_cachedrtt++; 633 } 634 if ((rt->rt_rmx.rmx_locks & RTV_RTTVAR) == 0) { 635 i = tp->t_rttvar * 636 (RTM_RTTUNIT / (hz * TCP_RTTVAR_SCALE)); 637 if (rt->rt_rmx.rmx_rttvar && i) 638 rt->rt_rmx.rmx_rttvar = 639 (rt->rt_rmx.rmx_rttvar + i) / 2; 640 else 641 rt->rt_rmx.rmx_rttvar = i; 642 tcpstat.tcps_cachedrttvar++; 643 } 644 /* 645 * The old comment here said: 646 * update the pipelimit (ssthresh) if it has been updated 647 * already or if a pipesize was specified & the threshhold 648 * got below half the pipesize. I.e., wait for bad news 649 * before we start updating, then update on both good 650 * and bad news. 651 * 652 * But we want to save the ssthresh even if no pipesize is 653 * specified explicitly in the route, because such 654 * connections still have an implicit pipesize specified 655 * by the global tcp_sendspace. In the absence of a reliable 656 * way to calculate the pipesize, it will have to do. 657 */ 658 i = tp->snd_ssthresh; 659 if (rt->rt_rmx.rmx_sendpipe != 0) 660 dosavessthresh = (i < rt->rt_rmx.rmx_sendpipe / 2); 661 else 662 dosavessthresh = (i < so->so_snd.sb_hiwat / 2); 663 if (((rt->rt_rmx.rmx_locks & RTV_SSTHRESH) == 0 && 664 i != 0 && rt->rt_rmx.rmx_ssthresh != 0) 665 || dosavessthresh) { 666 /* 667 * convert the limit from user data bytes to 668 * packets then to packet data bytes. 669 */ 670 i = (i + tp->t_maxseg / 2) / tp->t_maxseg; 671 if (i < 2) 672 i = 2; 673 i *= (u_long)(tp->t_maxseg + 674#ifdef INET6 675 (isipv6 ? sizeof (struct ip6_hdr) + 676 sizeof (struct tcphdr) : 677#endif 678 sizeof (struct tcpiphdr) 679#ifdef INET6 680 ) 681#endif 682 ); 683 if (rt->rt_rmx.rmx_ssthresh) 684 rt->rt_rmx.rmx_ssthresh = 685 (rt->rt_rmx.rmx_ssthresh + i) / 2; 686 else 687 rt->rt_rmx.rmx_ssthresh = i; 688 tcpstat.tcps_cachedssthresh++; 689 } 690 } 691 rt = inp->inp_route.ro_rt; 692 if (rt) { 693 /* 694 * mark route for deletion if no information is 695 * cached. 696 */ 697 if ((tp->t_flags & TF_LQ_OVERFLOW) && 698 ((rt->rt_rmx.rmx_locks & RTV_RTT) == 0)){ 699 if (rt->rt_rmx.rmx_rtt == 0) 700 rt->rt_flags |= RTF_DELCLONE; 701 } 702 } 703 no_valid_rt: 704 /* free the reassembly queue, if any */ 705 while((q = LIST_FIRST(&tp->t_segq)) != NULL) { 706 LIST_REMOVE(q, tqe_q); 707 m_freem(q->tqe_m); 708 FREE(q, M_TSEGQ); 709 } 710 if (tp->t_template) 711 (void) m_free(dtom(tp->t_template)); 712 inp->inp_ppcb = NULL; 713 soisdisconnected(so); 714#ifdef INET6 715 if (INP_CHECK_SOCKAF(so, AF_INET6)) 716 in6_pcbdetach(inp); 717 else 718#endif /* INET6 */ 719 in_pcbdetach(inp); 720 tcpstat.tcps_closed++; 721 return ((struct tcpcb *)0); 722} 723 724void 725tcp_drain() 726{ 727 if (do_tcpdrain) 728 { 729 struct inpcb *inpb; 730 struct tcpcb *tcpb; 731 struct tseg_qent *te; 732 733 /* 734 * Walk the tcpbs, if existing, and flush the reassembly queue, 735 * if there is one... 736 * XXX: The "Net/3" implementation doesn't imply that the TCP 737 * reassembly queue should be flushed, but in a situation 738 * where we're really low on mbufs, this is potentially 739 * usefull. 740 */ 741 for (inpb = tcbinfo.listhead->lh_first; inpb; 742 inpb = inpb->inp_list.le_next) { 743 if ((tcpb = intotcpcb(inpb))) { 744 while ((te = LIST_FIRST(&tcpb->t_segq)) 745 != NULL) { 746 LIST_REMOVE(te, tqe_q); 747 m_freem(te->tqe_m); 748 FREE(te, M_TSEGQ); 749 } 750 } 751 } 752 753 } 754} 755 756/* 757 * Notify a tcp user of an asynchronous error; 758 * store error as soft error, but wake up user 759 * (for now, won't do anything until can select for soft error). 760 */ 761static void 762tcp_notify(inp, error) 763 struct inpcb *inp; 764 int error; 765{ 766 register struct tcpcb *tp = (struct tcpcb *)inp->inp_ppcb; 767 register struct socket *so = inp->inp_socket; 768 769 /* 770 * Ignore some errors if we are hooked up. 771 * If connection hasn't completed, has retransmitted several times, 772 * and receives a second error, give up now. This is better 773 * than waiting a long time to establish a connection that 774 * can never complete. 775 */ 776 if (tp->t_state == TCPS_ESTABLISHED && 777 (error == EHOSTUNREACH || error == ENETUNREACH || 778 error == EHOSTDOWN)) { 779 return; 780 } else if (tp->t_state < TCPS_ESTABLISHED && tp->t_rxtshift > 3 && 781 tp->t_softerror) 782 so->so_error = error; 783 else 784 tp->t_softerror = error; 785 wakeup((caddr_t) &so->so_timeo); 786 sorwakeup(so); 787 sowwakeup(so); 788} 789 790static int 791tcp_pcblist(SYSCTL_HANDLER_ARGS) 792{ 793 int error, i, n, s; 794 struct inpcb *inp, **inp_list; 795 inp_gen_t gencnt; 796 struct xinpgen xig; 797 798 /* 799 * The process of preparing the TCB list is too time-consuming and 800 * resource-intensive to repeat twice on every request. 801 */ 802 if (req->oldptr == 0) { 803 n = tcbinfo.ipi_count; 804 req->oldidx = 2 * (sizeof xig) 805 + (n + n/8) * sizeof(struct xtcpcb); 806 return 0; 807 } 808 809 if (req->newptr != 0) 810 return EPERM; 811 812 /* 813 * OK, now we're committed to doing something. 814 */ 815 s = splnet(); 816 gencnt = tcbinfo.ipi_gencnt; 817 n = tcbinfo.ipi_count; 818 splx(s); 819 820 xig.xig_len = sizeof xig; 821 xig.xig_count = n; 822 xig.xig_gen = gencnt; 823 xig.xig_sogen = so_gencnt; 824 error = SYSCTL_OUT(req, &xig, sizeof xig); 825 if (error) 826 return error; 827 828 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 829 if (inp_list == 0) 830 return ENOMEM; 831 832 s = splnet(); 833 for (inp = tcbinfo.listhead->lh_first, i = 0; inp && i < n; 834 inp = inp->inp_list.le_next) { 835 if (inp->inp_gencnt <= gencnt && !prison_xinpcb(req->p, inp)) 836 inp_list[i++] = inp; 837 } 838 splx(s); 839 n = i; 840 841 error = 0; 842 for (i = 0; i < n; i++) { 843 inp = inp_list[i]; 844 if (inp->inp_gencnt <= gencnt) { 845 struct xtcpcb xt; 846 caddr_t inp_ppcb; 847 xt.xt_len = sizeof xt; 848 /* XXX should avoid extra copy */ 849 bcopy(inp, &xt.xt_inp, sizeof *inp); 850 inp_ppcb = inp->inp_ppcb; 851 if (inp_ppcb != NULL) 852 bcopy(inp_ppcb, &xt.xt_tp, sizeof xt.xt_tp); 853 else 854 bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); 855 if (inp->inp_socket) 856 sotoxsocket(inp->inp_socket, &xt.xt_socket); 857 error = SYSCTL_OUT(req, &xt, sizeof xt); 858 } 859 } 860 if (!error) { 861 /* 862 * Give the user an updated idea of our state. 863 * If the generation differs from what we told 864 * her before, she knows that something happened 865 * while we were processing this request, and it 866 * might be necessary to retry. 867 */ 868 s = splnet(); 869 xig.xig_gen = tcbinfo.ipi_gencnt; 870 xig.xig_sogen = so_gencnt; 871 xig.xig_count = tcbinfo.ipi_count; 872 splx(s); 873 error = SYSCTL_OUT(req, &xig, sizeof xig); 874 } 875 free(inp_list, M_TEMP); 876 return error; 877} 878 879SYSCTL_PROC(_net_inet_tcp, TCPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0, 880 tcp_pcblist, "S,xtcpcb", "List of active TCP connections"); 881 882static int 883tcp_getcred(SYSCTL_HANDLER_ARGS) 884{ 885 struct sockaddr_in addrs[2]; 886 struct inpcb *inp; 887 int error, s; 888 889 error = suser(req->p); 890 if (error) 891 return (error); 892 error = SYSCTL_IN(req, addrs, sizeof(addrs)); 893 if (error) 894 return (error); 895 s = splnet(); 896 inp = in_pcblookup_hash(&tcbinfo, addrs[1].sin_addr, addrs[1].sin_port, 897 addrs[0].sin_addr, addrs[0].sin_port, 0, NULL); 898 if (inp == NULL || inp->inp_socket == NULL) { 899 error = ENOENT; 900 goto out; 901 } 902 error = SYSCTL_OUT(req, inp->inp_socket->so_cred, sizeof(struct ucred)); 903out: 904 splx(s); 905 return (error); 906} 907 908SYSCTL_PROC(_net_inet_tcp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 909 0, 0, tcp_getcred, "S,ucred", "Get the ucred of a TCP connection"); 910 911#ifdef INET6 912static int 913tcp6_getcred(SYSCTL_HANDLER_ARGS) 914{ 915 struct sockaddr_in6 addrs[2]; 916 struct inpcb *inp; 917 int error, s, mapped = 0; 918 919 error = suser(req->p); 920 if (error) 921 return (error); 922 error = SYSCTL_IN(req, addrs, sizeof(addrs)); 923 if (error) 924 return (error); 925 if (IN6_IS_ADDR_V4MAPPED(&addrs[0].sin6_addr)) { 926 if (IN6_IS_ADDR_V4MAPPED(&addrs[1].sin6_addr)) 927 mapped = 1; 928 else 929 return (EINVAL); 930 } 931 s = splnet(); 932 if (mapped == 1) 933 inp = in_pcblookup_hash(&tcbinfo, 934 *(struct in_addr *)&addrs[1].sin6_addr.s6_addr[12], 935 addrs[1].sin6_port, 936 *(struct in_addr *)&addrs[0].sin6_addr.s6_addr[12], 937 addrs[0].sin6_port, 938 0, NULL); 939 else 940 inp = in6_pcblookup_hash(&tcbinfo, &addrs[1].sin6_addr, 941 addrs[1].sin6_port, 942 &addrs[0].sin6_addr, addrs[0].sin6_port, 943 0, NULL); 944 if (inp == NULL || inp->inp_socket == NULL) { 945 error = ENOENT; 946 goto out; 947 } 948 error = SYSCTL_OUT(req, inp->inp_socket->so_cred, 949 sizeof(struct ucred)); 950out: 951 splx(s); 952 return (error); 953} 954 955SYSCTL_PROC(_net_inet6_tcp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 956 0, 0, 957 tcp6_getcred, "S,ucred", "Get the ucred of a TCP6 connection"); 958#endif 959 960 961void 962tcp_ctlinput(cmd, sa, vip) 963 int cmd; 964 struct sockaddr *sa; 965 void *vip; 966{ 967 register struct ip *ip = vip; 968 register struct tcphdr *th; 969 void (*notify) __P((struct inpcb *, int)) = tcp_notify; 970 971 if (cmd == PRC_QUENCH) 972 notify = tcp_quench; 973 else if ((icmp_admin_prohib_like_rst == 1) && (cmd == PRC_UNREACH_PORT) && (ip)) 974 notify = tcp_drop_syn_sent; 975 else if (cmd == PRC_MSGSIZE) 976 notify = tcp_mtudisc; 977 else if (!PRC_IS_REDIRECT(cmd) && 978 ((unsigned)cmd > PRC_NCMDS || inetctlerrmap[cmd] == 0)) 979 return; 980 if (ip) { 981 th = (struct tcphdr *)((caddr_t)ip 982 + (IP_VHL_HL(ip->ip_vhl) << 2)); 983 in_pcbnotify(&tcb, sa, th->th_dport, ip->ip_src, th->th_sport, 984 cmd, notify); 985 } else 986 in_pcbnotify(&tcb, sa, 0, zeroin_addr, 0, cmd, notify); 987} 988 989#ifdef INET6 990void 991tcp6_ctlinput(cmd, sa, d) 992 int cmd; 993 struct sockaddr *sa; 994 void *d; 995{ 996 register struct tcphdr *thp; 997 struct tcphdr th; 998 void (*notify) __P((struct inpcb *, int)) = tcp_notify; 999 struct sockaddr_in6 sa6; 1000 struct ip6_hdr *ip6; 1001 struct mbuf *m; 1002 int off; 1003 1004 if (sa->sa_family != AF_INET6 || 1005 sa->sa_len != sizeof(struct sockaddr_in6)) 1006 return; 1007 1008 if (cmd == PRC_QUENCH) 1009 notify = tcp_quench; 1010 else if (cmd == PRC_MSGSIZE) 1011 notify = tcp_mtudisc; 1012 else if (!PRC_IS_REDIRECT(cmd) && 1013 ((unsigned)cmd > PRC_NCMDS || inet6ctlerrmap[cmd] == 0)) 1014 return; 1015 1016 /* if the parameter is from icmp6, decode it. */ 1017 if (d != NULL) { 1018 struct ip6ctlparam *ip6cp = (struct ip6ctlparam *)d; 1019 m = ip6cp->ip6c_m; 1020 ip6 = ip6cp->ip6c_ip6; 1021 off = ip6cp->ip6c_off; 1022 } else { 1023 m = NULL; 1024 ip6 = NULL; 1025 off = 0; /* fool gcc */ 1026 } 1027 1028 /* 1029 * Translate addresses into internal form. 1030 * Sa check if it is AF_INET6 is done at the top of this funciton. 1031 */ 1032 sa6 = *(struct sockaddr_in6 *)sa; 1033 if (IN6_IS_ADDR_LINKLOCAL(&sa6.sin6_addr) != 0 && m != NULL && 1034 m->m_pkthdr.rcvif != NULL) 1035 sa6.sin6_addr.s6_addr16[1] = htons(m->m_pkthdr.rcvif->if_index); 1036 1037 if (ip6) { 1038 /* 1039 * XXX: We assume that when IPV6 is non NULL, 1040 * M and OFF are valid. 1041 */ 1042 struct in6_addr s; 1043 1044 /* translate addresses into internal form */ 1045 memcpy(&s, &ip6->ip6_src, sizeof(s)); 1046 if (IN6_IS_ADDR_LINKLOCAL(&s) != 0 && m != NULL && 1047 m->m_pkthdr.rcvif != NULL) 1048 s.s6_addr16[1] = htons(m->m_pkthdr.rcvif->if_index); 1049 1050 /* check if we can safely examine src and dst ports */ 1051 if (m->m_pkthdr.len < off + sizeof(th)) 1052 return; 1053 1054 if (m->m_len < off + sizeof(th)) { 1055 /* 1056 * this should be rare case 1057 * because now MINCLSIZE is "(MHLEN + 1)", 1058 * so we compromise on this copy... 1059 */ 1060 m_copydata(m, off, sizeof(th), (caddr_t)&th); 1061 thp = &th; 1062 } else 1063 thp = (struct tcphdr *)(mtod(m, caddr_t) + off); 1064 in6_pcbnotify(&tcb, (struct sockaddr *)&sa6, thp->th_dport, 1065 &s, thp->th_sport, cmd, notify); 1066 } else 1067 in6_pcbnotify(&tcb, (struct sockaddr *)&sa6, 0, &zeroin6_addr, 1068 0, cmd, notify); 1069} 1070#endif /* INET6 */ 1071 1072/* 1073 * When a source quench is received, close congestion window 1074 * to one segment. We will gradually open it again as we proceed. 1075 */ 1076void 1077tcp_quench(inp, errno) 1078 struct inpcb *inp; 1079 int errno; 1080{ 1081 struct tcpcb *tp = intotcpcb(inp); 1082 1083 if (tp) 1084 tp->snd_cwnd = tp->t_maxseg; 1085} 1086 1087/* 1088 * When a ICMP unreachable is recieved, drop the 1089 * TCP connection, but only if in SYN_SENT 1090 */ 1091void 1092tcp_drop_syn_sent(inp, errno) 1093 struct inpcb *inp; 1094 int errno; 1095{ 1096 struct tcpcb *tp = intotcpcb(inp); 1097 if((tp) && (tp->t_state == TCPS_SYN_SENT)) 1098 tcp_drop(tp, errno); 1099} 1100 1101/* 1102 * When `need fragmentation' ICMP is received, update our idea of the MSS 1103 * based on the new value in the route. Also nudge TCP to send something, 1104 * since we know the packet we just sent was dropped. 1105 * This duplicates some code in the tcp_mss() function in tcp_input.c. 1106 */ 1107void 1108tcp_mtudisc(inp, errno) 1109 struct inpcb *inp; 1110 int errno; 1111{ 1112 struct tcpcb *tp = intotcpcb(inp); 1113 struct rtentry *rt; 1114 struct rmxp_tao *taop; 1115 struct socket *so = inp->inp_socket; 1116 int offered; 1117 int mss; 1118#ifdef INET6 1119 int isipv6 = (tp->t_inpcb->inp_vflag & INP_IPV6) != 0; 1120#endif /* INET6 */ 1121 1122 if (tp) { 1123#ifdef INET6 1124 if (isipv6) 1125 rt = tcp_rtlookup6(inp); 1126 else 1127#endif /* INET6 */ 1128 rt = tcp_rtlookup(inp); 1129 if (!rt || !rt->rt_rmx.rmx_mtu) { 1130 tp->t_maxopd = tp->t_maxseg = 1131#ifdef INET6 1132 isipv6 ? tcp_v6mssdflt : 1133#endif /* INET6 */ 1134 tcp_mssdflt; 1135 return; 1136 } 1137 taop = rmx_taop(rt->rt_rmx); 1138 offered = taop->tao_mssopt; 1139 mss = rt->rt_rmx.rmx_mtu - 1140#ifdef INET6 1141 (isipv6 ? 1142 sizeof(struct ip6_hdr) + sizeof(struct tcphdr) : 1143#endif /* INET6 */ 1144 sizeof(struct tcpiphdr) 1145#ifdef INET6 1146 ) 1147#endif /* INET6 */ 1148 ; 1149 1150 if (offered) 1151 mss = min(mss, offered); 1152 /* 1153 * XXX - The above conditional probably violates the TCP 1154 * spec. The problem is that, since we don't know the 1155 * other end's MSS, we are supposed to use a conservative 1156 * default. But, if we do that, then MTU discovery will 1157 * never actually take place, because the conservative 1158 * default is much less than the MTUs typically seen 1159 * on the Internet today. For the moment, we'll sweep 1160 * this under the carpet. 1161 * 1162 * The conservative default might not actually be a problem 1163 * if the only case this occurs is when sending an initial 1164 * SYN with options and data to a host we've never talked 1165 * to before. Then, they will reply with an MSS value which 1166 * will get recorded and the new parameters should get 1167 * recomputed. For Further Study. 1168 */ 1169 if (tp->t_maxopd <= mss) 1170 return; 1171 tp->t_maxopd = mss; 1172 1173 if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP && 1174 (tp->t_flags & TF_RCVD_TSTMP) == TF_RCVD_TSTMP) 1175 mss -= TCPOLEN_TSTAMP_APPA; 1176 if ((tp->t_flags & (TF_REQ_CC|TF_NOOPT)) == TF_REQ_CC && 1177 (tp->t_flags & TF_RCVD_CC) == TF_RCVD_CC) 1178 mss -= TCPOLEN_CC_APPA; 1179#if (MCLBYTES & (MCLBYTES - 1)) == 0 1180 if (mss > MCLBYTES) 1181 mss &= ~(MCLBYTES-1); 1182#else 1183 if (mss > MCLBYTES) 1184 mss = mss / MCLBYTES * MCLBYTES; 1185#endif 1186 if (so->so_snd.sb_hiwat < mss) 1187 mss = so->so_snd.sb_hiwat; 1188 1189 tp->t_maxseg = mss; 1190 1191 tcpstat.tcps_mturesent++; 1192 tp->t_rtttime = 0; 1193 tp->snd_nxt = tp->snd_una; 1194 tcp_output(tp); 1195 } 1196} 1197 1198/* 1199 * Look-up the routing entry to the peer of this inpcb. If no route 1200 * is found and it cannot be allocated the return NULL. This routine 1201 * is called by TCP routines that access the rmx structure and by tcp_mss 1202 * to get the interface MTU. 1203 */ 1204struct rtentry * 1205tcp_rtlookup(inp) 1206 struct inpcb *inp; 1207{ 1208 struct route *ro; 1209 struct rtentry *rt; 1210 1211 ro = &inp->inp_route; 1212 rt = ro->ro_rt; 1213 if (rt == NULL || !(rt->rt_flags & RTF_UP)) { 1214 /* No route yet, so try to acquire one */ 1215 if (inp->inp_faddr.s_addr != INADDR_ANY) { 1216 ro->ro_dst.sa_family = AF_INET; 1217 ro->ro_dst.sa_len = sizeof(ro->ro_dst); 1218 ((struct sockaddr_in *) &ro->ro_dst)->sin_addr = 1219 inp->inp_faddr; 1220 rtalloc(ro); 1221 rt = ro->ro_rt; 1222 } 1223 } 1224 return rt; 1225} 1226 1227#ifdef INET6 1228struct rtentry * 1229tcp_rtlookup6(inp) 1230 struct inpcb *inp; 1231{ 1232 struct route_in6 *ro6; 1233 struct rtentry *rt; 1234 1235 ro6 = &inp->in6p_route; 1236 rt = ro6->ro_rt; 1237 if (rt == NULL || !(rt->rt_flags & RTF_UP)) { 1238 /* No route yet, so try to acquire one */ 1239 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { 1240 ro6->ro_dst.sin6_family = AF_INET6; 1241 ro6->ro_dst.sin6_len = sizeof(ro6->ro_dst); 1242 ro6->ro_dst.sin6_addr = inp->in6p_faddr; 1243 rtalloc((struct route *)ro6); 1244 rt = ro6->ro_rt; 1245 } 1246 } 1247 return rt; 1248} 1249#endif /* INET6 */ 1250 1251#ifdef IPSEC 1252/* compute ESP/AH header size for TCP, including outer IP header. */ 1253size_t 1254ipsec_hdrsiz_tcp(tp) 1255 struct tcpcb *tp; 1256{ 1257 struct inpcb *inp; 1258 struct mbuf *m; 1259 size_t hdrsiz; 1260 struct ip *ip; 1261#ifdef INET6 1262 struct ip6_hdr *ip6; 1263#endif /* INET6 */ 1264 struct tcphdr *th; 1265 1266 if (!tp || !tp->t_template || !(inp = tp->t_inpcb)) 1267 return 0; 1268 MGETHDR(m, M_DONTWAIT, MT_DATA); 1269 if (!m) 1270 return 0; 1271 1272#ifdef INET6 1273 if ((inp->inp_vflag & INP_IPV6) != 0) { 1274 ip6 = mtod(m, struct ip6_hdr *); 1275 th = (struct tcphdr *)(ip6 + 1); 1276 m->m_pkthdr.len = m->m_len = 1277 sizeof(struct ip6_hdr) + sizeof(struct tcphdr); 1278 bcopy((caddr_t)tp->t_template->tt_ipgen, (caddr_t)ip6, 1279 sizeof(struct ip6_hdr)); 1280 bcopy((caddr_t)&tp->t_template->tt_t, (caddr_t)th, 1281 sizeof(struct tcphdr)); 1282 hdrsiz = ipsec6_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); 1283 } else 1284#endif /* INET6 */ 1285 { 1286 ip = mtod(m, struct ip *); 1287 th = (struct tcphdr *)(ip + 1); 1288 m->m_pkthdr.len = m->m_len = sizeof(struct tcpiphdr); 1289 bcopy((caddr_t)tp->t_template->tt_ipgen, (caddr_t)ip, 1290 sizeof(struct ip)); 1291 bcopy((caddr_t)&tp->t_template->tt_t, (caddr_t)th, 1292 sizeof(struct tcphdr)); 1293 hdrsiz = ipsec4_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); 1294 } 1295 1296 m_free(m); 1297 return hdrsiz; 1298} 1299#endif /*IPSEC*/ 1300 1301/* 1302 * Return a pointer to the cached information about the remote host. 1303 * The cached information is stored in the protocol specific part of 1304 * the route metrics. 1305 */ 1306struct rmxp_tao * 1307tcp_gettaocache(inp) 1308 struct inpcb *inp; 1309{ 1310 struct rtentry *rt; 1311 1312#ifdef INET6 1313 if ((inp->inp_vflag & INP_IPV6) != 0) 1314 rt = tcp_rtlookup6(inp); 1315 else 1316#endif /* INET6 */ 1317 rt = tcp_rtlookup(inp); 1318 1319 /* Make sure this is a host route and is up. */ 1320 if (rt == NULL || 1321 (rt->rt_flags & (RTF_UP|RTF_HOST)) != (RTF_UP|RTF_HOST)) 1322 return NULL; 1323 1324 return rmx_taop(rt->rt_rmx); 1325} 1326 1327/* 1328 * Clear all the TAO cache entries, called from tcp_init. 1329 * 1330 * XXX 1331 * This routine is just an empty one, because we assume that the routing 1332 * routing tables are initialized at the same time when TCP, so there is 1333 * nothing in the cache left over. 1334 */ 1335static void 1336tcp_cleartaocache() 1337{ 1338} 1339