pf.c revision 221132
1/* $OpenBSD: pf.c,v 1.527 2007/02/22 15:23:23 pyr Exp $ */ 2/* add: $OpenBSD: pf.c,v 1.559 2007/09/18 18:45:59 markus Exp $ */ 3 4/* 5 * Copyright (c) 2001 Daniel Hartmeier 6 * Copyright (c) 2002,2003 Henning Brauer 7 * All rights reserved. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 13 * - Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * - Redistributions in binary form must reproduce the above 16 * copyright notice, this list of conditions and the following 17 * disclaimer in the documentation and/or other materials provided 18 * with the distribution. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 25 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, 26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 28 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN 30 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 31 * POSSIBILITY OF SUCH DAMAGE. 32 * 33 * Effort sponsored in part by the Defense Advanced Research Projects 34 * Agency (DARPA) and Air Force Research Laboratory, Air Force 35 * Materiel Command, USAF, under agreement number F30602-01-2-0537. 36 * 37 */ 38 39#ifdef __FreeBSD__ 40#include "opt_inet.h" 41#include "opt_inet6.h" 42 43#include <sys/cdefs.h> 44__FBSDID("$FreeBSD: head/sys/contrib/pf/net/pf.c 221132 2011-04-27 19:34:01Z bz $"); 45#endif 46 47#ifdef __FreeBSD__ 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_tcp_track_full(struct pf_state_peer *, 257 struct pf_state_peer *, struct pf_state **, 258 struct pfi_kif *, struct mbuf *, int, 259 struct pf_pdesc *, u_short *, int *); 260int pf_tcp_track_sloppy(struct pf_state_peer *, 261 struct pf_state_peer *, struct pf_state **, 262 struct pf_pdesc *, u_short *); 263int pf_test_state_tcp(struct pf_state **, int, 264 struct pfi_kif *, struct mbuf *, int, 265 void *, struct pf_pdesc *, u_short *); 266int pf_test_state_udp(struct pf_state **, int, 267 struct pfi_kif *, struct mbuf *, int, 268 void *, struct pf_pdesc *); 269int pf_test_state_icmp(struct pf_state **, int, 270 struct pfi_kif *, struct mbuf *, int, 271 void *, struct pf_pdesc *, u_short *); 272int pf_test_state_other(struct pf_state **, int, 273 struct pfi_kif *, struct pf_pdesc *); 274int pf_match_tag(struct mbuf *, struct pf_rule *, 275 struct pf_mtag *, int *); 276int pf_step_out_of_anchor(int *, struct pf_ruleset **, 277 int, struct pf_rule **, struct pf_rule **, 278 int *); 279void pf_hash(struct pf_addr *, struct pf_addr *, 280 struct pf_poolhashkey *, sa_family_t); 281int pf_map_addr(u_int8_t, struct pf_rule *, 282 struct pf_addr *, struct pf_addr *, 283 struct pf_addr *, struct pf_src_node **); 284int pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *, 285 struct pf_addr *, struct pf_addr *, u_int16_t, 286 struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t, 287 struct pf_src_node **); 288void pf_route(struct mbuf **, struct pf_rule *, int, 289 struct ifnet *, struct pf_state *, 290 struct pf_pdesc *); 291void pf_route6(struct mbuf **, struct pf_rule *, int, 292 struct ifnet *, struct pf_state *, 293 struct pf_pdesc *); 294#ifdef __FreeBSD__ 295/* XXX: import */ 296#else 297int pf_socket_lookup(int, struct pf_pdesc *); 298#endif 299u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t, 300 sa_family_t); 301u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t, 302 sa_family_t); 303u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t, 304 u_int16_t); 305void pf_set_rt_ifp(struct pf_state *, 306 struct pf_addr *); 307int pf_check_proto_cksum(struct mbuf *, int, int, 308 u_int8_t, sa_family_t); 309int pf_addr_wrap_neq(struct pf_addr_wrap *, 310 struct pf_addr_wrap *); 311struct pf_state *pf_find_state_recurse(struct pfi_kif *, 312 struct pf_state_cmp *, u_int8_t); 313int pf_src_connlimit(struct pf_state **); 314int pf_check_congestion(struct ifqueue *); 315 316#ifdef __FreeBSD__ 317int in4_cksum(struct mbuf *m, u_int8_t nxt, int off, int len); 318 319extern int pf_end_threads; 320 321struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX]; 322#else 323extern struct pool pfr_ktable_pl; 324extern struct pool pfr_kentry_pl; 325 326struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = { 327 { &pf_state_pl, PFSTATE_HIWAT }, 328 { &pf_src_tree_pl, PFSNODE_HIWAT }, 329 { &pf_frent_pl, PFFRAG_FRENT_HIWAT }, 330 { &pfr_ktable_pl, PFR_KTABLE_HIWAT }, 331 { &pfr_kentry_pl, PFR_KENTRY_HIWAT } 332}; 333#endif 334 335#define STATE_LOOKUP() \ 336 do { \ 337 if (direction == PF_IN) \ 338 *state = pf_find_state_recurse( \ 339 kif, &key, PF_EXT_GWY); \ 340 else \ 341 *state = pf_find_state_recurse( \ 342 kif, &key, PF_LAN_EXT); \ 343 if (*state == NULL || (*state)->timeout == PFTM_PURGE) \ 344 return (PF_DROP); \ 345 if (direction == PF_OUT && \ 346 (((*state)->rule.ptr->rt == PF_ROUTETO && \ 347 (*state)->rule.ptr->direction == PF_OUT) || \ 348 ((*state)->rule.ptr->rt == PF_REPLYTO && \ 349 (*state)->rule.ptr->direction == PF_IN)) && \ 350 (*state)->rt_kif != NULL && \ 351 (*state)->rt_kif != kif) \ 352 return (PF_PASS); \ 353 } while (0) 354 355#define STATE_TRANSLATE(s) \ 356 (s)->lan.addr.addr32[0] != (s)->gwy.addr.addr32[0] || \ 357 ((s)->af == AF_INET6 && \ 358 ((s)->lan.addr.addr32[1] != (s)->gwy.addr.addr32[1] || \ 359 (s)->lan.addr.addr32[2] != (s)->gwy.addr.addr32[2] || \ 360 (s)->lan.addr.addr32[3] != (s)->gwy.addr.addr32[3])) || \ 361 (s)->lan.port != (s)->gwy.port 362 363#define BOUND_IFACE(r, k) \ 364 ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all 365 366#define STATE_INC_COUNTERS(s) \ 367 do { \ 368 s->rule.ptr->states++; \ 369 if (s->anchor.ptr != NULL) \ 370 s->anchor.ptr->states++; \ 371 if (s->nat_rule.ptr != NULL) \ 372 s->nat_rule.ptr->states++; \ 373 } while (0) 374 375#define STATE_DEC_COUNTERS(s) \ 376 do { \ 377 if (s->nat_rule.ptr != NULL) \ 378 s->nat_rule.ptr->states--; \ 379 if (s->anchor.ptr != NULL) \ 380 s->anchor.ptr->states--; \ 381 s->rule.ptr->states--; \ 382 } while (0) 383 384struct pf_src_tree tree_src_tracking; 385 386struct pf_state_tree_id tree_id; 387struct pf_state_queue state_list; 388 389#ifdef __FreeBSD__ 390static int pf_src_compare(struct pf_src_node *, struct pf_src_node *); 391static int pf_state_compare_lan_ext(struct pf_state *, struct pf_state *); 392static int pf_state_compare_ext_gwy(struct pf_state *, struct pf_state *); 393static int pf_state_compare_id(struct pf_state *, struct pf_state *); 394#endif 395 396RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare); 397RB_GENERATE(pf_state_tree_lan_ext, pf_state, 398 u.s.entry_lan_ext, pf_state_compare_lan_ext); 399RB_GENERATE(pf_state_tree_ext_gwy, pf_state, 400 u.s.entry_ext_gwy, pf_state_compare_ext_gwy); 401RB_GENERATE(pf_state_tree_id, pf_state, 402 u.s.entry_id, pf_state_compare_id); 403 404#ifdef __FreeBSD__ 405static int 406#else 407static __inline int 408#endif 409pf_src_compare(struct pf_src_node *a, struct pf_src_node *b) 410{ 411 int diff; 412 413 if (a->rule.ptr > b->rule.ptr) 414 return (1); 415 if (a->rule.ptr < b->rule.ptr) 416 return (-1); 417 if ((diff = a->af - b->af) != 0) 418 return (diff); 419 switch (a->af) { 420#ifdef INET 421 case AF_INET: 422 if (a->addr.addr32[0] > b->addr.addr32[0]) 423 return (1); 424 if (a->addr.addr32[0] < b->addr.addr32[0]) 425 return (-1); 426 break; 427#endif /* INET */ 428#ifdef INET6 429 case AF_INET6: 430 if (a->addr.addr32[3] > b->addr.addr32[3]) 431 return (1); 432 if (a->addr.addr32[3] < b->addr.addr32[3]) 433 return (-1); 434 if (a->addr.addr32[2] > b->addr.addr32[2]) 435 return (1); 436 if (a->addr.addr32[2] < b->addr.addr32[2]) 437 return (-1); 438 if (a->addr.addr32[1] > b->addr.addr32[1]) 439 return (1); 440 if (a->addr.addr32[1] < b->addr.addr32[1]) 441 return (-1); 442 if (a->addr.addr32[0] > b->addr.addr32[0]) 443 return (1); 444 if (a->addr.addr32[0] < b->addr.addr32[0]) 445 return (-1); 446 break; 447#endif /* INET6 */ 448 } 449 return (0); 450} 451 452#ifdef __FreeBSD__ 453static int 454#else 455static __inline int 456#endif 457pf_state_compare_lan_ext(struct pf_state *a, struct pf_state *b) 458{ 459 int diff; 460 461 if ((diff = a->proto - b->proto) != 0) 462 return (diff); 463 if ((diff = a->af - b->af) != 0) 464 return (diff); 465 switch (a->af) { 466#ifdef INET 467 case AF_INET: 468 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0]) 469 return (1); 470 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0]) 471 return (-1); 472 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 473 return (1); 474 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 475 return (-1); 476 break; 477#endif /* INET */ 478#ifdef INET6 479 case AF_INET6: 480 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3]) 481 return (1); 482 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3]) 483 return (-1); 484 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3]) 485 return (1); 486 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3]) 487 return (-1); 488 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2]) 489 return (1); 490 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2]) 491 return (-1); 492 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2]) 493 return (1); 494 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2]) 495 return (-1); 496 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1]) 497 return (1); 498 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1]) 499 return (-1); 500 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1]) 501 return (1); 502 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1]) 503 return (-1); 504 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0]) 505 return (1); 506 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0]) 507 return (-1); 508 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 509 return (1); 510 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 511 return (-1); 512 break; 513#endif /* INET6 */ 514 } 515 516 if ((diff = a->lan.port - b->lan.port) != 0) 517 return (diff); 518 if ((diff = a->ext.port - b->ext.port) != 0) 519 return (diff); 520 521 return (0); 522} 523 524#ifdef __FreeBSD__ 525static int 526#else 527static __inline int 528#endif 529pf_state_compare_ext_gwy(struct pf_state *a, struct pf_state *b) 530{ 531 int diff; 532 533 if ((diff = a->proto - b->proto) != 0) 534 return (diff); 535 if ((diff = a->af - b->af) != 0) 536 return (diff); 537 switch (a->af) { 538#ifdef INET 539 case AF_INET: 540 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 541 return (1); 542 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 543 return (-1); 544 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0]) 545 return (1); 546 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0]) 547 return (-1); 548 break; 549#endif /* INET */ 550#ifdef INET6 551 case AF_INET6: 552 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3]) 553 return (1); 554 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3]) 555 return (-1); 556 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3]) 557 return (1); 558 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3]) 559 return (-1); 560 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2]) 561 return (1); 562 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2]) 563 return (-1); 564 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2]) 565 return (1); 566 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2]) 567 return (-1); 568 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1]) 569 return (1); 570 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1]) 571 return (-1); 572 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1]) 573 return (1); 574 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1]) 575 return (-1); 576 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 577 return (1); 578 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 579 return (-1); 580 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0]) 581 return (1); 582 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0]) 583 return (-1); 584 break; 585#endif /* INET6 */ 586 } 587 588 if ((diff = a->ext.port - b->ext.port) != 0) 589 return (diff); 590 if ((diff = a->gwy.port - b->gwy.port) != 0) 591 return (diff); 592 593 return (0); 594} 595 596#ifdef __FreeBSD__ 597static int 598#else 599static __inline int 600#endif 601pf_state_compare_id(struct pf_state *a, struct pf_state *b) 602{ 603 if (a->id > b->id) 604 return (1); 605 if (a->id < b->id) 606 return (-1); 607 if (a->creatorid > b->creatorid) 608 return (1); 609 if (a->creatorid < b->creatorid) 610 return (-1); 611 612 return (0); 613} 614 615#ifdef INET6 616void 617pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af) 618{ 619 switch (af) { 620#ifdef INET 621 case AF_INET: 622 dst->addr32[0] = src->addr32[0]; 623 break; 624#endif /* INET */ 625 case AF_INET6: 626 dst->addr32[0] = src->addr32[0]; 627 dst->addr32[1] = src->addr32[1]; 628 dst->addr32[2] = src->addr32[2]; 629 dst->addr32[3] = src->addr32[3]; 630 break; 631 } 632} 633#endif /* INET6 */ 634 635struct pf_state * 636pf_find_state_byid(struct pf_state_cmp *key) 637{ 638 pf_status.fcounters[FCNT_STATE_SEARCH]++; 639 return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key)); 640} 641 642struct pf_state * 643pf_find_state_recurse(struct pfi_kif *kif, struct pf_state_cmp *key, u_int8_t tree) 644{ 645 struct pf_state *s; 646 647 pf_status.fcounters[FCNT_STATE_SEARCH]++; 648 649 switch (tree) { 650 case PF_LAN_EXT: 651 if ((s = RB_FIND(pf_state_tree_lan_ext, &kif->pfik_lan_ext, 652 (struct pf_state *)key)) != NULL) 653 return (s); 654 if ((s = RB_FIND(pf_state_tree_lan_ext, &pfi_all->pfik_lan_ext, 655 (struct pf_state *)key)) != NULL) 656 return (s); 657 return (NULL); 658 case PF_EXT_GWY: 659 if ((s = RB_FIND(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, 660 (struct pf_state *)key)) != NULL) 661 return (s); 662 if ((s = RB_FIND(pf_state_tree_ext_gwy, &pfi_all->pfik_ext_gwy, 663 (struct pf_state *)key)) != NULL) 664 return (s); 665 return (NULL); 666 default: 667 panic("pf_find_state_recurse"); 668 } 669} 670 671struct pf_state * 672pf_find_state_all(struct pf_state_cmp *key, u_int8_t tree, int *more) 673{ 674 struct pf_state *s, *ss = NULL; 675 struct pfi_kif *kif; 676 677 pf_status.fcounters[FCNT_STATE_SEARCH]++; 678 679 switch (tree) { 680 case PF_LAN_EXT: 681 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) { 682 s = RB_FIND(pf_state_tree_lan_ext, 683 &kif->pfik_lan_ext, (struct pf_state *)key); 684 if (s == NULL) 685 continue; 686 if (more == NULL) 687 return (s); 688 ss = s; 689 (*more)++; 690 } 691 return (ss); 692 case PF_EXT_GWY: 693 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) { 694 s = RB_FIND(pf_state_tree_ext_gwy, 695 &kif->pfik_ext_gwy, (struct pf_state *)key); 696 if (s == NULL) 697 continue; 698 if (more == NULL) 699 return (s); 700 ss = s; 701 (*more)++; 702 } 703 return (ss); 704 default: 705 panic("pf_find_state_all"); 706 } 707} 708 709void 710pf_init_threshold(struct pf_threshold *threshold, 711 u_int32_t limit, u_int32_t seconds) 712{ 713 threshold->limit = limit * PF_THRESHOLD_MULT; 714 threshold->seconds = seconds; 715 threshold->count = 0; 716 threshold->last = time_second; 717} 718 719void 720pf_add_threshold(struct pf_threshold *threshold) 721{ 722 u_int32_t t = time_second, diff = t - threshold->last; 723 724 if (diff >= threshold->seconds) 725 threshold->count = 0; 726 else 727 threshold->count -= threshold->count * diff / 728 threshold->seconds; 729 threshold->count += PF_THRESHOLD_MULT; 730 threshold->last = t; 731} 732 733int 734pf_check_threshold(struct pf_threshold *threshold) 735{ 736 return (threshold->count > threshold->limit); 737} 738 739int 740pf_src_connlimit(struct pf_state **state) 741{ 742 struct pf_state *s; 743 int bad = 0; 744 745 (*state)->src_node->conn++; 746 (*state)->src.tcp_est = 1; 747 pf_add_threshold(&(*state)->src_node->conn_rate); 748 749 if ((*state)->rule.ptr->max_src_conn && 750 (*state)->rule.ptr->max_src_conn < 751 (*state)->src_node->conn) { 752 pf_status.lcounters[LCNT_SRCCONN]++; 753 bad++; 754 } 755 756 if ((*state)->rule.ptr->max_src_conn_rate.limit && 757 pf_check_threshold(&(*state)->src_node->conn_rate)) { 758 pf_status.lcounters[LCNT_SRCCONNRATE]++; 759 bad++; 760 } 761 762 if (!bad) 763 return (0); 764 765 if ((*state)->rule.ptr->overload_tbl) { 766 struct pfr_addr p; 767 u_int32_t killed = 0; 768 769 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++; 770 if (pf_status.debug >= PF_DEBUG_MISC) { 771 printf("pf_src_connlimit: blocking address "); 772 pf_print_host(&(*state)->src_node->addr, 0, 773 (*state)->af); 774 } 775 776 bzero(&p, sizeof(p)); 777 p.pfra_af = (*state)->af; 778 switch ((*state)->af) { 779#ifdef INET 780 case AF_INET: 781 p.pfra_net = 32; 782 p.pfra_ip4addr = (*state)->src_node->addr.v4; 783 break; 784#endif /* INET */ 785#ifdef INET6 786 case AF_INET6: 787 p.pfra_net = 128; 788 p.pfra_ip6addr = (*state)->src_node->addr.v6; 789 break; 790#endif /* INET6 */ 791 } 792 793 pfr_insert_kentry((*state)->rule.ptr->overload_tbl, 794 &p, time_second); 795 796 /* kill existing states if that's required. */ 797 if ((*state)->rule.ptr->flush) { 798 pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++; 799 800 RB_FOREACH(s, pf_state_tree_id, &tree_id) { 801 /* 802 * Kill states from this source. (Only those 803 * from the same rule if PF_FLUSH_GLOBAL is not 804 * set) 805 */ 806 if (s->af == (*state)->af && 807 (((*state)->direction == PF_OUT && 808 PF_AEQ(&(*state)->src_node->addr, 809 &s->lan.addr, s->af)) || 810 ((*state)->direction == PF_IN && 811 PF_AEQ(&(*state)->src_node->addr, 812 &s->ext.addr, s->af))) && 813 ((*state)->rule.ptr->flush & 814 PF_FLUSH_GLOBAL || 815 (*state)->rule.ptr == s->rule.ptr)) { 816 s->timeout = PFTM_PURGE; 817 s->src.state = s->dst.state = 818 TCPS_CLOSED; 819 killed++; 820 } 821 } 822 if (pf_status.debug >= PF_DEBUG_MISC) 823 printf(", %u states killed", killed); 824 } 825 if (pf_status.debug >= PF_DEBUG_MISC) 826 printf("\n"); 827 } 828 829 /* kill this state */ 830 (*state)->timeout = PFTM_PURGE; 831 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED; 832 return (1); 833} 834 835int 836pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule, 837 struct pf_addr *src, sa_family_t af) 838{ 839 struct pf_src_node k; 840 841 if (*sn == NULL) { 842 k.af = af; 843 PF_ACPY(&k.addr, src, af); 844 if (rule->rule_flag & PFRULE_RULESRCTRACK || 845 rule->rpool.opts & PF_POOL_STICKYADDR) 846 k.rule.ptr = rule; 847 else 848 k.rule.ptr = NULL; 849 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++; 850 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k); 851 } 852 if (*sn == NULL) { 853 if (!rule->max_src_nodes || 854 rule->src_nodes < rule->max_src_nodes) 855 (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT); 856 else 857 pf_status.lcounters[LCNT_SRCNODES]++; 858 if ((*sn) == NULL) 859 return (-1); 860 bzero(*sn, sizeof(struct pf_src_node)); 861 862 pf_init_threshold(&(*sn)->conn_rate, 863 rule->max_src_conn_rate.limit, 864 rule->max_src_conn_rate.seconds); 865 866 (*sn)->af = af; 867 if (rule->rule_flag & PFRULE_RULESRCTRACK || 868 rule->rpool.opts & PF_POOL_STICKYADDR) 869 (*sn)->rule.ptr = rule; 870 else 871 (*sn)->rule.ptr = NULL; 872 PF_ACPY(&(*sn)->addr, src, af); 873 if (RB_INSERT(pf_src_tree, 874 &tree_src_tracking, *sn) != NULL) { 875 if (pf_status.debug >= PF_DEBUG_MISC) { 876 printf("pf: src_tree insert failed: "); 877 pf_print_host(&(*sn)->addr, 0, af); 878 printf("\n"); 879 } 880 pool_put(&pf_src_tree_pl, *sn); 881 return (-1); 882 } 883 (*sn)->creation = time_second; 884 (*sn)->ruletype = rule->action; 885 if ((*sn)->rule.ptr != NULL) 886 (*sn)->rule.ptr->src_nodes++; 887 pf_status.scounters[SCNT_SRC_NODE_INSERT]++; 888 pf_status.src_nodes++; 889 } else { 890 if (rule->max_src_states && 891 (*sn)->states >= rule->max_src_states) { 892 pf_status.lcounters[LCNT_SRCSTATES]++; 893 return (-1); 894 } 895 } 896 return (0); 897} 898 899int 900pf_insert_state(struct pfi_kif *kif, struct pf_state *state) 901{ 902 /* Thou MUST NOT insert multiple duplicate keys */ 903 state->u.s.kif = kif; 904 if (RB_INSERT(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state)) { 905 if (pf_status.debug >= PF_DEBUG_MISC) { 906 printf("pf: state insert failed: tree_lan_ext"); 907 printf(" lan: "); 908 pf_print_host(&state->lan.addr, state->lan.port, 909 state->af); 910 printf(" gwy: "); 911 pf_print_host(&state->gwy.addr, state->gwy.port, 912 state->af); 913 printf(" ext: "); 914 pf_print_host(&state->ext.addr, state->ext.port, 915 state->af); 916 if (state->sync_flags & PFSTATE_FROMSYNC) 917 printf(" (from sync)"); 918 printf("\n"); 919 } 920 return (-1); 921 } 922 923 if (RB_INSERT(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state)) { 924 if (pf_status.debug >= PF_DEBUG_MISC) { 925 printf("pf: state insert failed: tree_ext_gwy"); 926 printf(" lan: "); 927 pf_print_host(&state->lan.addr, state->lan.port, 928 state->af); 929 printf(" gwy: "); 930 pf_print_host(&state->gwy.addr, state->gwy.port, 931 state->af); 932 printf(" ext: "); 933 pf_print_host(&state->ext.addr, state->ext.port, 934 state->af); 935 if (state->sync_flags & PFSTATE_FROMSYNC) 936 printf(" (from sync)"); 937 printf("\n"); 938 } 939 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state); 940 return (-1); 941 } 942 943 if (state->id == 0 && state->creatorid == 0) { 944 state->id = htobe64(pf_status.stateid++); 945 state->creatorid = pf_status.hostid; 946 } 947 if (RB_INSERT(pf_state_tree_id, &tree_id, state) != NULL) { 948 if (pf_status.debug >= PF_DEBUG_MISC) { 949#ifdef __FreeBSD__ 950 printf("pf: state insert failed: " 951 "id: %016llx creatorid: %08x", 952 (long long)be64toh(state->id), 953 ntohl(state->creatorid)); 954#else 955 printf("pf: state insert failed: " 956 "id: %016llx creatorid: %08x", 957 betoh64(state->id), ntohl(state->creatorid)); 958#endif 959 if (state->sync_flags & PFSTATE_FROMSYNC) 960 printf(" (from sync)"); 961 printf("\n"); 962 } 963 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state); 964 RB_REMOVE(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state); 965 return (-1); 966 } 967 TAILQ_INSERT_TAIL(&state_list, state, u.s.entry_list); 968 pf_status.fcounters[FCNT_STATE_INSERT]++; 969 pf_status.states++; 970 pfi_kif_ref(kif, PFI_KIF_REF_STATE); 971#if NPFSYNC 972 pfsync_insert_state(state); 973#endif 974 return (0); 975} 976 977void 978pf_purge_thread(void *v) 979{ 980 int nloops = 0, s; 981#ifdef __FreeBSD__ 982 int locked; 983#endif 984 985 for (;;) { 986 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz); 987 988#ifdef __FreeBSD__ 989 sx_slock(&pf_consistency_lock); 990 PF_LOCK(); 991 locked = 0; 992 993 if (pf_end_threads) { 994 PF_UNLOCK(); 995 sx_sunlock(&pf_consistency_lock); 996 sx_xlock(&pf_consistency_lock); 997 PF_LOCK(); 998 pf_purge_expired_states(pf_status.states, 1); 999 pf_purge_expired_fragments(); 1000 pf_purge_expired_src_nodes(1); 1001 pf_end_threads++; 1002 1003 sx_xunlock(&pf_consistency_lock); 1004 PF_UNLOCK(); 1005 wakeup(pf_purge_thread); 1006 kproc_exit(0); 1007 } 1008#endif 1009 s = splsoftnet(); 1010 1011 /* process a fraction of the state table every second */ 1012#ifdef __FreeBSD__ 1013 if(!pf_purge_expired_states(1 + (pf_status.states 1014 / pf_default_rule.timeout[PFTM_INTERVAL]), 0)) { 1015 PF_UNLOCK(); 1016 sx_sunlock(&pf_consistency_lock); 1017 sx_xlock(&pf_consistency_lock); 1018 PF_LOCK(); 1019 locked = 1; 1020 1021 pf_purge_expired_states(1 + (pf_status.states 1022 / pf_default_rule.timeout[PFTM_INTERVAL]), 1); 1023 } 1024#else 1025 pf_purge_expired_states(1 + (pf_status.states 1026 / pf_default_rule.timeout[PFTM_INTERVAL])); 1027#endif 1028 1029 /* purge other expired types every PFTM_INTERVAL seconds */ 1030 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) { 1031 pf_purge_expired_fragments(); 1032 if (!pf_purge_expired_src_nodes(locked)) { 1033 PF_UNLOCK(); 1034 sx_sunlock(&pf_consistency_lock); 1035 sx_xlock(&pf_consistency_lock); 1036 PF_LOCK(); 1037 locked = 1; 1038 pf_purge_expired_src_nodes(1); 1039 } 1040 nloops = 0; 1041 } 1042 1043 splx(s); 1044#ifdef __FreeBSD__ 1045 PF_UNLOCK(); 1046 if (locked) 1047 sx_xunlock(&pf_consistency_lock); 1048 else 1049 sx_sunlock(&pf_consistency_lock); 1050#endif 1051 } 1052} 1053 1054u_int32_t 1055pf_state_expires(const struct pf_state *state) 1056{ 1057 u_int32_t timeout; 1058 u_int32_t start; 1059 u_int32_t end; 1060 u_int32_t states; 1061 1062 /* handle all PFTM_* > PFTM_MAX here */ 1063 if (state->timeout == PFTM_PURGE) 1064 return (time_second); 1065 if (state->timeout == PFTM_UNTIL_PACKET) 1066 return (0); 1067#ifdef __FreeBSD__ 1068 KASSERT(state->timeout != PFTM_UNLINKED, 1069 ("pf_state_expires: timeout == PFTM_UNLINKED")); 1070 KASSERT((state->timeout < PFTM_MAX), 1071 ("pf_state_expires: timeout > PFTM_MAX")); 1072#else 1073 KASSERT(state->timeout != PFTM_UNLINKED); 1074 KASSERT(state->timeout < PFTM_MAX); 1075#endif 1076 timeout = state->rule.ptr->timeout[state->timeout]; 1077 if (!timeout) 1078 timeout = pf_default_rule.timeout[state->timeout]; 1079 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START]; 1080 if (start) { 1081 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END]; 1082 states = state->rule.ptr->states; 1083 } else { 1084 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START]; 1085 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END]; 1086 states = pf_status.states; 1087 } 1088 if (end && states > start && start < end) { 1089 if (states < end) 1090 return (state->expire + timeout * (end - states) / 1091 (end - start)); 1092 else 1093 return (time_second); 1094 } 1095 return (state->expire + timeout); 1096} 1097 1098#ifdef __FreeBSD__ 1099int 1100pf_purge_expired_src_nodes(int waslocked) 1101#else 1102void 1103pf_purge_expired_src_nodes(int waslocked) 1104#endif 1105{ 1106 struct pf_src_node *cur, *next; 1107 int locked = waslocked; 1108 1109 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) { 1110 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur); 1111 1112 if (cur->states <= 0 && cur->expire <= time_second) { 1113 if (! locked) { 1114#ifdef __FreeBSD__ 1115 if (!sx_try_upgrade(&pf_consistency_lock)) 1116 return (0); 1117#else 1118 rw_enter_write(&pf_consistency_lock); 1119#endif 1120 next = RB_NEXT(pf_src_tree, 1121 &tree_src_tracking, cur); 1122 locked = 1; 1123 } 1124 if (cur->rule.ptr != NULL) { 1125 cur->rule.ptr->src_nodes--; 1126 if (cur->rule.ptr->states <= 0 && 1127 cur->rule.ptr->max_src_nodes <= 0) 1128 pf_rm_rule(NULL, cur->rule.ptr); 1129 } 1130 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur); 1131 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 1132 pf_status.src_nodes--; 1133 pool_put(&pf_src_tree_pl, cur); 1134 } 1135 } 1136 1137 if (locked && !waslocked) 1138#ifdef __FreeBSD__ 1139 sx_downgrade(&pf_consistency_lock); 1140#else 1141 rw_exit_write(&pf_consistency_lock); 1142#endif 1143 1144#ifdef __FreeBSD__ 1145 return (1); 1146#endif 1147} 1148 1149void 1150pf_src_tree_remove_state(struct pf_state *s) 1151{ 1152 u_int32_t timeout; 1153 1154 if (s->src_node != NULL) { 1155 if (s->proto == IPPROTO_TCP) { 1156 if (s->src.tcp_est) 1157 --s->src_node->conn; 1158 } 1159 if (--s->src_node->states <= 0) { 1160 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE]; 1161 if (!timeout) 1162 timeout = 1163 pf_default_rule.timeout[PFTM_SRC_NODE]; 1164 s->src_node->expire = time_second + timeout; 1165 } 1166 } 1167 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) { 1168 if (--s->nat_src_node->states <= 0) { 1169 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE]; 1170 if (!timeout) 1171 timeout = 1172 pf_default_rule.timeout[PFTM_SRC_NODE]; 1173 s->nat_src_node->expire = time_second + timeout; 1174 } 1175 } 1176 s->src_node = s->nat_src_node = NULL; 1177} 1178 1179/* callers should be at splsoftnet */ 1180void 1181pf_unlink_state(struct pf_state *cur) 1182{ 1183#ifdef __FreeBSD__ 1184 if (cur->local_flags & PFSTATE_EXPIRING) 1185 return; 1186 cur->local_flags |= PFSTATE_EXPIRING; 1187#endif 1188 if (cur->src.state == PF_TCPS_PROXY_DST) { 1189#ifdef __FreeBSD__ 1190 pf_send_tcp(NULL, cur->rule.ptr, cur->af, 1191#else 1192 pf_send_tcp(cur->rule.ptr, cur->af, 1193#endif 1194 &cur->ext.addr, &cur->lan.addr, 1195 cur->ext.port, cur->lan.port, 1196 cur->src.seqhi, cur->src.seqlo + 1, 1197 TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL); 1198 } 1199 RB_REMOVE(pf_state_tree_ext_gwy, 1200 &cur->u.s.kif->pfik_ext_gwy, cur); 1201 RB_REMOVE(pf_state_tree_lan_ext, 1202 &cur->u.s.kif->pfik_lan_ext, cur); 1203 RB_REMOVE(pf_state_tree_id, &tree_id, cur); 1204#if NPFSYNC 1205 if (cur->creatorid == pf_status.hostid) 1206 pfsync_delete_state(cur); 1207#endif 1208 cur->timeout = PFTM_UNLINKED; 1209 pf_src_tree_remove_state(cur); 1210} 1211 1212/* callers should be at splsoftnet and hold the 1213 * write_lock on pf_consistency_lock */ 1214void 1215pf_free_state(struct pf_state *cur) 1216{ 1217#if NPFSYNC 1218 if (pfsyncif != NULL && 1219 (pfsyncif->sc_bulk_send_next == cur || 1220 pfsyncif->sc_bulk_terminator == cur)) 1221 return; 1222#endif 1223#ifdef __FreeBSD__ 1224 KASSERT(cur->timeout == PFTM_UNLINKED, 1225 ("pf_free_state: cur->timeout != PFTM_UNLINKED")); 1226#else 1227 KASSERT(cur->timeout == PFTM_UNLINKED); 1228#endif 1229 if (--cur->rule.ptr->states <= 0 && 1230 cur->rule.ptr->src_nodes <= 0) 1231 pf_rm_rule(NULL, cur->rule.ptr); 1232 if (cur->nat_rule.ptr != NULL) 1233 if (--cur->nat_rule.ptr->states <= 0 && 1234 cur->nat_rule.ptr->src_nodes <= 0) 1235 pf_rm_rule(NULL, cur->nat_rule.ptr); 1236 if (cur->anchor.ptr != NULL) 1237 if (--cur->anchor.ptr->states <= 0) 1238 pf_rm_rule(NULL, cur->anchor.ptr); 1239 pf_normalize_tcp_cleanup(cur); 1240 pfi_kif_unref(cur->u.s.kif, PFI_KIF_REF_STATE); 1241 TAILQ_REMOVE(&state_list, cur, u.s.entry_list); 1242 if (cur->tag) 1243 pf_tag_unref(cur->tag); 1244 pool_put(&pf_state_pl, cur); 1245 pf_status.fcounters[FCNT_STATE_REMOVALS]++; 1246 pf_status.states--; 1247} 1248 1249#ifdef __FreeBSD__ 1250int 1251pf_purge_expired_states(u_int32_t maxcheck, int waslocked) 1252#else 1253void 1254pf_purge_expired_states(u_int32_t maxcheck) 1255#endif 1256{ 1257 static struct pf_state *cur = NULL; 1258 struct pf_state *next; 1259#ifdef __FreeBSD__ 1260 int locked = waslocked; 1261#else 1262 int locked = 0; 1263#endif 1264 1265 while (maxcheck--) { 1266 /* wrap to start of list when we hit the end */ 1267 if (cur == NULL) { 1268 cur = TAILQ_FIRST(&state_list); 1269 if (cur == NULL) 1270 break; /* list empty */ 1271 } 1272 1273 /* get next state, as cur may get deleted */ 1274 next = TAILQ_NEXT(cur, u.s.entry_list); 1275 1276 if (cur->timeout == PFTM_UNLINKED) { 1277 /* free unlinked state */ 1278 if (! locked) { 1279#ifdef __FreeBSD__ 1280 if (!sx_try_upgrade(&pf_consistency_lock)) 1281 return (0); 1282#else 1283 rw_enter_write(&pf_consistency_lock); 1284#endif 1285 locked = 1; 1286 } 1287 pf_free_state(cur); 1288 } else if (pf_state_expires(cur) <= time_second) { 1289 /* unlink and free expired state */ 1290 pf_unlink_state(cur); 1291 if (! locked) { 1292#ifdef __FreeBSD__ 1293 if (!sx_try_upgrade(&pf_consistency_lock)) 1294 return (0); 1295#else 1296 rw_enter_write(&pf_consistency_lock); 1297#endif 1298 locked = 1; 1299 } 1300 pf_free_state(cur); 1301 } 1302 cur = next; 1303 } 1304 1305#ifdef __FreeBSD__ 1306 if (!waslocked && locked) 1307 sx_downgrade(&pf_consistency_lock); 1308 1309 return (1); 1310#else 1311 if (locked) 1312 rw_exit_write(&pf_consistency_lock); 1313#endif 1314} 1315 1316int 1317pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw) 1318{ 1319 if (aw->type != PF_ADDR_TABLE) 1320 return (0); 1321 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL) 1322 return (1); 1323 return (0); 1324} 1325 1326void 1327pf_tbladdr_remove(struct pf_addr_wrap *aw) 1328{ 1329 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL) 1330 return; 1331 pfr_detach_table(aw->p.tbl); 1332 aw->p.tbl = NULL; 1333} 1334 1335void 1336pf_tbladdr_copyout(struct pf_addr_wrap *aw) 1337{ 1338 struct pfr_ktable *kt = aw->p.tbl; 1339 1340 if (aw->type != PF_ADDR_TABLE || kt == NULL) 1341 return; 1342 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) 1343 kt = kt->pfrkt_root; 1344 aw->p.tbl = NULL; 1345 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ? 1346 kt->pfrkt_cnt : -1; 1347} 1348 1349void 1350pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af) 1351{ 1352 switch (af) { 1353#ifdef INET 1354 case AF_INET: { 1355 u_int32_t a = ntohl(addr->addr32[0]); 1356 printf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255, 1357 (a>>8)&255, a&255); 1358 if (p) { 1359 p = ntohs(p); 1360 printf(":%u", p); 1361 } 1362 break; 1363 } 1364#endif /* INET */ 1365#ifdef INET6 1366 case AF_INET6: { 1367 u_int16_t b; 1368 u_int8_t i, curstart = 255, curend = 0, 1369 maxstart = 0, maxend = 0; 1370 for (i = 0; i < 8; i++) { 1371 if (!addr->addr16[i]) { 1372 if (curstart == 255) 1373 curstart = i; 1374 else 1375 curend = i; 1376 } else { 1377 if (curstart) { 1378 if ((curend - curstart) > 1379 (maxend - maxstart)) { 1380 maxstart = curstart; 1381 maxend = curend; 1382 curstart = 255; 1383 } 1384 } 1385 } 1386 } 1387 for (i = 0; i < 8; i++) { 1388 if (i >= maxstart && i <= maxend) { 1389 if (maxend != 7) { 1390 if (i == maxstart) 1391 printf(":"); 1392 } else { 1393 if (i == maxend) 1394 printf(":"); 1395 } 1396 } else { 1397 b = ntohs(addr->addr16[i]); 1398 printf("%x", b); 1399 if (i < 7) 1400 printf(":"); 1401 } 1402 } 1403 if (p) { 1404 p = ntohs(p); 1405 printf("[%u]", p); 1406 } 1407 break; 1408 } 1409#endif /* INET6 */ 1410 } 1411} 1412 1413void 1414pf_print_state(struct pf_state *s) 1415{ 1416 switch (s->proto) { 1417 case IPPROTO_TCP: 1418 printf("TCP "); 1419 break; 1420 case IPPROTO_UDP: 1421 printf("UDP "); 1422 break; 1423 case IPPROTO_ICMP: 1424 printf("ICMP "); 1425 break; 1426 case IPPROTO_ICMPV6: 1427 printf("ICMPV6 "); 1428 break; 1429 default: 1430 printf("%u ", s->proto); 1431 break; 1432 } 1433 pf_print_host(&s->lan.addr, s->lan.port, s->af); 1434 printf(" "); 1435 pf_print_host(&s->gwy.addr, s->gwy.port, s->af); 1436 printf(" "); 1437 pf_print_host(&s->ext.addr, s->ext.port, s->af); 1438 printf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo, 1439 s->src.seqhi, s->src.max_win, s->src.seqdiff); 1440 if (s->src.wscale && s->dst.wscale) 1441 printf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK); 1442 printf("]"); 1443 printf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo, 1444 s->dst.seqhi, s->dst.max_win, s->dst.seqdiff); 1445 if (s->src.wscale && s->dst.wscale) 1446 printf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK); 1447 printf("]"); 1448 printf(" %u:%u", s->src.state, s->dst.state); 1449} 1450 1451void 1452pf_print_flags(u_int8_t f) 1453{ 1454 if (f) 1455 printf(" "); 1456 if (f & TH_FIN) 1457 printf("F"); 1458 if (f & TH_SYN) 1459 printf("S"); 1460 if (f & TH_RST) 1461 printf("R"); 1462 if (f & TH_PUSH) 1463 printf("P"); 1464 if (f & TH_ACK) 1465 printf("A"); 1466 if (f & TH_URG) 1467 printf("U"); 1468 if (f & TH_ECE) 1469 printf("E"); 1470 if (f & TH_CWR) 1471 printf("W"); 1472} 1473 1474#define PF_SET_SKIP_STEPS(i) \ 1475 do { \ 1476 while (head[i] != cur) { \ 1477 head[i]->skip[i].ptr = cur; \ 1478 head[i] = TAILQ_NEXT(head[i], entries); \ 1479 } \ 1480 } while (0) 1481 1482void 1483pf_calc_skip_steps(struct pf_rulequeue *rules) 1484{ 1485 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT]; 1486 int i; 1487 1488 cur = TAILQ_FIRST(rules); 1489 prev = cur; 1490 for (i = 0; i < PF_SKIP_COUNT; ++i) 1491 head[i] = cur; 1492 while (cur != NULL) { 1493 1494 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot) 1495 PF_SET_SKIP_STEPS(PF_SKIP_IFP); 1496 if (cur->direction != prev->direction) 1497 PF_SET_SKIP_STEPS(PF_SKIP_DIR); 1498 if (cur->af != prev->af) 1499 PF_SET_SKIP_STEPS(PF_SKIP_AF); 1500 if (cur->proto != prev->proto) 1501 PF_SET_SKIP_STEPS(PF_SKIP_PROTO); 1502 if (cur->src.neg != prev->src.neg || 1503 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr)) 1504 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR); 1505 if (cur->src.port[0] != prev->src.port[0] || 1506 cur->src.port[1] != prev->src.port[1] || 1507 cur->src.port_op != prev->src.port_op) 1508 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT); 1509 if (cur->dst.neg != prev->dst.neg || 1510 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr)) 1511 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR); 1512 if (cur->dst.port[0] != prev->dst.port[0] || 1513 cur->dst.port[1] != prev->dst.port[1] || 1514 cur->dst.port_op != prev->dst.port_op) 1515 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT); 1516 1517 prev = cur; 1518 cur = TAILQ_NEXT(cur, entries); 1519 } 1520 for (i = 0; i < PF_SKIP_COUNT; ++i) 1521 PF_SET_SKIP_STEPS(i); 1522} 1523 1524int 1525pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2) 1526{ 1527 if (aw1->type != aw2->type) 1528 return (1); 1529 switch (aw1->type) { 1530 case PF_ADDR_ADDRMASK: 1531 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0)) 1532 return (1); 1533 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0)) 1534 return (1); 1535 return (0); 1536 case PF_ADDR_DYNIFTL: 1537 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt); 1538 case PF_ADDR_NOROUTE: 1539 case PF_ADDR_URPFFAILED: 1540 return (0); 1541 case PF_ADDR_TABLE: 1542 return (aw1->p.tbl != aw2->p.tbl); 1543 case PF_ADDR_RTLABEL: 1544 return (aw1->v.rtlabel != aw2->v.rtlabel); 1545 default: 1546 printf("invalid address type: %d\n", aw1->type); 1547 return (1); 1548 } 1549} 1550 1551u_int16_t 1552pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp) 1553{ 1554 u_int32_t l; 1555 1556 if (udp && !cksum) 1557 return (0x0000); 1558 l = cksum + old - new; 1559 l = (l >> 16) + (l & 65535); 1560 l = l & 65535; 1561 if (udp && !l) 1562 return (0xFFFF); 1563 return (l); 1564} 1565 1566void 1567pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc, 1568 struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af) 1569{ 1570 struct pf_addr ao; 1571 u_int16_t po = *p; 1572 1573 PF_ACPY(&ao, a, af); 1574 PF_ACPY(a, an, af); 1575 1576 *p = pn; 1577 1578 switch (af) { 1579#ifdef INET 1580 case AF_INET: 1581 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic, 1582 ao.addr16[0], an->addr16[0], 0), 1583 ao.addr16[1], an->addr16[1], 0); 1584 *p = pn; 1585 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc, 1586 ao.addr16[0], an->addr16[0], u), 1587 ao.addr16[1], an->addr16[1], u), 1588 po, pn, u); 1589 break; 1590#endif /* INET */ 1591#ifdef INET6 1592 case AF_INET6: 1593 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1594 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1595 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc, 1596 ao.addr16[0], an->addr16[0], u), 1597 ao.addr16[1], an->addr16[1], u), 1598 ao.addr16[2], an->addr16[2], u), 1599 ao.addr16[3], an->addr16[3], u), 1600 ao.addr16[4], an->addr16[4], u), 1601 ao.addr16[5], an->addr16[5], u), 1602 ao.addr16[6], an->addr16[6], u), 1603 ao.addr16[7], an->addr16[7], u), 1604 po, pn, u); 1605 break; 1606#endif /* INET6 */ 1607 } 1608} 1609 1610 1611/* Changes a u_int32_t. Uses a void * so there are no align restrictions */ 1612void 1613pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u) 1614{ 1615 u_int32_t ao; 1616 1617 memcpy(&ao, a, sizeof(ao)); 1618 memcpy(a, &an, sizeof(u_int32_t)); 1619 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u), 1620 ao % 65536, an % 65536, u); 1621} 1622 1623#ifdef INET6 1624void 1625pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u) 1626{ 1627 struct pf_addr ao; 1628 1629 PF_ACPY(&ao, a, AF_INET6); 1630 PF_ACPY(a, an, AF_INET6); 1631 1632 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1633 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1634 pf_cksum_fixup(pf_cksum_fixup(*c, 1635 ao.addr16[0], an->addr16[0], u), 1636 ao.addr16[1], an->addr16[1], u), 1637 ao.addr16[2], an->addr16[2], u), 1638 ao.addr16[3], an->addr16[3], u), 1639 ao.addr16[4], an->addr16[4], u), 1640 ao.addr16[5], an->addr16[5], u), 1641 ao.addr16[6], an->addr16[6], u), 1642 ao.addr16[7], an->addr16[7], u); 1643} 1644#endif /* INET6 */ 1645 1646void 1647pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa, 1648 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c, 1649 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af) 1650{ 1651 struct pf_addr oia, ooa; 1652 1653 PF_ACPY(&oia, ia, af); 1654 PF_ACPY(&ooa, oa, af); 1655 1656 /* Change inner protocol port, fix inner protocol checksum. */ 1657 if (ip != NULL) { 1658 u_int16_t oip = *ip; 1659 u_int32_t opc = 0; /* make the compiler happy */ 1660 1661 if (pc != NULL) 1662 opc = *pc; 1663 *ip = np; 1664 if (pc != NULL) 1665 *pc = pf_cksum_fixup(*pc, oip, *ip, u); 1666 *ic = pf_cksum_fixup(*ic, oip, *ip, 0); 1667 if (pc != NULL) 1668 *ic = pf_cksum_fixup(*ic, opc, *pc, 0); 1669 } 1670 /* Change inner ip address, fix inner ip and icmp checksums. */ 1671 PF_ACPY(ia, na, af); 1672 switch (af) { 1673#ifdef INET 1674 case AF_INET: { 1675 u_int32_t oh2c = *h2c; 1676 1677 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c, 1678 oia.addr16[0], ia->addr16[0], 0), 1679 oia.addr16[1], ia->addr16[1], 0); 1680 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic, 1681 oia.addr16[0], ia->addr16[0], 0), 1682 oia.addr16[1], ia->addr16[1], 0); 1683 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0); 1684 break; 1685 } 1686#endif /* INET */ 1687#ifdef INET6 1688 case AF_INET6: 1689 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1690 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1691 pf_cksum_fixup(pf_cksum_fixup(*ic, 1692 oia.addr16[0], ia->addr16[0], u), 1693 oia.addr16[1], ia->addr16[1], u), 1694 oia.addr16[2], ia->addr16[2], u), 1695 oia.addr16[3], ia->addr16[3], u), 1696 oia.addr16[4], ia->addr16[4], u), 1697 oia.addr16[5], ia->addr16[5], u), 1698 oia.addr16[6], ia->addr16[6], u), 1699 oia.addr16[7], ia->addr16[7], u); 1700 break; 1701#endif /* INET6 */ 1702 } 1703 /* Change outer ip address, fix outer ip or icmpv6 checksum. */ 1704 PF_ACPY(oa, na, af); 1705 switch (af) { 1706#ifdef INET 1707 case AF_INET: 1708 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc, 1709 ooa.addr16[0], oa->addr16[0], 0), 1710 ooa.addr16[1], oa->addr16[1], 0); 1711 break; 1712#endif /* INET */ 1713#ifdef INET6 1714 case AF_INET6: 1715 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1716 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1717 pf_cksum_fixup(pf_cksum_fixup(*ic, 1718 ooa.addr16[0], oa->addr16[0], u), 1719 ooa.addr16[1], oa->addr16[1], u), 1720 ooa.addr16[2], oa->addr16[2], u), 1721 ooa.addr16[3], oa->addr16[3], u), 1722 ooa.addr16[4], oa->addr16[4], u), 1723 ooa.addr16[5], oa->addr16[5], u), 1724 ooa.addr16[6], oa->addr16[6], u), 1725 ooa.addr16[7], oa->addr16[7], u); 1726 break; 1727#endif /* INET6 */ 1728 } 1729} 1730 1731 1732/* 1733 * Need to modulate the sequence numbers in the TCP SACK option 1734 * (credits to Krzysztof Pfaff for report and patch) 1735 */ 1736int 1737pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd, 1738 struct tcphdr *th, struct pf_state_peer *dst) 1739{ 1740 int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen; 1741#ifdef __FreeBSD__ 1742 u_int8_t opts[TCP_MAXOLEN], *opt = opts; 1743#else 1744 u_int8_t opts[MAX_TCPOPTLEN], *opt = opts; 1745#endif 1746 int copyback = 0, i, olen; 1747 struct sackblk sack; 1748 1749#define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2) 1750 if (hlen < TCPOLEN_SACKLEN || 1751 !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af)) 1752 return 0; 1753 1754 while (hlen >= TCPOLEN_SACKLEN) { 1755 olen = opt[1]; 1756 switch (*opt) { 1757 case TCPOPT_EOL: /* FALLTHROUGH */ 1758 case TCPOPT_NOP: 1759 opt++; 1760 hlen--; 1761 break; 1762 case TCPOPT_SACK: 1763 if (olen > hlen) 1764 olen = hlen; 1765 if (olen >= TCPOLEN_SACKLEN) { 1766 for (i = 2; i + TCPOLEN_SACK <= olen; 1767 i += TCPOLEN_SACK) { 1768 memcpy(&sack, &opt[i], sizeof(sack)); 1769 pf_change_a(&sack.start, &th->th_sum, 1770 htonl(ntohl(sack.start) - 1771 dst->seqdiff), 0); 1772 pf_change_a(&sack.end, &th->th_sum, 1773 htonl(ntohl(sack.end) - 1774 dst->seqdiff), 0); 1775 memcpy(&opt[i], &sack, sizeof(sack)); 1776 } 1777 copyback = 1; 1778 } 1779 /* FALLTHROUGH */ 1780 default: 1781 if (olen < 2) 1782 olen = 2; 1783 hlen -= olen; 1784 opt += olen; 1785 } 1786 } 1787 1788 if (copyback) 1789#ifdef __FreeBSD__ 1790 m_copyback(m, off + sizeof(*th), thoptlen, (caddr_t)opts); 1791#else 1792 m_copyback(m, off + sizeof(*th), thoptlen, opts); 1793#endif 1794 return (copyback); 1795} 1796 1797void 1798#ifdef __FreeBSD__ 1799pf_send_tcp(struct mbuf *replyto, const struct pf_rule *r, sa_family_t af, 1800#else 1801pf_send_tcp(const struct pf_rule *r, sa_family_t af, 1802#endif 1803 const struct pf_addr *saddr, const struct pf_addr *daddr, 1804 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack, 1805 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag, 1806 u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp) 1807{ 1808 struct mbuf *m; 1809 int len, tlen; 1810#ifdef INET 1811 struct ip *h; 1812#endif /* INET */ 1813#ifdef INET6 1814 struct ip6_hdr *h6; 1815#endif /* INET6 */ 1816 struct tcphdr *th; 1817 char *opt; 1818 struct pf_mtag *pf_mtag; 1819 1820#ifdef __FreeBSD__ 1821 KASSERT( 1822#ifdef INET 1823 af == AF_INET 1824#else 1825 0 1826#endif 1827 || 1828#ifdef INET6 1829 af == AF_INET6 1830#else 1831 0 1832#endif 1833 , ("Unsupported AF %d", af)); 1834 len = 0; 1835 th = NULL; 1836#ifdef INET 1837 h = NULL; 1838#endif 1839#ifdef INET6 1840 h6 = NULL; 1841#endif 1842#endif 1843 1844 /* maximum segment size tcp option */ 1845 tlen = sizeof(struct tcphdr); 1846 if (mss) 1847 tlen += 4; 1848 1849 switch (af) { 1850#ifdef INET 1851 case AF_INET: 1852 len = sizeof(struct ip) + tlen; 1853 break; 1854#endif /* INET */ 1855#ifdef INET6 1856 case AF_INET6: 1857 len = sizeof(struct ip6_hdr) + tlen; 1858 break; 1859#endif /* INET6 */ 1860 } 1861 1862 /* create outgoing mbuf */ 1863 m = m_gethdr(M_DONTWAIT, MT_HEADER); 1864 if (m == NULL) 1865 return; 1866#ifdef __FreeBSD__ 1867#ifdef MAC 1868 if (replyto) 1869 mac_netinet_firewall_reply(replyto, m); 1870 else 1871 mac_netinet_firewall_send(m); 1872#else 1873 (void)replyto; 1874#endif 1875#endif 1876 if ((pf_mtag = pf_get_mtag(m)) == NULL) { 1877 m_freem(m); 1878 return; 1879 } 1880 if (tag) 1881#ifdef __FreeBSD__ 1882 m->m_flags |= M_SKIP_FIREWALL; 1883#else 1884 pf_mtag->flags |= PF_TAG_GENERATED; 1885#endif 1886 1887 pf_mtag->tag = rtag; 1888 1889 if (r != NULL && r->rtableid >= 0) 1890#ifdef __FreeBSD__ 1891 { 1892 M_SETFIB(m, r->rtableid); 1893#endif 1894 pf_mtag->rtableid = r->rtableid; 1895#ifdef __FreeBSD__ 1896 } 1897#endif 1898#ifdef ALTQ 1899 if (r != NULL && r->qid) { 1900 pf_mtag->qid = r->qid; 1901 /* add hints for ecn */ 1902 pf_mtag->af = af; 1903 pf_mtag->hdr = mtod(m, struct ip *); 1904 } 1905#endif /* ALTQ */ 1906 m->m_data += max_linkhdr; 1907 m->m_pkthdr.len = m->m_len = len; 1908 m->m_pkthdr.rcvif = NULL; 1909 bzero(m->m_data, len); 1910 switch (af) { 1911#ifdef INET 1912 case AF_INET: 1913 h = mtod(m, struct ip *); 1914 1915 /* IP header fields included in the TCP checksum */ 1916 h->ip_p = IPPROTO_TCP; 1917 h->ip_len = htons(tlen); 1918 h->ip_src.s_addr = saddr->v4.s_addr; 1919 h->ip_dst.s_addr = daddr->v4.s_addr; 1920 1921 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip)); 1922 break; 1923#endif /* INET */ 1924#ifdef INET6 1925 case AF_INET6: 1926 h6 = mtod(m, struct ip6_hdr *); 1927 1928 /* IP header fields included in the TCP checksum */ 1929 h6->ip6_nxt = IPPROTO_TCP; 1930 h6->ip6_plen = htons(tlen); 1931 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr)); 1932 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr)); 1933 1934 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr)); 1935 break; 1936#endif /* INET6 */ 1937 } 1938 1939 /* TCP header */ 1940 th->th_sport = sport; 1941 th->th_dport = dport; 1942 th->th_seq = htonl(seq); 1943 th->th_ack = htonl(ack); 1944 th->th_off = tlen >> 2; 1945 th->th_flags = flags; 1946 th->th_win = htons(win); 1947 1948 if (mss) { 1949 opt = (char *)(th + 1); 1950 opt[0] = TCPOPT_MAXSEG; 1951 opt[1] = 4; 1952 HTONS(mss); 1953 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2); 1954 } 1955 1956 switch (af) { 1957#ifdef INET 1958 case AF_INET: 1959 /* TCP checksum */ 1960 th->th_sum = in_cksum(m, len); 1961 1962 /* Finish the IP header */ 1963 h->ip_v = 4; 1964 h->ip_hl = sizeof(*h) >> 2; 1965 h->ip_tos = IPTOS_LOWDELAY; 1966#ifdef __FreeBSD__ 1967 h->ip_off = V_path_mtu_discovery ? IP_DF : 0; 1968 h->ip_len = len; 1969#else 1970 h->ip_off = htons(ip_mtudisc ? IP_DF : 0); 1971 h->ip_len = htons(len); 1972#endif 1973 h->ip_ttl = ttl ? ttl : V_ip_defttl; 1974 h->ip_sum = 0; 1975 if (eh == NULL) { 1976#ifdef __FreeBSD__ 1977 PF_UNLOCK(); 1978 ip_output(m, (void *)NULL, (void *)NULL, 0, 1979 (void *)NULL, (void *)NULL); 1980 PF_LOCK(); 1981#else /* ! __FreeBSD__ */ 1982 ip_output(m, (void *)NULL, (void *)NULL, 0, 1983 (void *)NULL, (void *)NULL); 1984#endif 1985 } else { 1986 struct route ro; 1987 struct rtentry rt; 1988 struct ether_header *e = (void *)ro.ro_dst.sa_data; 1989 1990 if (ifp == NULL) { 1991 m_freem(m); 1992 return; 1993 } 1994 rt.rt_ifp = ifp; 1995 ro.ro_rt = &rt; 1996 ro.ro_dst.sa_len = sizeof(ro.ro_dst); 1997 ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT; 1998 bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN); 1999 bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN); 2000 e->ether_type = eh->ether_type; 2001#ifdef __FreeBSD__ 2002 PF_UNLOCK(); 2003 /* XXX_IMPORT: later */ 2004 ip_output(m, (void *)NULL, &ro, 0, 2005 (void *)NULL, (void *)NULL); 2006 PF_LOCK(); 2007#else /* ! __FreeBSD__ */ 2008 ip_output(m, (void *)NULL, &ro, IP_ROUTETOETHER, 2009 (void *)NULL, (void *)NULL); 2010#endif 2011 } 2012 break; 2013#endif /* INET */ 2014#ifdef INET6 2015 case AF_INET6: 2016 /* TCP checksum */ 2017 th->th_sum = in6_cksum(m, IPPROTO_TCP, 2018 sizeof(struct ip6_hdr), tlen); 2019 2020 h6->ip6_vfc |= IPV6_VERSION; 2021 h6->ip6_hlim = IPV6_DEFHLIM; 2022 2023#ifdef __FreeBSD__ 2024 PF_UNLOCK(); 2025 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL); 2026 PF_LOCK(); 2027#else 2028 ip6_output(m, NULL, NULL, 0, NULL, NULL); 2029#endif 2030 break; 2031#endif /* INET6 */ 2032 } 2033} 2034 2035void 2036pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af, 2037 struct pf_rule *r) 2038{ 2039 struct pf_mtag *pf_mtag; 2040 struct mbuf *m0; 2041#ifdef __FreeBSD__ 2042#ifdef INET 2043 struct ip *ip; 2044#endif 2045#endif 2046 2047#ifdef __FreeBSD__ 2048 m0 = m_copypacket(m, M_DONTWAIT); 2049 if (m0 == NULL) 2050 return; 2051#else 2052 m0 = m_copy(m, 0, M_COPYALL); 2053#endif 2054 if ((pf_mtag = pf_get_mtag(m0)) == NULL) 2055 return; 2056#ifdef __FreeBSD__ 2057 /* XXX: revisit */ 2058 m0->m_flags |= M_SKIP_FIREWALL; 2059#else 2060 pf_mtag->flags |= PF_TAG_GENERATED; 2061#endif 2062 2063 if (r->rtableid >= 0) 2064#ifdef __FreeBSD__ 2065 { 2066 M_SETFIB(m0, r->rtableid); 2067#endif 2068 pf_mtag->rtableid = r->rtableid; 2069#ifdef __FreeBSD__ 2070 } 2071#endif 2072 2073#ifdef ALTQ 2074 if (r->qid) { 2075 pf_mtag->qid = r->qid; 2076 /* add hints for ecn */ 2077 pf_mtag->af = af; 2078 pf_mtag->hdr = mtod(m0, struct ip *); 2079 } 2080#endif /* ALTQ */ 2081 2082 switch (af) { 2083#ifdef INET 2084 case AF_INET: 2085#ifdef __FreeBSD__ 2086 /* icmp_error() expects host byte ordering */ 2087 ip = mtod(m0, struct ip *); 2088 NTOHS(ip->ip_len); 2089 NTOHS(ip->ip_off); 2090 PF_UNLOCK(); 2091 icmp_error(m0, type, code, 0, 0); 2092 PF_LOCK(); 2093#else 2094 icmp_error(m0, type, code, 0, 0); 2095#endif 2096 break; 2097#endif /* INET */ 2098#ifdef INET6 2099 case AF_INET6: 2100#ifdef __FreeBSD__ 2101 PF_UNLOCK(); 2102#endif 2103 icmp6_error(m0, type, code, 0); 2104#ifdef __FreeBSD__ 2105 PF_LOCK(); 2106#endif 2107 break; 2108#endif /* INET6 */ 2109 } 2110} 2111 2112/* 2113 * Return 1 if the addresses a and b match (with mask m), otherwise return 0. 2114 * If n is 0, they match if they are equal. If n is != 0, they match if they 2115 * are different. 2116 */ 2117int 2118pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m, 2119 struct pf_addr *b, sa_family_t af) 2120{ 2121 int match = 0; 2122 2123 switch (af) { 2124#ifdef INET 2125 case AF_INET: 2126 if ((a->addr32[0] & m->addr32[0]) == 2127 (b->addr32[0] & m->addr32[0])) 2128 match++; 2129 break; 2130#endif /* INET */ 2131#ifdef INET6 2132 case AF_INET6: 2133 if (((a->addr32[0] & m->addr32[0]) == 2134 (b->addr32[0] & m->addr32[0])) && 2135 ((a->addr32[1] & m->addr32[1]) == 2136 (b->addr32[1] & m->addr32[1])) && 2137 ((a->addr32[2] & m->addr32[2]) == 2138 (b->addr32[2] & m->addr32[2])) && 2139 ((a->addr32[3] & m->addr32[3]) == 2140 (b->addr32[3] & m->addr32[3]))) 2141 match++; 2142 break; 2143#endif /* INET6 */ 2144 } 2145 if (match) { 2146 if (n) 2147 return (0); 2148 else 2149 return (1); 2150 } else { 2151 if (n) 2152 return (1); 2153 else 2154 return (0); 2155 } 2156} 2157 2158int 2159pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p) 2160{ 2161 switch (op) { 2162 case PF_OP_IRG: 2163 return ((p > a1) && (p < a2)); 2164 case PF_OP_XRG: 2165 return ((p < a1) || (p > a2)); 2166 case PF_OP_RRG: 2167 return ((p >= a1) && (p <= a2)); 2168 case PF_OP_EQ: 2169 return (p == a1); 2170 case PF_OP_NE: 2171 return (p != a1); 2172 case PF_OP_LT: 2173 return (p < a1); 2174 case PF_OP_LE: 2175 return (p <= a1); 2176 case PF_OP_GT: 2177 return (p > a1); 2178 case PF_OP_GE: 2179 return (p >= a1); 2180 } 2181 return (0); /* never reached */ 2182} 2183 2184int 2185pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p) 2186{ 2187 NTOHS(a1); 2188 NTOHS(a2); 2189 NTOHS(p); 2190 return (pf_match(op, a1, a2, p)); 2191} 2192 2193int 2194pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u) 2195{ 2196 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE) 2197 return (0); 2198 return (pf_match(op, a1, a2, u)); 2199} 2200 2201int 2202pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g) 2203{ 2204 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE) 2205 return (0); 2206 return (pf_match(op, a1, a2, g)); 2207} 2208 2209#ifndef __FreeBSD__ 2210struct pf_mtag * 2211pf_find_mtag(struct mbuf *m) 2212{ 2213 struct m_tag *mtag; 2214 2215 if ((mtag = m_tag_find(m, PACKET_TAG_PF, NULL)) == NULL) 2216 return (NULL); 2217 2218 return ((struct pf_mtag *)(mtag + 1)); 2219} 2220 2221struct pf_mtag * 2222pf_get_mtag(struct mbuf *m) 2223{ 2224 struct m_tag *mtag; 2225 2226 if ((mtag = m_tag_find(m, PACKET_TAG_PF, NULL)) == NULL) { 2227 mtag = m_tag_get(PACKET_TAG_PF, sizeof(struct pf_mtag), 2228 M_NOWAIT); 2229 if (mtag == NULL) 2230 return (NULL); 2231 bzero(mtag + 1, sizeof(struct pf_mtag)); 2232 m_tag_prepend(m, mtag); 2233 } 2234 2235 return ((struct pf_mtag *)(mtag + 1)); 2236} 2237#endif 2238 2239int 2240pf_match_tag(struct mbuf *m, struct pf_rule *r, struct pf_mtag *pf_mtag, 2241 int *tag) 2242{ 2243 if (*tag == -1) 2244 *tag = pf_mtag->tag; 2245 2246 return ((!r->match_tag_not && r->match_tag == *tag) || 2247 (r->match_tag_not && r->match_tag != *tag)); 2248} 2249 2250int 2251pf_tag_packet(struct mbuf *m, struct pf_mtag *pf_mtag, int tag, int rtableid) 2252{ 2253 if (tag <= 0 && rtableid < 0) 2254 return (0); 2255 2256 if (pf_mtag == NULL) 2257 if ((pf_mtag = pf_get_mtag(m)) == NULL) 2258 return (1); 2259 if (tag > 0) 2260 pf_mtag->tag = tag; 2261 if (rtableid >= 0) 2262#ifdef __FreeBSD__ 2263 { 2264 M_SETFIB(m, rtableid); 2265#endif 2266 pf_mtag->rtableid = rtableid; 2267#ifdef __FreeBSD__ 2268 } 2269#endif 2270 2271 return (0); 2272} 2273 2274static void 2275pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n, 2276 struct pf_rule **r, struct pf_rule **a, int *match) 2277{ 2278 struct pf_anchor_stackframe *f; 2279 2280 (*r)->anchor->match = 0; 2281 if (match) 2282 *match = 0; 2283 if (*depth >= sizeof(pf_anchor_stack) / 2284 sizeof(pf_anchor_stack[0])) { 2285 printf("pf_step_into_anchor: stack overflow\n"); 2286 *r = TAILQ_NEXT(*r, entries); 2287 return; 2288 } else if (*depth == 0 && a != NULL) 2289 *a = *r; 2290 f = pf_anchor_stack + (*depth)++; 2291 f->rs = *rs; 2292 f->r = *r; 2293 if ((*r)->anchor_wildcard) { 2294 f->parent = &(*r)->anchor->children; 2295 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) == 2296 NULL) { 2297 *r = NULL; 2298 return; 2299 } 2300 *rs = &f->child->ruleset; 2301 } else { 2302 f->parent = NULL; 2303 f->child = NULL; 2304 *rs = &(*r)->anchor->ruleset; 2305 } 2306 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr); 2307} 2308 2309int 2310pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n, 2311 struct pf_rule **r, struct pf_rule **a, int *match) 2312{ 2313 struct pf_anchor_stackframe *f; 2314 int quick = 0; 2315 2316 do { 2317 if (*depth <= 0) 2318 break; 2319 f = pf_anchor_stack + *depth - 1; 2320 if (f->parent != NULL && f->child != NULL) { 2321 if (f->child->match || 2322 (match != NULL && *match)) { 2323 f->r->anchor->match = 1; 2324 *match = 0; 2325 } 2326 f->child = RB_NEXT(pf_anchor_node, f->parent, f->child); 2327 if (f->child != NULL) { 2328 *rs = &f->child->ruleset; 2329 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr); 2330 if (*r == NULL) 2331 continue; 2332 else 2333 break; 2334 } 2335 } 2336 (*depth)--; 2337 if (*depth == 0 && a != NULL) 2338 *a = NULL; 2339 *rs = f->rs; 2340 if (f->r->anchor->match || (match != NULL && *match)) 2341 quick = f->r->quick; 2342 *r = TAILQ_NEXT(f->r, entries); 2343 } while (*r == NULL); 2344 2345 return (quick); 2346} 2347 2348#ifdef INET6 2349void 2350pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr, 2351 struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af) 2352{ 2353 switch (af) { 2354#ifdef INET 2355 case AF_INET: 2356 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) | 2357 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]); 2358 break; 2359#endif /* INET */ 2360 case AF_INET6: 2361 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) | 2362 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]); 2363 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) | 2364 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]); 2365 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) | 2366 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]); 2367 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) | 2368 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]); 2369 break; 2370 } 2371} 2372 2373void 2374pf_addr_inc(struct pf_addr *addr, sa_family_t af) 2375{ 2376 switch (af) { 2377#ifdef INET 2378 case AF_INET: 2379 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1); 2380 break; 2381#endif /* INET */ 2382 case AF_INET6: 2383 if (addr->addr32[3] == 0xffffffff) { 2384 addr->addr32[3] = 0; 2385 if (addr->addr32[2] == 0xffffffff) { 2386 addr->addr32[2] = 0; 2387 if (addr->addr32[1] == 0xffffffff) { 2388 addr->addr32[1] = 0; 2389 addr->addr32[0] = 2390 htonl(ntohl(addr->addr32[0]) + 1); 2391 } else 2392 addr->addr32[1] = 2393 htonl(ntohl(addr->addr32[1]) + 1); 2394 } else 2395 addr->addr32[2] = 2396 htonl(ntohl(addr->addr32[2]) + 1); 2397 } else 2398 addr->addr32[3] = 2399 htonl(ntohl(addr->addr32[3]) + 1); 2400 break; 2401 } 2402} 2403#endif /* INET6 */ 2404 2405#define mix(a,b,c) \ 2406 do { \ 2407 a -= b; a -= c; a ^= (c >> 13); \ 2408 b -= c; b -= a; b ^= (a << 8); \ 2409 c -= a; c -= b; c ^= (b >> 13); \ 2410 a -= b; a -= c; a ^= (c >> 12); \ 2411 b -= c; b -= a; b ^= (a << 16); \ 2412 c -= a; c -= b; c ^= (b >> 5); \ 2413 a -= b; a -= c; a ^= (c >> 3); \ 2414 b -= c; b -= a; b ^= (a << 10); \ 2415 c -= a; c -= b; c ^= (b >> 15); \ 2416 } while (0) 2417 2418/* 2419 * hash function based on bridge_hash in if_bridge.c 2420 */ 2421void 2422pf_hash(struct pf_addr *inaddr, struct pf_addr *hash, 2423 struct pf_poolhashkey *key, sa_family_t af) 2424{ 2425 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0]; 2426 2427 switch (af) { 2428#ifdef INET 2429 case AF_INET: 2430 a += inaddr->addr32[0]; 2431 b += key->key32[1]; 2432 mix(a, b, c); 2433 hash->addr32[0] = c + key->key32[2]; 2434 break; 2435#endif /* INET */ 2436#ifdef INET6 2437 case AF_INET6: 2438 a += inaddr->addr32[0]; 2439 b += inaddr->addr32[2]; 2440 mix(a, b, c); 2441 hash->addr32[0] = c; 2442 a += inaddr->addr32[1]; 2443 b += inaddr->addr32[3]; 2444 c += key->key32[1]; 2445 mix(a, b, c); 2446 hash->addr32[1] = c; 2447 a += inaddr->addr32[2]; 2448 b += inaddr->addr32[1]; 2449 c += key->key32[2]; 2450 mix(a, b, c); 2451 hash->addr32[2] = c; 2452 a += inaddr->addr32[3]; 2453 b += inaddr->addr32[0]; 2454 c += key->key32[3]; 2455 mix(a, b, c); 2456 hash->addr32[3] = c; 2457 break; 2458#endif /* INET6 */ 2459 } 2460} 2461 2462int 2463pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr, 2464 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn) 2465{ 2466 unsigned char hash[16]; 2467 struct pf_pool *rpool = &r->rpool; 2468 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr; 2469 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask; 2470 struct pf_pooladdr *acur = rpool->cur; 2471 struct pf_src_node k; 2472 2473 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR && 2474 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { 2475 k.af = af; 2476 PF_ACPY(&k.addr, saddr, af); 2477 if (r->rule_flag & PFRULE_RULESRCTRACK || 2478 r->rpool.opts & PF_POOL_STICKYADDR) 2479 k.rule.ptr = r; 2480 else 2481 k.rule.ptr = NULL; 2482 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++; 2483 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k); 2484 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) { 2485 PF_ACPY(naddr, &(*sn)->raddr, af); 2486 if (pf_status.debug >= PF_DEBUG_MISC) { 2487 printf("pf_map_addr: src tracking maps "); 2488 pf_print_host(&k.addr, 0, af); 2489 printf(" to "); 2490 pf_print_host(naddr, 0, af); 2491 printf("\n"); 2492 } 2493 return (0); 2494 } 2495 } 2496 2497 if (rpool->cur->addr.type == PF_ADDR_NOROUTE) 2498 return (1); 2499 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2500 switch (af) { 2501#ifdef INET 2502 case AF_INET: 2503 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 && 2504 (rpool->opts & PF_POOL_TYPEMASK) != 2505 PF_POOL_ROUNDROBIN) 2506 return (1); 2507 raddr = &rpool->cur->addr.p.dyn->pfid_addr4; 2508 rmask = &rpool->cur->addr.p.dyn->pfid_mask4; 2509 break; 2510#endif /* INET */ 2511#ifdef INET6 2512 case AF_INET6: 2513 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 && 2514 (rpool->opts & PF_POOL_TYPEMASK) != 2515 PF_POOL_ROUNDROBIN) 2516 return (1); 2517 raddr = &rpool->cur->addr.p.dyn->pfid_addr6; 2518 rmask = &rpool->cur->addr.p.dyn->pfid_mask6; 2519 break; 2520#endif /* INET6 */ 2521 } 2522 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2523 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN) 2524 return (1); /* unsupported */ 2525 } else { 2526 raddr = &rpool->cur->addr.v.a.addr; 2527 rmask = &rpool->cur->addr.v.a.mask; 2528 } 2529 2530 switch (rpool->opts & PF_POOL_TYPEMASK) { 2531 case PF_POOL_NONE: 2532 PF_ACPY(naddr, raddr, af); 2533 break; 2534 case PF_POOL_BITMASK: 2535 PF_POOLMASK(naddr, raddr, rmask, saddr, af); 2536 break; 2537 case PF_POOL_RANDOM: 2538 if (init_addr != NULL && PF_AZERO(init_addr, af)) { 2539 switch (af) { 2540#ifdef INET 2541 case AF_INET: 2542 rpool->counter.addr32[0] = htonl(arc4random()); 2543 break; 2544#endif /* INET */ 2545#ifdef INET6 2546 case AF_INET6: 2547 if (rmask->addr32[3] != 0xffffffff) 2548 rpool->counter.addr32[3] = 2549 htonl(arc4random()); 2550 else 2551 break; 2552 if (rmask->addr32[2] != 0xffffffff) 2553 rpool->counter.addr32[2] = 2554 htonl(arc4random()); 2555 else 2556 break; 2557 if (rmask->addr32[1] != 0xffffffff) 2558 rpool->counter.addr32[1] = 2559 htonl(arc4random()); 2560 else 2561 break; 2562 if (rmask->addr32[0] != 0xffffffff) 2563 rpool->counter.addr32[0] = 2564 htonl(arc4random()); 2565 break; 2566#endif /* INET6 */ 2567 } 2568 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af); 2569 PF_ACPY(init_addr, naddr, af); 2570 2571 } else { 2572 PF_AINC(&rpool->counter, af); 2573 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af); 2574 } 2575 break; 2576 case PF_POOL_SRCHASH: 2577 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af); 2578 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af); 2579 break; 2580 case PF_POOL_ROUNDROBIN: 2581 if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2582 if (!pfr_pool_get(rpool->cur->addr.p.tbl, 2583 &rpool->tblidx, &rpool->counter, 2584 &raddr, &rmask, af)) 2585 goto get_addr; 2586 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2587 if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, 2588 &rpool->tblidx, &rpool->counter, 2589 &raddr, &rmask, af)) 2590 goto get_addr; 2591 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af)) 2592 goto get_addr; 2593 2594 try_next: 2595 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL) 2596 rpool->cur = TAILQ_FIRST(&rpool->list); 2597 if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2598 rpool->tblidx = -1; 2599 if (pfr_pool_get(rpool->cur->addr.p.tbl, 2600 &rpool->tblidx, &rpool->counter, 2601 &raddr, &rmask, af)) { 2602 /* table contains no address of type 'af' */ 2603 if (rpool->cur != acur) 2604 goto try_next; 2605 return (1); 2606 } 2607 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2608 rpool->tblidx = -1; 2609 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, 2610 &rpool->tblidx, &rpool->counter, 2611 &raddr, &rmask, af)) { 2612 /* table contains no address of type 'af' */ 2613 if (rpool->cur != acur) 2614 goto try_next; 2615 return (1); 2616 } 2617 } else { 2618 raddr = &rpool->cur->addr.v.a.addr; 2619 rmask = &rpool->cur->addr.v.a.mask; 2620 PF_ACPY(&rpool->counter, raddr, af); 2621 } 2622 2623 get_addr: 2624 PF_ACPY(naddr, &rpool->counter, af); 2625 if (init_addr != NULL && PF_AZERO(init_addr, af)) 2626 PF_ACPY(init_addr, naddr, af); 2627 PF_AINC(&rpool->counter, af); 2628 break; 2629 } 2630 if (*sn != NULL) 2631 PF_ACPY(&(*sn)->raddr, naddr, af); 2632 2633 if (pf_status.debug >= PF_DEBUG_MISC && 2634 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { 2635 printf("pf_map_addr: selected address "); 2636 pf_print_host(naddr, 0, af); 2637 printf("\n"); 2638 } 2639 2640 return (0); 2641} 2642 2643int 2644pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r, 2645 struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport, 2646 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high, 2647 struct pf_src_node **sn) 2648{ 2649 struct pf_state_cmp key; 2650 struct pf_addr init_addr; 2651 u_int16_t cut; 2652 2653 bzero(&init_addr, sizeof(init_addr)); 2654 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn)) 2655 return (1); 2656 2657 if (proto == IPPROTO_ICMP) { 2658 low = 1; 2659 high = 65535; 2660 } 2661 2662 do { 2663 key.af = af; 2664 key.proto = proto; 2665 PF_ACPY(&key.ext.addr, daddr, key.af); 2666 PF_ACPY(&key.gwy.addr, naddr, key.af); 2667 key.ext.port = dport; 2668 2669 /* 2670 * port search; start random, step; 2671 * similar 2 portloop in in_pcbbind 2672 */ 2673 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP || 2674 proto == IPPROTO_ICMP)) { 2675 key.gwy.port = dport; 2676 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) 2677 return (0); 2678 } else if (low == 0 && high == 0) { 2679 key.gwy.port = *nport; 2680 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) 2681 return (0); 2682 } else if (low == high) { 2683 key.gwy.port = htons(low); 2684 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) { 2685 *nport = htons(low); 2686 return (0); 2687 } 2688 } else { 2689 u_int16_t tmp; 2690 2691 if (low > high) { 2692 tmp = low; 2693 low = high; 2694 high = tmp; 2695 } 2696 /* low < high */ 2697 cut = htonl(arc4random()) % (1 + high - low) + low; 2698 /* low <= cut <= high */ 2699 for (tmp = cut; tmp <= high; ++(tmp)) { 2700 key.gwy.port = htons(tmp); 2701 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == 2702 NULL) { 2703 *nport = htons(tmp); 2704 return (0); 2705 } 2706 } 2707 for (tmp = cut - 1; tmp >= low; --(tmp)) { 2708 key.gwy.port = htons(tmp); 2709 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == 2710 NULL) { 2711 *nport = htons(tmp); 2712 return (0); 2713 } 2714 } 2715 } 2716 2717 switch (r->rpool.opts & PF_POOL_TYPEMASK) { 2718 case PF_POOL_RANDOM: 2719 case PF_POOL_ROUNDROBIN: 2720 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn)) 2721 return (1); 2722 break; 2723 case PF_POOL_NONE: 2724 case PF_POOL_SRCHASH: 2725 case PF_POOL_BITMASK: 2726 default: 2727 return (1); 2728 } 2729 } while (! PF_AEQ(&init_addr, naddr, af) ); 2730 2731 return (1); /* none available */ 2732} 2733 2734struct pf_rule * 2735pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off, 2736 int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport, 2737 struct pf_addr *daddr, u_int16_t dport, int rs_num) 2738{ 2739 struct pf_rule *r, *rm = NULL; 2740 struct pf_ruleset *ruleset = NULL; 2741 int tag = -1; 2742 int rtableid = -1; 2743 int asd = 0; 2744 2745 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr); 2746 while (r && rm == NULL) { 2747 struct pf_rule_addr *src = NULL, *dst = NULL; 2748 struct pf_addr_wrap *xdst = NULL; 2749 2750 if (r->action == PF_BINAT && direction == PF_IN) { 2751 src = &r->dst; 2752 if (r->rpool.cur != NULL) 2753 xdst = &r->rpool.cur->addr; 2754 } else { 2755 src = &r->src; 2756 dst = &r->dst; 2757 } 2758 2759 r->evaluations++; 2760 if (pfi_kif_match(r->kif, kif) == r->ifnot) 2761 r = r->skip[PF_SKIP_IFP].ptr; 2762 else if (r->direction && r->direction != direction) 2763 r = r->skip[PF_SKIP_DIR].ptr; 2764 else if (r->af && r->af != pd->af) 2765 r = r->skip[PF_SKIP_AF].ptr; 2766 else if (r->proto && r->proto != pd->proto) 2767 r = r->skip[PF_SKIP_PROTO].ptr; 2768 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af, 2769 src->neg, kif)) 2770 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR : 2771 PF_SKIP_DST_ADDR].ptr; 2772 else if (src->port_op && !pf_match_port(src->port_op, 2773 src->port[0], src->port[1], sport)) 2774 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT : 2775 PF_SKIP_DST_PORT].ptr; 2776 else if (dst != NULL && 2777 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL)) 2778 r = r->skip[PF_SKIP_DST_ADDR].ptr; 2779 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af, 2780 0, NULL)) 2781 r = TAILQ_NEXT(r, entries); 2782 else if (dst != NULL && dst->port_op && 2783 !pf_match_port(dst->port_op, dst->port[0], 2784 dst->port[1], dport)) 2785 r = r->skip[PF_SKIP_DST_PORT].ptr; 2786 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) 2787 r = TAILQ_NEXT(r, entries); 2788 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto != 2789 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m, 2790 off, pd->hdr.tcp), r->os_fingerprint))) 2791 r = TAILQ_NEXT(r, entries); 2792 else { 2793 if (r->tag) 2794 tag = r->tag; 2795 if (r->rtableid >= 0) 2796 rtableid = r->rtableid; 2797 if (r->anchor == NULL) { 2798 rm = r; 2799 } else 2800 pf_step_into_anchor(&asd, &ruleset, rs_num, 2801 &r, NULL, NULL); 2802 } 2803 if (r == NULL) 2804 pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r, 2805 NULL, NULL); 2806 } 2807 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) 2808 return (NULL); 2809 if (rm != NULL && (rm->action == PF_NONAT || 2810 rm->action == PF_NORDR || rm->action == PF_NOBINAT)) 2811 return (NULL); 2812 return (rm); 2813} 2814 2815struct pf_rule * 2816pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction, 2817 struct pfi_kif *kif, struct pf_src_node **sn, 2818 struct pf_addr *saddr, u_int16_t sport, 2819 struct pf_addr *daddr, u_int16_t dport, 2820 struct pf_addr *naddr, u_int16_t *nport) 2821{ 2822 struct pf_rule *r = NULL; 2823 2824 if (direction == PF_OUT) { 2825 r = pf_match_translation(pd, m, off, direction, kif, saddr, 2826 sport, daddr, dport, PF_RULESET_BINAT); 2827 if (r == NULL) 2828 r = pf_match_translation(pd, m, off, direction, kif, 2829 saddr, sport, daddr, dport, PF_RULESET_NAT); 2830 } else { 2831 r = pf_match_translation(pd, m, off, direction, kif, saddr, 2832 sport, daddr, dport, PF_RULESET_RDR); 2833 if (r == NULL) 2834 r = pf_match_translation(pd, m, off, direction, kif, 2835 saddr, sport, daddr, dport, PF_RULESET_BINAT); 2836 } 2837 2838 if (r != NULL) { 2839 switch (r->action) { 2840 case PF_NONAT: 2841 case PF_NOBINAT: 2842 case PF_NORDR: 2843 return (NULL); 2844 case PF_NAT: 2845 if (pf_get_sport(pd->af, pd->proto, r, saddr, 2846 daddr, dport, naddr, nport, r->rpool.proxy_port[0], 2847 r->rpool.proxy_port[1], sn)) { 2848 DPFPRINTF(PF_DEBUG_MISC, 2849 ("pf: NAT proxy port allocation " 2850 "(%u-%u) failed\n", 2851 r->rpool.proxy_port[0], 2852 r->rpool.proxy_port[1])); 2853 return (NULL); 2854 } 2855 break; 2856 case PF_BINAT: 2857 switch (direction) { 2858 case PF_OUT: 2859 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){ 2860 switch (pd->af) { 2861#ifdef INET 2862 case AF_INET: 2863 if (r->rpool.cur->addr.p.dyn-> 2864 pfid_acnt4 < 1) 2865 return (NULL); 2866 PF_POOLMASK(naddr, 2867 &r->rpool.cur->addr.p.dyn-> 2868 pfid_addr4, 2869 &r->rpool.cur->addr.p.dyn-> 2870 pfid_mask4, 2871 saddr, AF_INET); 2872 break; 2873#endif /* INET */ 2874#ifdef INET6 2875 case AF_INET6: 2876 if (r->rpool.cur->addr.p.dyn-> 2877 pfid_acnt6 < 1) 2878 return (NULL); 2879 PF_POOLMASK(naddr, 2880 &r->rpool.cur->addr.p.dyn-> 2881 pfid_addr6, 2882 &r->rpool.cur->addr.p.dyn-> 2883 pfid_mask6, 2884 saddr, AF_INET6); 2885 break; 2886#endif /* INET6 */ 2887 } 2888 } else 2889 PF_POOLMASK(naddr, 2890 &r->rpool.cur->addr.v.a.addr, 2891 &r->rpool.cur->addr.v.a.mask, 2892 saddr, pd->af); 2893 break; 2894 case PF_IN: 2895 if (r->src.addr.type == PF_ADDR_DYNIFTL) { 2896 switch (pd->af) { 2897#ifdef INET 2898 case AF_INET: 2899 if (r->src.addr.p.dyn-> 2900 pfid_acnt4 < 1) 2901 return (NULL); 2902 PF_POOLMASK(naddr, 2903 &r->src.addr.p.dyn-> 2904 pfid_addr4, 2905 &r->src.addr.p.dyn-> 2906 pfid_mask4, 2907 daddr, AF_INET); 2908 break; 2909#endif /* INET */ 2910#ifdef INET6 2911 case AF_INET6: 2912 if (r->src.addr.p.dyn-> 2913 pfid_acnt6 < 1) 2914 return (NULL); 2915 PF_POOLMASK(naddr, 2916 &r->src.addr.p.dyn-> 2917 pfid_addr6, 2918 &r->src.addr.p.dyn-> 2919 pfid_mask6, 2920 daddr, AF_INET6); 2921 break; 2922#endif /* INET6 */ 2923 } 2924 } else 2925 PF_POOLMASK(naddr, 2926 &r->src.addr.v.a.addr, 2927 &r->src.addr.v.a.mask, daddr, 2928 pd->af); 2929 break; 2930 } 2931 break; 2932 case PF_RDR: { 2933 if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn)) 2934 return (NULL); 2935 if ((r->rpool.opts & PF_POOL_TYPEMASK) == 2936 PF_POOL_BITMASK) 2937 PF_POOLMASK(naddr, naddr, 2938 &r->rpool.cur->addr.v.a.mask, daddr, 2939 pd->af); 2940 2941 if (r->rpool.proxy_port[1]) { 2942 u_int32_t tmp_nport; 2943 2944 tmp_nport = ((ntohs(dport) - 2945 ntohs(r->dst.port[0])) % 2946 (r->rpool.proxy_port[1] - 2947 r->rpool.proxy_port[0] + 1)) + 2948 r->rpool.proxy_port[0]; 2949 2950 /* wrap around if necessary */ 2951 if (tmp_nport > 65535) 2952 tmp_nport -= 65535; 2953 *nport = htons((u_int16_t)tmp_nport); 2954 } else if (r->rpool.proxy_port[0]) 2955 *nport = htons(r->rpool.proxy_port[0]); 2956 break; 2957 } 2958 default: 2959 return (NULL); 2960 } 2961 } 2962 2963 return (r); 2964} 2965 2966int 2967#ifdef __FreeBSD__ 2968pf_socket_lookup(int direction, struct pf_pdesc *pd, struct inpcb *inp_arg) 2969#else 2970pf_socket_lookup(int direction, struct pf_pdesc *pd) 2971#endif 2972{ 2973 struct pf_addr *saddr, *daddr; 2974 u_int16_t sport, dport; 2975#ifdef __FreeBSD__ 2976 struct inpcbinfo *pi; 2977#else 2978 struct inpcbtable *tb; 2979#endif 2980 struct inpcb *inp; 2981 2982 if (pd == NULL) 2983 return (-1); 2984 pd->lookup.uid = UID_MAX; 2985 pd->lookup.gid = GID_MAX; 2986 pd->lookup.pid = NO_PID; /* XXX: revisit */ 2987#ifdef __FreeBSD__ 2988 if (inp_arg != NULL) { 2989 INP_LOCK_ASSERT(inp_arg); 2990 pd->lookup.uid = inp_arg->inp_cred->cr_uid; 2991 pd->lookup.gid = inp_arg->inp_cred->cr_groups[0]; 2992 return (1); 2993 } 2994#endif 2995 switch (pd->proto) { 2996 case IPPROTO_TCP: 2997 if (pd->hdr.tcp == NULL) 2998 return (-1); 2999 sport = pd->hdr.tcp->th_sport; 3000 dport = pd->hdr.tcp->th_dport; 3001#ifdef __FreeBSD__ 3002 pi = &V_tcbinfo; 3003#else 3004 tb = &tcbtable; 3005#endif 3006 break; 3007 case IPPROTO_UDP: 3008 if (pd->hdr.udp == NULL) 3009 return (-1); 3010 sport = pd->hdr.udp->uh_sport; 3011 dport = pd->hdr.udp->uh_dport; 3012#ifdef __FreeBSD__ 3013 pi = &V_udbinfo; 3014#else 3015 tb = &udbtable; 3016#endif 3017 break; 3018 default: 3019 return (-1); 3020 } 3021 if (direction == PF_IN) { 3022 saddr = pd->src; 3023 daddr = pd->dst; 3024 } else { 3025 u_int16_t p; 3026 3027 p = sport; 3028 sport = dport; 3029 dport = p; 3030 saddr = pd->dst; 3031 daddr = pd->src; 3032 } 3033 switch (pd->af) { 3034#ifdef INET 3035 case AF_INET: 3036#ifdef __FreeBSD__ 3037 INP_INFO_RLOCK(pi); /* XXX LOR */ 3038 inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4, 3039 dport, 0, NULL); 3040 if (inp == NULL) { 3041 inp = in_pcblookup_hash(pi, saddr->v4, sport, 3042 daddr->v4, dport, INPLOOKUP_WILDCARD, NULL); 3043 if(inp == NULL) { 3044 INP_INFO_RUNLOCK(pi); 3045 return (-1); 3046 } 3047 } 3048#else 3049 inp = in_pcbhashlookup(tb, saddr->v4, sport, daddr->v4, dport); 3050 if (inp == NULL) { 3051 inp = in_pcblookup_listen(tb, daddr->v4, dport, 0); 3052 if (inp == NULL) 3053 return (-1); 3054 } 3055#endif 3056 break; 3057#endif /* INET */ 3058#ifdef INET6 3059 case AF_INET6: 3060#ifdef __FreeBSD__ 3061 INP_INFO_RLOCK(pi); 3062 inp = in6_pcblookup_hash(pi, &saddr->v6, sport, 3063 &daddr->v6, dport, 0, NULL); 3064 if (inp == NULL) { 3065 inp = in6_pcblookup_hash(pi, &saddr->v6, sport, 3066 &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL); 3067 if (inp == NULL) { 3068 INP_INFO_RUNLOCK(pi); 3069 return (-1); 3070 } 3071 } 3072#else 3073 inp = in6_pcbhashlookup(tb, &saddr->v6, sport, &daddr->v6, 3074 dport); 3075 if (inp == NULL) { 3076 inp = in6_pcblookup_listen(tb, &daddr->v6, dport, 0); 3077 if (inp == NULL) 3078 return (-1); 3079 } 3080#endif 3081 break; 3082#endif /* INET6 */ 3083 3084 default: 3085 return (-1); 3086 } 3087#ifdef __FreeBSD__ 3088 pd->lookup.uid = inp->inp_cred->cr_uid; 3089 pd->lookup.gid = inp->inp_cred->cr_groups[0]; 3090 INP_INFO_RUNLOCK(pi); 3091#else 3092 pd->lookup.uid = inp->inp_socket->so_euid; 3093 pd->lookup.gid = inp->inp_socket->so_egid; 3094 pd->lookup.pid = inp->inp_socket->so_cpid; 3095#endif 3096 return (1); 3097} 3098 3099u_int8_t 3100pf_get_wscale(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_int8_t wscale = 0; 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 >= 3) { 3115 switch (*opt) { 3116 case TCPOPT_EOL: 3117 case TCPOPT_NOP: 3118 ++opt; 3119 --hlen; 3120 break; 3121 case TCPOPT_WINDOW: 3122 wscale = opt[2]; 3123 if (wscale > TCP_MAX_WINSHIFT) 3124 wscale = TCP_MAX_WINSHIFT; 3125 wscale |= PF_WSCALE_FLAG; 3126 /* FALLTHROUGH */ 3127 default: 3128 optlen = opt[1]; 3129 if (optlen < 2) 3130 optlen = 2; 3131 hlen -= optlen; 3132 opt += optlen; 3133 break; 3134 } 3135 } 3136 return (wscale); 3137} 3138 3139u_int16_t 3140pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af) 3141{ 3142 int hlen; 3143 u_int8_t hdr[60]; 3144 u_int8_t *opt, optlen; 3145 u_int16_t mss = V_tcp_mssdflt; 3146 3147 hlen = th_off << 2; /* hlen <= sizeof(hdr) */ 3148 if (hlen <= sizeof(struct tcphdr)) 3149 return (0); 3150 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af)) 3151 return (0); 3152 opt = hdr + sizeof(struct tcphdr); 3153 hlen -= sizeof(struct tcphdr); 3154 while (hlen >= TCPOLEN_MAXSEG) { 3155 switch (*opt) { 3156 case TCPOPT_EOL: 3157 case TCPOPT_NOP: 3158 ++opt; 3159 --hlen; 3160 break; 3161 case TCPOPT_MAXSEG: 3162 bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2); 3163 NTOHS(mss); 3164 /* FALLTHROUGH */ 3165 default: 3166 optlen = opt[1]; 3167 if (optlen < 2) 3168 optlen = 2; 3169 hlen -= optlen; 3170 opt += optlen; 3171 break; 3172 } 3173 } 3174 return (mss); 3175} 3176 3177u_int16_t 3178pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer) 3179{ 3180#ifdef INET 3181 struct sockaddr_in *dst; 3182 struct route ro; 3183#endif /* INET */ 3184#ifdef INET6 3185 struct sockaddr_in6 *dst6; 3186 struct route_in6 ro6; 3187#endif /* INET6 */ 3188 struct rtentry *rt = NULL; 3189 int hlen = 0; /* make the compiler happy */ 3190 u_int16_t mss = V_tcp_mssdflt; 3191 3192 switch (af) { 3193#ifdef INET 3194 case AF_INET: 3195 hlen = sizeof(struct ip); 3196 bzero(&ro, sizeof(ro)); 3197 dst = (struct sockaddr_in *)&ro.ro_dst; 3198 dst->sin_family = AF_INET; 3199 dst->sin_len = sizeof(*dst); 3200 dst->sin_addr = addr->v4; 3201#ifdef __FreeBSD__ 3202#ifdef RTF_PRCLONING 3203 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING)); 3204#else /* !RTF_PRCLONING */ 3205 in_rtalloc_ign(&ro, 0, 0); 3206#endif 3207#else /* ! __FreeBSD__ */ 3208 rtalloc_noclone(&ro, NO_CLONING); 3209#endif 3210 rt = ro.ro_rt; 3211 break; 3212#endif /* INET */ 3213#ifdef INET6 3214 case AF_INET6: 3215 hlen = sizeof(struct ip6_hdr); 3216 bzero(&ro6, sizeof(ro6)); 3217 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst; 3218 dst6->sin6_family = AF_INET6; 3219 dst6->sin6_len = sizeof(*dst6); 3220 dst6->sin6_addr = addr->v6; 3221#ifdef __FreeBSD__ 3222#ifdef RTF_PRCLONING 3223 rtalloc_ign((struct route *)&ro6, 3224 (RTF_CLONING | RTF_PRCLONING)); 3225#else /* !RTF_PRCLONING */ 3226 rtalloc_ign((struct route *)&ro6, 0); 3227#endif 3228#else /* ! __FreeBSD__ */ 3229 rtalloc_noclone((struct route *)&ro6, NO_CLONING); 3230#endif 3231 rt = ro6.ro_rt; 3232 break; 3233#endif /* INET6 */ 3234 } 3235 3236 if (rt && rt->rt_ifp) { 3237 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr); 3238 mss = max(V_tcp_mssdflt, mss); 3239 RTFREE(rt); 3240 } 3241 mss = min(mss, offer); 3242 mss = max(mss, 64); /* sanity - at least max opt space */ 3243 return (mss); 3244} 3245 3246void 3247pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr) 3248{ 3249 struct pf_rule *r = s->rule.ptr; 3250 3251 s->rt_kif = NULL; 3252 if (!r->rt || r->rt == PF_FASTROUTE) 3253 return; 3254 switch (s->af) { 3255#ifdef INET 3256 case AF_INET: 3257 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL, 3258 &s->nat_src_node); 3259 s->rt_kif = r->rpool.cur->kif; 3260 break; 3261#endif /* INET */ 3262#ifdef INET6 3263 case AF_INET6: 3264 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL, 3265 &s->nat_src_node); 3266 s->rt_kif = r->rpool.cur->kif; 3267 break; 3268#endif /* INET6 */ 3269 } 3270} 3271 3272int 3273pf_test_tcp(struct pf_rule **rm, struct pf_state **sm, int direction, 3274 struct pfi_kif *kif, struct mbuf *m, int off, void *h, 3275#ifdef __FreeBSD__ 3276 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3277 struct ifqueue *ifq, struct inpcb *inp) 3278#else 3279 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3280 struct ifqueue *ifq) 3281#endif 3282{ 3283 struct pf_rule *nr = NULL; 3284 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 3285 struct tcphdr *th = pd->hdr.tcp; 3286 u_int16_t bport, nport = 0; 3287 sa_family_t af = pd->af; 3288 struct pf_rule *r, *a = NULL; 3289 struct pf_ruleset *ruleset = NULL; 3290 struct pf_src_node *nsn = NULL; 3291 u_short reason; 3292 int rewrite = 0; 3293 int tag = -1, rtableid = -1; 3294 u_int16_t mss = V_tcp_mssdflt; 3295 int asd = 0; 3296 int match = 0; 3297 3298 if (pf_check_congestion(ifq)) { 3299 REASON_SET(&reason, PFRES_CONGEST); 3300 return (PF_DROP); 3301 } 3302 3303#ifdef __FreeBSD__ 3304 if (inp != NULL) 3305 pd->lookup.done = pf_socket_lookup(direction, pd, inp); 3306 else if (debug_pfugidhack) { 3307 PF_UNLOCK(); 3308 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n")); 3309 pd->lookup.done = pf_socket_lookup(direction, pd, inp); 3310 PF_LOCK(); 3311 } 3312#endif 3313 3314 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 3315 3316 if (direction == PF_OUT) { 3317 bport = nport = th->th_sport; 3318 /* check outgoing packet for BINAT/NAT */ 3319 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, 3320 saddr, th->th_sport, daddr, th->th_dport, 3321 &pd->naddr, &nport)) != NULL) { 3322 PF_ACPY(&pd->baddr, saddr, af); 3323 pf_change_ap(saddr, &th->th_sport, pd->ip_sum, 3324 &th->th_sum, &pd->naddr, nport, 0, af); 3325 rewrite++; 3326 if (nr->natpass) 3327 r = NULL; 3328 pd->nat_rule = nr; 3329 } 3330 } else { 3331 bport = nport = th->th_dport; 3332 /* check incoming packet for BINAT/RDR */ 3333 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, 3334 saddr, th->th_sport, daddr, th->th_dport, 3335 &pd->naddr, &nport)) != NULL) { 3336 PF_ACPY(&pd->baddr, daddr, af); 3337 pf_change_ap(daddr, &th->th_dport, pd->ip_sum, 3338 &th->th_sum, &pd->naddr, nport, 0, af); 3339 rewrite++; 3340 if (nr->natpass) 3341 r = NULL; 3342 pd->nat_rule = nr; 3343 } 3344 } 3345 3346 while (r != NULL) { 3347 r->evaluations++; 3348 if (pfi_kif_match(r->kif, kif) == r->ifnot) 3349 r = r->skip[PF_SKIP_IFP].ptr; 3350 else if (r->direction && r->direction != direction) 3351 r = r->skip[PF_SKIP_DIR].ptr; 3352 else if (r->af && r->af != af) 3353 r = r->skip[PF_SKIP_AF].ptr; 3354 else if (r->proto && r->proto != IPPROTO_TCP) 3355 r = r->skip[PF_SKIP_PROTO].ptr; 3356 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, 3357 r->src.neg, kif)) 3358 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 3359 else if (r->src.port_op && !pf_match_port(r->src.port_op, 3360 r->src.port[0], r->src.port[1], th->th_sport)) 3361 r = r->skip[PF_SKIP_SRC_PORT].ptr; 3362 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, 3363 r->dst.neg, NULL)) 3364 r = r->skip[PF_SKIP_DST_ADDR].ptr; 3365 else if (r->dst.port_op && !pf_match_port(r->dst.port_op, 3366 r->dst.port[0], r->dst.port[1], th->th_dport)) 3367 r = r->skip[PF_SKIP_DST_PORT].ptr; 3368 else if (r->tos && !(r->tos == pd->tos)) 3369 r = TAILQ_NEXT(r, entries); 3370 else if (r->rule_flag & PFRULE_FRAGMENT) 3371 r = TAILQ_NEXT(r, entries); 3372 else if ((r->flagset & th->th_flags) != r->flags) 3373 r = TAILQ_NEXT(r, entries); 3374 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done = 3375#ifdef __FreeBSD__ 3376 pf_socket_lookup(direction, pd, inp), 1)) && 3377#else 3378 pf_socket_lookup(direction, pd), 1)) && 3379#endif 3380 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1], 3381 pd->lookup.uid)) 3382 r = TAILQ_NEXT(r, entries); 3383 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done = 3384#ifdef __FreeBSD__ 3385 pf_socket_lookup(direction, pd, inp), 1)) && 3386#else 3387 pf_socket_lookup(direction, pd), 1)) && 3388#endif 3389 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1], 3390 pd->lookup.gid)) 3391 r = TAILQ_NEXT(r, entries); 3392 else if (r->prob && r->prob <= arc4random()) 3393 r = TAILQ_NEXT(r, entries); 3394 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) 3395 r = TAILQ_NEXT(r, entries); 3396 else if (r->os_fingerprint != PF_OSFP_ANY && !pf_osfp_match( 3397 pf_osfp_fingerprint(pd, m, off, th), r->os_fingerprint)) 3398 r = TAILQ_NEXT(r, entries); 3399 else { 3400 if (r->tag) 3401 tag = r->tag; 3402 if (r->rtableid >= 0) 3403 rtableid = r->rtableid; 3404 if (r->anchor == NULL) { 3405 match = 1; 3406 *rm = r; 3407 *am = a; 3408 *rsm = ruleset; 3409 if ((*rm)->quick) 3410 break; 3411 r = TAILQ_NEXT(r, entries); 3412 } else 3413 pf_step_into_anchor(&asd, &ruleset, 3414 PF_RULESET_FILTER, &r, &a, &match); 3415 } 3416 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 3417 PF_RULESET_FILTER, &r, &a, &match)) 3418 break; 3419 } 3420 r = *rm; 3421 a = *am; 3422 ruleset = *rsm; 3423 3424 REASON_SET(&reason, PFRES_MATCH); 3425 3426 if (r->log || (nr != NULL && nr->natpass && nr->log)) { 3427 if (rewrite) 3428#ifdef __FreeBSD__ 3429 m_copyback(m, off, sizeof(*th), (caddr_t)th); 3430#else 3431 m_copyback(m, off, sizeof(*th), th); 3432#endif 3433 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, 3434 a, ruleset, pd); 3435 } 3436 3437 if ((r->action == PF_DROP) && 3438 ((r->rule_flag & PFRULE_RETURNRST) || 3439 (r->rule_flag & PFRULE_RETURNICMP) || 3440 (r->rule_flag & PFRULE_RETURN))) { 3441 /* undo NAT changes, if they have taken place */ 3442 if (nr != NULL) { 3443 if (direction == PF_OUT) { 3444 pf_change_ap(saddr, &th->th_sport, pd->ip_sum, 3445 &th->th_sum, &pd->baddr, bport, 0, af); 3446 rewrite++; 3447 } else { 3448 pf_change_ap(daddr, &th->th_dport, pd->ip_sum, 3449 &th->th_sum, &pd->baddr, bport, 0, af); 3450 rewrite++; 3451 } 3452 } 3453 if (((r->rule_flag & PFRULE_RETURNRST) || 3454 (r->rule_flag & PFRULE_RETURN)) && 3455 !(th->th_flags & TH_RST)) { 3456 u_int32_t ack = ntohl(th->th_seq) + pd->p_len; 3457 3458 if (th->th_flags & TH_SYN) 3459 ack++; 3460 if (th->th_flags & TH_FIN) 3461 ack++; 3462#ifdef __FreeBSD__ 3463 pf_send_tcp(m, r, af, pd->dst, 3464#else 3465 pf_send_tcp(r, af, pd->dst, 3466#endif 3467 pd->src, th->th_dport, th->th_sport, 3468 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0, 3469 r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp); 3470 } else if ((af == AF_INET) && r->return_icmp) 3471 pf_send_icmp(m, r->return_icmp >> 8, 3472 r->return_icmp & 255, af, r); 3473 else if ((af == AF_INET6) && r->return_icmp6) 3474 pf_send_icmp(m, r->return_icmp6 >> 8, 3475 r->return_icmp6 & 255, af, r); 3476 } 3477 3478 if (r->action == PF_DROP) 3479 return (PF_DROP); 3480 3481 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) { 3482 REASON_SET(&reason, PFRES_MEMORY); 3483 return (PF_DROP); 3484 } 3485 3486 if (r->keep_state || nr != NULL || 3487 (pd->flags & PFDESC_TCP_NORM)) { 3488 /* create new state */ 3489 u_int16_t len; 3490 struct pf_state *s = NULL; 3491 struct pf_src_node *sn = NULL; 3492 3493 len = pd->tot_len - off - (th->th_off << 2); 3494 3495 /* check maximums */ 3496 if (r->max_states && (r->states >= r->max_states)) { 3497 pf_status.lcounters[LCNT_STATES]++; 3498 REASON_SET(&reason, PFRES_MAXSTATES); 3499 goto cleanup; 3500 } 3501 /* src node for filter rule */ 3502 if ((r->rule_flag & PFRULE_SRCTRACK || 3503 r->rpool.opts & PF_POOL_STICKYADDR) && 3504 pf_insert_src_node(&sn, r, saddr, af) != 0) { 3505 REASON_SET(&reason, PFRES_SRCLIMIT); 3506 goto cleanup; 3507 } 3508 /* src node for translation rule */ 3509 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 3510 ((direction == PF_OUT && 3511 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || 3512 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { 3513 REASON_SET(&reason, PFRES_SRCLIMIT); 3514 goto cleanup; 3515 } 3516 s = pool_get(&pf_state_pl, PR_NOWAIT); 3517 if (s == NULL) { 3518 REASON_SET(&reason, PFRES_MEMORY); 3519cleanup: 3520 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 3521 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 3522 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3523 pf_status.src_nodes--; 3524 pool_put(&pf_src_tree_pl, sn); 3525 } 3526 if (nsn != sn && nsn != NULL && nsn->states == 0 && 3527 nsn->expire == 0) { 3528 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 3529 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3530 pf_status.src_nodes--; 3531 pool_put(&pf_src_tree_pl, nsn); 3532 } 3533 return (PF_DROP); 3534 } 3535 bzero(s, sizeof(*s)); 3536 s->rule.ptr = r; 3537 s->nat_rule.ptr = nr; 3538 s->anchor.ptr = a; 3539 STATE_INC_COUNTERS(s); 3540 if (r->allow_opts) 3541 s->state_flags |= PFSTATE_ALLOWOPTS; 3542 if (r->rule_flag & PFRULE_STATESLOPPY) 3543 s->state_flags |= PFSTATE_SLOPPY; 3544 s->log = r->log & PF_LOG_ALL; 3545 if (nr != NULL) 3546 s->log |= nr->log & PF_LOG_ALL; 3547 s->proto = IPPROTO_TCP; 3548 s->direction = direction; 3549 s->af = af; 3550 if (direction == PF_OUT) { 3551 PF_ACPY(&s->gwy.addr, saddr, af); 3552 s->gwy.port = th->th_sport; /* sport */ 3553 PF_ACPY(&s->ext.addr, daddr, af); 3554 s->ext.port = th->th_dport; 3555 if (nr != NULL) { 3556 PF_ACPY(&s->lan.addr, &pd->baddr, af); 3557 s->lan.port = bport; 3558 } else { 3559 PF_ACPY(&s->lan.addr, &s->gwy.addr, af); 3560 s->lan.port = s->gwy.port; 3561 } 3562 } else { 3563 PF_ACPY(&s->lan.addr, daddr, af); 3564 s->lan.port = th->th_dport; 3565 PF_ACPY(&s->ext.addr, saddr, af); 3566 s->ext.port = th->th_sport; 3567 if (nr != NULL) { 3568 PF_ACPY(&s->gwy.addr, &pd->baddr, af); 3569 s->gwy.port = bport; 3570 } else { 3571 PF_ACPY(&s->gwy.addr, &s->lan.addr, af); 3572 s->gwy.port = s->lan.port; 3573 } 3574 } 3575 3576 s->src.seqlo = ntohl(th->th_seq); 3577 s->src.seqhi = s->src.seqlo + len + 1; 3578 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN && 3579 r->keep_state == PF_STATE_MODULATE) { 3580 /* Generate sequence number modulator */ 3581#ifdef __FreeBSD__ 3582 while ((s->src.seqdiff = 3583 pf_new_isn(s) - s->src.seqlo) == 0) 3584 ; 3585#else 3586 while ((s->src.seqdiff = 3587 tcp_rndiss_next() - s->src.seqlo) == 0) 3588 ; 3589#endif 3590 pf_change_a(&th->th_seq, &th->th_sum, 3591 htonl(s->src.seqlo + s->src.seqdiff), 0); 3592 rewrite = 1; 3593 } else 3594 s->src.seqdiff = 0; 3595 if (th->th_flags & TH_SYN) { 3596 s->src.seqhi++; 3597 s->src.wscale = pf_get_wscale(m, off, th->th_off, af); 3598 } 3599 s->src.max_win = MAX(ntohs(th->th_win), 1); 3600 if (s->src.wscale & PF_WSCALE_MASK) { 3601 /* Remove scale factor from initial window */ 3602 int win = s->src.max_win; 3603 win += 1 << (s->src.wscale & PF_WSCALE_MASK); 3604 s->src.max_win = (win - 1) >> 3605 (s->src.wscale & PF_WSCALE_MASK); 3606 } 3607 if (th->th_flags & TH_FIN) 3608 s->src.seqhi++; 3609 s->dst.seqhi = 1; 3610 s->dst.max_win = 1; 3611 s->src.state = TCPS_SYN_SENT; 3612 s->dst.state = TCPS_CLOSED; 3613 s->creation = time_second; 3614 s->expire = time_second; 3615 s->timeout = PFTM_TCP_FIRST_PACKET; 3616 pf_set_rt_ifp(s, saddr); 3617 if (sn != NULL) { 3618 s->src_node = sn; 3619 s->src_node->states++; 3620 } 3621 if (nsn != NULL) { 3622 PF_ACPY(&nsn->raddr, &pd->naddr, af); 3623 s->nat_src_node = nsn; 3624 s->nat_src_node->states++; 3625 } 3626 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m, 3627 off, pd, th, &s->src, &s->dst)) { 3628 REASON_SET(&reason, PFRES_MEMORY); 3629 pf_src_tree_remove_state(s); 3630 STATE_DEC_COUNTERS(s); 3631 pool_put(&pf_state_pl, s); 3632 return (PF_DROP); 3633 } 3634 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub && 3635 pf_normalize_tcp_stateful(m, off, pd, &reason, th, s, 3636 &s->src, &s->dst, &rewrite)) { 3637 /* This really shouldn't happen!!! */ 3638 DPFPRINTF(PF_DEBUG_URGENT, 3639 ("pf_normalize_tcp_stateful failed on first pkt")); 3640 pf_normalize_tcp_cleanup(s); 3641 pf_src_tree_remove_state(s); 3642 STATE_DEC_COUNTERS(s); 3643 pool_put(&pf_state_pl, s); 3644 return (PF_DROP); 3645 } 3646 if (pf_insert_state(BOUND_IFACE(r, kif), s)) { 3647 pf_normalize_tcp_cleanup(s); 3648 REASON_SET(&reason, PFRES_STATEINS); 3649 pf_src_tree_remove_state(s); 3650 STATE_DEC_COUNTERS(s); 3651 pool_put(&pf_state_pl, s); 3652 return (PF_DROP); 3653 } else 3654 *sm = s; 3655 if (tag > 0) { 3656 pf_tag_ref(tag); 3657 s->tag = tag; 3658 } 3659 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN && 3660 r->keep_state == PF_STATE_SYNPROXY) { 3661 s->src.state = PF_TCPS_PROXY_SRC; 3662 if (nr != NULL) { 3663 if (direction == PF_OUT) { 3664 pf_change_ap(saddr, &th->th_sport, 3665 pd->ip_sum, &th->th_sum, &pd->baddr, 3666 bport, 0, af); 3667 } else { 3668 pf_change_ap(daddr, &th->th_dport, 3669 pd->ip_sum, &th->th_sum, &pd->baddr, 3670 bport, 0, af); 3671 } 3672 } 3673 s->src.seqhi = htonl(arc4random()); 3674 /* Find mss option */ 3675 mss = pf_get_mss(m, off, th->th_off, af); 3676 mss = pf_calc_mss(saddr, af, mss); 3677 mss = pf_calc_mss(daddr, af, mss); 3678 s->src.mss = mss; 3679#ifdef __FreeBSD__ 3680 pf_send_tcp(NULL, r, af, daddr, saddr, th->th_dport, 3681#else 3682 pf_send_tcp(r, af, daddr, saddr, th->th_dport, 3683#endif 3684 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1, 3685 TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL); 3686 REASON_SET(&reason, PFRES_SYNPROXY); 3687 return (PF_SYNPROXY_DROP); 3688 } 3689 } 3690 3691 /* copy back packet headers if we performed NAT operations */ 3692 if (rewrite) 3693 m_copyback(m, off, sizeof(*th), (caddr_t)th); 3694 3695 return (PF_PASS); 3696} 3697 3698int 3699pf_test_udp(struct pf_rule **rm, struct pf_state **sm, int direction, 3700 struct pfi_kif *kif, struct mbuf *m, int off, void *h, 3701#ifdef __FreeBSD__ 3702 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3703 struct ifqueue *ifq, struct inpcb *inp) 3704#else 3705 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3706 struct ifqueue *ifq) 3707#endif 3708{ 3709 struct pf_rule *nr = NULL; 3710 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 3711 struct udphdr *uh = pd->hdr.udp; 3712 u_int16_t bport, nport = 0; 3713 sa_family_t af = pd->af; 3714 struct pf_rule *r, *a = NULL; 3715 struct pf_ruleset *ruleset = NULL; 3716 struct pf_src_node *nsn = NULL; 3717 u_short reason; 3718 int rewrite = 0; 3719 int tag = -1, rtableid = -1; 3720 int asd = 0; 3721 int match = 0; 3722 3723 if (pf_check_congestion(ifq)) { 3724 REASON_SET(&reason, PFRES_CONGEST); 3725 return (PF_DROP); 3726 } 3727 3728#ifdef __FreeBSD__ 3729 if (inp != NULL) 3730 pd->lookup.done = pf_socket_lookup(direction, pd, inp); 3731 else if (debug_pfugidhack) { 3732 PF_UNLOCK(); 3733 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n")); 3734 pd->lookup.done = pf_socket_lookup(direction, pd, inp); 3735 PF_LOCK(); 3736 } 3737#endif 3738 3739 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 3740 3741 if (direction == PF_OUT) { 3742 bport = nport = uh->uh_sport; 3743 /* check outgoing packet for BINAT/NAT */ 3744 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, 3745 saddr, uh->uh_sport, daddr, uh->uh_dport, 3746 &pd->naddr, &nport)) != NULL) { 3747 PF_ACPY(&pd->baddr, saddr, af); 3748 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum, 3749 &uh->uh_sum, &pd->naddr, nport, 1, af); 3750 rewrite++; 3751 if (nr->natpass) 3752 r = NULL; 3753 pd->nat_rule = nr; 3754 } 3755 } else { 3756 bport = nport = uh->uh_dport; 3757 /* check incoming packet for BINAT/RDR */ 3758 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, 3759 saddr, uh->uh_sport, daddr, uh->uh_dport, &pd->naddr, 3760 &nport)) != NULL) { 3761 PF_ACPY(&pd->baddr, daddr, af); 3762 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum, 3763 &uh->uh_sum, &pd->naddr, nport, 1, af); 3764 rewrite++; 3765 if (nr->natpass) 3766 r = NULL; 3767 pd->nat_rule = nr; 3768 } 3769 } 3770 3771 while (r != NULL) { 3772 r->evaluations++; 3773 if (pfi_kif_match(r->kif, kif) == r->ifnot) 3774 r = r->skip[PF_SKIP_IFP].ptr; 3775 else if (r->direction && r->direction != direction) 3776 r = r->skip[PF_SKIP_DIR].ptr; 3777 else if (r->af && r->af != af) 3778 r = r->skip[PF_SKIP_AF].ptr; 3779 else if (r->proto && r->proto != IPPROTO_UDP) 3780 r = r->skip[PF_SKIP_PROTO].ptr; 3781 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, 3782 r->src.neg, kif)) 3783 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 3784 else if (r->src.port_op && !pf_match_port(r->src.port_op, 3785 r->src.port[0], r->src.port[1], uh->uh_sport)) 3786 r = r->skip[PF_SKIP_SRC_PORT].ptr; 3787 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, 3788 r->dst.neg, NULL)) 3789 r = r->skip[PF_SKIP_DST_ADDR].ptr; 3790 else if (r->dst.port_op && !pf_match_port(r->dst.port_op, 3791 r->dst.port[0], r->dst.port[1], uh->uh_dport)) 3792 r = r->skip[PF_SKIP_DST_PORT].ptr; 3793 else if (r->tos && !(r->tos == pd->tos)) 3794 r = TAILQ_NEXT(r, entries); 3795 else if (r->rule_flag & PFRULE_FRAGMENT) 3796 r = TAILQ_NEXT(r, entries); 3797 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done = 3798#ifdef __FreeBSD__ 3799 pf_socket_lookup(direction, pd, inp), 1)) && 3800#else 3801 pf_socket_lookup(direction, pd), 1)) && 3802#endif 3803 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1], 3804 pd->lookup.uid)) 3805 r = TAILQ_NEXT(r, entries); 3806 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done = 3807#ifdef __FreeBSD__ 3808 pf_socket_lookup(direction, pd, inp), 1)) && 3809#else 3810 pf_socket_lookup(direction, pd), 1)) && 3811#endif 3812 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1], 3813 pd->lookup.gid)) 3814 r = TAILQ_NEXT(r, entries); 3815 else if (r->prob && r->prob <= arc4random()) 3816 r = TAILQ_NEXT(r, entries); 3817 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) 3818 r = TAILQ_NEXT(r, entries); 3819 else if (r->os_fingerprint != PF_OSFP_ANY) 3820 r = TAILQ_NEXT(r, entries); 3821 else { 3822 if (r->tag) 3823 tag = r->tag; 3824 if (r->rtableid >= 0) 3825 rtableid = r->rtableid; 3826 if (r->anchor == NULL) { 3827 match = 1; 3828 *rm = r; 3829 *am = a; 3830 *rsm = ruleset; 3831 if ((*rm)->quick) 3832 break; 3833 r = TAILQ_NEXT(r, entries); 3834 } else 3835 pf_step_into_anchor(&asd, &ruleset, 3836 PF_RULESET_FILTER, &r, &a, &match); 3837 } 3838 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 3839 PF_RULESET_FILTER, &r, &a, &match)) 3840 break; 3841 } 3842 r = *rm; 3843 a = *am; 3844 ruleset = *rsm; 3845 3846 REASON_SET(&reason, PFRES_MATCH); 3847 3848 if (r->log || (nr != NULL && nr->natpass && nr->log)) { 3849 if (rewrite) 3850#ifdef __FreeBSD__ 3851 m_copyback(m, off, sizeof(*uh), (caddr_t)uh); 3852#else 3853 m_copyback(m, off, sizeof(*uh), uh); 3854#endif 3855 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, 3856 a, ruleset, pd); 3857 } 3858 3859 if ((r->action == PF_DROP) && 3860 ((r->rule_flag & PFRULE_RETURNICMP) || 3861 (r->rule_flag & PFRULE_RETURN))) { 3862 /* undo NAT changes, if they have taken place */ 3863 if (nr != NULL) { 3864 if (direction == PF_OUT) { 3865 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum, 3866 &uh->uh_sum, &pd->baddr, bport, 1, af); 3867 rewrite++; 3868 } else { 3869 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum, 3870 &uh->uh_sum, &pd->baddr, bport, 1, af); 3871 rewrite++; 3872 } 3873 } 3874 if ((af == AF_INET) && r->return_icmp) 3875 pf_send_icmp(m, r->return_icmp >> 8, 3876 r->return_icmp & 255, af, r); 3877 else if ((af == AF_INET6) && r->return_icmp6) 3878 pf_send_icmp(m, r->return_icmp6 >> 8, 3879 r->return_icmp6 & 255, af, r); 3880 } 3881 3882 if (r->action == PF_DROP) 3883 return (PF_DROP); 3884 3885 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) { 3886 REASON_SET(&reason, PFRES_MEMORY); 3887 return (PF_DROP); 3888 } 3889 3890 if (r->keep_state || nr != NULL) { 3891 /* create new state */ 3892 struct pf_state *s = NULL; 3893 struct pf_src_node *sn = NULL; 3894 3895 /* check maximums */ 3896 if (r->max_states && (r->states >= r->max_states)) { 3897 pf_status.lcounters[LCNT_STATES]++; 3898 REASON_SET(&reason, PFRES_MAXSTATES); 3899 goto cleanup; 3900 } 3901 /* src node for filter rule */ 3902 if ((r->rule_flag & PFRULE_SRCTRACK || 3903 r->rpool.opts & PF_POOL_STICKYADDR) && 3904 pf_insert_src_node(&sn, r, saddr, af) != 0) { 3905 REASON_SET(&reason, PFRES_SRCLIMIT); 3906 goto cleanup; 3907 } 3908 /* src node for translation rule */ 3909 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 3910 ((direction == PF_OUT && 3911 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || 3912 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { 3913 REASON_SET(&reason, PFRES_SRCLIMIT); 3914 goto cleanup; 3915 } 3916 s = pool_get(&pf_state_pl, PR_NOWAIT); 3917 if (s == NULL) { 3918 REASON_SET(&reason, PFRES_MEMORY); 3919cleanup: 3920 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 3921 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 3922 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3923 pf_status.src_nodes--; 3924 pool_put(&pf_src_tree_pl, sn); 3925 } 3926 if (nsn != sn && nsn != NULL && nsn->states == 0 && 3927 nsn->expire == 0) { 3928 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 3929 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3930 pf_status.src_nodes--; 3931 pool_put(&pf_src_tree_pl, nsn); 3932 } 3933 return (PF_DROP); 3934 } 3935 bzero(s, sizeof(*s)); 3936 s->rule.ptr = r; 3937 s->nat_rule.ptr = nr; 3938 s->anchor.ptr = a; 3939 STATE_INC_COUNTERS(s); 3940 if (r->allow_opts) 3941 s->state_flags |= PFSTATE_ALLOWOPTS; 3942 if (r->rule_flag & PFRULE_STATESLOPPY) 3943 s->state_flags |= PFSTATE_SLOPPY; 3944 s->log = r->log & PF_LOG_ALL; 3945 if (nr != NULL) 3946 s->log |= nr->log & PF_LOG_ALL; 3947 s->proto = IPPROTO_UDP; 3948 s->direction = direction; 3949 s->af = af; 3950 if (direction == PF_OUT) { 3951 PF_ACPY(&s->gwy.addr, saddr, af); 3952 s->gwy.port = uh->uh_sport; 3953 PF_ACPY(&s->ext.addr, daddr, af); 3954 s->ext.port = uh->uh_dport; 3955 if (nr != NULL) { 3956 PF_ACPY(&s->lan.addr, &pd->baddr, af); 3957 s->lan.port = bport; 3958 } else { 3959 PF_ACPY(&s->lan.addr, &s->gwy.addr, af); 3960 s->lan.port = s->gwy.port; 3961 } 3962 } else { 3963 PF_ACPY(&s->lan.addr, daddr, af); 3964 s->lan.port = uh->uh_dport; 3965 PF_ACPY(&s->ext.addr, saddr, af); 3966 s->ext.port = uh->uh_sport; 3967 if (nr != NULL) { 3968 PF_ACPY(&s->gwy.addr, &pd->baddr, af); 3969 s->gwy.port = bport; 3970 } else { 3971 PF_ACPY(&s->gwy.addr, &s->lan.addr, af); 3972 s->gwy.port = s->lan.port; 3973 } 3974 } 3975 s->src.state = PFUDPS_SINGLE; 3976 s->dst.state = PFUDPS_NO_TRAFFIC; 3977 s->creation = time_second; 3978 s->expire = time_second; 3979 s->timeout = PFTM_UDP_FIRST_PACKET; 3980 pf_set_rt_ifp(s, saddr); 3981 if (sn != NULL) { 3982 s->src_node = sn; 3983 s->src_node->states++; 3984 } 3985 if (nsn != NULL) { 3986 PF_ACPY(&nsn->raddr, &pd->naddr, af); 3987 s->nat_src_node = nsn; 3988 s->nat_src_node->states++; 3989 } 3990 if (pf_insert_state(BOUND_IFACE(r, kif), s)) { 3991 REASON_SET(&reason, PFRES_STATEINS); 3992 pf_src_tree_remove_state(s); 3993 STATE_DEC_COUNTERS(s); 3994 pool_put(&pf_state_pl, s); 3995 return (PF_DROP); 3996 } else 3997 *sm = s; 3998 if (tag > 0) { 3999 pf_tag_ref(tag); 4000 s->tag = tag; 4001 } 4002 } 4003 4004 /* copy back packet headers if we performed NAT operations */ 4005 if (rewrite) 4006 m_copyback(m, off, sizeof(*uh), (caddr_t)uh); 4007 4008 return (PF_PASS); 4009} 4010 4011int 4012pf_test_icmp(struct pf_rule **rm, struct pf_state **sm, int direction, 4013 struct pfi_kif *kif, struct mbuf *m, int off, void *h, 4014 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 4015 struct ifqueue *ifq) 4016{ 4017 struct pf_rule *nr = NULL; 4018 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 4019 struct pf_rule *r, *a = NULL; 4020 struct pf_ruleset *ruleset = NULL; 4021 struct pf_src_node *nsn = NULL; 4022 u_short reason; 4023 u_int16_t icmpid = 0, bport, nport = 0; 4024 sa_family_t af = pd->af; 4025 u_int8_t icmptype = 0; /* make the compiler happy */ 4026 u_int8_t icmpcode = 0; /* make the compiler happy */ 4027 int state_icmp = 0; 4028 int tag = -1, rtableid = -1; 4029#ifdef INET6 4030 int rewrite = 0; 4031#endif /* INET6 */ 4032 int asd = 0; 4033 int match = 0; 4034 4035 if (pf_check_congestion(ifq)) { 4036 REASON_SET(&reason, PFRES_CONGEST); 4037 return (PF_DROP); 4038 } 4039 4040 switch (pd->proto) { 4041#ifdef INET 4042 case IPPROTO_ICMP: 4043 icmptype = pd->hdr.icmp->icmp_type; 4044 icmpcode = pd->hdr.icmp->icmp_code; 4045 icmpid = pd->hdr.icmp->icmp_id; 4046 4047 if (icmptype == ICMP_UNREACH || 4048 icmptype == ICMP_SOURCEQUENCH || 4049 icmptype == ICMP_REDIRECT || 4050 icmptype == ICMP_TIMXCEED || 4051 icmptype == ICMP_PARAMPROB) 4052 state_icmp++; 4053 break; 4054#endif /* INET */ 4055#ifdef INET6 4056 case IPPROTO_ICMPV6: 4057 icmptype = pd->hdr.icmp6->icmp6_type; 4058 icmpcode = pd->hdr.icmp6->icmp6_code; 4059 icmpid = pd->hdr.icmp6->icmp6_id; 4060 4061 if (icmptype == ICMP6_DST_UNREACH || 4062 icmptype == ICMP6_PACKET_TOO_BIG || 4063 icmptype == ICMP6_TIME_EXCEEDED || 4064 icmptype == ICMP6_PARAM_PROB) 4065 state_icmp++; 4066 break; 4067#endif /* INET6 */ 4068 } 4069 4070 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 4071 4072 if (direction == PF_OUT) { 4073 bport = nport = icmpid; 4074 /* check outgoing packet for BINAT/NAT */ 4075 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, 4076 saddr, icmpid, daddr, icmpid, &pd->naddr, &nport)) != 4077 NULL) { 4078 PF_ACPY(&pd->baddr, saddr, af); 4079 switch (af) { 4080#ifdef INET 4081 case AF_INET: 4082 pf_change_a(&saddr->v4.s_addr, pd->ip_sum, 4083 pd->naddr.v4.s_addr, 0); 4084 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup( 4085 pd->hdr.icmp->icmp_cksum, icmpid, nport, 0); 4086 pd->hdr.icmp->icmp_id = nport; 4087 m_copyback(m, off, ICMP_MINLEN, 4088 (caddr_t)pd->hdr.icmp); 4089 break; 4090#endif /* INET */ 4091#ifdef INET6 4092 case AF_INET6: 4093 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum, 4094 &pd->naddr, 0); 4095 rewrite++; 4096 break; 4097#endif /* INET6 */ 4098 } 4099 if (nr->natpass) 4100 r = NULL; 4101 pd->nat_rule = nr; 4102 } 4103 } else { 4104 bport = nport = icmpid; 4105 /* check incoming packet for BINAT/RDR */ 4106 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, 4107 saddr, icmpid, daddr, icmpid, &pd->naddr, &nport)) != 4108 NULL) { 4109 PF_ACPY(&pd->baddr, daddr, af); 4110 switch (af) { 4111#ifdef INET 4112 case AF_INET: 4113 pf_change_a(&daddr->v4.s_addr, 4114 pd->ip_sum, pd->naddr.v4.s_addr, 0); 4115 break; 4116#endif /* INET */ 4117#ifdef INET6 4118 case AF_INET6: 4119 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum, 4120 &pd->naddr, 0); 4121 rewrite++; 4122 break; 4123#endif /* INET6 */ 4124 } 4125 if (nr->natpass) 4126 r = NULL; 4127 pd->nat_rule = nr; 4128 } 4129 } 4130 4131 while (r != NULL) { 4132 r->evaluations++; 4133 if (pfi_kif_match(r->kif, kif) == r->ifnot) 4134 r = r->skip[PF_SKIP_IFP].ptr; 4135 else if (r->direction && r->direction != direction) 4136 r = r->skip[PF_SKIP_DIR].ptr; 4137 else if (r->af && r->af != af) 4138 r = r->skip[PF_SKIP_AF].ptr; 4139 else if (r->proto && r->proto != pd->proto) 4140 r = r->skip[PF_SKIP_PROTO].ptr; 4141 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, 4142 r->src.neg, kif)) 4143 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 4144 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, 4145 r->dst.neg, NULL)) 4146 r = r->skip[PF_SKIP_DST_ADDR].ptr; 4147 else if (r->type && r->type != icmptype + 1) 4148 r = TAILQ_NEXT(r, entries); 4149 else if (r->code && r->code != icmpcode + 1) 4150 r = TAILQ_NEXT(r, entries); 4151 else if (r->tos && !(r->tos == pd->tos)) 4152 r = TAILQ_NEXT(r, entries); 4153 else if (r->rule_flag & PFRULE_FRAGMENT) 4154 r = TAILQ_NEXT(r, entries); 4155 else if (r->prob && r->prob <= arc4random()) 4156 r = TAILQ_NEXT(r, entries); 4157 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) 4158 r = TAILQ_NEXT(r, entries); 4159 else if (r->os_fingerprint != PF_OSFP_ANY) 4160 r = TAILQ_NEXT(r, entries); 4161 else { 4162 if (r->tag) 4163 tag = r->tag; 4164 if (r->rtableid >= 0) 4165 rtableid = r->rtableid; 4166 if (r->anchor == NULL) { 4167 match = 1; 4168 *rm = r; 4169 *am = a; 4170 *rsm = ruleset; 4171 if ((*rm)->quick) 4172 break; 4173 r = TAILQ_NEXT(r, entries); 4174 } else 4175 pf_step_into_anchor(&asd, &ruleset, 4176 PF_RULESET_FILTER, &r, &a, &match); 4177 } 4178 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 4179 PF_RULESET_FILTER, &r, &a, &match)) 4180 break; 4181 } 4182 r = *rm; 4183 a = *am; 4184 ruleset = *rsm; 4185 4186 REASON_SET(&reason, PFRES_MATCH); 4187 4188 if (r->log || (nr != NULL && nr->natpass && nr->log)) { 4189#ifdef INET6 4190 if (rewrite) 4191 m_copyback(m, off, sizeof(struct icmp6_hdr), 4192 (caddr_t)pd->hdr.icmp6); 4193#endif /* INET6 */ 4194 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, 4195 a, ruleset, pd); 4196 } 4197 4198 if (r->action != PF_PASS) 4199 return (PF_DROP); 4200 4201 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) { 4202 REASON_SET(&reason, PFRES_MEMORY); 4203 return (PF_DROP); 4204 } 4205 4206 if (!state_icmp && (r->keep_state || nr != NULL)) { 4207 /* create new state */ 4208 struct pf_state *s = NULL; 4209 struct pf_src_node *sn = NULL; 4210 4211 /* check maximums */ 4212 if (r->max_states && (r->states >= r->max_states)) { 4213 pf_status.lcounters[LCNT_STATES]++; 4214 REASON_SET(&reason, PFRES_MAXSTATES); 4215 goto cleanup; 4216 } 4217 /* src node for filter rule */ 4218 if ((r->rule_flag & PFRULE_SRCTRACK || 4219 r->rpool.opts & PF_POOL_STICKYADDR) && 4220 pf_insert_src_node(&sn, r, saddr, af) != 0) { 4221 REASON_SET(&reason, PFRES_SRCLIMIT); 4222 goto cleanup; 4223 } 4224 /* src node for translation rule */ 4225 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 4226 ((direction == PF_OUT && 4227 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || 4228 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { 4229 REASON_SET(&reason, PFRES_SRCLIMIT); 4230 goto cleanup; 4231 } 4232 s = pool_get(&pf_state_pl, PR_NOWAIT); 4233 if (s == NULL) { 4234 REASON_SET(&reason, PFRES_MEMORY); 4235cleanup: 4236 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 4237 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 4238 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 4239 pf_status.src_nodes--; 4240 pool_put(&pf_src_tree_pl, sn); 4241 } 4242 if (nsn != sn && nsn != NULL && nsn->states == 0 && 4243 nsn->expire == 0) { 4244 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 4245 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 4246 pf_status.src_nodes--; 4247 pool_put(&pf_src_tree_pl, nsn); 4248 } 4249 return (PF_DROP); 4250 } 4251 bzero(s, sizeof(*s)); 4252 s->rule.ptr = r; 4253 s->nat_rule.ptr = nr; 4254 s->anchor.ptr = a; 4255 STATE_INC_COUNTERS(s); 4256 if (r->allow_opts) 4257 s->state_flags |= PFSTATE_ALLOWOPTS; 4258 if (r->rule_flag & PFRULE_STATESLOPPY) 4259 s->state_flags |= PFSTATE_SLOPPY; 4260 s->log = r->log & PF_LOG_ALL; 4261 if (nr != NULL) 4262 s->log |= nr->log & PF_LOG_ALL; 4263 s->proto = pd->proto; 4264 s->direction = direction; 4265 s->af = af; 4266 if (direction == PF_OUT) { 4267 PF_ACPY(&s->gwy.addr, saddr, af); 4268 s->gwy.port = nport; 4269 PF_ACPY(&s->ext.addr, daddr, af); 4270 s->ext.port = 0; 4271 if (nr != NULL) { 4272 PF_ACPY(&s->lan.addr, &pd->baddr, af); 4273 s->lan.port = bport; 4274 } else { 4275 PF_ACPY(&s->lan.addr, &s->gwy.addr, af); 4276 s->lan.port = s->gwy.port; 4277 } 4278 } else { 4279 PF_ACPY(&s->lan.addr, daddr, af); 4280 s->lan.port = nport; 4281 PF_ACPY(&s->ext.addr, saddr, af); 4282 s->ext.port = 0; 4283 if (nr != NULL) { 4284 PF_ACPY(&s->gwy.addr, &pd->baddr, af); 4285 s->gwy.port = bport; 4286 } else { 4287 PF_ACPY(&s->gwy.addr, &s->lan.addr, af); 4288 s->gwy.port = s->lan.port; 4289 } 4290 } 4291 s->creation = time_second; 4292 s->expire = time_second; 4293 s->timeout = PFTM_ICMP_FIRST_PACKET; 4294 pf_set_rt_ifp(s, saddr); 4295 if (sn != NULL) { 4296 s->src_node = sn; 4297 s->src_node->states++; 4298 } 4299 if (nsn != NULL) { 4300 PF_ACPY(&nsn->raddr, &pd->naddr, af); 4301 s->nat_src_node = nsn; 4302 s->nat_src_node->states++; 4303 } 4304 if (pf_insert_state(BOUND_IFACE(r, kif), s)) { 4305 REASON_SET(&reason, PFRES_STATEINS); 4306 pf_src_tree_remove_state(s); 4307 STATE_DEC_COUNTERS(s); 4308 pool_put(&pf_state_pl, s); 4309 return (PF_DROP); 4310 } else 4311 *sm = s; 4312 if (tag > 0) { 4313 pf_tag_ref(tag); 4314 s->tag = tag; 4315 } 4316 } 4317 4318#ifdef INET6 4319 /* copy back packet headers if we performed IPv6 NAT operations */ 4320 if (rewrite) 4321 m_copyback(m, off, sizeof(struct icmp6_hdr), 4322 (caddr_t)pd->hdr.icmp6); 4323#endif /* INET6 */ 4324 4325 return (PF_PASS); 4326} 4327 4328int 4329pf_test_other(struct pf_rule **rm, struct pf_state **sm, int direction, 4330 struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd, 4331 struct pf_rule **am, struct pf_ruleset **rsm, struct ifqueue *ifq) 4332{ 4333 struct pf_rule *nr = NULL; 4334 struct pf_rule *r, *a = NULL; 4335 struct pf_ruleset *ruleset = NULL; 4336 struct pf_src_node *nsn = NULL; 4337 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 4338 sa_family_t af = pd->af; 4339 u_short reason; 4340 int tag = -1, rtableid = -1; 4341 int asd = 0; 4342 int match = 0; 4343 4344 if (pf_check_congestion(ifq)) { 4345 REASON_SET(&reason, PFRES_CONGEST); 4346 return (PF_DROP); 4347 } 4348 4349 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 4350 4351 if (direction == PF_OUT) { 4352 /* check outgoing packet for BINAT/NAT */ 4353 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, 4354 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) { 4355 PF_ACPY(&pd->baddr, saddr, af); 4356 switch (af) { 4357#ifdef INET 4358 case AF_INET: 4359 pf_change_a(&saddr->v4.s_addr, pd->ip_sum, 4360 pd->naddr.v4.s_addr, 0); 4361 break; 4362#endif /* INET */ 4363#ifdef INET6 4364 case AF_INET6: 4365 PF_ACPY(saddr, &pd->naddr, af); 4366 break; 4367#endif /* INET6 */ 4368 } 4369 if (nr->natpass) 4370 r = NULL; 4371 pd->nat_rule = nr; 4372 } 4373 } else { 4374 /* check incoming packet for BINAT/RDR */ 4375 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, 4376 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) { 4377 PF_ACPY(&pd->baddr, daddr, af); 4378 switch (af) { 4379#ifdef INET 4380 case AF_INET: 4381 pf_change_a(&daddr->v4.s_addr, 4382 pd->ip_sum, pd->naddr.v4.s_addr, 0); 4383 break; 4384#endif /* INET */ 4385#ifdef INET6 4386 case AF_INET6: 4387 PF_ACPY(daddr, &pd->naddr, af); 4388 break; 4389#endif /* INET6 */ 4390 } 4391 if (nr->natpass) 4392 r = NULL; 4393 pd->nat_rule = nr; 4394 } 4395 } 4396 4397 while (r != NULL) { 4398 r->evaluations++; 4399 if (pfi_kif_match(r->kif, kif) == r->ifnot) 4400 r = r->skip[PF_SKIP_IFP].ptr; 4401 else if (r->direction && r->direction != direction) 4402 r = r->skip[PF_SKIP_DIR].ptr; 4403 else if (r->af && r->af != af) 4404 r = r->skip[PF_SKIP_AF].ptr; 4405 else if (r->proto && r->proto != pd->proto) 4406 r = r->skip[PF_SKIP_PROTO].ptr; 4407 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, 4408 r->src.neg, kif)) 4409 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 4410 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, 4411 r->dst.neg, NULL)) 4412 r = r->skip[PF_SKIP_DST_ADDR].ptr; 4413 else if (r->tos && !(r->tos == pd->tos)) 4414 r = TAILQ_NEXT(r, entries); 4415 else if (r->rule_flag & PFRULE_FRAGMENT) 4416 r = TAILQ_NEXT(r, entries); 4417 else if (r->prob && r->prob <= arc4random()) 4418 r = TAILQ_NEXT(r, entries); 4419 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) 4420 r = TAILQ_NEXT(r, entries); 4421 else if (r->os_fingerprint != PF_OSFP_ANY) 4422 r = TAILQ_NEXT(r, entries); 4423 else { 4424 if (r->tag) 4425 tag = r->tag; 4426 if (r->rtableid >= 0) 4427 rtableid = r->rtableid; 4428 if (r->anchor == NULL) { 4429 match = 1; 4430 *rm = r; 4431 *am = a; 4432 *rsm = ruleset; 4433 if ((*rm)->quick) 4434 break; 4435 r = TAILQ_NEXT(r, entries); 4436 } else 4437 pf_step_into_anchor(&asd, &ruleset, 4438 PF_RULESET_FILTER, &r, &a, &match); 4439 } 4440 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 4441 PF_RULESET_FILTER, &r, &a, &match)) 4442 break; 4443 } 4444 r = *rm; 4445 a = *am; 4446 ruleset = *rsm; 4447 4448 REASON_SET(&reason, PFRES_MATCH); 4449 4450 if (r->log || (nr != NULL && nr->natpass && nr->log)) 4451 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, 4452 a, ruleset, pd); 4453 4454 if ((r->action == PF_DROP) && 4455 ((r->rule_flag & PFRULE_RETURNICMP) || 4456 (r->rule_flag & PFRULE_RETURN))) { 4457 struct pf_addr *a = NULL; 4458 4459 if (nr != NULL) { 4460 if (direction == PF_OUT) 4461 a = saddr; 4462 else 4463 a = daddr; 4464 } 4465 if (a != NULL) { 4466 switch (af) { 4467#ifdef INET 4468 case AF_INET: 4469 pf_change_a(&a->v4.s_addr, pd->ip_sum, 4470 pd->baddr.v4.s_addr, 0); 4471 break; 4472#endif /* INET */ 4473#ifdef INET6 4474 case AF_INET6: 4475 PF_ACPY(a, &pd->baddr, af); 4476 break; 4477#endif /* INET6 */ 4478 } 4479 } 4480 if ((af == AF_INET) && r->return_icmp) 4481 pf_send_icmp(m, r->return_icmp >> 8, 4482 r->return_icmp & 255, af, r); 4483 else if ((af == AF_INET6) && r->return_icmp6) 4484 pf_send_icmp(m, r->return_icmp6 >> 8, 4485 r->return_icmp6 & 255, af, r); 4486 } 4487 4488 if (r->action != PF_PASS) 4489 return (PF_DROP); 4490 4491 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) { 4492 REASON_SET(&reason, PFRES_MEMORY); 4493 return (PF_DROP); 4494 } 4495 4496 if (r->keep_state || nr != NULL) { 4497 /* create new state */ 4498 struct pf_state *s = NULL; 4499 struct pf_src_node *sn = NULL; 4500 4501 /* check maximums */ 4502 if (r->max_states && (r->states >= r->max_states)) { 4503 pf_status.lcounters[LCNT_STATES]++; 4504 REASON_SET(&reason, PFRES_MAXSTATES); 4505 goto cleanup; 4506 } 4507 /* src node for filter rule */ 4508 if ((r->rule_flag & PFRULE_SRCTRACK || 4509 r->rpool.opts & PF_POOL_STICKYADDR) && 4510 pf_insert_src_node(&sn, r, saddr, af) != 0) { 4511 REASON_SET(&reason, PFRES_SRCLIMIT); 4512 goto cleanup; 4513 } 4514 /* src node for translation rule */ 4515 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 4516 ((direction == PF_OUT && 4517 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || 4518 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { 4519 REASON_SET(&reason, PFRES_SRCLIMIT); 4520 goto cleanup; 4521 } 4522 s = pool_get(&pf_state_pl, PR_NOWAIT); 4523 if (s == NULL) { 4524 REASON_SET(&reason, PFRES_MEMORY); 4525cleanup: 4526 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 4527 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 4528 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 4529 pf_status.src_nodes--; 4530 pool_put(&pf_src_tree_pl, sn); 4531 } 4532 if (nsn != sn && nsn != NULL && nsn->states == 0 && 4533 nsn->expire == 0) { 4534 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 4535 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 4536 pf_status.src_nodes--; 4537 pool_put(&pf_src_tree_pl, nsn); 4538 } 4539 return (PF_DROP); 4540 } 4541 bzero(s, sizeof(*s)); 4542 s->rule.ptr = r; 4543 s->nat_rule.ptr = nr; 4544 s->anchor.ptr = a; 4545 STATE_INC_COUNTERS(s); 4546 if (r->allow_opts) 4547 s->state_flags |= PFSTATE_ALLOWOPTS; 4548 if (r->rule_flag & PFRULE_STATESLOPPY) 4549 s->state_flags |= PFSTATE_SLOPPY; 4550 s->log = r->log & PF_LOG_ALL; 4551 if (nr != NULL) 4552 s->log |= nr->log & PF_LOG_ALL; 4553 s->proto = pd->proto; 4554 s->direction = direction; 4555 s->af = af; 4556 if (direction == PF_OUT) { 4557 PF_ACPY(&s->gwy.addr, saddr, af); 4558 PF_ACPY(&s->ext.addr, daddr, af); 4559 if (nr != NULL) 4560 PF_ACPY(&s->lan.addr, &pd->baddr, af); 4561 else 4562 PF_ACPY(&s->lan.addr, &s->gwy.addr, af); 4563 } else { 4564 PF_ACPY(&s->lan.addr, daddr, af); 4565 PF_ACPY(&s->ext.addr, saddr, af); 4566 if (nr != NULL) 4567 PF_ACPY(&s->gwy.addr, &pd->baddr, af); 4568 else 4569 PF_ACPY(&s->gwy.addr, &s->lan.addr, af); 4570 } 4571 s->src.state = PFOTHERS_SINGLE; 4572 s->dst.state = PFOTHERS_NO_TRAFFIC; 4573 s->creation = time_second; 4574 s->expire = time_second; 4575 s->timeout = PFTM_OTHER_FIRST_PACKET; 4576 pf_set_rt_ifp(s, saddr); 4577 if (sn != NULL) { 4578 s->src_node = sn; 4579 s->src_node->states++; 4580 } 4581 if (nsn != NULL) { 4582 PF_ACPY(&nsn->raddr, &pd->naddr, af); 4583 s->nat_src_node = nsn; 4584 s->nat_src_node->states++; 4585 } 4586 if (pf_insert_state(BOUND_IFACE(r, kif), s)) { 4587 REASON_SET(&reason, PFRES_STATEINS); 4588 pf_src_tree_remove_state(s); 4589 STATE_DEC_COUNTERS(s); 4590 pool_put(&pf_state_pl, s); 4591 return (PF_DROP); 4592 } else 4593 *sm = s; 4594 if (tag > 0) { 4595 pf_tag_ref(tag); 4596 s->tag = tag; 4597 } 4598 } 4599 4600 return (PF_PASS); 4601} 4602 4603int 4604pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif, 4605 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am, 4606 struct pf_ruleset **rsm) 4607{ 4608 struct pf_rule *r, *a = NULL; 4609 struct pf_ruleset *ruleset = NULL; 4610 sa_family_t af = pd->af; 4611 u_short reason; 4612 int tag = -1; 4613 int asd = 0; 4614 int match = 0; 4615 4616 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 4617 while (r != NULL) { 4618 r->evaluations++; 4619 if (pfi_kif_match(r->kif, kif) == r->ifnot) 4620 r = r->skip[PF_SKIP_IFP].ptr; 4621 else if (r->direction && r->direction != direction) 4622 r = r->skip[PF_SKIP_DIR].ptr; 4623 else if (r->af && r->af != af) 4624 r = r->skip[PF_SKIP_AF].ptr; 4625 else if (r->proto && r->proto != pd->proto) 4626 r = r->skip[PF_SKIP_PROTO].ptr; 4627 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, 4628 r->src.neg, kif)) 4629 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 4630 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, 4631 r->dst.neg, NULL)) 4632 r = r->skip[PF_SKIP_DST_ADDR].ptr; 4633 else if (r->tos && !(r->tos == pd->tos)) 4634 r = TAILQ_NEXT(r, entries); 4635 else if (r->os_fingerprint != PF_OSFP_ANY) 4636 r = TAILQ_NEXT(r, entries); 4637 else if (pd->proto == IPPROTO_UDP && 4638 (r->src.port_op || r->dst.port_op)) 4639 r = TAILQ_NEXT(r, entries); 4640 else if (pd->proto == IPPROTO_TCP && 4641 (r->src.port_op || r->dst.port_op || r->flagset)) 4642 r = TAILQ_NEXT(r, entries); 4643 else if ((pd->proto == IPPROTO_ICMP || 4644 pd->proto == IPPROTO_ICMPV6) && 4645 (r->type || r->code)) 4646 r = TAILQ_NEXT(r, entries); 4647 else if (r->prob && r->prob <= arc4random()) 4648 r = TAILQ_NEXT(r, entries); 4649 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag)) 4650 r = TAILQ_NEXT(r, entries); 4651 else { 4652 if (r->anchor == NULL) { 4653 match = 1; 4654 *rm = r; 4655 *am = a; 4656 *rsm = ruleset; 4657 if ((*rm)->quick) 4658 break; 4659 r = TAILQ_NEXT(r, entries); 4660 } else 4661 pf_step_into_anchor(&asd, &ruleset, 4662 PF_RULESET_FILTER, &r, &a, &match); 4663 } 4664 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 4665 PF_RULESET_FILTER, &r, &a, &match)) 4666 break; 4667 } 4668 r = *rm; 4669 a = *am; 4670 ruleset = *rsm; 4671 4672 REASON_SET(&reason, PFRES_MATCH); 4673 4674 if (r->log) 4675 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset, 4676 pd); 4677 4678 if (r->action != PF_PASS) 4679 return (PF_DROP); 4680 4681 if (pf_tag_packet(m, pd->pf_mtag, tag, -1)) { 4682 REASON_SET(&reason, PFRES_MEMORY); 4683 return (PF_DROP); 4684 } 4685 4686 return (PF_PASS); 4687} 4688 4689int 4690pf_tcp_track_full(struct pf_state_peer *src, struct pf_state_peer *dst, 4691 struct pf_state **state, struct pfi_kif *kif, struct mbuf *m, int off, 4692 struct pf_pdesc *pd, u_short *reason, int *copyback) 4693{ 4694 struct tcphdr *th = pd->hdr.tcp; 4695 u_int16_t win = ntohs(th->th_win); 4696 u_int32_t ack, end, seq, orig_seq; 4697 u_int8_t sws, dws; 4698 int ackskew; 4699 4700 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) { 4701 sws = src->wscale & PF_WSCALE_MASK; 4702 dws = dst->wscale & PF_WSCALE_MASK; 4703 } else 4704 sws = dws = 0; 4705 4706 /* 4707 * Sequence tracking algorithm from Guido van Rooij's paper: 4708 * http://www.madison-gurkha.com/publications/tcp_filtering/ 4709 * tcp_filtering.ps 4710 */ 4711 4712 orig_seq = seq = ntohl(th->th_seq); 4713 if (src->seqlo == 0) { 4714 /* First packet from this end. Set its state */ 4715 4716 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) && 4717 src->scrub == NULL) { 4718 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) { 4719 REASON_SET(reason, PFRES_MEMORY); 4720 return (PF_DROP); 4721 } 4722 } 4723 4724 /* Deferred generation of sequence number modulator */ 4725 if (dst->seqdiff && !src->seqdiff) { 4726#ifdef __FreeBSD__ 4727 while ((src->seqdiff = pf_new_isn(*state) - seq) == 0) 4728 ; 4729#else 4730 while ((src->seqdiff = tcp_rndiss_next() - seq) == 0) 4731 ; 4732#endif 4733 ack = ntohl(th->th_ack) - dst->seqdiff; 4734 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq + 4735 src->seqdiff), 0); 4736 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0); 4737 *copyback = 1; 4738 } else { 4739 ack = ntohl(th->th_ack); 4740 } 4741 4742 end = seq + pd->p_len; 4743 if (th->th_flags & TH_SYN) { 4744 end++; 4745 if (dst->wscale & PF_WSCALE_FLAG) { 4746 src->wscale = pf_get_wscale(m, off, th->th_off, 4747 pd->af); 4748 if (src->wscale & PF_WSCALE_FLAG) { 4749 /* Remove scale factor from initial 4750 * window */ 4751 sws = src->wscale & PF_WSCALE_MASK; 4752 win = ((u_int32_t)win + (1 << sws) - 1) 4753 >> sws; 4754 dws = dst->wscale & PF_WSCALE_MASK; 4755 } else { 4756 /* fixup other window */ 4757 dst->max_win <<= dst->wscale & 4758 PF_WSCALE_MASK; 4759 /* in case of a retrans SYN|ACK */ 4760 dst->wscale = 0; 4761 } 4762 } 4763 } 4764 if (th->th_flags & TH_FIN) 4765 end++; 4766 4767 src->seqlo = seq; 4768 if (src->state < TCPS_SYN_SENT) 4769 src->state = TCPS_SYN_SENT; 4770 4771 /* 4772 * May need to slide the window (seqhi may have been set by 4773 * the crappy stack check or if we picked up the connection 4774 * after establishment) 4775 */ 4776 if (src->seqhi == 1 || 4777 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi)) 4778 src->seqhi = end + MAX(1, dst->max_win << dws); 4779 if (win > src->max_win) 4780 src->max_win = win; 4781 4782 } else { 4783 ack = ntohl(th->th_ack) - dst->seqdiff; 4784 if (src->seqdiff) { 4785 /* Modulate sequence numbers */ 4786 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq + 4787 src->seqdiff), 0); 4788 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0); 4789 *copyback = 1; 4790 } 4791 end = seq + pd->p_len; 4792 if (th->th_flags & TH_SYN) 4793 end++; 4794 if (th->th_flags & TH_FIN) 4795 end++; 4796 } 4797 4798 if ((th->th_flags & TH_ACK) == 0) { 4799 /* Let it pass through the ack skew check */ 4800 ack = dst->seqlo; 4801 } else if ((ack == 0 && 4802 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) || 4803 /* broken tcp stacks do not set ack */ 4804 (dst->state < TCPS_SYN_SENT)) { 4805 /* 4806 * Many stacks (ours included) will set the ACK number in an 4807 * FIN|ACK if the SYN times out -- no sequence to ACK. 4808 */ 4809 ack = dst->seqlo; 4810 } 4811 4812 if (seq == end) { 4813 /* Ease sequencing restrictions on no data packets */ 4814 seq = src->seqlo; 4815 end = seq; 4816 } 4817 4818 ackskew = dst->seqlo - ack; 4819 4820 4821 /* 4822 * Need to demodulate the sequence numbers in any TCP SACK options 4823 * (Selective ACK). We could optionally validate the SACK values 4824 * against the current ACK window, either forwards or backwards, but 4825 * I'm not confident that SACK has been implemented properly 4826 * everywhere. It wouldn't surprise me if several stacks accidently 4827 * SACK too far backwards of previously ACKed data. There really aren't 4828 * any security implications of bad SACKing unless the target stack 4829 * doesn't validate the option length correctly. Someone trying to 4830 * spoof into a TCP connection won't bother blindly sending SACK 4831 * options anyway. 4832 */ 4833 if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) { 4834 if (pf_modulate_sack(m, off, pd, th, dst)) 4835 *copyback = 1; 4836 } 4837 4838 4839#define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */ 4840 if (SEQ_GEQ(src->seqhi, end) && 4841 /* Last octet inside other's window space */ 4842 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) && 4843 /* Retrans: not more than one window back */ 4844 (ackskew >= -MAXACKWINDOW) && 4845 /* Acking not more than one reassembled fragment backwards */ 4846 (ackskew <= (MAXACKWINDOW << sws)) && 4847 /* Acking not more than one window forward */ 4848 ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo || 4849 (orig_seq == src->seqlo + 1) || (pd->flags & PFDESC_IP_REAS) == 0)) { 4850 /* Require an exact/+1 sequence match on resets when possible */ 4851 4852 if (dst->scrub || src->scrub) { 4853 if (pf_normalize_tcp_stateful(m, off, pd, reason, th, 4854 *state, src, dst, copyback)) 4855 return (PF_DROP); 4856 } 4857 4858 /* update max window */ 4859 if (src->max_win < win) 4860 src->max_win = win; 4861 /* synchronize sequencing */ 4862 if (SEQ_GT(end, src->seqlo)) 4863 src->seqlo = end; 4864 /* slide the window of what the other end can send */ 4865 if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) 4866 dst->seqhi = ack + MAX((win << sws), 1); 4867 4868 4869 /* update states */ 4870 if (th->th_flags & TH_SYN) 4871 if (src->state < TCPS_SYN_SENT) 4872 src->state = TCPS_SYN_SENT; 4873 if (th->th_flags & TH_FIN) 4874 if (src->state < TCPS_CLOSING) 4875 src->state = TCPS_CLOSING; 4876 if (th->th_flags & TH_ACK) { 4877 if (dst->state == TCPS_SYN_SENT) { 4878 dst->state = TCPS_ESTABLISHED; 4879 if (src->state == TCPS_ESTABLISHED && 4880 (*state)->src_node != NULL && 4881 pf_src_connlimit(state)) { 4882 REASON_SET(reason, PFRES_SRCLIMIT); 4883 return (PF_DROP); 4884 } 4885 } else if (dst->state == TCPS_CLOSING) 4886 dst->state = TCPS_FIN_WAIT_2; 4887 } 4888 if (th->th_flags & TH_RST) 4889 src->state = dst->state = TCPS_TIME_WAIT; 4890 4891 /* update expire time */ 4892 (*state)->expire = time_second; 4893 if (src->state >= TCPS_FIN_WAIT_2 && 4894 dst->state >= TCPS_FIN_WAIT_2) 4895 (*state)->timeout = PFTM_TCP_CLOSED; 4896 else if (src->state >= TCPS_CLOSING && 4897 dst->state >= TCPS_CLOSING) 4898 (*state)->timeout = PFTM_TCP_FIN_WAIT; 4899 else if (src->state < TCPS_ESTABLISHED || 4900 dst->state < TCPS_ESTABLISHED) 4901 (*state)->timeout = PFTM_TCP_OPENING; 4902 else if (src->state >= TCPS_CLOSING || 4903 dst->state >= TCPS_CLOSING) 4904 (*state)->timeout = PFTM_TCP_CLOSING; 4905 else 4906 (*state)->timeout = PFTM_TCP_ESTABLISHED; 4907 4908 /* Fall through to PASS packet */ 4909 4910 } else if ((dst->state < TCPS_SYN_SENT || 4911 dst->state >= TCPS_FIN_WAIT_2 || 4912 src->state >= TCPS_FIN_WAIT_2) && 4913 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) && 4914 /* Within a window forward of the originating packet */ 4915 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) { 4916 /* Within a window backward of the originating packet */ 4917 4918 /* 4919 * This currently handles three situations: 4920 * 1) Stupid stacks will shotgun SYNs before their peer 4921 * replies. 4922 * 2) When PF catches an already established stream (the 4923 * firewall rebooted, the state table was flushed, routes 4924 * changed...) 4925 * 3) Packets get funky immediately after the connection 4926 * closes (this should catch Solaris spurious ACK|FINs 4927 * that web servers like to spew after a close) 4928 * 4929 * This must be a little more careful than the above code 4930 * since packet floods will also be caught here. We don't 4931 * update the TTL here to mitigate the damage of a packet 4932 * flood and so the same code can handle awkward establishment 4933 * and a loosened connection close. 4934 * In the establishment case, a correct peer response will 4935 * validate the connection, go through the normal state code 4936 * and keep updating the state TTL. 4937 */ 4938 4939 if (pf_status.debug >= PF_DEBUG_MISC) { 4940 printf("pf: loose state match: "); 4941 pf_print_state(*state); 4942 pf_print_flags(th->th_flags); 4943 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d " 4944 "pkts=%llu:%llu\n", seq, orig_seq, ack, pd->p_len, 4945#ifdef __FreeBSD__ 4946 ackskew, (unsigned long long)(*state)->packets[0], 4947 (unsigned long long)(*state)->packets[1]); 4948#else 4949 ackskew, (*state)->packets[0], 4950 (*state)->packets[1]); 4951#endif 4952 } 4953 4954 if (dst->scrub || src->scrub) { 4955 if (pf_normalize_tcp_stateful(m, off, pd, reason, th, 4956 *state, src, dst, copyback)) 4957 return (PF_DROP); 4958 } 4959 4960 /* update max window */ 4961 if (src->max_win < win) 4962 src->max_win = win; 4963 /* synchronize sequencing */ 4964 if (SEQ_GT(end, src->seqlo)) 4965 src->seqlo = end; 4966 /* slide the window of what the other end can send */ 4967 if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) 4968 dst->seqhi = ack + MAX((win << sws), 1); 4969 4970 /* 4971 * Cannot set dst->seqhi here since this could be a shotgunned 4972 * SYN and not an already established connection. 4973 */ 4974 4975 if (th->th_flags & TH_FIN) 4976 if (src->state < TCPS_CLOSING) 4977 src->state = TCPS_CLOSING; 4978 if (th->th_flags & TH_RST) 4979 src->state = dst->state = TCPS_TIME_WAIT; 4980 4981 /* Fall through to PASS packet */ 4982 4983 } else { 4984 if ((*state)->dst.state == TCPS_SYN_SENT && 4985 (*state)->src.state == TCPS_SYN_SENT) { 4986 /* Send RST for state mismatches during handshake */ 4987 if (!(th->th_flags & TH_RST)) 4988#ifdef __FreeBSD__ 4989 pf_send_tcp(m, (*state)->rule.ptr, pd->af, 4990#else 4991 pf_send_tcp((*state)->rule.ptr, pd->af, 4992#endif 4993 pd->dst, pd->src, th->th_dport, 4994 th->th_sport, ntohl(th->th_ack), 0, 4995 TH_RST, 0, 0, 4996 (*state)->rule.ptr->return_ttl, 1, 0, 4997 pd->eh, kif->pfik_ifp); 4998 src->seqlo = 0; 4999 src->seqhi = 1; 5000 src->max_win = 1; 5001 } else if (pf_status.debug >= PF_DEBUG_MISC) { 5002 printf("pf: BAD state: "); 5003 pf_print_state(*state); 5004 pf_print_flags(th->th_flags); 5005 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d " 5006#ifdef notyet 5007 "pkts=%llu:%llu dir=%s,%s\n", 5008#else 5009 "pkts=%llu:%llu%s\n", 5010#endif 5011 seq, orig_seq, ack, pd->p_len, ackskew, 5012#ifdef __FreeBSD__ 5013 (unsigned long long)(*state)->packets[0], 5014 (unsigned long long)(*state)->packets[1], 5015#else 5016 (*state)->packets[0], (*state)->packets[1], 5017#endif 5018#ifdef notyet 5019 direction == PF_IN ? "in" : "out", 5020 direction == (*state)->direction ? "fwd" : "rev"); 5021#else 5022 ""); 5023#endif 5024 printf("pf: State failure on: %c %c %c %c | %c %c\n", 5025 SEQ_GEQ(src->seqhi, end) ? ' ' : '1', 5026 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ? 5027 ' ': '2', 5028 (ackskew >= -MAXACKWINDOW) ? ' ' : '3', 5029 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4', 5030 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5', 5031 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6'); 5032 } 5033 REASON_SET(reason, PFRES_BADSTATE); 5034 return (PF_DROP); 5035 } 5036 5037 /* Any packets which have gotten here are to be passed */ 5038 return (PF_PASS); 5039} 5040 5041int 5042pf_tcp_track_sloppy(struct pf_state_peer *src, struct pf_state_peer *dst, 5043 struct pf_state **state, struct pf_pdesc *pd, u_short *reason) 5044{ 5045 struct tcphdr *th = pd->hdr.tcp; 5046 5047 if (th->th_flags & TH_SYN) 5048 if (src->state < TCPS_SYN_SENT) 5049 src->state = TCPS_SYN_SENT; 5050 if (th->th_flags & TH_FIN) 5051 if (src->state < TCPS_CLOSING) 5052 src->state = TCPS_CLOSING; 5053 if (th->th_flags & TH_ACK) { 5054 if (dst->state == TCPS_SYN_SENT) { 5055 dst->state = TCPS_ESTABLISHED; 5056 if (src->state == TCPS_ESTABLISHED && 5057 (*state)->src_node != NULL && 5058 pf_src_connlimit(state)) { 5059 REASON_SET(reason, PFRES_SRCLIMIT); 5060 return (PF_DROP); 5061 } 5062 } else if (dst->state == TCPS_CLOSING) { 5063 dst->state = TCPS_FIN_WAIT_2; 5064 } else if (src->state == TCPS_SYN_SENT && 5065 dst->state < TCPS_SYN_SENT) { 5066 /* 5067 * Handle a special sloppy case where we only see one 5068 * half of the connection. If there is a ACK after 5069 * the initial SYN without ever seeing a packet from 5070 * the destination, set the connection to established. 5071 */ 5072 dst->state = src->state = TCPS_ESTABLISHED; 5073 if ((*state)->src_node != NULL && 5074 pf_src_connlimit(state)) { 5075 REASON_SET(reason, PFRES_SRCLIMIT); 5076 return (PF_DROP); 5077 } 5078 } else if (src->state == TCPS_CLOSING && 5079 dst->state == TCPS_ESTABLISHED && 5080 dst->seqlo == 0) { 5081 /* 5082 * Handle the closing of half connections where we 5083 * don't see the full bidirectional FIN/ACK+ACK 5084 * handshake. 5085 */ 5086 dst->state = TCPS_CLOSING; 5087 } 5088 } 5089 if (th->th_flags & TH_RST) 5090 src->state = dst->state = TCPS_TIME_WAIT; 5091 5092 /* update expire time */ 5093 (*state)->expire = time_second; 5094 if (src->state >= TCPS_FIN_WAIT_2 && 5095 dst->state >= TCPS_FIN_WAIT_2) 5096 (*state)->timeout = PFTM_TCP_CLOSED; 5097 else if (src->state >= TCPS_CLOSING && 5098 dst->state >= TCPS_CLOSING) 5099 (*state)->timeout = PFTM_TCP_FIN_WAIT; 5100 else if (src->state < TCPS_ESTABLISHED || 5101 dst->state < TCPS_ESTABLISHED) 5102 (*state)->timeout = PFTM_TCP_OPENING; 5103 else if (src->state >= TCPS_CLOSING || 5104 dst->state >= TCPS_CLOSING) 5105 (*state)->timeout = PFTM_TCP_CLOSING; 5106 else 5107 (*state)->timeout = PFTM_TCP_ESTABLISHED; 5108 5109 return (PF_PASS); 5110} 5111 5112 5113int 5114pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif, 5115 struct mbuf *m, int off, void *h, struct pf_pdesc *pd, 5116 u_short *reason) 5117{ 5118 struct pf_state_cmp key; 5119 struct tcphdr *th = pd->hdr.tcp; 5120 int copyback = 0; 5121 struct pf_state_peer *src, *dst; 5122 5123 key.af = pd->af; 5124 key.proto = IPPROTO_TCP; 5125 if (direction == PF_IN) { 5126 PF_ACPY(&key.ext.addr, pd->src, key.af); 5127 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 5128 key.ext.port = th->th_sport; 5129 key.gwy.port = th->th_dport; 5130 } else { 5131 PF_ACPY(&key.lan.addr, pd->src, key.af); 5132 PF_ACPY(&key.ext.addr, pd->dst, key.af); 5133 key.lan.port = th->th_sport; 5134 key.ext.port = th->th_dport; 5135 } 5136 5137 STATE_LOOKUP(); 5138 5139 if (direction == (*state)->direction) { 5140 src = &(*state)->src; 5141 dst = &(*state)->dst; 5142 } else { 5143 src = &(*state)->dst; 5144 dst = &(*state)->src; 5145 } 5146 5147 if ((*state)->src.state == PF_TCPS_PROXY_SRC) { 5148 if (direction != (*state)->direction) { 5149 REASON_SET(reason, PFRES_SYNPROXY); 5150 return (PF_SYNPROXY_DROP); 5151 } 5152 if (th->th_flags & TH_SYN) { 5153 if (ntohl(th->th_seq) != (*state)->src.seqlo) { 5154 REASON_SET(reason, PFRES_SYNPROXY); 5155 return (PF_DROP); 5156 } 5157#ifdef __FreeBSD__ 5158 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst, 5159#else 5160 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, 5161#endif 5162 pd->src, th->th_dport, th->th_sport, 5163 (*state)->src.seqhi, ntohl(th->th_seq) + 1, 5164 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1, 5165 0, NULL, NULL); 5166 REASON_SET(reason, PFRES_SYNPROXY); 5167 return (PF_SYNPROXY_DROP); 5168 } else if (!(th->th_flags & TH_ACK) || 5169 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) || 5170 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) { 5171 REASON_SET(reason, PFRES_SYNPROXY); 5172 return (PF_DROP); 5173 } else if ((*state)->src_node != NULL && 5174 pf_src_connlimit(state)) { 5175 REASON_SET(reason, PFRES_SRCLIMIT); 5176 return (PF_DROP); 5177 } else 5178 (*state)->src.state = PF_TCPS_PROXY_DST; 5179 } 5180 if ((*state)->src.state == PF_TCPS_PROXY_DST) { 5181 struct pf_state_host *src, *dst; 5182 5183 if (direction == PF_OUT) { 5184 src = &(*state)->gwy; 5185 dst = &(*state)->ext; 5186 } else { 5187 src = &(*state)->ext; 5188 dst = &(*state)->lan; 5189 } 5190 if (direction == (*state)->direction) { 5191 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) || 5192 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) || 5193 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) { 5194 REASON_SET(reason, PFRES_SYNPROXY); 5195 return (PF_DROP); 5196 } 5197 (*state)->src.max_win = MAX(ntohs(th->th_win), 1); 5198 if ((*state)->dst.seqhi == 1) 5199 (*state)->dst.seqhi = htonl(arc4random()); 5200#ifdef __FreeBSD__ 5201 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, 5202 &src->addr, 5203#else 5204 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr, 5205#endif 5206 &dst->addr, src->port, dst->port, 5207 (*state)->dst.seqhi, 0, TH_SYN, 0, 5208 (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL); 5209 REASON_SET(reason, PFRES_SYNPROXY); 5210 return (PF_SYNPROXY_DROP); 5211 } else if (((th->th_flags & (TH_SYN|TH_ACK)) != 5212 (TH_SYN|TH_ACK)) || 5213 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) { 5214 REASON_SET(reason, PFRES_SYNPROXY); 5215 return (PF_DROP); 5216 } else { 5217 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1); 5218 (*state)->dst.seqlo = ntohl(th->th_seq); 5219#ifdef __FreeBSD__ 5220 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst, 5221#else 5222 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, 5223#endif 5224 pd->src, th->th_dport, th->th_sport, 5225 ntohl(th->th_ack), ntohl(th->th_seq) + 1, 5226 TH_ACK, (*state)->src.max_win, 0, 0, 0, 5227 (*state)->tag, NULL, NULL); 5228#ifdef __FreeBSD__ 5229 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, 5230 &src->addr, 5231#else 5232 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr, 5233#endif 5234 &dst->addr, src->port, dst->port, 5235 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1, 5236 TH_ACK, (*state)->dst.max_win, 0, 0, 1, 5237 0, NULL, NULL); 5238 (*state)->src.seqdiff = (*state)->dst.seqhi - 5239 (*state)->src.seqlo; 5240 (*state)->dst.seqdiff = (*state)->src.seqhi - 5241 (*state)->dst.seqlo; 5242 (*state)->src.seqhi = (*state)->src.seqlo + 5243 (*state)->dst.max_win; 5244 (*state)->dst.seqhi = (*state)->dst.seqlo + 5245 (*state)->src.max_win; 5246 (*state)->src.wscale = (*state)->dst.wscale = 0; 5247 (*state)->src.state = (*state)->dst.state = 5248 TCPS_ESTABLISHED; 5249 REASON_SET(reason, PFRES_SYNPROXY); 5250 return (PF_SYNPROXY_DROP); 5251 } 5252 } 5253 5254 if (((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN) && 5255 dst->state >= TCPS_FIN_WAIT_2 && 5256 src->state >= TCPS_FIN_WAIT_2) { 5257 if (pf_status.debug >= PF_DEBUG_MISC) { 5258 printf("pf: state reuse "); 5259 pf_print_state(*state); 5260 pf_print_flags(th->th_flags); 5261 printf("\n"); 5262 } 5263 /* XXX make sure it's the same direction ?? */ 5264 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED; 5265 pf_unlink_state(*state); 5266 *state = NULL; 5267 return (PF_DROP); 5268 } 5269 5270 if ((*state)->state_flags & PFSTATE_SLOPPY) { 5271 if (pf_tcp_track_sloppy(src, dst, state, pd, reason) == PF_DROP) 5272 return (PF_DROP); 5273 } else { 5274 if (pf_tcp_track_full(src, dst, state, kif, m, off, pd, reason, 5275 ©back) == PF_DROP) 5276 return (PF_DROP); 5277 } 5278 5279 /* translate source/destination address, if necessary */ 5280 if (STATE_TRANSLATE(*state)) { 5281 if (direction == PF_OUT) 5282 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum, 5283 &th->th_sum, &(*state)->gwy.addr, 5284 (*state)->gwy.port, 0, pd->af); 5285 else 5286 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum, 5287 &th->th_sum, &(*state)->lan.addr, 5288 (*state)->lan.port, 0, pd->af); 5289 m_copyback(m, off, sizeof(*th), (caddr_t)th); 5290 } else if (copyback) { 5291 /* Copyback sequence modulation or stateful scrub changes */ 5292 m_copyback(m, off, sizeof(*th), (caddr_t)th); 5293 } 5294 5295 return (PF_PASS); 5296} 5297 5298int 5299pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif, 5300 struct mbuf *m, int off, void *h, struct pf_pdesc *pd) 5301{ 5302 struct pf_state_peer *src, *dst; 5303 struct pf_state_cmp key; 5304 struct udphdr *uh = pd->hdr.udp; 5305 5306 key.af = pd->af; 5307 key.proto = IPPROTO_UDP; 5308 if (direction == PF_IN) { 5309 PF_ACPY(&key.ext.addr, pd->src, key.af); 5310 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 5311 key.ext.port = uh->uh_sport; 5312 key.gwy.port = uh->uh_dport; 5313 } else { 5314 PF_ACPY(&key.lan.addr, pd->src, key.af); 5315 PF_ACPY(&key.ext.addr, pd->dst, key.af); 5316 key.lan.port = uh->uh_sport; 5317 key.ext.port = uh->uh_dport; 5318 } 5319 5320 STATE_LOOKUP(); 5321 5322 if (direction == (*state)->direction) { 5323 src = &(*state)->src; 5324 dst = &(*state)->dst; 5325 } else { 5326 src = &(*state)->dst; 5327 dst = &(*state)->src; 5328 } 5329 5330 /* update states */ 5331 if (src->state < PFUDPS_SINGLE) 5332 src->state = PFUDPS_SINGLE; 5333 if (dst->state == PFUDPS_SINGLE) 5334 dst->state = PFUDPS_MULTIPLE; 5335 5336 /* update expire time */ 5337 (*state)->expire = time_second; 5338 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE) 5339 (*state)->timeout = PFTM_UDP_MULTIPLE; 5340 else 5341 (*state)->timeout = PFTM_UDP_SINGLE; 5342 5343 /* translate source/destination address, if necessary */ 5344 if (STATE_TRANSLATE(*state)) { 5345 if (direction == PF_OUT) 5346 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum, 5347 &uh->uh_sum, &(*state)->gwy.addr, 5348 (*state)->gwy.port, 1, pd->af); 5349 else 5350 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum, 5351 &uh->uh_sum, &(*state)->lan.addr, 5352 (*state)->lan.port, 1, pd->af); 5353 m_copyback(m, off, sizeof(*uh), (caddr_t)uh); 5354 } 5355 5356 return (PF_PASS); 5357} 5358 5359int 5360pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif, 5361 struct mbuf *m, int off, void *h, struct pf_pdesc *pd, u_short *reason) 5362{ 5363 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 5364 u_int16_t icmpid = 0; /* make the compiler happy */ 5365 u_int16_t *icmpsum = NULL; /* make the compiler happy */ 5366 u_int8_t icmptype = 0; /* make the compiler happy */ 5367 int state_icmp = 0; 5368 struct pf_state_cmp key; 5369 5370 switch (pd->proto) { 5371#ifdef INET 5372 case IPPROTO_ICMP: 5373 icmptype = pd->hdr.icmp->icmp_type; 5374 icmpid = pd->hdr.icmp->icmp_id; 5375 icmpsum = &pd->hdr.icmp->icmp_cksum; 5376 5377 if (icmptype == ICMP_UNREACH || 5378 icmptype == ICMP_SOURCEQUENCH || 5379 icmptype == ICMP_REDIRECT || 5380 icmptype == ICMP_TIMXCEED || 5381 icmptype == ICMP_PARAMPROB) 5382 state_icmp++; 5383 break; 5384#endif /* INET */ 5385#ifdef INET6 5386 case IPPROTO_ICMPV6: 5387 icmptype = pd->hdr.icmp6->icmp6_type; 5388 icmpid = pd->hdr.icmp6->icmp6_id; 5389 icmpsum = &pd->hdr.icmp6->icmp6_cksum; 5390 5391 if (icmptype == ICMP6_DST_UNREACH || 5392 icmptype == ICMP6_PACKET_TOO_BIG || 5393 icmptype == ICMP6_TIME_EXCEEDED || 5394 icmptype == ICMP6_PARAM_PROB) 5395 state_icmp++; 5396 break; 5397#endif /* INET6 */ 5398 } 5399 5400 if (!state_icmp) { 5401 5402 /* 5403 * ICMP query/reply message not related to a TCP/UDP packet. 5404 * Search for an ICMP state. 5405 */ 5406 key.af = pd->af; 5407 key.proto = pd->proto; 5408 if (direction == PF_IN) { 5409 PF_ACPY(&key.ext.addr, pd->src, key.af); 5410 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 5411 key.ext.port = 0; 5412 key.gwy.port = icmpid; 5413 } else { 5414 PF_ACPY(&key.lan.addr, pd->src, key.af); 5415 PF_ACPY(&key.ext.addr, pd->dst, key.af); 5416 key.lan.port = icmpid; 5417 key.ext.port = 0; 5418 } 5419 5420 STATE_LOOKUP(); 5421 5422 (*state)->expire = time_second; 5423 (*state)->timeout = PFTM_ICMP_ERROR_REPLY; 5424 5425 /* translate source/destination address, if necessary */ 5426 if (STATE_TRANSLATE(*state)) { 5427 if (direction == PF_OUT) { 5428 switch (pd->af) { 5429#ifdef INET 5430 case AF_INET: 5431 pf_change_a(&saddr->v4.s_addr, 5432 pd->ip_sum, 5433 (*state)->gwy.addr.v4.s_addr, 0); 5434 pd->hdr.icmp->icmp_cksum = 5435 pf_cksum_fixup( 5436 pd->hdr.icmp->icmp_cksum, icmpid, 5437 (*state)->gwy.port, 0); 5438 pd->hdr.icmp->icmp_id = 5439 (*state)->gwy.port; 5440 m_copyback(m, off, ICMP_MINLEN, 5441 (caddr_t)pd->hdr.icmp); 5442 break; 5443#endif /* INET */ 5444#ifdef INET6 5445 case AF_INET6: 5446 pf_change_a6(saddr, 5447 &pd->hdr.icmp6->icmp6_cksum, 5448 &(*state)->gwy.addr, 0); 5449 m_copyback(m, off, 5450 sizeof(struct icmp6_hdr), 5451 (caddr_t)pd->hdr.icmp6); 5452 break; 5453#endif /* INET6 */ 5454 } 5455 } else { 5456 switch (pd->af) { 5457#ifdef INET 5458 case AF_INET: 5459 pf_change_a(&daddr->v4.s_addr, 5460 pd->ip_sum, 5461 (*state)->lan.addr.v4.s_addr, 0); 5462 pd->hdr.icmp->icmp_cksum = 5463 pf_cksum_fixup( 5464 pd->hdr.icmp->icmp_cksum, icmpid, 5465 (*state)->lan.port, 0); 5466 pd->hdr.icmp->icmp_id = 5467 (*state)->lan.port; 5468 m_copyback(m, off, ICMP_MINLEN, 5469 (caddr_t)pd->hdr.icmp); 5470 break; 5471#endif /* INET */ 5472#ifdef INET6 5473 case AF_INET6: 5474 pf_change_a6(daddr, 5475 &pd->hdr.icmp6->icmp6_cksum, 5476 &(*state)->lan.addr, 0); 5477 m_copyback(m, off, 5478 sizeof(struct icmp6_hdr), 5479 (caddr_t)pd->hdr.icmp6); 5480 break; 5481#endif /* INET6 */ 5482 } 5483 } 5484 } 5485 5486 return (PF_PASS); 5487 5488 } else { 5489 /* 5490 * ICMP error message in response to a TCP/UDP packet. 5491 * Extract the inner TCP/UDP header and search for that state. 5492 */ 5493 5494 struct pf_pdesc pd2; 5495#ifdef INET 5496 struct ip h2; 5497#endif /* INET */ 5498#ifdef INET6 5499 struct ip6_hdr h2_6; 5500 int terminal = 0; 5501#endif /* INET6 */ 5502 int ipoff2 = 0; /* make the compiler happy */ 5503 int off2 = 0; /* make the compiler happy */ 5504 5505 pd2.af = pd->af; 5506 switch (pd->af) { 5507#ifdef INET 5508 case AF_INET: 5509 /* offset of h2 in mbuf chain */ 5510 ipoff2 = off + ICMP_MINLEN; 5511 5512 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2), 5513 NULL, reason, pd2.af)) { 5514 DPFPRINTF(PF_DEBUG_MISC, 5515 ("pf: ICMP error message too short " 5516 "(ip)\n")); 5517 return (PF_DROP); 5518 } 5519 /* 5520 * ICMP error messages don't refer to non-first 5521 * fragments 5522 */ 5523 if (h2.ip_off & htons(IP_OFFMASK)) { 5524 REASON_SET(reason, PFRES_FRAG); 5525 return (PF_DROP); 5526 } 5527 5528 /* offset of protocol header that follows h2 */ 5529 off2 = ipoff2 + (h2.ip_hl << 2); 5530 5531 pd2.proto = h2.ip_p; 5532 pd2.src = (struct pf_addr *)&h2.ip_src; 5533 pd2.dst = (struct pf_addr *)&h2.ip_dst; 5534 pd2.ip_sum = &h2.ip_sum; 5535 break; 5536#endif /* INET */ 5537#ifdef INET6 5538 case AF_INET6: 5539 ipoff2 = off + sizeof(struct icmp6_hdr); 5540 5541 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6), 5542 NULL, reason, pd2.af)) { 5543 DPFPRINTF(PF_DEBUG_MISC, 5544 ("pf: ICMP error message too short " 5545 "(ip6)\n")); 5546 return (PF_DROP); 5547 } 5548 pd2.proto = h2_6.ip6_nxt; 5549 pd2.src = (struct pf_addr *)&h2_6.ip6_src; 5550 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst; 5551 pd2.ip_sum = NULL; 5552 off2 = ipoff2 + sizeof(h2_6); 5553 do { 5554 switch (pd2.proto) { 5555 case IPPROTO_FRAGMENT: 5556 /* 5557 * ICMPv6 error messages for 5558 * non-first fragments 5559 */ 5560 REASON_SET(reason, PFRES_FRAG); 5561 return (PF_DROP); 5562 case IPPROTO_AH: 5563 case IPPROTO_HOPOPTS: 5564 case IPPROTO_ROUTING: 5565 case IPPROTO_DSTOPTS: { 5566 /* get next header and header length */ 5567 struct ip6_ext opt6; 5568 5569 if (!pf_pull_hdr(m, off2, &opt6, 5570 sizeof(opt6), NULL, reason, 5571 pd2.af)) { 5572 DPFPRINTF(PF_DEBUG_MISC, 5573 ("pf: ICMPv6 short opt\n")); 5574 return (PF_DROP); 5575 } 5576 if (pd2.proto == IPPROTO_AH) 5577 off2 += (opt6.ip6e_len + 2) * 4; 5578 else 5579 off2 += (opt6.ip6e_len + 1) * 8; 5580 pd2.proto = opt6.ip6e_nxt; 5581 /* goto the next header */ 5582 break; 5583 } 5584 default: 5585 terminal++; 5586 break; 5587 } 5588 } while (!terminal); 5589 break; 5590#endif /* INET6 */ 5591#ifdef __FreeBSD__ 5592 default: 5593 panic("AF not supported: %d", pd->af); 5594#endif 5595 } 5596 5597 switch (pd2.proto) { 5598 case IPPROTO_TCP: { 5599 struct tcphdr th; 5600 u_int32_t seq; 5601 struct pf_state_peer *src, *dst; 5602 u_int8_t dws; 5603 int copyback = 0; 5604 5605 /* 5606 * Only the first 8 bytes of the TCP header can be 5607 * expected. Don't access any TCP header fields after 5608 * th_seq, an ackskew test is not possible. 5609 */ 5610 if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason, 5611 pd2.af)) { 5612 DPFPRINTF(PF_DEBUG_MISC, 5613 ("pf: ICMP error message too short " 5614 "(tcp)\n")); 5615 return (PF_DROP); 5616 } 5617 5618 key.af = pd2.af; 5619 key.proto = IPPROTO_TCP; 5620 if (direction == PF_IN) { 5621 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 5622 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 5623 key.ext.port = th.th_dport; 5624 key.gwy.port = th.th_sport; 5625 } else { 5626 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 5627 PF_ACPY(&key.ext.addr, pd2.src, key.af); 5628 key.lan.port = th.th_dport; 5629 key.ext.port = th.th_sport; 5630 } 5631 5632 STATE_LOOKUP(); 5633 5634 if (direction == (*state)->direction) { 5635 src = &(*state)->dst; 5636 dst = &(*state)->src; 5637 } else { 5638 src = &(*state)->src; 5639 dst = &(*state)->dst; 5640 } 5641 5642 if (src->wscale && dst->wscale) 5643 dws = dst->wscale & PF_WSCALE_MASK; 5644 else 5645 dws = 0; 5646 5647 /* Demodulate sequence number */ 5648 seq = ntohl(th.th_seq) - src->seqdiff; 5649 if (src->seqdiff) { 5650 pf_change_a(&th.th_seq, icmpsum, 5651 htonl(seq), 0); 5652 copyback = 1; 5653 } 5654 5655 if (!((*state)->state_flags & PFSTATE_SLOPPY) && 5656 (!SEQ_GEQ(src->seqhi, seq) || 5657 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)))) { 5658 if (pf_status.debug >= PF_DEBUG_MISC) { 5659 printf("pf: BAD ICMP %d:%d ", 5660 icmptype, pd->hdr.icmp->icmp_code); 5661 pf_print_host(pd->src, 0, pd->af); 5662 printf(" -> "); 5663 pf_print_host(pd->dst, 0, pd->af); 5664 printf(" state: "); 5665 pf_print_state(*state); 5666 printf(" seq=%u\n", seq); 5667 } 5668 REASON_SET(reason, PFRES_BADSTATE); 5669 return (PF_DROP); 5670 } 5671 5672 if (STATE_TRANSLATE(*state)) { 5673 if (direction == PF_IN) { 5674 pf_change_icmp(pd2.src, &th.th_sport, 5675 daddr, &(*state)->lan.addr, 5676 (*state)->lan.port, NULL, 5677 pd2.ip_sum, icmpsum, 5678 pd->ip_sum, 0, pd2.af); 5679 } else { 5680 pf_change_icmp(pd2.dst, &th.th_dport, 5681 saddr, &(*state)->gwy.addr, 5682 (*state)->gwy.port, NULL, 5683 pd2.ip_sum, icmpsum, 5684 pd->ip_sum, 0, pd2.af); 5685 } 5686 copyback = 1; 5687 } 5688 5689 if (copyback) { 5690 switch (pd2.af) { 5691#ifdef INET 5692 case AF_INET: 5693 m_copyback(m, off, ICMP_MINLEN, 5694 (caddr_t)pd->hdr.icmp); 5695 m_copyback(m, ipoff2, sizeof(h2), 5696 (caddr_t)&h2); 5697 break; 5698#endif /* INET */ 5699#ifdef INET6 5700 case AF_INET6: 5701 m_copyback(m, off, 5702 sizeof(struct icmp6_hdr), 5703 (caddr_t)pd->hdr.icmp6); 5704 m_copyback(m, ipoff2, sizeof(h2_6), 5705 (caddr_t)&h2_6); 5706 break; 5707#endif /* INET6 */ 5708 } 5709 m_copyback(m, off2, 8, (caddr_t)&th); 5710 } 5711 5712 return (PF_PASS); 5713 break; 5714 } 5715 case IPPROTO_UDP: { 5716 struct udphdr uh; 5717 5718 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh), 5719 NULL, reason, pd2.af)) { 5720 DPFPRINTF(PF_DEBUG_MISC, 5721 ("pf: ICMP error message too short " 5722 "(udp)\n")); 5723 return (PF_DROP); 5724 } 5725 5726 key.af = pd2.af; 5727 key.proto = IPPROTO_UDP; 5728 if (direction == PF_IN) { 5729 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 5730 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 5731 key.ext.port = uh.uh_dport; 5732 key.gwy.port = uh.uh_sport; 5733 } else { 5734 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 5735 PF_ACPY(&key.ext.addr, pd2.src, key.af); 5736 key.lan.port = uh.uh_dport; 5737 key.ext.port = uh.uh_sport; 5738 } 5739 5740 STATE_LOOKUP(); 5741 5742 if (STATE_TRANSLATE(*state)) { 5743 if (direction == PF_IN) { 5744 pf_change_icmp(pd2.src, &uh.uh_sport, 5745 daddr, &(*state)->lan.addr, 5746 (*state)->lan.port, &uh.uh_sum, 5747 pd2.ip_sum, icmpsum, 5748 pd->ip_sum, 1, pd2.af); 5749 } else { 5750 pf_change_icmp(pd2.dst, &uh.uh_dport, 5751 saddr, &(*state)->gwy.addr, 5752 (*state)->gwy.port, &uh.uh_sum, 5753 pd2.ip_sum, icmpsum, 5754 pd->ip_sum, 1, pd2.af); 5755 } 5756 switch (pd2.af) { 5757#ifdef INET 5758 case AF_INET: 5759 m_copyback(m, off, ICMP_MINLEN, 5760 (caddr_t)pd->hdr.icmp); 5761 m_copyback(m, ipoff2, sizeof(h2), 5762 (caddr_t)&h2); 5763 break; 5764#endif /* INET */ 5765#ifdef INET6 5766 case AF_INET6: 5767 m_copyback(m, off, 5768 sizeof(struct icmp6_hdr), 5769 (caddr_t)pd->hdr.icmp6); 5770 m_copyback(m, ipoff2, sizeof(h2_6), 5771 (caddr_t)&h2_6); 5772 break; 5773#endif /* INET6 */ 5774 } 5775 m_copyback(m, off2, sizeof(uh), 5776 (caddr_t)&uh); 5777 } 5778 5779 return (PF_PASS); 5780 break; 5781 } 5782#ifdef INET 5783 case IPPROTO_ICMP: { 5784 struct icmp iih; 5785 5786 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN, 5787 NULL, reason, pd2.af)) { 5788 DPFPRINTF(PF_DEBUG_MISC, 5789 ("pf: ICMP error message too short i" 5790 "(icmp)\n")); 5791 return (PF_DROP); 5792 } 5793 5794 key.af = pd2.af; 5795 key.proto = IPPROTO_ICMP; 5796 if (direction == PF_IN) { 5797 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 5798 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 5799 key.ext.port = 0; 5800 key.gwy.port = iih.icmp_id; 5801 } else { 5802 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 5803 PF_ACPY(&key.ext.addr, pd2.src, key.af); 5804 key.lan.port = iih.icmp_id; 5805 key.ext.port = 0; 5806 } 5807 5808 STATE_LOOKUP(); 5809 5810 if (STATE_TRANSLATE(*state)) { 5811 if (direction == PF_IN) { 5812 pf_change_icmp(pd2.src, &iih.icmp_id, 5813 daddr, &(*state)->lan.addr, 5814 (*state)->lan.port, NULL, 5815 pd2.ip_sum, icmpsum, 5816 pd->ip_sum, 0, AF_INET); 5817 } else { 5818 pf_change_icmp(pd2.dst, &iih.icmp_id, 5819 saddr, &(*state)->gwy.addr, 5820 (*state)->gwy.port, NULL, 5821 pd2.ip_sum, icmpsum, 5822 pd->ip_sum, 0, AF_INET); 5823 } 5824 m_copyback(m, off, ICMP_MINLEN, 5825 (caddr_t)pd->hdr.icmp); 5826 m_copyback(m, ipoff2, sizeof(h2), 5827 (caddr_t)&h2); 5828 m_copyback(m, off2, ICMP_MINLEN, 5829 (caddr_t)&iih); 5830 } 5831 5832 return (PF_PASS); 5833 break; 5834 } 5835#endif /* INET */ 5836#ifdef INET6 5837 case IPPROTO_ICMPV6: { 5838 struct icmp6_hdr iih; 5839 5840 if (!pf_pull_hdr(m, off2, &iih, 5841 sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) { 5842 DPFPRINTF(PF_DEBUG_MISC, 5843 ("pf: ICMP error message too short " 5844 "(icmp6)\n")); 5845 return (PF_DROP); 5846 } 5847 5848 key.af = pd2.af; 5849 key.proto = IPPROTO_ICMPV6; 5850 if (direction == PF_IN) { 5851 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 5852 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 5853 key.ext.port = 0; 5854 key.gwy.port = iih.icmp6_id; 5855 } else { 5856 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 5857 PF_ACPY(&key.ext.addr, pd2.src, key.af); 5858 key.lan.port = iih.icmp6_id; 5859 key.ext.port = 0; 5860 } 5861 5862 STATE_LOOKUP(); 5863 5864 if (STATE_TRANSLATE(*state)) { 5865 if (direction == PF_IN) { 5866 pf_change_icmp(pd2.src, &iih.icmp6_id, 5867 daddr, &(*state)->lan.addr, 5868 (*state)->lan.port, NULL, 5869 pd2.ip_sum, icmpsum, 5870 pd->ip_sum, 0, AF_INET6); 5871 } else { 5872 pf_change_icmp(pd2.dst, &iih.icmp6_id, 5873 saddr, &(*state)->gwy.addr, 5874 (*state)->gwy.port, NULL, 5875 pd2.ip_sum, icmpsum, 5876 pd->ip_sum, 0, AF_INET6); 5877 } 5878 m_copyback(m, off, sizeof(struct icmp6_hdr), 5879 (caddr_t)pd->hdr.icmp6); 5880 m_copyback(m, ipoff2, sizeof(h2_6), 5881 (caddr_t)&h2_6); 5882 m_copyback(m, off2, sizeof(struct icmp6_hdr), 5883 (caddr_t)&iih); 5884 } 5885 5886 return (PF_PASS); 5887 break; 5888 } 5889#endif /* INET6 */ 5890 default: { 5891 key.af = pd2.af; 5892 key.proto = pd2.proto; 5893 if (direction == PF_IN) { 5894 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 5895 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 5896 key.ext.port = 0; 5897 key.gwy.port = 0; 5898 } else { 5899 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 5900 PF_ACPY(&key.ext.addr, pd2.src, key.af); 5901 key.lan.port = 0; 5902 key.ext.port = 0; 5903 } 5904 5905 STATE_LOOKUP(); 5906 5907 if (STATE_TRANSLATE(*state)) { 5908 if (direction == PF_IN) { 5909 pf_change_icmp(pd2.src, NULL, 5910 daddr, &(*state)->lan.addr, 5911 0, NULL, 5912 pd2.ip_sum, icmpsum, 5913 pd->ip_sum, 0, pd2.af); 5914 } else { 5915 pf_change_icmp(pd2.dst, NULL, 5916 saddr, &(*state)->gwy.addr, 5917 0, NULL, 5918 pd2.ip_sum, icmpsum, 5919 pd->ip_sum, 0, pd2.af); 5920 } 5921 switch (pd2.af) { 5922#ifdef INET 5923 case AF_INET: 5924 m_copyback(m, off, ICMP_MINLEN, 5925 (caddr_t)pd->hdr.icmp); 5926 m_copyback(m, ipoff2, sizeof(h2), 5927 (caddr_t)&h2); 5928 break; 5929#endif /* INET */ 5930#ifdef INET6 5931 case AF_INET6: 5932 m_copyback(m, off, 5933 sizeof(struct icmp6_hdr), 5934 (caddr_t)pd->hdr.icmp6); 5935 m_copyback(m, ipoff2, sizeof(h2_6), 5936 (caddr_t)&h2_6); 5937 break; 5938#endif /* INET6 */ 5939 } 5940 } 5941 5942 return (PF_PASS); 5943 break; 5944 } 5945 } 5946 } 5947} 5948 5949int 5950pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif, 5951 struct pf_pdesc *pd) 5952{ 5953 struct pf_state_peer *src, *dst; 5954 struct pf_state_cmp key; 5955 5956 key.af = pd->af; 5957 key.proto = pd->proto; 5958 if (direction == PF_IN) { 5959 PF_ACPY(&key.ext.addr, pd->src, key.af); 5960 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 5961 key.ext.port = 0; 5962 key.gwy.port = 0; 5963 } else { 5964 PF_ACPY(&key.lan.addr, pd->src, key.af); 5965 PF_ACPY(&key.ext.addr, pd->dst, key.af); 5966 key.lan.port = 0; 5967 key.ext.port = 0; 5968 } 5969 5970 STATE_LOOKUP(); 5971 5972 if (direction == (*state)->direction) { 5973 src = &(*state)->src; 5974 dst = &(*state)->dst; 5975 } else { 5976 src = &(*state)->dst; 5977 dst = &(*state)->src; 5978 } 5979 5980 /* update states */ 5981 if (src->state < PFOTHERS_SINGLE) 5982 src->state = PFOTHERS_SINGLE; 5983 if (dst->state == PFOTHERS_SINGLE) 5984 dst->state = PFOTHERS_MULTIPLE; 5985 5986 /* update expire time */ 5987 (*state)->expire = time_second; 5988 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE) 5989 (*state)->timeout = PFTM_OTHER_MULTIPLE; 5990 else 5991 (*state)->timeout = PFTM_OTHER_SINGLE; 5992 5993 /* translate source/destination address, if necessary */ 5994 if (STATE_TRANSLATE(*state)) { 5995 if (direction == PF_OUT) 5996 switch (pd->af) { 5997#ifdef INET 5998 case AF_INET: 5999 pf_change_a(&pd->src->v4.s_addr, 6000 pd->ip_sum, (*state)->gwy.addr.v4.s_addr, 6001 0); 6002 break; 6003#endif /* INET */ 6004#ifdef INET6 6005 case AF_INET6: 6006 PF_ACPY(pd->src, &(*state)->gwy.addr, pd->af); 6007 break; 6008#endif /* INET6 */ 6009 } 6010 else 6011 switch (pd->af) { 6012#ifdef INET 6013 case AF_INET: 6014 pf_change_a(&pd->dst->v4.s_addr, 6015 pd->ip_sum, (*state)->lan.addr.v4.s_addr, 6016 0); 6017 break; 6018#endif /* INET */ 6019#ifdef INET6 6020 case AF_INET6: 6021 PF_ACPY(pd->dst, &(*state)->lan.addr, pd->af); 6022 break; 6023#endif /* INET6 */ 6024 } 6025 } 6026 6027 return (PF_PASS); 6028} 6029 6030/* 6031 * ipoff and off are measured from the start of the mbuf chain. 6032 * h must be at "ipoff" on the mbuf chain. 6033 */ 6034void * 6035pf_pull_hdr(struct mbuf *m, int off, void *p, int len, 6036 u_short *actionp, u_short *reasonp, sa_family_t af) 6037{ 6038 switch (af) { 6039#ifdef INET 6040 case AF_INET: { 6041 struct ip *h = mtod(m, struct ip *); 6042 u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3; 6043 6044 if (fragoff) { 6045 if (fragoff >= len) 6046 ACTION_SET(actionp, PF_PASS); 6047 else { 6048 ACTION_SET(actionp, PF_DROP); 6049 REASON_SET(reasonp, PFRES_FRAG); 6050 } 6051 return (NULL); 6052 } 6053 if (m->m_pkthdr.len < off + len || 6054 ntohs(h->ip_len) < off + len) { 6055 ACTION_SET(actionp, PF_DROP); 6056 REASON_SET(reasonp, PFRES_SHORT); 6057 return (NULL); 6058 } 6059 break; 6060 } 6061#endif /* INET */ 6062#ifdef INET6 6063 case AF_INET6: { 6064 struct ip6_hdr *h = mtod(m, struct ip6_hdr *); 6065 6066 if (m->m_pkthdr.len < off + len || 6067 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) < 6068 (unsigned)(off + len)) { 6069 ACTION_SET(actionp, PF_DROP); 6070 REASON_SET(reasonp, PFRES_SHORT); 6071 return (NULL); 6072 } 6073 break; 6074 } 6075#endif /* INET6 */ 6076 } 6077 m_copydata(m, off, len, p); 6078 return (p); 6079} 6080 6081int 6082pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif) 6083{ 6084 struct sockaddr_in *dst; 6085 int ret = 1; 6086 int check_mpath; 6087#ifndef __FreeBSD__ 6088 extern int ipmultipath; 6089#endif 6090#ifdef INET6 6091#ifndef __FreeBSD__ 6092 extern int ip6_multipath; 6093#endif 6094 struct sockaddr_in6 *dst6; 6095 struct route_in6 ro; 6096#else 6097 struct route ro; 6098#endif 6099 struct radix_node *rn; 6100 struct rtentry *rt; 6101 struct ifnet *ifp; 6102 6103 check_mpath = 0; 6104 bzero(&ro, sizeof(ro)); 6105 switch (af) { 6106 case AF_INET: 6107 dst = satosin(&ro.ro_dst); 6108 dst->sin_family = AF_INET; 6109 dst->sin_len = sizeof(*dst); 6110 dst->sin_addr = addr->v4; 6111#ifndef __FreeBSD__ /* MULTIPATH_ROUTING */ 6112 if (ipmultipath) 6113 check_mpath = 1; 6114#endif 6115 break; 6116#ifdef INET6 6117 case AF_INET6: 6118 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 6119 dst6->sin6_family = AF_INET6; 6120 dst6->sin6_len = sizeof(*dst6); 6121 dst6->sin6_addr = addr->v6; 6122#ifndef __FreeBSD__ /* MULTIPATH_ROUTING */ 6123 if (ip6_multipath) 6124 check_mpath = 1; 6125#endif 6126 break; 6127#endif /* INET6 */ 6128 default: 6129 return (0); 6130 } 6131 6132 /* Skip checks for ipsec interfaces */ 6133 if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC) 6134 goto out; 6135 6136#ifdef __FreeBSD__ 6137/* XXX MRT not always INET */ /* stick with table 0 though */ 6138 if (af == AF_INET) 6139 in_rtalloc_ign((struct route *)&ro, 0, 0); 6140 else 6141 rtalloc_ign((struct route *)&ro, 0); 6142#else /* ! __FreeBSD__ */ 6143 rtalloc_noclone((struct route *)&ro, NO_CLONING); 6144#endif 6145 6146 if (ro.ro_rt != NULL) { 6147 /* No interface given, this is a no-route check */ 6148 if (kif == NULL) 6149 goto out; 6150 6151 if (kif->pfik_ifp == NULL) { 6152 ret = 0; 6153 goto out; 6154 } 6155 6156 /* Perform uRPF check if passed input interface */ 6157 ret = 0; 6158 rn = (struct radix_node *)ro.ro_rt; 6159 do { 6160 rt = (struct rtentry *)rn; 6161#ifndef __FreeBSD__ /* CARPDEV */ 6162 if (rt->rt_ifp->if_type == IFT_CARP) 6163 ifp = rt->rt_ifp->if_carpdev; 6164 else 6165#endif 6166 ifp = rt->rt_ifp; 6167 6168 if (kif->pfik_ifp == ifp) 6169 ret = 1; 6170#ifdef __FreeBSD__ /* MULTIPATH_ROUTING */ 6171 rn = NULL; 6172#else 6173 rn = rn_mpath_next(rn); 6174#endif 6175 } while (check_mpath == 1 && rn != NULL && ret == 0); 6176 } else 6177 ret = 0; 6178out: 6179 if (ro.ro_rt != NULL) 6180 RTFREE(ro.ro_rt); 6181 return (ret); 6182} 6183 6184int 6185pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw) 6186{ 6187 struct sockaddr_in *dst; 6188#ifdef INET6 6189 struct sockaddr_in6 *dst6; 6190 struct route_in6 ro; 6191#else 6192 struct route ro; 6193#endif 6194 int ret = 0; 6195 6196 bzero(&ro, sizeof(ro)); 6197 switch (af) { 6198 case AF_INET: 6199 dst = satosin(&ro.ro_dst); 6200 dst->sin_family = AF_INET; 6201 dst->sin_len = sizeof(*dst); 6202 dst->sin_addr = addr->v4; 6203 break; 6204#ifdef INET6 6205 case AF_INET6: 6206 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 6207 dst6->sin6_family = AF_INET6; 6208 dst6->sin6_len = sizeof(*dst6); 6209 dst6->sin6_addr = addr->v6; 6210 break; 6211#endif /* INET6 */ 6212 default: 6213 return (0); 6214 } 6215 6216#ifdef __FreeBSD__ 6217# ifdef RTF_PRCLONING 6218 rtalloc_ign((struct route *)&ro, (RTF_CLONING|RTF_PRCLONING)); 6219# else /* !RTF_PRCLONING */ 6220 if (af == AF_INET) 6221 in_rtalloc_ign((struct route *)&ro, 0, 0); 6222 else 6223 rtalloc_ign((struct route *)&ro, 0); 6224# endif 6225#else /* ! __FreeBSD__ */ 6226 rtalloc_noclone((struct route *)&ro, NO_CLONING); 6227#endif 6228 6229 if (ro.ro_rt != NULL) { 6230#ifdef __FreeBSD__ 6231 /* XXX_IMPORT: later */ 6232#else 6233 if (ro.ro_rt->rt_labelid == aw->v.rtlabel) 6234 ret = 1; 6235#endif 6236 RTFREE(ro.ro_rt); 6237 } 6238 6239 return (ret); 6240} 6241 6242#ifdef INET 6243 6244void 6245pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp, 6246 struct pf_state *s, struct pf_pdesc *pd) 6247{ 6248 struct mbuf *m0, *m1; 6249 struct route iproute; 6250 struct route *ro = NULL; 6251 struct sockaddr_in *dst; 6252 struct ip *ip; 6253 struct ifnet *ifp = NULL; 6254 struct pf_addr naddr; 6255 struct pf_src_node *sn = NULL; 6256 int error = 0; 6257#ifdef __FreeBSD__ 6258 int sw_csum; 6259#endif 6260#ifdef IPSEC 6261 struct m_tag *mtag; 6262#endif /* IPSEC */ 6263 6264 if (m == NULL || *m == NULL || r == NULL || 6265 (dir != PF_IN && dir != PF_OUT) || oifp == NULL) 6266 panic("pf_route: invalid parameters"); 6267 6268 if (pd->pf_mtag->routed++ > 3) { 6269 m0 = *m; 6270 *m = NULL; 6271 goto bad; 6272 } 6273 6274 if (r->rt == PF_DUPTO) { 6275#ifdef __FreeBSD__ 6276 if ((m0 = m_dup(*m, M_DONTWAIT)) == NULL) 6277#else 6278 if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL) 6279#endif 6280 return; 6281 } else { 6282 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) 6283 return; 6284 m0 = *m; 6285 } 6286 6287 if (m0->m_len < sizeof(struct ip)) { 6288 DPFPRINTF(PF_DEBUG_URGENT, 6289 ("pf_route: m0->m_len < sizeof(struct ip)\n")); 6290 goto bad; 6291 } 6292 6293 ip = mtod(m0, struct ip *); 6294 6295 ro = &iproute; 6296 bzero((caddr_t)ro, sizeof(*ro)); 6297 dst = satosin(&ro->ro_dst); 6298 dst->sin_family = AF_INET; 6299 dst->sin_len = sizeof(*dst); 6300 dst->sin_addr = ip->ip_dst; 6301 6302 if (r->rt == PF_FASTROUTE) { 6303 in_rtalloc(ro, 0); 6304 if (ro->ro_rt == 0) { 6305 KMOD_IPSTAT_INC(ips_noroute); 6306 goto bad; 6307 } 6308 6309 ifp = ro->ro_rt->rt_ifp; 6310 ro->ro_rt->rt_use++; 6311 6312 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 6313 dst = satosin(ro->ro_rt->rt_gateway); 6314 } else { 6315 if (TAILQ_EMPTY(&r->rpool.list)) { 6316 DPFPRINTF(PF_DEBUG_URGENT, 6317 ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n")); 6318 goto bad; 6319 } 6320 if (s == NULL) { 6321 pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src, 6322 &naddr, NULL, &sn); 6323 if (!PF_AZERO(&naddr, AF_INET)) 6324 dst->sin_addr.s_addr = naddr.v4.s_addr; 6325 ifp = r->rpool.cur->kif ? 6326 r->rpool.cur->kif->pfik_ifp : NULL; 6327 } else { 6328 if (!PF_AZERO(&s->rt_addr, AF_INET)) 6329 dst->sin_addr.s_addr = 6330 s->rt_addr.v4.s_addr; 6331 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; 6332 } 6333 } 6334 if (ifp == NULL) 6335 goto bad; 6336 6337 if (oifp != ifp) { 6338#ifdef __FreeBSD__ 6339 PF_UNLOCK(); 6340 if (pf_test(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) { 6341 PF_LOCK(); 6342 goto bad; 6343 } else if (m0 == NULL) { 6344 PF_LOCK(); 6345 goto done; 6346 } 6347 PF_LOCK(); 6348#else 6349 if (pf_test(PF_OUT, ifp, &m0, NULL) != PF_PASS) 6350 goto bad; 6351 else if (m0 == NULL) 6352 goto done; 6353#endif 6354 if (m0->m_len < sizeof(struct ip)) { 6355 DPFPRINTF(PF_DEBUG_URGENT, 6356 ("pf_route: m0->m_len < sizeof(struct ip)\n")); 6357 goto bad; 6358 } 6359 ip = mtod(m0, struct ip *); 6360 } 6361 6362#ifdef __FreeBSD__ 6363 /* Copied from FreeBSD 5.1-CURRENT ip_output. */ 6364 m0->m_pkthdr.csum_flags |= CSUM_IP; 6365 sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist; 6366 if (sw_csum & CSUM_DELAY_DATA) { 6367 /* 6368 * XXX: in_delayed_cksum assumes HBO for ip->ip_len (at least) 6369 */ 6370 NTOHS(ip->ip_len); 6371 NTOHS(ip->ip_off); /* XXX: needed? */ 6372 in_delayed_cksum(m0); 6373 HTONS(ip->ip_len); 6374 HTONS(ip->ip_off); 6375 sw_csum &= ~CSUM_DELAY_DATA; 6376 } 6377 m0->m_pkthdr.csum_flags &= ifp->if_hwassist; 6378 6379 if (ntohs(ip->ip_len) <= ifp->if_mtu || 6380 (m0->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 || 6381 (ifp->if_hwassist & CSUM_FRAGMENT && 6382 ((ip->ip_off & htons(IP_DF)) == 0))) { 6383 /* 6384 * ip->ip_len = htons(ip->ip_len); 6385 * ip->ip_off = htons(ip->ip_off); 6386 */ 6387 ip->ip_sum = 0; 6388 if (sw_csum & CSUM_DELAY_IP) { 6389 /* From KAME */ 6390 if (ip->ip_v == IPVERSION && 6391 (ip->ip_hl << 2) == sizeof(*ip)) { 6392 ip->ip_sum = in_cksum_hdr(ip); 6393 } else { 6394 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2); 6395 } 6396 } 6397 PF_UNLOCK(); 6398 error = (*ifp->if_output)(ifp, m0, sintosa(dst), ro); 6399 PF_LOCK(); 6400 goto done; 6401 } 6402 6403#else 6404 /* Copied from ip_output. */ 6405#ifdef IPSEC 6406 /* 6407 * If deferred crypto processing is needed, check that the 6408 * interface supports it. 6409 */ 6410 if ((mtag = m_tag_find(m0, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL)) 6411 != NULL && (ifp->if_capabilities & IFCAP_IPSEC) == 0) { 6412 /* Notify IPsec to do its own crypto. */ 6413 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 6414 goto bad; 6415 } 6416#endif /* IPSEC */ 6417 6418 /* Catch routing changes wrt. hardware checksumming for TCP or UDP. */ 6419 if (m0->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT) { 6420 if (!(ifp->if_capabilities & IFCAP_CSUM_TCPv4) || 6421 ifp->if_bridge != NULL) { 6422 in_delayed_cksum(m0); 6423 m0->m_pkthdr.csum_flags &= ~M_TCPV4_CSUM_OUT; /* Clear */ 6424 } 6425 } else if (m0->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT) { 6426 if (!(ifp->if_capabilities & IFCAP_CSUM_UDPv4) || 6427 ifp->if_bridge != NULL) { 6428 in_delayed_cksum(m0); 6429 m0->m_pkthdr.csum_flags &= ~M_UDPV4_CSUM_OUT; /* Clear */ 6430 } 6431 } 6432 6433 if (ntohs(ip->ip_len) <= ifp->if_mtu) { 6434 if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) && 6435 ifp->if_bridge == NULL) { 6436 m0->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT; 6437 KMOD_IPSTAT_INC(ips_outhwcsum); 6438 } else { 6439 ip->ip_sum = 0; 6440 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2); 6441 } 6442 /* Update relevant hardware checksum stats for TCP/UDP */ 6443 if (m0->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT) 6444 KMOD_TCPSTAT_INC(tcps_outhwcsum); 6445 else if (m0->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT) 6446 KMOD_UDPSTAT_INC(udps_outhwcsum); 6447 error = (*ifp->if_output)(ifp, m0, sintosa(dst), NULL); 6448 goto done; 6449 } 6450#endif 6451 /* 6452 * Too large for interface; fragment if possible. 6453 * Must be able to put at least 8 bytes per fragment. 6454 */ 6455 if (ip->ip_off & htons(IP_DF) || (m0->m_pkthdr.csum_flags & CSUM_TSO)) { 6456 KMOD_IPSTAT_INC(ips_cantfrag); 6457 if (r->rt != PF_DUPTO) { 6458#ifdef __FreeBSD__ 6459 /* icmp_error() expects host byte ordering */ 6460 NTOHS(ip->ip_len); 6461 NTOHS(ip->ip_off); 6462 PF_UNLOCK(); 6463 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0, 6464 ifp->if_mtu); 6465 PF_LOCK(); 6466#else 6467 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0, 6468 ifp->if_mtu); 6469#endif 6470 goto done; 6471 } else 6472 goto bad; 6473 } 6474 6475 m1 = m0; 6476#ifdef __FreeBSD__ 6477 /* 6478 * XXX: is cheaper + less error prone than own function 6479 */ 6480 NTOHS(ip->ip_len); 6481 NTOHS(ip->ip_off); 6482 error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum); 6483#else 6484 error = ip_fragment(m0, ifp, ifp->if_mtu); 6485#endif 6486 if (error) { 6487#ifndef __FreeBSD__ /* ip_fragment does not do m_freem() on FreeBSD */ 6488 m0 = NULL; 6489#endif 6490 goto bad; 6491 } 6492 6493 for (m0 = m1; m0; m0 = m1) { 6494 m1 = m0->m_nextpkt; 6495 m0->m_nextpkt = 0; 6496#ifdef __FreeBSD__ 6497 if (error == 0) { 6498 PF_UNLOCK(); 6499 error = (*ifp->if_output)(ifp, m0, sintosa(dst), 6500 NULL); 6501 PF_LOCK(); 6502 } else 6503#else 6504 if (error == 0) 6505 error = (*ifp->if_output)(ifp, m0, sintosa(dst), 6506 NULL); 6507 else 6508#endif 6509 m_freem(m0); 6510 } 6511 6512 if (error == 0) 6513 KMOD_IPSTAT_INC(ips_fragmented); 6514 6515done: 6516 if (r->rt != PF_DUPTO) 6517 *m = NULL; 6518 if (ro == &iproute && ro->ro_rt) 6519 RTFREE(ro->ro_rt); 6520 return; 6521 6522bad: 6523 m_freem(m0); 6524 goto done; 6525} 6526#endif /* INET */ 6527 6528#ifdef INET6 6529void 6530pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp, 6531 struct pf_state *s, struct pf_pdesc *pd) 6532{ 6533 struct mbuf *m0; 6534 struct route_in6 ip6route; 6535 struct route_in6 *ro; 6536 struct sockaddr_in6 *dst; 6537 struct ip6_hdr *ip6; 6538 struct ifnet *ifp = NULL; 6539 struct pf_addr naddr; 6540 struct pf_src_node *sn = NULL; 6541 int error = 0; 6542 6543 if (m == NULL || *m == NULL || r == NULL || 6544 (dir != PF_IN && dir != PF_OUT) || oifp == NULL) 6545 panic("pf_route6: invalid parameters"); 6546 6547 if (pd->pf_mtag->routed++ > 3) { 6548 m0 = *m; 6549 *m = NULL; 6550 goto bad; 6551 } 6552 6553 if (r->rt == PF_DUPTO) { 6554#ifdef __FreeBSD__ 6555 if ((m0 = m_dup(*m, M_DONTWAIT)) == NULL) 6556#else 6557 if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL) 6558#endif 6559 return; 6560 } else { 6561 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) 6562 return; 6563 m0 = *m; 6564 } 6565 6566 if (m0->m_len < sizeof(struct ip6_hdr)) { 6567 DPFPRINTF(PF_DEBUG_URGENT, 6568 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n")); 6569 goto bad; 6570 } 6571 ip6 = mtod(m0, struct ip6_hdr *); 6572 6573 ro = &ip6route; 6574 bzero((caddr_t)ro, sizeof(*ro)); 6575 dst = (struct sockaddr_in6 *)&ro->ro_dst; 6576 dst->sin6_family = AF_INET6; 6577 dst->sin6_len = sizeof(*dst); 6578 dst->sin6_addr = ip6->ip6_dst; 6579 6580 /* Cheat. XXX why only in the v6 case??? */ 6581 if (r->rt == PF_FASTROUTE) { 6582#ifdef __FreeBSD__ 6583 m0->m_flags |= M_SKIP_FIREWALL; 6584 PF_UNLOCK(); 6585 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL); 6586 PF_LOCK(); 6587#else 6588 mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT); 6589 if (mtag == NULL) 6590 goto bad; 6591 m_tag_prepend(m0, mtag); 6592 pd->pf_mtag->flags |= PF_TAG_GENERATED; 6593 ip6_output(m0, NULL, NULL, 0, NULL, NULL); 6594#endif 6595 return; 6596 } 6597 6598 if (TAILQ_EMPTY(&r->rpool.list)) { 6599 DPFPRINTF(PF_DEBUG_URGENT, 6600 ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n")); 6601 goto bad; 6602 } 6603 if (s == NULL) { 6604 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src, 6605 &naddr, NULL, &sn); 6606 if (!PF_AZERO(&naddr, AF_INET6)) 6607 PF_ACPY((struct pf_addr *)&dst->sin6_addr, 6608 &naddr, AF_INET6); 6609 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL; 6610 } else { 6611 if (!PF_AZERO(&s->rt_addr, AF_INET6)) 6612 PF_ACPY((struct pf_addr *)&dst->sin6_addr, 6613 &s->rt_addr, AF_INET6); 6614 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; 6615 } 6616 if (ifp == NULL) 6617 goto bad; 6618 6619 if (oifp != ifp) { 6620#ifdef __FreeBSD__ 6621 PF_UNLOCK(); 6622 if (pf_test6(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) { 6623 PF_LOCK(); 6624 goto bad; 6625 } else if (m0 == NULL) { 6626 PF_LOCK(); 6627 goto done; 6628 } 6629 PF_LOCK(); 6630#else 6631 if (pf_test6(PF_OUT, ifp, &m0, NULL) != PF_PASS) 6632 goto bad; 6633 else if (m0 == NULL) 6634 goto done; 6635#endif 6636 if (m0->m_len < sizeof(struct ip6_hdr)) { 6637 DPFPRINTF(PF_DEBUG_URGENT, 6638 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n")); 6639 goto bad; 6640 } 6641 ip6 = mtod(m0, struct ip6_hdr *); 6642 } 6643 6644 /* 6645 * If the packet is too large for the outgoing interface, 6646 * send back an icmp6 error. 6647 */ 6648 if (IN6_IS_SCOPE_EMBED(&dst->sin6_addr)) 6649 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index); 6650 if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) { 6651#ifdef __FreeBSD__ 6652 PF_UNLOCK(); 6653#endif 6654 error = nd6_output(ifp, ifp, m0, dst, NULL); 6655#ifdef __FreeBSD__ 6656 PF_LOCK(); 6657#endif 6658 } else { 6659 in6_ifstat_inc(ifp, ifs6_in_toobig); 6660#ifdef __FreeBSD__ 6661 if (r->rt != PF_DUPTO) { 6662 PF_UNLOCK(); 6663 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu); 6664 PF_LOCK(); 6665 } else 6666#else 6667 if (r->rt != PF_DUPTO) 6668 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu); 6669 else 6670#endif 6671 goto bad; 6672 } 6673 6674done: 6675 if (r->rt != PF_DUPTO) 6676 *m = NULL; 6677 return; 6678 6679bad: 6680 m_freem(m0); 6681 goto done; 6682} 6683#endif /* INET6 */ 6684 6685 6686#ifdef __FreeBSD__ 6687/* 6688 * FreeBSD supports cksum offloads for the following drivers. 6689 * em(4), fxp(4), ixgb(4), lge(4), ndis(4), nge(4), re(4), 6690 * ti(4), txp(4), xl(4) 6691 * 6692 * CSUM_DATA_VALID | CSUM_PSEUDO_HDR : 6693 * network driver performed cksum including pseudo header, need to verify 6694 * csum_data 6695 * CSUM_DATA_VALID : 6696 * network driver performed cksum, needs to additional pseudo header 6697 * cksum computation with partial csum_data(i.e. lack of H/W support for 6698 * pseudo header, for instance hme(4), sk(4) and possibly gem(4)) 6699 * 6700 * After validating the cksum of packet, set both flag CSUM_DATA_VALID and 6701 * CSUM_PSEUDO_HDR in order to avoid recomputation of the cksum in upper 6702 * TCP/UDP layer. 6703 * Also, set csum_data to 0xffff to force cksum validation. 6704 */ 6705int 6706pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p, sa_family_t af) 6707{ 6708 u_int16_t sum = 0; 6709 int hw_assist = 0; 6710 struct ip *ip; 6711 6712 if (off < sizeof(struct ip) || len < sizeof(struct udphdr)) 6713 return (1); 6714 if (m->m_pkthdr.len < off + len) 6715 return (1); 6716 6717 switch (p) { 6718 case IPPROTO_TCP: 6719 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { 6720 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) { 6721 sum = m->m_pkthdr.csum_data; 6722 } else { 6723 ip = mtod(m, struct ip *); 6724 sum = in_pseudo(ip->ip_src.s_addr, 6725 ip->ip_dst.s_addr, htonl((u_short)len + 6726 m->m_pkthdr.csum_data + IPPROTO_TCP)); 6727 } 6728 sum ^= 0xffff; 6729 ++hw_assist; 6730 } 6731 break; 6732 case IPPROTO_UDP: 6733 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { 6734 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) { 6735 sum = m->m_pkthdr.csum_data; 6736 } else { 6737 ip = mtod(m, struct ip *); 6738 sum = in_pseudo(ip->ip_src.s_addr, 6739 ip->ip_dst.s_addr, htonl((u_short)len + 6740 m->m_pkthdr.csum_data + IPPROTO_UDP)); 6741 } 6742 sum ^= 0xffff; 6743 ++hw_assist; 6744 } 6745 break; 6746 case IPPROTO_ICMP: 6747#ifdef INET6 6748 case IPPROTO_ICMPV6: 6749#endif /* INET6 */ 6750 break; 6751 default: 6752 return (1); 6753 } 6754 6755 if (!hw_assist) { 6756 switch (af) { 6757 case AF_INET: 6758 if (p == IPPROTO_ICMP) { 6759 if (m->m_len < off) 6760 return (1); 6761 m->m_data += off; 6762 m->m_len -= off; 6763 sum = in_cksum(m, len); 6764 m->m_data -= off; 6765 m->m_len += off; 6766 } else { 6767 if (m->m_len < sizeof(struct ip)) 6768 return (1); 6769 sum = in4_cksum(m, p, off, len); 6770 } 6771 break; 6772#ifdef INET6 6773 case AF_INET6: 6774 if (m->m_len < sizeof(struct ip6_hdr)) 6775 return (1); 6776 sum = in6_cksum(m, p, off, len); 6777 break; 6778#endif /* INET6 */ 6779 default: 6780 return (1); 6781 } 6782 } 6783 if (sum) { 6784 switch (p) { 6785 case IPPROTO_TCP: 6786 { 6787 KMOD_TCPSTAT_INC(tcps_rcvbadsum); 6788 break; 6789 } 6790 case IPPROTO_UDP: 6791 { 6792 KMOD_UDPSTAT_INC(udps_badsum); 6793 break; 6794 } 6795 case IPPROTO_ICMP: 6796 { 6797 KMOD_ICMPSTAT_INC(icps_checksum); 6798 break; 6799 } 6800#ifdef INET6 6801 case IPPROTO_ICMPV6: 6802 { 6803 KMOD_ICMP6STAT_INC(icp6s_checksum); 6804 break; 6805 } 6806#endif /* INET6 */ 6807 } 6808 return (1); 6809 } else { 6810 if (p == IPPROTO_TCP || p == IPPROTO_UDP) { 6811 m->m_pkthdr.csum_flags |= 6812 (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 6813 m->m_pkthdr.csum_data = 0xffff; 6814 } 6815 } 6816 return (0); 6817} 6818#else /* !__FreeBSD__ */ 6819/* 6820 * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag 6821 * off is the offset where the protocol header starts 6822 * len is the total length of protocol header plus payload 6823 * returns 0 when the checksum is valid, otherwise returns 1. 6824 */ 6825int 6826pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p, 6827 sa_family_t af) 6828{ 6829 u_int16_t flag_ok, flag_bad; 6830 u_int16_t sum; 6831 6832 switch (p) { 6833 case IPPROTO_TCP: 6834 flag_ok = M_TCP_CSUM_IN_OK; 6835 flag_bad = M_TCP_CSUM_IN_BAD; 6836 break; 6837 case IPPROTO_UDP: 6838 flag_ok = M_UDP_CSUM_IN_OK; 6839 flag_bad = M_UDP_CSUM_IN_BAD; 6840 break; 6841 case IPPROTO_ICMP: 6842#ifdef INET6 6843 case IPPROTO_ICMPV6: 6844#endif /* INET6 */ 6845 flag_ok = flag_bad = 0; 6846 break; 6847 default: 6848 return (1); 6849 } 6850 if (m->m_pkthdr.csum_flags & flag_ok) 6851 return (0); 6852 if (m->m_pkthdr.csum_flags & flag_bad) 6853 return (1); 6854 if (off < sizeof(struct ip) || len < sizeof(struct udphdr)) 6855 return (1); 6856 if (m->m_pkthdr.len < off + len) 6857 return (1); 6858 switch (af) { 6859#ifdef INET 6860 case AF_INET: 6861 if (p == IPPROTO_ICMP) { 6862 if (m->m_len < off) 6863 return (1); 6864 m->m_data += off; 6865 m->m_len -= off; 6866 sum = in_cksum(m, len); 6867 m->m_data -= off; 6868 m->m_len += off; 6869 } else { 6870 if (m->m_len < sizeof(struct ip)) 6871 return (1); 6872 sum = in4_cksum(m, p, off, len); 6873 } 6874 break; 6875#endif /* INET */ 6876#ifdef INET6 6877 case AF_INET6: 6878 if (m->m_len < sizeof(struct ip6_hdr)) 6879 return (1); 6880 sum = in6_cksum(m, p, off, len); 6881 break; 6882#endif /* INET6 */ 6883 default: 6884 return (1); 6885 } 6886 if (sum) { 6887 m->m_pkthdr.csum_flags |= flag_bad; 6888 switch (p) { 6889 case IPPROTO_TCP: 6890 KMOD_TCPSTAT_INC(tcps_rcvbadsum); 6891 break; 6892 case IPPROTO_UDP: 6893 KMOD_UDPSTAT_INC(udps_badsum); 6894 break; 6895 case IPPROTO_ICMP: 6896 KMOD_ICMPSTAT_INC(icps_checksum); 6897 break; 6898#ifdef INET6 6899 case IPPROTO_ICMPV6: 6900 KMOD_ICMP6STAT_INC(icp6s_checksum); 6901 break; 6902#endif /* INET6 */ 6903 } 6904 return (1); 6905 } 6906 m->m_pkthdr.csum_flags |= flag_ok; 6907 return (0); 6908} 6909#endif /* __FreeBSD__ */ 6910 6911#ifdef INET 6912int 6913#ifdef __FreeBSD__ 6914pf_test(int dir, struct ifnet *ifp, struct mbuf **m0, 6915 struct ether_header *eh, struct inpcb *inp) 6916#else 6917pf_test(int dir, struct ifnet *ifp, struct mbuf **m0, 6918 struct ether_header *eh) 6919#endif 6920{ 6921 struct pfi_kif *kif; 6922 u_short action, reason = 0, log = 0; 6923 struct mbuf *m = *m0; 6924 struct ip *h = NULL; /* make the compiler happy */ 6925 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr; 6926 struct pf_state *s = NULL; 6927 struct pf_ruleset *ruleset = NULL; 6928 struct pf_pdesc pd; 6929 int off, dirndx, pqid = 0; 6930 6931#ifdef __FreeBSD__ 6932 PF_LOCK(); 6933#endif 6934 if (!pf_status.running) 6935#ifdef __FreeBSD__ 6936 { 6937 PF_UNLOCK(); 6938#endif 6939 return (PF_PASS); 6940#ifdef __FreeBSD__ 6941 } 6942#endif 6943 6944 memset(&pd, 0, sizeof(pd)); 6945 if ((pd.pf_mtag = pf_get_mtag(m)) == NULL) { 6946#ifdef __FreeBSD__ 6947 PF_UNLOCK(); 6948#endif 6949 DPFPRINTF(PF_DEBUG_URGENT, 6950 ("pf_test: pf_get_mtag returned NULL\n")); 6951 return (PF_DROP); 6952 } 6953#ifdef __FreeBSD__ 6954 if (m->m_flags & M_SKIP_FIREWALL) { 6955 PF_UNLOCK(); 6956 return (PF_PASS); 6957 } 6958#else 6959 if (pd.pf_mtag->flags & PF_TAG_GENERATED) 6960 return (PF_PASS); 6961#endif 6962 6963#ifdef __FreeBSD__ 6964 /* XXX_IMPORT: later */ 6965#else 6966 if (ifp->if_type == IFT_CARP && ifp->if_carpdev) 6967 ifp = ifp->if_carpdev; 6968#endif 6969 6970 kif = (struct pfi_kif *)ifp->if_pf_kif; 6971 if (kif == NULL) { 6972#ifdef __FreeBSD__ 6973 PF_UNLOCK(); 6974#endif 6975 DPFPRINTF(PF_DEBUG_URGENT, 6976 ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname)); 6977 return (PF_DROP); 6978 } 6979 if (kif->pfik_flags & PFI_IFLAG_SKIP) { 6980#ifdef __FreeBSD__ 6981 PF_UNLOCK(); 6982#endif 6983 return (PF_PASS); 6984 } 6985 6986#ifdef __FreeBSD__ 6987 M_ASSERTPKTHDR(m); 6988#else 6989#ifdef DIAGNOSTIC 6990 if ((m->m_flags & M_PKTHDR) == 0) 6991 panic("non-M_PKTHDR is passed to pf_test"); 6992#endif /* DIAGNOSTIC */ 6993#endif /* __FreeBSD__ */ 6994 6995 if (m->m_pkthdr.len < (int)sizeof(*h)) { 6996 action = PF_DROP; 6997 REASON_SET(&reason, PFRES_SHORT); 6998 log = 1; 6999 goto done; 7000 } 7001 7002 /* We do IP header normalization and packet reassembly here */ 7003 if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) { 7004 action = PF_DROP; 7005 goto done; 7006 } 7007 m = *m0; 7008 h = mtod(m, struct ip *); 7009 7010 off = h->ip_hl << 2; 7011 if (off < (int)sizeof(*h)) { 7012 action = PF_DROP; 7013 REASON_SET(&reason, PFRES_SHORT); 7014 log = 1; 7015 goto done; 7016 } 7017 7018 pd.src = (struct pf_addr *)&h->ip_src; 7019 pd.dst = (struct pf_addr *)&h->ip_dst; 7020 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET); 7021 pd.ip_sum = &h->ip_sum; 7022 pd.proto = h->ip_p; 7023 pd.af = AF_INET; 7024 pd.tos = h->ip_tos; 7025 pd.tot_len = ntohs(h->ip_len); 7026 pd.eh = eh; 7027 7028 /* handle fragments that didn't get reassembled by normalization */ 7029 if (h->ip_off & htons(IP_MF | IP_OFFMASK)) { 7030 action = pf_test_fragment(&r, dir, kif, m, h, 7031 &pd, &a, &ruleset); 7032 goto done; 7033 } 7034 7035 switch (h->ip_p) { 7036 7037 case IPPROTO_TCP: { 7038 struct tcphdr th; 7039 7040 pd.hdr.tcp = &th; 7041 if (!pf_pull_hdr(m, off, &th, sizeof(th), 7042 &action, &reason, AF_INET)) { 7043 log = action != PF_PASS; 7044 goto done; 7045 } 7046 if (dir == PF_IN && pf_check_proto_cksum(m, off, 7047 ntohs(h->ip_len) - off, IPPROTO_TCP, AF_INET)) { 7048 REASON_SET(&reason, PFRES_PROTCKSUM); 7049 action = PF_DROP; 7050 goto done; 7051 } 7052 pd.p_len = pd.tot_len - off - (th.th_off << 2); 7053 if ((th.th_flags & TH_ACK) && pd.p_len == 0) 7054 pqid = 1; 7055 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd); 7056 if (action == PF_DROP) 7057 goto done; 7058 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd, 7059 &reason); 7060 if (action == PF_PASS) { 7061#if NPFSYNC 7062 pfsync_update_state(s); 7063#endif /* NPFSYNC */ 7064 r = s->rule.ptr; 7065 a = s->anchor.ptr; 7066 log = s->log; 7067 } else if (s == NULL) 7068#ifdef __FreeBSD__ 7069 action = pf_test_tcp(&r, &s, dir, kif, 7070 m, off, h, &pd, &a, &ruleset, NULL, inp); 7071#else 7072 action = pf_test_tcp(&r, &s, dir, kif, 7073 m, off, h, &pd, &a, &ruleset, &ipintrq); 7074#endif 7075 break; 7076 } 7077 7078 case IPPROTO_UDP: { 7079 struct udphdr uh; 7080 7081 pd.hdr.udp = &uh; 7082 if (!pf_pull_hdr(m, off, &uh, sizeof(uh), 7083 &action, &reason, AF_INET)) { 7084 log = action != PF_PASS; 7085 goto done; 7086 } 7087 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m, 7088 off, ntohs(h->ip_len) - off, IPPROTO_UDP, AF_INET)) { 7089 action = PF_DROP; 7090 REASON_SET(&reason, PFRES_PROTCKSUM); 7091 goto done; 7092 } 7093 if (uh.uh_dport == 0 || 7094 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off || 7095 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) { 7096 action = PF_DROP; 7097 REASON_SET(&reason, PFRES_SHORT); 7098 goto done; 7099 } 7100 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd); 7101 if (action == PF_PASS) { 7102#if NPFSYNC 7103 pfsync_update_state(s); 7104#endif /* NPFSYNC */ 7105 r = s->rule.ptr; 7106 a = s->anchor.ptr; 7107 log = s->log; 7108 } else if (s == NULL) 7109#ifdef __FreeBSD__ 7110 action = pf_test_udp(&r, &s, dir, kif, 7111 m, off, h, &pd, &a, &ruleset, NULL, inp); 7112#else 7113 action = pf_test_udp(&r, &s, dir, kif, 7114 m, off, h, &pd, &a, &ruleset, &ipintrq); 7115#endif 7116 break; 7117 } 7118 7119 case IPPROTO_ICMP: { 7120 struct icmp ih; 7121 7122 pd.hdr.icmp = &ih; 7123 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN, 7124 &action, &reason, AF_INET)) { 7125 log = action != PF_PASS; 7126 goto done; 7127 } 7128 if (dir == PF_IN && pf_check_proto_cksum(m, off, 7129 ntohs(h->ip_len) - off, IPPROTO_ICMP, AF_INET)) { 7130 action = PF_DROP; 7131 REASON_SET(&reason, PFRES_PROTCKSUM); 7132 goto done; 7133 } 7134 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd, 7135 &reason); 7136 if (action == PF_PASS) { 7137#if NPFSYNC 7138 pfsync_update_state(s); 7139#endif /* NPFSYNC */ 7140 r = s->rule.ptr; 7141 a = s->anchor.ptr; 7142 log = s->log; 7143 } else if (s == NULL) 7144#ifdef __FreeBSD__ 7145 action = pf_test_icmp(&r, &s, dir, kif, 7146 m, off, h, &pd, &a, &ruleset, NULL); 7147#else 7148 action = pf_test_icmp(&r, &s, dir, kif, 7149 m, off, h, &pd, &a, &ruleset, &ipintrq); 7150#endif 7151 break; 7152 } 7153 7154 default: 7155 action = pf_test_state_other(&s, dir, kif, &pd); 7156 if (action == PF_PASS) { 7157#if NPFSYNC 7158 pfsync_update_state(s); 7159#endif /* NPFSYNC */ 7160 r = s->rule.ptr; 7161 a = s->anchor.ptr; 7162 log = s->log; 7163 } else if (s == NULL) 7164#ifdef __FreeBSD__ 7165 action = pf_test_other(&r, &s, dir, kif, m, off, h, 7166 &pd, &a, &ruleset, NULL); 7167#else 7168 action = pf_test_other(&r, &s, dir, kif, m, off, h, 7169 &pd, &a, &ruleset, &ipintrq); 7170#endif 7171 break; 7172 } 7173 7174done: 7175 if (action == PF_PASS && h->ip_hl > 5 && 7176 !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) { 7177 action = PF_DROP; 7178 REASON_SET(&reason, PFRES_IPOPTIONS); 7179 log = 1; 7180 DPFPRINTF(PF_DEBUG_MISC, 7181 ("pf: dropping packet with ip options\n")); 7182 } 7183 7184 if ((s && s->tag) || r->rtableid) 7185 pf_tag_packet(m, pd.pf_mtag, s ? s->tag : 0, r->rtableid); 7186 7187#ifdef ALTQ 7188 if (action == PF_PASS && r->qid) { 7189 if (pqid || (pd.tos & IPTOS_LOWDELAY)) 7190 pd.pf_mtag->qid = r->pqid; 7191 else 7192 pd.pf_mtag->qid = r->qid; 7193 /* add hints for ecn */ 7194 pd.pf_mtag->af = AF_INET; 7195 pd.pf_mtag->hdr = h; 7196 } 7197#endif /* ALTQ */ 7198 7199 /* 7200 * connections redirected to loopback should not match sockets 7201 * bound specifically to loopback due to security implications, 7202 * see tcp_input() and in_pcblookup_listen(). 7203 */ 7204 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP || 7205 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL && 7206 (s->nat_rule.ptr->action == PF_RDR || 7207 s->nat_rule.ptr->action == PF_BINAT) && 7208 (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) 7209 pd.pf_mtag->flags |= PF_TAG_TRANSLATE_LOCALHOST; 7210 7211 if (log) { 7212 struct pf_rule *lr; 7213 7214 if (s != NULL && s->nat_rule.ptr != NULL && 7215 s->nat_rule.ptr->log & PF_LOG_ALL) 7216 lr = s->nat_rule.ptr; 7217 else 7218 lr = r; 7219 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset, 7220 &pd); 7221 } 7222 7223 kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len; 7224 kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++; 7225 7226 if (action == PF_PASS || r->action == PF_DROP) { 7227 dirndx = (dir == PF_OUT); 7228 r->packets[dirndx]++; 7229 r->bytes[dirndx] += pd.tot_len; 7230 if (a != NULL) { 7231 a->packets[dirndx]++; 7232 a->bytes[dirndx] += pd.tot_len; 7233 } 7234 if (s != NULL) { 7235 if (s->nat_rule.ptr != NULL) { 7236 s->nat_rule.ptr->packets[dirndx]++; 7237 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len; 7238 } 7239 if (s->src_node != NULL) { 7240 s->src_node->packets[dirndx]++; 7241 s->src_node->bytes[dirndx] += pd.tot_len; 7242 } 7243 if (s->nat_src_node != NULL) { 7244 s->nat_src_node->packets[dirndx]++; 7245 s->nat_src_node->bytes[dirndx] += pd.tot_len; 7246 } 7247 dirndx = (dir == s->direction) ? 0 : 1; 7248 s->packets[dirndx]++; 7249 s->bytes[dirndx] += pd.tot_len; 7250 } 7251 tr = r; 7252 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule; 7253 if (nr != NULL) { 7254 struct pf_addr *x; 7255 /* 7256 * XXX: we need to make sure that the addresses 7257 * passed to pfr_update_stats() are the same than 7258 * the addresses used during matching (pfr_match) 7259 */ 7260 if (r == &pf_default_rule) { 7261 tr = nr; 7262 x = (s == NULL || s->direction == dir) ? 7263 &pd.baddr : &pd.naddr; 7264 } else 7265 x = (s == NULL || s->direction == dir) ? 7266 &pd.naddr : &pd.baddr; 7267 if (x == &pd.baddr || s == NULL) { 7268 /* we need to change the address */ 7269 if (dir == PF_OUT) 7270 pd.src = x; 7271 else 7272 pd.dst = x; 7273 } 7274 } 7275 if (tr->src.addr.type == PF_ADDR_TABLE) 7276 pfr_update_stats(tr->src.addr.p.tbl, (s == NULL || 7277 s->direction == dir) ? pd.src : pd.dst, pd.af, 7278 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 7279 tr->src.neg); 7280 if (tr->dst.addr.type == PF_ADDR_TABLE) 7281 pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL || 7282 s->direction == dir) ? pd.dst : pd.src, pd.af, 7283 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 7284 tr->dst.neg); 7285 } 7286 7287 7288 if (action == PF_SYNPROXY_DROP) { 7289 m_freem(*m0); 7290 *m0 = NULL; 7291 action = PF_PASS; 7292 } else if (r->rt) 7293 /* pf_route can free the mbuf causing *m0 to become NULL */ 7294 pf_route(m0, r, dir, ifp, s, &pd); 7295 7296#ifdef __FreeBSD__ 7297 PF_UNLOCK(); 7298#endif 7299 7300 return (action); 7301} 7302#endif /* INET */ 7303 7304#ifdef INET6 7305int 7306#ifdef __FreeBSD__ 7307pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0, 7308 struct ether_header *eh, struct inpcb *inp) 7309#else 7310pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0, 7311 struct ether_header *eh) 7312#endif 7313{ 7314 struct pfi_kif *kif; 7315 u_short action, reason = 0, log = 0; 7316 struct mbuf *m = *m0, *n = NULL; 7317 struct ip6_hdr *h; 7318 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr; 7319 struct pf_state *s = NULL; 7320 struct pf_ruleset *ruleset = NULL; 7321 struct pf_pdesc pd; 7322 int off, terminal = 0, dirndx, rh_cnt = 0; 7323 7324#ifdef __FreeBSD__ 7325 PF_LOCK(); 7326#endif 7327 7328 if (!pf_status.running) 7329#ifdef __FreeBSD__ 7330 { 7331 PF_UNLOCK(); 7332#endif 7333 return (PF_PASS); 7334#ifdef __FreeBSD__ 7335 } 7336#endif 7337 7338 memset(&pd, 0, sizeof(pd)); 7339 if ((pd.pf_mtag = pf_get_mtag(m)) == NULL) { 7340#ifdef __FreeBSD__ 7341 PF_UNLOCK(); 7342#endif 7343 DPFPRINTF(PF_DEBUG_URGENT, 7344 ("pf_test6: pf_get_mtag returned NULL\n")); 7345 return (PF_DROP); 7346 } 7347 if (pd.pf_mtag->flags & PF_TAG_GENERATED) 7348 return (PF_PASS); 7349 7350#ifdef __FreeBSD__ 7351 /* XXX_IMPORT: later */ 7352#else 7353 if (ifp->if_type == IFT_CARP && ifp->if_carpdev) 7354 ifp = ifp->if_carpdev; 7355#endif 7356 7357 kif = (struct pfi_kif *)ifp->if_pf_kif; 7358 if (kif == NULL) { 7359#ifdef __FreeBSD__ 7360 PF_UNLOCK(); 7361#endif 7362 DPFPRINTF(PF_DEBUG_URGENT, 7363 ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname)); 7364 return (PF_DROP); 7365 } 7366 if (kif->pfik_flags & PFI_IFLAG_SKIP) { 7367#ifdef __FreeBSD__ 7368 PF_UNLOCK(); 7369#endif 7370 return (PF_PASS); 7371 } 7372 7373#ifdef __FreeBSD__ 7374 M_ASSERTPKTHDR(m); 7375#else 7376#ifdef DIAGNOSTIC 7377 if ((m->m_flags & M_PKTHDR) == 0) 7378 panic("non-M_PKTHDR is passed to pf_test6"); 7379#endif /* DIAGNOSTIC */ 7380#endif 7381 7382#ifdef __FreeBSD__ 7383 h = NULL; /* make the compiler happy */ 7384#endif 7385 7386 if (m->m_pkthdr.len < (int)sizeof(*h)) { 7387 action = PF_DROP; 7388 REASON_SET(&reason, PFRES_SHORT); 7389 log = 1; 7390 goto done; 7391 } 7392 7393 /* We do IP header normalization and packet reassembly here */ 7394 if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) { 7395 action = PF_DROP; 7396 goto done; 7397 } 7398 m = *m0; 7399 h = mtod(m, struct ip6_hdr *); 7400 7401#if 1 7402 /* 7403 * we do not support jumbogram yet. if we keep going, zero ip6_plen 7404 * will do something bad, so drop the packet for now. 7405 */ 7406 if (htons(h->ip6_plen) == 0) { 7407 action = PF_DROP; 7408 REASON_SET(&reason, PFRES_NORM); /*XXX*/ 7409 goto done; 7410 } 7411#endif 7412 7413 pd.src = (struct pf_addr *)&h->ip6_src; 7414 pd.dst = (struct pf_addr *)&h->ip6_dst; 7415 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6); 7416 pd.ip_sum = NULL; 7417 pd.af = AF_INET6; 7418 pd.tos = 0; 7419 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr); 7420 pd.eh = eh; 7421 7422 off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr); 7423 pd.proto = h->ip6_nxt; 7424 do { 7425 switch (pd.proto) { 7426 case IPPROTO_FRAGMENT: 7427 action = pf_test_fragment(&r, dir, kif, m, h, 7428 &pd, &a, &ruleset); 7429 if (action == PF_DROP) 7430 REASON_SET(&reason, PFRES_FRAG); 7431 goto done; 7432 case IPPROTO_ROUTING: { 7433 struct ip6_rthdr rthdr; 7434 7435 if (rh_cnt++) { 7436 DPFPRINTF(PF_DEBUG_MISC, 7437 ("pf: IPv6 more than one rthdr\n")); 7438 action = PF_DROP; 7439 REASON_SET(&reason, PFRES_IPOPTIONS); 7440 log = 1; 7441 goto done; 7442 } 7443 if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL, 7444 &reason, pd.af)) { 7445 DPFPRINTF(PF_DEBUG_MISC, 7446 ("pf: IPv6 short rthdr\n")); 7447 action = PF_DROP; 7448 REASON_SET(&reason, PFRES_SHORT); 7449 log = 1; 7450 goto done; 7451 } 7452 if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) { 7453 DPFPRINTF(PF_DEBUG_MISC, 7454 ("pf: IPv6 rthdr0\n")); 7455 action = PF_DROP; 7456 REASON_SET(&reason, PFRES_IPOPTIONS); 7457 log = 1; 7458 goto done; 7459 } 7460 /* fallthrough */ 7461 } 7462 case IPPROTO_AH: 7463 case IPPROTO_HOPOPTS: 7464 case IPPROTO_DSTOPTS: { 7465 /* get next header and header length */ 7466 struct ip6_ext opt6; 7467 7468 if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6), 7469 NULL, &reason, pd.af)) { 7470 DPFPRINTF(PF_DEBUG_MISC, 7471 ("pf: IPv6 short opt\n")); 7472 action = PF_DROP; 7473 log = 1; 7474 goto done; 7475 } 7476 if (pd.proto == IPPROTO_AH) 7477 off += (opt6.ip6e_len + 2) * 4; 7478 else 7479 off += (opt6.ip6e_len + 1) * 8; 7480 pd.proto = opt6.ip6e_nxt; 7481 /* goto the next header */ 7482 break; 7483 } 7484 default: 7485 terminal++; 7486 break; 7487 } 7488 } while (!terminal); 7489 7490 /* if there's no routing header, use unmodified mbuf for checksumming */ 7491 if (!n) 7492 n = m; 7493 7494 switch (pd.proto) { 7495 7496 case IPPROTO_TCP: { 7497 struct tcphdr th; 7498 7499 pd.hdr.tcp = &th; 7500 if (!pf_pull_hdr(m, off, &th, sizeof(th), 7501 &action, &reason, AF_INET6)) { 7502 log = action != PF_PASS; 7503 goto done; 7504 } 7505 if (dir == PF_IN && pf_check_proto_cksum(n, off, 7506 ntohs(h->ip6_plen) - (off - sizeof(struct ip6_hdr)), 7507 IPPROTO_TCP, AF_INET6)) { 7508 action = PF_DROP; 7509 REASON_SET(&reason, PFRES_PROTCKSUM); 7510 goto done; 7511 } 7512 pd.p_len = pd.tot_len - off - (th.th_off << 2); 7513 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd); 7514 if (action == PF_DROP) 7515 goto done; 7516 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd, 7517 &reason); 7518 if (action == PF_PASS) { 7519#if NPFSYNC 7520 pfsync_update_state(s); 7521#endif /* NPFSYNC */ 7522 r = s->rule.ptr; 7523 a = s->anchor.ptr; 7524 log = s->log; 7525 } else if (s == NULL) 7526#ifdef __FreeBSD__ 7527 action = pf_test_tcp(&r, &s, dir, kif, 7528 m, off, h, &pd, &a, &ruleset, NULL, inp); 7529#else 7530 action = pf_test_tcp(&r, &s, dir, kif, 7531 m, off, h, &pd, &a, &ruleset, &ip6intrq); 7532#endif 7533 break; 7534 } 7535 7536 case IPPROTO_UDP: { 7537 struct udphdr uh; 7538 7539 pd.hdr.udp = &uh; 7540 if (!pf_pull_hdr(m, off, &uh, sizeof(uh), 7541 &action, &reason, AF_INET6)) { 7542 log = action != PF_PASS; 7543 goto done; 7544 } 7545 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(n, 7546 off, ntohs(h->ip6_plen) - (off - sizeof(struct ip6_hdr)), 7547 IPPROTO_UDP, AF_INET6)) { 7548 action = PF_DROP; 7549 REASON_SET(&reason, PFRES_PROTCKSUM); 7550 goto done; 7551 } 7552 if (uh.uh_dport == 0 || 7553 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off || 7554 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) { 7555 action = PF_DROP; 7556 REASON_SET(&reason, PFRES_SHORT); 7557 goto done; 7558 } 7559 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd); 7560 if (action == PF_PASS) { 7561#if NPFSYNC 7562 pfsync_update_state(s); 7563#endif /* NPFSYNC */ 7564 r = s->rule.ptr; 7565 a = s->anchor.ptr; 7566 log = s->log; 7567 } else if (s == NULL) 7568#ifdef __FreeBSD__ 7569 action = pf_test_udp(&r, &s, dir, kif, 7570 m, off, h, &pd, &a, &ruleset, NULL, inp); 7571#else 7572 action = pf_test_udp(&r, &s, dir, kif, 7573 m, off, h, &pd, &a, &ruleset, &ip6intrq); 7574#endif 7575 break; 7576 } 7577 7578 case IPPROTO_ICMPV6: { 7579 struct icmp6_hdr ih; 7580 7581 pd.hdr.icmp6 = &ih; 7582 if (!pf_pull_hdr(m, off, &ih, sizeof(ih), 7583 &action, &reason, AF_INET6)) { 7584 log = action != PF_PASS; 7585 goto done; 7586 } 7587 if (dir == PF_IN && pf_check_proto_cksum(n, off, 7588 ntohs(h->ip6_plen) - (off - sizeof(struct ip6_hdr)), 7589 IPPROTO_ICMPV6, AF_INET6)) { 7590 action = PF_DROP; 7591 REASON_SET(&reason, PFRES_PROTCKSUM); 7592 goto done; 7593 } 7594 action = pf_test_state_icmp(&s, dir, kif, 7595 m, off, h, &pd, &reason); 7596 if (action == PF_PASS) { 7597#if NPFSYNC 7598 pfsync_update_state(s); 7599#endif /* NPFSYNC */ 7600 r = s->rule.ptr; 7601 a = s->anchor.ptr; 7602 log = s->log; 7603 } else if (s == NULL) 7604#ifdef __FreeBSD__ 7605 action = pf_test_icmp(&r, &s, dir, kif, 7606 m, off, h, &pd, &a, &ruleset, NULL); 7607#else 7608 action = pf_test_icmp(&r, &s, dir, kif, 7609 m, off, h, &pd, &a, &ruleset, &ip6intrq); 7610#endif 7611 break; 7612 } 7613 7614 default: 7615 action = pf_test_state_other(&s, dir, kif, &pd); 7616 if (action == PF_PASS) { 7617#if NPFSYNC 7618 pfsync_update_state(s); 7619#endif /* NPFSYNC */ 7620 r = s->rule.ptr; 7621 a = s->anchor.ptr; 7622 log = s->log; 7623 } else if (s == NULL) 7624#ifdef __FreeBSD__ 7625 action = pf_test_other(&r, &s, dir, kif, m, off, h, 7626 &pd, &a, &ruleset, NULL); 7627#else 7628 action = pf_test_other(&r, &s, dir, kif, m, off, h, 7629 &pd, &a, &ruleset, &ip6intrq); 7630#endif 7631 break; 7632 } 7633 7634done: 7635 /* handle dangerous IPv6 extension headers. */ 7636 if (action == PF_PASS && rh_cnt && 7637 !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) { 7638 action = PF_DROP; 7639 REASON_SET(&reason, PFRES_IPOPTIONS); 7640 log = 1; 7641 DPFPRINTF(PF_DEBUG_MISC, 7642 ("pf: dropping packet with dangerous v6 headers\n")); 7643 } 7644 7645 if ((s && s->tag) || r->rtableid) 7646 pf_tag_packet(m, pd.pf_mtag, s ? s->tag : 0, r->rtableid); 7647 7648#ifdef ALTQ 7649 if (action == PF_PASS && r->qid) { 7650 if (pd.tos & IPTOS_LOWDELAY) 7651 pd.pf_mtag->qid = r->pqid; 7652 else 7653 pd.pf_mtag->qid = r->qid; 7654 /* add hints for ecn */ 7655 pd.pf_mtag->af = AF_INET6; 7656 pd.pf_mtag->hdr = h; 7657 } 7658#endif /* ALTQ */ 7659 7660 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP || 7661 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL && 7662 (s->nat_rule.ptr->action == PF_RDR || 7663 s->nat_rule.ptr->action == PF_BINAT) && 7664 IN6_IS_ADDR_LOOPBACK(&pd.dst->v6)) 7665 pd.pf_mtag->flags |= PF_TAG_TRANSLATE_LOCALHOST; 7666 7667 if (log) { 7668 struct pf_rule *lr; 7669 7670 if (s != NULL && s->nat_rule.ptr != NULL && 7671 s->nat_rule.ptr->log & PF_LOG_ALL) 7672 lr = s->nat_rule.ptr; 7673 else 7674 lr = r; 7675 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset, 7676 &pd); 7677 } 7678 7679 kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len; 7680 kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++; 7681 7682 if (action == PF_PASS || r->action == PF_DROP) { 7683 dirndx = (dir == PF_OUT); 7684 r->packets[dirndx]++; 7685 r->bytes[dirndx] += pd.tot_len; 7686 if (a != NULL) { 7687 a->packets[dirndx]++; 7688 a->bytes[dirndx] += pd.tot_len; 7689 } 7690 if (s != NULL) { 7691 if (s->nat_rule.ptr != NULL) { 7692 s->nat_rule.ptr->packets[dirndx]++; 7693 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len; 7694 } 7695 if (s->src_node != NULL) { 7696 s->src_node->packets[dirndx]++; 7697 s->src_node->bytes[dirndx] += pd.tot_len; 7698 } 7699 if (s->nat_src_node != NULL) { 7700 s->nat_src_node->packets[dirndx]++; 7701 s->nat_src_node->bytes[dirndx] += pd.tot_len; 7702 } 7703 dirndx = (dir == s->direction) ? 0 : 1; 7704 s->packets[dirndx]++; 7705 s->bytes[dirndx] += pd.tot_len; 7706 } 7707 tr = r; 7708 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule; 7709 if (nr != NULL) { 7710 struct pf_addr *x; 7711 /* 7712 * XXX: we need to make sure that the addresses 7713 * passed to pfr_update_stats() are the same than 7714 * the addresses used during matching (pfr_match) 7715 */ 7716 if (r == &pf_default_rule) { 7717 tr = nr; 7718 x = (s == NULL || s->direction == dir) ? 7719 &pd.baddr : &pd.naddr; 7720 } else { 7721 x = (s == NULL || s->direction == dir) ? 7722 &pd.naddr : &pd.baddr; 7723 } 7724 if (x == &pd.baddr || s == NULL) { 7725 if (dir == PF_OUT) 7726 pd.src = x; 7727 else 7728 pd.dst = x; 7729 } 7730 } 7731 if (tr->src.addr.type == PF_ADDR_TABLE) 7732 pfr_update_stats(tr->src.addr.p.tbl, (s == NULL || 7733 s->direction == dir) ? pd.src : pd.dst, pd.af, 7734 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 7735 tr->src.neg); 7736 if (tr->dst.addr.type == PF_ADDR_TABLE) 7737 pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL || 7738 s->direction == dir) ? pd.dst : pd.src, pd.af, 7739 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 7740 tr->dst.neg); 7741 } 7742 7743 7744 if (action == PF_SYNPROXY_DROP) { 7745 m_freem(*m0); 7746 *m0 = NULL; 7747 action = PF_PASS; 7748 } else if (r->rt) 7749 /* pf_route6 can free the mbuf causing *m0 to become NULL */ 7750 pf_route6(m0, r, dir, ifp, s, &pd); 7751 7752#ifdef __FreeBSD__ 7753 PF_UNLOCK(); 7754#endif 7755 return (action); 7756} 7757#endif /* INET6 */ 7758 7759int 7760pf_check_congestion(struct ifqueue *ifq) 7761{ 7762#ifdef __FreeBSD__ 7763 /* XXX_IMPORT: later */ 7764 return (0); 7765#else 7766 if (ifq->ifq_congestion) 7767 return (1); 7768 else 7769 return (0); 7770#endif 7771} 7772