tcp_timewait.c revision 78100
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 78064 2001-06-11 12:39:29Z ume $ 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#include <sys/random.h> 57 58#include <vm/vm_zone.h> 59 60#include <net/route.h> 61#include <net/if.h> 62 63#define _IP_VHL 64#include <netinet/in.h> 65#include <netinet/in_systm.h> 66#include <netinet/ip.h> 67#ifdef INET6 68#include <netinet/ip6.h> 69#endif 70#include <netinet/in_pcb.h> 71#ifdef INET6 72#include <netinet6/in6_pcb.h> 73#endif 74#include <netinet/in_var.h> 75#include <netinet/ip_var.h> 76#ifdef INET6 77#include <netinet6/ip6_var.h> 78#endif 79#include <netinet/tcp.h> 80#include <netinet/tcp_fsm.h> 81#include <netinet/tcp_seq.h> 82#include <netinet/tcp_timer.h> 83#include <netinet/tcp_var.h> 84#ifdef INET6 85#include <netinet6/tcp6_var.h> 86#endif 87#include <netinet/tcpip.h> 88#ifdef TCPDEBUG 89#include <netinet/tcp_debug.h> 90#endif 91#include <netinet6/ip6protosw.h> 92 93#ifdef IPSEC 94#include <netinet6/ipsec.h> 95#ifdef INET6 96#include <netinet6/ipsec6.h> 97#endif 98#endif /*IPSEC*/ 99 100#include <machine/in_cksum.h> 101 102int tcp_mssdflt = TCP_MSS; 103SYSCTL_INT(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt, CTLFLAG_RW, 104 &tcp_mssdflt , 0, "Default TCP Maximum Segment Size"); 105 106#ifdef INET6 107int tcp_v6mssdflt = TCP6_MSS; 108SYSCTL_INT(_net_inet_tcp, TCPCTL_V6MSSDFLT, v6mssdflt, 109 CTLFLAG_RW, &tcp_v6mssdflt , 0, 110 "Default TCP Maximum Segment Size for IPv6"); 111#endif 112 113#if 0 114static int tcp_rttdflt = TCPTV_SRTTDFLT / PR_SLOWHZ; 115SYSCTL_INT(_net_inet_tcp, TCPCTL_RTTDFLT, rttdflt, CTLFLAG_RW, 116 &tcp_rttdflt , 0, "Default maximum TCP Round Trip Time"); 117#endif 118 119static int tcp_do_rfc1323 = 1; 120SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323, CTLFLAG_RW, 121 &tcp_do_rfc1323 , 0, "Enable rfc1323 (high performance TCP) extensions"); 122 123static int tcp_do_rfc1644 = 0; 124SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1644, rfc1644, CTLFLAG_RW, 125 &tcp_do_rfc1644 , 0, "Enable rfc1644 (TTCP) extensions"); 126 127static int tcp_tcbhashsize = 0; 128SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcbhashsize, CTLFLAG_RD, 129 &tcp_tcbhashsize, 0, "Size of TCP control-block hashtable"); 130 131static int do_tcpdrain = 1; 132SYSCTL_INT(_net_inet_tcp, OID_AUTO, do_tcpdrain, CTLFLAG_RW, &do_tcpdrain, 0, 133 "Enable tcp_drain routine for extra help when low on mbufs"); 134 135SYSCTL_INT(_net_inet_tcp, OID_AUTO, pcbcount, CTLFLAG_RD, 136 &tcbinfo.ipi_count, 0, "Number of active PCBs"); 137 138static int icmp_may_rst = 1; 139SYSCTL_INT(_net_inet_tcp, OID_AUTO, icmp_may_rst, CTLFLAG_RW, &icmp_may_rst, 0, 140 "Certain ICMP unreachable messages may abort connections in SYN_SENT"); 141 142static void tcp_cleartaocache __P((void)); 143static void tcp_notify __P((struct inpcb *, int)); 144 145/* 146 * Target size of TCP PCB hash tables. Must be a power of two. 147 * 148 * Note that this can be overridden by the kernel environment 149 * variable net.inet.tcp.tcbhashsize 150 */ 151#ifndef TCBHASHSIZE 152#define TCBHASHSIZE 512 153#endif 154 155/* 156 * This is the actual shape of what we allocate using the zone 157 * allocator. Doing it this way allows us to protect both structures 158 * using the same generation count, and also eliminates the overhead 159 * of allocating tcpcbs separately. By hiding the structure here, 160 * we avoid changing most of the rest of the code (although it needs 161 * to be changed, eventually, for greater efficiency). 162 */ 163#define ALIGNMENT 32 164#define ALIGNM1 (ALIGNMENT - 1) 165struct inp_tp { 166 union { 167 struct inpcb inp; 168 char align[(sizeof(struct inpcb) + ALIGNM1) & ~ALIGNM1]; 169 } inp_tp_u; 170 struct tcpcb tcb; 171 struct callout inp_tp_rexmt, inp_tp_persist, inp_tp_keep, inp_tp_2msl; 172 struct callout inp_tp_delack; 173}; 174#undef ALIGNMENT 175#undef ALIGNM1 176 177/* 178 * Tcp initialization 179 */ 180void 181tcp_init() 182{ 183 int hashsize = TCBHASHSIZE; 184 185 tcp_ccgen = 1; 186 tcp_cleartaocache(); 187 188 tcp_delacktime = TCPTV_DELACK; 189 tcp_keepinit = TCPTV_KEEP_INIT; 190 tcp_keepidle = TCPTV_KEEP_IDLE; 191 tcp_keepintvl = TCPTV_KEEPINTVL; 192 tcp_maxpersistidle = TCPTV_KEEP_IDLE; 193 tcp_msl = TCPTV_MSL; 194 195 LIST_INIT(&tcb); 196 tcbinfo.listhead = &tcb; 197 TUNABLE_INT_FETCH("net.inet.tcp.tcbhashsize", &hashsize); 198 if (!powerof2(hashsize)) { 199 printf("WARNING: TCB hash size not a power of 2\n"); 200 hashsize = 512; /* safe default */ 201 } 202 tcp_tcbhashsize = hashsize; 203 tcbinfo.hashbase = hashinit(hashsize, M_PCB, &tcbinfo.hashmask); 204 tcbinfo.porthashbase = hashinit(hashsize, M_PCB, 205 &tcbinfo.porthashmask); 206 tcbinfo.ipi_zone = zinit("tcpcb", sizeof(struct inp_tp), maxsockets, 207 ZONE_INTERRUPT, 0); 208#ifdef INET6 209#define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr)) 210#else /* INET6 */ 211#define TCP_MINPROTOHDR (sizeof(struct tcpiphdr)) 212#endif /* INET6 */ 213 if (max_protohdr < TCP_MINPROTOHDR) 214 max_protohdr = TCP_MINPROTOHDR; 215 if (max_linkhdr + TCP_MINPROTOHDR > MHLEN) 216 panic("tcp_init"); 217#undef TCP_MINPROTOHDR 218} 219 220/* 221 * Create template to be used to send tcp packets on a connection. 222 * Call after host entry created, allocates an mbuf and fills 223 * in a skeletal tcp/ip header, minimizing the amount of work 224 * necessary when the connection is used. 225 */ 226struct tcptemp * 227tcp_template(tp) 228 struct tcpcb *tp; 229{ 230 register struct inpcb *inp = tp->t_inpcb; 231 register struct mbuf *m; 232 register struct tcptemp *n; 233 234 if ((n = tp->t_template) == 0) { 235 m = m_get(M_DONTWAIT, MT_HEADER); 236 if (m == NULL) 237 return (0); 238 m->m_len = sizeof (struct tcptemp); 239 n = mtod(m, struct tcptemp *); 240 } 241#ifdef INET6 242 if ((inp->inp_vflag & INP_IPV6) != 0) { 243 register struct ip6_hdr *ip6; 244 245 ip6 = (struct ip6_hdr *)n->tt_ipgen; 246 ip6->ip6_flow = (ip6->ip6_flow & ~IPV6_FLOWINFO_MASK) | 247 (inp->in6p_flowinfo & IPV6_FLOWINFO_MASK); 248 ip6->ip6_vfc = (ip6->ip6_vfc & ~IPV6_VERSION_MASK) | 249 (IPV6_VERSION & IPV6_VERSION_MASK); 250 ip6->ip6_nxt = IPPROTO_TCP; 251 ip6->ip6_plen = sizeof(struct tcphdr); 252 ip6->ip6_src = inp->in6p_laddr; 253 ip6->ip6_dst = inp->in6p_faddr; 254 n->tt_t.th_sum = 0; 255 } else 256#endif 257 { 258 struct ip *ip = (struct ip *)n->tt_ipgen; 259 260 bzero(ip, sizeof(struct ip)); /* XXX overkill? */ 261 ip->ip_vhl = IP_VHL_BORING; 262 ip->ip_p = IPPROTO_TCP; 263 ip->ip_src = inp->inp_laddr; 264 ip->ip_dst = inp->inp_faddr; 265 n->tt_t.th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, 266 htons(sizeof(struct tcphdr) + IPPROTO_TCP)); 267 } 268 n->tt_t.th_sport = inp->inp_lport; 269 n->tt_t.th_dport = inp->inp_fport; 270 n->tt_t.th_seq = 0; 271 n->tt_t.th_ack = 0; 272 n->tt_t.th_x2 = 0; 273 n->tt_t.th_off = 5; 274 n->tt_t.th_flags = 0; 275 n->tt_t.th_win = 0; 276 n->tt_t.th_urp = 0; 277 return (n); 278} 279 280/* 281 * Send a single message to the TCP at address specified by 282 * the given TCP/IP header. If m == 0, then we make a copy 283 * of the tcpiphdr at ti and send directly to the addressed host. 284 * This is used to force keep alive messages out using the TCP 285 * template for a connection tp->t_template. If flags are given 286 * then we send a message back to the TCP which originated the 287 * segment ti, and discard the mbuf containing it and any other 288 * attached mbufs. 289 * 290 * In any case the ack and sequence number of the transmitted 291 * segment are as specified by the parameters. 292 * 293 * NOTE: If m != NULL, then ti must point to *inside* the mbuf. 294 */ 295void 296tcp_respond(tp, ipgen, th, m, ack, seq, flags) 297 struct tcpcb *tp; 298 void *ipgen; 299 register struct tcphdr *th; 300 register struct mbuf *m; 301 tcp_seq ack, seq; 302 int flags; 303{ 304 register int tlen; 305 int win = 0; 306 struct route *ro = 0; 307 struct route sro; 308 struct ip *ip; 309 struct tcphdr *nth; 310#ifdef INET6 311 struct route_in6 *ro6 = 0; 312 struct route_in6 sro6; 313 struct ip6_hdr *ip6; 314 int isipv6; 315#endif /* INET6 */ 316 int ipflags = 0; 317 318#ifdef INET6 319 isipv6 = IP_VHL_V(((struct ip *)ipgen)->ip_vhl) == 6; 320 ip6 = ipgen; 321#endif /* INET6 */ 322 ip = ipgen; 323 324 if (tp) { 325 if (!(flags & TH_RST)) { 326 win = sbspace(&tp->t_inpcb->inp_socket->so_rcv); 327 if (win > (long)TCP_MAXWIN << tp->rcv_scale) 328 win = (long)TCP_MAXWIN << tp->rcv_scale; 329 } 330#ifdef INET6 331 if (isipv6) 332 ro6 = &tp->t_inpcb->in6p_route; 333 else 334#endif /* INET6 */ 335 ro = &tp->t_inpcb->inp_route; 336 } else { 337#ifdef INET6 338 if (isipv6) { 339 ro6 = &sro6; 340 bzero(ro6, sizeof *ro6); 341 } else 342#endif /* INET6 */ 343 { 344 ro = &sro; 345 bzero(ro, sizeof *ro); 346 } 347 } 348 if (m == 0) { 349 m = m_gethdr(M_DONTWAIT, MT_HEADER); 350 if (m == NULL) 351 return; 352 tlen = 0; 353 m->m_data += max_linkhdr; 354#ifdef INET6 355 if (isipv6) { 356 bcopy((caddr_t)ip6, mtod(m, caddr_t), 357 sizeof(struct ip6_hdr)); 358 ip6 = mtod(m, struct ip6_hdr *); 359 nth = (struct tcphdr *)(ip6 + 1); 360 } else 361#endif /* INET6 */ 362 { 363 bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 364 ip = mtod(m, struct ip *); 365 nth = (struct tcphdr *)(ip + 1); 366 } 367 bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr)); 368 flags = TH_ACK; 369 } else { 370 m_freem(m->m_next); 371 m->m_next = 0; 372 m->m_data = (caddr_t)ipgen; 373 /* m_len is set later */ 374 tlen = 0; 375#define xchg(a,b,type) { type t; t=a; a=b; b=t; } 376#ifdef INET6 377 if (isipv6) { 378 xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr); 379 nth = (struct tcphdr *)(ip6 + 1); 380 } else 381#endif /* INET6 */ 382 { 383 xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, n_long); 384 nth = (struct tcphdr *)(ip + 1); 385 } 386 if (th != nth) { 387 /* 388 * this is usually a case when an extension header 389 * exists between the IPv6 header and the 390 * TCP header. 391 */ 392 nth->th_sport = th->th_sport; 393 nth->th_dport = th->th_dport; 394 } 395 xchg(nth->th_dport, nth->th_sport, n_short); 396#undef xchg 397 } 398#ifdef INET6 399 if (isipv6) { 400 ip6->ip6_plen = htons((u_short)(sizeof (struct tcphdr) + 401 tlen)); 402 tlen += sizeof (struct ip6_hdr) + sizeof (struct tcphdr); 403 } else 404#endif 405 { 406 tlen += sizeof (struct tcpiphdr); 407 ip->ip_len = tlen; 408 ip->ip_ttl = ip_defttl; 409 } 410 m->m_len = tlen; 411 m->m_pkthdr.len = tlen; 412 m->m_pkthdr.rcvif = (struct ifnet *) 0; 413 nth->th_seq = htonl(seq); 414 nth->th_ack = htonl(ack); 415 nth->th_x2 = 0; 416 nth->th_off = sizeof (struct tcphdr) >> 2; 417 nth->th_flags = flags; 418 if (tp) 419 nth->th_win = htons((u_short) (win >> tp->rcv_scale)); 420 else 421 nth->th_win = htons((u_short)win); 422 nth->th_urp = 0; 423#ifdef INET6 424 if (isipv6) { 425 nth->th_sum = 0; 426 nth->th_sum = in6_cksum(m, IPPROTO_TCP, 427 sizeof(struct ip6_hdr), 428 tlen - sizeof(struct ip6_hdr)); 429 ip6->ip6_hlim = in6_selecthlim(tp ? tp->t_inpcb : NULL, 430 ro6 && ro6->ro_rt ? 431 ro6->ro_rt->rt_ifp : 432 NULL); 433 } else 434#endif /* INET6 */ 435 { 436 nth->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, 437 htons((u_short)(tlen - sizeof(struct ip) + ip->ip_p))); 438 m->m_pkthdr.csum_flags = CSUM_TCP; 439 m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); 440 } 441#ifdef TCPDEBUG 442 if (tp == NULL || (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) 443 tcp_trace(TA_OUTPUT, 0, tp, mtod(m, void *), th, 0); 444#endif 445#ifdef IPSEC 446 if (ipsec_setsocket(m, tp ? tp->t_inpcb->inp_socket : NULL) != 0) { 447 m_freem(m); 448 return; 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, 0); 497 callout_init(tp->tt_persist = &it->inp_tp_persist, 0); 498 callout_init(tp->tt_keep = &it->inp_tp_keep, 0); 499 callout_init(tp->tt_2msl = &it->inp_tp_2msl, 0); 500 callout_init(tp->tt_delack = &it->inp_tp_delack, 0); 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 /* 520 * IPv4 TTL initialization is necessary for an IPv6 socket as well, 521 * because the socket may be bound to an IPv6 wildcard address, 522 * which may match an IPv4-mapped IPv6 address. 523 */ 524 inp->inp_ip_ttl = ip_defttl; 525 inp->inp_ppcb = (caddr_t)tp; 526 return (tp); /* XXX */ 527} 528 529/* 530 * Drop a TCP connection, reporting 531 * the specified error. If connection is synchronized, 532 * then send a RST to peer. 533 */ 534struct tcpcb * 535tcp_drop(tp, errno) 536 register struct tcpcb *tp; 537 int errno; 538{ 539 struct socket *so = tp->t_inpcb->inp_socket; 540 541 if (TCPS_HAVERCVDSYN(tp->t_state)) { 542 tp->t_state = TCPS_CLOSED; 543 (void) tcp_output(tp); 544 tcpstat.tcps_drops++; 545 } else 546 tcpstat.tcps_conndrops++; 547 if (errno == ETIMEDOUT && tp->t_softerror) 548 errno = tp->t_softerror; 549 so->so_error = errno; 550 return (tcp_close(tp)); 551} 552 553/* 554 * Close a TCP control block: 555 * discard all space held by the tcp 556 * discard internet protocol block 557 * wake up any sleepers 558 */ 559struct tcpcb * 560tcp_close(tp) 561 register struct tcpcb *tp; 562{ 563 register struct tseg_qent *q; 564 struct inpcb *inp = tp->t_inpcb; 565 struct socket *so = inp->inp_socket; 566#ifdef INET6 567 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; 568#endif /* INET6 */ 569 register struct rtentry *rt; 570 int dosavessthresh; 571 572 /* 573 * Make sure that all of our timers are stopped before we 574 * delete the PCB. 575 */ 576 callout_stop(tp->tt_rexmt); 577 callout_stop(tp->tt_persist); 578 callout_stop(tp->tt_keep); 579 callout_stop(tp->tt_2msl); 580 callout_stop(tp->tt_delack); 581 582 /* 583 * If we got enough samples through the srtt filter, 584 * save the rtt and rttvar in the routing entry. 585 * 'Enough' is arbitrarily defined as the 16 samples. 586 * 16 samples is enough for the srtt filter to converge 587 * to within 5% of the correct value; fewer samples and 588 * we could save a very bogus rtt. 589 * 590 * Don't update the default route's characteristics and don't 591 * update anything that the user "locked". 592 */ 593 if (tp->t_rttupdated >= 16) { 594 register u_long i = 0; 595#ifdef INET6 596 if (isipv6) { 597 struct sockaddr_in6 *sin6; 598 599 if ((rt = inp->in6p_route.ro_rt) == NULL) 600 goto no_valid_rt; 601 sin6 = (struct sockaddr_in6 *)rt_key(rt); 602 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) 603 goto no_valid_rt; 604 } 605 else 606#endif /* INET6 */ 607 if ((rt = inp->inp_route.ro_rt) == NULL || 608 ((struct sockaddr_in *)rt_key(rt))->sin_addr.s_addr 609 == INADDR_ANY) 610 goto no_valid_rt; 611 612 if ((rt->rt_rmx.rmx_locks & RTV_RTT) == 0) { 613 i = tp->t_srtt * 614 (RTM_RTTUNIT / (hz * TCP_RTT_SCALE)); 615 if (rt->rt_rmx.rmx_rtt && i) 616 /* 617 * filter this update to half the old & half 618 * the new values, converting scale. 619 * See route.h and tcp_var.h for a 620 * description of the scaling constants. 621 */ 622 rt->rt_rmx.rmx_rtt = 623 (rt->rt_rmx.rmx_rtt + i) / 2; 624 else 625 rt->rt_rmx.rmx_rtt = i; 626 tcpstat.tcps_cachedrtt++; 627 } 628 if ((rt->rt_rmx.rmx_locks & RTV_RTTVAR) == 0) { 629 i = tp->t_rttvar * 630 (RTM_RTTUNIT / (hz * TCP_RTTVAR_SCALE)); 631 if (rt->rt_rmx.rmx_rttvar && i) 632 rt->rt_rmx.rmx_rttvar = 633 (rt->rt_rmx.rmx_rttvar + i) / 2; 634 else 635 rt->rt_rmx.rmx_rttvar = i; 636 tcpstat.tcps_cachedrttvar++; 637 } 638 /* 639 * The old comment here said: 640 * update the pipelimit (ssthresh) if it has been updated 641 * already or if a pipesize was specified & the threshhold 642 * got below half the pipesize. I.e., wait for bad news 643 * before we start updating, then update on both good 644 * and bad news. 645 * 646 * But we want to save the ssthresh even if no pipesize is 647 * specified explicitly in the route, because such 648 * connections still have an implicit pipesize specified 649 * by the global tcp_sendspace. In the absence of a reliable 650 * way to calculate the pipesize, it will have to do. 651 */ 652 i = tp->snd_ssthresh; 653 if (rt->rt_rmx.rmx_sendpipe != 0) 654 dosavessthresh = (i < rt->rt_rmx.rmx_sendpipe / 2); 655 else 656 dosavessthresh = (i < so->so_snd.sb_hiwat / 2); 657 if (((rt->rt_rmx.rmx_locks & RTV_SSTHRESH) == 0 && 658 i != 0 && rt->rt_rmx.rmx_ssthresh != 0) 659 || dosavessthresh) { 660 /* 661 * convert the limit from user data bytes to 662 * packets then to packet data bytes. 663 */ 664 i = (i + tp->t_maxseg / 2) / tp->t_maxseg; 665 if (i < 2) 666 i = 2; 667 i *= (u_long)(tp->t_maxseg + 668#ifdef INET6 669 (isipv6 ? sizeof (struct ip6_hdr) + 670 sizeof (struct tcphdr) : 671#endif 672 sizeof (struct tcpiphdr) 673#ifdef INET6 674 ) 675#endif 676 ); 677 if (rt->rt_rmx.rmx_ssthresh) 678 rt->rt_rmx.rmx_ssthresh = 679 (rt->rt_rmx.rmx_ssthresh + i) / 2; 680 else 681 rt->rt_rmx.rmx_ssthresh = i; 682 tcpstat.tcps_cachedssthresh++; 683 } 684 } 685 rt = inp->inp_route.ro_rt; 686 if (rt) { 687 /* 688 * mark route for deletion if no information is 689 * cached. 690 */ 691 if ((tp->t_flags & TF_LQ_OVERFLOW) && 692 ((rt->rt_rmx.rmx_locks & RTV_RTT) == 0)){ 693 if (rt->rt_rmx.rmx_rtt == 0) 694 rt->rt_flags |= RTF_DELCLONE; 695 } 696 } 697 no_valid_rt: 698 /* free the reassembly queue, if any */ 699 while((q = LIST_FIRST(&tp->t_segq)) != NULL) { 700 LIST_REMOVE(q, tqe_q); 701 m_freem(q->tqe_m); 702 FREE(q, M_TSEGQ); 703 } 704 if (tp->t_template) 705 (void) m_free(dtom(tp->t_template)); 706 inp->inp_ppcb = NULL; 707 soisdisconnected(so); 708#ifdef INET6 709 if (INP_CHECK_SOCKAF(so, AF_INET6)) 710 in6_pcbdetach(inp); 711 else 712#endif /* INET6 */ 713 in_pcbdetach(inp); 714 tcpstat.tcps_closed++; 715 return ((struct tcpcb *)0); 716} 717 718void 719tcp_drain() 720{ 721 if (do_tcpdrain) 722 { 723 struct inpcb *inpb; 724 struct tcpcb *tcpb; 725 struct tseg_qent *te; 726 727 /* 728 * Walk the tcpbs, if existing, and flush the reassembly queue, 729 * if there is one... 730 * XXX: The "Net/3" implementation doesn't imply that the TCP 731 * reassembly queue should be flushed, but in a situation 732 * where we're really low on mbufs, this is potentially 733 * usefull. 734 */ 735 LIST_FOREACH(inpb, tcbinfo.listhead, inp_list) { 736 if ((tcpb = intotcpcb(inpb))) { 737 while ((te = LIST_FIRST(&tcpb->t_segq)) 738 != NULL) { 739 LIST_REMOVE(te, tqe_q); 740 m_freem(te->tqe_m); 741 FREE(te, M_TSEGQ); 742 } 743 } 744 } 745 } 746} 747 748/* 749 * Notify a tcp user of an asynchronous error; 750 * store error as soft error, but wake up user 751 * (for now, won't do anything until can select for soft error). 752 * 753 * Do not wake up user since there currently is no mechanism for 754 * reporting soft errors (yet - a kqueue filter may be added). 755 */ 756static void 757tcp_notify(inp, error) 758 struct inpcb *inp; 759 int error; 760{ 761 struct tcpcb *tp = (struct tcpcb *)inp->inp_ppcb; 762 763 /* 764 * Ignore some errors if we are hooked up. 765 * If connection hasn't completed, has retransmitted several times, 766 * and receives a second error, give up now. This is better 767 * than waiting a long time to establish a connection that 768 * can never complete. 769 */ 770 if (tp->t_state == TCPS_ESTABLISHED && 771 (error == EHOSTUNREACH || error == ENETUNREACH || 772 error == EHOSTDOWN)) { 773 return; 774 } else if (tp->t_state < TCPS_ESTABLISHED && tp->t_rxtshift > 3 && 775 tp->t_softerror) 776 tcp_drop(tp, error); 777 else 778 tp->t_softerror = error; 779#if 0 780 wakeup((caddr_t) &so->so_timeo); 781 sorwakeup(so); 782 sowwakeup(so); 783#endif 784} 785 786static int 787tcp_pcblist(SYSCTL_HANDLER_ARGS) 788{ 789 int error, i, n, s; 790 struct inpcb *inp, **inp_list; 791 inp_gen_t gencnt; 792 struct xinpgen xig; 793 794 /* 795 * The process of preparing the TCB list is too time-consuming and 796 * resource-intensive to repeat twice on every request. 797 */ 798 if (req->oldptr == 0) { 799 n = tcbinfo.ipi_count; 800 req->oldidx = 2 * (sizeof xig) 801 + (n + n/8) * sizeof(struct xtcpcb); 802 return 0; 803 } 804 805 if (req->newptr != 0) 806 return EPERM; 807 808 /* 809 * OK, now we're committed to doing something. 810 */ 811 s = splnet(); 812 gencnt = tcbinfo.ipi_gencnt; 813 n = tcbinfo.ipi_count; 814 splx(s); 815 816 xig.xig_len = sizeof xig; 817 xig.xig_count = n; 818 xig.xig_gen = gencnt; 819 xig.xig_sogen = so_gencnt; 820 error = SYSCTL_OUT(req, &xig, sizeof xig); 821 if (error) 822 return error; 823 824 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 825 if (inp_list == 0) 826 return ENOMEM; 827 828 s = splnet(); 829 for (inp = LIST_FIRST(tcbinfo.listhead), i = 0; inp && i < n; 830 inp = LIST_NEXT(inp, inp_list)) { 831 if (inp->inp_gencnt <= gencnt && !prison_xinpcb(req->p, inp)) 832 inp_list[i++] = inp; 833 } 834 splx(s); 835 n = i; 836 837 error = 0; 838 for (i = 0; i < n; i++) { 839 inp = inp_list[i]; 840 if (inp->inp_gencnt <= gencnt) { 841 struct xtcpcb xt; 842 caddr_t inp_ppcb; 843 xt.xt_len = sizeof xt; 844 /* XXX should avoid extra copy */ 845 bcopy(inp, &xt.xt_inp, sizeof *inp); 846 inp_ppcb = inp->inp_ppcb; 847 if (inp_ppcb != NULL) 848 bcopy(inp_ppcb, &xt.xt_tp, sizeof xt.xt_tp); 849 else 850 bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); 851 if (inp->inp_socket) 852 sotoxsocket(inp->inp_socket, &xt.xt_socket); 853 error = SYSCTL_OUT(req, &xt, sizeof xt); 854 } 855 } 856 if (!error) { 857 /* 858 * Give the user an updated idea of our state. 859 * If the generation differs from what we told 860 * her before, she knows that something happened 861 * while we were processing this request, and it 862 * might be necessary to retry. 863 */ 864 s = splnet(); 865 xig.xig_gen = tcbinfo.ipi_gencnt; 866 xig.xig_sogen = so_gencnt; 867 xig.xig_count = tcbinfo.ipi_count; 868 splx(s); 869 error = SYSCTL_OUT(req, &xig, sizeof xig); 870 } 871 free(inp_list, M_TEMP); 872 return error; 873} 874 875SYSCTL_PROC(_net_inet_tcp, TCPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0, 876 tcp_pcblist, "S,xtcpcb", "List of active TCP connections"); 877 878static int 879tcp_getcred(SYSCTL_HANDLER_ARGS) 880{ 881 struct xucred xuc; 882 struct sockaddr_in addrs[2]; 883 struct inpcb *inp; 884 int error, s; 885 886 error = suser(req->p); 887 if (error) 888 return (error); 889 error = SYSCTL_IN(req, addrs, sizeof(addrs)); 890 if (error) 891 return (error); 892 s = splnet(); 893 inp = in_pcblookup_hash(&tcbinfo, addrs[1].sin_addr, addrs[1].sin_port, 894 addrs[0].sin_addr, addrs[0].sin_port, 0, NULL); 895 if (inp == NULL || inp->inp_socket == NULL) { 896 error = ENOENT; 897 goto out; 898 } 899 bzero(&xuc, sizeof(xuc)); 900 xuc.cr_uid = inp->inp_socket->so_cred->cr_uid; 901 xuc.cr_ngroups = inp->inp_socket->so_cred->cr_ngroups; 902 bcopy(inp->inp_socket->so_cred->cr_groups, xuc.cr_groups, 903 sizeof(xuc.cr_groups)); 904 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); 905out: 906 splx(s); 907 return (error); 908} 909 910SYSCTL_PROC(_net_inet_tcp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 911 0, 0, tcp_getcred, "S,xucred", "Get the xucred of a TCP connection"); 912 913#ifdef INET6 914static int 915tcp6_getcred(SYSCTL_HANDLER_ARGS) 916{ 917 struct xucred xuc; 918 struct sockaddr_in6 addrs[2]; 919 struct inpcb *inp; 920 int error, s, mapped = 0; 921 922 error = suser(req->p); 923 if (error) 924 return (error); 925 error = SYSCTL_IN(req, addrs, sizeof(addrs)); 926 if (error) 927 return (error); 928 if (IN6_IS_ADDR_V4MAPPED(&addrs[0].sin6_addr)) { 929 if (IN6_IS_ADDR_V4MAPPED(&addrs[1].sin6_addr)) 930 mapped = 1; 931 else 932 return (EINVAL); 933 } 934 s = splnet(); 935 if (mapped == 1) 936 inp = in_pcblookup_hash(&tcbinfo, 937 *(struct in_addr *)&addrs[1].sin6_addr.s6_addr[12], 938 addrs[1].sin6_port, 939 *(struct in_addr *)&addrs[0].sin6_addr.s6_addr[12], 940 addrs[0].sin6_port, 941 0, NULL); 942 else 943 inp = in6_pcblookup_hash(&tcbinfo, &addrs[1].sin6_addr, 944 addrs[1].sin6_port, 945 &addrs[0].sin6_addr, addrs[0].sin6_port, 946 0, NULL); 947 if (inp == NULL || inp->inp_socket == NULL) { 948 error = ENOENT; 949 goto out; 950 } 951 bzero(&xuc, sizeof(xuc)); 952 xuc.cr_uid = inp->inp_socket->so_cred->cr_uid; 953 xuc.cr_ngroups = inp->inp_socket->so_cred->cr_ngroups; 954 bcopy(inp->inp_socket->so_cred->cr_groups, xuc.cr_groups, 955 sizeof(xuc.cr_groups)); 956 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); 957out: 958 splx(s); 959 return (error); 960} 961 962SYSCTL_PROC(_net_inet6_tcp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 963 0, 0, 964 tcp6_getcred, "S,xucred", "Get the xucred of a TCP6 connection"); 965#endif 966 967 968void 969tcp_ctlinput(cmd, sa, vip) 970 int cmd; 971 struct sockaddr *sa; 972 void *vip; 973{ 974 struct ip *ip = vip; 975 struct tcphdr *th; 976 struct in_addr faddr; 977 struct inpcb *inp; 978 struct tcpcb *tp; 979 void (*notify) __P((struct inpcb *, int)) = tcp_notify; 980 tcp_seq icmp_seq; 981 int s; 982 983 faddr = ((struct sockaddr_in *)sa)->sin_addr; 984 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) 985 return; 986 987 if (cmd == PRC_QUENCH) 988 notify = tcp_quench; 989 else if (icmp_may_rst && (cmd == PRC_UNREACH_ADMIN_PROHIB || 990 cmd == PRC_UNREACH_PORT) && ip) 991 notify = tcp_drop_syn_sent; 992 else if (cmd == PRC_MSGSIZE) 993 notify = tcp_mtudisc; 994 else if (PRC_IS_REDIRECT(cmd)) { 995 ip = 0; 996 notify = in_rtchange; 997 } else if (cmd == PRC_HOSTDEAD) 998 ip = 0; 999 else if ((unsigned)cmd > PRC_NCMDS || inetctlerrmap[cmd] == 0) 1000 return; 1001 if (ip) { 1002 s = splnet(); 1003 th = (struct tcphdr *)((caddr_t)ip 1004 + (IP_VHL_HL(ip->ip_vhl) << 2)); 1005 inp = in_pcblookup_hash(&tcbinfo, faddr, th->th_dport, 1006 ip->ip_src, th->th_sport, 0, NULL); 1007 if (inp != NULL && inp->inp_socket != NULL) { 1008 icmp_seq = htonl(th->th_seq); 1009 tp = intotcpcb(inp); 1010 if (SEQ_GEQ(icmp_seq, tp->snd_una) && 1011 SEQ_LT(icmp_seq, tp->snd_max)) 1012 (*notify)(inp, inetctlerrmap[cmd]); 1013 } 1014 splx(s); 1015 } else 1016 in_pcbnotifyall(&tcb, faddr, inetctlerrmap[cmd], notify); 1017} 1018 1019#ifdef INET6 1020void 1021tcp6_ctlinput(cmd, sa, d) 1022 int cmd; 1023 struct sockaddr *sa; 1024 void *d; 1025{ 1026 struct tcphdr th; 1027 void (*notify) __P((struct inpcb *, int)) = tcp_notify; 1028 struct ip6_hdr *ip6; 1029 struct mbuf *m; 1030 struct ip6ctlparam *ip6cp = NULL; 1031 const struct sockaddr_in6 *sa6_src = NULL; 1032 int off; 1033 struct tcp_portonly { 1034 u_int16_t th_sport; 1035 u_int16_t th_dport; 1036 } *thp; 1037 1038 if (sa->sa_family != AF_INET6 || 1039 sa->sa_len != sizeof(struct sockaddr_in6)) 1040 return; 1041 1042 if (cmd == PRC_QUENCH) 1043 notify = tcp_quench; 1044 else if (cmd == PRC_MSGSIZE) 1045 notify = tcp_mtudisc; 1046 else if (!PRC_IS_REDIRECT(cmd) && 1047 ((unsigned)cmd > PRC_NCMDS || inet6ctlerrmap[cmd] == 0)) 1048 return; 1049 1050 /* if the parameter is from icmp6, decode it. */ 1051 if (d != NULL) { 1052 ip6cp = (struct ip6ctlparam *)d; 1053 m = ip6cp->ip6c_m; 1054 ip6 = ip6cp->ip6c_ip6; 1055 off = ip6cp->ip6c_off; 1056 sa6_src = ip6cp->ip6c_src; 1057 } else { 1058 m = NULL; 1059 ip6 = NULL; 1060 off = 0; /* fool gcc */ 1061 sa6_src = &sa6_any; 1062 } 1063 1064 if (ip6) { 1065 /* 1066 * XXX: We assume that when IPV6 is non NULL, 1067 * M and OFF are valid. 1068 */ 1069 1070 /* check if we can safely examine src and dst ports */ 1071 if (m->m_pkthdr.len < off + sizeof(*thp)) 1072 return; 1073 1074 bzero(&th, sizeof(th)); 1075 m_copydata(m, off, sizeof(*thp), (caddr_t)&th); 1076 1077 in6_pcbnotify(&tcb, sa, th.th_dport, 1078 (struct sockaddr *)ip6cp->ip6c_src, 1079 th.th_sport, cmd, notify); 1080 } else 1081 in6_pcbnotify(&tcb, sa, 0, (struct sockaddr *)sa6_src, 1082 0, cmd, notify); 1083} 1084#endif /* INET6 */ 1085 1086#define TCP_RNDISS_ROUNDS 16 1087#define TCP_RNDISS_OUT 7200 1088#define TCP_RNDISS_MAX 30000 1089 1090u_int8_t tcp_rndiss_sbox[128]; 1091u_int16_t tcp_rndiss_msb; 1092u_int16_t tcp_rndiss_cnt; 1093long tcp_rndiss_reseed; 1094 1095u_int16_t 1096tcp_rndiss_encrypt(val) 1097 u_int16_t val; 1098{ 1099 u_int16_t sum = 0, i; 1100 1101 for (i = 0; i < TCP_RNDISS_ROUNDS; i++) { 1102 sum += 0x79b9; 1103 val ^= ((u_int16_t)tcp_rndiss_sbox[(val^sum) & 0x7f]) << 7; 1104 val = ((val & 0xff) << 7) | (val >> 8); 1105 } 1106 1107 return val; 1108} 1109 1110void 1111tcp_rndiss_init() 1112{ 1113 struct timeval time; 1114 1115 getmicrotime(&time); 1116 read_random(tcp_rndiss_sbox, sizeof(tcp_rndiss_sbox)); 1117 1118 tcp_rndiss_reseed = time.tv_sec + TCP_RNDISS_OUT; 1119 tcp_rndiss_msb = tcp_rndiss_msb == 0x8000 ? 0 : 0x8000; 1120 tcp_rndiss_cnt = 0; 1121} 1122 1123tcp_seq 1124tcp_rndiss_next() 1125{ 1126 u_int16_t tmp; 1127 struct timeval time; 1128 1129 getmicrotime(&time); 1130 1131 if (tcp_rndiss_cnt >= TCP_RNDISS_MAX || 1132 time.tv_sec > tcp_rndiss_reseed) 1133 tcp_rndiss_init(); 1134 1135 read_random(&tmp, sizeof(tmp)); 1136 1137 /* (tmp & 0x7fff) ensures a 32768 byte gap between ISS */ 1138 return ((tcp_rndiss_encrypt(tcp_rndiss_cnt++) | tcp_rndiss_msb) <<16) | 1139 (tmp & 0x7fff); 1140} 1141 1142 1143/* 1144 * When a source quench is received, close congestion window 1145 * to one segment. We will gradually open it again as we proceed. 1146 */ 1147void 1148tcp_quench(inp, errno) 1149 struct inpcb *inp; 1150 int errno; 1151{ 1152 struct tcpcb *tp = intotcpcb(inp); 1153 1154 if (tp) 1155 tp->snd_cwnd = tp->t_maxseg; 1156} 1157 1158/* 1159 * When a specific ICMP unreachable message is received and the 1160 * connection state is SYN-SENT, drop the connection. This behavior 1161 * is controlled by the icmp_may_rst sysctl. 1162 */ 1163void 1164tcp_drop_syn_sent(inp, errno) 1165 struct inpcb *inp; 1166 int errno; 1167{ 1168 struct tcpcb *tp = intotcpcb(inp); 1169 1170 if (tp && tp->t_state == TCPS_SYN_SENT) 1171 tcp_drop(tp, errno); 1172} 1173 1174/* 1175 * When `need fragmentation' ICMP is received, update our idea of the MSS 1176 * based on the new value in the route. Also nudge TCP to send something, 1177 * since we know the packet we just sent was dropped. 1178 * This duplicates some code in the tcp_mss() function in tcp_input.c. 1179 */ 1180void 1181tcp_mtudisc(inp, errno) 1182 struct inpcb *inp; 1183 int errno; 1184{ 1185 struct tcpcb *tp = intotcpcb(inp); 1186 struct rtentry *rt; 1187 struct rmxp_tao *taop; 1188 struct socket *so = inp->inp_socket; 1189 int offered; 1190 int mss; 1191#ifdef INET6 1192 int isipv6 = (tp->t_inpcb->inp_vflag & INP_IPV6) != 0; 1193#endif /* INET6 */ 1194 1195 if (tp) { 1196#ifdef INET6 1197 if (isipv6) 1198 rt = tcp_rtlookup6(inp); 1199 else 1200#endif /* INET6 */ 1201 rt = tcp_rtlookup(inp); 1202 if (!rt || !rt->rt_rmx.rmx_mtu) { 1203 tp->t_maxopd = tp->t_maxseg = 1204#ifdef INET6 1205 isipv6 ? tcp_v6mssdflt : 1206#endif /* INET6 */ 1207 tcp_mssdflt; 1208 return; 1209 } 1210 taop = rmx_taop(rt->rt_rmx); 1211 offered = taop->tao_mssopt; 1212 mss = rt->rt_rmx.rmx_mtu - 1213#ifdef INET6 1214 (isipv6 ? 1215 sizeof(struct ip6_hdr) + sizeof(struct tcphdr) : 1216#endif /* INET6 */ 1217 sizeof(struct tcpiphdr) 1218#ifdef INET6 1219 ) 1220#endif /* INET6 */ 1221 ; 1222 1223 if (offered) 1224 mss = min(mss, offered); 1225 /* 1226 * XXX - The above conditional probably violates the TCP 1227 * spec. The problem is that, since we don't know the 1228 * other end's MSS, we are supposed to use a conservative 1229 * default. But, if we do that, then MTU discovery will 1230 * never actually take place, because the conservative 1231 * default is much less than the MTUs typically seen 1232 * on the Internet today. For the moment, we'll sweep 1233 * this under the carpet. 1234 * 1235 * The conservative default might not actually be a problem 1236 * if the only case this occurs is when sending an initial 1237 * SYN with options and data to a host we've never talked 1238 * to before. Then, they will reply with an MSS value which 1239 * will get recorded and the new parameters should get 1240 * recomputed. For Further Study. 1241 */ 1242 if (tp->t_maxopd <= mss) 1243 return; 1244 tp->t_maxopd = mss; 1245 1246 if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP && 1247 (tp->t_flags & TF_RCVD_TSTMP) == TF_RCVD_TSTMP) 1248 mss -= TCPOLEN_TSTAMP_APPA; 1249 if ((tp->t_flags & (TF_REQ_CC|TF_NOOPT)) == TF_REQ_CC && 1250 (tp->t_flags & TF_RCVD_CC) == TF_RCVD_CC) 1251 mss -= TCPOLEN_CC_APPA; 1252#if (MCLBYTES & (MCLBYTES - 1)) == 0 1253 if (mss > MCLBYTES) 1254 mss &= ~(MCLBYTES-1); 1255#else 1256 if (mss > MCLBYTES) 1257 mss = mss / MCLBYTES * MCLBYTES; 1258#endif 1259 if (so->so_snd.sb_hiwat < mss) 1260 mss = so->so_snd.sb_hiwat; 1261 1262 tp->t_maxseg = mss; 1263 1264 tcpstat.tcps_mturesent++; 1265 tp->t_rtttime = 0; 1266 tp->snd_nxt = tp->snd_una; 1267 tcp_output(tp); 1268 } 1269} 1270 1271/* 1272 * Look-up the routing entry to the peer of this inpcb. If no route 1273 * is found and it cannot be allocated the return NULL. This routine 1274 * is called by TCP routines that access the rmx structure and by tcp_mss 1275 * to get the interface MTU. 1276 */ 1277struct rtentry * 1278tcp_rtlookup(inp) 1279 struct inpcb *inp; 1280{ 1281 struct route *ro; 1282 struct rtentry *rt; 1283 1284 ro = &inp->inp_route; 1285 rt = ro->ro_rt; 1286 if (rt == NULL || !(rt->rt_flags & RTF_UP)) { 1287 /* No route yet, so try to acquire one */ 1288 if (inp->inp_faddr.s_addr != INADDR_ANY) { 1289 ro->ro_dst.sa_family = AF_INET; 1290 ro->ro_dst.sa_len = sizeof(ro->ro_dst); 1291 ((struct sockaddr_in *) &ro->ro_dst)->sin_addr = 1292 inp->inp_faddr; 1293 rtalloc(ro); 1294 rt = ro->ro_rt; 1295 } 1296 } 1297 return rt; 1298} 1299 1300#ifdef INET6 1301struct rtentry * 1302tcp_rtlookup6(inp) 1303 struct inpcb *inp; 1304{ 1305 struct route_in6 *ro6; 1306 struct rtentry *rt; 1307 1308 ro6 = &inp->in6p_route; 1309 rt = ro6->ro_rt; 1310 if (rt == NULL || !(rt->rt_flags & RTF_UP)) { 1311 /* No route yet, so try to acquire one */ 1312 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) { 1313 struct sockaddr_in6 *dst6; 1314 1315 dst6 = (struct sockaddr_in6 *)&ro6->ro_dst; 1316 dst6->sin6_family = AF_INET6; 1317 dst6->sin6_len = sizeof(ro6->ro_dst); 1318 dst6->sin6_addr = inp->in6p_faddr; 1319 rtalloc((struct route *)ro6); 1320 rt = ro6->ro_rt; 1321 } 1322 } 1323 return rt; 1324} 1325#endif /* INET6 */ 1326 1327#ifdef IPSEC 1328/* compute ESP/AH header size for TCP, including outer IP header. */ 1329size_t 1330ipsec_hdrsiz_tcp(tp) 1331 struct tcpcb *tp; 1332{ 1333 struct inpcb *inp; 1334 struct mbuf *m; 1335 size_t hdrsiz; 1336 struct ip *ip; 1337#ifdef INET6 1338 struct ip6_hdr *ip6; 1339#endif /* INET6 */ 1340 struct tcphdr *th; 1341 1342 if (!tp || !tp->t_template || !(inp = tp->t_inpcb)) 1343 return 0; 1344 MGETHDR(m, M_DONTWAIT, MT_DATA); 1345 if (!m) 1346 return 0; 1347 1348#ifdef INET6 1349 if ((inp->inp_vflag & INP_IPV6) != 0) { 1350 ip6 = mtod(m, struct ip6_hdr *); 1351 th = (struct tcphdr *)(ip6 + 1); 1352 m->m_pkthdr.len = m->m_len = 1353 sizeof(struct ip6_hdr) + sizeof(struct tcphdr); 1354 bcopy((caddr_t)tp->t_template->tt_ipgen, (caddr_t)ip6, 1355 sizeof(struct ip6_hdr)); 1356 bcopy((caddr_t)&tp->t_template->tt_t, (caddr_t)th, 1357 sizeof(struct tcphdr)); 1358 hdrsiz = ipsec6_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); 1359 } else 1360#endif /* INET6 */ 1361 { 1362 ip = mtod(m, struct ip *); 1363 th = (struct tcphdr *)(ip + 1); 1364 m->m_pkthdr.len = m->m_len = sizeof(struct tcpiphdr); 1365 bcopy((caddr_t)tp->t_template->tt_ipgen, (caddr_t)ip, 1366 sizeof(struct ip)); 1367 bcopy((caddr_t)&tp->t_template->tt_t, (caddr_t)th, 1368 sizeof(struct tcphdr)); 1369 ip->ip_vhl = IP_VHL_BORING; 1370 hdrsiz = ipsec4_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); 1371 } 1372 1373 m_free(m); 1374 return hdrsiz; 1375} 1376#endif /*IPSEC*/ 1377 1378/* 1379 * Return a pointer to the cached information about the remote host. 1380 * The cached information is stored in the protocol specific part of 1381 * the route metrics. 1382 */ 1383struct rmxp_tao * 1384tcp_gettaocache(inp) 1385 struct inpcb *inp; 1386{ 1387 struct rtentry *rt; 1388 1389#ifdef INET6 1390 if ((inp->inp_vflag & INP_IPV6) != 0) 1391 rt = tcp_rtlookup6(inp); 1392 else 1393#endif /* INET6 */ 1394 rt = tcp_rtlookup(inp); 1395 1396 /* Make sure this is a host route and is up. */ 1397 if (rt == NULL || 1398 (rt->rt_flags & (RTF_UP|RTF_HOST)) != (RTF_UP|RTF_HOST)) 1399 return NULL; 1400 1401 return rmx_taop(rt->rt_rmx); 1402} 1403 1404/* 1405 * Clear all the TAO cache entries, called from tcp_init. 1406 * 1407 * XXX 1408 * This routine is just an empty one, because we assume that the routing 1409 * routing tables are initialized at the same time when TCP, so there is 1410 * nothing in the cache left over. 1411 */ 1412static void 1413tcp_cleartaocache() 1414{ 1415} 1416