tcp_input.c revision 1.284
1 /* $NetBSD: tcp_input.c,v 1.284 2008/04/12 05:58:22 thorpej Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 /* 33 * @(#)COPYRIGHT 1.1 (NRL) 17 January 1995 34 * 35 * NRL grants permission for redistribution and use in source and binary 36 * forms, with or without modification, of the software and documentation 37 * created at NRL provided that the following conditions are met: 38 * 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 3. All advertising materials mentioning features or use of this software 45 * must display the following acknowledgements: 46 * This product includes software developed by the University of 47 * California, Berkeley and its contributors. 48 * This product includes software developed at the Information 49 * Technology Division, US Naval Research Laboratory. 50 * 4. Neither the name of the NRL nor the names of its contributors 51 * may be used to endorse or promote products derived from this software 52 * without specific prior written permission. 53 * 54 * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS 55 * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 56 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A 57 * PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NRL OR 58 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 59 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 60 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 61 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 62 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 63 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 64 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 65 * 66 * The views and conclusions contained in the software and documentation 67 * are those of the authors and should not be interpreted as representing 68 * official policies, either expressed or implied, of the US Naval 69 * Research Laboratory (NRL). 70 */ 71 72 /*- 73 * Copyright (c) 1997, 1998, 1999, 2001, 2005, 2006 The NetBSD Foundation, Inc. 74 * All rights reserved. 75 * 76 * This code is derived from software contributed to The NetBSD Foundation 77 * by Jason R. Thorpe and Kevin M. Lahey of the Numerical Aerospace Simulation 78 * Facility, NASA Ames Research Center. 79 * This code is derived from software contributed to The NetBSD Foundation 80 * by Charles M. Hannum. 81 * This code is derived from software contributed to The NetBSD Foundation 82 * by Rui Paulo. 83 * 84 * Redistribution and use in source and binary forms, with or without 85 * modification, are permitted provided that the following conditions 86 * are met: 87 * 1. Redistributions of source code must retain the above copyright 88 * notice, this list of conditions and the following disclaimer. 89 * 2. Redistributions in binary form must reproduce the above copyright 90 * notice, this list of conditions and the following disclaimer in the 91 * documentation and/or other materials provided with the distribution. 92 * 3. All advertising materials mentioning features or use of this software 93 * must display the following acknowledgement: 94 * This product includes software developed by the NetBSD 95 * Foundation, Inc. and its contributors. 96 * 4. Neither the name of The NetBSD Foundation nor the names of its 97 * contributors may be used to endorse or promote products derived 98 * from this software without specific prior written permission. 99 * 100 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 101 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 102 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 103 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 104 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 105 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 106 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 107 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 108 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 109 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 110 * POSSIBILITY OF SUCH DAMAGE. 111 */ 112 113 /* 114 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995 115 * The Regents of the University of California. All rights reserved. 116 * 117 * Redistribution and use in source and binary forms, with or without 118 * modification, are permitted provided that the following conditions 119 * are met: 120 * 1. Redistributions of source code must retain the above copyright 121 * notice, this list of conditions and the following disclaimer. 122 * 2. Redistributions in binary form must reproduce the above copyright 123 * notice, this list of conditions and the following disclaimer in the 124 * documentation and/or other materials provided with the distribution. 125 * 3. Neither the name of the University nor the names of its contributors 126 * may be used to endorse or promote products derived from this software 127 * without specific prior written permission. 128 * 129 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 130 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 131 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 132 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 133 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 134 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 135 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 136 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 137 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 138 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 139 * SUCH DAMAGE. 140 * 141 * @(#)tcp_input.c 8.12 (Berkeley) 5/24/95 142 */ 143 144 /* 145 * TODO list for SYN cache stuff: 146 * 147 * Find room for a "state" field, which is needed to keep a 148 * compressed state for TIME_WAIT TCBs. It's been noted already 149 * that this is fairly important for very high-volume web and 150 * mail servers, which use a large number of short-lived 151 * connections. 152 */ 153 154 #include <sys/cdefs.h> 155 __KERNEL_RCSID(0, "$NetBSD: tcp_input.c,v 1.284 2008/04/12 05:58:22 thorpej Exp $"); 156 157 #include "opt_inet.h" 158 #include "opt_ipsec.h" 159 #include "opt_inet_csum.h" 160 #include "opt_tcp_debug.h" 161 162 #include <sys/param.h> 163 #include <sys/systm.h> 164 #include <sys/malloc.h> 165 #include <sys/mbuf.h> 166 #include <sys/protosw.h> 167 #include <sys/socket.h> 168 #include <sys/socketvar.h> 169 #include <sys/errno.h> 170 #include <sys/syslog.h> 171 #include <sys/pool.h> 172 #include <sys/domain.h> 173 #include <sys/kernel.h> 174 #ifdef TCP_SIGNATURE 175 #include <sys/md5.h> 176 #endif 177 #include <sys/lwp.h> /* for lwp0 */ 178 179 #include <net/if.h> 180 #include <net/route.h> 181 #include <net/if_types.h> 182 183 #include <netinet/in.h> 184 #include <netinet/in_systm.h> 185 #include <netinet/ip.h> 186 #include <netinet/in_pcb.h> 187 #include <netinet/in_var.h> 188 #include <netinet/ip_var.h> 189 #include <netinet/in_offload.h> 190 191 #ifdef INET6 192 #ifndef INET 193 #include <netinet/in.h> 194 #endif 195 #include <netinet/ip6.h> 196 #include <netinet6/ip6_var.h> 197 #include <netinet6/in6_pcb.h> 198 #include <netinet6/ip6_var.h> 199 #include <netinet6/in6_var.h> 200 #include <netinet/icmp6.h> 201 #include <netinet6/nd6.h> 202 #ifdef TCP_SIGNATURE 203 #include <netinet6/scope6_var.h> 204 #endif 205 #endif 206 207 #ifndef INET6 208 /* always need ip6.h for IP6_EXTHDR_GET */ 209 #include <netinet/ip6.h> 210 #endif 211 212 #include <netinet/tcp.h> 213 #include <netinet/tcp_fsm.h> 214 #include <netinet/tcp_seq.h> 215 #include <netinet/tcp_timer.h> 216 #include <netinet/tcp_var.h> 217 #include <netinet/tcp_private.h> 218 #include <netinet/tcpip.h> 219 #include <netinet/tcp_congctl.h> 220 #include <netinet/tcp_debug.h> 221 222 #include <machine/stdarg.h> 223 224 #ifdef IPSEC 225 #include <netinet6/ipsec.h> 226 #include <netkey/key.h> 227 #endif /*IPSEC*/ 228 #ifdef INET6 229 #include "faith.h" 230 #if defined(NFAITH) && NFAITH > 0 231 #include <net/if_faith.h> 232 #endif 233 #endif /* IPSEC */ 234 235 #ifdef FAST_IPSEC 236 #include <netipsec/ipsec.h> 237 #include <netipsec/ipsec_var.h> /* XXX ipsecstat namespace */ 238 #include <netipsec/key.h> 239 #ifdef INET6 240 #include <netipsec/ipsec6.h> 241 #endif 242 #endif /* FAST_IPSEC*/ 243 244 int tcprexmtthresh = 3; 245 int tcp_log_refused; 246 247 int tcp_do_autorcvbuf = 0; 248 int tcp_autorcvbuf_inc = 16 * 1024; 249 int tcp_autorcvbuf_max = 256 * 1024; 250 251 static int tcp_rst_ppslim_count = 0; 252 static struct timeval tcp_rst_ppslim_last; 253 static int tcp_ackdrop_ppslim_count = 0; 254 static struct timeval tcp_ackdrop_ppslim_last; 255 256 #define TCP_PAWS_IDLE (24U * 24 * 60 * 60 * PR_SLOWHZ) 257 258 /* for modulo comparisons of timestamps */ 259 #define TSTMP_LT(a,b) ((int)((a)-(b)) < 0) 260 #define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0) 261 262 /* 263 * Neighbor Discovery, Neighbor Unreachability Detection Upper layer hint. 264 */ 265 #ifdef INET6 266 static inline void 267 nd6_hint(struct tcpcb *tp) 268 { 269 struct rtentry *rt; 270 271 if (tp != NULL && tp->t_in6pcb != NULL && tp->t_family == AF_INET6 && 272 (rt = rtcache_validate(&tp->t_in6pcb->in6p_route)) != NULL) 273 nd6_nud_hint(rt, NULL, 0); 274 } 275 #else 276 static inline void 277 nd6_hint(struct tcpcb *tp) 278 { 279 } 280 #endif 281 282 /* 283 * Compute ACK transmission behavior. Delay the ACK unless 284 * we have already delayed an ACK (must send an ACK every two segments). 285 * We also ACK immediately if we received a PUSH and the ACK-on-PUSH 286 * option is enabled. 287 */ 288 static void 289 tcp_setup_ack(struct tcpcb *tp, const struct tcphdr *th) 290 { 291 292 if (tp->t_flags & TF_DELACK || 293 (tcp_ack_on_push && th->th_flags & TH_PUSH)) 294 tp->t_flags |= TF_ACKNOW; 295 else 296 TCP_SET_DELACK(tp); 297 } 298 299 static void 300 icmp_check(struct tcpcb *tp, const struct tcphdr *th, int acked) 301 { 302 303 /* 304 * If we had a pending ICMP message that refers to data that have 305 * just been acknowledged, disregard the recorded ICMP message. 306 */ 307 if ((tp->t_flags & TF_PMTUD_PEND) && 308 SEQ_GT(th->th_ack, tp->t_pmtud_th_seq)) 309 tp->t_flags &= ~TF_PMTUD_PEND; 310 311 /* 312 * Keep track of the largest chunk of data 313 * acknowledged since last PMTU update 314 */ 315 if (tp->t_pmtud_mss_acked < acked) 316 tp->t_pmtud_mss_acked = acked; 317 } 318 319 /* 320 * Convert TCP protocol fields to host order for easier processing. 321 */ 322 static void 323 tcp_fields_to_host(struct tcphdr *th) 324 { 325 326 NTOHL(th->th_seq); 327 NTOHL(th->th_ack); 328 NTOHS(th->th_win); 329 NTOHS(th->th_urp); 330 } 331 332 /* 333 * ... and reverse the above. 334 */ 335 static void 336 tcp_fields_to_net(struct tcphdr *th) 337 { 338 339 HTONL(th->th_seq); 340 HTONL(th->th_ack); 341 HTONS(th->th_win); 342 HTONS(th->th_urp); 343 } 344 345 #ifdef TCP_CSUM_COUNTERS 346 #include <sys/device.h> 347 348 #if defined(INET) 349 extern struct evcnt tcp_hwcsum_ok; 350 extern struct evcnt tcp_hwcsum_bad; 351 extern struct evcnt tcp_hwcsum_data; 352 extern struct evcnt tcp_swcsum; 353 #endif /* defined(INET) */ 354 #if defined(INET6) 355 extern struct evcnt tcp6_hwcsum_ok; 356 extern struct evcnt tcp6_hwcsum_bad; 357 extern struct evcnt tcp6_hwcsum_data; 358 extern struct evcnt tcp6_swcsum; 359 #endif /* defined(INET6) */ 360 361 #define TCP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 362 363 #else 364 365 #define TCP_CSUM_COUNTER_INCR(ev) /* nothing */ 366 367 #endif /* TCP_CSUM_COUNTERS */ 368 369 #ifdef TCP_REASS_COUNTERS 370 #include <sys/device.h> 371 372 extern struct evcnt tcp_reass_; 373 extern struct evcnt tcp_reass_empty; 374 extern struct evcnt tcp_reass_iteration[8]; 375 extern struct evcnt tcp_reass_prependfirst; 376 extern struct evcnt tcp_reass_prepend; 377 extern struct evcnt tcp_reass_insert; 378 extern struct evcnt tcp_reass_inserttail; 379 extern struct evcnt tcp_reass_append; 380 extern struct evcnt tcp_reass_appendtail; 381 extern struct evcnt tcp_reass_overlaptail; 382 extern struct evcnt tcp_reass_overlapfront; 383 extern struct evcnt tcp_reass_segdup; 384 extern struct evcnt tcp_reass_fragdup; 385 386 #define TCP_REASS_COUNTER_INCR(ev) (ev)->ev_count++ 387 388 #else 389 390 #define TCP_REASS_COUNTER_INCR(ev) /* nothing */ 391 392 #endif /* TCP_REASS_COUNTERS */ 393 394 static int tcp_reass(struct tcpcb *, const struct tcphdr *, struct mbuf *, 395 int *); 396 static int tcp_dooptions(struct tcpcb *, const u_char *, int, 397 struct tcphdr *, struct mbuf *, int, struct tcp_opt_info *); 398 399 #ifdef INET 400 static void tcp4_log_refused(const struct ip *, const struct tcphdr *); 401 #endif 402 #ifdef INET6 403 static void tcp6_log_refused(const struct ip6_hdr *, const struct tcphdr *); 404 #endif 405 406 #define TRAVERSE(x) while ((x)->m_next) (x) = (x)->m_next 407 408 #if defined(MBUFTRACE) 409 struct mowner tcp_reass_mowner = MOWNER_INIT("tcp", "reass"); 410 #endif /* defined(MBUFTRACE) */ 411 412 static POOL_INIT(tcpipqent_pool, sizeof(struct ipqent), 0, 0, 0, "tcpipqepl", 413 NULL, IPL_VM); 414 415 struct ipqent * 416 tcpipqent_alloc(void) 417 { 418 struct ipqent *ipqe; 419 int s; 420 421 s = splvm(); 422 ipqe = pool_get(&tcpipqent_pool, PR_NOWAIT); 423 splx(s); 424 425 return ipqe; 426 } 427 428 void 429 tcpipqent_free(struct ipqent *ipqe) 430 { 431 int s; 432 433 s = splvm(); 434 pool_put(&tcpipqent_pool, ipqe); 435 splx(s); 436 } 437 438 static int 439 tcp_reass(struct tcpcb *tp, const struct tcphdr *th, struct mbuf *m, int *tlen) 440 { 441 struct ipqent *p, *q, *nq, *tiqe = NULL; 442 struct socket *so = NULL; 443 int pkt_flags; 444 tcp_seq pkt_seq; 445 unsigned pkt_len; 446 u_long rcvpartdupbyte = 0; 447 u_long rcvoobyte; 448 #ifdef TCP_REASS_COUNTERS 449 u_int count = 0; 450 #endif 451 uint64_t *tcps; 452 453 if (tp->t_inpcb) 454 so = tp->t_inpcb->inp_socket; 455 #ifdef INET6 456 else if (tp->t_in6pcb) 457 so = tp->t_in6pcb->in6p_socket; 458 #endif 459 460 TCP_REASS_LOCK_CHECK(tp); 461 462 /* 463 * Call with th==0 after become established to 464 * force pre-ESTABLISHED data up to user socket. 465 */ 466 if (th == 0) 467 goto present; 468 469 m_claimm(m, &tcp_reass_mowner); 470 471 rcvoobyte = *tlen; 472 /* 473 * Copy these to local variables because the tcpiphdr 474 * gets munged while we are collapsing mbufs. 475 */ 476 pkt_seq = th->th_seq; 477 pkt_len = *tlen; 478 pkt_flags = th->th_flags; 479 480 TCP_REASS_COUNTER_INCR(&tcp_reass_); 481 482 if ((p = TAILQ_LAST(&tp->segq, ipqehead)) != NULL) { 483 /* 484 * When we miss a packet, the vast majority of time we get 485 * packets that follow it in order. So optimize for that. 486 */ 487 if (pkt_seq == p->ipqe_seq + p->ipqe_len) { 488 p->ipqe_len += pkt_len; 489 p->ipqe_flags |= pkt_flags; 490 m_cat(p->ipre_mlast, m); 491 TRAVERSE(p->ipre_mlast); 492 m = NULL; 493 tiqe = p; 494 TAILQ_REMOVE(&tp->timeq, p, ipqe_timeq); 495 TCP_REASS_COUNTER_INCR(&tcp_reass_appendtail); 496 goto skip_replacement; 497 } 498 /* 499 * While we're here, if the pkt is completely beyond 500 * anything we have, just insert it at the tail. 501 */ 502 if (SEQ_GT(pkt_seq, p->ipqe_seq + p->ipqe_len)) { 503 TCP_REASS_COUNTER_INCR(&tcp_reass_inserttail); 504 goto insert_it; 505 } 506 } 507 508 q = TAILQ_FIRST(&tp->segq); 509 510 if (q != NULL) { 511 /* 512 * If this segment immediately precedes the first out-of-order 513 * block, simply slap the segment in front of it and (mostly) 514 * skip the complicated logic. 515 */ 516 if (pkt_seq + pkt_len == q->ipqe_seq) { 517 q->ipqe_seq = pkt_seq; 518 q->ipqe_len += pkt_len; 519 q->ipqe_flags |= pkt_flags; 520 m_cat(m, q->ipqe_m); 521 q->ipqe_m = m; 522 q->ipre_mlast = m; /* last mbuf may have changed */ 523 TRAVERSE(q->ipre_mlast); 524 tiqe = q; 525 TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); 526 TCP_REASS_COUNTER_INCR(&tcp_reass_prependfirst); 527 goto skip_replacement; 528 } 529 } else { 530 TCP_REASS_COUNTER_INCR(&tcp_reass_empty); 531 } 532 533 /* 534 * Find a segment which begins after this one does. 535 */ 536 for (p = NULL; q != NULL; q = nq) { 537 nq = TAILQ_NEXT(q, ipqe_q); 538 #ifdef TCP_REASS_COUNTERS 539 count++; 540 #endif 541 /* 542 * If the received segment is just right after this 543 * fragment, merge the two together and then check 544 * for further overlaps. 545 */ 546 if (q->ipqe_seq + q->ipqe_len == pkt_seq) { 547 #ifdef TCPREASS_DEBUG 548 printf("tcp_reass[%p]: concat %u:%u(%u) to %u:%u(%u)\n", 549 tp, pkt_seq, pkt_seq + pkt_len, pkt_len, 550 q->ipqe_seq, q->ipqe_seq + q->ipqe_len, q->ipqe_len); 551 #endif 552 pkt_len += q->ipqe_len; 553 pkt_flags |= q->ipqe_flags; 554 pkt_seq = q->ipqe_seq; 555 m_cat(q->ipre_mlast, m); 556 TRAVERSE(q->ipre_mlast); 557 m = q->ipqe_m; 558 TCP_REASS_COUNTER_INCR(&tcp_reass_append); 559 goto free_ipqe; 560 } 561 /* 562 * If the received segment is completely past this 563 * fragment, we need to go the next fragment. 564 */ 565 if (SEQ_LT(q->ipqe_seq + q->ipqe_len, pkt_seq)) { 566 p = q; 567 continue; 568 } 569 /* 570 * If the fragment is past the received segment, 571 * it (or any following) can't be concatenated. 572 */ 573 if (SEQ_GT(q->ipqe_seq, pkt_seq + pkt_len)) { 574 TCP_REASS_COUNTER_INCR(&tcp_reass_insert); 575 break; 576 } 577 578 /* 579 * We've received all the data in this segment before. 580 * mark it as a duplicate and return. 581 */ 582 if (SEQ_LEQ(q->ipqe_seq, pkt_seq) && 583 SEQ_GEQ(q->ipqe_seq + q->ipqe_len, pkt_seq + pkt_len)) { 584 tcps = TCP_STAT_GETREF(); 585 tcps[TCP_STAT_RCVDUPPACK]++; 586 tcps[TCP_STAT_RCVDUPBYTE] += pkt_len; 587 TCP_STAT_PUTREF(); 588 tcp_new_dsack(tp, pkt_seq, pkt_len); 589 m_freem(m); 590 if (tiqe != NULL) { 591 tcpipqent_free(tiqe); 592 } 593 TCP_REASS_COUNTER_INCR(&tcp_reass_segdup); 594 return (0); 595 } 596 /* 597 * Received segment completely overlaps this fragment 598 * so we drop the fragment (this keeps the temporal 599 * ordering of segments correct). 600 */ 601 if (SEQ_GEQ(q->ipqe_seq, pkt_seq) && 602 SEQ_LEQ(q->ipqe_seq + q->ipqe_len, pkt_seq + pkt_len)) { 603 rcvpartdupbyte += q->ipqe_len; 604 m_freem(q->ipqe_m); 605 TCP_REASS_COUNTER_INCR(&tcp_reass_fragdup); 606 goto free_ipqe; 607 } 608 /* 609 * RX'ed segment extends past the end of the 610 * fragment. Drop the overlapping bytes. Then 611 * merge the fragment and segment then treat as 612 * a longer received packet. 613 */ 614 if (SEQ_LT(q->ipqe_seq, pkt_seq) && 615 SEQ_GT(q->ipqe_seq + q->ipqe_len, pkt_seq)) { 616 int overlap = q->ipqe_seq + q->ipqe_len - pkt_seq; 617 #ifdef TCPREASS_DEBUG 618 printf("tcp_reass[%p]: trim starting %d bytes of %u:%u(%u)\n", 619 tp, overlap, 620 pkt_seq, pkt_seq + pkt_len, pkt_len); 621 #endif 622 m_adj(m, overlap); 623 rcvpartdupbyte += overlap; 624 m_cat(q->ipre_mlast, m); 625 TRAVERSE(q->ipre_mlast); 626 m = q->ipqe_m; 627 pkt_seq = q->ipqe_seq; 628 pkt_len += q->ipqe_len - overlap; 629 rcvoobyte -= overlap; 630 TCP_REASS_COUNTER_INCR(&tcp_reass_overlaptail); 631 goto free_ipqe; 632 } 633 /* 634 * RX'ed segment extends past the front of the 635 * fragment. Drop the overlapping bytes on the 636 * received packet. The packet will then be 637 * contatentated with this fragment a bit later. 638 */ 639 if (SEQ_GT(q->ipqe_seq, pkt_seq) && 640 SEQ_LT(q->ipqe_seq, pkt_seq + pkt_len)) { 641 int overlap = pkt_seq + pkt_len - q->ipqe_seq; 642 #ifdef TCPREASS_DEBUG 643 printf("tcp_reass[%p]: trim trailing %d bytes of %u:%u(%u)\n", 644 tp, overlap, 645 pkt_seq, pkt_seq + pkt_len, pkt_len); 646 #endif 647 m_adj(m, -overlap); 648 pkt_len -= overlap; 649 rcvpartdupbyte += overlap; 650 TCP_REASS_COUNTER_INCR(&tcp_reass_overlapfront); 651 rcvoobyte -= overlap; 652 } 653 /* 654 * If the received segment immediates precedes this 655 * fragment then tack the fragment onto this segment 656 * and reinsert the data. 657 */ 658 if (q->ipqe_seq == pkt_seq + pkt_len) { 659 #ifdef TCPREASS_DEBUG 660 printf("tcp_reass[%p]: append %u:%u(%u) to %u:%u(%u)\n", 661 tp, q->ipqe_seq, q->ipqe_seq + q->ipqe_len, q->ipqe_len, 662 pkt_seq, pkt_seq + pkt_len, pkt_len); 663 #endif 664 pkt_len += q->ipqe_len; 665 pkt_flags |= q->ipqe_flags; 666 m_cat(m, q->ipqe_m); 667 TAILQ_REMOVE(&tp->segq, q, ipqe_q); 668 TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); 669 tp->t_segqlen--; 670 KASSERT(tp->t_segqlen >= 0); 671 KASSERT(tp->t_segqlen != 0 || 672 (TAILQ_EMPTY(&tp->segq) && 673 TAILQ_EMPTY(&tp->timeq))); 674 if (tiqe == NULL) { 675 tiqe = q; 676 } else { 677 tcpipqent_free(q); 678 } 679 TCP_REASS_COUNTER_INCR(&tcp_reass_prepend); 680 break; 681 } 682 /* 683 * If the fragment is before the segment, remember it. 684 * When this loop is terminated, p will contain the 685 * pointer to fragment that is right before the received 686 * segment. 687 */ 688 if (SEQ_LEQ(q->ipqe_seq, pkt_seq)) 689 p = q; 690 691 continue; 692 693 /* 694 * This is a common operation. It also will allow 695 * to save doing a malloc/free in most instances. 696 */ 697 free_ipqe: 698 TAILQ_REMOVE(&tp->segq, q, ipqe_q); 699 TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); 700 tp->t_segqlen--; 701 KASSERT(tp->t_segqlen >= 0); 702 KASSERT(tp->t_segqlen != 0 || 703 (TAILQ_EMPTY(&tp->segq) && TAILQ_EMPTY(&tp->timeq))); 704 if (tiqe == NULL) { 705 tiqe = q; 706 } else { 707 tcpipqent_free(q); 708 } 709 } 710 711 #ifdef TCP_REASS_COUNTERS 712 if (count > 7) 713 TCP_REASS_COUNTER_INCR(&tcp_reass_iteration[0]); 714 else if (count > 0) 715 TCP_REASS_COUNTER_INCR(&tcp_reass_iteration[count]); 716 #endif 717 718 insert_it: 719 720 /* 721 * Allocate a new queue entry since the received segment did not 722 * collapse onto any other out-of-order block; thus we are allocating 723 * a new block. If it had collapsed, tiqe would not be NULL and 724 * we would be reusing it. 725 * XXX If we can't, just drop the packet. XXX 726 */ 727 if (tiqe == NULL) { 728 tiqe = tcpipqent_alloc(); 729 if (tiqe == NULL) { 730 TCP_STATINC(TCP_STAT_RCVMEMDROP); 731 m_freem(m); 732 return (0); 733 } 734 } 735 736 /* 737 * Update the counters. 738 */ 739 tcps = TCP_STAT_GETREF(); 740 tcps[TCP_STAT_RCVOOPACK]++; 741 tcps[TCP_STAT_RCVOOBYTE] += rcvoobyte; 742 if (rcvpartdupbyte) { 743 tcps[TCP_STAT_RCVPARTDUPPACK]++; 744 tcps[TCP_STAT_RCVPARTDUPBYTE] += rcvpartdupbyte; 745 } 746 TCP_STAT_PUTREF(); 747 748 /* 749 * Insert the new fragment queue entry into both queues. 750 */ 751 tiqe->ipqe_m = m; 752 tiqe->ipre_mlast = m; 753 tiqe->ipqe_seq = pkt_seq; 754 tiqe->ipqe_len = pkt_len; 755 tiqe->ipqe_flags = pkt_flags; 756 if (p == NULL) { 757 TAILQ_INSERT_HEAD(&tp->segq, tiqe, ipqe_q); 758 #ifdef TCPREASS_DEBUG 759 if (tiqe->ipqe_seq != tp->rcv_nxt) 760 printf("tcp_reass[%p]: insert %u:%u(%u) at front\n", 761 tp, pkt_seq, pkt_seq + pkt_len, pkt_len); 762 #endif 763 } else { 764 TAILQ_INSERT_AFTER(&tp->segq, p, tiqe, ipqe_q); 765 #ifdef TCPREASS_DEBUG 766 printf("tcp_reass[%p]: insert %u:%u(%u) after %u:%u(%u)\n", 767 tp, pkt_seq, pkt_seq + pkt_len, pkt_len, 768 p->ipqe_seq, p->ipqe_seq + p->ipqe_len, p->ipqe_len); 769 #endif 770 } 771 tp->t_segqlen++; 772 773 skip_replacement: 774 775 TAILQ_INSERT_HEAD(&tp->timeq, tiqe, ipqe_timeq); 776 777 present: 778 /* 779 * Present data to user, advancing rcv_nxt through 780 * completed sequence space. 781 */ 782 if (TCPS_HAVEESTABLISHED(tp->t_state) == 0) 783 return (0); 784 q = TAILQ_FIRST(&tp->segq); 785 if (q == NULL || q->ipqe_seq != tp->rcv_nxt) 786 return (0); 787 if (tp->t_state == TCPS_SYN_RECEIVED && q->ipqe_len) 788 return (0); 789 790 tp->rcv_nxt += q->ipqe_len; 791 pkt_flags = q->ipqe_flags & TH_FIN; 792 nd6_hint(tp); 793 794 TAILQ_REMOVE(&tp->segq, q, ipqe_q); 795 TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); 796 tp->t_segqlen--; 797 KASSERT(tp->t_segqlen >= 0); 798 KASSERT(tp->t_segqlen != 0 || 799 (TAILQ_EMPTY(&tp->segq) && TAILQ_EMPTY(&tp->timeq))); 800 if (so->so_state & SS_CANTRCVMORE) 801 m_freem(q->ipqe_m); 802 else 803 sbappendstream(&so->so_rcv, q->ipqe_m); 804 tcpipqent_free(q); 805 sorwakeup(so); 806 return (pkt_flags); 807 } 808 809 #ifdef INET6 810 int 811 tcp6_input(struct mbuf **mp, int *offp, int proto) 812 { 813 struct mbuf *m = *mp; 814 815 /* 816 * draft-itojun-ipv6-tcp-to-anycast 817 * better place to put this in? 818 */ 819 if (m->m_flags & M_ANYCAST6) { 820 struct ip6_hdr *ip6; 821 if (m->m_len < sizeof(struct ip6_hdr)) { 822 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 823 TCP_STATINC(TCP_STAT_RCVSHORT); 824 return IPPROTO_DONE; 825 } 826 } 827 ip6 = mtod(m, struct ip6_hdr *); 828 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR, 829 (char *)&ip6->ip6_dst - (char *)ip6); 830 return IPPROTO_DONE; 831 } 832 833 tcp_input(m, *offp, proto); 834 return IPPROTO_DONE; 835 } 836 #endif 837 838 #ifdef INET 839 static void 840 tcp4_log_refused(const struct ip *ip, const struct tcphdr *th) 841 { 842 char src[4*sizeof "123"]; 843 char dst[4*sizeof "123"]; 844 845 if (ip) { 846 strlcpy(src, inet_ntoa(ip->ip_src), sizeof(src)); 847 strlcpy(dst, inet_ntoa(ip->ip_dst), sizeof(dst)); 848 } 849 else { 850 strlcpy(src, "(unknown)", sizeof(src)); 851 strlcpy(dst, "(unknown)", sizeof(dst)); 852 } 853 log(LOG_INFO, 854 "Connection attempt to TCP %s:%d from %s:%d\n", 855 dst, ntohs(th->th_dport), 856 src, ntohs(th->th_sport)); 857 } 858 #endif 859 860 #ifdef INET6 861 static void 862 tcp6_log_refused(const struct ip6_hdr *ip6, const struct tcphdr *th) 863 { 864 char src[INET6_ADDRSTRLEN]; 865 char dst[INET6_ADDRSTRLEN]; 866 867 if (ip6) { 868 strlcpy(src, ip6_sprintf(&ip6->ip6_src), sizeof(src)); 869 strlcpy(dst, ip6_sprintf(&ip6->ip6_dst), sizeof(dst)); 870 } 871 else { 872 strlcpy(src, "(unknown v6)", sizeof(src)); 873 strlcpy(dst, "(unknown v6)", sizeof(dst)); 874 } 875 log(LOG_INFO, 876 "Connection attempt to TCP [%s]:%d from [%s]:%d\n", 877 dst, ntohs(th->th_dport), 878 src, ntohs(th->th_sport)); 879 } 880 #endif 881 882 /* 883 * Checksum extended TCP header and data. 884 */ 885 int 886 tcp_input_checksum(int af, struct mbuf *m, const struct tcphdr *th, 887 int toff, int off, int tlen) 888 { 889 890 /* 891 * XXX it's better to record and check if this mbuf is 892 * already checked. 893 */ 894 895 switch (af) { 896 #ifdef INET 897 case AF_INET: 898 switch (m->m_pkthdr.csum_flags & 899 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_TCPv4) | 900 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { 901 case M_CSUM_TCPv4|M_CSUM_TCP_UDP_BAD: 902 TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_bad); 903 goto badcsum; 904 905 case M_CSUM_TCPv4|M_CSUM_DATA: { 906 u_int32_t hw_csum = m->m_pkthdr.csum_data; 907 908 TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_data); 909 if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) { 910 const struct ip *ip = 911 mtod(m, const struct ip *); 912 913 hw_csum = in_cksum_phdr(ip->ip_src.s_addr, 914 ip->ip_dst.s_addr, 915 htons(hw_csum + tlen + off + IPPROTO_TCP)); 916 } 917 if ((hw_csum ^ 0xffff) != 0) 918 goto badcsum; 919 break; 920 } 921 922 case M_CSUM_TCPv4: 923 /* Checksum was okay. */ 924 TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_ok); 925 break; 926 927 default: 928 /* 929 * Must compute it ourselves. Maybe skip checksum 930 * on loopback interfaces. 931 */ 932 if (__predict_true(!(m->m_pkthdr.rcvif->if_flags & 933 IFF_LOOPBACK) || 934 tcp_do_loopback_cksum)) { 935 TCP_CSUM_COUNTER_INCR(&tcp_swcsum); 936 if (in4_cksum(m, IPPROTO_TCP, toff, 937 tlen + off) != 0) 938 goto badcsum; 939 } 940 break; 941 } 942 break; 943 #endif /* INET4 */ 944 945 #ifdef INET6 946 case AF_INET6: 947 switch (m->m_pkthdr.csum_flags & 948 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_TCPv6) | 949 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { 950 case M_CSUM_TCPv6|M_CSUM_TCP_UDP_BAD: 951 TCP_CSUM_COUNTER_INCR(&tcp6_hwcsum_bad); 952 goto badcsum; 953 954 #if 0 /* notyet */ 955 case M_CSUM_TCPv6|M_CSUM_DATA: 956 #endif 957 958 case M_CSUM_TCPv6: 959 /* Checksum was okay. */ 960 TCP_CSUM_COUNTER_INCR(&tcp6_hwcsum_ok); 961 break; 962 963 default: 964 /* 965 * Must compute it ourselves. Maybe skip checksum 966 * on loopback interfaces. 967 */ 968 if (__predict_true((m->m_flags & M_LOOP) == 0 || 969 tcp_do_loopback_cksum)) { 970 TCP_CSUM_COUNTER_INCR(&tcp6_swcsum); 971 if (in6_cksum(m, IPPROTO_TCP, toff, 972 tlen + off) != 0) 973 goto badcsum; 974 } 975 } 976 break; 977 #endif /* INET6 */ 978 } 979 980 return 0; 981 982 badcsum: 983 TCP_STATINC(TCP_STAT_RCVBADSUM); 984 return -1; 985 } 986 987 /* 988 * TCP input routine, follows pages 65-76 of RFC 793 very closely. 989 */ 990 void 991 tcp_input(struct mbuf *m, ...) 992 { 993 struct tcphdr *th; 994 struct ip *ip; 995 struct inpcb *inp; 996 #ifdef INET6 997 struct ip6_hdr *ip6; 998 struct in6pcb *in6p; 999 #endif 1000 u_int8_t *optp = NULL; 1001 int optlen = 0; 1002 int len, tlen, toff, hdroptlen = 0; 1003 struct tcpcb *tp = 0; 1004 int tiflags; 1005 struct socket *so = NULL; 1006 int todrop, dupseg, acked, ourfinisacked, needoutput = 0; 1007 #ifdef TCP_DEBUG 1008 short ostate = 0; 1009 #endif 1010 u_long tiwin; 1011 struct tcp_opt_info opti; 1012 int off, iphlen; 1013 va_list ap; 1014 int af; /* af on the wire */ 1015 struct mbuf *tcp_saveti = NULL; 1016 uint32_t ts_rtt; 1017 uint8_t iptos; 1018 uint64_t *tcps; 1019 1020 MCLAIM(m, &tcp_rx_mowner); 1021 va_start(ap, m); 1022 toff = va_arg(ap, int); 1023 (void)va_arg(ap, int); /* ignore value, advance ap */ 1024 va_end(ap); 1025 1026 TCP_STATINC(TCP_STAT_RCVTOTAL); 1027 1028 bzero(&opti, sizeof(opti)); 1029 opti.ts_present = 0; 1030 opti.maxseg = 0; 1031 1032 /* 1033 * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN. 1034 * 1035 * TCP is, by definition, unicast, so we reject all 1036 * multicast outright. 1037 * 1038 * Note, there are additional src/dst address checks in 1039 * the AF-specific code below. 1040 */ 1041 if (m->m_flags & (M_BCAST|M_MCAST)) { 1042 /* XXX stat */ 1043 goto drop; 1044 } 1045 #ifdef INET6 1046 if (m->m_flags & M_ANYCAST6) { 1047 /* XXX stat */ 1048 goto drop; 1049 } 1050 #endif 1051 1052 /* 1053 * Get IP and TCP header. 1054 * Note: IP leaves IP header in first mbuf. 1055 */ 1056 ip = mtod(m, struct ip *); 1057 #ifdef INET6 1058 ip6 = NULL; 1059 #endif 1060 switch (ip->ip_v) { 1061 #ifdef INET 1062 case 4: 1063 af = AF_INET; 1064 iphlen = sizeof(struct ip); 1065 ip = mtod(m, struct ip *); 1066 IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, 1067 sizeof(struct tcphdr)); 1068 if (th == NULL) { 1069 TCP_STATINC(TCP_STAT_RCVSHORT); 1070 return; 1071 } 1072 /* We do the checksum after PCB lookup... */ 1073 len = ntohs(ip->ip_len); 1074 tlen = len - toff; 1075 iptos = ip->ip_tos; 1076 break; 1077 #endif 1078 #ifdef INET6 1079 case 6: 1080 ip = NULL; 1081 iphlen = sizeof(struct ip6_hdr); 1082 af = AF_INET6; 1083 ip6 = mtod(m, struct ip6_hdr *); 1084 IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, 1085 sizeof(struct tcphdr)); 1086 if (th == NULL) { 1087 TCP_STATINC(TCP_STAT_RCVSHORT); 1088 return; 1089 } 1090 1091 /* Be proactive about malicious use of IPv4 mapped address */ 1092 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 1093 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 1094 /* XXX stat */ 1095 goto drop; 1096 } 1097 1098 /* 1099 * Be proactive about unspecified IPv6 address in source. 1100 * As we use all-zero to indicate unbounded/unconnected pcb, 1101 * unspecified IPv6 address can be used to confuse us. 1102 * 1103 * Note that packets with unspecified IPv6 destination is 1104 * already dropped in ip6_input. 1105 */ 1106 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 1107 /* XXX stat */ 1108 goto drop; 1109 } 1110 1111 /* 1112 * Make sure destination address is not multicast. 1113 * Source address checked in ip6_input(). 1114 */ 1115 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 1116 /* XXX stat */ 1117 goto drop; 1118 } 1119 1120 /* We do the checksum after PCB lookup... */ 1121 len = m->m_pkthdr.len; 1122 tlen = len - toff; 1123 iptos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 1124 break; 1125 #endif 1126 default: 1127 m_freem(m); 1128 return; 1129 } 1130 1131 KASSERT(TCP_HDR_ALIGNED_P(th)); 1132 1133 /* 1134 * Check that TCP offset makes sense, 1135 * pull out TCP options and adjust length. XXX 1136 */ 1137 off = th->th_off << 2; 1138 if (off < sizeof (struct tcphdr) || off > tlen) { 1139 TCP_STATINC(TCP_STAT_RCVBADOFF); 1140 goto drop; 1141 } 1142 tlen -= off; 1143 1144 /* 1145 * tcp_input() has been modified to use tlen to mean the TCP data 1146 * length throughout the function. Other functions can use 1147 * m->m_pkthdr.len as the basis for calculating the TCP data length. 1148 * rja 1149 */ 1150 1151 if (off > sizeof (struct tcphdr)) { 1152 IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, off); 1153 if (th == NULL) { 1154 TCP_STATINC(TCP_STAT_RCVSHORT); 1155 return; 1156 } 1157 /* 1158 * NOTE: ip/ip6 will not be affected by m_pulldown() 1159 * (as they're before toff) and we don't need to update those. 1160 */ 1161 KASSERT(TCP_HDR_ALIGNED_P(th)); 1162 optlen = off - sizeof (struct tcphdr); 1163 optp = ((u_int8_t *)th) + sizeof(struct tcphdr); 1164 /* 1165 * Do quick retrieval of timestamp options ("options 1166 * prediction?"). If timestamp is the only option and it's 1167 * formatted as recommended in RFC 1323 appendix A, we 1168 * quickly get the values now and not bother calling 1169 * tcp_dooptions(), etc. 1170 */ 1171 if ((optlen == TCPOLEN_TSTAMP_APPA || 1172 (optlen > TCPOLEN_TSTAMP_APPA && 1173 optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) && 1174 *(u_int32_t *)optp == htonl(TCPOPT_TSTAMP_HDR) && 1175 (th->th_flags & TH_SYN) == 0) { 1176 opti.ts_present = 1; 1177 opti.ts_val = ntohl(*(u_int32_t *)(optp + 4)); 1178 opti.ts_ecr = ntohl(*(u_int32_t *)(optp + 8)); 1179 optp = NULL; /* we've parsed the options */ 1180 } 1181 } 1182 tiflags = th->th_flags; 1183 1184 /* 1185 * Locate pcb for segment. 1186 */ 1187 findpcb: 1188 inp = NULL; 1189 #ifdef INET6 1190 in6p = NULL; 1191 #endif 1192 switch (af) { 1193 #ifdef INET 1194 case AF_INET: 1195 inp = in_pcblookup_connect(&tcbtable, ip->ip_src, th->th_sport, 1196 ip->ip_dst, th->th_dport); 1197 if (inp == 0) { 1198 TCP_STATINC(TCP_STAT_PCBHASHMISS); 1199 inp = in_pcblookup_bind(&tcbtable, ip->ip_dst, th->th_dport); 1200 } 1201 #ifdef INET6 1202 if (inp == 0) { 1203 struct in6_addr s, d; 1204 1205 /* mapped addr case */ 1206 bzero(&s, sizeof(s)); 1207 s.s6_addr16[5] = htons(0xffff); 1208 bcopy(&ip->ip_src, &s.s6_addr32[3], sizeof(ip->ip_src)); 1209 bzero(&d, sizeof(d)); 1210 d.s6_addr16[5] = htons(0xffff); 1211 bcopy(&ip->ip_dst, &d.s6_addr32[3], sizeof(ip->ip_dst)); 1212 in6p = in6_pcblookup_connect(&tcbtable, &s, 1213 th->th_sport, &d, th->th_dport, 0); 1214 if (in6p == 0) { 1215 TCP_STATINC(TCP_STAT_PCBHASHMISS); 1216 in6p = in6_pcblookup_bind(&tcbtable, &d, 1217 th->th_dport, 0); 1218 } 1219 } 1220 #endif 1221 #ifndef INET6 1222 if (inp == 0) 1223 #else 1224 if (inp == 0 && in6p == 0) 1225 #endif 1226 { 1227 TCP_STATINC(TCP_STAT_NOPORT); 1228 if (tcp_log_refused && 1229 (tiflags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN) { 1230 tcp4_log_refused(ip, th); 1231 } 1232 tcp_fields_to_host(th); 1233 goto dropwithreset_ratelim; 1234 } 1235 #if defined(IPSEC) || defined(FAST_IPSEC) 1236 if (inp && (inp->inp_socket->so_options & SO_ACCEPTCONN) == 0 && 1237 ipsec4_in_reject(m, inp)) { 1238 ipsecstat.in_polvio++; 1239 goto drop; 1240 } 1241 #ifdef INET6 1242 else if (in6p && 1243 (in6p->in6p_socket->so_options & SO_ACCEPTCONN) == 0 && 1244 ipsec6_in_reject_so(m, in6p->in6p_socket)) { 1245 ipsecstat.in_polvio++; 1246 goto drop; 1247 } 1248 #endif 1249 #endif /*IPSEC*/ 1250 break; 1251 #endif /*INET*/ 1252 #ifdef INET6 1253 case AF_INET6: 1254 { 1255 int faith; 1256 1257 #if defined(NFAITH) && NFAITH > 0 1258 faith = faithprefix(&ip6->ip6_dst); 1259 #else 1260 faith = 0; 1261 #endif 1262 in6p = in6_pcblookup_connect(&tcbtable, &ip6->ip6_src, 1263 th->th_sport, &ip6->ip6_dst, th->th_dport, faith); 1264 if (in6p == NULL) { 1265 TCP_STATINC(TCP_STAT_PCBHASHMISS); 1266 in6p = in6_pcblookup_bind(&tcbtable, &ip6->ip6_dst, 1267 th->th_dport, faith); 1268 } 1269 if (in6p == NULL) { 1270 TCP_STATINC(TCP_STAT_NOPORT); 1271 if (tcp_log_refused && 1272 (tiflags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN) { 1273 tcp6_log_refused(ip6, th); 1274 } 1275 tcp_fields_to_host(th); 1276 goto dropwithreset_ratelim; 1277 } 1278 #if defined(IPSEC) || defined(FAST_IPSEC) 1279 if ((in6p->in6p_socket->so_options & SO_ACCEPTCONN) == 0 && 1280 ipsec6_in_reject(m, in6p)) { 1281 ipsec6stat.in_polvio++; 1282 goto drop; 1283 } 1284 #endif /*IPSEC*/ 1285 break; 1286 } 1287 #endif 1288 } 1289 1290 /* 1291 * If the state is CLOSED (i.e., TCB does not exist) then 1292 * all data in the incoming segment is discarded. 1293 * If the TCB exists but is in CLOSED state, it is embryonic, 1294 * but should either do a listen or a connect soon. 1295 */ 1296 tp = NULL; 1297 so = NULL; 1298 if (inp) { 1299 tp = intotcpcb(inp); 1300 so = inp->inp_socket; 1301 } 1302 #ifdef INET6 1303 else if (in6p) { 1304 tp = in6totcpcb(in6p); 1305 so = in6p->in6p_socket; 1306 } 1307 #endif 1308 if (tp == 0) { 1309 tcp_fields_to_host(th); 1310 goto dropwithreset_ratelim; 1311 } 1312 if (tp->t_state == TCPS_CLOSED) 1313 goto drop; 1314 1315 /* 1316 * Checksum extended TCP header and data. 1317 */ 1318 if (tcp_input_checksum(af, m, th, toff, off, tlen)) 1319 goto badcsum; 1320 1321 tcp_fields_to_host(th); 1322 1323 /* Unscale the window into a 32-bit value. */ 1324 if ((tiflags & TH_SYN) == 0) 1325 tiwin = th->th_win << tp->snd_scale; 1326 else 1327 tiwin = th->th_win; 1328 1329 #ifdef INET6 1330 /* save packet options if user wanted */ 1331 if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS)) { 1332 if (in6p->in6p_options) { 1333 m_freem(in6p->in6p_options); 1334 in6p->in6p_options = 0; 1335 } 1336 KASSERT(ip6 != NULL); 1337 ip6_savecontrol(in6p, &in6p->in6p_options, ip6, m); 1338 } 1339 #endif 1340 1341 if (so->so_options & (SO_DEBUG|SO_ACCEPTCONN)) { 1342 union syn_cache_sa src; 1343 union syn_cache_sa dst; 1344 1345 bzero(&src, sizeof(src)); 1346 bzero(&dst, sizeof(dst)); 1347 switch (af) { 1348 #ifdef INET 1349 case AF_INET: 1350 src.sin.sin_len = sizeof(struct sockaddr_in); 1351 src.sin.sin_family = AF_INET; 1352 src.sin.sin_addr = ip->ip_src; 1353 src.sin.sin_port = th->th_sport; 1354 1355 dst.sin.sin_len = sizeof(struct sockaddr_in); 1356 dst.sin.sin_family = AF_INET; 1357 dst.sin.sin_addr = ip->ip_dst; 1358 dst.sin.sin_port = th->th_dport; 1359 break; 1360 #endif 1361 #ifdef INET6 1362 case AF_INET6: 1363 src.sin6.sin6_len = sizeof(struct sockaddr_in6); 1364 src.sin6.sin6_family = AF_INET6; 1365 src.sin6.sin6_addr = ip6->ip6_src; 1366 src.sin6.sin6_port = th->th_sport; 1367 1368 dst.sin6.sin6_len = sizeof(struct sockaddr_in6); 1369 dst.sin6.sin6_family = AF_INET6; 1370 dst.sin6.sin6_addr = ip6->ip6_dst; 1371 dst.sin6.sin6_port = th->th_dport; 1372 break; 1373 #endif /* INET6 */ 1374 default: 1375 goto badsyn; /*sanity*/ 1376 } 1377 1378 if (so->so_options & SO_DEBUG) { 1379 #ifdef TCP_DEBUG 1380 ostate = tp->t_state; 1381 #endif 1382 1383 tcp_saveti = NULL; 1384 if (iphlen + sizeof(struct tcphdr) > MHLEN) 1385 goto nosave; 1386 1387 if (m->m_len > iphlen && (m->m_flags & M_EXT) == 0) { 1388 tcp_saveti = m_copym(m, 0, iphlen, M_DONTWAIT); 1389 if (!tcp_saveti) 1390 goto nosave; 1391 } else { 1392 MGETHDR(tcp_saveti, M_DONTWAIT, MT_HEADER); 1393 if (!tcp_saveti) 1394 goto nosave; 1395 MCLAIM(m, &tcp_mowner); 1396 tcp_saveti->m_len = iphlen; 1397 m_copydata(m, 0, iphlen, 1398 mtod(tcp_saveti, void *)); 1399 } 1400 1401 if (M_TRAILINGSPACE(tcp_saveti) < sizeof(struct tcphdr)) { 1402 m_freem(tcp_saveti); 1403 tcp_saveti = NULL; 1404 } else { 1405 tcp_saveti->m_len += sizeof(struct tcphdr); 1406 memcpy(mtod(tcp_saveti, char *) + iphlen, th, 1407 sizeof(struct tcphdr)); 1408 } 1409 nosave:; 1410 } 1411 if (so->so_options & SO_ACCEPTCONN) { 1412 if ((tiflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) { 1413 if (tiflags & TH_RST) { 1414 syn_cache_reset(&src.sa, &dst.sa, th); 1415 } else if ((tiflags & (TH_ACK|TH_SYN)) == 1416 (TH_ACK|TH_SYN)) { 1417 /* 1418 * Received a SYN,ACK. This should 1419 * never happen while we are in 1420 * LISTEN. Send an RST. 1421 */ 1422 goto badsyn; 1423 } else if (tiflags & TH_ACK) { 1424 so = syn_cache_get(&src.sa, &dst.sa, 1425 th, toff, tlen, so, m); 1426 if (so == NULL) { 1427 /* 1428 * We don't have a SYN for 1429 * this ACK; send an RST. 1430 */ 1431 goto badsyn; 1432 } else if (so == 1433 (struct socket *)(-1)) { 1434 /* 1435 * We were unable to create 1436 * the connection. If the 1437 * 3-way handshake was 1438 * completed, and RST has 1439 * been sent to the peer. 1440 * Since the mbuf might be 1441 * in use for the reply, 1442 * do not free it. 1443 */ 1444 m = NULL; 1445 } else { 1446 /* 1447 * We have created a 1448 * full-blown connection. 1449 */ 1450 tp = NULL; 1451 inp = NULL; 1452 #ifdef INET6 1453 in6p = NULL; 1454 #endif 1455 switch (so->so_proto->pr_domain->dom_family) { 1456 #ifdef INET 1457 case AF_INET: 1458 inp = sotoinpcb(so); 1459 tp = intotcpcb(inp); 1460 break; 1461 #endif 1462 #ifdef INET6 1463 case AF_INET6: 1464 in6p = sotoin6pcb(so); 1465 tp = in6totcpcb(in6p); 1466 break; 1467 #endif 1468 } 1469 if (tp == NULL) 1470 goto badsyn; /*XXX*/ 1471 tiwin <<= tp->snd_scale; 1472 goto after_listen; 1473 } 1474 } else { 1475 /* 1476 * None of RST, SYN or ACK was set. 1477 * This is an invalid packet for a 1478 * TCB in LISTEN state. Send a RST. 1479 */ 1480 goto badsyn; 1481 } 1482 } else { 1483 /* 1484 * Received a SYN. 1485 * 1486 * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN 1487 */ 1488 if (m->m_flags & (M_BCAST|M_MCAST)) 1489 goto drop; 1490 1491 switch (af) { 1492 #ifdef INET6 1493 case AF_INET6: 1494 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 1495 goto drop; 1496 break; 1497 #endif /* INET6 */ 1498 case AF_INET: 1499 if (IN_MULTICAST(ip->ip_dst.s_addr) || 1500 in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) 1501 goto drop; 1502 break; 1503 } 1504 1505 #ifdef INET6 1506 /* 1507 * If deprecated address is forbidden, we do 1508 * not accept SYN to deprecated interface 1509 * address to prevent any new inbound 1510 * connection from getting established. 1511 * When we do not accept SYN, we send a TCP 1512 * RST, with deprecated source address (instead 1513 * of dropping it). We compromise it as it is 1514 * much better for peer to send a RST, and 1515 * RST will be the final packet for the 1516 * exchange. 1517 * 1518 * If we do not forbid deprecated addresses, we 1519 * accept the SYN packet. RFC2462 does not 1520 * suggest dropping SYN in this case. 1521 * If we decipher RFC2462 5.5.4, it says like 1522 * this: 1523 * 1. use of deprecated addr with existing 1524 * communication is okay - "SHOULD continue 1525 * to be used" 1526 * 2. use of it with new communication: 1527 * (2a) "SHOULD NOT be used if alternate 1528 * address with sufficient scope is 1529 * available" 1530 * (2b) nothing mentioned otherwise. 1531 * Here we fall into (2b) case as we have no 1532 * choice in our source address selection - we 1533 * must obey the peer. 1534 * 1535 * The wording in RFC2462 is confusing, and 1536 * there are multiple description text for 1537 * deprecated address handling - worse, they 1538 * are not exactly the same. I believe 5.5.4 1539 * is the best one, so we follow 5.5.4. 1540 */ 1541 if (af == AF_INET6 && !ip6_use_deprecated) { 1542 struct in6_ifaddr *ia6; 1543 if ((ia6 = in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, 1544 &ip6->ip6_dst)) && 1545 (ia6->ia6_flags & IN6_IFF_DEPRECATED)) { 1546 tp = NULL; 1547 goto dropwithreset; 1548 } 1549 } 1550 #endif 1551 1552 #if defined(IPSEC) || defined(FAST_IPSEC) 1553 switch (af) { 1554 #ifdef INET 1555 case AF_INET: 1556 if (ipsec4_in_reject_so(m, so)) { 1557 ipsecstat.in_polvio++; 1558 tp = NULL; 1559 goto dropwithreset; 1560 } 1561 break; 1562 #endif 1563 #ifdef INET6 1564 case AF_INET6: 1565 if (ipsec6_in_reject_so(m, so)) { 1566 ipsec6stat.in_polvio++; 1567 tp = NULL; 1568 goto dropwithreset; 1569 } 1570 break; 1571 #endif /*INET6*/ 1572 } 1573 #endif /*IPSEC*/ 1574 1575 /* 1576 * LISTEN socket received a SYN 1577 * from itself? This can't possibly 1578 * be valid; drop the packet. 1579 */ 1580 if (th->th_sport == th->th_dport) { 1581 int i; 1582 1583 switch (af) { 1584 #ifdef INET 1585 case AF_INET: 1586 i = in_hosteq(ip->ip_src, ip->ip_dst); 1587 break; 1588 #endif 1589 #ifdef INET6 1590 case AF_INET6: 1591 i = IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &ip6->ip6_dst); 1592 break; 1593 #endif 1594 default: 1595 i = 1; 1596 } 1597 if (i) { 1598 TCP_STATINC(TCP_STAT_BADSYN); 1599 goto drop; 1600 } 1601 } 1602 1603 /* 1604 * SYN looks ok; create compressed TCP 1605 * state for it. 1606 */ 1607 if (so->so_qlen <= so->so_qlimit && 1608 syn_cache_add(&src.sa, &dst.sa, th, tlen, 1609 so, m, optp, optlen, &opti)) 1610 m = NULL; 1611 } 1612 goto drop; 1613 } 1614 } 1615 1616 after_listen: 1617 #ifdef DIAGNOSTIC 1618 /* 1619 * Should not happen now that all embryonic connections 1620 * are handled with compressed state. 1621 */ 1622 if (tp->t_state == TCPS_LISTEN) 1623 panic("tcp_input: TCPS_LISTEN"); 1624 #endif 1625 1626 /* 1627 * Segment received on connection. 1628 * Reset idle time and keep-alive timer. 1629 */ 1630 tp->t_rcvtime = tcp_now; 1631 if (TCPS_HAVEESTABLISHED(tp->t_state)) 1632 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle); 1633 1634 /* 1635 * Process options. 1636 */ 1637 #ifdef TCP_SIGNATURE 1638 if (optp || (tp->t_flags & TF_SIGNATURE)) 1639 #else 1640 if (optp) 1641 #endif 1642 if (tcp_dooptions(tp, optp, optlen, th, m, toff, &opti) < 0) 1643 goto drop; 1644 1645 if (TCP_SACK_ENABLED(tp)) { 1646 tcp_del_sackholes(tp, th); 1647 } 1648 1649 if (TCP_ECN_ALLOWED(tp)) { 1650 switch (iptos & IPTOS_ECN_MASK) { 1651 case IPTOS_ECN_CE: 1652 tp->t_flags |= TF_ECN_SND_ECE; 1653 TCP_STATINC(TCP_STAT_ECN_CE); 1654 break; 1655 case IPTOS_ECN_ECT0: 1656 TCP_STATINC(TCP_STAT_ECN_ECT); 1657 break; 1658 case IPTOS_ECN_ECT1: 1659 /* XXX: ignore for now -- rpaulo */ 1660 break; 1661 } 1662 1663 if (tiflags & TH_CWR) 1664 tp->t_flags &= ~TF_ECN_SND_ECE; 1665 1666 /* 1667 * Congestion experienced. 1668 * Ignore if we are already trying to recover. 1669 */ 1670 if ((tiflags & TH_ECE) && SEQ_GEQ(tp->snd_una, tp->snd_recover)) 1671 tp->t_congctl->cong_exp(tp); 1672 } 1673 1674 if (opti.ts_present && opti.ts_ecr) { 1675 /* 1676 * Calculate the RTT from the returned time stamp and the 1677 * connection's time base. If the time stamp is later than 1678 * the current time, or is extremely old, fall back to non-1323 1679 * RTT calculation. Since ts_ecr is unsigned, we can test both 1680 * at the same time. 1681 */ 1682 ts_rtt = TCP_TIMESTAMP(tp) - opti.ts_ecr + 1; 1683 if (ts_rtt > TCP_PAWS_IDLE) 1684 ts_rtt = 0; 1685 } else { 1686 ts_rtt = 0; 1687 } 1688 1689 /* 1690 * Header prediction: check for the two common cases 1691 * of a uni-directional data xfer. If the packet has 1692 * no control flags, is in-sequence, the window didn't 1693 * change and we're not retransmitting, it's a 1694 * candidate. If the length is zero and the ack moved 1695 * forward, we're the sender side of the xfer. Just 1696 * free the data acked & wake any higher level process 1697 * that was blocked waiting for space. If the length 1698 * is non-zero and the ack didn't move, we're the 1699 * receiver side. If we're getting packets in-order 1700 * (the reassembly queue is empty), add the data to 1701 * the socket buffer and note that we need a delayed ack. 1702 */ 1703 if (tp->t_state == TCPS_ESTABLISHED && 1704 (tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ECE|TH_CWR|TH_ACK)) 1705 == TH_ACK && 1706 (!opti.ts_present || TSTMP_GEQ(opti.ts_val, tp->ts_recent)) && 1707 th->th_seq == tp->rcv_nxt && 1708 tiwin && tiwin == tp->snd_wnd && 1709 tp->snd_nxt == tp->snd_max) { 1710 1711 /* 1712 * If last ACK falls within this segment's sequence numbers, 1713 * record the timestamp. 1714 * NOTE that the test is modified according to the latest 1715 * proposal of the tcplw (at) cray.com list (Braden 1993/04/26). 1716 * 1717 * note that we already know 1718 * TSTMP_GEQ(opti.ts_val, tp->ts_recent) 1719 */ 1720 if (opti.ts_present && 1721 SEQ_LEQ(th->th_seq, tp->last_ack_sent)) { 1722 tp->ts_recent_age = tcp_now; 1723 tp->ts_recent = opti.ts_val; 1724 } 1725 1726 if (tlen == 0) { 1727 /* Ack prediction. */ 1728 if (SEQ_GT(th->th_ack, tp->snd_una) && 1729 SEQ_LEQ(th->th_ack, tp->snd_max) && 1730 tp->snd_cwnd >= tp->snd_wnd && 1731 tp->t_partialacks < 0) { 1732 /* 1733 * this is a pure ack for outstanding data. 1734 */ 1735 if (ts_rtt) 1736 tcp_xmit_timer(tp, ts_rtt); 1737 else if (tp->t_rtttime && 1738 SEQ_GT(th->th_ack, tp->t_rtseq)) 1739 tcp_xmit_timer(tp, 1740 tcp_now - tp->t_rtttime); 1741 acked = th->th_ack - tp->snd_una; 1742 tcps = TCP_STAT_GETREF(); 1743 tcps[TCP_STAT_PREDACK]++; 1744 tcps[TCP_STAT_RCVACKPACK]++; 1745 tcps[TCP_STAT_RCVACKBYTE] += acked; 1746 TCP_STAT_PUTREF(); 1747 nd6_hint(tp); 1748 1749 if (acked > (tp->t_lastoff - tp->t_inoff)) 1750 tp->t_lastm = NULL; 1751 sbdrop(&so->so_snd, acked); 1752 tp->t_lastoff -= acked; 1753 1754 icmp_check(tp, th, acked); 1755 1756 tp->snd_una = th->th_ack; 1757 tp->snd_fack = tp->snd_una; 1758 if (SEQ_LT(tp->snd_high, tp->snd_una)) 1759 tp->snd_high = tp->snd_una; 1760 m_freem(m); 1761 1762 /* 1763 * If all outstanding data are acked, stop 1764 * retransmit timer, otherwise restart timer 1765 * using current (possibly backed-off) value. 1766 * If process is waiting for space, 1767 * wakeup/selnotify/signal. If data 1768 * are ready to send, let tcp_output 1769 * decide between more output or persist. 1770 */ 1771 if (tp->snd_una == tp->snd_max) 1772 TCP_TIMER_DISARM(tp, TCPT_REXMT); 1773 else if (TCP_TIMER_ISARMED(tp, 1774 TCPT_PERSIST) == 0) 1775 TCP_TIMER_ARM(tp, TCPT_REXMT, 1776 tp->t_rxtcur); 1777 1778 sowwakeup(so); 1779 if (so->so_snd.sb_cc) 1780 (void) tcp_output(tp); 1781 if (tcp_saveti) 1782 m_freem(tcp_saveti); 1783 return; 1784 } 1785 } else if (th->th_ack == tp->snd_una && 1786 TAILQ_FIRST(&tp->segq) == NULL && 1787 tlen <= sbspace(&so->so_rcv)) { 1788 int newsize = 0; /* automatic sockbuf scaling */ 1789 1790 /* 1791 * this is a pure, in-sequence data packet 1792 * with nothing on the reassembly queue and 1793 * we have enough buffer space to take it. 1794 */ 1795 tp->rcv_nxt += tlen; 1796 tcps = TCP_STAT_GETREF(); 1797 tcps[TCP_STAT_PREDDAT]++; 1798 tcps[TCP_STAT_RCVPACK]++; 1799 tcps[TCP_STAT_RCVBYTE] += tlen; 1800 TCP_STAT_PUTREF(); 1801 nd6_hint(tp); 1802 1803 /* 1804 * Automatic sizing enables the performance of large buffers 1805 * and most of the efficiency of small ones by only allocating 1806 * space when it is needed. 1807 * 1808 * On the receive side the socket buffer memory is only rarely 1809 * used to any significant extent. This allows us to be much 1810 * more aggressive in scaling the receive socket buffer. For 1811 * the case that the buffer space is actually used to a large 1812 * extent and we run out of kernel memory we can simply drop 1813 * the new segments; TCP on the sender will just retransmit it 1814 * later. Setting the buffer size too big may only consume too 1815 * much kernel memory if the application doesn't read() from 1816 * the socket or packet loss or reordering makes use of the 1817 * reassembly queue. 1818 * 1819 * The criteria to step up the receive buffer one notch are: 1820 * 1. the number of bytes received during the time it takes 1821 * one timestamp to be reflected back to us (the RTT); 1822 * 2. received bytes per RTT is within seven eighth of the 1823 * current socket buffer size; 1824 * 3. receive buffer size has not hit maximal automatic size; 1825 * 1826 * This algorithm does one step per RTT at most and only if 1827 * we receive a bulk stream w/o packet losses or reorderings. 1828 * Shrinking the buffer during idle times is not necessary as 1829 * it doesn't consume any memory when idle. 1830 * 1831 * TODO: Only step up if the application is actually serving 1832 * the buffer to better manage the socket buffer resources. 1833 */ 1834 if (tcp_do_autorcvbuf && 1835 opti.ts_ecr && 1836 (so->so_rcv.sb_flags & SB_AUTOSIZE)) { 1837 if (opti.ts_ecr > tp->rfbuf_ts && 1838 opti.ts_ecr - tp->rfbuf_ts < PR_SLOWHZ) { 1839 if (tp->rfbuf_cnt > 1840 (so->so_rcv.sb_hiwat / 8 * 7) && 1841 so->so_rcv.sb_hiwat < 1842 tcp_autorcvbuf_max) { 1843 newsize = 1844 min(so->so_rcv.sb_hiwat + 1845 tcp_autorcvbuf_inc, 1846 tcp_autorcvbuf_max); 1847 } 1848 /* Start over with next RTT. */ 1849 tp->rfbuf_ts = 0; 1850 tp->rfbuf_cnt = 0; 1851 } else 1852 tp->rfbuf_cnt += tlen; /* add up */ 1853 } 1854 1855 /* 1856 * Drop TCP, IP headers and TCP options then add data 1857 * to socket buffer. 1858 */ 1859 if (so->so_state & SS_CANTRCVMORE) 1860 m_freem(m); 1861 else { 1862 /* 1863 * Set new socket buffer size. 1864 * Give up when limit is reached. 1865 */ 1866 if (newsize) 1867 if (!sbreserve(&so->so_rcv, 1868 newsize, so)) 1869 so->so_rcv.sb_flags &= ~SB_AUTOSIZE; 1870 m_adj(m, toff + off); 1871 sbappendstream(&so->so_rcv, m); 1872 } 1873 sorwakeup(so); 1874 tcp_setup_ack(tp, th); 1875 if (tp->t_flags & TF_ACKNOW) 1876 (void) tcp_output(tp); 1877 if (tcp_saveti) 1878 m_freem(tcp_saveti); 1879 return; 1880 } 1881 } 1882 1883 /* 1884 * Compute mbuf offset to TCP data segment. 1885 */ 1886 hdroptlen = toff + off; 1887 1888 /* 1889 * Calculate amount of space in receive window, 1890 * and then do TCP input processing. 1891 * Receive window is amount of space in rcv queue, 1892 * but not less than advertised window. 1893 */ 1894 { int win; 1895 1896 win = sbspace(&so->so_rcv); 1897 if (win < 0) 1898 win = 0; 1899 tp->rcv_wnd = imax(win, (int)(tp->rcv_adv - tp->rcv_nxt)); 1900 } 1901 1902 /* Reset receive buffer auto scaling when not in bulk receive mode. */ 1903 tp->rfbuf_ts = 0; 1904 tp->rfbuf_cnt = 0; 1905 1906 switch (tp->t_state) { 1907 /* 1908 * If the state is SYN_SENT: 1909 * if seg contains an ACK, but not for our SYN, drop the input. 1910 * if seg contains a RST, then drop the connection. 1911 * if seg does not contain SYN, then drop it. 1912 * Otherwise this is an acceptable SYN segment 1913 * initialize tp->rcv_nxt and tp->irs 1914 * if seg contains ack then advance tp->snd_una 1915 * if seg contains a ECE and ECN support is enabled, the stream 1916 * is ECN capable. 1917 * if SYN has been acked change to ESTABLISHED else SYN_RCVD state 1918 * arrange for segment to be acked (eventually) 1919 * continue processing rest of data/controls, beginning with URG 1920 */ 1921 case TCPS_SYN_SENT: 1922 if ((tiflags & TH_ACK) && 1923 (SEQ_LEQ(th->th_ack, tp->iss) || 1924 SEQ_GT(th->th_ack, tp->snd_max))) 1925 goto dropwithreset; 1926 if (tiflags & TH_RST) { 1927 if (tiflags & TH_ACK) 1928 tp = tcp_drop(tp, ECONNREFUSED); 1929 goto drop; 1930 } 1931 if ((tiflags & TH_SYN) == 0) 1932 goto drop; 1933 if (tiflags & TH_ACK) { 1934 tp->snd_una = th->th_ack; 1935 if (SEQ_LT(tp->snd_nxt, tp->snd_una)) 1936 tp->snd_nxt = tp->snd_una; 1937 if (SEQ_LT(tp->snd_high, tp->snd_una)) 1938 tp->snd_high = tp->snd_una; 1939 TCP_TIMER_DISARM(tp, TCPT_REXMT); 1940 1941 if ((tiflags & TH_ECE) && tcp_do_ecn) { 1942 tp->t_flags |= TF_ECN_PERMIT; 1943 TCP_STATINC(TCP_STAT_ECN_SHS); 1944 } 1945 1946 } 1947 tp->irs = th->th_seq; 1948 tcp_rcvseqinit(tp); 1949 tp->t_flags |= TF_ACKNOW; 1950 tcp_mss_from_peer(tp, opti.maxseg); 1951 1952 /* 1953 * Initialize the initial congestion window. If we 1954 * had to retransmit the SYN, we must initialize cwnd 1955 * to 1 segment (i.e. the Loss Window). 1956 */ 1957 if (tp->t_flags & TF_SYN_REXMT) 1958 tp->snd_cwnd = tp->t_peermss; 1959 else { 1960 int ss = tcp_init_win; 1961 #ifdef INET 1962 if (inp != NULL && in_localaddr(inp->inp_faddr)) 1963 ss = tcp_init_win_local; 1964 #endif 1965 #ifdef INET6 1966 if (in6p != NULL && in6_localaddr(&in6p->in6p_faddr)) 1967 ss = tcp_init_win_local; 1968 #endif 1969 tp->snd_cwnd = TCP_INITIAL_WINDOW(ss, tp->t_peermss); 1970 } 1971 1972 tcp_rmx_rtt(tp); 1973 if (tiflags & TH_ACK) { 1974 TCP_STATINC(TCP_STAT_CONNECTS); 1975 soisconnected(so); 1976 tcp_established(tp); 1977 /* Do window scaling on this connection? */ 1978 if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == 1979 (TF_RCVD_SCALE|TF_REQ_SCALE)) { 1980 tp->snd_scale = tp->requested_s_scale; 1981 tp->rcv_scale = tp->request_r_scale; 1982 } 1983 TCP_REASS_LOCK(tp); 1984 (void) tcp_reass(tp, NULL, (struct mbuf *)0, &tlen); 1985 TCP_REASS_UNLOCK(tp); 1986 /* 1987 * if we didn't have to retransmit the SYN, 1988 * use its rtt as our initial srtt & rtt var. 1989 */ 1990 if (tp->t_rtttime) 1991 tcp_xmit_timer(tp, tcp_now - tp->t_rtttime); 1992 } else 1993 tp->t_state = TCPS_SYN_RECEIVED; 1994 1995 /* 1996 * Advance th->th_seq to correspond to first data byte. 1997 * If data, trim to stay within window, 1998 * dropping FIN if necessary. 1999 */ 2000 th->th_seq++; 2001 if (tlen > tp->rcv_wnd) { 2002 todrop = tlen - tp->rcv_wnd; 2003 m_adj(m, -todrop); 2004 tlen = tp->rcv_wnd; 2005 tiflags &= ~TH_FIN; 2006 tcps = TCP_STAT_GETREF(); 2007 tcps[TCP_STAT_RCVPACKAFTERWIN]++; 2008 tcps[TCP_STAT_RCVBYTEAFTERWIN] += todrop; 2009 TCP_STAT_PUTREF(); 2010 } 2011 tp->snd_wl1 = th->th_seq - 1; 2012 tp->rcv_up = th->th_seq; 2013 goto step6; 2014 2015 /* 2016 * If the state is SYN_RECEIVED: 2017 * If seg contains an ACK, but not for our SYN, drop the input 2018 * and generate an RST. See page 36, rfc793 2019 */ 2020 case TCPS_SYN_RECEIVED: 2021 if ((tiflags & TH_ACK) && 2022 (SEQ_LEQ(th->th_ack, tp->iss) || 2023 SEQ_GT(th->th_ack, tp->snd_max))) 2024 goto dropwithreset; 2025 break; 2026 } 2027 2028 /* 2029 * States other than LISTEN or SYN_SENT. 2030 * First check timestamp, if present. 2031 * Then check that at least some bytes of segment are within 2032 * receive window. If segment begins before rcv_nxt, 2033 * drop leading data (and SYN); if nothing left, just ack. 2034 * 2035 * RFC 1323 PAWS: If we have a timestamp reply on this segment 2036 * and it's less than ts_recent, drop it. 2037 */ 2038 if (opti.ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent && 2039 TSTMP_LT(opti.ts_val, tp->ts_recent)) { 2040 2041 /* Check to see if ts_recent is over 24 days old. */ 2042 if (tcp_now - tp->ts_recent_age > TCP_PAWS_IDLE) { 2043 /* 2044 * Invalidate ts_recent. If this segment updates 2045 * ts_recent, the age will be reset later and ts_recent 2046 * will get a valid value. If it does not, setting 2047 * ts_recent to zero will at least satisfy the 2048 * requirement that zero be placed in the timestamp 2049 * echo reply when ts_recent isn't valid. The 2050 * age isn't reset until we get a valid ts_recent 2051 * because we don't want out-of-order segments to be 2052 * dropped when ts_recent is old. 2053 */ 2054 tp->ts_recent = 0; 2055 } else { 2056 tcps = TCP_STAT_GETREF(); 2057 tcps[TCP_STAT_RCVDUPPACK]++; 2058 tcps[TCP_STAT_RCVDUPBYTE] += tlen; 2059 tcps[TCP_STAT_PAWSDROP]++; 2060 TCP_STAT_PUTREF(); 2061 tcp_new_dsack(tp, th->th_seq, tlen); 2062 goto dropafterack; 2063 } 2064 } 2065 2066 todrop = tp->rcv_nxt - th->th_seq; 2067 dupseg = false; 2068 if (todrop > 0) { 2069 if (tiflags & TH_SYN) { 2070 tiflags &= ~TH_SYN; 2071 th->th_seq++; 2072 if (th->th_urp > 1) 2073 th->th_urp--; 2074 else { 2075 tiflags &= ~TH_URG; 2076 th->th_urp = 0; 2077 } 2078 todrop--; 2079 } 2080 if (todrop > tlen || 2081 (todrop == tlen && (tiflags & TH_FIN) == 0)) { 2082 /* 2083 * Any valid FIN or RST must be to the left of the 2084 * window. At this point the FIN or RST must be a 2085 * duplicate or out of sequence; drop it. 2086 */ 2087 if (tiflags & TH_RST) 2088 goto drop; 2089 tiflags &= ~(TH_FIN|TH_RST); 2090 /* 2091 * Send an ACK to resynchronize and drop any data. 2092 * But keep on processing for RST or ACK. 2093 */ 2094 tp->t_flags |= TF_ACKNOW; 2095 todrop = tlen; 2096 dupseg = true; 2097 tcps = TCP_STAT_GETREF(); 2098 tcps[TCP_STAT_RCVDUPPACK]++; 2099 tcps[TCP_STAT_RCVDUPBYTE] += todrop; 2100 TCP_STAT_PUTREF(); 2101 } else if ((tiflags & TH_RST) && 2102 th->th_seq != tp->last_ack_sent) { 2103 /* 2104 * Test for reset before adjusting the sequence 2105 * number for overlapping data. 2106 */ 2107 goto dropafterack_ratelim; 2108 } else { 2109 tcps = TCP_STAT_GETREF(); 2110 tcps[TCP_STAT_RCVPARTDUPPACK]++; 2111 tcps[TCP_STAT_RCVPARTDUPBYTE] += todrop; 2112 TCP_STAT_PUTREF(); 2113 } 2114 tcp_new_dsack(tp, th->th_seq, todrop); 2115 hdroptlen += todrop; /*drop from head afterwards*/ 2116 th->th_seq += todrop; 2117 tlen -= todrop; 2118 if (th->th_urp > todrop) 2119 th->th_urp -= todrop; 2120 else { 2121 tiflags &= ~TH_URG; 2122 th->th_urp = 0; 2123 } 2124 } 2125 2126 /* 2127 * If new data are received on a connection after the 2128 * user processes are gone, then RST the other end. 2129 */ 2130 if ((so->so_state & SS_NOFDREF) && 2131 tp->t_state > TCPS_CLOSE_WAIT && tlen) { 2132 tp = tcp_close(tp); 2133 TCP_STATINC(TCP_STAT_RCVAFTERCLOSE); 2134 goto dropwithreset; 2135 } 2136 2137 /* 2138 * If segment ends after window, drop trailing data 2139 * (and PUSH and FIN); if nothing left, just ACK. 2140 */ 2141 todrop = (th->th_seq + tlen) - (tp->rcv_nxt+tp->rcv_wnd); 2142 if (todrop > 0) { 2143 TCP_STATINC(TCP_STAT_RCVPACKAFTERWIN); 2144 if (todrop >= tlen) { 2145 /* 2146 * The segment actually starts after the window. 2147 * th->th_seq + tlen - tp->rcv_nxt - tp->rcv_wnd >= tlen 2148 * th->th_seq - tp->rcv_nxt - tp->rcv_wnd >= 0 2149 * th->th_seq >= tp->rcv_nxt + tp->rcv_wnd 2150 */ 2151 TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, tlen); 2152 /* 2153 * If a new connection request is received 2154 * while in TIME_WAIT, drop the old connection 2155 * and start over if the sequence numbers 2156 * are above the previous ones. 2157 * 2158 * NOTE: We will checksum the packet again, and 2159 * so we need to put the header fields back into 2160 * network order! 2161 * XXX This kind of sucks, but we don't expect 2162 * XXX this to happen very often, so maybe it 2163 * XXX doesn't matter so much. 2164 */ 2165 if (tiflags & TH_SYN && 2166 tp->t_state == TCPS_TIME_WAIT && 2167 SEQ_GT(th->th_seq, tp->rcv_nxt)) { 2168 tp = tcp_close(tp); 2169 tcp_fields_to_net(th); 2170 goto findpcb; 2171 } 2172 /* 2173 * If window is closed can only take segments at 2174 * window edge, and have to drop data and PUSH from 2175 * incoming segments. Continue processing, but 2176 * remember to ack. Otherwise, drop segment 2177 * and (if not RST) ack. 2178 */ 2179 if (tp->rcv_wnd == 0 && th->th_seq == tp->rcv_nxt) { 2180 tp->t_flags |= TF_ACKNOW; 2181 TCP_STATINC(TCP_STAT_RCVWINPROBE); 2182 } else 2183 goto dropafterack; 2184 } else 2185 TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, todrop); 2186 m_adj(m, -todrop); 2187 tlen -= todrop; 2188 tiflags &= ~(TH_PUSH|TH_FIN); 2189 } 2190 2191 /* 2192 * If last ACK falls within this segment's sequence numbers, 2193 * record the timestamp. 2194 * NOTE: 2195 * 1) That the test incorporates suggestions from the latest 2196 * proposal of the tcplw (at) cray.com list (Braden 1993/04/26). 2197 * 2) That updating only on newer timestamps interferes with 2198 * our earlier PAWS tests, so this check should be solely 2199 * predicated on the sequence space of this segment. 2200 * 3) That we modify the segment boundary check to be 2201 * Last.ACK.Sent <= SEG.SEQ + SEG.Len 2202 * instead of RFC1323's 2203 * Last.ACK.Sent < SEG.SEQ + SEG.Len, 2204 * This modified check allows us to overcome RFC1323's 2205 * limitations as described in Stevens TCP/IP Illustrated 2206 * Vol. 2 p.869. In such cases, we can still calculate the 2207 * RTT correctly when RCV.NXT == Last.ACK.Sent. 2208 */ 2209 if (opti.ts_present && 2210 SEQ_LEQ(th->th_seq, tp->last_ack_sent) && 2211 SEQ_LEQ(tp->last_ack_sent, th->th_seq + tlen + 2212 ((tiflags & (TH_SYN|TH_FIN)) != 0))) { 2213 tp->ts_recent_age = tcp_now; 2214 tp->ts_recent = opti.ts_val; 2215 } 2216 2217 /* 2218 * If the RST bit is set examine the state: 2219 * SYN_RECEIVED STATE: 2220 * If passive open, return to LISTEN state. 2221 * If active open, inform user that connection was refused. 2222 * ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES: 2223 * Inform user that connection was reset, and close tcb. 2224 * CLOSING, LAST_ACK, TIME_WAIT STATES 2225 * Close the tcb. 2226 */ 2227 if (tiflags & TH_RST) { 2228 if (th->th_seq != tp->last_ack_sent) 2229 goto dropafterack_ratelim; 2230 2231 switch (tp->t_state) { 2232 case TCPS_SYN_RECEIVED: 2233 so->so_error = ECONNREFUSED; 2234 goto close; 2235 2236 case TCPS_ESTABLISHED: 2237 case TCPS_FIN_WAIT_1: 2238 case TCPS_FIN_WAIT_2: 2239 case TCPS_CLOSE_WAIT: 2240 so->so_error = ECONNRESET; 2241 close: 2242 tp->t_state = TCPS_CLOSED; 2243 TCP_STATINC(TCP_STAT_DROPS); 2244 tp = tcp_close(tp); 2245 goto drop; 2246 2247 case TCPS_CLOSING: 2248 case TCPS_LAST_ACK: 2249 case TCPS_TIME_WAIT: 2250 tp = tcp_close(tp); 2251 goto drop; 2252 } 2253 } 2254 2255 /* 2256 * Since we've covered the SYN-SENT and SYN-RECEIVED states above 2257 * we must be in a synchronized state. RFC791 states (under RST 2258 * generation) that any unacceptable segment (an out-of-order SYN 2259 * qualifies) received in a synchronized state must elicit only an 2260 * empty acknowledgment segment ... and the connection remains in 2261 * the same state. 2262 */ 2263 if (tiflags & TH_SYN) { 2264 if (tp->rcv_nxt == th->th_seq) { 2265 tcp_respond(tp, m, m, th, (tcp_seq)0, th->th_ack - 1, 2266 TH_ACK); 2267 if (tcp_saveti) 2268 m_freem(tcp_saveti); 2269 return; 2270 } 2271 2272 goto dropafterack_ratelim; 2273 } 2274 2275 /* 2276 * If the ACK bit is off we drop the segment and return. 2277 */ 2278 if ((tiflags & TH_ACK) == 0) { 2279 if (tp->t_flags & TF_ACKNOW) 2280 goto dropafterack; 2281 else 2282 goto drop; 2283 } 2284 2285 /* 2286 * Ack processing. 2287 */ 2288 switch (tp->t_state) { 2289 2290 /* 2291 * In SYN_RECEIVED state if the ack ACKs our SYN then enter 2292 * ESTABLISHED state and continue processing, otherwise 2293 * send an RST. 2294 */ 2295 case TCPS_SYN_RECEIVED: 2296 if (SEQ_GT(tp->snd_una, th->th_ack) || 2297 SEQ_GT(th->th_ack, tp->snd_max)) 2298 goto dropwithreset; 2299 TCP_STATINC(TCP_STAT_CONNECTS); 2300 soisconnected(so); 2301 tcp_established(tp); 2302 /* Do window scaling? */ 2303 if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == 2304 (TF_RCVD_SCALE|TF_REQ_SCALE)) { 2305 tp->snd_scale = tp->requested_s_scale; 2306 tp->rcv_scale = tp->request_r_scale; 2307 } 2308 TCP_REASS_LOCK(tp); 2309 (void) tcp_reass(tp, NULL, (struct mbuf *)0, &tlen); 2310 TCP_REASS_UNLOCK(tp); 2311 tp->snd_wl1 = th->th_seq - 1; 2312 /* fall into ... */ 2313 2314 /* 2315 * In ESTABLISHED state: drop duplicate ACKs; ACK out of range 2316 * ACKs. If the ack is in the range 2317 * tp->snd_una < th->th_ack <= tp->snd_max 2318 * then advance tp->snd_una to th->th_ack and drop 2319 * data from the retransmission queue. If this ACK reflects 2320 * more up to date window information we update our window information. 2321 */ 2322 case TCPS_ESTABLISHED: 2323 case TCPS_FIN_WAIT_1: 2324 case TCPS_FIN_WAIT_2: 2325 case TCPS_CLOSE_WAIT: 2326 case TCPS_CLOSING: 2327 case TCPS_LAST_ACK: 2328 case TCPS_TIME_WAIT: 2329 2330 if (SEQ_LEQ(th->th_ack, tp->snd_una)) { 2331 if (tlen == 0 && !dupseg && tiwin == tp->snd_wnd) { 2332 TCP_STATINC(TCP_STAT_RCVDUPPACK); 2333 /* 2334 * If we have outstanding data (other than 2335 * a window probe), this is a completely 2336 * duplicate ack (ie, window info didn't 2337 * change), the ack is the biggest we've 2338 * seen and we've seen exactly our rexmt 2339 * threshhold of them, assume a packet 2340 * has been dropped and retransmit it. 2341 * Kludge snd_nxt & the congestion 2342 * window so we send only this one 2343 * packet. 2344 */ 2345 if (TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 || 2346 th->th_ack != tp->snd_una) 2347 tp->t_dupacks = 0; 2348 else if (tp->t_partialacks < 0 && 2349 (++tp->t_dupacks == tcprexmtthresh || 2350 TCP_FACK_FASTRECOV(tp))) { 2351 /* 2352 * Do the fast retransmit, and adjust 2353 * congestion control paramenters. 2354 */ 2355 if (tp->t_congctl->fast_retransmit(tp, th)) { 2356 /* False fast retransmit */ 2357 break; 2358 } else 2359 goto drop; 2360 } else if (tp->t_dupacks > tcprexmtthresh) { 2361 tp->snd_cwnd += tp->t_segsz; 2362 (void) tcp_output(tp); 2363 goto drop; 2364 } 2365 } else { 2366 /* 2367 * If the ack appears to be very old, only 2368 * allow data that is in-sequence. This 2369 * makes it somewhat more difficult to insert 2370 * forged data by guessing sequence numbers. 2371 * Sent an ack to try to update the send 2372 * sequence number on the other side. 2373 */ 2374 if (tlen && th->th_seq != tp->rcv_nxt && 2375 SEQ_LT(th->th_ack, 2376 tp->snd_una - tp->max_sndwnd)) 2377 goto dropafterack; 2378 } 2379 break; 2380 } 2381 /* 2382 * If the congestion window was inflated to account 2383 * for the other side's cached packets, retract it. 2384 */ 2385 /* XXX: make SACK have his own congestion control 2386 * struct -- rpaulo */ 2387 if (TCP_SACK_ENABLED(tp)) 2388 tcp_sack_newack(tp, th); 2389 else 2390 tp->t_congctl->fast_retransmit_newack(tp, th); 2391 if (SEQ_GT(th->th_ack, tp->snd_max)) { 2392 TCP_STATINC(TCP_STAT_RCVACKTOOMUCH); 2393 goto dropafterack; 2394 } 2395 acked = th->th_ack - tp->snd_una; 2396 tcps = TCP_STAT_GETREF(); 2397 tcps[TCP_STAT_RCVACKPACK]++; 2398 tcps[TCP_STAT_RCVACKBYTE] += acked; 2399 TCP_STAT_PUTREF(); 2400 2401 /* 2402 * If we have a timestamp reply, update smoothed 2403 * round trip time. If no timestamp is present but 2404 * transmit timer is running and timed sequence 2405 * number was acked, update smoothed round trip time. 2406 * Since we now have an rtt measurement, cancel the 2407 * timer backoff (cf., Phil Karn's retransmit alg.). 2408 * Recompute the initial retransmit timer. 2409 */ 2410 if (ts_rtt) 2411 tcp_xmit_timer(tp, ts_rtt); 2412 else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq)) 2413 tcp_xmit_timer(tp, tcp_now - tp->t_rtttime); 2414 2415 /* 2416 * If all outstanding data is acked, stop retransmit 2417 * timer and remember to restart (more output or persist). 2418 * If there is more data to be acked, restart retransmit 2419 * timer, using current (possibly backed-off) value. 2420 */ 2421 if (th->th_ack == tp->snd_max) { 2422 TCP_TIMER_DISARM(tp, TCPT_REXMT); 2423 needoutput = 1; 2424 } else if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) 2425 TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur); 2426 2427 /* 2428 * New data has been acked, adjust the congestion window. 2429 */ 2430 tp->t_congctl->newack(tp, th); 2431 2432 nd6_hint(tp); 2433 if (acked > so->so_snd.sb_cc) { 2434 tp->snd_wnd -= so->so_snd.sb_cc; 2435 sbdrop(&so->so_snd, (int)so->so_snd.sb_cc); 2436 ourfinisacked = 1; 2437 } else { 2438 if (acked > (tp->t_lastoff - tp->t_inoff)) 2439 tp->t_lastm = NULL; 2440 sbdrop(&so->so_snd, acked); 2441 tp->t_lastoff -= acked; 2442 tp->snd_wnd -= acked; 2443 ourfinisacked = 0; 2444 } 2445 sowwakeup(so); 2446 2447 icmp_check(tp, th, acked); 2448 2449 tp->snd_una = th->th_ack; 2450 if (SEQ_GT(tp->snd_una, tp->snd_fack)) 2451 tp->snd_fack = tp->snd_una; 2452 if (SEQ_LT(tp->snd_nxt, tp->snd_una)) 2453 tp->snd_nxt = tp->snd_una; 2454 if (SEQ_LT(tp->snd_high, tp->snd_una)) 2455 tp->snd_high = tp->snd_una; 2456 2457 switch (tp->t_state) { 2458 2459 /* 2460 * In FIN_WAIT_1 STATE in addition to the processing 2461 * for the ESTABLISHED state if our FIN is now acknowledged 2462 * then enter FIN_WAIT_2. 2463 */ 2464 case TCPS_FIN_WAIT_1: 2465 if (ourfinisacked) { 2466 /* 2467 * If we can't receive any more 2468 * data, then closing user can proceed. 2469 * Starting the timer is contrary to the 2470 * specification, but if we don't get a FIN 2471 * we'll hang forever. 2472 */ 2473 if (so->so_state & SS_CANTRCVMORE) { 2474 soisdisconnected(so); 2475 if (tp->t_maxidle > 0) 2476 TCP_TIMER_ARM(tp, TCPT_2MSL, 2477 tp->t_maxidle); 2478 } 2479 tp->t_state = TCPS_FIN_WAIT_2; 2480 } 2481 break; 2482 2483 /* 2484 * In CLOSING STATE in addition to the processing for 2485 * the ESTABLISHED state if the ACK acknowledges our FIN 2486 * then enter the TIME-WAIT state, otherwise ignore 2487 * the segment. 2488 */ 2489 case TCPS_CLOSING: 2490 if (ourfinisacked) { 2491 tp->t_state = TCPS_TIME_WAIT; 2492 tcp_canceltimers(tp); 2493 TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * TCPTV_MSL); 2494 soisdisconnected(so); 2495 } 2496 break; 2497 2498 /* 2499 * In LAST_ACK, we may still be waiting for data to drain 2500 * and/or to be acked, as well as for the ack of our FIN. 2501 * If our FIN is now acknowledged, delete the TCB, 2502 * enter the closed state and return. 2503 */ 2504 case TCPS_LAST_ACK: 2505 if (ourfinisacked) { 2506 tp = tcp_close(tp); 2507 goto drop; 2508 } 2509 break; 2510 2511 /* 2512 * In TIME_WAIT state the only thing that should arrive 2513 * is a retransmission of the remote FIN. Acknowledge 2514 * it and restart the finack timer. 2515 */ 2516 case TCPS_TIME_WAIT: 2517 TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * TCPTV_MSL); 2518 goto dropafterack; 2519 } 2520 } 2521 2522 step6: 2523 /* 2524 * Update window information. 2525 * Don't look at window if no ACK: TAC's send garbage on first SYN. 2526 */ 2527 if ((tiflags & TH_ACK) && (SEQ_LT(tp->snd_wl1, th->th_seq) || 2528 (tp->snd_wl1 == th->th_seq && (SEQ_LT(tp->snd_wl2, th->th_ack) || 2529 (tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd))))) { 2530 /* keep track of pure window updates */ 2531 if (tlen == 0 && 2532 tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd) 2533 TCP_STATINC(TCP_STAT_RCVWINUPD); 2534 tp->snd_wnd = tiwin; 2535 tp->snd_wl1 = th->th_seq; 2536 tp->snd_wl2 = th->th_ack; 2537 if (tp->snd_wnd > tp->max_sndwnd) 2538 tp->max_sndwnd = tp->snd_wnd; 2539 needoutput = 1; 2540 } 2541 2542 /* 2543 * Process segments with URG. 2544 */ 2545 if ((tiflags & TH_URG) && th->th_urp && 2546 TCPS_HAVERCVDFIN(tp->t_state) == 0) { 2547 /* 2548 * This is a kludge, but if we receive and accept 2549 * random urgent pointers, we'll crash in 2550 * soreceive. It's hard to imagine someone 2551 * actually wanting to send this much urgent data. 2552 */ 2553 if (th->th_urp + so->so_rcv.sb_cc > sb_max) { 2554 th->th_urp = 0; /* XXX */ 2555 tiflags &= ~TH_URG; /* XXX */ 2556 goto dodata; /* XXX */ 2557 } 2558 /* 2559 * If this segment advances the known urgent pointer, 2560 * then mark the data stream. This should not happen 2561 * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since 2562 * a FIN has been received from the remote side. 2563 * In these states we ignore the URG. 2564 * 2565 * According to RFC961 (Assigned Protocols), 2566 * the urgent pointer points to the last octet 2567 * of urgent data. We continue, however, 2568 * to consider it to indicate the first octet 2569 * of data past the urgent section as the original 2570 * spec states (in one of two places). 2571 */ 2572 if (SEQ_GT(th->th_seq+th->th_urp, tp->rcv_up)) { 2573 tp->rcv_up = th->th_seq + th->th_urp; 2574 so->so_oobmark = so->so_rcv.sb_cc + 2575 (tp->rcv_up - tp->rcv_nxt) - 1; 2576 if (so->so_oobmark == 0) 2577 so->so_state |= SS_RCVATMARK; 2578 sohasoutofband(so); 2579 tp->t_oobflags &= ~(TCPOOB_HAVEDATA | TCPOOB_HADDATA); 2580 } 2581 /* 2582 * Remove out of band data so doesn't get presented to user. 2583 * This can happen independent of advancing the URG pointer, 2584 * but if two URG's are pending at once, some out-of-band 2585 * data may creep in... ick. 2586 */ 2587 if (th->th_urp <= (u_int16_t) tlen 2588 #ifdef SO_OOBINLINE 2589 && (so->so_options & SO_OOBINLINE) == 0 2590 #endif 2591 ) 2592 tcp_pulloutofband(so, th, m, hdroptlen); 2593 } else 2594 /* 2595 * If no out of band data is expected, 2596 * pull receive urgent pointer along 2597 * with the receive window. 2598 */ 2599 if (SEQ_GT(tp->rcv_nxt, tp->rcv_up)) 2600 tp->rcv_up = tp->rcv_nxt; 2601 dodata: /* XXX */ 2602 2603 /* 2604 * Process the segment text, merging it into the TCP sequencing queue, 2605 * and arranging for acknowledgement of receipt if necessary. 2606 * This process logically involves adjusting tp->rcv_wnd as data 2607 * is presented to the user (this happens in tcp_usrreq.c, 2608 * case PRU_RCVD). If a FIN has already been received on this 2609 * connection then we just ignore the text. 2610 */ 2611 if ((tlen || (tiflags & TH_FIN)) && 2612 TCPS_HAVERCVDFIN(tp->t_state) == 0) { 2613 /* 2614 * Insert segment ti into reassembly queue of tcp with 2615 * control block tp. Return TH_FIN if reassembly now includes 2616 * a segment with FIN. The macro form does the common case 2617 * inline (segment is the next to be received on an 2618 * established connection, and the queue is empty), 2619 * avoiding linkage into and removal from the queue and 2620 * repetition of various conversions. 2621 * Set DELACK for segments received in order, but ack 2622 * immediately when segments are out of order 2623 * (so fast retransmit can work). 2624 */ 2625 /* NOTE: this was TCP_REASS() macro, but used only once */ 2626 TCP_REASS_LOCK(tp); 2627 if (th->th_seq == tp->rcv_nxt && 2628 TAILQ_FIRST(&tp->segq) == NULL && 2629 tp->t_state == TCPS_ESTABLISHED) { 2630 tcp_setup_ack(tp, th); 2631 tp->rcv_nxt += tlen; 2632 tiflags = th->th_flags & TH_FIN; 2633 tcps = TCP_STAT_GETREF(); 2634 tcps[TCP_STAT_RCVPACK]++; 2635 tcps[TCP_STAT_RCVBYTE] += tlen; 2636 TCP_STAT_PUTREF(); 2637 nd6_hint(tp); 2638 if (so->so_state & SS_CANTRCVMORE) 2639 m_freem(m); 2640 else { 2641 m_adj(m, hdroptlen); 2642 sbappendstream(&(so)->so_rcv, m); 2643 } 2644 sorwakeup(so); 2645 } else { 2646 m_adj(m, hdroptlen); 2647 tiflags = tcp_reass(tp, th, m, &tlen); 2648 tp->t_flags |= TF_ACKNOW; 2649 } 2650 TCP_REASS_UNLOCK(tp); 2651 2652 /* 2653 * Note the amount of data that peer has sent into 2654 * our window, in order to estimate the sender's 2655 * buffer size. 2656 */ 2657 len = so->so_rcv.sb_hiwat - (tp->rcv_adv - tp->rcv_nxt); 2658 } else { 2659 m_freem(m); 2660 m = NULL; 2661 tiflags &= ~TH_FIN; 2662 } 2663 2664 /* 2665 * If FIN is received ACK the FIN and let the user know 2666 * that the connection is closing. Ignore a FIN received before 2667 * the connection is fully established. 2668 */ 2669 if ((tiflags & TH_FIN) && TCPS_HAVEESTABLISHED(tp->t_state)) { 2670 if (TCPS_HAVERCVDFIN(tp->t_state) == 0) { 2671 socantrcvmore(so); 2672 tp->t_flags |= TF_ACKNOW; 2673 tp->rcv_nxt++; 2674 } 2675 switch (tp->t_state) { 2676 2677 /* 2678 * In ESTABLISHED STATE enter the CLOSE_WAIT state. 2679 */ 2680 case TCPS_ESTABLISHED: 2681 tp->t_state = TCPS_CLOSE_WAIT; 2682 break; 2683 2684 /* 2685 * If still in FIN_WAIT_1 STATE FIN has not been acked so 2686 * enter the CLOSING state. 2687 */ 2688 case TCPS_FIN_WAIT_1: 2689 tp->t_state = TCPS_CLOSING; 2690 break; 2691 2692 /* 2693 * In FIN_WAIT_2 state enter the TIME_WAIT state, 2694 * starting the time-wait timer, turning off the other 2695 * standard timers. 2696 */ 2697 case TCPS_FIN_WAIT_2: 2698 tp->t_state = TCPS_TIME_WAIT; 2699 tcp_canceltimers(tp); 2700 TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * TCPTV_MSL); 2701 soisdisconnected(so); 2702 break; 2703 2704 /* 2705 * In TIME_WAIT state restart the 2 MSL time_wait timer. 2706 */ 2707 case TCPS_TIME_WAIT: 2708 TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * TCPTV_MSL); 2709 break; 2710 } 2711 } 2712 #ifdef TCP_DEBUG 2713 if (so->so_options & SO_DEBUG) 2714 tcp_trace(TA_INPUT, ostate, tp, tcp_saveti, 0); 2715 #endif 2716 2717 /* 2718 * Return any desired output. 2719 */ 2720 if (needoutput || (tp->t_flags & TF_ACKNOW)) { 2721 (void) tcp_output(tp); 2722 } 2723 if (tcp_saveti) 2724 m_freem(tcp_saveti); 2725 return; 2726 2727 badsyn: 2728 /* 2729 * Received a bad SYN. Increment counters and dropwithreset. 2730 */ 2731 TCP_STATINC(TCP_STAT_BADSYN); 2732 tp = NULL; 2733 goto dropwithreset; 2734 2735 dropafterack: 2736 /* 2737 * Generate an ACK dropping incoming segment if it occupies 2738 * sequence space, where the ACK reflects our state. 2739 */ 2740 if (tiflags & TH_RST) 2741 goto drop; 2742 goto dropafterack2; 2743 2744 dropafterack_ratelim: 2745 /* 2746 * We may want to rate-limit ACKs against SYN/RST attack. 2747 */ 2748 if (ppsratecheck(&tcp_ackdrop_ppslim_last, &tcp_ackdrop_ppslim_count, 2749 tcp_ackdrop_ppslim) == 0) { 2750 /* XXX stat */ 2751 goto drop; 2752 } 2753 /* ...fall into dropafterack2... */ 2754 2755 dropafterack2: 2756 m_freem(m); 2757 tp->t_flags |= TF_ACKNOW; 2758 (void) tcp_output(tp); 2759 if (tcp_saveti) 2760 m_freem(tcp_saveti); 2761 return; 2762 2763 dropwithreset_ratelim: 2764 /* 2765 * We may want to rate-limit RSTs in certain situations, 2766 * particularly if we are sending an RST in response to 2767 * an attempt to connect to or otherwise communicate with 2768 * a port for which we have no socket. 2769 */ 2770 if (ppsratecheck(&tcp_rst_ppslim_last, &tcp_rst_ppslim_count, 2771 tcp_rst_ppslim) == 0) { 2772 /* XXX stat */ 2773 goto drop; 2774 } 2775 /* ...fall into dropwithreset... */ 2776 2777 dropwithreset: 2778 /* 2779 * Generate a RST, dropping incoming segment. 2780 * Make ACK acceptable to originator of segment. 2781 */ 2782 if (tiflags & TH_RST) 2783 goto drop; 2784 2785 switch (af) { 2786 #ifdef INET6 2787 case AF_INET6: 2788 /* For following calls to tcp_respond */ 2789 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 2790 goto drop; 2791 break; 2792 #endif /* INET6 */ 2793 case AF_INET: 2794 if (IN_MULTICAST(ip->ip_dst.s_addr) || 2795 in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) 2796 goto drop; 2797 } 2798 2799 if (tiflags & TH_ACK) 2800 (void)tcp_respond(tp, m, m, th, (tcp_seq)0, th->th_ack, TH_RST); 2801 else { 2802 if (tiflags & TH_SYN) 2803 tlen++; 2804 (void)tcp_respond(tp, m, m, th, th->th_seq + tlen, (tcp_seq)0, 2805 TH_RST|TH_ACK); 2806 } 2807 if (tcp_saveti) 2808 m_freem(tcp_saveti); 2809 return; 2810 2811 badcsum: 2812 drop: 2813 /* 2814 * Drop space held by incoming segment and return. 2815 */ 2816 if (tp) { 2817 if (tp->t_inpcb) 2818 so = tp->t_inpcb->inp_socket; 2819 #ifdef INET6 2820 else if (tp->t_in6pcb) 2821 so = tp->t_in6pcb->in6p_socket; 2822 #endif 2823 else 2824 so = NULL; 2825 #ifdef TCP_DEBUG 2826 if (so && (so->so_options & SO_DEBUG) != 0) 2827 tcp_trace(TA_DROP, ostate, tp, tcp_saveti, 0); 2828 #endif 2829 } 2830 if (tcp_saveti) 2831 m_freem(tcp_saveti); 2832 m_freem(m); 2833 return; 2834 } 2835 2836 #ifdef TCP_SIGNATURE 2837 int 2838 tcp_signature_apply(void *fstate, void *data, u_int len) 2839 { 2840 2841 MD5Update(fstate, (u_char *)data, len); 2842 return (0); 2843 } 2844 2845 struct secasvar * 2846 tcp_signature_getsav(struct mbuf *m, struct tcphdr *th) 2847 { 2848 struct secasvar *sav; 2849 #ifdef FAST_IPSEC 2850 union sockaddr_union dst; 2851 #endif 2852 struct ip *ip; 2853 struct ip6_hdr *ip6; 2854 2855 ip = mtod(m, struct ip *); 2856 switch (ip->ip_v) { 2857 case 4: 2858 ip = mtod(m, struct ip *); 2859 ip6 = NULL; 2860 break; 2861 case 6: 2862 ip = NULL; 2863 ip6 = mtod(m, struct ip6_hdr *); 2864 break; 2865 default: 2866 return (NULL); 2867 } 2868 2869 #ifdef FAST_IPSEC 2870 /* Extract the destination from the IP header in the mbuf. */ 2871 bzero(&dst, sizeof(union sockaddr_union)); 2872 if (ip !=NULL) { 2873 dst.sa.sa_len = sizeof(struct sockaddr_in); 2874 dst.sa.sa_family = AF_INET; 2875 dst.sin.sin_addr = ip->ip_dst; 2876 } else { 2877 dst.sa.sa_len = sizeof(struct sockaddr_in6); 2878 dst.sa.sa_family = AF_INET6; 2879 dst.sin6.sin6_addr = ip6->ip6_dst; 2880 } 2881 2882 /* 2883 * Look up an SADB entry which matches the address of the peer. 2884 */ 2885 sav = KEY_ALLOCSA(&dst, IPPROTO_TCP, htonl(TCP_SIG_SPI)); 2886 #else 2887 if (ip) 2888 sav = key_allocsa(AF_INET, (void *)&ip->ip_src, 2889 (void *)&ip->ip_dst, IPPROTO_TCP, 2890 htonl(TCP_SIG_SPI), 0, 0); 2891 else 2892 sav = key_allocsa(AF_INET6, (void *)&ip6->ip6_src, 2893 (void *)&ip6->ip6_dst, IPPROTO_TCP, 2894 htonl(TCP_SIG_SPI), 0, 0); 2895 #endif 2896 2897 return (sav); /* freesav must be performed by caller */ 2898 } 2899 2900 int 2901 tcp_signature(struct mbuf *m, struct tcphdr *th, int thoff, 2902 struct secasvar *sav, char *sig) 2903 { 2904 MD5_CTX ctx; 2905 struct ip *ip; 2906 struct ipovly *ipovly; 2907 struct ip6_hdr *ip6; 2908 struct ippseudo ippseudo; 2909 struct ip6_hdr_pseudo ip6pseudo; 2910 struct tcphdr th0; 2911 int l, tcphdrlen; 2912 2913 if (sav == NULL) 2914 return (-1); 2915 2916 tcphdrlen = th->th_off * 4; 2917 2918 switch (mtod(m, struct ip *)->ip_v) { 2919 case 4: 2920 ip = mtod(m, struct ip *); 2921 ip6 = NULL; 2922 break; 2923 case 6: 2924 ip = NULL; 2925 ip6 = mtod(m, struct ip6_hdr *); 2926 break; 2927 default: 2928 return (-1); 2929 } 2930 2931 MD5Init(&ctx); 2932 2933 if (ip) { 2934 memset(&ippseudo, 0, sizeof(ippseudo)); 2935 ipovly = (struct ipovly *)ip; 2936 ippseudo.ippseudo_src = ipovly->ih_src; 2937 ippseudo.ippseudo_dst = ipovly->ih_dst; 2938 ippseudo.ippseudo_pad = 0; 2939 ippseudo.ippseudo_p = IPPROTO_TCP; 2940 ippseudo.ippseudo_len = htons(m->m_pkthdr.len - thoff); 2941 MD5Update(&ctx, (char *)&ippseudo, sizeof(ippseudo)); 2942 } else { 2943 memset(&ip6pseudo, 0, sizeof(ip6pseudo)); 2944 ip6pseudo.ip6ph_src = ip6->ip6_src; 2945 in6_clearscope(&ip6pseudo.ip6ph_src); 2946 ip6pseudo.ip6ph_dst = ip6->ip6_dst; 2947 in6_clearscope(&ip6pseudo.ip6ph_dst); 2948 ip6pseudo.ip6ph_len = htons(m->m_pkthdr.len - thoff); 2949 ip6pseudo.ip6ph_nxt = IPPROTO_TCP; 2950 MD5Update(&ctx, (char *)&ip6pseudo, sizeof(ip6pseudo)); 2951 } 2952 2953 th0 = *th; 2954 th0.th_sum = 0; 2955 MD5Update(&ctx, (char *)&th0, sizeof(th0)); 2956 2957 l = m->m_pkthdr.len - thoff - tcphdrlen; 2958 if (l > 0) 2959 m_apply(m, thoff + tcphdrlen, 2960 m->m_pkthdr.len - thoff - tcphdrlen, 2961 tcp_signature_apply, &ctx); 2962 2963 MD5Update(&ctx, _KEYBUF(sav->key_auth), _KEYLEN(sav->key_auth)); 2964 MD5Final(sig, &ctx); 2965 2966 return (0); 2967 } 2968 #endif 2969 2970 static int 2971 tcp_dooptions(struct tcpcb *tp, const u_char *cp, int cnt, 2972 struct tcphdr *th, 2973 struct mbuf *m, int toff, struct tcp_opt_info *oi) 2974 { 2975 u_int16_t mss; 2976 int opt, optlen = 0; 2977 #ifdef TCP_SIGNATURE 2978 void *sigp = NULL; 2979 char sigbuf[TCP_SIGLEN]; 2980 struct secasvar *sav = NULL; 2981 #endif 2982 2983 for (; cp && cnt > 0; cnt -= optlen, cp += optlen) { 2984 opt = cp[0]; 2985 if (opt == TCPOPT_EOL) 2986 break; 2987 if (opt == TCPOPT_NOP) 2988 optlen = 1; 2989 else { 2990 if (cnt < 2) 2991 break; 2992 optlen = cp[1]; 2993 if (optlen < 2 || optlen > cnt) 2994 break; 2995 } 2996 switch (opt) { 2997 2998 default: 2999 continue; 3000 3001 case TCPOPT_MAXSEG: 3002 if (optlen != TCPOLEN_MAXSEG) 3003 continue; 3004 if (!(th->th_flags & TH_SYN)) 3005 continue; 3006 if (TCPS_HAVERCVDSYN(tp->t_state)) 3007 continue; 3008 bcopy(cp + 2, &mss, sizeof(mss)); 3009 oi->maxseg = ntohs(mss); 3010 break; 3011 3012 case TCPOPT_WINDOW: 3013 if (optlen != TCPOLEN_WINDOW) 3014 continue; 3015 if (!(th->th_flags & TH_SYN)) 3016 continue; 3017 if (TCPS_HAVERCVDSYN(tp->t_state)) 3018 continue; 3019 tp->t_flags |= TF_RCVD_SCALE; 3020 tp->requested_s_scale = cp[2]; 3021 if (tp->requested_s_scale > TCP_MAX_WINSHIFT) { 3022 #if 0 /*XXX*/ 3023 char *p; 3024 3025 if (ip) 3026 p = ntohl(ip->ip_src); 3027 #ifdef INET6 3028 else if (ip6) 3029 p = ip6_sprintf(&ip6->ip6_src); 3030 #endif 3031 else 3032 p = "(unknown)"; 3033 log(LOG_ERR, "TCP: invalid wscale %d from %s, " 3034 "assuming %d\n", 3035 tp->requested_s_scale, p, 3036 TCP_MAX_WINSHIFT); 3037 #else 3038 log(LOG_ERR, "TCP: invalid wscale %d, " 3039 "assuming %d\n", 3040 tp->requested_s_scale, 3041 TCP_MAX_WINSHIFT); 3042 #endif 3043 tp->requested_s_scale = TCP_MAX_WINSHIFT; 3044 } 3045 break; 3046 3047 case TCPOPT_TIMESTAMP: 3048 if (optlen != TCPOLEN_TIMESTAMP) 3049 continue; 3050 oi->ts_present = 1; 3051 bcopy(cp + 2, &oi->ts_val, sizeof(oi->ts_val)); 3052 NTOHL(oi->ts_val); 3053 bcopy(cp + 6, &oi->ts_ecr, sizeof(oi->ts_ecr)); 3054 NTOHL(oi->ts_ecr); 3055 3056 if (!(th->th_flags & TH_SYN)) 3057 continue; 3058 if (TCPS_HAVERCVDSYN(tp->t_state)) 3059 continue; 3060 /* 3061 * A timestamp received in a SYN makes 3062 * it ok to send timestamp requests and replies. 3063 */ 3064 tp->t_flags |= TF_RCVD_TSTMP; 3065 tp->ts_recent = oi->ts_val; 3066 tp->ts_recent_age = tcp_now; 3067 break; 3068 3069 case TCPOPT_SACK_PERMITTED: 3070 if (optlen != TCPOLEN_SACK_PERMITTED) 3071 continue; 3072 if (!(th->th_flags & TH_SYN)) 3073 continue; 3074 if (TCPS_HAVERCVDSYN(tp->t_state)) 3075 continue; 3076 if (tcp_do_sack) { 3077 tp->t_flags |= TF_SACK_PERMIT; 3078 tp->t_flags |= TF_WILL_SACK; 3079 } 3080 break; 3081 3082 case TCPOPT_SACK: 3083 tcp_sack_option(tp, th, cp, optlen); 3084 break; 3085 #ifdef TCP_SIGNATURE 3086 case TCPOPT_SIGNATURE: 3087 if (optlen != TCPOLEN_SIGNATURE) 3088 continue; 3089 if (sigp && bcmp(sigp, cp + 2, TCP_SIGLEN)) 3090 return (-1); 3091 3092 sigp = sigbuf; 3093 memcpy(sigbuf, cp + 2, TCP_SIGLEN); 3094 tp->t_flags |= TF_SIGNATURE; 3095 break; 3096 #endif 3097 } 3098 } 3099 3100 #ifdef TCP_SIGNATURE 3101 if (tp->t_flags & TF_SIGNATURE) { 3102 3103 sav = tcp_signature_getsav(m, th); 3104 3105 if (sav == NULL && tp->t_state == TCPS_LISTEN) 3106 return (-1); 3107 } 3108 3109 if ((sigp ? TF_SIGNATURE : 0) ^ (tp->t_flags & TF_SIGNATURE)) { 3110 if (sav == NULL) 3111 return (-1); 3112 #ifdef FAST_IPSEC 3113 KEY_FREESAV(&sav); 3114 #else 3115 key_freesav(sav); 3116 #endif 3117 return (-1); 3118 } 3119 3120 if (sigp) { 3121 char sig[TCP_SIGLEN]; 3122 3123 tcp_fields_to_net(th); 3124 if (tcp_signature(m, th, toff, sav, sig) < 0) { 3125 tcp_fields_to_host(th); 3126 if (sav == NULL) 3127 return (-1); 3128 #ifdef FAST_IPSEC 3129 KEY_FREESAV(&sav); 3130 #else 3131 key_freesav(sav); 3132 #endif 3133 return (-1); 3134 } 3135 tcp_fields_to_host(th); 3136 3137 if (bcmp(sig, sigp, TCP_SIGLEN)) { 3138 TCP_STATINC(TCP_STAT_BADSIG); 3139 if (sav == NULL) 3140 return (-1); 3141 #ifdef FAST_IPSEC 3142 KEY_FREESAV(&sav); 3143 #else 3144 key_freesav(sav); 3145 #endif 3146 return (-1); 3147 } else 3148 TCP_STATINC(TCP_STAT_GOODSIG); 3149 3150 key_sa_recordxfer(sav, m); 3151 #ifdef FAST_IPSEC 3152 KEY_FREESAV(&sav); 3153 #else 3154 key_freesav(sav); 3155 #endif 3156 } 3157 #endif 3158 3159 return (0); 3160 } 3161 3162 /* 3163 * Pull out of band byte out of a segment so 3164 * it doesn't appear in the user's data queue. 3165 * It is still reflected in the segment length for 3166 * sequencing purposes. 3167 */ 3168 void 3169 tcp_pulloutofband(struct socket *so, struct tcphdr *th, 3170 struct mbuf *m, int off) 3171 { 3172 int cnt = off + th->th_urp - 1; 3173 3174 while (cnt >= 0) { 3175 if (m->m_len > cnt) { 3176 char *cp = mtod(m, char *) + cnt; 3177 struct tcpcb *tp = sototcpcb(so); 3178 3179 tp->t_iobc = *cp; 3180 tp->t_oobflags |= TCPOOB_HAVEDATA; 3181 bcopy(cp+1, cp, (unsigned)(m->m_len - cnt - 1)); 3182 m->m_len--; 3183 return; 3184 } 3185 cnt -= m->m_len; 3186 m = m->m_next; 3187 if (m == 0) 3188 break; 3189 } 3190 panic("tcp_pulloutofband"); 3191 } 3192 3193 /* 3194 * Collect new round-trip time estimate 3195 * and update averages and current timeout. 3196 */ 3197 void 3198 tcp_xmit_timer(struct tcpcb *tp, uint32_t rtt) 3199 { 3200 int32_t delta; 3201 3202 TCP_STATINC(TCP_STAT_RTTUPDATED); 3203 if (tp->t_srtt != 0) { 3204 /* 3205 * srtt is stored as fixed point with 3 bits after the 3206 * binary point (i.e., scaled by 8). The following magic 3207 * is equivalent to the smoothing algorithm in rfc793 with 3208 * an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed 3209 * point). Adjust rtt to origin 0. 3210 */ 3211 delta = (rtt << 2) - (tp->t_srtt >> TCP_RTT_SHIFT); 3212 if ((tp->t_srtt += delta) <= 0) 3213 tp->t_srtt = 1 << 2; 3214 /* 3215 * We accumulate a smoothed rtt variance (actually, a 3216 * smoothed mean difference), then set the retransmit 3217 * timer to smoothed rtt + 4 times the smoothed variance. 3218 * rttvar is stored as fixed point with 2 bits after the 3219 * binary point (scaled by 4). The following is 3220 * equivalent to rfc793 smoothing with an alpha of .75 3221 * (rttvar = rttvar*3/4 + |delta| / 4). This replaces 3222 * rfc793's wired-in beta. 3223 */ 3224 if (delta < 0) 3225 delta = -delta; 3226 delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT); 3227 if ((tp->t_rttvar += delta) <= 0) 3228 tp->t_rttvar = 1 << 2; 3229 } else { 3230 /* 3231 * No rtt measurement yet - use the unsmoothed rtt. 3232 * Set the variance to half the rtt (so our first 3233 * retransmit happens at 3*rtt). 3234 */ 3235 tp->t_srtt = rtt << (TCP_RTT_SHIFT + 2); 3236 tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT + 2 - 1); 3237 } 3238 tp->t_rtttime = 0; 3239 tp->t_rxtshift = 0; 3240 3241 /* 3242 * the retransmit should happen at rtt + 4 * rttvar. 3243 * Because of the way we do the smoothing, srtt and rttvar 3244 * will each average +1/2 tick of bias. When we compute 3245 * the retransmit timer, we want 1/2 tick of rounding and 3246 * 1 extra tick because of +-1/2 tick uncertainty in the 3247 * firing of the timer. The bias will give us exactly the 3248 * 1.5 tick we need. But, because the bias is 3249 * statistical, we have to test that we don't drop below 3250 * the minimum feasible timer (which is 2 ticks). 3251 */ 3252 TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp), 3253 max(tp->t_rttmin, rtt + 2), TCPTV_REXMTMAX); 3254 3255 /* 3256 * We received an ack for a packet that wasn't retransmitted; 3257 * it is probably safe to discard any error indications we've 3258 * received recently. This isn't quite right, but close enough 3259 * for now (a route might have failed after we sent a segment, 3260 * and the return path might not be symmetrical). 3261 */ 3262 tp->t_softerror = 0; 3263 } 3264 3265 3266 /* 3267 * TCP compressed state engine. Currently used to hold compressed 3268 * state for SYN_RECEIVED. 3269 */ 3270 3271 u_long syn_cache_count; 3272 u_int32_t syn_hash1, syn_hash2; 3273 3274 #define SYN_HASH(sa, sp, dp) \ 3275 ((((sa)->s_addr^syn_hash1)*(((((u_int32_t)(dp))<<16) + \ 3276 ((u_int32_t)(sp)))^syn_hash2))) 3277 #ifndef INET6 3278 #define SYN_HASHALL(hash, src, dst) \ 3279 do { \ 3280 hash = SYN_HASH(&((const struct sockaddr_in *)(src))->sin_addr, \ 3281 ((const struct sockaddr_in *)(src))->sin_port, \ 3282 ((const struct sockaddr_in *)(dst))->sin_port); \ 3283 } while (/*CONSTCOND*/ 0) 3284 #else 3285 #define SYN_HASH6(sa, sp, dp) \ 3286 ((((sa)->s6_addr32[0] ^ (sa)->s6_addr32[3] ^ syn_hash1) * \ 3287 (((((u_int32_t)(dp))<<16) + ((u_int32_t)(sp)))^syn_hash2)) \ 3288 & 0x7fffffff) 3289 3290 #define SYN_HASHALL(hash, src, dst) \ 3291 do { \ 3292 switch ((src)->sa_family) { \ 3293 case AF_INET: \ 3294 hash = SYN_HASH(&((const struct sockaddr_in *)(src))->sin_addr, \ 3295 ((const struct sockaddr_in *)(src))->sin_port, \ 3296 ((const struct sockaddr_in *)(dst))->sin_port); \ 3297 break; \ 3298 case AF_INET6: \ 3299 hash = SYN_HASH6(&((const struct sockaddr_in6 *)(src))->sin6_addr, \ 3300 ((const struct sockaddr_in6 *)(src))->sin6_port, \ 3301 ((const struct sockaddr_in6 *)(dst))->sin6_port); \ 3302 break; \ 3303 default: \ 3304 hash = 0; \ 3305 } \ 3306 } while (/*CONSTCOND*/0) 3307 #endif /* INET6 */ 3308 3309 POOL_INIT(syn_cache_pool, sizeof(struct syn_cache), 0, 0, 0, "synpl", NULL, 3310 IPL_SOFTNET); 3311 3312 /* 3313 * We don't estimate RTT with SYNs, so each packet starts with the default 3314 * RTT and each timer step has a fixed timeout value. 3315 */ 3316 #define SYN_CACHE_TIMER_ARM(sc) \ 3317 do { \ 3318 TCPT_RANGESET((sc)->sc_rxtcur, \ 3319 TCPTV_SRTTDFLT * tcp_backoff[(sc)->sc_rxtshift], TCPTV_MIN, \ 3320 TCPTV_REXMTMAX); \ 3321 callout_reset(&(sc)->sc_timer, \ 3322 (sc)->sc_rxtcur * (hz / PR_SLOWHZ), syn_cache_timer, (sc)); \ 3323 } while (/*CONSTCOND*/0) 3324 3325 #define SYN_CACHE_TIMESTAMP(sc) (tcp_now - (sc)->sc_timebase) 3326 3327 static inline void 3328 syn_cache_rm(struct syn_cache *sc) 3329 { 3330 TAILQ_REMOVE(&tcp_syn_cache[sc->sc_bucketidx].sch_bucket, 3331 sc, sc_bucketq); 3332 sc->sc_tp = NULL; 3333 LIST_REMOVE(sc, sc_tpq); 3334 tcp_syn_cache[sc->sc_bucketidx].sch_length--; 3335 callout_stop(&sc->sc_timer); 3336 syn_cache_count--; 3337 } 3338 3339 static inline void 3340 syn_cache_put(struct syn_cache *sc) 3341 { 3342 if (sc->sc_ipopts) 3343 (void) m_free(sc->sc_ipopts); 3344 rtcache_free(&sc->sc_route); 3345 if (callout_invoking(&sc->sc_timer)) 3346 sc->sc_flags |= SCF_DEAD; 3347 else { 3348 callout_destroy(&sc->sc_timer); 3349 pool_put(&syn_cache_pool, sc); 3350 } 3351 } 3352 3353 void 3354 syn_cache_init(void) 3355 { 3356 int i; 3357 3358 /* Initialize the hash buckets. */ 3359 for (i = 0; i < tcp_syn_cache_size; i++) 3360 TAILQ_INIT(&tcp_syn_cache[i].sch_bucket); 3361 } 3362 3363 void 3364 syn_cache_insert(struct syn_cache *sc, struct tcpcb *tp) 3365 { 3366 struct syn_cache_head *scp; 3367 struct syn_cache *sc2; 3368 int s; 3369 3370 /* 3371 * If there are no entries in the hash table, reinitialize 3372 * the hash secrets. 3373 */ 3374 if (syn_cache_count == 0) { 3375 syn_hash1 = arc4random(); 3376 syn_hash2 = arc4random(); 3377 } 3378 3379 SYN_HASHALL(sc->sc_hash, &sc->sc_src.sa, &sc->sc_dst.sa); 3380 sc->sc_bucketidx = sc->sc_hash % tcp_syn_cache_size; 3381 scp = &tcp_syn_cache[sc->sc_bucketidx]; 3382 3383 /* 3384 * Make sure that we don't overflow the per-bucket 3385 * limit or the total cache size limit. 3386 */ 3387 s = splsoftnet(); 3388 if (scp->sch_length >= tcp_syn_bucket_limit) { 3389 TCP_STATINC(TCP_STAT_SC_BUCKETOVERFLOW); 3390 /* 3391 * The bucket is full. Toss the oldest element in the 3392 * bucket. This will be the first entry in the bucket. 3393 */ 3394 sc2 = TAILQ_FIRST(&scp->sch_bucket); 3395 #ifdef DIAGNOSTIC 3396 /* 3397 * This should never happen; we should always find an 3398 * entry in our bucket. 3399 */ 3400 if (sc2 == NULL) 3401 panic("syn_cache_insert: bucketoverflow: impossible"); 3402 #endif 3403 syn_cache_rm(sc2); 3404 syn_cache_put(sc2); /* calls pool_put but see spl above */ 3405 } else if (syn_cache_count >= tcp_syn_cache_limit) { 3406 struct syn_cache_head *scp2, *sce; 3407 3408 TCP_STATINC(TCP_STAT_SC_OVERFLOWED); 3409 /* 3410 * The cache is full. Toss the oldest entry in the 3411 * first non-empty bucket we can find. 3412 * 3413 * XXX We would really like to toss the oldest 3414 * entry in the cache, but we hope that this 3415 * condition doesn't happen very often. 3416 */ 3417 scp2 = scp; 3418 if (TAILQ_EMPTY(&scp2->sch_bucket)) { 3419 sce = &tcp_syn_cache[tcp_syn_cache_size]; 3420 for (++scp2; scp2 != scp; scp2++) { 3421 if (scp2 >= sce) 3422 scp2 = &tcp_syn_cache[0]; 3423 if (! TAILQ_EMPTY(&scp2->sch_bucket)) 3424 break; 3425 } 3426 #ifdef DIAGNOSTIC 3427 /* 3428 * This should never happen; we should always find a 3429 * non-empty bucket. 3430 */ 3431 if (scp2 == scp) 3432 panic("syn_cache_insert: cacheoverflow: " 3433 "impossible"); 3434 #endif 3435 } 3436 sc2 = TAILQ_FIRST(&scp2->sch_bucket); 3437 syn_cache_rm(sc2); 3438 syn_cache_put(sc2); /* calls pool_put but see spl above */ 3439 } 3440 3441 /* 3442 * Initialize the entry's timer. 3443 */ 3444 sc->sc_rxttot = 0; 3445 sc->sc_rxtshift = 0; 3446 SYN_CACHE_TIMER_ARM(sc); 3447 3448 /* Link it from tcpcb entry */ 3449 LIST_INSERT_HEAD(&tp->t_sc, sc, sc_tpq); 3450 3451 /* Put it into the bucket. */ 3452 TAILQ_INSERT_TAIL(&scp->sch_bucket, sc, sc_bucketq); 3453 scp->sch_length++; 3454 syn_cache_count++; 3455 3456 TCP_STATINC(TCP_STAT_SC_ADDED); 3457 splx(s); 3458 } 3459 3460 /* 3461 * Walk the timer queues, looking for SYN,ACKs that need to be retransmitted. 3462 * If we have retransmitted an entry the maximum number of times, expire 3463 * that entry. 3464 */ 3465 void 3466 syn_cache_timer(void *arg) 3467 { 3468 struct syn_cache *sc = arg; 3469 int s; 3470 3471 s = splsoftnet(); 3472 callout_ack(&sc->sc_timer); 3473 3474 if (__predict_false(sc->sc_flags & SCF_DEAD)) { 3475 TCP_STATINC(TCP_STAT_SC_DELAYED_FREE); 3476 callout_destroy(&sc->sc_timer); 3477 pool_put(&syn_cache_pool, sc); 3478 splx(s); 3479 return; 3480 } 3481 3482 if (__predict_false(sc->sc_rxtshift == TCP_MAXRXTSHIFT)) { 3483 /* Drop it -- too many retransmissions. */ 3484 goto dropit; 3485 } 3486 3487 /* 3488 * Compute the total amount of time this entry has 3489 * been on a queue. If this entry has been on longer 3490 * than the keep alive timer would allow, expire it. 3491 */ 3492 sc->sc_rxttot += sc->sc_rxtcur; 3493 if (sc->sc_rxttot >= tcp_keepinit) 3494 goto dropit; 3495 3496 TCP_STATINC(TCP_STAT_SC_RETRANSMITTED); 3497 (void) syn_cache_respond(sc, NULL); 3498 3499 /* Advance the timer back-off. */ 3500 sc->sc_rxtshift++; 3501 SYN_CACHE_TIMER_ARM(sc); 3502 3503 splx(s); 3504 return; 3505 3506 dropit: 3507 TCP_STATINC(TCP_STAT_SC_TIMED_OUT); 3508 syn_cache_rm(sc); 3509 syn_cache_put(sc); /* calls pool_put but see spl above */ 3510 splx(s); 3511 } 3512 3513 /* 3514 * Remove syn cache created by the specified tcb entry, 3515 * because this does not make sense to keep them 3516 * (if there's no tcb entry, syn cache entry will never be used) 3517 */ 3518 void 3519 syn_cache_cleanup(struct tcpcb *tp) 3520 { 3521 struct syn_cache *sc, *nsc; 3522 int s; 3523 3524 s = splsoftnet(); 3525 3526 for (sc = LIST_FIRST(&tp->t_sc); sc != NULL; sc = nsc) { 3527 nsc = LIST_NEXT(sc, sc_tpq); 3528 3529 #ifdef DIAGNOSTIC 3530 if (sc->sc_tp != tp) 3531 panic("invalid sc_tp in syn_cache_cleanup"); 3532 #endif 3533 syn_cache_rm(sc); 3534 syn_cache_put(sc); /* calls pool_put but see spl above */ 3535 } 3536 /* just for safety */ 3537 LIST_INIT(&tp->t_sc); 3538 3539 splx(s); 3540 } 3541 3542 /* 3543 * Find an entry in the syn cache. 3544 */ 3545 struct syn_cache * 3546 syn_cache_lookup(const struct sockaddr *src, const struct sockaddr *dst, 3547 struct syn_cache_head **headp) 3548 { 3549 struct syn_cache *sc; 3550 struct syn_cache_head *scp; 3551 u_int32_t hash; 3552 int s; 3553 3554 SYN_HASHALL(hash, src, dst); 3555 3556 scp = &tcp_syn_cache[hash % tcp_syn_cache_size]; 3557 *headp = scp; 3558 s = splsoftnet(); 3559 for (sc = TAILQ_FIRST(&scp->sch_bucket); sc != NULL; 3560 sc = TAILQ_NEXT(sc, sc_bucketq)) { 3561 if (sc->sc_hash != hash) 3562 continue; 3563 if (!bcmp(&sc->sc_src, src, src->sa_len) && 3564 !bcmp(&sc->sc_dst, dst, dst->sa_len)) { 3565 splx(s); 3566 return (sc); 3567 } 3568 } 3569 splx(s); 3570 return (NULL); 3571 } 3572 3573 /* 3574 * This function gets called when we receive an ACK for a 3575 * socket in the LISTEN state. We look up the connection 3576 * in the syn cache, and if its there, we pull it out of 3577 * the cache and turn it into a full-blown connection in 3578 * the SYN-RECEIVED state. 3579 * 3580 * The return values may not be immediately obvious, and their effects 3581 * can be subtle, so here they are: 3582 * 3583 * NULL SYN was not found in cache; caller should drop the 3584 * packet and send an RST. 3585 * 3586 * -1 We were unable to create the new connection, and are 3587 * aborting it. An ACK,RST is being sent to the peer 3588 * (unless we got screwey sequence numbners; see below), 3589 * because the 3-way handshake has been completed. Caller 3590 * should not free the mbuf, since we may be using it. If 3591 * we are not, we will free it. 3592 * 3593 * Otherwise, the return value is a pointer to the new socket 3594 * associated with the connection. 3595 */ 3596 struct socket * 3597 syn_cache_get(struct sockaddr *src, struct sockaddr *dst, 3598 struct tcphdr *th, unsigned int hlen, unsigned int tlen, 3599 struct socket *so, struct mbuf *m) 3600 { 3601 struct syn_cache *sc; 3602 struct syn_cache_head *scp; 3603 struct inpcb *inp = NULL; 3604 #ifdef INET6 3605 struct in6pcb *in6p = NULL; 3606 #endif 3607 struct tcpcb *tp = 0; 3608 struct mbuf *am; 3609 int s; 3610 struct socket *oso; 3611 3612 s = splsoftnet(); 3613 if ((sc = syn_cache_lookup(src, dst, &scp)) == NULL) { 3614 splx(s); 3615 return (NULL); 3616 } 3617 3618 /* 3619 * Verify the sequence and ack numbers. Try getting the correct 3620 * response again. 3621 */ 3622 if ((th->th_ack != sc->sc_iss + 1) || 3623 SEQ_LEQ(th->th_seq, sc->sc_irs) || 3624 SEQ_GT(th->th_seq, sc->sc_irs + 1 + sc->sc_win)) { 3625 (void) syn_cache_respond(sc, m); 3626 splx(s); 3627 return ((struct socket *)(-1)); 3628 } 3629 3630 /* Remove this cache entry */ 3631 syn_cache_rm(sc); 3632 splx(s); 3633 3634 /* 3635 * Ok, create the full blown connection, and set things up 3636 * as they would have been set up if we had created the 3637 * connection when the SYN arrived. If we can't create 3638 * the connection, abort it. 3639 */ 3640 /* 3641 * inp still has the OLD in_pcb stuff, set the 3642 * v6-related flags on the new guy, too. This is 3643 * done particularly for the case where an AF_INET6 3644 * socket is bound only to a port, and a v4 connection 3645 * comes in on that port. 3646 * we also copy the flowinfo from the original pcb 3647 * to the new one. 3648 */ 3649 oso = so; 3650 so = sonewconn(so, SS_ISCONNECTED); 3651 if (so == NULL) 3652 goto resetandabort; 3653 3654 switch (so->so_proto->pr_domain->dom_family) { 3655 #ifdef INET 3656 case AF_INET: 3657 inp = sotoinpcb(so); 3658 break; 3659 #endif 3660 #ifdef INET6 3661 case AF_INET6: 3662 in6p = sotoin6pcb(so); 3663 break; 3664 #endif 3665 } 3666 switch (src->sa_family) { 3667 #ifdef INET 3668 case AF_INET: 3669 if (inp) { 3670 inp->inp_laddr = ((struct sockaddr_in *)dst)->sin_addr; 3671 inp->inp_lport = ((struct sockaddr_in *)dst)->sin_port; 3672 inp->inp_options = ip_srcroute(); 3673 in_pcbstate(inp, INP_BOUND); 3674 if (inp->inp_options == NULL) { 3675 inp->inp_options = sc->sc_ipopts; 3676 sc->sc_ipopts = NULL; 3677 } 3678 } 3679 #ifdef INET6 3680 else if (in6p) { 3681 /* IPv4 packet to AF_INET6 socket */ 3682 bzero(&in6p->in6p_laddr, sizeof(in6p->in6p_laddr)); 3683 in6p->in6p_laddr.s6_addr16[5] = htons(0xffff); 3684 bcopy(&((struct sockaddr_in *)dst)->sin_addr, 3685 &in6p->in6p_laddr.s6_addr32[3], 3686 sizeof(((struct sockaddr_in *)dst)->sin_addr)); 3687 in6p->in6p_lport = ((struct sockaddr_in *)dst)->sin_port; 3688 in6totcpcb(in6p)->t_family = AF_INET; 3689 if (sotoin6pcb(oso)->in6p_flags & IN6P_IPV6_V6ONLY) 3690 in6p->in6p_flags |= IN6P_IPV6_V6ONLY; 3691 else 3692 in6p->in6p_flags &= ~IN6P_IPV6_V6ONLY; 3693 in6_pcbstate(in6p, IN6P_BOUND); 3694 } 3695 #endif 3696 break; 3697 #endif 3698 #ifdef INET6 3699 case AF_INET6: 3700 if (in6p) { 3701 in6p->in6p_laddr = ((struct sockaddr_in6 *)dst)->sin6_addr; 3702 in6p->in6p_lport = ((struct sockaddr_in6 *)dst)->sin6_port; 3703 in6_pcbstate(in6p, IN6P_BOUND); 3704 } 3705 break; 3706 #endif 3707 } 3708 #ifdef INET6 3709 if (in6p && in6totcpcb(in6p)->t_family == AF_INET6 && sotoinpcb(oso)) { 3710 struct in6pcb *oin6p = sotoin6pcb(oso); 3711 /* inherit socket options from the listening socket */ 3712 in6p->in6p_flags |= (oin6p->in6p_flags & IN6P_CONTROLOPTS); 3713 if (in6p->in6p_flags & IN6P_CONTROLOPTS) { 3714 m_freem(in6p->in6p_options); 3715 in6p->in6p_options = 0; 3716 } 3717 ip6_savecontrol(in6p, &in6p->in6p_options, 3718 mtod(m, struct ip6_hdr *), m); 3719 } 3720 #endif 3721 3722 #if defined(IPSEC) || defined(FAST_IPSEC) 3723 /* 3724 * we make a copy of policy, instead of sharing the policy, 3725 * for better behavior in terms of SA lookup and dead SA removal. 3726 */ 3727 if (inp) { 3728 /* copy old policy into new socket's */ 3729 if (ipsec_copy_pcbpolicy(sotoinpcb(oso)->inp_sp, inp->inp_sp)) 3730 printf("tcp_input: could not copy policy\n"); 3731 } 3732 #ifdef INET6 3733 else if (in6p) { 3734 /* copy old policy into new socket's */ 3735 if (ipsec_copy_pcbpolicy(sotoin6pcb(oso)->in6p_sp, 3736 in6p->in6p_sp)) 3737 printf("tcp_input: could not copy policy\n"); 3738 } 3739 #endif 3740 #endif 3741 3742 /* 3743 * Give the new socket our cached route reference. 3744 */ 3745 if (inp) { 3746 rtcache_copy(&inp->inp_route, &sc->sc_route); 3747 rtcache_free(&sc->sc_route); 3748 } 3749 #ifdef INET6 3750 else { 3751 rtcache_copy(&in6p->in6p_route, &sc->sc_route); 3752 rtcache_free(&sc->sc_route); 3753 } 3754 #endif 3755 3756 am = m_get(M_DONTWAIT, MT_SONAME); /* XXX */ 3757 if (am == NULL) 3758 goto resetandabort; 3759 MCLAIM(am, &tcp_mowner); 3760 am->m_len = src->sa_len; 3761 bcopy(src, mtod(am, void *), src->sa_len); 3762 if (inp) { 3763 if (in_pcbconnect(inp, am, &lwp0)) { 3764 (void) m_free(am); 3765 goto resetandabort; 3766 } 3767 } 3768 #ifdef INET6 3769 else if (in6p) { 3770 if (src->sa_family == AF_INET) { 3771 /* IPv4 packet to AF_INET6 socket */ 3772 struct sockaddr_in6 *sin6; 3773 sin6 = mtod(am, struct sockaddr_in6 *); 3774 am->m_len = sizeof(*sin6); 3775 bzero(sin6, sizeof(*sin6)); 3776 sin6->sin6_family = AF_INET6; 3777 sin6->sin6_len = sizeof(*sin6); 3778 sin6->sin6_port = ((struct sockaddr_in *)src)->sin_port; 3779 sin6->sin6_addr.s6_addr16[5] = htons(0xffff); 3780 bcopy(&((struct sockaddr_in *)src)->sin_addr, 3781 &sin6->sin6_addr.s6_addr32[3], 3782 sizeof(sin6->sin6_addr.s6_addr32[3])); 3783 } 3784 if (in6_pcbconnect(in6p, am, NULL)) { 3785 (void) m_free(am); 3786 goto resetandabort; 3787 } 3788 } 3789 #endif 3790 else { 3791 (void) m_free(am); 3792 goto resetandabort; 3793 } 3794 (void) m_free(am); 3795 3796 if (inp) 3797 tp = intotcpcb(inp); 3798 #ifdef INET6 3799 else if (in6p) 3800 tp = in6totcpcb(in6p); 3801 #endif 3802 else 3803 tp = NULL; 3804 tp->t_flags = sototcpcb(oso)->t_flags & TF_NODELAY; 3805 if (sc->sc_request_r_scale != 15) { 3806 tp->requested_s_scale = sc->sc_requested_s_scale; 3807 tp->request_r_scale = sc->sc_request_r_scale; 3808 tp->snd_scale = sc->sc_requested_s_scale; 3809 tp->rcv_scale = sc->sc_request_r_scale; 3810 tp->t_flags |= TF_REQ_SCALE|TF_RCVD_SCALE; 3811 } 3812 if (sc->sc_flags & SCF_TIMESTAMP) 3813 tp->t_flags |= TF_REQ_TSTMP|TF_RCVD_TSTMP; 3814 tp->ts_timebase = sc->sc_timebase; 3815 3816 tp->t_template = tcp_template(tp); 3817 if (tp->t_template == 0) { 3818 tp = tcp_drop(tp, ENOBUFS); /* destroys socket */ 3819 so = NULL; 3820 m_freem(m); 3821 goto abort; 3822 } 3823 3824 tp->iss = sc->sc_iss; 3825 tp->irs = sc->sc_irs; 3826 tcp_sendseqinit(tp); 3827 tcp_rcvseqinit(tp); 3828 tp->t_state = TCPS_SYN_RECEIVED; 3829 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit); 3830 TCP_STATINC(TCP_STAT_ACCEPTS); 3831 3832 if ((sc->sc_flags & SCF_SACK_PERMIT) && tcp_do_sack) 3833 tp->t_flags |= TF_WILL_SACK; 3834 3835 if ((sc->sc_flags & SCF_ECN_PERMIT) && tcp_do_ecn) 3836 tp->t_flags |= TF_ECN_PERMIT; 3837 3838 #ifdef TCP_SIGNATURE 3839 if (sc->sc_flags & SCF_SIGNATURE) 3840 tp->t_flags |= TF_SIGNATURE; 3841 #endif 3842 3843 /* Initialize tp->t_ourmss before we deal with the peer's! */ 3844 tp->t_ourmss = sc->sc_ourmaxseg; 3845 tcp_mss_from_peer(tp, sc->sc_peermaxseg); 3846 3847 /* 3848 * Initialize the initial congestion window. If we 3849 * had to retransmit the SYN,ACK, we must initialize cwnd 3850 * to 1 segment (i.e. the Loss Window). 3851 */ 3852 if (sc->sc_rxtshift) 3853 tp->snd_cwnd = tp->t_peermss; 3854 else { 3855 int ss = tcp_init_win; 3856 #ifdef INET 3857 if (inp != NULL && in_localaddr(inp->inp_faddr)) 3858 ss = tcp_init_win_local; 3859 #endif 3860 #ifdef INET6 3861 if (in6p != NULL && in6_localaddr(&in6p->in6p_faddr)) 3862 ss = tcp_init_win_local; 3863 #endif 3864 tp->snd_cwnd = TCP_INITIAL_WINDOW(ss, tp->t_peermss); 3865 } 3866 3867 tcp_rmx_rtt(tp); 3868 tp->snd_wl1 = sc->sc_irs; 3869 tp->rcv_up = sc->sc_irs + 1; 3870 3871 /* 3872 * This is what whould have happened in tcp_output() when 3873 * the SYN,ACK was sent. 3874 */ 3875 tp->snd_up = tp->snd_una; 3876 tp->snd_max = tp->snd_nxt = tp->iss+1; 3877 TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur); 3878 if (sc->sc_win > 0 && SEQ_GT(tp->rcv_nxt + sc->sc_win, tp->rcv_adv)) 3879 tp->rcv_adv = tp->rcv_nxt + sc->sc_win; 3880 tp->last_ack_sent = tp->rcv_nxt; 3881 tp->t_partialacks = -1; 3882 tp->t_dupacks = 0; 3883 3884 TCP_STATINC(TCP_STAT_SC_COMPLETED); 3885 s = splsoftnet(); 3886 syn_cache_put(sc); 3887 splx(s); 3888 return (so); 3889 3890 resetandabort: 3891 (void)tcp_respond(NULL, m, m, th, (tcp_seq)0, th->th_ack, TH_RST); 3892 abort: 3893 if (so != NULL) 3894 (void) soabort(so); 3895 s = splsoftnet(); 3896 syn_cache_put(sc); 3897 splx(s); 3898 TCP_STATINC(TCP_STAT_SC_ABORTED); 3899 return ((struct socket *)(-1)); 3900 } 3901 3902 /* 3903 * This function is called when we get a RST for a 3904 * non-existent connection, so that we can see if the 3905 * connection is in the syn cache. If it is, zap it. 3906 */ 3907 3908 void 3909 syn_cache_reset(struct sockaddr *src, struct sockaddr *dst, struct tcphdr *th) 3910 { 3911 struct syn_cache *sc; 3912 struct syn_cache_head *scp; 3913 int s = splsoftnet(); 3914 3915 if ((sc = syn_cache_lookup(src, dst, &scp)) == NULL) { 3916 splx(s); 3917 return; 3918 } 3919 if (SEQ_LT(th->th_seq, sc->sc_irs) || 3920 SEQ_GT(th->th_seq, sc->sc_irs+1)) { 3921 splx(s); 3922 return; 3923 } 3924 syn_cache_rm(sc); 3925 TCP_STATINC(TCP_STAT_SC_RESET); 3926 syn_cache_put(sc); /* calls pool_put but see spl above */ 3927 splx(s); 3928 } 3929 3930 void 3931 syn_cache_unreach(const struct sockaddr *src, const struct sockaddr *dst, 3932 struct tcphdr *th) 3933 { 3934 struct syn_cache *sc; 3935 struct syn_cache_head *scp; 3936 int s; 3937 3938 s = splsoftnet(); 3939 if ((sc = syn_cache_lookup(src, dst, &scp)) == NULL) { 3940 splx(s); 3941 return; 3942 } 3943 /* If the sequence number != sc_iss, then it's a bogus ICMP msg */ 3944 if (ntohl (th->th_seq) != sc->sc_iss) { 3945 splx(s); 3946 return; 3947 } 3948 3949 /* 3950 * If we've retransmitted 3 times and this is our second error, 3951 * we remove the entry. Otherwise, we allow it to continue on. 3952 * This prevents us from incorrectly nuking an entry during a 3953 * spurious network outage. 3954 * 3955 * See tcp_notify(). 3956 */ 3957 if ((sc->sc_flags & SCF_UNREACH) == 0 || sc->sc_rxtshift < 3) { 3958 sc->sc_flags |= SCF_UNREACH; 3959 splx(s); 3960 return; 3961 } 3962 3963 syn_cache_rm(sc); 3964 TCP_STATINC(TCP_STAT_SC_UNREACH); 3965 syn_cache_put(sc); /* calls pool_put but see spl above */ 3966 splx(s); 3967 } 3968 3969 /* 3970 * Given a LISTEN socket and an inbound SYN request, add 3971 * this to the syn cache, and send back a segment: 3972 * <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK> 3973 * to the source. 3974 * 3975 * IMPORTANT NOTE: We do _NOT_ ACK data that might accompany the SYN. 3976 * Doing so would require that we hold onto the data and deliver it 3977 * to the application. However, if we are the target of a SYN-flood 3978 * DoS attack, an attacker could send data which would eventually 3979 * consume all available buffer space if it were ACKed. By not ACKing 3980 * the data, we avoid this DoS scenario. 3981 */ 3982 3983 int 3984 syn_cache_add(struct sockaddr *src, struct sockaddr *dst, struct tcphdr *th, 3985 unsigned int hlen, struct socket *so, struct mbuf *m, u_char *optp, 3986 int optlen, struct tcp_opt_info *oi) 3987 { 3988 struct tcpcb tb, *tp; 3989 long win; 3990 struct syn_cache *sc; 3991 struct syn_cache_head *scp; 3992 struct mbuf *ipopts; 3993 struct tcp_opt_info opti; 3994 int s; 3995 3996 tp = sototcpcb(so); 3997 3998 bzero(&opti, sizeof(opti)); 3999 4000 /* 4001 * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN 4002 * 4003 * Note this check is performed in tcp_input() very early on. 4004 */ 4005 4006 /* 4007 * Initialize some local state. 4008 */ 4009 win = sbspace(&so->so_rcv); 4010 if (win > TCP_MAXWIN) 4011 win = TCP_MAXWIN; 4012 4013 switch (src->sa_family) { 4014 #ifdef INET 4015 case AF_INET: 4016 /* 4017 * Remember the IP options, if any. 4018 */ 4019 ipopts = ip_srcroute(); 4020 break; 4021 #endif 4022 default: 4023 ipopts = NULL; 4024 } 4025 4026 #ifdef TCP_SIGNATURE 4027 if (optp || (tp->t_flags & TF_SIGNATURE)) 4028 #else 4029 if (optp) 4030 #endif 4031 { 4032 tb.t_flags = tcp_do_rfc1323 ? (TF_REQ_SCALE|TF_REQ_TSTMP) : 0; 4033 #ifdef TCP_SIGNATURE 4034 tb.t_flags |= (tp->t_flags & TF_SIGNATURE); 4035 #endif 4036 tb.t_state = TCPS_LISTEN; 4037 if (tcp_dooptions(&tb, optp, optlen, th, m, m->m_pkthdr.len - 4038 sizeof(struct tcphdr) - optlen - hlen, oi) < 0) 4039 return (0); 4040 } else 4041 tb.t_flags = 0; 4042 4043 /* 4044 * See if we already have an entry for this connection. 4045 * If we do, resend the SYN,ACK. We do not count this 4046 * as a retransmission (XXX though maybe we should). 4047 */ 4048 if ((sc = syn_cache_lookup(src, dst, &scp)) != NULL) { 4049 TCP_STATINC(TCP_STAT_SC_DUPESYN); 4050 if (ipopts) { 4051 /* 4052 * If we were remembering a previous source route, 4053 * forget it and use the new one we've been given. 4054 */ 4055 if (sc->sc_ipopts) 4056 (void) m_free(sc->sc_ipopts); 4057 sc->sc_ipopts = ipopts; 4058 } 4059 sc->sc_timestamp = tb.ts_recent; 4060 if (syn_cache_respond(sc, m) == 0) { 4061 uint64_t *tcps = TCP_STAT_GETREF(); 4062 tcps[TCP_STAT_SNDACKS]++; 4063 tcps[TCP_STAT_SNDTOTAL]++; 4064 TCP_STAT_PUTREF(); 4065 } 4066 return (1); 4067 } 4068 4069 s = splsoftnet(); 4070 sc = pool_get(&syn_cache_pool, PR_NOWAIT); 4071 splx(s); 4072 if (sc == NULL) { 4073 if (ipopts) 4074 (void) m_free(ipopts); 4075 return (0); 4076 } 4077 4078 /* 4079 * Fill in the cache, and put the necessary IP and TCP 4080 * options into the reply. 4081 */ 4082 bzero(sc, sizeof(struct syn_cache)); 4083 callout_init(&sc->sc_timer, 0); 4084 bcopy(src, &sc->sc_src, src->sa_len); 4085 bcopy(dst, &sc->sc_dst, dst->sa_len); 4086 sc->sc_flags = 0; 4087 sc->sc_ipopts = ipopts; 4088 sc->sc_irs = th->th_seq; 4089 switch (src->sa_family) { 4090 #ifdef INET 4091 case AF_INET: 4092 { 4093 struct sockaddr_in *srcin = (void *) src; 4094 struct sockaddr_in *dstin = (void *) dst; 4095 4096 sc->sc_iss = tcp_new_iss1(&dstin->sin_addr, 4097 &srcin->sin_addr, dstin->sin_port, 4098 srcin->sin_port, sizeof(dstin->sin_addr), 0); 4099 break; 4100 } 4101 #endif /* INET */ 4102 #ifdef INET6 4103 case AF_INET6: 4104 { 4105 struct sockaddr_in6 *srcin6 = (void *) src; 4106 struct sockaddr_in6 *dstin6 = (void *) dst; 4107 4108 sc->sc_iss = tcp_new_iss1(&dstin6->sin6_addr, 4109 &srcin6->sin6_addr, dstin6->sin6_port, 4110 srcin6->sin6_port, sizeof(dstin6->sin6_addr), 0); 4111 break; 4112 } 4113 #endif /* INET6 */ 4114 } 4115 sc->sc_peermaxseg = oi->maxseg; 4116 sc->sc_ourmaxseg = tcp_mss_to_advertise(m->m_flags & M_PKTHDR ? 4117 m->m_pkthdr.rcvif : NULL, 4118 sc->sc_src.sa.sa_family); 4119 sc->sc_win = win; 4120 sc->sc_timebase = tcp_now - 1; /* see tcp_newtcpcb() */ 4121 sc->sc_timestamp = tb.ts_recent; 4122 if ((tb.t_flags & (TF_REQ_TSTMP|TF_RCVD_TSTMP)) == 4123 (TF_REQ_TSTMP|TF_RCVD_TSTMP)) 4124 sc->sc_flags |= SCF_TIMESTAMP; 4125 if ((tb.t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == 4126 (TF_RCVD_SCALE|TF_REQ_SCALE)) { 4127 sc->sc_requested_s_scale = tb.requested_s_scale; 4128 sc->sc_request_r_scale = 0; 4129 /* 4130 * Pick the smallest possible scaling factor that 4131 * will still allow us to scale up to sb_max. 4132 * 4133 * We do this because there are broken firewalls that 4134 * will corrupt the window scale option, leading to 4135 * the other endpoint believing that our advertised 4136 * window is unscaled. At scale factors larger than 4137 * 5 the unscaled window will drop below 1500 bytes, 4138 * leading to serious problems when traversing these 4139 * broken firewalls. 4140 * 4141 * With the default sbmax of 256K, a scale factor 4142 * of 3 will be chosen by this algorithm. Those who 4143 * choose a larger sbmax should watch out 4144 * for the compatiblity problems mentioned above. 4145 * 4146 * RFC1323: The Window field in a SYN (i.e., a <SYN> 4147 * or <SYN,ACK>) segment itself is never scaled. 4148 */ 4149 while (sc->sc_request_r_scale < TCP_MAX_WINSHIFT && 4150 (TCP_MAXWIN << sc->sc_request_r_scale) < sb_max) 4151 sc->sc_request_r_scale++; 4152 } else { 4153 sc->sc_requested_s_scale = 15; 4154 sc->sc_request_r_scale = 15; 4155 } 4156 if ((tb.t_flags & TF_SACK_PERMIT) && tcp_do_sack) 4157 sc->sc_flags |= SCF_SACK_PERMIT; 4158 4159 /* 4160 * ECN setup packet recieved. 4161 */ 4162 if ((th->th_flags & (TH_ECE|TH_CWR)) && tcp_do_ecn) 4163 sc->sc_flags |= SCF_ECN_PERMIT; 4164 4165 #ifdef TCP_SIGNATURE 4166 if (tb.t_flags & TF_SIGNATURE) 4167 sc->sc_flags |= SCF_SIGNATURE; 4168 #endif 4169 sc->sc_tp = tp; 4170 if (syn_cache_respond(sc, m) == 0) { 4171 uint64_t *tcps = TCP_STAT_GETREF(); 4172 tcps[TCP_STAT_SNDACKS]++; 4173 tcps[TCP_STAT_SNDTOTAL]++; 4174 TCP_STAT_PUTREF(); 4175 syn_cache_insert(sc, tp); 4176 } else { 4177 s = splsoftnet(); 4178 syn_cache_put(sc); 4179 splx(s); 4180 TCP_STATINC(TCP_STAT_SC_DROPPED); 4181 } 4182 return (1); 4183 } 4184 4185 int 4186 syn_cache_respond(struct syn_cache *sc, struct mbuf *m) 4187 { 4188 #ifdef INET6 4189 struct rtentry *rt; 4190 #endif 4191 struct route *ro; 4192 u_int8_t *optp; 4193 int optlen, error; 4194 u_int16_t tlen; 4195 struct ip *ip = NULL; 4196 #ifdef INET6 4197 struct ip6_hdr *ip6 = NULL; 4198 #endif 4199 struct tcpcb *tp = NULL; 4200 struct tcphdr *th; 4201 u_int hlen; 4202 struct socket *so; 4203 4204 ro = &sc->sc_route; 4205 switch (sc->sc_src.sa.sa_family) { 4206 case AF_INET: 4207 hlen = sizeof(struct ip); 4208 break; 4209 #ifdef INET6 4210 case AF_INET6: 4211 hlen = sizeof(struct ip6_hdr); 4212 break; 4213 #endif 4214 default: 4215 if (m) 4216 m_freem(m); 4217 return (EAFNOSUPPORT); 4218 } 4219 4220 /* Compute the size of the TCP options. */ 4221 optlen = 4 + (sc->sc_request_r_scale != 15 ? 4 : 0) + 4222 ((sc->sc_flags & SCF_SACK_PERMIT) ? (TCPOLEN_SACK_PERMITTED + 2) : 0) + 4223 #ifdef TCP_SIGNATURE 4224 ((sc->sc_flags & SCF_SIGNATURE) ? (TCPOLEN_SIGNATURE + 2) : 0) + 4225 #endif 4226 ((sc->sc_flags & SCF_TIMESTAMP) ? TCPOLEN_TSTAMP_APPA : 0); 4227 4228 tlen = hlen + sizeof(struct tcphdr) + optlen; 4229 4230 /* 4231 * Create the IP+TCP header from scratch. 4232 */ 4233 if (m) 4234 m_freem(m); 4235 #ifdef DIAGNOSTIC 4236 if (max_linkhdr + tlen > MCLBYTES) 4237 return (ENOBUFS); 4238 #endif 4239 MGETHDR(m, M_DONTWAIT, MT_DATA); 4240 if (m && tlen > MHLEN) { 4241 MCLGET(m, M_DONTWAIT); 4242 if ((m->m_flags & M_EXT) == 0) { 4243 m_freem(m); 4244 m = NULL; 4245 } 4246 } 4247 if (m == NULL) 4248 return (ENOBUFS); 4249 MCLAIM(m, &tcp_tx_mowner); 4250 4251 /* Fixup the mbuf. */ 4252 m->m_data += max_linkhdr; 4253 m->m_len = m->m_pkthdr.len = tlen; 4254 if (sc->sc_tp) { 4255 tp = sc->sc_tp; 4256 if (tp->t_inpcb) 4257 so = tp->t_inpcb->inp_socket; 4258 #ifdef INET6 4259 else if (tp->t_in6pcb) 4260 so = tp->t_in6pcb->in6p_socket; 4261 #endif 4262 else 4263 so = NULL; 4264 } else 4265 so = NULL; 4266 m->m_pkthdr.rcvif = NULL; 4267 memset(mtod(m, u_char *), 0, tlen); 4268 4269 switch (sc->sc_src.sa.sa_family) { 4270 case AF_INET: 4271 ip = mtod(m, struct ip *); 4272 ip->ip_v = 4; 4273 ip->ip_dst = sc->sc_src.sin.sin_addr; 4274 ip->ip_src = sc->sc_dst.sin.sin_addr; 4275 ip->ip_p = IPPROTO_TCP; 4276 th = (struct tcphdr *)(ip + 1); 4277 th->th_dport = sc->sc_src.sin.sin_port; 4278 th->th_sport = sc->sc_dst.sin.sin_port; 4279 break; 4280 #ifdef INET6 4281 case AF_INET6: 4282 ip6 = mtod(m, struct ip6_hdr *); 4283 ip6->ip6_vfc = IPV6_VERSION; 4284 ip6->ip6_dst = sc->sc_src.sin6.sin6_addr; 4285 ip6->ip6_src = sc->sc_dst.sin6.sin6_addr; 4286 ip6->ip6_nxt = IPPROTO_TCP; 4287 /* ip6_plen will be updated in ip6_output() */ 4288 th = (struct tcphdr *)(ip6 + 1); 4289 th->th_dport = sc->sc_src.sin6.sin6_port; 4290 th->th_sport = sc->sc_dst.sin6.sin6_port; 4291 break; 4292 #endif 4293 default: 4294 th = NULL; 4295 } 4296 4297 th->th_seq = htonl(sc->sc_iss); 4298 th->th_ack = htonl(sc->sc_irs + 1); 4299 th->th_off = (sizeof(struct tcphdr) + optlen) >> 2; 4300 th->th_flags = TH_SYN|TH_ACK; 4301 th->th_win = htons(sc->sc_win); 4302 /* th_sum already 0 */ 4303 /* th_urp already 0 */ 4304 4305 /* Tack on the TCP options. */ 4306 optp = (u_int8_t *)(th + 1); 4307 *optp++ = TCPOPT_MAXSEG; 4308 *optp++ = 4; 4309 *optp++ = (sc->sc_ourmaxseg >> 8) & 0xff; 4310 *optp++ = sc->sc_ourmaxseg & 0xff; 4311 4312 if (sc->sc_request_r_scale != 15) { 4313 *((u_int32_t *)optp) = htonl(TCPOPT_NOP << 24 | 4314 TCPOPT_WINDOW << 16 | TCPOLEN_WINDOW << 8 | 4315 sc->sc_request_r_scale); 4316 optp += 4; 4317 } 4318 4319 if (sc->sc_flags & SCF_TIMESTAMP) { 4320 u_int32_t *lp = (u_int32_t *)(optp); 4321 /* Form timestamp option as shown in appendix A of RFC 1323. */ 4322 *lp++ = htonl(TCPOPT_TSTAMP_HDR); 4323 *lp++ = htonl(SYN_CACHE_TIMESTAMP(sc)); 4324 *lp = htonl(sc->sc_timestamp); 4325 optp += TCPOLEN_TSTAMP_APPA; 4326 } 4327 4328 if (sc->sc_flags & SCF_SACK_PERMIT) { 4329 u_int8_t *p = optp; 4330 4331 /* Let the peer know that we will SACK. */ 4332 p[0] = TCPOPT_SACK_PERMITTED; 4333 p[1] = 2; 4334 p[2] = TCPOPT_NOP; 4335 p[3] = TCPOPT_NOP; 4336 optp += 4; 4337 } 4338 4339 /* 4340 * Send ECN SYN-ACK setup packet. 4341 * Routes can be asymetric, so, even if we receive a packet 4342 * with ECE and CWR set, we must not assume no one will block 4343 * the ECE packet we are about to send. 4344 */ 4345 if ((sc->sc_flags & SCF_ECN_PERMIT) && tp && 4346 SEQ_GEQ(tp->snd_nxt, tp->snd_max)) { 4347 th->th_flags |= TH_ECE; 4348 TCP_STATINC(TCP_STAT_ECN_SHS); 4349 4350 /* 4351 * draft-ietf-tcpm-ecnsyn-00.txt 4352 * 4353 * "[...] a TCP node MAY respond to an ECN-setup 4354 * SYN packet by setting ECT in the responding 4355 * ECN-setup SYN/ACK packet, indicating to routers 4356 * that the SYN/ACK packet is ECN-Capable. 4357 * This allows a congested router along the path 4358 * to mark the packet instead of dropping the 4359 * packet as an indication of congestion." 4360 * 4361 * "[...] There can be a great benefit in setting 4362 * an ECN-capable codepoint in SYN/ACK packets [...] 4363 * Congestion is most likely to occur in 4364 * the server-to-client direction. As a result, 4365 * setting an ECN-capable codepoint in SYN/ACK 4366 * packets can reduce the occurence of three-second 4367 * retransmit timeouts resulting from the drop 4368 * of SYN/ACK packets." 4369 * 4370 * Page 4 and 6, January 2006. 4371 */ 4372 4373 switch (sc->sc_src.sa.sa_family) { 4374 #ifdef INET 4375 case AF_INET: 4376 ip->ip_tos |= IPTOS_ECN_ECT0; 4377 break; 4378 #endif 4379 #ifdef INET6 4380 case AF_INET6: 4381 ip6->ip6_flow |= htonl(IPTOS_ECN_ECT0 << 20); 4382 break; 4383 #endif 4384 } 4385 TCP_STATINC(TCP_STAT_ECN_ECT); 4386 } 4387 4388 #ifdef TCP_SIGNATURE 4389 if (sc->sc_flags & SCF_SIGNATURE) { 4390 struct secasvar *sav; 4391 u_int8_t *sigp; 4392 4393 sav = tcp_signature_getsav(m, th); 4394 4395 if (sav == NULL) { 4396 if (m) 4397 m_freem(m); 4398 return (EPERM); 4399 } 4400 4401 *optp++ = TCPOPT_SIGNATURE; 4402 *optp++ = TCPOLEN_SIGNATURE; 4403 sigp = optp; 4404 bzero(optp, TCP_SIGLEN); 4405 optp += TCP_SIGLEN; 4406 *optp++ = TCPOPT_NOP; 4407 *optp++ = TCPOPT_EOL; 4408 4409 (void)tcp_signature(m, th, hlen, sav, sigp); 4410 4411 key_sa_recordxfer(sav, m); 4412 #ifdef FAST_IPSEC 4413 KEY_FREESAV(&sav); 4414 #else 4415 key_freesav(sav); 4416 #endif 4417 } 4418 #endif 4419 4420 /* Compute the packet's checksum. */ 4421 switch (sc->sc_src.sa.sa_family) { 4422 case AF_INET: 4423 ip->ip_len = htons(tlen - hlen); 4424 th->th_sum = 0; 4425 th->th_sum = in4_cksum(m, IPPROTO_TCP, hlen, tlen - hlen); 4426 break; 4427 #ifdef INET6 4428 case AF_INET6: 4429 ip6->ip6_plen = htons(tlen - hlen); 4430 th->th_sum = 0; 4431 th->th_sum = in6_cksum(m, IPPROTO_TCP, hlen, tlen - hlen); 4432 break; 4433 #endif 4434 } 4435 4436 /* 4437 * Fill in some straggling IP bits. Note the stack expects 4438 * ip_len to be in host order, for convenience. 4439 */ 4440 switch (sc->sc_src.sa.sa_family) { 4441 #ifdef INET 4442 case AF_INET: 4443 ip->ip_len = htons(tlen); 4444 ip->ip_ttl = ip_defttl; 4445 /* XXX tos? */ 4446 break; 4447 #endif 4448 #ifdef INET6 4449 case AF_INET6: 4450 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 4451 ip6->ip6_vfc |= IPV6_VERSION; 4452 ip6->ip6_plen = htons(tlen - hlen); 4453 /* ip6_hlim will be initialized afterwards */ 4454 /* XXX flowlabel? */ 4455 break; 4456 #endif 4457 } 4458 4459 /* XXX use IPsec policy on listening socket, on SYN ACK */ 4460 tp = sc->sc_tp; 4461 4462 switch (sc->sc_src.sa.sa_family) { 4463 #ifdef INET 4464 case AF_INET: 4465 error = ip_output(m, sc->sc_ipopts, ro, 4466 (ip_mtudisc ? IP_MTUDISC : 0), 4467 (struct ip_moptions *)NULL, so); 4468 break; 4469 #endif 4470 #ifdef INET6 4471 case AF_INET6: 4472 ip6->ip6_hlim = in6_selecthlim(NULL, 4473 (rt = rtcache_validate(ro)) != NULL ? rt->rt_ifp 4474 : NULL); 4475 4476 error = ip6_output(m, NULL /*XXX*/, ro, 0, NULL, so, NULL); 4477 break; 4478 #endif 4479 default: 4480 error = EAFNOSUPPORT; 4481 break; 4482 } 4483 return (error); 4484 } 4485
Indexes created Mon Sep 22 10:09:38 GMT 2025