tcp_input.c revision 1.403
1 /* $NetBSD: tcp_input.c,v 1.403 2018/03/30 08:25:06 maxv 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, 74 * 2011 The NetBSD Foundation, Inc. 75 * All rights reserved. 76 * 77 * This code is derived from software contributed to The NetBSD Foundation 78 * by Coyote Point Systems, Inc. 79 * This code is derived from software contributed to The NetBSD Foundation 80 * by Jason R. Thorpe and Kevin M. Lahey of the Numerical Aerospace Simulation 81 * Facility, NASA Ames Research Center. 82 * This code is derived from software contributed to The NetBSD Foundation 83 * by Charles M. Hannum. 84 * This code is derived from software contributed to The NetBSD Foundation 85 * by Rui Paulo. 86 * 87 * Redistribution and use in source and binary forms, with or without 88 * modification, are permitted provided that the following conditions 89 * are met: 90 * 1. Redistributions of source code must retain the above copyright 91 * notice, this list of conditions and the following disclaimer. 92 * 2. Redistributions in binary form must reproduce the above copyright 93 * notice, this list of conditions and the following disclaimer in the 94 * documentation and/or other materials provided with the distribution. 95 * 96 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 97 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 98 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 99 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 100 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 101 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 102 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 103 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 104 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 105 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 106 * POSSIBILITY OF SUCH DAMAGE. 107 */ 108 109 /* 110 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995 111 * The Regents of the University of California. All rights reserved. 112 * 113 * Redistribution and use in source and binary forms, with or without 114 * modification, are permitted provided that the following conditions 115 * are met: 116 * 1. Redistributions of source code must retain the above copyright 117 * notice, this list of conditions and the following disclaimer. 118 * 2. Redistributions in binary form must reproduce the above copyright 119 * notice, this list of conditions and the following disclaimer in the 120 * documentation and/or other materials provided with the distribution. 121 * 3. Neither the name of the University nor the names of its contributors 122 * may be used to endorse or promote products derived from this software 123 * without specific prior written permission. 124 * 125 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 126 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 127 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 128 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 129 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 130 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 131 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 132 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 133 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 134 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 135 * SUCH DAMAGE. 136 * 137 * @(#)tcp_input.c 8.12 (Berkeley) 5/24/95 138 */ 139 140 /* 141 * TODO list for SYN cache stuff: 142 * 143 * Find room for a "state" field, which is needed to keep a 144 * compressed state for TIME_WAIT TCBs. It's been noted already 145 * that this is fairly important for very high-volume web and 146 * mail servers, which use a large number of short-lived 147 * connections. 148 */ 149 150 #include <sys/cdefs.h> 151 __KERNEL_RCSID(0, "$NetBSD: tcp_input.c,v 1.403 2018/03/30 08:25:06 maxv Exp $"); 152 153 #ifdef _KERNEL_OPT 154 #include "opt_inet.h" 155 #include "opt_ipsec.h" 156 #include "opt_inet_csum.h" 157 #include "opt_tcp_debug.h" 158 #endif 159 160 #include <sys/param.h> 161 #include <sys/systm.h> 162 #include <sys/malloc.h> 163 #include <sys/mbuf.h> 164 #include <sys/protosw.h> 165 #include <sys/socket.h> 166 #include <sys/socketvar.h> 167 #include <sys/errno.h> 168 #include <sys/syslog.h> 169 #include <sys/pool.h> 170 #include <sys/domain.h> 171 #include <sys/kernel.h> 172 #ifdef TCP_SIGNATURE 173 #include <sys/md5.h> 174 #endif 175 #include <sys/lwp.h> /* for lwp0 */ 176 #include <sys/cprng.h> 177 178 #include <net/if.h> 179 #include <net/if_types.h> 180 181 #include <netinet/in.h> 182 #include <netinet/in_systm.h> 183 #include <netinet/ip.h> 184 #include <netinet/in_pcb.h> 185 #include <netinet/in_var.h> 186 #include <netinet/ip_var.h> 187 #include <netinet/in_offload.h> 188 189 #ifdef INET6 190 #include <netinet/ip6.h> 191 #include <netinet6/ip6_var.h> 192 #include <netinet6/in6_pcb.h> 193 #include <netinet6/ip6_var.h> 194 #include <netinet6/in6_var.h> 195 #include <netinet/icmp6.h> 196 #include <netinet6/nd6.h> 197 #ifdef TCP_SIGNATURE 198 #include <netinet6/scope6_var.h> 199 #endif 200 #endif 201 202 #ifndef INET6 203 /* always need ip6.h for IP6_EXTHDR_GET */ 204 #include <netinet/ip6.h> 205 #endif 206 207 #include <netinet/tcp.h> 208 #include <netinet/tcp_fsm.h> 209 #include <netinet/tcp_seq.h> 210 #include <netinet/tcp_timer.h> 211 #include <netinet/tcp_var.h> 212 #include <netinet/tcp_private.h> 213 #include <netinet/tcpip.h> 214 #include <netinet/tcp_congctl.h> 215 #include <netinet/tcp_debug.h> 216 217 #ifdef INET6 218 #include "faith.h" 219 #if defined(NFAITH) && NFAITH > 0 220 #include <net/if_faith.h> 221 #endif 222 #endif 223 224 #ifdef IPSEC 225 #include <netipsec/ipsec.h> 226 #include <netipsec/ipsec_var.h> 227 #include <netipsec/key.h> 228 #ifdef INET6 229 #include <netipsec/ipsec6.h> 230 #endif 231 #endif /* IPSEC*/ 232 233 #include <netinet/tcp_vtw.h> 234 235 int tcprexmtthresh = 3; 236 int tcp_log_refused; 237 238 int tcp_do_autorcvbuf = 1; 239 int tcp_autorcvbuf_inc = 16 * 1024; 240 int tcp_autorcvbuf_max = 256 * 1024; 241 int tcp_msl = (TCPTV_MSL / PR_SLOWHZ); 242 243 static int tcp_rst_ppslim_count = 0; 244 static struct timeval tcp_rst_ppslim_last; 245 static int tcp_ackdrop_ppslim_count = 0; 246 static struct timeval tcp_ackdrop_ppslim_last; 247 248 static void syn_cache_timer(void *); 249 250 #define TCP_PAWS_IDLE (24U * 24 * 60 * 60 * PR_SLOWHZ) 251 252 /* for modulo comparisons of timestamps */ 253 #define TSTMP_LT(a,b) ((int)((a)-(b)) < 0) 254 #define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0) 255 256 /* 257 * Neighbor Discovery, Neighbor Unreachability Detection Upper layer hint. 258 */ 259 #ifdef INET6 260 static inline void 261 nd6_hint(struct tcpcb *tp) 262 { 263 struct rtentry *rt = NULL; 264 265 if (tp != NULL && tp->t_in6pcb != NULL && tp->t_family == AF_INET6 && 266 (rt = rtcache_validate(&tp->t_in6pcb->in6p_route)) != NULL) 267 nd6_nud_hint(rt); 268 rtcache_unref(rt, &tp->t_in6pcb->in6p_route); 269 } 270 #else 271 static inline void 272 nd6_hint(struct tcpcb *tp) 273 { 274 } 275 #endif 276 277 /* 278 * Compute ACK transmission behavior. Delay the ACK unless 279 * we have already delayed an ACK (must send an ACK every two segments). 280 * We also ACK immediately if we received a PUSH and the ACK-on-PUSH 281 * option is enabled. 282 */ 283 static void 284 tcp_setup_ack(struct tcpcb *tp, const struct tcphdr *th) 285 { 286 287 if (tp->t_flags & TF_DELACK || 288 (tcp_ack_on_push && th->th_flags & TH_PUSH)) 289 tp->t_flags |= TF_ACKNOW; 290 else 291 TCP_SET_DELACK(tp); 292 } 293 294 static void 295 icmp_check(struct tcpcb *tp, const struct tcphdr *th, int acked) 296 { 297 298 /* 299 * If we had a pending ICMP message that refers to data that have 300 * just been acknowledged, disregard the recorded ICMP message. 301 */ 302 if ((tp->t_flags & TF_PMTUD_PEND) && 303 SEQ_GT(th->th_ack, tp->t_pmtud_th_seq)) 304 tp->t_flags &= ~TF_PMTUD_PEND; 305 306 /* 307 * Keep track of the largest chunk of data 308 * acknowledged since last PMTU update 309 */ 310 if (tp->t_pmtud_mss_acked < acked) 311 tp->t_pmtud_mss_acked = acked; 312 } 313 314 /* 315 * Convert TCP protocol fields to host order for easier processing. 316 */ 317 static void 318 tcp_fields_to_host(struct tcphdr *th) 319 { 320 321 NTOHL(th->th_seq); 322 NTOHL(th->th_ack); 323 NTOHS(th->th_win); 324 NTOHS(th->th_urp); 325 } 326 327 /* 328 * ... and reverse the above. 329 */ 330 static void 331 tcp_fields_to_net(struct tcphdr *th) 332 { 333 334 HTONL(th->th_seq); 335 HTONL(th->th_ack); 336 HTONS(th->th_win); 337 HTONS(th->th_urp); 338 } 339 340 static void 341 tcp_urp_drop(struct tcphdr *th, int todrop, int *tiflags) 342 { 343 if (th->th_urp > todrop) { 344 th->th_urp -= todrop; 345 } else { 346 *tiflags &= ~TH_URG; 347 th->th_urp = 0; 348 } 349 } 350 351 #ifdef TCP_CSUM_COUNTERS 352 #include <sys/device.h> 353 354 extern struct evcnt tcp_hwcsum_ok; 355 extern struct evcnt tcp_hwcsum_bad; 356 extern struct evcnt tcp_hwcsum_data; 357 extern struct evcnt tcp_swcsum; 358 #if defined(INET6) 359 extern struct evcnt tcp6_hwcsum_ok; 360 extern struct evcnt tcp6_hwcsum_bad; 361 extern struct evcnt tcp6_hwcsum_data; 362 extern struct evcnt tcp6_swcsum; 363 #endif /* defined(INET6) */ 364 365 #define TCP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 366 367 #else 368 369 #define TCP_CSUM_COUNTER_INCR(ev) /* nothing */ 370 371 #endif /* TCP_CSUM_COUNTERS */ 372 373 #ifdef TCP_REASS_COUNTERS 374 #include <sys/device.h> 375 376 extern struct evcnt tcp_reass_; 377 extern struct evcnt tcp_reass_empty; 378 extern struct evcnt tcp_reass_iteration[8]; 379 extern struct evcnt tcp_reass_prependfirst; 380 extern struct evcnt tcp_reass_prepend; 381 extern struct evcnt tcp_reass_insert; 382 extern struct evcnt tcp_reass_inserttail; 383 extern struct evcnt tcp_reass_append; 384 extern struct evcnt tcp_reass_appendtail; 385 extern struct evcnt tcp_reass_overlaptail; 386 extern struct evcnt tcp_reass_overlapfront; 387 extern struct evcnt tcp_reass_segdup; 388 extern struct evcnt tcp_reass_fragdup; 389 390 #define TCP_REASS_COUNTER_INCR(ev) (ev)->ev_count++ 391 392 #else 393 394 #define TCP_REASS_COUNTER_INCR(ev) /* nothing */ 395 396 #endif /* TCP_REASS_COUNTERS */ 397 398 static int tcp_reass(struct tcpcb *, const struct tcphdr *, struct mbuf *, 399 int); 400 static int tcp_dooptions(struct tcpcb *, const u_char *, int, 401 struct tcphdr *, struct mbuf *, int, struct tcp_opt_info *); 402 403 static void tcp4_log_refused(const struct ip *, const struct tcphdr *); 404 #ifdef INET6 405 static void tcp6_log_refused(const struct ip6_hdr *, const struct tcphdr *); 406 #endif 407 408 #define TRAVERSE(x) while ((x)->m_next) (x) = (x)->m_next 409 410 #if defined(MBUFTRACE) 411 struct mowner tcp_reass_mowner = MOWNER_INIT("tcp", "reass"); 412 #endif /* defined(MBUFTRACE) */ 413 414 static struct pool tcpipqent_pool; 415 416 void 417 tcpipqent_init(void) 418 { 419 420 pool_init(&tcpipqent_pool, sizeof(struct ipqent), 0, 0, 0, "tcpipqepl", 421 NULL, IPL_VM); 422 } 423 424 struct ipqent * 425 tcpipqent_alloc(void) 426 { 427 struct ipqent *ipqe; 428 int s; 429 430 s = splvm(); 431 ipqe = pool_get(&tcpipqent_pool, PR_NOWAIT); 432 splx(s); 433 434 return ipqe; 435 } 436 437 void 438 tcpipqent_free(struct ipqent *ipqe) 439 { 440 int s; 441 442 s = splvm(); 443 pool_put(&tcpipqent_pool, ipqe); 444 splx(s); 445 } 446 447 /* 448 * Insert segment ti into reassembly queue of tcp with 449 * control block tp. Return TH_FIN if reassembly now includes 450 * a segment with FIN. 451 */ 452 static int 453 tcp_reass(struct tcpcb *tp, const struct tcphdr *th, struct mbuf *m, int tlen) 454 { 455 struct ipqent *p, *q, *nq, *tiqe = NULL; 456 struct socket *so = NULL; 457 int pkt_flags; 458 tcp_seq pkt_seq; 459 unsigned pkt_len; 460 u_long rcvpartdupbyte = 0; 461 u_long rcvoobyte; 462 #ifdef TCP_REASS_COUNTERS 463 u_int count = 0; 464 #endif 465 uint64_t *tcps; 466 467 if (tp->t_inpcb) 468 so = tp->t_inpcb->inp_socket; 469 #ifdef INET6 470 else if (tp->t_in6pcb) 471 so = tp->t_in6pcb->in6p_socket; 472 #endif 473 474 TCP_REASS_LOCK_CHECK(tp); 475 476 /* 477 * Call with th==NULL after become established to 478 * force pre-ESTABLISHED data up to user socket. 479 */ 480 if (th == NULL) 481 goto present; 482 483 m_claimm(m, &tcp_reass_mowner); 484 485 rcvoobyte = tlen; 486 /* 487 * Copy these to local variables because the TCP header gets munged 488 * while we are collapsing mbufs. 489 */ 490 pkt_seq = th->th_seq; 491 pkt_len = tlen; 492 pkt_flags = th->th_flags; 493 494 TCP_REASS_COUNTER_INCR(&tcp_reass_); 495 496 if ((p = TAILQ_LAST(&tp->segq, ipqehead)) != NULL) { 497 /* 498 * When we miss a packet, the vast majority of time we get 499 * packets that follow it in order. So optimize for that. 500 */ 501 if (pkt_seq == p->ipqe_seq + p->ipqe_len) { 502 p->ipqe_len += pkt_len; 503 p->ipqe_flags |= pkt_flags; 504 m_cat(p->ipre_mlast, m); 505 TRAVERSE(p->ipre_mlast); 506 m = NULL; 507 tiqe = p; 508 TAILQ_REMOVE(&tp->timeq, p, ipqe_timeq); 509 TCP_REASS_COUNTER_INCR(&tcp_reass_appendtail); 510 goto skip_replacement; 511 } 512 /* 513 * While we're here, if the pkt is completely beyond 514 * anything we have, just insert it at the tail. 515 */ 516 if (SEQ_GT(pkt_seq, p->ipqe_seq + p->ipqe_len)) { 517 TCP_REASS_COUNTER_INCR(&tcp_reass_inserttail); 518 goto insert_it; 519 } 520 } 521 522 q = TAILQ_FIRST(&tp->segq); 523 524 if (q != NULL) { 525 /* 526 * If this segment immediately precedes the first out-of-order 527 * block, simply slap the segment in front of it and (mostly) 528 * skip the complicated logic. 529 */ 530 if (pkt_seq + pkt_len == q->ipqe_seq) { 531 q->ipqe_seq = pkt_seq; 532 q->ipqe_len += pkt_len; 533 q->ipqe_flags |= pkt_flags; 534 m_cat(m, q->ipqe_m); 535 q->ipqe_m = m; 536 q->ipre_mlast = m; /* last mbuf may have changed */ 537 TRAVERSE(q->ipre_mlast); 538 tiqe = q; 539 TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); 540 TCP_REASS_COUNTER_INCR(&tcp_reass_prependfirst); 541 goto skip_replacement; 542 } 543 } else { 544 TCP_REASS_COUNTER_INCR(&tcp_reass_empty); 545 } 546 547 /* 548 * Find a segment which begins after this one does. 549 */ 550 for (p = NULL; q != NULL; q = nq) { 551 nq = TAILQ_NEXT(q, ipqe_q); 552 #ifdef TCP_REASS_COUNTERS 553 count++; 554 #endif 555 556 /* 557 * If the received segment is just right after this 558 * fragment, merge the two together and then check 559 * for further overlaps. 560 */ 561 if (q->ipqe_seq + q->ipqe_len == pkt_seq) { 562 pkt_len += q->ipqe_len; 563 pkt_flags |= q->ipqe_flags; 564 pkt_seq = q->ipqe_seq; 565 m_cat(q->ipre_mlast, m); 566 TRAVERSE(q->ipre_mlast); 567 m = q->ipqe_m; 568 TCP_REASS_COUNTER_INCR(&tcp_reass_append); 569 goto free_ipqe; 570 } 571 572 /* 573 * If the received segment is completely past this 574 * fragment, we need to go to the next fragment. 575 */ 576 if (SEQ_LT(q->ipqe_seq + q->ipqe_len, pkt_seq)) { 577 p = q; 578 continue; 579 } 580 581 /* 582 * If the fragment is past the received segment, 583 * it (or any following) can't be concatenated. 584 */ 585 if (SEQ_GT(q->ipqe_seq, pkt_seq + pkt_len)) { 586 TCP_REASS_COUNTER_INCR(&tcp_reass_insert); 587 break; 588 } 589 590 /* 591 * We've received all the data in this segment before. 592 * Mark it as a duplicate and return. 593 */ 594 if (SEQ_LEQ(q->ipqe_seq, pkt_seq) && 595 SEQ_GEQ(q->ipqe_seq + q->ipqe_len, pkt_seq + pkt_len)) { 596 tcps = TCP_STAT_GETREF(); 597 tcps[TCP_STAT_RCVDUPPACK]++; 598 tcps[TCP_STAT_RCVDUPBYTE] += pkt_len; 599 TCP_STAT_PUTREF(); 600 tcp_new_dsack(tp, pkt_seq, pkt_len); 601 m_freem(m); 602 if (tiqe != NULL) { 603 tcpipqent_free(tiqe); 604 } 605 TCP_REASS_COUNTER_INCR(&tcp_reass_segdup); 606 goto out; 607 } 608 609 /* 610 * Received segment completely overlaps this fragment 611 * so we drop the fragment (this keeps the temporal 612 * ordering of segments correct). 613 */ 614 if (SEQ_GEQ(q->ipqe_seq, pkt_seq) && 615 SEQ_LEQ(q->ipqe_seq + q->ipqe_len, pkt_seq + pkt_len)) { 616 rcvpartdupbyte += q->ipqe_len; 617 m_freem(q->ipqe_m); 618 TCP_REASS_COUNTER_INCR(&tcp_reass_fragdup); 619 goto free_ipqe; 620 } 621 622 /* 623 * Received segment extends past the end of the fragment. 624 * Drop the overlapping bytes, merge the fragment and 625 * segment, and treat as a longer received packet. 626 */ 627 if (SEQ_LT(q->ipqe_seq, pkt_seq) && 628 SEQ_GT(q->ipqe_seq + q->ipqe_len, pkt_seq)) { 629 int overlap = q->ipqe_seq + q->ipqe_len - pkt_seq; 630 m_adj(m, overlap); 631 rcvpartdupbyte += overlap; 632 m_cat(q->ipre_mlast, m); 633 TRAVERSE(q->ipre_mlast); 634 m = q->ipqe_m; 635 pkt_seq = q->ipqe_seq; 636 pkt_len += q->ipqe_len - overlap; 637 rcvoobyte -= overlap; 638 TCP_REASS_COUNTER_INCR(&tcp_reass_overlaptail); 639 goto free_ipqe; 640 } 641 642 /* 643 * Received segment extends past the front of the fragment. 644 * Drop the overlapping bytes on the received packet. The 645 * packet will then be concatenated with this fragment a 646 * bit later. 647 */ 648 if (SEQ_GT(q->ipqe_seq, pkt_seq) && 649 SEQ_LT(q->ipqe_seq, pkt_seq + pkt_len)) { 650 int overlap = pkt_seq + pkt_len - q->ipqe_seq; 651 m_adj(m, -overlap); 652 pkt_len -= overlap; 653 rcvpartdupbyte += overlap; 654 TCP_REASS_COUNTER_INCR(&tcp_reass_overlapfront); 655 rcvoobyte -= overlap; 656 } 657 658 /* 659 * If the received segment immediately precedes this 660 * fragment then tack the fragment onto this segment 661 * and reinsert the data. 662 */ 663 if (q->ipqe_seq == pkt_seq + pkt_len) { 664 pkt_len += q->ipqe_len; 665 pkt_flags |= q->ipqe_flags; 666 m_cat(m, q->ipqe_m); 667 TAILQ_REMOVE(&tp->segq, q, ipqe_q); 668 TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); 669 tp->t_segqlen--; 670 KASSERT(tp->t_segqlen >= 0); 671 KASSERT(tp->t_segqlen != 0 || 672 (TAILQ_EMPTY(&tp->segq) && 673 TAILQ_EMPTY(&tp->timeq))); 674 if (tiqe == NULL) { 675 tiqe = q; 676 } else { 677 tcpipqent_free(q); 678 } 679 TCP_REASS_COUNTER_INCR(&tcp_reass_prepend); 680 break; 681 } 682 683 /* 684 * If the fragment is before the segment, remember it. 685 * When this loop is terminated, p will contain the 686 * pointer to the fragment that is right before the 687 * received segment. 688 */ 689 if (SEQ_LEQ(q->ipqe_seq, pkt_seq)) 690 p = q; 691 692 continue; 693 694 /* 695 * This is a common operation. It also will allow 696 * to save doing a malloc/free in most instances. 697 */ 698 free_ipqe: 699 TAILQ_REMOVE(&tp->segq, q, ipqe_q); 700 TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); 701 tp->t_segqlen--; 702 KASSERT(tp->t_segqlen >= 0); 703 KASSERT(tp->t_segqlen != 0 || 704 (TAILQ_EMPTY(&tp->segq) && TAILQ_EMPTY(&tp->timeq))); 705 if (tiqe == NULL) { 706 tiqe = q; 707 } else { 708 tcpipqent_free(q); 709 } 710 } 711 712 #ifdef TCP_REASS_COUNTERS 713 if (count > 7) 714 TCP_REASS_COUNTER_INCR(&tcp_reass_iteration[0]); 715 else if (count > 0) 716 TCP_REASS_COUNTER_INCR(&tcp_reass_iteration[count]); 717 #endif 718 719 insert_it: 720 /* 721 * Allocate a new queue entry (block) since the received segment 722 * did not collapse onto any other out-of-order block. If it had 723 * collapsed, tiqe would not be NULL and we would be reusing it. 724 * 725 * If the allocation fails, drop the packet. 726 */ 727 if (tiqe == NULL) { 728 tiqe = tcpipqent_alloc(); 729 if (tiqe == NULL) { 730 TCP_STATINC(TCP_STAT_RCVMEMDROP); 731 m_freem(m); 732 goto out; 733 } 734 } 735 736 /* 737 * Update the counters. 738 */ 739 tp->t_rcvoopack++; 740 tcps = TCP_STAT_GETREF(); 741 tcps[TCP_STAT_RCVOOPACK]++; 742 tcps[TCP_STAT_RCVOOBYTE] += rcvoobyte; 743 if (rcvpartdupbyte) { 744 tcps[TCP_STAT_RCVPARTDUPPACK]++; 745 tcps[TCP_STAT_RCVPARTDUPBYTE] += rcvpartdupbyte; 746 } 747 TCP_STAT_PUTREF(); 748 749 /* 750 * Insert the new fragment queue entry into both queues. 751 */ 752 tiqe->ipqe_m = m; 753 tiqe->ipre_mlast = m; 754 tiqe->ipqe_seq = pkt_seq; 755 tiqe->ipqe_len = pkt_len; 756 tiqe->ipqe_flags = pkt_flags; 757 if (p == NULL) { 758 TAILQ_INSERT_HEAD(&tp->segq, tiqe, ipqe_q); 759 } else { 760 TAILQ_INSERT_AFTER(&tp->segq, p, tiqe, ipqe_q); 761 } 762 tp->t_segqlen++; 763 764 skip_replacement: 765 TAILQ_INSERT_HEAD(&tp->timeq, tiqe, ipqe_timeq); 766 767 present: 768 /* 769 * Present data to user, advancing rcv_nxt through 770 * completed sequence space. 771 */ 772 if (TCPS_HAVEESTABLISHED(tp->t_state) == 0) 773 goto out; 774 q = TAILQ_FIRST(&tp->segq); 775 if (q == NULL || q->ipqe_seq != tp->rcv_nxt) 776 goto out; 777 if (tp->t_state == TCPS_SYN_RECEIVED && q->ipqe_len) 778 goto out; 779 780 tp->rcv_nxt += q->ipqe_len; 781 pkt_flags = q->ipqe_flags & TH_FIN; 782 nd6_hint(tp); 783 784 TAILQ_REMOVE(&tp->segq, q, ipqe_q); 785 TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); 786 tp->t_segqlen--; 787 KASSERT(tp->t_segqlen >= 0); 788 KASSERT(tp->t_segqlen != 0 || 789 (TAILQ_EMPTY(&tp->segq) && TAILQ_EMPTY(&tp->timeq))); 790 if (so->so_state & SS_CANTRCVMORE) 791 m_freem(q->ipqe_m); 792 else 793 sbappendstream(&so->so_rcv, q->ipqe_m); 794 tcpipqent_free(q); 795 TCP_REASS_UNLOCK(tp); 796 sorwakeup(so); 797 return pkt_flags; 798 799 out: 800 TCP_REASS_UNLOCK(tp); 801 return 0; 802 } 803 804 #ifdef INET6 805 int 806 tcp6_input(struct mbuf **mp, int *offp, int proto) 807 { 808 struct mbuf *m = *mp; 809 810 /* 811 * draft-itojun-ipv6-tcp-to-anycast 812 * better place to put this in? 813 */ 814 if (m->m_flags & M_ANYCAST6) { 815 struct ip6_hdr *ip6; 816 if (m->m_len < sizeof(struct ip6_hdr)) { 817 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 818 TCP_STATINC(TCP_STAT_RCVSHORT); 819 return IPPROTO_DONE; 820 } 821 } 822 ip6 = mtod(m, struct ip6_hdr *); 823 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR, 824 (char *)&ip6->ip6_dst - (char *)ip6); 825 return IPPROTO_DONE; 826 } 827 828 tcp_input(m, *offp, proto); 829 return IPPROTO_DONE; 830 } 831 #endif 832 833 static void 834 tcp4_log_refused(const struct ip *ip, const struct tcphdr *th) 835 { 836 char src[INET_ADDRSTRLEN]; 837 char dst[INET_ADDRSTRLEN]; 838 839 if (ip) { 840 in_print(src, sizeof(src), &ip->ip_src); 841 in_print(dst, sizeof(dst), &ip->ip_dst); 842 } else { 843 strlcpy(src, "(unknown)", sizeof(src)); 844 strlcpy(dst, "(unknown)", sizeof(dst)); 845 } 846 log(LOG_INFO, 847 "Connection attempt to TCP %s:%d from %s:%d\n", 848 dst, ntohs(th->th_dport), 849 src, ntohs(th->th_sport)); 850 } 851 852 #ifdef INET6 853 static void 854 tcp6_log_refused(const struct ip6_hdr *ip6, const struct tcphdr *th) 855 { 856 char src[INET6_ADDRSTRLEN]; 857 char dst[INET6_ADDRSTRLEN]; 858 859 if (ip6) { 860 in6_print(src, sizeof(src), &ip6->ip6_src); 861 in6_print(dst, sizeof(dst), &ip6->ip6_dst); 862 } else { 863 strlcpy(src, "(unknown v6)", sizeof(src)); 864 strlcpy(dst, "(unknown v6)", sizeof(dst)); 865 } 866 log(LOG_INFO, 867 "Connection attempt to TCP [%s]:%d from [%s]:%d\n", 868 dst, ntohs(th->th_dport), 869 src, ntohs(th->th_sport)); 870 } 871 #endif 872 873 /* 874 * Checksum extended TCP header and data. 875 */ 876 int 877 tcp_input_checksum(int af, struct mbuf *m, const struct tcphdr *th, 878 int toff, int off, int tlen) 879 { 880 struct ifnet *rcvif; 881 int s; 882 883 /* 884 * XXX it's better to record and check if this mbuf is 885 * already checked. 886 */ 887 888 rcvif = m_get_rcvif(m, &s); 889 if (__predict_false(rcvif == NULL)) 890 goto badcsum; /* XXX */ 891 892 switch (af) { 893 case AF_INET: 894 switch (m->m_pkthdr.csum_flags & 895 ((rcvif->if_csum_flags_rx & M_CSUM_TCPv4) | 896 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { 897 case M_CSUM_TCPv4|M_CSUM_TCP_UDP_BAD: 898 TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_bad); 899 goto badcsum; 900 901 case M_CSUM_TCPv4|M_CSUM_DATA: { 902 u_int32_t hw_csum = m->m_pkthdr.csum_data; 903 904 TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_data); 905 if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) { 906 const struct ip *ip = 907 mtod(m, const struct ip *); 908 909 hw_csum = in_cksum_phdr(ip->ip_src.s_addr, 910 ip->ip_dst.s_addr, 911 htons(hw_csum + tlen + off + IPPROTO_TCP)); 912 } 913 if ((hw_csum ^ 0xffff) != 0) 914 goto badcsum; 915 break; 916 } 917 918 case M_CSUM_TCPv4: 919 /* Checksum was okay. */ 920 TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_ok); 921 break; 922 923 default: 924 /* 925 * Must compute it ourselves. Maybe skip checksum 926 * on loopback interfaces. 927 */ 928 if (__predict_true(!(rcvif->if_flags & IFF_LOOPBACK) || 929 tcp_do_loopback_cksum)) { 930 TCP_CSUM_COUNTER_INCR(&tcp_swcsum); 931 if (in4_cksum(m, IPPROTO_TCP, toff, 932 tlen + off) != 0) 933 goto badcsum; 934 } 935 break; 936 } 937 break; 938 939 #ifdef INET6 940 case AF_INET6: 941 switch (m->m_pkthdr.csum_flags & 942 ((rcvif->if_csum_flags_rx & M_CSUM_TCPv6) | 943 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { 944 case M_CSUM_TCPv6|M_CSUM_TCP_UDP_BAD: 945 TCP_CSUM_COUNTER_INCR(&tcp6_hwcsum_bad); 946 goto badcsum; 947 948 #if 0 /* notyet */ 949 case M_CSUM_TCPv6|M_CSUM_DATA: 950 #endif 951 952 case M_CSUM_TCPv6: 953 /* Checksum was okay. */ 954 TCP_CSUM_COUNTER_INCR(&tcp6_hwcsum_ok); 955 break; 956 957 default: 958 /* 959 * Must compute it ourselves. Maybe skip checksum 960 * on loopback interfaces. 961 */ 962 if (__predict_true((m->m_flags & M_LOOP) == 0 || 963 tcp_do_loopback_cksum)) { 964 TCP_CSUM_COUNTER_INCR(&tcp6_swcsum); 965 if (in6_cksum(m, IPPROTO_TCP, toff, 966 tlen + off) != 0) 967 goto badcsum; 968 } 969 } 970 break; 971 #endif /* INET6 */ 972 } 973 m_put_rcvif(rcvif, &s); 974 975 return 0; 976 977 badcsum: 978 m_put_rcvif(rcvif, &s); 979 TCP_STATINC(TCP_STAT_RCVBADSUM); 980 return -1; 981 } 982 983 /* 984 * When a packet arrives addressed to a vestigial tcpbp, we 985 * nevertheless have to respond to it per the spec. 986 * 987 * This code is duplicated from the one in tcp_input(). 988 */ 989 static void tcp_vtw_input(struct tcphdr *th, vestigial_inpcb_t *vp, 990 struct mbuf *m, int tlen) 991 { 992 int tiflags; 993 int todrop; 994 uint32_t t_flags = 0; 995 uint64_t *tcps; 996 997 tiflags = th->th_flags; 998 todrop = vp->rcv_nxt - th->th_seq; 999 1000 if (todrop > 0) { 1001 if (tiflags & TH_SYN) { 1002 tiflags &= ~TH_SYN; 1003 th->th_seq++; 1004 tcp_urp_drop(th, 1, &tiflags); 1005 todrop--; 1006 } 1007 if (todrop > tlen || 1008 (todrop == tlen && (tiflags & TH_FIN) == 0)) { 1009 /* 1010 * Any valid FIN or RST must be to the left of the 1011 * window. At this point the FIN or RST must be a 1012 * duplicate or out of sequence; drop it. 1013 */ 1014 if (tiflags & TH_RST) 1015 goto drop; 1016 tiflags &= ~(TH_FIN|TH_RST); 1017 1018 /* 1019 * Send an ACK to resynchronize and drop any data. 1020 * But keep on processing for RST or ACK. 1021 */ 1022 t_flags |= TF_ACKNOW; 1023 todrop = tlen; 1024 tcps = TCP_STAT_GETREF(); 1025 tcps[TCP_STAT_RCVDUPPACK] += 1; 1026 tcps[TCP_STAT_RCVDUPBYTE] += todrop; 1027 TCP_STAT_PUTREF(); 1028 } else if ((tiflags & TH_RST) && 1029 th->th_seq != vp->rcv_nxt) { 1030 /* 1031 * Test for reset before adjusting the sequence 1032 * number for overlapping data. 1033 */ 1034 goto dropafterack_ratelim; 1035 } else { 1036 tcps = TCP_STAT_GETREF(); 1037 tcps[TCP_STAT_RCVPARTDUPPACK] += 1; 1038 tcps[TCP_STAT_RCVPARTDUPBYTE] += todrop; 1039 TCP_STAT_PUTREF(); 1040 } 1041 1042 // tcp_new_dsack(tp, th->th_seq, todrop); 1043 // hdroptlen += todrop; /*drop from head afterwards*/ 1044 1045 th->th_seq += todrop; 1046 tlen -= todrop; 1047 tcp_urp_drop(th, todrop, &tiflags); 1048 } 1049 1050 /* 1051 * If new data are received on a connection after the 1052 * user processes are gone, then RST the other end. 1053 */ 1054 if (tlen) { 1055 TCP_STATINC(TCP_STAT_RCVAFTERCLOSE); 1056 goto dropwithreset; 1057 } 1058 1059 /* 1060 * If segment ends after window, drop trailing data 1061 * (and PUSH and FIN); if nothing left, just ACK. 1062 */ 1063 todrop = (th->th_seq + tlen) - (vp->rcv_nxt + vp->rcv_wnd); 1064 1065 if (todrop > 0) { 1066 TCP_STATINC(TCP_STAT_RCVPACKAFTERWIN); 1067 if (todrop >= tlen) { 1068 /* 1069 * The segment actually starts after the window. 1070 * th->th_seq + tlen - vp->rcv_nxt - vp->rcv_wnd >= tlen 1071 * th->th_seq - vp->rcv_nxt - vp->rcv_wnd >= 0 1072 * th->th_seq >= vp->rcv_nxt + vp->rcv_wnd 1073 */ 1074 TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, tlen); 1075 1076 /* 1077 * If a new connection request is received 1078 * while in TIME_WAIT, drop the old connection 1079 * and start over if the sequence numbers 1080 * are above the previous ones. 1081 */ 1082 if ((tiflags & TH_SYN) && 1083 SEQ_GT(th->th_seq, vp->rcv_nxt)) { 1084 /* 1085 * We only support this in the !NOFDREF case, which 1086 * is to say: not here. 1087 */ 1088 goto dropwithreset; 1089 } 1090 1091 /* 1092 * If window is closed can only take segments at 1093 * window edge, and have to drop data and PUSH from 1094 * incoming segments. Continue processing, but 1095 * remember to ack. Otherwise, drop segment 1096 * and (if not RST) ack. 1097 */ 1098 if (vp->rcv_wnd == 0 && th->th_seq == vp->rcv_nxt) { 1099 t_flags |= TF_ACKNOW; 1100 TCP_STATINC(TCP_STAT_RCVWINPROBE); 1101 } else { 1102 goto dropafterack; 1103 } 1104 } else { 1105 TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, todrop); 1106 } 1107 m_adj(m, -todrop); 1108 tlen -= todrop; 1109 tiflags &= ~(TH_PUSH|TH_FIN); 1110 } 1111 1112 if (tiflags & TH_RST) { 1113 if (th->th_seq != vp->rcv_nxt) 1114 goto dropafterack_ratelim; 1115 1116 vtw_del(vp->ctl, vp->vtw); 1117 goto drop; 1118 } 1119 1120 /* 1121 * If the ACK bit is off we drop the segment and return. 1122 */ 1123 if ((tiflags & TH_ACK) == 0) { 1124 if (t_flags & TF_ACKNOW) 1125 goto dropafterack; 1126 goto drop; 1127 } 1128 1129 /* 1130 * In TIME_WAIT state the only thing that should arrive 1131 * is a retransmission of the remote FIN. Acknowledge 1132 * it and restart the finack timer. 1133 */ 1134 vtw_restart(vp); 1135 goto dropafterack; 1136 1137 dropafterack: 1138 /* 1139 * Generate an ACK dropping incoming segment if it occupies 1140 * sequence space, where the ACK reflects our state. 1141 */ 1142 if (tiflags & TH_RST) 1143 goto drop; 1144 goto dropafterack2; 1145 1146 dropafterack_ratelim: 1147 /* 1148 * We may want to rate-limit ACKs against SYN/RST attack. 1149 */ 1150 if (ppsratecheck(&tcp_ackdrop_ppslim_last, &tcp_ackdrop_ppslim_count, 1151 tcp_ackdrop_ppslim) == 0) { 1152 /* XXX stat */ 1153 goto drop; 1154 } 1155 /* ...fall into dropafterack2... */ 1156 1157 dropafterack2: 1158 (void)tcp_respond(0, m, m, th, th->th_seq + tlen, th->th_ack, TH_ACK); 1159 return; 1160 1161 dropwithreset: 1162 /* 1163 * Generate a RST, dropping incoming segment. 1164 * Make ACK acceptable to originator of segment. 1165 */ 1166 if (tiflags & TH_RST) 1167 goto drop; 1168 1169 if (tiflags & TH_ACK) { 1170 tcp_respond(0, m, m, th, (tcp_seq)0, th->th_ack, TH_RST); 1171 } else { 1172 if (tiflags & TH_SYN) 1173 ++tlen; 1174 (void)tcp_respond(0, m, m, th, th->th_seq + tlen, (tcp_seq)0, 1175 TH_RST|TH_ACK); 1176 } 1177 return; 1178 drop: 1179 m_freem(m); 1180 } 1181 1182 /* 1183 * TCP input routine, follows pages 65-76 of RFC 793 very closely. 1184 */ 1185 void 1186 tcp_input(struct mbuf *m, ...) 1187 { 1188 struct tcphdr *th; 1189 struct ip *ip; 1190 struct inpcb *inp; 1191 #ifdef INET6 1192 struct ip6_hdr *ip6; 1193 struct in6pcb *in6p; 1194 #endif 1195 u_int8_t *optp = NULL; 1196 int optlen = 0; 1197 int len, tlen, toff, hdroptlen = 0; 1198 struct tcpcb *tp = NULL; 1199 int tiflags; 1200 struct socket *so = NULL; 1201 int todrop, acked, ourfinisacked, needoutput = 0; 1202 bool dupseg; 1203 #ifdef TCP_DEBUG 1204 short ostate = 0; 1205 #endif 1206 u_long tiwin; 1207 struct tcp_opt_info opti; 1208 int off, iphlen; 1209 va_list ap; 1210 int af; /* af on the wire */ 1211 struct mbuf *tcp_saveti = NULL; 1212 uint32_t ts_rtt; 1213 uint8_t iptos; 1214 uint64_t *tcps; 1215 vestigial_inpcb_t vestige; 1216 1217 vestige.valid = 0; 1218 1219 MCLAIM(m, &tcp_rx_mowner); 1220 va_start(ap, m); 1221 toff = va_arg(ap, int); 1222 (void)va_arg(ap, int); /* ignore value, advance ap */ 1223 va_end(ap); 1224 1225 TCP_STATINC(TCP_STAT_RCVTOTAL); 1226 1227 memset(&opti, 0, sizeof(opti)); 1228 opti.ts_present = 0; 1229 opti.maxseg = 0; 1230 1231 /* 1232 * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN. 1233 * 1234 * TCP is, by definition, unicast, so we reject all 1235 * multicast outright. 1236 * 1237 * Note, there are additional src/dst address checks in 1238 * the AF-specific code below. 1239 */ 1240 if (m->m_flags & (M_BCAST|M_MCAST)) { 1241 /* XXX stat */ 1242 goto drop; 1243 } 1244 #ifdef INET6 1245 if (m->m_flags & M_ANYCAST6) { 1246 /* XXX stat */ 1247 goto drop; 1248 } 1249 #endif 1250 1251 IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, sizeof(struct tcphdr)); 1252 if (th == NULL) { 1253 TCP_STATINC(TCP_STAT_RCVSHORT); 1254 return; 1255 } 1256 1257 /* 1258 * Get IP and TCP header. 1259 * Note: IP leaves IP header in first mbuf. 1260 */ 1261 ip = mtod(m, struct ip *); 1262 switch (ip->ip_v) { 1263 case 4: 1264 #ifdef INET6 1265 ip6 = NULL; 1266 #endif 1267 af = AF_INET; 1268 iphlen = sizeof(struct ip); 1269 1270 if (IN_MULTICAST(ip->ip_dst.s_addr) || 1271 in_broadcast(ip->ip_dst, m_get_rcvif_NOMPSAFE(m))) 1272 goto drop; 1273 1274 /* We do the checksum after PCB lookup... */ 1275 len = ntohs(ip->ip_len); 1276 tlen = len - toff; 1277 iptos = ip->ip_tos; 1278 break; 1279 #ifdef INET6 1280 case 6: 1281 ip = NULL; 1282 iphlen = sizeof(struct ip6_hdr); 1283 af = AF_INET6; 1284 ip6 = mtod(m, struct ip6_hdr *); 1285 1286 /* 1287 * Be proactive about unspecified IPv6 address in source. 1288 * As we use all-zero to indicate unbounded/unconnected pcb, 1289 * unspecified IPv6 address can be used to confuse us. 1290 * 1291 * Note that packets with unspecified IPv6 destination is 1292 * already dropped in ip6_input. 1293 */ 1294 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 1295 /* XXX stat */ 1296 goto drop; 1297 } 1298 1299 /* 1300 * Make sure destination address is not multicast. 1301 * Source address checked in ip6_input(). 1302 */ 1303 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 1304 /* XXX stat */ 1305 goto drop; 1306 } 1307 1308 /* We do the checksum after PCB lookup... */ 1309 len = m->m_pkthdr.len; 1310 tlen = len - toff; 1311 iptos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 1312 break; 1313 #endif 1314 default: 1315 m_freem(m); 1316 return; 1317 } 1318 1319 /* 1320 * Enforce alignment requirements that are violated in 1321 * some cases, see kern/50766 for details. 1322 */ 1323 if (TCP_HDR_ALIGNED_P(th) == 0) { 1324 m = m_copyup(m, toff + sizeof(struct tcphdr), 0); 1325 if (m == NULL) { 1326 TCP_STATINC(TCP_STAT_RCVSHORT); 1327 return; 1328 } 1329 ip = mtod(m, struct ip *); 1330 #ifdef INET6 1331 ip6 = mtod(m, struct ip6_hdr *); 1332 #endif 1333 th = (struct tcphdr *)(mtod(m, char *) + toff); 1334 } 1335 KASSERT(TCP_HDR_ALIGNED_P(th)); 1336 1337 /* 1338 * Check that TCP offset makes sense, pull out TCP options and 1339 * adjust length. 1340 */ 1341 off = th->th_off << 2; 1342 if (off < sizeof(struct tcphdr) || off > tlen) { 1343 TCP_STATINC(TCP_STAT_RCVBADOFF); 1344 goto drop; 1345 } 1346 tlen -= off; 1347 1348 if (off > sizeof(struct tcphdr)) { 1349 IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, off); 1350 if (th == NULL) { 1351 TCP_STATINC(TCP_STAT_RCVSHORT); 1352 return; 1353 } 1354 KASSERT(TCP_HDR_ALIGNED_P(th)); 1355 optlen = off - sizeof(struct tcphdr); 1356 optp = ((u_int8_t *)th) + sizeof(struct tcphdr); 1357 1358 /* 1359 * Do quick retrieval of timestamp options. 1360 * 1361 * If timestamp is the only option and it's formatted as 1362 * recommended in RFC 1323 appendix A, we quickly get the 1363 * values now and don't bother calling tcp_dooptions(), 1364 * etc. 1365 */ 1366 if ((optlen == TCPOLEN_TSTAMP_APPA || 1367 (optlen > TCPOLEN_TSTAMP_APPA && 1368 optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) && 1369 *(u_int32_t *)optp == htonl(TCPOPT_TSTAMP_HDR) && 1370 (th->th_flags & TH_SYN) == 0) { 1371 opti.ts_present = 1; 1372 opti.ts_val = ntohl(*(u_int32_t *)(optp + 4)); 1373 opti.ts_ecr = ntohl(*(u_int32_t *)(optp + 8)); 1374 optp = NULL; /* we've parsed the options */ 1375 } 1376 } 1377 tiflags = th->th_flags; 1378 1379 /* 1380 * Checksum extended TCP header and data 1381 */ 1382 if (tcp_input_checksum(af, m, th, toff, off, tlen)) 1383 goto badcsum; 1384 1385 /* 1386 * Locate pcb for segment. 1387 */ 1388 findpcb: 1389 inp = NULL; 1390 #ifdef INET6 1391 in6p = NULL; 1392 #endif 1393 switch (af) { 1394 case AF_INET: 1395 inp = in_pcblookup_connect(&tcbtable, ip->ip_src, th->th_sport, 1396 ip->ip_dst, th->th_dport, &vestige); 1397 if (inp == NULL && !vestige.valid) { 1398 TCP_STATINC(TCP_STAT_PCBHASHMISS); 1399 inp = in_pcblookup_bind(&tcbtable, ip->ip_dst, 1400 th->th_dport); 1401 } 1402 #ifdef INET6 1403 if (inp == NULL && !vestige.valid) { 1404 struct in6_addr s, d; 1405 1406 /* mapped addr case */ 1407 in6_in_2_v4mapin6(&ip->ip_src, &s); 1408 in6_in_2_v4mapin6(&ip->ip_dst, &d); 1409 in6p = in6_pcblookup_connect(&tcbtable, &s, 1410 th->th_sport, &d, th->th_dport, 0, &vestige); 1411 if (in6p == 0 && !vestige.valid) { 1412 TCP_STATINC(TCP_STAT_PCBHASHMISS); 1413 in6p = in6_pcblookup_bind(&tcbtable, &d, 1414 th->th_dport, 0); 1415 } 1416 } 1417 #endif 1418 #ifndef INET6 1419 if (inp == NULL && !vestige.valid) 1420 #else 1421 if (inp == NULL && in6p == NULL && !vestige.valid) 1422 #endif 1423 { 1424 TCP_STATINC(TCP_STAT_NOPORT); 1425 if (tcp_log_refused && 1426 (tiflags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN) { 1427 tcp4_log_refused(ip, th); 1428 } 1429 tcp_fields_to_host(th); 1430 goto dropwithreset_ratelim; 1431 } 1432 #if defined(IPSEC) 1433 if (ipsec_used) { 1434 if (inp && ipsec_in_reject(m, inp)) { 1435 goto drop; 1436 } 1437 #ifdef INET6 1438 else if (in6p && ipsec_in_reject(m, in6p)) { 1439 goto drop; 1440 } 1441 #endif 1442 } 1443 #endif /*IPSEC*/ 1444 break; 1445 #ifdef INET6 1446 case AF_INET6: 1447 { 1448 int faith; 1449 1450 #if defined(NFAITH) && NFAITH > 0 1451 faith = faithprefix(&ip6->ip6_dst); 1452 #else 1453 faith = 0; 1454 #endif 1455 in6p = in6_pcblookup_connect(&tcbtable, &ip6->ip6_src, 1456 th->th_sport, &ip6->ip6_dst, th->th_dport, faith, &vestige); 1457 if (!in6p && !vestige.valid) { 1458 TCP_STATINC(TCP_STAT_PCBHASHMISS); 1459 in6p = in6_pcblookup_bind(&tcbtable, &ip6->ip6_dst, 1460 th->th_dport, faith); 1461 } 1462 if (!in6p && !vestige.valid) { 1463 TCP_STATINC(TCP_STAT_NOPORT); 1464 if (tcp_log_refused && 1465 (tiflags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN) { 1466 tcp6_log_refused(ip6, th); 1467 } 1468 tcp_fields_to_host(th); 1469 goto dropwithreset_ratelim; 1470 } 1471 #if defined(IPSEC) 1472 if (ipsec_used && in6p && ipsec_in_reject(m, in6p)) { 1473 goto drop; 1474 } 1475 #endif 1476 break; 1477 } 1478 #endif 1479 } 1480 1481 tcp_fields_to_host(th); 1482 1483 /* 1484 * If the state is CLOSED (i.e., TCB does not exist) then 1485 * all data in the incoming segment is discarded. 1486 * If the TCB exists but is in CLOSED state, it is embryonic, 1487 * but should either do a listen or a connect soon. 1488 */ 1489 tp = NULL; 1490 so = NULL; 1491 if (inp) { 1492 /* Check the minimum TTL for socket. */ 1493 if (ip->ip_ttl < inp->inp_ip_minttl) 1494 goto drop; 1495 1496 tp = intotcpcb(inp); 1497 so = inp->inp_socket; 1498 } 1499 #ifdef INET6 1500 else if (in6p) { 1501 tp = in6totcpcb(in6p); 1502 so = in6p->in6p_socket; 1503 } 1504 #endif 1505 else if (vestige.valid) { 1506 /* We do not support the resurrection of vtw tcpcps. */ 1507 tcp_vtw_input(th, &vestige, m, tlen); 1508 m = NULL; 1509 goto drop; 1510 } 1511 1512 if (tp == NULL) 1513 goto dropwithreset_ratelim; 1514 if (tp->t_state == TCPS_CLOSED) 1515 goto drop; 1516 1517 KASSERT(so->so_lock == softnet_lock); 1518 KASSERT(solocked(so)); 1519 1520 /* Unscale the window into a 32-bit value. */ 1521 if ((tiflags & TH_SYN) == 0) 1522 tiwin = th->th_win << tp->snd_scale; 1523 else 1524 tiwin = th->th_win; 1525 1526 #ifdef INET6 1527 /* save packet options if user wanted */ 1528 if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS)) { 1529 if (in6p->in6p_options) { 1530 m_freem(in6p->in6p_options); 1531 in6p->in6p_options = NULL; 1532 } 1533 KASSERT(ip6 != NULL); 1534 ip6_savecontrol(in6p, &in6p->in6p_options, ip6, m); 1535 } 1536 #endif 1537 1538 if (so->so_options & SO_DEBUG) { 1539 #ifdef TCP_DEBUG 1540 ostate = tp->t_state; 1541 #endif 1542 1543 tcp_saveti = NULL; 1544 if (iphlen + sizeof(struct tcphdr) > MHLEN) 1545 goto nosave; 1546 1547 if (m->m_len > iphlen && (m->m_flags & M_EXT) == 0) { 1548 tcp_saveti = m_copym(m, 0, iphlen, M_DONTWAIT); 1549 if (tcp_saveti == NULL) 1550 goto nosave; 1551 } else { 1552 MGETHDR(tcp_saveti, M_DONTWAIT, MT_HEADER); 1553 if (tcp_saveti == NULL) 1554 goto nosave; 1555 MCLAIM(m, &tcp_mowner); 1556 tcp_saveti->m_len = iphlen; 1557 m_copydata(m, 0, iphlen, 1558 mtod(tcp_saveti, void *)); 1559 } 1560 1561 if (M_TRAILINGSPACE(tcp_saveti) < sizeof(struct tcphdr)) { 1562 m_freem(tcp_saveti); 1563 tcp_saveti = NULL; 1564 } else { 1565 tcp_saveti->m_len += sizeof(struct tcphdr); 1566 memcpy(mtod(tcp_saveti, char *) + iphlen, th, 1567 sizeof(struct tcphdr)); 1568 } 1569 nosave:; 1570 } 1571 1572 if (so->so_options & SO_ACCEPTCONN) { 1573 union syn_cache_sa src; 1574 union syn_cache_sa dst; 1575 1576 KASSERT(tp->t_state == TCPS_LISTEN); 1577 1578 memset(&src, 0, sizeof(src)); 1579 memset(&dst, 0, sizeof(dst)); 1580 switch (af) { 1581 case AF_INET: 1582 src.sin.sin_len = sizeof(struct sockaddr_in); 1583 src.sin.sin_family = AF_INET; 1584 src.sin.sin_addr = ip->ip_src; 1585 src.sin.sin_port = th->th_sport; 1586 1587 dst.sin.sin_len = sizeof(struct sockaddr_in); 1588 dst.sin.sin_family = AF_INET; 1589 dst.sin.sin_addr = ip->ip_dst; 1590 dst.sin.sin_port = th->th_dport; 1591 break; 1592 #ifdef INET6 1593 case AF_INET6: 1594 src.sin6.sin6_len = sizeof(struct sockaddr_in6); 1595 src.sin6.sin6_family = AF_INET6; 1596 src.sin6.sin6_addr = ip6->ip6_src; 1597 src.sin6.sin6_port = th->th_sport; 1598 1599 dst.sin6.sin6_len = sizeof(struct sockaddr_in6); 1600 dst.sin6.sin6_family = AF_INET6; 1601 dst.sin6.sin6_addr = ip6->ip6_dst; 1602 dst.sin6.sin6_port = th->th_dport; 1603 break; 1604 #endif 1605 } 1606 1607 if ((tiflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) { 1608 if (tiflags & TH_RST) { 1609 syn_cache_reset(&src.sa, &dst.sa, th); 1610 } else if ((tiflags & (TH_ACK|TH_SYN)) == 1611 (TH_ACK|TH_SYN)) { 1612 /* 1613 * Received a SYN,ACK. This should never 1614 * happen while we are in LISTEN. Send an RST. 1615 */ 1616 goto badsyn; 1617 } else if (tiflags & TH_ACK) { 1618 so = syn_cache_get(&src.sa, &dst.sa, th, so, m); 1619 if (so == NULL) { 1620 /* 1621 * We don't have a SYN for this ACK; 1622 * send an RST. 1623 */ 1624 goto badsyn; 1625 } else if (so == (struct socket *)(-1)) { 1626 /* 1627 * We were unable to create the 1628 * connection. If the 3-way handshake 1629 * was completed, and RST has been 1630 * sent to the peer. Since the mbuf 1631 * might be in use for the reply, do 1632 * not free it. 1633 */ 1634 m = NULL; 1635 } else { 1636 /* 1637 * We have created a full-blown 1638 * connection. 1639 */ 1640 tp = NULL; 1641 inp = NULL; 1642 #ifdef INET6 1643 in6p = NULL; 1644 #endif 1645 switch (so->so_proto->pr_domain->dom_family) { 1646 case AF_INET: 1647 inp = sotoinpcb(so); 1648 tp = intotcpcb(inp); 1649 break; 1650 #ifdef INET6 1651 case AF_INET6: 1652 in6p = sotoin6pcb(so); 1653 tp = in6totcpcb(in6p); 1654 break; 1655 #endif 1656 } 1657 if (tp == NULL) 1658 goto badsyn; /*XXX*/ 1659 tiwin <<= tp->snd_scale; 1660 goto after_listen; 1661 } 1662 } else { 1663 /* 1664 * None of RST, SYN or ACK was set. 1665 * This is an invalid packet for a 1666 * TCB in LISTEN state. Send a RST. 1667 */ 1668 goto badsyn; 1669 } 1670 } else { 1671 /* 1672 * Received a SYN. 1673 */ 1674 1675 #ifdef INET6 1676 /* 1677 * If deprecated address is forbidden, we do 1678 * not accept SYN to deprecated interface 1679 * address to prevent any new inbound 1680 * connection from getting established. 1681 * When we do not accept SYN, we send a TCP 1682 * RST, with deprecated source address (instead 1683 * of dropping it). We compromise it as it is 1684 * much better for peer to send a RST, and 1685 * RST will be the final packet for the 1686 * exchange. 1687 * 1688 * If we do not forbid deprecated addresses, we 1689 * accept the SYN packet. RFC2462 does not 1690 * suggest dropping SYN in this case. 1691 * If we decipher RFC2462 5.5.4, it says like 1692 * this: 1693 * 1. use of deprecated addr with existing 1694 * communication is okay - "SHOULD continue 1695 * to be used" 1696 * 2. use of it with new communication: 1697 * (2a) "SHOULD NOT be used if alternate 1698 * address with sufficient scope is 1699 * available" 1700 * (2b) nothing mentioned otherwise. 1701 * Here we fall into (2b) case as we have no 1702 * choice in our source address selection - we 1703 * must obey the peer. 1704 * 1705 * The wording in RFC2462 is confusing, and 1706 * there are multiple description text for 1707 * deprecated address handling - worse, they 1708 * are not exactly the same. I believe 5.5.4 1709 * is the best one, so we follow 5.5.4. 1710 */ 1711 if (af == AF_INET6 && !ip6_use_deprecated) { 1712 struct in6_ifaddr *ia6; 1713 int s; 1714 struct ifnet *rcvif = m_get_rcvif(m, &s); 1715 if (rcvif == NULL) 1716 goto dropwithreset; /* XXX */ 1717 if ((ia6 = in6ifa_ifpwithaddr(rcvif, 1718 &ip6->ip6_dst)) && 1719 (ia6->ia6_flags & IN6_IFF_DEPRECATED)) { 1720 tp = NULL; 1721 m_put_rcvif(rcvif, &s); 1722 goto dropwithreset; 1723 } 1724 m_put_rcvif(rcvif, &s); 1725 } 1726 #endif 1727 1728 /* 1729 * LISTEN socket received a SYN from itself? This 1730 * can't possibly be valid; drop the packet. 1731 */ 1732 if (th->th_sport == th->th_dport) { 1733 int eq = 0; 1734 1735 switch (af) { 1736 case AF_INET: 1737 eq = in_hosteq(ip->ip_src, ip->ip_dst); 1738 break; 1739 #ifdef INET6 1740 case AF_INET6: 1741 eq = IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, 1742 &ip6->ip6_dst); 1743 break; 1744 #endif 1745 } 1746 if (eq) { 1747 TCP_STATINC(TCP_STAT_BADSYN); 1748 goto drop; 1749 } 1750 } 1751 1752 /* 1753 * SYN looks ok; create compressed TCP 1754 * state for it. 1755 */ 1756 if (so->so_qlen <= so->so_qlimit && 1757 syn_cache_add(&src.sa, &dst.sa, th, toff, 1758 so, m, optp, optlen, &opti)) 1759 m = NULL; 1760 } 1761 1762 goto drop; 1763 } 1764 1765 after_listen: 1766 /* 1767 * From here on, we're dealing with !LISTEN. 1768 */ 1769 KASSERT(tp->t_state != TCPS_LISTEN); 1770 1771 /* 1772 * Segment received on connection. 1773 * Reset idle time and keep-alive timer. 1774 */ 1775 tp->t_rcvtime = tcp_now; 1776 if (TCPS_HAVEESTABLISHED(tp->t_state)) 1777 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle); 1778 1779 /* 1780 * Process options. 1781 */ 1782 #ifdef TCP_SIGNATURE 1783 if (optp || (tp->t_flags & TF_SIGNATURE)) 1784 #else 1785 if (optp) 1786 #endif 1787 if (tcp_dooptions(tp, optp, optlen, th, m, toff, &opti) < 0) 1788 goto drop; 1789 1790 if (TCP_SACK_ENABLED(tp)) { 1791 tcp_del_sackholes(tp, th); 1792 } 1793 1794 if (TCP_ECN_ALLOWED(tp)) { 1795 if (tiflags & TH_CWR) { 1796 tp->t_flags &= ~TF_ECN_SND_ECE; 1797 } 1798 switch (iptos & IPTOS_ECN_MASK) { 1799 case IPTOS_ECN_CE: 1800 tp->t_flags |= TF_ECN_SND_ECE; 1801 TCP_STATINC(TCP_STAT_ECN_CE); 1802 break; 1803 case IPTOS_ECN_ECT0: 1804 TCP_STATINC(TCP_STAT_ECN_ECT); 1805 break; 1806 case IPTOS_ECN_ECT1: 1807 /* XXX: ignore for now -- rpaulo */ 1808 break; 1809 } 1810 /* 1811 * Congestion experienced. 1812 * Ignore if we are already trying to recover. 1813 */ 1814 if ((tiflags & TH_ECE) && SEQ_GEQ(tp->snd_una, tp->snd_recover)) 1815 tp->t_congctl->cong_exp(tp); 1816 } 1817 1818 if (opti.ts_present && opti.ts_ecr) { 1819 /* 1820 * Calculate the RTT from the returned time stamp and the 1821 * connection's time base. If the time stamp is later than 1822 * the current time, or is extremely old, fall back to non-1323 1823 * RTT calculation. Since ts_rtt is unsigned, we can test both 1824 * at the same time. 1825 * 1826 * Note that ts_rtt is in units of slow ticks (500 1827 * ms). Since most earthbound RTTs are < 500 ms, 1828 * observed values will have large quantization noise. 1829 * Our smoothed RTT is then the fraction of observed 1830 * samples that are 1 tick instead of 0 (times 500 1831 * ms). 1832 * 1833 * ts_rtt is increased by 1 to denote a valid sample, 1834 * with 0 indicating an invalid measurement. This 1835 * extra 1 must be removed when ts_rtt is used, or 1836 * else an an erroneous extra 500 ms will result. 1837 */ 1838 ts_rtt = TCP_TIMESTAMP(tp) - opti.ts_ecr + 1; 1839 if (ts_rtt > TCP_PAWS_IDLE) 1840 ts_rtt = 0; 1841 } else { 1842 ts_rtt = 0; 1843 } 1844 1845 /* 1846 * Fast path: check for the two common cases of a uni-directional 1847 * data transfer. If: 1848 * o We are in the ESTABLISHED state, and 1849 * o The packet has no control flags, and 1850 * o The packet is in-sequence, and 1851 * o The window didn't change, and 1852 * o We are not retransmitting 1853 * It's a candidate. 1854 * 1855 * If the length (tlen) is zero and the ack moved forward, we're 1856 * the sender side of the transfer. Just free the data acked and 1857 * wake any higher level process that was blocked waiting for 1858 * space. 1859 * 1860 * If the length is non-zero and the ack didn't move, we're the 1861 * receiver side. If we're getting packets in-order (the reassembly 1862 * queue is empty), add the data to the socket buffer and note 1863 * that we need a delayed ack. 1864 */ 1865 if (tp->t_state == TCPS_ESTABLISHED && 1866 (tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ECE|TH_CWR|TH_ACK)) 1867 == TH_ACK && 1868 (!opti.ts_present || TSTMP_GEQ(opti.ts_val, tp->ts_recent)) && 1869 th->th_seq == tp->rcv_nxt && 1870 tiwin && tiwin == tp->snd_wnd && 1871 tp->snd_nxt == tp->snd_max) { 1872 1873 /* 1874 * If last ACK falls within this segment's sequence numbers, 1875 * record the timestamp. 1876 * NOTE that the test is modified according to the latest 1877 * proposal of the tcplw (at) cray.com list (Braden 1993/04/26). 1878 * 1879 * note that we already know 1880 * TSTMP_GEQ(opti.ts_val, tp->ts_recent) 1881 */ 1882 if (opti.ts_present && SEQ_LEQ(th->th_seq, tp->last_ack_sent)) { 1883 tp->ts_recent_age = tcp_now; 1884 tp->ts_recent = opti.ts_val; 1885 } 1886 1887 if (tlen == 0) { 1888 /* Ack prediction. */ 1889 if (SEQ_GT(th->th_ack, tp->snd_una) && 1890 SEQ_LEQ(th->th_ack, tp->snd_max) && 1891 tp->snd_cwnd >= tp->snd_wnd && 1892 tp->t_partialacks < 0) { 1893 /* 1894 * this is a pure ack for outstanding data. 1895 */ 1896 if (ts_rtt) 1897 tcp_xmit_timer(tp, ts_rtt - 1); 1898 else if (tp->t_rtttime && 1899 SEQ_GT(th->th_ack, tp->t_rtseq)) 1900 tcp_xmit_timer(tp, 1901 tcp_now - tp->t_rtttime); 1902 acked = th->th_ack - tp->snd_una; 1903 tcps = TCP_STAT_GETREF(); 1904 tcps[TCP_STAT_PREDACK]++; 1905 tcps[TCP_STAT_RCVACKPACK]++; 1906 tcps[TCP_STAT_RCVACKBYTE] += acked; 1907 TCP_STAT_PUTREF(); 1908 nd6_hint(tp); 1909 1910 if (acked > (tp->t_lastoff - tp->t_inoff)) 1911 tp->t_lastm = NULL; 1912 sbdrop(&so->so_snd, acked); 1913 tp->t_lastoff -= acked; 1914 1915 icmp_check(tp, th, acked); 1916 1917 tp->snd_una = th->th_ack; 1918 tp->snd_fack = tp->snd_una; 1919 if (SEQ_LT(tp->snd_high, tp->snd_una)) 1920 tp->snd_high = tp->snd_una; 1921 m_freem(m); 1922 1923 /* 1924 * If all outstanding data are acked, stop 1925 * retransmit timer, otherwise restart timer 1926 * using current (possibly backed-off) value. 1927 * If process is waiting for space, 1928 * wakeup/selnotify/signal. If data 1929 * are ready to send, let tcp_output 1930 * decide between more output or persist. 1931 */ 1932 if (tp->snd_una == tp->snd_max) 1933 TCP_TIMER_DISARM(tp, TCPT_REXMT); 1934 else if (TCP_TIMER_ISARMED(tp, 1935 TCPT_PERSIST) == 0) 1936 TCP_TIMER_ARM(tp, TCPT_REXMT, 1937 tp->t_rxtcur); 1938 1939 sowwakeup(so); 1940 if (so->so_snd.sb_cc) { 1941 KERNEL_LOCK(1, NULL); 1942 (void)tcp_output(tp); 1943 KERNEL_UNLOCK_ONE(NULL); 1944 } 1945 if (tcp_saveti) 1946 m_freem(tcp_saveti); 1947 return; 1948 } 1949 } else if (th->th_ack == tp->snd_una && 1950 TAILQ_FIRST(&tp->segq) == NULL && 1951 tlen <= sbspace(&so->so_rcv)) { 1952 int newsize = 0; 1953 1954 /* 1955 * this is a pure, in-sequence data packet 1956 * with nothing on the reassembly queue and 1957 * we have enough buffer space to take it. 1958 */ 1959 tp->rcv_nxt += tlen; 1960 tcps = TCP_STAT_GETREF(); 1961 tcps[TCP_STAT_PREDDAT]++; 1962 tcps[TCP_STAT_RCVPACK]++; 1963 tcps[TCP_STAT_RCVBYTE] += tlen; 1964 TCP_STAT_PUTREF(); 1965 nd6_hint(tp); 1966 1967 /* 1968 * Automatic sizing enables the performance of large buffers 1969 * and most of the efficiency of small ones by only allocating 1970 * space when it is needed. 1971 * 1972 * On the receive side the socket buffer memory is only rarely 1973 * used to any significant extent. This allows us to be much 1974 * more aggressive in scaling the receive socket buffer. For 1975 * the case that the buffer space is actually used to a large 1976 * extent and we run out of kernel memory we can simply drop 1977 * the new segments; TCP on the sender will just retransmit it 1978 * later. Setting the buffer size too big may only consume too 1979 * much kernel memory if the application doesn't read() from 1980 * the socket or packet loss or reordering makes use of the 1981 * reassembly queue. 1982 * 1983 * The criteria to step up the receive buffer one notch are: 1984 * 1. the number of bytes received during the time it takes 1985 * one timestamp to be reflected back to us (the RTT); 1986 * 2. received bytes per RTT is within seven eighth of the 1987 * current socket buffer size; 1988 * 3. receive buffer size has not hit maximal automatic size; 1989 * 1990 * This algorithm does one step per RTT at most and only if 1991 * we receive a bulk stream w/o packet losses or reorderings. 1992 * Shrinking the buffer during idle times is not necessary as 1993 * it doesn't consume any memory when idle. 1994 * 1995 * TODO: Only step up if the application is actually serving 1996 * the buffer to better manage the socket buffer resources. 1997 */ 1998 if (tcp_do_autorcvbuf && 1999 opti.ts_ecr && 2000 (so->so_rcv.sb_flags & SB_AUTOSIZE)) { 2001 if (opti.ts_ecr > tp->rfbuf_ts && 2002 opti.ts_ecr - tp->rfbuf_ts < PR_SLOWHZ) { 2003 if (tp->rfbuf_cnt > 2004 (so->so_rcv.sb_hiwat / 8 * 7) && 2005 so->so_rcv.sb_hiwat < 2006 tcp_autorcvbuf_max) { 2007 newsize = 2008 min(so->so_rcv.sb_hiwat + 2009 tcp_autorcvbuf_inc, 2010 tcp_autorcvbuf_max); 2011 } 2012 /* Start over with next RTT. */ 2013 tp->rfbuf_ts = 0; 2014 tp->rfbuf_cnt = 0; 2015 } else 2016 tp->rfbuf_cnt += tlen; /* add up */ 2017 } 2018 2019 /* 2020 * Drop TCP, IP headers and TCP options then add data 2021 * to socket buffer. 2022 */ 2023 if (so->so_state & SS_CANTRCVMORE) { 2024 m_freem(m); 2025 } else { 2026 /* 2027 * Set new socket buffer size. 2028 * Give up when limit is reached. 2029 */ 2030 if (newsize) 2031 if (!sbreserve(&so->so_rcv, 2032 newsize, so)) 2033 so->so_rcv.sb_flags &= ~SB_AUTOSIZE; 2034 m_adj(m, toff + off); 2035 sbappendstream(&so->so_rcv, m); 2036 } 2037 sorwakeup(so); 2038 tcp_setup_ack(tp, th); 2039 if (tp->t_flags & TF_ACKNOW) { 2040 KERNEL_LOCK(1, NULL); 2041 (void)tcp_output(tp); 2042 KERNEL_UNLOCK_ONE(NULL); 2043 } 2044 if (tcp_saveti) 2045 m_freem(tcp_saveti); 2046 return; 2047 } 2048 } 2049 2050 /* 2051 * Compute mbuf offset to TCP data segment. 2052 */ 2053 hdroptlen = toff + off; 2054 2055 /* 2056 * Calculate amount of space in receive window. Receive window is 2057 * amount of space in rcv queue, but not less than advertised 2058 * window. 2059 */ 2060 { 2061 int win; 2062 win = sbspace(&so->so_rcv); 2063 if (win < 0) 2064 win = 0; 2065 tp->rcv_wnd = imax(win, (int)(tp->rcv_adv - tp->rcv_nxt)); 2066 } 2067 2068 /* Reset receive buffer auto scaling when not in bulk receive mode. */ 2069 tp->rfbuf_ts = 0; 2070 tp->rfbuf_cnt = 0; 2071 2072 switch (tp->t_state) { 2073 /* 2074 * If the state is SYN_SENT: 2075 * if seg contains an ACK, but not for our SYN, drop the input. 2076 * if seg contains a RST, then drop the connection. 2077 * if seg does not contain SYN, then drop it. 2078 * Otherwise this is an acceptable SYN segment 2079 * initialize tp->rcv_nxt and tp->irs 2080 * if seg contains ack then advance tp->snd_una 2081 * if seg contains a ECE and ECN support is enabled, the stream 2082 * is ECN capable. 2083 * if SYN has been acked change to ESTABLISHED else SYN_RCVD state 2084 * arrange for segment to be acked (eventually) 2085 * continue processing rest of data/controls, beginning with URG 2086 */ 2087 case TCPS_SYN_SENT: 2088 if ((tiflags & TH_ACK) && 2089 (SEQ_LEQ(th->th_ack, tp->iss) || 2090 SEQ_GT(th->th_ack, tp->snd_max))) 2091 goto dropwithreset; 2092 if (tiflags & TH_RST) { 2093 if (tiflags & TH_ACK) 2094 tp = tcp_drop(tp, ECONNREFUSED); 2095 goto drop; 2096 } 2097 if ((tiflags & TH_SYN) == 0) 2098 goto drop; 2099 if (tiflags & TH_ACK) { 2100 tp->snd_una = th->th_ack; 2101 if (SEQ_LT(tp->snd_nxt, tp->snd_una)) 2102 tp->snd_nxt = tp->snd_una; 2103 if (SEQ_LT(tp->snd_high, tp->snd_una)) 2104 tp->snd_high = tp->snd_una; 2105 TCP_TIMER_DISARM(tp, TCPT_REXMT); 2106 2107 if ((tiflags & TH_ECE) && tcp_do_ecn) { 2108 tp->t_flags |= TF_ECN_PERMIT; 2109 TCP_STATINC(TCP_STAT_ECN_SHS); 2110 } 2111 } 2112 tp->irs = th->th_seq; 2113 tcp_rcvseqinit(tp); 2114 tp->t_flags |= TF_ACKNOW; 2115 tcp_mss_from_peer(tp, opti.maxseg); 2116 2117 /* 2118 * Initialize the initial congestion window. If we 2119 * had to retransmit the SYN, we must initialize cwnd 2120 * to 1 segment (i.e. the Loss Window). 2121 */ 2122 if (tp->t_flags & TF_SYN_REXMT) 2123 tp->snd_cwnd = tp->t_peermss; 2124 else { 2125 int ss = tcp_init_win; 2126 if (inp != NULL && in_localaddr(inp->inp_faddr)) 2127 ss = tcp_init_win_local; 2128 #ifdef INET6 2129 if (in6p != NULL && in6_localaddr(&in6p->in6p_faddr)) 2130 ss = tcp_init_win_local; 2131 #endif 2132 tp->snd_cwnd = TCP_INITIAL_WINDOW(ss, tp->t_peermss); 2133 } 2134 2135 tcp_rmx_rtt(tp); 2136 if (tiflags & TH_ACK) { 2137 TCP_STATINC(TCP_STAT_CONNECTS); 2138 /* 2139 * move tcp_established before soisconnected 2140 * because upcall handler can drive tcp_output 2141 * functionality. 2142 * XXX we might call soisconnected at the end of 2143 * all processing 2144 */ 2145 tcp_established(tp); 2146 soisconnected(so); 2147 /* Do window scaling on this connection? */ 2148 if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == 2149 (TF_RCVD_SCALE|TF_REQ_SCALE)) { 2150 tp->snd_scale = tp->requested_s_scale; 2151 tp->rcv_scale = tp->request_r_scale; 2152 } 2153 TCP_REASS_LOCK(tp); 2154 (void)tcp_reass(tp, NULL, NULL, tlen); 2155 /* 2156 * if we didn't have to retransmit the SYN, 2157 * use its rtt as our initial srtt & rtt var. 2158 */ 2159 if (tp->t_rtttime) 2160 tcp_xmit_timer(tp, tcp_now - tp->t_rtttime); 2161 } else { 2162 tp->t_state = TCPS_SYN_RECEIVED; 2163 } 2164 2165 /* 2166 * Advance th->th_seq to correspond to first data byte. 2167 * If data, trim to stay within window, 2168 * dropping FIN if necessary. 2169 */ 2170 th->th_seq++; 2171 if (tlen > tp->rcv_wnd) { 2172 todrop = tlen - tp->rcv_wnd; 2173 m_adj(m, -todrop); 2174 tlen = tp->rcv_wnd; 2175 tiflags &= ~TH_FIN; 2176 tcps = TCP_STAT_GETREF(); 2177 tcps[TCP_STAT_RCVPACKAFTERWIN]++; 2178 tcps[TCP_STAT_RCVBYTEAFTERWIN] += todrop; 2179 TCP_STAT_PUTREF(); 2180 } 2181 tp->snd_wl1 = th->th_seq - 1; 2182 tp->rcv_up = th->th_seq; 2183 goto step6; 2184 2185 /* 2186 * If the state is SYN_RECEIVED: 2187 * If seg contains an ACK, but not for our SYN, drop the input 2188 * and generate an RST. See page 36, rfc793 2189 */ 2190 case TCPS_SYN_RECEIVED: 2191 if ((tiflags & TH_ACK) && 2192 (SEQ_LEQ(th->th_ack, tp->iss) || 2193 SEQ_GT(th->th_ack, tp->snd_max))) 2194 goto dropwithreset; 2195 break; 2196 } 2197 2198 /* 2199 * From here on, we're dealing with !LISTEN and !SYN_SENT. 2200 */ 2201 KASSERT(tp->t_state != TCPS_LISTEN && 2202 tp->t_state != TCPS_SYN_SENT); 2203 2204 /* 2205 * RFC1323 PAWS: if we have a timestamp reply on this segment and 2206 * it's less than ts_recent, drop it. 2207 */ 2208 if (opti.ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent && 2209 TSTMP_LT(opti.ts_val, tp->ts_recent)) { 2210 /* Check to see if ts_recent is over 24 days old. */ 2211 if (tcp_now - tp->ts_recent_age > TCP_PAWS_IDLE) { 2212 /* 2213 * Invalidate ts_recent. If this segment updates 2214 * ts_recent, the age will be reset later and ts_recent 2215 * will get a valid value. If it does not, setting 2216 * ts_recent to zero will at least satisfy the 2217 * requirement that zero be placed in the timestamp 2218 * echo reply when ts_recent isn't valid. The 2219 * age isn't reset until we get a valid ts_recent 2220 * because we don't want out-of-order segments to be 2221 * dropped when ts_recent is old. 2222 */ 2223 tp->ts_recent = 0; 2224 } else { 2225 tcps = TCP_STAT_GETREF(); 2226 tcps[TCP_STAT_RCVDUPPACK]++; 2227 tcps[TCP_STAT_RCVDUPBYTE] += tlen; 2228 tcps[TCP_STAT_PAWSDROP]++; 2229 TCP_STAT_PUTREF(); 2230 tcp_new_dsack(tp, th->th_seq, tlen); 2231 goto dropafterack; 2232 } 2233 } 2234 2235 /* 2236 * Check that at least some bytes of the segment are within the 2237 * receive window. If segment begins before rcv_nxt, drop leading 2238 * data (and SYN); if nothing left, just ack. 2239 */ 2240 todrop = tp->rcv_nxt - th->th_seq; 2241 dupseg = false; 2242 if (todrop > 0) { 2243 if (tiflags & TH_SYN) { 2244 tiflags &= ~TH_SYN; 2245 th->th_seq++; 2246 tcp_urp_drop(th, 1, &tiflags); 2247 todrop--; 2248 } 2249 if (todrop > tlen || 2250 (todrop == tlen && (tiflags & TH_FIN) == 0)) { 2251 /* 2252 * Any valid FIN or RST must be to the left of the 2253 * window. At this point the FIN or RST must be a 2254 * duplicate or out of sequence; drop it. 2255 */ 2256 if (tiflags & TH_RST) 2257 goto drop; 2258 tiflags &= ~(TH_FIN|TH_RST); 2259 2260 /* 2261 * Send an ACK to resynchronize and drop any data. 2262 * But keep on processing for RST or ACK. 2263 */ 2264 tp->t_flags |= TF_ACKNOW; 2265 todrop = tlen; 2266 dupseg = true; 2267 tcps = TCP_STAT_GETREF(); 2268 tcps[TCP_STAT_RCVDUPPACK]++; 2269 tcps[TCP_STAT_RCVDUPBYTE] += todrop; 2270 TCP_STAT_PUTREF(); 2271 } else if ((tiflags & TH_RST) && th->th_seq != tp->rcv_nxt) { 2272 /* 2273 * Test for reset before adjusting the sequence 2274 * number for overlapping data. 2275 */ 2276 goto dropafterack_ratelim; 2277 } else { 2278 tcps = TCP_STAT_GETREF(); 2279 tcps[TCP_STAT_RCVPARTDUPPACK]++; 2280 tcps[TCP_STAT_RCVPARTDUPBYTE] += todrop; 2281 TCP_STAT_PUTREF(); 2282 } 2283 tcp_new_dsack(tp, th->th_seq, todrop); 2284 hdroptlen += todrop; /* drop from head afterwards (m_adj) */ 2285 th->th_seq += todrop; 2286 tlen -= todrop; 2287 tcp_urp_drop(th, todrop, &tiflags); 2288 } 2289 2290 /* 2291 * If new data is received on a connection after the user processes 2292 * are gone, then RST the other end. 2293 */ 2294 if ((so->so_state & SS_NOFDREF) && 2295 tp->t_state > TCPS_CLOSE_WAIT && tlen) { 2296 tp = tcp_close(tp); 2297 TCP_STATINC(TCP_STAT_RCVAFTERCLOSE); 2298 goto dropwithreset; 2299 } 2300 2301 /* 2302 * If the segment ends after the window, drop trailing data (and 2303 * PUSH and FIN); if nothing left, just ACK. 2304 */ 2305 todrop = (th->th_seq + tlen) - (tp->rcv_nxt + tp->rcv_wnd); 2306 if (todrop > 0) { 2307 TCP_STATINC(TCP_STAT_RCVPACKAFTERWIN); 2308 if (todrop >= tlen) { 2309 /* 2310 * The segment actually starts after the window. 2311 * th->th_seq + tlen - tp->rcv_nxt - tp->rcv_wnd >= tlen 2312 * th->th_seq - tp->rcv_nxt - tp->rcv_wnd >= 0 2313 * th->th_seq >= tp->rcv_nxt + tp->rcv_wnd 2314 */ 2315 TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, tlen); 2316 2317 /* 2318 * If a new connection request is received while in 2319 * TIME_WAIT, drop the old connection and start over 2320 * if the sequence numbers are above the previous 2321 * ones. 2322 * 2323 * NOTE: We need to put the header fields back into 2324 * network order. 2325 */ 2326 if ((tiflags & TH_SYN) && 2327 tp->t_state == TCPS_TIME_WAIT && 2328 SEQ_GT(th->th_seq, tp->rcv_nxt)) { 2329 tp = tcp_close(tp); 2330 tcp_fields_to_net(th); 2331 m_freem(tcp_saveti); 2332 tcp_saveti = NULL; 2333 goto findpcb; 2334 } 2335 2336 /* 2337 * If window is closed can only take segments at 2338 * window edge, and have to drop data and PUSH from 2339 * incoming segments. Continue processing, but 2340 * remember to ack. Otherwise, drop segment 2341 * and (if not RST) ack. 2342 */ 2343 if (tp->rcv_wnd == 0 && th->th_seq == tp->rcv_nxt) { 2344 KASSERT(todrop == tlen); 2345 tp->t_flags |= TF_ACKNOW; 2346 TCP_STATINC(TCP_STAT_RCVWINPROBE); 2347 } else { 2348 goto dropafterack; 2349 } 2350 } else { 2351 TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, todrop); 2352 } 2353 m_adj(m, -todrop); 2354 tlen -= todrop; 2355 tiflags &= ~(TH_PUSH|TH_FIN); 2356 } 2357 2358 /* 2359 * If last ACK falls within this segment's sequence numbers, 2360 * record the timestamp. 2361 * NOTE: 2362 * 1) That the test incorporates suggestions from the latest 2363 * proposal of the tcplw (at) cray.com list (Braden 1993/04/26). 2364 * 2) That updating only on newer timestamps interferes with 2365 * our earlier PAWS tests, so this check should be solely 2366 * predicated on the sequence space of this segment. 2367 * 3) That we modify the segment boundary check to be 2368 * Last.ACK.Sent <= SEG.SEQ + SEG.Len 2369 * instead of RFC1323's 2370 * Last.ACK.Sent < SEG.SEQ + SEG.Len, 2371 * This modified check allows us to overcome RFC1323's 2372 * limitations as described in Stevens TCP/IP Illustrated 2373 * Vol. 2 p.869. In such cases, we can still calculate the 2374 * RTT correctly when RCV.NXT == Last.ACK.Sent. 2375 */ 2376 if (opti.ts_present && 2377 SEQ_LEQ(th->th_seq, tp->last_ack_sent) && 2378 SEQ_LEQ(tp->last_ack_sent, th->th_seq + tlen + 2379 ((tiflags & (TH_SYN|TH_FIN)) != 0))) { 2380 tp->ts_recent_age = tcp_now; 2381 tp->ts_recent = opti.ts_val; 2382 } 2383 2384 /* 2385 * If the RST bit is set examine the state: 2386 * RECEIVED state: 2387 * If passive open, return to LISTEN state. 2388 * If active open, inform user that connection was refused. 2389 * ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT states: 2390 * Inform user that connection was reset, and close tcb. 2391 * CLOSING, LAST_ACK, TIME_WAIT states: 2392 * Close the tcb. 2393 */ 2394 if (tiflags & TH_RST) { 2395 if (th->th_seq != tp->rcv_nxt) 2396 goto dropafterack_ratelim; 2397 2398 switch (tp->t_state) { 2399 case TCPS_SYN_RECEIVED: 2400 so->so_error = ECONNREFUSED; 2401 goto close; 2402 2403 case TCPS_ESTABLISHED: 2404 case TCPS_FIN_WAIT_1: 2405 case TCPS_FIN_WAIT_2: 2406 case TCPS_CLOSE_WAIT: 2407 so->so_error = ECONNRESET; 2408 close: 2409 tp->t_state = TCPS_CLOSED; 2410 TCP_STATINC(TCP_STAT_DROPS); 2411 tp = tcp_close(tp); 2412 goto drop; 2413 2414 case TCPS_CLOSING: 2415 case TCPS_LAST_ACK: 2416 case TCPS_TIME_WAIT: 2417 tp = tcp_close(tp); 2418 goto drop; 2419 } 2420 } 2421 2422 /* 2423 * Since we've covered the SYN-SENT and SYN-RECEIVED states above 2424 * we must be in a synchronized state. RFC791 states (under RST 2425 * generation) that any unacceptable segment (an out-of-order SYN 2426 * qualifies) received in a synchronized state must elicit only an 2427 * empty acknowledgment segment ... and the connection remains in 2428 * the same state. 2429 */ 2430 if (tiflags & TH_SYN) { 2431 if (tp->rcv_nxt == th->th_seq) { 2432 tcp_respond(tp, m, m, th, (tcp_seq)0, th->th_ack - 1, 2433 TH_ACK); 2434 if (tcp_saveti) 2435 m_freem(tcp_saveti); 2436 return; 2437 } 2438 2439 goto dropafterack_ratelim; 2440 } 2441 2442 /* 2443 * If the ACK bit is off we drop the segment and return. 2444 */ 2445 if ((tiflags & TH_ACK) == 0) { 2446 if (tp->t_flags & TF_ACKNOW) 2447 goto dropafterack; 2448 goto drop; 2449 } 2450 2451 /* 2452 * From here on, we're doing ACK processing. 2453 */ 2454 2455 switch (tp->t_state) { 2456 /* 2457 * In SYN_RECEIVED state if the ack ACKs our SYN then enter 2458 * ESTABLISHED state and continue processing, otherwise 2459 * send an RST. 2460 */ 2461 case TCPS_SYN_RECEIVED: 2462 if (SEQ_GT(tp->snd_una, th->th_ack) || 2463 SEQ_GT(th->th_ack, tp->snd_max)) 2464 goto dropwithreset; 2465 TCP_STATINC(TCP_STAT_CONNECTS); 2466 soisconnected(so); 2467 tcp_established(tp); 2468 /* Do window scaling? */ 2469 if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == 2470 (TF_RCVD_SCALE|TF_REQ_SCALE)) { 2471 tp->snd_scale = tp->requested_s_scale; 2472 tp->rcv_scale = tp->request_r_scale; 2473 } 2474 TCP_REASS_LOCK(tp); 2475 (void)tcp_reass(tp, NULL, NULL, tlen); 2476 tp->snd_wl1 = th->th_seq - 1; 2477 /* FALLTHROUGH */ 2478 2479 /* 2480 * In ESTABLISHED state: drop duplicate ACKs; ACK out of range 2481 * ACKs. If the ack is in the range 2482 * tp->snd_una < th->th_ack <= tp->snd_max 2483 * then advance tp->snd_una to th->th_ack and drop 2484 * data from the retransmission queue. If this ACK reflects 2485 * more up to date window information we update our window information. 2486 */ 2487 case TCPS_ESTABLISHED: 2488 case TCPS_FIN_WAIT_1: 2489 case TCPS_FIN_WAIT_2: 2490 case TCPS_CLOSE_WAIT: 2491 case TCPS_CLOSING: 2492 case TCPS_LAST_ACK: 2493 case TCPS_TIME_WAIT: 2494 if (SEQ_LEQ(th->th_ack, tp->snd_una)) { 2495 if (tlen == 0 && !dupseg && tiwin == tp->snd_wnd) { 2496 TCP_STATINC(TCP_STAT_RCVDUPACK); 2497 /* 2498 * If we have outstanding data (other than 2499 * a window probe), this is a completely 2500 * duplicate ack (ie, window info didn't 2501 * change), the ack is the biggest we've 2502 * seen and we've seen exactly our rexmt 2503 * threshhold of them, assume a packet 2504 * has been dropped and retransmit it. 2505 * Kludge snd_nxt & the congestion 2506 * window so we send only this one 2507 * packet. 2508 */ 2509 if (TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 || 2510 th->th_ack != tp->snd_una) 2511 tp->t_dupacks = 0; 2512 else if (tp->t_partialacks < 0 && 2513 (++tp->t_dupacks == tcprexmtthresh || 2514 TCP_FACK_FASTRECOV(tp))) { 2515 /* 2516 * Do the fast retransmit, and adjust 2517 * congestion control paramenters. 2518 */ 2519 if (tp->t_congctl->fast_retransmit(tp, th)) { 2520 /* False fast retransmit */ 2521 break; 2522 } 2523 goto drop; 2524 } else if (tp->t_dupacks > tcprexmtthresh) { 2525 tp->snd_cwnd += tp->t_segsz; 2526 KERNEL_LOCK(1, NULL); 2527 (void)tcp_output(tp); 2528 KERNEL_UNLOCK_ONE(NULL); 2529 goto drop; 2530 } 2531 } else { 2532 /* 2533 * If the ack appears to be very old, only 2534 * allow data that is in-sequence. This 2535 * makes it somewhat more difficult to insert 2536 * forged data by guessing sequence numbers. 2537 * Sent an ack to try to update the send 2538 * sequence number on the other side. 2539 */ 2540 if (tlen && th->th_seq != tp->rcv_nxt && 2541 SEQ_LT(th->th_ack, 2542 tp->snd_una - tp->max_sndwnd)) 2543 goto dropafterack; 2544 } 2545 break; 2546 } 2547 /* 2548 * If the congestion window was inflated to account 2549 * for the other side's cached packets, retract it. 2550 */ 2551 tp->t_congctl->fast_retransmit_newack(tp, th); 2552 2553 if (SEQ_GT(th->th_ack, tp->snd_max)) { 2554 TCP_STATINC(TCP_STAT_RCVACKTOOMUCH); 2555 goto dropafterack; 2556 } 2557 acked = th->th_ack - tp->snd_una; 2558 tcps = TCP_STAT_GETREF(); 2559 tcps[TCP_STAT_RCVACKPACK]++; 2560 tcps[TCP_STAT_RCVACKBYTE] += acked; 2561 TCP_STAT_PUTREF(); 2562 2563 /* 2564 * If we have a timestamp reply, update smoothed 2565 * round trip time. If no timestamp is present but 2566 * transmit timer is running and timed sequence 2567 * number was acked, update smoothed round trip time. 2568 * Since we now have an rtt measurement, cancel the 2569 * timer backoff (cf., Phil Karn's retransmit alg.). 2570 * Recompute the initial retransmit timer. 2571 */ 2572 if (ts_rtt) 2573 tcp_xmit_timer(tp, ts_rtt - 1); 2574 else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq)) 2575 tcp_xmit_timer(tp, tcp_now - tp->t_rtttime); 2576 2577 /* 2578 * If all outstanding data is acked, stop retransmit 2579 * timer and remember to restart (more output or persist). 2580 * If there is more data to be acked, restart retransmit 2581 * timer, using current (possibly backed-off) value. 2582 */ 2583 if (th->th_ack == tp->snd_max) { 2584 TCP_TIMER_DISARM(tp, TCPT_REXMT); 2585 needoutput = 1; 2586 } else if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) 2587 TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur); 2588 2589 /* 2590 * New data has been acked, adjust the congestion window. 2591 */ 2592 tp->t_congctl->newack(tp, th); 2593 2594 nd6_hint(tp); 2595 if (acked > so->so_snd.sb_cc) { 2596 tp->snd_wnd -= so->so_snd.sb_cc; 2597 sbdrop(&so->so_snd, (int)so->so_snd.sb_cc); 2598 ourfinisacked = 1; 2599 } else { 2600 if (acked > (tp->t_lastoff - tp->t_inoff)) 2601 tp->t_lastm = NULL; 2602 sbdrop(&so->so_snd, acked); 2603 tp->t_lastoff -= acked; 2604 if (tp->snd_wnd > acked) 2605 tp->snd_wnd -= acked; 2606 else 2607 tp->snd_wnd = 0; 2608 ourfinisacked = 0; 2609 } 2610 sowwakeup(so); 2611 2612 icmp_check(tp, th, acked); 2613 2614 tp->snd_una = th->th_ack; 2615 if (SEQ_GT(tp->snd_una, tp->snd_fack)) 2616 tp->snd_fack = tp->snd_una; 2617 if (SEQ_LT(tp->snd_nxt, tp->snd_una)) 2618 tp->snd_nxt = tp->snd_una; 2619 if (SEQ_LT(tp->snd_high, tp->snd_una)) 2620 tp->snd_high = tp->snd_una; 2621 2622 switch (tp->t_state) { 2623 2624 /* 2625 * In FIN_WAIT_1 STATE in addition to the processing 2626 * for the ESTABLISHED state if our FIN is now acknowledged 2627 * then enter FIN_WAIT_2. 2628 */ 2629 case TCPS_FIN_WAIT_1: 2630 if (ourfinisacked) { 2631 /* 2632 * If we can't receive any more 2633 * data, then closing user can proceed. 2634 * Starting the timer is contrary to the 2635 * specification, but if we don't get a FIN 2636 * we'll hang forever. 2637 */ 2638 if (so->so_state & SS_CANTRCVMORE) { 2639 soisdisconnected(so); 2640 if (tp->t_maxidle > 0) 2641 TCP_TIMER_ARM(tp, TCPT_2MSL, 2642 tp->t_maxidle); 2643 } 2644 tp->t_state = TCPS_FIN_WAIT_2; 2645 } 2646 break; 2647 2648 /* 2649 * In CLOSING STATE in addition to the processing for 2650 * the ESTABLISHED state if the ACK acknowledges our FIN 2651 * then enter the TIME-WAIT state, otherwise ignore 2652 * the segment. 2653 */ 2654 case TCPS_CLOSING: 2655 if (ourfinisacked) { 2656 tp->t_state = TCPS_TIME_WAIT; 2657 tcp_canceltimers(tp); 2658 TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * tp->t_msl); 2659 soisdisconnected(so); 2660 } 2661 break; 2662 2663 /* 2664 * In LAST_ACK, we may still be waiting for data to drain 2665 * and/or to be acked, as well as for the ack of our FIN. 2666 * If our FIN is now acknowledged, delete the TCB, 2667 * enter the closed state and return. 2668 */ 2669 case TCPS_LAST_ACK: 2670 if (ourfinisacked) { 2671 tp = tcp_close(tp); 2672 goto drop; 2673 } 2674 break; 2675 2676 /* 2677 * In TIME_WAIT state the only thing that should arrive 2678 * is a retransmission of the remote FIN. Acknowledge 2679 * it and restart the finack timer. 2680 */ 2681 case TCPS_TIME_WAIT: 2682 TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * tp->t_msl); 2683 goto dropafterack; 2684 } 2685 } 2686 2687 step6: 2688 /* 2689 * Update window information. 2690 * Don't look at window if no ACK: TAC's send garbage on first SYN. 2691 */ 2692 if ((tiflags & TH_ACK) && (SEQ_LT(tp->snd_wl1, th->th_seq) || 2693 (tp->snd_wl1 == th->th_seq && (SEQ_LT(tp->snd_wl2, th->th_ack) || 2694 (tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd))))) { 2695 /* keep track of pure window updates */ 2696 if (tlen == 0 && 2697 tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd) 2698 TCP_STATINC(TCP_STAT_RCVWINUPD); 2699 tp->snd_wnd = tiwin; 2700 tp->snd_wl1 = th->th_seq; 2701 tp->snd_wl2 = th->th_ack; 2702 if (tp->snd_wnd > tp->max_sndwnd) 2703 tp->max_sndwnd = tp->snd_wnd; 2704 needoutput = 1; 2705 } 2706 2707 /* 2708 * Process segments with URG. 2709 */ 2710 if ((tiflags & TH_URG) && th->th_urp && 2711 TCPS_HAVERCVDFIN(tp->t_state) == 0) { 2712 /* 2713 * This is a kludge, but if we receive and accept 2714 * random urgent pointers, we'll crash in 2715 * soreceive. It's hard to imagine someone 2716 * actually wanting to send this much urgent data. 2717 */ 2718 if (th->th_urp + so->so_rcv.sb_cc > sb_max) { 2719 th->th_urp = 0; /* XXX */ 2720 tiflags &= ~TH_URG; /* XXX */ 2721 goto dodata; /* XXX */ 2722 } 2723 2724 /* 2725 * If this segment advances the known urgent pointer, 2726 * then mark the data stream. This should not happen 2727 * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since 2728 * a FIN has been received from the remote side. 2729 * In these states we ignore the URG. 2730 * 2731 * According to RFC961 (Assigned Protocols), 2732 * the urgent pointer points to the last octet 2733 * of urgent data. We continue, however, 2734 * to consider it to indicate the first octet 2735 * of data past the urgent section as the original 2736 * spec states (in one of two places). 2737 */ 2738 if (SEQ_GT(th->th_seq+th->th_urp, tp->rcv_up)) { 2739 tp->rcv_up = th->th_seq + th->th_urp; 2740 so->so_oobmark = so->so_rcv.sb_cc + 2741 (tp->rcv_up - tp->rcv_nxt) - 1; 2742 if (so->so_oobmark == 0) 2743 so->so_state |= SS_RCVATMARK; 2744 sohasoutofband(so); 2745 tp->t_oobflags &= ~(TCPOOB_HAVEDATA | TCPOOB_HADDATA); 2746 } 2747 2748 /* 2749 * Remove out of band data so doesn't get presented to user. 2750 * This can happen independent of advancing the URG pointer, 2751 * but if two URG's are pending at once, some out-of-band 2752 * data may creep in... ick. 2753 */ 2754 if (th->th_urp <= (u_int16_t)tlen && 2755 (so->so_options & SO_OOBINLINE) == 0) 2756 tcp_pulloutofband(so, th, m, hdroptlen); 2757 } else { 2758 /* 2759 * If no out of band data is expected, 2760 * pull receive urgent pointer along 2761 * with the receive window. 2762 */ 2763 if (SEQ_GT(tp->rcv_nxt, tp->rcv_up)) 2764 tp->rcv_up = tp->rcv_nxt; 2765 } 2766 dodata: 2767 2768 /* 2769 * Process the segment text, merging it into the TCP sequencing queue, 2770 * and arranging for acknowledgement of receipt if necessary. 2771 * This process logically involves adjusting tp->rcv_wnd as data 2772 * is presented to the user (this happens in tcp_usrreq.c, 2773 * tcp_rcvd()). If a FIN has already been received on this 2774 * connection then we just ignore the text. 2775 */ 2776 if ((tlen || (tiflags & TH_FIN)) && 2777 TCPS_HAVERCVDFIN(tp->t_state) == 0) { 2778 /* 2779 * Handle the common case: 2780 * o Segment is the next to be received, and 2781 * o The queue is empty, and 2782 * o The connection is established 2783 * In this case, we avoid calling tcp_reass. 2784 * 2785 * tcp_setup_ack: set DELACK for segments received in order, 2786 * but ack immediately when segments are out of order (so that 2787 * fast retransmit can work). 2788 */ 2789 TCP_REASS_LOCK(tp); 2790 if (th->th_seq == tp->rcv_nxt && 2791 TAILQ_FIRST(&tp->segq) == NULL && 2792 tp->t_state == TCPS_ESTABLISHED) { 2793 tcp_setup_ack(tp, th); 2794 tp->rcv_nxt += tlen; 2795 tiflags = th->th_flags & TH_FIN; 2796 tcps = TCP_STAT_GETREF(); 2797 tcps[TCP_STAT_RCVPACK]++; 2798 tcps[TCP_STAT_RCVBYTE] += tlen; 2799 TCP_STAT_PUTREF(); 2800 nd6_hint(tp); 2801 if (so->so_state & SS_CANTRCVMORE) { 2802 m_freem(m); 2803 } else { 2804 m_adj(m, hdroptlen); 2805 sbappendstream(&(so)->so_rcv, m); 2806 } 2807 TCP_REASS_UNLOCK(tp); 2808 sorwakeup(so); 2809 } else { 2810 m_adj(m, hdroptlen); 2811 tiflags = tcp_reass(tp, th, m, tlen); 2812 tp->t_flags |= TF_ACKNOW; 2813 } 2814 2815 /* 2816 * Note the amount of data that peer has sent into 2817 * our window, in order to estimate the sender's 2818 * buffer size. 2819 */ 2820 len = so->so_rcv.sb_hiwat - (tp->rcv_adv - tp->rcv_nxt); 2821 } else { 2822 m_freem(m); 2823 m = NULL; 2824 tiflags &= ~TH_FIN; 2825 } 2826 2827 /* 2828 * If FIN is received ACK the FIN and let the user know 2829 * that the connection is closing. Ignore a FIN received before 2830 * the connection is fully established. 2831 */ 2832 if ((tiflags & TH_FIN) && TCPS_HAVEESTABLISHED(tp->t_state)) { 2833 if (TCPS_HAVERCVDFIN(tp->t_state) == 0) { 2834 socantrcvmore(so); 2835 tp->t_flags |= TF_ACKNOW; 2836 tp->rcv_nxt++; 2837 } 2838 switch (tp->t_state) { 2839 2840 /* 2841 * In ESTABLISHED STATE enter the CLOSE_WAIT state. 2842 */ 2843 case TCPS_ESTABLISHED: 2844 tp->t_state = TCPS_CLOSE_WAIT; 2845 break; 2846 2847 /* 2848 * If still in FIN_WAIT_1 STATE FIN has not been acked so 2849 * enter the CLOSING state. 2850 */ 2851 case TCPS_FIN_WAIT_1: 2852 tp->t_state = TCPS_CLOSING; 2853 break; 2854 2855 /* 2856 * In FIN_WAIT_2 state enter the TIME_WAIT state, 2857 * starting the time-wait timer, turning off the other 2858 * standard timers. 2859 */ 2860 case TCPS_FIN_WAIT_2: 2861 tp->t_state = TCPS_TIME_WAIT; 2862 tcp_canceltimers(tp); 2863 TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * tp->t_msl); 2864 soisdisconnected(so); 2865 break; 2866 2867 /* 2868 * In TIME_WAIT state restart the 2 MSL time_wait timer. 2869 */ 2870 case TCPS_TIME_WAIT: 2871 TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * tp->t_msl); 2872 break; 2873 } 2874 } 2875 #ifdef TCP_DEBUG 2876 if (so->so_options & SO_DEBUG) 2877 tcp_trace(TA_INPUT, ostate, tp, tcp_saveti, 0); 2878 #endif 2879 2880 /* 2881 * Return any desired output. 2882 */ 2883 if (needoutput || (tp->t_flags & TF_ACKNOW)) { 2884 KERNEL_LOCK(1, NULL); 2885 (void)tcp_output(tp); 2886 KERNEL_UNLOCK_ONE(NULL); 2887 } 2888 if (tcp_saveti) 2889 m_freem(tcp_saveti); 2890 2891 if (tp->t_state == TCPS_TIME_WAIT 2892 && (so->so_state & SS_NOFDREF) 2893 && (tp->t_inpcb || af != AF_INET) 2894 && (tp->t_in6pcb || af != AF_INET6) 2895 && ((af == AF_INET ? tcp4_vtw_enable : tcp6_vtw_enable) & 1) != 0 2896 && TAILQ_EMPTY(&tp->segq) 2897 && vtw_add(af, tp)) { 2898 ; 2899 } 2900 return; 2901 2902 badsyn: 2903 /* 2904 * Received a bad SYN. Increment counters and dropwithreset. 2905 */ 2906 TCP_STATINC(TCP_STAT_BADSYN); 2907 tp = NULL; 2908 goto dropwithreset; 2909 2910 dropafterack: 2911 /* 2912 * Generate an ACK dropping incoming segment if it occupies 2913 * sequence space, where the ACK reflects our state. 2914 */ 2915 if (tiflags & TH_RST) 2916 goto drop; 2917 goto dropafterack2; 2918 2919 dropafterack_ratelim: 2920 /* 2921 * We may want to rate-limit ACKs against SYN/RST attack. 2922 */ 2923 if (ppsratecheck(&tcp_ackdrop_ppslim_last, &tcp_ackdrop_ppslim_count, 2924 tcp_ackdrop_ppslim) == 0) { 2925 /* XXX stat */ 2926 goto drop; 2927 } 2928 2929 dropafterack2: 2930 m_freem(m); 2931 tp->t_flags |= TF_ACKNOW; 2932 KERNEL_LOCK(1, NULL); 2933 (void)tcp_output(tp); 2934 KERNEL_UNLOCK_ONE(NULL); 2935 if (tcp_saveti) 2936 m_freem(tcp_saveti); 2937 return; 2938 2939 dropwithreset_ratelim: 2940 /* 2941 * We may want to rate-limit RSTs in certain situations, 2942 * particularly if we are sending an RST in response to 2943 * an attempt to connect to or otherwise communicate with 2944 * a port for which we have no socket. 2945 */ 2946 if (ppsratecheck(&tcp_rst_ppslim_last, &tcp_rst_ppslim_count, 2947 tcp_rst_ppslim) == 0) { 2948 /* XXX stat */ 2949 goto drop; 2950 } 2951 2952 dropwithreset: 2953 /* 2954 * Generate a RST, dropping incoming segment. 2955 * Make ACK acceptable to originator of segment. 2956 */ 2957 if (tiflags & TH_RST) 2958 goto drop; 2959 if (tiflags & TH_ACK) { 2960 (void)tcp_respond(tp, m, m, th, (tcp_seq)0, th->th_ack, TH_RST); 2961 } else { 2962 if (tiflags & TH_SYN) 2963 tlen++; 2964 (void)tcp_respond(tp, m, m, th, th->th_seq + tlen, (tcp_seq)0, 2965 TH_RST|TH_ACK); 2966 } 2967 if (tcp_saveti) 2968 m_freem(tcp_saveti); 2969 return; 2970 2971 badcsum: 2972 drop: 2973 /* 2974 * Drop space held by incoming segment and return. 2975 */ 2976 if (tp) { 2977 if (tp->t_inpcb) 2978 so = tp->t_inpcb->inp_socket; 2979 #ifdef INET6 2980 else if (tp->t_in6pcb) 2981 so = tp->t_in6pcb->in6p_socket; 2982 #endif 2983 else 2984 so = NULL; 2985 #ifdef TCP_DEBUG 2986 if (so && (so->so_options & SO_DEBUG) != 0) 2987 tcp_trace(TA_DROP, ostate, tp, tcp_saveti, 0); 2988 #endif 2989 } 2990 if (tcp_saveti) 2991 m_freem(tcp_saveti); 2992 m_freem(m); 2993 return; 2994 } 2995 2996 #ifdef TCP_SIGNATURE 2997 int 2998 tcp_signature_apply(void *fstate, void *data, u_int len) 2999 { 3000 3001 MD5Update(fstate, (u_char *)data, len); 3002 return (0); 3003 } 3004 3005 struct secasvar * 3006 tcp_signature_getsav(struct mbuf *m) 3007 { 3008 struct ip *ip; 3009 struct ip6_hdr *ip6; 3010 3011 ip = mtod(m, struct ip *); 3012 switch (ip->ip_v) { 3013 case 4: 3014 ip = mtod(m, struct ip *); 3015 ip6 = NULL; 3016 break; 3017 case 6: 3018 ip = NULL; 3019 ip6 = mtod(m, struct ip6_hdr *); 3020 break; 3021 default: 3022 return (NULL); 3023 } 3024 3025 #ifdef IPSEC 3026 union sockaddr_union dst; 3027 3028 /* Extract the destination from the IP header in the mbuf. */ 3029 memset(&dst, 0, sizeof(union sockaddr_union)); 3030 if (ip != NULL) { 3031 dst.sa.sa_len = sizeof(struct sockaddr_in); 3032 dst.sa.sa_family = AF_INET; 3033 dst.sin.sin_addr = ip->ip_dst; 3034 } else { 3035 dst.sa.sa_len = sizeof(struct sockaddr_in6); 3036 dst.sa.sa_family = AF_INET6; 3037 dst.sin6.sin6_addr = ip6->ip6_dst; 3038 } 3039 3040 /* 3041 * Look up an SADB entry which matches the address of the peer. 3042 */ 3043 return KEY_LOOKUP_SA(&dst, IPPROTO_TCP, htonl(TCP_SIG_SPI), 0, 0); 3044 #else 3045 return NULL; 3046 #endif 3047 } 3048 3049 int 3050 tcp_signature(struct mbuf *m, struct tcphdr *th, int thoff, 3051 struct secasvar *sav, char *sig) 3052 { 3053 MD5_CTX ctx; 3054 struct ip *ip; 3055 struct ipovly *ipovly; 3056 #ifdef INET6 3057 struct ip6_hdr *ip6; 3058 struct ip6_hdr_pseudo ip6pseudo; 3059 #endif 3060 struct ippseudo ippseudo; 3061 struct tcphdr th0; 3062 int l, tcphdrlen; 3063 3064 if (sav == NULL) 3065 return (-1); 3066 3067 tcphdrlen = th->th_off * 4; 3068 3069 switch (mtod(m, struct ip *)->ip_v) { 3070 case 4: 3071 MD5Init(&ctx); 3072 ip = mtod(m, struct ip *); 3073 memset(&ippseudo, 0, sizeof(ippseudo)); 3074 ipovly = (struct ipovly *)ip; 3075 ippseudo.ippseudo_src = ipovly->ih_src; 3076 ippseudo.ippseudo_dst = ipovly->ih_dst; 3077 ippseudo.ippseudo_pad = 0; 3078 ippseudo.ippseudo_p = IPPROTO_TCP; 3079 ippseudo.ippseudo_len = htons(m->m_pkthdr.len - thoff); 3080 MD5Update(&ctx, (char *)&ippseudo, sizeof(ippseudo)); 3081 break; 3082 #if INET6 3083 case 6: 3084 MD5Init(&ctx); 3085 ip6 = mtod(m, struct ip6_hdr *); 3086 memset(&ip6pseudo, 0, sizeof(ip6pseudo)); 3087 ip6pseudo.ip6ph_src = ip6->ip6_src; 3088 in6_clearscope(&ip6pseudo.ip6ph_src); 3089 ip6pseudo.ip6ph_dst = ip6->ip6_dst; 3090 in6_clearscope(&ip6pseudo.ip6ph_dst); 3091 ip6pseudo.ip6ph_len = htons(m->m_pkthdr.len - thoff); 3092 ip6pseudo.ip6ph_nxt = IPPROTO_TCP; 3093 MD5Update(&ctx, (char *)&ip6pseudo, sizeof(ip6pseudo)); 3094 break; 3095 #endif 3096 default: 3097 return (-1); 3098 } 3099 3100 th0 = *th; 3101 th0.th_sum = 0; 3102 MD5Update(&ctx, (char *)&th0, sizeof(th0)); 3103 3104 l = m->m_pkthdr.len - thoff - tcphdrlen; 3105 if (l > 0) 3106 m_apply(m, thoff + tcphdrlen, 3107 m->m_pkthdr.len - thoff - tcphdrlen, 3108 tcp_signature_apply, &ctx); 3109 3110 MD5Update(&ctx, _KEYBUF(sav->key_auth), _KEYLEN(sav->key_auth)); 3111 MD5Final(sig, &ctx); 3112 3113 return (0); 3114 } 3115 #endif 3116 3117 /* 3118 * Parse and process tcp options. 3119 * 3120 * Returns -1 if this segment should be dropped. (eg. wrong signature) 3121 * Otherwise returns 0. 3122 */ 3123 static int 3124 tcp_dooptions(struct tcpcb *tp, const u_char *cp, int cnt, struct tcphdr *th, 3125 struct mbuf *m, int toff, struct tcp_opt_info *oi) 3126 { 3127 u_int16_t mss; 3128 int opt, optlen = 0; 3129 #ifdef TCP_SIGNATURE 3130 void *sigp = NULL; 3131 char sigbuf[TCP_SIGLEN]; 3132 struct secasvar *sav = NULL; 3133 #endif 3134 3135 for (; cp && cnt > 0; cnt -= optlen, cp += optlen) { 3136 opt = cp[0]; 3137 if (opt == TCPOPT_EOL) 3138 break; 3139 if (opt == TCPOPT_NOP) 3140 optlen = 1; 3141 else { 3142 if (cnt < 2) 3143 break; 3144 optlen = cp[1]; 3145 if (optlen < 2 || optlen > cnt) 3146 break; 3147 } 3148 switch (opt) { 3149 3150 default: 3151 continue; 3152 3153 case TCPOPT_MAXSEG: 3154 if (optlen != TCPOLEN_MAXSEG) 3155 continue; 3156 if (!(th->th_flags & TH_SYN)) 3157 continue; 3158 if (TCPS_HAVERCVDSYN(tp->t_state)) 3159 continue; 3160 memcpy(&mss, cp + 2, sizeof(mss)); 3161 oi->maxseg = ntohs(mss); 3162 break; 3163 3164 case TCPOPT_WINDOW: 3165 if (optlen != TCPOLEN_WINDOW) 3166 continue; 3167 if (!(th->th_flags & TH_SYN)) 3168 continue; 3169 if (TCPS_HAVERCVDSYN(tp->t_state)) 3170 continue; 3171 tp->t_flags |= TF_RCVD_SCALE; 3172 tp->requested_s_scale = cp[2]; 3173 if (tp->requested_s_scale > TCP_MAX_WINSHIFT) { 3174 char buf[INET6_ADDRSTRLEN]; 3175 struct ip *ip = mtod(m, struct ip *); 3176 #ifdef INET6 3177 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 3178 #endif 3179 3180 switch (ip->ip_v) { 3181 case 4: 3182 in_print(buf, sizeof(buf), 3183 &ip->ip_src); 3184 break; 3185 #ifdef INET6 3186 case 6: 3187 in6_print(buf, sizeof(buf), 3188 &ip6->ip6_src); 3189 break; 3190 #endif 3191 default: 3192 strlcpy(buf, "(unknown)", sizeof(buf)); 3193 break; 3194 } 3195 3196 log(LOG_ERR, "TCP: invalid wscale %d from %s, " 3197 "assuming %d\n", 3198 tp->requested_s_scale, buf, 3199 TCP_MAX_WINSHIFT); 3200 tp->requested_s_scale = TCP_MAX_WINSHIFT; 3201 } 3202 break; 3203 3204 case TCPOPT_TIMESTAMP: 3205 if (optlen != TCPOLEN_TIMESTAMP) 3206 continue; 3207 oi->ts_present = 1; 3208 memcpy(&oi->ts_val, cp + 2, sizeof(oi->ts_val)); 3209 NTOHL(oi->ts_val); 3210 memcpy(&oi->ts_ecr, cp + 6, sizeof(oi->ts_ecr)); 3211 NTOHL(oi->ts_ecr); 3212 3213 if (!(th->th_flags & TH_SYN)) 3214 continue; 3215 if (TCPS_HAVERCVDSYN(tp->t_state)) 3216 continue; 3217 /* 3218 * A timestamp received in a SYN makes 3219 * it ok to send timestamp requests and replies. 3220 */ 3221 tp->t_flags |= TF_RCVD_TSTMP; 3222 tp->ts_recent = oi->ts_val; 3223 tp->ts_recent_age = tcp_now; 3224 break; 3225 3226 case TCPOPT_SACK_PERMITTED: 3227 if (optlen != TCPOLEN_SACK_PERMITTED) 3228 continue; 3229 if (!(th->th_flags & TH_SYN)) 3230 continue; 3231 if (TCPS_HAVERCVDSYN(tp->t_state)) 3232 continue; 3233 if (tcp_do_sack) { 3234 tp->t_flags |= TF_SACK_PERMIT; 3235 tp->t_flags |= TF_WILL_SACK; 3236 } 3237 break; 3238 3239 case TCPOPT_SACK: 3240 tcp_sack_option(tp, th, cp, optlen); 3241 break; 3242 #ifdef TCP_SIGNATURE 3243 case TCPOPT_SIGNATURE: 3244 if (optlen != TCPOLEN_SIGNATURE) 3245 continue; 3246 if (sigp && 3247 !consttime_memequal(sigp, cp + 2, TCP_SIGLEN)) 3248 return (-1); 3249 3250 sigp = sigbuf; 3251 memcpy(sigbuf, cp + 2, TCP_SIGLEN); 3252 tp->t_flags |= TF_SIGNATURE; 3253 break; 3254 #endif 3255 } 3256 } 3257 3258 #ifndef TCP_SIGNATURE 3259 return 0; 3260 #else 3261 if (tp->t_flags & TF_SIGNATURE) { 3262 sav = tcp_signature_getsav(m); 3263 if (sav == NULL && tp->t_state == TCPS_LISTEN) 3264 return (-1); 3265 } 3266 3267 if ((sigp ? TF_SIGNATURE : 0) ^ (tp->t_flags & TF_SIGNATURE)) 3268 goto out; 3269 3270 if (sigp) { 3271 char sig[TCP_SIGLEN]; 3272 3273 tcp_fields_to_net(th); 3274 if (tcp_signature(m, th, toff, sav, sig) < 0) { 3275 tcp_fields_to_host(th); 3276 goto out; 3277 } 3278 tcp_fields_to_host(th); 3279 3280 if (!consttime_memequal(sig, sigp, TCP_SIGLEN)) { 3281 TCP_STATINC(TCP_STAT_BADSIG); 3282 goto out; 3283 } else 3284 TCP_STATINC(TCP_STAT_GOODSIG); 3285 3286 key_sa_recordxfer(sav, m); 3287 KEY_SA_UNREF(&sav); 3288 } 3289 return 0; 3290 out: 3291 if (sav != NULL) 3292 KEY_SA_UNREF(&sav); 3293 return -1; 3294 #endif 3295 } 3296 3297 /* 3298 * Pull out of band byte out of a segment so 3299 * it doesn't appear in the user's data queue. 3300 * It is still reflected in the segment length for 3301 * sequencing purposes. 3302 */ 3303 void 3304 tcp_pulloutofband(struct socket *so, struct tcphdr *th, 3305 struct mbuf *m, int off) 3306 { 3307 int cnt = off + th->th_urp - 1; 3308 3309 while (cnt >= 0) { 3310 if (m->m_len > cnt) { 3311 char *cp = mtod(m, char *) + cnt; 3312 struct tcpcb *tp = sototcpcb(so); 3313 3314 tp->t_iobc = *cp; 3315 tp->t_oobflags |= TCPOOB_HAVEDATA; 3316 memmove(cp, cp + 1, (unsigned)(m->m_len - cnt - 1)); 3317 m->m_len--; 3318 return; 3319 } 3320 cnt -= m->m_len; 3321 m = m->m_next; 3322 if (m == NULL) 3323 break; 3324 } 3325 panic("tcp_pulloutofband"); 3326 } 3327 3328 /* 3329 * Collect new round-trip time estimate 3330 * and update averages and current timeout. 3331 * 3332 * rtt is in units of slow ticks (typically 500 ms) -- essentially the 3333 * difference of two timestamps. 3334 */ 3335 void 3336 tcp_xmit_timer(struct tcpcb *tp, uint32_t rtt) 3337 { 3338 int32_t delta; 3339 3340 TCP_STATINC(TCP_STAT_RTTUPDATED); 3341 if (tp->t_srtt != 0) { 3342 /* 3343 * Compute the amount to add to srtt for smoothing, 3344 * *alpha, or 2^(-TCP_RTT_SHIFT). Because 3345 * srtt is stored in 1/32 slow ticks, we conceptually 3346 * shift left 5 bits, subtract srtt to get the 3347 * diference, and then shift right by TCP_RTT_SHIFT 3348 * (3) to obtain 1/8 of the difference. 3349 */ 3350 delta = (rtt << 2) - (tp->t_srtt >> TCP_RTT_SHIFT); 3351 /* 3352 * This can never happen, because delta's lowest 3353 * possible value is 1/8 of t_srtt. But if it does, 3354 * set srtt to some reasonable value, here chosen 3355 * as 1/8 tick. 3356 */ 3357 if ((tp->t_srtt += delta) <= 0) 3358 tp->t_srtt = 1 << 2; 3359 /* 3360 * RFC2988 requires that rttvar be updated first. 3361 * This code is compliant because "delta" is the old 3362 * srtt minus the new observation (scaled). 3363 * 3364 * RFC2988 says: 3365 * rttvar = (1-beta) * rttvar + beta * |srtt-observed| 3366 * 3367 * delta is in units of 1/32 ticks, and has then been 3368 * divided by 8. This is equivalent to being in 1/16s 3369 * units and divided by 4. Subtract from it 1/4 of 3370 * the existing rttvar to form the (signed) amount to 3371 * adjust. 3372 */ 3373 if (delta < 0) 3374 delta = -delta; 3375 delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT); 3376 /* 3377 * As with srtt, this should never happen. There is 3378 * no support in RFC2988 for this operation. But 1/4s 3379 * as rttvar when faced with something arguably wrong 3380 * is ok. 3381 */ 3382 if ((tp->t_rttvar += delta) <= 0) 3383 tp->t_rttvar = 1 << 2; 3384 3385 /* 3386 * If srtt exceeds .01 second, ensure we use the 'remote' MSL 3387 * Problem is: it doesn't work. Disabled by defaulting 3388 * tcp_rttlocal to 0; see corresponding code in 3389 * tcp_subr that selects local vs remote in a different way. 3390 * 3391 * The static branch prediction hint here should be removed 3392 * when the rtt estimator is fixed and the rtt_enable code 3393 * is turned back on. 3394 */ 3395 if (__predict_false(tcp_rttlocal) && tcp_msl_enable 3396 && tp->t_srtt > tcp_msl_remote_threshold 3397 && tp->t_msl < tcp_msl_remote) { 3398 tp->t_msl = tcp_msl_remote; 3399 } 3400 } else { 3401 /* 3402 * This is the first measurement. Per RFC2988, 2.2, 3403 * set rtt=R and srtt=R/2. 3404 * For srtt, storage representation is 1/32 ticks, 3405 * so shift left by 5. 3406 * For rttvar, storage representation is 1/16 ticks, 3407 * So shift left by 4, but then right by 1 to halve. 3408 */ 3409 tp->t_srtt = rtt << (TCP_RTT_SHIFT + 2); 3410 tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT + 2 - 1); 3411 } 3412 tp->t_rtttime = 0; 3413 tp->t_rxtshift = 0; 3414 3415 /* 3416 * the retransmit should happen at rtt + 4 * rttvar. 3417 * Because of the way we do the smoothing, srtt and rttvar 3418 * will each average +1/2 tick of bias. When we compute 3419 * the retransmit timer, we want 1/2 tick of rounding and 3420 * 1 extra tick because of +-1/2 tick uncertainty in the 3421 * firing of the timer. The bias will give us exactly the 3422 * 1.5 tick we need. But, because the bias is 3423 * statistical, we have to test that we don't drop below 3424 * the minimum feasible timer (which is 2 ticks). 3425 */ 3426 TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp), 3427 max(tp->t_rttmin, rtt + 2), TCPTV_REXMTMAX); 3428 3429 /* 3430 * We received an ack for a packet that wasn't retransmitted; 3431 * it is probably safe to discard any error indications we've 3432 * received recently. This isn't quite right, but close enough 3433 * for now (a route might have failed after we sent a segment, 3434 * and the return path might not be symmetrical). 3435 */ 3436 tp->t_softerror = 0; 3437 } 3438 3439 3440 /* 3441 * TCP compressed state engine. Currently used to hold compressed 3442 * state for SYN_RECEIVED. 3443 */ 3444 3445 u_long syn_cache_count; 3446 u_int32_t syn_hash1, syn_hash2; 3447 3448 #define SYN_HASH(sa, sp, dp) \ 3449 ((((sa)->s_addr^syn_hash1)*(((((u_int32_t)(dp))<<16) + \ 3450 ((u_int32_t)(sp)))^syn_hash2))) 3451 #ifndef INET6 3452 #define SYN_HASHALL(hash, src, dst) \ 3453 do { \ 3454 hash = SYN_HASH(&((const struct sockaddr_in *)(src))->sin_addr, \ 3455 ((const struct sockaddr_in *)(src))->sin_port, \ 3456 ((const struct sockaddr_in *)(dst))->sin_port); \ 3457 } while (/*CONSTCOND*/ 0) 3458 #else 3459 #define SYN_HASH6(sa, sp, dp) \ 3460 ((((sa)->s6_addr32[0] ^ (sa)->s6_addr32[3] ^ syn_hash1) * \ 3461 (((((u_int32_t)(dp))<<16) + ((u_int32_t)(sp)))^syn_hash2)) \ 3462 & 0x7fffffff) 3463 3464 #define SYN_HASHALL(hash, src, dst) \ 3465 do { \ 3466 switch ((src)->sa_family) { \ 3467 case AF_INET: \ 3468 hash = SYN_HASH(&((const struct sockaddr_in *)(src))->sin_addr, \ 3469 ((const struct sockaddr_in *)(src))->sin_port, \ 3470 ((const struct sockaddr_in *)(dst))->sin_port); \ 3471 break; \ 3472 case AF_INET6: \ 3473 hash = SYN_HASH6(&((const struct sockaddr_in6 *)(src))->sin6_addr, \ 3474 ((const struct sockaddr_in6 *)(src))->sin6_port, \ 3475 ((const struct sockaddr_in6 *)(dst))->sin6_port); \ 3476 break; \ 3477 default: \ 3478 hash = 0; \ 3479 } \ 3480 } while (/*CONSTCOND*/0) 3481 #endif /* INET6 */ 3482 3483 static struct pool syn_cache_pool; 3484 3485 /* 3486 * We don't estimate RTT with SYNs, so each packet starts with the default 3487 * RTT and each timer step has a fixed timeout value. 3488 */ 3489 static inline void 3490 syn_cache_timer_arm(struct syn_cache *sc) 3491 { 3492 3493 TCPT_RANGESET(sc->sc_rxtcur, 3494 TCPTV_SRTTDFLT * tcp_backoff[sc->sc_rxtshift], TCPTV_MIN, 3495 TCPTV_REXMTMAX); 3496 callout_reset(&sc->sc_timer, 3497 sc->sc_rxtcur * (hz / PR_SLOWHZ), syn_cache_timer, sc); 3498 } 3499 3500 #define SYN_CACHE_TIMESTAMP(sc) (tcp_now - (sc)->sc_timebase) 3501 3502 static inline void 3503 syn_cache_rm(struct syn_cache *sc) 3504 { 3505 TAILQ_REMOVE(&tcp_syn_cache[sc->sc_bucketidx].sch_bucket, 3506 sc, sc_bucketq); 3507 sc->sc_tp = NULL; 3508 LIST_REMOVE(sc, sc_tpq); 3509 tcp_syn_cache[sc->sc_bucketidx].sch_length--; 3510 callout_stop(&sc->sc_timer); 3511 syn_cache_count--; 3512 } 3513 3514 static inline void 3515 syn_cache_put(struct syn_cache *sc) 3516 { 3517 if (sc->sc_ipopts) 3518 (void) m_free(sc->sc_ipopts); 3519 rtcache_free(&sc->sc_route); 3520 sc->sc_flags |= SCF_DEAD; 3521 if (!callout_invoking(&sc->sc_timer)) 3522 callout_schedule(&(sc)->sc_timer, 1); 3523 } 3524 3525 void 3526 syn_cache_init(void) 3527 { 3528 int i; 3529 3530 pool_init(&syn_cache_pool, sizeof(struct syn_cache), 0, 0, 0, 3531 "synpl", NULL, IPL_SOFTNET); 3532 3533 /* Initialize the hash buckets. */ 3534 for (i = 0; i < tcp_syn_cache_size; i++) 3535 TAILQ_INIT(&tcp_syn_cache[i].sch_bucket); 3536 } 3537 3538 void 3539 syn_cache_insert(struct syn_cache *sc, struct tcpcb *tp) 3540 { 3541 struct syn_cache_head *scp; 3542 struct syn_cache *sc2; 3543 int s; 3544 3545 /* 3546 * If there are no entries in the hash table, reinitialize 3547 * the hash secrets. 3548 */ 3549 if (syn_cache_count == 0) { 3550 syn_hash1 = cprng_fast32(); 3551 syn_hash2 = cprng_fast32(); 3552 } 3553 3554 SYN_HASHALL(sc->sc_hash, &sc->sc_src.sa, &sc->sc_dst.sa); 3555 sc->sc_bucketidx = sc->sc_hash % tcp_syn_cache_size; 3556 scp = &tcp_syn_cache[sc->sc_bucketidx]; 3557 3558 /* 3559 * Make sure that we don't overflow the per-bucket 3560 * limit or the total cache size limit. 3561 */ 3562 s = splsoftnet(); 3563 if (scp->sch_length >= tcp_syn_bucket_limit) { 3564 TCP_STATINC(TCP_STAT_SC_BUCKETOVERFLOW); 3565 /* 3566 * The bucket is full. Toss the oldest element in the 3567 * bucket. This will be the first entry in the bucket. 3568 */ 3569 sc2 = TAILQ_FIRST(&scp->sch_bucket); 3570 #ifdef DIAGNOSTIC 3571 /* 3572 * This should never happen; we should always find an 3573 * entry in our bucket. 3574 */ 3575 if (sc2 == NULL) 3576 panic("syn_cache_insert: bucketoverflow: impossible"); 3577 #endif 3578 syn_cache_rm(sc2); 3579 syn_cache_put(sc2); /* calls pool_put but see spl above */ 3580 } else if (syn_cache_count >= tcp_syn_cache_limit) { 3581 struct syn_cache_head *scp2, *sce; 3582 3583 TCP_STATINC(TCP_STAT_SC_OVERFLOWED); 3584 /* 3585 * The cache is full. Toss the oldest entry in the 3586 * first non-empty bucket we can find. 3587 * 3588 * XXX We would really like to toss the oldest 3589 * entry in the cache, but we hope that this 3590 * condition doesn't happen very often. 3591 */ 3592 scp2 = scp; 3593 if (TAILQ_EMPTY(&scp2->sch_bucket)) { 3594 sce = &tcp_syn_cache[tcp_syn_cache_size]; 3595 for (++scp2; scp2 != scp; scp2++) { 3596 if (scp2 >= sce) 3597 scp2 = &tcp_syn_cache[0]; 3598 if (! TAILQ_EMPTY(&scp2->sch_bucket)) 3599 break; 3600 } 3601 #ifdef DIAGNOSTIC 3602 /* 3603 * This should never happen; we should always find a 3604 * non-empty bucket. 3605 */ 3606 if (scp2 == scp) 3607 panic("syn_cache_insert: cacheoverflow: " 3608 "impossible"); 3609 #endif 3610 } 3611 sc2 = TAILQ_FIRST(&scp2->sch_bucket); 3612 syn_cache_rm(sc2); 3613 syn_cache_put(sc2); /* calls pool_put but see spl above */ 3614 } 3615 3616 /* 3617 * Initialize the entry's timer. 3618 */ 3619 sc->sc_rxttot = 0; 3620 sc->sc_rxtshift = 0; 3621 syn_cache_timer_arm(sc); 3622 3623 /* Link it from tcpcb entry */ 3624 LIST_INSERT_HEAD(&tp->t_sc, sc, sc_tpq); 3625 3626 /* Put it into the bucket. */ 3627 TAILQ_INSERT_TAIL(&scp->sch_bucket, sc, sc_bucketq); 3628 scp->sch_length++; 3629 syn_cache_count++; 3630 3631 TCP_STATINC(TCP_STAT_SC_ADDED); 3632 splx(s); 3633 } 3634 3635 /* 3636 * Walk the timer queues, looking for SYN,ACKs that need to be retransmitted. 3637 * If we have retransmitted an entry the maximum number of times, expire 3638 * that entry. 3639 */ 3640 static void 3641 syn_cache_timer(void *arg) 3642 { 3643 struct syn_cache *sc = arg; 3644 3645 mutex_enter(softnet_lock); 3646 KERNEL_LOCK(1, NULL); 3647 3648 callout_ack(&sc->sc_timer); 3649 3650 if (__predict_false(sc->sc_flags & SCF_DEAD)) { 3651 TCP_STATINC(TCP_STAT_SC_DELAYED_FREE); 3652 goto free; 3653 } 3654 3655 if (__predict_false(sc->sc_rxtshift == TCP_MAXRXTSHIFT)) { 3656 /* Drop it -- too many retransmissions. */ 3657 goto dropit; 3658 } 3659 3660 /* 3661 * Compute the total amount of time this entry has 3662 * been on a queue. If this entry has been on longer 3663 * than the keep alive timer would allow, expire it. 3664 */ 3665 sc->sc_rxttot += sc->sc_rxtcur; 3666 if (sc->sc_rxttot >= tcp_keepinit) 3667 goto dropit; 3668 3669 TCP_STATINC(TCP_STAT_SC_RETRANSMITTED); 3670 (void)syn_cache_respond(sc); 3671 3672 /* Advance the timer back-off. */ 3673 sc->sc_rxtshift++; 3674 syn_cache_timer_arm(sc); 3675 3676 goto out; 3677 3678 dropit: 3679 TCP_STATINC(TCP_STAT_SC_TIMED_OUT); 3680 syn_cache_rm(sc); 3681 if (sc->sc_ipopts) 3682 (void) m_free(sc->sc_ipopts); 3683 rtcache_free(&sc->sc_route); 3684 3685 free: 3686 callout_destroy(&sc->sc_timer); 3687 pool_put(&syn_cache_pool, sc); 3688 3689 out: 3690 KERNEL_UNLOCK_ONE(NULL); 3691 mutex_exit(softnet_lock); 3692 } 3693 3694 /* 3695 * Remove syn cache created by the specified tcb entry, 3696 * because this does not make sense to keep them 3697 * (if there's no tcb entry, syn cache entry will never be used) 3698 */ 3699 void 3700 syn_cache_cleanup(struct tcpcb *tp) 3701 { 3702 struct syn_cache *sc, *nsc; 3703 int s; 3704 3705 s = splsoftnet(); 3706 3707 for (sc = LIST_FIRST(&tp->t_sc); sc != NULL; sc = nsc) { 3708 nsc = LIST_NEXT(sc, sc_tpq); 3709 3710 #ifdef DIAGNOSTIC 3711 if (sc->sc_tp != tp) 3712 panic("invalid sc_tp in syn_cache_cleanup"); 3713 #endif 3714 syn_cache_rm(sc); 3715 syn_cache_put(sc); /* calls pool_put but see spl above */ 3716 } 3717 /* just for safety */ 3718 LIST_INIT(&tp->t_sc); 3719 3720 splx(s); 3721 } 3722 3723 /* 3724 * Find an entry in the syn cache. 3725 */ 3726 struct syn_cache * 3727 syn_cache_lookup(const struct sockaddr *src, const struct sockaddr *dst, 3728 struct syn_cache_head **headp) 3729 { 3730 struct syn_cache *sc; 3731 struct syn_cache_head *scp; 3732 u_int32_t hash; 3733 int s; 3734 3735 SYN_HASHALL(hash, src, dst); 3736 3737 scp = &tcp_syn_cache[hash % tcp_syn_cache_size]; 3738 *headp = scp; 3739 s = splsoftnet(); 3740 for (sc = TAILQ_FIRST(&scp->sch_bucket); sc != NULL; 3741 sc = TAILQ_NEXT(sc, sc_bucketq)) { 3742 if (sc->sc_hash != hash) 3743 continue; 3744 if (!memcmp(&sc->sc_src, src, src->sa_len) && 3745 !memcmp(&sc->sc_dst, dst, dst->sa_len)) { 3746 splx(s); 3747 return (sc); 3748 } 3749 } 3750 splx(s); 3751 return (NULL); 3752 } 3753 3754 /* 3755 * This function gets called when we receive an ACK for a socket in the 3756 * LISTEN state. We look up the connection in the syn cache, and if it's 3757 * there, we pull it out of the cache and turn it into a full-blown 3758 * connection in the SYN-RECEIVED state. 3759 * 3760 * The return values may not be immediately obvious, and their effects 3761 * can be subtle, so here they are: 3762 * 3763 * NULL SYN was not found in cache; caller should drop the 3764 * packet and send an RST. 3765 * 3766 * -1 We were unable to create the new connection, and are 3767 * aborting it. An ACK,RST is being sent to the peer 3768 * (unless we got screwey sequence numbers; see below), 3769 * because the 3-way handshake has been completed. Caller 3770 * should not free the mbuf, since we may be using it. If 3771 * we are not, we will free it. 3772 * 3773 * Otherwise, the return value is a pointer to the new socket 3774 * associated with the connection. 3775 */ 3776 struct socket * 3777 syn_cache_get(struct sockaddr *src, struct sockaddr *dst, 3778 struct tcphdr *th, struct socket *so, struct mbuf *m) 3779 { 3780 struct syn_cache *sc; 3781 struct syn_cache_head *scp; 3782 struct inpcb *inp = NULL; 3783 #ifdef INET6 3784 struct in6pcb *in6p = NULL; 3785 #endif 3786 struct tcpcb *tp; 3787 int s; 3788 struct socket *oso; 3789 3790 s = splsoftnet(); 3791 if ((sc = syn_cache_lookup(src, dst, &scp)) == NULL) { 3792 splx(s); 3793 return NULL; 3794 } 3795 3796 /* 3797 * Verify the sequence and ack numbers. Try getting the correct 3798 * response again. 3799 */ 3800 if ((th->th_ack != sc->sc_iss + 1) || 3801 SEQ_LEQ(th->th_seq, sc->sc_irs) || 3802 SEQ_GT(th->th_seq, sc->sc_irs + 1 + sc->sc_win)) { 3803 m_freem(m); 3804 (void)syn_cache_respond(sc); 3805 splx(s); 3806 return ((struct socket *)(-1)); 3807 } 3808 3809 /* Remove this cache entry */ 3810 syn_cache_rm(sc); 3811 splx(s); 3812 3813 /* 3814 * Ok, create the full blown connection, and set things up 3815 * as they would have been set up if we had created the 3816 * connection when the SYN arrived. If we can't create 3817 * the connection, abort it. 3818 */ 3819 /* 3820 * inp still has the OLD in_pcb stuff, set the 3821 * v6-related flags on the new guy, too. This is 3822 * done particularly for the case where an AF_INET6 3823 * socket is bound only to a port, and a v4 connection 3824 * comes in on that port. 3825 * we also copy the flowinfo from the original pcb 3826 * to the new one. 3827 */ 3828 oso = so; 3829 so = sonewconn(so, true); 3830 if (so == NULL) 3831 goto resetandabort; 3832 3833 switch (so->so_proto->pr_domain->dom_family) { 3834 case AF_INET: 3835 inp = sotoinpcb(so); 3836 break; 3837 #ifdef INET6 3838 case AF_INET6: 3839 in6p = sotoin6pcb(so); 3840 break; 3841 #endif 3842 } 3843 3844 switch (src->sa_family) { 3845 case AF_INET: 3846 if (inp) { 3847 inp->inp_laddr = ((struct sockaddr_in *)dst)->sin_addr; 3848 inp->inp_lport = ((struct sockaddr_in *)dst)->sin_port; 3849 inp->inp_options = ip_srcroute(m); 3850 in_pcbstate(inp, INP_BOUND); 3851 if (inp->inp_options == NULL) { 3852 inp->inp_options = sc->sc_ipopts; 3853 sc->sc_ipopts = NULL; 3854 } 3855 } 3856 #ifdef INET6 3857 else if (in6p) { 3858 /* IPv4 packet to AF_INET6 socket */ 3859 memset(&in6p->in6p_laddr, 0, sizeof(in6p->in6p_laddr)); 3860 in6p->in6p_laddr.s6_addr16[5] = htons(0xffff); 3861 bcopy(&((struct sockaddr_in *)dst)->sin_addr, 3862 &in6p->in6p_laddr.s6_addr32[3], 3863 sizeof(((struct sockaddr_in *)dst)->sin_addr)); 3864 in6p->in6p_lport = ((struct sockaddr_in *)dst)->sin_port; 3865 in6totcpcb(in6p)->t_family = AF_INET; 3866 if (sotoin6pcb(oso)->in6p_flags & IN6P_IPV6_V6ONLY) 3867 in6p->in6p_flags |= IN6P_IPV6_V6ONLY; 3868 else 3869 in6p->in6p_flags &= ~IN6P_IPV6_V6ONLY; 3870 in6_pcbstate(in6p, IN6P_BOUND); 3871 } 3872 #endif 3873 break; 3874 #ifdef INET6 3875 case AF_INET6: 3876 if (in6p) { 3877 in6p->in6p_laddr = ((struct sockaddr_in6 *)dst)->sin6_addr; 3878 in6p->in6p_lport = ((struct sockaddr_in6 *)dst)->sin6_port; 3879 in6_pcbstate(in6p, IN6P_BOUND); 3880 } 3881 break; 3882 #endif 3883 } 3884 3885 #ifdef INET6 3886 if (in6p && in6totcpcb(in6p)->t_family == AF_INET6 && sotoinpcb(oso)) { 3887 struct in6pcb *oin6p = sotoin6pcb(oso); 3888 /* inherit socket options from the listening socket */ 3889 in6p->in6p_flags |= (oin6p->in6p_flags & IN6P_CONTROLOPTS); 3890 if (in6p->in6p_flags & IN6P_CONTROLOPTS) { 3891 m_freem(in6p->in6p_options); 3892 in6p->in6p_options = NULL; 3893 } 3894 ip6_savecontrol(in6p, &in6p->in6p_options, 3895 mtod(m, struct ip6_hdr *), m); 3896 } 3897 #endif 3898 3899 #if defined(IPSEC) 3900 if (ipsec_used) { 3901 /* 3902 * we make a copy of policy, instead of sharing the policy, for 3903 * better behavior in terms of SA lookup and dead SA removal. 3904 */ 3905 if (inp) { 3906 /* copy old policy into new socket's */ 3907 if (ipsec_copy_pcbpolicy(sotoinpcb(oso)->inp_sp, 3908 inp->inp_sp)) 3909 printf("tcp_input: could not copy policy\n"); 3910 } 3911 #ifdef INET6 3912 else if (in6p) { 3913 /* copy old policy into new socket's */ 3914 if (ipsec_copy_pcbpolicy(sotoin6pcb(oso)->in6p_sp, 3915 in6p->in6p_sp)) 3916 printf("tcp_input: could not copy policy\n"); 3917 } 3918 #endif 3919 } 3920 #endif 3921 3922 /* 3923 * Give the new socket our cached route reference. 3924 */ 3925 if (inp) { 3926 rtcache_copy(&inp->inp_route, &sc->sc_route); 3927 rtcache_free(&sc->sc_route); 3928 } 3929 #ifdef INET6 3930 else { 3931 rtcache_copy(&in6p->in6p_route, &sc->sc_route); 3932 rtcache_free(&sc->sc_route); 3933 } 3934 #endif 3935 3936 if (inp) { 3937 struct sockaddr_in sin; 3938 memcpy(&sin, src, src->sa_len); 3939 if (in_pcbconnect(inp, &sin, &lwp0)) { 3940 goto resetandabort; 3941 } 3942 } 3943 #ifdef INET6 3944 else if (in6p) { 3945 struct sockaddr_in6 sin6; 3946 memcpy(&sin6, src, src->sa_len); 3947 if (src->sa_family == AF_INET) { 3948 /* IPv4 packet to AF_INET6 socket */ 3949 in6_sin_2_v4mapsin6((struct sockaddr_in *)src, &sin6); 3950 } 3951 if (in6_pcbconnect(in6p, &sin6, NULL)) { 3952 goto resetandabort; 3953 } 3954 } 3955 #endif 3956 else { 3957 goto resetandabort; 3958 } 3959 3960 if (inp) 3961 tp = intotcpcb(inp); 3962 #ifdef INET6 3963 else if (in6p) 3964 tp = in6totcpcb(in6p); 3965 #endif 3966 else 3967 tp = NULL; 3968 3969 tp->t_flags = sototcpcb(oso)->t_flags & TF_NODELAY; 3970 if (sc->sc_request_r_scale != 15) { 3971 tp->requested_s_scale = sc->sc_requested_s_scale; 3972 tp->request_r_scale = sc->sc_request_r_scale; 3973 tp->snd_scale = sc->sc_requested_s_scale; 3974 tp->rcv_scale = sc->sc_request_r_scale; 3975 tp->t_flags |= TF_REQ_SCALE|TF_RCVD_SCALE; 3976 } 3977 if (sc->sc_flags & SCF_TIMESTAMP) 3978 tp->t_flags |= TF_REQ_TSTMP|TF_RCVD_TSTMP; 3979 tp->ts_timebase = sc->sc_timebase; 3980 3981 tp->t_template = tcp_template(tp); 3982 if (tp->t_template == 0) { 3983 tp = tcp_drop(tp, ENOBUFS); /* destroys socket */ 3984 so = NULL; 3985 m_freem(m); 3986 goto abort; 3987 } 3988 3989 tp->iss = sc->sc_iss; 3990 tp->irs = sc->sc_irs; 3991 tcp_sendseqinit(tp); 3992 tcp_rcvseqinit(tp); 3993 tp->t_state = TCPS_SYN_RECEIVED; 3994 TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit); 3995 TCP_STATINC(TCP_STAT_ACCEPTS); 3996 3997 if ((sc->sc_flags & SCF_SACK_PERMIT) && tcp_do_sack) 3998 tp->t_flags |= TF_WILL_SACK; 3999 4000 if ((sc->sc_flags & SCF_ECN_PERMIT) && tcp_do_ecn) 4001 tp->t_flags |= TF_ECN_PERMIT; 4002 4003 #ifdef TCP_SIGNATURE 4004 if (sc->sc_flags & SCF_SIGNATURE) 4005 tp->t_flags |= TF_SIGNATURE; 4006 #endif 4007 4008 /* Initialize tp->t_ourmss before we deal with the peer's! */ 4009 tp->t_ourmss = sc->sc_ourmaxseg; 4010 tcp_mss_from_peer(tp, sc->sc_peermaxseg); 4011 4012 /* 4013 * Initialize the initial congestion window. If we 4014 * had to retransmit the SYN,ACK, we must initialize cwnd 4015 * to 1 segment (i.e. the Loss Window). 4016 */ 4017 if (sc->sc_rxtshift) 4018 tp->snd_cwnd = tp->t_peermss; 4019 else { 4020 int ss = tcp_init_win; 4021 if (inp != NULL && in_localaddr(inp->inp_faddr)) 4022 ss = tcp_init_win_local; 4023 #ifdef INET6 4024 if (in6p != NULL && in6_localaddr(&in6p->in6p_faddr)) 4025 ss = tcp_init_win_local; 4026 #endif 4027 tp->snd_cwnd = TCP_INITIAL_WINDOW(ss, tp->t_peermss); 4028 } 4029 4030 tcp_rmx_rtt(tp); 4031 tp->snd_wl1 = sc->sc_irs; 4032 tp->rcv_up = sc->sc_irs + 1; 4033 4034 /* 4035 * This is what whould have happened in tcp_output() when 4036 * the SYN,ACK was sent. 4037 */ 4038 tp->snd_up = tp->snd_una; 4039 tp->snd_max = tp->snd_nxt = tp->iss+1; 4040 TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur); 4041 if (sc->sc_win > 0 && SEQ_GT(tp->rcv_nxt + sc->sc_win, tp->rcv_adv)) 4042 tp->rcv_adv = tp->rcv_nxt + sc->sc_win; 4043 tp->last_ack_sent = tp->rcv_nxt; 4044 tp->t_partialacks = -1; 4045 tp->t_dupacks = 0; 4046 4047 TCP_STATINC(TCP_STAT_SC_COMPLETED); 4048 s = splsoftnet(); 4049 syn_cache_put(sc); 4050 splx(s); 4051 return so; 4052 4053 resetandabort: 4054 (void)tcp_respond(NULL, m, m, th, (tcp_seq)0, th->th_ack, TH_RST); 4055 abort: 4056 if (so != NULL) { 4057 (void) soqremque(so, 1); 4058 (void) soabort(so); 4059 mutex_enter(softnet_lock); 4060 } 4061 s = splsoftnet(); 4062 syn_cache_put(sc); 4063 splx(s); 4064 TCP_STATINC(TCP_STAT_SC_ABORTED); 4065 return ((struct socket *)(-1)); 4066 } 4067 4068 /* 4069 * This function is called when we get a RST for a 4070 * non-existent connection, so that we can see if the 4071 * connection is in the syn cache. If it is, zap it. 4072 */ 4073 4074 void 4075 syn_cache_reset(struct sockaddr *src, struct sockaddr *dst, struct tcphdr *th) 4076 { 4077 struct syn_cache *sc; 4078 struct syn_cache_head *scp; 4079 int s = splsoftnet(); 4080 4081 if ((sc = syn_cache_lookup(src, dst, &scp)) == NULL) { 4082 splx(s); 4083 return; 4084 } 4085 if (SEQ_LT(th->th_seq, sc->sc_irs) || 4086 SEQ_GT(th->th_seq, sc->sc_irs+1)) { 4087 splx(s); 4088 return; 4089 } 4090 syn_cache_rm(sc); 4091 TCP_STATINC(TCP_STAT_SC_RESET); 4092 syn_cache_put(sc); /* calls pool_put but see spl above */ 4093 splx(s); 4094 } 4095 4096 void 4097 syn_cache_unreach(const struct sockaddr *src, const struct sockaddr *dst, 4098 struct tcphdr *th) 4099 { 4100 struct syn_cache *sc; 4101 struct syn_cache_head *scp; 4102 int s; 4103 4104 s = splsoftnet(); 4105 if ((sc = syn_cache_lookup(src, dst, &scp)) == NULL) { 4106 splx(s); 4107 return; 4108 } 4109 /* If the sequence number != sc_iss, then it's a bogus ICMP msg */ 4110 if (ntohl(th->th_seq) != sc->sc_iss) { 4111 splx(s); 4112 return; 4113 } 4114 4115 /* 4116 * If we've retransmitted 3 times and this is our second error, 4117 * we remove the entry. Otherwise, we allow it to continue on. 4118 * This prevents us from incorrectly nuking an entry during a 4119 * spurious network outage. 4120 * 4121 * See tcp_notify(). 4122 */ 4123 if ((sc->sc_flags & SCF_UNREACH) == 0 || sc->sc_rxtshift < 3) { 4124 sc->sc_flags |= SCF_UNREACH; 4125 splx(s); 4126 return; 4127 } 4128 4129 syn_cache_rm(sc); 4130 TCP_STATINC(TCP_STAT_SC_UNREACH); 4131 syn_cache_put(sc); /* calls pool_put but see spl above */ 4132 splx(s); 4133 } 4134 4135 /* 4136 * Given a LISTEN socket and an inbound SYN request, add this to the syn 4137 * cache, and send back a segment: 4138 * <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK> 4139 * to the source. 4140 * 4141 * IMPORTANT NOTE: We do _NOT_ ACK data that might accompany the SYN. 4142 * Doing so would require that we hold onto the data and deliver it 4143 * to the application. However, if we are the target of a SYN-flood 4144 * DoS attack, an attacker could send data which would eventually 4145 * consume all available buffer space if it were ACKed. By not ACKing 4146 * the data, we avoid this DoS scenario. 4147 */ 4148 int 4149 syn_cache_add(struct sockaddr *src, struct sockaddr *dst, struct tcphdr *th, 4150 unsigned int toff, struct socket *so, struct mbuf *m, u_char *optp, 4151 int optlen, struct tcp_opt_info *oi) 4152 { 4153 struct tcpcb tb, *tp; 4154 long win; 4155 struct syn_cache *sc; 4156 struct syn_cache_head *scp; 4157 struct mbuf *ipopts; 4158 int s; 4159 4160 tp = sototcpcb(so); 4161 4162 /* 4163 * Initialize some local state. 4164 */ 4165 win = sbspace(&so->so_rcv); 4166 if (win > TCP_MAXWIN) 4167 win = TCP_MAXWIN; 4168 4169 #ifdef TCP_SIGNATURE 4170 if (optp || (tp->t_flags & TF_SIGNATURE)) 4171 #else 4172 if (optp) 4173 #endif 4174 { 4175 tb.t_flags = tcp_do_rfc1323 ? (TF_REQ_SCALE|TF_REQ_TSTMP) : 0; 4176 #ifdef TCP_SIGNATURE 4177 tb.t_flags |= (tp->t_flags & TF_SIGNATURE); 4178 #endif 4179 tb.t_state = TCPS_LISTEN; 4180 if (tcp_dooptions(&tb, optp, optlen, th, m, toff, oi) < 0) 4181 return 0; 4182 } else 4183 tb.t_flags = 0; 4184 4185 switch (src->sa_family) { 4186 case AF_INET: 4187 /* Remember the IP options, if any. */ 4188 ipopts = ip_srcroute(m); 4189 break; 4190 default: 4191 ipopts = NULL; 4192 } 4193 4194 /* 4195 * See if we already have an entry for this connection. 4196 * If we do, resend the SYN,ACK. We do not count this 4197 * as a retransmission (XXX though maybe we should). 4198 */ 4199 if ((sc = syn_cache_lookup(src, dst, &scp)) != NULL) { 4200 TCP_STATINC(TCP_STAT_SC_DUPESYN); 4201 if (ipopts) { 4202 /* 4203 * If we were remembering a previous source route, 4204 * forget it and use the new one we've been given. 4205 */ 4206 if (sc->sc_ipopts) 4207 (void)m_free(sc->sc_ipopts); 4208 sc->sc_ipopts = ipopts; 4209 } 4210 sc->sc_timestamp = tb.ts_recent; 4211 m_freem(m); 4212 if (syn_cache_respond(sc) == 0) { 4213 uint64_t *tcps = TCP_STAT_GETREF(); 4214 tcps[TCP_STAT_SNDACKS]++; 4215 tcps[TCP_STAT_SNDTOTAL]++; 4216 TCP_STAT_PUTREF(); 4217 } 4218 return 1; 4219 } 4220 4221 s = splsoftnet(); 4222 sc = pool_get(&syn_cache_pool, PR_NOWAIT); 4223 splx(s); 4224 if (sc == NULL) { 4225 if (ipopts) 4226 (void)m_free(ipopts); 4227 return 0; 4228 } 4229 4230 /* 4231 * Fill in the cache, and put the necessary IP and TCP 4232 * options into the reply. 4233 */ 4234 memset(sc, 0, sizeof(struct syn_cache)); 4235 callout_init(&sc->sc_timer, CALLOUT_MPSAFE); 4236 memcpy(&sc->sc_src, src, src->sa_len); 4237 memcpy(&sc->sc_dst, dst, dst->sa_len); 4238 sc->sc_flags = 0; 4239 sc->sc_ipopts = ipopts; 4240 sc->sc_irs = th->th_seq; 4241 switch (src->sa_family) { 4242 case AF_INET: 4243 { 4244 struct sockaddr_in *srcin = (void *)src; 4245 struct sockaddr_in *dstin = (void *)dst; 4246 4247 sc->sc_iss = tcp_new_iss1(&dstin->sin_addr, 4248 &srcin->sin_addr, dstin->sin_port, 4249 srcin->sin_port, sizeof(dstin->sin_addr), 0); 4250 break; 4251 } 4252 #ifdef INET6 4253 case AF_INET6: 4254 { 4255 struct sockaddr_in6 *srcin6 = (void *)src; 4256 struct sockaddr_in6 *dstin6 = (void *)dst; 4257 4258 sc->sc_iss = tcp_new_iss1(&dstin6->sin6_addr, 4259 &srcin6->sin6_addr, dstin6->sin6_port, 4260 srcin6->sin6_port, sizeof(dstin6->sin6_addr), 0); 4261 break; 4262 } 4263 #endif 4264 } 4265 sc->sc_peermaxseg = oi->maxseg; 4266 sc->sc_ourmaxseg = tcp_mss_to_advertise(m->m_flags & M_PKTHDR ? 4267 m_get_rcvif_NOMPSAFE(m) : NULL, sc->sc_src.sa.sa_family); 4268 sc->sc_win = win; 4269 sc->sc_timebase = tcp_now - 1; /* see tcp_newtcpcb() */ 4270 sc->sc_timestamp = tb.ts_recent; 4271 if ((tb.t_flags & (TF_REQ_TSTMP|TF_RCVD_TSTMP)) == 4272 (TF_REQ_TSTMP|TF_RCVD_TSTMP)) 4273 sc->sc_flags |= SCF_TIMESTAMP; 4274 if ((tb.t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == 4275 (TF_RCVD_SCALE|TF_REQ_SCALE)) { 4276 sc->sc_requested_s_scale = tb.requested_s_scale; 4277 sc->sc_request_r_scale = 0; 4278 /* 4279 * Pick the smallest possible scaling factor that 4280 * will still allow us to scale up to sb_max. 4281 * 4282 * We do this because there are broken firewalls that 4283 * will corrupt the window scale option, leading to 4284 * the other endpoint believing that our advertised 4285 * window is unscaled. At scale factors larger than 4286 * 5 the unscaled window will drop below 1500 bytes, 4287 * leading to serious problems when traversing these 4288 * broken firewalls. 4289 * 4290 * With the default sbmax of 256K, a scale factor 4291 * of 3 will be chosen by this algorithm. Those who 4292 * choose a larger sbmax should watch out 4293 * for the compatiblity problems mentioned above. 4294 * 4295 * RFC1323: The Window field in a SYN (i.e., a <SYN> 4296 * or <SYN,ACK>) segment itself is never scaled. 4297 */ 4298 while (sc->sc_request_r_scale < TCP_MAX_WINSHIFT && 4299 (TCP_MAXWIN << sc->sc_request_r_scale) < sb_max) 4300 sc->sc_request_r_scale++; 4301 } else { 4302 sc->sc_requested_s_scale = 15; 4303 sc->sc_request_r_scale = 15; 4304 } 4305 if ((tb.t_flags & TF_SACK_PERMIT) && tcp_do_sack) 4306 sc->sc_flags |= SCF_SACK_PERMIT; 4307 4308 /* 4309 * ECN setup packet received. 4310 */ 4311 if ((th->th_flags & (TH_ECE|TH_CWR)) && tcp_do_ecn) 4312 sc->sc_flags |= SCF_ECN_PERMIT; 4313 4314 #ifdef TCP_SIGNATURE 4315 if (tb.t_flags & TF_SIGNATURE) 4316 sc->sc_flags |= SCF_SIGNATURE; 4317 #endif 4318 sc->sc_tp = tp; 4319 m_freem(m); 4320 if (syn_cache_respond(sc) == 0) { 4321 uint64_t *tcps = TCP_STAT_GETREF(); 4322 tcps[TCP_STAT_SNDACKS]++; 4323 tcps[TCP_STAT_SNDTOTAL]++; 4324 TCP_STAT_PUTREF(); 4325 syn_cache_insert(sc, tp); 4326 } else { 4327 s = splsoftnet(); 4328 /* 4329 * syn_cache_put() will try to schedule the timer, so 4330 * we need to initialize it 4331 */ 4332 syn_cache_timer_arm(sc); 4333 syn_cache_put(sc); 4334 splx(s); 4335 TCP_STATINC(TCP_STAT_SC_DROPPED); 4336 } 4337 return 1; 4338 } 4339 4340 /* 4341 * syn_cache_respond: (re)send SYN+ACK. 4342 * 4343 * Returns 0 on success. 4344 */ 4345 4346 int 4347 syn_cache_respond(struct syn_cache *sc) 4348 { 4349 #ifdef INET6 4350 struct rtentry *rt = NULL; 4351 #endif 4352 struct route *ro; 4353 u_int8_t *optp; 4354 int optlen, error; 4355 u_int16_t tlen; 4356 struct ip *ip = NULL; 4357 #ifdef INET6 4358 struct ip6_hdr *ip6 = NULL; 4359 #endif 4360 struct tcpcb *tp; 4361 struct tcphdr *th; 4362 struct mbuf *m; 4363 u_int hlen; 4364 #ifdef TCP_SIGNATURE 4365 struct secasvar *sav = NULL; 4366 u_int8_t *sigp = NULL; 4367 #endif 4368 4369 ro = &sc->sc_route; 4370 switch (sc->sc_src.sa.sa_family) { 4371 case AF_INET: 4372 hlen = sizeof(struct ip); 4373 break; 4374 #ifdef INET6 4375 case AF_INET6: 4376 hlen = sizeof(struct ip6_hdr); 4377 break; 4378 #endif 4379 default: 4380 return EAFNOSUPPORT; 4381 } 4382 4383 /* Worst case scanario, since we don't know the option size yet. */ 4384 tlen = hlen + sizeof(struct tcphdr) + MAX_TCPOPTLEN; 4385 KASSERT(max_linkhdr + tlen <= MCLBYTES); 4386 4387 /* 4388 * Create the IP+TCP header from scratch. 4389 */ 4390 MGETHDR(m, M_DONTWAIT, MT_DATA); 4391 if (m && (max_linkhdr + tlen) > MHLEN) { 4392 MCLGET(m, M_DONTWAIT); 4393 if ((m->m_flags & M_EXT) == 0) { 4394 m_freem(m); 4395 m = NULL; 4396 } 4397 } 4398 if (m == NULL) 4399 return ENOBUFS; 4400 MCLAIM(m, &tcp_tx_mowner); 4401 4402 tp = sc->sc_tp; 4403 4404 /* Fixup the mbuf. */ 4405 m->m_data += max_linkhdr; 4406 m_reset_rcvif(m); 4407 memset(mtod(m, void *), 0, tlen); 4408 4409 switch (sc->sc_src.sa.sa_family) { 4410 case AF_INET: 4411 ip = mtod(m, struct ip *); 4412 ip->ip_v = 4; 4413 ip->ip_dst = sc->sc_src.sin.sin_addr; 4414 ip->ip_src = sc->sc_dst.sin.sin_addr; 4415 ip->ip_p = IPPROTO_TCP; 4416 th = (struct tcphdr *)(ip + 1); 4417 th->th_dport = sc->sc_src.sin.sin_port; 4418 th->th_sport = sc->sc_dst.sin.sin_port; 4419 break; 4420 #ifdef INET6 4421 case AF_INET6: 4422 ip6 = mtod(m, struct ip6_hdr *); 4423 ip6->ip6_vfc = IPV6_VERSION; 4424 ip6->ip6_dst = sc->sc_src.sin6.sin6_addr; 4425 ip6->ip6_src = sc->sc_dst.sin6.sin6_addr; 4426 ip6->ip6_nxt = IPPROTO_TCP; 4427 /* ip6_plen will be updated in ip6_output() */ 4428 th = (struct tcphdr *)(ip6 + 1); 4429 th->th_dport = sc->sc_src.sin6.sin6_port; 4430 th->th_sport = sc->sc_dst.sin6.sin6_port; 4431 break; 4432 #endif 4433 default: 4434 panic("%s: impossible (1)", __func__); 4435 } 4436 4437 th->th_seq = htonl(sc->sc_iss); 4438 th->th_ack = htonl(sc->sc_irs + 1); 4439 th->th_flags = TH_SYN|TH_ACK; 4440 th->th_win = htons(sc->sc_win); 4441 /* th_x2, th_sum, th_urp already 0 from memset */ 4442 4443 /* Tack on the TCP options. */ 4444 optp = (u_int8_t *)(th + 1); 4445 optlen = 0; 4446 *optp++ = TCPOPT_MAXSEG; 4447 *optp++ = TCPOLEN_MAXSEG; 4448 *optp++ = (sc->sc_ourmaxseg >> 8) & 0xff; 4449 *optp++ = sc->sc_ourmaxseg & 0xff; 4450 optlen += TCPOLEN_MAXSEG; 4451 4452 if (sc->sc_request_r_scale != 15) { 4453 *((u_int32_t *)optp) = htonl(TCPOPT_NOP << 24 | 4454 TCPOPT_WINDOW << 16 | TCPOLEN_WINDOW << 8 | 4455 sc->sc_request_r_scale); 4456 optp += TCPOLEN_WINDOW + TCPOLEN_NOP; 4457 optlen += TCPOLEN_WINDOW + TCPOLEN_NOP; 4458 } 4459 4460 if (sc->sc_flags & SCF_SACK_PERMIT) { 4461 /* Let the peer know that we will SACK. */ 4462 *optp++ = TCPOPT_SACK_PERMITTED; 4463 *optp++ = TCPOLEN_SACK_PERMITTED; 4464 optlen += TCPOLEN_SACK_PERMITTED; 4465 } 4466 4467 if (sc->sc_flags & SCF_TIMESTAMP) { 4468 while (optlen % 4 != 2) { 4469 optlen += TCPOLEN_NOP; 4470 *optp++ = TCPOPT_NOP; 4471 } 4472 *optp++ = TCPOPT_TIMESTAMP; 4473 *optp++ = TCPOLEN_TIMESTAMP; 4474 u_int32_t *lp = (u_int32_t *)(optp); 4475 /* Form timestamp option as shown in appendix A of RFC 1323. */ 4476 *lp++ = htonl(SYN_CACHE_TIMESTAMP(sc)); 4477 *lp = htonl(sc->sc_timestamp); 4478 optp += TCPOLEN_TIMESTAMP - 2; 4479 optlen += TCPOLEN_TIMESTAMP; 4480 } 4481 4482 #ifdef TCP_SIGNATURE 4483 if (sc->sc_flags & SCF_SIGNATURE) { 4484 sav = tcp_signature_getsav(m); 4485 if (sav == NULL) { 4486 m_freem(m); 4487 return EPERM; 4488 } 4489 4490 *optp++ = TCPOPT_SIGNATURE; 4491 *optp++ = TCPOLEN_SIGNATURE; 4492 sigp = optp; 4493 memset(optp, 0, TCP_SIGLEN); 4494 optp += TCP_SIGLEN; 4495 optlen += TCPOLEN_SIGNATURE; 4496 } 4497 #endif 4498 4499 /* 4500 * Terminate and pad TCP options to a 4 byte boundary. 4501 * 4502 * According to RFC793: "The content of the header beyond the 4503 * End-of-Option option must be header padding (i.e., zero)." 4504 * And later: "The padding is composed of zeros." 4505 */ 4506 if (optlen % 4) { 4507 optlen += TCPOLEN_EOL; 4508 *optp++ = TCPOPT_EOL; 4509 } 4510 while (optlen % 4) { 4511 optlen += TCPOLEN_PAD; 4512 *optp++ = TCPOPT_PAD; 4513 } 4514 4515 /* Compute the actual values now that we've added the options. */ 4516 tlen = hlen + sizeof(struct tcphdr) + optlen; 4517 m->m_len = m->m_pkthdr.len = tlen; 4518 th->th_off = (sizeof(struct tcphdr) + optlen) >> 2; 4519 4520 #ifdef TCP_SIGNATURE 4521 if (sav) { 4522 (void)tcp_signature(m, th, hlen, sav, sigp); 4523 key_sa_recordxfer(sav, m); 4524 KEY_SA_UNREF(&sav); 4525 } 4526 #endif 4527 4528 /* 4529 * Send ECN SYN-ACK setup packet. 4530 * Routes can be asymetric, so, even if we receive a packet 4531 * with ECE and CWR set, we must not assume no one will block 4532 * the ECE packet we are about to send. 4533 */ 4534 if ((sc->sc_flags & SCF_ECN_PERMIT) && tp && 4535 SEQ_GEQ(tp->snd_nxt, tp->snd_max)) { 4536 th->th_flags |= TH_ECE; 4537 TCP_STATINC(TCP_STAT_ECN_SHS); 4538 4539 /* 4540 * draft-ietf-tcpm-ecnsyn-00.txt 4541 * 4542 * "[...] a TCP node MAY respond to an ECN-setup 4543 * SYN packet by setting ECT in the responding 4544 * ECN-setup SYN/ACK packet, indicating to routers 4545 * that the SYN/ACK packet is ECN-Capable. 4546 * This allows a congested router along the path 4547 * to mark the packet instead of dropping the 4548 * packet as an indication of congestion." 4549 * 4550 * "[...] There can be a great benefit in setting 4551 * an ECN-capable codepoint in SYN/ACK packets [...] 4552 * Congestion is most likely to occur in 4553 * the server-to-client direction. As a result, 4554 * setting an ECN-capable codepoint in SYN/ACK 4555 * packets can reduce the occurence of three-second 4556 * retransmit timeouts resulting from the drop 4557 * of SYN/ACK packets." 4558 * 4559 * Page 4 and 6, January 2006. 4560 */ 4561 4562 switch (sc->sc_src.sa.sa_family) { 4563 case AF_INET: 4564 ip->ip_tos |= IPTOS_ECN_ECT0; 4565 break; 4566 #ifdef INET6 4567 case AF_INET6: 4568 ip6->ip6_flow |= htonl(IPTOS_ECN_ECT0 << 20); 4569 break; 4570 #endif 4571 } 4572 TCP_STATINC(TCP_STAT_ECN_ECT); 4573 } 4574 4575 4576 /* 4577 * Compute the packet's checksum. 4578 * 4579 * Fill in some straggling IP bits. Note the stack expects 4580 * ip_len to be in host order, for convenience. 4581 */ 4582 switch (sc->sc_src.sa.sa_family) { 4583 case AF_INET: 4584 ip->ip_len = htons(tlen - hlen); 4585 th->th_sum = 0; 4586 th->th_sum = in4_cksum(m, IPPROTO_TCP, hlen, tlen - hlen); 4587 ip->ip_len = htons(tlen); 4588 ip->ip_ttl = ip_defttl; 4589 /* XXX tos? */ 4590 break; 4591 #ifdef INET6 4592 case AF_INET6: 4593 ip6->ip6_plen = htons(tlen - hlen); 4594 th->th_sum = 0; 4595 th->th_sum = in6_cksum(m, IPPROTO_TCP, hlen, tlen - hlen); 4596 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 4597 ip6->ip6_vfc |= IPV6_VERSION; 4598 ip6->ip6_plen = htons(tlen - hlen); 4599 /* ip6_hlim will be initialized afterwards */ 4600 /* XXX flowlabel? */ 4601 break; 4602 #endif 4603 } 4604 4605 /* XXX use IPsec policy on listening socket, on SYN ACK */ 4606 tp = sc->sc_tp; 4607 4608 switch (sc->sc_src.sa.sa_family) { 4609 case AF_INET: 4610 error = ip_output(m, sc->sc_ipopts, ro, 4611 (ip_mtudisc ? IP_MTUDISC : 0), 4612 NULL, tp ? tp->t_inpcb : NULL); 4613 break; 4614 #ifdef INET6 4615 case AF_INET6: 4616 ip6->ip6_hlim = in6_selecthlim(NULL, 4617 (rt = rtcache_validate(ro)) != NULL ? rt->rt_ifp : NULL); 4618 rtcache_unref(rt, ro); 4619 4620 error = ip6_output(m, NULL /*XXX*/, ro, 0, NULL, 4621 tp ? tp->t_in6pcb : NULL, NULL); 4622 break; 4623 #endif 4624 default: 4625 panic("%s: impossible (2)", __func__); 4626 } 4627 4628 return error; 4629 } 4630
Indexes created Wed Sep 24 05:09:52 GMT 2025