tcp_timewait.c revision 264342
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 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)tcp_subr.c 8.2 (Berkeley) 5/24/95 30 */ 31 32#include <sys/cdefs.h> 33__FBSDID("$FreeBSD: head/sys/netinet/tcp_timewait.c 264342 2014-04-11 13:11:43Z jhb $"); 34 35#include "opt_inet.h" 36#include "opt_inet6.h" 37#include "opt_tcpdebug.h" 38 39#include <sys/param.h> 40#include <sys/systm.h> 41#include <sys/callout.h> 42#include <sys/kernel.h> 43#include <sys/sysctl.h> 44#include <sys/malloc.h> 45#include <sys/mbuf.h> 46#include <sys/priv.h> 47#include <sys/proc.h> 48#include <sys/socket.h> 49#include <sys/socketvar.h> 50#include <sys/protosw.h> 51#include <sys/random.h> 52#include <sys/refcount.h> 53 54#include <vm/uma.h> 55 56#include <net/route.h> 57#include <net/if.h> 58#include <net/if_var.h> 59#include <net/vnet.h> 60 61#include <netinet/in.h> 62#include <netinet/in_pcb.h> 63#include <netinet/in_systm.h> 64#include <netinet/in_var.h> 65#include <netinet/ip.h> 66#include <netinet/ip_icmp.h> 67#include <netinet/ip_var.h> 68#ifdef INET6 69#include <netinet/ip6.h> 70#include <netinet6/in6_pcb.h> 71#include <netinet6/ip6_var.h> 72#include <netinet6/scope6_var.h> 73#include <netinet6/nd6.h> 74#endif 75#include <netinet/tcp.h> 76#include <netinet/tcp_fsm.h> 77#include <netinet/tcp_seq.h> 78#include <netinet/tcp_timer.h> 79#include <netinet/tcp_var.h> 80#ifdef INET6 81#include <netinet6/tcp6_var.h> 82#endif 83#include <netinet/tcpip.h> 84#ifdef TCPDEBUG 85#include <netinet/tcp_debug.h> 86#endif 87#ifdef INET6 88#include <netinet6/ip6protosw.h> 89#endif 90 91#include <machine/in_cksum.h> 92 93#include <security/mac/mac_framework.h> 94 95static VNET_DEFINE(uma_zone_t, tcptw_zone); 96#define V_tcptw_zone VNET(tcptw_zone) 97static int maxtcptw; 98 99/* 100 * The timed wait queue contains references to each of the TCP sessions 101 * currently in the TIME_WAIT state. The queue pointers, including the 102 * queue pointers in each tcptw structure, are protected using the global 103 * timewait lock, which must be held over queue iteration and modification. 104 */ 105static VNET_DEFINE(TAILQ_HEAD(, tcptw), twq_2msl); 106#define V_twq_2msl VNET(twq_2msl) 107 108/* Global timewait lock */ 109static VNET_DEFINE(struct rwlock, tw_lock); 110#define V_tw_lock VNET(tw_lock) 111 112#define TW_LOCK_INIT(tw, d) rw_init_flags(&(tw), (d), 0) 113#define TW_LOCK_DESTROY(tw) rw_destroy(&(tw)) 114#define TW_RLOCK(tw) rw_rlock(&(tw)) 115#define TW_WLOCK(tw) rw_wlock(&(tw)) 116#define TW_RUNLOCK(tw) rw_runlock(&(tw)) 117#define TW_WUNLOCK(tw) rw_wunlock(&(tw)) 118#define TW_LOCK_ASSERT(tw) rw_assert(&(tw), RA_LOCKED) 119#define TW_RLOCK_ASSERT(tw) rw_assert(&(tw), RA_RLOCKED) 120#define TW_WLOCK_ASSERT(tw) rw_assert(&(tw), RA_WLOCKED) 121#define TW_UNLOCK_ASSERT(tw) rw_assert(&(tw), RA_UNLOCKED) 122 123/* 124 * tw_pcbref() bumps the reference count on an tw in order to maintain 125 * stability of an tw pointer despite the tw lock being released. 126 */ 127static void 128tw_pcbref(struct tcptw *tw) 129{ 130 131 KASSERT(tw->tw_refcount > 0, ("%s: refcount 0", __func__)); 132 refcount_acquire(&tw->tw_refcount); 133} 134 135/* 136 * Drop a refcount on an tw elevated using tw_pcbref(). Return 137 * the tw lock released. 138 */ 139static int 140tw_pcbrele(struct tcptw *tw) 141{ 142 143 TW_WLOCK_ASSERT(V_tw_lock); 144 KASSERT(tw->tw_refcount > 0, ("%s: refcount 0", __func__)); 145 146 if (!refcount_release(&tw->tw_refcount)) { 147 TW_WUNLOCK(V_tw_lock); 148 return (0); 149 } 150 151 uma_zfree(V_tcptw_zone, tw); 152 TW_WUNLOCK(V_tw_lock); 153 return (1); 154} 155 156static void tcp_tw_2msl_reset(struct tcptw *, int ream); 157static void tcp_tw_2msl_stop(struct tcptw *, int reuse); 158 159static int 160tcptw_auto_size(void) 161{ 162 int halfrange; 163 164 /* 165 * Max out at half the ephemeral port range so that TIME_WAIT 166 * sockets don't tie up too many ephemeral ports. 167 */ 168 if (V_ipport_lastauto > V_ipport_firstauto) 169 halfrange = (V_ipport_lastauto - V_ipport_firstauto) / 2; 170 else 171 halfrange = (V_ipport_firstauto - V_ipport_lastauto) / 2; 172 /* Protect against goofy port ranges smaller than 32. */ 173 return (imin(imax(halfrange, 32), maxsockets / 5)); 174} 175 176static int 177sysctl_maxtcptw(SYSCTL_HANDLER_ARGS) 178{ 179 int error, new; 180 181 if (maxtcptw == 0) 182 new = tcptw_auto_size(); 183 else 184 new = maxtcptw; 185 error = sysctl_handle_int(oidp, &new, 0, req); 186 if (error == 0 && req->newptr) 187 if (new >= 32) { 188 maxtcptw = new; 189 uma_zone_set_max(V_tcptw_zone, maxtcptw); 190 } 191 return (error); 192} 193 194SYSCTL_PROC(_net_inet_tcp, OID_AUTO, maxtcptw, CTLTYPE_INT|CTLFLAG_RW, 195 &maxtcptw, 0, sysctl_maxtcptw, "IU", 196 "Maximum number of compressed TCP TIME_WAIT entries"); 197 198VNET_DEFINE(int, nolocaltimewait) = 0; 199#define V_nolocaltimewait VNET(nolocaltimewait) 200SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, nolocaltimewait, CTLFLAG_RW, 201 &VNET_NAME(nolocaltimewait), 0, 202 "Do not create compressed TCP TIME_WAIT entries for local connections"); 203 204void 205tcp_tw_zone_change(void) 206{ 207 208 if (maxtcptw == 0) 209 uma_zone_set_max(V_tcptw_zone, tcptw_auto_size()); 210} 211 212void 213tcp_tw_init(void) 214{ 215 216 V_tcptw_zone = uma_zcreate("tcptw", sizeof(struct tcptw), 217 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 218 TUNABLE_INT_FETCH("net.inet.tcp.maxtcptw", &maxtcptw); 219 if (maxtcptw == 0) 220 uma_zone_set_max(V_tcptw_zone, tcptw_auto_size()); 221 else 222 uma_zone_set_max(V_tcptw_zone, maxtcptw); 223 TAILQ_INIT(&V_twq_2msl); 224 TW_LOCK_INIT(V_tw_lock, "tcptw"); 225} 226 227#ifdef VIMAGE 228void 229tcp_tw_destroy(void) 230{ 231 struct tcptw *tw; 232 233 INP_INFO_WLOCK(&V_tcbinfo); 234 while ((tw = TAILQ_FIRST(&V_twq_2msl)) != NULL) 235 tcp_twclose(tw, 0); 236 INP_INFO_WUNLOCK(&V_tcbinfo); 237 238 TW_LOCK_DESTROY(V_tw_lock); 239 uma_zdestroy(V_tcptw_zone); 240} 241#endif 242 243/* 244 * Move a TCP connection into TIME_WAIT state. 245 * tcbinfo is locked. 246 * inp is locked, and is unlocked before returning. 247 */ 248void 249tcp_twstart(struct tcpcb *tp) 250{ 251 struct tcptw *tw; 252 struct inpcb *inp = tp->t_inpcb; 253 int acknow; 254 struct socket *so; 255#ifdef INET6 256 int isipv6 = inp->inp_inc.inc_flags & INC_ISIPV6; 257#endif 258 259 INP_INFO_WLOCK_ASSERT(&V_tcbinfo); 260 INP_WLOCK_ASSERT(inp); 261 262 if (V_nolocaltimewait) { 263 int error = 0; 264#ifdef INET6 265 if (isipv6) 266 error = in6_localaddr(&inp->in6p_faddr); 267#endif 268#if defined(INET6) && defined(INET) 269 else 270#endif 271#ifdef INET 272 error = in_localip(inp->inp_faddr); 273#endif 274 if (error) { 275 tp = tcp_close(tp); 276 if (tp != NULL) 277 INP_WUNLOCK(inp); 278 return; 279 } 280 } 281 282 tw = uma_zalloc(V_tcptw_zone, M_NOWAIT); 283 if (tw == NULL) { 284 tw = tcp_tw_2msl_reuse(); 285 if (tw == NULL) { 286 tp = tcp_close(tp); 287 if (tp != NULL) 288 INP_WUNLOCK(inp); 289 return; 290 } 291 } 292 tw->tw_inpcb = inp; 293 refcount_init(&tw->tw_refcount, 1); 294 295 /* 296 * Recover last window size sent. 297 */ 298 if (SEQ_GT(tp->rcv_adv, tp->rcv_nxt)) 299 tw->last_win = (tp->rcv_adv - tp->rcv_nxt) >> tp->rcv_scale; 300 else 301 tw->last_win = 0; 302 303 /* 304 * Set t_recent if timestamps are used on the connection. 305 */ 306 if ((tp->t_flags & (TF_REQ_TSTMP|TF_RCVD_TSTMP|TF_NOOPT)) == 307 (TF_REQ_TSTMP|TF_RCVD_TSTMP)) { 308 tw->t_recent = tp->ts_recent; 309 tw->ts_offset = tp->ts_offset; 310 } else { 311 tw->t_recent = 0; 312 tw->ts_offset = 0; 313 } 314 315 tw->snd_nxt = tp->snd_nxt; 316 tw->rcv_nxt = tp->rcv_nxt; 317 tw->iss = tp->iss; 318 tw->irs = tp->irs; 319 tw->t_starttime = tp->t_starttime; 320 tw->tw_time = 0; 321 322/* XXX 323 * If this code will 324 * be used for fin-wait-2 state also, then we may need 325 * a ts_recent from the last segment. 326 */ 327 acknow = tp->t_flags & TF_ACKNOW; 328 329 /* 330 * First, discard tcpcb state, which includes stopping its timers and 331 * freeing it. tcp_discardcb() used to also release the inpcb, but 332 * that work is now done in the caller. 333 * 334 * Note: soisdisconnected() call used to be made in tcp_discardcb(), 335 * and might not be needed here any longer. 336 */ 337 tcp_discardcb(tp); 338 so = inp->inp_socket; 339 soisdisconnected(so); 340 tw->tw_cred = crhold(so->so_cred); 341 SOCK_LOCK(so); 342 tw->tw_so_options = so->so_options; 343 SOCK_UNLOCK(so); 344 if (acknow) 345 tcp_twrespond(tw, TH_ACK); 346 inp->inp_ppcb = tw; 347 inp->inp_flags |= INP_TIMEWAIT; 348 tcp_tw_2msl_reset(tw, 0); 349 350 /* 351 * If the inpcb owns the sole reference to the socket, then we can 352 * detach and free the socket as it is not needed in time wait. 353 */ 354 if (inp->inp_flags & INP_SOCKREF) { 355 KASSERT(so->so_state & SS_PROTOREF, 356 ("tcp_twstart: !SS_PROTOREF")); 357 inp->inp_flags &= ~INP_SOCKREF; 358 INP_WUNLOCK(inp); 359 ACCEPT_LOCK(); 360 SOCK_LOCK(so); 361 so->so_state &= ~SS_PROTOREF; 362 sofree(so); 363 } else 364 INP_WUNLOCK(inp); 365} 366 367#if 0 368/* 369 * The appromixate rate of ISN increase of Microsoft TCP stacks; 370 * the actual rate is slightly higher due to the addition of 371 * random positive increments. 372 * 373 * Most other new OSes use semi-randomized ISN values, so we 374 * do not need to worry about them. 375 */ 376#define MS_ISN_BYTES_PER_SECOND 250000 377 378/* 379 * Determine if the ISN we will generate has advanced beyond the last 380 * sequence number used by the previous connection. If so, indicate 381 * that it is safe to recycle this tw socket by returning 1. 382 */ 383int 384tcp_twrecycleable(struct tcptw *tw) 385{ 386 tcp_seq new_iss = tw->iss; 387 tcp_seq new_irs = tw->irs; 388 389 INP_INFO_WLOCK_ASSERT(&V_tcbinfo); 390 new_iss += (ticks - tw->t_starttime) * (ISN_BYTES_PER_SECOND / hz); 391 new_irs += (ticks - tw->t_starttime) * (MS_ISN_BYTES_PER_SECOND / hz); 392 393 if (SEQ_GT(new_iss, tw->snd_nxt) && SEQ_GT(new_irs, tw->rcv_nxt)) 394 return (1); 395 else 396 return (0); 397} 398#endif 399 400/* 401 * Returns 1 if the TIME_WAIT state was killed and we should start over, 402 * looking for a pcb in the listen state. Returns 0 otherwise. 403 */ 404int 405tcp_twcheck(struct inpcb *inp, struct tcpopt *to, struct tcphdr *th, 406 struct mbuf *m, int tlen) 407{ 408 struct tcptw *tw; 409 int thflags; 410 tcp_seq seq; 411 412 INP_INFO_WLOCK_ASSERT(&V_tcbinfo); 413 INP_WLOCK_ASSERT(inp); 414 415 /* 416 * XXXRW: Time wait state for inpcb has been recycled, but inpcb is 417 * still present. This is undesirable, but temporarily necessary 418 * until we work out how to handle inpcb's who's timewait state has 419 * been removed. 420 */ 421 tw = intotw(inp); 422 if (tw == NULL) 423 goto drop; 424 425 thflags = th->th_flags; 426 427 /* 428 * NOTE: for FIN_WAIT_2 (to be added later), 429 * must validate sequence number before accepting RST 430 */ 431 432 /* 433 * If the segment contains RST: 434 * Drop the segment - see Stevens, vol. 2, p. 964 and 435 * RFC 1337. 436 */ 437 if (thflags & TH_RST) 438 goto drop; 439 440#if 0 441/* PAWS not needed at the moment */ 442 /* 443 * RFC 1323 PAWS: If we have a timestamp reply on this segment 444 * and it's less than ts_recent, drop it. 445 */ 446 if ((to.to_flags & TOF_TS) != 0 && tp->ts_recent && 447 TSTMP_LT(to.to_tsval, tp->ts_recent)) { 448 if ((thflags & TH_ACK) == 0) 449 goto drop; 450 goto ack; 451 } 452 /* 453 * ts_recent is never updated because we never accept new segments. 454 */ 455#endif 456 457 /* 458 * If a new connection request is received 459 * while in TIME_WAIT, drop the old connection 460 * and start over if the sequence numbers 461 * are above the previous ones. 462 */ 463 if ((thflags & TH_SYN) && SEQ_GT(th->th_seq, tw->rcv_nxt)) { 464 tcp_twclose(tw, 0); 465 return (1); 466 } 467 468 /* 469 * Drop the segment if it does not contain an ACK. 470 */ 471 if ((thflags & TH_ACK) == 0) 472 goto drop; 473 474 /* 475 * Reset the 2MSL timer if this is a duplicate FIN. 476 */ 477 if (thflags & TH_FIN) { 478 seq = th->th_seq + tlen + (thflags & TH_SYN ? 1 : 0); 479 if (seq + 1 == tw->rcv_nxt) 480 tcp_tw_2msl_reset(tw, 1); 481 } 482 483 /* 484 * Acknowledge the segment if it has data or is not a duplicate ACK. 485 */ 486 if (thflags != TH_ACK || tlen != 0 || 487 th->th_seq != tw->rcv_nxt || th->th_ack != tw->snd_nxt) 488 tcp_twrespond(tw, TH_ACK); 489drop: 490 INP_WUNLOCK(inp); 491 m_freem(m); 492 return (0); 493} 494 495void 496tcp_twclose(struct tcptw *tw, int reuse) 497{ 498 struct socket *so; 499 struct inpcb *inp; 500 501 /* 502 * At this point, we are in one of two situations: 503 * 504 * (1) We have no socket, just an inpcb<->twtcp pair. We can free 505 * all state. 506 * 507 * (2) We have a socket -- if we own a reference, release it and 508 * notify the socket layer. 509 */ 510 inp = tw->tw_inpcb; 511 KASSERT((inp->inp_flags & INP_TIMEWAIT), ("tcp_twclose: !timewait")); 512 KASSERT(intotw(inp) == tw, ("tcp_twclose: inp_ppcb != tw")); 513 INP_INFO_WLOCK_ASSERT(&V_tcbinfo); /* in_pcbfree() */ 514 INP_WLOCK_ASSERT(inp); 515 516 tw->tw_inpcb = NULL; 517 tcp_tw_2msl_stop(tw, reuse); 518 inp->inp_ppcb = NULL; 519 in_pcbdrop(inp); 520 521 so = inp->inp_socket; 522 if (so != NULL) { 523 /* 524 * If there's a socket, handle two cases: first, we own a 525 * strong reference, which we will now release, or we don't 526 * in which case another reference exists (XXXRW: think 527 * about this more), and we don't need to take action. 528 */ 529 if (inp->inp_flags & INP_SOCKREF) { 530 inp->inp_flags &= ~INP_SOCKREF; 531 INP_WUNLOCK(inp); 532 ACCEPT_LOCK(); 533 SOCK_LOCK(so); 534 KASSERT(so->so_state & SS_PROTOREF, 535 ("tcp_twclose: INP_SOCKREF && !SS_PROTOREF")); 536 so->so_state &= ~SS_PROTOREF; 537 sofree(so); 538 } else { 539 /* 540 * If we don't own the only reference, the socket and 541 * inpcb need to be left around to be handled by 542 * tcp_usr_detach() later. 543 */ 544 INP_WUNLOCK(inp); 545 } 546 } else 547 in_pcbfree(inp); 548 TCPSTAT_INC(tcps_closed); 549} 550 551int 552tcp_twrespond(struct tcptw *tw, int flags) 553{ 554 struct inpcb *inp = tw->tw_inpcb; 555#if defined(INET6) || defined(INET) 556 struct tcphdr *th = NULL; 557#endif 558 struct mbuf *m; 559#ifdef INET 560 struct ip *ip = NULL; 561#endif 562 u_int hdrlen, optlen; 563 int error = 0; /* Keep compiler happy */ 564 struct tcpopt to; 565#ifdef INET6 566 struct ip6_hdr *ip6 = NULL; 567 int isipv6 = inp->inp_inc.inc_flags & INC_ISIPV6; 568#endif 569 hdrlen = 0; /* Keep compiler happy */ 570 571 INP_WLOCK_ASSERT(inp); 572 573 m = m_gethdr(M_NOWAIT, MT_DATA); 574 if (m == NULL) 575 return (ENOBUFS); 576 m->m_data += max_linkhdr; 577 578#ifdef MAC 579 mac_inpcb_create_mbuf(inp, m); 580#endif 581 582#ifdef INET6 583 if (isipv6) { 584 hdrlen = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); 585 ip6 = mtod(m, struct ip6_hdr *); 586 th = (struct tcphdr *)(ip6 + 1); 587 tcpip_fillheaders(inp, ip6, th); 588 } 589#endif 590#if defined(INET6) && defined(INET) 591 else 592#endif 593#ifdef INET 594 { 595 hdrlen = sizeof(struct tcpiphdr); 596 ip = mtod(m, struct ip *); 597 th = (struct tcphdr *)(ip + 1); 598 tcpip_fillheaders(inp, ip, th); 599 } 600#endif 601 to.to_flags = 0; 602 603 /* 604 * Send a timestamp and echo-reply if both our side and our peer 605 * have sent timestamps in our SYN's and this is not a RST. 606 */ 607 if (tw->t_recent && flags == TH_ACK) { 608 to.to_flags |= TOF_TS; 609 to.to_tsval = tcp_ts_getticks() + tw->ts_offset; 610 to.to_tsecr = tw->t_recent; 611 } 612 optlen = tcp_addoptions(&to, (u_char *)(th + 1)); 613 614 m->m_len = hdrlen + optlen; 615 m->m_pkthdr.len = m->m_len; 616 617 KASSERT(max_linkhdr + m->m_len <= MHLEN, ("tcptw: mbuf too small")); 618 619 th->th_seq = htonl(tw->snd_nxt); 620 th->th_ack = htonl(tw->rcv_nxt); 621 th->th_off = (sizeof(struct tcphdr) + optlen) >> 2; 622 th->th_flags = flags; 623 th->th_win = htons(tw->last_win); 624 625 m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); 626#ifdef INET6 627 if (isipv6) { 628 m->m_pkthdr.csum_flags = CSUM_TCP_IPV6; 629 th->th_sum = in6_cksum_pseudo(ip6, 630 sizeof(struct tcphdr) + optlen, IPPROTO_TCP, 0); 631 ip6->ip6_hlim = in6_selecthlim(inp, NULL); 632 error = ip6_output(m, inp->in6p_outputopts, NULL, 633 (tw->tw_so_options & SO_DONTROUTE), NULL, NULL, inp); 634 } 635#endif 636#if defined(INET6) && defined(INET) 637 else 638#endif 639#ifdef INET 640 { 641 m->m_pkthdr.csum_flags = CSUM_TCP; 642 th->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, 643 htons(sizeof(struct tcphdr) + optlen + IPPROTO_TCP)); 644 ip->ip_len = htons(m->m_pkthdr.len); 645 if (V_path_mtu_discovery) 646 ip->ip_off |= htons(IP_DF); 647 error = ip_output(m, inp->inp_options, NULL, 648 ((tw->tw_so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0), 649 NULL, inp); 650 } 651#endif 652 if (flags & TH_ACK) 653 TCPSTAT_INC(tcps_sndacks); 654 else 655 TCPSTAT_INC(tcps_sndctrl); 656 TCPSTAT_INC(tcps_sndtotal); 657 return (error); 658} 659 660static void 661tcp_tw_2msl_reset(struct tcptw *tw, int rearm) 662{ 663 664 INP_INFO_WLOCK_ASSERT(&V_tcbinfo); 665 INP_WLOCK_ASSERT(tw->tw_inpcb); 666 667 TW_WLOCK(V_tw_lock); 668 if (rearm) 669 TAILQ_REMOVE(&V_twq_2msl, tw, tw_2msl); 670 tw->tw_time = ticks + 2 * tcp_msl; 671 TAILQ_INSERT_TAIL(&V_twq_2msl, tw, tw_2msl); 672 TW_WUNLOCK(V_tw_lock); 673} 674 675static void 676tcp_tw_2msl_stop(struct tcptw *tw, int reuse) 677{ 678 679 INP_INFO_WLOCK_ASSERT(&V_tcbinfo); 680 681 TW_WLOCK(V_tw_lock); 682 TAILQ_REMOVE(&V_twq_2msl, tw, tw_2msl); 683 crfree(tw->tw_cred); 684 tw->tw_cred = NULL; 685 686 if (!reuse) { 687 tw_pcbrele(tw); 688 return; 689 } 690 691 TW_WUNLOCK(V_tw_lock); 692} 693 694struct tcptw * 695tcp_tw_2msl_reuse(void) 696{ 697 698 INP_INFO_WLOCK_ASSERT(&V_tcbinfo); 699 700 struct tcptw *tw; 701 702 TW_WLOCK(V_tw_lock); 703 tw = TAILQ_FIRST(&V_twq_2msl); 704 if (tw == NULL) { 705 TW_WUNLOCK(V_tw_lock); 706 return NULL; 707 } 708 TW_WUNLOCK(V_tw_lock); 709 710 INP_WLOCK(tw->tw_inpcb); 711 tcp_twclose(tw, 1); 712 713 return (tw); 714} 715 716void 717tcp_tw_2msl_scan(void) 718{ 719 720 struct tcptw *tw; 721 for (;;) { 722 TW_RLOCK(V_tw_lock); 723 tw = TAILQ_FIRST(&V_twq_2msl); 724 if (tw == NULL || ((tw->tw_time - ticks) > 0)) { 725 TW_RUNLOCK(V_tw_lock); 726 break; 727 } 728 tw_pcbref(tw); 729 TW_RUNLOCK(V_tw_lock); 730 731 /* Close timewait state */ 732 if (INP_INFO_TRY_WLOCK(&V_tcbinfo)) { 733 TW_WLOCK(V_tw_lock); 734 if (tw_pcbrele(tw)) { 735 INP_INFO_WUNLOCK(&V_tcbinfo); 736 continue; 737 } 738 739 KASSERT(tw->tw_inpcb != NULL, 740 ("%s: tw->tw_inpcb == NULL", __func__)); 741 742 INP_WLOCK(tw->tw_inpcb); 743 tcp_twclose(tw, 0); 744 INP_INFO_WUNLOCK(&V_tcbinfo); 745 } else { 746 /* INP_INFO lock is busy, continue later */ 747 TW_WLOCK(V_tw_lock); 748 tw_pcbrele(tw); 749 break; 750 } 751 } 752} 753