tcp_timewait.c revision 56039
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 56039 2000-01-15 14:34:56Z shin $ 35 */ 36 37#include "opt_compat.h" 38#include "opt_inet6.h" 39#include "opt_tcpdebug.h" 40 41#include <sys/param.h> 42#include <sys/systm.h> 43#include <sys/callout.h> 44#include <sys/kernel.h> 45#include <sys/sysctl.h> 46#include <sys/malloc.h> 47#include <sys/mbuf.h> 48#ifdef INET6 49#include <sys/domain.h> 50#endif 51#include <sys/proc.h> 52#include <sys/socket.h> 53#include <sys/socketvar.h> 54#include <sys/protosw.h> 55 56#include <vm/vm_zone.h> 57 58#include <net/route.h> 59#include <net/if.h> 60 61#define _IP_VHL 62#include <netinet/in.h> 63#include <netinet/in_systm.h> 64#include <netinet/ip.h> 65#ifdef INET6 66#include <netinet/ip6.h> 67#endif 68#include <netinet/in_pcb.h> 69#ifdef INET6 70#include <netinet6/in6_pcb.h> 71#endif 72#include <netinet/in_var.h> 73#include <netinet/ip_var.h> 74#ifdef INET6 75#include <netinet6/ip6_var.h> 76#endif 77#include <netinet/tcp.h> 78#include <netinet/tcp_fsm.h> 79#include <netinet/tcp_seq.h> 80#include <netinet/tcp_timer.h> 81#include <netinet/tcp_var.h> 82#ifdef INET6 83#include <netinet6/tcp6_var.h> 84#endif 85#include <netinet/tcpip.h> 86#ifdef TCPDEBUG 87#include <netinet/tcp_debug.h> 88#endif 89#include <netinet6/ip6protosw.h> 90 91#ifdef IPSEC 92#include <netinet6/ipsec.h> 93#endif /*IPSEC*/ 94 95int tcp_mssdflt = TCP_MSS; 96SYSCTL_INT(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt, CTLFLAG_RW, 97 &tcp_mssdflt , 0, "Default TCP Maximum Segment Size"); 98 99#ifdef INET6 100int tcp_v6mssdflt = TCP6_MSS; 101SYSCTL_INT(_net_inet_tcp, TCPCTL_V6MSSDFLT, v6mssdflt, 102 CTLFLAG_RW, &tcp_v6mssdflt , 0, 103 "Default TCP Maximum Segment Size for IPv6"); 104#endif 105 106#if 0 107static int tcp_rttdflt = TCPTV_SRTTDFLT / PR_SLOWHZ; 108SYSCTL_INT(_net_inet_tcp, TCPCTL_RTTDFLT, rttdflt, CTLFLAG_RW, 109 &tcp_rttdflt , 0, "Default maximum TCP Round Trip Time"); 110#endif 111 112static int tcp_do_rfc1323 = 1; 113SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323, CTLFLAG_RW, 114 &tcp_do_rfc1323 , 0, "Enable rfc1323 (high performance TCP) extensions"); 115 116static int tcp_do_rfc1644 = 0; 117SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1644, rfc1644, CTLFLAG_RW, 118 &tcp_do_rfc1644 , 0, "Enable rfc1644 (TTCP) extensions"); 119 120static int tcp_tcbhashsize = 0; 121SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcbhashsize, CTLFLAG_RD, 122 &tcp_tcbhashsize, 0, "Size of TCP control-block hashtable"); 123 124static int do_tcpdrain = 1; 125SYSCTL_INT(_debug, OID_AUTO, do_tcpdrain, CTLFLAG_RW, &do_tcpdrain, 0, 126 "Enable non Net3 compliant tcp_drain"); 127 128SYSCTL_INT(_net_inet_tcp, OID_AUTO, pcbcount, CTLFLAG_RD, 129 &tcbinfo.ipi_count, 0, "Number of active PCBs"); 130 131static void tcp_cleartaocache __P((void)); 132static void tcp_notify __P((struct inpcb *, int)); 133 134/* 135 * Target size of TCP PCB hash tables. Must be a power of two. 136 * 137 * Note that this can be overridden by the kernel environment 138 * variable net.inet.tcp.tcbhashsize 139 */ 140#ifndef TCBHASHSIZE 141#define TCBHASHSIZE 512 142#endif 143 144/* 145 * This is the actual shape of what we allocate using the zone 146 * allocator. Doing it this way allows us to protect both structures 147 * using the same generation count, and also eliminates the overhead 148 * of allocating tcpcbs separately. By hiding the structure here, 149 * we avoid changing most of the rest of the code (although it needs 150 * to be changed, eventually, for greater efficiency). 151 */ 152#define ALIGNMENT 32 153#define ALIGNM1 (ALIGNMENT - 1) 154struct inp_tp { 155 union { 156 struct inpcb inp; 157 char align[(sizeof(struct inpcb) + ALIGNM1) & ~ALIGNM1]; 158 } inp_tp_u; 159 struct tcpcb tcb; 160 struct callout inp_tp_rexmt, inp_tp_persist, inp_tp_keep, inp_tp_2msl; 161 struct callout inp_tp_delack; 162}; 163#undef ALIGNMENT 164#undef ALIGNM1 165 166/* 167 * Tcp initialization 168 */ 169void 170tcp_init() 171{ 172 int hashsize; 173 174 tcp_iss = random(); /* wrong, but better than a constant */ 175 tcp_ccgen = 1; 176 tcp_cleartaocache(); 177 178 tcp_delacktime = TCPTV_DELACK; 179 tcp_keepinit = TCPTV_KEEP_INIT; 180 tcp_keepidle = TCPTV_KEEP_IDLE; 181 tcp_keepintvl = TCPTV_KEEPINTVL; 182 tcp_maxpersistidle = TCPTV_KEEP_IDLE; 183 tcp_msl = TCPTV_MSL; 184 185 LIST_INIT(&tcb); 186 tcbinfo.listhead = &tcb; 187 TUNABLE_INT_FETCH("net.inet.tcp.tcbhashsize", TCBHASHSIZE, hashsize); 188 if (!powerof2(hashsize)) { 189 printf("WARNING: TCB hash size not a power of 2\n"); 190 hashsize = 512; /* safe default */ 191 } 192 tcp_tcbhashsize = hashsize; 193 tcbinfo.hashbase = hashinit(hashsize, M_PCB, &tcbinfo.hashmask); 194 tcbinfo.porthashbase = hashinit(hashsize, M_PCB, 195 &tcbinfo.porthashmask); 196 tcbinfo.ipi_zone = zinit("tcpcb", sizeof(struct inp_tp), maxsockets, 197 ZONE_INTERRUPT, 0); 198#ifdef INET6 199#define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr)) 200#else /* INET6 */ 201#define TCP_MINPROTOHDR (sizeof(struct tcpiphdr)) 202#endif /* INET6 */ 203 if (max_protohdr < TCP_MINPROTOHDR) 204 max_protohdr = TCP_MINPROTOHDR; 205 if (max_linkhdr + TCP_MINPROTOHDR > MHLEN) 206 panic("tcp_init"); 207#undef TCP_MINPROTOHDR 208} 209 210/* 211 * Create template to be used to send tcp packets on a connection. 212 * Call after host entry created, allocates an mbuf and fills 213 * in a skeletal tcp/ip header, minimizing the amount of work 214 * necessary when the connection is used. 215 */ 216struct tcptemp * 217tcp_template(tp) 218 struct tcpcb *tp; 219{ 220 register struct inpcb *inp = tp->t_inpcb; 221 register struct mbuf *m; 222 register struct tcptemp *n; 223 224 if ((n = tp->t_template) == 0) { 225 m = m_get(M_DONTWAIT, MT_HEADER); 226 if (m == NULL) 227 return (0); 228 m->m_len = sizeof (struct tcptemp); 229 n = mtod(m, struct tcptemp *); 230 } 231#ifdef INET6 232 if ((inp->inp_vflag & INP_IPV6) != 0) { 233 register struct ip6_hdr *ip6; 234 235 ip6 = (struct ip6_hdr *)n->tt_ipgen; 236 ip6->ip6_flow = (ip6->ip6_flow & ~IPV6_FLOWINFO_MASK) | 237 (inp->in6p_flowinfo & IPV6_FLOWINFO_MASK); 238 ip6->ip6_vfc = (ip6->ip6_vfc & ~IPV6_VERSION_MASK) | 239 (IPV6_VERSION & IPV6_VERSION_MASK); 240 ip6->ip6_nxt = IPPROTO_TCP; 241 ip6->ip6_plen = sizeof(struct tcphdr); 242 ip6->ip6_src = inp->in6p_laddr; 243 ip6->ip6_dst = inp->in6p_faddr; 244 } else 245#endif 246 { 247 register struct ipovly *ipov; 248 249 ipov = (struct ipovly *)n->tt_ipgen; 250 bzero(ipov->ih_x1, sizeof(ipov->ih_x1)); 251 ipov->ih_pr = IPPROTO_TCP; 252 ipov->ih_len = htons(sizeof (struct tcpiphdr) - sizeof (struct ip)); 253 ipov->ih_src = inp->inp_laddr; 254 ipov->ih_dst = inp->inp_faddr; 255 } 256 n->tt_t.th_sport = inp->inp_lport; 257 n->tt_t.th_dport = inp->inp_fport; 258 n->tt_t.th_seq = 0; 259 n->tt_t.th_ack = 0; 260 n->tt_t.th_x2 = 0; 261 n->tt_t.th_off = 5; 262 n->tt_t.th_flags = 0; 263 n->tt_t.th_win = 0; 264 n->tt_t.th_sum = 0; 265 n->tt_t.th_urp = 0; 266 return (n); 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 * NOTE: If m != NULL, then ti must point to *inside* the mbuf. 283 */ 284void 285tcp_respond(tp, ipgen, th, m, ack, seq, flags) 286 struct tcpcb *tp; 287 void *ipgen; 288 register struct tcphdr *th; 289 register struct mbuf *m; 290 tcp_seq ack, seq; 291 int flags; 292{ 293 register int tlen; 294 int win = 0; 295 struct route *ro = 0; 296 struct route sro; 297 struct ip *ip; 298 struct ipovly *ipov; 299 struct tcphdr *nth; 300#ifdef INET6 301 struct route_in6 *ro6 = 0; 302 struct route_in6 sro6; 303 struct ip6_hdr *ip6; 304 int isipv6; 305#endif /* INET6 */ 306 int ipflags = 0; 307 308#ifdef INET6 309 isipv6 = IP_VHL_V(((struct ip *)ipgen)->ip_vhl) == 6; 310 ip6 = ipgen; 311#endif /* INET6 */ 312 ip = ipgen; 313 ipov = ipgen; 314 315 if (tp) { 316 if (!(flags & TH_RST)) 317 win = sbspace(&tp->t_inpcb->inp_socket->so_rcv); 318#ifdef INET6 319 if (isipv6) 320 ro6 = &tp->t_inpcb->in6p_route; 321 else 322#endif /* INET6 */ 323 ro = &tp->t_inpcb->inp_route; 324 } else { 325#ifdef INET6 326 if (isipv6) { 327 ro6 = &sro6; 328 bzero(ro6, sizeof *ro6); 329 } else 330#endif /* INET6 */ 331 { 332 ro = &sro; 333 bzero(ro, sizeof *ro); 334 } 335 } 336 if (m == 0) { 337 m = m_gethdr(M_DONTWAIT, MT_HEADER); 338 if (m == NULL) 339 return; 340#ifdef TCP_COMPAT_42 341 tlen = 1; 342#else 343 tlen = 0; 344#endif 345 m->m_data += max_linkhdr; 346#ifdef INET6 347 if (isipv6) { 348 bcopy((caddr_t)ip6, mtod(m, caddr_t), 349 sizeof(struct ip6_hdr)); 350 ip6 = mtod(m, struct ip6_hdr *); 351 nth = (struct tcphdr *)(ip6 + 1); 352 } else 353#endif /* INET6 */ 354 { 355 bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 356 ip = mtod(m, struct ip *); 357 ipov = mtod(m, struct ipovly *); 358 nth = (struct tcphdr *)(ip + 1); 359 } 360 bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr)); 361 flags = TH_ACK; 362 } else { 363 m_freem(m->m_next); 364 m->m_next = 0; 365 m->m_data = (caddr_t)ipgen; 366 /* m_len is set later */ 367 tlen = 0; 368#define xchg(a,b,type) { type t; t=a; a=b; b=t; } 369#ifdef INET6 370 if (isipv6) { 371 xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr); 372 nth = (struct tcphdr *)(ip6 + 1); 373 } else 374#endif /* INET6 */ 375 { 376 xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, n_long); 377 nth = (struct tcphdr *)(ip + 1); 378 } 379 if (th != nth) { 380 /* 381 * this is usually a case when an extension header 382 * exists between the IPv6 header and the 383 * TCP header. 384 */ 385 nth->th_sport = th->th_sport; 386 nth->th_dport = th->th_dport; 387 } 388 xchg(nth->th_dport, nth->th_sport, n_short); 389#undef xchg 390 } 391#ifdef INET6 392 if (isipv6) { 393 ip6->ip6_plen = htons((u_short)(sizeof (struct tcphdr) + 394 tlen)); 395 tlen += sizeof (struct ip6_hdr) + sizeof (struct tcphdr); 396 } else 397#endif 398 { 399 ipov->ih_len = htons((u_short)(sizeof (struct tcphdr) + tlen)); 400 tlen += sizeof (struct tcpiphdr); 401 } 402 m->m_len = tlen; 403 m->m_pkthdr.len = tlen; 404 m->m_pkthdr.rcvif = (struct ifnet *) 0; 405 nth->th_seq = htonl(seq); 406 nth->th_ack = htonl(ack); 407 nth->th_x2 = 0; 408 nth->th_off = sizeof (struct tcphdr) >> 2; 409 nth->th_flags = flags; 410 if (tp) 411 nth->th_win = htons((u_short) (win >> tp->rcv_scale)); 412 else 413 nth->th_win = htons((u_short)win); 414 nth->th_urp = 0; 415 nth->th_sum = 0; 416#ifdef INET6 417 if (isipv6) { 418 nth->th_sum = in6_cksum(m, IPPROTO_TCP, 419 sizeof(struct ip6_hdr), 420 tlen - sizeof(struct ip6_hdr)); 421 ip6->ip6_hlim = in6_selecthlim(tp ? tp->t_inpcb : NULL, 422 ro6 && ro6->ro_rt ? 423 ro6->ro_rt->rt_ifp : 424 NULL); 425 } else 426#endif /* INET6 */ 427 { 428 bzero(ipov->ih_x1, sizeof(ipov->ih_x1)); 429 nth->th_sum = in_cksum(m, tlen); 430#ifdef INET6 431 /* Re-initialization for later version check */ 432 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, 0); 433#endif /* INET6 */ 434 ip->ip_len = tlen; 435 ip->ip_ttl = ip_defttl; 436 } 437#ifdef TCPDEBUG 438 if (tp == NULL || (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) 439 tcp_trace(TA_OUTPUT, 0, tp, mtod(m, void *), th, 0); 440#endif 441#ifdef IPSEC 442 if (tp != NULL) { 443 m->m_pkthdr.rcvif = (struct ifnet *)tp->t_inpcb->inp_socket; 444 ipflags |= 445#ifdef INET6 446 isipv6 ? IPV6_SOCKINMRCVIF : 447#endif 448 IP_SOCKINMRCVIF; 449 } 450#endif 451#ifdef INET6 452 if (isipv6) { 453 (void)ip6_output(m, NULL, ro6, ipflags, NULL, NULL); 454 if (ro6 == &sro6 && ro6->ro_rt) { 455 RTFREE(ro6->ro_rt); 456 ro6->ro_rt = NULL; 457 } 458 } else 459#endif /* INET6 */ 460 { 461 (void) ip_output(m, NULL, ro, ipflags, NULL); 462 if (ro == &sro && ro->ro_rt) { 463 RTFREE(ro->ro_rt); 464 ro->ro_rt = NULL; 465 } 466 } 467} 468 469/* 470 * Create a new TCP control block, making an 471 * empty reassembly queue and hooking it to the argument 472 * protocol control block. The `inp' parameter must have 473 * come from the zone allocator set up in tcp_init(). 474 */ 475struct tcpcb * 476tcp_newtcpcb(inp) 477 struct inpcb *inp; 478{ 479 struct inp_tp *it; 480 register struct tcpcb *tp; 481#ifdef INET6 482 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; 483#endif /* INET6 */ 484 485 it = (struct inp_tp *)inp; 486 tp = &it->tcb; 487 bzero((char *) tp, sizeof(struct tcpcb)); 488 LIST_INIT(&tp->t_segq); 489 tp->t_maxseg = tp->t_maxopd = 490#ifdef INET6 491 isipv6 ? tcp_v6mssdflt : 492#endif /* INET6 */ 493 tcp_mssdflt; 494 495 /* Set up our timeouts. */ 496 callout_init(tp->tt_rexmt = &it->inp_tp_rexmt); 497 callout_init(tp->tt_persist = &it->inp_tp_persist); 498 callout_init(tp->tt_keep = &it->inp_tp_keep); 499 callout_init(tp->tt_2msl = &it->inp_tp_2msl); 500 callout_init(tp->tt_delack = &it->inp_tp_delack); 501 502 if (tcp_do_rfc1323) 503 tp->t_flags = (TF_REQ_SCALE|TF_REQ_TSTMP); 504 if (tcp_do_rfc1644) 505 tp->t_flags |= TF_REQ_CC; 506 tp->t_inpcb = inp; /* XXX */ 507 /* 508 * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no 509 * rtt estimate. Set rttvar so that srtt + 4 * rttvar gives 510 * reasonable initial retransmit time. 511 */ 512 tp->t_srtt = TCPTV_SRTTBASE; 513 tp->t_rttvar = ((TCPTV_RTOBASE - TCPTV_SRTTBASE) << TCP_RTTVAR_SHIFT) / 4; 514 tp->t_rttmin = TCPTV_MIN; 515 tp->t_rxtcur = TCPTV_RTOBASE; 516 tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; 517 tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT; 518 tp->t_rcvtime = ticks; 519#ifdef INET6 520 if (isipv6 != 0) 521 inp->in6p_ip6_hlim = in6_selecthlim(inp, 522 inp->in6p_route.ro_rt ? 523 inp->in6p_route.ro_rt->rt_ifp : 524 NULL); 525 else 526#endif 527 inp->inp_ip_ttl = ip_defttl; 528 inp->inp_ppcb = (caddr_t)tp; 529 return (tp); /* XXX */ 530} 531 532/* 533 * Drop a TCP connection, reporting 534 * the specified error. If connection is synchronized, 535 * then send a RST to peer. 536 */ 537struct tcpcb * 538tcp_drop(tp, errno) 539 register struct tcpcb *tp; 540 int errno; 541{ 542 struct socket *so = tp->t_inpcb->inp_socket; 543 544 if (TCPS_HAVERCVDSYN(tp->t_state)) { 545 tp->t_state = TCPS_CLOSED; 546 (void) tcp_output(tp); 547 tcpstat.tcps_drops++; 548 } else 549 tcpstat.tcps_conndrops++; 550 if (errno == ETIMEDOUT && tp->t_softerror) 551 errno = tp->t_softerror; 552 so->so_error = errno; 553 return (tcp_close(tp)); 554} 555 556/* 557 * Close a TCP control block: 558 * discard all space held by the tcp 559 * discard internet protocol block 560 * wake up any sleepers 561 */ 562struct tcpcb * 563tcp_close(tp) 564 register struct tcpcb *tp; 565{ 566 register struct tseg_qent *q; 567 struct inpcb *inp = tp->t_inpcb; 568 struct socket *so = inp->inp_socket; 569#ifdef INET6 570 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; 571#endif /* INET6 */ 572 register struct rtentry *rt; 573 int dosavessthresh; 574 575 /* 576 * Make sure that all of our timers are stopped before we 577 * delete the PCB. 578 */ 579 callout_stop(tp->tt_rexmt); 580 callout_stop(tp->tt_persist); 581 callout_stop(tp->tt_keep); 582 callout_stop(tp->tt_2msl); 583 callout_stop(tp->tt_delack); 584 585 /* 586 * If we got enough samples through the srtt filter, 587 * save the rtt and rttvar in the routing entry. 588 * 'Enough' is arbitrarily defined as the 16 samples. 589 * 16 samples is enough for the srtt filter to converge 590 * to within 5% of the correct value; fewer samples and 591 * we could save a very bogus rtt. 592 * 593 * Don't update the default route's characteristics and don't 594 * update anything that the user "locked". 595 */ 596 if (tp->t_rttupdated >= 16) { 597 register u_long i = 0; 598#ifdef INET6 599 if (isipv6) { 600 struct sockaddr_in6 *sin6; 601 602 if ((rt = inp->in6p_route.ro_rt) == NULL) 603 goto no_valid_rt; 604 sin6 = (struct sockaddr_in6 *)rt_key(rt); 605 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) 606 goto no_valid_rt; 607 } 608 else 609#endif /* INET6 */ 610 if ((rt = inp->inp_route.ro_rt) == NULL || 611 ((struct sockaddr_in *)rt_key(rt))->sin_addr.s_addr 612 == INADDR_ANY) 613 goto no_valid_rt; 614 615 if ((rt->rt_rmx.rmx_locks & RTV_RTT) == 0) { 616 i = tp->t_srtt * 617 (RTM_RTTUNIT / (hz * TCP_RTT_SCALE)); 618 if (rt->rt_rmx.rmx_rtt && i) 619 /* 620 * filter this update to half the old & half 621 * the new values, converting scale. 622 * See route.h and tcp_var.h for a 623 * description of the scaling constants. 624 */ 625 rt->rt_rmx.rmx_rtt = 626 (rt->rt_rmx.rmx_rtt + i) / 2; 627 else 628 rt->rt_rmx.rmx_rtt = i; 629 tcpstat.tcps_cachedrtt++; 630 } 631 if ((rt->rt_rmx.rmx_locks & RTV_RTTVAR) == 0) { 632 i = tp->t_rttvar * 633 (RTM_RTTUNIT / (hz * TCP_RTTVAR_SCALE)); 634 if (rt->rt_rmx.rmx_rttvar && i) 635 rt->rt_rmx.rmx_rttvar = 636 (rt->rt_rmx.rmx_rttvar + i) / 2; 637 else 638 rt->rt_rmx.rmx_rttvar = i; 639 tcpstat.tcps_cachedrttvar++; 640 } 641 /* 642 * The old comment here said: 643 * update the pipelimit (ssthresh) if it has been updated 644 * already or if a pipesize was specified & the threshhold 645 * got below half the pipesize. I.e., wait for bad news 646 * before we start updating, then update on both good 647 * and bad news. 648 * 649 * But we want to save the ssthresh even if no pipesize is 650 * specified explicitly in the route, because such 651 * connections still have an implicit pipesize specified 652 * by the global tcp_sendspace. In the absence of a reliable 653 * way to calculate the pipesize, it will have to do. 654 */ 655 i = tp->snd_ssthresh; 656 if (rt->rt_rmx.rmx_sendpipe != 0) 657 dosavessthresh = (i < rt->rt_rmx.rmx_sendpipe / 2); 658 else 659 dosavessthresh = (i < so->so_snd.sb_hiwat / 2); 660 if (((rt->rt_rmx.rmx_locks & RTV_SSTHRESH) == 0 && 661 i != 0 && rt->rt_rmx.rmx_ssthresh != 0) 662 || dosavessthresh) { 663 /* 664 * convert the limit from user data bytes to 665 * packets then to packet data bytes. 666 */ 667 i = (i + tp->t_maxseg / 2) / tp->t_maxseg; 668 if (i < 2) 669 i = 2; 670 i *= (u_long)(tp->t_maxseg + 671#ifdef INET6 672 (isipv6 ? sizeof (struct ip6_hdr) + 673 sizeof (struct tcphdr) : 674#endif 675 sizeof (struct tcpiphdr) 676#ifdef INET6 677 ) 678#endif 679 ); 680 if (rt->rt_rmx.rmx_ssthresh) 681 rt->rt_rmx.rmx_ssthresh = 682 (rt->rt_rmx.rmx_ssthresh + i) / 2; 683 else 684 rt->rt_rmx.rmx_ssthresh = i; 685 tcpstat.tcps_cachedssthresh++; 686 } 687 } 688 no_valid_rt: 689 /* free the reassembly queue, if any */ 690 while((q = LIST_FIRST(&tp->t_segq)) != NULL) { 691 LIST_REMOVE(q, tqe_q); 692 m_freem(q->tqe_m); 693 FREE(q, M_TSEGQ); 694 } 695 if (tp->t_template) 696 (void) m_free(dtom(tp->t_template)); 697 inp->inp_ppcb = NULL; 698 soisdisconnected(so); 699#ifdef INET6 700 if (INP_CHECK_SOCKAF(so, AF_INET6)) 701 in6_pcbdetach(inp); 702 else 703#endif /* INET6 */ 704 in_pcbdetach(inp); 705 tcpstat.tcps_closed++; 706 return ((struct tcpcb *)0); 707} 708 709void 710tcp_drain() 711{ 712 if (do_tcpdrain) 713 { 714 struct inpcb *inpb; 715 struct tcpcb *tcpb; 716 struct tseg_qent *te; 717 718 /* 719 * Walk the tcpbs, if existing, and flush the reassembly queue, 720 * if there is one... 721 * XXX: The "Net/3" implementation doesn't imply that the TCP 722 * reassembly queue should be flushed, but in a situation 723 * where we're really low on mbufs, this is potentially 724 * usefull. 725 */ 726 for (inpb = tcbinfo.listhead->lh_first; inpb; 727 inpb = inpb->inp_list.le_next) { 728 if ((tcpb = intotcpcb(inpb))) { 729 while ((te = LIST_FIRST(&tcpb->t_segq)) 730 != NULL) { 731 LIST_REMOVE(te, tqe_q); 732 m_freem(te->tqe_m); 733 FREE(te, M_TSEGQ); 734 } 735 } 736 } 737 738 } 739} 740 741/* 742 * Notify a tcp user of an asynchronous error; 743 * store error as soft error, but wake up user 744 * (for now, won't do anything until can select for soft error). 745 */ 746static void 747tcp_notify(inp, error) 748 struct inpcb *inp; 749 int error; 750{ 751 register struct tcpcb *tp = (struct tcpcb *)inp->inp_ppcb; 752 register struct socket *so = inp->inp_socket; 753 754 /* 755 * Ignore some errors if we are hooked up. 756 * If connection hasn't completed, has retransmitted several times, 757 * and receives a second error, give up now. This is better 758 * than waiting a long time to establish a connection that 759 * can never complete. 760 */ 761 if (tp->t_state == TCPS_ESTABLISHED && 762 (error == EHOSTUNREACH || error == ENETUNREACH || 763 error == EHOSTDOWN)) { 764 return; 765 } else if (tp->t_state < TCPS_ESTABLISHED && tp->t_rxtshift > 3 && 766 tp->t_softerror) 767 so->so_error = error; 768 else 769 tp->t_softerror = error; 770 wakeup((caddr_t) &so->so_timeo); 771 sorwakeup(so); 772 sowwakeup(so); 773} 774 775static int 776tcp_pcblist SYSCTL_HANDLER_ARGS 777{ 778 int error, i, n, s; 779 struct inpcb *inp, **inp_list; 780 inp_gen_t gencnt; 781 struct xinpgen xig; 782 783 /* 784 * The process of preparing the TCB list is too time-consuming and 785 * resource-intensive to repeat twice on every request. 786 */ 787 if (req->oldptr == 0) { 788 n = tcbinfo.ipi_count; 789 req->oldidx = 2 * (sizeof xig) 790 + (n + n/8) * sizeof(struct xtcpcb); 791 return 0; 792 } 793 794 if (req->newptr != 0) 795 return EPERM; 796 797 /* 798 * OK, now we're committed to doing something. 799 */ 800 s = splnet(); 801 gencnt = tcbinfo.ipi_gencnt; 802 n = tcbinfo.ipi_count; 803 splx(s); 804 805 xig.xig_len = sizeof xig; 806 xig.xig_count = n; 807 xig.xig_gen = gencnt; 808 xig.xig_sogen = so_gencnt; 809 error = SYSCTL_OUT(req, &xig, sizeof xig); 810 if (error) 811 return error; 812 813 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 814 if (inp_list == 0) 815 return ENOMEM; 816 817 s = splnet(); 818 for (inp = tcbinfo.listhead->lh_first, i = 0; inp && i < n; 819 inp = inp->inp_list.le_next) { 820 if (inp->inp_gencnt <= gencnt && !prison_xinpcb(req->p, inp)) 821 inp_list[i++] = inp; 822 } 823 splx(s); 824 n = i; 825 826 error = 0; 827 for (i = 0; i < n; i++) { 828 inp = inp_list[i]; 829 if (inp->inp_gencnt <= gencnt) { 830 struct xtcpcb xt; 831 caddr_t inp_ppcb; 832 xt.xt_len = sizeof xt; 833 /* XXX should avoid extra copy */ 834 bcopy(inp, &xt.xt_inp, sizeof *inp); 835 inp_ppcb = inp->inp_ppcb; 836 if (inp_ppcb != NULL) 837 bcopy(inp_ppcb, &xt.xt_tp, sizeof xt.xt_tp); 838 else 839 bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); 840 if (inp->inp_socket) 841 sotoxsocket(inp->inp_socket, &xt.xt_socket); 842 error = SYSCTL_OUT(req, &xt, sizeof xt); 843 } 844 } 845 if (!error) { 846 /* 847 * Give the user an updated idea of our state. 848 * If the generation differs from what we told 849 * her before, she knows that something happened 850 * while we were processing this request, and it 851 * might be necessary to retry. 852 */ 853 s = splnet(); 854 xig.xig_gen = tcbinfo.ipi_gencnt; 855 xig.xig_sogen = so_gencnt; 856 xig.xig_count = tcbinfo.ipi_count; 857 splx(s); 858 error = SYSCTL_OUT(req, &xig, sizeof xig); 859 } 860 free(inp_list, M_TEMP); 861 return error; 862} 863 864SYSCTL_PROC(_net_inet_tcp, TCPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0, 865 tcp_pcblist, "S,xtcpcb", "List of active TCP connections"); 866 867static int 868tcp_getcred SYSCTL_HANDLER_ARGS 869{ 870 struct sockaddr_in addrs[2]; 871 struct inpcb *inp; 872 int error, s; 873 874 error = suser(req->p); 875 if (error) 876 return (error); 877 error = SYSCTL_IN(req, addrs, sizeof(addrs)); 878 if (error) 879 return (error); 880 s = splnet(); 881 inp = in_pcblookup_hash(&tcbinfo, addrs[1].sin_addr, addrs[1].sin_port, 882 addrs[0].sin_addr, addrs[0].sin_port, 0, NULL); 883 if (inp == NULL || inp->inp_socket == NULL) { 884 error = ENOENT; 885 goto out; 886 } 887 error = SYSCTL_OUT(req, inp->inp_socket->so_cred, sizeof(struct ucred)); 888out: 889 splx(s); 890 return (error); 891} 892 893SYSCTL_PROC(_net_inet_tcp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 894 0, 0, tcp_getcred, "S,ucred", "Get the ucred of a TCP connection"); 895 896#ifdef INET6 897static int 898tcp6_getcred SYSCTL_HANDLER_ARGS 899{ 900 struct sockaddr_in6 addrs[2]; 901 struct inpcb *inp; 902 int error, s, mapped = 0; 903 904 error = suser(req->p); 905 if (error) 906 return (error); 907 error = SYSCTL_IN(req, addrs, sizeof(addrs)); 908 if (error) 909 return (error); 910 if (IN6_IS_ADDR_V4MAPPED(&addrs[0].sin6_addr)) { 911 if (IN6_IS_ADDR_V4MAPPED(&addrs[1].sin6_addr)) 912 mapped = 1; 913 else 914 return (EINVAL); 915 } 916 s = splnet(); 917 if (mapped == 1) 918 inp = in_pcblookup_hash(&tcbinfo, 919 *(struct in_addr *)&addrs[1].sin6_addr.s6_addr[12], 920 addrs[1].sin6_port, 921 *(struct in_addr *)&addrs[0].sin6_addr.s6_addr[12], 922 addrs[0].sin6_port, 923 0, NULL); 924 else 925 inp = in6_pcblookup_hash(&tcbinfo, &addrs[1].sin6_addr, 926 addrs[1].sin6_port, 927 &addrs[0].sin6_addr, addrs[0].sin6_port, 928 0, NULL); 929 if (inp == NULL || inp->inp_socket == NULL) { 930 error = ENOENT; 931 goto out; 932 } 933 error = SYSCTL_OUT(req, inp->inp_socket->so_cred, 934 sizeof(struct ucred)); 935out: 936 splx(s); 937 return (error); 938} 939 940SYSCTL_PROC(_net_inet6_tcp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 941 0, 0, 942 tcp6_getcred, "S,ucred", "Get the ucred of a TCP6 connection"); 943#endif 944 945 946void 947tcp_ctlinput(cmd, sa, vip) 948 int cmd; 949 struct sockaddr *sa; 950 void *vip; 951{ 952 register struct ip *ip = vip; 953 register struct tcphdr *th; 954 void (*notify) __P((struct inpcb *, int)) = tcp_notify; 955 956 if (cmd == PRC_QUENCH) 957 notify = tcp_quench; 958 else if (cmd == PRC_MSGSIZE) 959 notify = tcp_mtudisc; 960 else if (!PRC_IS_REDIRECT(cmd) && 961 ((unsigned)cmd > PRC_NCMDS || inetctlerrmap[cmd] == 0)) 962 return; 963 if (ip) { 964 th = (struct tcphdr *)((caddr_t)ip 965 + (IP_VHL_HL(ip->ip_vhl) << 2)); 966 in_pcbnotify(&tcb, sa, th->th_dport, ip->ip_src, th->th_sport, 967 cmd, notify); 968 } else 969 in_pcbnotify(&tcb, sa, 0, zeroin_addr, 0, cmd, notify); 970} 971 972#ifdef INET6 973void 974tcp6_ctlinput(cmd, sa, d) 975 int cmd; 976 struct sockaddr *sa; 977 void *d; 978{ 979 register struct tcphdr *thp; 980 struct tcphdr th; 981 void (*notify) __P((struct inpcb *, int)) = tcp_notify; 982 struct sockaddr_in6 sa6; 983 struct ip6_hdr *ip6; 984 struct mbuf *m; 985 int off; 986 987 if (sa->sa_family != AF_INET6 || 988 sa->sa_len != sizeof(struct sockaddr_in6)) 989 return; 990 991 if (cmd == PRC_QUENCH) 992 notify = tcp_quench; 993 else if (cmd == PRC_MSGSIZE) 994 notify = tcp_mtudisc; 995 else if (!PRC_IS_REDIRECT(cmd) && 996 ((unsigned)cmd > PRC_NCMDS || inet6ctlerrmap[cmd] == 0)) 997 return; 998 999 /* if the parameter is from icmp6, decode it. */ 1000 if (d != NULL) { 1001 struct ip6ctlparam *ip6cp = (struct ip6ctlparam *)d; 1002 m = ip6cp->ip6c_m; 1003 ip6 = ip6cp->ip6c_ip6; 1004 off = ip6cp->ip6c_off; 1005 } else { 1006 m = NULL; 1007 ip6 = NULL; 1008 } 1009 1010 /* 1011 * Translate addresses into internal form. 1012 * Sa check if it is AF_INET6 is done at the top of this funciton. 1013 */ 1014 sa6 = *(struct sockaddr_in6 *)sa; 1015 if (IN6_IS_ADDR_LINKLOCAL(&sa6.sin6_addr) != 0 && m != NULL && 1016 m->m_pkthdr.rcvif != NULL) 1017 sa6.sin6_addr.s6_addr16[1] = htons(m->m_pkthdr.rcvif->if_index); 1018 1019 if (ip6) { 1020 /* 1021 * XXX: We assume that when IPV6 is non NULL, 1022 * M and OFF are valid. 1023 */ 1024 struct in6_addr s; 1025 1026 /* translate addresses into internal form */ 1027 memcpy(&s, &ip6->ip6_src, sizeof(s)); 1028 if (IN6_IS_ADDR_LINKLOCAL(&s) != 0 && m != NULL && 1029 m->m_pkthdr.rcvif != NULL) 1030 s.s6_addr16[1] = htons(m->m_pkthdr.rcvif->if_index); 1031 1032 if (m->m_len < off + sizeof(*thp)) { 1033 /* 1034 * this should be rare case 1035 * because now MINCLSIZE is "(MHLEN + 1)", 1036 * so we compromise on this copy... 1037 */ 1038 m_copydata(m, off, sizeof(th), (caddr_t)&th); 1039 thp = &th; 1040 } else 1041 thp = (struct tcphdr *)(mtod(m, caddr_t) + off); 1042 in6_pcbnotify(&tcb, (struct sockaddr *)&sa6, thp->th_dport, 1043 &s, thp->th_sport, cmd, notify); 1044 } else 1045 in6_pcbnotify(&tcb, (struct sockaddr *)&sa6, 0, &zeroin6_addr, 1046 0, cmd, notify); 1047} 1048#endif /* INET6 */ 1049 1050/* 1051 * When a source quench is received, close congestion window 1052 * to one segment. We will gradually open it again as we proceed. 1053 */ 1054void 1055tcp_quench(inp, errno) 1056 struct inpcb *inp; 1057 int errno; 1058{ 1059 struct tcpcb *tp = intotcpcb(inp); 1060 1061 if (tp) 1062 tp->snd_cwnd = tp->t_maxseg; 1063} 1064 1065/* 1066 * When `need fragmentation' ICMP is received, update our idea of the MSS 1067 * based on the new value in the route. Also nudge TCP to send something, 1068 * since we know the packet we just sent was dropped. 1069 * This duplicates some code in the tcp_mss() function in tcp_input.c. 1070 */ 1071void 1072tcp_mtudisc(inp, errno) 1073 struct inpcb *inp; 1074 int errno; 1075{ 1076 struct tcpcb *tp = intotcpcb(inp); 1077 struct rtentry *rt; 1078 struct rmxp_tao *taop; 1079 struct socket *so = inp->inp_socket; 1080 int offered; 1081 int mss; 1082#ifdef INET6 1083 int isipv6 = (tp->t_inpcb->inp_vflag & INP_IPV6) != 0; 1084#endif /* INET6 */ 1085 1086 if (tp) { 1087#ifdef INET6 1088 if (isipv6) 1089 rt = tcp_rtlookup6(inp); 1090 else 1091#endif /* INET6 */ 1092 rt = tcp_rtlookup(inp); 1093 if (!rt || !rt->rt_rmx.rmx_mtu) { 1094 tp->t_maxopd = tp->t_maxseg = 1095#ifdef INET6 1096 isipv6 ? tcp_v6mssdflt : 1097#endif /* INET6 */ 1098 tcp_mssdflt; 1099 return; 1100 } 1101 taop = rmx_taop(rt->rt_rmx); 1102 offered = taop->tao_mssopt; 1103 mss = rt->rt_rmx.rmx_mtu - 1104#ifdef INET6 1105 (isipv6 ? 1106 sizeof(struct ip6_hdr) + sizeof(struct tcphdr) : 1107#endif /* INET6 */ 1108 sizeof(struct tcpiphdr) 1109#ifdef INET6 1110 ) 1111#endif /* INET6 */ 1112 ; 1113 1114 if (offered) 1115 mss = min(mss, offered); 1116 /* 1117 * XXX - The above conditional probably violates the TCP 1118 * spec. The problem is that, since we don't know the 1119 * other end's MSS, we are supposed to use a conservative 1120 * default. But, if we do that, then MTU discovery will 1121 * never actually take place, because the conservative 1122 * default is much less than the MTUs typically seen 1123 * on the Internet today. For the moment, we'll sweep 1124 * this under the carpet. 1125 * 1126 * The conservative default might not actually be a problem 1127 * if the only case this occurs is when sending an initial 1128 * SYN with options and data to a host we've never talked 1129 * to before. Then, they will reply with an MSS value which 1130 * will get recorded and the new parameters should get 1131 * recomputed. For Further Study. 1132 */ 1133 if (tp->t_maxopd <= mss) 1134 return; 1135 tp->t_maxopd = mss; 1136 1137 if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP && 1138 (tp->t_flags & TF_RCVD_TSTMP) == TF_RCVD_TSTMP) 1139 mss -= TCPOLEN_TSTAMP_APPA; 1140 if ((tp->t_flags & (TF_REQ_CC|TF_NOOPT)) == TF_REQ_CC && 1141 (tp->t_flags & TF_RCVD_CC) == TF_RCVD_CC) 1142 mss -= TCPOLEN_CC_APPA; 1143#if (MCLBYTES & (MCLBYTES - 1)) == 0 1144 if (mss > MCLBYTES) 1145 mss &= ~(MCLBYTES-1); 1146#else 1147 if (mss > MCLBYTES) 1148 mss = mss / MCLBYTES * MCLBYTES; 1149#endif 1150 if (so->so_snd.sb_hiwat < mss) 1151 mss = so->so_snd.sb_hiwat; 1152 1153 tp->t_maxseg = mss; 1154 1155 tcpstat.tcps_mturesent++; 1156 tp->t_rtttime = 0; 1157 tp->snd_nxt = tp->snd_una; 1158 tcp_output(tp); 1159 } 1160} 1161 1162/* 1163 * Look-up the routing entry to the peer of this inpcb. If no route 1164 * is found and it cannot be allocated the return NULL. This routine 1165 * is called by TCP routines that access the rmx structure and by tcp_mss 1166 * to get the interface MTU. 1167 */ 1168struct rtentry * 1169tcp_rtlookup(inp) 1170 struct inpcb *inp; 1171{ 1172 struct route *ro; 1173 struct rtentry *rt; 1174 1175 ro = &inp->inp_route; 1176 rt = ro->ro_rt; 1177 if (rt == NULL || !(rt->rt_flags & RTF_UP)) { 1178 /* No route yet, so try to acquire one */ 1179 if (inp->inp_faddr.s_addr != INADDR_ANY) { 1180 ro->ro_dst.sa_family = AF_INET; 1181 ro->ro_dst.sa_len = sizeof(ro->ro_dst); 1182 ((struct sockaddr_in *) &ro->ro_dst)->sin_addr = 1183 inp->inp_faddr; 1184 rtalloc(ro); 1185 rt = ro->ro_rt; 1186 } 1187 } 1188 return rt; 1189} 1190 1191#ifdef INET6 1192struct rtentry * 1193tcp_rtlookup6(inp) 1194 struct inpcb *inp; 1195{ 1196 struct route_in6 *ro6; 1197 struct rtentry *rt; 1198 1199 ro6 = &inp->in6p_route; 1200 rt = ro6->ro_rt; 1201 if (rt == NULL || !(rt->rt_flags & RTF_UP)) { 1202 /* No route yet, so try to acquire one */ 1203 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { 1204 ro6->ro_dst.sin6_family = AF_INET6; 1205 ro6->ro_dst.sin6_len = sizeof(ro6->ro_dst); 1206 ro6->ro_dst.sin6_addr = inp->in6p_faddr; 1207 rtalloc((struct route *)ro6); 1208 rt = ro6->ro_rt; 1209 } 1210 } 1211 return rt; 1212} 1213#endif /* INET6 */ 1214 1215#ifdef IPSEC 1216/* compute ESP/AH header size for TCP, including outer IP header. */ 1217size_t 1218ipsec_hdrsiz_tcp(tp) 1219 struct tcpcb *tp; 1220{ 1221 struct inpcb *inp; 1222 struct mbuf *m; 1223 size_t hdrsiz; 1224 struct ip *ip; 1225#ifdef INET6 1226 struct ip6_hdr *ip6; 1227#endif /* INET6 */ 1228 struct tcphdr *th; 1229 1230 if (!tp || !tp->t_template || !(inp = tp->t_inpcb)) 1231 return 0; 1232 MGETHDR(m, M_DONTWAIT, MT_DATA); 1233 if (!m) 1234 return 0; 1235 1236#ifdef INET6 1237 if ((inp->inp_vflag & INP_IPV6) != 0) { 1238 ip6 = mtod(m, struct ip6_hdr *); 1239 th = (struct tcphdr *)(ip6 + 1); 1240 m->m_pkthdr.len = m->m_len = 1241 sizeof(struct ip6_hdr) + sizeof(struct tcphdr); 1242 bcopy((caddr_t)tp->t_template->tt_ipgen, (caddr_t)ip6, 1243 sizeof(struct ip6_hdr)); 1244 bcopy((caddr_t)&tp->t_template->tt_t, (caddr_t)th, 1245 sizeof(struct tcphdr)); 1246 hdrsiz = ipsec6_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); 1247 } else 1248#endif /* INET6 */ 1249 { 1250 ip = mtod(m, struct ip *); 1251 th = (struct tcphdr *)(ip + 1); 1252 m->m_pkthdr.len = m->m_len = sizeof(struct tcpiphdr); 1253 bcopy((caddr_t)tp->t_template->tt_ipgen, (caddr_t)ip, 1254 sizeof(struct ip)); 1255 bcopy((caddr_t)&tp->t_template->tt_t, (caddr_t)th, 1256 sizeof(struct tcphdr)); 1257 hdrsiz = ipsec4_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); 1258 } 1259 1260 m_free(m); 1261 return hdrsiz; 1262} 1263#endif /*IPSEC*/ 1264 1265/* 1266 * Return a pointer to the cached information about the remote host. 1267 * The cached information is stored in the protocol specific part of 1268 * the route metrics. 1269 */ 1270struct rmxp_tao * 1271tcp_gettaocache(inp) 1272 struct inpcb *inp; 1273{ 1274 struct rtentry *rt; 1275 1276#ifdef INET6 1277 if ((inp->inp_vflag & INP_IPV6) != 0) 1278 rt = tcp_rtlookup6(inp); 1279 else 1280#endif /* INET6 */ 1281 rt = tcp_rtlookup(inp); 1282 1283 /* Make sure this is a host route and is up. */ 1284 if (rt == NULL || 1285 (rt->rt_flags & (RTF_UP|RTF_HOST)) != (RTF_UP|RTF_HOST)) 1286 return NULL; 1287 1288 return rmx_taop(rt->rt_rmx); 1289} 1290 1291/* 1292 * Clear all the TAO cache entries, called from tcp_init. 1293 * 1294 * XXX 1295 * This routine is just an empty one, because we assume that the routing 1296 * routing tables are initialized at the same time when TCP, so there is 1297 * nothing in the cache left over. 1298 */ 1299static void 1300tcp_cleartaocache() 1301{ 1302} 1303