pf.c revision 178888
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 178888 2008-05-09 23:03:00Z julian $"); 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 kproc_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_netinet_firewall_reply(replyto, m); 1822 else 1823 mac_netinet_firewall_send(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#ifdef __FreeBSD__ 1843 { 1844 M_SETFIB(m, r->rtableid); 1845#endif 1846 pf_mtag->rtableid = r->rtableid; 1847#ifdef __FreeBSD__ 1848 } 1849#endif 1850#ifdef ALTQ 1851 if (r != NULL && r->qid) { 1852 pf_mtag->qid = r->qid; 1853 /* add hints for ecn */ 1854 pf_mtag->af = af; 1855 pf_mtag->hdr = mtod(m, struct ip *); 1856 } 1857#endif /* ALTQ */ 1858 m->m_data += max_linkhdr; 1859 m->m_pkthdr.len = m->m_len = len; 1860 m->m_pkthdr.rcvif = NULL; 1861 bzero(m->m_data, len); 1862 switch (af) { 1863#ifdef INET 1864 case AF_INET: 1865 h = mtod(m, struct ip *); 1866 1867 /* IP header fields included in the TCP checksum */ 1868 h->ip_p = IPPROTO_TCP; 1869 h->ip_len = htons(tlen); 1870 h->ip_src.s_addr = saddr->v4.s_addr; 1871 h->ip_dst.s_addr = daddr->v4.s_addr; 1872 1873 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip)); 1874 break; 1875#endif /* INET */ 1876#ifdef INET6 1877 case AF_INET6: 1878 h6 = mtod(m, struct ip6_hdr *); 1879 1880 /* IP header fields included in the TCP checksum */ 1881 h6->ip6_nxt = IPPROTO_TCP; 1882 h6->ip6_plen = htons(tlen); 1883 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr)); 1884 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr)); 1885 1886 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr)); 1887 break; 1888#endif /* INET6 */ 1889 } 1890 1891 /* TCP header */ 1892 th->th_sport = sport; 1893 th->th_dport = dport; 1894 th->th_seq = htonl(seq); 1895 th->th_ack = htonl(ack); 1896 th->th_off = tlen >> 2; 1897 th->th_flags = flags; 1898 th->th_win = htons(win); 1899 1900 if (mss) { 1901 opt = (char *)(th + 1); 1902 opt[0] = TCPOPT_MAXSEG; 1903 opt[1] = 4; 1904 HTONS(mss); 1905 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2); 1906 } 1907 1908 switch (af) { 1909#ifdef INET 1910 case AF_INET: 1911 /* TCP checksum */ 1912 th->th_sum = in_cksum(m, len); 1913 1914 /* Finish the IP header */ 1915 h->ip_v = 4; 1916 h->ip_hl = sizeof(*h) >> 2; 1917 h->ip_tos = IPTOS_LOWDELAY; 1918#ifdef __FreeBSD__ 1919 h->ip_off = path_mtu_discovery ? IP_DF : 0; 1920 h->ip_len = len; 1921#else 1922 h->ip_off = htons(ip_mtudisc ? IP_DF : 0); 1923 h->ip_len = htons(len); 1924#endif 1925 h->ip_ttl = ttl ? ttl : ip_defttl; 1926 h->ip_sum = 0; 1927 if (eh == NULL) { 1928#ifdef __FreeBSD__ 1929 PF_UNLOCK(); 1930 ip_output(m, (void *)NULL, (void *)NULL, 0, 1931 (void *)NULL, (void *)NULL); 1932 PF_LOCK(); 1933#else /* ! __FreeBSD__ */ 1934 ip_output(m, (void *)NULL, (void *)NULL, 0, 1935 (void *)NULL, (void *)NULL); 1936#endif 1937 } else { 1938 struct route ro; 1939 struct rtentry rt; 1940 struct ether_header *e = (void *)ro.ro_dst.sa_data; 1941 1942 if (ifp == NULL) { 1943 m_freem(m); 1944 return; 1945 } 1946 rt.rt_ifp = ifp; 1947 ro.ro_rt = &rt; 1948 ro.ro_dst.sa_len = sizeof(ro.ro_dst); 1949 ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT; 1950 bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN); 1951 bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN); 1952 e->ether_type = eh->ether_type; 1953#ifdef __FreeBSD__ 1954 PF_UNLOCK(); 1955 /* XXX_IMPORT: later */ 1956 ip_output(m, (void *)NULL, &ro, 0, 1957 (void *)NULL, (void *)NULL); 1958 PF_LOCK(); 1959#else /* ! __FreeBSD__ */ 1960 ip_output(m, (void *)NULL, &ro, IP_ROUTETOETHER, 1961 (void *)NULL, (void *)NULL); 1962#endif 1963 } 1964 break; 1965#endif /* INET */ 1966#ifdef INET6 1967 case AF_INET6: 1968 /* TCP checksum */ 1969 th->th_sum = in6_cksum(m, IPPROTO_TCP, 1970 sizeof(struct ip6_hdr), tlen); 1971 1972 h6->ip6_vfc |= IPV6_VERSION; 1973 h6->ip6_hlim = IPV6_DEFHLIM; 1974 1975#ifdef __FreeBSD__ 1976 PF_UNLOCK(); 1977 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL); 1978 PF_LOCK(); 1979#else 1980 ip6_output(m, NULL, NULL, 0, NULL, NULL); 1981#endif 1982 break; 1983#endif /* INET6 */ 1984 } 1985} 1986 1987void 1988pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af, 1989 struct pf_rule *r) 1990{ 1991 struct pf_mtag *pf_mtag; 1992 struct mbuf *m0; 1993#ifdef __FreeBSD__ 1994 struct ip *ip; 1995#endif 1996 1997#ifdef __FreeBSD__ 1998 m0 = m_copypacket(m, M_DONTWAIT); 1999 if (m0 == NULL) 2000 return; 2001#else 2002 m0 = m_copy(m, 0, M_COPYALL); 2003#endif 2004 if ((pf_mtag = pf_get_mtag(m0)) == NULL) 2005 return; 2006#ifdef __FreeBSD__ 2007 /* XXX: revisit */ 2008 m0->m_flags |= M_SKIP_FIREWALL; 2009#else 2010 pf_mtag->flags |= PF_TAG_GENERATED; 2011#endif 2012 2013 if (r->rtableid >= 0) 2014#ifdef __FreeBSD__ 2015 { 2016 M_SETFIB(m0, r->rtableid); 2017#endif 2018 pf_mtag->rtableid = r->rtableid; 2019#ifdef __FreeBSD__ 2020 } 2021#endif 2022 2023#ifdef ALTQ 2024 if (r->qid) { 2025 pf_mtag->qid = r->qid; 2026 /* add hints for ecn */ 2027 pf_mtag->af = af; 2028 pf_mtag->hdr = mtod(m0, struct ip *); 2029 } 2030#endif /* ALTQ */ 2031 2032 switch (af) { 2033#ifdef INET 2034 case AF_INET: 2035#ifdef __FreeBSD__ 2036 /* icmp_error() expects host byte ordering */ 2037 ip = mtod(m0, struct ip *); 2038 NTOHS(ip->ip_len); 2039 NTOHS(ip->ip_off); 2040 PF_UNLOCK(); 2041 icmp_error(m0, type, code, 0, 0); 2042 PF_LOCK(); 2043#else 2044 icmp_error(m0, type, code, 0, 0); 2045#endif 2046 break; 2047#endif /* INET */ 2048#ifdef INET6 2049 case AF_INET6: 2050#ifdef __FreeBSD__ 2051 PF_UNLOCK(); 2052#endif 2053 icmp6_error(m0, type, code, 0); 2054#ifdef __FreeBSD__ 2055 PF_LOCK(); 2056#endif 2057 break; 2058#endif /* INET6 */ 2059 } 2060} 2061 2062/* 2063 * Return 1 if the addresses a and b match (with mask m), otherwise return 0. 2064 * If n is 0, they match if they are equal. If n is != 0, they match if they 2065 * are different. 2066 */ 2067int 2068pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m, 2069 struct pf_addr *b, sa_family_t af) 2070{ 2071 int match = 0; 2072 2073 switch (af) { 2074#ifdef INET 2075 case AF_INET: 2076 if ((a->addr32[0] & m->addr32[0]) == 2077 (b->addr32[0] & m->addr32[0])) 2078 match++; 2079 break; 2080#endif /* INET */ 2081#ifdef INET6 2082 case AF_INET6: 2083 if (((a->addr32[0] & m->addr32[0]) == 2084 (b->addr32[0] & m->addr32[0])) && 2085 ((a->addr32[1] & m->addr32[1]) == 2086 (b->addr32[1] & m->addr32[1])) && 2087 ((a->addr32[2] & m->addr32[2]) == 2088 (b->addr32[2] & m->addr32[2])) && 2089 ((a->addr32[3] & m->addr32[3]) == 2090 (b->addr32[3] & m->addr32[3]))) 2091 match++; 2092 break; 2093#endif /* INET6 */ 2094 } 2095 if (match) { 2096 if (n) 2097 return (0); 2098 else 2099 return (1); 2100 } else { 2101 if (n) 2102 return (1); 2103 else 2104 return (0); 2105 } 2106} 2107 2108int 2109pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p) 2110{ 2111 switch (op) { 2112 case PF_OP_IRG: 2113 return ((p > a1) && (p < a2)); 2114 case PF_OP_XRG: 2115 return ((p < a1) || (p > a2)); 2116 case PF_OP_RRG: 2117 return ((p >= a1) && (p <= a2)); 2118 case PF_OP_EQ: 2119 return (p == a1); 2120 case PF_OP_NE: 2121 return (p != a1); 2122 case PF_OP_LT: 2123 return (p < a1); 2124 case PF_OP_LE: 2125 return (p <= a1); 2126 case PF_OP_GT: 2127 return (p > a1); 2128 case PF_OP_GE: 2129 return (p >= a1); 2130 } 2131 return (0); /* never reached */ 2132} 2133 2134int 2135pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p) 2136{ 2137 NTOHS(a1); 2138 NTOHS(a2); 2139 NTOHS(p); 2140 return (pf_match(op, a1, a2, p)); 2141} 2142 2143int 2144pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u) 2145{ 2146 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE) 2147 return (0); 2148 return (pf_match(op, a1, a2, u)); 2149} 2150 2151int 2152pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g) 2153{ 2154 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE) 2155 return (0); 2156 return (pf_match(op, a1, a2, g)); 2157} 2158 2159#ifndef __FreeBSD__ 2160struct pf_mtag * 2161pf_find_mtag(struct mbuf *m) 2162{ 2163 struct m_tag *mtag; 2164 2165 if ((mtag = m_tag_find(m, PACKET_TAG_PF, NULL)) == NULL) 2166 return (NULL); 2167 2168 return ((struct pf_mtag *)(mtag + 1)); 2169} 2170 2171struct pf_mtag * 2172pf_get_mtag(struct mbuf *m) 2173{ 2174 struct m_tag *mtag; 2175 2176 if ((mtag = m_tag_find(m, PACKET_TAG_PF, NULL)) == NULL) { 2177 mtag = m_tag_get(PACKET_TAG_PF, sizeof(struct pf_mtag), 2178 M_NOWAIT); 2179 if (mtag == NULL) 2180 return (NULL); 2181 bzero(mtag + 1, sizeof(struct pf_mtag)); 2182 m_tag_prepend(m, mtag); 2183 } 2184 2185 return ((struct pf_mtag *)(mtag + 1)); 2186} 2187#endif 2188 2189int 2190pf_match_tag(struct mbuf *m, struct pf_rule *r, struct pf_mtag *pf_mtag, 2191 int *tag) 2192{ 2193 if (*tag == -1) 2194 *tag = pf_mtag->tag; 2195 2196 return ((!r->match_tag_not && r->match_tag == *tag) || 2197 (r->match_tag_not && r->match_tag != *tag)); 2198} 2199 2200int 2201pf_tag_packet(struct mbuf *m, struct pf_mtag *pf_mtag, int tag, int rtableid) 2202{ 2203 if (tag <= 0 && rtableid < 0) 2204 return (0); 2205 2206 if (pf_mtag == NULL) 2207 if ((pf_mtag = pf_get_mtag(m)) == NULL) 2208 return (1); 2209 if (tag > 0) 2210 pf_mtag->tag = tag; 2211 if (rtableid >= 0) 2212#ifdef __FreeBSD__ 2213 { 2214 M_SETFIB(m, rtableid); 2215#endif 2216 pf_mtag->rtableid = rtableid; 2217#ifdef __FreeBSD__ 2218 } 2219#endif 2220 2221 return (0); 2222} 2223 2224static void 2225pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n, 2226 struct pf_rule **r, struct pf_rule **a, int *match) 2227{ 2228 struct pf_anchor_stackframe *f; 2229 2230 (*r)->anchor->match = 0; 2231 if (match) 2232 *match = 0; 2233 if (*depth >= sizeof(pf_anchor_stack) / 2234 sizeof(pf_anchor_stack[0])) { 2235 printf("pf_step_into_anchor: stack overflow\n"); 2236 *r = TAILQ_NEXT(*r, entries); 2237 return; 2238 } else if (*depth == 0 && a != NULL) 2239 *a = *r; 2240 f = pf_anchor_stack + (*depth)++; 2241 f->rs = *rs; 2242 f->r = *r; 2243 if ((*r)->anchor_wildcard) { 2244 f->parent = &(*r)->anchor->children; 2245 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) == 2246 NULL) { 2247 *r = NULL; 2248 return; 2249 } 2250 *rs = &f->child->ruleset; 2251 } else { 2252 f->parent = NULL; 2253 f->child = NULL; 2254 *rs = &(*r)->anchor->ruleset; 2255 } 2256 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr); 2257} 2258 2259int 2260pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n, 2261 struct pf_rule **r, struct pf_rule **a, int *match) 2262{ 2263 struct pf_anchor_stackframe *f; 2264 int quick = 0; 2265 2266 do { 2267 if (*depth <= 0) 2268 break; 2269 f = pf_anchor_stack + *depth - 1; 2270 if (f->parent != NULL && f->child != NULL) { 2271 if (f->child->match || 2272 (match != NULL && *match)) { 2273 f->r->anchor->match = 1; 2274 *match = 0; 2275 } 2276 f->child = RB_NEXT(pf_anchor_node, f->parent, f->child); 2277 if (f->child != NULL) { 2278 *rs = &f->child->ruleset; 2279 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr); 2280 if (*r == NULL) 2281 continue; 2282 else 2283 break; 2284 } 2285 } 2286 (*depth)--; 2287 if (*depth == 0 && a != NULL) 2288 *a = NULL; 2289 *rs = f->rs; 2290 if (f->r->anchor->match || (match != NULL && *match)) 2291 quick = f->r->quick; 2292 *r = TAILQ_NEXT(f->r, entries); 2293 } while (*r == NULL); 2294 2295 return (quick); 2296} 2297 2298#ifdef INET6 2299void 2300pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr, 2301 struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af) 2302{ 2303 switch (af) { 2304#ifdef INET 2305 case AF_INET: 2306 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) | 2307 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]); 2308 break; 2309#endif /* INET */ 2310 case AF_INET6: 2311 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) | 2312 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]); 2313 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) | 2314 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]); 2315 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) | 2316 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]); 2317 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) | 2318 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]); 2319 break; 2320 } 2321} 2322 2323void 2324pf_addr_inc(struct pf_addr *addr, sa_family_t af) 2325{ 2326 switch (af) { 2327#ifdef INET 2328 case AF_INET: 2329 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1); 2330 break; 2331#endif /* INET */ 2332 case AF_INET6: 2333 if (addr->addr32[3] == 0xffffffff) { 2334 addr->addr32[3] = 0; 2335 if (addr->addr32[2] == 0xffffffff) { 2336 addr->addr32[2] = 0; 2337 if (addr->addr32[1] == 0xffffffff) { 2338 addr->addr32[1] = 0; 2339 addr->addr32[0] = 2340 htonl(ntohl(addr->addr32[0]) + 1); 2341 } else 2342 addr->addr32[1] = 2343 htonl(ntohl(addr->addr32[1]) + 1); 2344 } else 2345 addr->addr32[2] = 2346 htonl(ntohl(addr->addr32[2]) + 1); 2347 } else 2348 addr->addr32[3] = 2349 htonl(ntohl(addr->addr32[3]) + 1); 2350 break; 2351 } 2352} 2353#endif /* INET6 */ 2354 2355#define mix(a,b,c) \ 2356 do { \ 2357 a -= b; a -= c; a ^= (c >> 13); \ 2358 b -= c; b -= a; b ^= (a << 8); \ 2359 c -= a; c -= b; c ^= (b >> 13); \ 2360 a -= b; a -= c; a ^= (c >> 12); \ 2361 b -= c; b -= a; b ^= (a << 16); \ 2362 c -= a; c -= b; c ^= (b >> 5); \ 2363 a -= b; a -= c; a ^= (c >> 3); \ 2364 b -= c; b -= a; b ^= (a << 10); \ 2365 c -= a; c -= b; c ^= (b >> 15); \ 2366 } while (0) 2367 2368/* 2369 * hash function based on bridge_hash in if_bridge.c 2370 */ 2371void 2372pf_hash(struct pf_addr *inaddr, struct pf_addr *hash, 2373 struct pf_poolhashkey *key, sa_family_t af) 2374{ 2375 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0]; 2376 2377 switch (af) { 2378#ifdef INET 2379 case AF_INET: 2380 a += inaddr->addr32[0]; 2381 b += key->key32[1]; 2382 mix(a, b, c); 2383 hash->addr32[0] = c + key->key32[2]; 2384 break; 2385#endif /* INET */ 2386#ifdef INET6 2387 case AF_INET6: 2388 a += inaddr->addr32[0]; 2389 b += inaddr->addr32[2]; 2390 mix(a, b, c); 2391 hash->addr32[0] = c; 2392 a += inaddr->addr32[1]; 2393 b += inaddr->addr32[3]; 2394 c += key->key32[1]; 2395 mix(a, b, c); 2396 hash->addr32[1] = c; 2397 a += inaddr->addr32[2]; 2398 b += inaddr->addr32[1]; 2399 c += key->key32[2]; 2400 mix(a, b, c); 2401 hash->addr32[2] = c; 2402 a += inaddr->addr32[3]; 2403 b += inaddr->addr32[0]; 2404 c += key->key32[3]; 2405 mix(a, b, c); 2406 hash->addr32[3] = c; 2407 break; 2408#endif /* INET6 */ 2409 } 2410} 2411 2412int 2413pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr, 2414 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn) 2415{ 2416 unsigned char hash[16]; 2417 struct pf_pool *rpool = &r->rpool; 2418 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr; 2419 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask; 2420 struct pf_pooladdr *acur = rpool->cur; 2421 struct pf_src_node k; 2422 2423 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR && 2424 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { 2425 k.af = af; 2426 PF_ACPY(&k.addr, saddr, af); 2427 if (r->rule_flag & PFRULE_RULESRCTRACK || 2428 r->rpool.opts & PF_POOL_STICKYADDR) 2429 k.rule.ptr = r; 2430 else 2431 k.rule.ptr = NULL; 2432 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++; 2433 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k); 2434 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) { 2435 PF_ACPY(naddr, &(*sn)->raddr, af); 2436 if (pf_status.debug >= PF_DEBUG_MISC) { 2437 printf("pf_map_addr: src tracking maps "); 2438 pf_print_host(&k.addr, 0, af); 2439 printf(" to "); 2440 pf_print_host(naddr, 0, af); 2441 printf("\n"); 2442 } 2443 return (0); 2444 } 2445 } 2446 2447 if (rpool->cur->addr.type == PF_ADDR_NOROUTE) 2448 return (1); 2449 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2450 switch (af) { 2451#ifdef INET 2452 case AF_INET: 2453 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 && 2454 (rpool->opts & PF_POOL_TYPEMASK) != 2455 PF_POOL_ROUNDROBIN) 2456 return (1); 2457 raddr = &rpool->cur->addr.p.dyn->pfid_addr4; 2458 rmask = &rpool->cur->addr.p.dyn->pfid_mask4; 2459 break; 2460#endif /* INET */ 2461#ifdef INET6 2462 case AF_INET6: 2463 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 && 2464 (rpool->opts & PF_POOL_TYPEMASK) != 2465 PF_POOL_ROUNDROBIN) 2466 return (1); 2467 raddr = &rpool->cur->addr.p.dyn->pfid_addr6; 2468 rmask = &rpool->cur->addr.p.dyn->pfid_mask6; 2469 break; 2470#endif /* INET6 */ 2471 } 2472 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2473 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN) 2474 return (1); /* unsupported */ 2475 } else { 2476 raddr = &rpool->cur->addr.v.a.addr; 2477 rmask = &rpool->cur->addr.v.a.mask; 2478 } 2479 2480 switch (rpool->opts & PF_POOL_TYPEMASK) { 2481 case PF_POOL_NONE: 2482 PF_ACPY(naddr, raddr, af); 2483 break; 2484 case PF_POOL_BITMASK: 2485 PF_POOLMASK(naddr, raddr, rmask, saddr, af); 2486 break; 2487 case PF_POOL_RANDOM: 2488 if (init_addr != NULL && PF_AZERO(init_addr, af)) { 2489 switch (af) { 2490#ifdef INET 2491 case AF_INET: 2492 rpool->counter.addr32[0] = htonl(arc4random()); 2493 break; 2494#endif /* INET */ 2495#ifdef INET6 2496 case AF_INET6: 2497 if (rmask->addr32[3] != 0xffffffff) 2498 rpool->counter.addr32[3] = 2499 htonl(arc4random()); 2500 else 2501 break; 2502 if (rmask->addr32[2] != 0xffffffff) 2503 rpool->counter.addr32[2] = 2504 htonl(arc4random()); 2505 else 2506 break; 2507 if (rmask->addr32[1] != 0xffffffff) 2508 rpool->counter.addr32[1] = 2509 htonl(arc4random()); 2510 else 2511 break; 2512 if (rmask->addr32[0] != 0xffffffff) 2513 rpool->counter.addr32[0] = 2514 htonl(arc4random()); 2515 break; 2516#endif /* INET6 */ 2517 } 2518 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af); 2519 PF_ACPY(init_addr, naddr, af); 2520 2521 } else { 2522 PF_AINC(&rpool->counter, af); 2523 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af); 2524 } 2525 break; 2526 case PF_POOL_SRCHASH: 2527 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af); 2528 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af); 2529 break; 2530 case PF_POOL_ROUNDROBIN: 2531 if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2532 if (!pfr_pool_get(rpool->cur->addr.p.tbl, 2533 &rpool->tblidx, &rpool->counter, 2534 &raddr, &rmask, af)) 2535 goto get_addr; 2536 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2537 if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, 2538 &rpool->tblidx, &rpool->counter, 2539 &raddr, &rmask, af)) 2540 goto get_addr; 2541 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af)) 2542 goto get_addr; 2543 2544 try_next: 2545 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL) 2546 rpool->cur = TAILQ_FIRST(&rpool->list); 2547 if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2548 rpool->tblidx = -1; 2549 if (pfr_pool_get(rpool->cur->addr.p.tbl, 2550 &rpool->tblidx, &rpool->counter, 2551 &raddr, &rmask, af)) { 2552 /* table contains no address of type 'af' */ 2553 if (rpool->cur != acur) 2554 goto try_next; 2555 return (1); 2556 } 2557 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2558 rpool->tblidx = -1; 2559 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, 2560 &rpool->tblidx, &rpool->counter, 2561 &raddr, &rmask, af)) { 2562 /* table contains no address of type 'af' */ 2563 if (rpool->cur != acur) 2564 goto try_next; 2565 return (1); 2566 } 2567 } else { 2568 raddr = &rpool->cur->addr.v.a.addr; 2569 rmask = &rpool->cur->addr.v.a.mask; 2570 PF_ACPY(&rpool->counter, raddr, af); 2571 } 2572 2573 get_addr: 2574 PF_ACPY(naddr, &rpool->counter, af); 2575 if (init_addr != NULL && PF_AZERO(init_addr, af)) 2576 PF_ACPY(init_addr, naddr, af); 2577 PF_AINC(&rpool->counter, af); 2578 break; 2579 } 2580 if (*sn != NULL) 2581 PF_ACPY(&(*sn)->raddr, naddr, af); 2582 2583 if (pf_status.debug >= PF_DEBUG_MISC && 2584 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { 2585 printf("pf_map_addr: selected address "); 2586 pf_print_host(naddr, 0, af); 2587 printf("\n"); 2588 } 2589 2590 return (0); 2591} 2592 2593int 2594pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r, 2595 struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport, 2596 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high, 2597 struct pf_src_node **sn) 2598{ 2599 struct pf_state_cmp key; 2600 struct pf_addr init_addr; 2601 u_int16_t cut; 2602 2603 bzero(&init_addr, sizeof(init_addr)); 2604 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn)) 2605 return (1); 2606 2607 if (proto == IPPROTO_ICMP) { 2608 low = 1; 2609 high = 65535; 2610 } 2611 2612 do { 2613 key.af = af; 2614 key.proto = proto; 2615 PF_ACPY(&key.ext.addr, daddr, key.af); 2616 PF_ACPY(&key.gwy.addr, naddr, key.af); 2617 key.ext.port = dport; 2618 2619 /* 2620 * port search; start random, step; 2621 * similar 2 portloop in in_pcbbind 2622 */ 2623 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP || 2624 proto == IPPROTO_ICMP)) { 2625 key.gwy.port = dport; 2626 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) 2627 return (0); 2628 } else if (low == 0 && high == 0) { 2629 key.gwy.port = *nport; 2630 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) 2631 return (0); 2632 } else if (low == high) { 2633 key.gwy.port = htons(low); 2634 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) { 2635 *nport = htons(low); 2636 return (0); 2637 } 2638 } else { 2639 u_int16_t tmp; 2640 2641 if (low > high) { 2642 tmp = low; 2643 low = high; 2644 high = tmp; 2645 } 2646 /* low < high */ 2647 cut = htonl(arc4random()) % (1 + high - low) + low; 2648 /* low <= cut <= high */ 2649 for (tmp = cut; tmp <= high; ++(tmp)) { 2650 key.gwy.port = htons(tmp); 2651 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == 2652 NULL) { 2653 *nport = htons(tmp); 2654 return (0); 2655 } 2656 } 2657 for (tmp = cut - 1; tmp >= low; --(tmp)) { 2658 key.gwy.port = htons(tmp); 2659 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == 2660 NULL) { 2661 *nport = htons(tmp); 2662 return (0); 2663 } 2664 } 2665 } 2666 2667 switch (r->rpool.opts & PF_POOL_TYPEMASK) { 2668 case PF_POOL_RANDOM: 2669 case PF_POOL_ROUNDROBIN: 2670 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn)) 2671 return (1); 2672 break; 2673 case PF_POOL_NONE: 2674 case PF_POOL_SRCHASH: 2675 case PF_POOL_BITMASK: 2676 default: 2677 return (1); 2678 } 2679 } while (! PF_AEQ(&init_addr, naddr, af) ); 2680 2681 return (1); /* none available */ 2682} 2683 2684struct pf_rule * 2685pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off, 2686 int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport, 2687 struct pf_addr *daddr, u_int16_t dport, int rs_num) 2688{ 2689 struct pf_rule *r, *rm = NULL; 2690 struct pf_ruleset *ruleset = NULL; 2691 int tag = -1; 2692 int rtableid = -1; 2693 int asd = 0; 2694 2695 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr); 2696 while (r && rm == NULL) { 2697 struct pf_rule_addr *src = NULL, *dst = NULL; 2698 struct pf_addr_wrap *xdst = NULL; 2699 2700 if (r->action == PF_BINAT && direction == PF_IN) { 2701 src = &r->dst; 2702 if (r->rpool.cur != NULL) 2703 xdst = &r->rpool.cur->addr; 2704 } else { 2705 src = &r->src; 2706 dst = &r->dst; 2707 } 2708 2709 r->evaluations++; 2710 if (pfi_kif_match(r->kif, kif) == r->ifnot) 2711 r = r->skip[PF_SKIP_IFP].ptr; 2712 else if (r->direction && r->direction != direction) 2713 r = r->skip[PF_SKIP_DIR].ptr; 2714 else if (r->af && r->af != pd->af) 2715 r = r->skip[PF_SKIP_AF].ptr; 2716 else if (r->proto && r->proto != pd->proto) 2717 r = r->skip[PF_SKIP_PROTO].ptr; 2718 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af, 2719 src->neg, kif)) 2720 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR : 2721 PF_SKIP_DST_ADDR].ptr; 2722 else if (src->port_op && !pf_match_port(src->port_op, 2723 src->port[0], src->port[1], sport)) 2724 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT : 2725 PF_SKIP_DST_PORT].ptr; 2726 else if (dst != NULL && 2727 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL)) 2728 r = r->skip[PF_SKIP_DST_ADDR].ptr; 2729 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af, 2730 0, NULL)) 2731 r = TAILQ_NEXT(r, entries); 2732 else if (dst != NULL && dst->port_op && 2733 !pf_match_port(dst->port_op, dst->port[0], 2734 dst->port[1], dport)) 2735 r = r->skip[PF_SKIP_DST_PORT].ptr; 2736 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) 2737 r = TAILQ_NEXT(r, entries); 2738 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto != 2739 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m, 2740 off, pd->hdr.tcp), r->os_fingerprint))) 2741 r = TAILQ_NEXT(r, entries); 2742 else { 2743 if (r->tag) 2744 tag = r->tag; 2745 if (r->rtableid >= 0) 2746 rtableid = r->rtableid; 2747 if (r->anchor == NULL) { 2748 rm = r; 2749 } else 2750 pf_step_into_anchor(&asd, &ruleset, rs_num, 2751 &r, NULL, NULL); 2752 } 2753 if (r == NULL) 2754 pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r, 2755 NULL, NULL); 2756 } 2757 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) 2758 return (NULL); 2759 if (rm != NULL && (rm->action == PF_NONAT || 2760 rm->action == PF_NORDR || rm->action == PF_NOBINAT)) 2761 return (NULL); 2762 return (rm); 2763} 2764 2765struct pf_rule * 2766pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction, 2767 struct pfi_kif *kif, struct pf_src_node **sn, 2768 struct pf_addr *saddr, u_int16_t sport, 2769 struct pf_addr *daddr, u_int16_t dport, 2770 struct pf_addr *naddr, u_int16_t *nport) 2771{ 2772 struct pf_rule *r = NULL; 2773 2774 if (direction == PF_OUT) { 2775 r = pf_match_translation(pd, m, off, direction, kif, saddr, 2776 sport, daddr, dport, PF_RULESET_BINAT); 2777 if (r == NULL) 2778 r = pf_match_translation(pd, m, off, direction, kif, 2779 saddr, sport, daddr, dport, PF_RULESET_NAT); 2780 } else { 2781 r = pf_match_translation(pd, m, off, direction, kif, saddr, 2782 sport, daddr, dport, PF_RULESET_RDR); 2783 if (r == NULL) 2784 r = pf_match_translation(pd, m, off, direction, kif, 2785 saddr, sport, daddr, dport, PF_RULESET_BINAT); 2786 } 2787 2788 if (r != NULL) { 2789 switch (r->action) { 2790 case PF_NONAT: 2791 case PF_NOBINAT: 2792 case PF_NORDR: 2793 return (NULL); 2794 case PF_NAT: 2795 if (pf_get_sport(pd->af, pd->proto, r, saddr, 2796 daddr, dport, naddr, nport, r->rpool.proxy_port[0], 2797 r->rpool.proxy_port[1], sn)) { 2798 DPFPRINTF(PF_DEBUG_MISC, 2799 ("pf: NAT proxy port allocation " 2800 "(%u-%u) failed\n", 2801 r->rpool.proxy_port[0], 2802 r->rpool.proxy_port[1])); 2803 return (NULL); 2804 } 2805 break; 2806 case PF_BINAT: 2807 switch (direction) { 2808 case PF_OUT: 2809 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){ 2810 switch (pd->af) { 2811#ifdef INET 2812 case AF_INET: 2813 if (r->rpool.cur->addr.p.dyn-> 2814 pfid_acnt4 < 1) 2815 return (NULL); 2816 PF_POOLMASK(naddr, 2817 &r->rpool.cur->addr.p.dyn-> 2818 pfid_addr4, 2819 &r->rpool.cur->addr.p.dyn-> 2820 pfid_mask4, 2821 saddr, AF_INET); 2822 break; 2823#endif /* INET */ 2824#ifdef INET6 2825 case AF_INET6: 2826 if (r->rpool.cur->addr.p.dyn-> 2827 pfid_acnt6 < 1) 2828 return (NULL); 2829 PF_POOLMASK(naddr, 2830 &r->rpool.cur->addr.p.dyn-> 2831 pfid_addr6, 2832 &r->rpool.cur->addr.p.dyn-> 2833 pfid_mask6, 2834 saddr, AF_INET6); 2835 break; 2836#endif /* INET6 */ 2837 } 2838 } else 2839 PF_POOLMASK(naddr, 2840 &r->rpool.cur->addr.v.a.addr, 2841 &r->rpool.cur->addr.v.a.mask, 2842 saddr, pd->af); 2843 break; 2844 case PF_IN: 2845 if (r->src.addr.type == PF_ADDR_DYNIFTL) { 2846 switch (pd->af) { 2847#ifdef INET 2848 case AF_INET: 2849 if (r->src.addr.p.dyn-> 2850 pfid_acnt4 < 1) 2851 return (NULL); 2852 PF_POOLMASK(naddr, 2853 &r->src.addr.p.dyn-> 2854 pfid_addr4, 2855 &r->src.addr.p.dyn-> 2856 pfid_mask4, 2857 daddr, AF_INET); 2858 break; 2859#endif /* INET */ 2860#ifdef INET6 2861 case AF_INET6: 2862 if (r->src.addr.p.dyn-> 2863 pfid_acnt6 < 1) 2864 return (NULL); 2865 PF_POOLMASK(naddr, 2866 &r->src.addr.p.dyn-> 2867 pfid_addr6, 2868 &r->src.addr.p.dyn-> 2869 pfid_mask6, 2870 daddr, AF_INET6); 2871 break; 2872#endif /* INET6 */ 2873 } 2874 } else 2875 PF_POOLMASK(naddr, 2876 &r->src.addr.v.a.addr, 2877 &r->src.addr.v.a.mask, daddr, 2878 pd->af); 2879 break; 2880 } 2881 break; 2882 case PF_RDR: { 2883 if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn)) 2884 return (NULL); 2885 if ((r->rpool.opts & PF_POOL_TYPEMASK) == 2886 PF_POOL_BITMASK) 2887 PF_POOLMASK(naddr, naddr, 2888 &r->rpool.cur->addr.v.a.mask, daddr, 2889 pd->af); 2890 2891 if (r->rpool.proxy_port[1]) { 2892 u_int32_t tmp_nport; 2893 2894 tmp_nport = ((ntohs(dport) - 2895 ntohs(r->dst.port[0])) % 2896 (r->rpool.proxy_port[1] - 2897 r->rpool.proxy_port[0] + 1)) + 2898 r->rpool.proxy_port[0]; 2899 2900 /* wrap around if necessary */ 2901 if (tmp_nport > 65535) 2902 tmp_nport -= 65535; 2903 *nport = htons((u_int16_t)tmp_nport); 2904 } else if (r->rpool.proxy_port[0]) 2905 *nport = htons(r->rpool.proxy_port[0]); 2906 break; 2907 } 2908 default: 2909 return (NULL); 2910 } 2911 } 2912 2913 return (r); 2914} 2915 2916int 2917#ifdef __FreeBSD__ 2918pf_socket_lookup(int direction, struct pf_pdesc *pd, struct inpcb *inp_arg) 2919#else 2920pf_socket_lookup(int direction, struct pf_pdesc *pd) 2921#endif 2922{ 2923 struct pf_addr *saddr, *daddr; 2924 u_int16_t sport, dport; 2925#ifdef __FreeBSD__ 2926 struct inpcbinfo *pi; 2927#else 2928 struct inpcbtable *tb; 2929#endif 2930 struct inpcb *inp; 2931 2932 if (pd == NULL) 2933 return (-1); 2934 pd->lookup.uid = UID_MAX; 2935 pd->lookup.gid = GID_MAX; 2936 pd->lookup.pid = NO_PID; /* XXX: revisit */ 2937#ifdef __FreeBSD__ 2938 if (inp_arg != NULL) { 2939 INP_LOCK_ASSERT(inp_arg); 2940 if (inp_arg->inp_socket) { 2941 pd->lookup.uid = inp_arg->inp_socket->so_cred->cr_uid; 2942 pd->lookup.gid = 2943 inp_arg->inp_socket->so_cred->cr_groups[0]; 2944 return (1); 2945 } else 2946 return (-1); 2947 } 2948#endif 2949 switch (pd->proto) { 2950 case IPPROTO_TCP: 2951 if (pd->hdr.tcp == NULL) 2952 return (-1); 2953 sport = pd->hdr.tcp->th_sport; 2954 dport = pd->hdr.tcp->th_dport; 2955#ifdef __FreeBSD__ 2956 pi = &tcbinfo; 2957#else 2958 tb = &tcbtable; 2959#endif 2960 break; 2961 case IPPROTO_UDP: 2962 if (pd->hdr.udp == NULL) 2963 return (-1); 2964 sport = pd->hdr.udp->uh_sport; 2965 dport = pd->hdr.udp->uh_dport; 2966#ifdef __FreeBSD__ 2967 pi = &udbinfo; 2968#else 2969 tb = &udbtable; 2970#endif 2971 break; 2972 default: 2973 return (-1); 2974 } 2975 if (direction == PF_IN) { 2976 saddr = pd->src; 2977 daddr = pd->dst; 2978 } else { 2979 u_int16_t p; 2980 2981 p = sport; 2982 sport = dport; 2983 dport = p; 2984 saddr = pd->dst; 2985 daddr = pd->src; 2986 } 2987 switch (pd->af) { 2988#ifdef INET 2989 case AF_INET: 2990#ifdef __FreeBSD__ 2991 INP_INFO_RLOCK(pi); /* XXX LOR */ 2992 inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4, 2993 dport, 0, NULL); 2994 if (inp == NULL) { 2995 inp = in_pcblookup_hash(pi, saddr->v4, sport, 2996 daddr->v4, dport, INPLOOKUP_WILDCARD, NULL); 2997 if(inp == NULL) { 2998 INP_INFO_RUNLOCK(pi); 2999 return (-1); 3000 } 3001 } 3002#else 3003 inp = in_pcbhashlookup(tb, saddr->v4, sport, daddr->v4, dport); 3004 if (inp == NULL) { 3005 inp = in_pcblookup_listen(tb, daddr->v4, dport, 0); 3006 if (inp == NULL) 3007 return (-1); 3008 } 3009#endif 3010 break; 3011#endif /* INET */ 3012#ifdef INET6 3013 case AF_INET6: 3014#ifdef __FreeBSD__ 3015 INP_INFO_RLOCK(pi); 3016 inp = in6_pcblookup_hash(pi, &saddr->v6, sport, 3017 &daddr->v6, dport, 0, NULL); 3018 if (inp == NULL) { 3019 inp = in6_pcblookup_hash(pi, &saddr->v6, sport, 3020 &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL); 3021 if (inp == NULL) { 3022 INP_INFO_RUNLOCK(pi); 3023 return (-1); 3024 } 3025 } 3026#else 3027 inp = in6_pcbhashlookup(tb, &saddr->v6, sport, &daddr->v6, 3028 dport); 3029 if (inp == NULL) { 3030 inp = in6_pcblookup_listen(tb, &daddr->v6, dport, 0); 3031 if (inp == NULL) 3032 return (-1); 3033 } 3034#endif 3035 break; 3036#endif /* INET6 */ 3037 3038 default: 3039 return (-1); 3040 } 3041#ifdef __FreeBSD__ 3042 INP_RLOCK(inp); 3043 INP_INFO_RUNLOCK(pi); 3044 if ((inp->inp_socket == NULL) || (inp->inp_socket->so_cred == NULL)) { 3045 INP_RUNLOCK(inp); 3046 return (-1); 3047 } 3048 pd->lookup.uid = inp->inp_socket->so_cred->cr_uid; 3049 pd->lookup.gid = inp->inp_socket->so_cred->cr_groups[0]; 3050 INP_RUNLOCK(inp); 3051#else 3052 pd->lookup.uid = inp->inp_socket->so_euid; 3053 pd->lookup.gid = inp->inp_socket->so_egid; 3054 pd->lookup.pid = inp->inp_socket->so_cpid; 3055#endif 3056 return (1); 3057} 3058 3059u_int8_t 3060pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af) 3061{ 3062 int hlen; 3063 u_int8_t hdr[60]; 3064 u_int8_t *opt, optlen; 3065 u_int8_t wscale = 0; 3066 3067 hlen = th_off << 2; /* hlen <= sizeof(hdr) */ 3068 if (hlen <= sizeof(struct tcphdr)) 3069 return (0); 3070 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af)) 3071 return (0); 3072 opt = hdr + sizeof(struct tcphdr); 3073 hlen -= sizeof(struct tcphdr); 3074 while (hlen >= 3) { 3075 switch (*opt) { 3076 case TCPOPT_EOL: 3077 case TCPOPT_NOP: 3078 ++opt; 3079 --hlen; 3080 break; 3081 case TCPOPT_WINDOW: 3082 wscale = opt[2]; 3083 if (wscale > TCP_MAX_WINSHIFT) 3084 wscale = TCP_MAX_WINSHIFT; 3085 wscale |= PF_WSCALE_FLAG; 3086 /* FALLTHROUGH */ 3087 default: 3088 optlen = opt[1]; 3089 if (optlen < 2) 3090 optlen = 2; 3091 hlen -= optlen; 3092 opt += optlen; 3093 break; 3094 } 3095 } 3096 return (wscale); 3097} 3098 3099u_int16_t 3100pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af) 3101{ 3102 int hlen; 3103 u_int8_t hdr[60]; 3104 u_int8_t *opt, optlen; 3105 u_int16_t mss = tcp_mssdflt; 3106 3107 hlen = th_off << 2; /* hlen <= sizeof(hdr) */ 3108 if (hlen <= sizeof(struct tcphdr)) 3109 return (0); 3110 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af)) 3111 return (0); 3112 opt = hdr + sizeof(struct tcphdr); 3113 hlen -= sizeof(struct tcphdr); 3114 while (hlen >= TCPOLEN_MAXSEG) { 3115 switch (*opt) { 3116 case TCPOPT_EOL: 3117 case TCPOPT_NOP: 3118 ++opt; 3119 --hlen; 3120 break; 3121 case TCPOPT_MAXSEG: 3122 bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2); 3123 NTOHS(mss); 3124 /* FALLTHROUGH */ 3125 default: 3126 optlen = opt[1]; 3127 if (optlen < 2) 3128 optlen = 2; 3129 hlen -= optlen; 3130 opt += optlen; 3131 break; 3132 } 3133 } 3134 return (mss); 3135} 3136 3137u_int16_t 3138pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer) 3139{ 3140#ifdef INET 3141 struct sockaddr_in *dst; 3142 struct route ro; 3143#endif /* INET */ 3144#ifdef INET6 3145 struct sockaddr_in6 *dst6; 3146 struct route_in6 ro6; 3147#endif /* INET6 */ 3148 struct rtentry *rt = NULL; 3149 int hlen = 0; /* make the compiler happy */ 3150 u_int16_t mss = tcp_mssdflt; 3151 3152 switch (af) { 3153#ifdef INET 3154 case AF_INET: 3155 hlen = sizeof(struct ip); 3156 bzero(&ro, sizeof(ro)); 3157 dst = (struct sockaddr_in *)&ro.ro_dst; 3158 dst->sin_family = AF_INET; 3159 dst->sin_len = sizeof(*dst); 3160 dst->sin_addr = addr->v4; 3161#ifdef __FreeBSD__ 3162#ifdef RTF_PRCLONING 3163 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING)); 3164#else /* !RTF_PRCLONING */ 3165 in_rtalloc_ign(&ro, RTF_CLONING, 0); 3166#endif 3167#else /* ! __FreeBSD__ */ 3168 rtalloc_noclone(&ro, NO_CLONING); 3169#endif 3170 rt = ro.ro_rt; 3171 break; 3172#endif /* INET */ 3173#ifdef INET6 3174 case AF_INET6: 3175 hlen = sizeof(struct ip6_hdr); 3176 bzero(&ro6, sizeof(ro6)); 3177 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst; 3178 dst6->sin6_family = AF_INET6; 3179 dst6->sin6_len = sizeof(*dst6); 3180 dst6->sin6_addr = addr->v6; 3181#ifdef __FreeBSD__ 3182#ifdef RTF_PRCLONING 3183 rtalloc_ign((struct route *)&ro6, 3184 (RTF_CLONING | RTF_PRCLONING)); 3185#else /* !RTF_PRCLONING */ 3186 rtalloc_ign((struct route *)&ro6, RTF_CLONING); 3187#endif 3188#else /* ! __FreeBSD__ */ 3189 rtalloc_noclone((struct route *)&ro6, NO_CLONING); 3190#endif 3191 rt = ro6.ro_rt; 3192 break; 3193#endif /* INET6 */ 3194 } 3195 3196 if (rt && rt->rt_ifp) { 3197 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr); 3198 mss = max(tcp_mssdflt, mss); 3199 RTFREE(rt); 3200 } 3201 mss = min(mss, offer); 3202 mss = max(mss, 64); /* sanity - at least max opt space */ 3203 return (mss); 3204} 3205 3206void 3207pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr) 3208{ 3209 struct pf_rule *r = s->rule.ptr; 3210 3211 s->rt_kif = NULL; 3212 if (!r->rt || r->rt == PF_FASTROUTE) 3213 return; 3214 switch (s->af) { 3215#ifdef INET 3216 case AF_INET: 3217 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL, 3218 &s->nat_src_node); 3219 s->rt_kif = r->rpool.cur->kif; 3220 break; 3221#endif /* INET */ 3222#ifdef INET6 3223 case AF_INET6: 3224 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL, 3225 &s->nat_src_node); 3226 s->rt_kif = r->rpool.cur->kif; 3227 break; 3228#endif /* INET6 */ 3229 } 3230} 3231 3232int 3233pf_test_tcp(struct pf_rule **rm, struct pf_state **sm, int direction, 3234 struct pfi_kif *kif, struct mbuf *m, int off, void *h, 3235#ifdef __FreeBSD__ 3236 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3237 struct ifqueue *ifq, struct inpcb *inp) 3238#else 3239 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3240 struct ifqueue *ifq) 3241#endif 3242{ 3243 struct pf_rule *nr = NULL; 3244 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 3245 struct tcphdr *th = pd->hdr.tcp; 3246 u_int16_t bport, nport = 0; 3247 sa_family_t af = pd->af; 3248 struct pf_rule *r, *a = NULL; 3249 struct pf_ruleset *ruleset = NULL; 3250 struct pf_src_node *nsn = NULL; 3251 u_short reason; 3252 int rewrite = 0; 3253 int tag = -1, rtableid = -1; 3254 u_int16_t mss = tcp_mssdflt; 3255 int asd = 0; 3256 int match = 0; 3257 3258 if (pf_check_congestion(ifq)) { 3259 REASON_SET(&reason, PFRES_CONGEST); 3260 return (PF_DROP); 3261 } 3262 3263#ifdef __FreeBSD__ 3264 if (inp != NULL) 3265 pd->lookup.done = pf_socket_lookup(direction, pd, inp); 3266 else if (debug_pfugidhack) { 3267 PF_UNLOCK(); 3268 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n")); 3269 pd->lookup.done = pf_socket_lookup(direction, pd, inp); 3270 PF_LOCK(); 3271 } 3272#endif 3273 3274 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 3275 3276 if (direction == PF_OUT) { 3277 bport = nport = th->th_sport; 3278 /* check outgoing packet for BINAT/NAT */ 3279 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, 3280 saddr, th->th_sport, daddr, th->th_dport, 3281 &pd->naddr, &nport)) != NULL) { 3282 PF_ACPY(&pd->baddr, saddr, af); 3283 pf_change_ap(saddr, &th->th_sport, pd->ip_sum, 3284 &th->th_sum, &pd->naddr, nport, 0, af); 3285 rewrite++; 3286 if (nr->natpass) 3287 r = NULL; 3288 pd->nat_rule = nr; 3289 } 3290 } else { 3291 bport = nport = th->th_dport; 3292 /* check incoming packet for BINAT/RDR */ 3293 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, 3294 saddr, th->th_sport, daddr, th->th_dport, 3295 &pd->naddr, &nport)) != NULL) { 3296 PF_ACPY(&pd->baddr, daddr, af); 3297 pf_change_ap(daddr, &th->th_dport, pd->ip_sum, 3298 &th->th_sum, &pd->naddr, nport, 0, af); 3299 rewrite++; 3300 if (nr->natpass) 3301 r = NULL; 3302 pd->nat_rule = nr; 3303 } 3304 } 3305 3306 while (r != NULL) { 3307 r->evaluations++; 3308 if (pfi_kif_match(r->kif, kif) == r->ifnot) 3309 r = r->skip[PF_SKIP_IFP].ptr; 3310 else if (r->direction && r->direction != direction) 3311 r = r->skip[PF_SKIP_DIR].ptr; 3312 else if (r->af && r->af != af) 3313 r = r->skip[PF_SKIP_AF].ptr; 3314 else if (r->proto && r->proto != IPPROTO_TCP) 3315 r = r->skip[PF_SKIP_PROTO].ptr; 3316 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, 3317 r->src.neg, kif)) 3318 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 3319 else if (r->src.port_op && !pf_match_port(r->src.port_op, 3320 r->src.port[0], r->src.port[1], th->th_sport)) 3321 r = r->skip[PF_SKIP_SRC_PORT].ptr; 3322 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, 3323 r->dst.neg, NULL)) 3324 r = r->skip[PF_SKIP_DST_ADDR].ptr; 3325 else if (r->dst.port_op && !pf_match_port(r->dst.port_op, 3326 r->dst.port[0], r->dst.port[1], th->th_dport)) 3327 r = r->skip[PF_SKIP_DST_PORT].ptr; 3328 else if (r->tos && !(r->tos == pd->tos)) 3329 r = TAILQ_NEXT(r, entries); 3330 else if (r->rule_flag & PFRULE_FRAGMENT) 3331 r = TAILQ_NEXT(r, entries); 3332 else if ((r->flagset & th->th_flags) != r->flags) 3333 r = TAILQ_NEXT(r, entries); 3334 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done = 3335#ifdef __FreeBSD__ 3336 pf_socket_lookup(direction, pd, inp), 1)) && 3337#else 3338 pf_socket_lookup(direction, pd), 1)) && 3339#endif 3340 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1], 3341 pd->lookup.uid)) 3342 r = TAILQ_NEXT(r, entries); 3343 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done = 3344#ifdef __FreeBSD__ 3345 pf_socket_lookup(direction, pd, inp), 1)) && 3346#else 3347 pf_socket_lookup(direction, pd), 1)) && 3348#endif 3349 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1], 3350 pd->lookup.gid)) 3351 r = TAILQ_NEXT(r, entries); 3352 else if (r->prob && r->prob <= arc4random()) 3353 r = TAILQ_NEXT(r, entries); 3354 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) 3355 r = TAILQ_NEXT(r, entries); 3356 else if (r->os_fingerprint != PF_OSFP_ANY && !pf_osfp_match( 3357 pf_osfp_fingerprint(pd, m, off, th), r->os_fingerprint)) 3358 r = TAILQ_NEXT(r, entries); 3359 else { 3360 if (r->tag) 3361 tag = r->tag; 3362 if (r->rtableid >= 0) 3363 rtableid = r->rtableid; 3364 if (r->anchor == NULL) { 3365 match = 1; 3366 *rm = r; 3367 *am = a; 3368 *rsm = ruleset; 3369 if ((*rm)->quick) 3370 break; 3371 r = TAILQ_NEXT(r, entries); 3372 } else 3373 pf_step_into_anchor(&asd, &ruleset, 3374 PF_RULESET_FILTER, &r, &a, &match); 3375 } 3376 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 3377 PF_RULESET_FILTER, &r, &a, &match)) 3378 break; 3379 } 3380 r = *rm; 3381 a = *am; 3382 ruleset = *rsm; 3383 3384 REASON_SET(&reason, PFRES_MATCH); 3385 3386 if (r->log || (nr != NULL && nr->natpass && nr->log)) { 3387 if (rewrite) 3388#ifdef __FreeBSD__ 3389 m_copyback(m, off, sizeof(*th), (caddr_t)th); 3390#else 3391 m_copyback(m, off, sizeof(*th), th); 3392#endif 3393 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, 3394 a, ruleset, pd); 3395 } 3396 3397 if ((r->action == PF_DROP) && 3398 ((r->rule_flag & PFRULE_RETURNRST) || 3399 (r->rule_flag & PFRULE_RETURNICMP) || 3400 (r->rule_flag & PFRULE_RETURN))) { 3401 /* undo NAT changes, if they have taken place */ 3402 if (nr != NULL) { 3403 if (direction == PF_OUT) { 3404 pf_change_ap(saddr, &th->th_sport, pd->ip_sum, 3405 &th->th_sum, &pd->baddr, bport, 0, af); 3406 rewrite++; 3407 } else { 3408 pf_change_ap(daddr, &th->th_dport, pd->ip_sum, 3409 &th->th_sum, &pd->baddr, bport, 0, af); 3410 rewrite++; 3411 } 3412 } 3413 if (((r->rule_flag & PFRULE_RETURNRST) || 3414 (r->rule_flag & PFRULE_RETURN)) && 3415 !(th->th_flags & TH_RST)) { 3416 u_int32_t ack = ntohl(th->th_seq) + pd->p_len; 3417 3418 if (th->th_flags & TH_SYN) 3419 ack++; 3420 if (th->th_flags & TH_FIN) 3421 ack++; 3422#ifdef __FreeBSD__ 3423 pf_send_tcp(m, r, af, pd->dst, 3424#else 3425 pf_send_tcp(r, af, pd->dst, 3426#endif 3427 pd->src, th->th_dport, th->th_sport, 3428 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0, 3429 r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp); 3430 } else if ((af == AF_INET) && r->return_icmp) 3431 pf_send_icmp(m, r->return_icmp >> 8, 3432 r->return_icmp & 255, af, r); 3433 else if ((af == AF_INET6) && r->return_icmp6) 3434 pf_send_icmp(m, r->return_icmp6 >> 8, 3435 r->return_icmp6 & 255, af, r); 3436 } 3437 3438 if (r->action == PF_DROP) 3439 return (PF_DROP); 3440 3441 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) { 3442 REASON_SET(&reason, PFRES_MEMORY); 3443 return (PF_DROP); 3444 } 3445 3446 if (r->keep_state || nr != NULL || 3447 (pd->flags & PFDESC_TCP_NORM)) { 3448 /* create new state */ 3449 u_int16_t len; 3450 struct pf_state *s = NULL; 3451 struct pf_src_node *sn = NULL; 3452 3453 len = pd->tot_len - off - (th->th_off << 2); 3454 3455 /* check maximums */ 3456 if (r->max_states && (r->states >= r->max_states)) { 3457 pf_status.lcounters[LCNT_STATES]++; 3458 REASON_SET(&reason, PFRES_MAXSTATES); 3459 goto cleanup; 3460 } 3461 /* src node for filter rule */ 3462 if ((r->rule_flag & PFRULE_SRCTRACK || 3463 r->rpool.opts & PF_POOL_STICKYADDR) && 3464 pf_insert_src_node(&sn, r, saddr, af) != 0) { 3465 REASON_SET(&reason, PFRES_SRCLIMIT); 3466 goto cleanup; 3467 } 3468 /* src node for translation rule */ 3469 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 3470 ((direction == PF_OUT && 3471 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || 3472 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { 3473 REASON_SET(&reason, PFRES_SRCLIMIT); 3474 goto cleanup; 3475 } 3476 s = pool_get(&pf_state_pl, PR_NOWAIT); 3477 if (s == NULL) { 3478 REASON_SET(&reason, PFRES_MEMORY); 3479cleanup: 3480 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 3481 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 3482 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3483 pf_status.src_nodes--; 3484 pool_put(&pf_src_tree_pl, sn); 3485 } 3486 if (nsn != sn && nsn != NULL && nsn->states == 0 && 3487 nsn->expire == 0) { 3488 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 3489 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3490 pf_status.src_nodes--; 3491 pool_put(&pf_src_tree_pl, nsn); 3492 } 3493 return (PF_DROP); 3494 } 3495 bzero(s, sizeof(*s)); 3496 s->rule.ptr = r; 3497 s->nat_rule.ptr = nr; 3498 s->anchor.ptr = a; 3499 STATE_INC_COUNTERS(s); 3500 s->allow_opts = r->allow_opts; 3501 s->log = r->log & PF_LOG_ALL; 3502 if (nr != NULL) 3503 s->log |= nr->log & PF_LOG_ALL; 3504 s->proto = IPPROTO_TCP; 3505 s->direction = direction; 3506 s->af = af; 3507 if (direction == PF_OUT) { 3508 PF_ACPY(&s->gwy.addr, saddr, af); 3509 s->gwy.port = th->th_sport; /* sport */ 3510 PF_ACPY(&s->ext.addr, daddr, af); 3511 s->ext.port = th->th_dport; 3512 if (nr != NULL) { 3513 PF_ACPY(&s->lan.addr, &pd->baddr, af); 3514 s->lan.port = bport; 3515 } else { 3516 PF_ACPY(&s->lan.addr, &s->gwy.addr, af); 3517 s->lan.port = s->gwy.port; 3518 } 3519 } else { 3520 PF_ACPY(&s->lan.addr, daddr, af); 3521 s->lan.port = th->th_dport; 3522 PF_ACPY(&s->ext.addr, saddr, af); 3523 s->ext.port = th->th_sport; 3524 if (nr != NULL) { 3525 PF_ACPY(&s->gwy.addr, &pd->baddr, af); 3526 s->gwy.port = bport; 3527 } else { 3528 PF_ACPY(&s->gwy.addr, &s->lan.addr, af); 3529 s->gwy.port = s->lan.port; 3530 } 3531 } 3532 3533 s->src.seqlo = ntohl(th->th_seq); 3534 s->src.seqhi = s->src.seqlo + len + 1; 3535 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN && 3536 r->keep_state == PF_STATE_MODULATE) { 3537 /* Generate sequence number modulator */ 3538#ifdef __FreeBSD__ 3539 while ((s->src.seqdiff = 3540 pf_new_isn(s) - s->src.seqlo) == 0) 3541 ; 3542#else 3543 while ((s->src.seqdiff = 3544 tcp_rndiss_next() - s->src.seqlo) == 0) 3545 ; 3546#endif 3547 pf_change_a(&th->th_seq, &th->th_sum, 3548 htonl(s->src.seqlo + s->src.seqdiff), 0); 3549 rewrite = 1; 3550 } else 3551 s->src.seqdiff = 0; 3552 if (th->th_flags & TH_SYN) { 3553 s->src.seqhi++; 3554 s->src.wscale = pf_get_wscale(m, off, th->th_off, af); 3555 } 3556 s->src.max_win = MAX(ntohs(th->th_win), 1); 3557 if (s->src.wscale & PF_WSCALE_MASK) { 3558 /* Remove scale factor from initial window */ 3559 int win = s->src.max_win; 3560 win += 1 << (s->src.wscale & PF_WSCALE_MASK); 3561 s->src.max_win = (win - 1) >> 3562 (s->src.wscale & PF_WSCALE_MASK); 3563 } 3564 if (th->th_flags & TH_FIN) 3565 s->src.seqhi++; 3566 s->dst.seqhi = 1; 3567 s->dst.max_win = 1; 3568 s->src.state = TCPS_SYN_SENT; 3569 s->dst.state = TCPS_CLOSED; 3570 s->creation = time_second; 3571 s->expire = time_second; 3572 s->timeout = PFTM_TCP_FIRST_PACKET; 3573 pf_set_rt_ifp(s, saddr); 3574 if (sn != NULL) { 3575 s->src_node = sn; 3576 s->src_node->states++; 3577 } 3578 if (nsn != NULL) { 3579 PF_ACPY(&nsn->raddr, &pd->naddr, af); 3580 s->nat_src_node = nsn; 3581 s->nat_src_node->states++; 3582 } 3583 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m, 3584 off, pd, th, &s->src, &s->dst)) { 3585 REASON_SET(&reason, PFRES_MEMORY); 3586 pf_src_tree_remove_state(s); 3587 STATE_DEC_COUNTERS(s); 3588 pool_put(&pf_state_pl, s); 3589 return (PF_DROP); 3590 } 3591 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub && 3592 pf_normalize_tcp_stateful(m, off, pd, &reason, th, s, 3593 &s->src, &s->dst, &rewrite)) { 3594 /* This really shouldn't happen!!! */ 3595 DPFPRINTF(PF_DEBUG_URGENT, 3596 ("pf_normalize_tcp_stateful failed on first pkt")); 3597 pf_normalize_tcp_cleanup(s); 3598 pf_src_tree_remove_state(s); 3599 STATE_DEC_COUNTERS(s); 3600 pool_put(&pf_state_pl, s); 3601 return (PF_DROP); 3602 } 3603 if (pf_insert_state(BOUND_IFACE(r, kif), s)) { 3604 pf_normalize_tcp_cleanup(s); 3605 REASON_SET(&reason, PFRES_STATEINS); 3606 pf_src_tree_remove_state(s); 3607 STATE_DEC_COUNTERS(s); 3608 pool_put(&pf_state_pl, s); 3609 return (PF_DROP); 3610 } else 3611 *sm = s; 3612 if (tag > 0) { 3613 pf_tag_ref(tag); 3614 s->tag = tag; 3615 } 3616 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN && 3617 r->keep_state == PF_STATE_SYNPROXY) { 3618 s->src.state = PF_TCPS_PROXY_SRC; 3619 if (nr != NULL) { 3620 if (direction == PF_OUT) { 3621 pf_change_ap(saddr, &th->th_sport, 3622 pd->ip_sum, &th->th_sum, &pd->baddr, 3623 bport, 0, af); 3624 } else { 3625 pf_change_ap(daddr, &th->th_dport, 3626 pd->ip_sum, &th->th_sum, &pd->baddr, 3627 bport, 0, af); 3628 } 3629 } 3630 s->src.seqhi = htonl(arc4random()); 3631 /* Find mss option */ 3632 mss = pf_get_mss(m, off, th->th_off, af); 3633 mss = pf_calc_mss(saddr, af, mss); 3634 mss = pf_calc_mss(daddr, af, mss); 3635 s->src.mss = mss; 3636#ifdef __FreeBSD__ 3637 pf_send_tcp(NULL, r, af, daddr, saddr, th->th_dport, 3638#else 3639 pf_send_tcp(r, af, daddr, saddr, th->th_dport, 3640#endif 3641 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1, 3642 TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL); 3643 REASON_SET(&reason, PFRES_SYNPROXY); 3644 return (PF_SYNPROXY_DROP); 3645 } 3646 } 3647 3648 /* copy back packet headers if we performed NAT operations */ 3649 if (rewrite) 3650 m_copyback(m, off, sizeof(*th), (caddr_t)th); 3651 3652 return (PF_PASS); 3653} 3654 3655int 3656pf_test_udp(struct pf_rule **rm, struct pf_state **sm, int direction, 3657 struct pfi_kif *kif, struct mbuf *m, int off, void *h, 3658#ifdef __FreeBSD__ 3659 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3660 struct ifqueue *ifq, struct inpcb *inp) 3661#else 3662 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3663 struct ifqueue *ifq) 3664#endif 3665{ 3666 struct pf_rule *nr = NULL; 3667 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 3668 struct udphdr *uh = pd->hdr.udp; 3669 u_int16_t bport, nport = 0; 3670 sa_family_t af = pd->af; 3671 struct pf_rule *r, *a = NULL; 3672 struct pf_ruleset *ruleset = NULL; 3673 struct pf_src_node *nsn = NULL; 3674 u_short reason; 3675 int rewrite = 0; 3676 int tag = -1, rtableid = -1; 3677 int asd = 0; 3678 int match = 0; 3679 3680 if (pf_check_congestion(ifq)) { 3681 REASON_SET(&reason, PFRES_CONGEST); 3682 return (PF_DROP); 3683 } 3684 3685#ifdef __FreeBSD__ 3686 if (inp != NULL) 3687 pd->lookup.done = pf_socket_lookup(direction, pd, inp); 3688 else if (debug_pfugidhack) { 3689 PF_UNLOCK(); 3690 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n")); 3691 pd->lookup.done = pf_socket_lookup(direction, pd, inp); 3692 PF_LOCK(); 3693 } 3694#endif 3695 3696 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 3697 3698 if (direction == PF_OUT) { 3699 bport = nport = uh->uh_sport; 3700 /* check outgoing packet for BINAT/NAT */ 3701 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, 3702 saddr, uh->uh_sport, daddr, uh->uh_dport, 3703 &pd->naddr, &nport)) != NULL) { 3704 PF_ACPY(&pd->baddr, saddr, af); 3705 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum, 3706 &uh->uh_sum, &pd->naddr, nport, 1, af); 3707 rewrite++; 3708 if (nr->natpass) 3709 r = NULL; 3710 pd->nat_rule = nr; 3711 } 3712 } else { 3713 bport = nport = uh->uh_dport; 3714 /* check incoming packet for BINAT/RDR */ 3715 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, 3716 saddr, uh->uh_sport, daddr, uh->uh_dport, &pd->naddr, 3717 &nport)) != NULL) { 3718 PF_ACPY(&pd->baddr, daddr, af); 3719 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum, 3720 &uh->uh_sum, &pd->naddr, nport, 1, af); 3721 rewrite++; 3722 if (nr->natpass) 3723 r = NULL; 3724 pd->nat_rule = nr; 3725 } 3726 } 3727 3728 while (r != NULL) { 3729 r->evaluations++; 3730 if (pfi_kif_match(r->kif, kif) == r->ifnot) 3731 r = r->skip[PF_SKIP_IFP].ptr; 3732 else if (r->direction && r->direction != direction) 3733 r = r->skip[PF_SKIP_DIR].ptr; 3734 else if (r->af && r->af != af) 3735 r = r->skip[PF_SKIP_AF].ptr; 3736 else if (r->proto && r->proto != IPPROTO_UDP) 3737 r = r->skip[PF_SKIP_PROTO].ptr; 3738 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, 3739 r->src.neg, kif)) 3740 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 3741 else if (r->src.port_op && !pf_match_port(r->src.port_op, 3742 r->src.port[0], r->src.port[1], uh->uh_sport)) 3743 r = r->skip[PF_SKIP_SRC_PORT].ptr; 3744 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, 3745 r->dst.neg, NULL)) 3746 r = r->skip[PF_SKIP_DST_ADDR].ptr; 3747 else if (r->dst.port_op && !pf_match_port(r->dst.port_op, 3748 r->dst.port[0], r->dst.port[1], uh->uh_dport)) 3749 r = r->skip[PF_SKIP_DST_PORT].ptr; 3750 else if (r->tos && !(r->tos == pd->tos)) 3751 r = TAILQ_NEXT(r, entries); 3752 else if (r->rule_flag & PFRULE_FRAGMENT) 3753 r = TAILQ_NEXT(r, entries); 3754 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done = 3755#ifdef __FreeBSD__ 3756 pf_socket_lookup(direction, pd, inp), 1)) && 3757#else 3758 pf_socket_lookup(direction, pd), 1)) && 3759#endif 3760 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1], 3761 pd->lookup.uid)) 3762 r = TAILQ_NEXT(r, entries); 3763 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done = 3764#ifdef __FreeBSD__ 3765 pf_socket_lookup(direction, pd, inp), 1)) && 3766#else 3767 pf_socket_lookup(direction, pd), 1)) && 3768#endif 3769 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1], 3770 pd->lookup.gid)) 3771 r = TAILQ_NEXT(r, entries); 3772 else if (r->prob && r->prob <= arc4random()) 3773 r = TAILQ_NEXT(r, entries); 3774 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) 3775 r = TAILQ_NEXT(r, entries); 3776 else if (r->os_fingerprint != PF_OSFP_ANY) 3777 r = TAILQ_NEXT(r, entries); 3778 else { 3779 if (r->tag) 3780 tag = r->tag; 3781 if (r->rtableid >= 0) 3782 rtableid = r->rtableid; 3783 if (r->anchor == NULL) { 3784 match = 1; 3785 *rm = r; 3786 *am = a; 3787 *rsm = ruleset; 3788 if ((*rm)->quick) 3789 break; 3790 r = TAILQ_NEXT(r, entries); 3791 } else 3792 pf_step_into_anchor(&asd, &ruleset, 3793 PF_RULESET_FILTER, &r, &a, &match); 3794 } 3795 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 3796 PF_RULESET_FILTER, &r, &a, &match)) 3797 break; 3798 } 3799 r = *rm; 3800 a = *am; 3801 ruleset = *rsm; 3802 3803 REASON_SET(&reason, PFRES_MATCH); 3804 3805 if (r->log || (nr != NULL && nr->natpass && nr->log)) { 3806 if (rewrite) 3807#ifdef __FreeBSD__ 3808 m_copyback(m, off, sizeof(*uh), (caddr_t)uh); 3809#else 3810 m_copyback(m, off, sizeof(*uh), uh); 3811#endif 3812 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, 3813 a, ruleset, pd); 3814 } 3815 3816 if ((r->action == PF_DROP) && 3817 ((r->rule_flag & PFRULE_RETURNICMP) || 3818 (r->rule_flag & PFRULE_RETURN))) { 3819 /* undo NAT changes, if they have taken place */ 3820 if (nr != NULL) { 3821 if (direction == PF_OUT) { 3822 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum, 3823 &uh->uh_sum, &pd->baddr, bport, 1, af); 3824 rewrite++; 3825 } else { 3826 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum, 3827 &uh->uh_sum, &pd->baddr, bport, 1, af); 3828 rewrite++; 3829 } 3830 } 3831 if ((af == AF_INET) && r->return_icmp) 3832 pf_send_icmp(m, r->return_icmp >> 8, 3833 r->return_icmp & 255, af, r); 3834 else if ((af == AF_INET6) && r->return_icmp6) 3835 pf_send_icmp(m, r->return_icmp6 >> 8, 3836 r->return_icmp6 & 255, af, r); 3837 } 3838 3839 if (r->action == PF_DROP) 3840 return (PF_DROP); 3841 3842 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) { 3843 REASON_SET(&reason, PFRES_MEMORY); 3844 return (PF_DROP); 3845 } 3846 3847 if (r->keep_state || nr != NULL) { 3848 /* create new state */ 3849 struct pf_state *s = NULL; 3850 struct pf_src_node *sn = NULL; 3851 3852 /* check maximums */ 3853 if (r->max_states && (r->states >= r->max_states)) { 3854 pf_status.lcounters[LCNT_STATES]++; 3855 REASON_SET(&reason, PFRES_MAXSTATES); 3856 goto cleanup; 3857 } 3858 /* src node for filter rule */ 3859 if ((r->rule_flag & PFRULE_SRCTRACK || 3860 r->rpool.opts & PF_POOL_STICKYADDR) && 3861 pf_insert_src_node(&sn, r, saddr, af) != 0) { 3862 REASON_SET(&reason, PFRES_SRCLIMIT); 3863 goto cleanup; 3864 } 3865 /* src node for translation rule */ 3866 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 3867 ((direction == PF_OUT && 3868 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || 3869 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { 3870 REASON_SET(&reason, PFRES_SRCLIMIT); 3871 goto cleanup; 3872 } 3873 s = pool_get(&pf_state_pl, PR_NOWAIT); 3874 if (s == NULL) { 3875 REASON_SET(&reason, PFRES_MEMORY); 3876cleanup: 3877 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 3878 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 3879 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3880 pf_status.src_nodes--; 3881 pool_put(&pf_src_tree_pl, sn); 3882 } 3883 if (nsn != sn && nsn != NULL && nsn->states == 0 && 3884 nsn->expire == 0) { 3885 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 3886 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3887 pf_status.src_nodes--; 3888 pool_put(&pf_src_tree_pl, nsn); 3889 } 3890 return (PF_DROP); 3891 } 3892 bzero(s, sizeof(*s)); 3893 s->rule.ptr = r; 3894 s->nat_rule.ptr = nr; 3895 s->anchor.ptr = a; 3896 STATE_INC_COUNTERS(s); 3897 s->allow_opts = r->allow_opts; 3898 s->log = r->log & PF_LOG_ALL; 3899 if (nr != NULL) 3900 s->log |= nr->log & PF_LOG_ALL; 3901 s->proto = IPPROTO_UDP; 3902 s->direction = direction; 3903 s->af = af; 3904 if (direction == PF_OUT) { 3905 PF_ACPY(&s->gwy.addr, saddr, af); 3906 s->gwy.port = uh->uh_sport; 3907 PF_ACPY(&s->ext.addr, daddr, af); 3908 s->ext.port = uh->uh_dport; 3909 if (nr != NULL) { 3910 PF_ACPY(&s->lan.addr, &pd->baddr, af); 3911 s->lan.port = bport; 3912 } else { 3913 PF_ACPY(&s->lan.addr, &s->gwy.addr, af); 3914 s->lan.port = s->gwy.port; 3915 } 3916 } else { 3917 PF_ACPY(&s->lan.addr, daddr, af); 3918 s->lan.port = uh->uh_dport; 3919 PF_ACPY(&s->ext.addr, saddr, af); 3920 s->ext.port = uh->uh_sport; 3921 if (nr != NULL) { 3922 PF_ACPY(&s->gwy.addr, &pd->baddr, af); 3923 s->gwy.port = bport; 3924 } else { 3925 PF_ACPY(&s->gwy.addr, &s->lan.addr, af); 3926 s->gwy.port = s->lan.port; 3927 } 3928 } 3929 s->src.state = PFUDPS_SINGLE; 3930 s->dst.state = PFUDPS_NO_TRAFFIC; 3931 s->creation = time_second; 3932 s->expire = time_second; 3933 s->timeout = PFTM_UDP_FIRST_PACKET; 3934 pf_set_rt_ifp(s, saddr); 3935 if (sn != NULL) { 3936 s->src_node = sn; 3937 s->src_node->states++; 3938 } 3939 if (nsn != NULL) { 3940 PF_ACPY(&nsn->raddr, &pd->naddr, af); 3941 s->nat_src_node = nsn; 3942 s->nat_src_node->states++; 3943 } 3944 if (pf_insert_state(BOUND_IFACE(r, kif), s)) { 3945 REASON_SET(&reason, PFRES_STATEINS); 3946 pf_src_tree_remove_state(s); 3947 STATE_DEC_COUNTERS(s); 3948 pool_put(&pf_state_pl, s); 3949 return (PF_DROP); 3950 } else 3951 *sm = s; 3952 if (tag > 0) { 3953 pf_tag_ref(tag); 3954 s->tag = tag; 3955 } 3956 } 3957 3958 /* copy back packet headers if we performed NAT operations */ 3959 if (rewrite) 3960 m_copyback(m, off, sizeof(*uh), (caddr_t)uh); 3961 3962 return (PF_PASS); 3963} 3964 3965int 3966pf_test_icmp(struct pf_rule **rm, struct pf_state **sm, int direction, 3967 struct pfi_kif *kif, struct mbuf *m, int off, void *h, 3968 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3969 struct ifqueue *ifq) 3970{ 3971 struct pf_rule *nr = NULL; 3972 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 3973 struct pf_rule *r, *a = NULL; 3974 struct pf_ruleset *ruleset = NULL; 3975 struct pf_src_node *nsn = NULL; 3976 u_short reason; 3977 u_int16_t icmpid = 0, bport, nport = 0; 3978 sa_family_t af = pd->af; 3979 u_int8_t icmptype = 0; /* make the compiler happy */ 3980 u_int8_t icmpcode = 0; /* make the compiler happy */ 3981 int state_icmp = 0; 3982 int tag = -1, rtableid = -1; 3983#ifdef INET6 3984 int rewrite = 0; 3985#endif /* INET6 */ 3986 int asd = 0; 3987 int match = 0; 3988 3989 if (pf_check_congestion(ifq)) { 3990 REASON_SET(&reason, PFRES_CONGEST); 3991 return (PF_DROP); 3992 } 3993 3994 switch (pd->proto) { 3995#ifdef INET 3996 case IPPROTO_ICMP: 3997 icmptype = pd->hdr.icmp->icmp_type; 3998 icmpcode = pd->hdr.icmp->icmp_code; 3999 icmpid = pd->hdr.icmp->icmp_id; 4000 4001 if (icmptype == ICMP_UNREACH || 4002 icmptype == ICMP_SOURCEQUENCH || 4003 icmptype == ICMP_REDIRECT || 4004 icmptype == ICMP_TIMXCEED || 4005 icmptype == ICMP_PARAMPROB) 4006 state_icmp++; 4007 break; 4008#endif /* INET */ 4009#ifdef INET6 4010 case IPPROTO_ICMPV6: 4011 icmptype = pd->hdr.icmp6->icmp6_type; 4012 icmpcode = pd->hdr.icmp6->icmp6_code; 4013 icmpid = pd->hdr.icmp6->icmp6_id; 4014 4015 if (icmptype == ICMP6_DST_UNREACH || 4016 icmptype == ICMP6_PACKET_TOO_BIG || 4017 icmptype == ICMP6_TIME_EXCEEDED || 4018 icmptype == ICMP6_PARAM_PROB) 4019 state_icmp++; 4020 break; 4021#endif /* INET6 */ 4022 } 4023 4024 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 4025 4026 if (direction == PF_OUT) { 4027 bport = nport = icmpid; 4028 /* check outgoing packet for BINAT/NAT */ 4029 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, 4030 saddr, icmpid, daddr, icmpid, &pd->naddr, &nport)) != 4031 NULL) { 4032 PF_ACPY(&pd->baddr, saddr, af); 4033 switch (af) { 4034#ifdef INET 4035 case AF_INET: 4036 pf_change_a(&saddr->v4.s_addr, pd->ip_sum, 4037 pd->naddr.v4.s_addr, 0); 4038 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup( 4039 pd->hdr.icmp->icmp_cksum, icmpid, nport, 0); 4040 pd->hdr.icmp->icmp_id = nport; 4041 m_copyback(m, off, ICMP_MINLEN, 4042 (caddr_t)pd->hdr.icmp); 4043 break; 4044#endif /* INET */ 4045#ifdef INET6 4046 case AF_INET6: 4047 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum, 4048 &pd->naddr, 0); 4049 rewrite++; 4050 break; 4051#endif /* INET6 */ 4052 } 4053 if (nr->natpass) 4054 r = NULL; 4055 pd->nat_rule = nr; 4056 } 4057 } else { 4058 bport = nport = icmpid; 4059 /* check incoming packet for BINAT/RDR */ 4060 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, 4061 saddr, icmpid, daddr, icmpid, &pd->naddr, &nport)) != 4062 NULL) { 4063 PF_ACPY(&pd->baddr, daddr, af); 4064 switch (af) { 4065#ifdef INET 4066 case AF_INET: 4067 pf_change_a(&daddr->v4.s_addr, 4068 pd->ip_sum, pd->naddr.v4.s_addr, 0); 4069 break; 4070#endif /* INET */ 4071#ifdef INET6 4072 case AF_INET6: 4073 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum, 4074 &pd->naddr, 0); 4075 rewrite++; 4076 break; 4077#endif /* INET6 */ 4078 } 4079 if (nr->natpass) 4080 r = NULL; 4081 pd->nat_rule = nr; 4082 } 4083 } 4084 4085 while (r != NULL) { 4086 r->evaluations++; 4087 if (pfi_kif_match(r->kif, kif) == r->ifnot) 4088 r = r->skip[PF_SKIP_IFP].ptr; 4089 else if (r->direction && r->direction != direction) 4090 r = r->skip[PF_SKIP_DIR].ptr; 4091 else if (r->af && r->af != af) 4092 r = r->skip[PF_SKIP_AF].ptr; 4093 else if (r->proto && r->proto != pd->proto) 4094 r = r->skip[PF_SKIP_PROTO].ptr; 4095 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, 4096 r->src.neg, kif)) 4097 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 4098 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, 4099 r->dst.neg, NULL)) 4100 r = r->skip[PF_SKIP_DST_ADDR].ptr; 4101 else if (r->type && r->type != icmptype + 1) 4102 r = TAILQ_NEXT(r, entries); 4103 else if (r->code && r->code != icmpcode + 1) 4104 r = TAILQ_NEXT(r, entries); 4105 else if (r->tos && !(r->tos == pd->tos)) 4106 r = TAILQ_NEXT(r, entries); 4107 else if (r->rule_flag & PFRULE_FRAGMENT) 4108 r = TAILQ_NEXT(r, entries); 4109 else if (r->prob && r->prob <= arc4random()) 4110 r = TAILQ_NEXT(r, entries); 4111 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) 4112 r = TAILQ_NEXT(r, entries); 4113 else if (r->os_fingerprint != PF_OSFP_ANY) 4114 r = TAILQ_NEXT(r, entries); 4115 else { 4116 if (r->tag) 4117 tag = r->tag; 4118 if (r->rtableid >= 0) 4119 rtableid = r->rtableid; 4120 if (r->anchor == NULL) { 4121 match = 1; 4122 *rm = r; 4123 *am = a; 4124 *rsm = ruleset; 4125 if ((*rm)->quick) 4126 break; 4127 r = TAILQ_NEXT(r, entries); 4128 } else 4129 pf_step_into_anchor(&asd, &ruleset, 4130 PF_RULESET_FILTER, &r, &a, &match); 4131 } 4132 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 4133 PF_RULESET_FILTER, &r, &a, &match)) 4134 break; 4135 } 4136 r = *rm; 4137 a = *am; 4138 ruleset = *rsm; 4139 4140 REASON_SET(&reason, PFRES_MATCH); 4141 4142 if (r->log || (nr != NULL && nr->natpass && nr->log)) { 4143#ifdef INET6 4144 if (rewrite) 4145 m_copyback(m, off, sizeof(struct icmp6_hdr), 4146 (caddr_t)pd->hdr.icmp6); 4147#endif /* INET6 */ 4148 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, 4149 a, ruleset, pd); 4150 } 4151 4152 if (r->action != PF_PASS) 4153 return (PF_DROP); 4154 4155 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) { 4156 REASON_SET(&reason, PFRES_MEMORY); 4157 return (PF_DROP); 4158 } 4159 4160 if (!state_icmp && (r->keep_state || nr != NULL)) { 4161 /* create new state */ 4162 struct pf_state *s = NULL; 4163 struct pf_src_node *sn = NULL; 4164 4165 /* check maximums */ 4166 if (r->max_states && (r->states >= r->max_states)) { 4167 pf_status.lcounters[LCNT_STATES]++; 4168 REASON_SET(&reason, PFRES_MAXSTATES); 4169 goto cleanup; 4170 } 4171 /* src node for filter rule */ 4172 if ((r->rule_flag & PFRULE_SRCTRACK || 4173 r->rpool.opts & PF_POOL_STICKYADDR) && 4174 pf_insert_src_node(&sn, r, saddr, af) != 0) { 4175 REASON_SET(&reason, PFRES_SRCLIMIT); 4176 goto cleanup; 4177 } 4178 /* src node for translation rule */ 4179 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 4180 ((direction == PF_OUT && 4181 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || 4182 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { 4183 REASON_SET(&reason, PFRES_SRCLIMIT); 4184 goto cleanup; 4185 } 4186 s = pool_get(&pf_state_pl, PR_NOWAIT); 4187 if (s == NULL) { 4188 REASON_SET(&reason, PFRES_MEMORY); 4189cleanup: 4190 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 4191 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 4192 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 4193 pf_status.src_nodes--; 4194 pool_put(&pf_src_tree_pl, sn); 4195 } 4196 if (nsn != sn && nsn != NULL && nsn->states == 0 && 4197 nsn->expire == 0) { 4198 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 4199 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 4200 pf_status.src_nodes--; 4201 pool_put(&pf_src_tree_pl, nsn); 4202 } 4203 return (PF_DROP); 4204 } 4205 bzero(s, sizeof(*s)); 4206 s->rule.ptr = r; 4207 s->nat_rule.ptr = nr; 4208 s->anchor.ptr = a; 4209 STATE_INC_COUNTERS(s); 4210 s->allow_opts = r->allow_opts; 4211 s->log = r->log & PF_LOG_ALL; 4212 if (nr != NULL) 4213 s->log |= nr->log & PF_LOG_ALL; 4214 s->proto = pd->proto; 4215 s->direction = direction; 4216 s->af = af; 4217 if (direction == PF_OUT) { 4218 PF_ACPY(&s->gwy.addr, saddr, af); 4219 s->gwy.port = nport; 4220 PF_ACPY(&s->ext.addr, daddr, af); 4221 s->ext.port = 0; 4222 if (nr != NULL) { 4223 PF_ACPY(&s->lan.addr, &pd->baddr, af); 4224 s->lan.port = bport; 4225 } else { 4226 PF_ACPY(&s->lan.addr, &s->gwy.addr, af); 4227 s->lan.port = s->gwy.port; 4228 } 4229 } else { 4230 PF_ACPY(&s->lan.addr, daddr, af); 4231 s->lan.port = nport; 4232 PF_ACPY(&s->ext.addr, saddr, af); 4233 s->ext.port = 0; 4234 if (nr != NULL) { 4235 PF_ACPY(&s->gwy.addr, &pd->baddr, af); 4236 s->gwy.port = bport; 4237 } else { 4238 PF_ACPY(&s->gwy.addr, &s->lan.addr, af); 4239 s->gwy.port = s->lan.port; 4240 } 4241 } 4242 s->creation = time_second; 4243 s->expire = time_second; 4244 s->timeout = PFTM_ICMP_FIRST_PACKET; 4245 pf_set_rt_ifp(s, saddr); 4246 if (sn != NULL) { 4247 s->src_node = sn; 4248 s->src_node->states++; 4249 } 4250 if (nsn != NULL) { 4251 PF_ACPY(&nsn->raddr, &pd->naddr, af); 4252 s->nat_src_node = nsn; 4253 s->nat_src_node->states++; 4254 } 4255 if (pf_insert_state(BOUND_IFACE(r, kif), s)) { 4256 REASON_SET(&reason, PFRES_STATEINS); 4257 pf_src_tree_remove_state(s); 4258 STATE_DEC_COUNTERS(s); 4259 pool_put(&pf_state_pl, s); 4260 return (PF_DROP); 4261 } else 4262 *sm = s; 4263 if (tag > 0) { 4264 pf_tag_ref(tag); 4265 s->tag = tag; 4266 } 4267 } 4268 4269#ifdef INET6 4270 /* copy back packet headers if we performed IPv6 NAT operations */ 4271 if (rewrite) 4272 m_copyback(m, off, sizeof(struct icmp6_hdr), 4273 (caddr_t)pd->hdr.icmp6); 4274#endif /* INET6 */ 4275 4276 return (PF_PASS); 4277} 4278 4279int 4280pf_test_other(struct pf_rule **rm, struct pf_state **sm, int direction, 4281 struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd, 4282 struct pf_rule **am, struct pf_ruleset **rsm, struct ifqueue *ifq) 4283{ 4284 struct pf_rule *nr = NULL; 4285 struct pf_rule *r, *a = NULL; 4286 struct pf_ruleset *ruleset = NULL; 4287 struct pf_src_node *nsn = NULL; 4288 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 4289 sa_family_t af = pd->af; 4290 u_short reason; 4291 int tag = -1, rtableid = -1; 4292 int asd = 0; 4293 int match = 0; 4294 4295 if (pf_check_congestion(ifq)) { 4296 REASON_SET(&reason, PFRES_CONGEST); 4297 return (PF_DROP); 4298 } 4299 4300 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 4301 4302 if (direction == PF_OUT) { 4303 /* check outgoing packet for BINAT/NAT */ 4304 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, 4305 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) { 4306 PF_ACPY(&pd->baddr, saddr, af); 4307 switch (af) { 4308#ifdef INET 4309 case AF_INET: 4310 pf_change_a(&saddr->v4.s_addr, pd->ip_sum, 4311 pd->naddr.v4.s_addr, 0); 4312 break; 4313#endif /* INET */ 4314#ifdef INET6 4315 case AF_INET6: 4316 PF_ACPY(saddr, &pd->naddr, af); 4317 break; 4318#endif /* INET6 */ 4319 } 4320 if (nr->natpass) 4321 r = NULL; 4322 pd->nat_rule = nr; 4323 } 4324 } else { 4325 /* check incoming packet for BINAT/RDR */ 4326 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, 4327 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) { 4328 PF_ACPY(&pd->baddr, daddr, af); 4329 switch (af) { 4330#ifdef INET 4331 case AF_INET: 4332 pf_change_a(&daddr->v4.s_addr, 4333 pd->ip_sum, pd->naddr.v4.s_addr, 0); 4334 break; 4335#endif /* INET */ 4336#ifdef INET6 4337 case AF_INET6: 4338 PF_ACPY(daddr, &pd->naddr, af); 4339 break; 4340#endif /* INET6 */ 4341 } 4342 if (nr->natpass) 4343 r = NULL; 4344 pd->nat_rule = nr; 4345 } 4346 } 4347 4348 while (r != NULL) { 4349 r->evaluations++; 4350 if (pfi_kif_match(r->kif, kif) == r->ifnot) 4351 r = r->skip[PF_SKIP_IFP].ptr; 4352 else if (r->direction && r->direction != direction) 4353 r = r->skip[PF_SKIP_DIR].ptr; 4354 else if (r->af && r->af != af) 4355 r = r->skip[PF_SKIP_AF].ptr; 4356 else if (r->proto && r->proto != pd->proto) 4357 r = r->skip[PF_SKIP_PROTO].ptr; 4358 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, 4359 r->src.neg, kif)) 4360 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 4361 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, 4362 r->dst.neg, NULL)) 4363 r = r->skip[PF_SKIP_DST_ADDR].ptr; 4364 else if (r->tos && !(r->tos == pd->tos)) 4365 r = TAILQ_NEXT(r, entries); 4366 else if (r->rule_flag & PFRULE_FRAGMENT) 4367 r = TAILQ_NEXT(r, entries); 4368 else if (r->prob && r->prob <= arc4random()) 4369 r = TAILQ_NEXT(r, entries); 4370 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) 4371 r = TAILQ_NEXT(r, entries); 4372 else if (r->os_fingerprint != PF_OSFP_ANY) 4373 r = TAILQ_NEXT(r, entries); 4374 else { 4375 if (r->tag) 4376 tag = r->tag; 4377 if (r->rtableid >= 0) 4378 rtableid = r->rtableid; 4379 if (r->anchor == NULL) { 4380 match = 1; 4381 *rm = r; 4382 *am = a; 4383 *rsm = ruleset; 4384 if ((*rm)->quick) 4385 break; 4386 r = TAILQ_NEXT(r, entries); 4387 } else 4388 pf_step_into_anchor(&asd, &ruleset, 4389 PF_RULESET_FILTER, &r, &a, &match); 4390 } 4391 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 4392 PF_RULESET_FILTER, &r, &a, &match)) 4393 break; 4394 } 4395 r = *rm; 4396 a = *am; 4397 ruleset = *rsm; 4398 4399 REASON_SET(&reason, PFRES_MATCH); 4400 4401 if (r->log || (nr != NULL && nr->natpass && nr->log)) 4402 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, 4403 a, ruleset, pd); 4404 4405 if ((r->action == PF_DROP) && 4406 ((r->rule_flag & PFRULE_RETURNICMP) || 4407 (r->rule_flag & PFRULE_RETURN))) { 4408 struct pf_addr *a = NULL; 4409 4410 if (nr != NULL) { 4411 if (direction == PF_OUT) 4412 a = saddr; 4413 else 4414 a = daddr; 4415 } 4416 if (a != NULL) { 4417 switch (af) { 4418#ifdef INET 4419 case AF_INET: 4420 pf_change_a(&a->v4.s_addr, pd->ip_sum, 4421 pd->baddr.v4.s_addr, 0); 4422 break; 4423#endif /* INET */ 4424#ifdef INET6 4425 case AF_INET6: 4426 PF_ACPY(a, &pd->baddr, af); 4427 break; 4428#endif /* INET6 */ 4429 } 4430 } 4431 if ((af == AF_INET) && r->return_icmp) 4432 pf_send_icmp(m, r->return_icmp >> 8, 4433 r->return_icmp & 255, af, r); 4434 else if ((af == AF_INET6) && r->return_icmp6) 4435 pf_send_icmp(m, r->return_icmp6 >> 8, 4436 r->return_icmp6 & 255, af, r); 4437 } 4438 4439 if (r->action != PF_PASS) 4440 return (PF_DROP); 4441 4442 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) { 4443 REASON_SET(&reason, PFRES_MEMORY); 4444 return (PF_DROP); 4445 } 4446 4447 if (r->keep_state || nr != NULL) { 4448 /* create new state */ 4449 struct pf_state *s = NULL; 4450 struct pf_src_node *sn = NULL; 4451 4452 /* check maximums */ 4453 if (r->max_states && (r->states >= r->max_states)) { 4454 pf_status.lcounters[LCNT_STATES]++; 4455 REASON_SET(&reason, PFRES_MAXSTATES); 4456 goto cleanup; 4457 } 4458 /* src node for filter rule */ 4459 if ((r->rule_flag & PFRULE_SRCTRACK || 4460 r->rpool.opts & PF_POOL_STICKYADDR) && 4461 pf_insert_src_node(&sn, r, saddr, af) != 0) { 4462 REASON_SET(&reason, PFRES_SRCLIMIT); 4463 goto cleanup; 4464 } 4465 /* src node for translation rule */ 4466 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 4467 ((direction == PF_OUT && 4468 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || 4469 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { 4470 REASON_SET(&reason, PFRES_SRCLIMIT); 4471 goto cleanup; 4472 } 4473 s = pool_get(&pf_state_pl, PR_NOWAIT); 4474 if (s == NULL) { 4475 REASON_SET(&reason, PFRES_MEMORY); 4476cleanup: 4477 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 4478 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 4479 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 4480 pf_status.src_nodes--; 4481 pool_put(&pf_src_tree_pl, sn); 4482 } 4483 if (nsn != sn && nsn != NULL && nsn->states == 0 && 4484 nsn->expire == 0) { 4485 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 4486 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 4487 pf_status.src_nodes--; 4488 pool_put(&pf_src_tree_pl, nsn); 4489 } 4490 return (PF_DROP); 4491 } 4492 bzero(s, sizeof(*s)); 4493 s->rule.ptr = r; 4494 s->nat_rule.ptr = nr; 4495 s->anchor.ptr = a; 4496 STATE_INC_COUNTERS(s); 4497 s->allow_opts = r->allow_opts; 4498 s->log = r->log & PF_LOG_ALL; 4499 if (nr != NULL) 4500 s->log |= nr->log & PF_LOG_ALL; 4501 s->proto = pd->proto; 4502 s->direction = direction; 4503 s->af = af; 4504 if (direction == PF_OUT) { 4505 PF_ACPY(&s->gwy.addr, saddr, af); 4506 PF_ACPY(&s->ext.addr, daddr, af); 4507 if (nr != NULL) 4508 PF_ACPY(&s->lan.addr, &pd->baddr, af); 4509 else 4510 PF_ACPY(&s->lan.addr, &s->gwy.addr, af); 4511 } else { 4512 PF_ACPY(&s->lan.addr, daddr, af); 4513 PF_ACPY(&s->ext.addr, saddr, af); 4514 if (nr != NULL) 4515 PF_ACPY(&s->gwy.addr, &pd->baddr, af); 4516 else 4517 PF_ACPY(&s->gwy.addr, &s->lan.addr, af); 4518 } 4519 s->src.state = PFOTHERS_SINGLE; 4520 s->dst.state = PFOTHERS_NO_TRAFFIC; 4521 s->creation = time_second; 4522 s->expire = time_second; 4523 s->timeout = PFTM_OTHER_FIRST_PACKET; 4524 pf_set_rt_ifp(s, saddr); 4525 if (sn != NULL) { 4526 s->src_node = sn; 4527 s->src_node->states++; 4528 } 4529 if (nsn != NULL) { 4530 PF_ACPY(&nsn->raddr, &pd->naddr, af); 4531 s->nat_src_node = nsn; 4532 s->nat_src_node->states++; 4533 } 4534 if (pf_insert_state(BOUND_IFACE(r, kif), s)) { 4535 REASON_SET(&reason, PFRES_STATEINS); 4536 pf_src_tree_remove_state(s); 4537 STATE_DEC_COUNTERS(s); 4538 pool_put(&pf_state_pl, s); 4539 return (PF_DROP); 4540 } else 4541 *sm = s; 4542 if (tag > 0) { 4543 pf_tag_ref(tag); 4544 s->tag = tag; 4545 } 4546 } 4547 4548 return (PF_PASS); 4549} 4550 4551int 4552pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif, 4553 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am, 4554 struct pf_ruleset **rsm) 4555{ 4556 struct pf_rule *r, *a = NULL; 4557 struct pf_ruleset *ruleset = NULL; 4558 sa_family_t af = pd->af; 4559 u_short reason; 4560 int tag = -1; 4561 int asd = 0; 4562 int match = 0; 4563 4564 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 4565 while (r != NULL) { 4566 r->evaluations++; 4567 if (pfi_kif_match(r->kif, kif) == r->ifnot) 4568 r = r->skip[PF_SKIP_IFP].ptr; 4569 else if (r->direction && r->direction != direction) 4570 r = r->skip[PF_SKIP_DIR].ptr; 4571 else if (r->af && r->af != af) 4572 r = r->skip[PF_SKIP_AF].ptr; 4573 else if (r->proto && r->proto != pd->proto) 4574 r = r->skip[PF_SKIP_PROTO].ptr; 4575 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, 4576 r->src.neg, kif)) 4577 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 4578 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, 4579 r->dst.neg, NULL)) 4580 r = r->skip[PF_SKIP_DST_ADDR].ptr; 4581 else if (r->tos && !(r->tos == pd->tos)) 4582 r = TAILQ_NEXT(r, entries); 4583 else if (r->os_fingerprint != PF_OSFP_ANY) 4584 r = TAILQ_NEXT(r, entries); 4585 else if (pd->proto == IPPROTO_UDP && 4586 (r->src.port_op || r->dst.port_op)) 4587 r = TAILQ_NEXT(r, entries); 4588 else if (pd->proto == IPPROTO_TCP && 4589 (r->src.port_op || r->dst.port_op || r->flagset)) 4590 r = TAILQ_NEXT(r, entries); 4591 else if ((pd->proto == IPPROTO_ICMP || 4592 pd->proto == IPPROTO_ICMPV6) && 4593 (r->type || r->code)) 4594 r = TAILQ_NEXT(r, entries); 4595 else if (r->prob && r->prob <= arc4random()) 4596 r = TAILQ_NEXT(r, entries); 4597 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) 4598 r = TAILQ_NEXT(r, entries); 4599 else { 4600 if (r->anchor == NULL) { 4601 match = 1; 4602 *rm = r; 4603 *am = a; 4604 *rsm = ruleset; 4605 if ((*rm)->quick) 4606 break; 4607 r = TAILQ_NEXT(r, entries); 4608 } else 4609 pf_step_into_anchor(&asd, &ruleset, 4610 PF_RULESET_FILTER, &r, &a, &match); 4611 } 4612 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 4613 PF_RULESET_FILTER, &r, &a, &match)) 4614 break; 4615 } 4616 r = *rm; 4617 a = *am; 4618 ruleset = *rsm; 4619 4620 REASON_SET(&reason, PFRES_MATCH); 4621 4622 if (r->log) 4623 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset, 4624 pd); 4625 4626 if (r->action != PF_PASS) 4627 return (PF_DROP); 4628 4629 if (pf_tag_packet(m, pd->pf_mtag, tag, -1)) { 4630 REASON_SET(&reason, PFRES_MEMORY); 4631 return (PF_DROP); 4632 } 4633 4634 return (PF_PASS); 4635} 4636 4637int 4638pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif, 4639 struct mbuf *m, int off, void *h, struct pf_pdesc *pd, 4640 u_short *reason) 4641{ 4642 struct pf_state_cmp key; 4643 struct tcphdr *th = pd->hdr.tcp; 4644 u_int16_t win = ntohs(th->th_win); 4645 u_int32_t ack, end, seq, orig_seq; 4646 u_int8_t sws, dws; 4647 int ackskew; 4648 int copyback = 0; 4649 struct pf_state_peer *src, *dst; 4650 4651 key.af = pd->af; 4652 key.proto = IPPROTO_TCP; 4653 if (direction == PF_IN) { 4654 PF_ACPY(&key.ext.addr, pd->src, key.af); 4655 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 4656 key.ext.port = th->th_sport; 4657 key.gwy.port = th->th_dport; 4658 } else { 4659 PF_ACPY(&key.lan.addr, pd->src, key.af); 4660 PF_ACPY(&key.ext.addr, pd->dst, key.af); 4661 key.lan.port = th->th_sport; 4662 key.ext.port = th->th_dport; 4663 } 4664 4665 STATE_LOOKUP(); 4666 4667 if (direction == (*state)->direction) { 4668 src = &(*state)->src; 4669 dst = &(*state)->dst; 4670 } else { 4671 src = &(*state)->dst; 4672 dst = &(*state)->src; 4673 } 4674 4675 if ((*state)->src.state == PF_TCPS_PROXY_SRC) { 4676 if (direction != (*state)->direction) { 4677 REASON_SET(reason, PFRES_SYNPROXY); 4678 return (PF_SYNPROXY_DROP); 4679 } 4680 if (th->th_flags & TH_SYN) { 4681 if (ntohl(th->th_seq) != (*state)->src.seqlo) { 4682 REASON_SET(reason, PFRES_SYNPROXY); 4683 return (PF_DROP); 4684 } 4685#ifdef __FreeBSD__ 4686 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst, 4687#else 4688 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, 4689#endif 4690 pd->src, th->th_dport, th->th_sport, 4691 (*state)->src.seqhi, ntohl(th->th_seq) + 1, 4692 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1, 4693 0, NULL, NULL); 4694 REASON_SET(reason, PFRES_SYNPROXY); 4695 return (PF_SYNPROXY_DROP); 4696 } else if (!(th->th_flags & TH_ACK) || 4697 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) || 4698 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) { 4699 REASON_SET(reason, PFRES_SYNPROXY); 4700 return (PF_DROP); 4701 } else if ((*state)->src_node != NULL && 4702 pf_src_connlimit(state)) { 4703 REASON_SET(reason, PFRES_SRCLIMIT); 4704 return (PF_DROP); 4705 } else 4706 (*state)->src.state = PF_TCPS_PROXY_DST; 4707 } 4708 if ((*state)->src.state == PF_TCPS_PROXY_DST) { 4709 struct pf_state_host *src, *dst; 4710 4711 if (direction == PF_OUT) { 4712 src = &(*state)->gwy; 4713 dst = &(*state)->ext; 4714 } else { 4715 src = &(*state)->ext; 4716 dst = &(*state)->lan; 4717 } 4718 if (direction == (*state)->direction) { 4719 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) || 4720 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) || 4721 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) { 4722 REASON_SET(reason, PFRES_SYNPROXY); 4723 return (PF_DROP); 4724 } 4725 (*state)->src.max_win = MAX(ntohs(th->th_win), 1); 4726 if ((*state)->dst.seqhi == 1) 4727 (*state)->dst.seqhi = htonl(arc4random()); 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)->dst.seqhi, 0, TH_SYN, 0, 4736 (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL); 4737 REASON_SET(reason, PFRES_SYNPROXY); 4738 return (PF_SYNPROXY_DROP); 4739 } else if (((th->th_flags & (TH_SYN|TH_ACK)) != 4740 (TH_SYN|TH_ACK)) || 4741 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) { 4742 REASON_SET(reason, PFRES_SYNPROXY); 4743 return (PF_DROP); 4744 } else { 4745 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1); 4746 (*state)->dst.seqlo = ntohl(th->th_seq); 4747#ifdef __FreeBSD__ 4748 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst, 4749#else 4750 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, 4751#endif 4752 pd->src, th->th_dport, th->th_sport, 4753 ntohl(th->th_ack), ntohl(th->th_seq) + 1, 4754 TH_ACK, (*state)->src.max_win, 0, 0, 0, 4755 (*state)->tag, NULL, NULL); 4756#ifdef __FreeBSD__ 4757 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, 4758 &src->addr, 4759#else 4760 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr, 4761#endif 4762 &dst->addr, src->port, dst->port, 4763 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1, 4764 TH_ACK, (*state)->dst.max_win, 0, 0, 1, 4765 0, NULL, NULL); 4766 (*state)->src.seqdiff = (*state)->dst.seqhi - 4767 (*state)->src.seqlo; 4768 (*state)->dst.seqdiff = (*state)->src.seqhi - 4769 (*state)->dst.seqlo; 4770 (*state)->src.seqhi = (*state)->src.seqlo + 4771 (*state)->dst.max_win; 4772 (*state)->dst.seqhi = (*state)->dst.seqlo + 4773 (*state)->src.max_win; 4774 (*state)->src.wscale = (*state)->dst.wscale = 0; 4775 (*state)->src.state = (*state)->dst.state = 4776 TCPS_ESTABLISHED; 4777 REASON_SET(reason, PFRES_SYNPROXY); 4778 return (PF_SYNPROXY_DROP); 4779 } 4780 } 4781 4782 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) { 4783 sws = src->wscale & PF_WSCALE_MASK; 4784 dws = dst->wscale & PF_WSCALE_MASK; 4785 } else 4786 sws = dws = 0; 4787 4788 /* 4789 * Sequence tracking algorithm from Guido van Rooij's paper: 4790 * http://www.madison-gurkha.com/publications/tcp_filtering/ 4791 * tcp_filtering.ps 4792 */ 4793 4794 orig_seq = seq = ntohl(th->th_seq); 4795 if (src->seqlo == 0) { 4796 /* First packet from this end. Set its state */ 4797 4798 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) && 4799 src->scrub == NULL) { 4800 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) { 4801 REASON_SET(reason, PFRES_MEMORY); 4802 return (PF_DROP); 4803 } 4804 } 4805 4806 /* Deferred generation of sequence number modulator */ 4807 if (dst->seqdiff && !src->seqdiff) { 4808#ifdef __FreeBSD__ 4809 while ((src->seqdiff = pf_new_isn(*state) - seq) == 0) 4810 ; 4811#else 4812 while ((src->seqdiff = tcp_rndiss_next() - seq) == 0) 4813 ; 4814#endif 4815 ack = ntohl(th->th_ack) - dst->seqdiff; 4816 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq + 4817 src->seqdiff), 0); 4818 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0); 4819 copyback = 1; 4820 } else { 4821 ack = ntohl(th->th_ack); 4822 } 4823 4824 end = seq + pd->p_len; 4825 if (th->th_flags & TH_SYN) { 4826 end++; 4827 if (dst->wscale & PF_WSCALE_FLAG) { 4828 src->wscale = pf_get_wscale(m, off, th->th_off, 4829 pd->af); 4830 if (src->wscale & PF_WSCALE_FLAG) { 4831 /* Remove scale factor from initial 4832 * window */ 4833 sws = src->wscale & PF_WSCALE_MASK; 4834 win = ((u_int32_t)win + (1 << sws) - 1) 4835 >> sws; 4836 dws = dst->wscale & PF_WSCALE_MASK; 4837 } else { 4838 /* fixup other window */ 4839 dst->max_win <<= dst->wscale & 4840 PF_WSCALE_MASK; 4841 /* in case of a retrans SYN|ACK */ 4842 dst->wscale = 0; 4843 } 4844 } 4845 } 4846 if (th->th_flags & TH_FIN) 4847 end++; 4848 4849 src->seqlo = seq; 4850 if (src->state < TCPS_SYN_SENT) 4851 src->state = TCPS_SYN_SENT; 4852 4853 /* 4854 * May need to slide the window (seqhi may have been set by 4855 * the crappy stack check or if we picked up the connection 4856 * after establishment) 4857 */ 4858 if (src->seqhi == 1 || 4859 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi)) 4860 src->seqhi = end + MAX(1, dst->max_win << dws); 4861 if (win > src->max_win) 4862 src->max_win = win; 4863 4864 } else { 4865 ack = ntohl(th->th_ack) - dst->seqdiff; 4866 if (src->seqdiff) { 4867 /* Modulate sequence numbers */ 4868 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq + 4869 src->seqdiff), 0); 4870 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0); 4871 copyback = 1; 4872 } 4873 end = seq + pd->p_len; 4874 if (th->th_flags & TH_SYN) 4875 end++; 4876 if (th->th_flags & TH_FIN) 4877 end++; 4878 } 4879 4880 if ((th->th_flags & TH_ACK) == 0) { 4881 /* Let it pass through the ack skew check */ 4882 ack = dst->seqlo; 4883 } else if ((ack == 0 && 4884 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) || 4885 /* broken tcp stacks do not set ack */ 4886 (dst->state < TCPS_SYN_SENT)) { 4887 /* 4888 * Many stacks (ours included) will set the ACK number in an 4889 * FIN|ACK if the SYN times out -- no sequence to ACK. 4890 */ 4891 ack = dst->seqlo; 4892 } 4893 4894 if (seq == end) { 4895 /* Ease sequencing restrictions on no data packets */ 4896 seq = src->seqlo; 4897 end = seq; 4898 } 4899 4900 ackskew = dst->seqlo - ack; 4901 4902 4903 /* 4904 * Need to demodulate the sequence numbers in any TCP SACK options 4905 * (Selective ACK). We could optionally validate the SACK values 4906 * against the current ACK window, either forwards or backwards, but 4907 * I'm not confident that SACK has been implemented properly 4908 * everywhere. It wouldn't surprise me if several stacks accidently 4909 * SACK too far backwards of previously ACKed data. There really aren't 4910 * any security implications of bad SACKing unless the target stack 4911 * doesn't validate the option length correctly. Someone trying to 4912 * spoof into a TCP connection won't bother blindly sending SACK 4913 * options anyway. 4914 */ 4915 if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) { 4916 if (pf_modulate_sack(m, off, pd, th, dst)) 4917 copyback = 1; 4918 } 4919 4920 4921#define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */ 4922 if (SEQ_GEQ(src->seqhi, end) && 4923 /* Last octet inside other's window space */ 4924 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) && 4925 /* Retrans: not more than one window back */ 4926 (ackskew >= -MAXACKWINDOW) && 4927 /* Acking not more than one reassembled fragment backwards */ 4928 (ackskew <= (MAXACKWINDOW << sws)) && 4929 /* Acking not more than one window forward */ 4930 ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo || 4931 (orig_seq == src->seqlo + 1) || (pd->flags & PFDESC_IP_REAS) == 0)) { 4932 /* Require an exact/+1 sequence match on resets when possible */ 4933 4934 if (dst->scrub || src->scrub) { 4935 if (pf_normalize_tcp_stateful(m, off, pd, reason, th, 4936 *state, src, dst, ©back)) 4937 return (PF_DROP); 4938 } 4939 4940 /* update max window */ 4941 if (src->max_win < win) 4942 src->max_win = win; 4943 /* synchronize sequencing */ 4944 if (SEQ_GT(end, src->seqlo)) 4945 src->seqlo = end; 4946 /* slide the window of what the other end can send */ 4947 if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) 4948 dst->seqhi = ack + MAX((win << sws), 1); 4949 4950 4951 /* update states */ 4952 if (th->th_flags & TH_SYN) 4953 if (src->state < TCPS_SYN_SENT) 4954 src->state = TCPS_SYN_SENT; 4955 if (th->th_flags & TH_FIN) 4956 if (src->state < TCPS_CLOSING) 4957 src->state = TCPS_CLOSING; 4958 if (th->th_flags & TH_ACK) { 4959 if (dst->state == TCPS_SYN_SENT) { 4960 dst->state = TCPS_ESTABLISHED; 4961 if (src->state == TCPS_ESTABLISHED && 4962 (*state)->src_node != NULL && 4963 pf_src_connlimit(state)) { 4964 REASON_SET(reason, PFRES_SRCLIMIT); 4965 return (PF_DROP); 4966 } 4967 } else if (dst->state == TCPS_CLOSING) 4968 dst->state = TCPS_FIN_WAIT_2; 4969 } 4970 if (th->th_flags & TH_RST) 4971 src->state = dst->state = TCPS_TIME_WAIT; 4972 4973 /* update expire time */ 4974 (*state)->expire = time_second; 4975 if (src->state >= TCPS_FIN_WAIT_2 && 4976 dst->state >= TCPS_FIN_WAIT_2) 4977 (*state)->timeout = PFTM_TCP_CLOSED; 4978 else if (src->state >= TCPS_CLOSING && 4979 dst->state >= TCPS_CLOSING) 4980 (*state)->timeout = PFTM_TCP_FIN_WAIT; 4981 else if (src->state < TCPS_ESTABLISHED || 4982 dst->state < TCPS_ESTABLISHED) 4983 (*state)->timeout = PFTM_TCP_OPENING; 4984 else if (src->state >= TCPS_CLOSING || 4985 dst->state >= TCPS_CLOSING) 4986 (*state)->timeout = PFTM_TCP_CLOSING; 4987 else 4988 (*state)->timeout = PFTM_TCP_ESTABLISHED; 4989 4990 /* Fall through to PASS packet */ 4991 4992 } else if ((dst->state < TCPS_SYN_SENT || 4993 dst->state >= TCPS_FIN_WAIT_2 || 4994 src->state >= TCPS_FIN_WAIT_2) && 4995 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) && 4996 /* Within a window forward of the originating packet */ 4997 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) { 4998 /* Within a window backward of the originating packet */ 4999 5000 /* 5001 * This currently handles three situations: 5002 * 1) Stupid stacks will shotgun SYNs before their peer 5003 * replies. 5004 * 2) When PF catches an already established stream (the 5005 * firewall rebooted, the state table was flushed, routes 5006 * changed...) 5007 * 3) Packets get funky immediately after the connection 5008 * closes (this should catch Solaris spurious ACK|FINs 5009 * that web servers like to spew after a close) 5010 * 5011 * This must be a little more careful than the above code 5012 * since packet floods will also be caught here. We don't 5013 * update the TTL here to mitigate the damage of a packet 5014 * flood and so the same code can handle awkward establishment 5015 * and a loosened connection close. 5016 * In the establishment case, a correct peer response will 5017 * validate the connection, go through the normal state code 5018 * and keep updating the state TTL. 5019 */ 5020 5021 if (pf_status.debug >= PF_DEBUG_MISC) { 5022 printf("pf: loose state match: "); 5023 pf_print_state(*state); 5024 pf_print_flags(th->th_flags); 5025 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d " 5026 "pkts=%llu:%llu\n", seq, orig_seq, ack, pd->p_len, 5027#ifdef __FreeBSD__ 5028 ackskew, (unsigned long long)(*state)->packets[0], 5029 (unsigned long long)(*state)->packets[1]); 5030#else 5031 ackskew, (*state)->packets[0], 5032 (*state)->packets[1]); 5033#endif 5034 } 5035 5036 if (dst->scrub || src->scrub) { 5037 if (pf_normalize_tcp_stateful(m, off, pd, reason, th, 5038 *state, src, dst, ©back)) 5039 return (PF_DROP); 5040 } 5041 5042 /* update max window */ 5043 if (src->max_win < win) 5044 src->max_win = win; 5045 /* synchronize sequencing */ 5046 if (SEQ_GT(end, src->seqlo)) 5047 src->seqlo = end; 5048 /* slide the window of what the other end can send */ 5049 if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) 5050 dst->seqhi = ack + MAX((win << sws), 1); 5051 5052 /* 5053 * Cannot set dst->seqhi here since this could be a shotgunned 5054 * SYN and not an already established connection. 5055 */ 5056 5057 if (th->th_flags & TH_FIN) 5058 if (src->state < TCPS_CLOSING) 5059 src->state = TCPS_CLOSING; 5060 if (th->th_flags & TH_RST) 5061 src->state = dst->state = TCPS_TIME_WAIT; 5062 5063 /* Fall through to PASS packet */ 5064 5065 } else { 5066 if ((*state)->dst.state == TCPS_SYN_SENT && 5067 (*state)->src.state == TCPS_SYN_SENT) { 5068 /* Send RST for state mismatches during handshake */ 5069 if (!(th->th_flags & TH_RST)) 5070#ifdef __FreeBSD__ 5071 pf_send_tcp(m, (*state)->rule.ptr, pd->af, 5072#else 5073 pf_send_tcp((*state)->rule.ptr, pd->af, 5074#endif 5075 pd->dst, pd->src, th->th_dport, 5076 th->th_sport, ntohl(th->th_ack), 0, 5077 TH_RST, 0, 0, 5078 (*state)->rule.ptr->return_ttl, 1, 0, 5079 pd->eh, kif->pfik_ifp); 5080 src->seqlo = 0; 5081 src->seqhi = 1; 5082 src->max_win = 1; 5083 } else if (pf_status.debug >= PF_DEBUG_MISC) { 5084 printf("pf: BAD state: "); 5085 pf_print_state(*state); 5086 pf_print_flags(th->th_flags); 5087 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d " 5088 "pkts=%llu:%llu dir=%s,%s\n", 5089 seq, orig_seq, ack, pd->p_len, ackskew, 5090#ifdef __FreeBSD__ 5091 (unsigned long long)(*state)->packets[0], 5092 (unsigned long long)(*state)->packets[1], 5093#else 5094 (*state)->packets[0], (*state)->packets[1], 5095#endif 5096 direction == PF_IN ? "in" : "out", 5097 direction == (*state)->direction ? "fwd" : "rev"); 5098 printf("pf: State failure on: %c %c %c %c | %c %c\n", 5099 SEQ_GEQ(src->seqhi, end) ? ' ' : '1', 5100 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ? 5101 ' ': '2', 5102 (ackskew >= -MAXACKWINDOW) ? ' ' : '3', 5103 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4', 5104 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5', 5105 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6'); 5106 } 5107 REASON_SET(reason, PFRES_BADSTATE); 5108 return (PF_DROP); 5109 } 5110 5111 /* Any packets which have gotten here are to be passed */ 5112 5113 /* translate source/destination address, if necessary */ 5114 if (STATE_TRANSLATE(*state)) { 5115 if (direction == PF_OUT) 5116 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum, 5117 &th->th_sum, &(*state)->gwy.addr, 5118 (*state)->gwy.port, 0, pd->af); 5119 else 5120 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum, 5121 &th->th_sum, &(*state)->lan.addr, 5122 (*state)->lan.port, 0, pd->af); 5123 m_copyback(m, off, sizeof(*th), (caddr_t)th); 5124 } else if (copyback) { 5125 /* Copyback sequence modulation or stateful scrub changes */ 5126 m_copyback(m, off, sizeof(*th), (caddr_t)th); 5127 } 5128 5129 return (PF_PASS); 5130} 5131 5132int 5133pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif, 5134 struct mbuf *m, int off, void *h, struct pf_pdesc *pd) 5135{ 5136 struct pf_state_peer *src, *dst; 5137 struct pf_state_cmp key; 5138 struct udphdr *uh = pd->hdr.udp; 5139 5140 key.af = pd->af; 5141 key.proto = IPPROTO_UDP; 5142 if (direction == PF_IN) { 5143 PF_ACPY(&key.ext.addr, pd->src, key.af); 5144 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 5145 key.ext.port = uh->uh_sport; 5146 key.gwy.port = uh->uh_dport; 5147 } else { 5148 PF_ACPY(&key.lan.addr, pd->src, key.af); 5149 PF_ACPY(&key.ext.addr, pd->dst, key.af); 5150 key.lan.port = uh->uh_sport; 5151 key.ext.port = uh->uh_dport; 5152 } 5153 5154 STATE_LOOKUP(); 5155 5156 if (direction == (*state)->direction) { 5157 src = &(*state)->src; 5158 dst = &(*state)->dst; 5159 } else { 5160 src = &(*state)->dst; 5161 dst = &(*state)->src; 5162 } 5163 5164 /* update states */ 5165 if (src->state < PFUDPS_SINGLE) 5166 src->state = PFUDPS_SINGLE; 5167 if (dst->state == PFUDPS_SINGLE) 5168 dst->state = PFUDPS_MULTIPLE; 5169 5170 /* update expire time */ 5171 (*state)->expire = time_second; 5172 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE) 5173 (*state)->timeout = PFTM_UDP_MULTIPLE; 5174 else 5175 (*state)->timeout = PFTM_UDP_SINGLE; 5176 5177 /* translate source/destination address, if necessary */ 5178 if (STATE_TRANSLATE(*state)) { 5179 if (direction == PF_OUT) 5180 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum, 5181 &uh->uh_sum, &(*state)->gwy.addr, 5182 (*state)->gwy.port, 1, pd->af); 5183 else 5184 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum, 5185 &uh->uh_sum, &(*state)->lan.addr, 5186 (*state)->lan.port, 1, pd->af); 5187 m_copyback(m, off, sizeof(*uh), (caddr_t)uh); 5188 } 5189 5190 return (PF_PASS); 5191} 5192 5193int 5194pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif, 5195 struct mbuf *m, int off, void *h, struct pf_pdesc *pd, u_short *reason) 5196{ 5197 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 5198 u_int16_t icmpid = 0; /* make the compiler happy */ 5199 u_int16_t *icmpsum = NULL; /* make the compiler happy */ 5200 u_int8_t icmptype = 0; /* make the compiler happy */ 5201 int state_icmp = 0; 5202 struct pf_state_cmp key; 5203 5204 switch (pd->proto) { 5205#ifdef INET 5206 case IPPROTO_ICMP: 5207 icmptype = pd->hdr.icmp->icmp_type; 5208 icmpid = pd->hdr.icmp->icmp_id; 5209 icmpsum = &pd->hdr.icmp->icmp_cksum; 5210 5211 if (icmptype == ICMP_UNREACH || 5212 icmptype == ICMP_SOURCEQUENCH || 5213 icmptype == ICMP_REDIRECT || 5214 icmptype == ICMP_TIMXCEED || 5215 icmptype == ICMP_PARAMPROB) 5216 state_icmp++; 5217 break; 5218#endif /* INET */ 5219#ifdef INET6 5220 case IPPROTO_ICMPV6: 5221 icmptype = pd->hdr.icmp6->icmp6_type; 5222 icmpid = pd->hdr.icmp6->icmp6_id; 5223 icmpsum = &pd->hdr.icmp6->icmp6_cksum; 5224 5225 if (icmptype == ICMP6_DST_UNREACH || 5226 icmptype == ICMP6_PACKET_TOO_BIG || 5227 icmptype == ICMP6_TIME_EXCEEDED || 5228 icmptype == ICMP6_PARAM_PROB) 5229 state_icmp++; 5230 break; 5231#endif /* INET6 */ 5232 } 5233 5234 if (!state_icmp) { 5235 5236 /* 5237 * ICMP query/reply message not related to a TCP/UDP packet. 5238 * Search for an ICMP state. 5239 */ 5240 key.af = pd->af; 5241 key.proto = pd->proto; 5242 if (direction == PF_IN) { 5243 PF_ACPY(&key.ext.addr, pd->src, key.af); 5244 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 5245 key.ext.port = 0; 5246 key.gwy.port = icmpid; 5247 } else { 5248 PF_ACPY(&key.lan.addr, pd->src, key.af); 5249 PF_ACPY(&key.ext.addr, pd->dst, key.af); 5250 key.lan.port = icmpid; 5251 key.ext.port = 0; 5252 } 5253 5254 STATE_LOOKUP(); 5255 5256 (*state)->expire = time_second; 5257 (*state)->timeout = PFTM_ICMP_ERROR_REPLY; 5258 5259 /* translate source/destination address, if necessary */ 5260 if (STATE_TRANSLATE(*state)) { 5261 if (direction == PF_OUT) { 5262 switch (pd->af) { 5263#ifdef INET 5264 case AF_INET: 5265 pf_change_a(&saddr->v4.s_addr, 5266 pd->ip_sum, 5267 (*state)->gwy.addr.v4.s_addr, 0); 5268 pd->hdr.icmp->icmp_cksum = 5269 pf_cksum_fixup( 5270 pd->hdr.icmp->icmp_cksum, icmpid, 5271 (*state)->gwy.port, 0); 5272 pd->hdr.icmp->icmp_id = 5273 (*state)->gwy.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(saddr, 5281 &pd->hdr.icmp6->icmp6_cksum, 5282 &(*state)->gwy.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 } else { 5290 switch (pd->af) { 5291#ifdef INET 5292 case AF_INET: 5293 pf_change_a(&daddr->v4.s_addr, 5294 pd->ip_sum, 5295 (*state)->lan.addr.v4.s_addr, 0); 5296 pd->hdr.icmp->icmp_cksum = 5297 pf_cksum_fixup( 5298 pd->hdr.icmp->icmp_cksum, icmpid, 5299 (*state)->lan.port, 0); 5300 pd->hdr.icmp->icmp_id = 5301 (*state)->lan.port; 5302 m_copyback(m, off, ICMP_MINLEN, 5303 (caddr_t)pd->hdr.icmp); 5304 break; 5305#endif /* INET */ 5306#ifdef INET6 5307 case AF_INET6: 5308 pf_change_a6(daddr, 5309 &pd->hdr.icmp6->icmp6_cksum, 5310 &(*state)->lan.addr, 0); 5311 m_copyback(m, off, 5312 sizeof(struct icmp6_hdr), 5313 (caddr_t)pd->hdr.icmp6); 5314 break; 5315#endif /* INET6 */ 5316 } 5317 } 5318 } 5319 5320 return (PF_PASS); 5321 5322 } else { 5323 /* 5324 * ICMP error message in response to a TCP/UDP packet. 5325 * Extract the inner TCP/UDP header and search for that state. 5326 */ 5327 5328 struct pf_pdesc pd2; 5329#ifdef INET 5330 struct ip h2; 5331#endif /* INET */ 5332#ifdef INET6 5333 struct ip6_hdr h2_6; 5334 int terminal = 0; 5335#endif /* INET6 */ 5336 int ipoff2 = 0; /* make the compiler happy */ 5337 int off2 = 0; /* make the compiler happy */ 5338 5339 pd2.af = pd->af; 5340 switch (pd->af) { 5341#ifdef INET 5342 case AF_INET: 5343 /* offset of h2 in mbuf chain */ 5344 ipoff2 = off + ICMP_MINLEN; 5345 5346 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2), 5347 NULL, reason, pd2.af)) { 5348 DPFPRINTF(PF_DEBUG_MISC, 5349 ("pf: ICMP error message too short " 5350 "(ip)\n")); 5351 return (PF_DROP); 5352 } 5353 /* 5354 * ICMP error messages don't refer to non-first 5355 * fragments 5356 */ 5357 if (h2.ip_off & htons(IP_OFFMASK)) { 5358 REASON_SET(reason, PFRES_FRAG); 5359 return (PF_DROP); 5360 } 5361 5362 /* offset of protocol header that follows h2 */ 5363 off2 = ipoff2 + (h2.ip_hl << 2); 5364 5365 pd2.proto = h2.ip_p; 5366 pd2.src = (struct pf_addr *)&h2.ip_src; 5367 pd2.dst = (struct pf_addr *)&h2.ip_dst; 5368 pd2.ip_sum = &h2.ip_sum; 5369 break; 5370#endif /* INET */ 5371#ifdef INET6 5372 case AF_INET6: 5373 ipoff2 = off + sizeof(struct icmp6_hdr); 5374 5375 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6), 5376 NULL, reason, pd2.af)) { 5377 DPFPRINTF(PF_DEBUG_MISC, 5378 ("pf: ICMP error message too short " 5379 "(ip6)\n")); 5380 return (PF_DROP); 5381 } 5382 pd2.proto = h2_6.ip6_nxt; 5383 pd2.src = (struct pf_addr *)&h2_6.ip6_src; 5384 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst; 5385 pd2.ip_sum = NULL; 5386 off2 = ipoff2 + sizeof(h2_6); 5387 do { 5388 switch (pd2.proto) { 5389 case IPPROTO_FRAGMENT: 5390 /* 5391 * ICMPv6 error messages for 5392 * non-first fragments 5393 */ 5394 REASON_SET(reason, PFRES_FRAG); 5395 return (PF_DROP); 5396 case IPPROTO_AH: 5397 case IPPROTO_HOPOPTS: 5398 case IPPROTO_ROUTING: 5399 case IPPROTO_DSTOPTS: { 5400 /* get next header and header length */ 5401 struct ip6_ext opt6; 5402 5403 if (!pf_pull_hdr(m, off2, &opt6, 5404 sizeof(opt6), NULL, reason, 5405 pd2.af)) { 5406 DPFPRINTF(PF_DEBUG_MISC, 5407 ("pf: ICMPv6 short opt\n")); 5408 return (PF_DROP); 5409 } 5410 if (pd2.proto == IPPROTO_AH) 5411 off2 += (opt6.ip6e_len + 2) * 4; 5412 else 5413 off2 += (opt6.ip6e_len + 1) * 8; 5414 pd2.proto = opt6.ip6e_nxt; 5415 /* goto the next header */ 5416 break; 5417 } 5418 default: 5419 terminal++; 5420 break; 5421 } 5422 } while (!terminal); 5423 break; 5424#endif /* INET6 */ 5425#ifdef __FreeBSD__ 5426 default: 5427 panic("AF not supported: %d", pd->af); 5428#endif 5429 } 5430 5431 switch (pd2.proto) { 5432 case IPPROTO_TCP: { 5433 struct tcphdr th; 5434 u_int32_t seq; 5435 struct pf_state_peer *src, *dst; 5436 u_int8_t dws; 5437 int copyback = 0; 5438 5439 /* 5440 * Only the first 8 bytes of the TCP header can be 5441 * expected. Don't access any TCP header fields after 5442 * th_seq, an ackskew test is not possible. 5443 */ 5444 if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason, 5445 pd2.af)) { 5446 DPFPRINTF(PF_DEBUG_MISC, 5447 ("pf: ICMP error message too short " 5448 "(tcp)\n")); 5449 return (PF_DROP); 5450 } 5451 5452 key.af = pd2.af; 5453 key.proto = IPPROTO_TCP; 5454 if (direction == PF_IN) { 5455 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 5456 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 5457 key.ext.port = th.th_dport; 5458 key.gwy.port = th.th_sport; 5459 } else { 5460 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 5461 PF_ACPY(&key.ext.addr, pd2.src, key.af); 5462 key.lan.port = th.th_dport; 5463 key.ext.port = th.th_sport; 5464 } 5465 5466 STATE_LOOKUP(); 5467 5468 if (direction == (*state)->direction) { 5469 src = &(*state)->dst; 5470 dst = &(*state)->src; 5471 } else { 5472 src = &(*state)->src; 5473 dst = &(*state)->dst; 5474 } 5475 5476 if (src->wscale && dst->wscale) 5477 dws = dst->wscale & PF_WSCALE_MASK; 5478 else 5479 dws = 0; 5480 5481 /* Demodulate sequence number */ 5482 seq = ntohl(th.th_seq) - src->seqdiff; 5483 if (src->seqdiff) { 5484 pf_change_a(&th.th_seq, icmpsum, 5485 htonl(seq), 0); 5486 copyback = 1; 5487 } 5488 5489 if (!SEQ_GEQ(src->seqhi, seq) || 5490 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) { 5491 if (pf_status.debug >= PF_DEBUG_MISC) { 5492 printf("pf: BAD ICMP %d:%d ", 5493 icmptype, pd->hdr.icmp->icmp_code); 5494 pf_print_host(pd->src, 0, pd->af); 5495 printf(" -> "); 5496 pf_print_host(pd->dst, 0, pd->af); 5497 printf(" state: "); 5498 pf_print_state(*state); 5499 printf(" seq=%u\n", seq); 5500 } 5501 REASON_SET(reason, PFRES_BADSTATE); 5502 return (PF_DROP); 5503 } 5504 5505 if (STATE_TRANSLATE(*state)) { 5506 if (direction == PF_IN) { 5507 pf_change_icmp(pd2.src, &th.th_sport, 5508 daddr, &(*state)->lan.addr, 5509 (*state)->lan.port, NULL, 5510 pd2.ip_sum, icmpsum, 5511 pd->ip_sum, 0, pd2.af); 5512 } else { 5513 pf_change_icmp(pd2.dst, &th.th_dport, 5514 saddr, &(*state)->gwy.addr, 5515 (*state)->gwy.port, NULL, 5516 pd2.ip_sum, icmpsum, 5517 pd->ip_sum, 0, pd2.af); 5518 } 5519 copyback = 1; 5520 } 5521 5522 if (copyback) { 5523 switch (pd2.af) { 5524#ifdef INET 5525 case AF_INET: 5526 m_copyback(m, off, ICMP_MINLEN, 5527 (caddr_t)pd->hdr.icmp); 5528 m_copyback(m, ipoff2, sizeof(h2), 5529 (caddr_t)&h2); 5530 break; 5531#endif /* INET */ 5532#ifdef INET6 5533 case AF_INET6: 5534 m_copyback(m, off, 5535 sizeof(struct icmp6_hdr), 5536 (caddr_t)pd->hdr.icmp6); 5537 m_copyback(m, ipoff2, sizeof(h2_6), 5538 (caddr_t)&h2_6); 5539 break; 5540#endif /* INET6 */ 5541 } 5542 m_copyback(m, off2, 8, (caddr_t)&th); 5543 } 5544 5545 return (PF_PASS); 5546 break; 5547 } 5548 case IPPROTO_UDP: { 5549 struct udphdr uh; 5550 5551 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh), 5552 NULL, reason, pd2.af)) { 5553 DPFPRINTF(PF_DEBUG_MISC, 5554 ("pf: ICMP error message too short " 5555 "(udp)\n")); 5556 return (PF_DROP); 5557 } 5558 5559 key.af = pd2.af; 5560 key.proto = IPPROTO_UDP; 5561 if (direction == PF_IN) { 5562 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 5563 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 5564 key.ext.port = uh.uh_dport; 5565 key.gwy.port = uh.uh_sport; 5566 } else { 5567 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 5568 PF_ACPY(&key.ext.addr, pd2.src, key.af); 5569 key.lan.port = uh.uh_dport; 5570 key.ext.port = uh.uh_sport; 5571 } 5572 5573 STATE_LOOKUP(); 5574 5575 if (STATE_TRANSLATE(*state)) { 5576 if (direction == PF_IN) { 5577 pf_change_icmp(pd2.src, &uh.uh_sport, 5578 daddr, &(*state)->lan.addr, 5579 (*state)->lan.port, &uh.uh_sum, 5580 pd2.ip_sum, icmpsum, 5581 pd->ip_sum, 1, pd2.af); 5582 } else { 5583 pf_change_icmp(pd2.dst, &uh.uh_dport, 5584 saddr, &(*state)->gwy.addr, 5585 (*state)->gwy.port, &uh.uh_sum, 5586 pd2.ip_sum, icmpsum, 5587 pd->ip_sum, 1, pd2.af); 5588 } 5589 switch (pd2.af) { 5590#ifdef INET 5591 case AF_INET: 5592 m_copyback(m, off, ICMP_MINLEN, 5593 (caddr_t)pd->hdr.icmp); 5594 m_copyback(m, ipoff2, sizeof(h2), 5595 (caddr_t)&h2); 5596 break; 5597#endif /* INET */ 5598#ifdef INET6 5599 case AF_INET6: 5600 m_copyback(m, off, 5601 sizeof(struct icmp6_hdr), 5602 (caddr_t)pd->hdr.icmp6); 5603 m_copyback(m, ipoff2, sizeof(h2_6), 5604 (caddr_t)&h2_6); 5605 break; 5606#endif /* INET6 */ 5607 } 5608 m_copyback(m, off2, sizeof(uh), 5609 (caddr_t)&uh); 5610 } 5611 5612 return (PF_PASS); 5613 break; 5614 } 5615#ifdef INET 5616 case IPPROTO_ICMP: { 5617 struct icmp iih; 5618 5619 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN, 5620 NULL, reason, pd2.af)) { 5621 DPFPRINTF(PF_DEBUG_MISC, 5622 ("pf: ICMP error message too short i" 5623 "(icmp)\n")); 5624 return (PF_DROP); 5625 } 5626 5627 key.af = pd2.af; 5628 key.proto = IPPROTO_ICMP; 5629 if (direction == PF_IN) { 5630 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 5631 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 5632 key.ext.port = 0; 5633 key.gwy.port = iih.icmp_id; 5634 } else { 5635 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 5636 PF_ACPY(&key.ext.addr, pd2.src, key.af); 5637 key.lan.port = iih.icmp_id; 5638 key.ext.port = 0; 5639 } 5640 5641 STATE_LOOKUP(); 5642 5643 if (STATE_TRANSLATE(*state)) { 5644 if (direction == PF_IN) { 5645 pf_change_icmp(pd2.src, &iih.icmp_id, 5646 daddr, &(*state)->lan.addr, 5647 (*state)->lan.port, NULL, 5648 pd2.ip_sum, icmpsum, 5649 pd->ip_sum, 0, AF_INET); 5650 } else { 5651 pf_change_icmp(pd2.dst, &iih.icmp_id, 5652 saddr, &(*state)->gwy.addr, 5653 (*state)->gwy.port, NULL, 5654 pd2.ip_sum, icmpsum, 5655 pd->ip_sum, 0, AF_INET); 5656 } 5657 m_copyback(m, off, ICMP_MINLEN, 5658 (caddr_t)pd->hdr.icmp); 5659 m_copyback(m, ipoff2, sizeof(h2), 5660 (caddr_t)&h2); 5661 m_copyback(m, off2, ICMP_MINLEN, 5662 (caddr_t)&iih); 5663 } 5664 5665 return (PF_PASS); 5666 break; 5667 } 5668#endif /* INET */ 5669#ifdef INET6 5670 case IPPROTO_ICMPV6: { 5671 struct icmp6_hdr iih; 5672 5673 if (!pf_pull_hdr(m, off2, &iih, 5674 sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) { 5675 DPFPRINTF(PF_DEBUG_MISC, 5676 ("pf: ICMP error message too short " 5677 "(icmp6)\n")); 5678 return (PF_DROP); 5679 } 5680 5681 key.af = pd2.af; 5682 key.proto = IPPROTO_ICMPV6; 5683 if (direction == PF_IN) { 5684 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 5685 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 5686 key.ext.port = 0; 5687 key.gwy.port = iih.icmp6_id; 5688 } else { 5689 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 5690 PF_ACPY(&key.ext.addr, pd2.src, key.af); 5691 key.lan.port = iih.icmp6_id; 5692 key.ext.port = 0; 5693 } 5694 5695 STATE_LOOKUP(); 5696 5697 if (STATE_TRANSLATE(*state)) { 5698 if (direction == PF_IN) { 5699 pf_change_icmp(pd2.src, &iih.icmp6_id, 5700 daddr, &(*state)->lan.addr, 5701 (*state)->lan.port, NULL, 5702 pd2.ip_sum, icmpsum, 5703 pd->ip_sum, 0, AF_INET6); 5704 } else { 5705 pf_change_icmp(pd2.dst, &iih.icmp6_id, 5706 saddr, &(*state)->gwy.addr, 5707 (*state)->gwy.port, NULL, 5708 pd2.ip_sum, icmpsum, 5709 pd->ip_sum, 0, AF_INET6); 5710 } 5711 m_copyback(m, off, sizeof(struct icmp6_hdr), 5712 (caddr_t)pd->hdr.icmp6); 5713 m_copyback(m, ipoff2, sizeof(h2_6), 5714 (caddr_t)&h2_6); 5715 m_copyback(m, off2, sizeof(struct icmp6_hdr), 5716 (caddr_t)&iih); 5717 } 5718 5719 return (PF_PASS); 5720 break; 5721 } 5722#endif /* INET6 */ 5723 default: { 5724 key.af = pd2.af; 5725 key.proto = pd2.proto; 5726 if (direction == PF_IN) { 5727 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 5728 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 5729 key.ext.port = 0; 5730 key.gwy.port = 0; 5731 } else { 5732 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 5733 PF_ACPY(&key.ext.addr, pd2.src, key.af); 5734 key.lan.port = 0; 5735 key.ext.port = 0; 5736 } 5737 5738 STATE_LOOKUP(); 5739 5740 if (STATE_TRANSLATE(*state)) { 5741 if (direction == PF_IN) { 5742 pf_change_icmp(pd2.src, NULL, 5743 daddr, &(*state)->lan.addr, 5744 0, NULL, 5745 pd2.ip_sum, icmpsum, 5746 pd->ip_sum, 0, pd2.af); 5747 } else { 5748 pf_change_icmp(pd2.dst, NULL, 5749 saddr, &(*state)->gwy.addr, 5750 0, NULL, 5751 pd2.ip_sum, icmpsum, 5752 pd->ip_sum, 0, pd2.af); 5753 } 5754 switch (pd2.af) { 5755#ifdef INET 5756 case AF_INET: 5757 m_copyback(m, off, ICMP_MINLEN, 5758 (caddr_t)pd->hdr.icmp); 5759 m_copyback(m, ipoff2, sizeof(h2), 5760 (caddr_t)&h2); 5761 break; 5762#endif /* INET */ 5763#ifdef INET6 5764 case AF_INET6: 5765 m_copyback(m, off, 5766 sizeof(struct icmp6_hdr), 5767 (caddr_t)pd->hdr.icmp6); 5768 m_copyback(m, ipoff2, sizeof(h2_6), 5769 (caddr_t)&h2_6); 5770 break; 5771#endif /* INET6 */ 5772 } 5773 } 5774 5775 return (PF_PASS); 5776 break; 5777 } 5778 } 5779 } 5780} 5781 5782int 5783pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif, 5784 struct pf_pdesc *pd) 5785{ 5786 struct pf_state_peer *src, *dst; 5787 struct pf_state_cmp key; 5788 5789 key.af = pd->af; 5790 key.proto = pd->proto; 5791 if (direction == PF_IN) { 5792 PF_ACPY(&key.ext.addr, pd->src, key.af); 5793 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 5794 key.ext.port = 0; 5795 key.gwy.port = 0; 5796 } else { 5797 PF_ACPY(&key.lan.addr, pd->src, key.af); 5798 PF_ACPY(&key.ext.addr, pd->dst, key.af); 5799 key.lan.port = 0; 5800 key.ext.port = 0; 5801 } 5802 5803 STATE_LOOKUP(); 5804 5805 if (direction == (*state)->direction) { 5806 src = &(*state)->src; 5807 dst = &(*state)->dst; 5808 } else { 5809 src = &(*state)->dst; 5810 dst = &(*state)->src; 5811 } 5812 5813 /* update states */ 5814 if (src->state < PFOTHERS_SINGLE) 5815 src->state = PFOTHERS_SINGLE; 5816 if (dst->state == PFOTHERS_SINGLE) 5817 dst->state = PFOTHERS_MULTIPLE; 5818 5819 /* update expire time */ 5820 (*state)->expire = time_second; 5821 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE) 5822 (*state)->timeout = PFTM_OTHER_MULTIPLE; 5823 else 5824 (*state)->timeout = PFTM_OTHER_SINGLE; 5825 5826 /* translate source/destination address, if necessary */ 5827 if (STATE_TRANSLATE(*state)) { 5828 if (direction == PF_OUT) 5829 switch (pd->af) { 5830#ifdef INET 5831 case AF_INET: 5832 pf_change_a(&pd->src->v4.s_addr, 5833 pd->ip_sum, (*state)->gwy.addr.v4.s_addr, 5834 0); 5835 break; 5836#endif /* INET */ 5837#ifdef INET6 5838 case AF_INET6: 5839 PF_ACPY(pd->src, &(*state)->gwy.addr, pd->af); 5840 break; 5841#endif /* INET6 */ 5842 } 5843 else 5844 switch (pd->af) { 5845#ifdef INET 5846 case AF_INET: 5847 pf_change_a(&pd->dst->v4.s_addr, 5848 pd->ip_sum, (*state)->lan.addr.v4.s_addr, 5849 0); 5850 break; 5851#endif /* INET */ 5852#ifdef INET6 5853 case AF_INET6: 5854 PF_ACPY(pd->dst, &(*state)->lan.addr, pd->af); 5855 break; 5856#endif /* INET6 */ 5857 } 5858 } 5859 5860 return (PF_PASS); 5861} 5862 5863/* 5864 * ipoff and off are measured from the start of the mbuf chain. 5865 * h must be at "ipoff" on the mbuf chain. 5866 */ 5867void * 5868pf_pull_hdr(struct mbuf *m, int off, void *p, int len, 5869 u_short *actionp, u_short *reasonp, sa_family_t af) 5870{ 5871 switch (af) { 5872#ifdef INET 5873 case AF_INET: { 5874 struct ip *h = mtod(m, struct ip *); 5875 u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3; 5876 5877 if (fragoff) { 5878 if (fragoff >= len) 5879 ACTION_SET(actionp, PF_PASS); 5880 else { 5881 ACTION_SET(actionp, PF_DROP); 5882 REASON_SET(reasonp, PFRES_FRAG); 5883 } 5884 return (NULL); 5885 } 5886 if (m->m_pkthdr.len < off + len || 5887 ntohs(h->ip_len) < off + len) { 5888 ACTION_SET(actionp, PF_DROP); 5889 REASON_SET(reasonp, PFRES_SHORT); 5890 return (NULL); 5891 } 5892 break; 5893 } 5894#endif /* INET */ 5895#ifdef INET6 5896 case AF_INET6: { 5897 struct ip6_hdr *h = mtod(m, struct ip6_hdr *); 5898 5899 if (m->m_pkthdr.len < off + len || 5900 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) < 5901 (unsigned)(off + len)) { 5902 ACTION_SET(actionp, PF_DROP); 5903 REASON_SET(reasonp, PFRES_SHORT); 5904 return (NULL); 5905 } 5906 break; 5907 } 5908#endif /* INET6 */ 5909 } 5910 m_copydata(m, off, len, p); 5911 return (p); 5912} 5913 5914int 5915pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif) 5916{ 5917 struct sockaddr_in *dst; 5918 int ret = 1; 5919 int check_mpath; 5920#ifndef __FreeBSD__ 5921 extern int ipmultipath; 5922#endif 5923#ifdef INET6 5924#ifndef __FreeBSD__ 5925 extern int ip6_multipath; 5926#endif 5927 struct sockaddr_in6 *dst6; 5928 struct route_in6 ro; 5929#else 5930 struct route ro; 5931#endif 5932 struct radix_node *rn; 5933 struct rtentry *rt; 5934 struct ifnet *ifp; 5935 5936 check_mpath = 0; 5937 bzero(&ro, sizeof(ro)); 5938 switch (af) { 5939 case AF_INET: 5940 dst = satosin(&ro.ro_dst); 5941 dst->sin_family = AF_INET; 5942 dst->sin_len = sizeof(*dst); 5943 dst->sin_addr = addr->v4; 5944#ifndef __FreeBSD__ /* MULTIPATH_ROUTING */ 5945 if (ipmultipath) 5946 check_mpath = 1; 5947#endif 5948 break; 5949#ifdef INET6 5950 case AF_INET6: 5951 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 5952 dst6->sin6_family = AF_INET6; 5953 dst6->sin6_len = sizeof(*dst6); 5954 dst6->sin6_addr = addr->v6; 5955#ifndef __FreeBSD__ /* MULTIPATH_ROUTING */ 5956 if (ip6_multipath) 5957 check_mpath = 1; 5958#endif 5959 break; 5960#endif /* INET6 */ 5961 default: 5962 return (0); 5963 } 5964 5965 /* Skip checks for ipsec interfaces */ 5966 if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC) 5967 goto out; 5968 5969#ifdef __FreeBSD__ 5970/* XXX MRT not always INET */ /* stick with table 0 though */ 5971 if (af == AF_INET) 5972 in_rtalloc_ign((struct route *)&ro, RTF_CLONING, 0); 5973 else 5974 rtalloc_ign((struct route *)&ro, RTF_CLONING); 5975#else /* ! __FreeBSD__ */ 5976 rtalloc_noclone((struct route *)&ro, NO_CLONING); 5977#endif 5978 5979 if (ro.ro_rt != NULL) { 5980 /* No interface given, this is a no-route check */ 5981 if (kif == NULL) 5982 goto out; 5983 5984 if (kif->pfik_ifp == NULL) { 5985 ret = 0; 5986 goto out; 5987 } 5988 5989 /* Perform uRPF check if passed input interface */ 5990 ret = 0; 5991 rn = (struct radix_node *)ro.ro_rt; 5992 do { 5993 rt = (struct rtentry *)rn; 5994#ifndef __FreeBSD__ /* CARPDEV */ 5995 if (rt->rt_ifp->if_type == IFT_CARP) 5996 ifp = rt->rt_ifp->if_carpdev; 5997 else 5998#endif 5999 ifp = rt->rt_ifp; 6000 6001 if (kif->pfik_ifp == ifp) 6002 ret = 1; 6003#ifdef __FreeBSD__ /* MULTIPATH_ROUTING */ 6004 rn = NULL; 6005#else 6006 rn = rn_mpath_next(rn); 6007#endif 6008 } while (check_mpath == 1 && rn != NULL && ret == 0); 6009 } else 6010 ret = 0; 6011out: 6012 if (ro.ro_rt != NULL) 6013 RTFREE(ro.ro_rt); 6014 return (ret); 6015} 6016 6017int 6018pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw) 6019{ 6020 struct sockaddr_in *dst; 6021#ifdef INET6 6022 struct sockaddr_in6 *dst6; 6023 struct route_in6 ro; 6024#else 6025 struct route ro; 6026#endif 6027 int ret = 0; 6028 6029 bzero(&ro, sizeof(ro)); 6030 switch (af) { 6031 case AF_INET: 6032 dst = satosin(&ro.ro_dst); 6033 dst->sin_family = AF_INET; 6034 dst->sin_len = sizeof(*dst); 6035 dst->sin_addr = addr->v4; 6036 break; 6037#ifdef INET6 6038 case AF_INET6: 6039 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 6040 dst6->sin6_family = AF_INET6; 6041 dst6->sin6_len = sizeof(*dst6); 6042 dst6->sin6_addr = addr->v6; 6043 break; 6044#endif /* INET6 */ 6045 default: 6046 return (0); 6047 } 6048 6049#ifdef __FreeBSD__ 6050# ifdef RTF_PRCLONING 6051 rtalloc_ign((struct route *)&ro, (RTF_CLONING|RTF_PRCLONING)); 6052# else /* !RTF_PRCLONING */ 6053 if (af == AF_INET) 6054 in_rtalloc_ign((struct route *)&ro, RTF_CLONING, 0); 6055 else 6056 rtalloc_ign((struct route *)&ro, RTF_CLONING); 6057# endif 6058#else /* ! __FreeBSD__ */ 6059 rtalloc_noclone((struct route *)&ro, NO_CLONING); 6060#endif 6061 6062 if (ro.ro_rt != NULL) { 6063#ifdef __FreeBSD__ 6064 /* XXX_IMPORT: later */ 6065#else 6066 if (ro.ro_rt->rt_labelid == aw->v.rtlabel) 6067 ret = 1; 6068#endif 6069 RTFREE(ro.ro_rt); 6070 } 6071 6072 return (ret); 6073} 6074 6075#ifdef INET 6076 6077void 6078pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp, 6079 struct pf_state *s, struct pf_pdesc *pd) 6080{ 6081 struct mbuf *m0, *m1; 6082 struct route iproute; 6083 struct route *ro = NULL; 6084 struct sockaddr_in *dst; 6085 struct ip *ip; 6086 struct ifnet *ifp = NULL; 6087 struct pf_addr naddr; 6088 struct pf_src_node *sn = NULL; 6089 int error = 0; 6090#ifdef __FreeBSD__ 6091 int sw_csum; 6092#endif 6093#ifdef IPSEC 6094 struct m_tag *mtag; 6095#endif /* IPSEC */ 6096 6097 if (m == NULL || *m == NULL || r == NULL || 6098 (dir != PF_IN && dir != PF_OUT) || oifp == NULL) 6099 panic("pf_route: invalid parameters"); 6100 6101 if (pd->pf_mtag->routed++ > 3) { 6102 m0 = *m; 6103 *m = NULL; 6104 goto bad; 6105 } 6106 6107 if (r->rt == PF_DUPTO) { 6108#ifdef __FreeBSD__ 6109 if ((m0 = m_dup(*m, M_DONTWAIT)) == NULL) 6110#else 6111 if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL) 6112#endif 6113 return; 6114 } else { 6115 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) 6116 return; 6117 m0 = *m; 6118 } 6119 6120 if (m0->m_len < sizeof(struct ip)) { 6121 DPFPRINTF(PF_DEBUG_URGENT, 6122 ("pf_route: m0->m_len < sizeof(struct ip)\n")); 6123 goto bad; 6124 } 6125 6126 ip = mtod(m0, struct ip *); 6127 6128 ro = &iproute; 6129 bzero((caddr_t)ro, sizeof(*ro)); 6130 dst = satosin(&ro->ro_dst); 6131 dst->sin_family = AF_INET; 6132 dst->sin_len = sizeof(*dst); 6133 dst->sin_addr = ip->ip_dst; 6134 6135 if (r->rt == PF_FASTROUTE) { 6136 in_rtalloc(ro, 0); 6137 if (ro->ro_rt == 0) { 6138 ipstat.ips_noroute++; 6139 goto bad; 6140 } 6141 6142 ifp = ro->ro_rt->rt_ifp; 6143 ro->ro_rt->rt_use++; 6144 6145 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 6146 dst = satosin(ro->ro_rt->rt_gateway); 6147 } else { 6148 if (TAILQ_EMPTY(&r->rpool.list)) { 6149 DPFPRINTF(PF_DEBUG_URGENT, 6150 ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n")); 6151 goto bad; 6152 } 6153 if (s == NULL) { 6154 pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src, 6155 &naddr, NULL, &sn); 6156 if (!PF_AZERO(&naddr, AF_INET)) 6157 dst->sin_addr.s_addr = naddr.v4.s_addr; 6158 ifp = r->rpool.cur->kif ? 6159 r->rpool.cur->kif->pfik_ifp : NULL; 6160 } else { 6161 if (!PF_AZERO(&s->rt_addr, AF_INET)) 6162 dst->sin_addr.s_addr = 6163 s->rt_addr.v4.s_addr; 6164 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; 6165 } 6166 } 6167 if (ifp == NULL) 6168 goto bad; 6169 6170 if (oifp != ifp) { 6171#ifdef __FreeBSD__ 6172 PF_UNLOCK(); 6173 if (pf_test(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) { 6174 PF_LOCK(); 6175 goto bad; 6176 } else if (m0 == NULL) { 6177 PF_LOCK(); 6178 goto done; 6179 } 6180 PF_LOCK(); 6181#else 6182 if (pf_test(PF_OUT, ifp, &m0, NULL) != PF_PASS) 6183 goto bad; 6184 else if (m0 == NULL) 6185 goto done; 6186#endif 6187 if (m0->m_len < sizeof(struct ip)) { 6188 DPFPRINTF(PF_DEBUG_URGENT, 6189 ("pf_route: m0->m_len < sizeof(struct ip)\n")); 6190 goto bad; 6191 } 6192 ip = mtod(m0, struct ip *); 6193 } 6194 6195#ifdef __FreeBSD__ 6196 /* Copied from FreeBSD 5.1-CURRENT ip_output. */ 6197 m0->m_pkthdr.csum_flags |= CSUM_IP; 6198 sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist; 6199 if (sw_csum & CSUM_DELAY_DATA) { 6200 /* 6201 * XXX: in_delayed_cksum assumes HBO for ip->ip_len (at least) 6202 */ 6203 NTOHS(ip->ip_len); 6204 NTOHS(ip->ip_off); /* XXX: needed? */ 6205 in_delayed_cksum(m0); 6206 HTONS(ip->ip_len); 6207 HTONS(ip->ip_off); 6208 sw_csum &= ~CSUM_DELAY_DATA; 6209 } 6210 m0->m_pkthdr.csum_flags &= ifp->if_hwassist; 6211 6212 if (ntohs(ip->ip_len) <= ifp->if_mtu || 6213 (ifp->if_hwassist & CSUM_FRAGMENT && 6214 ((ip->ip_off & htons(IP_DF)) == 0))) { 6215 /* 6216 * ip->ip_len = htons(ip->ip_len); 6217 * ip->ip_off = htons(ip->ip_off); 6218 */ 6219 ip->ip_sum = 0; 6220 if (sw_csum & CSUM_DELAY_IP) { 6221 /* From KAME */ 6222 if (ip->ip_v == IPVERSION && 6223 (ip->ip_hl << 2) == sizeof(*ip)) { 6224 ip->ip_sum = in_cksum_hdr(ip); 6225 } else { 6226 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2); 6227 } 6228 } 6229 PF_UNLOCK(); 6230 error = (*ifp->if_output)(ifp, m0, sintosa(dst), ro->ro_rt); 6231 PF_LOCK(); 6232 goto done; 6233 } 6234 6235#else 6236 /* Copied from ip_output. */ 6237#ifdef IPSEC 6238 /* 6239 * If deferred crypto processing is needed, check that the 6240 * interface supports it. 6241 */ 6242 if ((mtag = m_tag_find(m0, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL)) 6243 != NULL && (ifp->if_capabilities & IFCAP_IPSEC) == 0) { 6244 /* Notify IPsec to do its own crypto. */ 6245 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 6246 goto bad; 6247 } 6248#endif /* IPSEC */ 6249 6250 /* Catch routing changes wrt. hardware checksumming for TCP or UDP. */ 6251 if (m0->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT) { 6252 if (!(ifp->if_capabilities & IFCAP_CSUM_TCPv4) || 6253 ifp->if_bridge != NULL) { 6254 in_delayed_cksum(m0); 6255 m0->m_pkthdr.csum_flags &= ~M_TCPV4_CSUM_OUT; /* Clear */ 6256 } 6257 } else if (m0->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT) { 6258 if (!(ifp->if_capabilities & IFCAP_CSUM_UDPv4) || 6259 ifp->if_bridge != NULL) { 6260 in_delayed_cksum(m0); 6261 m0->m_pkthdr.csum_flags &= ~M_UDPV4_CSUM_OUT; /* Clear */ 6262 } 6263 } 6264 6265 if (ntohs(ip->ip_len) <= ifp->if_mtu) { 6266 if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) && 6267 ifp->if_bridge == NULL) { 6268 m0->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT; 6269 ipstat.ips_outhwcsum++; 6270 } else { 6271 ip->ip_sum = 0; 6272 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2); 6273 } 6274 /* Update relevant hardware checksum stats for TCP/UDP */ 6275 if (m0->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT) 6276 tcpstat.tcps_outhwcsum++; 6277 else if (m0->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT) 6278 udpstat.udps_outhwcsum++; 6279 error = (*ifp->if_output)(ifp, m0, sintosa(dst), NULL); 6280 goto done; 6281 } 6282#endif 6283 /* 6284 * Too large for interface; fragment if possible. 6285 * Must be able to put at least 8 bytes per fragment. 6286 */ 6287 if (ip->ip_off & htons(IP_DF)) { 6288 ipstat.ips_cantfrag++; 6289 if (r->rt != PF_DUPTO) { 6290#ifdef __FreeBSD__ 6291 /* icmp_error() expects host byte ordering */ 6292 NTOHS(ip->ip_len); 6293 NTOHS(ip->ip_off); 6294 PF_UNLOCK(); 6295 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0, 6296 ifp->if_mtu); 6297 PF_LOCK(); 6298#else 6299 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0, 6300 ifp->if_mtu); 6301#endif 6302 goto done; 6303 } else 6304 goto bad; 6305 } 6306 6307 m1 = m0; 6308#ifdef __FreeBSD__ 6309 /* 6310 * XXX: is cheaper + less error prone than own function 6311 */ 6312 NTOHS(ip->ip_len); 6313 NTOHS(ip->ip_off); 6314 error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum); 6315#else 6316 error = ip_fragment(m0, ifp, ifp->if_mtu); 6317#endif 6318 if (error) { 6319#ifndef __FreeBSD__ /* ip_fragment does not do m_freem() on FreeBSD */ 6320 m0 = NULL; 6321#endif 6322 goto bad; 6323 } 6324 6325 for (m0 = m1; m0; m0 = m1) { 6326 m1 = m0->m_nextpkt; 6327 m0->m_nextpkt = 0; 6328#ifdef __FreeBSD__ 6329 if (error == 0) { 6330 PF_UNLOCK(); 6331 error = (*ifp->if_output)(ifp, m0, sintosa(dst), 6332 NULL); 6333 PF_LOCK(); 6334 } else 6335#else 6336 if (error == 0) 6337 error = (*ifp->if_output)(ifp, m0, sintosa(dst), 6338 NULL); 6339 else 6340#endif 6341 m_freem(m0); 6342 } 6343 6344 if (error == 0) 6345 ipstat.ips_fragmented++; 6346 6347done: 6348 if (r->rt != PF_DUPTO) 6349 *m = NULL; 6350 if (ro == &iproute && ro->ro_rt) 6351 RTFREE(ro->ro_rt); 6352 return; 6353 6354bad: 6355 m_freem(m0); 6356 goto done; 6357} 6358#endif /* INET */ 6359 6360#ifdef INET6 6361void 6362pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp, 6363 struct pf_state *s, struct pf_pdesc *pd) 6364{ 6365 struct mbuf *m0; 6366 struct route_in6 ip6route; 6367 struct route_in6 *ro; 6368 struct sockaddr_in6 *dst; 6369 struct ip6_hdr *ip6; 6370 struct ifnet *ifp = NULL; 6371 struct pf_addr naddr; 6372 struct pf_src_node *sn = NULL; 6373 int error = 0; 6374 6375 if (m == NULL || *m == NULL || r == NULL || 6376 (dir != PF_IN && dir != PF_OUT) || oifp == NULL) 6377 panic("pf_route6: invalid parameters"); 6378 6379 if (pd->pf_mtag->routed++ > 3) { 6380 m0 = *m; 6381 *m = NULL; 6382 goto bad; 6383 } 6384 6385 if (r->rt == PF_DUPTO) { 6386#ifdef __FreeBSD__ 6387 if ((m0 = m_dup(*m, M_DONTWAIT)) == NULL) 6388#else 6389 if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL) 6390#endif 6391 return; 6392 } else { 6393 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) 6394 return; 6395 m0 = *m; 6396 } 6397 6398 if (m0->m_len < sizeof(struct ip6_hdr)) { 6399 DPFPRINTF(PF_DEBUG_URGENT, 6400 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n")); 6401 goto bad; 6402 } 6403 ip6 = mtod(m0, struct ip6_hdr *); 6404 6405 ro = &ip6route; 6406 bzero((caddr_t)ro, sizeof(*ro)); 6407 dst = (struct sockaddr_in6 *)&ro->ro_dst; 6408 dst->sin6_family = AF_INET6; 6409 dst->sin6_len = sizeof(*dst); 6410 dst->sin6_addr = ip6->ip6_dst; 6411 6412 /* Cheat. XXX why only in the v6 case??? */ 6413 if (r->rt == PF_FASTROUTE) { 6414#ifdef __FreeBSD__ 6415 m0->m_flags |= M_SKIP_FIREWALL; 6416 PF_UNLOCK(); 6417 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL); 6418 PF_LOCK(); 6419#else 6420 mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT); 6421 if (mtag == NULL) 6422 goto bad; 6423 m_tag_prepend(m0, mtag); 6424 pd->pf_mtag->flags |= PF_TAG_GENERATED; 6425 ip6_output(m0, NULL, NULL, 0, NULL, NULL); 6426#endif 6427 return; 6428 } 6429 6430 if (TAILQ_EMPTY(&r->rpool.list)) { 6431 DPFPRINTF(PF_DEBUG_URGENT, 6432 ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n")); 6433 goto bad; 6434 } 6435 if (s == NULL) { 6436 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src, 6437 &naddr, NULL, &sn); 6438 if (!PF_AZERO(&naddr, AF_INET6)) 6439 PF_ACPY((struct pf_addr *)&dst->sin6_addr, 6440 &naddr, AF_INET6); 6441 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL; 6442 } else { 6443 if (!PF_AZERO(&s->rt_addr, AF_INET6)) 6444 PF_ACPY((struct pf_addr *)&dst->sin6_addr, 6445 &s->rt_addr, AF_INET6); 6446 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; 6447 } 6448 if (ifp == NULL) 6449 goto bad; 6450 6451 if (oifp != ifp) { 6452#ifdef __FreeBSD__ 6453 PF_UNLOCK(); 6454 if (pf_test6(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) { 6455 PF_LOCK(); 6456 goto bad; 6457 } else if (m0 == NULL) { 6458 PF_LOCK(); 6459 goto done; 6460 } 6461 PF_LOCK(); 6462#else 6463 if (pf_test6(PF_OUT, ifp, &m0, NULL) != PF_PASS) 6464 goto bad; 6465 else if (m0 == NULL) 6466 goto done; 6467#endif 6468 if (m0->m_len < sizeof(struct ip6_hdr)) { 6469 DPFPRINTF(PF_DEBUG_URGENT, 6470 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n")); 6471 goto bad; 6472 } 6473 ip6 = mtod(m0, struct ip6_hdr *); 6474 } 6475 6476 /* 6477 * If the packet is too large for the outgoing interface, 6478 * send back an icmp6 error. 6479 */ 6480 if (IN6_IS_SCOPE_EMBED(&dst->sin6_addr)) 6481 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index); 6482 if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) { 6483#ifdef __FreeBSD__ 6484 PF_UNLOCK(); 6485#endif 6486 error = nd6_output(ifp, ifp, m0, dst, NULL); 6487#ifdef __FreeBSD__ 6488 PF_LOCK(); 6489#endif 6490 } else { 6491 in6_ifstat_inc(ifp, ifs6_in_toobig); 6492#ifdef __FreeBSD__ 6493 if (r->rt != PF_DUPTO) { 6494 PF_UNLOCK(); 6495 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu); 6496 PF_LOCK(); 6497 } else 6498#else 6499 if (r->rt != PF_DUPTO) 6500 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu); 6501 else 6502#endif 6503 goto bad; 6504 } 6505 6506done: 6507 if (r->rt != PF_DUPTO) 6508 *m = NULL; 6509 return; 6510 6511bad: 6512 m_freem(m0); 6513 goto done; 6514} 6515#endif /* INET6 */ 6516 6517 6518#ifdef __FreeBSD__ 6519/* 6520 * FreeBSD supports cksum offloads for the following drivers. 6521 * em(4), fxp(4), ixgb(4), lge(4), ndis(4), nge(4), re(4), 6522 * ti(4), txp(4), xl(4) 6523 * 6524 * CSUM_DATA_VALID | CSUM_PSEUDO_HDR : 6525 * network driver performed cksum including pseudo header, need to verify 6526 * csum_data 6527 * CSUM_DATA_VALID : 6528 * network driver performed cksum, needs to additional pseudo header 6529 * cksum computation with partial csum_data(i.e. lack of H/W support for 6530 * pseudo header, for instance hme(4), sk(4) and possibly gem(4)) 6531 * 6532 * After validating the cksum of packet, set both flag CSUM_DATA_VALID and 6533 * CSUM_PSEUDO_HDR in order to avoid recomputation of the cksum in upper 6534 * TCP/UDP layer. 6535 * Also, set csum_data to 0xffff to force cksum validation. 6536 */ 6537int 6538pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p, sa_family_t af) 6539{ 6540 u_int16_t sum = 0; 6541 int hw_assist = 0; 6542 struct ip *ip; 6543 6544 if (off < sizeof(struct ip) || len < sizeof(struct udphdr)) 6545 return (1); 6546 if (m->m_pkthdr.len < off + len) 6547 return (1); 6548 6549 switch (p) { 6550 case IPPROTO_TCP: 6551 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { 6552 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) { 6553 sum = m->m_pkthdr.csum_data; 6554 } else { 6555 ip = mtod(m, struct ip *); 6556 sum = in_pseudo(ip->ip_src.s_addr, 6557 ip->ip_dst.s_addr, htonl((u_short)len + 6558 m->m_pkthdr.csum_data + IPPROTO_TCP)); 6559 } 6560 sum ^= 0xffff; 6561 ++hw_assist; 6562 } 6563 break; 6564 case IPPROTO_UDP: 6565 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { 6566 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) { 6567 sum = m->m_pkthdr.csum_data; 6568 } else { 6569 ip = mtod(m, struct ip *); 6570 sum = in_pseudo(ip->ip_src.s_addr, 6571 ip->ip_dst.s_addr, htonl((u_short)len + 6572 m->m_pkthdr.csum_data + IPPROTO_UDP)); 6573 } 6574 sum ^= 0xffff; 6575 ++hw_assist; 6576 } 6577 break; 6578 case IPPROTO_ICMP: 6579#ifdef INET6 6580 case IPPROTO_ICMPV6: 6581#endif /* INET6 */ 6582 break; 6583 default: 6584 return (1); 6585 } 6586 6587 if (!hw_assist) { 6588 switch (af) { 6589 case AF_INET: 6590 if (p == IPPROTO_ICMP) { 6591 if (m->m_len < off) 6592 return (1); 6593 m->m_data += off; 6594 m->m_len -= off; 6595 sum = in_cksum(m, len); 6596 m->m_data -= off; 6597 m->m_len += off; 6598 } else { 6599 if (m->m_len < sizeof(struct ip)) 6600 return (1); 6601 sum = in4_cksum(m, p, off, len); 6602 } 6603 break; 6604#ifdef INET6 6605 case AF_INET6: 6606 if (m->m_len < sizeof(struct ip6_hdr)) 6607 return (1); 6608 sum = in6_cksum(m, p, off, len); 6609 break; 6610#endif /* INET6 */ 6611 default: 6612 return (1); 6613 } 6614 } 6615 if (sum) { 6616 switch (p) { 6617 case IPPROTO_TCP: 6618 tcpstat.tcps_rcvbadsum++; 6619 break; 6620 case IPPROTO_UDP: 6621 udpstat.udps_badsum++; 6622 break; 6623 case IPPROTO_ICMP: 6624 icmpstat.icps_checksum++; 6625 break; 6626#ifdef INET6 6627 case IPPROTO_ICMPV6: 6628 icmp6stat.icp6s_checksum++; 6629 break; 6630#endif /* INET6 */ 6631 } 6632 return (1); 6633 } else { 6634 if (p == IPPROTO_TCP || p == IPPROTO_UDP) { 6635 m->m_pkthdr.csum_flags |= 6636 (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 6637 m->m_pkthdr.csum_data = 0xffff; 6638 } 6639 } 6640 return (0); 6641} 6642#else /* !__FreeBSD__ */ 6643/* 6644 * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag 6645 * off is the offset where the protocol header starts 6646 * len is the total length of protocol header plus payload 6647 * returns 0 when the checksum is valid, otherwise returns 1. 6648 */ 6649int 6650pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p, 6651 sa_family_t af) 6652{ 6653 u_int16_t flag_ok, flag_bad; 6654 u_int16_t sum; 6655 6656 switch (p) { 6657 case IPPROTO_TCP: 6658 flag_ok = M_TCP_CSUM_IN_OK; 6659 flag_bad = M_TCP_CSUM_IN_BAD; 6660 break; 6661 case IPPROTO_UDP: 6662 flag_ok = M_UDP_CSUM_IN_OK; 6663 flag_bad = M_UDP_CSUM_IN_BAD; 6664 break; 6665 case IPPROTO_ICMP: 6666#ifdef INET6 6667 case IPPROTO_ICMPV6: 6668#endif /* INET6 */ 6669 flag_ok = flag_bad = 0; 6670 break; 6671 default: 6672 return (1); 6673 } 6674 if (m->m_pkthdr.csum_flags & flag_ok) 6675 return (0); 6676 if (m->m_pkthdr.csum_flags & flag_bad) 6677 return (1); 6678 if (off < sizeof(struct ip) || len < sizeof(struct udphdr)) 6679 return (1); 6680 if (m->m_pkthdr.len < off + len) 6681 return (1); 6682 switch (af) { 6683#ifdef INET 6684 case AF_INET: 6685 if (p == IPPROTO_ICMP) { 6686 if (m->m_len < off) 6687 return (1); 6688 m->m_data += off; 6689 m->m_len -= off; 6690 sum = in_cksum(m, len); 6691 m->m_data -= off; 6692 m->m_len += off; 6693 } else { 6694 if (m->m_len < sizeof(struct ip)) 6695 return (1); 6696 sum = in4_cksum(m, p, off, len); 6697 } 6698 break; 6699#endif /* INET */ 6700#ifdef INET6 6701 case AF_INET6: 6702 if (m->m_len < sizeof(struct ip6_hdr)) 6703 return (1); 6704 sum = in6_cksum(m, p, off, len); 6705 break; 6706#endif /* INET6 */ 6707 default: 6708 return (1); 6709 } 6710 if (sum) { 6711 m->m_pkthdr.csum_flags |= flag_bad; 6712 switch (p) { 6713 case IPPROTO_TCP: 6714 tcpstat.tcps_rcvbadsum++; 6715 break; 6716 case IPPROTO_UDP: 6717 udpstat.udps_badsum++; 6718 break; 6719 case IPPROTO_ICMP: 6720 icmpstat.icps_checksum++; 6721 break; 6722#ifdef INET6 6723 case IPPROTO_ICMPV6: 6724 icmp6stat.icp6s_checksum++; 6725 break; 6726#endif /* INET6 */ 6727 } 6728 return (1); 6729 } 6730 m->m_pkthdr.csum_flags |= flag_ok; 6731 return (0); 6732} 6733#endif /* __FreeBSD__ */ 6734 6735#ifdef INET 6736int 6737#ifdef __FreeBSD__ 6738pf_test(int dir, struct ifnet *ifp, struct mbuf **m0, 6739 struct ether_header *eh, struct inpcb *inp) 6740#else 6741pf_test(int dir, struct ifnet *ifp, struct mbuf **m0, 6742 struct ether_header *eh) 6743#endif 6744{ 6745 struct pfi_kif *kif; 6746 u_short action, reason = 0, log = 0; 6747 struct mbuf *m = *m0; 6748 struct ip *h = NULL; /* make the compiler happy */ 6749 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr; 6750 struct pf_state *s = NULL; 6751 struct pf_ruleset *ruleset = NULL; 6752 struct pf_pdesc pd; 6753 int off, dirndx, pqid = 0; 6754 6755#ifdef __FreeBSD__ 6756 PF_LOCK(); 6757#endif 6758 if (!pf_status.running) 6759#ifdef __FreeBSD__ 6760 { 6761 PF_UNLOCK(); 6762#endif 6763 return (PF_PASS); 6764#ifdef __FreeBSD__ 6765 } 6766#endif 6767 6768 memset(&pd, 0, sizeof(pd)); 6769 if ((pd.pf_mtag = pf_get_mtag(m)) == NULL) { 6770#ifdef __FreeBSD__ 6771 PF_UNLOCK(); 6772#endif 6773 DPFPRINTF(PF_DEBUG_URGENT, 6774 ("pf_test: pf_get_mtag returned NULL\n")); 6775 return (PF_DROP); 6776 } 6777#ifdef __FreeBSD__ 6778 if (m->m_flags & M_SKIP_FIREWALL) { 6779 PF_UNLOCK(); 6780 return (PF_PASS); 6781 } 6782#else 6783 if (pd.pf_mtag->flags & PF_TAG_GENERATED) 6784 return (PF_PASS); 6785#endif 6786 6787#ifdef __FreeBSD__ 6788 /* XXX_IMPORT: later */ 6789#else 6790 if (ifp->if_type == IFT_CARP && ifp->if_carpdev) 6791 ifp = ifp->if_carpdev; 6792#endif 6793 6794 kif = (struct pfi_kif *)ifp->if_pf_kif; 6795 if (kif == NULL) { 6796#ifdef __FreeBSD__ 6797 PF_UNLOCK(); 6798#endif 6799 DPFPRINTF(PF_DEBUG_URGENT, 6800 ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname)); 6801 return (PF_DROP); 6802 } 6803 if (kif->pfik_flags & PFI_IFLAG_SKIP) { 6804#ifdef __FreeBSD__ 6805 PF_UNLOCK(); 6806#endif 6807 return (PF_PASS); 6808 } 6809 6810#ifdef __FreeBSD__ 6811 M_ASSERTPKTHDR(m); 6812#else 6813#ifdef DIAGNOSTIC 6814 if ((m->m_flags & M_PKTHDR) == 0) 6815 panic("non-M_PKTHDR is passed to pf_test"); 6816#endif /* DIAGNOSTIC */ 6817#endif /* __FreeBSD__ */ 6818 6819 if (m->m_pkthdr.len < (int)sizeof(*h)) { 6820 action = PF_DROP; 6821 REASON_SET(&reason, PFRES_SHORT); 6822 log = 1; 6823 goto done; 6824 } 6825 6826 /* We do IP header normalization and packet reassembly here */ 6827 if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) { 6828 action = PF_DROP; 6829 goto done; 6830 } 6831 m = *m0; 6832 h = mtod(m, struct ip *); 6833 6834 off = h->ip_hl << 2; 6835 if (off < (int)sizeof(*h)) { 6836 action = PF_DROP; 6837 REASON_SET(&reason, PFRES_SHORT); 6838 log = 1; 6839 goto done; 6840 } 6841 6842 pd.src = (struct pf_addr *)&h->ip_src; 6843 pd.dst = (struct pf_addr *)&h->ip_dst; 6844 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET); 6845 pd.ip_sum = &h->ip_sum; 6846 pd.proto = h->ip_p; 6847 pd.af = AF_INET; 6848 pd.tos = h->ip_tos; 6849 pd.tot_len = ntohs(h->ip_len); 6850 pd.eh = eh; 6851 6852 /* handle fragments that didn't get reassembled by normalization */ 6853 if (h->ip_off & htons(IP_MF | IP_OFFMASK)) { 6854 action = pf_test_fragment(&r, dir, kif, m, h, 6855 &pd, &a, &ruleset); 6856 goto done; 6857 } 6858 6859 switch (h->ip_p) { 6860 6861 case IPPROTO_TCP: { 6862 struct tcphdr th; 6863 6864 pd.hdr.tcp = &th; 6865 if (!pf_pull_hdr(m, off, &th, sizeof(th), 6866 &action, &reason, AF_INET)) { 6867 log = action != PF_PASS; 6868 goto done; 6869 } 6870 if (dir == PF_IN && pf_check_proto_cksum(m, off, 6871 ntohs(h->ip_len) - off, IPPROTO_TCP, AF_INET)) { 6872 REASON_SET(&reason, PFRES_PROTCKSUM); 6873 action = PF_DROP; 6874 goto done; 6875 } 6876 pd.p_len = pd.tot_len - off - (th.th_off << 2); 6877 if ((th.th_flags & TH_ACK) && pd.p_len == 0) 6878 pqid = 1; 6879 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd); 6880 if (action == PF_DROP) 6881 goto done; 6882 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd, 6883 &reason); 6884 if (action == PF_PASS) { 6885#if NPFSYNC 6886 pfsync_update_state(s); 6887#endif /* NPFSYNC */ 6888 r = s->rule.ptr; 6889 a = s->anchor.ptr; 6890 log = s->log; 6891 } else if (s == NULL) 6892#ifdef __FreeBSD__ 6893 action = pf_test_tcp(&r, &s, dir, kif, 6894 m, off, h, &pd, &a, &ruleset, NULL, inp); 6895#else 6896 action = pf_test_tcp(&r, &s, dir, kif, 6897 m, off, h, &pd, &a, &ruleset, &ipintrq); 6898#endif 6899 break; 6900 } 6901 6902 case IPPROTO_UDP: { 6903 struct udphdr uh; 6904 6905 pd.hdr.udp = &uh; 6906 if (!pf_pull_hdr(m, off, &uh, sizeof(uh), 6907 &action, &reason, AF_INET)) { 6908 log = action != PF_PASS; 6909 goto done; 6910 } 6911 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m, 6912 off, ntohs(h->ip_len) - off, IPPROTO_UDP, AF_INET)) { 6913 action = PF_DROP; 6914 REASON_SET(&reason, PFRES_PROTCKSUM); 6915 goto done; 6916 } 6917 if (uh.uh_dport == 0 || 6918 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off || 6919 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) { 6920 action = PF_DROP; 6921 REASON_SET(&reason, PFRES_SHORT); 6922 goto done; 6923 } 6924 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd); 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_udp(&r, &s, dir, kif, 6935 m, off, h, &pd, &a, &ruleset, NULL, inp); 6936#else 6937 action = pf_test_udp(&r, &s, dir, kif, 6938 m, off, h, &pd, &a, &ruleset, &ipintrq); 6939#endif 6940 break; 6941 } 6942 6943 case IPPROTO_ICMP: { 6944 struct icmp ih; 6945 6946 pd.hdr.icmp = &ih; 6947 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN, 6948 &action, &reason, AF_INET)) { 6949 log = action != PF_PASS; 6950 goto done; 6951 } 6952 if (dir == PF_IN && pf_check_proto_cksum(m, off, 6953 ntohs(h->ip_len) - off, IPPROTO_ICMP, AF_INET)) { 6954 action = PF_DROP; 6955 REASON_SET(&reason, PFRES_PROTCKSUM); 6956 goto done; 6957 } 6958 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd, 6959 &reason); 6960 if (action == PF_PASS) { 6961#if NPFSYNC 6962 pfsync_update_state(s); 6963#endif /* NPFSYNC */ 6964 r = s->rule.ptr; 6965 a = s->anchor.ptr; 6966 log = s->log; 6967 } else if (s == NULL) 6968#ifdef __FreeBSD__ 6969 action = pf_test_icmp(&r, &s, dir, kif, 6970 m, off, h, &pd, &a, &ruleset, NULL); 6971#else 6972 action = pf_test_icmp(&r, &s, dir, kif, 6973 m, off, h, &pd, &a, &ruleset, &ipintrq); 6974#endif 6975 break; 6976 } 6977 6978 default: 6979 action = pf_test_state_other(&s, dir, kif, &pd); 6980 if (action == PF_PASS) { 6981#if NPFSYNC 6982 pfsync_update_state(s); 6983#endif /* NPFSYNC */ 6984 r = s->rule.ptr; 6985 a = s->anchor.ptr; 6986 log = s->log; 6987 } else if (s == NULL) 6988#ifdef __FreeBSD__ 6989 action = pf_test_other(&r, &s, dir, kif, m, off, h, 6990 &pd, &a, &ruleset, NULL); 6991#else 6992 action = pf_test_other(&r, &s, dir, kif, m, off, h, 6993 &pd, &a, &ruleset, &ipintrq); 6994#endif 6995 break; 6996 } 6997 6998done: 6999 if (action == PF_PASS && h->ip_hl > 5 && 7000 !((s && s->allow_opts) || r->allow_opts)) { 7001 action = PF_DROP; 7002 REASON_SET(&reason, PFRES_IPOPTIONS); 7003 log = 1; 7004 DPFPRINTF(PF_DEBUG_MISC, 7005 ("pf: dropping packet with ip options\n")); 7006 } 7007 7008 if ((s && s->tag) || r->rtableid) 7009 pf_tag_packet(m, pd.pf_mtag, s ? s->tag : 0, r->rtableid); 7010 7011#ifdef ALTQ 7012 if (action == PF_PASS && r->qid) { 7013 if (pqid || (pd.tos & IPTOS_LOWDELAY)) 7014 pd.pf_mtag->qid = r->pqid; 7015 else 7016 pd.pf_mtag->qid = r->qid; 7017 /* add hints for ecn */ 7018 pd.pf_mtag->af = AF_INET; 7019 pd.pf_mtag->hdr = h; 7020 } 7021#endif /* ALTQ */ 7022 7023 /* 7024 * connections redirected to loopback should not match sockets 7025 * bound specifically to loopback due to security implications, 7026 * see tcp_input() and in_pcblookup_listen(). 7027 */ 7028 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP || 7029 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL && 7030 (s->nat_rule.ptr->action == PF_RDR || 7031 s->nat_rule.ptr->action == PF_BINAT) && 7032 (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) 7033 pd.pf_mtag->flags |= PF_TAG_TRANSLATE_LOCALHOST; 7034 7035 if (log) { 7036 struct pf_rule *lr; 7037 7038 if (s != NULL && s->nat_rule.ptr != NULL && 7039 s->nat_rule.ptr->log & PF_LOG_ALL) 7040 lr = s->nat_rule.ptr; 7041 else 7042 lr = r; 7043 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset, 7044 &pd); 7045 } 7046 7047 kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len; 7048 kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++; 7049 7050 if (action == PF_PASS || r->action == PF_DROP) { 7051 dirndx = (dir == PF_OUT); 7052 r->packets[dirndx]++; 7053 r->bytes[dirndx] += pd.tot_len; 7054 if (a != NULL) { 7055 a->packets[dirndx]++; 7056 a->bytes[dirndx] += pd.tot_len; 7057 } 7058 if (s != NULL) { 7059 if (s->nat_rule.ptr != NULL) { 7060 s->nat_rule.ptr->packets[dirndx]++; 7061 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len; 7062 } 7063 if (s->src_node != NULL) { 7064 s->src_node->packets[dirndx]++; 7065 s->src_node->bytes[dirndx] += pd.tot_len; 7066 } 7067 if (s->nat_src_node != NULL) { 7068 s->nat_src_node->packets[dirndx]++; 7069 s->nat_src_node->bytes[dirndx] += pd.tot_len; 7070 } 7071 dirndx = (dir == s->direction) ? 0 : 1; 7072 s->packets[dirndx]++; 7073 s->bytes[dirndx] += pd.tot_len; 7074 } 7075 tr = r; 7076 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule; 7077 if (nr != NULL) { 7078 struct pf_addr *x; 7079 /* 7080 * XXX: we need to make sure that the addresses 7081 * passed to pfr_update_stats() are the same than 7082 * the addresses used during matching (pfr_match) 7083 */ 7084 if (r == &pf_default_rule) { 7085 tr = nr; 7086 x = (s == NULL || s->direction == dir) ? 7087 &pd.baddr : &pd.naddr; 7088 } else 7089 x = (s == NULL || s->direction == dir) ? 7090 &pd.naddr : &pd.baddr; 7091 if (x == &pd.baddr || s == NULL) { 7092 /* we need to change the address */ 7093 if (dir == PF_OUT) 7094 pd.src = x; 7095 else 7096 pd.dst = x; 7097 } 7098 } 7099 if (tr->src.addr.type == PF_ADDR_TABLE) 7100 pfr_update_stats(tr->src.addr.p.tbl, (s == NULL || 7101 s->direction == dir) ? pd.src : pd.dst, pd.af, 7102 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 7103 tr->src.neg); 7104 if (tr->dst.addr.type == PF_ADDR_TABLE) 7105 pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL || 7106 s->direction == dir) ? pd.dst : pd.src, pd.af, 7107 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 7108 tr->dst.neg); 7109 } 7110 7111 7112 if (action == PF_SYNPROXY_DROP) { 7113 m_freem(*m0); 7114 *m0 = NULL; 7115 action = PF_PASS; 7116 } else if (r->rt) 7117 /* pf_route can free the mbuf causing *m0 to become NULL */ 7118 pf_route(m0, r, dir, ifp, s, &pd); 7119 7120#ifdef __FreeBSD__ 7121 PF_UNLOCK(); 7122#endif 7123 7124 return (action); 7125} 7126#endif /* INET */ 7127 7128#ifdef INET6 7129int 7130#ifdef __FreeBSD__ 7131pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0, 7132 struct ether_header *eh, struct inpcb *inp) 7133#else 7134pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0, 7135 struct ether_header *eh) 7136#endif 7137{ 7138 struct pfi_kif *kif; 7139 u_short action, reason = 0, log = 0; 7140 struct mbuf *m = *m0, *n = NULL; 7141 struct ip6_hdr *h; 7142 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr; 7143 struct pf_state *s = NULL; 7144 struct pf_ruleset *ruleset = NULL; 7145 struct pf_pdesc pd; 7146 int off, terminal = 0, dirndx, rh_cnt = 0; 7147 7148#ifdef __FreeBSD__ 7149 PF_LOCK(); 7150#endif 7151 7152 if (!pf_status.running) 7153#ifdef __FreeBSD__ 7154 { 7155 PF_UNLOCK(); 7156#endif 7157 return (PF_PASS); 7158#ifdef __FreeBSD__ 7159 } 7160#endif 7161 7162 memset(&pd, 0, sizeof(pd)); 7163 if ((pd.pf_mtag = pf_get_mtag(m)) == NULL) { 7164#ifdef __FreeBSD__ 7165 PF_UNLOCK(); 7166#endif 7167 DPFPRINTF(PF_DEBUG_URGENT, 7168 ("pf_test6: pf_get_mtag returned NULL\n")); 7169 return (PF_DROP); 7170 } 7171 if (pd.pf_mtag->flags & PF_TAG_GENERATED) 7172 return (PF_PASS); 7173 7174#ifdef __FreeBSD__ 7175 /* XXX_IMPORT: later */ 7176#else 7177 if (ifp->if_type == IFT_CARP && ifp->if_carpdev) 7178 ifp = ifp->if_carpdev; 7179#endif 7180 7181 kif = (struct pfi_kif *)ifp->if_pf_kif; 7182 if (kif == NULL) { 7183#ifdef __FreeBSD__ 7184 PF_UNLOCK(); 7185#endif 7186 DPFPRINTF(PF_DEBUG_URGENT, 7187 ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname)); 7188 return (PF_DROP); 7189 } 7190 if (kif->pfik_flags & PFI_IFLAG_SKIP) { 7191#ifdef __FreeBSD__ 7192 PF_UNLOCK(); 7193#endif 7194 return (PF_PASS); 7195 } 7196 7197#ifdef __FreeBSD__ 7198 M_ASSERTPKTHDR(m); 7199#else 7200#ifdef DIAGNOSTIC 7201 if ((m->m_flags & M_PKTHDR) == 0) 7202 panic("non-M_PKTHDR is passed to pf_test6"); 7203#endif /* DIAGNOSTIC */ 7204#endif 7205 7206#ifdef __FreeBSD__ 7207 h = NULL; /* make the compiler happy */ 7208#endif 7209 7210 if (m->m_pkthdr.len < (int)sizeof(*h)) { 7211 action = PF_DROP; 7212 REASON_SET(&reason, PFRES_SHORT); 7213 log = 1; 7214 goto done; 7215 } 7216 7217 /* We do IP header normalization and packet reassembly here */ 7218 if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) { 7219 action = PF_DROP; 7220 goto done; 7221 } 7222 m = *m0; 7223 h = mtod(m, struct ip6_hdr *); 7224 7225#if 1 7226 /* 7227 * we do not support jumbogram yet. if we keep going, zero ip6_plen 7228 * will do something bad, so drop the packet for now. 7229 */ 7230 if (htons(h->ip6_plen) == 0) { 7231 action = PF_DROP; 7232 REASON_SET(&reason, PFRES_NORM); /*XXX*/ 7233 goto done; 7234 } 7235#endif 7236 7237 pd.src = (struct pf_addr *)&h->ip6_src; 7238 pd.dst = (struct pf_addr *)&h->ip6_dst; 7239 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6); 7240 pd.ip_sum = NULL; 7241 pd.af = AF_INET6; 7242 pd.tos = 0; 7243 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr); 7244 pd.eh = eh; 7245 7246 off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr); 7247 pd.proto = h->ip6_nxt; 7248 do { 7249 switch (pd.proto) { 7250 case IPPROTO_FRAGMENT: 7251 action = pf_test_fragment(&r, dir, kif, m, h, 7252 &pd, &a, &ruleset); 7253 if (action == PF_DROP) 7254 REASON_SET(&reason, PFRES_FRAG); 7255 goto done; 7256 case IPPROTO_ROUTING: { 7257 struct ip6_rthdr rthdr; 7258 7259 if (rh_cnt++) { 7260 DPFPRINTF(PF_DEBUG_MISC, 7261 ("pf: IPv6 more than one rthdr\n")); 7262 action = PF_DROP; 7263 REASON_SET(&reason, PFRES_IPOPTIONS); 7264 log = 1; 7265 goto done; 7266 } 7267 if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL, 7268 &reason, pd.af)) { 7269 DPFPRINTF(PF_DEBUG_MISC, 7270 ("pf: IPv6 short rthdr\n")); 7271 action = PF_DROP; 7272 REASON_SET(&reason, PFRES_SHORT); 7273 log = 1; 7274 goto done; 7275 } 7276 if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) { 7277 DPFPRINTF(PF_DEBUG_MISC, 7278 ("pf: IPv6 rthdr0\n")); 7279 action = PF_DROP; 7280 REASON_SET(&reason, PFRES_IPOPTIONS); 7281 log = 1; 7282 goto done; 7283 } 7284 /* fallthrough */ 7285 } 7286 case IPPROTO_AH: 7287 case IPPROTO_HOPOPTS: 7288 case IPPROTO_DSTOPTS: { 7289 /* get next header and header length */ 7290 struct ip6_ext opt6; 7291 7292 if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6), 7293 NULL, &reason, pd.af)) { 7294 DPFPRINTF(PF_DEBUG_MISC, 7295 ("pf: IPv6 short opt\n")); 7296 action = PF_DROP; 7297 log = 1; 7298 goto done; 7299 } 7300 if (pd.proto == IPPROTO_AH) 7301 off += (opt6.ip6e_len + 2) * 4; 7302 else 7303 off += (opt6.ip6e_len + 1) * 8; 7304 pd.proto = opt6.ip6e_nxt; 7305 /* goto the next header */ 7306 break; 7307 } 7308 default: 7309 terminal++; 7310 break; 7311 } 7312 } while (!terminal); 7313 7314 /* if there's no routing header, use unmodified mbuf for checksumming */ 7315 if (!n) 7316 n = m; 7317 7318 switch (pd.proto) { 7319 7320 case IPPROTO_TCP: { 7321 struct tcphdr th; 7322 7323 pd.hdr.tcp = &th; 7324 if (!pf_pull_hdr(m, off, &th, sizeof(th), 7325 &action, &reason, AF_INET6)) { 7326 log = action != PF_PASS; 7327 goto done; 7328 } 7329 if (dir == PF_IN && pf_check_proto_cksum(n, off, 7330 ntohs(h->ip6_plen) - (off - sizeof(struct ip6_hdr)), 7331 IPPROTO_TCP, AF_INET6)) { 7332 action = PF_DROP; 7333 REASON_SET(&reason, PFRES_PROTCKSUM); 7334 goto done; 7335 } 7336 pd.p_len = pd.tot_len - off - (th.th_off << 2); 7337 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd); 7338 if (action == PF_DROP) 7339 goto done; 7340 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd, 7341 &reason); 7342 if (action == PF_PASS) { 7343#if NPFSYNC 7344 pfsync_update_state(s); 7345#endif /* NPFSYNC */ 7346 r = s->rule.ptr; 7347 a = s->anchor.ptr; 7348 log = s->log; 7349 } else if (s == NULL) 7350#ifdef __FreeBSD__ 7351 action = pf_test_tcp(&r, &s, dir, kif, 7352 m, off, h, &pd, &a, &ruleset, NULL, inp); 7353#else 7354 action = pf_test_tcp(&r, &s, dir, kif, 7355 m, off, h, &pd, &a, &ruleset, &ip6intrq); 7356#endif 7357 break; 7358 } 7359 7360 case IPPROTO_UDP: { 7361 struct udphdr uh; 7362 7363 pd.hdr.udp = &uh; 7364 if (!pf_pull_hdr(m, off, &uh, sizeof(uh), 7365 &action, &reason, AF_INET6)) { 7366 log = action != PF_PASS; 7367 goto done; 7368 } 7369 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(n, 7370 off, ntohs(h->ip6_plen) - (off - sizeof(struct ip6_hdr)), 7371 IPPROTO_UDP, AF_INET6)) { 7372 action = PF_DROP; 7373 REASON_SET(&reason, PFRES_PROTCKSUM); 7374 goto done; 7375 } 7376 if (uh.uh_dport == 0 || 7377 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off || 7378 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) { 7379 action = PF_DROP; 7380 REASON_SET(&reason, PFRES_SHORT); 7381 goto done; 7382 } 7383 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd); 7384 if (action == PF_PASS) { 7385#if NPFSYNC 7386 pfsync_update_state(s); 7387#endif /* NPFSYNC */ 7388 r = s->rule.ptr; 7389 a = s->anchor.ptr; 7390 log = s->log; 7391 } else if (s == NULL) 7392#ifdef __FreeBSD__ 7393 action = pf_test_udp(&r, &s, dir, kif, 7394 m, off, h, &pd, &a, &ruleset, NULL, inp); 7395#else 7396 action = pf_test_udp(&r, &s, dir, kif, 7397 m, off, h, &pd, &a, &ruleset, &ip6intrq); 7398#endif 7399 break; 7400 } 7401 7402 case IPPROTO_ICMPV6: { 7403 struct icmp6_hdr ih; 7404 7405 pd.hdr.icmp6 = &ih; 7406 if (!pf_pull_hdr(m, off, &ih, sizeof(ih), 7407 &action, &reason, AF_INET6)) { 7408 log = action != PF_PASS; 7409 goto done; 7410 } 7411 if (dir == PF_IN && pf_check_proto_cksum(n, off, 7412 ntohs(h->ip6_plen) - (off - sizeof(struct ip6_hdr)), 7413 IPPROTO_ICMPV6, AF_INET6)) { 7414 action = PF_DROP; 7415 REASON_SET(&reason, PFRES_PROTCKSUM); 7416 goto done; 7417 } 7418 action = pf_test_state_icmp(&s, dir, kif, 7419 m, off, h, &pd, &reason); 7420 if (action == PF_PASS) { 7421#if NPFSYNC 7422 pfsync_update_state(s); 7423#endif /* NPFSYNC */ 7424 r = s->rule.ptr; 7425 a = s->anchor.ptr; 7426 log = s->log; 7427 } else if (s == NULL) 7428#ifdef __FreeBSD__ 7429 action = pf_test_icmp(&r, &s, dir, kif, 7430 m, off, h, &pd, &a, &ruleset, NULL); 7431#else 7432 action = pf_test_icmp(&r, &s, dir, kif, 7433 m, off, h, &pd, &a, &ruleset, &ip6intrq); 7434#endif 7435 break; 7436 } 7437 7438 default: 7439 action = pf_test_state_other(&s, dir, kif, &pd); 7440 if (action == PF_PASS) { 7441#if NPFSYNC 7442 pfsync_update_state(s); 7443#endif /* NPFSYNC */ 7444 r = s->rule.ptr; 7445 a = s->anchor.ptr; 7446 log = s->log; 7447 } else if (s == NULL) 7448#ifdef __FreeBSD__ 7449 action = pf_test_other(&r, &s, dir, kif, m, off, h, 7450 &pd, &a, &ruleset, NULL); 7451#else 7452 action = pf_test_other(&r, &s, dir, kif, m, off, h, 7453 &pd, &a, &ruleset, &ip6intrq); 7454#endif 7455 break; 7456 } 7457 7458done: 7459 /* handle dangerous IPv6 extension headers. */ 7460 if (action == PF_PASS && rh_cnt && 7461 !((s && s->allow_opts) || r->allow_opts)) { 7462 action = PF_DROP; 7463 REASON_SET(&reason, PFRES_IPOPTIONS); 7464 log = 1; 7465 DPFPRINTF(PF_DEBUG_MISC, 7466 ("pf: dropping packet with dangerous v6 headers\n")); 7467 } 7468 7469 if ((s && s->tag) || r->rtableid) 7470 pf_tag_packet(m, pd.pf_mtag, s ? s->tag : 0, r->rtableid); 7471 7472#ifdef ALTQ 7473 if (action == PF_PASS && r->qid) { 7474 if (pd.tos & IPTOS_LOWDELAY) 7475 pd.pf_mtag->qid = r->pqid; 7476 else 7477 pd.pf_mtag->qid = r->qid; 7478 /* add hints for ecn */ 7479 pd.pf_mtag->af = AF_INET6; 7480 pd.pf_mtag->hdr = h; 7481 } 7482#endif /* ALTQ */ 7483 7484 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP || 7485 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL && 7486 (s->nat_rule.ptr->action == PF_RDR || 7487 s->nat_rule.ptr->action == PF_BINAT) && 7488 IN6_IS_ADDR_LOOPBACK(&pd.dst->v6)) 7489 pd.pf_mtag->flags |= PF_TAG_TRANSLATE_LOCALHOST; 7490 7491 if (log) { 7492 struct pf_rule *lr; 7493 7494 if (s != NULL && s->nat_rule.ptr != NULL && 7495 s->nat_rule.ptr->log & PF_LOG_ALL) 7496 lr = s->nat_rule.ptr; 7497 else 7498 lr = r; 7499 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset, 7500 &pd); 7501 } 7502 7503 kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len; 7504 kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++; 7505 7506 if (action == PF_PASS || r->action == PF_DROP) { 7507 dirndx = (dir == PF_OUT); 7508 r->packets[dirndx]++; 7509 r->bytes[dirndx] += pd.tot_len; 7510 if (a != NULL) { 7511 a->packets[dirndx]++; 7512 a->bytes[dirndx] += pd.tot_len; 7513 } 7514 if (s != NULL) { 7515 if (s->nat_rule.ptr != NULL) { 7516 s->nat_rule.ptr->packets[dirndx]++; 7517 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len; 7518 } 7519 if (s->src_node != NULL) { 7520 s->src_node->packets[dirndx]++; 7521 s->src_node->bytes[dirndx] += pd.tot_len; 7522 } 7523 if (s->nat_src_node != NULL) { 7524 s->nat_src_node->packets[dirndx]++; 7525 s->nat_src_node->bytes[dirndx] += pd.tot_len; 7526 } 7527 dirndx = (dir == s->direction) ? 0 : 1; 7528 s->packets[dirndx]++; 7529 s->bytes[dirndx] += pd.tot_len; 7530 } 7531 tr = r; 7532 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule; 7533 if (nr != NULL) { 7534 struct pf_addr *x; 7535 /* 7536 * XXX: we need to make sure that the addresses 7537 * passed to pfr_update_stats() are the same than 7538 * the addresses used during matching (pfr_match) 7539 */ 7540 if (r == &pf_default_rule) { 7541 tr = nr; 7542 x = (s == NULL || s->direction == dir) ? 7543 &pd.baddr : &pd.naddr; 7544 } else { 7545 x = (s == NULL || s->direction == dir) ? 7546 &pd.naddr : &pd.baddr; 7547 } 7548 if (x == &pd.baddr || s == NULL) { 7549 if (dir == PF_OUT) 7550 pd.src = x; 7551 else 7552 pd.dst = x; 7553 } 7554 } 7555 if (tr->src.addr.type == PF_ADDR_TABLE) 7556 pfr_update_stats(tr->src.addr.p.tbl, (s == NULL || 7557 s->direction == dir) ? pd.src : pd.dst, pd.af, 7558 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 7559 tr->src.neg); 7560 if (tr->dst.addr.type == PF_ADDR_TABLE) 7561 pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL || 7562 s->direction == dir) ? pd.dst : pd.src, pd.af, 7563 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 7564 tr->dst.neg); 7565 } 7566 7567 7568 if (action == PF_SYNPROXY_DROP) { 7569 m_freem(*m0); 7570 *m0 = NULL; 7571 action = PF_PASS; 7572 } else if (r->rt) 7573 /* pf_route6 can free the mbuf causing *m0 to become NULL */ 7574 pf_route6(m0, r, dir, ifp, s, &pd); 7575 7576#ifdef __FreeBSD__ 7577 PF_UNLOCK(); 7578#endif 7579 return (action); 7580} 7581#endif /* INET6 */ 7582 7583int 7584pf_check_congestion(struct ifqueue *ifq) 7585{ 7586#ifdef __FreeBSD__ 7587 /* XXX_IMPORT: later */ 7588 return (0); 7589#else 7590 if (ifq->ifq_congestion) 7591 return (1); 7592 else 7593 return (0); 7594#endif 7595} 7596