pf.c revision 171929
1/* $OpenBSD: pf.c,v 1.527 2007/02/22 15:23:23 pyr Exp $ */ 2 3/* 4 * Copyright (c) 2001 Daniel Hartmeier 5 * Copyright (c) 2002,2003 Henning Brauer 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 12 * - Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * - Redistributions in binary form must reproduce the above 15 * copyright notice, this list of conditions and the following 16 * disclaimer in the documentation and/or other materials provided 17 * with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 20 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 21 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 22 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 23 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 24 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, 25 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 26 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 27 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN 29 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 * POSSIBILITY OF SUCH DAMAGE. 31 * 32 * Effort sponsored in part by the Defense Advanced Research Projects 33 * Agency (DARPA) and Air Force Research Laboratory, Air Force 34 * Materiel Command, USAF, under agreement number F30602-01-2-0537. 35 * 36 */ 37 38#ifdef __FreeBSD__ 39#include "opt_inet.h" 40#include "opt_inet6.h" 41 42#include <sys/cdefs.h> 43__FBSDID("$FreeBSD: head/sys/contrib/pf/net/pf.c 171929 2007-08-23 09:30:58Z dhartmei $"); 44#endif 45 46#ifdef __FreeBSD__ 47#include "opt_mac.h" 48#include "opt_bpf.h" 49#include "opt_pf.h" 50 51#ifdef DEV_BPF 52#define NBPFILTER DEV_BPF 53#else 54#define NBPFILTER 0 55#endif 56 57#ifdef DEV_PFLOG 58#define NPFLOG DEV_PFLOG 59#else 60#define NPFLOG 0 61#endif 62 63#ifdef DEV_PFSYNC 64#define NPFSYNC DEV_PFSYNC 65#else 66#define NPFSYNC 0 67#endif 68 69#else 70#include "bpfilter.h" 71#include "pflog.h" 72#include "pfsync.h" 73#endif 74 75#include <sys/param.h> 76#include <sys/systm.h> 77#include <sys/mbuf.h> 78#include <sys/filio.h> 79#include <sys/socket.h> 80#include <sys/socketvar.h> 81#include <sys/kernel.h> 82#include <sys/time.h> 83#ifdef __FreeBSD__ 84#include <sys/sysctl.h> 85#include <sys/endian.h> 86#else 87#include <sys/pool.h> 88#endif 89#include <sys/proc.h> 90#ifdef __FreeBSD__ 91#include <sys/kthread.h> 92#include <sys/lock.h> 93#include <sys/sx.h> 94#else 95#include <sys/rwlock.h> 96#endif 97 98#include <net/if.h> 99#include <net/if_types.h> 100#include <net/bpf.h> 101#include <net/route.h> 102#ifndef __FreeBSD__ 103#include <net/radix_mpath.h> 104#endif 105 106#include <netinet/in.h> 107#include <netinet/in_var.h> 108#include <netinet/in_systm.h> 109#include <netinet/ip.h> 110#include <netinet/ip_var.h> 111#include <netinet/tcp.h> 112#include <netinet/tcp_seq.h> 113#include <netinet/udp.h> 114#include <netinet/ip_icmp.h> 115#include <netinet/in_pcb.h> 116#include <netinet/tcp_timer.h> 117#include <netinet/tcp_var.h> 118#include <netinet/udp_var.h> 119#include <netinet/icmp_var.h> 120#include <netinet/if_ether.h> 121 122#ifndef __FreeBSD__ 123#include <dev/rndvar.h> 124#endif 125#include <net/pfvar.h> 126#include <net/if_pflog.h> 127 128#if NPFSYNC > 0 129#include <net/if_pfsync.h> 130#endif /* NPFSYNC > 0 */ 131 132#ifdef INET6 133#include <netinet/ip6.h> 134#include <netinet/in_pcb.h> 135#include <netinet/icmp6.h> 136#include <netinet6/nd6.h> 137#ifdef __FreeBSD__ 138#include <netinet6/ip6_var.h> 139#include <netinet6/in6_pcb.h> 140#endif 141#endif /* INET6 */ 142 143#ifdef __FreeBSD__ 144#include <machine/in_cksum.h> 145#include <sys/limits.h> 146#include <sys/ucred.h> 147#include <security/mac/mac_framework.h> 148 149extern int ip_optcopy(struct ip *, struct ip *); 150extern int debug_pfugidhack; 151#endif 152 153#define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x 154 155/* 156 * Global variables 157 */ 158 159struct pf_altqqueue pf_altqs[2]; 160struct pf_palist pf_pabuf; 161struct pf_altqqueue *pf_altqs_active; 162struct pf_altqqueue *pf_altqs_inactive; 163struct pf_status pf_status; 164 165u_int32_t ticket_altqs_active; 166u_int32_t ticket_altqs_inactive; 167int altqs_inactive_open; 168u_int32_t ticket_pabuf; 169 170struct pf_anchor_stackframe { 171 struct pf_ruleset *rs; 172 struct pf_rule *r; 173 struct pf_anchor_node *parent; 174 struct pf_anchor *child; 175} pf_anchor_stack[64]; 176 177#ifdef __FreeBSD__ 178uma_zone_t pf_src_tree_pl, pf_rule_pl; 179uma_zone_t pf_state_pl, pf_altq_pl, pf_pooladdr_pl; 180#else 181struct pool pf_src_tree_pl, pf_rule_pl; 182struct pool pf_state_pl, pf_altq_pl, pf_pooladdr_pl; 183#endif 184 185void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t); 186 187void pf_init_threshold(struct pf_threshold *, u_int32_t, 188 u_int32_t); 189void pf_add_threshold(struct pf_threshold *); 190int pf_check_threshold(struct pf_threshold *); 191 192void pf_change_ap(struct pf_addr *, u_int16_t *, 193 u_int16_t *, u_int16_t *, struct pf_addr *, 194 u_int16_t, u_int8_t, sa_family_t); 195int pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *, 196 struct tcphdr *, struct pf_state_peer *); 197#ifdef INET6 198void pf_change_a6(struct pf_addr *, u_int16_t *, 199 struct pf_addr *, u_int8_t); 200#endif /* INET6 */ 201void pf_change_icmp(struct pf_addr *, u_int16_t *, 202 struct pf_addr *, struct pf_addr *, u_int16_t, 203 u_int16_t *, u_int16_t *, u_int16_t *, 204 u_int16_t *, u_int8_t, sa_family_t); 205#ifdef __FreeBSD__ 206void pf_send_tcp(struct mbuf *, 207 const struct pf_rule *, sa_family_t, 208#else 209void pf_send_tcp(const struct pf_rule *, sa_family_t, 210#endif 211 const struct pf_addr *, const struct pf_addr *, 212 u_int16_t, u_int16_t, u_int32_t, u_int32_t, 213 u_int8_t, u_int16_t, u_int16_t, u_int8_t, int, 214 u_int16_t, struct ether_header *, struct ifnet *); 215void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t, 216 sa_family_t, struct pf_rule *); 217struct pf_rule *pf_match_translation(struct pf_pdesc *, struct mbuf *, 218 int, int, struct pfi_kif *, 219 struct pf_addr *, u_int16_t, struct pf_addr *, 220 u_int16_t, int); 221struct pf_rule *pf_get_translation(struct pf_pdesc *, struct mbuf *, 222 int, int, struct pfi_kif *, struct pf_src_node **, 223 struct pf_addr *, u_int16_t, 224 struct pf_addr *, u_int16_t, 225 struct pf_addr *, u_int16_t *); 226int pf_test_tcp(struct pf_rule **, struct pf_state **, 227 int, struct pfi_kif *, struct mbuf *, int, 228 void *, struct pf_pdesc *, struct pf_rule **, 229#ifdef __FreeBSD__ 230 struct pf_ruleset **, struct ifqueue *, 231 struct inpcb *); 232#else 233 struct pf_ruleset **, struct ifqueue *); 234#endif 235int pf_test_udp(struct pf_rule **, struct pf_state **, 236 int, struct pfi_kif *, struct mbuf *, int, 237 void *, struct pf_pdesc *, struct pf_rule **, 238#ifdef __FreeBSD__ 239 struct pf_ruleset **, struct ifqueue *, 240 struct inpcb *); 241#else 242 struct pf_ruleset **, struct ifqueue *); 243#endif 244int pf_test_icmp(struct pf_rule **, struct pf_state **, 245 int, struct pfi_kif *, struct mbuf *, int, 246 void *, struct pf_pdesc *, struct pf_rule **, 247 struct pf_ruleset **, struct ifqueue *); 248int pf_test_other(struct pf_rule **, struct pf_state **, 249 int, struct pfi_kif *, struct mbuf *, int, void *, 250 struct pf_pdesc *, struct pf_rule **, 251 struct pf_ruleset **, struct ifqueue *); 252int pf_test_fragment(struct pf_rule **, int, 253 struct pfi_kif *, struct mbuf *, void *, 254 struct pf_pdesc *, struct pf_rule **, 255 struct pf_ruleset **); 256int pf_test_state_tcp(struct pf_state **, int, 257 struct pfi_kif *, struct mbuf *, int, 258 void *, struct pf_pdesc *, u_short *); 259int pf_test_state_udp(struct pf_state **, int, 260 struct pfi_kif *, struct mbuf *, int, 261 void *, struct pf_pdesc *); 262int pf_test_state_icmp(struct pf_state **, int, 263 struct pfi_kif *, struct mbuf *, int, 264 void *, struct pf_pdesc *, u_short *); 265int pf_test_state_other(struct pf_state **, int, 266 struct pfi_kif *, struct pf_pdesc *); 267int pf_match_tag(struct mbuf *, struct pf_rule *, 268 struct pf_mtag *, int *); 269int pf_step_out_of_anchor(int *, struct pf_ruleset **, 270 int, struct pf_rule **, struct pf_rule **, 271 int *); 272void pf_hash(struct pf_addr *, struct pf_addr *, 273 struct pf_poolhashkey *, sa_family_t); 274int pf_map_addr(u_int8_t, struct pf_rule *, 275 struct pf_addr *, struct pf_addr *, 276 struct pf_addr *, struct pf_src_node **); 277int pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *, 278 struct pf_addr *, struct pf_addr *, u_int16_t, 279 struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t, 280 struct pf_src_node **); 281void pf_route(struct mbuf **, struct pf_rule *, int, 282 struct ifnet *, struct pf_state *, 283 struct pf_pdesc *); 284void pf_route6(struct mbuf **, struct pf_rule *, int, 285 struct ifnet *, struct pf_state *, 286 struct pf_pdesc *); 287#ifdef __FreeBSD__ 288/* XXX: import */ 289#else 290int pf_socket_lookup(int, struct pf_pdesc *); 291#endif 292u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t, 293 sa_family_t); 294u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t, 295 sa_family_t); 296u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t, 297 u_int16_t); 298void pf_set_rt_ifp(struct pf_state *, 299 struct pf_addr *); 300int pf_check_proto_cksum(struct mbuf *, int, int, 301 u_int8_t, sa_family_t); 302int pf_addr_wrap_neq(struct pf_addr_wrap *, 303 struct pf_addr_wrap *); 304struct pf_state *pf_find_state_recurse(struct pfi_kif *, 305 struct pf_state_cmp *, u_int8_t); 306int pf_src_connlimit(struct pf_state **); 307int pf_check_congestion(struct ifqueue *); 308 309#ifdef __FreeBSD__ 310int in4_cksum(struct mbuf *m, u_int8_t nxt, int off, int len); 311 312extern int pf_end_threads; 313 314struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX]; 315#else 316extern struct pool pfr_ktable_pl; 317extern struct pool pfr_kentry_pl; 318 319struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = { 320 { &pf_state_pl, PFSTATE_HIWAT }, 321 { &pf_src_tree_pl, PFSNODE_HIWAT }, 322 { &pf_frent_pl, PFFRAG_FRENT_HIWAT }, 323 { &pfr_ktable_pl, PFR_KTABLE_HIWAT }, 324 { &pfr_kentry_pl, PFR_KENTRY_HIWAT } 325}; 326#endif 327 328#define STATE_LOOKUP() \ 329 do { \ 330 if (direction == PF_IN) \ 331 *state = pf_find_state_recurse( \ 332 kif, &key, PF_EXT_GWY); \ 333 else \ 334 *state = pf_find_state_recurse( \ 335 kif, &key, PF_LAN_EXT); \ 336 if (*state == NULL || (*state)->timeout == PFTM_PURGE) \ 337 return (PF_DROP); \ 338 if (direction == PF_OUT && \ 339 (((*state)->rule.ptr->rt == PF_ROUTETO && \ 340 (*state)->rule.ptr->direction == PF_OUT) || \ 341 ((*state)->rule.ptr->rt == PF_REPLYTO && \ 342 (*state)->rule.ptr->direction == PF_IN)) && \ 343 (*state)->rt_kif != NULL && \ 344 (*state)->rt_kif != kif) \ 345 return (PF_PASS); \ 346 } while (0) 347 348#define STATE_TRANSLATE(s) \ 349 (s)->lan.addr.addr32[0] != (s)->gwy.addr.addr32[0] || \ 350 ((s)->af == AF_INET6 && \ 351 ((s)->lan.addr.addr32[1] != (s)->gwy.addr.addr32[1] || \ 352 (s)->lan.addr.addr32[2] != (s)->gwy.addr.addr32[2] || \ 353 (s)->lan.addr.addr32[3] != (s)->gwy.addr.addr32[3])) || \ 354 (s)->lan.port != (s)->gwy.port 355 356#define BOUND_IFACE(r, k) \ 357 ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all 358 359#define STATE_INC_COUNTERS(s) \ 360 do { \ 361 s->rule.ptr->states++; \ 362 if (s->anchor.ptr != NULL) \ 363 s->anchor.ptr->states++; \ 364 if (s->nat_rule.ptr != NULL) \ 365 s->nat_rule.ptr->states++; \ 366 } while (0) 367 368#define STATE_DEC_COUNTERS(s) \ 369 do { \ 370 if (s->nat_rule.ptr != NULL) \ 371 s->nat_rule.ptr->states--; \ 372 if (s->anchor.ptr != NULL) \ 373 s->anchor.ptr->states--; \ 374 s->rule.ptr->states--; \ 375 } while (0) 376 377struct pf_src_tree tree_src_tracking; 378 379struct pf_state_tree_id tree_id; 380struct pf_state_queue state_list; 381 382#ifdef __FreeBSD__ 383static int pf_src_compare(struct pf_src_node *, struct pf_src_node *); 384static int pf_state_compare_lan_ext(struct pf_state *, struct pf_state *); 385static int pf_state_compare_ext_gwy(struct pf_state *, struct pf_state *); 386static int pf_state_compare_id(struct pf_state *, struct pf_state *); 387#endif 388 389RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare); 390RB_GENERATE(pf_state_tree_lan_ext, pf_state, 391 u.s.entry_lan_ext, pf_state_compare_lan_ext); 392RB_GENERATE(pf_state_tree_ext_gwy, pf_state, 393 u.s.entry_ext_gwy, pf_state_compare_ext_gwy); 394RB_GENERATE(pf_state_tree_id, pf_state, 395 u.s.entry_id, pf_state_compare_id); 396 397#ifdef __FreeBSD__ 398static int 399#else 400static __inline int 401#endif 402pf_src_compare(struct pf_src_node *a, struct pf_src_node *b) 403{ 404 int diff; 405 406 if (a->rule.ptr > b->rule.ptr) 407 return (1); 408 if (a->rule.ptr < b->rule.ptr) 409 return (-1); 410 if ((diff = a->af - b->af) != 0) 411 return (diff); 412 switch (a->af) { 413#ifdef INET 414 case AF_INET: 415 if (a->addr.addr32[0] > b->addr.addr32[0]) 416 return (1); 417 if (a->addr.addr32[0] < b->addr.addr32[0]) 418 return (-1); 419 break; 420#endif /* INET */ 421#ifdef INET6 422 case AF_INET6: 423 if (a->addr.addr32[3] > b->addr.addr32[3]) 424 return (1); 425 if (a->addr.addr32[3] < b->addr.addr32[3]) 426 return (-1); 427 if (a->addr.addr32[2] > b->addr.addr32[2]) 428 return (1); 429 if (a->addr.addr32[2] < b->addr.addr32[2]) 430 return (-1); 431 if (a->addr.addr32[1] > b->addr.addr32[1]) 432 return (1); 433 if (a->addr.addr32[1] < b->addr.addr32[1]) 434 return (-1); 435 if (a->addr.addr32[0] > b->addr.addr32[0]) 436 return (1); 437 if (a->addr.addr32[0] < b->addr.addr32[0]) 438 return (-1); 439 break; 440#endif /* INET6 */ 441 } 442 return (0); 443} 444 445#ifdef __FreeBSD__ 446static int 447#else 448static __inline int 449#endif 450pf_state_compare_lan_ext(struct pf_state *a, struct pf_state *b) 451{ 452 int diff; 453 454 if ((diff = a->proto - b->proto) != 0) 455 return (diff); 456 if ((diff = a->af - b->af) != 0) 457 return (diff); 458 switch (a->af) { 459#ifdef INET 460 case AF_INET: 461 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0]) 462 return (1); 463 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0]) 464 return (-1); 465 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 466 return (1); 467 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 468 return (-1); 469 break; 470#endif /* INET */ 471#ifdef INET6 472 case AF_INET6: 473 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3]) 474 return (1); 475 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3]) 476 return (-1); 477 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3]) 478 return (1); 479 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3]) 480 return (-1); 481 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2]) 482 return (1); 483 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2]) 484 return (-1); 485 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2]) 486 return (1); 487 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2]) 488 return (-1); 489 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1]) 490 return (1); 491 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1]) 492 return (-1); 493 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1]) 494 return (1); 495 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1]) 496 return (-1); 497 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0]) 498 return (1); 499 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0]) 500 return (-1); 501 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 502 return (1); 503 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 504 return (-1); 505 break; 506#endif /* INET6 */ 507 } 508 509 if ((diff = a->lan.port - b->lan.port) != 0) 510 return (diff); 511 if ((diff = a->ext.port - b->ext.port) != 0) 512 return (diff); 513 514 return (0); 515} 516 517#ifdef __FreeBSD__ 518static int 519#else 520static __inline int 521#endif 522pf_state_compare_ext_gwy(struct pf_state *a, struct pf_state *b) 523{ 524 int diff; 525 526 if ((diff = a->proto - b->proto) != 0) 527 return (diff); 528 if ((diff = a->af - b->af) != 0) 529 return (diff); 530 switch (a->af) { 531#ifdef INET 532 case AF_INET: 533 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 534 return (1); 535 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 536 return (-1); 537 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0]) 538 return (1); 539 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0]) 540 return (-1); 541 break; 542#endif /* INET */ 543#ifdef INET6 544 case AF_INET6: 545 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3]) 546 return (1); 547 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3]) 548 return (-1); 549 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3]) 550 return (1); 551 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3]) 552 return (-1); 553 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2]) 554 return (1); 555 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2]) 556 return (-1); 557 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2]) 558 return (1); 559 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2]) 560 return (-1); 561 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1]) 562 return (1); 563 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1]) 564 return (-1); 565 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1]) 566 return (1); 567 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1]) 568 return (-1); 569 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 570 return (1); 571 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 572 return (-1); 573 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0]) 574 return (1); 575 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0]) 576 return (-1); 577 break; 578#endif /* INET6 */ 579 } 580 581 if ((diff = a->ext.port - b->ext.port) != 0) 582 return (diff); 583 if ((diff = a->gwy.port - b->gwy.port) != 0) 584 return (diff); 585 586 return (0); 587} 588 589#ifdef __FreeBSD__ 590static int 591#else 592static __inline int 593#endif 594pf_state_compare_id(struct pf_state *a, struct pf_state *b) 595{ 596 if (a->id > b->id) 597 return (1); 598 if (a->id < b->id) 599 return (-1); 600 if (a->creatorid > b->creatorid) 601 return (1); 602 if (a->creatorid < b->creatorid) 603 return (-1); 604 605 return (0); 606} 607 608#ifdef INET6 609void 610pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af) 611{ 612 switch (af) { 613#ifdef INET 614 case AF_INET: 615 dst->addr32[0] = src->addr32[0]; 616 break; 617#endif /* INET */ 618 case AF_INET6: 619 dst->addr32[0] = src->addr32[0]; 620 dst->addr32[1] = src->addr32[1]; 621 dst->addr32[2] = src->addr32[2]; 622 dst->addr32[3] = src->addr32[3]; 623 break; 624 } 625} 626#endif /* INET6 */ 627 628struct pf_state * 629pf_find_state_byid(struct pf_state_cmp *key) 630{ 631 pf_status.fcounters[FCNT_STATE_SEARCH]++; 632 return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key)); 633} 634 635struct pf_state * 636pf_find_state_recurse(struct pfi_kif *kif, struct pf_state_cmp *key, u_int8_t tree) 637{ 638 struct pf_state *s; 639 640 pf_status.fcounters[FCNT_STATE_SEARCH]++; 641 642 switch (tree) { 643 case PF_LAN_EXT: 644 if ((s = RB_FIND(pf_state_tree_lan_ext, &kif->pfik_lan_ext, 645 (struct pf_state *)key)) != NULL) 646 return (s); 647 if ((s = RB_FIND(pf_state_tree_lan_ext, &pfi_all->pfik_lan_ext, 648 (struct pf_state *)key)) != NULL) 649 return (s); 650 return (NULL); 651 case PF_EXT_GWY: 652 if ((s = RB_FIND(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, 653 (struct pf_state *)key)) != NULL) 654 return (s); 655 if ((s = RB_FIND(pf_state_tree_ext_gwy, &pfi_all->pfik_ext_gwy, 656 (struct pf_state *)key)) != NULL) 657 return (s); 658 return (NULL); 659 default: 660 panic("pf_find_state_recurse"); 661 } 662} 663 664struct pf_state * 665pf_find_state_all(struct pf_state_cmp *key, u_int8_t tree, int *more) 666{ 667 struct pf_state *s, *ss = NULL; 668 struct pfi_kif *kif; 669 670 pf_status.fcounters[FCNT_STATE_SEARCH]++; 671 672 switch (tree) { 673 case PF_LAN_EXT: 674 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) { 675 s = RB_FIND(pf_state_tree_lan_ext, 676 &kif->pfik_lan_ext, (struct pf_state *)key); 677 if (s == NULL) 678 continue; 679 if (more == NULL) 680 return (s); 681 ss = s; 682 (*more)++; 683 } 684 return (ss); 685 case PF_EXT_GWY: 686 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) { 687 s = RB_FIND(pf_state_tree_ext_gwy, 688 &kif->pfik_ext_gwy, (struct pf_state *)key); 689 if (s == NULL) 690 continue; 691 if (more == NULL) 692 return (s); 693 ss = s; 694 (*more)++; 695 } 696 return (ss); 697 default: 698 panic("pf_find_state_all"); 699 } 700} 701 702void 703pf_init_threshold(struct pf_threshold *threshold, 704 u_int32_t limit, u_int32_t seconds) 705{ 706 threshold->limit = limit * PF_THRESHOLD_MULT; 707 threshold->seconds = seconds; 708 threshold->count = 0; 709 threshold->last = time_second; 710} 711 712void 713pf_add_threshold(struct pf_threshold *threshold) 714{ 715 u_int32_t t = time_second, diff = t - threshold->last; 716 717 if (diff >= threshold->seconds) 718 threshold->count = 0; 719 else 720 threshold->count -= threshold->count * diff / 721 threshold->seconds; 722 threshold->count += PF_THRESHOLD_MULT; 723 threshold->last = t; 724} 725 726int 727pf_check_threshold(struct pf_threshold *threshold) 728{ 729 return (threshold->count > threshold->limit); 730} 731 732int 733pf_src_connlimit(struct pf_state **state) 734{ 735 struct pf_state *s; 736 int bad = 0; 737 738 (*state)->src_node->conn++; 739 (*state)->src.tcp_est = 1; 740 pf_add_threshold(&(*state)->src_node->conn_rate); 741 742 if ((*state)->rule.ptr->max_src_conn && 743 (*state)->rule.ptr->max_src_conn < 744 (*state)->src_node->conn) { 745 pf_status.lcounters[LCNT_SRCCONN]++; 746 bad++; 747 } 748 749 if ((*state)->rule.ptr->max_src_conn_rate.limit && 750 pf_check_threshold(&(*state)->src_node->conn_rate)) { 751 pf_status.lcounters[LCNT_SRCCONNRATE]++; 752 bad++; 753 } 754 755 if (!bad) 756 return (0); 757 758 if ((*state)->rule.ptr->overload_tbl) { 759 struct pfr_addr p; 760 u_int32_t killed = 0; 761 762 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++; 763 if (pf_status.debug >= PF_DEBUG_MISC) { 764 printf("pf_src_connlimit: blocking address "); 765 pf_print_host(&(*state)->src_node->addr, 0, 766 (*state)->af); 767 } 768 769 bzero(&p, sizeof(p)); 770 p.pfra_af = (*state)->af; 771 switch ((*state)->af) { 772#ifdef INET 773 case AF_INET: 774 p.pfra_net = 32; 775 p.pfra_ip4addr = (*state)->src_node->addr.v4; 776 break; 777#endif /* INET */ 778#ifdef INET6 779 case AF_INET6: 780 p.pfra_net = 128; 781 p.pfra_ip6addr = (*state)->src_node->addr.v6; 782 break; 783#endif /* INET6 */ 784 } 785 786 pfr_insert_kentry((*state)->rule.ptr->overload_tbl, 787 &p, time_second); 788 789 /* kill existing states if that's required. */ 790 if ((*state)->rule.ptr->flush) { 791 pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++; 792 793 RB_FOREACH(s, pf_state_tree_id, &tree_id) { 794 /* 795 * Kill states from this source. (Only those 796 * from the same rule if PF_FLUSH_GLOBAL is not 797 * set) 798 */ 799 if (s->af == (*state)->af && 800 (((*state)->direction == PF_OUT && 801 PF_AEQ(&(*state)->src_node->addr, 802 &s->lan.addr, s->af)) || 803 ((*state)->direction == PF_IN && 804 PF_AEQ(&(*state)->src_node->addr, 805 &s->ext.addr, s->af))) && 806 ((*state)->rule.ptr->flush & 807 PF_FLUSH_GLOBAL || 808 (*state)->rule.ptr == s->rule.ptr)) { 809 s->timeout = PFTM_PURGE; 810 s->src.state = s->dst.state = 811 TCPS_CLOSED; 812 killed++; 813 } 814 } 815 if (pf_status.debug >= PF_DEBUG_MISC) 816 printf(", %u states killed", killed); 817 } 818 if (pf_status.debug >= PF_DEBUG_MISC) 819 printf("\n"); 820 } 821 822 /* kill this state */ 823 (*state)->timeout = PFTM_PURGE; 824 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED; 825 return (1); 826} 827 828int 829pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule, 830 struct pf_addr *src, sa_family_t af) 831{ 832 struct pf_src_node k; 833 834 if (*sn == NULL) { 835 k.af = af; 836 PF_ACPY(&k.addr, src, af); 837 if (rule->rule_flag & PFRULE_RULESRCTRACK || 838 rule->rpool.opts & PF_POOL_STICKYADDR) 839 k.rule.ptr = rule; 840 else 841 k.rule.ptr = NULL; 842 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++; 843 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k); 844 } 845 if (*sn == NULL) { 846 if (!rule->max_src_nodes || 847 rule->src_nodes < rule->max_src_nodes) 848 (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT); 849 else 850 pf_status.lcounters[LCNT_SRCNODES]++; 851 if ((*sn) == NULL) 852 return (-1); 853 bzero(*sn, sizeof(struct pf_src_node)); 854 855 pf_init_threshold(&(*sn)->conn_rate, 856 rule->max_src_conn_rate.limit, 857 rule->max_src_conn_rate.seconds); 858 859 (*sn)->af = af; 860 if (rule->rule_flag & PFRULE_RULESRCTRACK || 861 rule->rpool.opts & PF_POOL_STICKYADDR) 862 (*sn)->rule.ptr = rule; 863 else 864 (*sn)->rule.ptr = NULL; 865 PF_ACPY(&(*sn)->addr, src, af); 866 if (RB_INSERT(pf_src_tree, 867 &tree_src_tracking, *sn) != NULL) { 868 if (pf_status.debug >= PF_DEBUG_MISC) { 869 printf("pf: src_tree insert failed: "); 870 pf_print_host(&(*sn)->addr, 0, af); 871 printf("\n"); 872 } 873 pool_put(&pf_src_tree_pl, *sn); 874 return (-1); 875 } 876 (*sn)->creation = time_second; 877 (*sn)->ruletype = rule->action; 878 if ((*sn)->rule.ptr != NULL) 879 (*sn)->rule.ptr->src_nodes++; 880 pf_status.scounters[SCNT_SRC_NODE_INSERT]++; 881 pf_status.src_nodes++; 882 } else { 883 if (rule->max_src_states && 884 (*sn)->states >= rule->max_src_states) { 885 pf_status.lcounters[LCNT_SRCSTATES]++; 886 return (-1); 887 } 888 } 889 return (0); 890} 891 892int 893pf_insert_state(struct pfi_kif *kif, struct pf_state *state) 894{ 895 /* Thou MUST NOT insert multiple duplicate keys */ 896 state->u.s.kif = kif; 897 if (RB_INSERT(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state)) { 898 if (pf_status.debug >= PF_DEBUG_MISC) { 899 printf("pf: state insert failed: tree_lan_ext"); 900 printf(" lan: "); 901 pf_print_host(&state->lan.addr, state->lan.port, 902 state->af); 903 printf(" gwy: "); 904 pf_print_host(&state->gwy.addr, state->gwy.port, 905 state->af); 906 printf(" ext: "); 907 pf_print_host(&state->ext.addr, state->ext.port, 908 state->af); 909 if (state->sync_flags & PFSTATE_FROMSYNC) 910 printf(" (from sync)"); 911 printf("\n"); 912 } 913 return (-1); 914 } 915 916 if (RB_INSERT(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state)) { 917 if (pf_status.debug >= PF_DEBUG_MISC) { 918 printf("pf: state insert failed: tree_ext_gwy"); 919 printf(" lan: "); 920 pf_print_host(&state->lan.addr, state->lan.port, 921 state->af); 922 printf(" gwy: "); 923 pf_print_host(&state->gwy.addr, state->gwy.port, 924 state->af); 925 printf(" ext: "); 926 pf_print_host(&state->ext.addr, state->ext.port, 927 state->af); 928 if (state->sync_flags & PFSTATE_FROMSYNC) 929 printf(" (from sync)"); 930 printf("\n"); 931 } 932 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state); 933 return (-1); 934 } 935 936 if (state->id == 0 && state->creatorid == 0) { 937 state->id = htobe64(pf_status.stateid++); 938 state->creatorid = pf_status.hostid; 939 } 940 if (RB_INSERT(pf_state_tree_id, &tree_id, state) != NULL) { 941 if (pf_status.debug >= PF_DEBUG_MISC) { 942#ifdef __FreeBSD__ 943 printf("pf: state insert failed: " 944 "id: %016llx creatorid: %08x", 945 (long long)be64toh(state->id), 946 ntohl(state->creatorid)); 947#else 948 printf("pf: state insert failed: " 949 "id: %016llx creatorid: %08x", 950 betoh64(state->id), ntohl(state->creatorid)); 951#endif 952 if (state->sync_flags & PFSTATE_FROMSYNC) 953 printf(" (from sync)"); 954 printf("\n"); 955 } 956 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state); 957 RB_REMOVE(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state); 958 return (-1); 959 } 960 TAILQ_INSERT_TAIL(&state_list, state, u.s.entry_list); 961 pf_status.fcounters[FCNT_STATE_INSERT]++; 962 pf_status.states++; 963 pfi_kif_ref(kif, PFI_KIF_REF_STATE); 964#if NPFSYNC 965 pfsync_insert_state(state); 966#endif 967 return (0); 968} 969 970void 971pf_purge_thread(void *v) 972{ 973 int nloops = 0, s; 974 975 for (;;) { 976 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz); 977 978#ifdef __FreeBSD__ 979 sx_slock(&pf_consistency_lock); 980 PF_LOCK(); 981 982 if (pf_end_threads) { 983 pf_purge_expired_states(pf_status.states); 984 pf_purge_expired_fragments(); 985 pf_purge_expired_src_nodes(0); 986 pf_end_threads++; 987 988 sx_sunlock(&pf_consistency_lock); 989 PF_UNLOCK(); 990 wakeup(pf_purge_thread); 991 kthread_exit(0); 992 } 993#endif 994 s = splsoftnet(); 995 996 /* process a fraction of the state table every second */ 997 pf_purge_expired_states(1 + (pf_status.states 998 / pf_default_rule.timeout[PFTM_INTERVAL])); 999 1000 /* purge other expired types every PFTM_INTERVAL seconds */ 1001 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) { 1002 pf_purge_expired_fragments(); 1003 pf_purge_expired_src_nodes(0); 1004 nloops = 0; 1005 } 1006 1007 splx(s); 1008#ifdef __FreeBSD__ 1009 PF_UNLOCK(); 1010 sx_sunlock(&pf_consistency_lock); 1011#endif 1012 } 1013} 1014 1015u_int32_t 1016pf_state_expires(const struct pf_state *state) 1017{ 1018 u_int32_t timeout; 1019 u_int32_t start; 1020 u_int32_t end; 1021 u_int32_t states; 1022 1023 /* handle all PFTM_* > PFTM_MAX here */ 1024 if (state->timeout == PFTM_PURGE) 1025 return (time_second); 1026 if (state->timeout == PFTM_UNTIL_PACKET) 1027 return (0); 1028#ifdef __FreeBSD__ 1029 KASSERT(state->timeout != PFTM_UNLINKED, 1030 ("pf_state_expires: timeout == PFTM_UNLINKED")); 1031 KASSERT((state->timeout < PFTM_MAX), 1032 ("pf_state_expires: timeout > PFTM_MAX")); 1033#else 1034 KASSERT(state->timeout != PFTM_UNLINKED); 1035 KASSERT(state->timeout < PFTM_MAX); 1036#endif 1037 timeout = state->rule.ptr->timeout[state->timeout]; 1038 if (!timeout) 1039 timeout = pf_default_rule.timeout[state->timeout]; 1040 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START]; 1041 if (start) { 1042 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END]; 1043 states = state->rule.ptr->states; 1044 } else { 1045 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START]; 1046 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END]; 1047 states = pf_status.states; 1048 } 1049 if (end && states > start && start < end) { 1050 if (states < end) 1051 return (state->expire + timeout * (end - states) / 1052 (end - start)); 1053 else 1054 return (time_second); 1055 } 1056 return (state->expire + timeout); 1057} 1058 1059void 1060pf_purge_expired_src_nodes(int waslocked) 1061{ 1062 struct pf_src_node *cur, *next; 1063 int locked = waslocked; 1064 1065 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) { 1066 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur); 1067 1068 if (cur->states <= 0 && cur->expire <= time_second) { 1069 if (! locked) { 1070#ifdef __FreeBSD__ 1071 if (!sx_try_upgrade(&pf_consistency_lock)) { 1072 PF_UNLOCK(); 1073 sx_sunlock(&pf_consistency_lock); 1074 sx_xlock(&pf_consistency_lock); 1075 PF_LOCK(); 1076 } 1077#else 1078 rw_enter_write(&pf_consistency_lock); 1079#endif 1080 next = RB_NEXT(pf_src_tree, 1081 &tree_src_tracking, cur); 1082 locked = 1; 1083 } 1084 if (cur->rule.ptr != NULL) { 1085 cur->rule.ptr->src_nodes--; 1086 if (cur->rule.ptr->states <= 0 && 1087 cur->rule.ptr->max_src_nodes <= 0) 1088 pf_rm_rule(NULL, cur->rule.ptr); 1089 } 1090 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur); 1091 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 1092 pf_status.src_nodes--; 1093 pool_put(&pf_src_tree_pl, cur); 1094 } 1095 } 1096 1097 if (locked && !waslocked) 1098#ifdef __FreeBSD__ 1099 sx_downgrade(&pf_consistency_lock); 1100#else 1101 rw_exit_write(&pf_consistency_lock); 1102#endif 1103} 1104 1105void 1106pf_src_tree_remove_state(struct pf_state *s) 1107{ 1108 u_int32_t timeout; 1109 1110 if (s->src_node != NULL) { 1111 if (s->proto == IPPROTO_TCP) { 1112 if (s->src.tcp_est) 1113 --s->src_node->conn; 1114 } 1115 if (--s->src_node->states <= 0) { 1116 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE]; 1117 if (!timeout) 1118 timeout = 1119 pf_default_rule.timeout[PFTM_SRC_NODE]; 1120 s->src_node->expire = time_second + timeout; 1121 } 1122 } 1123 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) { 1124 if (--s->nat_src_node->states <= 0) { 1125 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE]; 1126 if (!timeout) 1127 timeout = 1128 pf_default_rule.timeout[PFTM_SRC_NODE]; 1129 s->nat_src_node->expire = time_second + timeout; 1130 } 1131 } 1132 s->src_node = s->nat_src_node = NULL; 1133} 1134 1135/* callers should be at splsoftnet */ 1136void 1137pf_unlink_state(struct pf_state *cur) 1138{ 1139#ifdef __FreeBSD__ 1140 if (cur->local_flags & PFSTATE_EXPIRING) 1141 return; 1142 cur->local_flags |= PFSTATE_EXPIRING; 1143#endif 1144 if (cur->src.state == PF_TCPS_PROXY_DST) { 1145#ifdef __FreeBSD__ 1146 pf_send_tcp(NULL, cur->rule.ptr, cur->af, 1147#else 1148 pf_send_tcp(cur->rule.ptr, cur->af, 1149#endif 1150 &cur->ext.addr, &cur->lan.addr, 1151 cur->ext.port, cur->lan.port, 1152 cur->src.seqhi, cur->src.seqlo + 1, 1153 TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL); 1154 } 1155 RB_REMOVE(pf_state_tree_ext_gwy, 1156 &cur->u.s.kif->pfik_ext_gwy, cur); 1157 RB_REMOVE(pf_state_tree_lan_ext, 1158 &cur->u.s.kif->pfik_lan_ext, cur); 1159 RB_REMOVE(pf_state_tree_id, &tree_id, cur); 1160#if NPFSYNC 1161 if (cur->creatorid == pf_status.hostid) 1162 pfsync_delete_state(cur); 1163#endif 1164 cur->timeout = PFTM_UNLINKED; 1165 pf_src_tree_remove_state(cur); 1166} 1167 1168/* callers should be at splsoftnet and hold the 1169 * write_lock on pf_consistency_lock */ 1170void 1171pf_free_state(struct pf_state *cur) 1172{ 1173#if NPFSYNC 1174 if (pfsyncif != NULL && 1175 (pfsyncif->sc_bulk_send_next == cur || 1176 pfsyncif->sc_bulk_terminator == cur)) 1177 return; 1178#endif 1179#ifdef __FreeBSD__ 1180 KASSERT(cur->timeout == PFTM_UNLINKED, 1181 ("pf_free_state: cur->timeout != PFTM_UNLINKED")); 1182#else 1183 KASSERT(cur->timeout == PFTM_UNLINKED); 1184#endif 1185 if (--cur->rule.ptr->states <= 0 && 1186 cur->rule.ptr->src_nodes <= 0) 1187 pf_rm_rule(NULL, cur->rule.ptr); 1188 if (cur->nat_rule.ptr != NULL) 1189 if (--cur->nat_rule.ptr->states <= 0 && 1190 cur->nat_rule.ptr->src_nodes <= 0) 1191 pf_rm_rule(NULL, cur->nat_rule.ptr); 1192 if (cur->anchor.ptr != NULL) 1193 if (--cur->anchor.ptr->states <= 0) 1194 pf_rm_rule(NULL, cur->anchor.ptr); 1195 pf_normalize_tcp_cleanup(cur); 1196 pfi_kif_unref(cur->u.s.kif, PFI_KIF_REF_STATE); 1197 TAILQ_REMOVE(&state_list, cur, u.s.entry_list); 1198 if (cur->tag) 1199 pf_tag_unref(cur->tag); 1200 pool_put(&pf_state_pl, cur); 1201 pf_status.fcounters[FCNT_STATE_REMOVALS]++; 1202 pf_status.states--; 1203} 1204 1205void 1206pf_purge_expired_states(u_int32_t maxcheck) 1207{ 1208 static struct pf_state *cur = NULL; 1209 struct pf_state *next; 1210 int locked = 0; 1211 1212 while (maxcheck--) { 1213 /* wrap to start of list when we hit the end */ 1214 if (cur == NULL) { 1215 cur = TAILQ_FIRST(&state_list); 1216 if (cur == NULL) 1217 break; /* list empty */ 1218 } 1219 1220 /* get next state, as cur may get deleted */ 1221 next = TAILQ_NEXT(cur, u.s.entry_list); 1222 1223 if (cur->timeout == PFTM_UNLINKED) { 1224 /* free unlinked state */ 1225 if (! locked) { 1226#ifdef __FreeBSD__ 1227 if (!sx_try_upgrade(&pf_consistency_lock)) { 1228 PF_UNLOCK(); 1229 sx_sunlock(&pf_consistency_lock); 1230 sx_xlock(&pf_consistency_lock); 1231 PF_LOCK(); 1232 } 1233#else 1234 rw_enter_write(&pf_consistency_lock); 1235#endif 1236 locked = 1; 1237 } 1238 pf_free_state(cur); 1239 } else if (pf_state_expires(cur) <= time_second) { 1240 /* unlink and free expired state */ 1241 pf_unlink_state(cur); 1242 if (! locked) { 1243#ifdef __FreeBSD__ 1244 if (!sx_try_upgrade(&pf_consistency_lock)) { 1245 PF_UNLOCK(); 1246 sx_sunlock(&pf_consistency_lock); 1247 sx_xlock(&pf_consistency_lock); 1248 PF_LOCK(); 1249 } 1250#else 1251 rw_enter_write(&pf_consistency_lock); 1252#endif 1253 locked = 1; 1254 } 1255 pf_free_state(cur); 1256 } 1257 cur = next; 1258 } 1259 1260 if (locked) 1261#ifdef __FreeBSD__ 1262 sx_downgrade(&pf_consistency_lock); 1263#else 1264 rw_exit_write(&pf_consistency_lock); 1265#endif 1266} 1267 1268int 1269pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw) 1270{ 1271 if (aw->type != PF_ADDR_TABLE) 1272 return (0); 1273 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL) 1274 return (1); 1275 return (0); 1276} 1277 1278void 1279pf_tbladdr_remove(struct pf_addr_wrap *aw) 1280{ 1281 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL) 1282 return; 1283 pfr_detach_table(aw->p.tbl); 1284 aw->p.tbl = NULL; 1285} 1286 1287void 1288pf_tbladdr_copyout(struct pf_addr_wrap *aw) 1289{ 1290 struct pfr_ktable *kt = aw->p.tbl; 1291 1292 if (aw->type != PF_ADDR_TABLE || kt == NULL) 1293 return; 1294 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) 1295 kt = kt->pfrkt_root; 1296 aw->p.tbl = NULL; 1297 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ? 1298 kt->pfrkt_cnt : -1; 1299} 1300 1301void 1302pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af) 1303{ 1304 switch (af) { 1305#ifdef INET 1306 case AF_INET: { 1307 u_int32_t a = ntohl(addr->addr32[0]); 1308 printf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255, 1309 (a>>8)&255, a&255); 1310 if (p) { 1311 p = ntohs(p); 1312 printf(":%u", p); 1313 } 1314 break; 1315 } 1316#endif /* INET */ 1317#ifdef INET6 1318 case AF_INET6: { 1319 u_int16_t b; 1320 u_int8_t i, curstart = 255, curend = 0, 1321 maxstart = 0, maxend = 0; 1322 for (i = 0; i < 8; i++) { 1323 if (!addr->addr16[i]) { 1324 if (curstart == 255) 1325 curstart = i; 1326 else 1327 curend = i; 1328 } else { 1329 if (curstart) { 1330 if ((curend - curstart) > 1331 (maxend - maxstart)) { 1332 maxstart = curstart; 1333 maxend = curend; 1334 curstart = 255; 1335 } 1336 } 1337 } 1338 } 1339 for (i = 0; i < 8; i++) { 1340 if (i >= maxstart && i <= maxend) { 1341 if (maxend != 7) { 1342 if (i == maxstart) 1343 printf(":"); 1344 } else { 1345 if (i == maxend) 1346 printf(":"); 1347 } 1348 } else { 1349 b = ntohs(addr->addr16[i]); 1350 printf("%x", b); 1351 if (i < 7) 1352 printf(":"); 1353 } 1354 } 1355 if (p) { 1356 p = ntohs(p); 1357 printf("[%u]", p); 1358 } 1359 break; 1360 } 1361#endif /* INET6 */ 1362 } 1363} 1364 1365void 1366pf_print_state(struct pf_state *s) 1367{ 1368 switch (s->proto) { 1369 case IPPROTO_TCP: 1370 printf("TCP "); 1371 break; 1372 case IPPROTO_UDP: 1373 printf("UDP "); 1374 break; 1375 case IPPROTO_ICMP: 1376 printf("ICMP "); 1377 break; 1378 case IPPROTO_ICMPV6: 1379 printf("ICMPV6 "); 1380 break; 1381 default: 1382 printf("%u ", s->proto); 1383 break; 1384 } 1385 pf_print_host(&s->lan.addr, s->lan.port, s->af); 1386 printf(" "); 1387 pf_print_host(&s->gwy.addr, s->gwy.port, s->af); 1388 printf(" "); 1389 pf_print_host(&s->ext.addr, s->ext.port, s->af); 1390 printf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo, 1391 s->src.seqhi, s->src.max_win, s->src.seqdiff); 1392 if (s->src.wscale && s->dst.wscale) 1393 printf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK); 1394 printf("]"); 1395 printf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo, 1396 s->dst.seqhi, s->dst.max_win, s->dst.seqdiff); 1397 if (s->src.wscale && s->dst.wscale) 1398 printf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK); 1399 printf("]"); 1400 printf(" %u:%u", s->src.state, s->dst.state); 1401} 1402 1403void 1404pf_print_flags(u_int8_t f) 1405{ 1406 if (f) 1407 printf(" "); 1408 if (f & TH_FIN) 1409 printf("F"); 1410 if (f & TH_SYN) 1411 printf("S"); 1412 if (f & TH_RST) 1413 printf("R"); 1414 if (f & TH_PUSH) 1415 printf("P"); 1416 if (f & TH_ACK) 1417 printf("A"); 1418 if (f & TH_URG) 1419 printf("U"); 1420 if (f & TH_ECE) 1421 printf("E"); 1422 if (f & TH_CWR) 1423 printf("W"); 1424} 1425 1426#define PF_SET_SKIP_STEPS(i) \ 1427 do { \ 1428 while (head[i] != cur) { \ 1429 head[i]->skip[i].ptr = cur; \ 1430 head[i] = TAILQ_NEXT(head[i], entries); \ 1431 } \ 1432 } while (0) 1433 1434void 1435pf_calc_skip_steps(struct pf_rulequeue *rules) 1436{ 1437 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT]; 1438 int i; 1439 1440 cur = TAILQ_FIRST(rules); 1441 prev = cur; 1442 for (i = 0; i < PF_SKIP_COUNT; ++i) 1443 head[i] = cur; 1444 while (cur != NULL) { 1445 1446 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot) 1447 PF_SET_SKIP_STEPS(PF_SKIP_IFP); 1448 if (cur->direction != prev->direction) 1449 PF_SET_SKIP_STEPS(PF_SKIP_DIR); 1450 if (cur->af != prev->af) 1451 PF_SET_SKIP_STEPS(PF_SKIP_AF); 1452 if (cur->proto != prev->proto) 1453 PF_SET_SKIP_STEPS(PF_SKIP_PROTO); 1454 if (cur->src.neg != prev->src.neg || 1455 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr)) 1456 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR); 1457 if (cur->src.port[0] != prev->src.port[0] || 1458 cur->src.port[1] != prev->src.port[1] || 1459 cur->src.port_op != prev->src.port_op) 1460 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT); 1461 if (cur->dst.neg != prev->dst.neg || 1462 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr)) 1463 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR); 1464 if (cur->dst.port[0] != prev->dst.port[0] || 1465 cur->dst.port[1] != prev->dst.port[1] || 1466 cur->dst.port_op != prev->dst.port_op) 1467 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT); 1468 1469 prev = cur; 1470 cur = TAILQ_NEXT(cur, entries); 1471 } 1472 for (i = 0; i < PF_SKIP_COUNT; ++i) 1473 PF_SET_SKIP_STEPS(i); 1474} 1475 1476int 1477pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2) 1478{ 1479 if (aw1->type != aw2->type) 1480 return (1); 1481 switch (aw1->type) { 1482 case PF_ADDR_ADDRMASK: 1483 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0)) 1484 return (1); 1485 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0)) 1486 return (1); 1487 return (0); 1488 case PF_ADDR_DYNIFTL: 1489 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt); 1490 case PF_ADDR_NOROUTE: 1491 case PF_ADDR_URPFFAILED: 1492 return (0); 1493 case PF_ADDR_TABLE: 1494 return (aw1->p.tbl != aw2->p.tbl); 1495 case PF_ADDR_RTLABEL: 1496 return (aw1->v.rtlabel != aw2->v.rtlabel); 1497 default: 1498 printf("invalid address type: %d\n", aw1->type); 1499 return (1); 1500 } 1501} 1502 1503u_int16_t 1504pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp) 1505{ 1506 u_int32_t l; 1507 1508 if (udp && !cksum) 1509 return (0x0000); 1510 l = cksum + old - new; 1511 l = (l >> 16) + (l & 65535); 1512 l = l & 65535; 1513 if (udp && !l) 1514 return (0xFFFF); 1515 return (l); 1516} 1517 1518void 1519pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc, 1520 struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af) 1521{ 1522 struct pf_addr ao; 1523 u_int16_t po = *p; 1524 1525 PF_ACPY(&ao, a, af); 1526 PF_ACPY(a, an, af); 1527 1528 *p = pn; 1529 1530 switch (af) { 1531#ifdef INET 1532 case AF_INET: 1533 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic, 1534 ao.addr16[0], an->addr16[0], 0), 1535 ao.addr16[1], an->addr16[1], 0); 1536 *p = pn; 1537 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc, 1538 ao.addr16[0], an->addr16[0], u), 1539 ao.addr16[1], an->addr16[1], u), 1540 po, pn, u); 1541 break; 1542#endif /* INET */ 1543#ifdef INET6 1544 case AF_INET6: 1545 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1546 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1547 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc, 1548 ao.addr16[0], an->addr16[0], u), 1549 ao.addr16[1], an->addr16[1], u), 1550 ao.addr16[2], an->addr16[2], u), 1551 ao.addr16[3], an->addr16[3], u), 1552 ao.addr16[4], an->addr16[4], u), 1553 ao.addr16[5], an->addr16[5], u), 1554 ao.addr16[6], an->addr16[6], u), 1555 ao.addr16[7], an->addr16[7], u), 1556 po, pn, u); 1557 break; 1558#endif /* INET6 */ 1559 } 1560} 1561 1562 1563/* Changes a u_int32_t. Uses a void * so there are no align restrictions */ 1564void 1565pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u) 1566{ 1567 u_int32_t ao; 1568 1569 memcpy(&ao, a, sizeof(ao)); 1570 memcpy(a, &an, sizeof(u_int32_t)); 1571 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u), 1572 ao % 65536, an % 65536, u); 1573} 1574 1575#ifdef INET6 1576void 1577pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u) 1578{ 1579 struct pf_addr ao; 1580 1581 PF_ACPY(&ao, a, AF_INET6); 1582 PF_ACPY(a, an, AF_INET6); 1583 1584 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1585 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1586 pf_cksum_fixup(pf_cksum_fixup(*c, 1587 ao.addr16[0], an->addr16[0], u), 1588 ao.addr16[1], an->addr16[1], u), 1589 ao.addr16[2], an->addr16[2], u), 1590 ao.addr16[3], an->addr16[3], u), 1591 ao.addr16[4], an->addr16[4], u), 1592 ao.addr16[5], an->addr16[5], u), 1593 ao.addr16[6], an->addr16[6], u), 1594 ao.addr16[7], an->addr16[7], u); 1595} 1596#endif /* INET6 */ 1597 1598void 1599pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa, 1600 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c, 1601 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af) 1602{ 1603 struct pf_addr oia, ooa; 1604 1605 PF_ACPY(&oia, ia, af); 1606 PF_ACPY(&ooa, oa, af); 1607 1608 /* Change inner protocol port, fix inner protocol checksum. */ 1609 if (ip != NULL) { 1610 u_int16_t oip = *ip; 1611 u_int32_t opc = 0; /* make the compiler happy */ 1612 1613 if (pc != NULL) 1614 opc = *pc; 1615 *ip = np; 1616 if (pc != NULL) 1617 *pc = pf_cksum_fixup(*pc, oip, *ip, u); 1618 *ic = pf_cksum_fixup(*ic, oip, *ip, 0); 1619 if (pc != NULL) 1620 *ic = pf_cksum_fixup(*ic, opc, *pc, 0); 1621 } 1622 /* Change inner ip address, fix inner ip and icmp checksums. */ 1623 PF_ACPY(ia, na, af); 1624 switch (af) { 1625#ifdef INET 1626 case AF_INET: { 1627 u_int32_t oh2c = *h2c; 1628 1629 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c, 1630 oia.addr16[0], ia->addr16[0], 0), 1631 oia.addr16[1], ia->addr16[1], 0); 1632 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic, 1633 oia.addr16[0], ia->addr16[0], 0), 1634 oia.addr16[1], ia->addr16[1], 0); 1635 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0); 1636 break; 1637 } 1638#endif /* INET */ 1639#ifdef INET6 1640 case AF_INET6: 1641 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1642 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1643 pf_cksum_fixup(pf_cksum_fixup(*ic, 1644 oia.addr16[0], ia->addr16[0], u), 1645 oia.addr16[1], ia->addr16[1], u), 1646 oia.addr16[2], ia->addr16[2], u), 1647 oia.addr16[3], ia->addr16[3], u), 1648 oia.addr16[4], ia->addr16[4], u), 1649 oia.addr16[5], ia->addr16[5], u), 1650 oia.addr16[6], ia->addr16[6], u), 1651 oia.addr16[7], ia->addr16[7], u); 1652 break; 1653#endif /* INET6 */ 1654 } 1655 /* Change outer ip address, fix outer ip or icmpv6 checksum. */ 1656 PF_ACPY(oa, na, af); 1657 switch (af) { 1658#ifdef INET 1659 case AF_INET: 1660 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc, 1661 ooa.addr16[0], oa->addr16[0], 0), 1662 ooa.addr16[1], oa->addr16[1], 0); 1663 break; 1664#endif /* INET */ 1665#ifdef INET6 1666 case AF_INET6: 1667 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1668 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1669 pf_cksum_fixup(pf_cksum_fixup(*ic, 1670 ooa.addr16[0], oa->addr16[0], u), 1671 ooa.addr16[1], oa->addr16[1], u), 1672 ooa.addr16[2], oa->addr16[2], u), 1673 ooa.addr16[3], oa->addr16[3], u), 1674 ooa.addr16[4], oa->addr16[4], u), 1675 ooa.addr16[5], oa->addr16[5], u), 1676 ooa.addr16[6], oa->addr16[6], u), 1677 ooa.addr16[7], oa->addr16[7], u); 1678 break; 1679#endif /* INET6 */ 1680 } 1681} 1682 1683 1684/* 1685 * Need to modulate the sequence numbers in the TCP SACK option 1686 * (credits to Krzysztof Pfaff for report and patch) 1687 */ 1688int 1689pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd, 1690 struct tcphdr *th, struct pf_state_peer *dst) 1691{ 1692 int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen; 1693#ifdef __FreeBSD__ 1694 u_int8_t opts[TCP_MAXOLEN], *opt = opts; 1695#else 1696 u_int8_t opts[MAX_TCPOPTLEN], *opt = opts; 1697#endif 1698 int copyback = 0, i, olen; 1699 struct sackblk sack; 1700 1701#define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2) 1702 if (hlen < TCPOLEN_SACKLEN || 1703 !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af)) 1704 return 0; 1705 1706 while (hlen >= TCPOLEN_SACKLEN) { 1707 olen = opt[1]; 1708 switch (*opt) { 1709 case TCPOPT_EOL: /* FALLTHROUGH */ 1710 case TCPOPT_NOP: 1711 opt++; 1712 hlen--; 1713 break; 1714 case TCPOPT_SACK: 1715 if (olen > hlen) 1716 olen = hlen; 1717 if (olen >= TCPOLEN_SACKLEN) { 1718 for (i = 2; i + TCPOLEN_SACK <= olen; 1719 i += TCPOLEN_SACK) { 1720 memcpy(&sack, &opt[i], sizeof(sack)); 1721 pf_change_a(&sack.start, &th->th_sum, 1722 htonl(ntohl(sack.start) - 1723 dst->seqdiff), 0); 1724 pf_change_a(&sack.end, &th->th_sum, 1725 htonl(ntohl(sack.end) - 1726 dst->seqdiff), 0); 1727 memcpy(&opt[i], &sack, sizeof(sack)); 1728 } 1729 copyback = 1; 1730 } 1731 /* FALLTHROUGH */ 1732 default: 1733 if (olen < 2) 1734 olen = 2; 1735 hlen -= olen; 1736 opt += olen; 1737 } 1738 } 1739 1740 if (copyback) 1741#ifdef __FreeBSD__ 1742 m_copyback(m, off + sizeof(*th), thoptlen, (caddr_t)opts); 1743#else 1744 m_copyback(m, off + sizeof(*th), thoptlen, opts); 1745#endif 1746 return (copyback); 1747} 1748 1749void 1750#ifdef __FreeBSD__ 1751pf_send_tcp(struct mbuf *replyto, const struct pf_rule *r, sa_family_t af, 1752#else 1753pf_send_tcp(const struct pf_rule *r, sa_family_t af, 1754#endif 1755 const struct pf_addr *saddr, const struct pf_addr *daddr, 1756 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack, 1757 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag, 1758 u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp) 1759{ 1760 struct mbuf *m; 1761 int len, tlen; 1762#ifdef INET 1763 struct ip *h; 1764#endif /* INET */ 1765#ifdef INET6 1766 struct ip6_hdr *h6; 1767#endif /* INET6 */ 1768 struct tcphdr *th; 1769 char *opt; 1770 struct pf_mtag *pf_mtag; 1771 1772#ifdef __FreeBSD__ 1773 KASSERT( 1774#ifdef INET 1775 af == AF_INET 1776#else 1777 0 1778#endif 1779 || 1780#ifdef INET6 1781 af == AF_INET6 1782#else 1783 0 1784#endif 1785 , ("Unsupported AF %d", af)); 1786 len = 0; 1787 th = NULL; 1788#ifdef INET 1789 h = NULL; 1790#endif 1791#ifdef INET6 1792 h6 = NULL; 1793#endif 1794#endif 1795 1796 /* maximum segment size tcp option */ 1797 tlen = sizeof(struct tcphdr); 1798 if (mss) 1799 tlen += 4; 1800 1801 switch (af) { 1802#ifdef INET 1803 case AF_INET: 1804 len = sizeof(struct ip) + tlen; 1805 break; 1806#endif /* INET */ 1807#ifdef INET6 1808 case AF_INET6: 1809 len = sizeof(struct ip6_hdr) + tlen; 1810 break; 1811#endif /* INET6 */ 1812 } 1813 1814 /* create outgoing mbuf */ 1815 m = m_gethdr(M_DONTWAIT, MT_HEADER); 1816 if (m == NULL) 1817 return; 1818#ifdef __FreeBSD__ 1819#ifdef MAC 1820 if (replyto) 1821 mac_create_mbuf_netlayer(replyto, m); 1822 else 1823 mac_create_mbuf_from_firewall(m); 1824#else 1825 (void)replyto; 1826#endif 1827#endif 1828 if ((pf_mtag = pf_get_mtag(m)) == NULL) { 1829 m_freem(m); 1830 return; 1831 } 1832 if (tag) 1833#ifdef __FreeBSD__ 1834 m->m_flags |= M_SKIP_FIREWALL; 1835#else 1836 pf_mtag->flags |= PF_TAG_GENERATED; 1837#endif 1838 1839 pf_mtag->tag = rtag; 1840 1841 if (r != NULL && r->rtableid >= 0) 1842 pf_mtag->rtableid = r->rtableid; 1843#ifdef ALTQ 1844 if (r != NULL && r->qid) { 1845 pf_mtag->qid = r->qid; 1846 /* add hints for ecn */ 1847 pf_mtag->af = af; 1848 pf_mtag->hdr = mtod(m, struct ip *); 1849 } 1850#endif /* ALTQ */ 1851 m->m_data += max_linkhdr; 1852 m->m_pkthdr.len = m->m_len = len; 1853 m->m_pkthdr.rcvif = NULL; 1854 bzero(m->m_data, len); 1855 switch (af) { 1856#ifdef INET 1857 case AF_INET: 1858 h = mtod(m, struct ip *); 1859 1860 /* IP header fields included in the TCP checksum */ 1861 h->ip_p = IPPROTO_TCP; 1862 h->ip_len = htons(tlen); 1863 h->ip_src.s_addr = saddr->v4.s_addr; 1864 h->ip_dst.s_addr = daddr->v4.s_addr; 1865 1866 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip)); 1867 break; 1868#endif /* INET */ 1869#ifdef INET6 1870 case AF_INET6: 1871 h6 = mtod(m, struct ip6_hdr *); 1872 1873 /* IP header fields included in the TCP checksum */ 1874 h6->ip6_nxt = IPPROTO_TCP; 1875 h6->ip6_plen = htons(tlen); 1876 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr)); 1877 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr)); 1878 1879 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr)); 1880 break; 1881#endif /* INET6 */ 1882 } 1883 1884 /* TCP header */ 1885 th->th_sport = sport; 1886 th->th_dport = dport; 1887 th->th_seq = htonl(seq); 1888 th->th_ack = htonl(ack); 1889 th->th_off = tlen >> 2; 1890 th->th_flags = flags; 1891 th->th_win = htons(win); 1892 1893 if (mss) { 1894 opt = (char *)(th + 1); 1895 opt[0] = TCPOPT_MAXSEG; 1896 opt[1] = 4; 1897 HTONS(mss); 1898 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2); 1899 } 1900 1901 switch (af) { 1902#ifdef INET 1903 case AF_INET: 1904 /* TCP checksum */ 1905 th->th_sum = in_cksum(m, len); 1906 1907 /* Finish the IP header */ 1908 h->ip_v = 4; 1909 h->ip_hl = sizeof(*h) >> 2; 1910 h->ip_tos = IPTOS_LOWDELAY; 1911#ifdef __FreeBSD__ 1912 h->ip_off = path_mtu_discovery ? IP_DF : 0; 1913 h->ip_len = len; 1914#else 1915 h->ip_off = htons(ip_mtudisc ? IP_DF : 0); 1916 h->ip_len = htons(len); 1917#endif 1918 h->ip_ttl = ttl ? ttl : ip_defttl; 1919 h->ip_sum = 0; 1920 if (eh == NULL) { 1921#ifdef __FreeBSD__ 1922 PF_UNLOCK(); 1923 ip_output(m, (void *)NULL, (void *)NULL, 0, 1924 (void *)NULL, (void *)NULL); 1925 PF_LOCK(); 1926#else /* ! __FreeBSD__ */ 1927 ip_output(m, (void *)NULL, (void *)NULL, 0, 1928 (void *)NULL, (void *)NULL); 1929#endif 1930 } else { 1931 struct route ro; 1932 struct rtentry rt; 1933 struct ether_header *e = (void *)ro.ro_dst.sa_data; 1934 1935 if (ifp == NULL) { 1936 m_freem(m); 1937 return; 1938 } 1939 rt.rt_ifp = ifp; 1940 ro.ro_rt = &rt; 1941 ro.ro_dst.sa_len = sizeof(ro.ro_dst); 1942 ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT; 1943 bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN); 1944 bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN); 1945 e->ether_type = eh->ether_type; 1946#ifdef __FreeBSD__ 1947 PF_UNLOCK(); 1948 /* XXX_IMPORT: later */ 1949 ip_output(m, (void *)NULL, &ro, 0, 1950 (void *)NULL, (void *)NULL); 1951 PF_LOCK(); 1952#else /* ! __FreeBSD__ */ 1953 ip_output(m, (void *)NULL, &ro, IP_ROUTETOETHER, 1954 (void *)NULL, (void *)NULL); 1955#endif 1956 } 1957 break; 1958#endif /* INET */ 1959#ifdef INET6 1960 case AF_INET6: 1961 /* TCP checksum */ 1962 th->th_sum = in6_cksum(m, IPPROTO_TCP, 1963 sizeof(struct ip6_hdr), tlen); 1964 1965 h6->ip6_vfc |= IPV6_VERSION; 1966 h6->ip6_hlim = IPV6_DEFHLIM; 1967 1968#ifdef __FreeBSD__ 1969 PF_UNLOCK(); 1970 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL); 1971 PF_LOCK(); 1972#else 1973 ip6_output(m, NULL, NULL, 0, NULL, NULL); 1974#endif 1975 break; 1976#endif /* INET6 */ 1977 } 1978} 1979 1980void 1981pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af, 1982 struct pf_rule *r) 1983{ 1984 struct pf_mtag *pf_mtag; 1985 struct mbuf *m0; 1986#ifdef __FreeBSD__ 1987 struct ip *ip; 1988#endif 1989 1990#ifdef __FreeBSD__ 1991 m0 = m_copypacket(m, M_DONTWAIT); 1992 if (m0 == NULL) 1993 return; 1994#else 1995 m0 = m_copy(m, 0, M_COPYALL); 1996#endif 1997 if ((pf_mtag = pf_get_mtag(m0)) == NULL) 1998 return; 1999#ifdef __FreeBSD__ 2000 /* XXX: revisit */ 2001 m0->m_flags |= M_SKIP_FIREWALL; 2002#else 2003 pf_mtag->flags |= PF_TAG_GENERATED; 2004#endif 2005 2006 if (r->rtableid >= 0) 2007 pf_mtag->rtableid = r->rtableid; 2008 2009#ifdef ALTQ 2010 if (r->qid) { 2011 pf_mtag->qid = r->qid; 2012 /* add hints for ecn */ 2013 pf_mtag->af = af; 2014 pf_mtag->hdr = mtod(m0, struct ip *); 2015 } 2016#endif /* ALTQ */ 2017 2018 switch (af) { 2019#ifdef INET 2020 case AF_INET: 2021#ifdef __FreeBSD__ 2022 /* icmp_error() expects host byte ordering */ 2023 ip = mtod(m0, struct ip *); 2024 NTOHS(ip->ip_len); 2025 NTOHS(ip->ip_off); 2026 PF_UNLOCK(); 2027 icmp_error(m0, type, code, 0, 0); 2028 PF_LOCK(); 2029#else 2030 icmp_error(m0, type, code, 0, 0); 2031#endif 2032 break; 2033#endif /* INET */ 2034#ifdef INET6 2035 case AF_INET6: 2036#ifdef __FreeBSD__ 2037 PF_UNLOCK(); 2038#endif 2039 icmp6_error(m0, type, code, 0); 2040#ifdef __FreeBSD__ 2041 PF_LOCK(); 2042#endif 2043 break; 2044#endif /* INET6 */ 2045 } 2046} 2047 2048/* 2049 * Return 1 if the addresses a and b match (with mask m), otherwise return 0. 2050 * If n is 0, they match if they are equal. If n is != 0, they match if they 2051 * are different. 2052 */ 2053int 2054pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m, 2055 struct pf_addr *b, sa_family_t af) 2056{ 2057 int match = 0; 2058 2059 switch (af) { 2060#ifdef INET 2061 case AF_INET: 2062 if ((a->addr32[0] & m->addr32[0]) == 2063 (b->addr32[0] & m->addr32[0])) 2064 match++; 2065 break; 2066#endif /* INET */ 2067#ifdef INET6 2068 case AF_INET6: 2069 if (((a->addr32[0] & m->addr32[0]) == 2070 (b->addr32[0] & m->addr32[0])) && 2071 ((a->addr32[1] & m->addr32[1]) == 2072 (b->addr32[1] & m->addr32[1])) && 2073 ((a->addr32[2] & m->addr32[2]) == 2074 (b->addr32[2] & m->addr32[2])) && 2075 ((a->addr32[3] & m->addr32[3]) == 2076 (b->addr32[3] & m->addr32[3]))) 2077 match++; 2078 break; 2079#endif /* INET6 */ 2080 } 2081 if (match) { 2082 if (n) 2083 return (0); 2084 else 2085 return (1); 2086 } else { 2087 if (n) 2088 return (1); 2089 else 2090 return (0); 2091 } 2092} 2093 2094int 2095pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p) 2096{ 2097 switch (op) { 2098 case PF_OP_IRG: 2099 return ((p > a1) && (p < a2)); 2100 case PF_OP_XRG: 2101 return ((p < a1) || (p > a2)); 2102 case PF_OP_RRG: 2103 return ((p >= a1) && (p <= a2)); 2104 case PF_OP_EQ: 2105 return (p == a1); 2106 case PF_OP_NE: 2107 return (p != a1); 2108 case PF_OP_LT: 2109 return (p < a1); 2110 case PF_OP_LE: 2111 return (p <= a1); 2112 case PF_OP_GT: 2113 return (p > a1); 2114 case PF_OP_GE: 2115 return (p >= a1); 2116 } 2117 return (0); /* never reached */ 2118} 2119 2120int 2121pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p) 2122{ 2123 NTOHS(a1); 2124 NTOHS(a2); 2125 NTOHS(p); 2126 return (pf_match(op, a1, a2, p)); 2127} 2128 2129int 2130pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u) 2131{ 2132 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE) 2133 return (0); 2134 return (pf_match(op, a1, a2, u)); 2135} 2136 2137int 2138pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g) 2139{ 2140 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE) 2141 return (0); 2142 return (pf_match(op, a1, a2, g)); 2143} 2144 2145#ifndef __FreeBSD__ 2146struct pf_mtag * 2147pf_find_mtag(struct mbuf *m) 2148{ 2149 struct m_tag *mtag; 2150 2151 if ((mtag = m_tag_find(m, PACKET_TAG_PF, NULL)) == NULL) 2152 return (NULL); 2153 2154 return ((struct pf_mtag *)(mtag + 1)); 2155} 2156 2157struct pf_mtag * 2158pf_get_mtag(struct mbuf *m) 2159{ 2160 struct m_tag *mtag; 2161 2162 if ((mtag = m_tag_find(m, PACKET_TAG_PF, NULL)) == NULL) { 2163 mtag = m_tag_get(PACKET_TAG_PF, sizeof(struct pf_mtag), 2164 M_NOWAIT); 2165 if (mtag == NULL) 2166 return (NULL); 2167 bzero(mtag + 1, sizeof(struct pf_mtag)); 2168 m_tag_prepend(m, mtag); 2169 } 2170 2171 return ((struct pf_mtag *)(mtag + 1)); 2172} 2173#endif 2174 2175int 2176pf_match_tag(struct mbuf *m, struct pf_rule *r, struct pf_mtag *pf_mtag, 2177 int *tag) 2178{ 2179 if (*tag == -1) 2180 *tag = pf_mtag->tag; 2181 2182 return ((!r->match_tag_not && r->match_tag == *tag) || 2183 (r->match_tag_not && r->match_tag != *tag)); 2184} 2185 2186int 2187pf_tag_packet(struct mbuf *m, struct pf_mtag *pf_mtag, int tag, int rtableid) 2188{ 2189 if (tag <= 0 && rtableid < 0) 2190 return (0); 2191 2192 if (pf_mtag == NULL) 2193 if ((pf_mtag = pf_get_mtag(m)) == NULL) 2194 return (1); 2195 if (tag > 0) 2196 pf_mtag->tag = tag; 2197 if (rtableid >= 0) 2198 pf_mtag->rtableid = rtableid; 2199 2200 return (0); 2201} 2202 2203static void 2204pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n, 2205 struct pf_rule **r, struct pf_rule **a, int *match) 2206{ 2207 struct pf_anchor_stackframe *f; 2208 2209 (*r)->anchor->match = 0; 2210 if (match) 2211 *match = 0; 2212 if (*depth >= sizeof(pf_anchor_stack) / 2213 sizeof(pf_anchor_stack[0])) { 2214 printf("pf_step_into_anchor: stack overflow\n"); 2215 *r = TAILQ_NEXT(*r, entries); 2216 return; 2217 } else if (*depth == 0 && a != NULL) 2218 *a = *r; 2219 f = pf_anchor_stack + (*depth)++; 2220 f->rs = *rs; 2221 f->r = *r; 2222 if ((*r)->anchor_wildcard) { 2223 f->parent = &(*r)->anchor->children; 2224 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) == 2225 NULL) { 2226 *r = NULL; 2227 return; 2228 } 2229 *rs = &f->child->ruleset; 2230 } else { 2231 f->parent = NULL; 2232 f->child = NULL; 2233 *rs = &(*r)->anchor->ruleset; 2234 } 2235 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr); 2236} 2237 2238int 2239pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n, 2240 struct pf_rule **r, struct pf_rule **a, int *match) 2241{ 2242 struct pf_anchor_stackframe *f; 2243 int quick = 0; 2244 2245 do { 2246 if (*depth <= 0) 2247 break; 2248 f = pf_anchor_stack + *depth - 1; 2249 if (f->parent != NULL && f->child != NULL) { 2250 if (f->child->match || 2251 (match != NULL && *match)) { 2252 f->r->anchor->match = 1; 2253 *match = 0; 2254 } 2255 f->child = RB_NEXT(pf_anchor_node, f->parent, f->child); 2256 if (f->child != NULL) { 2257 *rs = &f->child->ruleset; 2258 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr); 2259 if (*r == NULL) 2260 continue; 2261 else 2262 break; 2263 } 2264 } 2265 (*depth)--; 2266 if (*depth == 0 && a != NULL) 2267 *a = NULL; 2268 *rs = f->rs; 2269 if (f->r->anchor->match || (match != NULL && *match)) 2270 quick = f->r->quick; 2271 *r = TAILQ_NEXT(f->r, entries); 2272 } while (*r == NULL); 2273 2274 return (quick); 2275} 2276 2277#ifdef INET6 2278void 2279pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr, 2280 struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af) 2281{ 2282 switch (af) { 2283#ifdef INET 2284 case AF_INET: 2285 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) | 2286 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]); 2287 break; 2288#endif /* INET */ 2289 case AF_INET6: 2290 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) | 2291 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]); 2292 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) | 2293 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]); 2294 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) | 2295 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]); 2296 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) | 2297 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]); 2298 break; 2299 } 2300} 2301 2302void 2303pf_addr_inc(struct pf_addr *addr, sa_family_t af) 2304{ 2305 switch (af) { 2306#ifdef INET 2307 case AF_INET: 2308 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1); 2309 break; 2310#endif /* INET */ 2311 case AF_INET6: 2312 if (addr->addr32[3] == 0xffffffff) { 2313 addr->addr32[3] = 0; 2314 if (addr->addr32[2] == 0xffffffff) { 2315 addr->addr32[2] = 0; 2316 if (addr->addr32[1] == 0xffffffff) { 2317 addr->addr32[1] = 0; 2318 addr->addr32[0] = 2319 htonl(ntohl(addr->addr32[0]) + 1); 2320 } else 2321 addr->addr32[1] = 2322 htonl(ntohl(addr->addr32[1]) + 1); 2323 } else 2324 addr->addr32[2] = 2325 htonl(ntohl(addr->addr32[2]) + 1); 2326 } else 2327 addr->addr32[3] = 2328 htonl(ntohl(addr->addr32[3]) + 1); 2329 break; 2330 } 2331} 2332#endif /* INET6 */ 2333 2334#define mix(a,b,c) \ 2335 do { \ 2336 a -= b; a -= c; a ^= (c >> 13); \ 2337 b -= c; b -= a; b ^= (a << 8); \ 2338 c -= a; c -= b; c ^= (b >> 13); \ 2339 a -= b; a -= c; a ^= (c >> 12); \ 2340 b -= c; b -= a; b ^= (a << 16); \ 2341 c -= a; c -= b; c ^= (b >> 5); \ 2342 a -= b; a -= c; a ^= (c >> 3); \ 2343 b -= c; b -= a; b ^= (a << 10); \ 2344 c -= a; c -= b; c ^= (b >> 15); \ 2345 } while (0) 2346 2347/* 2348 * hash function based on bridge_hash in if_bridge.c 2349 */ 2350void 2351pf_hash(struct pf_addr *inaddr, struct pf_addr *hash, 2352 struct pf_poolhashkey *key, sa_family_t af) 2353{ 2354 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0]; 2355 2356 switch (af) { 2357#ifdef INET 2358 case AF_INET: 2359 a += inaddr->addr32[0]; 2360 b += key->key32[1]; 2361 mix(a, b, c); 2362 hash->addr32[0] = c + key->key32[2]; 2363 break; 2364#endif /* INET */ 2365#ifdef INET6 2366 case AF_INET6: 2367 a += inaddr->addr32[0]; 2368 b += inaddr->addr32[2]; 2369 mix(a, b, c); 2370 hash->addr32[0] = c; 2371 a += inaddr->addr32[1]; 2372 b += inaddr->addr32[3]; 2373 c += key->key32[1]; 2374 mix(a, b, c); 2375 hash->addr32[1] = c; 2376 a += inaddr->addr32[2]; 2377 b += inaddr->addr32[1]; 2378 c += key->key32[2]; 2379 mix(a, b, c); 2380 hash->addr32[2] = c; 2381 a += inaddr->addr32[3]; 2382 b += inaddr->addr32[0]; 2383 c += key->key32[3]; 2384 mix(a, b, c); 2385 hash->addr32[3] = c; 2386 break; 2387#endif /* INET6 */ 2388 } 2389} 2390 2391int 2392pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr, 2393 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn) 2394{ 2395 unsigned char hash[16]; 2396 struct pf_pool *rpool = &r->rpool; 2397 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr; 2398 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask; 2399 struct pf_pooladdr *acur = rpool->cur; 2400 struct pf_src_node k; 2401 2402 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR && 2403 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { 2404 k.af = af; 2405 PF_ACPY(&k.addr, saddr, af); 2406 if (r->rule_flag & PFRULE_RULESRCTRACK || 2407 r->rpool.opts & PF_POOL_STICKYADDR) 2408 k.rule.ptr = r; 2409 else 2410 k.rule.ptr = NULL; 2411 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++; 2412 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k); 2413 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) { 2414 PF_ACPY(naddr, &(*sn)->raddr, af); 2415 if (pf_status.debug >= PF_DEBUG_MISC) { 2416 printf("pf_map_addr: src tracking maps "); 2417 pf_print_host(&k.addr, 0, af); 2418 printf(" to "); 2419 pf_print_host(naddr, 0, af); 2420 printf("\n"); 2421 } 2422 return (0); 2423 } 2424 } 2425 2426 if (rpool->cur->addr.type == PF_ADDR_NOROUTE) 2427 return (1); 2428 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2429 switch (af) { 2430#ifdef INET 2431 case AF_INET: 2432 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 && 2433 (rpool->opts & PF_POOL_TYPEMASK) != 2434 PF_POOL_ROUNDROBIN) 2435 return (1); 2436 raddr = &rpool->cur->addr.p.dyn->pfid_addr4; 2437 rmask = &rpool->cur->addr.p.dyn->pfid_mask4; 2438 break; 2439#endif /* INET */ 2440#ifdef INET6 2441 case AF_INET6: 2442 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 && 2443 (rpool->opts & PF_POOL_TYPEMASK) != 2444 PF_POOL_ROUNDROBIN) 2445 return (1); 2446 raddr = &rpool->cur->addr.p.dyn->pfid_addr6; 2447 rmask = &rpool->cur->addr.p.dyn->pfid_mask6; 2448 break; 2449#endif /* INET6 */ 2450 } 2451 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2452 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN) 2453 return (1); /* unsupported */ 2454 } else { 2455 raddr = &rpool->cur->addr.v.a.addr; 2456 rmask = &rpool->cur->addr.v.a.mask; 2457 } 2458 2459 switch (rpool->opts & PF_POOL_TYPEMASK) { 2460 case PF_POOL_NONE: 2461 PF_ACPY(naddr, raddr, af); 2462 break; 2463 case PF_POOL_BITMASK: 2464 PF_POOLMASK(naddr, raddr, rmask, saddr, af); 2465 break; 2466 case PF_POOL_RANDOM: 2467 if (init_addr != NULL && PF_AZERO(init_addr, af)) { 2468 switch (af) { 2469#ifdef INET 2470 case AF_INET: 2471 rpool->counter.addr32[0] = htonl(arc4random()); 2472 break; 2473#endif /* INET */ 2474#ifdef INET6 2475 case AF_INET6: 2476 if (rmask->addr32[3] != 0xffffffff) 2477 rpool->counter.addr32[3] = 2478 htonl(arc4random()); 2479 else 2480 break; 2481 if (rmask->addr32[2] != 0xffffffff) 2482 rpool->counter.addr32[2] = 2483 htonl(arc4random()); 2484 else 2485 break; 2486 if (rmask->addr32[1] != 0xffffffff) 2487 rpool->counter.addr32[1] = 2488 htonl(arc4random()); 2489 else 2490 break; 2491 if (rmask->addr32[0] != 0xffffffff) 2492 rpool->counter.addr32[0] = 2493 htonl(arc4random()); 2494 break; 2495#endif /* INET6 */ 2496 } 2497 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af); 2498 PF_ACPY(init_addr, naddr, af); 2499 2500 } else { 2501 PF_AINC(&rpool->counter, af); 2502 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af); 2503 } 2504 break; 2505 case PF_POOL_SRCHASH: 2506 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af); 2507 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af); 2508 break; 2509 case PF_POOL_ROUNDROBIN: 2510 if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2511 if (!pfr_pool_get(rpool->cur->addr.p.tbl, 2512 &rpool->tblidx, &rpool->counter, 2513 &raddr, &rmask, af)) 2514 goto get_addr; 2515 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2516 if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, 2517 &rpool->tblidx, &rpool->counter, 2518 &raddr, &rmask, af)) 2519 goto get_addr; 2520 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af)) 2521 goto get_addr; 2522 2523 try_next: 2524 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL) 2525 rpool->cur = TAILQ_FIRST(&rpool->list); 2526 if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2527 rpool->tblidx = -1; 2528 if (pfr_pool_get(rpool->cur->addr.p.tbl, 2529 &rpool->tblidx, &rpool->counter, 2530 &raddr, &rmask, af)) { 2531 /* table contains no address of type 'af' */ 2532 if (rpool->cur != acur) 2533 goto try_next; 2534 return (1); 2535 } 2536 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2537 rpool->tblidx = -1; 2538 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, 2539 &rpool->tblidx, &rpool->counter, 2540 &raddr, &rmask, af)) { 2541 /* table contains no address of type 'af' */ 2542 if (rpool->cur != acur) 2543 goto try_next; 2544 return (1); 2545 } 2546 } else { 2547 raddr = &rpool->cur->addr.v.a.addr; 2548 rmask = &rpool->cur->addr.v.a.mask; 2549 PF_ACPY(&rpool->counter, raddr, af); 2550 } 2551 2552 get_addr: 2553 PF_ACPY(naddr, &rpool->counter, af); 2554 if (init_addr != NULL && PF_AZERO(init_addr, af)) 2555 PF_ACPY(init_addr, naddr, af); 2556 PF_AINC(&rpool->counter, af); 2557 break; 2558 } 2559 if (*sn != NULL) 2560 PF_ACPY(&(*sn)->raddr, naddr, af); 2561 2562 if (pf_status.debug >= PF_DEBUG_MISC && 2563 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { 2564 printf("pf_map_addr: selected address "); 2565 pf_print_host(naddr, 0, af); 2566 printf("\n"); 2567 } 2568 2569 return (0); 2570} 2571 2572int 2573pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r, 2574 struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport, 2575 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high, 2576 struct pf_src_node **sn) 2577{ 2578 struct pf_state_cmp key; 2579 struct pf_addr init_addr; 2580 u_int16_t cut; 2581 2582 bzero(&init_addr, sizeof(init_addr)); 2583 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn)) 2584 return (1); 2585 2586 if (proto == IPPROTO_ICMP) { 2587 low = 1; 2588 high = 65535; 2589 } 2590 2591 do { 2592 key.af = af; 2593 key.proto = proto; 2594 PF_ACPY(&key.ext.addr, daddr, key.af); 2595 PF_ACPY(&key.gwy.addr, naddr, key.af); 2596 key.ext.port = dport; 2597 2598 /* 2599 * port search; start random, step; 2600 * similar 2 portloop in in_pcbbind 2601 */ 2602 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP || 2603 proto == IPPROTO_ICMP)) { 2604 key.gwy.port = dport; 2605 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) 2606 return (0); 2607 } else if (low == 0 && high == 0) { 2608 key.gwy.port = *nport; 2609 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) 2610 return (0); 2611 } else if (low == high) { 2612 key.gwy.port = htons(low); 2613 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) { 2614 *nport = htons(low); 2615 return (0); 2616 } 2617 } else { 2618 u_int16_t tmp; 2619 2620 if (low > high) { 2621 tmp = low; 2622 low = high; 2623 high = tmp; 2624 } 2625 /* low < high */ 2626 cut = htonl(arc4random()) % (1 + high - low) + low; 2627 /* low <= cut <= high */ 2628 for (tmp = cut; tmp <= high; ++(tmp)) { 2629 key.gwy.port = htons(tmp); 2630 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == 2631 NULL) { 2632 *nport = htons(tmp); 2633 return (0); 2634 } 2635 } 2636 for (tmp = cut - 1; tmp >= low; --(tmp)) { 2637 key.gwy.port = htons(tmp); 2638 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == 2639 NULL) { 2640 *nport = htons(tmp); 2641 return (0); 2642 } 2643 } 2644 } 2645 2646 switch (r->rpool.opts & PF_POOL_TYPEMASK) { 2647 case PF_POOL_RANDOM: 2648 case PF_POOL_ROUNDROBIN: 2649 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn)) 2650 return (1); 2651 break; 2652 case PF_POOL_NONE: 2653 case PF_POOL_SRCHASH: 2654 case PF_POOL_BITMASK: 2655 default: 2656 return (1); 2657 } 2658 } while (! PF_AEQ(&init_addr, naddr, af) ); 2659 2660 return (1); /* none available */ 2661} 2662 2663struct pf_rule * 2664pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off, 2665 int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport, 2666 struct pf_addr *daddr, u_int16_t dport, int rs_num) 2667{ 2668 struct pf_rule *r, *rm = NULL; 2669 struct pf_ruleset *ruleset = NULL; 2670 int tag = -1; 2671 int rtableid = -1; 2672 int asd = 0; 2673 2674 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr); 2675 while (r && rm == NULL) { 2676 struct pf_rule_addr *src = NULL, *dst = NULL; 2677 struct pf_addr_wrap *xdst = NULL; 2678 2679 if (r->action == PF_BINAT && direction == PF_IN) { 2680 src = &r->dst; 2681 if (r->rpool.cur != NULL) 2682 xdst = &r->rpool.cur->addr; 2683 } else { 2684 src = &r->src; 2685 dst = &r->dst; 2686 } 2687 2688 r->evaluations++; 2689 if (pfi_kif_match(r->kif, kif) == r->ifnot) 2690 r = r->skip[PF_SKIP_IFP].ptr; 2691 else if (r->direction && r->direction != direction) 2692 r = r->skip[PF_SKIP_DIR].ptr; 2693 else if (r->af && r->af != pd->af) 2694 r = r->skip[PF_SKIP_AF].ptr; 2695 else if (r->proto && r->proto != pd->proto) 2696 r = r->skip[PF_SKIP_PROTO].ptr; 2697 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af, 2698 src->neg, kif)) 2699 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR : 2700 PF_SKIP_DST_ADDR].ptr; 2701 else if (src->port_op && !pf_match_port(src->port_op, 2702 src->port[0], src->port[1], sport)) 2703 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT : 2704 PF_SKIP_DST_PORT].ptr; 2705 else if (dst != NULL && 2706 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL)) 2707 r = r->skip[PF_SKIP_DST_ADDR].ptr; 2708 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af, 2709 0, NULL)) 2710 r = TAILQ_NEXT(r, entries); 2711 else if (dst != NULL && dst->port_op && 2712 !pf_match_port(dst->port_op, dst->port[0], 2713 dst->port[1], dport)) 2714 r = r->skip[PF_SKIP_DST_PORT].ptr; 2715 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) 2716 r = TAILQ_NEXT(r, entries); 2717 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto != 2718 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m, 2719 off, pd->hdr.tcp), r->os_fingerprint))) 2720 r = TAILQ_NEXT(r, entries); 2721 else { 2722 if (r->tag) 2723 tag = r->tag; 2724 if (r->rtableid >= 0) 2725 rtableid = r->rtableid; 2726 if (r->anchor == NULL) { 2727 rm = r; 2728 } else 2729 pf_step_into_anchor(&asd, &ruleset, rs_num, 2730 &r, NULL, NULL); 2731 } 2732 if (r == NULL) 2733 pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r, 2734 NULL, NULL); 2735 } 2736 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) 2737 return (NULL); 2738 if (rm != NULL && (rm->action == PF_NONAT || 2739 rm->action == PF_NORDR || rm->action == PF_NOBINAT)) 2740 return (NULL); 2741 return (rm); 2742} 2743 2744struct pf_rule * 2745pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction, 2746 struct pfi_kif *kif, struct pf_src_node **sn, 2747 struct pf_addr *saddr, u_int16_t sport, 2748 struct pf_addr *daddr, u_int16_t dport, 2749 struct pf_addr *naddr, u_int16_t *nport) 2750{ 2751 struct pf_rule *r = NULL; 2752 2753 if (direction == PF_OUT) { 2754 r = pf_match_translation(pd, m, off, direction, kif, saddr, 2755 sport, daddr, dport, PF_RULESET_BINAT); 2756 if (r == NULL) 2757 r = pf_match_translation(pd, m, off, direction, kif, 2758 saddr, sport, daddr, dport, PF_RULESET_NAT); 2759 } else { 2760 r = pf_match_translation(pd, m, off, direction, kif, saddr, 2761 sport, daddr, dport, PF_RULESET_RDR); 2762 if (r == NULL) 2763 r = pf_match_translation(pd, m, off, direction, kif, 2764 saddr, sport, daddr, dport, PF_RULESET_BINAT); 2765 } 2766 2767 if (r != NULL) { 2768 switch (r->action) { 2769 case PF_NONAT: 2770 case PF_NOBINAT: 2771 case PF_NORDR: 2772 return (NULL); 2773 case PF_NAT: 2774 if (pf_get_sport(pd->af, pd->proto, r, saddr, 2775 daddr, dport, naddr, nport, r->rpool.proxy_port[0], 2776 r->rpool.proxy_port[1], sn)) { 2777 DPFPRINTF(PF_DEBUG_MISC, 2778 ("pf: NAT proxy port allocation " 2779 "(%u-%u) failed\n", 2780 r->rpool.proxy_port[0], 2781 r->rpool.proxy_port[1])); 2782 return (NULL); 2783 } 2784 break; 2785 case PF_BINAT: 2786 switch (direction) { 2787 case PF_OUT: 2788 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){ 2789 switch (pd->af) { 2790#ifdef INET 2791 case AF_INET: 2792 if (r->rpool.cur->addr.p.dyn-> 2793 pfid_acnt4 < 1) 2794 return (NULL); 2795 PF_POOLMASK(naddr, 2796 &r->rpool.cur->addr.p.dyn-> 2797 pfid_addr4, 2798 &r->rpool.cur->addr.p.dyn-> 2799 pfid_mask4, 2800 saddr, AF_INET); 2801 break; 2802#endif /* INET */ 2803#ifdef INET6 2804 case AF_INET6: 2805 if (r->rpool.cur->addr.p.dyn-> 2806 pfid_acnt6 < 1) 2807 return (NULL); 2808 PF_POOLMASK(naddr, 2809 &r->rpool.cur->addr.p.dyn-> 2810 pfid_addr6, 2811 &r->rpool.cur->addr.p.dyn-> 2812 pfid_mask6, 2813 saddr, AF_INET6); 2814 break; 2815#endif /* INET6 */ 2816 } 2817 } else 2818 PF_POOLMASK(naddr, 2819 &r->rpool.cur->addr.v.a.addr, 2820 &r->rpool.cur->addr.v.a.mask, 2821 saddr, pd->af); 2822 break; 2823 case PF_IN: 2824 if (r->src.addr.type == PF_ADDR_DYNIFTL) { 2825 switch (pd->af) { 2826#ifdef INET 2827 case AF_INET: 2828 if (r->src.addr.p.dyn-> 2829 pfid_acnt4 < 1) 2830 return (NULL); 2831 PF_POOLMASK(naddr, 2832 &r->src.addr.p.dyn-> 2833 pfid_addr4, 2834 &r->src.addr.p.dyn-> 2835 pfid_mask4, 2836 daddr, AF_INET); 2837 break; 2838#endif /* INET */ 2839#ifdef INET6 2840 case AF_INET6: 2841 if (r->src.addr.p.dyn-> 2842 pfid_acnt6 < 1) 2843 return (NULL); 2844 PF_POOLMASK(naddr, 2845 &r->src.addr.p.dyn-> 2846 pfid_addr6, 2847 &r->src.addr.p.dyn-> 2848 pfid_mask6, 2849 daddr, AF_INET6); 2850 break; 2851#endif /* INET6 */ 2852 } 2853 } else 2854 PF_POOLMASK(naddr, 2855 &r->src.addr.v.a.addr, 2856 &r->src.addr.v.a.mask, daddr, 2857 pd->af); 2858 break; 2859 } 2860 break; 2861 case PF_RDR: { 2862 if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn)) 2863 return (NULL); 2864 if ((r->rpool.opts & PF_POOL_TYPEMASK) == 2865 PF_POOL_BITMASK) 2866 PF_POOLMASK(naddr, naddr, 2867 &r->rpool.cur->addr.v.a.mask, daddr, 2868 pd->af); 2869 2870 if (r->rpool.proxy_port[1]) { 2871 u_int32_t tmp_nport; 2872 2873 tmp_nport = ((ntohs(dport) - 2874 ntohs(r->dst.port[0])) % 2875 (r->rpool.proxy_port[1] - 2876 r->rpool.proxy_port[0] + 1)) + 2877 r->rpool.proxy_port[0]; 2878 2879 /* wrap around if necessary */ 2880 if (tmp_nport > 65535) 2881 tmp_nport -= 65535; 2882 *nport = htons((u_int16_t)tmp_nport); 2883 } else if (r->rpool.proxy_port[0]) 2884 *nport = htons(r->rpool.proxy_port[0]); 2885 break; 2886 } 2887 default: 2888 return (NULL); 2889 } 2890 } 2891 2892 return (r); 2893} 2894 2895int 2896#ifdef __FreeBSD__ 2897pf_socket_lookup(int direction, struct pf_pdesc *pd, struct inpcb *inp_arg) 2898#else 2899pf_socket_lookup(int direction, struct pf_pdesc *pd) 2900#endif 2901{ 2902 struct pf_addr *saddr, *daddr; 2903 u_int16_t sport, dport; 2904#ifdef __FreeBSD__ 2905 struct inpcbinfo *pi; 2906#else 2907 struct inpcbtable *tb; 2908#endif 2909 struct inpcb *inp; 2910 2911 if (pd == NULL) 2912 return (-1); 2913 pd->lookup.uid = UID_MAX; 2914 pd->lookup.gid = GID_MAX; 2915 pd->lookup.pid = NO_PID; /* XXX: revisit */ 2916#ifdef __FreeBSD__ 2917 if (inp_arg != NULL) { 2918 INP_LOCK_ASSERT(inp_arg); 2919 if (inp_arg->inp_socket) { 2920 pd->lookup.uid = inp_arg->inp_socket->so_cred->cr_uid; 2921 pd->lookup.gid = 2922 inp_arg->inp_socket->so_cred->cr_groups[0]; 2923 return (1); 2924 } else 2925 return (-1); 2926 } 2927#endif 2928 switch (pd->proto) { 2929 case IPPROTO_TCP: 2930 if (pd->hdr.tcp == NULL) 2931 return (-1); 2932 sport = pd->hdr.tcp->th_sport; 2933 dport = pd->hdr.tcp->th_dport; 2934#ifdef __FreeBSD__ 2935 pi = &tcbinfo; 2936#else 2937 tb = &tcbtable; 2938#endif 2939 break; 2940 case IPPROTO_UDP: 2941 if (pd->hdr.udp == NULL) 2942 return (-1); 2943 sport = pd->hdr.udp->uh_sport; 2944 dport = pd->hdr.udp->uh_dport; 2945#ifdef __FreeBSD__ 2946 pi = &udbinfo; 2947#else 2948 tb = &udbtable; 2949#endif 2950 break; 2951 default: 2952 return (-1); 2953 } 2954 if (direction == PF_IN) { 2955 saddr = pd->src; 2956 daddr = pd->dst; 2957 } else { 2958 u_int16_t p; 2959 2960 p = sport; 2961 sport = dport; 2962 dport = p; 2963 saddr = pd->dst; 2964 daddr = pd->src; 2965 } 2966 switch (pd->af) { 2967#ifdef INET 2968 case AF_INET: 2969#ifdef __FreeBSD__ 2970 INP_INFO_RLOCK(pi); /* XXX LOR */ 2971 inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4, 2972 dport, 0, NULL); 2973 if (inp == NULL) { 2974 inp = in_pcblookup_hash(pi, saddr->v4, sport, 2975 daddr->v4, dport, INPLOOKUP_WILDCARD, NULL); 2976 if(inp == NULL) { 2977 INP_INFO_RUNLOCK(pi); 2978 return (-1); 2979 } 2980 } 2981#else 2982 inp = in_pcbhashlookup(tb, saddr->v4, sport, daddr->v4, dport); 2983 if (inp == NULL) { 2984 inp = in_pcblookup_listen(tb, daddr->v4, dport, 0); 2985 if (inp == NULL) 2986 return (-1); 2987 } 2988#endif 2989 break; 2990#endif /* INET */ 2991#ifdef INET6 2992 case AF_INET6: 2993#ifdef __FreeBSD__ 2994 INP_INFO_RLOCK(pi); 2995 inp = in6_pcblookup_hash(pi, &saddr->v6, sport, 2996 &daddr->v6, dport, 0, NULL); 2997 if (inp == NULL) { 2998 inp = in6_pcblookup_hash(pi, &saddr->v6, sport, 2999 &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL); 3000 if (inp == NULL) { 3001 INP_INFO_RUNLOCK(pi); 3002 return (-1); 3003 } 3004 } 3005#else 3006 inp = in6_pcbhashlookup(tb, &saddr->v6, sport, &daddr->v6, 3007 dport); 3008 if (inp == NULL) { 3009 inp = in6_pcblookup_listen(tb, &daddr->v6, dport, 0); 3010 if (inp == NULL) 3011 return (-1); 3012 } 3013#endif 3014 break; 3015#endif /* INET6 */ 3016 3017 default: 3018 return (-1); 3019 } 3020#ifdef __FreeBSD__ 3021 INP_LOCK(inp); 3022 if ((inp->inp_socket == NULL) || (inp->inp_socket->so_cred == NULL)) { 3023 INP_UNLOCK(inp); 3024 INP_INFO_RUNLOCK(pi); 3025 return (-1); 3026 } 3027 pd->lookup.uid = inp->inp_socket->so_cred->cr_uid; 3028 pd->lookup.gid = inp->inp_socket->so_cred->cr_groups[0]; 3029 INP_UNLOCK(inp); 3030 INP_INFO_RUNLOCK(pi); 3031#else 3032 pd->lookup.uid = inp->inp_socket->so_euid; 3033 pd->lookup.gid = inp->inp_socket->so_egid; 3034 pd->lookup.pid = inp->inp_socket->so_cpid; 3035#endif 3036 return (1); 3037} 3038 3039u_int8_t 3040pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af) 3041{ 3042 int hlen; 3043 u_int8_t hdr[60]; 3044 u_int8_t *opt, optlen; 3045 u_int8_t wscale = 0; 3046 3047 hlen = th_off << 2; /* hlen <= sizeof(hdr) */ 3048 if (hlen <= sizeof(struct tcphdr)) 3049 return (0); 3050 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af)) 3051 return (0); 3052 opt = hdr + sizeof(struct tcphdr); 3053 hlen -= sizeof(struct tcphdr); 3054 while (hlen >= 3) { 3055 switch (*opt) { 3056 case TCPOPT_EOL: 3057 case TCPOPT_NOP: 3058 ++opt; 3059 --hlen; 3060 break; 3061 case TCPOPT_WINDOW: 3062 wscale = opt[2]; 3063 if (wscale > TCP_MAX_WINSHIFT) 3064 wscale = TCP_MAX_WINSHIFT; 3065 wscale |= PF_WSCALE_FLAG; 3066 /* FALLTHROUGH */ 3067 default: 3068 optlen = opt[1]; 3069 if (optlen < 2) 3070 optlen = 2; 3071 hlen -= optlen; 3072 opt += optlen; 3073 break; 3074 } 3075 } 3076 return (wscale); 3077} 3078 3079u_int16_t 3080pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af) 3081{ 3082 int hlen; 3083 u_int8_t hdr[60]; 3084 u_int8_t *opt, optlen; 3085 u_int16_t mss = tcp_mssdflt; 3086 3087 hlen = th_off << 2; /* hlen <= sizeof(hdr) */ 3088 if (hlen <= sizeof(struct tcphdr)) 3089 return (0); 3090 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af)) 3091 return (0); 3092 opt = hdr + sizeof(struct tcphdr); 3093 hlen -= sizeof(struct tcphdr); 3094 while (hlen >= TCPOLEN_MAXSEG) { 3095 switch (*opt) { 3096 case TCPOPT_EOL: 3097 case TCPOPT_NOP: 3098 ++opt; 3099 --hlen; 3100 break; 3101 case TCPOPT_MAXSEG: 3102 bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2); 3103 NTOHS(mss); 3104 /* FALLTHROUGH */ 3105 default: 3106 optlen = opt[1]; 3107 if (optlen < 2) 3108 optlen = 2; 3109 hlen -= optlen; 3110 opt += optlen; 3111 break; 3112 } 3113 } 3114 return (mss); 3115} 3116 3117u_int16_t 3118pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer) 3119{ 3120#ifdef INET 3121 struct sockaddr_in *dst; 3122 struct route ro; 3123#endif /* INET */ 3124#ifdef INET6 3125 struct sockaddr_in6 *dst6; 3126 struct route_in6 ro6; 3127#endif /* INET6 */ 3128 struct rtentry *rt = NULL; 3129 int hlen = 0; /* make the compiler happy */ 3130 u_int16_t mss = tcp_mssdflt; 3131 3132 switch (af) { 3133#ifdef INET 3134 case AF_INET: 3135 hlen = sizeof(struct ip); 3136 bzero(&ro, sizeof(ro)); 3137 dst = (struct sockaddr_in *)&ro.ro_dst; 3138 dst->sin_family = AF_INET; 3139 dst->sin_len = sizeof(*dst); 3140 dst->sin_addr = addr->v4; 3141#ifdef __FreeBSD__ 3142#ifdef RTF_PRCLONING 3143 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING)); 3144#else /* !RTF_PRCLONING */ 3145 rtalloc_ign(&ro, RTF_CLONING); 3146#endif 3147#else /* ! __FreeBSD__ */ 3148 rtalloc_noclone(&ro, NO_CLONING); 3149#endif 3150 rt = ro.ro_rt; 3151 break; 3152#endif /* INET */ 3153#ifdef INET6 3154 case AF_INET6: 3155 hlen = sizeof(struct ip6_hdr); 3156 bzero(&ro6, sizeof(ro6)); 3157 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst; 3158 dst6->sin6_family = AF_INET6; 3159 dst6->sin6_len = sizeof(*dst6); 3160 dst6->sin6_addr = addr->v6; 3161#ifdef __FreeBSD__ 3162#ifdef RTF_PRCLONING 3163 rtalloc_ign((struct route *)&ro6, 3164 (RTF_CLONING | RTF_PRCLONING)); 3165#else /* !RTF_PRCLONING */ 3166 rtalloc_ign((struct route *)&ro6, RTF_CLONING); 3167#endif 3168#else /* ! __FreeBSD__ */ 3169 rtalloc_noclone((struct route *)&ro6, NO_CLONING); 3170#endif 3171 rt = ro6.ro_rt; 3172 break; 3173#endif /* INET6 */ 3174 } 3175 3176 if (rt && rt->rt_ifp) { 3177 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr); 3178 mss = max(tcp_mssdflt, mss); 3179 RTFREE(rt); 3180 } 3181 mss = min(mss, offer); 3182 mss = max(mss, 64); /* sanity - at least max opt space */ 3183 return (mss); 3184} 3185 3186void 3187pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr) 3188{ 3189 struct pf_rule *r = s->rule.ptr; 3190 3191 s->rt_kif = NULL; 3192 if (!r->rt || r->rt == PF_FASTROUTE) 3193 return; 3194 switch (s->af) { 3195#ifdef INET 3196 case AF_INET: 3197 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL, 3198 &s->nat_src_node); 3199 s->rt_kif = r->rpool.cur->kif; 3200 break; 3201#endif /* INET */ 3202#ifdef INET6 3203 case AF_INET6: 3204 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL, 3205 &s->nat_src_node); 3206 s->rt_kif = r->rpool.cur->kif; 3207 break; 3208#endif /* INET6 */ 3209 } 3210} 3211 3212int 3213pf_test_tcp(struct pf_rule **rm, struct pf_state **sm, int direction, 3214 struct pfi_kif *kif, struct mbuf *m, int off, void *h, 3215#ifdef __FreeBSD__ 3216 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3217 struct ifqueue *ifq, struct inpcb *inp) 3218#else 3219 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3220 struct ifqueue *ifq) 3221#endif 3222{ 3223 struct pf_rule *nr = NULL; 3224 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 3225 struct tcphdr *th = pd->hdr.tcp; 3226 u_int16_t bport, nport = 0; 3227 sa_family_t af = pd->af; 3228 struct pf_rule *r, *a = NULL; 3229 struct pf_ruleset *ruleset = NULL; 3230 struct pf_src_node *nsn = NULL; 3231 u_short reason; 3232 int rewrite = 0; 3233 int tag = -1, rtableid = -1; 3234 u_int16_t mss = tcp_mssdflt; 3235 int asd = 0; 3236 int match = 0; 3237 3238 if (pf_check_congestion(ifq)) { 3239 REASON_SET(&reason, PFRES_CONGEST); 3240 return (PF_DROP); 3241 } 3242 3243#ifdef __FreeBSD__ 3244 if (inp != NULL) 3245 pd->lookup.done = pf_socket_lookup(direction, pd, inp); 3246 else if (debug_pfugidhack) { 3247 PF_UNLOCK(); 3248 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n")); 3249 pd->lookup.done = pf_socket_lookup(direction, pd, inp); 3250 PF_LOCK(); 3251 } 3252#endif 3253 3254 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 3255 3256 if (direction == PF_OUT) { 3257 bport = nport = th->th_sport; 3258 /* check outgoing packet for BINAT/NAT */ 3259 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, 3260 saddr, th->th_sport, daddr, th->th_dport, 3261 &pd->naddr, &nport)) != NULL) { 3262 PF_ACPY(&pd->baddr, saddr, af); 3263 pf_change_ap(saddr, &th->th_sport, pd->ip_sum, 3264 &th->th_sum, &pd->naddr, nport, 0, af); 3265 rewrite++; 3266 if (nr->natpass) 3267 r = NULL; 3268 pd->nat_rule = nr; 3269 } 3270 } else { 3271 bport = nport = th->th_dport; 3272 /* check incoming packet for BINAT/RDR */ 3273 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, 3274 saddr, th->th_sport, daddr, th->th_dport, 3275 &pd->naddr, &nport)) != NULL) { 3276 PF_ACPY(&pd->baddr, daddr, af); 3277 pf_change_ap(daddr, &th->th_dport, pd->ip_sum, 3278 &th->th_sum, &pd->naddr, nport, 0, af); 3279 rewrite++; 3280 if (nr->natpass) 3281 r = NULL; 3282 pd->nat_rule = nr; 3283 } 3284 } 3285 3286 while (r != NULL) { 3287 r->evaluations++; 3288 if (pfi_kif_match(r->kif, kif) == r->ifnot) 3289 r = r->skip[PF_SKIP_IFP].ptr; 3290 else if (r->direction && r->direction != direction) 3291 r = r->skip[PF_SKIP_DIR].ptr; 3292 else if (r->af && r->af != af) 3293 r = r->skip[PF_SKIP_AF].ptr; 3294 else if (r->proto && r->proto != IPPROTO_TCP) 3295 r = r->skip[PF_SKIP_PROTO].ptr; 3296 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, 3297 r->src.neg, kif)) 3298 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 3299 else if (r->src.port_op && !pf_match_port(r->src.port_op, 3300 r->src.port[0], r->src.port[1], th->th_sport)) 3301 r = r->skip[PF_SKIP_SRC_PORT].ptr; 3302 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, 3303 r->dst.neg, NULL)) 3304 r = r->skip[PF_SKIP_DST_ADDR].ptr; 3305 else if (r->dst.port_op && !pf_match_port(r->dst.port_op, 3306 r->dst.port[0], r->dst.port[1], th->th_dport)) 3307 r = r->skip[PF_SKIP_DST_PORT].ptr; 3308 else if (r->tos && !(r->tos == pd->tos)) 3309 r = TAILQ_NEXT(r, entries); 3310 else if (r->rule_flag & PFRULE_FRAGMENT) 3311 r = TAILQ_NEXT(r, entries); 3312 else if ((r->flagset & th->th_flags) != r->flags) 3313 r = TAILQ_NEXT(r, entries); 3314 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done = 3315#ifdef __FreeBSD__ 3316 pf_socket_lookup(direction, pd, inp), 1)) && 3317#else 3318 pf_socket_lookup(direction, pd), 1)) && 3319#endif 3320 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1], 3321 pd->lookup.uid)) 3322 r = TAILQ_NEXT(r, entries); 3323 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done = 3324#ifdef __FreeBSD__ 3325 pf_socket_lookup(direction, pd, inp), 1)) && 3326#else 3327 pf_socket_lookup(direction, pd), 1)) && 3328#endif 3329 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1], 3330 pd->lookup.gid)) 3331 r = TAILQ_NEXT(r, entries); 3332 else if (r->prob && r->prob <= arc4random()) 3333 r = TAILQ_NEXT(r, entries); 3334 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) 3335 r = TAILQ_NEXT(r, entries); 3336 else if (r->os_fingerprint != PF_OSFP_ANY && !pf_osfp_match( 3337 pf_osfp_fingerprint(pd, m, off, th), r->os_fingerprint)) 3338 r = TAILQ_NEXT(r, entries); 3339 else { 3340 if (r->tag) 3341 tag = r->tag; 3342 if (r->rtableid >= 0) 3343 rtableid = r->rtableid; 3344 if (r->anchor == NULL) { 3345 match = 1; 3346 *rm = r; 3347 *am = a; 3348 *rsm = ruleset; 3349 if ((*rm)->quick) 3350 break; 3351 r = TAILQ_NEXT(r, entries); 3352 } else 3353 pf_step_into_anchor(&asd, &ruleset, 3354 PF_RULESET_FILTER, &r, &a, &match); 3355 } 3356 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 3357 PF_RULESET_FILTER, &r, &a, &match)) 3358 break; 3359 } 3360 r = *rm; 3361 a = *am; 3362 ruleset = *rsm; 3363 3364 REASON_SET(&reason, PFRES_MATCH); 3365 3366 if (r->log || (nr != NULL && nr->natpass && nr->log)) { 3367 if (rewrite) 3368#ifdef __FreeBSD__ 3369 m_copyback(m, off, sizeof(*th), (caddr_t)th); 3370#else 3371 m_copyback(m, off, sizeof(*th), th); 3372#endif 3373 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, 3374 a, ruleset, pd); 3375 } 3376 3377 if ((r->action == PF_DROP) && 3378 ((r->rule_flag & PFRULE_RETURNRST) || 3379 (r->rule_flag & PFRULE_RETURNICMP) || 3380 (r->rule_flag & PFRULE_RETURN))) { 3381 /* undo NAT changes, if they have taken place */ 3382 if (nr != NULL) { 3383 if (direction == PF_OUT) { 3384 pf_change_ap(saddr, &th->th_sport, pd->ip_sum, 3385 &th->th_sum, &pd->baddr, bport, 0, af); 3386 rewrite++; 3387 } else { 3388 pf_change_ap(daddr, &th->th_dport, pd->ip_sum, 3389 &th->th_sum, &pd->baddr, bport, 0, af); 3390 rewrite++; 3391 } 3392 } 3393 if (((r->rule_flag & PFRULE_RETURNRST) || 3394 (r->rule_flag & PFRULE_RETURN)) && 3395 !(th->th_flags & TH_RST)) { 3396 u_int32_t ack = ntohl(th->th_seq) + pd->p_len; 3397 3398 if (th->th_flags & TH_SYN) 3399 ack++; 3400 if (th->th_flags & TH_FIN) 3401 ack++; 3402#ifdef __FreeBSD__ 3403 pf_send_tcp(m, r, af, pd->dst, 3404#else 3405 pf_send_tcp(r, af, pd->dst, 3406#endif 3407 pd->src, th->th_dport, th->th_sport, 3408 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0, 3409 r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp); 3410 } else if ((af == AF_INET) && r->return_icmp) 3411 pf_send_icmp(m, r->return_icmp >> 8, 3412 r->return_icmp & 255, af, r); 3413 else if ((af == AF_INET6) && r->return_icmp6) 3414 pf_send_icmp(m, r->return_icmp6 >> 8, 3415 r->return_icmp6 & 255, af, r); 3416 } 3417 3418 if (r->action == PF_DROP) 3419 return (PF_DROP); 3420 3421 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) { 3422 REASON_SET(&reason, PFRES_MEMORY); 3423 return (PF_DROP); 3424 } 3425 3426 if (r->keep_state || nr != NULL || 3427 (pd->flags & PFDESC_TCP_NORM)) { 3428 /* create new state */ 3429 u_int16_t len; 3430 struct pf_state *s = NULL; 3431 struct pf_src_node *sn = NULL; 3432 3433 len = pd->tot_len - off - (th->th_off << 2); 3434 3435 /* check maximums */ 3436 if (r->max_states && (r->states >= r->max_states)) { 3437 pf_status.lcounters[LCNT_STATES]++; 3438 REASON_SET(&reason, PFRES_MAXSTATES); 3439 goto cleanup; 3440 } 3441 /* src node for filter rule */ 3442 if ((r->rule_flag & PFRULE_SRCTRACK || 3443 r->rpool.opts & PF_POOL_STICKYADDR) && 3444 pf_insert_src_node(&sn, r, saddr, af) != 0) { 3445 REASON_SET(&reason, PFRES_SRCLIMIT); 3446 goto cleanup; 3447 } 3448 /* src node for translation rule */ 3449 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 3450 ((direction == PF_OUT && 3451 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || 3452 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { 3453 REASON_SET(&reason, PFRES_SRCLIMIT); 3454 goto cleanup; 3455 } 3456 s = pool_get(&pf_state_pl, PR_NOWAIT); 3457 if (s == NULL) { 3458 REASON_SET(&reason, PFRES_MEMORY); 3459cleanup: 3460 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 3461 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 3462 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3463 pf_status.src_nodes--; 3464 pool_put(&pf_src_tree_pl, sn); 3465 } 3466 if (nsn != sn && nsn != NULL && nsn->states == 0 && 3467 nsn->expire == 0) { 3468 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 3469 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3470 pf_status.src_nodes--; 3471 pool_put(&pf_src_tree_pl, nsn); 3472 } 3473 return (PF_DROP); 3474 } 3475 bzero(s, sizeof(*s)); 3476 s->rule.ptr = r; 3477 s->nat_rule.ptr = nr; 3478 s->anchor.ptr = a; 3479 STATE_INC_COUNTERS(s); 3480 s->allow_opts = r->allow_opts; 3481 s->log = r->log & PF_LOG_ALL; 3482 if (nr != NULL) 3483 s->log |= nr->log & PF_LOG_ALL; 3484 s->proto = IPPROTO_TCP; 3485 s->direction = direction; 3486 s->af = af; 3487 if (direction == PF_OUT) { 3488 PF_ACPY(&s->gwy.addr, saddr, af); 3489 s->gwy.port = th->th_sport; /* sport */ 3490 PF_ACPY(&s->ext.addr, daddr, af); 3491 s->ext.port = th->th_dport; 3492 if (nr != NULL) { 3493 PF_ACPY(&s->lan.addr, &pd->baddr, af); 3494 s->lan.port = bport; 3495 } else { 3496 PF_ACPY(&s->lan.addr, &s->gwy.addr, af); 3497 s->lan.port = s->gwy.port; 3498 } 3499 } else { 3500 PF_ACPY(&s->lan.addr, daddr, af); 3501 s->lan.port = th->th_dport; 3502 PF_ACPY(&s->ext.addr, saddr, af); 3503 s->ext.port = th->th_sport; 3504 if (nr != NULL) { 3505 PF_ACPY(&s->gwy.addr, &pd->baddr, af); 3506 s->gwy.port = bport; 3507 } else { 3508 PF_ACPY(&s->gwy.addr, &s->lan.addr, af); 3509 s->gwy.port = s->lan.port; 3510 } 3511 } 3512 3513 s->src.seqlo = ntohl(th->th_seq); 3514 s->src.seqhi = s->src.seqlo + len + 1; 3515 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN && 3516 r->keep_state == PF_STATE_MODULATE) { 3517 /* Generate sequence number modulator */ 3518#ifdef __FreeBSD__ 3519 while ((s->src.seqdiff = 3520 pf_new_isn(s) - s->src.seqlo) == 0) 3521 ; 3522#else 3523 while ((s->src.seqdiff = 3524 tcp_rndiss_next() - s->src.seqlo) == 0) 3525 ; 3526#endif 3527 pf_change_a(&th->th_seq, &th->th_sum, 3528 htonl(s->src.seqlo + s->src.seqdiff), 0); 3529 rewrite = 1; 3530 } else 3531 s->src.seqdiff = 0; 3532 if (th->th_flags & TH_SYN) { 3533 s->src.seqhi++; 3534 s->src.wscale = pf_get_wscale(m, off, th->th_off, af); 3535 } 3536 s->src.max_win = MAX(ntohs(th->th_win), 1); 3537 if (s->src.wscale & PF_WSCALE_MASK) { 3538 /* Remove scale factor from initial window */ 3539 int win = s->src.max_win; 3540 win += 1 << (s->src.wscale & PF_WSCALE_MASK); 3541 s->src.max_win = (win - 1) >> 3542 (s->src.wscale & PF_WSCALE_MASK); 3543 } 3544 if (th->th_flags & TH_FIN) 3545 s->src.seqhi++; 3546 s->dst.seqhi = 1; 3547 s->dst.max_win = 1; 3548 s->src.state = TCPS_SYN_SENT; 3549 s->dst.state = TCPS_CLOSED; 3550 s->creation = time_second; 3551 s->expire = time_second; 3552 s->timeout = PFTM_TCP_FIRST_PACKET; 3553 pf_set_rt_ifp(s, saddr); 3554 if (sn != NULL) { 3555 s->src_node = sn; 3556 s->src_node->states++; 3557 } 3558 if (nsn != NULL) { 3559 PF_ACPY(&nsn->raddr, &pd->naddr, af); 3560 s->nat_src_node = nsn; 3561 s->nat_src_node->states++; 3562 } 3563 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m, 3564 off, pd, th, &s->src, &s->dst)) { 3565 REASON_SET(&reason, PFRES_MEMORY); 3566 pf_src_tree_remove_state(s); 3567 STATE_DEC_COUNTERS(s); 3568 pool_put(&pf_state_pl, s); 3569 return (PF_DROP); 3570 } 3571 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub && 3572 pf_normalize_tcp_stateful(m, off, pd, &reason, th, s, 3573 &s->src, &s->dst, &rewrite)) { 3574 /* This really shouldn't happen!!! */ 3575 DPFPRINTF(PF_DEBUG_URGENT, 3576 ("pf_normalize_tcp_stateful failed on first pkt")); 3577 pf_normalize_tcp_cleanup(s); 3578 pf_src_tree_remove_state(s); 3579 STATE_DEC_COUNTERS(s); 3580 pool_put(&pf_state_pl, s); 3581 return (PF_DROP); 3582 } 3583 if (pf_insert_state(BOUND_IFACE(r, kif), s)) { 3584 pf_normalize_tcp_cleanup(s); 3585 REASON_SET(&reason, PFRES_STATEINS); 3586 pf_src_tree_remove_state(s); 3587 STATE_DEC_COUNTERS(s); 3588 pool_put(&pf_state_pl, s); 3589 return (PF_DROP); 3590 } else 3591 *sm = s; 3592 if (tag > 0) { 3593 pf_tag_ref(tag); 3594 s->tag = tag; 3595 } 3596 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN && 3597 r->keep_state == PF_STATE_SYNPROXY) { 3598 s->src.state = PF_TCPS_PROXY_SRC; 3599 if (nr != NULL) { 3600 if (direction == PF_OUT) { 3601 pf_change_ap(saddr, &th->th_sport, 3602 pd->ip_sum, &th->th_sum, &pd->baddr, 3603 bport, 0, af); 3604 } else { 3605 pf_change_ap(daddr, &th->th_dport, 3606 pd->ip_sum, &th->th_sum, &pd->baddr, 3607 bport, 0, af); 3608 } 3609 } 3610 s->src.seqhi = htonl(arc4random()); 3611 /* Find mss option */ 3612 mss = pf_get_mss(m, off, th->th_off, af); 3613 mss = pf_calc_mss(saddr, af, mss); 3614 mss = pf_calc_mss(daddr, af, mss); 3615 s->src.mss = mss; 3616#ifdef __FreeBSD__ 3617 pf_send_tcp(NULL, r, af, daddr, saddr, th->th_dport, 3618#else 3619 pf_send_tcp(r, af, daddr, saddr, th->th_dport, 3620#endif 3621 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1, 3622 TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL); 3623 REASON_SET(&reason, PFRES_SYNPROXY); 3624 return (PF_SYNPROXY_DROP); 3625 } 3626 } 3627 3628 /* copy back packet headers if we performed NAT operations */ 3629 if (rewrite) 3630 m_copyback(m, off, sizeof(*th), (caddr_t)th); 3631 3632 return (PF_PASS); 3633} 3634 3635int 3636pf_test_udp(struct pf_rule **rm, struct pf_state **sm, int direction, 3637 struct pfi_kif *kif, struct mbuf *m, int off, void *h, 3638#ifdef __FreeBSD__ 3639 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3640 struct ifqueue *ifq, struct inpcb *inp) 3641#else 3642 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3643 struct ifqueue *ifq) 3644#endif 3645{ 3646 struct pf_rule *nr = NULL; 3647 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 3648 struct udphdr *uh = pd->hdr.udp; 3649 u_int16_t bport, nport = 0; 3650 sa_family_t af = pd->af; 3651 struct pf_rule *r, *a = NULL; 3652 struct pf_ruleset *ruleset = NULL; 3653 struct pf_src_node *nsn = NULL; 3654 u_short reason; 3655 int rewrite = 0; 3656 int tag = -1, rtableid = -1; 3657 int asd = 0; 3658 int match = 0; 3659 3660 if (pf_check_congestion(ifq)) { 3661 REASON_SET(&reason, PFRES_CONGEST); 3662 return (PF_DROP); 3663 } 3664 3665#ifdef __FreeBSD__ 3666 if (inp != NULL) 3667 pd->lookup.done = pf_socket_lookup(direction, pd, inp); 3668 else if (debug_pfugidhack) { 3669 PF_UNLOCK(); 3670 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n")); 3671 pd->lookup.done = pf_socket_lookup(direction, pd, inp); 3672 PF_LOCK(); 3673 } 3674#endif 3675 3676 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 3677 3678 if (direction == PF_OUT) { 3679 bport = nport = uh->uh_sport; 3680 /* check outgoing packet for BINAT/NAT */ 3681 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, 3682 saddr, uh->uh_sport, daddr, uh->uh_dport, 3683 &pd->naddr, &nport)) != NULL) { 3684 PF_ACPY(&pd->baddr, saddr, af); 3685 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum, 3686 &uh->uh_sum, &pd->naddr, nport, 1, af); 3687 rewrite++; 3688 if (nr->natpass) 3689 r = NULL; 3690 pd->nat_rule = nr; 3691 } 3692 } else { 3693 bport = nport = uh->uh_dport; 3694 /* check incoming packet for BINAT/RDR */ 3695 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, 3696 saddr, uh->uh_sport, daddr, uh->uh_dport, &pd->naddr, 3697 &nport)) != NULL) { 3698 PF_ACPY(&pd->baddr, daddr, af); 3699 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum, 3700 &uh->uh_sum, &pd->naddr, nport, 1, af); 3701 rewrite++; 3702 if (nr->natpass) 3703 r = NULL; 3704 pd->nat_rule = nr; 3705 } 3706 } 3707 3708 while (r != NULL) { 3709 r->evaluations++; 3710 if (pfi_kif_match(r->kif, kif) == r->ifnot) 3711 r = r->skip[PF_SKIP_IFP].ptr; 3712 else if (r->direction && r->direction != direction) 3713 r = r->skip[PF_SKIP_DIR].ptr; 3714 else if (r->af && r->af != af) 3715 r = r->skip[PF_SKIP_AF].ptr; 3716 else if (r->proto && r->proto != IPPROTO_UDP) 3717 r = r->skip[PF_SKIP_PROTO].ptr; 3718 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, 3719 r->src.neg, kif)) 3720 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 3721 else if (r->src.port_op && !pf_match_port(r->src.port_op, 3722 r->src.port[0], r->src.port[1], uh->uh_sport)) 3723 r = r->skip[PF_SKIP_SRC_PORT].ptr; 3724 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, 3725 r->dst.neg, NULL)) 3726 r = r->skip[PF_SKIP_DST_ADDR].ptr; 3727 else if (r->dst.port_op && !pf_match_port(r->dst.port_op, 3728 r->dst.port[0], r->dst.port[1], uh->uh_dport)) 3729 r = r->skip[PF_SKIP_DST_PORT].ptr; 3730 else if (r->tos && !(r->tos == pd->tos)) 3731 r = TAILQ_NEXT(r, entries); 3732 else if (r->rule_flag & PFRULE_FRAGMENT) 3733 r = TAILQ_NEXT(r, entries); 3734 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done = 3735#ifdef __FreeBSD__ 3736 pf_socket_lookup(direction, pd, inp), 1)) && 3737#else 3738 pf_socket_lookup(direction, pd), 1)) && 3739#endif 3740 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1], 3741 pd->lookup.uid)) 3742 r = TAILQ_NEXT(r, entries); 3743 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done = 3744#ifdef __FreeBSD__ 3745 pf_socket_lookup(direction, pd, inp), 1)) && 3746#else 3747 pf_socket_lookup(direction, pd), 1)) && 3748#endif 3749 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1], 3750 pd->lookup.gid)) 3751 r = TAILQ_NEXT(r, entries); 3752 else if (r->prob && r->prob <= arc4random()) 3753 r = TAILQ_NEXT(r, entries); 3754 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) 3755 r = TAILQ_NEXT(r, entries); 3756 else if (r->os_fingerprint != PF_OSFP_ANY) 3757 r = TAILQ_NEXT(r, entries); 3758 else { 3759 if (r->tag) 3760 tag = r->tag; 3761 if (r->rtableid >= 0) 3762 rtableid = r->rtableid; 3763 if (r->anchor == NULL) { 3764 match = 1; 3765 *rm = r; 3766 *am = a; 3767 *rsm = ruleset; 3768 if ((*rm)->quick) 3769 break; 3770 r = TAILQ_NEXT(r, entries); 3771 } else 3772 pf_step_into_anchor(&asd, &ruleset, 3773 PF_RULESET_FILTER, &r, &a, &match); 3774 } 3775 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 3776 PF_RULESET_FILTER, &r, &a, &match)) 3777 break; 3778 } 3779 r = *rm; 3780 a = *am; 3781 ruleset = *rsm; 3782 3783 REASON_SET(&reason, PFRES_MATCH); 3784 3785 if (r->log || (nr != NULL && nr->natpass && nr->log)) { 3786 if (rewrite) 3787#ifdef __FreeBSD__ 3788 m_copyback(m, off, sizeof(*uh), (caddr_t)uh); 3789#else 3790 m_copyback(m, off, sizeof(*uh), uh); 3791#endif 3792 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, 3793 a, ruleset, pd); 3794 } 3795 3796 if ((r->action == PF_DROP) && 3797 ((r->rule_flag & PFRULE_RETURNICMP) || 3798 (r->rule_flag & PFRULE_RETURN))) { 3799 /* undo NAT changes, if they have taken place */ 3800 if (nr != NULL) { 3801 if (direction == PF_OUT) { 3802 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum, 3803 &uh->uh_sum, &pd->baddr, bport, 1, af); 3804 rewrite++; 3805 } else { 3806 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum, 3807 &uh->uh_sum, &pd->baddr, bport, 1, af); 3808 rewrite++; 3809 } 3810 } 3811 if ((af == AF_INET) && r->return_icmp) 3812 pf_send_icmp(m, r->return_icmp >> 8, 3813 r->return_icmp & 255, af, r); 3814 else if ((af == AF_INET6) && r->return_icmp6) 3815 pf_send_icmp(m, r->return_icmp6 >> 8, 3816 r->return_icmp6 & 255, af, r); 3817 } 3818 3819 if (r->action == PF_DROP) 3820 return (PF_DROP); 3821 3822 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) { 3823 REASON_SET(&reason, PFRES_MEMORY); 3824 return (PF_DROP); 3825 } 3826 3827 if (r->keep_state || nr != NULL) { 3828 /* create new state */ 3829 struct pf_state *s = NULL; 3830 struct pf_src_node *sn = NULL; 3831 3832 /* check maximums */ 3833 if (r->max_states && (r->states >= r->max_states)) { 3834 pf_status.lcounters[LCNT_STATES]++; 3835 REASON_SET(&reason, PFRES_MAXSTATES); 3836 goto cleanup; 3837 } 3838 /* src node for filter rule */ 3839 if ((r->rule_flag & PFRULE_SRCTRACK || 3840 r->rpool.opts & PF_POOL_STICKYADDR) && 3841 pf_insert_src_node(&sn, r, saddr, af) != 0) { 3842 REASON_SET(&reason, PFRES_SRCLIMIT); 3843 goto cleanup; 3844 } 3845 /* src node for translation rule */ 3846 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 3847 ((direction == PF_OUT && 3848 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || 3849 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { 3850 REASON_SET(&reason, PFRES_SRCLIMIT); 3851 goto cleanup; 3852 } 3853 s = pool_get(&pf_state_pl, PR_NOWAIT); 3854 if (s == NULL) { 3855 REASON_SET(&reason, PFRES_MEMORY); 3856cleanup: 3857 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 3858 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 3859 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3860 pf_status.src_nodes--; 3861 pool_put(&pf_src_tree_pl, sn); 3862 } 3863 if (nsn != sn && nsn != NULL && nsn->states == 0 && 3864 nsn->expire == 0) { 3865 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 3866 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3867 pf_status.src_nodes--; 3868 pool_put(&pf_src_tree_pl, nsn); 3869 } 3870 return (PF_DROP); 3871 } 3872 bzero(s, sizeof(*s)); 3873 s->rule.ptr = r; 3874 s->nat_rule.ptr = nr; 3875 s->anchor.ptr = a; 3876 STATE_INC_COUNTERS(s); 3877 s->allow_opts = r->allow_opts; 3878 s->log = r->log & PF_LOG_ALL; 3879 if (nr != NULL) 3880 s->log |= nr->log & PF_LOG_ALL; 3881 s->proto = IPPROTO_UDP; 3882 s->direction = direction; 3883 s->af = af; 3884 if (direction == PF_OUT) { 3885 PF_ACPY(&s->gwy.addr, saddr, af); 3886 s->gwy.port = uh->uh_sport; 3887 PF_ACPY(&s->ext.addr, daddr, af); 3888 s->ext.port = uh->uh_dport; 3889 if (nr != NULL) { 3890 PF_ACPY(&s->lan.addr, &pd->baddr, af); 3891 s->lan.port = bport; 3892 } else { 3893 PF_ACPY(&s->lan.addr, &s->gwy.addr, af); 3894 s->lan.port = s->gwy.port; 3895 } 3896 } else { 3897 PF_ACPY(&s->lan.addr, daddr, af); 3898 s->lan.port = uh->uh_dport; 3899 PF_ACPY(&s->ext.addr, saddr, af); 3900 s->ext.port = uh->uh_sport; 3901 if (nr != NULL) { 3902 PF_ACPY(&s->gwy.addr, &pd->baddr, af); 3903 s->gwy.port = bport; 3904 } else { 3905 PF_ACPY(&s->gwy.addr, &s->lan.addr, af); 3906 s->gwy.port = s->lan.port; 3907 } 3908 } 3909 s->src.state = PFUDPS_SINGLE; 3910 s->dst.state = PFUDPS_NO_TRAFFIC; 3911 s->creation = time_second; 3912 s->expire = time_second; 3913 s->timeout = PFTM_UDP_FIRST_PACKET; 3914 pf_set_rt_ifp(s, saddr); 3915 if (sn != NULL) { 3916 s->src_node = sn; 3917 s->src_node->states++; 3918 } 3919 if (nsn != NULL) { 3920 PF_ACPY(&nsn->raddr, &pd->naddr, af); 3921 s->nat_src_node = nsn; 3922 s->nat_src_node->states++; 3923 } 3924 if (pf_insert_state(BOUND_IFACE(r, kif), s)) { 3925 REASON_SET(&reason, PFRES_STATEINS); 3926 pf_src_tree_remove_state(s); 3927 STATE_DEC_COUNTERS(s); 3928 pool_put(&pf_state_pl, s); 3929 return (PF_DROP); 3930 } else 3931 *sm = s; 3932 if (tag > 0) { 3933 pf_tag_ref(tag); 3934 s->tag = tag; 3935 } 3936 } 3937 3938 /* copy back packet headers if we performed NAT operations */ 3939 if (rewrite) 3940 m_copyback(m, off, sizeof(*uh), (caddr_t)uh); 3941 3942 return (PF_PASS); 3943} 3944 3945int 3946pf_test_icmp(struct pf_rule **rm, struct pf_state **sm, int direction, 3947 struct pfi_kif *kif, struct mbuf *m, int off, void *h, 3948 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3949 struct ifqueue *ifq) 3950{ 3951 struct pf_rule *nr = NULL; 3952 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 3953 struct pf_rule *r, *a = NULL; 3954 struct pf_ruleset *ruleset = NULL; 3955 struct pf_src_node *nsn = NULL; 3956 u_short reason; 3957 u_int16_t icmpid = 0, bport, nport = 0; 3958 sa_family_t af = pd->af; 3959 u_int8_t icmptype = 0; /* make the compiler happy */ 3960 u_int8_t icmpcode = 0; /* make the compiler happy */ 3961 int state_icmp = 0; 3962 int tag = -1, rtableid = -1; 3963#ifdef INET6 3964 int rewrite = 0; 3965#endif /* INET6 */ 3966 int asd = 0; 3967 int match = 0; 3968 3969 if (pf_check_congestion(ifq)) { 3970 REASON_SET(&reason, PFRES_CONGEST); 3971 return (PF_DROP); 3972 } 3973 3974 switch (pd->proto) { 3975#ifdef INET 3976 case IPPROTO_ICMP: 3977 icmptype = pd->hdr.icmp->icmp_type; 3978 icmpcode = pd->hdr.icmp->icmp_code; 3979 icmpid = pd->hdr.icmp->icmp_id; 3980 3981 if (icmptype == ICMP_UNREACH || 3982 icmptype == ICMP_SOURCEQUENCH || 3983 icmptype == ICMP_REDIRECT || 3984 icmptype == ICMP_TIMXCEED || 3985 icmptype == ICMP_PARAMPROB) 3986 state_icmp++; 3987 break; 3988#endif /* INET */ 3989#ifdef INET6 3990 case IPPROTO_ICMPV6: 3991 icmptype = pd->hdr.icmp6->icmp6_type; 3992 icmpcode = pd->hdr.icmp6->icmp6_code; 3993 icmpid = pd->hdr.icmp6->icmp6_id; 3994 3995 if (icmptype == ICMP6_DST_UNREACH || 3996 icmptype == ICMP6_PACKET_TOO_BIG || 3997 icmptype == ICMP6_TIME_EXCEEDED || 3998 icmptype == ICMP6_PARAM_PROB) 3999 state_icmp++; 4000 break; 4001#endif /* INET6 */ 4002 } 4003 4004 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 4005 4006 if (direction == PF_OUT) { 4007 bport = nport = icmpid; 4008 /* check outgoing packet for BINAT/NAT */ 4009 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, 4010 saddr, icmpid, daddr, icmpid, &pd->naddr, &nport)) != 4011 NULL) { 4012 PF_ACPY(&pd->baddr, saddr, af); 4013 switch (af) { 4014#ifdef INET 4015 case AF_INET: 4016 pf_change_a(&saddr->v4.s_addr, pd->ip_sum, 4017 pd->naddr.v4.s_addr, 0); 4018 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup( 4019 pd->hdr.icmp->icmp_cksum, icmpid, nport, 0); 4020 pd->hdr.icmp->icmp_id = nport; 4021 m_copyback(m, off, ICMP_MINLEN, 4022 (caddr_t)pd->hdr.icmp); 4023 break; 4024#endif /* INET */ 4025#ifdef INET6 4026 case AF_INET6: 4027 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum, 4028 &pd->naddr, 0); 4029 rewrite++; 4030 break; 4031#endif /* INET6 */ 4032 } 4033 if (nr->natpass) 4034 r = NULL; 4035 pd->nat_rule = nr; 4036 } 4037 } else { 4038 bport = nport = icmpid; 4039 /* check incoming packet for BINAT/RDR */ 4040 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, 4041 saddr, icmpid, daddr, icmpid, &pd->naddr, &nport)) != 4042 NULL) { 4043 PF_ACPY(&pd->baddr, daddr, af); 4044 switch (af) { 4045#ifdef INET 4046 case AF_INET: 4047 pf_change_a(&daddr->v4.s_addr, 4048 pd->ip_sum, pd->naddr.v4.s_addr, 0); 4049 break; 4050#endif /* INET */ 4051#ifdef INET6 4052 case AF_INET6: 4053 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum, 4054 &pd->naddr, 0); 4055 rewrite++; 4056 break; 4057#endif /* INET6 */ 4058 } 4059 if (nr->natpass) 4060 r = NULL; 4061 pd->nat_rule = nr; 4062 } 4063 } 4064 4065 while (r != NULL) { 4066 r->evaluations++; 4067 if (pfi_kif_match(r->kif, kif) == r->ifnot) 4068 r = r->skip[PF_SKIP_IFP].ptr; 4069 else if (r->direction && r->direction != direction) 4070 r = r->skip[PF_SKIP_DIR].ptr; 4071 else if (r->af && r->af != af) 4072 r = r->skip[PF_SKIP_AF].ptr; 4073 else if (r->proto && r->proto != pd->proto) 4074 r = r->skip[PF_SKIP_PROTO].ptr; 4075 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, 4076 r->src.neg, kif)) 4077 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 4078 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, 4079 r->dst.neg, NULL)) 4080 r = r->skip[PF_SKIP_DST_ADDR].ptr; 4081 else if (r->type && r->type != icmptype + 1) 4082 r = TAILQ_NEXT(r, entries); 4083 else if (r->code && r->code != icmpcode + 1) 4084 r = TAILQ_NEXT(r, entries); 4085 else if (r->tos && !(r->tos == pd->tos)) 4086 r = TAILQ_NEXT(r, entries); 4087 else if (r->rule_flag & PFRULE_FRAGMENT) 4088 r = TAILQ_NEXT(r, entries); 4089 else if (r->prob && r->prob <= arc4random()) 4090 r = TAILQ_NEXT(r, entries); 4091 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) 4092 r = TAILQ_NEXT(r, entries); 4093 else if (r->os_fingerprint != PF_OSFP_ANY) 4094 r = TAILQ_NEXT(r, entries); 4095 else { 4096 if (r->tag) 4097 tag = r->tag; 4098 if (r->rtableid >= 0) 4099 rtableid = r->rtableid; 4100 if (r->anchor == NULL) { 4101 match = 1; 4102 *rm = r; 4103 *am = a; 4104 *rsm = ruleset; 4105 if ((*rm)->quick) 4106 break; 4107 r = TAILQ_NEXT(r, entries); 4108 } else 4109 pf_step_into_anchor(&asd, &ruleset, 4110 PF_RULESET_FILTER, &r, &a, &match); 4111 } 4112 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 4113 PF_RULESET_FILTER, &r, &a, &match)) 4114 break; 4115 } 4116 r = *rm; 4117 a = *am; 4118 ruleset = *rsm; 4119 4120 REASON_SET(&reason, PFRES_MATCH); 4121 4122 if (r->log || (nr != NULL && nr->natpass && nr->log)) { 4123#ifdef INET6 4124 if (rewrite) 4125 m_copyback(m, off, sizeof(struct icmp6_hdr), 4126 (caddr_t)pd->hdr.icmp6); 4127#endif /* INET6 */ 4128 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, 4129 a, ruleset, pd); 4130 } 4131 4132 if (r->action != PF_PASS) 4133 return (PF_DROP); 4134 4135 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) { 4136 REASON_SET(&reason, PFRES_MEMORY); 4137 return (PF_DROP); 4138 } 4139 4140 if (!state_icmp && (r->keep_state || nr != NULL)) { 4141 /* create new state */ 4142 struct pf_state *s = NULL; 4143 struct pf_src_node *sn = NULL; 4144 4145 /* check maximums */ 4146 if (r->max_states && (r->states >= r->max_states)) { 4147 pf_status.lcounters[LCNT_STATES]++; 4148 REASON_SET(&reason, PFRES_MAXSTATES); 4149 goto cleanup; 4150 } 4151 /* src node for filter rule */ 4152 if ((r->rule_flag & PFRULE_SRCTRACK || 4153 r->rpool.opts & PF_POOL_STICKYADDR) && 4154 pf_insert_src_node(&sn, r, saddr, af) != 0) { 4155 REASON_SET(&reason, PFRES_SRCLIMIT); 4156 goto cleanup; 4157 } 4158 /* src node for translation rule */ 4159 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 4160 ((direction == PF_OUT && 4161 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || 4162 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { 4163 REASON_SET(&reason, PFRES_SRCLIMIT); 4164 goto cleanup; 4165 } 4166 s = pool_get(&pf_state_pl, PR_NOWAIT); 4167 if (s == NULL) { 4168 REASON_SET(&reason, PFRES_MEMORY); 4169cleanup: 4170 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 4171 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 4172 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 4173 pf_status.src_nodes--; 4174 pool_put(&pf_src_tree_pl, sn); 4175 } 4176 if (nsn != sn && nsn != NULL && nsn->states == 0 && 4177 nsn->expire == 0) { 4178 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 4179 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 4180 pf_status.src_nodes--; 4181 pool_put(&pf_src_tree_pl, nsn); 4182 } 4183 return (PF_DROP); 4184 } 4185 bzero(s, sizeof(*s)); 4186 s->rule.ptr = r; 4187 s->nat_rule.ptr = nr; 4188 s->anchor.ptr = a; 4189 STATE_INC_COUNTERS(s); 4190 s->allow_opts = r->allow_opts; 4191 s->log = r->log & PF_LOG_ALL; 4192 if (nr != NULL) 4193 s->log |= nr->log & PF_LOG_ALL; 4194 s->proto = pd->proto; 4195 s->direction = direction; 4196 s->af = af; 4197 if (direction == PF_OUT) { 4198 PF_ACPY(&s->gwy.addr, saddr, af); 4199 s->gwy.port = nport; 4200 PF_ACPY(&s->ext.addr, daddr, af); 4201 s->ext.port = 0; 4202 if (nr != NULL) { 4203 PF_ACPY(&s->lan.addr, &pd->baddr, af); 4204 s->lan.port = bport; 4205 } else { 4206 PF_ACPY(&s->lan.addr, &s->gwy.addr, af); 4207 s->lan.port = s->gwy.port; 4208 } 4209 } else { 4210 PF_ACPY(&s->lan.addr, daddr, af); 4211 s->lan.port = nport; 4212 PF_ACPY(&s->ext.addr, saddr, af); 4213 s->ext.port = 0; 4214 if (nr != NULL) { 4215 PF_ACPY(&s->gwy.addr, &pd->baddr, af); 4216 s->gwy.port = bport; 4217 } else { 4218 PF_ACPY(&s->gwy.addr, &s->lan.addr, af); 4219 s->gwy.port = s->lan.port; 4220 } 4221 } 4222 s->creation = time_second; 4223 s->expire = time_second; 4224 s->timeout = PFTM_ICMP_FIRST_PACKET; 4225 pf_set_rt_ifp(s, saddr); 4226 if (sn != NULL) { 4227 s->src_node = sn; 4228 s->src_node->states++; 4229 } 4230 if (nsn != NULL) { 4231 PF_ACPY(&nsn->raddr, &pd->naddr, af); 4232 s->nat_src_node = nsn; 4233 s->nat_src_node->states++; 4234 } 4235 if (pf_insert_state(BOUND_IFACE(r, kif), s)) { 4236 REASON_SET(&reason, PFRES_STATEINS); 4237 pf_src_tree_remove_state(s); 4238 STATE_DEC_COUNTERS(s); 4239 pool_put(&pf_state_pl, s); 4240 return (PF_DROP); 4241 } else 4242 *sm = s; 4243 if (tag > 0) { 4244 pf_tag_ref(tag); 4245 s->tag = tag; 4246 } 4247 } 4248 4249#ifdef INET6 4250 /* copy back packet headers if we performed IPv6 NAT operations */ 4251 if (rewrite) 4252 m_copyback(m, off, sizeof(struct icmp6_hdr), 4253 (caddr_t)pd->hdr.icmp6); 4254#endif /* INET6 */ 4255 4256 return (PF_PASS); 4257} 4258 4259int 4260pf_test_other(struct pf_rule **rm, struct pf_state **sm, int direction, 4261 struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd, 4262 struct pf_rule **am, struct pf_ruleset **rsm, struct ifqueue *ifq) 4263{ 4264 struct pf_rule *nr = NULL; 4265 struct pf_rule *r, *a = NULL; 4266 struct pf_ruleset *ruleset = NULL; 4267 struct pf_src_node *nsn = NULL; 4268 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 4269 sa_family_t af = pd->af; 4270 u_short reason; 4271 int tag = -1, rtableid = -1; 4272 int asd = 0; 4273 int match = 0; 4274 4275 if (pf_check_congestion(ifq)) { 4276 REASON_SET(&reason, PFRES_CONGEST); 4277 return (PF_DROP); 4278 } 4279 4280 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 4281 4282 if (direction == PF_OUT) { 4283 /* check outgoing packet for BINAT/NAT */ 4284 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, 4285 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) { 4286 PF_ACPY(&pd->baddr, saddr, af); 4287 switch (af) { 4288#ifdef INET 4289 case AF_INET: 4290 pf_change_a(&saddr->v4.s_addr, pd->ip_sum, 4291 pd->naddr.v4.s_addr, 0); 4292 break; 4293#endif /* INET */ 4294#ifdef INET6 4295 case AF_INET6: 4296 PF_ACPY(saddr, &pd->naddr, af); 4297 break; 4298#endif /* INET6 */ 4299 } 4300 if (nr->natpass) 4301 r = NULL; 4302 pd->nat_rule = nr; 4303 } 4304 } else { 4305 /* check incoming packet for BINAT/RDR */ 4306 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, 4307 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) { 4308 PF_ACPY(&pd->baddr, daddr, af); 4309 switch (af) { 4310#ifdef INET 4311 case AF_INET: 4312 pf_change_a(&daddr->v4.s_addr, 4313 pd->ip_sum, pd->naddr.v4.s_addr, 0); 4314 break; 4315#endif /* INET */ 4316#ifdef INET6 4317 case AF_INET6: 4318 PF_ACPY(daddr, &pd->naddr, af); 4319 break; 4320#endif /* INET6 */ 4321 } 4322 if (nr->natpass) 4323 r = NULL; 4324 pd->nat_rule = nr; 4325 } 4326 } 4327 4328 while (r != NULL) { 4329 r->evaluations++; 4330 if (pfi_kif_match(r->kif, kif) == r->ifnot) 4331 r = r->skip[PF_SKIP_IFP].ptr; 4332 else if (r->direction && r->direction != direction) 4333 r = r->skip[PF_SKIP_DIR].ptr; 4334 else if (r->af && r->af != af) 4335 r = r->skip[PF_SKIP_AF].ptr; 4336 else if (r->proto && r->proto != pd->proto) 4337 r = r->skip[PF_SKIP_PROTO].ptr; 4338 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, 4339 r->src.neg, kif)) 4340 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 4341 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, 4342 r->dst.neg, NULL)) 4343 r = r->skip[PF_SKIP_DST_ADDR].ptr; 4344 else if (r->tos && !(r->tos == pd->tos)) 4345 r = TAILQ_NEXT(r, entries); 4346 else if (r->rule_flag & PFRULE_FRAGMENT) 4347 r = TAILQ_NEXT(r, entries); 4348 else if (r->prob && r->prob <= arc4random()) 4349 r = TAILQ_NEXT(r, entries); 4350 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) 4351 r = TAILQ_NEXT(r, entries); 4352 else if (r->os_fingerprint != PF_OSFP_ANY) 4353 r = TAILQ_NEXT(r, entries); 4354 else { 4355 if (r->tag) 4356 tag = r->tag; 4357 if (r->rtableid >= 0) 4358 rtableid = r->rtableid; 4359 if (r->anchor == NULL) { 4360 match = 1; 4361 *rm = r; 4362 *am = a; 4363 *rsm = ruleset; 4364 if ((*rm)->quick) 4365 break; 4366 r = TAILQ_NEXT(r, entries); 4367 } else 4368 pf_step_into_anchor(&asd, &ruleset, 4369 PF_RULESET_FILTER, &r, &a, &match); 4370 } 4371 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 4372 PF_RULESET_FILTER, &r, &a, &match)) 4373 break; 4374 } 4375 r = *rm; 4376 a = *am; 4377 ruleset = *rsm; 4378 4379 REASON_SET(&reason, PFRES_MATCH); 4380 4381 if (r->log || (nr != NULL && nr->natpass && nr->log)) 4382 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, 4383 a, ruleset, pd); 4384 4385 if ((r->action == PF_DROP) && 4386 ((r->rule_flag & PFRULE_RETURNICMP) || 4387 (r->rule_flag & PFRULE_RETURN))) { 4388 struct pf_addr *a = NULL; 4389 4390 if (nr != NULL) { 4391 if (direction == PF_OUT) 4392 a = saddr; 4393 else 4394 a = daddr; 4395 } 4396 if (a != NULL) { 4397 switch (af) { 4398#ifdef INET 4399 case AF_INET: 4400 pf_change_a(&a->v4.s_addr, pd->ip_sum, 4401 pd->baddr.v4.s_addr, 0); 4402 break; 4403#endif /* INET */ 4404#ifdef INET6 4405 case AF_INET6: 4406 PF_ACPY(a, &pd->baddr, af); 4407 break; 4408#endif /* INET6 */ 4409 } 4410 } 4411 if ((af == AF_INET) && r->return_icmp) 4412 pf_send_icmp(m, r->return_icmp >> 8, 4413 r->return_icmp & 255, af, r); 4414 else if ((af == AF_INET6) && r->return_icmp6) 4415 pf_send_icmp(m, r->return_icmp6 >> 8, 4416 r->return_icmp6 & 255, af, r); 4417 } 4418 4419 if (r->action != PF_PASS) 4420 return (PF_DROP); 4421 4422 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) { 4423 REASON_SET(&reason, PFRES_MEMORY); 4424 return (PF_DROP); 4425 } 4426 4427 if (r->keep_state || nr != NULL) { 4428 /* create new state */ 4429 struct pf_state *s = NULL; 4430 struct pf_src_node *sn = NULL; 4431 4432 /* check maximums */ 4433 if (r->max_states && (r->states >= r->max_states)) { 4434 pf_status.lcounters[LCNT_STATES]++; 4435 REASON_SET(&reason, PFRES_MAXSTATES); 4436 goto cleanup; 4437 } 4438 /* src node for filter rule */ 4439 if ((r->rule_flag & PFRULE_SRCTRACK || 4440 r->rpool.opts & PF_POOL_STICKYADDR) && 4441 pf_insert_src_node(&sn, r, saddr, af) != 0) { 4442 REASON_SET(&reason, PFRES_SRCLIMIT); 4443 goto cleanup; 4444 } 4445 /* src node for translation rule */ 4446 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 4447 ((direction == PF_OUT && 4448 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || 4449 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { 4450 REASON_SET(&reason, PFRES_SRCLIMIT); 4451 goto cleanup; 4452 } 4453 s = pool_get(&pf_state_pl, PR_NOWAIT); 4454 if (s == NULL) { 4455 REASON_SET(&reason, PFRES_MEMORY); 4456cleanup: 4457 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 4458 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 4459 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 4460 pf_status.src_nodes--; 4461 pool_put(&pf_src_tree_pl, sn); 4462 } 4463 if (nsn != sn && nsn != NULL && nsn->states == 0 && 4464 nsn->expire == 0) { 4465 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 4466 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 4467 pf_status.src_nodes--; 4468 pool_put(&pf_src_tree_pl, nsn); 4469 } 4470 return (PF_DROP); 4471 } 4472 bzero(s, sizeof(*s)); 4473 s->rule.ptr = r; 4474 s->nat_rule.ptr = nr; 4475 s->anchor.ptr = a; 4476 STATE_INC_COUNTERS(s); 4477 s->allow_opts = r->allow_opts; 4478 s->log = r->log & PF_LOG_ALL; 4479 if (nr != NULL) 4480 s->log |= nr->log & PF_LOG_ALL; 4481 s->proto = pd->proto; 4482 s->direction = direction; 4483 s->af = af; 4484 if (direction == PF_OUT) { 4485 PF_ACPY(&s->gwy.addr, saddr, af); 4486 PF_ACPY(&s->ext.addr, daddr, af); 4487 if (nr != NULL) 4488 PF_ACPY(&s->lan.addr, &pd->baddr, af); 4489 else 4490 PF_ACPY(&s->lan.addr, &s->gwy.addr, af); 4491 } else { 4492 PF_ACPY(&s->lan.addr, daddr, af); 4493 PF_ACPY(&s->ext.addr, saddr, af); 4494 if (nr != NULL) 4495 PF_ACPY(&s->gwy.addr, &pd->baddr, af); 4496 else 4497 PF_ACPY(&s->gwy.addr, &s->lan.addr, af); 4498 } 4499 s->src.state = PFOTHERS_SINGLE; 4500 s->dst.state = PFOTHERS_NO_TRAFFIC; 4501 s->creation = time_second; 4502 s->expire = time_second; 4503 s->timeout = PFTM_OTHER_FIRST_PACKET; 4504 pf_set_rt_ifp(s, saddr); 4505 if (sn != NULL) { 4506 s->src_node = sn; 4507 s->src_node->states++; 4508 } 4509 if (nsn != NULL) { 4510 PF_ACPY(&nsn->raddr, &pd->naddr, af); 4511 s->nat_src_node = nsn; 4512 s->nat_src_node->states++; 4513 } 4514 if (pf_insert_state(BOUND_IFACE(r, kif), s)) { 4515 REASON_SET(&reason, PFRES_STATEINS); 4516 pf_src_tree_remove_state(s); 4517 STATE_DEC_COUNTERS(s); 4518 pool_put(&pf_state_pl, s); 4519 return (PF_DROP); 4520 } else 4521 *sm = s; 4522 if (tag > 0) { 4523 pf_tag_ref(tag); 4524 s->tag = tag; 4525 } 4526 } 4527 4528 return (PF_PASS); 4529} 4530 4531int 4532pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif, 4533 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am, 4534 struct pf_ruleset **rsm) 4535{ 4536 struct pf_rule *r, *a = NULL; 4537 struct pf_ruleset *ruleset = NULL; 4538 sa_family_t af = pd->af; 4539 u_short reason; 4540 int tag = -1; 4541 int asd = 0; 4542 int match = 0; 4543 4544 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 4545 while (r != NULL) { 4546 r->evaluations++; 4547 if (pfi_kif_match(r->kif, kif) == r->ifnot) 4548 r = r->skip[PF_SKIP_IFP].ptr; 4549 else if (r->direction && r->direction != direction) 4550 r = r->skip[PF_SKIP_DIR].ptr; 4551 else if (r->af && r->af != af) 4552 r = r->skip[PF_SKIP_AF].ptr; 4553 else if (r->proto && r->proto != pd->proto) 4554 r = r->skip[PF_SKIP_PROTO].ptr; 4555 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, 4556 r->src.neg, kif)) 4557 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 4558 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, 4559 r->dst.neg, NULL)) 4560 r = r->skip[PF_SKIP_DST_ADDR].ptr; 4561 else if (r->tos && !(r->tos == pd->tos)) 4562 r = TAILQ_NEXT(r, entries); 4563 else if (r->src.port_op || r->dst.port_op || 4564 r->flagset || r->type || r->code || 4565 r->os_fingerprint != PF_OSFP_ANY) 4566 r = TAILQ_NEXT(r, entries); 4567 else if (r->prob && r->prob <= arc4random()) 4568 r = TAILQ_NEXT(r, entries); 4569 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) 4570 r = TAILQ_NEXT(r, entries); 4571 else { 4572 if (r->anchor == NULL) { 4573 match = 1; 4574 *rm = r; 4575 *am = a; 4576 *rsm = ruleset; 4577 if ((*rm)->quick) 4578 break; 4579 r = TAILQ_NEXT(r, entries); 4580 } else 4581 pf_step_into_anchor(&asd, &ruleset, 4582 PF_RULESET_FILTER, &r, &a, &match); 4583 } 4584 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 4585 PF_RULESET_FILTER, &r, &a, &match)) 4586 break; 4587 } 4588 r = *rm; 4589 a = *am; 4590 ruleset = *rsm; 4591 4592 REASON_SET(&reason, PFRES_MATCH); 4593 4594 if (r->log) 4595 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset, 4596 pd); 4597 4598 if (r->action != PF_PASS) 4599 return (PF_DROP); 4600 4601 if (pf_tag_packet(m, pd->pf_mtag, tag, -1)) { 4602 REASON_SET(&reason, PFRES_MEMORY); 4603 return (PF_DROP); 4604 } 4605 4606 return (PF_PASS); 4607} 4608 4609int 4610pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif, 4611 struct mbuf *m, int off, void *h, struct pf_pdesc *pd, 4612 u_short *reason) 4613{ 4614 struct pf_state_cmp key; 4615 struct tcphdr *th = pd->hdr.tcp; 4616 u_int16_t win = ntohs(th->th_win); 4617 u_int32_t ack, end, seq, orig_seq; 4618 u_int8_t sws, dws; 4619 int ackskew; 4620 int copyback = 0; 4621 struct pf_state_peer *src, *dst; 4622 4623 key.af = pd->af; 4624 key.proto = IPPROTO_TCP; 4625 if (direction == PF_IN) { 4626 PF_ACPY(&key.ext.addr, pd->src, key.af); 4627 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 4628 key.ext.port = th->th_sport; 4629 key.gwy.port = th->th_dport; 4630 } else { 4631 PF_ACPY(&key.lan.addr, pd->src, key.af); 4632 PF_ACPY(&key.ext.addr, pd->dst, key.af); 4633 key.lan.port = th->th_sport; 4634 key.ext.port = th->th_dport; 4635 } 4636 4637 STATE_LOOKUP(); 4638 4639 if (direction == (*state)->direction) { 4640 src = &(*state)->src; 4641 dst = &(*state)->dst; 4642 } else { 4643 src = &(*state)->dst; 4644 dst = &(*state)->src; 4645 } 4646 4647 if ((*state)->src.state == PF_TCPS_PROXY_SRC) { 4648 if (direction != (*state)->direction) { 4649 REASON_SET(reason, PFRES_SYNPROXY); 4650 return (PF_SYNPROXY_DROP); 4651 } 4652 if (th->th_flags & TH_SYN) { 4653 if (ntohl(th->th_seq) != (*state)->src.seqlo) { 4654 REASON_SET(reason, PFRES_SYNPROXY); 4655 return (PF_DROP); 4656 } 4657#ifdef __FreeBSD__ 4658 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst, 4659#else 4660 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, 4661#endif 4662 pd->src, th->th_dport, th->th_sport, 4663 (*state)->src.seqhi, ntohl(th->th_seq) + 1, 4664 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1, 4665 0, NULL, NULL); 4666 REASON_SET(reason, PFRES_SYNPROXY); 4667 return (PF_SYNPROXY_DROP); 4668 } else if (!(th->th_flags & TH_ACK) || 4669 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) || 4670 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) { 4671 REASON_SET(reason, PFRES_SYNPROXY); 4672 return (PF_DROP); 4673 } else if ((*state)->src_node != NULL && 4674 pf_src_connlimit(state)) { 4675 REASON_SET(reason, PFRES_SRCLIMIT); 4676 return (PF_DROP); 4677 } else 4678 (*state)->src.state = PF_TCPS_PROXY_DST; 4679 } 4680 if ((*state)->src.state == PF_TCPS_PROXY_DST) { 4681 struct pf_state_host *src, *dst; 4682 4683 if (direction == PF_OUT) { 4684 src = &(*state)->gwy; 4685 dst = &(*state)->ext; 4686 } else { 4687 src = &(*state)->ext; 4688 dst = &(*state)->lan; 4689 } 4690 if (direction == (*state)->direction) { 4691 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) || 4692 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) || 4693 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) { 4694 REASON_SET(reason, PFRES_SYNPROXY); 4695 return (PF_DROP); 4696 } 4697 (*state)->src.max_win = MAX(ntohs(th->th_win), 1); 4698 if ((*state)->dst.seqhi == 1) 4699 (*state)->dst.seqhi = htonl(arc4random()); 4700#ifdef __FreeBSD__ 4701 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, 4702 &src->addr, 4703#else 4704 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr, 4705#endif 4706 &dst->addr, src->port, dst->port, 4707 (*state)->dst.seqhi, 0, TH_SYN, 0, 4708 (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL); 4709 REASON_SET(reason, PFRES_SYNPROXY); 4710 return (PF_SYNPROXY_DROP); 4711 } else if (((th->th_flags & (TH_SYN|TH_ACK)) != 4712 (TH_SYN|TH_ACK)) || 4713 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) { 4714 REASON_SET(reason, PFRES_SYNPROXY); 4715 return (PF_DROP); 4716 } else { 4717 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1); 4718 (*state)->dst.seqlo = ntohl(th->th_seq); 4719#ifdef __FreeBSD__ 4720 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst, 4721#else 4722 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, 4723#endif 4724 pd->src, th->th_dport, th->th_sport, 4725 ntohl(th->th_ack), ntohl(th->th_seq) + 1, 4726 TH_ACK, (*state)->src.max_win, 0, 0, 0, 4727 (*state)->tag, NULL, NULL); 4728#ifdef __FreeBSD__ 4729 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, 4730 &src->addr, 4731#else 4732 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr, 4733#endif 4734 &dst->addr, src->port, dst->port, 4735 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1, 4736 TH_ACK, (*state)->dst.max_win, 0, 0, 1, 4737 0, NULL, NULL); 4738 (*state)->src.seqdiff = (*state)->dst.seqhi - 4739 (*state)->src.seqlo; 4740 (*state)->dst.seqdiff = (*state)->src.seqhi - 4741 (*state)->dst.seqlo; 4742 (*state)->src.seqhi = (*state)->src.seqlo + 4743 (*state)->dst.max_win; 4744 (*state)->dst.seqhi = (*state)->dst.seqlo + 4745 (*state)->src.max_win; 4746 (*state)->src.wscale = (*state)->dst.wscale = 0; 4747 (*state)->src.state = (*state)->dst.state = 4748 TCPS_ESTABLISHED; 4749 REASON_SET(reason, PFRES_SYNPROXY); 4750 return (PF_SYNPROXY_DROP); 4751 } 4752 } 4753 4754 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) { 4755 sws = src->wscale & PF_WSCALE_MASK; 4756 dws = dst->wscale & PF_WSCALE_MASK; 4757 } else 4758 sws = dws = 0; 4759 4760 /* 4761 * Sequence tracking algorithm from Guido van Rooij's paper: 4762 * http://www.madison-gurkha.com/publications/tcp_filtering/ 4763 * tcp_filtering.ps 4764 */ 4765 4766 orig_seq = seq = ntohl(th->th_seq); 4767 if (src->seqlo == 0) { 4768 /* First packet from this end. Set its state */ 4769 4770 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) && 4771 src->scrub == NULL) { 4772 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) { 4773 REASON_SET(reason, PFRES_MEMORY); 4774 return (PF_DROP); 4775 } 4776 } 4777 4778 /* Deferred generation of sequence number modulator */ 4779 if (dst->seqdiff && !src->seqdiff) { 4780#ifdef __FreeBSD__ 4781 while ((src->seqdiff = pf_new_isn(*state) - seq) == 0) 4782 ; 4783#else 4784 while ((src->seqdiff = tcp_rndiss_next() - seq) == 0) 4785 ; 4786#endif 4787 ack = ntohl(th->th_ack) - dst->seqdiff; 4788 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq + 4789 src->seqdiff), 0); 4790 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0); 4791 copyback = 1; 4792 } else { 4793 ack = ntohl(th->th_ack); 4794 } 4795 4796 end = seq + pd->p_len; 4797 if (th->th_flags & TH_SYN) { 4798 end++; 4799 if (dst->wscale & PF_WSCALE_FLAG) { 4800 src->wscale = pf_get_wscale(m, off, th->th_off, 4801 pd->af); 4802 if (src->wscale & PF_WSCALE_FLAG) { 4803 /* Remove scale factor from initial 4804 * window */ 4805 sws = src->wscale & PF_WSCALE_MASK; 4806 win = ((u_int32_t)win + (1 << sws) - 1) 4807 >> sws; 4808 dws = dst->wscale & PF_WSCALE_MASK; 4809 } else { 4810 /* fixup other window */ 4811 dst->max_win <<= dst->wscale & 4812 PF_WSCALE_MASK; 4813 /* in case of a retrans SYN|ACK */ 4814 dst->wscale = 0; 4815 } 4816 } 4817 } 4818 if (th->th_flags & TH_FIN) 4819 end++; 4820 4821 src->seqlo = seq; 4822 if (src->state < TCPS_SYN_SENT) 4823 src->state = TCPS_SYN_SENT; 4824 4825 /* 4826 * May need to slide the window (seqhi may have been set by 4827 * the crappy stack check or if we picked up the connection 4828 * after establishment) 4829 */ 4830 if (src->seqhi == 1 || 4831 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi)) 4832 src->seqhi = end + MAX(1, dst->max_win << dws); 4833 if (win > src->max_win) 4834 src->max_win = win; 4835 4836 } else { 4837 ack = ntohl(th->th_ack) - dst->seqdiff; 4838 if (src->seqdiff) { 4839 /* Modulate sequence numbers */ 4840 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq + 4841 src->seqdiff), 0); 4842 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0); 4843 copyback = 1; 4844 } 4845 end = seq + pd->p_len; 4846 if (th->th_flags & TH_SYN) 4847 end++; 4848 if (th->th_flags & TH_FIN) 4849 end++; 4850 } 4851 4852 if ((th->th_flags & TH_ACK) == 0) { 4853 /* Let it pass through the ack skew check */ 4854 ack = dst->seqlo; 4855 } else if ((ack == 0 && 4856 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) || 4857 /* broken tcp stacks do not set ack */ 4858 (dst->state < TCPS_SYN_SENT)) { 4859 /* 4860 * Many stacks (ours included) will set the ACK number in an 4861 * FIN|ACK if the SYN times out -- no sequence to ACK. 4862 */ 4863 ack = dst->seqlo; 4864 } 4865 4866 if (seq == end) { 4867 /* Ease sequencing restrictions on no data packets */ 4868 seq = src->seqlo; 4869 end = seq; 4870 } 4871 4872 ackskew = dst->seqlo - ack; 4873 4874 4875 /* 4876 * Need to demodulate the sequence numbers in any TCP SACK options 4877 * (Selective ACK). We could optionally validate the SACK values 4878 * against the current ACK window, either forwards or backwards, but 4879 * I'm not confident that SACK has been implemented properly 4880 * everywhere. It wouldn't surprise me if several stacks accidently 4881 * SACK too far backwards of previously ACKed data. There really aren't 4882 * any security implications of bad SACKing unless the target stack 4883 * doesn't validate the option length correctly. Someone trying to 4884 * spoof into a TCP connection won't bother blindly sending SACK 4885 * options anyway. 4886 */ 4887 if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) { 4888 if (pf_modulate_sack(m, off, pd, th, dst)) 4889 copyback = 1; 4890 } 4891 4892 4893#define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */ 4894 if (SEQ_GEQ(src->seqhi, end) && 4895 /* Last octet inside other's window space */ 4896 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) && 4897 /* Retrans: not more than one window back */ 4898 (ackskew >= -MAXACKWINDOW) && 4899 /* Acking not more than one reassembled fragment backwards */ 4900 (ackskew <= (MAXACKWINDOW << sws)) && 4901 /* Acking not more than one window forward */ 4902 ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo || 4903 (orig_seq == src->seqlo + 1) || (pd->flags & PFDESC_IP_REAS) == 0)) { 4904 /* Require an exact/+1 sequence match on resets when possible */ 4905 4906 if (dst->scrub || src->scrub) { 4907 if (pf_normalize_tcp_stateful(m, off, pd, reason, th, 4908 *state, src, dst, ©back)) 4909 return (PF_DROP); 4910 } 4911 4912 /* update max window */ 4913 if (src->max_win < win) 4914 src->max_win = win; 4915 /* synchronize sequencing */ 4916 if (SEQ_GT(end, src->seqlo)) 4917 src->seqlo = end; 4918 /* slide the window of what the other end can send */ 4919 if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) 4920 dst->seqhi = ack + MAX((win << sws), 1); 4921 4922 4923 /* update states */ 4924 if (th->th_flags & TH_SYN) 4925 if (src->state < TCPS_SYN_SENT) 4926 src->state = TCPS_SYN_SENT; 4927 if (th->th_flags & TH_FIN) 4928 if (src->state < TCPS_CLOSING) 4929 src->state = TCPS_CLOSING; 4930 if (th->th_flags & TH_ACK) { 4931 if (dst->state == TCPS_SYN_SENT) { 4932 dst->state = TCPS_ESTABLISHED; 4933 if (src->state == TCPS_ESTABLISHED && 4934 (*state)->src_node != NULL && 4935 pf_src_connlimit(state)) { 4936 REASON_SET(reason, PFRES_SRCLIMIT); 4937 return (PF_DROP); 4938 } 4939 } else if (dst->state == TCPS_CLOSING) 4940 dst->state = TCPS_FIN_WAIT_2; 4941 } 4942 if (th->th_flags & TH_RST) 4943 src->state = dst->state = TCPS_TIME_WAIT; 4944 4945 /* update expire time */ 4946 (*state)->expire = time_second; 4947 if (src->state >= TCPS_FIN_WAIT_2 && 4948 dst->state >= TCPS_FIN_WAIT_2) 4949 (*state)->timeout = PFTM_TCP_CLOSED; 4950 else if (src->state >= TCPS_CLOSING && 4951 dst->state >= TCPS_CLOSING) 4952 (*state)->timeout = PFTM_TCP_FIN_WAIT; 4953 else if (src->state < TCPS_ESTABLISHED || 4954 dst->state < TCPS_ESTABLISHED) 4955 (*state)->timeout = PFTM_TCP_OPENING; 4956 else if (src->state >= TCPS_CLOSING || 4957 dst->state >= TCPS_CLOSING) 4958 (*state)->timeout = PFTM_TCP_CLOSING; 4959 else 4960 (*state)->timeout = PFTM_TCP_ESTABLISHED; 4961 4962 /* Fall through to PASS packet */ 4963 4964 } else if ((dst->state < TCPS_SYN_SENT || 4965 dst->state >= TCPS_FIN_WAIT_2 || 4966 src->state >= TCPS_FIN_WAIT_2) && 4967 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) && 4968 /* Within a window forward of the originating packet */ 4969 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) { 4970 /* Within a window backward of the originating packet */ 4971 4972 /* 4973 * This currently handles three situations: 4974 * 1) Stupid stacks will shotgun SYNs before their peer 4975 * replies. 4976 * 2) When PF catches an already established stream (the 4977 * firewall rebooted, the state table was flushed, routes 4978 * changed...) 4979 * 3) Packets get funky immediately after the connection 4980 * closes (this should catch Solaris spurious ACK|FINs 4981 * that web servers like to spew after a close) 4982 * 4983 * This must be a little more careful than the above code 4984 * since packet floods will also be caught here. We don't 4985 * update the TTL here to mitigate the damage of a packet 4986 * flood and so the same code can handle awkward establishment 4987 * and a loosened connection close. 4988 * In the establishment case, a correct peer response will 4989 * validate the connection, go through the normal state code 4990 * and keep updating the state TTL. 4991 */ 4992 4993 if (pf_status.debug >= PF_DEBUG_MISC) { 4994 printf("pf: loose state match: "); 4995 pf_print_state(*state); 4996 pf_print_flags(th->th_flags); 4997 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d " 4998 "pkts=%llu:%llu\n", seq, orig_seq, ack, pd->p_len, 4999#ifdef __FreeBSD__ 5000 ackskew, (unsigned long long)(*state)->packets[0], 5001 (unsigned long long)(*state)->packets[1]); 5002#else 5003 ackskew, (*state)->packets[0], 5004 (*state)->packets[1]); 5005#endif 5006 } 5007 5008 if (dst->scrub || src->scrub) { 5009 if (pf_normalize_tcp_stateful(m, off, pd, reason, th, 5010 *state, src, dst, ©back)) 5011 return (PF_DROP); 5012 } 5013 5014 /* update max window */ 5015 if (src->max_win < win) 5016 src->max_win = win; 5017 /* synchronize sequencing */ 5018 if (SEQ_GT(end, src->seqlo)) 5019 src->seqlo = end; 5020 /* slide the window of what the other end can send */ 5021 if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) 5022 dst->seqhi = ack + MAX((win << sws), 1); 5023 5024 /* 5025 * Cannot set dst->seqhi here since this could be a shotgunned 5026 * SYN and not an already established connection. 5027 */ 5028 5029 if (th->th_flags & TH_FIN) 5030 if (src->state < TCPS_CLOSING) 5031 src->state = TCPS_CLOSING; 5032 if (th->th_flags & TH_RST) 5033 src->state = dst->state = TCPS_TIME_WAIT; 5034 5035 /* Fall through to PASS packet */ 5036 5037 } else { 5038 if ((*state)->dst.state == TCPS_SYN_SENT && 5039 (*state)->src.state == TCPS_SYN_SENT) { 5040 /* Send RST for state mismatches during handshake */ 5041 if (!(th->th_flags & TH_RST)) 5042#ifdef __FreeBSD__ 5043 pf_send_tcp(m, (*state)->rule.ptr, pd->af, 5044#else 5045 pf_send_tcp((*state)->rule.ptr, pd->af, 5046#endif 5047 pd->dst, pd->src, th->th_dport, 5048 th->th_sport, ntohl(th->th_ack), 0, 5049 TH_RST, 0, 0, 5050 (*state)->rule.ptr->return_ttl, 1, 0, 5051 pd->eh, kif->pfik_ifp); 5052 src->seqlo = 0; 5053 src->seqhi = 1; 5054 src->max_win = 1; 5055 } else if (pf_status.debug >= PF_DEBUG_MISC) { 5056 printf("pf: BAD state: "); 5057 pf_print_state(*state); 5058 pf_print_flags(th->th_flags); 5059 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d " 5060 "pkts=%llu:%llu dir=%s,%s\n", 5061 seq, orig_seq, ack, pd->p_len, ackskew, 5062#ifdef __FreeBSD__ 5063 (unsigned long long)(*state)->packets[0], 5064 (unsigned long long)(*state)->packets[1], 5065#else 5066 (*state)->packets[0], (*state)->packets[1], 5067#endif 5068 direction == PF_IN ? "in" : "out", 5069 direction == (*state)->direction ? "fwd" : "rev"); 5070 printf("pf: State failure on: %c %c %c %c | %c %c\n", 5071 SEQ_GEQ(src->seqhi, end) ? ' ' : '1', 5072 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ? 5073 ' ': '2', 5074 (ackskew >= -MAXACKWINDOW) ? ' ' : '3', 5075 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4', 5076 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5', 5077 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6'); 5078 } 5079 REASON_SET(reason, PFRES_BADSTATE); 5080 return (PF_DROP); 5081 } 5082 5083 /* Any packets which have gotten here are to be passed */ 5084 5085 /* translate source/destination address, if necessary */ 5086 if (STATE_TRANSLATE(*state)) { 5087 if (direction == PF_OUT) 5088 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum, 5089 &th->th_sum, &(*state)->gwy.addr, 5090 (*state)->gwy.port, 0, pd->af); 5091 else 5092 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum, 5093 &th->th_sum, &(*state)->lan.addr, 5094 (*state)->lan.port, 0, pd->af); 5095 m_copyback(m, off, sizeof(*th), (caddr_t)th); 5096 } else if (copyback) { 5097 /* Copyback sequence modulation or stateful scrub changes */ 5098 m_copyback(m, off, sizeof(*th), (caddr_t)th); 5099 } 5100 5101 return (PF_PASS); 5102} 5103 5104int 5105pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif, 5106 struct mbuf *m, int off, void *h, struct pf_pdesc *pd) 5107{ 5108 struct pf_state_peer *src, *dst; 5109 struct pf_state_cmp key; 5110 struct udphdr *uh = pd->hdr.udp; 5111 5112 key.af = pd->af; 5113 key.proto = IPPROTO_UDP; 5114 if (direction == PF_IN) { 5115 PF_ACPY(&key.ext.addr, pd->src, key.af); 5116 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 5117 key.ext.port = uh->uh_sport; 5118 key.gwy.port = uh->uh_dport; 5119 } else { 5120 PF_ACPY(&key.lan.addr, pd->src, key.af); 5121 PF_ACPY(&key.ext.addr, pd->dst, key.af); 5122 key.lan.port = uh->uh_sport; 5123 key.ext.port = uh->uh_dport; 5124 } 5125 5126 STATE_LOOKUP(); 5127 5128 if (direction == (*state)->direction) { 5129 src = &(*state)->src; 5130 dst = &(*state)->dst; 5131 } else { 5132 src = &(*state)->dst; 5133 dst = &(*state)->src; 5134 } 5135 5136 /* update states */ 5137 if (src->state < PFUDPS_SINGLE) 5138 src->state = PFUDPS_SINGLE; 5139 if (dst->state == PFUDPS_SINGLE) 5140 dst->state = PFUDPS_MULTIPLE; 5141 5142 /* update expire time */ 5143 (*state)->expire = time_second; 5144 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE) 5145 (*state)->timeout = PFTM_UDP_MULTIPLE; 5146 else 5147 (*state)->timeout = PFTM_UDP_SINGLE; 5148 5149 /* translate source/destination address, if necessary */ 5150 if (STATE_TRANSLATE(*state)) { 5151 if (direction == PF_OUT) 5152 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum, 5153 &uh->uh_sum, &(*state)->gwy.addr, 5154 (*state)->gwy.port, 1, pd->af); 5155 else 5156 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum, 5157 &uh->uh_sum, &(*state)->lan.addr, 5158 (*state)->lan.port, 1, pd->af); 5159 m_copyback(m, off, sizeof(*uh), (caddr_t)uh); 5160 } 5161 5162 return (PF_PASS); 5163} 5164 5165int 5166pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif, 5167 struct mbuf *m, int off, void *h, struct pf_pdesc *pd, u_short *reason) 5168{ 5169 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 5170 u_int16_t icmpid = 0; /* make the compiler happy */ 5171 u_int16_t *icmpsum = NULL; /* make the compiler happy */ 5172 u_int8_t icmptype = 0; /* make the compiler happy */ 5173 int state_icmp = 0; 5174 struct pf_state_cmp key; 5175 5176 switch (pd->proto) { 5177#ifdef INET 5178 case IPPROTO_ICMP: 5179 icmptype = pd->hdr.icmp->icmp_type; 5180 icmpid = pd->hdr.icmp->icmp_id; 5181 icmpsum = &pd->hdr.icmp->icmp_cksum; 5182 5183 if (icmptype == ICMP_UNREACH || 5184 icmptype == ICMP_SOURCEQUENCH || 5185 icmptype == ICMP_REDIRECT || 5186 icmptype == ICMP_TIMXCEED || 5187 icmptype == ICMP_PARAMPROB) 5188 state_icmp++; 5189 break; 5190#endif /* INET */ 5191#ifdef INET6 5192 case IPPROTO_ICMPV6: 5193 icmptype = pd->hdr.icmp6->icmp6_type; 5194 icmpid = pd->hdr.icmp6->icmp6_id; 5195 icmpsum = &pd->hdr.icmp6->icmp6_cksum; 5196 5197 if (icmptype == ICMP6_DST_UNREACH || 5198 icmptype == ICMP6_PACKET_TOO_BIG || 5199 icmptype == ICMP6_TIME_EXCEEDED || 5200 icmptype == ICMP6_PARAM_PROB) 5201 state_icmp++; 5202 break; 5203#endif /* INET6 */ 5204 } 5205 5206 if (!state_icmp) { 5207 5208 /* 5209 * ICMP query/reply message not related to a TCP/UDP packet. 5210 * Search for an ICMP state. 5211 */ 5212 key.af = pd->af; 5213 key.proto = pd->proto; 5214 if (direction == PF_IN) { 5215 PF_ACPY(&key.ext.addr, pd->src, key.af); 5216 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 5217 key.ext.port = 0; 5218 key.gwy.port = icmpid; 5219 } else { 5220 PF_ACPY(&key.lan.addr, pd->src, key.af); 5221 PF_ACPY(&key.ext.addr, pd->dst, key.af); 5222 key.lan.port = icmpid; 5223 key.ext.port = 0; 5224 } 5225 5226 STATE_LOOKUP(); 5227 5228 (*state)->expire = time_second; 5229 (*state)->timeout = PFTM_ICMP_ERROR_REPLY; 5230 5231 /* translate source/destination address, if necessary */ 5232 if (STATE_TRANSLATE(*state)) { 5233 if (direction == PF_OUT) { 5234 switch (pd->af) { 5235#ifdef INET 5236 case AF_INET: 5237 pf_change_a(&saddr->v4.s_addr, 5238 pd->ip_sum, 5239 (*state)->gwy.addr.v4.s_addr, 0); 5240 pd->hdr.icmp->icmp_cksum = 5241 pf_cksum_fixup( 5242 pd->hdr.icmp->icmp_cksum, icmpid, 5243 (*state)->gwy.port, 0); 5244 pd->hdr.icmp->icmp_id = 5245 (*state)->gwy.port; 5246 m_copyback(m, off, ICMP_MINLEN, 5247 (caddr_t)pd->hdr.icmp); 5248 break; 5249#endif /* INET */ 5250#ifdef INET6 5251 case AF_INET6: 5252 pf_change_a6(saddr, 5253 &pd->hdr.icmp6->icmp6_cksum, 5254 &(*state)->gwy.addr, 0); 5255 m_copyback(m, off, 5256 sizeof(struct icmp6_hdr), 5257 (caddr_t)pd->hdr.icmp6); 5258 break; 5259#endif /* INET6 */ 5260 } 5261 } else { 5262 switch (pd->af) { 5263#ifdef INET 5264 case AF_INET: 5265 pf_change_a(&daddr->v4.s_addr, 5266 pd->ip_sum, 5267 (*state)->lan.addr.v4.s_addr, 0); 5268 pd->hdr.icmp->icmp_cksum = 5269 pf_cksum_fixup( 5270 pd->hdr.icmp->icmp_cksum, icmpid, 5271 (*state)->lan.port, 0); 5272 pd->hdr.icmp->icmp_id = 5273 (*state)->lan.port; 5274 m_copyback(m, off, ICMP_MINLEN, 5275 (caddr_t)pd->hdr.icmp); 5276 break; 5277#endif /* INET */ 5278#ifdef INET6 5279 case AF_INET6: 5280 pf_change_a6(daddr, 5281 &pd->hdr.icmp6->icmp6_cksum, 5282 &(*state)->lan.addr, 0); 5283 m_copyback(m, off, 5284 sizeof(struct icmp6_hdr), 5285 (caddr_t)pd->hdr.icmp6); 5286 break; 5287#endif /* INET6 */ 5288 } 5289 } 5290 } 5291 5292 return (PF_PASS); 5293 5294 } else { 5295 /* 5296 * ICMP error message in response to a TCP/UDP packet. 5297 * Extract the inner TCP/UDP header and search for that state. 5298 */ 5299 5300 struct pf_pdesc pd2; 5301#ifdef INET 5302 struct ip h2; 5303#endif /* INET */ 5304#ifdef INET6 5305 struct ip6_hdr h2_6; 5306 int terminal = 0; 5307#endif /* INET6 */ 5308 int ipoff2 = 0; /* make the compiler happy */ 5309 int off2 = 0; /* make the compiler happy */ 5310 5311 pd2.af = pd->af; 5312 switch (pd->af) { 5313#ifdef INET 5314 case AF_INET: 5315 /* offset of h2 in mbuf chain */ 5316 ipoff2 = off + ICMP_MINLEN; 5317 5318 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2), 5319 NULL, reason, pd2.af)) { 5320 DPFPRINTF(PF_DEBUG_MISC, 5321 ("pf: ICMP error message too short " 5322 "(ip)\n")); 5323 return (PF_DROP); 5324 } 5325 /* 5326 * ICMP error messages don't refer to non-first 5327 * fragments 5328 */ 5329 if (h2.ip_off & htons(IP_OFFMASK)) { 5330 REASON_SET(reason, PFRES_FRAG); 5331 return (PF_DROP); 5332 } 5333 5334 /* offset of protocol header that follows h2 */ 5335 off2 = ipoff2 + (h2.ip_hl << 2); 5336 5337 pd2.proto = h2.ip_p; 5338 pd2.src = (struct pf_addr *)&h2.ip_src; 5339 pd2.dst = (struct pf_addr *)&h2.ip_dst; 5340 pd2.ip_sum = &h2.ip_sum; 5341 break; 5342#endif /* INET */ 5343#ifdef INET6 5344 case AF_INET6: 5345 ipoff2 = off + sizeof(struct icmp6_hdr); 5346 5347 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6), 5348 NULL, reason, pd2.af)) { 5349 DPFPRINTF(PF_DEBUG_MISC, 5350 ("pf: ICMP error message too short " 5351 "(ip6)\n")); 5352 return (PF_DROP); 5353 } 5354 pd2.proto = h2_6.ip6_nxt; 5355 pd2.src = (struct pf_addr *)&h2_6.ip6_src; 5356 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst; 5357 pd2.ip_sum = NULL; 5358 off2 = ipoff2 + sizeof(h2_6); 5359 do { 5360 switch (pd2.proto) { 5361 case IPPROTO_FRAGMENT: 5362 /* 5363 * ICMPv6 error messages for 5364 * non-first fragments 5365 */ 5366 REASON_SET(reason, PFRES_FRAG); 5367 return (PF_DROP); 5368 case IPPROTO_AH: 5369 case IPPROTO_HOPOPTS: 5370 case IPPROTO_ROUTING: 5371 case IPPROTO_DSTOPTS: { 5372 /* get next header and header length */ 5373 struct ip6_ext opt6; 5374 5375 if (!pf_pull_hdr(m, off2, &opt6, 5376 sizeof(opt6), NULL, reason, 5377 pd2.af)) { 5378 DPFPRINTF(PF_DEBUG_MISC, 5379 ("pf: ICMPv6 short opt\n")); 5380 return (PF_DROP); 5381 } 5382 if (pd2.proto == IPPROTO_AH) 5383 off2 += (opt6.ip6e_len + 2) * 4; 5384 else 5385 off2 += (opt6.ip6e_len + 1) * 8; 5386 pd2.proto = opt6.ip6e_nxt; 5387 /* goto the next header */ 5388 break; 5389 } 5390 default: 5391 terminal++; 5392 break; 5393 } 5394 } while (!terminal); 5395 break; 5396#endif /* INET6 */ 5397#ifdef __FreeBSD__ 5398 default: 5399 panic("AF not supported: %d", pd->af); 5400#endif 5401 } 5402 5403 switch (pd2.proto) { 5404 case IPPROTO_TCP: { 5405 struct tcphdr th; 5406 u_int32_t seq; 5407 struct pf_state_peer *src, *dst; 5408 u_int8_t dws; 5409 int copyback = 0; 5410 5411 /* 5412 * Only the first 8 bytes of the TCP header can be 5413 * expected. Don't access any TCP header fields after 5414 * th_seq, an ackskew test is not possible. 5415 */ 5416 if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason, 5417 pd2.af)) { 5418 DPFPRINTF(PF_DEBUG_MISC, 5419 ("pf: ICMP error message too short " 5420 "(tcp)\n")); 5421 return (PF_DROP); 5422 } 5423 5424 key.af = pd2.af; 5425 key.proto = IPPROTO_TCP; 5426 if (direction == PF_IN) { 5427 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 5428 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 5429 key.ext.port = th.th_dport; 5430 key.gwy.port = th.th_sport; 5431 } else { 5432 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 5433 PF_ACPY(&key.ext.addr, pd2.src, key.af); 5434 key.lan.port = th.th_dport; 5435 key.ext.port = th.th_sport; 5436 } 5437 5438 STATE_LOOKUP(); 5439 5440 if (direction == (*state)->direction) { 5441 src = &(*state)->dst; 5442 dst = &(*state)->src; 5443 } else { 5444 src = &(*state)->src; 5445 dst = &(*state)->dst; 5446 } 5447 5448 if (src->wscale && dst->wscale) 5449 dws = dst->wscale & PF_WSCALE_MASK; 5450 else 5451 dws = 0; 5452 5453 /* Demodulate sequence number */ 5454 seq = ntohl(th.th_seq) - src->seqdiff; 5455 if (src->seqdiff) { 5456 pf_change_a(&th.th_seq, icmpsum, 5457 htonl(seq), 0); 5458 copyback = 1; 5459 } 5460 5461 if (!SEQ_GEQ(src->seqhi, seq) || 5462 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) { 5463 if (pf_status.debug >= PF_DEBUG_MISC) { 5464 printf("pf: BAD ICMP %d:%d ", 5465 icmptype, pd->hdr.icmp->icmp_code); 5466 pf_print_host(pd->src, 0, pd->af); 5467 printf(" -> "); 5468 pf_print_host(pd->dst, 0, pd->af); 5469 printf(" state: "); 5470 pf_print_state(*state); 5471 printf(" seq=%u\n", seq); 5472 } 5473 REASON_SET(reason, PFRES_BADSTATE); 5474 return (PF_DROP); 5475 } 5476 5477 if (STATE_TRANSLATE(*state)) { 5478 if (direction == PF_IN) { 5479 pf_change_icmp(pd2.src, &th.th_sport, 5480 daddr, &(*state)->lan.addr, 5481 (*state)->lan.port, NULL, 5482 pd2.ip_sum, icmpsum, 5483 pd->ip_sum, 0, pd2.af); 5484 } else { 5485 pf_change_icmp(pd2.dst, &th.th_dport, 5486 saddr, &(*state)->gwy.addr, 5487 (*state)->gwy.port, NULL, 5488 pd2.ip_sum, icmpsum, 5489 pd->ip_sum, 0, pd2.af); 5490 } 5491 copyback = 1; 5492 } 5493 5494 if (copyback) { 5495 switch (pd2.af) { 5496#ifdef INET 5497 case AF_INET: 5498 m_copyback(m, off, ICMP_MINLEN, 5499 (caddr_t)pd->hdr.icmp); 5500 m_copyback(m, ipoff2, sizeof(h2), 5501 (caddr_t)&h2); 5502 break; 5503#endif /* INET */ 5504#ifdef INET6 5505 case AF_INET6: 5506 m_copyback(m, off, 5507 sizeof(struct icmp6_hdr), 5508 (caddr_t)pd->hdr.icmp6); 5509 m_copyback(m, ipoff2, sizeof(h2_6), 5510 (caddr_t)&h2_6); 5511 break; 5512#endif /* INET6 */ 5513 } 5514 m_copyback(m, off2, 8, (caddr_t)&th); 5515 } 5516 5517 return (PF_PASS); 5518 break; 5519 } 5520 case IPPROTO_UDP: { 5521 struct udphdr uh; 5522 5523 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh), 5524 NULL, reason, pd2.af)) { 5525 DPFPRINTF(PF_DEBUG_MISC, 5526 ("pf: ICMP error message too short " 5527 "(udp)\n")); 5528 return (PF_DROP); 5529 } 5530 5531 key.af = pd2.af; 5532 key.proto = IPPROTO_UDP; 5533 if (direction == PF_IN) { 5534 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 5535 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 5536 key.ext.port = uh.uh_dport; 5537 key.gwy.port = uh.uh_sport; 5538 } else { 5539 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 5540 PF_ACPY(&key.ext.addr, pd2.src, key.af); 5541 key.lan.port = uh.uh_dport; 5542 key.ext.port = uh.uh_sport; 5543 } 5544 5545 STATE_LOOKUP(); 5546 5547 if (STATE_TRANSLATE(*state)) { 5548 if (direction == PF_IN) { 5549 pf_change_icmp(pd2.src, &uh.uh_sport, 5550 daddr, &(*state)->lan.addr, 5551 (*state)->lan.port, &uh.uh_sum, 5552 pd2.ip_sum, icmpsum, 5553 pd->ip_sum, 1, pd2.af); 5554 } else { 5555 pf_change_icmp(pd2.dst, &uh.uh_dport, 5556 saddr, &(*state)->gwy.addr, 5557 (*state)->gwy.port, &uh.uh_sum, 5558 pd2.ip_sum, icmpsum, 5559 pd->ip_sum, 1, pd2.af); 5560 } 5561 switch (pd2.af) { 5562#ifdef INET 5563 case AF_INET: 5564 m_copyback(m, off, ICMP_MINLEN, 5565 (caddr_t)pd->hdr.icmp); 5566 m_copyback(m, ipoff2, sizeof(h2), 5567 (caddr_t)&h2); 5568 break; 5569#endif /* INET */ 5570#ifdef INET6 5571 case AF_INET6: 5572 m_copyback(m, off, 5573 sizeof(struct icmp6_hdr), 5574 (caddr_t)pd->hdr.icmp6); 5575 m_copyback(m, ipoff2, sizeof(h2_6), 5576 (caddr_t)&h2_6); 5577 break; 5578#endif /* INET6 */ 5579 } 5580 m_copyback(m, off2, sizeof(uh), 5581 (caddr_t)&uh); 5582 } 5583 5584 return (PF_PASS); 5585 break; 5586 } 5587#ifdef INET 5588 case IPPROTO_ICMP: { 5589 struct icmp iih; 5590 5591 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN, 5592 NULL, reason, pd2.af)) { 5593 DPFPRINTF(PF_DEBUG_MISC, 5594 ("pf: ICMP error message too short i" 5595 "(icmp)\n")); 5596 return (PF_DROP); 5597 } 5598 5599 key.af = pd2.af; 5600 key.proto = IPPROTO_ICMP; 5601 if (direction == PF_IN) { 5602 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 5603 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 5604 key.ext.port = 0; 5605 key.gwy.port = iih.icmp_id; 5606 } else { 5607 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 5608 PF_ACPY(&key.ext.addr, pd2.src, key.af); 5609 key.lan.port = iih.icmp_id; 5610 key.ext.port = 0; 5611 } 5612 5613 STATE_LOOKUP(); 5614 5615 if (STATE_TRANSLATE(*state)) { 5616 if (direction == PF_IN) { 5617 pf_change_icmp(pd2.src, &iih.icmp_id, 5618 daddr, &(*state)->lan.addr, 5619 (*state)->lan.port, NULL, 5620 pd2.ip_sum, icmpsum, 5621 pd->ip_sum, 0, AF_INET); 5622 } else { 5623 pf_change_icmp(pd2.dst, &iih.icmp_id, 5624 saddr, &(*state)->gwy.addr, 5625 (*state)->gwy.port, NULL, 5626 pd2.ip_sum, icmpsum, 5627 pd->ip_sum, 0, AF_INET); 5628 } 5629 m_copyback(m, off, ICMP_MINLEN, 5630 (caddr_t)pd->hdr.icmp); 5631 m_copyback(m, ipoff2, sizeof(h2), 5632 (caddr_t)&h2); 5633 m_copyback(m, off2, ICMP_MINLEN, 5634 (caddr_t)&iih); 5635 } 5636 5637 return (PF_PASS); 5638 break; 5639 } 5640#endif /* INET */ 5641#ifdef INET6 5642 case IPPROTO_ICMPV6: { 5643 struct icmp6_hdr iih; 5644 5645 if (!pf_pull_hdr(m, off2, &iih, 5646 sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) { 5647 DPFPRINTF(PF_DEBUG_MISC, 5648 ("pf: ICMP error message too short " 5649 "(icmp6)\n")); 5650 return (PF_DROP); 5651 } 5652 5653 key.af = pd2.af; 5654 key.proto = IPPROTO_ICMPV6; 5655 if (direction == PF_IN) { 5656 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 5657 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 5658 key.ext.port = 0; 5659 key.gwy.port = iih.icmp6_id; 5660 } else { 5661 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 5662 PF_ACPY(&key.ext.addr, pd2.src, key.af); 5663 key.lan.port = iih.icmp6_id; 5664 key.ext.port = 0; 5665 } 5666 5667 STATE_LOOKUP(); 5668 5669 if (STATE_TRANSLATE(*state)) { 5670 if (direction == PF_IN) { 5671 pf_change_icmp(pd2.src, &iih.icmp6_id, 5672 daddr, &(*state)->lan.addr, 5673 (*state)->lan.port, NULL, 5674 pd2.ip_sum, icmpsum, 5675 pd->ip_sum, 0, AF_INET6); 5676 } else { 5677 pf_change_icmp(pd2.dst, &iih.icmp6_id, 5678 saddr, &(*state)->gwy.addr, 5679 (*state)->gwy.port, NULL, 5680 pd2.ip_sum, icmpsum, 5681 pd->ip_sum, 0, AF_INET6); 5682 } 5683 m_copyback(m, off, sizeof(struct icmp6_hdr), 5684 (caddr_t)pd->hdr.icmp6); 5685 m_copyback(m, ipoff2, sizeof(h2_6), 5686 (caddr_t)&h2_6); 5687 m_copyback(m, off2, sizeof(struct icmp6_hdr), 5688 (caddr_t)&iih); 5689 } 5690 5691 return (PF_PASS); 5692 break; 5693 } 5694#endif /* INET6 */ 5695 default: { 5696 key.af = pd2.af; 5697 key.proto = pd2.proto; 5698 if (direction == PF_IN) { 5699 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 5700 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 5701 key.ext.port = 0; 5702 key.gwy.port = 0; 5703 } else { 5704 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 5705 PF_ACPY(&key.ext.addr, pd2.src, key.af); 5706 key.lan.port = 0; 5707 key.ext.port = 0; 5708 } 5709 5710 STATE_LOOKUP(); 5711 5712 if (STATE_TRANSLATE(*state)) { 5713 if (direction == PF_IN) { 5714 pf_change_icmp(pd2.src, NULL, 5715 daddr, &(*state)->lan.addr, 5716 0, NULL, 5717 pd2.ip_sum, icmpsum, 5718 pd->ip_sum, 0, pd2.af); 5719 } else { 5720 pf_change_icmp(pd2.dst, NULL, 5721 saddr, &(*state)->gwy.addr, 5722 0, NULL, 5723 pd2.ip_sum, icmpsum, 5724 pd->ip_sum, 0, pd2.af); 5725 } 5726 switch (pd2.af) { 5727#ifdef INET 5728 case AF_INET: 5729 m_copyback(m, off, ICMP_MINLEN, 5730 (caddr_t)pd->hdr.icmp); 5731 m_copyback(m, ipoff2, sizeof(h2), 5732 (caddr_t)&h2); 5733 break; 5734#endif /* INET */ 5735#ifdef INET6 5736 case AF_INET6: 5737 m_copyback(m, off, 5738 sizeof(struct icmp6_hdr), 5739 (caddr_t)pd->hdr.icmp6); 5740 m_copyback(m, ipoff2, sizeof(h2_6), 5741 (caddr_t)&h2_6); 5742 break; 5743#endif /* INET6 */ 5744 } 5745 } 5746 5747 return (PF_PASS); 5748 break; 5749 } 5750 } 5751 } 5752} 5753 5754int 5755pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif, 5756 struct pf_pdesc *pd) 5757{ 5758 struct pf_state_peer *src, *dst; 5759 struct pf_state_cmp key; 5760 5761 key.af = pd->af; 5762 key.proto = pd->proto; 5763 if (direction == PF_IN) { 5764 PF_ACPY(&key.ext.addr, pd->src, key.af); 5765 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 5766 key.ext.port = 0; 5767 key.gwy.port = 0; 5768 } else { 5769 PF_ACPY(&key.lan.addr, pd->src, key.af); 5770 PF_ACPY(&key.ext.addr, pd->dst, key.af); 5771 key.lan.port = 0; 5772 key.ext.port = 0; 5773 } 5774 5775 STATE_LOOKUP(); 5776 5777 if (direction == (*state)->direction) { 5778 src = &(*state)->src; 5779 dst = &(*state)->dst; 5780 } else { 5781 src = &(*state)->dst; 5782 dst = &(*state)->src; 5783 } 5784 5785 /* update states */ 5786 if (src->state < PFOTHERS_SINGLE) 5787 src->state = PFOTHERS_SINGLE; 5788 if (dst->state == PFOTHERS_SINGLE) 5789 dst->state = PFOTHERS_MULTIPLE; 5790 5791 /* update expire time */ 5792 (*state)->expire = time_second; 5793 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE) 5794 (*state)->timeout = PFTM_OTHER_MULTIPLE; 5795 else 5796 (*state)->timeout = PFTM_OTHER_SINGLE; 5797 5798 /* translate source/destination address, if necessary */ 5799 if (STATE_TRANSLATE(*state)) { 5800 if (direction == PF_OUT) 5801 switch (pd->af) { 5802#ifdef INET 5803 case AF_INET: 5804 pf_change_a(&pd->src->v4.s_addr, 5805 pd->ip_sum, (*state)->gwy.addr.v4.s_addr, 5806 0); 5807 break; 5808#endif /* INET */ 5809#ifdef INET6 5810 case AF_INET6: 5811 PF_ACPY(pd->src, &(*state)->gwy.addr, pd->af); 5812 break; 5813#endif /* INET6 */ 5814 } 5815 else 5816 switch (pd->af) { 5817#ifdef INET 5818 case AF_INET: 5819 pf_change_a(&pd->dst->v4.s_addr, 5820 pd->ip_sum, (*state)->lan.addr.v4.s_addr, 5821 0); 5822 break; 5823#endif /* INET */ 5824#ifdef INET6 5825 case AF_INET6: 5826 PF_ACPY(pd->dst, &(*state)->lan.addr, pd->af); 5827 break; 5828#endif /* INET6 */ 5829 } 5830 } 5831 5832 return (PF_PASS); 5833} 5834 5835/* 5836 * ipoff and off are measured from the start of the mbuf chain. 5837 * h must be at "ipoff" on the mbuf chain. 5838 */ 5839void * 5840pf_pull_hdr(struct mbuf *m, int off, void *p, int len, 5841 u_short *actionp, u_short *reasonp, sa_family_t af) 5842{ 5843 switch (af) { 5844#ifdef INET 5845 case AF_INET: { 5846 struct ip *h = mtod(m, struct ip *); 5847 u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3; 5848 5849 if (fragoff) { 5850 if (fragoff >= len) 5851 ACTION_SET(actionp, PF_PASS); 5852 else { 5853 ACTION_SET(actionp, PF_DROP); 5854 REASON_SET(reasonp, PFRES_FRAG); 5855 } 5856 return (NULL); 5857 } 5858 if (m->m_pkthdr.len < off + len || 5859 ntohs(h->ip_len) < off + len) { 5860 ACTION_SET(actionp, PF_DROP); 5861 REASON_SET(reasonp, PFRES_SHORT); 5862 return (NULL); 5863 } 5864 break; 5865 } 5866#endif /* INET */ 5867#ifdef INET6 5868 case AF_INET6: { 5869 struct ip6_hdr *h = mtod(m, struct ip6_hdr *); 5870 5871 if (m->m_pkthdr.len < off + len || 5872 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) < 5873 (unsigned)(off + len)) { 5874 ACTION_SET(actionp, PF_DROP); 5875 REASON_SET(reasonp, PFRES_SHORT); 5876 return (NULL); 5877 } 5878 break; 5879 } 5880#endif /* INET6 */ 5881 } 5882 m_copydata(m, off, len, p); 5883 return (p); 5884} 5885 5886int 5887pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif) 5888{ 5889 struct sockaddr_in *dst; 5890 int ret = 1; 5891 int check_mpath; 5892#ifndef __FreeBSD__ 5893 extern int ipmultipath; 5894#endif 5895#ifdef INET6 5896#ifndef __FreeBSD__ 5897 extern int ip6_multipath; 5898#endif 5899 struct sockaddr_in6 *dst6; 5900 struct route_in6 ro; 5901#else 5902 struct route ro; 5903#endif 5904 struct radix_node *rn; 5905 struct rtentry *rt; 5906 struct ifnet *ifp; 5907 5908 check_mpath = 0; 5909 bzero(&ro, sizeof(ro)); 5910 switch (af) { 5911 case AF_INET: 5912 dst = satosin(&ro.ro_dst); 5913 dst->sin_family = AF_INET; 5914 dst->sin_len = sizeof(*dst); 5915 dst->sin_addr = addr->v4; 5916#ifndef __FreeBSD__ /* MULTIPATH_ROUTING */ 5917 if (ipmultipath) 5918 check_mpath = 1; 5919#endif 5920 break; 5921#ifdef INET6 5922 case AF_INET6: 5923 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 5924 dst6->sin6_family = AF_INET6; 5925 dst6->sin6_len = sizeof(*dst6); 5926 dst6->sin6_addr = addr->v6; 5927#ifndef __FreeBSD__ /* MULTIPATH_ROUTING */ 5928 if (ip6_multipath) 5929 check_mpath = 1; 5930#endif 5931 break; 5932#endif /* INET6 */ 5933 default: 5934 return (0); 5935 } 5936 5937 /* Skip checks for ipsec interfaces */ 5938 if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC) 5939 goto out; 5940 5941#ifdef __FreeBSD__ 5942 rtalloc_ign((struct route *)&ro, RTF_CLONING); 5943#else /* ! __FreeBSD__ */ 5944 rtalloc_noclone((struct route *)&ro, NO_CLONING); 5945#endif 5946 5947 if (ro.ro_rt != NULL) { 5948 /* No interface given, this is a no-route check */ 5949 if (kif == NULL) 5950 goto out; 5951 5952 if (kif->pfik_ifp == NULL) { 5953 ret = 0; 5954 goto out; 5955 } 5956 5957 /* Perform uRPF check if passed input interface */ 5958 ret = 0; 5959 rn = (struct radix_node *)ro.ro_rt; 5960 do { 5961 rt = (struct rtentry *)rn; 5962#ifndef __FreeBSD__ /* CARPDEV */ 5963 if (rt->rt_ifp->if_type == IFT_CARP) 5964 ifp = rt->rt_ifp->if_carpdev; 5965 else 5966#endif 5967 ifp = rt->rt_ifp; 5968 5969 if (kif->pfik_ifp == ifp) 5970 ret = 1; 5971#ifdef __FreeBSD__ /* MULTIPATH_ROUTING */ 5972 rn = NULL; 5973#else 5974 rn = rn_mpath_next(rn); 5975#endif 5976 } while (check_mpath == 1 && rn != NULL && ret == 0); 5977 } else 5978 ret = 0; 5979out: 5980 if (ro.ro_rt != NULL) 5981 RTFREE(ro.ro_rt); 5982 return (ret); 5983} 5984 5985int 5986pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw) 5987{ 5988 struct sockaddr_in *dst; 5989#ifdef INET6 5990 struct sockaddr_in6 *dst6; 5991 struct route_in6 ro; 5992#else 5993 struct route ro; 5994#endif 5995 int ret = 0; 5996 5997 bzero(&ro, sizeof(ro)); 5998 switch (af) { 5999 case AF_INET: 6000 dst = satosin(&ro.ro_dst); 6001 dst->sin_family = AF_INET; 6002 dst->sin_len = sizeof(*dst); 6003 dst->sin_addr = addr->v4; 6004 break; 6005#ifdef INET6 6006 case AF_INET6: 6007 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 6008 dst6->sin6_family = AF_INET6; 6009 dst6->sin6_len = sizeof(*dst6); 6010 dst6->sin6_addr = addr->v6; 6011 break; 6012#endif /* INET6 */ 6013 default: 6014 return (0); 6015 } 6016 6017#ifdef __FreeBSD__ 6018# ifdef RTF_PRCLONING 6019 rtalloc_ign((struct route *)&ro, (RTF_CLONING|RTF_PRCLONING)); 6020# else /* !RTF_PRCLONING */ 6021 rtalloc_ign((struct route *)&ro, RTF_CLONING); 6022# endif 6023#else /* ! __FreeBSD__ */ 6024 rtalloc_noclone((struct route *)&ro, NO_CLONING); 6025#endif 6026 6027 if (ro.ro_rt != NULL) { 6028#ifdef __FreeBSD__ 6029 /* XXX_IMPORT: later */ 6030#else 6031 if (ro.ro_rt->rt_labelid == aw->v.rtlabel) 6032 ret = 1; 6033#endif 6034 RTFREE(ro.ro_rt); 6035 } 6036 6037 return (ret); 6038} 6039 6040#ifdef INET 6041 6042void 6043pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp, 6044 struct pf_state *s, struct pf_pdesc *pd) 6045{ 6046 struct mbuf *m0, *m1; 6047 struct route iproute; 6048 struct route *ro = NULL; 6049 struct sockaddr_in *dst; 6050 struct ip *ip; 6051 struct ifnet *ifp = NULL; 6052 struct pf_addr naddr; 6053 struct pf_src_node *sn = NULL; 6054 int error = 0; 6055#ifdef __FreeBSD__ 6056 int sw_csum; 6057#endif 6058#ifdef IPSEC 6059 struct m_tag *mtag; 6060#endif /* IPSEC */ 6061 6062 if (m == NULL || *m == NULL || r == NULL || 6063 (dir != PF_IN && dir != PF_OUT) || oifp == NULL) 6064 panic("pf_route: invalid parameters"); 6065 6066 if (pd->pf_mtag->routed++ > 3) { 6067 m0 = *m; 6068 *m = NULL; 6069 goto bad; 6070 } 6071 6072 if (r->rt == PF_DUPTO) { 6073#ifdef __FreeBSD__ 6074 if ((m0 = m_dup(*m, M_DONTWAIT)) == NULL) 6075#else 6076 if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL) 6077#endif 6078 return; 6079 } else { 6080 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) 6081 return; 6082 m0 = *m; 6083 } 6084 6085 if (m0->m_len < sizeof(struct ip)) { 6086 DPFPRINTF(PF_DEBUG_URGENT, 6087 ("pf_route: m0->m_len < sizeof(struct ip)\n")); 6088 goto bad; 6089 } 6090 6091 ip = mtod(m0, struct ip *); 6092 6093 ro = &iproute; 6094 bzero((caddr_t)ro, sizeof(*ro)); 6095 dst = satosin(&ro->ro_dst); 6096 dst->sin_family = AF_INET; 6097 dst->sin_len = sizeof(*dst); 6098 dst->sin_addr = ip->ip_dst; 6099 6100 if (r->rt == PF_FASTROUTE) { 6101 rtalloc(ro); 6102 if (ro->ro_rt == 0) { 6103 ipstat.ips_noroute++; 6104 goto bad; 6105 } 6106 6107 ifp = ro->ro_rt->rt_ifp; 6108 ro->ro_rt->rt_use++; 6109 6110 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 6111 dst = satosin(ro->ro_rt->rt_gateway); 6112 } else { 6113 if (TAILQ_EMPTY(&r->rpool.list)) { 6114 DPFPRINTF(PF_DEBUG_URGENT, 6115 ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n")); 6116 goto bad; 6117 } 6118 if (s == NULL) { 6119 pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src, 6120 &naddr, NULL, &sn); 6121 if (!PF_AZERO(&naddr, AF_INET)) 6122 dst->sin_addr.s_addr = naddr.v4.s_addr; 6123 ifp = r->rpool.cur->kif ? 6124 r->rpool.cur->kif->pfik_ifp : NULL; 6125 } else { 6126 if (!PF_AZERO(&s->rt_addr, AF_INET)) 6127 dst->sin_addr.s_addr = 6128 s->rt_addr.v4.s_addr; 6129 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; 6130 } 6131 } 6132 if (ifp == NULL) 6133 goto bad; 6134 6135 if (oifp != ifp) { 6136#ifdef __FreeBSD__ 6137 PF_UNLOCK(); 6138 if (pf_test(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) { 6139 PF_LOCK(); 6140 goto bad; 6141 } else if (m0 == NULL) { 6142 PF_LOCK(); 6143 goto done; 6144 } 6145 PF_LOCK(); 6146#else 6147 if (pf_test(PF_OUT, ifp, &m0, NULL) != PF_PASS) 6148 goto bad; 6149 else if (m0 == NULL) 6150 goto done; 6151#endif 6152 if (m0->m_len < sizeof(struct ip)) { 6153 DPFPRINTF(PF_DEBUG_URGENT, 6154 ("pf_route: m0->m_len < sizeof(struct ip)\n")); 6155 goto bad; 6156 } 6157 ip = mtod(m0, struct ip *); 6158 } 6159 6160#ifdef __FreeBSD__ 6161 /* Copied from FreeBSD 5.1-CURRENT ip_output. */ 6162 m0->m_pkthdr.csum_flags |= CSUM_IP; 6163 sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist; 6164 if (sw_csum & CSUM_DELAY_DATA) { 6165 /* 6166 * XXX: in_delayed_cksum assumes HBO for ip->ip_len (at least) 6167 */ 6168 NTOHS(ip->ip_len); 6169 NTOHS(ip->ip_off); /* XXX: needed? */ 6170 in_delayed_cksum(m0); 6171 HTONS(ip->ip_len); 6172 HTONS(ip->ip_off); 6173 sw_csum &= ~CSUM_DELAY_DATA; 6174 } 6175 m0->m_pkthdr.csum_flags &= ifp->if_hwassist; 6176 6177 if (ntohs(ip->ip_len) <= ifp->if_mtu || 6178 (ifp->if_hwassist & CSUM_FRAGMENT && 6179 ((ip->ip_off & htons(IP_DF)) == 0))) { 6180 /* 6181 * ip->ip_len = htons(ip->ip_len); 6182 * ip->ip_off = htons(ip->ip_off); 6183 */ 6184 ip->ip_sum = 0; 6185 if (sw_csum & CSUM_DELAY_IP) { 6186 /* From KAME */ 6187 if (ip->ip_v == IPVERSION && 6188 (ip->ip_hl << 2) == sizeof(*ip)) { 6189 ip->ip_sum = in_cksum_hdr(ip); 6190 } else { 6191 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2); 6192 } 6193 } 6194 PF_UNLOCK(); 6195 error = (*ifp->if_output)(ifp, m0, sintosa(dst), ro->ro_rt); 6196 PF_LOCK(); 6197 goto done; 6198 } 6199 6200#else 6201 /* Copied from ip_output. */ 6202#ifdef IPSEC 6203 /* 6204 * If deferred crypto processing is needed, check that the 6205 * interface supports it. 6206 */ 6207 if ((mtag = m_tag_find(m0, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL)) 6208 != NULL && (ifp->if_capabilities & IFCAP_IPSEC) == 0) { 6209 /* Notify IPsec to do its own crypto. */ 6210 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 6211 goto bad; 6212 } 6213#endif /* IPSEC */ 6214 6215 /* Catch routing changes wrt. hardware checksumming for TCP or UDP. */ 6216 if (m0->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT) { 6217 if (!(ifp->if_capabilities & IFCAP_CSUM_TCPv4) || 6218 ifp->if_bridge != NULL) { 6219 in_delayed_cksum(m0); 6220 m0->m_pkthdr.csum_flags &= ~M_TCPV4_CSUM_OUT; /* Clear */ 6221 } 6222 } else if (m0->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT) { 6223 if (!(ifp->if_capabilities & IFCAP_CSUM_UDPv4) || 6224 ifp->if_bridge != NULL) { 6225 in_delayed_cksum(m0); 6226 m0->m_pkthdr.csum_flags &= ~M_UDPV4_CSUM_OUT; /* Clear */ 6227 } 6228 } 6229 6230 if (ntohs(ip->ip_len) <= ifp->if_mtu) { 6231 if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) && 6232 ifp->if_bridge == NULL) { 6233 m0->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT; 6234 ipstat.ips_outhwcsum++; 6235 } else { 6236 ip->ip_sum = 0; 6237 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2); 6238 } 6239 /* Update relevant hardware checksum stats for TCP/UDP */ 6240 if (m0->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT) 6241 tcpstat.tcps_outhwcsum++; 6242 else if (m0->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT) 6243 udpstat.udps_outhwcsum++; 6244 error = (*ifp->if_output)(ifp, m0, sintosa(dst), NULL); 6245 goto done; 6246 } 6247#endif 6248 /* 6249 * Too large for interface; fragment if possible. 6250 * Must be able to put at least 8 bytes per fragment. 6251 */ 6252 if (ip->ip_off & htons(IP_DF)) { 6253 ipstat.ips_cantfrag++; 6254 if (r->rt != PF_DUPTO) { 6255#ifdef __FreeBSD__ 6256 /* icmp_error() expects host byte ordering */ 6257 NTOHS(ip->ip_len); 6258 NTOHS(ip->ip_off); 6259 PF_UNLOCK(); 6260 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0, 6261 ifp->if_mtu); 6262 PF_LOCK(); 6263#else 6264 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0, 6265 ifp->if_mtu); 6266#endif 6267 goto done; 6268 } else 6269 goto bad; 6270 } 6271 6272 m1 = m0; 6273#ifdef __FreeBSD__ 6274 /* 6275 * XXX: is cheaper + less error prone than own function 6276 */ 6277 NTOHS(ip->ip_len); 6278 NTOHS(ip->ip_off); 6279 error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum); 6280#else 6281 error = ip_fragment(m0, ifp, ifp->if_mtu); 6282#endif 6283 if (error) { 6284#ifndef __FreeBSD__ /* ip_fragment does not do m_freem() on FreeBSD */ 6285 m0 = NULL; 6286#endif 6287 goto bad; 6288 } 6289 6290 for (m0 = m1; m0; m0 = m1) { 6291 m1 = m0->m_nextpkt; 6292 m0->m_nextpkt = 0; 6293#ifdef __FreeBSD__ 6294 if (error == 0) { 6295 PF_UNLOCK(); 6296 error = (*ifp->if_output)(ifp, m0, sintosa(dst), 6297 NULL); 6298 PF_LOCK(); 6299 } else 6300#else 6301 if (error == 0) 6302 error = (*ifp->if_output)(ifp, m0, sintosa(dst), 6303 NULL); 6304 else 6305#endif 6306 m_freem(m0); 6307 } 6308 6309 if (error == 0) 6310 ipstat.ips_fragmented++; 6311 6312done: 6313 if (r->rt != PF_DUPTO) 6314 *m = NULL; 6315 if (ro == &iproute && ro->ro_rt) 6316 RTFREE(ro->ro_rt); 6317 return; 6318 6319bad: 6320 m_freem(m0); 6321 goto done; 6322} 6323#endif /* INET */ 6324 6325#ifdef INET6 6326void 6327pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp, 6328 struct pf_state *s, struct pf_pdesc *pd) 6329{ 6330 struct mbuf *m0; 6331 struct route_in6 ip6route; 6332 struct route_in6 *ro; 6333 struct sockaddr_in6 *dst; 6334 struct ip6_hdr *ip6; 6335 struct ifnet *ifp = NULL; 6336 struct pf_addr naddr; 6337 struct pf_src_node *sn = NULL; 6338 int error = 0; 6339 6340 if (m == NULL || *m == NULL || r == NULL || 6341 (dir != PF_IN && dir != PF_OUT) || oifp == NULL) 6342 panic("pf_route6: invalid parameters"); 6343 6344 if (pd->pf_mtag->routed++ > 3) { 6345 m0 = *m; 6346 *m = NULL; 6347 goto bad; 6348 } 6349 6350 if (r->rt == PF_DUPTO) { 6351#ifdef __FreeBSD__ 6352 if ((m0 = m_dup(*m, M_DONTWAIT)) == NULL) 6353#else 6354 if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL) 6355#endif 6356 return; 6357 } else { 6358 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) 6359 return; 6360 m0 = *m; 6361 } 6362 6363 if (m0->m_len < sizeof(struct ip6_hdr)) { 6364 DPFPRINTF(PF_DEBUG_URGENT, 6365 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n")); 6366 goto bad; 6367 } 6368 ip6 = mtod(m0, struct ip6_hdr *); 6369 6370 ro = &ip6route; 6371 bzero((caddr_t)ro, sizeof(*ro)); 6372 dst = (struct sockaddr_in6 *)&ro->ro_dst; 6373 dst->sin6_family = AF_INET6; 6374 dst->sin6_len = sizeof(*dst); 6375 dst->sin6_addr = ip6->ip6_dst; 6376 6377 /* Cheat. XXX why only in the v6 case??? */ 6378 if (r->rt == PF_FASTROUTE) { 6379#ifdef __FreeBSD__ 6380 m0->m_flags |= M_SKIP_FIREWALL; 6381 PF_UNLOCK(); 6382 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL); 6383 PF_LOCK(); 6384#else 6385 mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT); 6386 if (mtag == NULL) 6387 goto bad; 6388 m_tag_prepend(m0, mtag); 6389 pd->pf_mtag->flags |= PF_TAG_GENERATED; 6390 ip6_output(m0, NULL, NULL, 0, NULL, NULL); 6391#endif 6392 return; 6393 } 6394 6395 if (TAILQ_EMPTY(&r->rpool.list)) { 6396 DPFPRINTF(PF_DEBUG_URGENT, 6397 ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n")); 6398 goto bad; 6399 } 6400 if (s == NULL) { 6401 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src, 6402 &naddr, NULL, &sn); 6403 if (!PF_AZERO(&naddr, AF_INET6)) 6404 PF_ACPY((struct pf_addr *)&dst->sin6_addr, 6405 &naddr, AF_INET6); 6406 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL; 6407 } else { 6408 if (!PF_AZERO(&s->rt_addr, AF_INET6)) 6409 PF_ACPY((struct pf_addr *)&dst->sin6_addr, 6410 &s->rt_addr, AF_INET6); 6411 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; 6412 } 6413 if (ifp == NULL) 6414 goto bad; 6415 6416 if (oifp != ifp) { 6417#ifdef __FreeBSD__ 6418 PF_UNLOCK(); 6419 if (pf_test6(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) { 6420 PF_LOCK(); 6421 goto bad; 6422 } else if (m0 == NULL) { 6423 PF_LOCK(); 6424 goto done; 6425 } 6426 PF_LOCK(); 6427#else 6428 if (pf_test6(PF_OUT, ifp, &m0, NULL) != PF_PASS) 6429 goto bad; 6430 else if (m0 == NULL) 6431 goto done; 6432#endif 6433 if (m0->m_len < sizeof(struct ip6_hdr)) { 6434 DPFPRINTF(PF_DEBUG_URGENT, 6435 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n")); 6436 goto bad; 6437 } 6438 ip6 = mtod(m0, struct ip6_hdr *); 6439 } 6440 6441 /* 6442 * If the packet is too large for the outgoing interface, 6443 * send back an icmp6 error. 6444 */ 6445 if (IN6_IS_SCOPE_EMBED(&dst->sin6_addr)) 6446 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index); 6447 if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) { 6448#ifdef __FreeBSD__ 6449 PF_UNLOCK(); 6450#endif 6451 error = nd6_output(ifp, ifp, m0, dst, NULL); 6452#ifdef __FreeBSD__ 6453 PF_LOCK(); 6454#endif 6455 } else { 6456 in6_ifstat_inc(ifp, ifs6_in_toobig); 6457#ifdef __FreeBSD__ 6458 if (r->rt != PF_DUPTO) { 6459 PF_UNLOCK(); 6460 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu); 6461 PF_LOCK(); 6462 } else 6463#else 6464 if (r->rt != PF_DUPTO) 6465 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu); 6466 else 6467#endif 6468 goto bad; 6469 } 6470 6471done: 6472 if (r->rt != PF_DUPTO) 6473 *m = NULL; 6474 return; 6475 6476bad: 6477 m_freem(m0); 6478 goto done; 6479} 6480#endif /* INET6 */ 6481 6482 6483#ifdef __FreeBSD__ 6484/* 6485 * FreeBSD supports cksum offloads for the following drivers. 6486 * em(4), fxp(4), ixgb(4), lge(4), ndis(4), nge(4), re(4), 6487 * ti(4), txp(4), xl(4) 6488 * 6489 * CSUM_DATA_VALID | CSUM_PSEUDO_HDR : 6490 * network driver performed cksum including pseudo header, need to verify 6491 * csum_data 6492 * CSUM_DATA_VALID : 6493 * network driver performed cksum, needs to additional pseudo header 6494 * cksum computation with partial csum_data(i.e. lack of H/W support for 6495 * pseudo header, for instance hme(4), sk(4) and possibly gem(4)) 6496 * 6497 * After validating the cksum of packet, set both flag CSUM_DATA_VALID and 6498 * CSUM_PSEUDO_HDR in order to avoid recomputation of the cksum in upper 6499 * TCP/UDP layer. 6500 * Also, set csum_data to 0xffff to force cksum validation. 6501 */ 6502int 6503pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p, sa_family_t af) 6504{ 6505 u_int16_t sum = 0; 6506 int hw_assist = 0; 6507 struct ip *ip; 6508 6509 if (off < sizeof(struct ip) || len < sizeof(struct udphdr)) 6510 return (1); 6511 if (m->m_pkthdr.len < off + len) 6512 return (1); 6513 6514 switch (p) { 6515 case IPPROTO_TCP: 6516 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { 6517 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) { 6518 sum = m->m_pkthdr.csum_data; 6519 } else { 6520 ip = mtod(m, struct ip *); 6521 sum = in_pseudo(ip->ip_src.s_addr, 6522 ip->ip_dst.s_addr, htonl((u_short)len + 6523 m->m_pkthdr.csum_data + IPPROTO_TCP)); 6524 } 6525 sum ^= 0xffff; 6526 ++hw_assist; 6527 } 6528 break; 6529 case IPPROTO_UDP: 6530 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { 6531 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) { 6532 sum = m->m_pkthdr.csum_data; 6533 } else { 6534 ip = mtod(m, struct ip *); 6535 sum = in_pseudo(ip->ip_src.s_addr, 6536 ip->ip_dst.s_addr, htonl((u_short)len + 6537 m->m_pkthdr.csum_data + IPPROTO_UDP)); 6538 } 6539 sum ^= 0xffff; 6540 ++hw_assist; 6541 } 6542 break; 6543 case IPPROTO_ICMP: 6544#ifdef INET6 6545 case IPPROTO_ICMPV6: 6546#endif /* INET6 */ 6547 break; 6548 default: 6549 return (1); 6550 } 6551 6552 if (!hw_assist) { 6553 switch (af) { 6554 case AF_INET: 6555 if (p == IPPROTO_ICMP) { 6556 if (m->m_len < off) 6557 return (1); 6558 m->m_data += off; 6559 m->m_len -= off; 6560 sum = in_cksum(m, len); 6561 m->m_data -= off; 6562 m->m_len += off; 6563 } else { 6564 if (m->m_len < sizeof(struct ip)) 6565 return (1); 6566 sum = in4_cksum(m, p, off, len); 6567 } 6568 break; 6569#ifdef INET6 6570 case AF_INET6: 6571 if (m->m_len < sizeof(struct ip6_hdr)) 6572 return (1); 6573 sum = in6_cksum(m, p, off, len); 6574 break; 6575#endif /* INET6 */ 6576 default: 6577 return (1); 6578 } 6579 } 6580 if (sum) { 6581 switch (p) { 6582 case IPPROTO_TCP: 6583 tcpstat.tcps_rcvbadsum++; 6584 break; 6585 case IPPROTO_UDP: 6586 udpstat.udps_badsum++; 6587 break; 6588 case IPPROTO_ICMP: 6589 icmpstat.icps_checksum++; 6590 break; 6591#ifdef INET6 6592 case IPPROTO_ICMPV6: 6593 icmp6stat.icp6s_checksum++; 6594 break; 6595#endif /* INET6 */ 6596 } 6597 return (1); 6598 } else { 6599 if (p == IPPROTO_TCP || p == IPPROTO_UDP) { 6600 m->m_pkthdr.csum_flags |= 6601 (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 6602 m->m_pkthdr.csum_data = 0xffff; 6603 } 6604 } 6605 return (0); 6606} 6607#else /* !__FreeBSD__ */ 6608/* 6609 * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag 6610 * off is the offset where the protocol header starts 6611 * len is the total length of protocol header plus payload 6612 * returns 0 when the checksum is valid, otherwise returns 1. 6613 */ 6614int 6615pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p, 6616 sa_family_t af) 6617{ 6618 u_int16_t flag_ok, flag_bad; 6619 u_int16_t sum; 6620 6621 switch (p) { 6622 case IPPROTO_TCP: 6623 flag_ok = M_TCP_CSUM_IN_OK; 6624 flag_bad = M_TCP_CSUM_IN_BAD; 6625 break; 6626 case IPPROTO_UDP: 6627 flag_ok = M_UDP_CSUM_IN_OK; 6628 flag_bad = M_UDP_CSUM_IN_BAD; 6629 break; 6630 case IPPROTO_ICMP: 6631#ifdef INET6 6632 case IPPROTO_ICMPV6: 6633#endif /* INET6 */ 6634 flag_ok = flag_bad = 0; 6635 break; 6636 default: 6637 return (1); 6638 } 6639 if (m->m_pkthdr.csum_flags & flag_ok) 6640 return (0); 6641 if (m->m_pkthdr.csum_flags & flag_bad) 6642 return (1); 6643 if (off < sizeof(struct ip) || len < sizeof(struct udphdr)) 6644 return (1); 6645 if (m->m_pkthdr.len < off + len) 6646 return (1); 6647 switch (af) { 6648#ifdef INET 6649 case AF_INET: 6650 if (p == IPPROTO_ICMP) { 6651 if (m->m_len < off) 6652 return (1); 6653 m->m_data += off; 6654 m->m_len -= off; 6655 sum = in_cksum(m, len); 6656 m->m_data -= off; 6657 m->m_len += off; 6658 } else { 6659 if (m->m_len < sizeof(struct ip)) 6660 return (1); 6661 sum = in4_cksum(m, p, off, len); 6662 } 6663 break; 6664#endif /* INET */ 6665#ifdef INET6 6666 case AF_INET6: 6667 if (m->m_len < sizeof(struct ip6_hdr)) 6668 return (1); 6669 sum = in6_cksum(m, p, off, len); 6670 break; 6671#endif /* INET6 */ 6672 default: 6673 return (1); 6674 } 6675 if (sum) { 6676 m->m_pkthdr.csum_flags |= flag_bad; 6677 switch (p) { 6678 case IPPROTO_TCP: 6679 tcpstat.tcps_rcvbadsum++; 6680 break; 6681 case IPPROTO_UDP: 6682 udpstat.udps_badsum++; 6683 break; 6684 case IPPROTO_ICMP: 6685 icmpstat.icps_checksum++; 6686 break; 6687#ifdef INET6 6688 case IPPROTO_ICMPV6: 6689 icmp6stat.icp6s_checksum++; 6690 break; 6691#endif /* INET6 */ 6692 } 6693 return (1); 6694 } 6695 m->m_pkthdr.csum_flags |= flag_ok; 6696 return (0); 6697} 6698#endif /* __FreeBSD__ */ 6699 6700#ifdef INET 6701int 6702#ifdef __FreeBSD__ 6703pf_test(int dir, struct ifnet *ifp, struct mbuf **m0, 6704 struct ether_header *eh, struct inpcb *inp) 6705#else 6706pf_test(int dir, struct ifnet *ifp, struct mbuf **m0, 6707 struct ether_header *eh) 6708#endif 6709{ 6710 struct pfi_kif *kif; 6711 u_short action, reason = 0, log = 0; 6712 struct mbuf *m = *m0; 6713 struct ip *h = NULL; /* make the compiler happy */ 6714 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr; 6715 struct pf_state *s = NULL; 6716 struct pf_ruleset *ruleset = NULL; 6717 struct pf_pdesc pd; 6718 int off, dirndx, pqid = 0; 6719 6720#ifdef __FreeBSD__ 6721 PF_LOCK(); 6722#endif 6723 if (!pf_status.running) 6724#ifdef __FreeBSD__ 6725 { 6726 PF_UNLOCK(); 6727#endif 6728 return (PF_PASS); 6729#ifdef __FreeBSD__ 6730 } 6731#endif 6732 6733 memset(&pd, 0, sizeof(pd)); 6734 if ((pd.pf_mtag = pf_get_mtag(m)) == NULL) { 6735#ifdef __FreeBSD__ 6736 PF_UNLOCK(); 6737#endif 6738 DPFPRINTF(PF_DEBUG_URGENT, 6739 ("pf_test: pf_get_mtag returned NULL\n")); 6740 return (PF_DROP); 6741 } 6742#ifdef __FreeBSD__ 6743 if (m->m_flags & M_SKIP_FIREWALL) { 6744 PF_UNLOCK(); 6745 return (PF_PASS); 6746 } 6747#else 6748 if (pd.pf_mtag->flags & PF_TAG_GENERATED) 6749 return (PF_PASS); 6750#endif 6751 6752#ifdef __FreeBSD__ 6753 /* XXX_IMPORT: later */ 6754#else 6755 if (ifp->if_type == IFT_CARP && ifp->if_carpdev) 6756 ifp = ifp->if_carpdev; 6757#endif 6758 6759 kif = (struct pfi_kif *)ifp->if_pf_kif; 6760 if (kif == NULL) { 6761#ifdef __FreeBSD__ 6762 PF_UNLOCK(); 6763#endif 6764 DPFPRINTF(PF_DEBUG_URGENT, 6765 ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname)); 6766 return (PF_DROP); 6767 } 6768 if (kif->pfik_flags & PFI_IFLAG_SKIP) { 6769#ifdef __FreeBSD__ 6770 PF_UNLOCK(); 6771#endif 6772 return (PF_PASS); 6773 } 6774 6775#ifdef __FreeBSD__ 6776 M_ASSERTPKTHDR(m); 6777#else 6778#ifdef DIAGNOSTIC 6779 if ((m->m_flags & M_PKTHDR) == 0) 6780 panic("non-M_PKTHDR is passed to pf_test"); 6781#endif /* DIAGNOSTIC */ 6782#endif /* __FreeBSD__ */ 6783 6784 if (m->m_pkthdr.len < (int)sizeof(*h)) { 6785 action = PF_DROP; 6786 REASON_SET(&reason, PFRES_SHORT); 6787 log = 1; 6788 goto done; 6789 } 6790 6791 /* We do IP header normalization and packet reassembly here */ 6792 if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) { 6793 action = PF_DROP; 6794 goto done; 6795 } 6796 m = *m0; 6797 h = mtod(m, struct ip *); 6798 6799 off = h->ip_hl << 2; 6800 if (off < (int)sizeof(*h)) { 6801 action = PF_DROP; 6802 REASON_SET(&reason, PFRES_SHORT); 6803 log = 1; 6804 goto done; 6805 } 6806 6807 pd.src = (struct pf_addr *)&h->ip_src; 6808 pd.dst = (struct pf_addr *)&h->ip_dst; 6809 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET); 6810 pd.ip_sum = &h->ip_sum; 6811 pd.proto = h->ip_p; 6812 pd.af = AF_INET; 6813 pd.tos = h->ip_tos; 6814 pd.tot_len = ntohs(h->ip_len); 6815 pd.eh = eh; 6816 6817 /* handle fragments that didn't get reassembled by normalization */ 6818 if (h->ip_off & htons(IP_MF | IP_OFFMASK)) { 6819 action = pf_test_fragment(&r, dir, kif, m, h, 6820 &pd, &a, &ruleset); 6821 goto done; 6822 } 6823 6824 switch (h->ip_p) { 6825 6826 case IPPROTO_TCP: { 6827 struct tcphdr th; 6828 6829 pd.hdr.tcp = &th; 6830 if (!pf_pull_hdr(m, off, &th, sizeof(th), 6831 &action, &reason, AF_INET)) { 6832 log = action != PF_PASS; 6833 goto done; 6834 } 6835 if (dir == PF_IN && pf_check_proto_cksum(m, off, 6836 ntohs(h->ip_len) - off, IPPROTO_TCP, AF_INET)) { 6837 REASON_SET(&reason, PFRES_PROTCKSUM); 6838 action = PF_DROP; 6839 goto done; 6840 } 6841 pd.p_len = pd.tot_len - off - (th.th_off << 2); 6842 if ((th.th_flags & TH_ACK) && pd.p_len == 0) 6843 pqid = 1; 6844 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd); 6845 if (action == PF_DROP) 6846 goto done; 6847 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd, 6848 &reason); 6849 if (action == PF_PASS) { 6850#if NPFSYNC 6851 pfsync_update_state(s); 6852#endif /* NPFSYNC */ 6853 r = s->rule.ptr; 6854 a = s->anchor.ptr; 6855 log = s->log; 6856 } else if (s == NULL) 6857#ifdef __FreeBSD__ 6858 action = pf_test_tcp(&r, &s, dir, kif, 6859 m, off, h, &pd, &a, &ruleset, NULL, inp); 6860#else 6861 action = pf_test_tcp(&r, &s, dir, kif, 6862 m, off, h, &pd, &a, &ruleset, &ipintrq); 6863#endif 6864 break; 6865 } 6866 6867 case IPPROTO_UDP: { 6868 struct udphdr uh; 6869 6870 pd.hdr.udp = &uh; 6871 if (!pf_pull_hdr(m, off, &uh, sizeof(uh), 6872 &action, &reason, AF_INET)) { 6873 log = action != PF_PASS; 6874 goto done; 6875 } 6876 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m, 6877 off, ntohs(h->ip_len) - off, IPPROTO_UDP, AF_INET)) { 6878 action = PF_DROP; 6879 REASON_SET(&reason, PFRES_PROTCKSUM); 6880 goto done; 6881 } 6882 if (uh.uh_dport == 0 || 6883 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off || 6884 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) { 6885 action = PF_DROP; 6886 REASON_SET(&reason, PFRES_SHORT); 6887 goto done; 6888 } 6889 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd); 6890 if (action == PF_PASS) { 6891#if NPFSYNC 6892 pfsync_update_state(s); 6893#endif /* NPFSYNC */ 6894 r = s->rule.ptr; 6895 a = s->anchor.ptr; 6896 log = s->log; 6897 } else if (s == NULL) 6898#ifdef __FreeBSD__ 6899 action = pf_test_udp(&r, &s, dir, kif, 6900 m, off, h, &pd, &a, &ruleset, NULL, inp); 6901#else 6902 action = pf_test_udp(&r, &s, dir, kif, 6903 m, off, h, &pd, &a, &ruleset, &ipintrq); 6904#endif 6905 break; 6906 } 6907 6908 case IPPROTO_ICMP: { 6909 struct icmp ih; 6910 6911 pd.hdr.icmp = &ih; 6912 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN, 6913 &action, &reason, AF_INET)) { 6914 log = action != PF_PASS; 6915 goto done; 6916 } 6917 if (dir == PF_IN && pf_check_proto_cksum(m, off, 6918 ntohs(h->ip_len) - off, IPPROTO_ICMP, AF_INET)) { 6919 action = PF_DROP; 6920 REASON_SET(&reason, PFRES_PROTCKSUM); 6921 goto done; 6922 } 6923 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd, 6924 &reason); 6925 if (action == PF_PASS) { 6926#if NPFSYNC 6927 pfsync_update_state(s); 6928#endif /* NPFSYNC */ 6929 r = s->rule.ptr; 6930 a = s->anchor.ptr; 6931 log = s->log; 6932 } else if (s == NULL) 6933#ifdef __FreeBSD__ 6934 action = pf_test_icmp(&r, &s, dir, kif, 6935 m, off, h, &pd, &a, &ruleset, NULL); 6936#else 6937 action = pf_test_icmp(&r, &s, dir, kif, 6938 m, off, h, &pd, &a, &ruleset, &ipintrq); 6939#endif 6940 break; 6941 } 6942 6943 default: 6944 action = pf_test_state_other(&s, dir, kif, &pd); 6945 if (action == PF_PASS) { 6946#if NPFSYNC 6947 pfsync_update_state(s); 6948#endif /* NPFSYNC */ 6949 r = s->rule.ptr; 6950 a = s->anchor.ptr; 6951 log = s->log; 6952 } else if (s == NULL) 6953#ifdef __FreeBSD__ 6954 action = pf_test_other(&r, &s, dir, kif, m, off, h, 6955 &pd, &a, &ruleset, NULL); 6956#else 6957 action = pf_test_other(&r, &s, dir, kif, m, off, h, 6958 &pd, &a, &ruleset, &ipintrq); 6959#endif 6960 break; 6961 } 6962 6963done: 6964 if (action == PF_PASS && h->ip_hl > 5 && 6965 !((s && s->allow_opts) || r->allow_opts)) { 6966 action = PF_DROP; 6967 REASON_SET(&reason, PFRES_IPOPTIONS); 6968 log = 1; 6969 DPFPRINTF(PF_DEBUG_MISC, 6970 ("pf: dropping packet with ip options\n")); 6971 } 6972 6973 if ((s && s->tag) || r->rtableid) 6974 pf_tag_packet(m, pd.pf_mtag, s ? s->tag : 0, r->rtableid); 6975 6976#ifdef ALTQ 6977 if (action == PF_PASS && r->qid) { 6978 if (pqid || (pd.tos & IPTOS_LOWDELAY)) 6979 pd.pf_mtag->qid = r->pqid; 6980 else 6981 pd.pf_mtag->qid = r->qid; 6982 /* add hints for ecn */ 6983 pd.pf_mtag->af = AF_INET; 6984 pd.pf_mtag->hdr = h; 6985 } 6986#endif /* ALTQ */ 6987 6988 /* 6989 * connections redirected to loopback should not match sockets 6990 * bound specifically to loopback due to security implications, 6991 * see tcp_input() and in_pcblookup_listen(). 6992 */ 6993 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP || 6994 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL && 6995 (s->nat_rule.ptr->action == PF_RDR || 6996 s->nat_rule.ptr->action == PF_BINAT) && 6997 (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) 6998 pd.pf_mtag->flags |= PF_TAG_TRANSLATE_LOCALHOST; 6999 7000 if (log) { 7001 struct pf_rule *lr; 7002 7003 if (s != NULL && s->nat_rule.ptr != NULL && 7004 s->nat_rule.ptr->log & PF_LOG_ALL) 7005 lr = s->nat_rule.ptr; 7006 else 7007 lr = r; 7008 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset, 7009 &pd); 7010 } 7011 7012 kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len; 7013 kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++; 7014 7015 if (action == PF_PASS || r->action == PF_DROP) { 7016 dirndx = (dir == PF_OUT); 7017 r->packets[dirndx]++; 7018 r->bytes[dirndx] += pd.tot_len; 7019 if (a != NULL) { 7020 a->packets[dirndx]++; 7021 a->bytes[dirndx] += pd.tot_len; 7022 } 7023 if (s != NULL) { 7024 if (s->nat_rule.ptr != NULL) { 7025 s->nat_rule.ptr->packets[dirndx]++; 7026 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len; 7027 } 7028 if (s->src_node != NULL) { 7029 s->src_node->packets[dirndx]++; 7030 s->src_node->bytes[dirndx] += pd.tot_len; 7031 } 7032 if (s->nat_src_node != NULL) { 7033 s->nat_src_node->packets[dirndx]++; 7034 s->nat_src_node->bytes[dirndx] += pd.tot_len; 7035 } 7036 dirndx = (dir == s->direction) ? 0 : 1; 7037 s->packets[dirndx]++; 7038 s->bytes[dirndx] += pd.tot_len; 7039 } 7040 tr = r; 7041 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule; 7042 if (nr != NULL) { 7043 struct pf_addr *x; 7044 /* 7045 * XXX: we need to make sure that the addresses 7046 * passed to pfr_update_stats() are the same than 7047 * the addresses used during matching (pfr_match) 7048 */ 7049 if (r == &pf_default_rule) { 7050 tr = nr; 7051 x = (s == NULL || s->direction == dir) ? 7052 &pd.baddr : &pd.naddr; 7053 } else 7054 x = (s == NULL || s->direction == dir) ? 7055 &pd.naddr : &pd.baddr; 7056 if (x == &pd.baddr || s == NULL) { 7057 /* we need to change the address */ 7058 if (dir == PF_OUT) 7059 pd.src = x; 7060 else 7061 pd.dst = x; 7062 } 7063 } 7064 if (tr->src.addr.type == PF_ADDR_TABLE) 7065 pfr_update_stats(tr->src.addr.p.tbl, (s == NULL || 7066 s->direction == dir) ? pd.src : pd.dst, pd.af, 7067 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 7068 tr->src.neg); 7069 if (tr->dst.addr.type == PF_ADDR_TABLE) 7070 pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL || 7071 s->direction == dir) ? pd.dst : pd.src, pd.af, 7072 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 7073 tr->dst.neg); 7074 } 7075 7076 7077 if (action == PF_SYNPROXY_DROP) { 7078 m_freem(*m0); 7079 *m0 = NULL; 7080 action = PF_PASS; 7081 } else if (r->rt) 7082 /* pf_route can free the mbuf causing *m0 to become NULL */ 7083 pf_route(m0, r, dir, ifp, s, &pd); 7084 7085#ifdef __FreeBSD__ 7086 PF_UNLOCK(); 7087#endif 7088 7089 return (action); 7090} 7091#endif /* INET */ 7092 7093#ifdef INET6 7094int 7095#ifdef __FreeBSD__ 7096pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0, 7097 struct ether_header *eh, struct inpcb *inp) 7098#else 7099pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0, 7100 struct ether_header *eh) 7101#endif 7102{ 7103 struct pfi_kif *kif; 7104 u_short action, reason = 0, log = 0; 7105 struct mbuf *m = *m0, *n = NULL; 7106 struct ip6_hdr *h; 7107 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr; 7108 struct pf_state *s = NULL; 7109 struct pf_ruleset *ruleset = NULL; 7110 struct pf_pdesc pd; 7111 int off, terminal = 0, dirndx, rh_cnt = 0; 7112 7113#ifdef __FreeBSD__ 7114 PF_LOCK(); 7115#endif 7116 7117 if (!pf_status.running) 7118#ifdef __FreeBSD__ 7119 { 7120 PF_UNLOCK(); 7121#endif 7122 return (PF_PASS); 7123#ifdef __FreeBSD__ 7124 } 7125#endif 7126 7127 memset(&pd, 0, sizeof(pd)); 7128 if ((pd.pf_mtag = pf_get_mtag(m)) == NULL) { 7129#ifdef __FreeBSD__ 7130 PF_UNLOCK(); 7131#endif 7132 DPFPRINTF(PF_DEBUG_URGENT, 7133 ("pf_test6: pf_get_mtag returned NULL\n")); 7134 return (PF_DROP); 7135 } 7136 if (pd.pf_mtag->flags & PF_TAG_GENERATED) 7137 return (PF_PASS); 7138 7139#ifdef __FreeBSD__ 7140 /* XXX_IMPORT: later */ 7141#else 7142 if (ifp->if_type == IFT_CARP && ifp->if_carpdev) 7143 ifp = ifp->if_carpdev; 7144#endif 7145 7146 kif = (struct pfi_kif *)ifp->if_pf_kif; 7147 if (kif == NULL) { 7148#ifdef __FreeBSD__ 7149 PF_UNLOCK(); 7150#endif 7151 DPFPRINTF(PF_DEBUG_URGENT, 7152 ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname)); 7153 return (PF_DROP); 7154 } 7155 if (kif->pfik_flags & PFI_IFLAG_SKIP) { 7156#ifdef __FreeBSD__ 7157 PF_UNLOCK(); 7158#endif 7159 return (PF_PASS); 7160 } 7161 7162#ifdef __FreeBSD__ 7163 M_ASSERTPKTHDR(m); 7164#else 7165#ifdef DIAGNOSTIC 7166 if ((m->m_flags & M_PKTHDR) == 0) 7167 panic("non-M_PKTHDR is passed to pf_test6"); 7168#endif /* DIAGNOSTIC */ 7169#endif 7170 7171#ifdef __FreeBSD__ 7172 h = NULL; /* make the compiler happy */ 7173#endif 7174 7175 if (m->m_pkthdr.len < (int)sizeof(*h)) { 7176 action = PF_DROP; 7177 REASON_SET(&reason, PFRES_SHORT); 7178 log = 1; 7179 goto done; 7180 } 7181 7182 /* We do IP header normalization and packet reassembly here */ 7183 if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) { 7184 action = PF_DROP; 7185 goto done; 7186 } 7187 m = *m0; 7188 h = mtod(m, struct ip6_hdr *); 7189 7190#if 1 7191 /* 7192 * we do not support jumbogram yet. if we keep going, zero ip6_plen 7193 * will do something bad, so drop the packet for now. 7194 */ 7195 if (htons(h->ip6_plen) == 0) { 7196 action = PF_DROP; 7197 REASON_SET(&reason, PFRES_NORM); /*XXX*/ 7198 goto done; 7199 } 7200#endif 7201 7202 pd.src = (struct pf_addr *)&h->ip6_src; 7203 pd.dst = (struct pf_addr *)&h->ip6_dst; 7204 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6); 7205 pd.ip_sum = NULL; 7206 pd.af = AF_INET6; 7207 pd.tos = 0; 7208 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr); 7209 pd.eh = eh; 7210 7211 off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr); 7212 pd.proto = h->ip6_nxt; 7213 do { 7214 switch (pd.proto) { 7215 case IPPROTO_FRAGMENT: 7216 action = pf_test_fragment(&r, dir, kif, m, h, 7217 &pd, &a, &ruleset); 7218 if (action == PF_DROP) 7219 REASON_SET(&reason, PFRES_FRAG); 7220 goto done; 7221 case IPPROTO_ROUTING: { 7222 struct ip6_rthdr rthdr; 7223 7224 if (rh_cnt++) { 7225 DPFPRINTF(PF_DEBUG_MISC, 7226 ("pf: IPv6 more than one rthdr\n")); 7227 action = PF_DROP; 7228 REASON_SET(&reason, PFRES_IPOPTIONS); 7229 log = 1; 7230 goto done; 7231 } 7232 if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL, 7233 &reason, pd.af)) { 7234 DPFPRINTF(PF_DEBUG_MISC, 7235 ("pf: IPv6 short rthdr\n")); 7236 action = PF_DROP; 7237 REASON_SET(&reason, PFRES_SHORT); 7238 log = 1; 7239 goto done; 7240 } 7241 if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) { 7242 DPFPRINTF(PF_DEBUG_MISC, 7243 ("pf: IPv6 rthdr0\n")); 7244 action = PF_DROP; 7245 REASON_SET(&reason, PFRES_IPOPTIONS); 7246 log = 1; 7247 goto done; 7248 } 7249 /* fallthrough */ 7250 } 7251 case IPPROTO_AH: 7252 case IPPROTO_HOPOPTS: 7253 case IPPROTO_DSTOPTS: { 7254 /* get next header and header length */ 7255 struct ip6_ext opt6; 7256 7257 if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6), 7258 NULL, &reason, pd.af)) { 7259 DPFPRINTF(PF_DEBUG_MISC, 7260 ("pf: IPv6 short opt\n")); 7261 action = PF_DROP; 7262 log = 1; 7263 goto done; 7264 } 7265 if (pd.proto == IPPROTO_AH) 7266 off += (opt6.ip6e_len + 2) * 4; 7267 else 7268 off += (opt6.ip6e_len + 1) * 8; 7269 pd.proto = opt6.ip6e_nxt; 7270 /* goto the next header */ 7271 break; 7272 } 7273 default: 7274 terminal++; 7275 break; 7276 } 7277 } while (!terminal); 7278 7279 /* if there's no routing header, use unmodified mbuf for checksumming */ 7280 if (!n) 7281 n = m; 7282 7283 switch (pd.proto) { 7284 7285 case IPPROTO_TCP: { 7286 struct tcphdr th; 7287 7288 pd.hdr.tcp = &th; 7289 if (!pf_pull_hdr(m, off, &th, sizeof(th), 7290 &action, &reason, AF_INET6)) { 7291 log = action != PF_PASS; 7292 goto done; 7293 } 7294 if (dir == PF_IN && pf_check_proto_cksum(n, off, 7295 ntohs(h->ip6_plen) - (off - sizeof(struct ip6_hdr)), 7296 IPPROTO_TCP, AF_INET6)) { 7297 action = PF_DROP; 7298 REASON_SET(&reason, PFRES_PROTCKSUM); 7299 goto done; 7300 } 7301 pd.p_len = pd.tot_len - off - (th.th_off << 2); 7302 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd); 7303 if (action == PF_DROP) 7304 goto done; 7305 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd, 7306 &reason); 7307 if (action == PF_PASS) { 7308#if NPFSYNC 7309 pfsync_update_state(s); 7310#endif /* NPFSYNC */ 7311 r = s->rule.ptr; 7312 a = s->anchor.ptr; 7313 log = s->log; 7314 } else if (s == NULL) 7315#ifdef __FreeBSD__ 7316 action = pf_test_tcp(&r, &s, dir, kif, 7317 m, off, h, &pd, &a, &ruleset, NULL, inp); 7318#else 7319 action = pf_test_tcp(&r, &s, dir, kif, 7320 m, off, h, &pd, &a, &ruleset, &ip6intrq); 7321#endif 7322 break; 7323 } 7324 7325 case IPPROTO_UDP: { 7326 struct udphdr uh; 7327 7328 pd.hdr.udp = &uh; 7329 if (!pf_pull_hdr(m, off, &uh, sizeof(uh), 7330 &action, &reason, AF_INET6)) { 7331 log = action != PF_PASS; 7332 goto done; 7333 } 7334 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(n, 7335 off, ntohs(h->ip6_plen) - (off - sizeof(struct ip6_hdr)), 7336 IPPROTO_UDP, AF_INET6)) { 7337 action = PF_DROP; 7338 REASON_SET(&reason, PFRES_PROTCKSUM); 7339 goto done; 7340 } 7341 if (uh.uh_dport == 0 || 7342 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off || 7343 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) { 7344 action = PF_DROP; 7345 REASON_SET(&reason, PFRES_SHORT); 7346 goto done; 7347 } 7348 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd); 7349 if (action == PF_PASS) { 7350#if NPFSYNC 7351 pfsync_update_state(s); 7352#endif /* NPFSYNC */ 7353 r = s->rule.ptr; 7354 a = s->anchor.ptr; 7355 log = s->log; 7356 } else if (s == NULL) 7357#ifdef __FreeBSD__ 7358 action = pf_test_udp(&r, &s, dir, kif, 7359 m, off, h, &pd, &a, &ruleset, NULL, inp); 7360#else 7361 action = pf_test_udp(&r, &s, dir, kif, 7362 m, off, h, &pd, &a, &ruleset, &ip6intrq); 7363#endif 7364 break; 7365 } 7366 7367 case IPPROTO_ICMPV6: { 7368 struct icmp6_hdr ih; 7369 7370 pd.hdr.icmp6 = &ih; 7371 if (!pf_pull_hdr(m, off, &ih, sizeof(ih), 7372 &action, &reason, AF_INET6)) { 7373 log = action != PF_PASS; 7374 goto done; 7375 } 7376 if (dir == PF_IN && pf_check_proto_cksum(n, off, 7377 ntohs(h->ip6_plen) - (off - sizeof(struct ip6_hdr)), 7378 IPPROTO_ICMPV6, AF_INET6)) { 7379 action = PF_DROP; 7380 REASON_SET(&reason, PFRES_PROTCKSUM); 7381 goto done; 7382 } 7383 action = pf_test_state_icmp(&s, dir, kif, 7384 m, off, h, &pd, &reason); 7385 if (action == PF_PASS) { 7386#if NPFSYNC 7387 pfsync_update_state(s); 7388#endif /* NPFSYNC */ 7389 r = s->rule.ptr; 7390 a = s->anchor.ptr; 7391 log = s->log; 7392 } else if (s == NULL) 7393#ifdef __FreeBSD__ 7394 action = pf_test_icmp(&r, &s, dir, kif, 7395 m, off, h, &pd, &a, &ruleset, NULL); 7396#else 7397 action = pf_test_icmp(&r, &s, dir, kif, 7398 m, off, h, &pd, &a, &ruleset, &ip6intrq); 7399#endif 7400 break; 7401 } 7402 7403 default: 7404 action = pf_test_state_other(&s, dir, kif, &pd); 7405 if (action == PF_PASS) { 7406#if NPFSYNC 7407 pfsync_update_state(s); 7408#endif /* NPFSYNC */ 7409 r = s->rule.ptr; 7410 a = s->anchor.ptr; 7411 log = s->log; 7412 } else if (s == NULL) 7413#ifdef __FreeBSD__ 7414 action = pf_test_other(&r, &s, dir, kif, m, off, h, 7415 &pd, &a, &ruleset, NULL); 7416#else 7417 action = pf_test_other(&r, &s, dir, kif, m, off, h, 7418 &pd, &a, &ruleset, &ip6intrq); 7419#endif 7420 break; 7421 } 7422 7423done: 7424 /* handle dangerous IPv6 extension headers. */ 7425 if (action == PF_PASS && rh_cnt && 7426 !((s && s->allow_opts) || r->allow_opts)) { 7427 action = PF_DROP; 7428 REASON_SET(&reason, PFRES_IPOPTIONS); 7429 log = 1; 7430 DPFPRINTF(PF_DEBUG_MISC, 7431 ("pf: dropping packet with dangerous v6 headers\n")); 7432 } 7433 7434 if ((s && s->tag) || r->rtableid) 7435 pf_tag_packet(m, pd.pf_mtag, s ? s->tag : 0, r->rtableid); 7436 7437#ifdef ALTQ 7438 if (action == PF_PASS && r->qid) { 7439 if (pd.tos & IPTOS_LOWDELAY) 7440 pd.pf_mtag->qid = r->pqid; 7441 else 7442 pd.pf_mtag->qid = r->qid; 7443 /* add hints for ecn */ 7444 pd.pf_mtag->af = AF_INET6; 7445 pd.pf_mtag->hdr = h; 7446 } 7447#endif /* ALTQ */ 7448 7449 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP || 7450 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL && 7451 (s->nat_rule.ptr->action == PF_RDR || 7452 s->nat_rule.ptr->action == PF_BINAT) && 7453 IN6_IS_ADDR_LOOPBACK(&pd.dst->v6)) 7454 pd.pf_mtag->flags |= PF_TAG_TRANSLATE_LOCALHOST; 7455 7456 if (log) { 7457 struct pf_rule *lr; 7458 7459 if (s != NULL && s->nat_rule.ptr != NULL && 7460 s->nat_rule.ptr->log & PF_LOG_ALL) 7461 lr = s->nat_rule.ptr; 7462 else 7463 lr = r; 7464 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset, 7465 &pd); 7466 } 7467 7468 kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len; 7469 kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++; 7470 7471 if (action == PF_PASS || r->action == PF_DROP) { 7472 dirndx = (dir == PF_OUT); 7473 r->packets[dirndx]++; 7474 r->bytes[dirndx] += pd.tot_len; 7475 if (a != NULL) { 7476 a->packets[dirndx]++; 7477 a->bytes[dirndx] += pd.tot_len; 7478 } 7479 if (s != NULL) { 7480 if (s->nat_rule.ptr != NULL) { 7481 s->nat_rule.ptr->packets[dirndx]++; 7482 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len; 7483 } 7484 if (s->src_node != NULL) { 7485 s->src_node->packets[dirndx]++; 7486 s->src_node->bytes[dirndx] += pd.tot_len; 7487 } 7488 if (s->nat_src_node != NULL) { 7489 s->nat_src_node->packets[dirndx]++; 7490 s->nat_src_node->bytes[dirndx] += pd.tot_len; 7491 } 7492 dirndx = (dir == s->direction) ? 0 : 1; 7493 s->packets[dirndx]++; 7494 s->bytes[dirndx] += pd.tot_len; 7495 } 7496 tr = r; 7497 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule; 7498 if (nr != NULL) { 7499 struct pf_addr *x; 7500 /* 7501 * XXX: we need to make sure that the addresses 7502 * passed to pfr_update_stats() are the same than 7503 * the addresses used during matching (pfr_match) 7504 */ 7505 if (r == &pf_default_rule) { 7506 tr = nr; 7507 x = (s == NULL || s->direction == dir) ? 7508 &pd.baddr : &pd.naddr; 7509 } else { 7510 x = (s == NULL || s->direction == dir) ? 7511 &pd.naddr : &pd.baddr; 7512 } 7513 if (x == &pd.baddr || s == NULL) { 7514 if (dir == PF_OUT) 7515 pd.src = x; 7516 else 7517 pd.dst = x; 7518 } 7519 } 7520 if (tr->src.addr.type == PF_ADDR_TABLE) 7521 pfr_update_stats(tr->src.addr.p.tbl, (s == NULL || 7522 s->direction == dir) ? pd.src : pd.dst, pd.af, 7523 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 7524 tr->src.neg); 7525 if (tr->dst.addr.type == PF_ADDR_TABLE) 7526 pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL || 7527 s->direction == dir) ? pd.dst : pd.src, pd.af, 7528 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 7529 tr->dst.neg); 7530 } 7531 7532 7533 if (action == PF_SYNPROXY_DROP) { 7534 m_freem(*m0); 7535 *m0 = NULL; 7536 action = PF_PASS; 7537 } else if (r->rt) 7538 /* pf_route6 can free the mbuf causing *m0 to become NULL */ 7539 pf_route6(m0, r, dir, ifp, s, &pd); 7540 7541#ifdef __FreeBSD__ 7542 PF_UNLOCK(); 7543#endif 7544 return (action); 7545} 7546#endif /* INET6 */ 7547 7548int 7549pf_check_congestion(struct ifqueue *ifq) 7550{ 7551#ifdef __FreeBSD__ 7552 /* XXX_IMPORT: later */ 7553 return (0); 7554#else 7555 if (ifq->ifq_congestion) 7556 return (1); 7557 else 7558 return (0); 7559#endif 7560} 7561