tcp_input.c revision 1.270
1 /* $NetBSD: tcp_input.c,v 1.270 2007/08/02 13:06:30 yamt 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 1.270 2007/08/02 13:06:30 yamt 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 177 178 #include <net/if.h> 179 #include <net/route.h> 180 #include <net/if_types.h> 181 182 #include <netinet/in.h> 183 #include <netinet/in_systm.h> 184 #include <netinet/ip.h> 185 #include <netinet/in_pcb.h> 186 #include <netinet/in_var.h> 187 #include <netinet/ip_var.h> 188 #include <netinet/in_offload.h> 189 190 #ifdef INET6 191 #ifndef INET 192 #include <netinet/in.h> 193 #endif 194 #include <netinet/ip6.h> 195 #include <netinet6/ip6_var.h> 196 #include <netinet6/in6_pcb.h> 197 #include <netinet6/ip6_var.h> 198 #include <netinet6/in6_var.h> 199 #include <netinet/icmp6.h> 200 #include <netinet6/nd6.h> 201 #ifdef TCP_SIGNATURE 202 #include <netinet6/scope6_var.h> 203 #endif 204 #endif 205 206 #ifndef INET6 207 /* always need ip6.h for IP6_EXTHDR_GET */ 208 #include <netinet/ip6.h> 209 #endif 210 211 #include <netinet/tcp.h> 212 #include <netinet/tcp_fsm.h> 213 #include <netinet/tcp_seq.h> 214 #include <netinet/tcp_timer.h> 215 #include <netinet/tcp_var.h> 216 #include <netinet/tcpip.h> 217 #include <netinet/tcp_congctl.h> 218 #include <netinet/tcp_debug.h> 219 220 #include <machine/stdarg.h> 221 222 #ifdef IPSEC 223 #include <netinet6/ipsec.h> 224 #include <netkey/key.h> 225 #endif /*IPSEC*/ 226 #ifdef INET6 227 #include "faith.h" 228 #if defined(NFAITH) && NFAITH > 0 229 #include <net/if_faith.h> 230 #endif 231 #endif /* IPSEC */ 232 233 #ifdef FAST_IPSEC 234 #include <netipsec/ipsec.h> 235 #include <netipsec/ipsec_var.h> /* XXX ipsecstat namespace */ 236 #include <netipsec/key.h> 237 #ifdef INET6 238 #include <netipsec/ipsec6.h> 239 #endif 240 #endif /* FAST_IPSEC*/ 241 242 int tcprexmtthresh = 3; 243 int tcp_log_refused; 244 245 int tcp_do_autorcvbuf = 0; 246 int tcp_autorcvbuf_inc = 16 * 1024; 247 int tcp_autorcvbuf_max = 256 * 1024; 248 249 static int tcp_rst_ppslim_count = 0; 250 static struct timeval tcp_rst_ppslim_last; 251 static int tcp_ackdrop_ppslim_count = 0; 252 static struct timeval tcp_ackdrop_ppslim_last; 253 254 #define TCP_PAWS_IDLE (24U * 24 * 60 * 60 * PR_SLOWHZ) 255 256 /* for modulo comparisons of timestamps */ 257 #define TSTMP_LT(a,b) ((int)((a)-(b)) < 0) 258 #define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0) 259 260 /* 261 * Neighbor Discovery, Neighbor Unreachability Detection Upper layer hint. 262 */ 263 #ifdef INET6 264 #define ND6_HINT(tp) \ 265 do { \ 266 if (tp && tp->t_in6pcb && tp->t_family == AF_INET6 && \ 267 tp->t_in6pcb->in6p_route.ro_rt) { \ 268 nd6_nud_hint(tp->t_in6pcb->in6p_route.ro_rt, NULL, 0); \ 269 } \ 270 } while (/*CONSTCOND*/ 0) 271 #else 272 #define ND6_HINT(tp) 273 #endif 274 275 /* 276 * Macro to compute ACK transmission behavior. Delay the ACK unless 277 * we have already delayed an ACK (must send an ACK every two segments). 278 * We also ACK immediately if we received a PUSH and the ACK-on-PUSH 279 * option is enabled. 280 */ 281 #define TCP_SETUP_ACK(tp, th) \ 282 do { \ 283 if ((tp)->t_flags & TF_DELACK || \ 284 (tcp_ack_on_push && (th)->th_flags & TH_PUSH)) \ 285 tp->t_flags |= TF_ACKNOW; \ 286 else \ 287 TCP_SET_DELACK(tp); \ 288 } while (/*CONSTCOND*/ 0) 289 290 #define ICMP_CHECK(tp, th, acked) \ 291 do { \ 292 /* \ 293 * If we had a pending ICMP message that \ 294 * refers to data that have just been \ 295 * acknowledged, disregard the recorded ICMP \ 296 * message. \ 297 */ \ 298 if (((tp)->t_flags & TF_PMTUD_PEND) && \ 299 SEQ_GT((th)->th_ack, (tp)->t_pmtud_th_seq)) \ 300 (tp)->t_flags &= ~TF_PMTUD_PEND; \ 301 \ 302 /* \ 303 * Keep track of the largest chunk of data \ 304 * acknowledged since last PMTU update \ 305 */ \ 306 if ((tp)->t_pmtud_mss_acked < (acked)) \ 307 (tp)->t_pmtud_mss_acked = (acked); \ 308 } while (/*CONSTCOND*/ 0) 309 310 /* 311 * Convert TCP protocol fields to host order for easier processing. 312 */ 313 #define TCP_FIELDS_TO_HOST(th) \ 314 do { \ 315 NTOHL((th)->th_seq); \ 316 NTOHL((th)->th_ack); \ 317 NTOHS((th)->th_win); \ 318 NTOHS((th)->th_urp); \ 319 } while (/*CONSTCOND*/ 0) 320 321 /* 322 * ... and reverse the above. 323 */ 324 #define TCP_FIELDS_TO_NET(th) \ 325 do { \ 326 HTONL((th)->th_seq); \ 327 HTONL((th)->th_ack); \ 328 HTONS((th)->th_win); \ 329 HTONS((th)->th_urp); \ 330 } while (/*CONSTCOND*/ 0) 331 332 #ifdef TCP_CSUM_COUNTERS 333 #include <sys/device.h> 334 335 #if defined(INET) 336 extern struct evcnt tcp_hwcsum_ok; 337 extern struct evcnt tcp_hwcsum_bad; 338 extern struct evcnt tcp_hwcsum_data; 339 extern struct evcnt tcp_swcsum; 340 #endif /* defined(INET) */ 341 #if defined(INET6) 342 extern struct evcnt tcp6_hwcsum_ok; 343 extern struct evcnt tcp6_hwcsum_bad; 344 extern struct evcnt tcp6_hwcsum_data; 345 extern struct evcnt tcp6_swcsum; 346 #endif /* defined(INET6) */ 347 348 #define TCP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 349 350 #else 351 352 #define TCP_CSUM_COUNTER_INCR(ev) /* nothing */ 353 354 #endif /* TCP_CSUM_COUNTERS */ 355 356 #ifdef TCP_REASS_COUNTERS 357 #include <sys/device.h> 358 359 extern struct evcnt tcp_reass_; 360 extern struct evcnt tcp_reass_empty; 361 extern struct evcnt tcp_reass_iteration[8]; 362 extern struct evcnt tcp_reass_prependfirst; 363 extern struct evcnt tcp_reass_prepend; 364 extern struct evcnt tcp_reass_insert; 365 extern struct evcnt tcp_reass_inserttail; 366 extern struct evcnt tcp_reass_append; 367 extern struct evcnt tcp_reass_appendtail; 368 extern struct evcnt tcp_reass_overlaptail; 369 extern struct evcnt tcp_reass_overlapfront; 370 extern struct evcnt tcp_reass_segdup; 371 extern struct evcnt tcp_reass_fragdup; 372 373 #define TCP_REASS_COUNTER_INCR(ev) (ev)->ev_count++ 374 375 #else 376 377 #define TCP_REASS_COUNTER_INCR(ev) /* nothing */ 378 379 #endif /* TCP_REASS_COUNTERS */ 380 381 static int tcp_reass(struct tcpcb *, const struct tcphdr *, struct mbuf *, 382 int *); 383 static int tcp_dooptions(struct tcpcb *, const u_char *, int, 384 struct tcphdr *, struct mbuf *, int, struct tcp_opt_info *); 385 386 #ifdef INET 387 static void tcp4_log_refused(const struct ip *, const struct tcphdr *); 388 #endif 389 #ifdef INET6 390 static void tcp6_log_refused(const struct ip6_hdr *, const struct tcphdr *); 391 #endif 392 393 #define TRAVERSE(x) while ((x)->m_next) (x) = (x)->m_next 394 395 #if defined(MBUFTRACE) 396 struct mowner tcp_reass_mowner = MOWNER_INIT("tcp", "reass"); 397 #endif /* defined(MBUFTRACE) */ 398 399 static POOL_INIT(tcpipqent_pool, sizeof(struct ipqent), 0, 0, 0, "tcpipqepl", 400 NULL, IPL_VM); 401 402 struct ipqent * 403 tcpipqent_alloc() 404 { 405 struct ipqent *ipqe; 406 int s; 407 408 s = splvm(); 409 ipqe = pool_get(&tcpipqent_pool, PR_NOWAIT); 410 splx(s); 411 412 return ipqe; 413 } 414 415 void 416 tcpipqent_free(struct ipqent *ipqe) 417 { 418 int s; 419 420 s = splvm(); 421 pool_put(&tcpipqent_pool, ipqe); 422 splx(s); 423 } 424 425 static int 426 tcp_reass(struct tcpcb *tp, const struct tcphdr *th, struct mbuf *m, int *tlen) 427 { 428 struct ipqent *p, *q, *nq, *tiqe = NULL; 429 struct socket *so = NULL; 430 int pkt_flags; 431 tcp_seq pkt_seq; 432 unsigned pkt_len; 433 u_long rcvpartdupbyte = 0; 434 u_long rcvoobyte; 435 #ifdef TCP_REASS_COUNTERS 436 u_int count = 0; 437 #endif 438 439 if (tp->t_inpcb) 440 so = tp->t_inpcb->inp_socket; 441 #ifdef INET6 442 else if (tp->t_in6pcb) 443 so = tp->t_in6pcb->in6p_socket; 444 #endif 445 446 TCP_REASS_LOCK_CHECK(tp); 447 448 /* 449 * Call with th==0 after become established to 450 * force pre-ESTABLISHED data up to user socket. 451 */ 452 if (th == 0) 453 goto present; 454 455 m_claimm(m, &tcp_reass_mowner); 456 457 rcvoobyte = *tlen; 458 /* 459 * Copy these to local variables because the tcpiphdr 460 * gets munged while we are collapsing mbufs. 461 */ 462 pkt_seq = th->th_seq; 463 pkt_len = *tlen; 464 pkt_flags = th->th_flags; 465 466 TCP_REASS_COUNTER_INCR(&tcp_reass_); 467 468 if ((p = TAILQ_LAST(&tp->segq, ipqehead)) != NULL) { 469 /* 470 * When we miss a packet, the vast majority of time we get 471 * packets that follow it in order. So optimize for that. 472 */ 473 if (pkt_seq == p->ipqe_seq + p->ipqe_len) { 474 p->ipqe_len += pkt_len; 475 p->ipqe_flags |= pkt_flags; 476 m_cat(p->ipre_mlast, m); 477 TRAVERSE(p->ipre_mlast); 478 m = NULL; 479 tiqe = p; 480 TAILQ_REMOVE(&tp->timeq, p, ipqe_timeq); 481 TCP_REASS_COUNTER_INCR(&tcp_reass_appendtail); 482 goto skip_replacement; 483 } 484 /* 485 * While we're here, if the pkt is completely beyond 486 * anything we have, just insert it at the tail. 487 */ 488 if (SEQ_GT(pkt_seq, p->ipqe_seq + p->ipqe_len)) { 489 TCP_REASS_COUNTER_INCR(&tcp_reass_inserttail); 490 goto insert_it; 491 } 492 } 493 494 q = TAILQ_FIRST(&tp->segq); 495 496 if (q != NULL) { 497 /* 498 * If this segment immediately precedes the first out-of-order 499 * block, simply slap the segment in front of it and (mostly) 500 * skip the complicated logic. 501 */ 502 if (pkt_seq + pkt_len == q->ipqe_seq) { 503 q->ipqe_seq = pkt_seq; 504 q->ipqe_len += pkt_len; 505 q->ipqe_flags |= pkt_flags; 506 m_cat(m, q->ipqe_m); 507 q->ipqe_m = m; 508 q->ipre_mlast = m; /* last mbuf may have changed */ 509 TRAVERSE(q->ipre_mlast); 510 tiqe = q; 511 TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); 512 TCP_REASS_COUNTER_INCR(&tcp_reass_prependfirst); 513 goto skip_replacement; 514 } 515 } else { 516 TCP_REASS_COUNTER_INCR(&tcp_reass_empty); 517 } 518 519 /* 520 * Find a segment which begins after this one does. 521 */ 522 for (p = NULL; q != NULL; q = nq) { 523 nq = TAILQ_NEXT(q, ipqe_q); 524 #ifdef TCP_REASS_COUNTERS 525 count++; 526 #endif 527 /* 528 * If the received segment is just right after this 529 * fragment, merge the two together and then check 530 * for further overlaps. 531 */ 532 if (q->ipqe_seq + q->ipqe_len == pkt_seq) { 533 #ifdef TCPREASS_DEBUG 534 printf("tcp_reass[%p]: concat %u:%u(%u) to %u:%u(%u)\n", 535 tp, pkt_seq, pkt_seq + pkt_len, pkt_len, 536 q->ipqe_seq, q->ipqe_seq + q->ipqe_len, q->ipqe_len); 537 #endif 538 pkt_len += q->ipqe_len; 539 pkt_flags |= q->ipqe_flags; 540 pkt_seq = q->ipqe_seq; 541 m_cat(q->ipre_mlast, m); 542 TRAVERSE(q->ipre_mlast); 543 m = q->ipqe_m; 544 TCP_REASS_COUNTER_INCR(&tcp_reass_append); 545 goto free_ipqe; 546 } 547 /* 548 * If the received segment is completely past this 549 * fragment, we need to go the next fragment. 550 */ 551 if (SEQ_LT(q->ipqe_seq + q->ipqe_len, pkt_seq)) { 552 p = q; 553 continue; 554 } 555 /* 556 * If the fragment is past the received segment, 557 * it (or any following) can't be concatenated. 558 */ 559 if (SEQ_GT(q->ipqe_seq, pkt_seq + pkt_len)) { 560 TCP_REASS_COUNTER_INCR(&tcp_reass_insert); 561 break; 562 } 563 564 /* 565 * We've received all the data in this segment before. 566 * mark it as a duplicate and return. 567 */ 568 if (SEQ_LEQ(q->ipqe_seq, pkt_seq) && 569 SEQ_GEQ(q->ipqe_seq + q->ipqe_len, pkt_seq + pkt_len)) { 570 tcpstat.tcps_rcvduppack++; 571 tcpstat.tcps_rcvdupbyte += pkt_len; 572 tcp_new_dsack(tp, pkt_seq, pkt_len); 573 m_freem(m); 574 if (tiqe != NULL) { 575 tcpipqent_free(tiqe); 576 } 577 TCP_REASS_COUNTER_INCR(&tcp_reass_segdup); 578 return (0); 579 } 580 /* 581 * Received segment completely overlaps this fragment 582 * so we drop the fragment (this keeps the temporal 583 * ordering of segments correct). 584 */ 585 if (SEQ_GEQ(q->ipqe_seq, pkt_seq) && 586 SEQ_LEQ(q->ipqe_seq + q->ipqe_len, pkt_seq + pkt_len)) { 587 rcvpartdupbyte += q->ipqe_len; 588 m_freem(q->ipqe_m); 589 TCP_REASS_COUNTER_INCR(&tcp_reass_fragdup); 590 goto free_ipqe; 591 } 592 /* 593 * RX'ed segment extends past the end of the 594 * fragment. Drop the overlapping bytes. Then 595 * merge the fragment and segment then treat as 596 * a longer received packet. 597 */ 598 if (SEQ_LT(q->ipqe_seq, pkt_seq) && 599 SEQ_GT(q->ipqe_seq + q->ipqe_len, pkt_seq)) { 600 int overlap = q->ipqe_seq + q->ipqe_len - pkt_seq; 601 #ifdef TCPREASS_DEBUG 602 printf("tcp_reass[%p]: trim starting %d bytes of %u:%u(%u)\n", 603 tp, overlap, 604 pkt_seq, pkt_seq + pkt_len, pkt_len); 605 #endif 606 m_adj(m, overlap); 607 rcvpartdupbyte += overlap; 608 m_cat(q->ipre_mlast, m); 609 TRAVERSE(q->ipre_mlast); 610 m = q->ipqe_m; 611 pkt_seq = q->ipqe_seq; 612 pkt_len += q->ipqe_len - overlap; 613 rcvoobyte -= overlap; 614 TCP_REASS_COUNTER_INCR(&tcp_reass_overlaptail); 615 goto free_ipqe; 616 } 617 /* 618 * RX'ed segment extends past the front of the 619 * fragment. Drop the overlapping bytes on the 620 * received packet. The packet will then be 621 * contatentated with this fragment a bit later. 622 */ 623 if (SEQ_GT(q->ipqe_seq, pkt_seq) && 624 SEQ_LT(q->ipqe_seq, pkt_seq + pkt_len)) { 625 int overlap = pkt_seq + pkt_len - q->ipqe_seq; 626 #ifdef TCPREASS_DEBUG 627 printf("tcp_reass[%p]: trim trailing %d bytes of %u:%u(%u)\n", 628 tp, overlap, 629 pkt_seq, pkt_seq + pkt_len, pkt_len); 630 #endif 631 m_adj(m, -overlap); 632 pkt_len -= overlap; 633 rcvpartdupbyte += overlap; 634 TCP_REASS_COUNTER_INCR(&tcp_reass_overlapfront); 635 rcvoobyte -= overlap; 636 } 637 /* 638 * If the received segment immediates precedes this 639 * fragment then tack the fragment onto this segment 640 * and reinsert the data. 641 */ 642 if (q->ipqe_seq == pkt_seq + pkt_len) { 643 #ifdef TCPREASS_DEBUG 644 printf("tcp_reass[%p]: append %u:%u(%u) to %u:%u(%u)\n", 645 tp, q->ipqe_seq, q->ipqe_seq + q->ipqe_len, q->ipqe_len, 646 pkt_seq, pkt_seq + pkt_len, pkt_len); 647 #endif 648 pkt_len += q->ipqe_len; 649 pkt_flags |= q->ipqe_flags; 650 m_cat(m, q->ipqe_m); 651 TAILQ_REMOVE(&tp->segq, q, ipqe_q); 652 TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); 653 tp->t_segqlen--; 654 KASSERT(tp->t_segqlen >= 0); 655 KASSERT(tp->t_segqlen != 0 || 656 (TAILQ_EMPTY(&tp->segq) && 657 TAILQ_EMPTY(&tp->timeq))); 658 if (tiqe == NULL) { 659 tiqe = q; 660 } else { 661 tcpipqent_free(q); 662 } 663 TCP_REASS_COUNTER_INCR(&tcp_reass_prepend); 664 break; 665 } 666 /* 667 * If the fragment is before the segment, remember it. 668 * When this loop is terminated, p will contain the 669 * pointer to fragment that is right before the received 670 * segment. 671 */ 672 if (SEQ_LEQ(q->ipqe_seq, pkt_seq)) 673 p = q; 674 675 continue; 676 677 /* 678 * This is a common operation. It also will allow 679 * to save doing a malloc/free in most instances. 680 */ 681 free_ipqe: 682 TAILQ_REMOVE(&tp->segq, q, ipqe_q); 683 TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); 684 tp->t_segqlen--; 685 KASSERT(tp->t_segqlen >= 0); 686 KASSERT(tp->t_segqlen != 0 || 687 (TAILQ_EMPTY(&tp->segq) && TAILQ_EMPTY(&tp->timeq))); 688 if (tiqe == NULL) { 689 tiqe = q; 690 } else { 691 tcpipqent_free(q); 692 } 693 } 694 695 #ifdef TCP_REASS_COUNTERS 696 if (count > 7) 697 TCP_REASS_COUNTER_INCR(&tcp_reass_iteration[0]); 698 else if (count > 0) 699 TCP_REASS_COUNTER_INCR(&tcp_reass_iteration[count]); 700 #endif 701 702 insert_it: 703 704 /* 705 * Allocate a new queue entry since the received segment did not 706 * collapse onto any other out-of-order block; thus we are allocating 707 * a new block. If it had collapsed, tiqe would not be NULL and 708 * we would be reusing it. 709 * XXX If we can't, just drop the packet. XXX 710 */ 711 if (tiqe == NULL) { 712 tiqe = tcpipqent_alloc(); 713 if (tiqe == NULL) { 714 tcpstat.tcps_rcvmemdrop++; 715 m_freem(m); 716 return (0); 717 } 718 } 719 720 /* 721 * Update the counters. 722 */ 723 tcpstat.tcps_rcvoopack++; 724 tcpstat.tcps_rcvoobyte += rcvoobyte; 725 if (rcvpartdupbyte) { 726 tcpstat.tcps_rcvpartduppack++; 727 tcpstat.tcps_rcvpartdupbyte += rcvpartdupbyte; 728 } 729 730 /* 731 * Insert the new fragment queue entry into both queues. 732 */ 733 tiqe->ipqe_m = m; 734 tiqe->ipre_mlast = m; 735 tiqe->ipqe_seq = pkt_seq; 736 tiqe->ipqe_len = pkt_len; 737 tiqe->ipqe_flags = pkt_flags; 738 if (p == NULL) { 739 TAILQ_INSERT_HEAD(&tp->segq, tiqe, ipqe_q); 740 #ifdef TCPREASS_DEBUG 741 if (tiqe->ipqe_seq != tp->rcv_nxt) 742 printf("tcp_reass[%p]: insert %u:%u(%u) at front\n", 743 tp, pkt_seq, pkt_seq + pkt_len, pkt_len); 744 #endif 745 } else { 746 TAILQ_INSERT_AFTER(&tp->segq, p, tiqe, ipqe_q); 747 #ifdef TCPREASS_DEBUG 748 printf("tcp_reass[%p]: insert %u:%u(%u) after %u:%u(%u)\n", 749 tp, pkt_seq, pkt_seq + pkt_len, pkt_len, 750 p->ipqe_seq, p->ipqe_seq + p->ipqe_len, p->ipqe_len); 751 #endif 752 } 753 tp->t_segqlen++; 754 755 skip_replacement: 756 757 TAILQ_INSERT_HEAD(&tp->timeq, tiqe, ipqe_timeq); 758 759 present: 760 /* 761 * Present data to user, advancing rcv_nxt through 762 * completed sequence space. 763 */ 764 if (TCPS_HAVEESTABLISHED(tp->t_state) == 0) 765 return (0); 766 q = TAILQ_FIRST(&tp->segq); 767 if (q == NULL || q->ipqe_seq != tp->rcv_nxt) 768 return (0); 769 if (tp->t_state == TCPS_SYN_RECEIVED && q->ipqe_len) 770 return (0); 771 772 tp->rcv_nxt += q->ipqe_len; 773 pkt_flags = q->ipqe_flags & TH_FIN; 774 ND6_HINT(tp); 775 776 TAILQ_REMOVE(&tp->segq, q, ipqe_q); 777 TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); 778 tp->t_segqlen--; 779 KASSERT(tp->t_segqlen >= 0); 780 KASSERT(tp->t_segqlen != 0 || 781 (TAILQ_EMPTY(&tp->segq) && TAILQ_EMPTY(&tp->timeq))); 782 if (so->so_state & SS_CANTRCVMORE) 783 m_freem(q->ipqe_m); 784 else 785 sbappendstream(&so->so_rcv, q->ipqe_m); 786 tcpipqent_free(q); 787 sorwakeup(so); 788 return (pkt_flags); 789 } 790 791 #ifdef INET6 792 int 793 tcp6_input(struct mbuf **mp, int *offp, int proto) 794 { 795 struct mbuf *m = *mp; 796 797 /* 798 * draft-itojun-ipv6-tcp-to-anycast 799 * better place to put this in? 800 */ 801 if (m->m_flags & M_ANYCAST6) { 802 struct ip6_hdr *ip6; 803 if (m->m_len < sizeof(struct ip6_hdr)) { 804 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 805 tcpstat.tcps_rcvshort++; 806 return IPPROTO_DONE; 807 } 808 } 809 ip6 = mtod(m, struct ip6_hdr *); 810 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR, 811 (char *)&ip6->ip6_dst - (char *)ip6); 812 return IPPROTO_DONE; 813 } 814 815 tcp_input(m, *offp, proto); 816 return IPPROTO_DONE; 817 } 818 #endif 819 820 #ifdef INET 821 static void 822 tcp4_log_refused(const struct ip *ip, const struct tcphdr *th) 823 { 824 char src[4*sizeof "123"]; 825 char dst[4*sizeof "123"]; 826 827 if (ip) { 828 strlcpy(src, inet_ntoa(ip->ip_src), sizeof(src)); 829 strlcpy(dst, inet_ntoa(ip->ip_dst), sizeof(dst)); 830 } 831 else { 832 strlcpy(src, "(unknown)", sizeof(src)); 833 strlcpy(dst, "(unknown)", sizeof(dst)); 834 } 835 log(LOG_INFO, 836 "Connection attempt to TCP %s:%d from %s:%d\n", 837 dst, ntohs(th->th_dport), 838 src, ntohs(th->th_sport)); 839 } 840 #endif 841 842 #ifdef INET6 843 static void 844 tcp6_log_refused(const struct ip6_hdr *ip6, const struct tcphdr *th) 845 { 846 char src[INET6_ADDRSTRLEN]; 847 char dst[INET6_ADDRSTRLEN]; 848 849 if (ip6) { 850 strlcpy(src, ip6_sprintf(&ip6->ip6_src), sizeof(src)); 851 strlcpy(dst, ip6_sprintf(&ip6->ip6_dst), sizeof(dst)); 852 } 853 else { 854 strlcpy(src, "(unknown v6)", sizeof(src)); 855 strlcpy(dst, "(unknown v6)", sizeof(dst)); 856 } 857 log(LOG_INFO, 858 "Connection attempt to TCP [%s]:%d from [%s]:%d\n", 859 dst, ntohs(th->th_dport), 860 src, ntohs(th->th_sport)); 861 } 862 #endif 863 864 /* 865 * Checksum extended TCP header and data. 866 */ 867 int 868 tcp_input_checksum(int af, struct mbuf *m, const struct tcphdr *th, 869 int toff, int off, int tlen) 870 { 871 872 /* 873 * XXX it's better to record and check if this mbuf is 874 * already checked. 875 */ 876 877 switch (af) { 878 #ifdef INET 879 case AF_INET: 880 switch (m->m_pkthdr.csum_flags & 881 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_TCPv4) | 882 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { 883 case M_CSUM_TCPv4|M_CSUM_TCP_UDP_BAD: 884 TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_bad); 885 goto badcsum; 886 887 case M_CSUM_TCPv4|M_CSUM_DATA: { 888 u_int32_t hw_csum = m->m_pkthdr.csum_data; 889 890 TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_data); 891 if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) { 892 const struct ip *ip = 893 mtod(m, const struct ip *); 894 895 hw_csum = in_cksum_phdr(ip->ip_src.s_addr, 896 ip->ip_dst.s_addr, 897 htons(hw_csum + tlen + off + IPPROTO_TCP)); 898 } 899 if ((hw_csum ^ 0xffff) != 0) 900 goto badcsum; 901 break; 902 } 903 904 case M_CSUM_TCPv4: 905 /* Checksum was okay. */ 906 TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_ok); 907 break; 908 909 default: 910 /* 911 * Must compute it ourselves. Maybe skip checksum 912 * on loopback interfaces. 913 */ 914 if (__predict_true(!(m->m_pkthdr.rcvif->if_flags & 915 IFF_LOOPBACK) || 916 tcp_do_loopback_cksum)) { 917 TCP_CSUM_COUNTER_INCR(&tcp_swcsum); 918 if (in4_cksum(m, IPPROTO_TCP, toff, 919 tlen + off) != 0) 920 goto badcsum; 921 } 922 break; 923 } 924 break; 925 #endif /* INET4 */ 926 927 #ifdef INET6 928 case AF_INET6: 929 switch (m->m_pkthdr.csum_flags & 930 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_TCPv6) | 931 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { 932 case M_CSUM_TCPv6|M_CSUM_TCP_UDP_BAD: 933 TCP_CSUM_COUNTER_INCR(&tcp6_hwcsum_bad); 934 goto badcsum; 935 936 #if 0 /* notyet */ 937 case M_CSUM_TCPv6|M_CSUM_DATA: 938 #endif 939 940 case M_CSUM_TCPv6: 941 /* Checksum was okay. */ 942 TCP_CSUM_COUNTER_INCR(&tcp6_hwcsum_ok); 943 break; 944 945 default: 946 /* 947 * Must compute it ourselves. Maybe skip checksum 948 * on loopback interfaces. 949 */ 950 if (__predict_true((m->m_flags & M_LOOP) == 0 || 951 tcp_do_loopback_cksum)) { 952 TCP_CSUM_COUNTER_INCR(&tcp6_swcsum); 953 if (in6_cksum(m, IPPROTO_TCP, toff, 954 tlen + off) != 0) 955 goto badcsum; 956 } 957 } 958 break; 959 #endif /* INET6 */ 960 } 961 962 return 0; 963 964 badcsum: 965 tcpstat.tcps_rcvbadsum++; 966 return -1; 967 } 968 969 /* 970 * TCP input routine, follows pages 65-76 of RFC 793 very closely. 971 */ 972 void 973 tcp_input(struct mbuf *m, ...) 974 { 975 struct tcphdr *th; 976 struct ip *ip; 977 struct inpcb *inp; 978 #ifdef INET6 979 struct ip6_hdr *ip6; 980 struct in6pcb *in6p; 981 #endif 982 u_int8_t *optp = NULL; 983 int optlen = 0; 984 int len, tlen, toff, hdroptlen = 0; 985 struct tcpcb *tp = 0; 986 int tiflags; 987 struct socket *so = NULL; 988 int todrop, dupseg, acked, ourfinisacked, needoutput = 0; 989 #ifdef TCP_DEBUG 990 short ostate = 0; 991 #endif 992 u_long tiwin; 993 struct tcp_opt_info opti; 994 int off, iphlen; 995 va_list ap; 996 int af; /* af on the wire */ 997 struct mbuf *tcp_saveti = NULL; 998 uint32_t ts_rtt; 999 uint8_t iptos; 1000 1001 MCLAIM(m, &tcp_rx_mowner); 1002 va_start(ap, m); 1003 toff = va_arg(ap, int); 1004 (void)va_arg(ap, int); /* ignore value, advance ap */ 1005 va_end(ap); 1006 1007 tcpstat.tcps_rcvtotal++; 1008 1009 bzero(&opti, sizeof(opti)); 1010 opti.ts_present = 0; 1011 opti.maxseg = 0; 1012 1013 /* 1014 * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN. 1015 * 1016 * TCP is, by definition, unicast, so we reject all 1017 * multicast outright. 1018 * 1019 * Note, there are additional src/dst address checks in 1020 * the AF-specific code below. 1021 */ 1022 if (m->m_flags & (M_BCAST|M_MCAST)) { 1023 /* XXX stat */ 1024 goto drop; 1025 } 1026 #ifdef INET6 1027 if (m->m_flags & M_ANYCAST6) { 1028 /* XXX stat */ 1029 goto drop; 1030 } 1031 #endif 1032 1033 /* 1034 * Get IP and TCP header. 1035 * Note: IP leaves IP header in first mbuf. 1036 */ 1037 ip = mtod(m, struct ip *); 1038 #ifdef INET6 1039 ip6 = NULL; 1040 #endif 1041 switch (ip->ip_v) { 1042 #ifdef INET 1043 case 4: 1044 af = AF_INET; 1045 iphlen = sizeof(struct ip); 1046 ip = mtod(m, struct ip *); 1047 IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, 1048 sizeof(struct tcphdr)); 1049 if (th == NULL) { 1050 tcpstat.tcps_rcvshort++; 1051 return; 1052 } 1053 /* We do the checksum after PCB lookup... */ 1054 len = ntohs(ip->ip_len); 1055 tlen = len - toff; 1056 iptos = ip->ip_tos; 1057 break; 1058 #endif 1059 #ifdef INET6 1060 case 6: 1061 ip = NULL; 1062 iphlen = sizeof(struct ip6_hdr); 1063 af = AF_INET6; 1064 ip6 = mtod(m, struct ip6_hdr *); 1065 IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, 1066 sizeof(struct tcphdr)); 1067 if (th == NULL) { 1068 tcpstat.tcps_rcvshort++; 1069 return; 1070 } 1071 1072 /* Be proactive about malicious use of IPv4 mapped address */ 1073 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 1074 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 1075 /* XXX stat */ 1076 goto drop; 1077 } 1078 1079 /* 1080 * Be proactive about unspecified IPv6 address in source. 1081 * As we use all-zero to indicate unbounded/unconnected pcb, 1082 * unspecified IPv6 address can be used to confuse us. 1083 * 1084 * Note that packets with unspecified IPv6 destination is 1085 * already dropped in ip6_input. 1086 */ 1087 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 1088 /* XXX stat */ 1089 goto drop; 1090 } 1091 1092 /* 1093 * Make sure destination address is not multicast. 1094 * Source address checked in ip6_input(). 1095 */ 1096 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 1097 /* XXX stat */ 1098 goto drop; 1099 } 1100 1101 /* We do the checksum after PCB lookup... */ 1102 len = m->m_pkthdr.len; 1103 tlen = len - toff; 1104 iptos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 1105 break; 1106 #endif 1107 default: 1108 m_freem(m); 1109 return; 1110 } 1111 1112 KASSERT(TCP_HDR_ALIGNED_P(th)); 1113 1114 /* 1115 * Check that TCP offset makes sense, 1116 * pull out TCP options and adjust length. XXX 1117 */ 1118 off = th->th_off << 2; 1119 if (off < sizeof (struct tcphdr) || off > tlen) { 1120 tcpstat.tcps_rcvbadoff++; 1121 goto drop; 1122 } 1123 tlen -= off; 1124 1125 /* 1126 * tcp_input() has been modified to use tlen to mean the TCP data 1127 * length throughout the function. Other functions can use 1128 * m->m_pkthdr.len as the basis for calculating the TCP data length. 1129 * rja 1130 */ 1131 1132 if (off > sizeof (struct tcphdr)) { 1133 IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, off); 1134 if (th == NULL) { 1135 tcpstat.tcps_rcvshort++; 1136 return; 1137 } 1138 /* 1139 * NOTE: ip/ip6 will not be affected by m_pulldown() 1140 * (as they're before toff) and we don't need to update those. 1141 */ 1142 KASSERT(TCP_HDR_ALIGNED_P(th)); 1143 optlen = off - sizeof (struct tcphdr); 1144 optp = ((u_int8_t *)th) + sizeof(struct tcphdr); 1145 /* 1146 * Do quick retrieval of timestamp options ("options 1147 * prediction?"). If timestamp is the only option and it's 1148 * formatted as recommended in RFC 1323 appendix A, we 1149 * quickly get the values now and not bother calling 1150 * tcp_dooptions(), etc. 1151 */ 1152 if ((optlen == TCPOLEN_TSTAMP_APPA || 1153 (optlen > TCPOLEN_TSTAMP_APPA && 1154 optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) && 1155 *(u_int32_t *)optp == htonl(TCPOPT_TSTAMP_HDR) && 1156 (th->th_flags & TH_SYN) == 0) { 1157 opti.ts_present = 1; 1158 opti.ts_val = ntohl(*(u_int32_t *)(optp + 4)); 1159 opti.ts_ecr = ntohl(*(u_int32_t *)(optp + 8)); 1160 optp = NULL; /* we've parsed the options */ 1161 } 1162 } 1163 tiflags = th->th_flags; 1164 1165 /* 1166 * Locate pcb for segment. 1167 */ 1168 findpcb: 1169 inp = NULL; 1170 #ifdef INET6 1171 in6p = NULL; 1172 #endif 1173 switch (af) { 1174 #ifdef INET 1175 case AF_INET: 1176 inp = in_pcblookup_connect(&tcbtable, ip->ip_src, th->th_sport, 1177 ip->ip_dst, th->th_dport); 1178 if (inp == 0) { 1179 ++tcpstat.tcps_pcbhashmiss; 1180 inp = in_pcblookup_bind(&tcbtable, ip->ip_dst, th->th_dport); 1181 } 1182 #ifdef INET6 1183 if (inp == 0) { 1184 struct in6_addr s, d; 1185 1186 /* mapped addr case */ 1187 bzero(&s, sizeof(s)); 1188 s.s6_addr16[5] = htons(0xffff); 1189 bcopy(&ip->ip_src, &s.s6_addr32[3], sizeof(ip->ip_src)); 1190 bzero(&d, sizeof(d)); 1191 d.s6_addr16[5] = htons(0xffff); 1192 bcopy(&ip->ip_dst, &d.s6_addr32[3], sizeof(ip->ip_dst)); 1193 in6p = in6_pcblookup_connect(&tcbtable, &s, 1194 th->th_sport, &d, th->th_dport, 0); 1195 if (in6p == 0) { 1196 ++tcpstat.tcps_pcbhashmiss; 1197 in6p = in6_pcblookup_bind(&tcbtable, &d, 1198 th->th_dport, 0); 1199 } 1200 } 1201 #endif 1202 #ifndef INET6 1203 if (inp == 0) 1204 #else 1205 if (inp == 0 && in6p == 0) 1206 #endif 1207 { 1208 ++tcpstat.tcps_noport; 1209 if (tcp_log_refused && 1210 (tiflags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN) { 1211 tcp4_log_refused(ip, th); 1212 } 1213 TCP_FIELDS_TO_HOST(th); 1214 goto dropwithreset_ratelim; 1215 } 1216 #if defined(IPSEC) || defined(FAST_IPSEC) 1217 if (inp && (inp->inp_socket->so_options & SO_ACCEPTCONN) == 0 && 1218 ipsec4_in_reject(m, inp)) { 1219 ipsecstat.in_polvio++; 1220 goto drop; 1221 } 1222 #ifdef INET6 1223 else if (in6p && 1224 (in6p->in6p_socket->so_options & SO_ACCEPTCONN) == 0 && 1225 ipsec6_in_reject_so(m, in6p->in6p_socket)) { 1226 ipsecstat.in_polvio++; 1227 goto drop; 1228 } 1229 #endif 1230 #endif /*IPSEC*/ 1231 break; 1232 #endif /*INET*/ 1233 #ifdef INET6 1234 case AF_INET6: 1235 { 1236 int faith; 1237 1238 #if defined(NFAITH) && NFAITH > 0 1239 faith = faithprefix(&ip6->ip6_dst); 1240 #else 1241 faith = 0; 1242 #endif 1243 in6p = in6_pcblookup_connect(&tcbtable, &ip6->ip6_src, 1244 th->th_sport, &ip6->ip6_dst, th->th_dport, faith); 1245 if (in6p == NULL) { 1246 ++tcpstat.tcps_pcbhashmiss; 1247 in6p = in6_pcblookup_bind(&tcbtable, &ip6->ip6_dst, 1248 th->th_dport, faith); 1249 } 1250 if (in6p == NULL) { 1251 ++tcpstat.tcps_noport; 1252 if (tcp_log_refused && 1253 (tiflags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN) { 1254 tcp6_log_refused(ip6, th); 1255 } 1256 TCP_FIELDS_TO_HOST(th); 1257 goto dropwithreset_ratelim; 1258 } 1259 #if defined(IPSEC) || defined(FAST_IPSEC) 1260 if ((in6p->in6p_socket->so_options & SO_ACCEPTCONN) == 0 && 1261 ipsec6_in_reject(m, in6p)) { 1262 ipsec6stat.in_polvio++; 1263 goto drop; 1264 } 1265 #endif /*IPSEC*/ 1266 break; 1267 } 1268 #endif 1269 } 1270 1271 /* 1272 * If the state is CLOSED (i.e., TCB does not exist) then 1273 * all data in the incoming segment is discarded. 1274 * If the TCB exists but is in CLOSED state, it is embryonic, 1275 * but should either do a listen or a connect soon. 1276 */ 1277 tp = NULL; 1278 so = NULL; 1279 if (inp) { 1280 tp = intotcpcb(inp); 1281 so = inp->inp_socket; 1282 } 1283 #ifdef INET6 1284 else if (in6p) { 1285 tp = in6totcpcb(in6p); 1286 so = in6p->in6p_socket; 1287 } 1288 #endif 1289 if (tp == 0) { 1290 TCP_FIELDS_TO_HOST(th); 1291 goto dropwithreset_ratelim; 1292 } 1293 if (tp->t_state == TCPS_CLOSED) 1294 goto drop; 1295 1296 /* 1297 * Checksum extended TCP header and data. 1298 */ 1299 if (tcp_input_checksum(af, m, th, toff, off, tlen)) 1300 goto badcsum; 1301 1302 TCP_FIELDS_TO_HOST(th); 1303 1304 /* Unscale the window into a 32-bit value. */ 1305 if ((tiflags & TH_SYN) == 0) 1306 tiwin = th->th_win << tp->snd_scale; 1307 else 1308 tiwin = th->th_win; 1309 1310 #ifdef INET6 1311 /* save packet options if user wanted */ 1312 if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS)) { 1313 if (in6p->in6p_options) { 1314 m_freem(in6p->in6p_options); 1315 in6p->in6p_options = 0; 1316 } 1317 KASSERT(ip6 != NULL); 1318 ip6_savecontrol(in6p, &in6p->in6p_options, ip6, m); 1319 } 1320 #endif 1321 1322 if (so->so_options & (SO_DEBUG|SO_ACCEPTCONN)) { 1323 union syn_cache_sa src; 1324 union syn_cache_sa dst; 1325 1326 bzero(&src, sizeof(src)); 1327 bzero(&dst, sizeof(dst)); 1328 switch (af) { 1329 #ifdef INET 1330 case AF_INET: 1331 src.sin.sin_len = sizeof(struct sockaddr_in); 1332 src.sin.sin_family = AF_INET; 1333 src.sin.sin_addr = ip->ip_src; 1334 src.sin.sin_port = th->th_sport; 1335 1336 dst.sin.sin_len = sizeof(struct sockaddr_in); 1337 dst.sin.sin_family = AF_INET; 1338 dst.sin.sin_addr = ip->ip_dst; 1339 dst.sin.sin_port = th->th_dport; 1340 break; 1341 #endif 1342 #ifdef INET6 1343 case AF_INET6: 1344 src.sin6.sin6_len = sizeof(struct sockaddr_in6); 1345 src.sin6.sin6_family = AF_INET6; 1346 src.sin6.sin6_addr = ip6->ip6_src; 1347 src.sin6.sin6_port = th->th_sport; 1348 1349 dst.sin6.sin6_len = sizeof(struct sockaddr_in6); 1350 dst.sin6.sin6_family = AF_INET6; 1351 dst.sin6.sin6_addr = ip6->ip6_dst; 1352 dst.sin6.sin6_port = th->th_dport; 1353 break; 1354 #endif /* INET6 */ 1355 default: 1356 goto badsyn; /*sanity*/ 1357 } 1358 1359 if (so->so_options & SO_DEBUG) { 1360 #ifdef TCP_DEBUG 1361 ostate = tp->t_state; 1362 #endif 1363 1364 tcp_saveti = NULL; 1365 if (iphlen + sizeof(struct tcphdr) > MHLEN) 1366 goto nosave; 1367 1368 if (m->m_len > iphlen && (m->m_flags & M_EXT) == 0) { 1369 tcp_saveti = m_copym(m, 0, iphlen, M_DONTWAIT); 1370 if (!tcp_saveti) 1371 goto nosave; 1372 } else { 1373 MGETHDR(tcp_saveti, M_DONTWAIT, MT_HEADER); 1374 if (!tcp_saveti) 1375 goto nosave; 1376 MCLAIM(m, &tcp_mowner); 1377 tcp_saveti->m_len = iphlen; 1378 m_copydata(m, 0, iphlen, 1379 mtod(tcp_saveti, void *)); 1380 } 1381 1382 if (M_TRAILINGSPACE(tcp_saveti) < sizeof(struct tcphdr)) { 1383 m_freem(tcp_saveti); 1384 tcp_saveti = NULL; 1385 } else { 1386 tcp_saveti->m_len += sizeof(struct tcphdr); 1387 memcpy(mtod(tcp_saveti, char *) + iphlen, th, 1388 sizeof(struct tcphdr)); 1389 } 1390 nosave:; 1391 } 1392 if (so->so_options & SO_ACCEPTCONN) { 1393 if ((tiflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) { 1394 if (tiflags & TH_RST) { 1395 syn_cache_reset(&src.sa, &dst.sa, th); 1396 } else if ((tiflags & (TH_ACK|TH_SYN)) == 1397 (TH_ACK|TH_SYN)) { 1398 /* 1399 * Received a SYN,ACK. This should 1400 * never happen while we are in 1401 * LISTEN. Send an RST. 1402 */ 1403 goto badsyn; 1404 } else if (tiflags & TH_ACK) { 1405 so = syn_cache_get(&src.sa, &dst.sa, 1406 th, toff, tlen, so, m); 1407 if (so == NULL) { 1408 /* 1409 * We don't have a SYN for 1410 * this ACK; send an RST. 1411 */ 1412 goto badsyn; 1413 } else if (so == 1414 (struct socket *)(-1)) { 1415 /* 1416 * We were unable to create 1417 * the connection. If the 1418 * 3-way handshake was 1419 * completed, and RST has 1420 * been sent to the peer. 1421 * Since the mbuf might be 1422 * in use for the reply, 1423 * do not free it. 1424 */ 1425 m = NULL; 1426 } else { 1427 /* 1428 * We have created a 1429 * full-blown connection. 1430 */ 1431 tp = NULL; 1432 inp = NULL; 1433 #ifdef INET6 1434 in6p = NULL; 1435 #endif 1436 switch (so->so_proto->pr_domain->dom_family) { 1437 #ifdef INET 1438 case AF_INET: 1439 inp = sotoinpcb(so); 1440 tp = intotcpcb(inp); 1441 break; 1442 #endif 1443 #ifdef INET6 1444 case AF_INET6: 1445 in6p = sotoin6pcb(so); 1446 tp = in6totcpcb(in6p); 1447 break; 1448 #endif 1449 } 1450 if (tp == NULL) 1451 goto badsyn; /*XXX*/ 1452 tiwin <<= tp->snd_scale; 1453 goto after_listen; 1454 } 1455 } else { 1456 /* 1457 * None of RST, SYN or ACK was set. 1458 * This is an invalid packet for a 1459 * TCB in LISTEN state. Send a RST. 1460 */ 1461 goto badsyn; 1462 } 1463 } else { 1464 /* 1465 * Received a SYN. 1466 * 1467 * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN 1468 */ 1469 if (m->m_flags & (M_BCAST|M_MCAST)) 1470 goto drop; 1471 1472 switch (af) { 1473 #ifdef INET6 1474 case AF_INET6: 1475 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 1476 goto drop; 1477 break; 1478 #endif /* INET6 */ 1479 case AF_INET: 1480 if (IN_MULTICAST(ip->ip_dst.s_addr) || 1481 in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) 1482 goto drop; 1483 break; 1484 } 1485 1486 #ifdef INET6 1487 /* 1488 * If deprecated address is forbidden, we do 1489 * not accept SYN to deprecated interface 1490 * address to prevent any new inbound 1491 * connection from getting established. 1492 * When we do not accept SYN, we send a TCP 1493 * RST, with deprecated source address (instead 1494 * of dropping it). We compromise it as it is 1495 * much better for peer to send a RST, and 1496 * RST will be the final packet for the 1497 * exchange. 1498 * 1499 * If we do not forbid deprecated addresses, we 1500 * accept the SYN packet. RFC2462 does not 1501 * suggest dropping SYN in this case. 1502 * If we decipher RFC2462 5.5.4, it says like 1503 * this: 1504 * 1. use of deprecated addr with existing 1505 * communication is okay - "SHOULD continue 1506 * to be used" 1507 * 2. use of it with new communication: 1508 * (2a) "SHOULD NOT be used if alternate 1509 * address with sufficient scope is 1510 * available" 1511 * (2b) nothing mentioned otherwise. 1512 * Here we fall into (2b) case as we have no 1513 * choice in our source address selection - we 1514 * must obey the peer. 1515 * 1516 * The wording in RFC2462 is confusing, and 1517 * there are multiple description text for 1518 * deprecated address handling - worse, they 1519 * are not exactly the same. I believe 5.5.4 1520 * is the best one, so we follow 5.5.4. 1521 */ 1522 if (af == AF_INET6 && !ip6_use_deprecated) { 1523 struct in6_ifaddr *ia6; 1524 if ((ia6 = in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, 1525 &ip6->ip6_dst)) && 1526 (ia6->ia6_flags & IN6_IFF_DEPRECATED)) { 1527 tp = NULL; 1528 goto dropwithreset; 1529 } 1530 } 1531 #endif 1532 1533 #if defined(IPSEC) || defined(FAST_IPSEC) 1534 switch (af) { 1535 #ifdef INET 1536 case AF_INET: 1537 if (ipsec4_in_reject_so(m, so)) { 1538 ipsecstat.in_polvio++; 1539 tp = NULL; 1540 goto dropwithreset; 1541 } 1542 break; 1543 #endif 1544 #ifdef INET6 1545 case AF_INET6: 1546 if (ipsec6_in_reject_so(m, so)) { 1547 ipsec6stat.in_polvio++; 1548 tp = NULL; 1549 goto dropwithreset; 1550 } 1551 break; 1552 #endif /*INET6*/ 1553 } 1554 #endif /*IPSEC*/ 1555 1556 /* 1557 * LISTEN socket received a SYN 1558 * from itself? This can't possibly 1559 * be valid; drop the packet. 1560 */ 1561 if (th->th_sport == th->th_dport) { 1562 int i; 1563 1564 switch (af) { 1565 #ifdef INET 1566 case AF_INET: 1567 i = in_hosteq(ip->ip_src, ip->ip_dst); 1568 break; 1569 #endif 1570 #ifdef INET6 1571 case AF_INET6: 1572 i = IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &ip6->ip6_dst); 1573 break; 1574 #endif 1575 default: 1576 i = 1; 1577 } 1578 if (i) { 1579 tcpstat.tcps_badsyn++; 1580 goto drop; 1581 } 1582 } 1583 1584 /* 1585 * SYN looks ok; create compressed TCP 1586 * state for it. 1587 */ 1588 if (so->so_qlen <= so->so_qlimit && 1589 syn_cache_add(&src.sa, &dst.sa, th, tlen, 1590 so, m, optp, optlen, &opti)) 1591 m = NULL; 1592 } 1593 goto drop; 1594 } 1595 } 1596 1597 after_listen: 1598 #ifdef DIAGNOSTIC 1599 /* 1600 * Should not happen now that all embryonic connections 1601 * are handled with compressed state. 1602 */ 1603 if (tp->t_state == TCPS_LISTEN) 1604 panic("tcp_input: TCPS_LISTEN"); 1605 #endif 1606 1607 /* 1608 * Segment received on connection. 1609 * Reset idle time and keep-alive timer. 1610 */ 1611 tp->t_rcvtime = tcp_now; 1612 if (TCPS_HAVEESTABLISHED(tp->t_state)) 1613 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle); 1614 1615 /* 1616 * Process options. 1617 */ 1618 #ifdef TCP_SIGNATURE 1619 if (optp || (tp->t_flags & TF_SIGNATURE)) 1620 #else 1621 if (optp) 1622 #endif 1623 if (tcp_dooptions(tp, optp, optlen, th, m, toff, &opti) < 0) 1624 goto drop; 1625 1626 if (TCP_SACK_ENABLED(tp)) { 1627 tcp_del_sackholes(tp, th); 1628 } 1629 1630 if (TCP_ECN_ALLOWED(tp)) { 1631 switch (iptos & IPTOS_ECN_MASK) { 1632 case IPTOS_ECN_CE: 1633 tp->t_flags |= TF_ECN_SND_ECE; 1634 tcpstat.tcps_ecn_ce++; 1635 break; 1636 case IPTOS_ECN_ECT0: 1637 tcpstat.tcps_ecn_ect++; 1638 break; 1639 case IPTOS_ECN_ECT1: 1640 /* XXX: ignore for now -- rpaulo */ 1641 break; 1642 } 1643 1644 if (tiflags & TH_CWR) 1645 tp->t_flags &= ~TF_ECN_SND_ECE; 1646 1647 /* 1648 * Congestion experienced. 1649 * Ignore if we are already trying to recover. 1650 */ 1651 if ((tiflags & TH_ECE) && SEQ_GEQ(tp->snd_una, tp->snd_recover)) 1652 tp->t_congctl->cong_exp(tp); 1653 } 1654 1655 if (opti.ts_present && opti.ts_ecr) { 1656 /* 1657 * Calculate the RTT from the returned time stamp and the 1658 * connection's time base. If the time stamp is later than 1659 * the current time, or is extremely old, fall back to non-1323 1660 * RTT calculation. Since ts_ecr is unsigned, we can test both 1661 * at the same time. 1662 */ 1663 ts_rtt = TCP_TIMESTAMP(tp) - opti.ts_ecr + 1; 1664 if (ts_rtt > TCP_PAWS_IDLE) 1665 ts_rtt = 0; 1666 } else { 1667 ts_rtt = 0; 1668 } 1669 1670 /* 1671 * Header prediction: check for the two common cases 1672 * of a uni-directional data xfer. If the packet has 1673 * no control flags, is in-sequence, the window didn't 1674 * change and we're not retransmitting, it's a 1675 * candidate. If the length is zero and the ack moved 1676 * forward, we're the sender side of the xfer. Just 1677 * free the data acked & wake any higher level process 1678 * that was blocked waiting for space. If the length 1679 * is non-zero and the ack didn't move, we're the 1680 * receiver side. If we're getting packets in-order 1681 * (the reassembly queue is empty), add the data to 1682 * the socket buffer and note that we need a delayed ack. 1683 */ 1684 if (tp->t_state == TCPS_ESTABLISHED && 1685 (tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ECE|TH_CWR|TH_ACK)) 1686 == TH_ACK && 1687 (!opti.ts_present || TSTMP_GEQ(opti.ts_val, tp->ts_recent)) && 1688 th->th_seq == tp->rcv_nxt && 1689 tiwin && tiwin == tp->snd_wnd && 1690 tp->snd_nxt == tp->snd_max) { 1691 1692 /* 1693 * If last ACK falls within this segment's sequence numbers, 1694 * record the timestamp. 1695 * NOTE: 1696 * 1) That the test incorporates suggestions from the latest 1697 * proposal of the tcplw (at) cray.com list (Braden 1993/04/26). 1698 * 2) That updating only on newer timestamps interferes with 1699 * our earlier PAWS tests, so this check should be solely 1700 * predicated on the sequence space of this segment. 1701 * 3) That we modify the segment boundary check to be 1702 * Last.ACK.Sent <= SEG.SEQ + SEG.Len 1703 * instead of RFC1323's 1704 * Last.ACK.Sent < SEG.SEQ + SEG.Len, 1705 * This modified check allows us to overcome RFC1323's 1706 * limitations as described in Stevens TCP/IP Illustrated 1707 * Vol. 2 p.869. In such cases, we can still calculate the 1708 * RTT correctly when RCV.NXT == Last.ACK.Sent. 1709 */ 1710 if (opti.ts_present && 1711 SEQ_LEQ(th->th_seq, tp->last_ack_sent) && 1712 SEQ_LEQ(tp->last_ack_sent, th->th_seq + tlen + 1713 ((tiflags & (TH_SYN|TH_FIN)) != 0))) { 1714 tp->ts_recent_age = tcp_now; 1715 tp->ts_recent = opti.ts_val; 1716 } 1717 1718 if (tlen == 0) { 1719 /* Ack prediction. */ 1720 if (SEQ_GT(th->th_ack, tp->snd_una) && 1721 SEQ_LEQ(th->th_ack, tp->snd_max) && 1722 tp->snd_cwnd >= tp->snd_wnd && 1723 tp->t_partialacks < 0) { 1724 /* 1725 * this is a pure ack for outstanding data. 1726 */ 1727 ++tcpstat.tcps_predack; 1728 if (ts_rtt) 1729 tcp_xmit_timer(tp, ts_rtt); 1730 else if (tp->t_rtttime && 1731 SEQ_GT(th->th_ack, tp->t_rtseq)) 1732 tcp_xmit_timer(tp, 1733 tcp_now - tp->t_rtttime); 1734 acked = th->th_ack - tp->snd_una; 1735 tcpstat.tcps_rcvackpack++; 1736 tcpstat.tcps_rcvackbyte += acked; 1737 ND6_HINT(tp); 1738 1739 if (acked > (tp->t_lastoff - tp->t_inoff)) 1740 tp->t_lastm = NULL; 1741 sbdrop(&so->so_snd, acked); 1742 tp->t_lastoff -= acked; 1743 1744 ICMP_CHECK(tp, th, acked); 1745 1746 tp->snd_una = th->th_ack; 1747 tp->snd_fack = tp->snd_una; 1748 if (SEQ_LT(tp->snd_high, tp->snd_una)) 1749 tp->snd_high = tp->snd_una; 1750 m_freem(m); 1751 1752 /* 1753 * If all outstanding data are acked, stop 1754 * retransmit timer, otherwise restart timer 1755 * using current (possibly backed-off) value. 1756 * If process is waiting for space, 1757 * wakeup/selwakeup/signal. If data 1758 * are ready to send, let tcp_output 1759 * decide between more output or persist. 1760 */ 1761 if (tp->snd_una == tp->snd_max) 1762 TCP_TIMER_DISARM(tp, TCPT_REXMT); 1763 else if (TCP_TIMER_ISARMED(tp, 1764 TCPT_PERSIST) == 0) 1765 TCP_TIMER_ARM(tp, TCPT_REXMT, 1766 tp->t_rxtcur); 1767 1768 sowwakeup(so); 1769 if (so->so_snd.sb_cc) 1770 (void) tcp_output(tp); 1771 if (tcp_saveti) 1772 m_freem(tcp_saveti); 1773 return; 1774 } 1775 } else if (th->th_ack == tp->snd_una && 1776 TAILQ_FIRST(&tp->segq) == NULL && 1777 tlen <= sbspace(&so->so_rcv)) { 1778 int newsize = 0; /* automatic sockbuf scaling */ 1779 1780 /* 1781 * this is a pure, in-sequence data packet 1782 * with nothing on the reassembly queue and 1783 * we have enough buffer space to take it. 1784 */ 1785 ++tcpstat.tcps_preddat; 1786 tp->rcv_nxt += tlen; 1787 tcpstat.tcps_rcvpack++; 1788 tcpstat.tcps_rcvbyte += tlen; 1789 ND6_HINT(tp); 1790 1791 /* 1792 * Automatic sizing enables the performance of large buffers 1793 * and most of the efficiency of small ones by only allocating 1794 * space when it is needed. 1795 * 1796 * On the receive side the socket buffer memory is only rarely 1797 * used to any significant extent. This allows us to be much 1798 * more aggressive in scaling the receive socket buffer. For 1799 * the case that the buffer space is actually used to a large 1800 * extent and we run out of kernel memory we can simply drop 1801 * the new segments; TCP on the sender will just retransmit it 1802 * later. Setting the buffer size too big may only consume too 1803 * much kernel memory if the application doesn't read() from 1804 * the socket or packet loss or reordering makes use of the 1805 * reassembly queue. 1806 * 1807 * The criteria to step up the receive buffer one notch are: 1808 * 1. the number of bytes received during the time it takes 1809 * one timestamp to be reflected back to us (the RTT); 1810 * 2. received bytes per RTT is within seven eighth of the 1811 * current socket buffer size; 1812 * 3. receive buffer size has not hit maximal automatic size; 1813 * 1814 * This algorithm does one step per RTT at most and only if 1815 * we receive a bulk stream w/o packet losses or reorderings. 1816 * Shrinking the buffer during idle times is not necessary as 1817 * it doesn't consume any memory when idle. 1818 * 1819 * TODO: Only step up if the application is actually serving 1820 * the buffer to better manage the socket buffer resources. 1821 */ 1822 if (tcp_do_autorcvbuf && 1823 opti.ts_ecr && 1824 (so->so_rcv.sb_flags & SB_AUTOSIZE)) { 1825 if (opti.ts_ecr > tp->rfbuf_ts && 1826 opti.ts_ecr - tp->rfbuf_ts < PR_SLOWHZ) { 1827 if (tp->rfbuf_cnt > 1828 (so->so_rcv.sb_hiwat / 8 * 7) && 1829 so->so_rcv.sb_hiwat < 1830 tcp_autorcvbuf_max) { 1831 newsize = 1832 min(so->so_rcv.sb_hiwat + 1833 tcp_autorcvbuf_inc, 1834 tcp_autorcvbuf_max); 1835 } 1836 /* Start over with next RTT. */ 1837 tp->rfbuf_ts = 0; 1838 tp->rfbuf_cnt = 0; 1839 } else 1840 tp->rfbuf_cnt += tlen; /* add up */ 1841 } 1842 1843 /* 1844 * Drop TCP, IP headers and TCP options then add data 1845 * to socket buffer. 1846 */ 1847 if (so->so_state & SS_CANTRCVMORE) 1848 m_freem(m); 1849 else { 1850 /* 1851 * Set new socket buffer size. 1852 * Give up when limit is reached. 1853 */ 1854 if (newsize) 1855 if (!sbreserve(&so->so_rcv, 1856 newsize, so)) 1857 so->so_rcv.sb_flags &= ~SB_AUTOSIZE; 1858 m_adj(m, toff + off); 1859 sbappendstream(&so->so_rcv, m); 1860 } 1861 sorwakeup(so); 1862 TCP_SETUP_ACK(tp, th); 1863 if (tp->t_flags & TF_ACKNOW) 1864 (void) tcp_output(tp); 1865 if (tcp_saveti) 1866 m_freem(tcp_saveti); 1867 return; 1868 } 1869 } 1870 1871 /* 1872 * Compute mbuf offset to TCP data segment. 1873 */ 1874 hdroptlen = toff + off; 1875 1876 /* 1877 * Calculate amount of space in receive window, 1878 * and then do TCP input processing. 1879 * Receive window is amount of space in rcv queue, 1880 * but not less than advertised window. 1881 */ 1882 { int win; 1883 1884 win = sbspace(&so->so_rcv); 1885 if (win < 0) 1886 win = 0; 1887 tp->rcv_wnd = imax(win, (int)(tp->rcv_adv - tp->rcv_nxt)); 1888 } 1889 1890 /* Reset receive buffer auto scaling when not in bulk receive mode. */ 1891 tp->rfbuf_ts = 0; 1892 tp->rfbuf_cnt = 0; 1893 1894 switch (tp->t_state) { 1895 /* 1896 * If the state is SYN_SENT: 1897 * if seg contains an ACK, but not for our SYN, drop the input. 1898 * if seg contains a RST, then drop the connection. 1899 * if seg does not contain SYN, then drop it. 1900 * Otherwise this is an acceptable SYN segment 1901 * initialize tp->rcv_nxt and tp->irs 1902 * if seg contains ack then advance tp->snd_una 1903 * if seg contains a ECE and ECN support is enabled, the stream 1904 * is ECN capable. 1905 * if SYN has been acked change to ESTABLISHED else SYN_RCVD state 1906 * arrange for segment to be acked (eventually) 1907 * continue processing rest of data/controls, beginning with URG 1908 */ 1909 case TCPS_SYN_SENT: 1910 if ((tiflags & TH_ACK) && 1911 (SEQ_LEQ(th->th_ack, tp->iss) || 1912 SEQ_GT(th->th_ack, tp->snd_max))) 1913 goto dropwithreset; 1914 if (tiflags & TH_RST) { 1915 if (tiflags & TH_ACK) 1916 tp = tcp_drop(tp, ECONNREFUSED); 1917 goto drop; 1918 } 1919 if ((tiflags & TH_SYN) == 0) 1920 goto drop; 1921 if (tiflags & TH_ACK) { 1922 tp->snd_una = th->th_ack; 1923 if (SEQ_LT(tp->snd_nxt, tp->snd_una)) 1924 tp->snd_nxt = tp->snd_una; 1925 if (SEQ_LT(tp->snd_high, tp->snd_una)) 1926 tp->snd_high = tp->snd_una; 1927 TCP_TIMER_DISARM(tp, TCPT_REXMT); 1928 1929 if ((tiflags & TH_ECE) && tcp_do_ecn) { 1930 tp->t_flags |= TF_ECN_PERMIT; 1931 tcpstat.tcps_ecn_shs++; 1932 } 1933 1934 } 1935 tp->irs = th->th_seq; 1936 tcp_rcvseqinit(tp); 1937 tp->t_flags |= TF_ACKNOW; 1938 tcp_mss_from_peer(tp, opti.maxseg); 1939 1940 /* 1941 * Initialize the initial congestion window. If we 1942 * had to retransmit the SYN, we must initialize cwnd 1943 * to 1 segment (i.e. the Loss Window). 1944 */ 1945 if (tp->t_flags & TF_SYN_REXMT) 1946 tp->snd_cwnd = tp->t_peermss; 1947 else { 1948 int ss = tcp_init_win; 1949 #ifdef INET 1950 if (inp != NULL && in_localaddr(inp->inp_faddr)) 1951 ss = tcp_init_win_local; 1952 #endif 1953 #ifdef INET6 1954 if (in6p != NULL && in6_localaddr(&in6p->in6p_faddr)) 1955 ss = tcp_init_win_local; 1956 #endif 1957 tp->snd_cwnd = TCP_INITIAL_WINDOW(ss, tp->t_peermss); 1958 } 1959 1960 tcp_rmx_rtt(tp); 1961 if (tiflags & TH_ACK) { 1962 tcpstat.tcps_connects++; 1963 soisconnected(so); 1964 tcp_established(tp); 1965 /* Do window scaling on this connection? */ 1966 if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == 1967 (TF_RCVD_SCALE|TF_REQ_SCALE)) { 1968 tp->snd_scale = tp->requested_s_scale; 1969 tp->rcv_scale = tp->request_r_scale; 1970 } 1971 TCP_REASS_LOCK(tp); 1972 (void) tcp_reass(tp, NULL, (struct mbuf *)0, &tlen); 1973 TCP_REASS_UNLOCK(tp); 1974 /* 1975 * if we didn't have to retransmit the SYN, 1976 * use its rtt as our initial srtt & rtt var. 1977 */ 1978 if (tp->t_rtttime) 1979 tcp_xmit_timer(tp, tcp_now - tp->t_rtttime); 1980 } else 1981 tp->t_state = TCPS_SYN_RECEIVED; 1982 1983 /* 1984 * Advance th->th_seq to correspond to first data byte. 1985 * If data, trim to stay within window, 1986 * dropping FIN if necessary. 1987 */ 1988 th->th_seq++; 1989 if (tlen > tp->rcv_wnd) { 1990 todrop = tlen - tp->rcv_wnd; 1991 m_adj(m, -todrop); 1992 tlen = tp->rcv_wnd; 1993 tiflags &= ~TH_FIN; 1994 tcpstat.tcps_rcvpackafterwin++; 1995 tcpstat.tcps_rcvbyteafterwin += todrop; 1996 } 1997 tp->snd_wl1 = th->th_seq - 1; 1998 tp->rcv_up = th->th_seq; 1999 goto step6; 2000 2001 /* 2002 * If the state is SYN_RECEIVED: 2003 * If seg contains an ACK, but not for our SYN, drop the input 2004 * and generate an RST. See page 36, rfc793 2005 */ 2006 case TCPS_SYN_RECEIVED: 2007 if ((tiflags & TH_ACK) && 2008 (SEQ_LEQ(th->th_ack, tp->iss) || 2009 SEQ_GT(th->th_ack, tp->snd_max))) 2010 goto dropwithreset; 2011 break; 2012 } 2013 2014 /* 2015 * States other than LISTEN or SYN_SENT. 2016 * First check timestamp, if present. 2017 * Then check that at least some bytes of segment are within 2018 * receive window. If segment begins before rcv_nxt, 2019 * drop leading data (and SYN); if nothing left, just ack. 2020 * 2021 * RFC 1323 PAWS: If we have a timestamp reply on this segment 2022 * and it's less than ts_recent, drop it. 2023 */ 2024 if (opti.ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent && 2025 TSTMP_LT(opti.ts_val, tp->ts_recent)) { 2026 2027 /* Check to see if ts_recent is over 24 days old. */ 2028 if (tcp_now - tp->ts_recent_age > TCP_PAWS_IDLE) { 2029 /* 2030 * Invalidate ts_recent. If this segment updates 2031 * ts_recent, the age will be reset later and ts_recent 2032 * will get a valid value. If it does not, setting 2033 * ts_recent to zero will at least satisfy the 2034 * requirement that zero be placed in the timestamp 2035 * echo reply when ts_recent isn't valid. The 2036 * age isn't reset until we get a valid ts_recent 2037 * because we don't want out-of-order segments to be 2038 * dropped when ts_recent is old. 2039 */ 2040 tp->ts_recent = 0; 2041 } else { 2042 tcpstat.tcps_rcvduppack++; 2043 tcpstat.tcps_rcvdupbyte += tlen; 2044 tcpstat.tcps_pawsdrop++; 2045 tcp_new_dsack(tp, th->th_seq, tlen); 2046 goto dropafterack; 2047 } 2048 } 2049 2050 todrop = tp->rcv_nxt - th->th_seq; 2051 dupseg = false; 2052 if (todrop > 0) { 2053 if (tiflags & TH_SYN) { 2054 tiflags &= ~TH_SYN; 2055 th->th_seq++; 2056 if (th->th_urp > 1) 2057 th->th_urp--; 2058 else { 2059 tiflags &= ~TH_URG; 2060 th->th_urp = 0; 2061 } 2062 todrop--; 2063 } 2064 if (todrop > tlen || 2065 (todrop == tlen && (tiflags & TH_FIN) == 0)) { 2066 /* 2067 * Any valid FIN or RST must be to the left of the 2068 * window. At this point the FIN or RST must be a 2069 * duplicate or out of sequence; drop it. 2070 */ 2071 if (tiflags & TH_RST) 2072 goto drop; 2073 tiflags &= ~(TH_FIN|TH_RST); 2074 /* 2075 * Send an ACK to resynchronize and drop any data. 2076 * But keep on processing for RST or ACK. 2077 */ 2078 tp->t_flags |= TF_ACKNOW; 2079 todrop = tlen; 2080 dupseg = true; 2081 tcpstat.tcps_rcvdupbyte += todrop; 2082 tcpstat.tcps_rcvduppack++; 2083 } else if ((tiflags & TH_RST) && 2084 th->th_seq != tp->last_ack_sent) { 2085 /* 2086 * Test for reset before adjusting the sequence 2087 * number for overlapping data. 2088 */ 2089 goto dropafterack_ratelim; 2090 } else { 2091 tcpstat.tcps_rcvpartduppack++; 2092 tcpstat.tcps_rcvpartdupbyte += todrop; 2093 } 2094 tcp_new_dsack(tp, th->th_seq, todrop); 2095 hdroptlen += todrop; /*drop from head afterwards*/ 2096 th->th_seq += todrop; 2097 tlen -= todrop; 2098 if (th->th_urp > todrop) 2099 th->th_urp -= todrop; 2100 else { 2101 tiflags &= ~TH_URG; 2102 th->th_urp = 0; 2103 } 2104 } 2105 2106 /* 2107 * If new data are received on a connection after the 2108 * user processes are gone, then RST the other end. 2109 */ 2110 if ((so->so_state & SS_NOFDREF) && 2111 tp->t_state > TCPS_CLOSE_WAIT && tlen) { 2112 tp = tcp_close(tp); 2113 tcpstat.tcps_rcvafterclose++; 2114 goto dropwithreset; 2115 } 2116 2117 /* 2118 * If segment ends after window, drop trailing data 2119 * (and PUSH and FIN); if nothing left, just ACK. 2120 */ 2121 todrop = (th->th_seq + tlen) - (tp->rcv_nxt+tp->rcv_wnd); 2122 if (todrop > 0) { 2123 tcpstat.tcps_rcvpackafterwin++; 2124 if (todrop >= tlen) { 2125 /* 2126 * The segment actually starts after the window. 2127 * th->th_seq + tlen - tp->rcv_nxt - tp->rcv_wnd >= tlen 2128 * th->th_seq - tp->rcv_nxt - tp->rcv_wnd >= 0 2129 * th->th_seq >= tp->rcv_nxt + tp->rcv_wnd 2130 */ 2131 tcpstat.tcps_rcvbyteafterwin += tlen; 2132 /* 2133 * If a new connection request is received 2134 * while in TIME_WAIT, drop the old connection 2135 * and start over if the sequence numbers 2136 * are above the previous ones. 2137 * 2138 * NOTE: We will checksum the packet again, and 2139 * so we need to put the header fields back into 2140 * network order! 2141 * XXX This kind of sucks, but we don't expect 2142 * XXX this to happen very often, so maybe it 2143 * XXX doesn't matter so much. 2144 */ 2145 if (tiflags & TH_SYN && 2146 tp->t_state == TCPS_TIME_WAIT && 2147 SEQ_GT(th->th_seq, tp->rcv_nxt)) { 2148 tp = tcp_close(tp); 2149 TCP_FIELDS_TO_NET(th); 2150 goto findpcb; 2151 } 2152 /* 2153 * If window is closed can only take segments at 2154 * window edge, and have to drop data and PUSH from 2155 * incoming segments. Continue processing, but 2156 * remember to ack. Otherwise, drop segment 2157 * and (if not RST) ack. 2158 */ 2159 if (tp->rcv_wnd == 0 && th->th_seq == tp->rcv_nxt) { 2160 tp->t_flags |= TF_ACKNOW; 2161 tcpstat.tcps_rcvwinprobe++; 2162 } else 2163 goto dropafterack; 2164 } else 2165 tcpstat.tcps_rcvbyteafterwin += todrop; 2166 m_adj(m, -todrop); 2167 tlen -= todrop; 2168 tiflags &= ~(TH_PUSH|TH_FIN); 2169 } 2170 2171 /* 2172 * If last ACK falls within this segment's sequence numbers, 2173 * and the timestamp is newer, record it. 2174 */ 2175 if (opti.ts_present && TSTMP_GEQ(opti.ts_val, tp->ts_recent) && 2176 SEQ_LEQ(th->th_seq, tp->last_ack_sent) && 2177 SEQ_LT(tp->last_ack_sent, th->th_seq + tlen + 2178 ((tiflags & (TH_SYN|TH_FIN)) != 0))) { 2179 tp->ts_recent_age = tcp_now; 2180 tp->ts_recent = opti.ts_val; 2181 } 2182 2183 /* 2184 * If the RST bit is set examine the state: 2185 * SYN_RECEIVED STATE: 2186 * If passive open, return to LISTEN state. 2187 * If active open, inform user that connection was refused. 2188 * ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES: 2189 * Inform user that connection was reset, and close tcb. 2190 * CLOSING, LAST_ACK, TIME_WAIT STATES 2191 * Close the tcb. 2192 */ 2193 if (tiflags & TH_RST) { 2194 if (th->th_seq != tp->last_ack_sent) 2195 goto dropafterack_ratelim; 2196 2197 switch (tp->t_state) { 2198 case TCPS_SYN_RECEIVED: 2199 so->so_error = ECONNREFUSED; 2200 goto close; 2201 2202 case TCPS_ESTABLISHED: 2203 case TCPS_FIN_WAIT_1: 2204 case TCPS_FIN_WAIT_2: 2205 case TCPS_CLOSE_WAIT: 2206 so->so_error = ECONNRESET; 2207 close: 2208 tp->t_state = TCPS_CLOSED; 2209 tcpstat.tcps_drops++; 2210 tp = tcp_close(tp); 2211 goto drop; 2212 2213 case TCPS_CLOSING: 2214 case TCPS_LAST_ACK: 2215 case TCPS_TIME_WAIT: 2216 tp = tcp_close(tp); 2217 goto drop; 2218 } 2219 } 2220 2221 /* 2222 * Since we've covered the SYN-SENT and SYN-RECEIVED states above 2223 * we must be in a synchronized state. RFC791 states (under RST 2224 * generation) that any unacceptable segment (an out-of-order SYN 2225 * qualifies) received in a synchronized state must elicit only an 2226 * empty acknowledgment segment ... and the connection remains in 2227 * the same state. 2228 */ 2229 if (tiflags & TH_SYN) { 2230 if (tp->rcv_nxt == th->th_seq) { 2231 tcp_respond(tp, m, m, th, (tcp_seq)0, th->th_ack - 1, 2232 TH_ACK); 2233 if (tcp_saveti) 2234 m_freem(tcp_saveti); 2235 return; 2236 } 2237 2238 goto dropafterack_ratelim; 2239 } 2240 2241 /* 2242 * If the ACK bit is off we drop the segment and return. 2243 */ 2244 if ((tiflags & TH_ACK) == 0) { 2245 if (tp->t_flags & TF_ACKNOW) 2246 goto dropafterack; 2247 else 2248 goto drop; 2249 } 2250 2251 /* 2252 * Ack processing. 2253 */ 2254 switch (tp->t_state) { 2255 2256 /* 2257 * In SYN_RECEIVED state if the ack ACKs our SYN then enter 2258 * ESTABLISHED state and continue processing, otherwise 2259 * send an RST. 2260 */ 2261 case TCPS_SYN_RECEIVED: 2262 if (SEQ_GT(tp->snd_una, th->th_ack) || 2263 SEQ_GT(th->th_ack, tp->snd_max)) 2264 goto dropwithreset; 2265 tcpstat.tcps_connects++; 2266 soisconnected(so); 2267 tcp_established(tp); 2268 /* Do window scaling? */ 2269 if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == 2270 (TF_RCVD_SCALE|TF_REQ_SCALE)) { 2271 tp->snd_scale = tp->requested_s_scale; 2272 tp->rcv_scale = tp->request_r_scale; 2273 } 2274 TCP_REASS_LOCK(tp); 2275 (void) tcp_reass(tp, NULL, (struct mbuf *)0, &tlen); 2276 TCP_REASS_UNLOCK(tp); 2277 tp->snd_wl1 = th->th_seq - 1; 2278 /* fall into ... */ 2279 2280 /* 2281 * In ESTABLISHED state: drop duplicate ACKs; ACK out of range 2282 * ACKs. If the ack is in the range 2283 * tp->snd_una < th->th_ack <= tp->snd_max 2284 * then advance tp->snd_una to th->th_ack and drop 2285 * data from the retransmission queue. If this ACK reflects 2286 * more up to date window information we update our window information. 2287 */ 2288 case TCPS_ESTABLISHED: 2289 case TCPS_FIN_WAIT_1: 2290 case TCPS_FIN_WAIT_2: 2291 case TCPS_CLOSE_WAIT: 2292 case TCPS_CLOSING: 2293 case TCPS_LAST_ACK: 2294 case TCPS_TIME_WAIT: 2295 2296 if (SEQ_LEQ(th->th_ack, tp->snd_una)) { 2297 if (tlen == 0 && !dupseg && tiwin == tp->snd_wnd) { 2298 tcpstat.tcps_rcvdupack++; 2299 /* 2300 * If we have outstanding data (other than 2301 * a window probe), this is a completely 2302 * duplicate ack (ie, window info didn't 2303 * change), the ack is the biggest we've 2304 * seen and we've seen exactly our rexmt 2305 * threshhold of them, assume a packet 2306 * has been dropped and retransmit it. 2307 * Kludge snd_nxt & the congestion 2308 * window so we send only this one 2309 * packet. 2310 */ 2311 if (TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 || 2312 th->th_ack != tp->snd_una) 2313 tp->t_dupacks = 0; 2314 else if (tp->t_partialacks < 0 && 2315 ((!TCP_SACK_ENABLED(tp) && 2316 ++tp->t_dupacks == tcprexmtthresh) || 2317 TCP_FACK_FASTRECOV(tp))) { 2318 /* 2319 * Do the fast retransmit, and adjust 2320 * congestion control paramenters. 2321 */ 2322 if (tp->t_congctl->fast_retransmit(tp, th)) { 2323 /* False fast retransmit */ 2324 break; 2325 } else 2326 goto drop; 2327 } else if (tp->t_dupacks > tcprexmtthresh) { 2328 tp->snd_cwnd += tp->t_segsz; 2329 (void) tcp_output(tp); 2330 goto drop; 2331 } 2332 } else { 2333 /* 2334 * If the ack appears to be very old, only 2335 * allow data that is in-sequence. This 2336 * makes it somewhat more difficult to insert 2337 * forged data by guessing sequence numbers. 2338 * Sent an ack to try to update the send 2339 * sequence number on the other side. 2340 */ 2341 if (tlen && th->th_seq != tp->rcv_nxt && 2342 SEQ_LT(th->th_ack, 2343 tp->snd_una - tp->max_sndwnd)) 2344 goto dropafterack; 2345 } 2346 break; 2347 } 2348 /* 2349 * If the congestion window was inflated to account 2350 * for the other side's cached packets, retract it. 2351 */ 2352 /* XXX: make SACK have his own congestion control 2353 * struct -- rpaulo */ 2354 if (TCP_SACK_ENABLED(tp)) 2355 tcp_sack_newack(tp, th); 2356 else 2357 tp->t_congctl->fast_retransmit_newack(tp, th); 2358 if (SEQ_GT(th->th_ack, tp->snd_max)) { 2359 tcpstat.tcps_rcvacktoomuch++; 2360 goto dropafterack; 2361 } 2362 acked = th->th_ack - tp->snd_una; 2363 tcpstat.tcps_rcvackpack++; 2364 tcpstat.tcps_rcvackbyte += acked; 2365 2366 /* 2367 * If we have a timestamp reply, update smoothed 2368 * round trip time. If no timestamp is present but 2369 * transmit timer is running and timed sequence 2370 * number was acked, update smoothed round trip time. 2371 * Since we now have an rtt measurement, cancel the 2372 * timer backoff (cf., Phil Karn's retransmit alg.). 2373 * Recompute the initial retransmit timer. 2374 */ 2375 if (ts_rtt) 2376 tcp_xmit_timer(tp, ts_rtt); 2377 else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq)) 2378 tcp_xmit_timer(tp, tcp_now - tp->t_rtttime); 2379 2380 /* 2381 * If all outstanding data is acked, stop retransmit 2382 * timer and remember to restart (more output or persist). 2383 * If there is more data to be acked, restart retransmit 2384 * timer, using current (possibly backed-off) value. 2385 */ 2386 if (th->th_ack == tp->snd_max) { 2387 TCP_TIMER_DISARM(tp, TCPT_REXMT); 2388 needoutput = 1; 2389 } else if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) 2390 TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur); 2391 2392 /* 2393 * New data has been acked, adjust the congestion window. 2394 */ 2395 tp->t_congctl->newack(tp, th); 2396 2397 ND6_HINT(tp); 2398 if (acked > so->so_snd.sb_cc) { 2399 tp->snd_wnd -= so->so_snd.sb_cc; 2400 sbdrop(&so->so_snd, (int)so->so_snd.sb_cc); 2401 ourfinisacked = 1; 2402 } else { 2403 if (acked > (tp->t_lastoff - tp->t_inoff)) 2404 tp->t_lastm = NULL; 2405 sbdrop(&so->so_snd, acked); 2406 tp->t_lastoff -= acked; 2407 tp->snd_wnd -= acked; 2408 ourfinisacked = 0; 2409 } 2410 sowwakeup(so); 2411 2412 ICMP_CHECK(tp, th, acked); 2413 2414 tp->snd_una = th->th_ack; 2415 if (SEQ_GT(tp->snd_una, tp->snd_fack)) 2416 tp->snd_fack = tp->snd_una; 2417 if (SEQ_LT(tp->snd_nxt, tp->snd_una)) 2418 tp->snd_nxt = tp->snd_una; 2419 if (SEQ_LT(tp->snd_high, tp->snd_una)) 2420 tp->snd_high = tp->snd_una; 2421 2422 switch (tp->t_state) { 2423 2424 /* 2425 * In FIN_WAIT_1 STATE in addition to the processing 2426 * for the ESTABLISHED state if our FIN is now acknowledged 2427 * then enter FIN_WAIT_2. 2428 */ 2429 case TCPS_FIN_WAIT_1: 2430 if (ourfinisacked) { 2431 /* 2432 * If we can't receive any more 2433 * data, then closing user can proceed. 2434 * Starting the timer is contrary to the 2435 * specification, but if we don't get a FIN 2436 * we'll hang forever. 2437 */ 2438 if (so->so_state & SS_CANTRCVMORE) { 2439 soisdisconnected(so); 2440 if (tp->t_maxidle > 0) 2441 TCP_TIMER_ARM(tp, TCPT_2MSL, 2442 tp->t_maxidle); 2443 } 2444 tp->t_state = TCPS_FIN_WAIT_2; 2445 } 2446 break; 2447 2448 /* 2449 * In CLOSING STATE in addition to the processing for 2450 * the ESTABLISHED state if the ACK acknowledges our FIN 2451 * then enter the TIME-WAIT state, otherwise ignore 2452 * the segment. 2453 */ 2454 case TCPS_CLOSING: 2455 if (ourfinisacked) { 2456 tp->t_state = TCPS_TIME_WAIT; 2457 tcp_canceltimers(tp); 2458 TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * TCPTV_MSL); 2459 soisdisconnected(so); 2460 } 2461 break; 2462 2463 /* 2464 * In LAST_ACK, we may still be waiting for data to drain 2465 * and/or to be acked, as well as for the ack of our FIN. 2466 * If our FIN is now acknowledged, delete the TCB, 2467 * enter the closed state and return. 2468 */ 2469 case TCPS_LAST_ACK: 2470 if (ourfinisacked) { 2471 tp = tcp_close(tp); 2472 goto drop; 2473 } 2474 break; 2475 2476 /* 2477 * In TIME_WAIT state the only thing that should arrive 2478 * is a retransmission of the remote FIN. Acknowledge 2479 * it and restart the finack timer. 2480 */ 2481 case TCPS_TIME_WAIT: 2482 TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * TCPTV_MSL); 2483 goto dropafterack; 2484 } 2485 } 2486 2487 step6: 2488 /* 2489 * Update window information. 2490 * Don't look at window if no ACK: TAC's send garbage on first SYN. 2491 */ 2492 if ((tiflags & TH_ACK) && (SEQ_LT(tp->snd_wl1, th->th_seq) || 2493 (tp->snd_wl1 == th->th_seq && (SEQ_LT(tp->snd_wl2, th->th_ack) || 2494 (tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd))))) { 2495 /* keep track of pure window updates */ 2496 if (tlen == 0 && 2497 tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd) 2498 tcpstat.tcps_rcvwinupd++; 2499 tp->snd_wnd = tiwin; 2500 tp->snd_wl1 = th->th_seq; 2501 tp->snd_wl2 = th->th_ack; 2502 if (tp->snd_wnd > tp->max_sndwnd) 2503 tp->max_sndwnd = tp->snd_wnd; 2504 needoutput = 1; 2505 } 2506 2507 /* 2508 * Process segments with URG. 2509 */ 2510 if ((tiflags & TH_URG) && th->th_urp && 2511 TCPS_HAVERCVDFIN(tp->t_state) == 0) { 2512 /* 2513 * This is a kludge, but if we receive and accept 2514 * random urgent pointers, we'll crash in 2515 * soreceive. It's hard to imagine someone 2516 * actually wanting to send this much urgent data. 2517 */ 2518 if (th->th_urp + so->so_rcv.sb_cc > sb_max) { 2519 th->th_urp = 0; /* XXX */ 2520 tiflags &= ~TH_URG; /* XXX */ 2521 goto dodata; /* XXX */ 2522 } 2523 /* 2524 * If this segment advances the known urgent pointer, 2525 * then mark the data stream. This should not happen 2526 * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since 2527 * a FIN has been received from the remote side. 2528 * In these states we ignore the URG. 2529 * 2530 * According to RFC961 (Assigned Protocols), 2531 * the urgent pointer points to the last octet 2532 * of urgent data. We continue, however, 2533 * to consider it to indicate the first octet 2534 * of data past the urgent section as the original 2535 * spec states (in one of two places). 2536 */ 2537 if (SEQ_GT(th->th_seq+th->th_urp, tp->rcv_up)) { 2538 tp->rcv_up = th->th_seq + th->th_urp; 2539 so->so_oobmark = so->so_rcv.sb_cc + 2540 (tp->rcv_up - tp->rcv_nxt) - 1; 2541 if (so->so_oobmark == 0) 2542 so->so_state |= SS_RCVATMARK; 2543 sohasoutofband(so); 2544 tp->t_oobflags &= ~(TCPOOB_HAVEDATA | TCPOOB_HADDATA); 2545 } 2546 /* 2547 * Remove out of band data so doesn't get presented to user. 2548 * This can happen independent of advancing the URG pointer, 2549 * but if two URG's are pending at once, some out-of-band 2550 * data may creep in... ick. 2551 */ 2552 if (th->th_urp <= (u_int16_t) tlen 2553 #ifdef SO_OOBINLINE 2554 && (so->so_options & SO_OOBINLINE) == 0 2555 #endif 2556 ) 2557 tcp_pulloutofband(so, th, m, hdroptlen); 2558 } else 2559 /* 2560 * If no out of band data is expected, 2561 * pull receive urgent pointer along 2562 * with the receive window. 2563 */ 2564 if (SEQ_GT(tp->rcv_nxt, tp->rcv_up)) 2565 tp->rcv_up = tp->rcv_nxt; 2566 dodata: /* XXX */ 2567 2568 /* 2569 * Process the segment text, merging it into the TCP sequencing queue, 2570 * and arranging for acknowledgement of receipt if necessary. 2571 * This process logically involves adjusting tp->rcv_wnd as data 2572 * is presented to the user (this happens in tcp_usrreq.c, 2573 * case PRU_RCVD). If a FIN has already been received on this 2574 * connection then we just ignore the text. 2575 */ 2576 if ((tlen || (tiflags & TH_FIN)) && 2577 TCPS_HAVERCVDFIN(tp->t_state) == 0) { 2578 /* 2579 * Insert segment ti into reassembly queue of tcp with 2580 * control block tp. Return TH_FIN if reassembly now includes 2581 * a segment with FIN. The macro form does the common case 2582 * inline (segment is the next to be received on an 2583 * established connection, and the queue is empty), 2584 * avoiding linkage into and removal from the queue and 2585 * repetition of various conversions. 2586 * Set DELACK for segments received in order, but ack 2587 * immediately when segments are out of order 2588 * (so fast retransmit can work). 2589 */ 2590 /* NOTE: this was TCP_REASS() macro, but used only once */ 2591 TCP_REASS_LOCK(tp); 2592 if (th->th_seq == tp->rcv_nxt && 2593 TAILQ_FIRST(&tp->segq) == NULL && 2594 tp->t_state == TCPS_ESTABLISHED) { 2595 TCP_SETUP_ACK(tp, th); 2596 tp->rcv_nxt += tlen; 2597 tiflags = th->th_flags & TH_FIN; 2598 tcpstat.tcps_rcvpack++; 2599 tcpstat.tcps_rcvbyte += tlen; 2600 ND6_HINT(tp); 2601 if (so->so_state & SS_CANTRCVMORE) 2602 m_freem(m); 2603 else { 2604 m_adj(m, hdroptlen); 2605 sbappendstream(&(so)->so_rcv, m); 2606 } 2607 sorwakeup(so); 2608 } else { 2609 m_adj(m, hdroptlen); 2610 tiflags = tcp_reass(tp, th, m, &tlen); 2611 tp->t_flags |= TF_ACKNOW; 2612 } 2613 TCP_REASS_UNLOCK(tp); 2614 2615 /* 2616 * Note the amount of data that peer has sent into 2617 * our window, in order to estimate the sender's 2618 * buffer size. 2619 */ 2620 len = so->so_rcv.sb_hiwat - (tp->rcv_adv - tp->rcv_nxt); 2621 } else { 2622 m_freem(m); 2623 m = NULL; 2624 tiflags &= ~TH_FIN; 2625 } 2626 2627 /* 2628 * If FIN is received ACK the FIN and let the user know 2629 * that the connection is closing. Ignore a FIN received before 2630 * the connection is fully established. 2631 */ 2632 if ((tiflags & TH_FIN) && TCPS_HAVEESTABLISHED(tp->t_state)) { 2633 if (TCPS_HAVERCVDFIN(tp->t_state) == 0) { 2634 socantrcvmore(so); 2635 tp->t_flags |= TF_ACKNOW; 2636 tp->rcv_nxt++; 2637 } 2638 switch (tp->t_state) { 2639 2640 /* 2641 * In ESTABLISHED STATE enter the CLOSE_WAIT state. 2642 */ 2643 case TCPS_ESTABLISHED: 2644 tp->t_state = TCPS_CLOSE_WAIT; 2645 break; 2646 2647 /* 2648 * If still in FIN_WAIT_1 STATE FIN has not been acked so 2649 * enter the CLOSING state. 2650 */ 2651 case TCPS_FIN_WAIT_1: 2652 tp->t_state = TCPS_CLOSING; 2653 break; 2654 2655 /* 2656 * In FIN_WAIT_2 state enter the TIME_WAIT state, 2657 * starting the time-wait timer, turning off the other 2658 * standard timers. 2659 */ 2660 case TCPS_FIN_WAIT_2: 2661 tp->t_state = TCPS_TIME_WAIT; 2662 tcp_canceltimers(tp); 2663 TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * TCPTV_MSL); 2664 soisdisconnected(so); 2665 break; 2666 2667 /* 2668 * In TIME_WAIT state restart the 2 MSL time_wait timer. 2669 */ 2670 case TCPS_TIME_WAIT: 2671 TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * TCPTV_MSL); 2672 break; 2673 } 2674 } 2675 #ifdef TCP_DEBUG 2676 if (so->so_options & SO_DEBUG) 2677 tcp_trace(TA_INPUT, ostate, tp, tcp_saveti, 0); 2678 #endif 2679 2680 /* 2681 * Return any desired output. 2682 */ 2683 if (needoutput || (tp->t_flags & TF_ACKNOW)) { 2684 (void) tcp_output(tp); 2685 } 2686 if (tcp_saveti) 2687 m_freem(tcp_saveti); 2688 return; 2689 2690 badsyn: 2691 /* 2692 * Received a bad SYN. Increment counters and dropwithreset. 2693 */ 2694 tcpstat.tcps_badsyn++; 2695 tp = NULL; 2696 goto dropwithreset; 2697 2698 dropafterack: 2699 /* 2700 * Generate an ACK dropping incoming segment if it occupies 2701 * sequence space, where the ACK reflects our state. 2702 */ 2703 if (tiflags & TH_RST) 2704 goto drop; 2705 goto dropafterack2; 2706 2707 dropafterack_ratelim: 2708 /* 2709 * We may want to rate-limit ACKs against SYN/RST attack. 2710 */ 2711 if (ppsratecheck(&tcp_ackdrop_ppslim_last, &tcp_ackdrop_ppslim_count, 2712 tcp_ackdrop_ppslim) == 0) { 2713 /* XXX stat */ 2714 goto drop; 2715 } 2716 /* ...fall into dropafterack2... */ 2717 2718 dropafterack2: 2719 m_freem(m); 2720 tp->t_flags |= TF_ACKNOW; 2721 (void) tcp_output(tp); 2722 if (tcp_saveti) 2723 m_freem(tcp_saveti); 2724 return; 2725 2726 dropwithreset_ratelim: 2727 /* 2728 * We may want to rate-limit RSTs in certain situations, 2729 * particularly if we are sending an RST in response to 2730 * an attempt to connect to or otherwise communicate with 2731 * a port for which we have no socket. 2732 */ 2733 if (ppsratecheck(&tcp_rst_ppslim_last, &tcp_rst_ppslim_count, 2734 tcp_rst_ppslim) == 0) { 2735 /* XXX stat */ 2736 goto drop; 2737 } 2738 /* ...fall into dropwithreset... */ 2739 2740 dropwithreset: 2741 /* 2742 * Generate a RST, dropping incoming segment. 2743 * Make ACK acceptable to originator of segment. 2744 */ 2745 if (tiflags & TH_RST) 2746 goto drop; 2747 2748 switch (af) { 2749 #ifdef INET6 2750 case AF_INET6: 2751 /* For following calls to tcp_respond */ 2752 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) 2753 goto drop; 2754 break; 2755 #endif /* INET6 */ 2756 case AF_INET: 2757 if (IN_MULTICAST(ip->ip_dst.s_addr) || 2758 in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) 2759 goto drop; 2760 } 2761 2762 if (tiflags & TH_ACK) 2763 (void)tcp_respond(tp, m, m, th, (tcp_seq)0, th->th_ack, TH_RST); 2764 else { 2765 if (tiflags & TH_SYN) 2766 tlen++; 2767 (void)tcp_respond(tp, m, m, th, th->th_seq + tlen, (tcp_seq)0, 2768 TH_RST|TH_ACK); 2769 } 2770 if (tcp_saveti) 2771 m_freem(tcp_saveti); 2772 return; 2773 2774 badcsum: 2775 drop: 2776 /* 2777 * Drop space held by incoming segment and return. 2778 */ 2779 if (tp) { 2780 if (tp->t_inpcb) 2781 so = tp->t_inpcb->inp_socket; 2782 #ifdef INET6 2783 else if (tp->t_in6pcb) 2784 so = tp->t_in6pcb->in6p_socket; 2785 #endif 2786 else 2787 so = NULL; 2788 #ifdef TCP_DEBUG 2789 if (so && (so->so_options & SO_DEBUG) != 0) 2790 tcp_trace(TA_DROP, ostate, tp, tcp_saveti, 0); 2791 #endif 2792 } 2793 if (tcp_saveti) 2794 m_freem(tcp_saveti); 2795 m_freem(m); 2796 return; 2797 } 2798 2799 #ifdef TCP_SIGNATURE 2800 int 2801 tcp_signature_apply(void *fstate, void *data, u_int len) 2802 { 2803 2804 MD5Update(fstate, (u_char *)data, len); 2805 return (0); 2806 } 2807 2808 struct secasvar * 2809 tcp_signature_getsav(struct mbuf *m, struct tcphdr *th) 2810 { 2811 struct secasvar *sav; 2812 #ifdef FAST_IPSEC 2813 union sockaddr_union dst; 2814 #endif 2815 struct ip *ip; 2816 struct ip6_hdr *ip6; 2817 2818 ip = mtod(m, struct ip *); 2819 switch (ip->ip_v) { 2820 case 4: 2821 ip = mtod(m, struct ip *); 2822 ip6 = NULL; 2823 break; 2824 case 6: 2825 ip = NULL; 2826 ip6 = mtod(m, struct ip6_hdr *); 2827 break; 2828 default: 2829 return (NULL); 2830 } 2831 2832 #ifdef FAST_IPSEC 2833 /* Extract the destination from the IP header in the mbuf. */ 2834 bzero(&dst, sizeof(union sockaddr_union)); 2835 if (ip !=NULL) { 2836 dst.sa.sa_len = sizeof(struct sockaddr_in); 2837 dst.sa.sa_family = AF_INET; 2838 dst.sin.sin_addr = ip->ip_dst; 2839 } else { 2840 dst.sa.sa_len = sizeof(struct sockaddr_in6); 2841 dst.sa.sa_family = AF_INET6; 2842 dst.sin6.sin6_addr = ip6->ip6_dst; 2843 } 2844 2845 /* 2846 * Look up an SADB entry which matches the address of the peer. 2847 */ 2848 sav = KEY_ALLOCSA(&dst, IPPROTO_TCP, htonl(TCP_SIG_SPI)); 2849 #else 2850 if (ip) 2851 sav = key_allocsa(AF_INET, (void *)&ip->ip_src, 2852 (void *)&ip->ip_dst, IPPROTO_TCP, 2853 htonl(TCP_SIG_SPI), 0, 0); 2854 else 2855 sav = key_allocsa(AF_INET6, (void *)&ip6->ip6_src, 2856 (void *)&ip6->ip6_dst, IPPROTO_TCP, 2857 htonl(TCP_SIG_SPI), 0, 0); 2858 #endif 2859 2860 return (sav); /* freesav must be performed by caller */ 2861 } 2862 2863 int 2864 tcp_signature(struct mbuf *m, struct tcphdr *th, int thoff, 2865 struct secasvar *sav, char *sig) 2866 { 2867 MD5_CTX ctx; 2868 struct ip *ip; 2869 struct ipovly *ipovly; 2870 struct ip6_hdr *ip6; 2871 struct ippseudo ippseudo; 2872 struct ip6_hdr_pseudo ip6pseudo; 2873 struct tcphdr th0; 2874 int l, tcphdrlen; 2875 2876 if (sav == NULL) 2877 return (-1); 2878 2879 tcphdrlen = th->th_off * 4; 2880 2881 switch (mtod(m, struct ip *)->ip_v) { 2882 case 4: 2883 ip = mtod(m, struct ip *); 2884 ip6 = NULL; 2885 break; 2886 case 6: 2887 ip = NULL; 2888 ip6 = mtod(m, struct ip6_hdr *); 2889 break; 2890 default: 2891 return (-1); 2892 } 2893 2894 MD5Init(&ctx); 2895 2896 if (ip) { 2897 memset(&ippseudo, 0, sizeof(ippseudo)); 2898 ipovly = (struct ipovly *)ip; 2899 ippseudo.ippseudo_src = ipovly->ih_src; 2900 ippseudo.ippseudo_dst = ipovly->ih_dst; 2901 ippseudo.ippseudo_pad = 0; 2902 ippseudo.ippseudo_p = IPPROTO_TCP; 2903 ippseudo.ippseudo_len = htons(m->m_pkthdr.len - thoff); 2904 MD5Update(&ctx, (char *)&ippseudo, sizeof(ippseudo)); 2905 } else { 2906 memset(&ip6pseudo, 0, sizeof(ip6pseudo)); 2907 ip6pseudo.ip6ph_src = ip6->ip6_src; 2908 in6_clearscope(&ip6pseudo.ip6ph_src); 2909 ip6pseudo.ip6ph_dst = ip6->ip6_dst; 2910 in6_clearscope(&ip6pseudo.ip6ph_dst); 2911 ip6pseudo.ip6ph_len = htons(m->m_pkthdr.len - thoff); 2912 ip6pseudo.ip6ph_nxt = IPPROTO_TCP; 2913 MD5Update(&ctx, (char *)&ip6pseudo, sizeof(ip6pseudo)); 2914 } 2915 2916 th0 = *th; 2917 th0.th_sum = 0; 2918 MD5Update(&ctx, (char *)&th0, sizeof(th0)); 2919 2920 l = m->m_pkthdr.len - thoff - tcphdrlen; 2921 if (l > 0) 2922 m_apply(m, thoff + tcphdrlen, 2923 m->m_pkthdr.len - thoff - tcphdrlen, 2924 tcp_signature_apply, &ctx); 2925 2926 MD5Update(&ctx, _KEYBUF(sav->key_auth), _KEYLEN(sav->key_auth)); 2927 MD5Final(sig, &ctx); 2928 2929 return (0); 2930 } 2931 #endif 2932 2933 static int 2934 tcp_dooptions(struct tcpcb *tp, const u_char *cp, int cnt, 2935 struct tcphdr *th, 2936 struct mbuf *m, int toff, struct tcp_opt_info *oi) 2937 { 2938 u_int16_t mss; 2939 int opt, optlen = 0; 2940 #ifdef TCP_SIGNATURE 2941 void *sigp = NULL; 2942 char sigbuf[TCP_SIGLEN]; 2943 struct secasvar *sav = NULL; 2944 #endif 2945 2946 for (; cp && cnt > 0; cnt -= optlen, cp += optlen) { 2947 opt = cp[0]; 2948 if (opt == TCPOPT_EOL) 2949 break; 2950 if (opt == TCPOPT_NOP) 2951 optlen = 1; 2952 else { 2953 if (cnt < 2) 2954 break; 2955 optlen = cp[1]; 2956 if (optlen < 2 || optlen > cnt) 2957 break; 2958 } 2959 switch (opt) { 2960 2961 default: 2962 continue; 2963 2964 case TCPOPT_MAXSEG: 2965 if (optlen != TCPOLEN_MAXSEG) 2966 continue; 2967 if (!(th->th_flags & TH_SYN)) 2968 continue; 2969 if (TCPS_HAVERCVDSYN(tp->t_state)) 2970 continue; 2971 bcopy(cp + 2, &mss, sizeof(mss)); 2972 oi->maxseg = ntohs(mss); 2973 break; 2974 2975 case TCPOPT_WINDOW: 2976 if (optlen != TCPOLEN_WINDOW) 2977 continue; 2978 if (!(th->th_flags & TH_SYN)) 2979 continue; 2980 if (TCPS_HAVERCVDSYN(tp->t_state)) 2981 continue; 2982 tp->t_flags |= TF_RCVD_SCALE; 2983 tp->requested_s_scale = cp[2]; 2984 if (tp->requested_s_scale > TCP_MAX_WINSHIFT) { 2985 #if 0 /*XXX*/ 2986 char *p; 2987 2988 if (ip) 2989 p = ntohl(ip->ip_src); 2990 #ifdef INET6 2991 else if (ip6) 2992 p = ip6_sprintf(&ip6->ip6_src); 2993 #endif 2994 else 2995 p = "(unknown)"; 2996 log(LOG_ERR, "TCP: invalid wscale %d from %s, " 2997 "assuming %d\n", 2998 tp->requested_s_scale, p, 2999 TCP_MAX_WINSHIFT); 3000 #else 3001 log(LOG_ERR, "TCP: invalid wscale %d, " 3002 "assuming %d\n", 3003 tp->requested_s_scale, 3004 TCP_MAX_WINSHIFT); 3005 #endif 3006 tp->requested_s_scale = TCP_MAX_WINSHIFT; 3007 } 3008 break; 3009 3010 case TCPOPT_TIMESTAMP: 3011 if (optlen != TCPOLEN_TIMESTAMP) 3012 continue; 3013 oi->ts_present = 1; 3014 bcopy(cp + 2, &oi->ts_val, sizeof(oi->ts_val)); 3015 NTOHL(oi->ts_val); 3016 bcopy(cp + 6, &oi->ts_ecr, sizeof(oi->ts_ecr)); 3017 NTOHL(oi->ts_ecr); 3018 3019 if (!(th->th_flags & TH_SYN)) 3020 continue; 3021 if (TCPS_HAVERCVDSYN(tp->t_state)) 3022 continue; 3023 /* 3024 * A timestamp received in a SYN makes 3025 * it ok to send timestamp requests and replies. 3026 */ 3027 tp->t_flags |= TF_RCVD_TSTMP; 3028 tp->ts_recent = oi->ts_val; 3029 tp->ts_recent_age = tcp_now; 3030 break; 3031 3032 case TCPOPT_SACK_PERMITTED: 3033 if (optlen != TCPOLEN_SACK_PERMITTED) 3034 continue; 3035 if (!(th->th_flags & TH_SYN)) 3036 continue; 3037 if (TCPS_HAVERCVDSYN(tp->t_state)) 3038 continue; 3039 if (tcp_do_sack) { 3040 tp->t_flags |= TF_SACK_PERMIT; 3041 tp->t_flags |= TF_WILL_SACK; 3042 } 3043 break; 3044 3045 case TCPOPT_SACK: 3046 tcp_sack_option(tp, th, cp, optlen); 3047 break; 3048 #ifdef TCP_SIGNATURE 3049 case TCPOPT_SIGNATURE: 3050 if (optlen != TCPOLEN_SIGNATURE) 3051 continue; 3052 if (sigp && bcmp(sigp, cp + 2, TCP_SIGLEN)) 3053 return (-1); 3054 3055 sigp = sigbuf; 3056 memcpy(sigbuf, cp + 2, TCP_SIGLEN); 3057 tp->t_flags |= TF_SIGNATURE; 3058 break; 3059 #endif 3060 } 3061 } 3062 3063 #ifdef TCP_SIGNATURE 3064 if (tp->t_flags & TF_SIGNATURE) { 3065 3066 sav = tcp_signature_getsav(m, th); 3067 3068 if (sav == NULL && tp->t_state == TCPS_LISTEN) 3069 return (-1); 3070 } 3071 3072 if ((sigp ? TF_SIGNATURE : 0) ^ (tp->t_flags & TF_SIGNATURE)) { 3073 if (sav == NULL) 3074 return (-1); 3075 #ifdef FAST_IPSEC 3076 KEY_FREESAV(&sav); 3077 #else 3078 key_freesav(sav); 3079 #endif 3080 return (-1); 3081 } 3082 3083 if (sigp) { 3084 char sig[TCP_SIGLEN]; 3085 3086 TCP_FIELDS_TO_NET(th); 3087 if (tcp_signature(m, th, toff, sav, sig) < 0) { 3088 TCP_FIELDS_TO_HOST(th); 3089 if (sav == NULL) 3090 return (-1); 3091 #ifdef FAST_IPSEC 3092 KEY_FREESAV(&sav); 3093 #else 3094 key_freesav(sav); 3095 #endif 3096 return (-1); 3097 } 3098 TCP_FIELDS_TO_HOST(th); 3099 3100 if (bcmp(sig, sigp, TCP_SIGLEN)) { 3101 tcpstat.tcps_badsig++; 3102 if (sav == NULL) 3103 return (-1); 3104 #ifdef FAST_IPSEC 3105 KEY_FREESAV(&sav); 3106 #else 3107 key_freesav(sav); 3108 #endif 3109 return (-1); 3110 } else 3111 tcpstat.tcps_goodsig++; 3112 3113 key_sa_recordxfer(sav, m); 3114 #ifdef FAST_IPSEC 3115 KEY_FREESAV(&sav); 3116 #else 3117 key_freesav(sav); 3118 #endif 3119 } 3120 #endif 3121 3122 return (0); 3123 } 3124 3125 /* 3126 * Pull out of band byte out of a segment so 3127 * it doesn't appear in the user's data queue. 3128 * It is still reflected in the segment length for 3129 * sequencing purposes. 3130 */ 3131 void 3132 tcp_pulloutofband(struct socket *so, struct tcphdr *th, 3133 struct mbuf *m, int off) 3134 { 3135 int cnt = off + th->th_urp - 1; 3136 3137 while (cnt >= 0) { 3138 if (m->m_len > cnt) { 3139 char *cp = mtod(m, char *) + cnt; 3140 struct tcpcb *tp = sototcpcb(so); 3141 3142 tp->t_iobc = *cp; 3143 tp->t_oobflags |= TCPOOB_HAVEDATA; 3144 bcopy(cp+1, cp, (unsigned)(m->m_len - cnt - 1)); 3145 m->m_len--; 3146 return; 3147 } 3148 cnt -= m->m_len; 3149 m = m->m_next; 3150 if (m == 0) 3151 break; 3152 } 3153 panic("tcp_pulloutofband"); 3154 } 3155 3156 /* 3157 * Collect new round-trip time estimate 3158 * and update averages and current timeout. 3159 */ 3160 void 3161 tcp_xmit_timer(struct tcpcb *tp, uint32_t rtt) 3162 { 3163 int32_t delta; 3164 3165 tcpstat.tcps_rttupdated++; 3166 if (tp->t_srtt != 0) { 3167 /* 3168 * srtt is stored as fixed point with 3 bits after the 3169 * binary point (i.e., scaled by 8). The following magic 3170 * is equivalent to the smoothing algorithm in rfc793 with 3171 * an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed 3172 * point). Adjust rtt to origin 0. 3173 */ 3174 delta = (rtt << 2) - (tp->t_srtt >> TCP_RTT_SHIFT); 3175 if ((tp->t_srtt += delta) <= 0) 3176 tp->t_srtt = 1 << 2; 3177 /* 3178 * We accumulate a smoothed rtt variance (actually, a 3179 * smoothed mean difference), then set the retransmit 3180 * timer to smoothed rtt + 4 times the smoothed variance. 3181 * rttvar is stored as fixed point with 2 bits after the 3182 * binary point (scaled by 4). The following is 3183 * equivalent to rfc793 smoothing with an alpha of .75 3184 * (rttvar = rttvar*3/4 + |delta| / 4). This replaces 3185 * rfc793's wired-in beta. 3186 */ 3187 if (delta < 0) 3188 delta = -delta; 3189 delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT); 3190 if ((tp->t_rttvar += delta) <= 0) 3191 tp->t_rttvar = 1 << 2; 3192 } else { 3193 /* 3194 * No rtt measurement yet - use the unsmoothed rtt. 3195 * Set the variance to half the rtt (so our first 3196 * retransmit happens at 3*rtt). 3197 */ 3198 tp->t_srtt = rtt << (TCP_RTT_SHIFT + 2); 3199 tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT + 2 - 1); 3200 } 3201 tp->t_rtttime = 0; 3202 tp->t_rxtshift = 0; 3203 3204 /* 3205 * the retransmit should happen at rtt + 4 * rttvar. 3206 * Because of the way we do the smoothing, srtt and rttvar 3207 * will each average +1/2 tick of bias. When we compute 3208 * the retransmit timer, we want 1/2 tick of rounding and 3209 * 1 extra tick because of +-1/2 tick uncertainty in the 3210 * firing of the timer. The bias will give us exactly the 3211 * 1.5 tick we need. But, because the bias is 3212 * statistical, we have to test that we don't drop below 3213 * the minimum feasible timer (which is 2 ticks). 3214 */ 3215 TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp), 3216 max(tp->t_rttmin, rtt + 2), TCPTV_REXMTMAX); 3217 3218 /* 3219 * We received an ack for a packet that wasn't retransmitted; 3220 * it is probably safe to discard any error indications we've 3221 * received recently. This isn't quite right, but close enough 3222 * for now (a route might have failed after we sent a segment, 3223 * and the return path might not be symmetrical). 3224 */ 3225 tp->t_softerror = 0; 3226 } 3227 3228 3229 /* 3230 * TCP compressed state engine. Currently used to hold compressed 3231 * state for SYN_RECEIVED. 3232 */ 3233 3234 u_long syn_cache_count; 3235 u_int32_t syn_hash1, syn_hash2; 3236 3237 #define SYN_HASH(sa, sp, dp) \ 3238 ((((sa)->s_addr^syn_hash1)*(((((u_int32_t)(dp))<<16) + \ 3239 ((u_int32_t)(sp)))^syn_hash2))) 3240 #ifndef INET6 3241 #define SYN_HASHALL(hash, src, dst) \ 3242 do { \ 3243 hash = SYN_HASH(&((const struct sockaddr_in *)(src))->sin_addr, \ 3244 ((const struct sockaddr_in *)(src))->sin_port, \ 3245 ((const struct sockaddr_in *)(dst))->sin_port); \ 3246 } while (/*CONSTCOND*/ 0) 3247 #else 3248 #define SYN_HASH6(sa, sp, dp) \ 3249 ((((sa)->s6_addr32[0] ^ (sa)->s6_addr32[3] ^ syn_hash1) * \ 3250 (((((u_int32_t)(dp))<<16) + ((u_int32_t)(sp)))^syn_hash2)) \ 3251 & 0x7fffffff) 3252 3253 #define SYN_HASHALL(hash, src, dst) \ 3254 do { \ 3255 switch ((src)->sa_family) { \ 3256 case AF_INET: \ 3257 hash = SYN_HASH(&((const struct sockaddr_in *)(src))->sin_addr, \ 3258 ((const struct sockaddr_in *)(src))->sin_port, \ 3259 ((const struct sockaddr_in *)(dst))->sin_port); \ 3260 break; \ 3261 case AF_INET6: \ 3262 hash = SYN_HASH6(&((const struct sockaddr_in6 *)(src))->sin6_addr, \ 3263 ((const struct sockaddr_in6 *)(src))->sin6_port, \ 3264 ((const struct sockaddr_in6 *)(dst))->sin6_port); \ 3265 break; \ 3266 default: \ 3267 hash = 0; \ 3268 } \ 3269 } while (/*CONSTCOND*/0) 3270 #endif /* INET6 */ 3271 3272 #define SYN_CACHE_RM(sc) \ 3273 do { \ 3274 TAILQ_REMOVE(&tcp_syn_cache[(sc)->sc_bucketidx].sch_bucket, \ 3275 (sc), sc_bucketq); \ 3276 (sc)->sc_tp = NULL; \ 3277 LIST_REMOVE((sc), sc_tpq); \ 3278 tcp_syn_cache[(sc)->sc_bucketidx].sch_length--; \ 3279 callout_stop(&(sc)->sc_timer); \ 3280 syn_cache_count--; \ 3281 } while (/*CONSTCOND*/0) 3282 3283 #define SYN_CACHE_PUT(sc) \ 3284 do { \ 3285 if ((sc)->sc_ipopts) \ 3286 (void) m_free((sc)->sc_ipopts); \ 3287 rtcache_free(&(sc)->sc_route); \ 3288 if (callout_invoking(&(sc)->sc_timer)) \ 3289 (sc)->sc_flags |= SCF_DEAD; \ 3290 else { \ 3291 callout_destroy(&sc->sc_timer); \ 3292 pool_put(&syn_cache_pool, (sc)); \ 3293 } \ 3294 } while (/*CONSTCOND*/0) 3295 3296 POOL_INIT(syn_cache_pool, sizeof(struct syn_cache), 0, 0, 0, "synpl", NULL, 3297 IPL_SOFTNET); 3298 3299 /* 3300 * We don't estimate RTT with SYNs, so each packet starts with the default 3301 * RTT and each timer step has a fixed timeout value. 3302 */ 3303 #define SYN_CACHE_TIMER_ARM(sc) \ 3304 do { \ 3305 TCPT_RANGESET((sc)->sc_rxtcur, \ 3306 TCPTV_SRTTDFLT * tcp_backoff[(sc)->sc_rxtshift], TCPTV_MIN, \ 3307 TCPTV_REXMTMAX); \ 3308 callout_reset(&(sc)->sc_timer, \ 3309 (sc)->sc_rxtcur * (hz / PR_SLOWHZ), syn_cache_timer, (sc)); \ 3310 } while (/*CONSTCOND*/0) 3311 3312 #define SYN_CACHE_TIMESTAMP(sc) (tcp_now - (sc)->sc_timebase) 3313 3314 void 3315 syn_cache_init(void) 3316 { 3317 int i; 3318 3319 /* Initialize the hash buckets. */ 3320 for (i = 0; i < tcp_syn_cache_size; i++) 3321 TAILQ_INIT(&tcp_syn_cache[i].sch_bucket); 3322 } 3323 3324 void 3325 syn_cache_insert(struct syn_cache *sc, struct tcpcb *tp) 3326 { 3327 struct syn_cache_head *scp; 3328 struct syn_cache *sc2; 3329 int s; 3330 3331 /* 3332 * If there are no entries in the hash table, reinitialize 3333 * the hash secrets. 3334 */ 3335 if (syn_cache_count == 0) { 3336 syn_hash1 = arc4random(); 3337 syn_hash2 = arc4random(); 3338 } 3339 3340 SYN_HASHALL(sc->sc_hash, &sc->sc_src.sa, &sc->sc_dst.sa); 3341 sc->sc_bucketidx = sc->sc_hash % tcp_syn_cache_size; 3342 scp = &tcp_syn_cache[sc->sc_bucketidx]; 3343 3344 /* 3345 * Make sure that we don't overflow the per-bucket 3346 * limit or the total cache size limit. 3347 */ 3348 s = splsoftnet(); 3349 if (scp->sch_length >= tcp_syn_bucket_limit) { 3350 tcpstat.tcps_sc_bucketoverflow++; 3351 /* 3352 * The bucket is full. Toss the oldest element in the 3353 * bucket. This will be the first entry in the bucket. 3354 */ 3355 sc2 = TAILQ_FIRST(&scp->sch_bucket); 3356 #ifdef DIAGNOSTIC 3357 /* 3358 * This should never happen; we should always find an 3359 * entry in our bucket. 3360 */ 3361 if (sc2 == NULL) 3362 panic("syn_cache_insert: bucketoverflow: impossible"); 3363 #endif 3364 SYN_CACHE_RM(sc2); 3365 SYN_CACHE_PUT(sc2); /* calls pool_put but see spl above */ 3366 } else if (syn_cache_count >= tcp_syn_cache_limit) { 3367 struct syn_cache_head *scp2, *sce; 3368 3369 tcpstat.tcps_sc_overflowed++; 3370 /* 3371 * The cache is full. Toss the oldest entry in the 3372 * first non-empty bucket we can find. 3373 * 3374 * XXX We would really like to toss the oldest 3375 * entry in the cache, but we hope that this 3376 * condition doesn't happen very often. 3377 */ 3378 scp2 = scp; 3379 if (TAILQ_EMPTY(&scp2->sch_bucket)) { 3380 sce = &tcp_syn_cache[tcp_syn_cache_size]; 3381 for (++scp2; scp2 != scp; scp2++) { 3382 if (scp2 >= sce) 3383 scp2 = &tcp_syn_cache[0]; 3384 if (! TAILQ_EMPTY(&scp2->sch_bucket)) 3385 break; 3386 } 3387 #ifdef DIAGNOSTIC 3388 /* 3389 * This should never happen; we should always find a 3390 * non-empty bucket. 3391 */ 3392 if (scp2 == scp) 3393 panic("syn_cache_insert: cacheoverflow: " 3394 "impossible"); 3395 #endif 3396 } 3397 sc2 = TAILQ_FIRST(&scp2->sch_bucket); 3398 SYN_CACHE_RM(sc2); 3399 SYN_CACHE_PUT(sc2); /* calls pool_put but see spl above */ 3400 } 3401 3402 /* 3403 * Initialize the entry's timer. 3404 */ 3405 sc->sc_rxttot = 0; 3406 sc->sc_rxtshift = 0; 3407 SYN_CACHE_TIMER_ARM(sc); 3408 3409 /* Link it from tcpcb entry */ 3410 LIST_INSERT_HEAD(&tp->t_sc, sc, sc_tpq); 3411 3412 /* Put it into the bucket. */ 3413 TAILQ_INSERT_TAIL(&scp->sch_bucket, sc, sc_bucketq); 3414 scp->sch_length++; 3415 syn_cache_count++; 3416 3417 tcpstat.tcps_sc_added++; 3418 splx(s); 3419 } 3420 3421 /* 3422 * Walk the timer queues, looking for SYN,ACKs that need to be retransmitted. 3423 * If we have retransmitted an entry the maximum number of times, expire 3424 * that entry. 3425 */ 3426 void 3427 syn_cache_timer(void *arg) 3428 { 3429 struct syn_cache *sc = arg; 3430 int s; 3431 3432 s = splsoftnet(); 3433 callout_ack(&sc->sc_timer); 3434 3435 if (__predict_false(sc->sc_flags & SCF_DEAD)) { 3436 tcpstat.tcps_sc_delayed_free++; 3437 callout_destroy(&sc->sc_timer); 3438 pool_put(&syn_cache_pool, sc); 3439 splx(s); 3440 return; 3441 } 3442 3443 if (__predict_false(sc->sc_rxtshift == TCP_MAXRXTSHIFT)) { 3444 /* Drop it -- too many retransmissions. */ 3445 goto dropit; 3446 } 3447 3448 /* 3449 * Compute the total amount of time this entry has 3450 * been on a queue. If this entry has been on longer 3451 * than the keep alive timer would allow, expire it. 3452 */ 3453 sc->sc_rxttot += sc->sc_rxtcur; 3454 if (sc->sc_rxttot >= tcp_keepinit) 3455 goto dropit; 3456 3457 tcpstat.tcps_sc_retransmitted++; 3458 (void) syn_cache_respond(sc, NULL); 3459 3460 /* Advance the timer back-off. */ 3461 sc->sc_rxtshift++; 3462 SYN_CACHE_TIMER_ARM(sc); 3463 3464 splx(s); 3465 return; 3466 3467 dropit: 3468 tcpstat.tcps_sc_timed_out++; 3469 SYN_CACHE_RM(sc); 3470 SYN_CACHE_PUT(sc); /* calls pool_put but see spl above */ 3471 splx(s); 3472 } 3473 3474 /* 3475 * Remove syn cache created by the specified tcb entry, 3476 * because this does not make sense to keep them 3477 * (if there's no tcb entry, syn cache entry will never be used) 3478 */ 3479 void 3480 syn_cache_cleanup(struct tcpcb *tp) 3481 { 3482 struct syn_cache *sc, *nsc; 3483 int s; 3484 3485 s = splsoftnet(); 3486 3487 for (sc = LIST_FIRST(&tp->t_sc); sc != NULL; sc = nsc) { 3488 nsc = LIST_NEXT(sc, sc_tpq); 3489 3490 #ifdef DIAGNOSTIC 3491 if (sc->sc_tp != tp) 3492 panic("invalid sc_tp in syn_cache_cleanup"); 3493 #endif 3494 SYN_CACHE_RM(sc); 3495 SYN_CACHE_PUT(sc); /* calls pool_put but see spl above */ 3496 } 3497 /* just for safety */ 3498 LIST_INIT(&tp->t_sc); 3499 3500 splx(s); 3501 } 3502 3503 /* 3504 * Find an entry in the syn cache. 3505 */ 3506 struct syn_cache * 3507 syn_cache_lookup(const struct sockaddr *src, const struct sockaddr *dst, 3508 struct syn_cache_head **headp) 3509 { 3510 struct syn_cache *sc; 3511 struct syn_cache_head *scp; 3512 u_int32_t hash; 3513 int s; 3514 3515 SYN_HASHALL(hash, src, dst); 3516 3517 scp = &tcp_syn_cache[hash % tcp_syn_cache_size]; 3518 *headp = scp; 3519 s = splsoftnet(); 3520 for (sc = TAILQ_FIRST(&scp->sch_bucket); sc != NULL; 3521 sc = TAILQ_NEXT(sc, sc_bucketq)) { 3522 if (sc->sc_hash != hash) 3523 continue; 3524 if (!bcmp(&sc->sc_src, src, src->sa_len) && 3525 !bcmp(&sc->sc_dst, dst, dst->sa_len)) { 3526 splx(s); 3527 return (sc); 3528 } 3529 } 3530 splx(s); 3531 return (NULL); 3532 } 3533 3534 /* 3535 * This function gets called when we receive an ACK for a 3536 * socket in the LISTEN state. We look up the connection 3537 * in the syn cache, and if its there, we pull it out of 3538 * the cache and turn it into a full-blown connection in 3539 * the SYN-RECEIVED state. 3540 * 3541 * The return values may not be immediately obvious, and their effects 3542 * can be subtle, so here they are: 3543 * 3544 * NULL SYN was not found in cache; caller should drop the 3545 * packet and send an RST. 3546 * 3547 * -1 We were unable to create the new connection, and are 3548 * aborting it. An ACK,RST is being sent to the peer 3549 * (unless we got screwey sequence numbners; see below), 3550 * because the 3-way handshake has been completed. Caller 3551 * should not free the mbuf, since we may be using it. If 3552 * we are not, we will free it. 3553 * 3554 * Otherwise, the return value is a pointer to the new socket 3555 * associated with the connection. 3556 */ 3557 struct socket * 3558 syn_cache_get(struct sockaddr *src, struct sockaddr *dst, 3559 struct tcphdr *th, unsigned int hlen, unsigned int tlen, 3560 struct socket *so, struct mbuf *m) 3561 { 3562 struct syn_cache *sc; 3563 struct syn_cache_head *scp; 3564 struct inpcb *inp = NULL; 3565 #ifdef INET6 3566 struct in6pcb *in6p = NULL; 3567 #endif 3568 struct tcpcb *tp = 0; 3569 struct mbuf *am; 3570 int s; 3571 struct socket *oso; 3572 3573 s = splsoftnet(); 3574 if ((sc = syn_cache_lookup(src, dst, &scp)) == NULL) { 3575 splx(s); 3576 return (NULL); 3577 } 3578 3579 /* 3580 * Verify the sequence and ack numbers. Try getting the correct 3581 * response again. 3582 */ 3583 if ((th->th_ack != sc->sc_iss + 1) || 3584 SEQ_LEQ(th->th_seq, sc->sc_irs) || 3585 SEQ_GT(th->th_seq, sc->sc_irs + 1 + sc->sc_win)) { 3586 (void) syn_cache_respond(sc, m); 3587 splx(s); 3588 return ((struct socket *)(-1)); 3589 } 3590 3591 /* Remove this cache entry */ 3592 SYN_CACHE_RM(sc); 3593 splx(s); 3594 3595 /* 3596 * Ok, create the full blown connection, and set things up 3597 * as they would have been set up if we had created the 3598 * connection when the SYN arrived. If we can't create 3599 * the connection, abort it. 3600 */ 3601 /* 3602 * inp still has the OLD in_pcb stuff, set the 3603 * v6-related flags on the new guy, too. This is 3604 * done particularly for the case where an AF_INET6 3605 * socket is bound only to a port, and a v4 connection 3606 * comes in on that port. 3607 * we also copy the flowinfo from the original pcb 3608 * to the new one. 3609 */ 3610 oso = so; 3611 so = sonewconn(so, SS_ISCONNECTED); 3612 if (so == NULL) 3613 goto resetandabort; 3614 3615 switch (so->so_proto->pr_domain->dom_family) { 3616 #ifdef INET 3617 case AF_INET: 3618 inp = sotoinpcb(so); 3619 break; 3620 #endif 3621 #ifdef INET6 3622 case AF_INET6: 3623 in6p = sotoin6pcb(so); 3624 break; 3625 #endif 3626 } 3627 switch (src->sa_family) { 3628 #ifdef INET 3629 case AF_INET: 3630 if (inp) { 3631 inp->inp_laddr = ((struct sockaddr_in *)dst)->sin_addr; 3632 inp->inp_lport = ((struct sockaddr_in *)dst)->sin_port; 3633 inp->inp_options = ip_srcroute(); 3634 in_pcbstate(inp, INP_BOUND); 3635 if (inp->inp_options == NULL) { 3636 inp->inp_options = sc->sc_ipopts; 3637 sc->sc_ipopts = NULL; 3638 } 3639 } 3640 #ifdef INET6 3641 else if (in6p) { 3642 /* IPv4 packet to AF_INET6 socket */ 3643 bzero(&in6p->in6p_laddr, sizeof(in6p->in6p_laddr)); 3644 in6p->in6p_laddr.s6_addr16[5] = htons(0xffff); 3645 bcopy(&((struct sockaddr_in *)dst)->sin_addr, 3646 &in6p->in6p_laddr.s6_addr32[3], 3647 sizeof(((struct sockaddr_in *)dst)->sin_addr)); 3648 in6p->in6p_lport = ((struct sockaddr_in *)dst)->sin_port; 3649 in6totcpcb(in6p)->t_family = AF_INET; 3650 if (sotoin6pcb(oso)->in6p_flags & IN6P_IPV6_V6ONLY) 3651 in6p->in6p_flags |= IN6P_IPV6_V6ONLY; 3652 else 3653 in6p->in6p_flags &= ~IN6P_IPV6_V6ONLY; 3654 in6_pcbstate(in6p, IN6P_BOUND); 3655 } 3656 #endif 3657 break; 3658 #endif 3659 #ifdef INET6 3660 case AF_INET6: 3661 if (in6p) { 3662 in6p->in6p_laddr = ((struct sockaddr_in6 *)dst)->sin6_addr; 3663 in6p->in6p_lport = ((struct sockaddr_in6 *)dst)->sin6_port; 3664 in6_pcbstate(in6p, IN6P_BOUND); 3665 } 3666 break; 3667 #endif 3668 } 3669 #ifdef INET6 3670 if (in6p && in6totcpcb(in6p)->t_family == AF_INET6 && sotoinpcb(oso)) { 3671 struct in6pcb *oin6p = sotoin6pcb(oso); 3672 /* inherit socket options from the listening socket */ 3673 in6p->in6p_flags |= (oin6p->in6p_flags & IN6P_CONTROLOPTS); 3674 if (in6p->in6p_flags & IN6P_CONTROLOPTS) { 3675 m_freem(in6p->in6p_options); 3676 in6p->in6p_options = 0; 3677 } 3678 ip6_savecontrol(in6p, &in6p->in6p_options, 3679 mtod(m, struct ip6_hdr *), m); 3680 } 3681 #endif 3682 3683 #if defined(IPSEC) || defined(FAST_IPSEC) 3684 /* 3685 * we make a copy of policy, instead of sharing the policy, 3686 * for better behavior in terms of SA lookup and dead SA removal. 3687 */ 3688 if (inp) { 3689 /* copy old policy into new socket's */ 3690 if (ipsec_copy_pcbpolicy(sotoinpcb(oso)->inp_sp, inp->inp_sp)) 3691 printf("tcp_input: could not copy policy\n"); 3692 } 3693 #ifdef INET6 3694 else if (in6p) { 3695 /* copy old policy into new socket's */ 3696 if (ipsec_copy_pcbpolicy(sotoin6pcb(oso)->in6p_sp, 3697 in6p->in6p_sp)) 3698 printf("tcp_input: could not copy policy\n"); 3699 } 3700 #endif 3701 #endif 3702 3703 /* 3704 * Give the new socket our cached route reference. 3705 */ 3706 if (inp) { 3707 rtcache_copy(&inp->inp_route, &sc->sc_route); 3708 rtcache_free(&sc->sc_route); 3709 } 3710 #ifdef INET6 3711 else { 3712 rtcache_copy(&in6p->in6p_route, &sc->sc_route); 3713 rtcache_free(&sc->sc_route); 3714 } 3715 #endif 3716 3717 am = m_get(M_DONTWAIT, MT_SONAME); /* XXX */ 3718 if (am == NULL) 3719 goto resetandabort; 3720 MCLAIM(am, &tcp_mowner); 3721 am->m_len = src->sa_len; 3722 bcopy(src, mtod(am, void *), src->sa_len); 3723 if (inp) { 3724 if (in_pcbconnect(inp, am, NULL)) { 3725 (void) m_free(am); 3726 goto resetandabort; 3727 } 3728 } 3729 #ifdef INET6 3730 else if (in6p) { 3731 if (src->sa_family == AF_INET) { 3732 /* IPv4 packet to AF_INET6 socket */ 3733 struct sockaddr_in6 *sin6; 3734 sin6 = mtod(am, struct sockaddr_in6 *); 3735 am->m_len = sizeof(*sin6); 3736 bzero(sin6, sizeof(*sin6)); 3737 sin6->sin6_family = AF_INET6; 3738 sin6->sin6_len = sizeof(*sin6); 3739 sin6->sin6_port = ((struct sockaddr_in *)src)->sin_port; 3740 sin6->sin6_addr.s6_addr16[5] = htons(0xffff); 3741 bcopy(&((struct sockaddr_in *)src)->sin_addr, 3742 &sin6->sin6_addr.s6_addr32[3], 3743 sizeof(sin6->sin6_addr.s6_addr32[3])); 3744 } 3745 if (in6_pcbconnect(in6p, am, NULL)) { 3746 (void) m_free(am); 3747 goto resetandabort; 3748 } 3749 } 3750 #endif 3751 else { 3752 (void) m_free(am); 3753 goto resetandabort; 3754 } 3755 (void) m_free(am); 3756 3757 if (inp) 3758 tp = intotcpcb(inp); 3759 #ifdef INET6 3760 else if (in6p) 3761 tp = in6totcpcb(in6p); 3762 #endif 3763 else 3764 tp = NULL; 3765 tp->t_flags = sototcpcb(oso)->t_flags & TF_NODELAY; 3766 if (sc->sc_request_r_scale != 15) { 3767 tp->requested_s_scale = sc->sc_requested_s_scale; 3768 tp->request_r_scale = sc->sc_request_r_scale; 3769 tp->snd_scale = sc->sc_requested_s_scale; 3770 tp->rcv_scale = sc->sc_request_r_scale; 3771 tp->t_flags |= TF_REQ_SCALE|TF_RCVD_SCALE; 3772 } 3773 if (sc->sc_flags & SCF_TIMESTAMP) 3774 tp->t_flags |= TF_REQ_TSTMP|TF_RCVD_TSTMP; 3775 tp->ts_timebase = sc->sc_timebase; 3776 3777 tp->t_template = tcp_template(tp); 3778 if (tp->t_template == 0) { 3779 tp = tcp_drop(tp, ENOBUFS); /* destroys socket */ 3780 so = NULL; 3781 m_freem(m); 3782 goto abort; 3783 } 3784 3785 tp->iss = sc->sc_iss; 3786 tp->irs = sc->sc_irs; 3787 tcp_sendseqinit(tp); 3788 tcp_rcvseqinit(tp); 3789 tp->t_state = TCPS_SYN_RECEIVED; 3790 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit); 3791 tcpstat.tcps_accepts++; 3792 3793 if ((sc->sc_flags & SCF_SACK_PERMIT) && tcp_do_sack) 3794 tp->t_flags |= TF_WILL_SACK; 3795 3796 if ((sc->sc_flags & SCF_ECN_PERMIT) && tcp_do_ecn) 3797 tp->t_flags |= TF_ECN_PERMIT; 3798 3799 #ifdef TCP_SIGNATURE 3800 if (sc->sc_flags & SCF_SIGNATURE) 3801 tp->t_flags |= TF_SIGNATURE; 3802 #endif 3803 3804 /* Initialize tp->t_ourmss before we deal with the peer's! */ 3805 tp->t_ourmss = sc->sc_ourmaxseg; 3806 tcp_mss_from_peer(tp, sc->sc_peermaxseg); 3807 3808 /* 3809 * Initialize the initial congestion window. If we 3810 * had to retransmit the SYN,ACK, we must initialize cwnd 3811 * to 1 segment (i.e. the Loss Window). 3812 */ 3813 if (sc->sc_rxtshift) 3814 tp->snd_cwnd = tp->t_peermss; 3815 else { 3816 int ss = tcp_init_win; 3817 #ifdef INET 3818 if (inp != NULL && in_localaddr(inp->inp_faddr)) 3819 ss = tcp_init_win_local; 3820 #endif 3821 #ifdef INET6 3822 if (in6p != NULL && in6_localaddr(&in6p->in6p_faddr)) 3823 ss = tcp_init_win_local; 3824 #endif 3825 tp->snd_cwnd = TCP_INITIAL_WINDOW(ss, tp->t_peermss); 3826 } 3827 3828 tcp_rmx_rtt(tp); 3829 tp->snd_wl1 = sc->sc_irs; 3830 tp->rcv_up = sc->sc_irs + 1; 3831 3832 /* 3833 * This is what whould have happened in tcp_output() when 3834 * the SYN,ACK was sent. 3835 */ 3836 tp->snd_up = tp->snd_una; 3837 tp->snd_max = tp->snd_nxt = tp->iss+1; 3838 TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur); 3839 if (sc->sc_win > 0 && SEQ_GT(tp->rcv_nxt + sc->sc_win, tp->rcv_adv)) 3840 tp->rcv_adv = tp->rcv_nxt + sc->sc_win; 3841 tp->last_ack_sent = tp->rcv_nxt; 3842 tp->t_partialacks = -1; 3843 tp->t_dupacks = 0; 3844 3845 tcpstat.tcps_sc_completed++; 3846 s = splsoftnet(); 3847 SYN_CACHE_PUT(sc); 3848 splx(s); 3849 return (so); 3850 3851 resetandabort: 3852 (void)tcp_respond(NULL, m, m, th, (tcp_seq)0, th->th_ack, TH_RST); 3853 abort: 3854 if (so != NULL) 3855 (void) soabort(so); 3856 s = splsoftnet(); 3857 SYN_CACHE_PUT(sc); 3858 splx(s); 3859 tcpstat.tcps_sc_aborted++; 3860 return ((struct socket *)(-1)); 3861 } 3862 3863 /* 3864 * This function is called when we get a RST for a 3865 * non-existent connection, so that we can see if the 3866 * connection is in the syn cache. If it is, zap it. 3867 */ 3868 3869 void 3870 syn_cache_reset(struct sockaddr *src, struct sockaddr *dst, struct tcphdr *th) 3871 { 3872 struct syn_cache *sc; 3873 struct syn_cache_head *scp; 3874 int s = splsoftnet(); 3875 3876 if ((sc = syn_cache_lookup(src, dst, &scp)) == NULL) { 3877 splx(s); 3878 return; 3879 } 3880 if (SEQ_LT(th->th_seq, sc->sc_irs) || 3881 SEQ_GT(th->th_seq, sc->sc_irs+1)) { 3882 splx(s); 3883 return; 3884 } 3885 SYN_CACHE_RM(sc); 3886 tcpstat.tcps_sc_reset++; 3887 SYN_CACHE_PUT(sc); /* calls pool_put but see spl above */ 3888 splx(s); 3889 } 3890 3891 void 3892 syn_cache_unreach(const struct sockaddr *src, const struct sockaddr *dst, 3893 struct tcphdr *th) 3894 { 3895 struct syn_cache *sc; 3896 struct syn_cache_head *scp; 3897 int s; 3898 3899 s = splsoftnet(); 3900 if ((sc = syn_cache_lookup(src, dst, &scp)) == NULL) { 3901 splx(s); 3902 return; 3903 } 3904 /* If the sequence number != sc_iss, then it's a bogus ICMP msg */ 3905 if (ntohl (th->th_seq) != sc->sc_iss) { 3906 splx(s); 3907 return; 3908 } 3909 3910 /* 3911 * If we've retransmitted 3 times and this is our second error, 3912 * we remove the entry. Otherwise, we allow it to continue on. 3913 * This prevents us from incorrectly nuking an entry during a 3914 * spurious network outage. 3915 * 3916 * See tcp_notify(). 3917 */ 3918 if ((sc->sc_flags & SCF_UNREACH) == 0 || sc->sc_rxtshift < 3) { 3919 sc->sc_flags |= SCF_UNREACH; 3920 splx(s); 3921 return; 3922 } 3923 3924 SYN_CACHE_RM(sc); 3925 tcpstat.tcps_sc_unreach++; 3926 SYN_CACHE_PUT(sc); /* calls pool_put but see spl above */ 3927 splx(s); 3928 } 3929 3930 /* 3931 * Given a LISTEN socket and an inbound SYN request, add 3932 * this to the syn cache, and send back a segment: 3933 * <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK> 3934 * to the source. 3935 * 3936 * IMPORTANT NOTE: We do _NOT_ ACK data that might accompany the SYN. 3937 * Doing so would require that we hold onto the data and deliver it 3938 * to the application. However, if we are the target of a SYN-flood 3939 * DoS attack, an attacker could send data which would eventually 3940 * consume all available buffer space if it were ACKed. By not ACKing 3941 * the data, we avoid this DoS scenario. 3942 */ 3943 3944 int 3945 syn_cache_add(struct sockaddr *src, struct sockaddr *dst, struct tcphdr *th, 3946 unsigned int hlen, struct socket *so, struct mbuf *m, u_char *optp, 3947 int optlen, struct tcp_opt_info *oi) 3948 { 3949 struct tcpcb tb, *tp; 3950 long win; 3951 struct syn_cache *sc; 3952 struct syn_cache_head *scp; 3953 struct mbuf *ipopts; 3954 struct tcp_opt_info opti; 3955 int s; 3956 3957 tp = sototcpcb(so); 3958 3959 bzero(&opti, sizeof(opti)); 3960 3961 /* 3962 * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN 3963 * 3964 * Note this check is performed in tcp_input() very early on. 3965 */ 3966 3967 /* 3968 * Initialize some local state. 3969 */ 3970 win = sbspace(&so->so_rcv); 3971 if (win > TCP_MAXWIN) 3972 win = TCP_MAXWIN; 3973 3974 switch (src->sa_family) { 3975 #ifdef INET 3976 case AF_INET: 3977 /* 3978 * Remember the IP options, if any. 3979 */ 3980 ipopts = ip_srcroute(); 3981 break; 3982 #endif 3983 default: 3984 ipopts = NULL; 3985 } 3986 3987 #ifdef TCP_SIGNATURE 3988 if (optp || (tp->t_flags & TF_SIGNATURE)) 3989 #else 3990 if (optp) 3991 #endif 3992 { 3993 tb.t_flags = tcp_do_rfc1323 ? (TF_REQ_SCALE|TF_REQ_TSTMP) : 0; 3994 #ifdef TCP_SIGNATURE 3995 tb.t_flags |= (tp->t_flags & TF_SIGNATURE); 3996 #endif 3997 tb.t_state = TCPS_LISTEN; 3998 if (tcp_dooptions(&tb, optp, optlen, th, m, m->m_pkthdr.len - 3999 sizeof(struct tcphdr) - optlen - hlen, oi) < 0) 4000 return (0); 4001 } else 4002 tb.t_flags = 0; 4003 4004 /* 4005 * See if we already have an entry for this connection. 4006 * If we do, resend the SYN,ACK. We do not count this 4007 * as a retransmission (XXX though maybe we should). 4008 */ 4009 if ((sc = syn_cache_lookup(src, dst, &scp)) != NULL) { 4010 tcpstat.tcps_sc_dupesyn++; 4011 if (ipopts) { 4012 /* 4013 * If we were remembering a previous source route, 4014 * forget it and use the new one we've been given. 4015 */ 4016 if (sc->sc_ipopts) 4017 (void) m_free(sc->sc_ipopts); 4018 sc->sc_ipopts = ipopts; 4019 } 4020 sc->sc_timestamp = tb.ts_recent; 4021 if (syn_cache_respond(sc, m) == 0) { 4022 tcpstat.tcps_sndacks++; 4023 tcpstat.tcps_sndtotal++; 4024 } 4025 return (1); 4026 } 4027 4028 s = splsoftnet(); 4029 sc = pool_get(&syn_cache_pool, PR_NOWAIT); 4030 splx(s); 4031 if (sc == NULL) { 4032 if (ipopts) 4033 (void) m_free(ipopts); 4034 return (0); 4035 } 4036 4037 /* 4038 * Fill in the cache, and put the necessary IP and TCP 4039 * options into the reply. 4040 */ 4041 bzero(sc, sizeof(struct syn_cache)); 4042 callout_init(&sc->sc_timer, 0); 4043 bcopy(src, &sc->sc_src, src->sa_len); 4044 bcopy(dst, &sc->sc_dst, dst->sa_len); 4045 sc->sc_flags = 0; 4046 sc->sc_ipopts = ipopts; 4047 sc->sc_irs = th->th_seq; 4048 switch (src->sa_family) { 4049 #ifdef INET 4050 case AF_INET: 4051 { 4052 struct sockaddr_in *srcin = (void *) src; 4053 struct sockaddr_in *dstin = (void *) dst; 4054 4055 sc->sc_iss = tcp_new_iss1(&dstin->sin_addr, 4056 &srcin->sin_addr, dstin->sin_port, 4057 srcin->sin_port, sizeof(dstin->sin_addr), 0); 4058 break; 4059 } 4060 #endif /* INET */ 4061 #ifdef INET6 4062 case AF_INET6: 4063 { 4064 struct sockaddr_in6 *srcin6 = (void *) src; 4065 struct sockaddr_in6 *dstin6 = (void *) dst; 4066 4067 sc->sc_iss = tcp_new_iss1(&dstin6->sin6_addr, 4068 &srcin6->sin6_addr, dstin6->sin6_port, 4069 srcin6->sin6_port, sizeof(dstin6->sin6_addr), 0); 4070 break; 4071 } 4072 #endif /* INET6 */ 4073 } 4074 sc->sc_peermaxseg = oi->maxseg; 4075 sc->sc_ourmaxseg = tcp_mss_to_advertise(m->m_flags & M_PKTHDR ? 4076 m->m_pkthdr.rcvif : NULL, 4077 sc->sc_src.sa.sa_family); 4078 sc->sc_win = win; 4079 sc->sc_timebase = tcp_now; /* see tcp_newtcpcb() */ 4080 sc->sc_timestamp = tb.ts_recent; 4081 if ((tb.t_flags & (TF_REQ_TSTMP|TF_RCVD_TSTMP)) == 4082 (TF_REQ_TSTMP|TF_RCVD_TSTMP)) 4083 sc->sc_flags |= SCF_TIMESTAMP; 4084 if ((tb.t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == 4085 (TF_RCVD_SCALE|TF_REQ_SCALE)) { 4086 sc->sc_requested_s_scale = tb.requested_s_scale; 4087 sc->sc_request_r_scale = 0; 4088 /* 4089 * Compute proper scaling value from buffer space. 4090 * Leave enough room for the socket buffer to grow 4091 * with auto sizing. This allows us to scale the 4092 * receive buffer over a wide range while not losing 4093 * any efficiency or fine granularity. 4094 * 4095 * RFC1323: The Window field in a SYN (i.e., a <SYN> 4096 * or <SYN,ACK>) segment itself is never scaled. 4097 */ 4098 while (sc->sc_request_r_scale < TCP_MAX_WINSHIFT && 4099 (0x1 << sc->sc_request_r_scale) < tcp_minmss) 4100 sc->sc_request_r_scale++; 4101 } else { 4102 sc->sc_requested_s_scale = 15; 4103 sc->sc_request_r_scale = 15; 4104 } 4105 if ((tb.t_flags & TF_SACK_PERMIT) && tcp_do_sack) 4106 sc->sc_flags |= SCF_SACK_PERMIT; 4107 4108 /* 4109 * ECN setup packet recieved. 4110 */ 4111 if ((th->th_flags & (TH_ECE|TH_CWR)) && tcp_do_ecn) 4112 sc->sc_flags |= SCF_ECN_PERMIT; 4113 4114 #ifdef TCP_SIGNATURE 4115 if (tb.t_flags & TF_SIGNATURE) 4116 sc->sc_flags |= SCF_SIGNATURE; 4117 #endif 4118 sc->sc_tp = tp; 4119 if (syn_cache_respond(sc, m) == 0) { 4120 syn_cache_insert(sc, tp); 4121 tcpstat.tcps_sndacks++; 4122 tcpstat.tcps_sndtotal++; 4123 } else { 4124 s = splsoftnet(); 4125 SYN_CACHE_PUT(sc); 4126 splx(s); 4127 tcpstat.tcps_sc_dropped++; 4128 } 4129 return (1); 4130 } 4131 4132 int 4133 syn_cache_respond(struct syn_cache *sc, struct mbuf *m) 4134 { 4135 struct route *ro; 4136 u_int8_t *optp; 4137 int optlen, error; 4138 u_int16_t tlen; 4139 struct ip *ip = NULL; 4140 #ifdef INET6 4141 struct ip6_hdr *ip6 = NULL; 4142 #endif 4143 struct tcpcb *tp = NULL; 4144 struct tcphdr *th; 4145 u_int hlen; 4146 struct socket *so; 4147 4148 ro = &sc->sc_route; 4149 switch (sc->sc_src.sa.sa_family) { 4150 case AF_INET: 4151 hlen = sizeof(struct ip); 4152 break; 4153 #ifdef INET6 4154 case AF_INET6: 4155 hlen = sizeof(struct ip6_hdr); 4156 break; 4157 #endif 4158 default: 4159 if (m) 4160 m_freem(m); 4161 return (EAFNOSUPPORT); 4162 } 4163 4164 /* Compute the size of the TCP options. */ 4165 optlen = 4 + (sc->sc_request_r_scale != 15 ? 4 : 0) + 4166 ((sc->sc_flags & SCF_SACK_PERMIT) ? (TCPOLEN_SACK_PERMITTED + 2) : 0) + 4167 #ifdef TCP_SIGNATURE 4168 ((sc->sc_flags & SCF_SIGNATURE) ? (TCPOLEN_SIGNATURE + 2) : 0) + 4169 #endif 4170 ((sc->sc_flags & SCF_TIMESTAMP) ? TCPOLEN_TSTAMP_APPA : 0); 4171 4172 tlen = hlen + sizeof(struct tcphdr) + optlen; 4173 4174 /* 4175 * Create the IP+TCP header from scratch. 4176 */ 4177 if (m) 4178 m_freem(m); 4179 #ifdef DIAGNOSTIC 4180 if (max_linkhdr + tlen > MCLBYTES) 4181 return (ENOBUFS); 4182 #endif 4183 MGETHDR(m, M_DONTWAIT, MT_DATA); 4184 if (m && tlen > MHLEN) { 4185 MCLGET(m, M_DONTWAIT); 4186 if ((m->m_flags & M_EXT) == 0) { 4187 m_freem(m); 4188 m = NULL; 4189 } 4190 } 4191 if (m == NULL) 4192 return (ENOBUFS); 4193 MCLAIM(m, &tcp_tx_mowner); 4194 4195 /* Fixup the mbuf. */ 4196 m->m_data += max_linkhdr; 4197 m->m_len = m->m_pkthdr.len = tlen; 4198 if (sc->sc_tp) { 4199 tp = sc->sc_tp; 4200 if (tp->t_inpcb) 4201 so = tp->t_inpcb->inp_socket; 4202 #ifdef INET6 4203 else if (tp->t_in6pcb) 4204 so = tp->t_in6pcb->in6p_socket; 4205 #endif 4206 else 4207 so = NULL; 4208 } else 4209 so = NULL; 4210 m->m_pkthdr.rcvif = NULL; 4211 memset(mtod(m, u_char *), 0, tlen); 4212 4213 switch (sc->sc_src.sa.sa_family) { 4214 case AF_INET: 4215 ip = mtod(m, struct ip *); 4216 ip->ip_v = 4; 4217 ip->ip_dst = sc->sc_src.sin.sin_addr; 4218 ip->ip_src = sc->sc_dst.sin.sin_addr; 4219 ip->ip_p = IPPROTO_TCP; 4220 th = (struct tcphdr *)(ip + 1); 4221 th->th_dport = sc->sc_src.sin.sin_port; 4222 th->th_sport = sc->sc_dst.sin.sin_port; 4223 break; 4224 #ifdef INET6 4225 case AF_INET6: 4226 ip6 = mtod(m, struct ip6_hdr *); 4227 ip6->ip6_vfc = IPV6_VERSION; 4228 ip6->ip6_dst = sc->sc_src.sin6.sin6_addr; 4229 ip6->ip6_src = sc->sc_dst.sin6.sin6_addr; 4230 ip6->ip6_nxt = IPPROTO_TCP; 4231 /* ip6_plen will be updated in ip6_output() */ 4232 th = (struct tcphdr *)(ip6 + 1); 4233 th->th_dport = sc->sc_src.sin6.sin6_port; 4234 th->th_sport = sc->sc_dst.sin6.sin6_port; 4235 break; 4236 #endif 4237 default: 4238 th = NULL; 4239 } 4240 4241 th->th_seq = htonl(sc->sc_iss); 4242 th->th_ack = htonl(sc->sc_irs + 1); 4243 th->th_off = (sizeof(struct tcphdr) + optlen) >> 2; 4244 th->th_flags = TH_SYN|TH_ACK; 4245 th->th_win = htons(sc->sc_win); 4246 /* th_sum already 0 */ 4247 /* th_urp already 0 */ 4248 4249 /* Tack on the TCP options. */ 4250 optp = (u_int8_t *)(th + 1); 4251 *optp++ = TCPOPT_MAXSEG; 4252 *optp++ = 4; 4253 *optp++ = (sc->sc_ourmaxseg >> 8) & 0xff; 4254 *optp++ = sc->sc_ourmaxseg & 0xff; 4255 4256 if (sc->sc_request_r_scale != 15) { 4257 *((u_int32_t *)optp) = htonl(TCPOPT_NOP << 24 | 4258 TCPOPT_WINDOW << 16 | TCPOLEN_WINDOW << 8 | 4259 sc->sc_request_r_scale); 4260 optp += 4; 4261 } 4262 4263 if (sc->sc_flags & SCF_TIMESTAMP) { 4264 u_int32_t *lp = (u_int32_t *)(optp); 4265 /* Form timestamp option as shown in appendix A of RFC 1323. */ 4266 *lp++ = htonl(TCPOPT_TSTAMP_HDR); 4267 *lp++ = htonl(SYN_CACHE_TIMESTAMP(sc)); 4268 *lp = htonl(sc->sc_timestamp); 4269 optp += TCPOLEN_TSTAMP_APPA; 4270 } 4271 4272 if (sc->sc_flags & SCF_SACK_PERMIT) { 4273 u_int8_t *p = optp; 4274 4275 /* Let the peer know that we will SACK. */ 4276 p[0] = TCPOPT_SACK_PERMITTED; 4277 p[1] = 2; 4278 p[2] = TCPOPT_NOP; 4279 p[3] = TCPOPT_NOP; 4280 optp += 4; 4281 } 4282 4283 /* 4284 * Send ECN SYN-ACK setup packet. 4285 * Routes can be asymetric, so, even if we receive a packet 4286 * with ECE and CWR set, we must not assume no one will block 4287 * the ECE packet we are about to send. 4288 */ 4289 if ((sc->sc_flags & SCF_ECN_PERMIT) && tp && 4290 SEQ_GEQ(tp->snd_nxt, tp->snd_max)) { 4291 th->th_flags |= TH_ECE; 4292 tcpstat.tcps_ecn_shs++; 4293 4294 /* 4295 * draft-ietf-tcpm-ecnsyn-00.txt 4296 * 4297 * "[...] a TCP node MAY respond to an ECN-setup 4298 * SYN packet by setting ECT in the responding 4299 * ECN-setup SYN/ACK packet, indicating to routers 4300 * that the SYN/ACK packet is ECN-Capable. 4301 * This allows a congested router along the path 4302 * to mark the packet instead of dropping the 4303 * packet as an indication of congestion." 4304 * 4305 * "[...] There can be a great benefit in setting 4306 * an ECN-capable codepoint in SYN/ACK packets [...] 4307 * Congestion is most likely to occur in 4308 * the server-to-client direction. As a result, 4309 * setting an ECN-capable codepoint in SYN/ACK 4310 * packets can reduce the occurence of three-second 4311 * retransmit timeouts resulting from the drop 4312 * of SYN/ACK packets." 4313 * 4314 * Page 4 and 6, January 2006. 4315 */ 4316 4317 switch (sc->sc_src.sa.sa_family) { 4318 #ifdef INET 4319 case AF_INET: 4320 ip->ip_tos |= IPTOS_ECN_ECT0; 4321 break; 4322 #endif 4323 #ifdef INET6 4324 case AF_INET6: 4325 ip6->ip6_flow |= htonl(IPTOS_ECN_ECT0 << 20); 4326 break; 4327 #endif 4328 } 4329 tcpstat.tcps_ecn_ect++; 4330 } 4331 4332 #ifdef TCP_SIGNATURE 4333 if (sc->sc_flags & SCF_SIGNATURE) { 4334 struct secasvar *sav; 4335 u_int8_t *sigp; 4336 4337 sav = tcp_signature_getsav(m, th); 4338 4339 if (sav == NULL) { 4340 if (m) 4341 m_freem(m); 4342 return (EPERM); 4343 } 4344 4345 *optp++ = TCPOPT_SIGNATURE; 4346 *optp++ = TCPOLEN_SIGNATURE; 4347 sigp = optp; 4348 bzero(optp, TCP_SIGLEN); 4349 optp += TCP_SIGLEN; 4350 *optp++ = TCPOPT_NOP; 4351 *optp++ = TCPOPT_EOL; 4352 4353 (void)tcp_signature(m, th, hlen, sav, sigp); 4354 4355 key_sa_recordxfer(sav, m); 4356 #ifdef FAST_IPSEC 4357 KEY_FREESAV(&sav); 4358 #else 4359 key_freesav(sav); 4360 #endif 4361 } 4362 #endif 4363 4364 /* Compute the packet's checksum. */ 4365 switch (sc->sc_src.sa.sa_family) { 4366 case AF_INET: 4367 ip->ip_len = htons(tlen - hlen); 4368 th->th_sum = 0; 4369 th->th_sum = in4_cksum(m, IPPROTO_TCP, hlen, tlen - hlen); 4370 break; 4371 #ifdef INET6 4372 case AF_INET6: 4373 ip6->ip6_plen = htons(tlen - hlen); 4374 th->th_sum = 0; 4375 th->th_sum = in6_cksum(m, IPPROTO_TCP, hlen, tlen - hlen); 4376 break; 4377 #endif 4378 } 4379 4380 /* 4381 * Fill in some straggling IP bits. Note the stack expects 4382 * ip_len to be in host order, for convenience. 4383 */ 4384 switch (sc->sc_src.sa.sa_family) { 4385 #ifdef INET 4386 case AF_INET: 4387 ip->ip_len = htons(tlen); 4388 ip->ip_ttl = ip_defttl; 4389 /* XXX tos? */ 4390 break; 4391 #endif 4392 #ifdef INET6 4393 case AF_INET6: 4394 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 4395 ip6->ip6_vfc |= IPV6_VERSION; 4396 ip6->ip6_plen = htons(tlen - hlen); 4397 /* ip6_hlim will be initialized afterwards */ 4398 /* XXX flowlabel? */ 4399 break; 4400 #endif 4401 } 4402 4403 /* XXX use IPsec policy on listening socket, on SYN ACK */ 4404 tp = sc->sc_tp; 4405 4406 switch (sc->sc_src.sa.sa_family) { 4407 #ifdef INET 4408 case AF_INET: 4409 error = ip_output(m, sc->sc_ipopts, ro, 4410 (ip_mtudisc ? IP_MTUDISC : 0), 4411 (struct ip_moptions *)NULL, so); 4412 break; 4413 #endif 4414 #ifdef INET6 4415 case AF_INET6: 4416 ip6->ip6_hlim = in6_selecthlim(NULL, 4417 ro->ro_rt ? ro->ro_rt->rt_ifp : NULL); 4418 4419 error = ip6_output(m, NULL /*XXX*/, ro, 0, NULL, so, NULL); 4420 break; 4421 #endif 4422 default: 4423 error = EAFNOSUPPORT; 4424 break; 4425 } 4426 return (error); 4427 } 4428
Indexes created Mon Sep 22 13:09:51 GMT 2025