1 /* $NetBSD: pf.c,v 1.53 2008/10/11 13:40:57 pooka Exp $ */ 2 /* $OpenBSD: pf.c,v 1.552.2.1 2007/11/27 16:37:57 henning 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 #include <sys/cdefs.h> 40 __KERNEL_RCSID(0, "$NetBSD: pf.c,v 1.53 2008/10/11 13:40:57 pooka Exp $"); 41 42 #include "bpfilter.h" 43 #include "pflog.h" 44 45 #ifndef __NetBSD__ 46 #include "pfsync.h" 47 #else 48 #define NPFSYNC 0 49 #endif /* __NetBSD__ */ 50 51 #include <sys/param.h> 52 #include <sys/systm.h> 53 #include <sys/mbuf.h> 54 #include <sys/filio.h> 55 #include <sys/socket.h> 56 #include <sys/socketvar.h> 57 #include <sys/kernel.h> 58 #include <sys/time.h> 59 #include <sys/pool.h> 60 #include <sys/proc.h> 61 #include <sys/rwlock.h> 62 #ifdef __NetBSD__ 63 #include <sys/kthread.h> 64 #include <sys/uidinfo.h> 65 #endif /* __NetBSD__ */ 66 67 #include <net/if.h> 68 #include <net/if_types.h> 69 #include <net/bpf.h> 70 #include <net/route.h> 71 #ifndef __NetBSD__ 72 #include <net/radix_mpath.h> 73 #endif /* !__NetBSD__ */ 74 75 #include <netinet/in.h> 76 #ifdef __NetBSD__ 77 #include <netinet/in_offload.h> 78 #endif /* __NetBSD__ */ 79 #include <netinet/in_var.h> 80 #include <netinet/in_systm.h> 81 #include <netinet/ip.h> 82 #include <netinet/ip_var.h> 83 #include <netinet/tcp.h> 84 #include <netinet/tcp_seq.h> 85 #include <netinet/udp.h> 86 #include <netinet/ip_icmp.h> 87 #include <netinet/in_pcb.h> 88 #include <netinet/tcp_timer.h> 89 #include <netinet/tcp_var.h> 90 #include <netinet/udp_var.h> 91 #include <netinet/icmp_var.h> 92 #ifndef __NetBSD__ 93 #include <netinet/if_ether.h> 94 #else 95 #include <net/if_ether.h> 96 #endif /* __NetBSD__ */ 97 98 #ifndef __NetBSD__ 99 #include <dev/rndvar.h> 100 #else 101 #include <sys/rnd.h> 102 #endif /* __NetBSD__ */ 103 104 #include <net/pfvar.h> 105 #include <net/if_pflog.h> 106 107 #if NPFSYNC > 0 108 #include <net/if_pfsync.h> 109 #endif /* NPFSYNC > 0 */ 110 111 #ifdef INET6 112 #include <netinet/ip6.h> 113 #include <netinet6/ip6_var.h> 114 #ifdef __NetBSD__ 115 #include <netinet6/in6_pcb.h> 116 #endif /* __NetBSD__ */ 117 #include <netinet/icmp6.h> 118 #include <netinet6/nd6.h> 119 #endif /* INET6 */ 120 121 #ifdef __NetBSD__ 122 #include <netinet/tcp_rndiss.h> 123 #endif /* __NetBSD__ */ 124 125 126 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x 127 128 /* 129 * Global variables 130 */ 131 132 /* state tables */ 133 struct pf_state_tree_lan_ext pf_statetbl_lan_ext; 134 struct pf_state_tree_ext_gwy pf_statetbl_ext_gwy; 135 136 struct pf_altqqueue pf_altqs[2]; 137 struct pf_palist pf_pabuf; 138 struct pf_altqqueue *pf_altqs_active; 139 struct pf_altqqueue *pf_altqs_inactive; 140 struct pf_status pf_status; 141 142 u_int32_t ticket_altqs_active; 143 u_int32_t ticket_altqs_inactive; 144 int altqs_inactive_open; 145 u_int32_t ticket_pabuf; 146 147 struct pf_anchor_stackframe { 148 struct pf_ruleset *rs; 149 struct pf_rule *r; 150 struct pf_anchor_node *parent; 151 struct pf_anchor *child; 152 } pf_anchor_stack[64]; 153 154 struct pool pf_src_tree_pl, pf_rule_pl, pf_pooladdr_pl; 155 struct pool pf_state_pl, pf_state_key_pl; 156 struct pool pf_altq_pl; 157 158 void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t); 159 160 void pf_init_threshold(struct pf_threshold *, u_int32_t, 161 u_int32_t); 162 void pf_add_threshold(struct pf_threshold *); 163 int pf_check_threshold(struct pf_threshold *); 164 165 void pf_change_ap(struct pf_addr *, u_int16_t *, 166 u_int16_t *, u_int16_t *, struct pf_addr *, 167 u_int16_t, u_int8_t, sa_family_t); 168 int pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *, 169 struct tcphdr *, struct pf_state_peer *); 170 #ifdef INET6 171 void pf_change_a6(struct pf_addr *, u_int16_t *, 172 struct pf_addr *, u_int8_t); 173 #endif /* INET6 */ 174 void pf_change_icmp(struct pf_addr *, u_int16_t *, 175 struct pf_addr *, struct pf_addr *, u_int16_t, 176 u_int16_t *, u_int16_t *, u_int16_t *, 177 u_int16_t *, u_int8_t, sa_family_t); 178 void pf_send_tcp(const struct pf_rule *, sa_family_t, 179 const struct pf_addr *, const struct pf_addr *, 180 u_int16_t, u_int16_t, u_int32_t, u_int32_t, 181 u_int8_t, u_int16_t, u_int16_t, u_int8_t, int, 182 u_int16_t, struct ether_header *, struct ifnet *); 183 void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t, 184 sa_family_t, struct pf_rule *); 185 struct pf_rule *pf_match_translation(struct pf_pdesc *, struct mbuf *, 186 int, int, struct pfi_kif *, 187 struct pf_addr *, u_int16_t, struct pf_addr *, 188 u_int16_t, int); 189 struct pf_rule *pf_get_translation(struct pf_pdesc *, struct mbuf *, 190 int, int, struct pfi_kif *, struct pf_src_node **, 191 struct pf_addr *, u_int16_t, 192 struct pf_addr *, u_int16_t, 193 struct pf_addr *, u_int16_t *); 194 void pf_attach_state(struct pf_state_key *, 195 struct pf_state *, int); 196 void pf_detach_state(struct pf_state *, int); 197 int pf_test_rule(struct pf_rule **, struct pf_state **, 198 int, struct pfi_kif *, struct mbuf *, int, 199 void *, struct pf_pdesc *, struct pf_rule **, 200 struct pf_ruleset **, struct ifqueue *); 201 int pf_test_fragment(struct pf_rule **, int, 202 struct pfi_kif *, struct mbuf *, void *, 203 struct pf_pdesc *, struct pf_rule **, 204 struct pf_ruleset **); 205 int pf_test_state_tcp(struct pf_state **, int, 206 struct pfi_kif *, struct mbuf *, int, 207 void *, struct pf_pdesc *, u_short *); 208 int pf_test_state_udp(struct pf_state **, int, 209 struct pfi_kif *, struct mbuf *, int, 210 void *, struct pf_pdesc *); 211 int pf_test_state_icmp(struct pf_state **, int, 212 struct pfi_kif *, struct mbuf *, int, 213 void *, struct pf_pdesc *, u_short *); 214 int pf_test_state_other(struct pf_state **, int, 215 struct pfi_kif *, struct pf_pdesc *); 216 int pf_match_tag(struct mbuf *, struct pf_rule *, int *); 217 void pf_step_into_anchor(int *, struct pf_ruleset **, int, 218 struct pf_rule **, struct pf_rule **, int *); 219 int pf_step_out_of_anchor(int *, struct pf_ruleset **, 220 int, struct pf_rule **, struct pf_rule **, 221 int *); 222 void pf_hash(const struct pf_addr *, struct pf_addr *, 223 struct pf_poolhashkey *, sa_family_t); 224 int pf_map_addr(u_int8_t, struct pf_rule *, 225 const struct pf_addr *, struct pf_addr *, 226 struct pf_addr *, struct pf_src_node **); 227 int pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *, 228 struct pf_addr *, struct pf_addr *, u_int16_t, 229 struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t, 230 struct pf_src_node **); 231 void pf_route(struct mbuf **, struct pf_rule *, int, 232 struct ifnet *, struct pf_state *, 233 struct pf_pdesc *); 234 void pf_route6(struct mbuf **, struct pf_rule *, int, 235 struct ifnet *, struct pf_state *, 236 struct pf_pdesc *); 237 int pf_socket_lookup(int, struct pf_pdesc *); 238 u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t, 239 sa_family_t); 240 u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t, 241 sa_family_t); 242 u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t, 243 u_int16_t); 244 void pf_set_rt_ifp(struct pf_state *, 245 struct pf_addr *); 246 #ifdef __NetBSD__ 247 int pf_check_proto_cksum(struct mbuf *, int, int, int, 248 u_int8_t, sa_family_t); 249 #else 250 int pf_check_proto_cksum(struct mbuf *, int, int, 251 u_int8_t, sa_family_t); 252 #endif /* !__NetBSD__ */ 253 int pf_addr_wrap_neq(struct pf_addr_wrap *, 254 struct pf_addr_wrap *); 255 struct pf_state *pf_find_state(struct pfi_kif *, 256 struct pf_state_key_cmp *, u_int8_t); 257 int pf_src_connlimit(struct pf_state **); 258 void pf_stateins_err(const char *, struct pf_state *, 259 struct pfi_kif *); 260 int pf_check_congestion(struct ifqueue *); 261 262 extern struct pool pfr_ktable_pl; 263 extern struct pool pfr_kentry_pl; 264 265 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = { 266 { &pf_state_pl, PFSTATE_HIWAT }, 267 { &pf_src_tree_pl, PFSNODE_HIWAT }, 268 { &pf_frent_pl, PFFRAG_FRENT_HIWAT }, 269 { &pfr_ktable_pl, PFR_KTABLE_HIWAT }, 270 { &pfr_kentry_pl, PFR_KENTRY_HIWAT } 271 }; 272 273 #define STATE_LOOKUP() \ 274 do { \ 275 if (direction == PF_IN) \ 276 *state = pf_find_state(kif, &key, PF_EXT_GWY); \ 277 else \ 278 *state = pf_find_state(kif, &key, PF_LAN_EXT); \ 279 if (*state == NULL || (*state)->timeout == PFTM_PURGE) \ 280 return (PF_DROP); \ 281 if (direction == PF_OUT && \ 282 (((*state)->rule.ptr->rt == PF_ROUTETO && \ 283 (*state)->rule.ptr->direction == PF_OUT) || \ 284 ((*state)->rule.ptr->rt == PF_REPLYTO && \ 285 (*state)->rule.ptr->direction == PF_IN)) && \ 286 (*state)->rt_kif != NULL && \ 287 (*state)->rt_kif != kif) \ 288 return (PF_PASS); \ 289 } while (0) 290 291 #define STATE_TRANSLATE(sk) \ 292 (sk)->lan.addr.addr32[0] != (sk)->gwy.addr.addr32[0] || \ 293 ((sk)->af == AF_INET6 && \ 294 ((sk)->lan.addr.addr32[1] != (sk)->gwy.addr.addr32[1] || \ 295 (sk)->lan.addr.addr32[2] != (sk)->gwy.addr.addr32[2] || \ 296 (sk)->lan.addr.addr32[3] != (sk)->gwy.addr.addr32[3])) || \ 297 (sk)->lan.port != (sk)->gwy.port 298 299 #define BOUND_IFACE(r, k) \ 300 ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all 301 302 #define STATE_INC_COUNTERS(s) \ 303 do { \ 304 s->rule.ptr->states++; \ 305 if (s->anchor.ptr != NULL) \ 306 s->anchor.ptr->states++; \ 307 if (s->nat_rule.ptr != NULL) \ 308 s->nat_rule.ptr->states++; \ 309 } while (0) 310 311 #define STATE_DEC_COUNTERS(s) \ 312 do { \ 313 if (s->nat_rule.ptr != NULL) \ 314 s->nat_rule.ptr->states--; \ 315 if (s->anchor.ptr != NULL) \ 316 s->anchor.ptr->states--; \ 317 s->rule.ptr->states--; \ 318 } while (0) 319 320 static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *); 321 static __inline int pf_state_compare_lan_ext(struct pf_state_key *, 322 struct pf_state_key *); 323 static __inline int pf_state_compare_ext_gwy(struct pf_state_key *, 324 struct pf_state_key *); 325 static __inline int pf_state_compare_id(struct pf_state *, 326 struct pf_state *); 327 328 struct pf_src_tree tree_src_tracking; 329 330 struct pf_state_tree_id tree_id; 331 struct pf_state_queue state_list; 332 333 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare); 334 RB_GENERATE(pf_state_tree_lan_ext, pf_state_key, 335 entry_lan_ext, pf_state_compare_lan_ext); 336 RB_GENERATE(pf_state_tree_ext_gwy, pf_state_key, 337 entry_ext_gwy, pf_state_compare_ext_gwy); 338 RB_GENERATE(pf_state_tree_id, pf_state, 339 entry_id, pf_state_compare_id); 340 341 #define PF_DT_SKIP_LANEXT 0x01 342 #define PF_DT_SKIP_EXTGWY 0x02 343 344 #ifdef __NetBSD__ 345 static __inline struct pfi_kif * 346 bound_iface(const struct pf_rule *r, const struct pf_rule *nr, 347 struct pfi_kif *k) 348 { 349 uint32_t rule_flag; 350 351 rule_flag = r->rule_flag; 352 if (nr != NULL) 353 rule_flag |= nr->rule_flag; 354 355 return ((rule_flag & PFRULE_IFBOUND) != 0) ? k : pfi_all; 356 } 357 #endif /* __NetBSD__ */ 358 359 static __inline int 360 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b) 361 { 362 int diff; 363 364 if (a->rule.ptr > b->rule.ptr) 365 return (1); 366 if (a->rule.ptr < b->rule.ptr) 367 return (-1); 368 if ((diff = a->af - b->af) != 0) 369 return (diff); 370 switch (a->af) { 371 #ifdef INET 372 case AF_INET: 373 if (a->addr.addr32[0] > b->addr.addr32[0]) 374 return (1); 375 if (a->addr.addr32[0] < b->addr.addr32[0]) 376 return (-1); 377 break; 378 #endif /* INET */ 379 #ifdef INET6 380 case AF_INET6: 381 if (a->addr.addr32[3] > b->addr.addr32[3]) 382 return (1); 383 if (a->addr.addr32[3] < b->addr.addr32[3]) 384 return (-1); 385 if (a->addr.addr32[2] > b->addr.addr32[2]) 386 return (1); 387 if (a->addr.addr32[2] < b->addr.addr32[2]) 388 return (-1); 389 if (a->addr.addr32[1] > b->addr.addr32[1]) 390 return (1); 391 if (a->addr.addr32[1] < b->addr.addr32[1]) 392 return (-1); 393 if (a->addr.addr32[0] > b->addr.addr32[0]) 394 return (1); 395 if (a->addr.addr32[0] < b->addr.addr32[0]) 396 return (-1); 397 break; 398 #endif /* INET6 */ 399 } 400 return (0); 401 } 402 403 static __inline int 404 pf_state_compare_lan_ext(struct pf_state_key *a, struct pf_state_key *b) 405 { 406 int diff; 407 408 if ((diff = a->proto - b->proto) != 0) 409 return (diff); 410 if ((diff = a->af - b->af) != 0) 411 return (diff); 412 switch (a->af) { 413 #ifdef INET 414 case AF_INET: 415 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0]) 416 return (1); 417 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0]) 418 return (-1); 419 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 420 return (1); 421 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 422 return (-1); 423 break; 424 #endif /* INET */ 425 #ifdef INET6 426 case AF_INET6: 427 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3]) 428 return (1); 429 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3]) 430 return (-1); 431 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3]) 432 return (1); 433 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3]) 434 return (-1); 435 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2]) 436 return (1); 437 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2]) 438 return (-1); 439 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2]) 440 return (1); 441 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2]) 442 return (-1); 443 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1]) 444 return (1); 445 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1]) 446 return (-1); 447 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1]) 448 return (1); 449 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1]) 450 return (-1); 451 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0]) 452 return (1); 453 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0]) 454 return (-1); 455 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 456 return (1); 457 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 458 return (-1); 459 break; 460 #endif /* INET6 */ 461 } 462 463 if ((diff = a->lan.port - b->lan.port) != 0) 464 return (diff); 465 if ((diff = a->ext.port - b->ext.port) != 0) 466 return (diff); 467 468 return (0); 469 } 470 471 static __inline int 472 pf_state_compare_ext_gwy(struct pf_state_key *a, struct pf_state_key *b) 473 { 474 int diff; 475 476 if ((diff = a->proto - b->proto) != 0) 477 return (diff); 478 if ((diff = a->af - b->af) != 0) 479 return (diff); 480 switch (a->af) { 481 #ifdef INET 482 case AF_INET: 483 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 484 return (1); 485 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 486 return (-1); 487 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0]) 488 return (1); 489 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0]) 490 return (-1); 491 break; 492 #endif /* INET */ 493 #ifdef INET6 494 case AF_INET6: 495 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3]) 496 return (1); 497 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3]) 498 return (-1); 499 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3]) 500 return (1); 501 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3]) 502 return (-1); 503 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2]) 504 return (1); 505 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2]) 506 return (-1); 507 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2]) 508 return (1); 509 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2]) 510 return (-1); 511 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1]) 512 return (1); 513 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1]) 514 return (-1); 515 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1]) 516 return (1); 517 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1]) 518 return (-1); 519 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 520 return (1); 521 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 522 return (-1); 523 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0]) 524 return (1); 525 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0]) 526 return (-1); 527 break; 528 #endif /* INET6 */ 529 } 530 531 if ((diff = a->ext.port - b->ext.port) != 0) 532 return (diff); 533 if ((diff = a->gwy.port - b->gwy.port) != 0) 534 return (diff); 535 536 return (0); 537 } 538 539 static __inline int 540 pf_state_compare_id(struct pf_state *a, struct pf_state *b) 541 { 542 if (a->id > b->id) 543 return (1); 544 if (a->id < b->id) 545 return (-1); 546 if (a->creatorid > b->creatorid) 547 return (1); 548 if (a->creatorid < b->creatorid) 549 return (-1); 550 551 return (0); 552 } 553 554 #ifdef INET6 555 void 556 pf_addrcpy(struct pf_addr *dst, const struct pf_addr *src, sa_family_t af) 557 { 558 switch (af) { 559 #ifdef INET 560 case AF_INET: 561 dst->addr32[0] = src->addr32[0]; 562 break; 563 #endif /* INET */ 564 case AF_INET6: 565 dst->addr32[0] = src->addr32[0]; 566 dst->addr32[1] = src->addr32[1]; 567 dst->addr32[2] = src->addr32[2]; 568 dst->addr32[3] = src->addr32[3]; 569 break; 570 } 571 } 572 #endif /* INET6 */ 573 574 struct pf_state * 575 pf_find_state_byid(struct pf_state_cmp *key) 576 { 577 pf_status.fcounters[FCNT_STATE_SEARCH]++; 578 579 return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key)); 580 } 581 582 struct pf_state * 583 pf_find_state(struct pfi_kif *kif, struct pf_state_key_cmp *key, u_int8_t tree) 584 { 585 struct pf_state_key *sk; 586 struct pf_state *s; 587 588 pf_status.fcounters[FCNT_STATE_SEARCH]++; 589 590 switch (tree) { 591 case PF_LAN_EXT: 592 sk = RB_FIND(pf_state_tree_lan_ext, &pf_statetbl_lan_ext, 593 (struct pf_state_key *)key); 594 break; 595 case PF_EXT_GWY: 596 sk = RB_FIND(pf_state_tree_ext_gwy, &pf_statetbl_ext_gwy, 597 (struct pf_state_key *)key); 598 break; 599 default: 600 panic("pf_find_state"); 601 } 602 603 /* list is sorted, if-bound states before floating ones */ 604 if (sk != NULL) 605 TAILQ_FOREACH(s, &sk->states, next) 606 if (s->kif == pfi_all || s->kif == kif) 607 return (s); 608 609 return (NULL); 610 } 611 612 struct pf_state * 613 pf_find_state_all(struct pf_state_key_cmp *key, u_int8_t tree, int *more) 614 { 615 struct pf_state_key *sk; 616 struct pf_state *s, *ret = NULL; 617 618 pf_status.fcounters[FCNT_STATE_SEARCH]++; 619 620 switch (tree) { 621 case PF_LAN_EXT: 622 sk = RB_FIND(pf_state_tree_lan_ext, 623 &pf_statetbl_lan_ext, (struct pf_state_key *)key); 624 break; 625 case PF_EXT_GWY: 626 sk = RB_FIND(pf_state_tree_ext_gwy, 627 &pf_statetbl_ext_gwy, (struct pf_state_key *)key); 628 break; 629 default: 630 panic("pf_find_state_all"); 631 } 632 633 if (sk != NULL) { 634 ret = TAILQ_FIRST(&sk->states); 635 if (more == NULL) 636 return (ret); 637 638 TAILQ_FOREACH(s, &sk->states, next) 639 (*more)++; 640 } 641 642 return (ret); 643 } 644 645 void 646 pf_init_threshold(struct pf_threshold *threshold, 647 u_int32_t limit, u_int32_t seconds) 648 { 649 threshold->limit = limit * PF_THRESHOLD_MULT; 650 threshold->seconds = seconds; 651 threshold->count = 0; 652 threshold->last = time_second; 653 } 654 655 void 656 pf_add_threshold(struct pf_threshold *threshold) 657 { 658 u_int32_t t = time_second, diff = t - threshold->last; 659 660 if (diff >= threshold->seconds) 661 threshold->count = 0; 662 else 663 threshold->count -= threshold->count * diff / 664 threshold->seconds; 665 threshold->count += PF_THRESHOLD_MULT; 666 threshold->last = t; 667 } 668 669 int 670 pf_check_threshold(struct pf_threshold *threshold) 671 { 672 return (threshold->count > threshold->limit); 673 } 674 675 int 676 pf_src_connlimit(struct pf_state **state) 677 { 678 int bad = 0; 679 680 (*state)->src_node->conn++; 681 (*state)->src.tcp_est = 1; 682 pf_add_threshold(&(*state)->src_node->conn_rate); 683 684 if ((*state)->rule.ptr->max_src_conn && 685 (*state)->rule.ptr->max_src_conn < 686 (*state)->src_node->conn) { 687 pf_status.lcounters[LCNT_SRCCONN]++; 688 bad++; 689 } 690 691 if ((*state)->rule.ptr->max_src_conn_rate.limit && 692 pf_check_threshold(&(*state)->src_node->conn_rate)) { 693 pf_status.lcounters[LCNT_SRCCONNRATE]++; 694 bad++; 695 } 696 697 if (!bad) 698 return (0); 699 700 if ((*state)->rule.ptr->overload_tbl) { 701 struct pfr_addr p; 702 u_int32_t killed = 0; 703 704 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++; 705 if (pf_status.debug >= PF_DEBUG_MISC) { 706 printf("pf_src_connlimit: blocking address "); 707 pf_print_host(&(*state)->src_node->addr, 0, 708 (*state)->state_key->af); 709 } 710 711 bzero(&p, sizeof(p)); 712 p.pfra_af = (*state)->state_key->af; 713 switch ((*state)->state_key->af) { 714 #ifdef INET 715 case AF_INET: 716 p.pfra_net = 32; 717 p.pfra_ip4addr = (*state)->src_node->addr.v4; 718 break; 719 #endif /* INET */ 720 #ifdef INET6 721 case AF_INET6: 722 p.pfra_net = 128; 723 p.pfra_ip6addr = (*state)->src_node->addr.v6; 724 break; 725 #endif /* INET6 */ 726 } 727 728 pfr_insert_kentry((*state)->rule.ptr->overload_tbl, 729 &p, time_second); 730 731 /* kill existing states if that's required. */ 732 if ((*state)->rule.ptr->flush) { 733 struct pf_state_key *sk; 734 struct pf_state *st; 735 736 pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++; 737 RB_FOREACH(st, pf_state_tree_id, &tree_id) { 738 sk = st->state_key; 739 /* 740 * Kill states from this source. (Only those 741 * from the same rule if PF_FLUSH_GLOBAL is not 742 * set) 743 */ 744 if (sk->af == 745 (*state)->state_key->af && 746 (((*state)->state_key->direction == 747 PF_OUT && 748 PF_AEQ(&(*state)->src_node->addr, 749 &sk->lan.addr, sk->af)) || 750 ((*state)->state_key->direction == PF_IN && 751 PF_AEQ(&(*state)->src_node->addr, 752 &sk->ext.addr, sk->af))) && 753 ((*state)->rule.ptr->flush & 754 PF_FLUSH_GLOBAL || 755 (*state)->rule.ptr == st->rule.ptr)) { 756 st->timeout = PFTM_PURGE; 757 st->src.state = st->dst.state = 758 TCPS_CLOSED; 759 killed++; 760 } 761 } 762 if (pf_status.debug >= PF_DEBUG_MISC) 763 printf(", %u states killed", killed); 764 } 765 if (pf_status.debug >= PF_DEBUG_MISC) 766 printf("\n"); 767 } 768 769 /* kill this state */ 770 (*state)->timeout = PFTM_PURGE; 771 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED; 772 return (1); 773 } 774 775 int 776 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule, 777 struct pf_addr *src, sa_family_t af) 778 { 779 struct pf_src_node k; 780 781 if (*sn == NULL) { 782 k.af = af; 783 PF_ACPY(&k.addr, src, af); 784 if (rule->rule_flag & PFRULE_RULESRCTRACK || 785 rule->rpool.opts & PF_POOL_STICKYADDR) 786 k.rule.ptr = rule; 787 else 788 k.rule.ptr = NULL; 789 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++; 790 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k); 791 } 792 if (*sn == NULL) { 793 if (!rule->max_src_nodes || 794 rule->src_nodes < rule->max_src_nodes) 795 (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT); 796 else 797 pf_status.lcounters[LCNT_SRCNODES]++; 798 if ((*sn) == NULL) 799 return (-1); 800 bzero(*sn, sizeof(struct pf_src_node)); 801 802 pf_init_threshold(&(*sn)->conn_rate, 803 rule->max_src_conn_rate.limit, 804 rule->max_src_conn_rate.seconds); 805 806 (*sn)->af = af; 807 if (rule->rule_flag & PFRULE_RULESRCTRACK || 808 rule->rpool.opts & PF_POOL_STICKYADDR) 809 (*sn)->rule.ptr = rule; 810 else 811 (*sn)->rule.ptr = NULL; 812 PF_ACPY(&(*sn)->addr, src, af); 813 if (RB_INSERT(pf_src_tree, 814 &tree_src_tracking, *sn) != NULL) { 815 if (pf_status.debug >= PF_DEBUG_MISC) { 816 printf("pf: src_tree insert failed: "); 817 pf_print_host(&(*sn)->addr, 0, af); 818 printf("\n"); 819 } 820 pool_put(&pf_src_tree_pl, *sn); 821 return (-1); 822 } 823 (*sn)->creation = time_second; 824 (*sn)->ruletype = rule->action; 825 if ((*sn)->rule.ptr != NULL) 826 (*sn)->rule.ptr->src_nodes++; 827 pf_status.scounters[SCNT_SRC_NODE_INSERT]++; 828 pf_status.src_nodes++; 829 } else { 830 if (rule->max_src_states && 831 (*sn)->states >= rule->max_src_states) { 832 pf_status.lcounters[LCNT_SRCSTATES]++; 833 return (-1); 834 } 835 } 836 return (0); 837 } 838 839 void 840 pf_stateins_err(const char *tree, struct pf_state *s, struct pfi_kif *kif) 841 { 842 struct pf_state_key *sk = s->state_key; 843 844 if (pf_status.debug >= PF_DEBUG_MISC) { 845 printf("pf: state insert failed: %s %s", tree, kif->pfik_name); 846 printf(" lan: "); 847 pf_print_host(&sk->lan.addr, sk->lan.port, 848 sk->af); 849 printf(" gwy: "); 850 pf_print_host(&sk->gwy.addr, sk->gwy.port, 851 sk->af); 852 printf(" ext: "); 853 pf_print_host(&sk->ext.addr, sk->ext.port, 854 sk->af); 855 if (s->sync_flags & PFSTATE_FROMSYNC) 856 printf(" (from sync)"); 857 printf("\n"); 858 } 859 } 860 861 int 862 pf_insert_state(struct pfi_kif *kif, struct pf_state *s) 863 { 864 struct pf_state_key *cur; 865 struct pf_state *sp; 866 867 KASSERT(s->state_key != NULL); 868 s->kif = kif; 869 870 if ((cur = RB_INSERT(pf_state_tree_lan_ext, &pf_statetbl_lan_ext, 871 s->state_key)) != NULL) { 872 /* key exists. check for same kif, if none, add to key */ 873 TAILQ_FOREACH(sp, &cur->states, next) 874 if (sp->kif == kif) { /* collision! */ 875 pf_stateins_err("tree_lan_ext", s, kif); 876 pf_detach_state(s, 877 PF_DT_SKIP_LANEXT|PF_DT_SKIP_EXTGWY); 878 return (-1); 879 } 880 pf_detach_state(s, PF_DT_SKIP_LANEXT|PF_DT_SKIP_EXTGWY); 881 pf_attach_state(cur, s, kif == pfi_all ? 1 : 0); 882 } 883 884 /* if cur != NULL, we already found a state key and attached to it */ 885 if (cur == NULL && (cur = RB_INSERT(pf_state_tree_ext_gwy, 886 &pf_statetbl_ext_gwy, s->state_key)) != NULL) { 887 /* must not happen. we must have found the sk above! */ 888 pf_stateins_err("tree_ext_gwy", s, kif); 889 pf_detach_state(s, PF_DT_SKIP_EXTGWY); 890 return (-1); 891 } 892 893 if (s->id == 0 && s->creatorid == 0) { 894 s->id = htobe64(pf_status.stateid++); 895 s->creatorid = pf_status.hostid; 896 } 897 if (RB_INSERT(pf_state_tree_id, &tree_id, s) != NULL) { 898 if (pf_status.debug >= PF_DEBUG_MISC) { 899 #ifdef __NetBSD__ 900 printf("pf: state insert failed: " 901 "id: %016" PRIx64 " creatorid: %08x", 902 be64toh(s->id), ntohl(s->creatorid)); 903 #else 904 printf("pf: state insert failed: " 905 "id: %016llx creatorid: %08x", 906 betoh64(s->id), ntohl(s->creatorid)); 907 #endif /* !__NetBSD__ */ 908 if (s->sync_flags & PFSTATE_FROMSYNC) 909 printf(" (from sync)"); 910 printf("\n"); 911 } 912 pf_detach_state(s, 0); 913 return (-1); 914 } 915 TAILQ_INSERT_TAIL(&state_list, s, entry_list); 916 pf_status.fcounters[FCNT_STATE_INSERT]++; 917 pf_status.states++; 918 pfi_kif_ref(kif, PFI_KIF_REF_STATE); 919 #if NPFSYNC 920 pfsync_insert_state(s); 921 #endif 922 return (0); 923 } 924 925 #ifdef _LKM 926 volatile int pf_purge_thread_stop; 927 volatile int pf_purge_thread_running; 928 #endif 929 930 void 931 pf_purge_thread(void *v) 932 { 933 int nloops = 0, s; 934 935 #ifdef _LKM 936 pf_purge_thread_running = 1; 937 pf_purge_thread_stop = 0; 938 939 while (!pf_purge_thread_stop) { 940 #else 941 for (;;) { 942 #endif /* !_LKM */ 943 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz); 944 945 s = splsoftnet(); 946 947 /* process a fraction of the state table every second */ 948 pf_purge_expired_states(1 + (pf_status.states 949 / pf_default_rule.timeout[PFTM_INTERVAL])); 950 951 /* purge other expired types every PFTM_INTERVAL seconds */ 952 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) { 953 pf_purge_expired_fragments(); 954 pf_purge_expired_src_nodes(0); 955 nloops = 0; 956 } 957 958 splx(s); 959 } 960 961 #ifdef _LKM 962 pf_purge_thread_running = 0; 963 wakeup(&pf_purge_thread_running); 964 kthread_exit(0); 965 #endif /* _LKM */ 966 } 967 968 u_int32_t 969 pf_state_expires(const struct pf_state *state) 970 { 971 u_int32_t timeout; 972 u_int32_t start; 973 u_int32_t end; 974 u_int32_t states; 975 976 /* handle all PFTM_* > PFTM_MAX here */ 977 if (state->timeout == PFTM_PURGE) 978 return (time_second); 979 if (state->timeout == PFTM_UNTIL_PACKET) 980 return (0); 981 KASSERT(state->timeout != PFTM_UNLINKED); 982 KASSERT(state->timeout < PFTM_MAX); 983 timeout = state->rule.ptr->timeout[state->timeout]; 984 if (!timeout) 985 timeout = pf_default_rule.timeout[state->timeout]; 986 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START]; 987 if (start) { 988 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END]; 989 states = state->rule.ptr->states; 990 } else { 991 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START]; 992 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END]; 993 states = pf_status.states; 994 } 995 if (end && states > start && start < end) { 996 if (states < end) 997 return (state->expire + timeout * (end - states) / 998 (end - start)); 999 else 1000 return (time_second); 1001 } 1002 return (state->expire + timeout); 1003 } 1004 1005 void 1006 pf_purge_expired_src_nodes(int waslocked) 1007 { 1008 struct pf_src_node *cur, *next; 1009 int locked = waslocked; 1010 1011 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) { 1012 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur); 1013 1014 if (cur->states <= 0 && cur->expire <= time_second) { 1015 if (! locked) { 1016 rw_enter_write(&pf_consistency_lock); 1017 next = RB_NEXT(pf_src_tree, 1018 &tree_src_tracking, cur); 1019 locked = 1; 1020 } 1021 if (cur->rule.ptr != NULL) { 1022 cur->rule.ptr->src_nodes--; 1023 if (cur->rule.ptr->states <= 0 && 1024 cur->rule.ptr->max_src_nodes <= 0) 1025 pf_rm_rule(NULL, cur->rule.ptr); 1026 } 1027 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur); 1028 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 1029 pf_status.src_nodes--; 1030 pool_put(&pf_src_tree_pl, cur); 1031 } 1032 } 1033 1034 if (locked && !waslocked) 1035 rw_exit_write(&pf_consistency_lock); 1036 } 1037 1038 void 1039 pf_src_tree_remove_state(struct pf_state *s) 1040 { 1041 u_int32_t timeout; 1042 1043 if (s->src_node != NULL) { 1044 if (s->src.tcp_est) 1045 --s->src_node->conn; 1046 if (--s->src_node->states <= 0) { 1047 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE]; 1048 if (!timeout) 1049 timeout = 1050 pf_default_rule.timeout[PFTM_SRC_NODE]; 1051 s->src_node->expire = time_second + timeout; 1052 } 1053 } 1054 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) { 1055 if (--s->nat_src_node->states <= 0) { 1056 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE]; 1057 if (!timeout) 1058 timeout = 1059 pf_default_rule.timeout[PFTM_SRC_NODE]; 1060 s->nat_src_node->expire = time_second + timeout; 1061 } 1062 } 1063 s->src_node = s->nat_src_node = NULL; 1064 } 1065 1066 /* callers should be at splsoftnet */ 1067 void 1068 pf_unlink_state(struct pf_state *cur) 1069 { 1070 if (cur->src.state == PF_TCPS_PROXY_DST) { 1071 pf_send_tcp(cur->rule.ptr, cur->state_key->af, 1072 &cur->state_key->ext.addr, &cur->state_key->lan.addr, 1073 cur->state_key->ext.port, cur->state_key->lan.port, 1074 cur->src.seqhi, cur->src.seqlo + 1, 1075 TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL); 1076 } 1077 RB_REMOVE(pf_state_tree_id, &tree_id, cur); 1078 #if NPFSYNC 1079 if (cur->creatorid == pf_status.hostid) 1080 pfsync_delete_state(cur); 1081 #endif 1082 cur->timeout = PFTM_UNLINKED; 1083 pf_src_tree_remove_state(cur); 1084 pf_detach_state(cur, 0); 1085 } 1086 1087 /* callers should be at splsoftnet and hold the 1088 * write_lock on pf_consistency_lock */ 1089 void 1090 pf_free_state(struct pf_state *cur) 1091 { 1092 #if NPFSYNC 1093 if (pfsyncif != NULL && 1094 (pfsyncif->sc_bulk_send_next == cur || 1095 pfsyncif->sc_bulk_terminator == cur)) 1096 return; 1097 #endif 1098 KASSERT(cur->timeout == PFTM_UNLINKED); 1099 if (--cur->rule.ptr->states <= 0 && 1100 cur->rule.ptr->src_nodes <= 0) 1101 pf_rm_rule(NULL, cur->rule.ptr); 1102 if (cur->nat_rule.ptr != NULL) 1103 if (--cur->nat_rule.ptr->states <= 0 && 1104 cur->nat_rule.ptr->src_nodes <= 0) 1105 pf_rm_rule(NULL, cur->nat_rule.ptr); 1106 if (cur->anchor.ptr != NULL) 1107 if (--cur->anchor.ptr->states <= 0) 1108 pf_rm_rule(NULL, cur->anchor.ptr); 1109 pf_normalize_tcp_cleanup(cur); 1110 pfi_kif_unref(cur->kif, PFI_KIF_REF_STATE); 1111 TAILQ_REMOVE(&state_list, cur, entry_list); 1112 if (cur->tag) 1113 pf_tag_unref(cur->tag); 1114 pool_put(&pf_state_pl, cur); 1115 pf_status.fcounters[FCNT_STATE_REMOVALS]++; 1116 pf_status.states--; 1117 } 1118 1119 void 1120 pf_purge_expired_states(u_int32_t maxcheck) 1121 { 1122 static struct pf_state *cur = NULL; 1123 struct pf_state *next; 1124 int locked = 0; 1125 1126 while (maxcheck--) { 1127 /* wrap to start of list when we hit the end */ 1128 if (cur == NULL) { 1129 cur = TAILQ_FIRST(&state_list); 1130 if (cur == NULL) 1131 break; /* list empty */ 1132 } 1133 1134 /* get next state, as cur may get deleted */ 1135 next = TAILQ_NEXT(cur, entry_list); 1136 1137 if (cur->timeout == PFTM_UNLINKED) { 1138 /* free unlinked state */ 1139 if (! locked) { 1140 rw_enter_write(&pf_consistency_lock); 1141 locked = 1; 1142 } 1143 pf_free_state(cur); 1144 } else if (pf_state_expires(cur) <= time_second) { 1145 /* unlink and free expired state */ 1146 pf_unlink_state(cur); 1147 if (! locked) { 1148 rw_enter_write(&pf_consistency_lock); 1149 locked = 1; 1150 } 1151 pf_free_state(cur); 1152 } 1153 cur = next; 1154 } 1155 1156 if (locked) 1157 rw_exit_write(&pf_consistency_lock); 1158 } 1159 1160 int 1161 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw) 1162 { 1163 if (aw->type != PF_ADDR_TABLE) 1164 return (0); 1165 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL) 1166 return (1); 1167 return (0); 1168 } 1169 1170 void 1171 pf_tbladdr_remove(struct pf_addr_wrap *aw) 1172 { 1173 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL) 1174 return; 1175 pfr_detach_table(aw->p.tbl); 1176 aw->p.tbl = NULL; 1177 } 1178 1179 void 1180 pf_tbladdr_copyout(struct pf_addr_wrap *aw) 1181 { 1182 struct pfr_ktable *kt = aw->p.tbl; 1183 1184 if (aw->type != PF_ADDR_TABLE || kt == NULL) 1185 return; 1186 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) 1187 kt = kt->pfrkt_root; 1188 aw->p.tbl = NULL; 1189 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ? 1190 kt->pfrkt_cnt : -1; 1191 } 1192 1193 void 1194 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af) 1195 { 1196 switch (af) { 1197 #ifdef INET 1198 case AF_INET: { 1199 u_int32_t a = ntohl(addr->addr32[0]); 1200 printf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255, 1201 (a>>8)&255, a&255); 1202 if (p) { 1203 p = ntohs(p); 1204 printf(":%u", p); 1205 } 1206 break; 1207 } 1208 #endif /* INET */ 1209 #ifdef INET6 1210 case AF_INET6: { 1211 u_int16_t b; 1212 u_int8_t i, curstart = 255, curend = 0, 1213 maxstart = 0, maxend = 0; 1214 for (i = 0; i < 8; i++) { 1215 if (!addr->addr16[i]) { 1216 if (curstart == 255) 1217 curstart = i; 1218 else 1219 curend = i; 1220 } else { 1221 if (curstart) { 1222 if ((curend - curstart) > 1223 (maxend - maxstart)) { 1224 maxstart = curstart; 1225 maxend = curend; 1226 curstart = 255; 1227 } 1228 } 1229 } 1230 } 1231 for (i = 0; i < 8; i++) { 1232 if (i >= maxstart && i <= maxend) { 1233 if (maxend != 7) { 1234 if (i == maxstart) 1235 printf(":"); 1236 } else { 1237 if (i == maxend) 1238 printf(":"); 1239 } 1240 } else { 1241 b = ntohs(addr->addr16[i]); 1242 printf("%x", b); 1243 if (i < 7) 1244 printf(":"); 1245 } 1246 } 1247 if (p) { 1248 p = ntohs(p); 1249 printf("[%u]", p); 1250 } 1251 break; 1252 } 1253 #endif /* INET6 */ 1254 } 1255 } 1256 1257 void 1258 pf_print_state(struct pf_state *s) 1259 { 1260 struct pf_state_key *sk = s->state_key; 1261 switch (sk->proto) { 1262 case IPPROTO_TCP: 1263 printf("TCP "); 1264 break; 1265 case IPPROTO_UDP: 1266 printf("UDP "); 1267 break; 1268 case IPPROTO_ICMP: 1269 printf("ICMP "); 1270 break; 1271 case IPPROTO_ICMPV6: 1272 printf("ICMPV6 "); 1273 break; 1274 default: 1275 printf("%u ", sk->proto); 1276 break; 1277 } 1278 pf_print_host(&sk->lan.addr, sk->lan.port, sk->af); 1279 printf(" "); 1280 pf_print_host(&sk->gwy.addr, sk->gwy.port, sk->af); 1281 printf(" "); 1282 pf_print_host(&sk->ext.addr, sk->ext.port, sk->af); 1283 printf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo, 1284 s->src.seqhi, s->src.max_win, s->src.seqdiff); 1285 if (s->src.wscale && s->dst.wscale) 1286 printf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK); 1287 printf("]"); 1288 printf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo, 1289 s->dst.seqhi, s->dst.max_win, s->dst.seqdiff); 1290 if (s->src.wscale && s->dst.wscale) 1291 printf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK); 1292 printf("]"); 1293 printf(" %u:%u", s->src.state, s->dst.state); 1294 } 1295 1296 void 1297 pf_print_flags(u_int8_t f) 1298 { 1299 if (f) 1300 printf(" "); 1301 if (f & TH_FIN) 1302 printf("F"); 1303 if (f & TH_SYN) 1304 printf("S"); 1305 if (f & TH_RST) 1306 printf("R"); 1307 if (f & TH_PUSH) 1308 printf("P"); 1309 if (f & TH_ACK) 1310 printf("A"); 1311 if (f & TH_URG) 1312 printf("U"); 1313 if (f & TH_ECE) 1314 printf("E"); 1315 if (f & TH_CWR) 1316 printf("W"); 1317 } 1318 1319 #define PF_SET_SKIP_STEPS(i) \ 1320 do { \ 1321 while (head[i] != cur) { \ 1322 head[i]->skip[i].ptr = cur; \ 1323 head[i] = TAILQ_NEXT(head[i], entries); \ 1324 } \ 1325 } while (0) 1326 1327 void 1328 pf_calc_skip_steps(struct pf_rulequeue *rules) 1329 { 1330 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT]; 1331 int i; 1332 1333 cur = TAILQ_FIRST(rules); 1334 prev = cur; 1335 for (i = 0; i < PF_SKIP_COUNT; ++i) 1336 head[i] = cur; 1337 while (cur != NULL) { 1338 1339 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot) 1340 PF_SET_SKIP_STEPS(PF_SKIP_IFP); 1341 if (cur->direction != prev->direction) 1342 PF_SET_SKIP_STEPS(PF_SKIP_DIR); 1343 if (cur->af != prev->af) 1344 PF_SET_SKIP_STEPS(PF_SKIP_AF); 1345 if (cur->proto != prev->proto) 1346 PF_SET_SKIP_STEPS(PF_SKIP_PROTO); 1347 if (cur->src.neg != prev->src.neg || 1348 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr)) 1349 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR); 1350 if (cur->src.port[0] != prev->src.port[0] || 1351 cur->src.port[1] != prev->src.port[1] || 1352 cur->src.port_op != prev->src.port_op) 1353 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT); 1354 if (cur->dst.neg != prev->dst.neg || 1355 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr)) 1356 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR); 1357 if (cur->dst.port[0] != prev->dst.port[0] || 1358 cur->dst.port[1] != prev->dst.port[1] || 1359 cur->dst.port_op != prev->dst.port_op) 1360 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT); 1361 1362 prev = cur; 1363 cur = TAILQ_NEXT(cur, entries); 1364 } 1365 for (i = 0; i < PF_SKIP_COUNT; ++i) 1366 PF_SET_SKIP_STEPS(i); 1367 } 1368 1369 int 1370 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2) 1371 { 1372 if (aw1->type != aw2->type) 1373 return (1); 1374 switch (aw1->type) { 1375 case PF_ADDR_ADDRMASK: 1376 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0)) 1377 return (1); 1378 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0)) 1379 return (1); 1380 return (0); 1381 case PF_ADDR_DYNIFTL: 1382 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt); 1383 case PF_ADDR_NOROUTE: 1384 case PF_ADDR_URPFFAILED: 1385 return (0); 1386 case PF_ADDR_TABLE: 1387 return (aw1->p.tbl != aw2->p.tbl); 1388 case PF_ADDR_RTLABEL: 1389 return (aw1->v.rtlabel != aw2->v.rtlabel); 1390 default: 1391 printf("invalid address type: %d\n", aw1->type); 1392 return (1); 1393 } 1394 } 1395 1396 u_int16_t 1397 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp) 1398 { 1399 u_int32_t l; 1400 1401 if (udp && !cksum) 1402 return (0x0000); 1403 l = cksum + old - new; 1404 l = (l >> 16) + (l & 65535); 1405 l = l & 65535; 1406 if (udp && !l) 1407 return (0xFFFF); 1408 return (l); 1409 } 1410 1411 void 1412 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc, 1413 struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af) 1414 { 1415 struct pf_addr ao; 1416 u_int16_t po = *p; 1417 1418 PF_ACPY(&ao, a, af); 1419 PF_ACPY(a, an, af); 1420 1421 *p = pn; 1422 1423 switch (af) { 1424 #ifdef INET 1425 case AF_INET: 1426 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic, 1427 ao.addr16[0], an->addr16[0], 0), 1428 ao.addr16[1], an->addr16[1], 0); 1429 *p = pn; 1430 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc, 1431 ao.addr16[0], an->addr16[0], u), 1432 ao.addr16[1], an->addr16[1], u), 1433 po, pn, u); 1434 break; 1435 #endif /* INET */ 1436 #ifdef INET6 1437 case AF_INET6: 1438 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1439 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1440 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc, 1441 ao.addr16[0], an->addr16[0], u), 1442 ao.addr16[1], an->addr16[1], u), 1443 ao.addr16[2], an->addr16[2], u), 1444 ao.addr16[3], an->addr16[3], u), 1445 ao.addr16[4], an->addr16[4], u), 1446 ao.addr16[5], an->addr16[5], u), 1447 ao.addr16[6], an->addr16[6], u), 1448 ao.addr16[7], an->addr16[7], u), 1449 po, pn, u); 1450 break; 1451 #endif /* INET6 */ 1452 } 1453 } 1454 1455 1456 /* Changes a u_int32_t. Uses a void *so there are no align restrictions */ 1457 void 1458 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u) 1459 { 1460 u_int32_t ao; 1461 1462 memcpy(&ao, a, sizeof(ao)); 1463 memcpy(a, &an, sizeof(u_int32_t)); 1464 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u), 1465 ao % 65536, an % 65536, u); 1466 } 1467 1468 #ifdef INET6 1469 void 1470 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u) 1471 { 1472 struct pf_addr ao; 1473 1474 PF_ACPY(&ao, a, AF_INET6); 1475 PF_ACPY(a, an, AF_INET6); 1476 1477 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1478 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1479 pf_cksum_fixup(pf_cksum_fixup(*c, 1480 ao.addr16[0], an->addr16[0], u), 1481 ao.addr16[1], an->addr16[1], u), 1482 ao.addr16[2], an->addr16[2], u), 1483 ao.addr16[3], an->addr16[3], u), 1484 ao.addr16[4], an->addr16[4], u), 1485 ao.addr16[5], an->addr16[5], u), 1486 ao.addr16[6], an->addr16[6], u), 1487 ao.addr16[7], an->addr16[7], u); 1488 } 1489 #endif /* INET6 */ 1490 1491 void 1492 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa, 1493 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c, 1494 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af) 1495 { 1496 struct pf_addr oia, ooa; 1497 1498 PF_ACPY(&oia, ia, af); 1499 PF_ACPY(&ooa, oa, af); 1500 1501 /* Change inner protocol port, fix inner protocol checksum. */ 1502 if (ip != NULL) { 1503 u_int16_t oip = *ip; 1504 u_int32_t opc = 0; 1505 1506 if (pc != NULL) 1507 opc = *pc; 1508 *ip = np; 1509 if (pc != NULL) 1510 *pc = pf_cksum_fixup(*pc, oip, *ip, u); 1511 *ic = pf_cksum_fixup(*ic, oip, *ip, 0); 1512 if (pc != NULL) 1513 *ic = pf_cksum_fixup(*ic, opc, *pc, 0); 1514 } 1515 /* Change inner ip address, fix inner ip and icmp checksums. */ 1516 PF_ACPY(ia, na, af); 1517 switch (af) { 1518 #ifdef INET 1519 case AF_INET: { 1520 u_int32_t oh2c = *h2c; 1521 1522 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c, 1523 oia.addr16[0], ia->addr16[0], 0), 1524 oia.addr16[1], ia->addr16[1], 0); 1525 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic, 1526 oia.addr16[0], ia->addr16[0], 0), 1527 oia.addr16[1], ia->addr16[1], 0); 1528 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0); 1529 break; 1530 } 1531 #endif /* INET */ 1532 #ifdef INET6 1533 case AF_INET6: 1534 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1535 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1536 pf_cksum_fixup(pf_cksum_fixup(*ic, 1537 oia.addr16[0], ia->addr16[0], u), 1538 oia.addr16[1], ia->addr16[1], u), 1539 oia.addr16[2], ia->addr16[2], u), 1540 oia.addr16[3], ia->addr16[3], u), 1541 oia.addr16[4], ia->addr16[4], u), 1542 oia.addr16[5], ia->addr16[5], u), 1543 oia.addr16[6], ia->addr16[6], u), 1544 oia.addr16[7], ia->addr16[7], u); 1545 break; 1546 #endif /* INET6 */ 1547 } 1548 /* Change outer ip address, fix outer ip or icmpv6 checksum. */ 1549 PF_ACPY(oa, na, af); 1550 switch (af) { 1551 #ifdef INET 1552 case AF_INET: 1553 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc, 1554 ooa.addr16[0], oa->addr16[0], 0), 1555 ooa.addr16[1], oa->addr16[1], 0); 1556 break; 1557 #endif /* INET */ 1558 #ifdef INET6 1559 case AF_INET6: 1560 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1561 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1562 pf_cksum_fixup(pf_cksum_fixup(*ic, 1563 ooa.addr16[0], oa->addr16[0], u), 1564 ooa.addr16[1], oa->addr16[1], u), 1565 ooa.addr16[2], oa->addr16[2], u), 1566 ooa.addr16[3], oa->addr16[3], u), 1567 ooa.addr16[4], oa->addr16[4], u), 1568 ooa.addr16[5], oa->addr16[5], u), 1569 ooa.addr16[6], oa->addr16[6], u), 1570 ooa.addr16[7], oa->addr16[7], u); 1571 break; 1572 #endif /* INET6 */ 1573 } 1574 } 1575 1576 1577 /* 1578 * Need to modulate the sequence numbers in the TCP SACK option 1579 * (credits to Krzysztof Pfaff for report and patch) 1580 */ 1581 int 1582 pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd, 1583 struct tcphdr *th, struct pf_state_peer *dst) 1584 { 1585 int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen; 1586 u_int8_t opts[MAX_TCPOPTLEN], *opt = opts; 1587 int copyback = 0, i, olen; 1588 struct sackblk sack; 1589 1590 #ifdef __NetBSD__ 1591 #define TCPOLEN_SACK (2 * sizeof(uint32_t)) 1592 #endif 1593 1594 #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2) 1595 if (hlen < TCPOLEN_SACKLEN || 1596 !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af)) 1597 return 0; 1598 1599 while (hlen >= TCPOLEN_SACKLEN) { 1600 olen = opt[1]; 1601 switch (*opt) { 1602 case TCPOPT_EOL: /* FALLTHROUGH */ 1603 case TCPOPT_NOP: 1604 opt++; 1605 hlen--; 1606 break; 1607 case TCPOPT_SACK: 1608 if (olen > hlen) 1609 olen = hlen; 1610 if (olen >= TCPOLEN_SACKLEN) { 1611 for (i = 2; i + TCPOLEN_SACK <= olen; 1612 i += TCPOLEN_SACK) { 1613 memcpy(&sack, &opt[i], sizeof(sack)); 1614 #ifdef __NetBSD__ 1615 #define SACK_START sack.left 1616 #define SACK_END sack.right 1617 #else 1618 #define SACK_START sack.start 1619 #define SACK_END sack.end 1620 #endif 1621 pf_change_a(&SACK_START, &th->th_sum, 1622 htonl(ntohl(SACK_START) - 1623 dst->seqdiff), 0); 1624 pf_change_a(&SACK_END, &th->th_sum, 1625 htonl(ntohl(SACK_END) - 1626 dst->seqdiff), 0); 1627 #undef SACK_START 1628 #undef SACK_END 1629 memcpy(&opt[i], &sack, sizeof(sack)); 1630 } 1631 copyback = 1; 1632 } 1633 /* FALLTHROUGH */ 1634 default: 1635 if (olen < 2) 1636 olen = 2; 1637 hlen -= olen; 1638 opt += olen; 1639 } 1640 } 1641 1642 if (copyback) 1643 m_copyback(m, off + sizeof(*th), thoptlen, opts); 1644 return (copyback); 1645 } 1646 1647 void 1648 pf_send_tcp(const struct pf_rule *r, sa_family_t af, 1649 const struct pf_addr *saddr, const struct pf_addr *daddr, 1650 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack, 1651 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag, 1652 u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp) 1653 { 1654 struct mbuf *m; 1655 int len, tlen; 1656 #ifdef INET 1657 struct ip *h = NULL; 1658 #endif /* INET */ 1659 #ifdef INET6 1660 struct ip6_hdr *h6 = NULL; 1661 #endif /* INET6 */ 1662 struct tcphdr *th; 1663 char *opt; 1664 #ifdef __NetBSD__ 1665 struct pf_mtag *pf_mtag; 1666 #endif /* __NetBSD__ */ 1667 1668 /* maximum segment size tcp option */ 1669 tlen = sizeof(struct tcphdr); 1670 if (mss) 1671 tlen += 4; 1672 1673 switch (af) { 1674 #ifdef INET 1675 case AF_INET: 1676 len = sizeof(struct ip) + tlen; 1677 break; 1678 #endif /* INET */ 1679 #ifdef INET6 1680 case AF_INET6: 1681 len = sizeof(struct ip6_hdr) + tlen; 1682 break; 1683 #endif /* INET6 */ 1684 default: 1685 return; 1686 } 1687 1688 /* create outgoing mbuf */ 1689 m = m_gethdr(M_DONTWAIT, MT_HEADER); 1690 if (m == NULL) 1691 return; 1692 #ifdef __NetBSD__ 1693 if ((pf_mtag = pf_get_mtag(m)) == NULL) { 1694 m_freem(m); 1695 return; 1696 } 1697 if (tag) 1698 pf_mtag->flags |= PF_TAG_GENERATED; 1699 pf_mtag->tag = rtag; 1700 1701 if (r != NULL && r->rtableid >= 0) 1702 pf_mtag->rtableid = r->rtableid; 1703 #else 1704 if (tag) 1705 m->m_pkthdr.pf.flags |= PF_TAG_GENERATED; 1706 m->m_pkthdr.pf.tag = rtag; 1707 1708 if (r != NULL && r->rtableid >= 0) 1709 m->m_pkthdr.pf.rtableid = m->m_pkthdr.pf.rtableid; 1710 #endif /* !__NetBSD__ */ 1711 1712 #ifdef ALTQ 1713 if (r != NULL && r->qid) { 1714 #ifdef __NetBSD__ 1715 struct m_tag *mtag; 1716 struct altq_tag *atag; 1717 1718 mtag = m_tag_get(PACKET_TAG_ALTQ_QID, sizeof(*atag), M_NOWAIT); 1719 if (mtag != NULL) { 1720 atag = (struct altq_tag *)(mtag + 1); 1721 atag->qid = r->qid; 1722 /* add hints for ecn */ 1723 atag->af = af; 1724 atag->hdr = mtod(m, struct ip *); 1725 m_tag_prepend(m, mtag); 1726 } 1727 #else 1728 m->m_pkthdr.pf.qid = r->qid; 1729 /* add hints for ecn */ 1730 m->m_pkthdr.pf.hdr = mtod(m, struct ip *); 1731 #endif /* !__NetBSD__ */ 1732 } 1733 #endif /* ALTQ */ 1734 m->m_data += max_linkhdr; 1735 m->m_pkthdr.len = m->m_len = len; 1736 m->m_pkthdr.rcvif = NULL; 1737 bzero(m->m_data, len); 1738 switch (af) { 1739 #ifdef INET 1740 case AF_INET: 1741 h = mtod(m, struct ip *); 1742 1743 /* IP header fields included in the TCP checksum */ 1744 h->ip_p = IPPROTO_TCP; 1745 h->ip_len = htons(tlen); 1746 h->ip_src.s_addr = saddr->v4.s_addr; 1747 h->ip_dst.s_addr = daddr->v4.s_addr; 1748 1749 th = (struct tcphdr *)((char *)h + sizeof(struct ip)); 1750 break; 1751 #endif /* INET */ 1752 #ifdef INET6 1753 case AF_INET6: 1754 h6 = mtod(m, struct ip6_hdr *); 1755 1756 /* IP header fields included in the TCP checksum */ 1757 h6->ip6_nxt = IPPROTO_TCP; 1758 h6->ip6_plen = htons(tlen); 1759 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr)); 1760 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr)); 1761 1762 th = (struct tcphdr *)((char *)h6 + sizeof(struct ip6_hdr)); 1763 break; 1764 #endif /* INET6 */ 1765 default: 1766 m_freem(m); 1767 return; 1768 } 1769 1770 /* TCP header */ 1771 th->th_sport = sport; 1772 th->th_dport = dport; 1773 th->th_seq = htonl(seq); 1774 th->th_ack = htonl(ack); 1775 th->th_off = tlen >> 2; 1776 th->th_flags = flags; 1777 th->th_win = htons(win); 1778 1779 if (mss) { 1780 opt = (char *)(th + 1); 1781 opt[0] = TCPOPT_MAXSEG; 1782 opt[1] = 4; 1783 HTONS(mss); 1784 bcopy((void *)&mss, (void *)(opt + 2), 2); 1785 } 1786 1787 switch (af) { 1788 #ifdef INET 1789 case AF_INET: 1790 /* TCP checksum */ 1791 th->th_sum = in_cksum(m, len); 1792 1793 /* Finish the IP header */ 1794 h->ip_v = 4; 1795 h->ip_hl = sizeof(*h) >> 2; 1796 h->ip_tos = IPTOS_LOWDELAY; 1797 h->ip_len = htons(len); 1798 h->ip_off = htons(ip_mtudisc ? IP_DF : 0); 1799 h->ip_ttl = ttl ? ttl : ip_defttl; 1800 h->ip_sum = 0; 1801 if (eh == NULL) { 1802 ip_output(m, (void *)NULL, (void *)NULL, 0, 1803 (void *)NULL, (void *)NULL); 1804 } else { 1805 #ifdef __NetBSD__ 1806 /* 1807 * On netbsd, pf_test and pf_test6 are always called 1808 * with eh == NULL. 1809 */ 1810 panic("pf_send_tcp: eh != NULL"); 1811 #else 1812 struct route ro; 1813 struct rtentry rt; 1814 struct ether_header *e = (void *)ro.ro_dst.sa_data; 1815 1816 if (ifp == NULL) { 1817 m_freem(m); 1818 return; 1819 } 1820 rt.rt_ifp = ifp; 1821 ro.ro_rt = &rt; 1822 ro.ro_dst.sa_len = sizeof(ro.ro_dst); 1823 ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT; 1824 bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN); 1825 bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN); 1826 e->ether_type = eh->ether_type; 1827 ip_output(m, (void *)NULL, &ro, IP_ROUTETOETHER, 1828 (void *)NULL, (void *)NULL); 1829 #endif /* !__NetBSD__ */ 1830 } 1831 break; 1832 #endif /* INET */ 1833 #ifdef INET6 1834 case AF_INET6: 1835 /* TCP checksum */ 1836 th->th_sum = in6_cksum(m, IPPROTO_TCP, 1837 sizeof(struct ip6_hdr), tlen); 1838 1839 h6->ip6_vfc |= IPV6_VERSION; 1840 h6->ip6_hlim = IPV6_DEFHLIM; 1841 1842 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL); 1843 break; 1844 #endif /* INET6 */ 1845 } 1846 } 1847 1848 void 1849 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af, 1850 struct pf_rule *r) 1851 { 1852 struct mbuf *m0; 1853 #ifdef __NetBSD__ 1854 struct pf_mtag *pf_mtag; 1855 #endif /* __NetBSD__ */ 1856 1857 m0 = m_copy(m, 0, M_COPYALL); 1858 1859 #ifdef __NetBSD__ 1860 if ((pf_mtag = pf_get_mtag(m0)) == NULL) 1861 return; 1862 pf_mtag->flags |= PF_TAG_GENERATED; 1863 1864 if (r->rtableid >= 0) 1865 pf_mtag->rtableid = r->rtableid; 1866 #else 1867 m0->m_pkthdr.pf.flags |= PF_TAG_GENERATED; 1868 1869 if (r->rtableid >= 0) 1870 m0->m_pkthdr.pf.rtableid = r->rtableid; 1871 #endif /* !__NetBSD__ */ 1872 1873 #ifdef ALTQ 1874 if (r->qid) { 1875 #ifdef __NetBSD__ 1876 struct m_tag *mtag; 1877 struct altq_tag *atag; 1878 1879 mtag = m_tag_get(PACKET_TAG_ALTQ_QID, sizeof(*atag), M_NOWAIT); 1880 if (mtag != NULL) { 1881 atag = (struct altq_tag *)(mtag + 1); 1882 atag->qid = r->qid; 1883 /* add hints for ecn */ 1884 atag->af = af; 1885 atag->hdr = mtod(m0, struct ip *); 1886 m_tag_prepend(m0, mtag); 1887 } 1888 #else 1889 m0->m_pkthdr.pf.qid = r->qid; 1890 /* add hints for ecn */ 1891 m0->m_pkthdr.pf.hdr = mtod(m0, struct ip *); 1892 #endif /* !__NetBSD__ */ 1893 } 1894 #endif /* ALTQ */ 1895 1896 switch (af) { 1897 #ifdef INET 1898 case AF_INET: 1899 icmp_error(m0, type, code, 0, 0); 1900 break; 1901 #endif /* INET */ 1902 #ifdef INET6 1903 case AF_INET6: 1904 icmp6_error(m0, type, code, 0); 1905 break; 1906 #endif /* INET6 */ 1907 } 1908 } 1909 1910 /* 1911 * Return 1 if the addresses a and b match (with mask m), otherwise return 0. 1912 * If n is 0, they match if they are equal. If n is != 0, they match if they 1913 * are different. 1914 */ 1915 int 1916 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m, 1917 struct pf_addr *b, sa_family_t af) 1918 { 1919 int match = 0; 1920 1921 switch (af) { 1922 #ifdef INET 1923 case AF_INET: 1924 if ((a->addr32[0] & m->addr32[0]) == 1925 (b->addr32[0] & m->addr32[0])) 1926 match++; 1927 break; 1928 #endif /* INET */ 1929 #ifdef INET6 1930 case AF_INET6: 1931 if (((a->addr32[0] & m->addr32[0]) == 1932 (b->addr32[0] & m->addr32[0])) && 1933 ((a->addr32[1] & m->addr32[1]) == 1934 (b->addr32[1] & m->addr32[1])) && 1935 ((a->addr32[2] & m->addr32[2]) == 1936 (b->addr32[2] & m->addr32[2])) && 1937 ((a->addr32[3] & m->addr32[3]) == 1938 (b->addr32[3] & m->addr32[3]))) 1939 match++; 1940 break; 1941 #endif /* INET6 */ 1942 } 1943 if (match) { 1944 if (n) 1945 return (0); 1946 else 1947 return (1); 1948 } else { 1949 if (n) 1950 return (1); 1951 else 1952 return (0); 1953 } 1954 } 1955 1956 int 1957 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p) 1958 { 1959 switch (op) { 1960 case PF_OP_IRG: 1961 return ((p > a1) && (p < a2)); 1962 case PF_OP_XRG: 1963 return ((p < a1) || (p > a2)); 1964 case PF_OP_RRG: 1965 return ((p >= a1) && (p <= a2)); 1966 case PF_OP_EQ: 1967 return (p == a1); 1968 case PF_OP_NE: 1969 return (p != a1); 1970 case PF_OP_LT: 1971 return (p < a1); 1972 case PF_OP_LE: 1973 return (p <= a1); 1974 case PF_OP_GT: 1975 return (p > a1); 1976 case PF_OP_GE: 1977 return (p >= a1); 1978 } 1979 return (0); /* never reached */ 1980 } 1981 1982 int 1983 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p) 1984 { 1985 NTOHS(a1); 1986 NTOHS(a2); 1987 NTOHS(p); 1988 return (pf_match(op, a1, a2, p)); 1989 } 1990 1991 int 1992 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u) 1993 { 1994 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE) 1995 return (0); 1996 return (pf_match(op, a1, a2, u)); 1997 } 1998 1999 int 2000 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g) 2001 { 2002 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE) 2003 return (0); 2004 return (pf_match(op, a1, a2, g)); 2005 } 2006 2007 int 2008 pf_match_tag(struct mbuf *m, struct pf_rule *r, int *tag) 2009 { 2010 #ifdef __NetBSD__ 2011 if (*tag == -1) { 2012 struct pf_mtag *pf_mtag = pf_get_mtag(m); 2013 if (pf_mtag == NULL) 2014 return (0); 2015 2016 *tag = pf_mtag->tag; 2017 } 2018 #else 2019 if (*tag == -1) 2020 *tag = m->m_pkthdr.pf.tag; 2021 #endif /* !__NetBSD__ */ 2022 2023 return ((!r->match_tag_not && r->match_tag == *tag) || 2024 (r->match_tag_not && r->match_tag != *tag)); 2025 } 2026 2027 int 2028 pf_tag_packet(struct mbuf *m, int tag, int rtableid) 2029 { 2030 if (tag <= 0 && rtableid < 0) 2031 return (0); 2032 2033 #ifdef __NetBSD__ 2034 if (tag > 0 || rtableid > 0) { 2035 struct pf_mtag *pf_mtag = pf_get_mtag(m); 2036 if (pf_mtag == NULL) 2037 return (1); 2038 2039 if (tag > 0) 2040 pf_mtag->tag = tag; 2041 if (rtableid > 0) 2042 pf_mtag->rtableid = rtableid; 2043 } 2044 #else 2045 if (tag > 0) 2046 m->m_pkthdr.pf.tag = tag; 2047 if (rtableid >= 0) 2048 m->m_pkthdr.pf.rtableid = rtableid; 2049 #endif /* !__NetBSD__ */ 2050 2051 return (0); 2052 } 2053 2054 void 2055 pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n, 2056 struct pf_rule **r, struct pf_rule **a, int *match) 2057 { 2058 struct pf_anchor_stackframe *f; 2059 2060 (*r)->anchor->match = 0; 2061 if (match) 2062 *match = 0; 2063 if (*depth >= sizeof(pf_anchor_stack) / 2064 sizeof(pf_anchor_stack[0])) { 2065 printf("pf_step_into_anchor: stack overflow\n"); 2066 *r = TAILQ_NEXT(*r, entries); 2067 return; 2068 } else if (*depth == 0 && a != NULL) 2069 *a = *r; 2070 f = pf_anchor_stack + (*depth)++; 2071 f->rs = *rs; 2072 f->r = *r; 2073 if ((*r)->anchor_wildcard) { 2074 f->parent = &(*r)->anchor->children; 2075 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) == 2076 NULL) { 2077 *r = NULL; 2078 return; 2079 } 2080 *rs = &f->child->ruleset; 2081 } else { 2082 f->parent = NULL; 2083 f->child = NULL; 2084 *rs = &(*r)->anchor->ruleset; 2085 } 2086 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr); 2087 } 2088 2089 int 2090 pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n, 2091 struct pf_rule **r, struct pf_rule **a, int *match) 2092 { 2093 struct pf_anchor_stackframe *f; 2094 int quick = 0; 2095 2096 do { 2097 if (*depth <= 0) 2098 break; 2099 f = pf_anchor_stack + *depth - 1; 2100 if (f->parent != NULL && f->child != NULL) { 2101 if (f->child->match || 2102 (match != NULL && *match)) { 2103 f->r->anchor->match = 1; 2104 *match = 0; 2105 } 2106 f->child = RB_NEXT(pf_anchor_node, f->parent, f->child); 2107 if (f->child != NULL) { 2108 *rs = &f->child->ruleset; 2109 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr); 2110 if (*r == NULL) 2111 continue; 2112 else 2113 break; 2114 } 2115 } 2116 (*depth)--; 2117 if (*depth == 0 && a != NULL) 2118 *a = NULL; 2119 *rs = f->rs; 2120 if (f->r->anchor->match || (match != NULL && *match)) 2121 quick = f->r->quick; 2122 *r = TAILQ_NEXT(f->r, entries); 2123 } while (*r == NULL); 2124 2125 return (quick); 2126 } 2127 2128 #ifdef INET6 2129 void 2130 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr, 2131 struct pf_addr *rmask, const struct pf_addr *saddr, sa_family_t af) 2132 { 2133 switch (af) { 2134 #ifdef INET 2135 case AF_INET: 2136 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) | 2137 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]); 2138 break; 2139 #endif /* INET */ 2140 case AF_INET6: 2141 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) | 2142 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]); 2143 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) | 2144 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]); 2145 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) | 2146 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]); 2147 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) | 2148 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]); 2149 break; 2150 } 2151 } 2152 2153 void 2154 pf_addr_inc(struct pf_addr *addr, sa_family_t af) 2155 { 2156 switch (af) { 2157 #ifdef INET 2158 case AF_INET: 2159 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1); 2160 break; 2161 #endif /* INET */ 2162 case AF_INET6: 2163 if (addr->addr32[3] == 0xffffffff) { 2164 addr->addr32[3] = 0; 2165 if (addr->addr32[2] == 0xffffffff) { 2166 addr->addr32[2] = 0; 2167 if (addr->addr32[1] == 0xffffffff) { 2168 addr->addr32[1] = 0; 2169 addr->addr32[0] = 2170 htonl(ntohl(addr->addr32[0]) + 1); 2171 } else 2172 addr->addr32[1] = 2173 htonl(ntohl(addr->addr32[1]) + 1); 2174 } else 2175 addr->addr32[2] = 2176 htonl(ntohl(addr->addr32[2]) + 1); 2177 } else 2178 addr->addr32[3] = 2179 htonl(ntohl(addr->addr32[3]) + 1); 2180 break; 2181 } 2182 } 2183 #endif /* INET6 */ 2184 2185 #define mix(a,b,c) \ 2186 do { \ 2187 a -= b; a -= c; a ^= (c >> 13); \ 2188 b -= c; b -= a; b ^= (a << 8); \ 2189 c -= a; c -= b; c ^= (b >> 13); \ 2190 a -= b; a -= c; a ^= (c >> 12); \ 2191 b -= c; b -= a; b ^= (a << 16); \ 2192 c -= a; c -= b; c ^= (b >> 5); \ 2193 a -= b; a -= c; a ^= (c >> 3); \ 2194 b -= c; b -= a; b ^= (a << 10); \ 2195 c -= a; c -= b; c ^= (b >> 15); \ 2196 } while (0) 2197 2198 /* 2199 * hash function based on bridge_hash in if_bridge.c 2200 */ 2201 void 2202 pf_hash(const struct pf_addr *inaddr, struct pf_addr *hash, 2203 struct pf_poolhashkey *key, sa_family_t af) 2204 { 2205 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0]; 2206 2207 switch (af) { 2208 #ifdef INET 2209 case AF_INET: 2210 a += inaddr->addr32[0]; 2211 b += key->key32[1]; 2212 mix(a, b, c); 2213 hash->addr32[0] = c + key->key32[2]; 2214 break; 2215 #endif /* INET */ 2216 #ifdef INET6 2217 case AF_INET6: 2218 a += inaddr->addr32[0]; 2219 b += inaddr->addr32[2]; 2220 mix(a, b, c); 2221 hash->addr32[0] = c; 2222 a += inaddr->addr32[1]; 2223 b += inaddr->addr32[3]; 2224 c += key->key32[1]; 2225 mix(a, b, c); 2226 hash->addr32[1] = c; 2227 a += inaddr->addr32[2]; 2228 b += inaddr->addr32[1]; 2229 c += key->key32[2]; 2230 mix(a, b, c); 2231 hash->addr32[2] = c; 2232 a += inaddr->addr32[3]; 2233 b += inaddr->addr32[0]; 2234 c += key->key32[3]; 2235 mix(a, b, c); 2236 hash->addr32[3] = c; 2237 break; 2238 #endif /* INET6 */ 2239 } 2240 } 2241 2242 int 2243 pf_map_addr(sa_family_t af, struct pf_rule *r, const struct pf_addr *saddr, 2244 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn) 2245 { 2246 unsigned char hash[16]; 2247 struct pf_pool *rpool = &r->rpool; 2248 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr; 2249 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask; 2250 struct pf_pooladdr *acur = rpool->cur; 2251 struct pf_src_node k; 2252 2253 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR && 2254 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { 2255 k.af = af; 2256 PF_ACPY(&k.addr, saddr, af); 2257 if (r->rule_flag & PFRULE_RULESRCTRACK || 2258 r->rpool.opts & PF_POOL_STICKYADDR) 2259 k.rule.ptr = r; 2260 else 2261 k.rule.ptr = NULL; 2262 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++; 2263 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k); 2264 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) { 2265 PF_ACPY(naddr, &(*sn)->raddr, af); 2266 if (pf_status.debug >= PF_DEBUG_MISC) { 2267 printf("pf_map_addr: src tracking maps "); 2268 pf_print_host(&k.addr, 0, af); 2269 printf(" to "); 2270 pf_print_host(naddr, 0, af); 2271 printf("\n"); 2272 } 2273 return (0); 2274 } 2275 } 2276 2277 if (rpool->cur->addr.type == PF_ADDR_NOROUTE) 2278 return (1); 2279 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2280 switch (af) { 2281 #ifdef INET 2282 case AF_INET: 2283 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 && 2284 (rpool->opts & PF_POOL_TYPEMASK) != 2285 PF_POOL_ROUNDROBIN) 2286 return (1); 2287 raddr = &rpool->cur->addr.p.dyn->pfid_addr4; 2288 rmask = &rpool->cur->addr.p.dyn->pfid_mask4; 2289 break; 2290 #endif /* INET */ 2291 #ifdef INET6 2292 case AF_INET6: 2293 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 && 2294 (rpool->opts & PF_POOL_TYPEMASK) != 2295 PF_POOL_ROUNDROBIN) 2296 return (1); 2297 raddr = &rpool->cur->addr.p.dyn->pfid_addr6; 2298 rmask = &rpool->cur->addr.p.dyn->pfid_mask6; 2299 break; 2300 #endif /* INET6 */ 2301 } 2302 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2303 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN) 2304 return (1); /* unsupported */ 2305 } else { 2306 raddr = &rpool->cur->addr.v.a.addr; 2307 rmask = &rpool->cur->addr.v.a.mask; 2308 } 2309 2310 switch (rpool->opts & PF_POOL_TYPEMASK) { 2311 case PF_POOL_NONE: 2312 PF_ACPY(naddr, raddr, af); 2313 break; 2314 case PF_POOL_BITMASK: 2315 PF_POOLMASK(naddr, raddr, rmask, saddr, af); 2316 break; 2317 case PF_POOL_RANDOM: 2318 if (init_addr != NULL && PF_AZERO(init_addr, af)) { 2319 switch (af) { 2320 #ifdef INET 2321 case AF_INET: 2322 rpool->counter.addr32[0] = htonl(arc4random()); 2323 break; 2324 #endif /* INET */ 2325 #ifdef INET6 2326 case AF_INET6: 2327 if (rmask->addr32[3] != 0xffffffff) 2328 rpool->counter.addr32[3] = 2329 htonl(arc4random()); 2330 else 2331 break; 2332 if (rmask->addr32[2] != 0xffffffff) 2333 rpool->counter.addr32[2] = 2334 htonl(arc4random()); 2335 else 2336 break; 2337 if (rmask->addr32[1] != 0xffffffff) 2338 rpool->counter.addr32[1] = 2339 htonl(arc4random()); 2340 else 2341 break; 2342 if (rmask->addr32[0] != 0xffffffff) 2343 rpool->counter.addr32[0] = 2344 htonl(arc4random()); 2345 break; 2346 #endif /* INET6 */ 2347 } 2348 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af); 2349 PF_ACPY(init_addr, naddr, af); 2350 2351 } else { 2352 PF_AINC(&rpool->counter, af); 2353 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af); 2354 } 2355 break; 2356 case PF_POOL_SRCHASH: 2357 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af); 2358 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af); 2359 break; 2360 case PF_POOL_ROUNDROBIN: 2361 if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2362 if (!pfr_pool_get(rpool->cur->addr.p.tbl, 2363 &rpool->tblidx, &rpool->counter, 2364 &raddr, &rmask, af)) 2365 goto get_addr; 2366 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2367 if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, 2368 &rpool->tblidx, &rpool->counter, 2369 &raddr, &rmask, af)) 2370 goto get_addr; 2371 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af)) 2372 goto get_addr; 2373 2374 try_next: 2375 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL) 2376 rpool->cur = TAILQ_FIRST(&rpool->list); 2377 if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2378 rpool->tblidx = -1; 2379 if (pfr_pool_get(rpool->cur->addr.p.tbl, 2380 &rpool->tblidx, &rpool->counter, 2381 &raddr, &rmask, af)) { 2382 /* table contains no address of type 'af' */ 2383 if (rpool->cur != acur) 2384 goto try_next; 2385 return (1); 2386 } 2387 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2388 rpool->tblidx = -1; 2389 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, 2390 &rpool->tblidx, &rpool->counter, 2391 &raddr, &rmask, af)) { 2392 /* table contains no address of type 'af' */ 2393 if (rpool->cur != acur) 2394 goto try_next; 2395 return (1); 2396 } 2397 } else { 2398 raddr = &rpool->cur->addr.v.a.addr; 2399 rmask = &rpool->cur->addr.v.a.mask; 2400 PF_ACPY(&rpool->counter, raddr, af); 2401 } 2402 2403 get_addr: 2404 PF_ACPY(naddr, &rpool->counter, af); 2405 if (init_addr != NULL && PF_AZERO(init_addr, af)) 2406 PF_ACPY(init_addr, naddr, af); 2407 PF_AINC(&rpool->counter, af); 2408 break; 2409 } 2410 if (*sn != NULL) 2411 PF_ACPY(&(*sn)->raddr, naddr, af); 2412 2413 if (pf_status.debug >= PF_DEBUG_MISC && 2414 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { 2415 printf("pf_map_addr: selected address "); 2416 pf_print_host(naddr, 0, af); 2417 printf("\n"); 2418 } 2419 2420 return (0); 2421 } 2422 2423 int 2424 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r, 2425 struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport, 2426 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high, 2427 struct pf_src_node **sn) 2428 { 2429 struct pf_state_key_cmp key; 2430 struct pf_addr init_addr; 2431 u_int16_t cut; 2432 2433 bzero(&init_addr, sizeof(init_addr)); 2434 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn)) 2435 return (1); 2436 2437 if (proto == IPPROTO_ICMP) { 2438 low = 1; 2439 high = 65535; 2440 } 2441 2442 do { 2443 key.af = af; 2444 key.proto = proto; 2445 PF_ACPY(&key.ext.addr, daddr, key.af); 2446 PF_ACPY(&key.gwy.addr, naddr, key.af); 2447 key.ext.port = dport; 2448 2449 /* 2450 * port search; start random, step; 2451 * similar 2 portloop in in_pcbbind 2452 */ 2453 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP || 2454 proto == IPPROTO_ICMP)) { 2455 key.gwy.port = dport; 2456 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) 2457 return (0); 2458 } else if (low == 0 && high == 0) { 2459 key.gwy.port = *nport; 2460 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) 2461 return (0); 2462 } else if (low == high) { 2463 key.gwy.port = htons(low); 2464 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) { 2465 *nport = htons(low); 2466 return (0); 2467 } 2468 } else { 2469 u_int16_t tmp; 2470 2471 if (low > high) { 2472 tmp = low; 2473 low = high; 2474 high = tmp; 2475 } 2476 /* low < high */ 2477 cut = htonl(arc4random()) % (1 + high - low) + low; 2478 /* low <= cut <= high */ 2479 for (tmp = cut; tmp <= high; ++(tmp)) { 2480 key.gwy.port = htons(tmp); 2481 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == 2482 NULL) { 2483 *nport = htons(tmp); 2484 return (0); 2485 } 2486 } 2487 for (tmp = cut - 1; tmp >= low; --(tmp)) { 2488 key.gwy.port = htons(tmp); 2489 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == 2490 NULL) { 2491 *nport = htons(tmp); 2492 return (0); 2493 } 2494 } 2495 } 2496 2497 switch (r->rpool.opts & PF_POOL_TYPEMASK) { 2498 case PF_POOL_RANDOM: 2499 case PF_POOL_ROUNDROBIN: 2500 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn)) 2501 return (1); 2502 break; 2503 case PF_POOL_NONE: 2504 case PF_POOL_SRCHASH: 2505 case PF_POOL_BITMASK: 2506 default: 2507 return (1); 2508 } 2509 } while (! PF_AEQ(&init_addr, naddr, af) ); 2510 2511 return (1); /* none available */ 2512 } 2513 2514 struct pf_rule * 2515 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off, 2516 int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport, 2517 struct pf_addr *daddr, u_int16_t dport, int rs_num) 2518 { 2519 struct pf_rule *r, *rm = NULL; 2520 struct pf_ruleset *ruleset = NULL; 2521 int tag = -1; 2522 int rtableid = -1; 2523 int asd = 0; 2524 2525 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr); 2526 while (r && rm == NULL) { 2527 struct pf_rule_addr *src = NULL, *dst = NULL; 2528 struct pf_addr_wrap *xdst = NULL; 2529 2530 if (r->action == PF_BINAT && direction == PF_IN) { 2531 src = &r->dst; 2532 if (r->rpool.cur != NULL) 2533 xdst = &r->rpool.cur->addr; 2534 } else { 2535 src = &r->src; 2536 dst = &r->dst; 2537 } 2538 2539 r->evaluations++; 2540 if (pfi_kif_match(r->kif, kif) == r->ifnot) 2541 r = r->skip[PF_SKIP_IFP].ptr; 2542 else if (r->direction && r->direction != direction) 2543 r = r->skip[PF_SKIP_DIR].ptr; 2544 else if (r->af && r->af != pd->af) 2545 r = r->skip[PF_SKIP_AF].ptr; 2546 else if (r->proto && r->proto != pd->proto) 2547 r = r->skip[PF_SKIP_PROTO].ptr; 2548 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af, 2549 src->neg, kif)) 2550 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR : 2551 PF_SKIP_DST_ADDR].ptr; 2552 else if (src->port_op && !pf_match_port(src->port_op, 2553 src->port[0], src->port[1], sport)) 2554 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT : 2555 PF_SKIP_DST_PORT].ptr; 2556 else if (dst != NULL && 2557 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL)) 2558 r = r->skip[PF_SKIP_DST_ADDR].ptr; 2559 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af, 2560 0, NULL)) 2561 r = TAILQ_NEXT(r, entries); 2562 else if (dst != NULL && dst->port_op && 2563 !pf_match_port(dst->port_op, dst->port[0], 2564 dst->port[1], dport)) 2565 r = r->skip[PF_SKIP_DST_PORT].ptr; 2566 else if (r->match_tag && !pf_match_tag(m, r, &tag)) 2567 r = TAILQ_NEXT(r, entries); 2568 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto != 2569 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m, 2570 off, pd->hdr.tcp), r->os_fingerprint))) 2571 r = TAILQ_NEXT(r, entries); 2572 else { 2573 if (r->tag) 2574 tag = r->tag; 2575 if (r->rtableid >= 0) 2576 rtableid = r->rtableid; 2577 if (r->anchor == NULL) { 2578 rm = r; 2579 } else 2580 pf_step_into_anchor(&asd, &ruleset, rs_num, 2581 &r, NULL, NULL); 2582 } 2583 if (r == NULL) 2584 pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r, 2585 NULL, NULL); 2586 } 2587 if (pf_tag_packet(m, tag, rtableid)) 2588 return (NULL); 2589 if (rm != NULL && (rm->action == PF_NONAT || 2590 rm->action == PF_NORDR || rm->action == PF_NOBINAT)) 2591 return (NULL); 2592 return (rm); 2593 } 2594 2595 struct pf_rule * 2596 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction, 2597 struct pfi_kif *kif, struct pf_src_node **sn, 2598 struct pf_addr *saddr, u_int16_t sport, 2599 struct pf_addr *daddr, u_int16_t dport, 2600 struct pf_addr *naddr, u_int16_t *nport) 2601 { 2602 struct pf_rule *r = NULL; 2603 2604 if (direction == PF_OUT) { 2605 r = pf_match_translation(pd, m, off, direction, kif, saddr, 2606 sport, daddr, dport, PF_RULESET_BINAT); 2607 if (r == NULL) 2608 r = pf_match_translation(pd, m, off, direction, kif, 2609 saddr, sport, daddr, dport, PF_RULESET_NAT); 2610 } else { 2611 r = pf_match_translation(pd, m, off, direction, kif, saddr, 2612 sport, daddr, dport, PF_RULESET_RDR); 2613 if (r == NULL) 2614 r = pf_match_translation(pd, m, off, direction, kif, 2615 saddr, sport, daddr, dport, PF_RULESET_BINAT); 2616 } 2617 2618 if (r != NULL) { 2619 switch (r->action) { 2620 case PF_NONAT: 2621 case PF_NOBINAT: 2622 case PF_NORDR: 2623 return (NULL); 2624 case PF_NAT: 2625 if (pf_get_sport(pd->af, pd->proto, r, saddr, 2626 daddr, dport, naddr, nport, r->rpool.proxy_port[0], 2627 r->rpool.proxy_port[1], sn)) { 2628 DPFPRINTF(PF_DEBUG_MISC, 2629 ("pf: NAT proxy port allocation " 2630 "(%u-%u) failed\n", 2631 r->rpool.proxy_port[0], 2632 r->rpool.proxy_port[1])); 2633 return (NULL); 2634 } 2635 break; 2636 case PF_BINAT: 2637 switch (direction) { 2638 case PF_OUT: 2639 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){ 2640 switch (pd->af) { 2641 #ifdef INET 2642 case AF_INET: 2643 if (r->rpool.cur->addr.p.dyn-> 2644 pfid_acnt4 < 1) 2645 return (NULL); 2646 PF_POOLMASK(naddr, 2647 &r->rpool.cur->addr.p.dyn-> 2648 pfid_addr4, 2649 &r->rpool.cur->addr.p.dyn-> 2650 pfid_mask4, 2651 saddr, AF_INET); 2652 break; 2653 #endif /* INET */ 2654 #ifdef INET6 2655 case AF_INET6: 2656 if (r->rpool.cur->addr.p.dyn-> 2657 pfid_acnt6 < 1) 2658 return (NULL); 2659 PF_POOLMASK(naddr, 2660 &r->rpool.cur->addr.p.dyn-> 2661 pfid_addr6, 2662 &r->rpool.cur->addr.p.dyn-> 2663 pfid_mask6, 2664 saddr, AF_INET6); 2665 break; 2666 #endif /* INET6 */ 2667 } 2668 } else 2669 PF_POOLMASK(naddr, 2670 &r->rpool.cur->addr.v.a.addr, 2671 &r->rpool.cur->addr.v.a.mask, 2672 saddr, pd->af); 2673 break; 2674 case PF_IN: 2675 if (r->src.addr.type == PF_ADDR_DYNIFTL) { 2676 switch (pd->af) { 2677 #ifdef INET 2678 case AF_INET: 2679 if (r->src.addr.p.dyn-> 2680 pfid_acnt4 < 1) 2681 return (NULL); 2682 PF_POOLMASK(naddr, 2683 &r->src.addr.p.dyn-> 2684 pfid_addr4, 2685 &r->src.addr.p.dyn-> 2686 pfid_mask4, 2687 daddr, AF_INET); 2688 break; 2689 #endif /* INET */ 2690 #ifdef INET6 2691 case AF_INET6: 2692 if (r->src.addr.p.dyn-> 2693 pfid_acnt6 < 1) 2694 return (NULL); 2695 PF_POOLMASK(naddr, 2696 &r->src.addr.p.dyn-> 2697 pfid_addr6, 2698 &r->src.addr.p.dyn-> 2699 pfid_mask6, 2700 daddr, AF_INET6); 2701 break; 2702 #endif /* INET6 */ 2703 } 2704 } else 2705 PF_POOLMASK(naddr, 2706 &r->src.addr.v.a.addr, 2707 &r->src.addr.v.a.mask, daddr, 2708 pd->af); 2709 break; 2710 } 2711 break; 2712 case PF_RDR: { 2713 if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn)) 2714 return (NULL); 2715 if ((r->rpool.opts & PF_POOL_TYPEMASK) == 2716 PF_POOL_BITMASK) 2717 PF_POOLMASK(naddr, naddr, 2718 &r->rpool.cur->addr.v.a.mask, daddr, 2719 pd->af); 2720 2721 if (r->rpool.proxy_port[1]) { 2722 u_int32_t tmp_nport; 2723 2724 tmp_nport = ((ntohs(dport) - 2725 ntohs(r->dst.port[0])) % 2726 (r->rpool.proxy_port[1] - 2727 r->rpool.proxy_port[0] + 1)) + 2728 r->rpool.proxy_port[0]; 2729 2730 /* wrap around if necessary */ 2731 if (tmp_nport > 65535) 2732 tmp_nport -= 65535; 2733 *nport = htons((u_int16_t)tmp_nport); 2734 } else if (r->rpool.proxy_port[0]) 2735 *nport = htons(r->rpool.proxy_port[0]); 2736 break; 2737 } 2738 default: 2739 return (NULL); 2740 } 2741 } 2742 2743 return (r); 2744 } 2745 2746 int 2747 pf_socket_lookup(int direction, struct pf_pdesc *pd) 2748 { 2749 struct pf_addr *saddr, *daddr; 2750 u_int16_t sport, dport; 2751 struct inpcbtable *tb; 2752 struct inpcb *inp = NULL; 2753 struct socket *so = NULL; 2754 #if defined(__NetBSD__) && defined(INET6) 2755 struct in6pcb *in6p = NULL; 2756 #else 2757 #define in6p inp 2758 #endif /* __NetBSD__ && INET6 */ 2759 2760 if (pd == NULL) 2761 return (-1); 2762 pd->lookup.uid = UID_MAX; 2763 pd->lookup.gid = GID_MAX; 2764 pd->lookup.pid = NO_PID; 2765 switch (pd->proto) { 2766 case IPPROTO_TCP: 2767 if (pd->hdr.tcp == NULL) 2768 return (-1); 2769 sport = pd->hdr.tcp->th_sport; 2770 dport = pd->hdr.tcp->th_dport; 2771 tb = &tcbtable; 2772 break; 2773 case IPPROTO_UDP: 2774 if (pd->hdr.udp == NULL) 2775 return (-1); 2776 sport = pd->hdr.udp->uh_sport; 2777 dport = pd->hdr.udp->uh_dport; 2778 tb = &udbtable; 2779 break; 2780 default: 2781 return (-1); 2782 } 2783 if (direction == PF_IN) { 2784 saddr = pd->src; 2785 daddr = pd->dst; 2786 } else { 2787 u_int16_t p; 2788 2789 p = sport; 2790 sport = dport; 2791 dport = p; 2792 saddr = pd->dst; 2793 daddr = pd->src; 2794 } 2795 switch (pd->af) { 2796 2797 #ifdef __NetBSD__ 2798 #define in_pcbhashlookup(tbl, saddr, sport, daddr, dport) \ 2799 in_pcblookup_connect(tbl, saddr, sport, daddr, dport) 2800 #define in6_pcbhashlookup(tbl, saddr, sport, daddr, dport) \ 2801 in6_pcblookup_connect(tbl, saddr, sport, daddr, dport, 0) 2802 #define in_pcblookup_listen(tbl, addr, port, zero) \ 2803 in_pcblookup_bind(tbl, addr, port) 2804 #define in6_pcblookup_listen(tbl, addr, port, zero) \ 2805 in6_pcblookup_bind(tbl, addr, port, zero) 2806 #endif 2807 2808 #ifdef INET 2809 case AF_INET: 2810 inp = in_pcbhashlookup(tb, saddr->v4, sport, daddr->v4, dport); 2811 if (inp == NULL) { 2812 inp = in_pcblookup_listen(tb, daddr->v4, dport, 0); 2813 if (inp == NULL) 2814 return (-1); 2815 } 2816 break; 2817 #endif /* INET */ 2818 #ifdef INET6 2819 case AF_INET6: 2820 /*###2817 [cc] warning: assignment from incompatible pointer type%%%*/ 2821 in6p = in6_pcbhashlookup(tb, &saddr->v6, sport, &daddr->v6, 2822 dport); 2823 if (inp == NULL) { 2824 in6p = in6_pcblookup_listen(tb, &daddr->v6, dport, 0); 2825 if (inp == NULL) 2826 return (-1); 2827 } 2828 break; 2829 #endif /* INET6 */ 2830 2831 default: 2832 return (-1); 2833 } 2834 2835 #ifdef __NetBSD__ 2836 switch (pd->af) { 2837 #ifdef INET 2838 case AF_INET: 2839 so = inp->inp_socket; 2840 break; 2841 #endif 2842 #ifdef INET6 2843 case AF_INET6: 2844 /*###2840 [cc] error: 'struct inpcb' has no member named 'in6p_head'%%%*/ 2845 so = in6p->in6p_socket; 2846 break; 2847 #endif /* INET6 */ 2848 } 2849 pd->lookup.uid = so->so_uidinfo->ui_uid; 2850 #else 2851 so = inp->inp_socket; 2852 pd->lookup.uid = so->so_euid; 2853 #endif /* !__NetBSD__ */ 2854 pd->lookup.gid = so->so_egid; 2855 pd->lookup.pid = so->so_cpid; 2856 return (1); 2857 } 2858 2859 u_int8_t 2860 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af) 2861 { 2862 int hlen; 2863 u_int8_t hdr[60]; 2864 u_int8_t *opt, optlen; 2865 u_int8_t wscale = 0; 2866 2867 hlen = th_off << 2; /* hlen <= sizeof(hdr) */ 2868 if (hlen <= sizeof(struct tcphdr)) 2869 return (0); 2870 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af)) 2871 return (0); 2872 opt = hdr + sizeof(struct tcphdr); 2873 hlen -= sizeof(struct tcphdr); 2874 while (hlen >= 3) { 2875 switch (*opt) { 2876 case TCPOPT_EOL: 2877 case TCPOPT_NOP: 2878 ++opt; 2879 --hlen; 2880 break; 2881 case TCPOPT_WINDOW: 2882 wscale = opt[2]; 2883 if (wscale > TCP_MAX_WINSHIFT) 2884 wscale = TCP_MAX_WINSHIFT; 2885 wscale |= PF_WSCALE_FLAG; 2886 /* FALLTHROUGH */ 2887 default: 2888 optlen = opt[1]; 2889 if (optlen < 2) 2890 optlen = 2; 2891 hlen -= optlen; 2892 opt += optlen; 2893 break; 2894 } 2895 } 2896 return (wscale); 2897 } 2898 2899 u_int16_t 2900 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af) 2901 { 2902 int hlen; 2903 u_int8_t hdr[60]; 2904 u_int8_t *opt, optlen; 2905 u_int16_t mss = tcp_mssdflt; 2906 2907 hlen = th_off << 2; /* hlen <= sizeof(hdr) */ 2908 if (hlen <= sizeof(struct tcphdr)) 2909 return (0); 2910 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af)) 2911 return (0); 2912 opt = hdr + sizeof(struct tcphdr); 2913 hlen -= sizeof(struct tcphdr); 2914 while (hlen >= TCPOLEN_MAXSEG) { 2915 switch (*opt) { 2916 case TCPOPT_EOL: 2917 case TCPOPT_NOP: 2918 ++opt; 2919 --hlen; 2920 break; 2921 case TCPOPT_MAXSEG: 2922 bcopy((void *)(opt + 2), (void *)&mss, 2); 2923 NTOHS(mss); 2924 /* FALLTHROUGH */ 2925 default: 2926 optlen = opt[1]; 2927 if (optlen < 2) 2928 optlen = 2; 2929 hlen -= optlen; 2930 opt += optlen; 2931 break; 2932 } 2933 } 2934 return (mss); 2935 } 2936 2937 u_int16_t 2938 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer) 2939 { 2940 union { 2941 struct sockaddr dst; 2942 struct sockaddr_in dst4; 2943 struct sockaddr_in6 dst6; 2944 } u; 2945 struct route ro; 2946 struct route *rop = &ro; 2947 struct rtentry *rt; 2948 int hlen; 2949 u_int16_t mss = tcp_mssdflt; 2950 2951 hlen = 0; /* XXXGCC -Wunitialized m68k */ 2952 2953 memset(&ro, 0, sizeof(ro)); 2954 switch (af) { 2955 #ifdef INET 2956 case AF_INET: 2957 hlen = sizeof(struct ip); 2958 sockaddr_in_init(&u.dst4, &addr->v4, 0); 2959 rtcache_setdst(rop, &u.dst); 2960 break; 2961 #endif /* INET */ 2962 #ifdef INET6 2963 case AF_INET6: 2964 hlen = sizeof(struct ip6_hdr); 2965 sockaddr_in6_init(&u.dst6, &addr->v6, 0, 0, 0); 2966 rtcache_setdst(rop, &u.dst); 2967 break; 2968 #endif /* INET6 */ 2969 } 2970 2971 #ifndef __NetBSD__ 2972 rtalloc_noclone(rop, NO_CLONING); 2973 if ((rt = ro->ro_rt) != NULL) { 2974 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr); 2975 mss = max(tcp_mssdflt, mss); 2976 } 2977 #else 2978 if ((rt = rtcache_init_noclone(rop)) != NULL) { 2979 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr); 2980 mss = max(tcp_mssdflt, mss); 2981 } 2982 rtcache_free(rop); 2983 #endif 2984 mss = min(mss, offer); 2985 mss = max(mss, 64); /* sanity - at least max opt space */ 2986 return (mss); 2987 } 2988 2989 void 2990 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr) 2991 { 2992 struct pf_rule *r = s->rule.ptr; 2993 2994 s->rt_kif = NULL; 2995 if (!r->rt || r->rt == PF_FASTROUTE) 2996 return; 2997 switch (s->state_key->af) { 2998 #ifdef INET 2999 case AF_INET: 3000 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL, 3001 &s->nat_src_node); 3002 s->rt_kif = r->rpool.cur->kif; 3003 break; 3004 #endif /* INET */ 3005 #ifdef INET6 3006 case AF_INET6: 3007 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL, 3008 &s->nat_src_node); 3009 s->rt_kif = r->rpool.cur->kif; 3010 break; 3011 #endif /* INET6 */ 3012 } 3013 } 3014 3015 void 3016 pf_attach_state(struct pf_state_key *sk, struct pf_state *s, int tail) 3017 { 3018 s->state_key = sk; 3019 sk->refcnt++; 3020 3021 /* list is sorted, if-bound states before floating */ 3022 if (tail) 3023 TAILQ_INSERT_TAIL(&sk->states, s, next); 3024 else 3025 TAILQ_INSERT_HEAD(&sk->states, s, next); 3026 } 3027 3028 void 3029 pf_detach_state(struct pf_state *s, int flags) 3030 { 3031 struct pf_state_key *sk = s->state_key; 3032 3033 if (sk == NULL) 3034 return; 3035 3036 s->state_key = NULL; 3037 TAILQ_REMOVE(&sk->states, s, next); 3038 if (--sk->refcnt == 0) { 3039 if (!(flags & PF_DT_SKIP_EXTGWY)) 3040 RB_REMOVE(pf_state_tree_ext_gwy, 3041 &pf_statetbl_ext_gwy, sk); 3042 if (!(flags & PF_DT_SKIP_LANEXT)) 3043 RB_REMOVE(pf_state_tree_lan_ext, 3044 &pf_statetbl_lan_ext, sk); 3045 pool_put(&pf_state_key_pl, sk); 3046 } 3047 } 3048 3049 struct pf_state_key * 3050 pf_alloc_state_key(struct pf_state *s) 3051 { 3052 struct pf_state_key *sk; 3053 3054 if ((sk = pool_get(&pf_state_key_pl, PR_NOWAIT)) == NULL) 3055 return (NULL); 3056 bzero(sk, sizeof(*sk)); 3057 TAILQ_INIT(&sk->states); 3058 pf_attach_state(sk, s, 0); 3059 3060 return (sk); 3061 } 3062 3063 int 3064 pf_test_rule(struct pf_rule **rm, struct pf_state **sm, int direction, 3065 struct pfi_kif *kif, struct mbuf *m, int off, void *h, 3066 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3067 struct ifqueue *ifq) 3068 { 3069 struct pf_rule *nr = NULL; 3070 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 3071 u_int16_t bport, nport = 0; 3072 sa_family_t af = pd->af; 3073 struct pf_rule *r, *a = NULL; 3074 struct pf_ruleset *ruleset = NULL; 3075 struct pf_src_node *nsn = NULL; 3076 struct tcphdr *th = pd->hdr.tcp; 3077 u_short reason; 3078 int rewrite = 0, hdrlen = 0; 3079 int tag = -1, rtableid = -1; 3080 int asd = 0; 3081 int match = 0; 3082 int state_icmp = 0; 3083 u_int16_t mss = tcp_mssdflt; 3084 u_int16_t sport, dport; 3085 u_int8_t icmptype = 0, icmpcode = 0; 3086 3087 if (direction == PF_IN && pf_check_congestion(ifq)) { 3088 REASON_SET(&reason, PFRES_CONGEST); 3089 return (PF_DROP); 3090 } 3091 3092 sport = dport = hdrlen = 0; 3093 3094 switch (pd->proto) { 3095 case IPPROTO_TCP: 3096 sport = th->th_sport; 3097 dport = th->th_dport; 3098 hdrlen = sizeof(*th); 3099 break; 3100 case IPPROTO_UDP: 3101 sport = pd->hdr.udp->uh_sport; 3102 dport = pd->hdr.udp->uh_dport; 3103 hdrlen = sizeof(*pd->hdr.udp); 3104 break; 3105 #ifdef INET 3106 case IPPROTO_ICMP: 3107 if (pd->af != AF_INET) 3108 break; 3109 sport = dport = pd->hdr.icmp->icmp_id; 3110 icmptype = pd->hdr.icmp->icmp_type; 3111 icmpcode = pd->hdr.icmp->icmp_code; 3112 3113 if (icmptype == ICMP_UNREACH || 3114 icmptype == ICMP_SOURCEQUENCH || 3115 icmptype == ICMP_REDIRECT || 3116 icmptype == ICMP_TIMXCEED || 3117 icmptype == ICMP_PARAMPROB) 3118 state_icmp++; 3119 break; 3120 #endif /* INET */ 3121 #ifdef INET6 3122 case IPPROTO_ICMPV6: 3123 if (pd->af != AF_INET6) 3124 break; 3125 sport = dport = pd->hdr.icmp6->icmp6_id; 3126 hdrlen = sizeof(*pd->hdr.icmp6); 3127 icmptype = pd->hdr.icmp6->icmp6_type; 3128 icmpcode = pd->hdr.icmp6->icmp6_code; 3129 3130 if (icmptype == ICMP6_DST_UNREACH || 3131 icmptype == ICMP6_PACKET_TOO_BIG || 3132 icmptype == ICMP6_TIME_EXCEEDED || 3133 icmptype == ICMP6_PARAM_PROB) 3134 state_icmp++; 3135 break; 3136 #endif /* INET6 */ 3137 } 3138 3139 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 3140 3141 if (direction == PF_OUT) { 3142 bport = nport = sport; 3143 /* check outgoing packet for BINAT/NAT */ 3144 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, 3145 saddr, sport, daddr, dport, &pd->naddr, &nport)) != NULL) { 3146 PF_ACPY(&pd->baddr, saddr, af); 3147 switch (pd->proto) { 3148 case IPPROTO_TCP: 3149 pf_change_ap(saddr, &th->th_sport, pd->ip_sum, 3150 &th->th_sum, &pd->naddr, nport, 0, af); 3151 sport = th->th_sport; 3152 rewrite++; 3153 break; 3154 case IPPROTO_UDP: 3155 pf_change_ap(saddr, &pd->hdr.udp->uh_sport, 3156 pd->ip_sum, &pd->hdr.udp->uh_sum, 3157 &pd->naddr, nport, 1, af); 3158 sport = pd->hdr.udp->uh_sport; 3159 rewrite++; 3160 break; 3161 #ifdef INET 3162 case IPPROTO_ICMP: 3163 pf_change_a(&saddr->v4.s_addr, pd->ip_sum, 3164 pd->naddr.v4.s_addr, 0); 3165 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup( 3166 pd->hdr.icmp->icmp_cksum, sport, nport, 0); 3167 pd->hdr.icmp->icmp_id = nport; 3168 m_copyback(m, off, ICMP_MINLEN, pd->hdr.icmp); 3169 break; 3170 #endif /* INET */ 3171 #ifdef INET6 3172 case IPPROTO_ICMPV6: 3173 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum, 3174 &pd->naddr, 0); 3175 rewrite++; 3176 break; 3177 #endif /* INET */ 3178 default: 3179 switch (af) { 3180 #ifdef INET 3181 case AF_INET: 3182 pf_change_a(&saddr->v4.s_addr, 3183 pd->ip_sum, pd->naddr.v4.s_addr, 0); 3184 break; 3185 #endif /* INET */ 3186 #ifdef INET6 3187 case AF_INET6: 3188 PF_ACPY(saddr, &pd->naddr, af); 3189 break; 3190 #endif /* INET */ 3191 } 3192 break; 3193 } 3194 3195 if (nr->natpass) 3196 r = NULL; 3197 pd->nat_rule = nr; 3198 } 3199 } else { 3200 bport = nport = dport; 3201 /* check incoming packet for BINAT/RDR */ 3202 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, 3203 saddr, sport, daddr, dport, &pd->naddr, &nport)) != NULL) { 3204 PF_ACPY(&pd->baddr, daddr, af); 3205 switch (pd->proto) { 3206 case IPPROTO_TCP: 3207 pf_change_ap(daddr, &th->th_dport, pd->ip_sum, 3208 &th->th_sum, &pd->naddr, nport, 0, af); 3209 dport = th->th_dport; 3210 rewrite++; 3211 break; 3212 case IPPROTO_UDP: 3213 pf_change_ap(daddr, &pd->hdr.udp->uh_dport, 3214 pd->ip_sum, &pd->hdr.udp->uh_sum, 3215 &pd->naddr, nport, 1, af); 3216 dport = pd->hdr.udp->uh_dport; 3217 rewrite++; 3218 break; 3219 #ifdef INET 3220 case IPPROTO_ICMP: 3221 pf_change_a(&daddr->v4.s_addr, pd->ip_sum, 3222 pd->naddr.v4.s_addr, 0); 3223 break; 3224 #endif /* INET */ 3225 #ifdef INET6 3226 case IPPROTO_ICMPV6: 3227 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum, 3228 &pd->naddr, 0); 3229 rewrite++; 3230 break; 3231 #endif /* INET6 */ 3232 default: 3233 switch (af) { 3234 #ifdef INET 3235 case AF_INET: 3236 pf_change_a(&daddr->v4.s_addr, 3237 pd->ip_sum, pd->naddr.v4.s_addr, 0); 3238 break; 3239 #endif /* INET */ 3240 #ifdef INET6 3241 case AF_INET6: 3242 PF_ACPY(daddr, &pd->naddr, af); 3243 break; 3244 #endif /* INET */ 3245 } 3246 break; 3247 } 3248 3249 if (nr->natpass) 3250 r = NULL; 3251 pd->nat_rule = nr; 3252 } 3253 } 3254 3255 while (r != NULL) { 3256 r->evaluations++; 3257 if (pfi_kif_match(r->kif, kif) == r->ifnot) 3258 r = r->skip[PF_SKIP_IFP].ptr; 3259 else if (r->direction && r->direction != direction) 3260 r = r->skip[PF_SKIP_DIR].ptr; 3261 else if (r->af && r->af != af) 3262 r = r->skip[PF_SKIP_AF].ptr; 3263 else if (r->proto && r->proto != pd->proto) 3264 r = r->skip[PF_SKIP_PROTO].ptr; 3265 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, 3266 r->src.neg, kif)) 3267 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 3268 /* tcp/udp only. port_op always 0 in other cases */ 3269 else if (r->src.port_op && !pf_match_port(r->src.port_op, 3270 r->src.port[0], r->src.port[1], sport)) 3271 r = r->skip[PF_SKIP_SRC_PORT].ptr; 3272 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, 3273 r->dst.neg, NULL)) 3274 r = r->skip[PF_SKIP_DST_ADDR].ptr; 3275 /* tcp/udp only. port_op always 0 in other cases */ 3276 else if (r->dst.port_op && !pf_match_port(r->dst.port_op, 3277 r->dst.port[0], r->dst.port[1], dport)) 3278 r = r->skip[PF_SKIP_DST_PORT].ptr; 3279 /* icmp only. type always 0 in other cases */ 3280 else if (r->type && r->type != icmptype + 1) 3281 r = TAILQ_NEXT(r, entries); 3282 /* icmp only. type always 0 in other cases */ 3283 else if (r->code && r->code != icmpcode + 1) 3284 r = TAILQ_NEXT(r, entries); 3285 else if (r->tos && !(r->tos == pd->tos)) 3286 r = TAILQ_NEXT(r, entries); 3287 else if (r->rule_flag & PFRULE_FRAGMENT) 3288 r = TAILQ_NEXT(r, entries); 3289 else if (pd->proto == IPPROTO_TCP && 3290 (r->flagset & th->th_flags) != r->flags) 3291 r = TAILQ_NEXT(r, entries); 3292 /* tcp/udp only. uid.op always 0 in other cases */ 3293 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done = 3294 pf_socket_lookup(direction, pd), 1)) && 3295 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1], 3296 pd->lookup.uid)) 3297 r = TAILQ_NEXT(r, entries); 3298 /* tcp/udp only. gid.op always 0 in other cases */ 3299 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done = 3300 pf_socket_lookup(direction, pd), 1)) && 3301 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1], 3302 pd->lookup.gid)) 3303 r = TAILQ_NEXT(r, entries); 3304 else if (r->prob && r->prob <= arc4random()) 3305 r = TAILQ_NEXT(r, entries); 3306 else if (r->match_tag && !pf_match_tag(m, r, &tag)) 3307 r = TAILQ_NEXT(r, entries); 3308 else if (r->os_fingerprint != PF_OSFP_ANY && 3309 (pd->proto != IPPROTO_TCP || !pf_osfp_match( 3310 pf_osfp_fingerprint(pd, m, off, th), 3311 r->os_fingerprint))) 3312 r = TAILQ_NEXT(r, entries); 3313 else { 3314 if (r->tag) 3315 tag = r->tag; 3316 if (r->rtableid >= 0) 3317 rtableid = r->rtableid; 3318 if (r->anchor == NULL) { 3319 match = 1; 3320 *rm = r; 3321 *am = a; 3322 *rsm = ruleset; 3323 if ((*rm)->quick) 3324 break; 3325 r = TAILQ_NEXT(r, entries); 3326 } else 3327 pf_step_into_anchor(&asd, &ruleset, 3328 PF_RULESET_FILTER, &r, &a, &match); 3329 } 3330 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 3331 PF_RULESET_FILTER, &r, &a, &match)) 3332 break; 3333 } 3334 r = *rm; 3335 a = *am; 3336 ruleset = *rsm; 3337 3338 REASON_SET(&reason, PFRES_MATCH); 3339 3340 if (r->log || (nr != NULL && nr->log)) { 3341 if (rewrite) 3342 m_copyback(m, off, hdrlen, pd->hdr.any); 3343 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, 3344 a, ruleset, pd); 3345 } 3346 3347 if ((r->action == PF_DROP) && 3348 ((r->rule_flag & PFRULE_RETURNRST) || 3349 (r->rule_flag & PFRULE_RETURNICMP) || 3350 (r->rule_flag & PFRULE_RETURN))) { 3351 /* undo NAT changes, if they have taken place */ 3352 if (nr != NULL) { 3353 if (direction == PF_OUT) { 3354 switch (pd->proto) { 3355 case IPPROTO_TCP: 3356 pf_change_ap(saddr, &th->th_sport, 3357 pd->ip_sum, &th->th_sum, 3358 &pd->baddr, bport, 0, af); 3359 sport = th->th_sport; 3360 rewrite++; 3361 break; 3362 case IPPROTO_UDP: 3363 pf_change_ap(saddr, 3364 &pd->hdr.udp->uh_sport, pd->ip_sum, 3365 &pd->hdr.udp->uh_sum, &pd->baddr, 3366 bport, 1, af); 3367 sport = pd->hdr.udp->uh_sport; 3368 rewrite++; 3369 break; 3370 case IPPROTO_ICMP: 3371 #ifdef INET6 3372 case IPPROTO_ICMPV6: 3373 #endif 3374 /* nothing! */ 3375 break; 3376 default: 3377 switch (af) { 3378 case AF_INET: 3379 pf_change_a(&saddr->v4.s_addr, 3380 pd->ip_sum, 3381 pd->baddr.v4.s_addr, 0); 3382 break; 3383 case AF_INET6: 3384 PF_ACPY(saddr, &pd->baddr, af); 3385 break; 3386 } 3387 } 3388 } else { 3389 switch (pd->proto) { 3390 case IPPROTO_TCP: 3391 pf_change_ap(daddr, &th->th_dport, 3392 pd->ip_sum, &th->th_sum, 3393 &pd->baddr, bport, 0, af); 3394 dport = th->th_dport; 3395 rewrite++; 3396 break; 3397 case IPPROTO_UDP: 3398 pf_change_ap(daddr, 3399 &pd->hdr.udp->uh_dport, pd->ip_sum, 3400 &pd->hdr.udp->uh_sum, &pd->baddr, 3401 bport, 1, af); 3402 dport = pd->hdr.udp->uh_dport; 3403 rewrite++; 3404 break; 3405 case IPPROTO_ICMP: 3406 #ifdef INET6 3407 case IPPROTO_ICMPV6: 3408 #endif 3409 /* nothing! */ 3410 break; 3411 default: 3412 switch (af) { 3413 case AF_INET: 3414 pf_change_a(&daddr->v4.s_addr, 3415 pd->ip_sum, 3416 pd->baddr.v4.s_addr, 0); 3417 break; 3418 case AF_INET6: 3419 PF_ACPY(daddr, &pd->baddr, af); 3420 break; 3421 } 3422 } 3423 } 3424 } 3425 if (pd->proto == IPPROTO_TCP && 3426 ((r->rule_flag & PFRULE_RETURNRST) || 3427 (r->rule_flag & PFRULE_RETURN)) && 3428 !(th->th_flags & TH_RST)) { 3429 u_int32_t ack = ntohl(th->th_seq) + pd->p_len; 3430 struct ip *h = mtod(m, struct ip *); 3431 3432 #ifdef __NetBSD__ 3433 if (pf_check_proto_cksum(m, direction, off, 3434 ntohs(h->ip_len) - off, IPPROTO_TCP, AF_INET)) 3435 #else 3436 if (pf_check_proto_cksum(m, off, 3437 ntohs(h->ip_len) - off, IPPROTO_TCP, AF_INET)) 3438 #endif /* !__NetBSD__ */ 3439 REASON_SET(&reason, PFRES_PROTCKSUM); 3440 else { 3441 if (th->th_flags & TH_SYN) 3442 ack++; 3443 if (th->th_flags & TH_FIN) 3444 ack++; 3445 pf_send_tcp(r, af, pd->dst, 3446 pd->src, th->th_dport, th->th_sport, 3447 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0, 3448 r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp); 3449 } 3450 } else if ((af == AF_INET) && r->return_icmp) 3451 pf_send_icmp(m, r->return_icmp >> 8, 3452 r->return_icmp & 255, af, r); 3453 else if ((af == AF_INET6) && r->return_icmp6) 3454 pf_send_icmp(m, r->return_icmp6 >> 8, 3455 r->return_icmp6 & 255, af, r); 3456 } 3457 3458 if (r->action == PF_DROP) 3459 return (PF_DROP); 3460 3461 if (pf_tag_packet(m, tag, rtableid)) { 3462 REASON_SET(&reason, PFRES_MEMORY); 3463 return (PF_DROP); 3464 } 3465 3466 if (!state_icmp && (r->keep_state || nr != NULL || 3467 (pd->flags & PFDESC_TCP_NORM))) { 3468 /* create new state */ 3469 u_int16_t len; 3470 struct pf_state *s = NULL; 3471 struct pf_state_key *sk = NULL; 3472 struct pf_src_node *sn = NULL; 3473 3474 /* check maximums */ 3475 if (r->max_states && (r->states >= r->max_states)) { 3476 pf_status.lcounters[LCNT_STATES]++; 3477 REASON_SET(&reason, PFRES_MAXSTATES); 3478 goto cleanup; 3479 } 3480 /* src node for filter rule */ 3481 if ((r->rule_flag & PFRULE_SRCTRACK || 3482 r->rpool.opts & PF_POOL_STICKYADDR) && 3483 pf_insert_src_node(&sn, r, saddr, af) != 0) { 3484 REASON_SET(&reason, PFRES_SRCLIMIT); 3485 goto cleanup; 3486 } 3487 /* src node for translation rule */ 3488 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 3489 ((direction == PF_OUT && 3490 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || 3491 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { 3492 REASON_SET(&reason, PFRES_SRCLIMIT); 3493 goto cleanup; 3494 } 3495 s = pool_get(&pf_state_pl, PR_NOWAIT); 3496 if (s == NULL) { 3497 REASON_SET(&reason, PFRES_MEMORY); 3498 cleanup: 3499 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 3500 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 3501 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3502 pf_status.src_nodes--; 3503 pool_put(&pf_src_tree_pl, sn); 3504 } 3505 if (nsn != sn && nsn != NULL && nsn->states == 0 && 3506 nsn->expire == 0) { 3507 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 3508 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3509 pf_status.src_nodes--; 3510 pool_put(&pf_src_tree_pl, nsn); 3511 } 3512 if (sk != NULL) { 3513 pool_put(&pf_state_key_pl, sk); 3514 } 3515 return (PF_DROP); 3516 } 3517 bzero(s, sizeof(*s)); 3518 s->rule.ptr = r; 3519 s->nat_rule.ptr = nr; 3520 s->anchor.ptr = a; 3521 STATE_INC_COUNTERS(s); 3522 s->allow_opts = r->allow_opts; 3523 s->log = r->log & PF_LOG_ALL; 3524 if (nr != NULL) 3525 s->log |= nr->log & PF_LOG_ALL; 3526 switch (pd->proto) { 3527 case IPPROTO_TCP: 3528 len = pd->tot_len - off - (th->th_off << 2); 3529 s->src.seqlo = ntohl(th->th_seq); 3530 s->src.seqhi = s->src.seqlo + len + 1; 3531 if ((th->th_flags & (TH_SYN|TH_ACK)) == 3532 TH_SYN && r->keep_state == PF_STATE_MODULATE) { 3533 /* Generate sequence number modulator */ 3534 while ((s->src.seqdiff = 3535 tcp_rndiss_next() - s->src.seqlo) == 0) 3536 ; 3537 pf_change_a(&th->th_seq, &th->th_sum, 3538 htonl(s->src.seqlo + s->src.seqdiff), 0); 3539 rewrite = 1; 3540 } else 3541 s->src.seqdiff = 0; 3542 if (th->th_flags & TH_SYN) { 3543 s->src.seqhi++; 3544 s->src.wscale = pf_get_wscale(m, off, 3545 th->th_off, af); 3546 } 3547 s->src.max_win = MAX(ntohs(th->th_win), 1); 3548 if (s->src.wscale & PF_WSCALE_MASK) { 3549 /* Remove scale factor from initial window */ 3550 int win = s->src.max_win; 3551 win += 1 << (s->src.wscale & PF_WSCALE_MASK); 3552 s->src.max_win = (win - 1) >> 3553 (s->src.wscale & PF_WSCALE_MASK); 3554 } 3555 if (th->th_flags & TH_FIN) 3556 s->src.seqhi++; 3557 s->dst.seqhi = 1; 3558 s->dst.max_win = 1; 3559 s->src.state = TCPS_SYN_SENT; 3560 s->dst.state = TCPS_CLOSED; 3561 s->timeout = PFTM_TCP_FIRST_PACKET; 3562 break; 3563 case IPPROTO_UDP: 3564 s->src.state = PFUDPS_SINGLE; 3565 s->dst.state = PFUDPS_NO_TRAFFIC; 3566 s->timeout = PFTM_UDP_FIRST_PACKET; 3567 break; 3568 case IPPROTO_ICMP: 3569 #ifdef INET6 3570 case IPPROTO_ICMPV6: 3571 #endif 3572 s->timeout = PFTM_ICMP_FIRST_PACKET; 3573 break; 3574 default: 3575 s->src.state = PFOTHERS_SINGLE; 3576 s->dst.state = PFOTHERS_NO_TRAFFIC; 3577 s->timeout = PFTM_OTHER_FIRST_PACKET; 3578 } 3579 3580 s->creation = time_second; 3581 s->expire = time_second; 3582 3583 if (sn != NULL) { 3584 s->src_node = sn; 3585 s->src_node->states++; 3586 } 3587 if (nsn != NULL) { 3588 PF_ACPY(&nsn->raddr, &pd->naddr, af); 3589 s->nat_src_node = nsn; 3590 s->nat_src_node->states++; 3591 } 3592 if (pd->proto == IPPROTO_TCP) { 3593 if ((pd->flags & PFDESC_TCP_NORM) && 3594 pf_normalize_tcp_init(m, off, pd, th, &s->src, 3595 &s->dst)) { 3596 REASON_SET(&reason, PFRES_MEMORY); 3597 pf_src_tree_remove_state(s); 3598 STATE_DEC_COUNTERS(s); 3599 pool_put(&pf_state_pl, s); 3600 return (PF_DROP); 3601 } 3602 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub && 3603 pf_normalize_tcp_stateful(m, off, pd, &reason, 3604 th, s, &s->src, &s->dst, &rewrite)) { 3605 /* This really shouldn't happen!!! */ 3606 DPFPRINTF(PF_DEBUG_URGENT, 3607 ("pf_normalize_tcp_stateful failed on " 3608 "first pkt")); 3609 pf_normalize_tcp_cleanup(s); 3610 pf_src_tree_remove_state(s); 3611 STATE_DEC_COUNTERS(s); 3612 pool_put(&pf_state_pl, s); 3613 return (PF_DROP); 3614 } 3615 } 3616 3617 if ((sk = pf_alloc_state_key(s)) == NULL) { 3618 REASON_SET(&reason, PFRES_MEMORY); 3619 goto cleanup; 3620 } 3621 3622 sk->proto = pd->proto; 3623 sk->direction = direction; 3624 sk->af = af; 3625 if (direction == PF_OUT) { 3626 PF_ACPY(&sk->gwy.addr, saddr, af); 3627 PF_ACPY(&sk->ext.addr, daddr, af); 3628 switch (pd->proto) { 3629 case IPPROTO_ICMP: 3630 #ifdef INET6 3631 case IPPROTO_ICMPV6: 3632 #endif 3633 sk->gwy.port = nport; 3634 sk->ext.port = 0; 3635 break; 3636 default: 3637 sk->gwy.port = sport; 3638 sk->ext.port = dport; 3639 } 3640 if (nr != NULL) { 3641 PF_ACPY(&sk->lan.addr, &pd->baddr, af); 3642 sk->lan.port = bport; 3643 } else { 3644 PF_ACPY(&sk->lan.addr, &sk->gwy.addr, af); 3645 sk->lan.port = sk->gwy.port; 3646 } 3647 } else { 3648 PF_ACPY(&sk->lan.addr, daddr, af); 3649 PF_ACPY(&sk->ext.addr, saddr, af); 3650 switch (pd->proto) { 3651 case IPPROTO_ICMP: 3652 #ifdef INET6 3653 case IPPROTO_ICMPV6: 3654 #endif 3655 sk->lan.port = nport; 3656 sk->ext.port = 0; 3657 break; 3658 default: 3659 sk->lan.port = dport; 3660 sk->ext.port = sport; 3661 } 3662 if (nr != NULL) { 3663 PF_ACPY(&sk->gwy.addr, &pd->baddr, af); 3664 sk->gwy.port = bport; 3665 } else { 3666 PF_ACPY(&sk->gwy.addr, &sk->lan.addr, af); 3667 sk->gwy.port = sk->lan.port; 3668 } 3669 } 3670 3671 pf_set_rt_ifp(s, saddr); /* needs s->state_key set */ 3672 3673 if (pf_insert_state(bound_iface(r, nr, kif), s)) { 3674 if (pd->proto == IPPROTO_TCP) 3675 pf_normalize_tcp_cleanup(s); 3676 REASON_SET(&reason, PFRES_STATEINS); 3677 pf_src_tree_remove_state(s); 3678 STATE_DEC_COUNTERS(s); 3679 pool_put(&pf_state_pl, s); 3680 return (PF_DROP); 3681 } else 3682 *sm = s; 3683 if (tag > 0) { 3684 pf_tag_ref(tag); 3685 s->tag = tag; 3686 } 3687 if (pd->proto == IPPROTO_TCP && 3688 (th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN && 3689 r->keep_state == PF_STATE_SYNPROXY) { 3690 s->src.state = PF_TCPS_PROXY_SRC; 3691 if (nr != NULL) { 3692 if (direction == PF_OUT) { 3693 pf_change_ap(saddr, &th->th_sport, 3694 pd->ip_sum, &th->th_sum, &pd->baddr, 3695 bport, 0, af); 3696 sport = th->th_sport; 3697 } else { 3698 pf_change_ap(daddr, &th->th_dport, 3699 pd->ip_sum, &th->th_sum, &pd->baddr, 3700 bport, 0, af); 3701 sport = th->th_dport; 3702 } 3703 } 3704 s->src.seqhi = htonl(arc4random()); 3705 /* Find mss option */ 3706 mss = pf_get_mss(m, off, th->th_off, af); 3707 mss = pf_calc_mss(saddr, af, mss); 3708 mss = pf_calc_mss(daddr, af, mss); 3709 s->src.mss = mss; 3710 pf_send_tcp(r, af, daddr, saddr, th->th_dport, 3711 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1, 3712 TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL); 3713 REASON_SET(&reason, PFRES_SYNPROXY); 3714 return (PF_SYNPROXY_DROP); 3715 } 3716 } 3717 3718 /* copy back packet headers if we performed NAT operations */ 3719 if (rewrite) 3720 m_copyback(m, off, hdrlen, pd->hdr.any); 3721 3722 return (PF_PASS); 3723 } 3724 3725 int 3726 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif, 3727 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am, 3728 struct pf_ruleset **rsm) 3729 { 3730 struct pf_rule *r, *a = NULL; 3731 struct pf_ruleset *ruleset = NULL; 3732 sa_family_t af = pd->af; 3733 u_short reason; 3734 int tag = -1; 3735 int asd = 0; 3736 int match = 0; 3737 3738 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 3739 while (r != NULL) { 3740 r->evaluations++; 3741 if (pfi_kif_match(r->kif, kif) == r->ifnot) 3742 r = r->skip[PF_SKIP_IFP].ptr; 3743 else if (r->direction && r->direction != direction) 3744 r = r->skip[PF_SKIP_DIR].ptr; 3745 else if (r->af && r->af != af) 3746 r = r->skip[PF_SKIP_AF].ptr; 3747 else if (r->proto && r->proto != pd->proto) 3748 r = r->skip[PF_SKIP_PROTO].ptr; 3749 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, 3750 r->src.neg, kif)) 3751 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 3752 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, 3753 r->dst.neg, NULL)) 3754 r = r->skip[PF_SKIP_DST_ADDR].ptr; 3755 else if (r->tos && !(r->tos == pd->tos)) 3756 r = TAILQ_NEXT(r, entries); 3757 else if (r->src.port_op || r->dst.port_op || 3758 r->flagset || r->type || r->code || 3759 r->os_fingerprint != PF_OSFP_ANY) 3760 r = TAILQ_NEXT(r, entries); 3761 else if (r->prob && r->prob <= arc4random()) 3762 r = TAILQ_NEXT(r, entries); 3763 else if (r->match_tag && !pf_match_tag(m, r, &tag)) 3764 r = TAILQ_NEXT(r, entries); 3765 else { 3766 if (r->anchor == NULL) { 3767 match = 1; 3768 *rm = r; 3769 *am = a; 3770 *rsm = ruleset; 3771 if ((*rm)->quick) 3772 break; 3773 r = TAILQ_NEXT(r, entries); 3774 } else 3775 pf_step_into_anchor(&asd, &ruleset, 3776 PF_RULESET_FILTER, &r, &a, &match); 3777 } 3778 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 3779 PF_RULESET_FILTER, &r, &a, &match)) 3780 break; 3781 } 3782 r = *rm; 3783 a = *am; 3784 ruleset = *rsm; 3785 3786 REASON_SET(&reason, PFRES_MATCH); 3787 3788 if (r->log) 3789 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset, 3790 pd); 3791 3792 if (r->action != PF_PASS) 3793 return (PF_DROP); 3794 3795 if (pf_tag_packet(m, tag, -1)) { 3796 REASON_SET(&reason, PFRES_MEMORY); 3797 return (PF_DROP); 3798 } 3799 3800 return (PF_PASS); 3801 } 3802 3803 int 3804 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif, 3805 struct mbuf *m, int off, void *h, struct pf_pdesc *pd, 3806 u_short *reason) 3807 { 3808 struct pf_state_key_cmp key; 3809 struct tcphdr *th = pd->hdr.tcp; 3810 u_int16_t win = ntohs(th->th_win); 3811 u_int32_t ack, end, seq, orig_seq; 3812 u_int8_t sws, dws; 3813 int ackskew; 3814 int copyback = 0; 3815 struct pf_state_peer *src, *dst; 3816 3817 key.af = pd->af; 3818 key.proto = IPPROTO_TCP; 3819 if (direction == PF_IN) { 3820 PF_ACPY(&key.ext.addr, pd->src, key.af); 3821 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 3822 key.ext.port = th->th_sport; 3823 key.gwy.port = th->th_dport; 3824 } else { 3825 PF_ACPY(&key.lan.addr, pd->src, key.af); 3826 PF_ACPY(&key.ext.addr, pd->dst, key.af); 3827 key.lan.port = th->th_sport; 3828 key.ext.port = th->th_dport; 3829 } 3830 3831 STATE_LOOKUP(); 3832 3833 if (direction == (*state)->state_key->direction) { 3834 src = &(*state)->src; 3835 dst = &(*state)->dst; 3836 } else { 3837 src = &(*state)->dst; 3838 dst = &(*state)->src; 3839 } 3840 3841 if ((*state)->src.state == PF_TCPS_PROXY_SRC) { 3842 if (direction != (*state)->state_key->direction) { 3843 REASON_SET(reason, PFRES_SYNPROXY); 3844 return (PF_SYNPROXY_DROP); 3845 } 3846 if (th->th_flags & TH_SYN) { 3847 if (ntohl(th->th_seq) != (*state)->src.seqlo) { 3848 REASON_SET(reason, PFRES_SYNPROXY); 3849 return (PF_DROP); 3850 } 3851 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, 3852 pd->src, th->th_dport, th->th_sport, 3853 (*state)->src.seqhi, ntohl(th->th_seq) + 1, 3854 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1, 3855 0, NULL, NULL); 3856 REASON_SET(reason, PFRES_SYNPROXY); 3857 return (PF_SYNPROXY_DROP); 3858 } else if (!(th->th_flags & TH_ACK) || 3859 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) || 3860 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) { 3861 REASON_SET(reason, PFRES_SYNPROXY); 3862 return (PF_DROP); 3863 } else if ((*state)->src_node != NULL && 3864 pf_src_connlimit(state)) { 3865 REASON_SET(reason, PFRES_SRCLIMIT); 3866 return (PF_DROP); 3867 } else 3868 (*state)->src.state = PF_TCPS_PROXY_DST; 3869 } 3870 if ((*state)->src.state == PF_TCPS_PROXY_DST) { 3871 struct pf_state_host *src, *dst; 3872 3873 if (direction == PF_OUT) { 3874 src = &(*state)->state_key->gwy; 3875 dst = &(*state)->state_key->ext; 3876 } else { 3877 src = &(*state)->state_key->ext; 3878 dst = &(*state)->state_key->lan; 3879 } 3880 if (direction == (*state)->state_key->direction) { 3881 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) || 3882 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) || 3883 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) { 3884 REASON_SET(reason, PFRES_SYNPROXY); 3885 return (PF_DROP); 3886 } 3887 (*state)->src.max_win = MAX(ntohs(th->th_win), 1); 3888 if ((*state)->dst.seqhi == 1) 3889 (*state)->dst.seqhi = htonl(arc4random()); 3890 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr, 3891 &dst->addr, src->port, dst->port, 3892 (*state)->dst.seqhi, 0, TH_SYN, 0, 3893 (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL); 3894 REASON_SET(reason, PFRES_SYNPROXY); 3895 return (PF_SYNPROXY_DROP); 3896 } else if (((th->th_flags & (TH_SYN|TH_ACK)) != 3897 (TH_SYN|TH_ACK)) || 3898 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) { 3899 REASON_SET(reason, PFRES_SYNPROXY); 3900 return (PF_DROP); 3901 } else { 3902 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1); 3903 (*state)->dst.seqlo = ntohl(th->th_seq); 3904 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, 3905 pd->src, th->th_dport, th->th_sport, 3906 ntohl(th->th_ack), ntohl(th->th_seq) + 1, 3907 TH_ACK, (*state)->src.max_win, 0, 0, 0, 3908 (*state)->tag, NULL, NULL); 3909 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr, 3910 &dst->addr, src->port, dst->port, 3911 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1, 3912 TH_ACK, (*state)->dst.max_win, 0, 0, 1, 3913 0, NULL, NULL); 3914 (*state)->src.seqdiff = (*state)->dst.seqhi - 3915 (*state)->src.seqlo; 3916 (*state)->dst.seqdiff = (*state)->src.seqhi - 3917 (*state)->dst.seqlo; 3918 (*state)->src.seqhi = (*state)->src.seqlo + 3919 (*state)->dst.max_win; 3920 (*state)->dst.seqhi = (*state)->dst.seqlo + 3921 (*state)->src.max_win; 3922 (*state)->src.wscale = (*state)->dst.wscale = 0; 3923 (*state)->src.state = (*state)->dst.state = 3924 TCPS_ESTABLISHED; 3925 REASON_SET(reason, PFRES_SYNPROXY); 3926 return (PF_SYNPROXY_DROP); 3927 } 3928 } 3929 3930 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) { 3931 sws = src->wscale & PF_WSCALE_MASK; 3932 dws = dst->wscale & PF_WSCALE_MASK; 3933 } else 3934 sws = dws = 0; 3935 3936 /* 3937 * Sequence tracking algorithm from Guido van Rooij's paper: 3938 * http://www.madison-gurkha.com/publications/tcp_filtering/ 3939 * tcp_filtering.ps 3940 */ 3941 3942 orig_seq = seq = ntohl(th->th_seq); 3943 if (src->seqlo == 0) { 3944 /* First packet from this end. Set its state */ 3945 3946 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) && 3947 src->scrub == NULL) { 3948 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) { 3949 REASON_SET(reason, PFRES_MEMORY); 3950 return (PF_DROP); 3951 } 3952 } 3953 3954 /* Deferred generation of sequence number modulator */ 3955 if (dst->seqdiff && !src->seqdiff) { 3956 while ((src->seqdiff = tcp_rndiss_next() - seq) == 0) 3957 ; 3958 ack = ntohl(th->th_ack) - dst->seqdiff; 3959 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq + 3960 src->seqdiff), 0); 3961 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0); 3962 copyback = 1; 3963 } else { 3964 ack = ntohl(th->th_ack); 3965 } 3966 3967 end = seq + pd->p_len; 3968 if (th->th_flags & TH_SYN) { 3969 end++; 3970 if (dst->wscale & PF_WSCALE_FLAG) { 3971 src->wscale = pf_get_wscale(m, off, th->th_off, 3972 pd->af); 3973 if (src->wscale & PF_WSCALE_FLAG) { 3974 /* Remove scale factor from initial 3975 * window */ 3976 sws = src->wscale & PF_WSCALE_MASK; 3977 win = ((u_int32_t)win + (1 << sws) - 1) 3978 >> sws; 3979 dws = dst->wscale & PF_WSCALE_MASK; 3980 } else { 3981 /* fixup other window */ 3982 dst->max_win <<= dst->wscale & 3983 PF_WSCALE_MASK; 3984 /* in case of a retrans SYN|ACK */ 3985 dst->wscale = 0; 3986 } 3987 } 3988 } 3989 if (th->th_flags & TH_FIN) 3990 end++; 3991 3992 src->seqlo = seq; 3993 if (src->state < TCPS_SYN_SENT) 3994 src->state = TCPS_SYN_SENT; 3995 3996 /* 3997 * May need to slide the window (seqhi may have been set by 3998 * the crappy stack check or if we picked up the connection 3999 * after establishment) 4000 */ 4001 if (src->seqhi == 1 || 4002 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi)) 4003 src->seqhi = end + MAX(1, dst->max_win << dws); 4004 if (win > src->max_win) 4005 src->max_win = win; 4006 4007 } else { 4008 ack = ntohl(th->th_ack) - dst->seqdiff; 4009 if (src->seqdiff) { 4010 /* Modulate sequence numbers */ 4011 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq + 4012 src->seqdiff), 0); 4013 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0); 4014 copyback = 1; 4015 } 4016 end = seq + pd->p_len; 4017 if (th->th_flags & TH_SYN) 4018 end++; 4019 if (th->th_flags & TH_FIN) 4020 end++; 4021 } 4022 4023 if ((th->th_flags & TH_ACK) == 0) { 4024 /* Let it pass through the ack skew check */ 4025 ack = dst->seqlo; 4026 } else if ((ack == 0 && 4027 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) || 4028 /* broken tcp stacks do not set ack */ 4029 (dst->state < TCPS_SYN_SENT)) { 4030 /* 4031 * Many stacks (ours included) will set the ACK number in an 4032 * FIN|ACK if the SYN times out -- no sequence to ACK. 4033 */ 4034 ack = dst->seqlo; 4035 } 4036 4037 if (seq == end) { 4038 /* Ease sequencing restrictions on no data packets */ 4039 seq = src->seqlo; 4040 end = seq; 4041 } 4042 4043 ackskew = dst->seqlo - ack; 4044 4045 4046 /* 4047 * Need to demodulate the sequence numbers in any TCP SACK options 4048 * (Selective ACK). We could optionally validate the SACK values 4049 * against the current ACK window, either forwards or backwards, but 4050 * I'm not confident that SACK has been implemented properly 4051 * everywhere. It wouldn't surprise me if several stacks accidently 4052 * SACK too far backwards of previously ACKed data. There really aren't 4053 * any security implications of bad SACKing unless the target stack 4054 * doesn't validate the option length correctly. Someone trying to 4055 * spoof into a TCP connection won't bother blindly sending SACK 4056 * options anyway. 4057 */ 4058 if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) { 4059 if (pf_modulate_sack(m, off, pd, th, dst)) 4060 copyback = 1; 4061 } 4062 4063 4064 #define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */ 4065 if (SEQ_GEQ(src->seqhi, end) && 4066 /* Last octet inside other's window space */ 4067 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) && 4068 /* Retrans: not more than one window back */ 4069 (ackskew >= -MAXACKWINDOW) && 4070 /* Acking not more than one reassembled fragment backwards */ 4071 (ackskew <= (MAXACKWINDOW << sws)) && 4072 /* Acking not more than one window forward */ 4073 ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo || 4074 (orig_seq == src->seqlo + 1) || (pd->flags & PFDESC_IP_REAS) == 0)) { 4075 /* Require an exact/+1 sequence match on resets when possible */ 4076 4077 if (dst->scrub || src->scrub) { 4078 if (pf_normalize_tcp_stateful(m, off, pd, reason, th, 4079 *state, src, dst, ©back)) 4080 return (PF_DROP); 4081 } 4082 4083 /* update max window */ 4084 if (src->max_win < win) 4085 src->max_win = win; 4086 /* synchronize sequencing */ 4087 if (SEQ_GT(end, src->seqlo)) 4088 src->seqlo = end; 4089 /* slide the window of what the other end can send */ 4090 if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) 4091 dst->seqhi = ack + MAX((win << sws), 1); 4092 4093 4094 /* update states */ 4095 if (th->th_flags & TH_SYN) 4096 if (src->state < TCPS_SYN_SENT) 4097 src->state = TCPS_SYN_SENT; 4098 if (th->th_flags & TH_FIN) 4099 if (src->state < TCPS_CLOSING) 4100 src->state = TCPS_CLOSING; 4101 if (th->th_flags & TH_ACK) { 4102 if (dst->state == TCPS_SYN_SENT) { 4103 dst->state = TCPS_ESTABLISHED; 4104 if (src->state == TCPS_ESTABLISHED && 4105 (*state)->src_node != NULL && 4106 pf_src_connlimit(state)) { 4107 REASON_SET(reason, PFRES_SRCLIMIT); 4108 return (PF_DROP); 4109 } 4110 } else if (dst->state == TCPS_CLOSING) 4111 dst->state = TCPS_FIN_WAIT_2; 4112 } 4113 if (th->th_flags & TH_RST) 4114 src->state = dst->state = TCPS_TIME_WAIT; 4115 4116 /* update expire time */ 4117 (*state)->expire = time_second; 4118 if (src->state >= TCPS_FIN_WAIT_2 && 4119 dst->state >= TCPS_FIN_WAIT_2) 4120 (*state)->timeout = PFTM_TCP_CLOSED; 4121 else if (src->state >= TCPS_CLOSING && 4122 dst->state >= TCPS_CLOSING) 4123 (*state)->timeout = PFTM_TCP_FIN_WAIT; 4124 else if (src->state < TCPS_ESTABLISHED || 4125 dst->state < TCPS_ESTABLISHED) 4126 (*state)->timeout = PFTM_TCP_OPENING; 4127 else if (src->state >= TCPS_CLOSING || 4128 dst->state >= TCPS_CLOSING) 4129 (*state)->timeout = PFTM_TCP_CLOSING; 4130 else 4131 (*state)->timeout = PFTM_TCP_ESTABLISHED; 4132 4133 /* Fall through to PASS packet */ 4134 4135 } else if ((dst->state < TCPS_SYN_SENT || 4136 dst->state >= TCPS_FIN_WAIT_2 || 4137 src->state >= TCPS_FIN_WAIT_2) && 4138 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) && 4139 /* Within a window forward of the originating packet */ 4140 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) { 4141 /* Within a window backward of the originating packet */ 4142 4143 /* 4144 * This currently handles three situations: 4145 * 1) Stupid stacks will shotgun SYNs before their peer 4146 * replies. 4147 * 2) When PF catches an already established stream (the 4148 * firewall rebooted, the state table was flushed, routes 4149 * changed...) 4150 * 3) Packets get funky immediately after the connection 4151 * closes (this should catch Solaris spurious ACK|FINs 4152 * that web servers like to spew after a close) 4153 * 4154 * This must be a little more careful than the above code 4155 * since packet floods will also be caught here. We don't 4156 * update the TTL here to mitigate the damage of a packet 4157 * flood and so the same code can handle awkward establishment 4158 * and a loosened connection close. 4159 * In the establishment case, a correct peer response will 4160 * validate the connection, go through the normal state code 4161 * and keep updating the state TTL. 4162 */ 4163 4164 if (pf_status.debug >= PF_DEBUG_MISC) { 4165 printf("pf: loose state match: "); 4166 pf_print_state(*state); 4167 pf_print_flags(th->th_flags); 4168 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d " 4169 "pkts=%llu:%llu\n", seq, orig_seq, ack, pd->p_len, 4170 ackskew, 4171 (unsigned long long int)(*state)->packets[0], 4172 (unsigned long long int)(*state)->packets[1]); 4173 } 4174 4175 if (dst->scrub || src->scrub) { 4176 if (pf_normalize_tcp_stateful(m, off, pd, reason, th, 4177 *state, src, dst, ©back)) 4178 return (PF_DROP); 4179 } 4180 4181 /* update max window */ 4182 if (src->max_win < win) 4183 src->max_win = win; 4184 /* synchronize sequencing */ 4185 if (SEQ_GT(end, src->seqlo)) 4186 src->seqlo = end; 4187 /* slide the window of what the other end can send */ 4188 if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) 4189 dst->seqhi = ack + MAX((win << sws), 1); 4190 4191 /* 4192 * Cannot set dst->seqhi here since this could be a shotgunned 4193 * SYN and not an already established connection. 4194 */ 4195 4196 if (th->th_flags & TH_FIN) 4197 if (src->state < TCPS_CLOSING) 4198 src->state = TCPS_CLOSING; 4199 if (th->th_flags & TH_RST) 4200 src->state = dst->state = TCPS_TIME_WAIT; 4201 4202 /* Fall through to PASS packet */ 4203 4204 } else { 4205 if ((*state)->dst.state == TCPS_SYN_SENT && 4206 (*state)->src.state == TCPS_SYN_SENT) { 4207 /* Send RST for state mismatches during handshake */ 4208 if (!(th->th_flags & TH_RST)) 4209 pf_send_tcp((*state)->rule.ptr, pd->af, 4210 pd->dst, pd->src, th->th_dport, 4211 th->th_sport, ntohl(th->th_ack), 0, 4212 TH_RST, 0, 0, 4213 (*state)->rule.ptr->return_ttl, 1, 0, 4214 pd->eh, kif->pfik_ifp); 4215 src->seqlo = 0; 4216 src->seqhi = 1; 4217 src->max_win = 1; 4218 } else if (pf_status.debug >= PF_DEBUG_MISC) { 4219 printf("pf: BAD state: "); 4220 pf_print_state(*state); 4221 pf_print_flags(th->th_flags); 4222 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d " 4223 "pkts=%llu:%llu dir=%s,%s\n", 4224 seq, orig_seq, ack, pd->p_len, ackskew, 4225 (unsigned long long int)(*state)->packets[0], 4226 (unsigned long long int)(*state)->packets[1], 4227 direction == PF_IN ? "in" : "out", 4228 direction == (*state)->state_key->direction ? 4229 "fwd" : "rev"); 4230 printf("pf: State failure on: %c %c %c %c | %c %c\n", 4231 SEQ_GEQ(src->seqhi, end) ? ' ' : '1', 4232 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ? 4233 ' ': '2', 4234 (ackskew >= -MAXACKWINDOW) ? ' ' : '3', 4235 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4', 4236 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5', 4237 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6'); 4238 } 4239 REASON_SET(reason, PFRES_BADSTATE); 4240 return (PF_DROP); 4241 } 4242 4243 /* Any packets which have gotten here are to be passed */ 4244 4245 /* translate source/destination address, if necessary */ 4246 if (STATE_TRANSLATE((*state)->state_key)) { 4247 if (direction == PF_OUT) 4248 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum, 4249 &th->th_sum, &(*state)->state_key->gwy.addr, 4250 (*state)->state_key->gwy.port, 0, pd->af); 4251 else 4252 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum, 4253 &th->th_sum, &(*state)->state_key->lan.addr, 4254 (*state)->state_key->lan.port, 0, pd->af); 4255 m_copyback(m, off, sizeof(*th), th); 4256 } else if (copyback) { 4257 /* Copyback sequence modulation or stateful scrub changes */ 4258 m_copyback(m, off, sizeof(*th), th); 4259 } 4260 4261 return (PF_PASS); 4262 } 4263 4264 int 4265 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif, 4266 struct mbuf *m, int off, void *h, struct pf_pdesc *pd) 4267 { 4268 struct pf_state_peer *src, *dst; 4269 struct pf_state_key_cmp key; 4270 struct udphdr *uh = pd->hdr.udp; 4271 4272 key.af = pd->af; 4273 key.proto = IPPROTO_UDP; 4274 if (direction == PF_IN) { 4275 PF_ACPY(&key.ext.addr, pd->src, key.af); 4276 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 4277 key.ext.port = uh->uh_sport; 4278 key.gwy.port = uh->uh_dport; 4279 } else { 4280 PF_ACPY(&key.lan.addr, pd->src, key.af); 4281 PF_ACPY(&key.ext.addr, pd->dst, key.af); 4282 key.lan.port = uh->uh_sport; 4283 key.ext.port = uh->uh_dport; 4284 } 4285 4286 STATE_LOOKUP(); 4287 4288 if (direction == (*state)->state_key->direction) { 4289 src = &(*state)->src; 4290 dst = &(*state)->dst; 4291 } else { 4292 src = &(*state)->dst; 4293 dst = &(*state)->src; 4294 } 4295 4296 /* update states */ 4297 if (src->state < PFUDPS_SINGLE) 4298 src->state = PFUDPS_SINGLE; 4299 if (dst->state == PFUDPS_SINGLE) 4300 dst->state = PFUDPS_MULTIPLE; 4301 4302 /* update expire time */ 4303 (*state)->expire = time_second; 4304 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE) 4305 (*state)->timeout = PFTM_UDP_MULTIPLE; 4306 else 4307 (*state)->timeout = PFTM_UDP_SINGLE; 4308 4309 /* translate source/destination address, if necessary */ 4310 if (STATE_TRANSLATE((*state)->state_key)) { 4311 if (direction == PF_OUT) 4312 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum, 4313 &uh->uh_sum, &(*state)->state_key->gwy.addr, 4314 (*state)->state_key->gwy.port, 1, pd->af); 4315 else 4316 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum, 4317 &uh->uh_sum, &(*state)->state_key->lan.addr, 4318 (*state)->state_key->lan.port, 1, pd->af); 4319 m_copyback(m, off, sizeof(*uh), uh); 4320 } 4321 4322 return (PF_PASS); 4323 } 4324 4325 int 4326 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif, 4327 struct mbuf *m, int off, void *h, struct pf_pdesc *pd, 4328 u_short *reason) 4329 { 4330 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 4331 u_int16_t icmpid = 0, *icmpsum; 4332 u_int8_t icmptype; 4333 int state_icmp = 0; 4334 struct pf_state_key_cmp key; 4335 4336 icmpsum = NULL; /* XXXGCC -Wunitialized m68k */ 4337 icmptype = 0; /* XXXGCC -Wunitialized m68k */ 4338 4339 switch (pd->proto) { 4340 #ifdef INET 4341 case IPPROTO_ICMP: 4342 icmptype = pd->hdr.icmp->icmp_type; 4343 icmpid = pd->hdr.icmp->icmp_id; 4344 icmpsum = &pd->hdr.icmp->icmp_cksum; 4345 4346 if (icmptype == ICMP_UNREACH || 4347 icmptype == ICMP_SOURCEQUENCH || 4348 icmptype == ICMP_REDIRECT || 4349 icmptype == ICMP_TIMXCEED || 4350 icmptype == ICMP_PARAMPROB) 4351 state_icmp++; 4352 break; 4353 #endif /* INET */ 4354 #ifdef INET6 4355 case IPPROTO_ICMPV6: 4356 icmptype = pd->hdr.icmp6->icmp6_type; 4357 icmpid = pd->hdr.icmp6->icmp6_id; 4358 icmpsum = &pd->hdr.icmp6->icmp6_cksum; 4359 4360 if (icmptype == ICMP6_DST_UNREACH || 4361 icmptype == ICMP6_PACKET_TOO_BIG || 4362 icmptype == ICMP6_TIME_EXCEEDED || 4363 icmptype == ICMP6_PARAM_PROB) 4364 state_icmp++; 4365 break; 4366 #endif /* INET6 */ 4367 } 4368 4369 if (!state_icmp) { 4370 4371 /* 4372 * ICMP query/reply message not related to a TCP/UDP packet. 4373 * Search for an ICMP state. 4374 */ 4375 key.af = pd->af; 4376 key.proto = pd->proto; 4377 if (direction == PF_IN) { 4378 PF_ACPY(&key.ext.addr, pd->src, key.af); 4379 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 4380 key.ext.port = 0; 4381 key.gwy.port = icmpid; 4382 } else { 4383 PF_ACPY(&key.lan.addr, pd->src, key.af); 4384 PF_ACPY(&key.ext.addr, pd->dst, key.af); 4385 key.lan.port = icmpid; 4386 key.ext.port = 0; 4387 } 4388 4389 STATE_LOOKUP(); 4390 4391 (*state)->expire = time_second; 4392 (*state)->timeout = PFTM_ICMP_ERROR_REPLY; 4393 4394 /* translate source/destination address, if necessary */ 4395 if (STATE_TRANSLATE((*state)->state_key)) { 4396 if (direction == PF_OUT) { 4397 switch (pd->af) { 4398 #ifdef INET 4399 case AF_INET: 4400 pf_change_a(&saddr->v4.s_addr, 4401 pd->ip_sum, 4402 (*state)->state_key->gwy.addr.v4.s_addr, 0); 4403 pd->hdr.icmp->icmp_cksum = 4404 pf_cksum_fixup( 4405 pd->hdr.icmp->icmp_cksum, icmpid, 4406 (*state)->state_key->gwy.port, 0); 4407 pd->hdr.icmp->icmp_id = 4408 (*state)->state_key->gwy.port; 4409 m_copyback(m, off, ICMP_MINLEN, 4410 pd->hdr.icmp); 4411 break; 4412 #endif /* INET */ 4413 #ifdef INET6 4414 case AF_INET6: 4415 pf_change_a6(saddr, 4416 &pd->hdr.icmp6->icmp6_cksum, 4417 &(*state)->state_key->gwy.addr, 0); 4418 m_copyback(m, off, 4419 sizeof(struct icmp6_hdr), 4420 pd->hdr.icmp6); 4421 break; 4422 #endif /* INET6 */ 4423 } 4424 } else { 4425 switch (pd->af) { 4426 #ifdef INET 4427 case AF_INET: 4428 pf_change_a(&daddr->v4.s_addr, 4429 pd->ip_sum, 4430 (*state)->state_key->lan.addr.v4.s_addr, 0); 4431 pd->hdr.icmp->icmp_cksum = 4432 pf_cksum_fixup( 4433 pd->hdr.icmp->icmp_cksum, icmpid, 4434 (*state)->state_key->lan.port, 0); 4435 pd->hdr.icmp->icmp_id = 4436 (*state)->state_key->lan.port; 4437 m_copyback(m, off, ICMP_MINLEN, 4438 pd->hdr.icmp); 4439 break; 4440 #endif /* INET */ 4441 #ifdef INET6 4442 case AF_INET6: 4443 pf_change_a6(daddr, 4444 &pd->hdr.icmp6->icmp6_cksum, 4445 &(*state)->state_key->lan.addr, 0); 4446 m_copyback(m, off, 4447 sizeof(struct icmp6_hdr), 4448 pd->hdr.icmp6); 4449 break; 4450 #endif /* INET6 */ 4451 } 4452 } 4453 } 4454 4455 return (PF_PASS); 4456 4457 } else { 4458 /* 4459 * ICMP error message in response to a TCP/UDP packet. 4460 * Extract the inner TCP/UDP header and search for that state. 4461 */ 4462 4463 struct pf_pdesc pd2; 4464 #ifdef INET 4465 struct ip h2; 4466 #endif /* INET */ 4467 #ifdef INET6 4468 struct ip6_hdr h2_6; 4469 int terminal = 0; 4470 #endif /* INET6 */ 4471 int ipoff2 = 0; 4472 int off2 = 0; 4473 4474 memset(&pd2, 0, sizeof pd2); /* XXX gcc */ 4475 4476 pd2.af = pd->af; 4477 switch (pd->af) { 4478 #ifdef INET 4479 case AF_INET: 4480 /* offset of h2 in mbuf chain */ 4481 ipoff2 = off + ICMP_MINLEN; 4482 4483 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2), 4484 NULL, reason, pd2.af)) { 4485 DPFPRINTF(PF_DEBUG_MISC, 4486 ("pf: ICMP error message too short " 4487 "(ip)\n")); 4488 return (PF_DROP); 4489 } 4490 /* 4491 * ICMP error messages don't refer to non-first 4492 * fragments 4493 */ 4494 if (h2.ip_off & htons(IP_OFFMASK)) { 4495 REASON_SET(reason, PFRES_FRAG); 4496 return (PF_DROP); 4497 } 4498 4499 /* offset of protocol header that follows h2 */ 4500 off2 = ipoff2 + (h2.ip_hl << 2); 4501 4502 pd2.proto = h2.ip_p; 4503 pd2.src = (struct pf_addr *)&h2.ip_src; 4504 pd2.dst = (struct pf_addr *)&h2.ip_dst; 4505 pd2.ip_sum = &h2.ip_sum; 4506 break; 4507 #endif /* INET */ 4508 #ifdef INET6 4509 case AF_INET6: 4510 ipoff2 = off + sizeof(struct icmp6_hdr); 4511 4512 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6), 4513 NULL, reason, pd2.af)) { 4514 DPFPRINTF(PF_DEBUG_MISC, 4515 ("pf: ICMP error message too short " 4516 "(ip6)\n")); 4517 return (PF_DROP); 4518 } 4519 pd2.proto = h2_6.ip6_nxt; 4520 pd2.src = (struct pf_addr *)&h2_6.ip6_src; 4521 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst; 4522 pd2.ip_sum = NULL; 4523 off2 = ipoff2 + sizeof(h2_6); 4524 do { 4525 switch (pd2.proto) { 4526 case IPPROTO_FRAGMENT: 4527 /* 4528 * ICMPv6 error messages for 4529 * non-first fragments 4530 */ 4531 REASON_SET(reason, PFRES_FRAG); 4532 return (PF_DROP); 4533 case IPPROTO_AH: 4534 case IPPROTO_HOPOPTS: 4535 case IPPROTO_ROUTING: 4536 case IPPROTO_DSTOPTS: { 4537 /* get next header and header length */ 4538 struct ip6_ext opt6; 4539 4540 if (!pf_pull_hdr(m, off2, &opt6, 4541 sizeof(opt6), NULL, reason, 4542 pd2.af)) { 4543 DPFPRINTF(PF_DEBUG_MISC, 4544 ("pf: ICMPv6 short opt\n")); 4545 return (PF_DROP); 4546 } 4547 if (pd2.proto == IPPROTO_AH) 4548 off2 += (opt6.ip6e_len + 2) * 4; 4549 else 4550 off2 += (opt6.ip6e_len + 1) * 8; 4551 pd2.proto = opt6.ip6e_nxt; 4552 /* goto the next header */ 4553 break; 4554 } 4555 default: 4556 terminal++; 4557 break; 4558 } 4559 } while (!terminal); 4560 break; 4561 #endif /* INET6 */ 4562 } 4563 4564 switch (pd2.proto) { 4565 case IPPROTO_TCP: { 4566 struct tcphdr th; 4567 u_int32_t seq; 4568 struct pf_state_peer *src, *dst; 4569 u_int8_t dws; 4570 int copyback = 0; 4571 4572 /* 4573 * Only the first 8 bytes of the TCP header can be 4574 * expected. Don't access any TCP header fields after 4575 * th_seq, an ackskew test is not possible. 4576 */ 4577 if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason, 4578 pd2.af)) { 4579 DPFPRINTF(PF_DEBUG_MISC, 4580 ("pf: ICMP error message too short " 4581 "(tcp)\n")); 4582 return (PF_DROP); 4583 } 4584 4585 key.af = pd2.af; 4586 key.proto = IPPROTO_TCP; 4587 if (direction == PF_IN) { 4588 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 4589 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 4590 key.ext.port = th.th_dport; 4591 key.gwy.port = th.th_sport; 4592 } else { 4593 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 4594 PF_ACPY(&key.ext.addr, pd2.src, key.af); 4595 key.lan.port = th.th_dport; 4596 key.ext.port = th.th_sport; 4597 } 4598 4599 STATE_LOOKUP(); 4600 4601 if (direction == (*state)->state_key->direction) { 4602 src = &(*state)->dst; 4603 dst = &(*state)->src; 4604 } else { 4605 src = &(*state)->src; 4606 dst = &(*state)->dst; 4607 } 4608 4609 if (src->wscale && dst->wscale) 4610 dws = dst->wscale & PF_WSCALE_MASK; 4611 else 4612 dws = 0; 4613 4614 /* Demodulate sequence number */ 4615 seq = ntohl(th.th_seq) - src->seqdiff; 4616 if (src->seqdiff) { 4617 pf_change_a(&th.th_seq, icmpsum, 4618 htonl(seq), 0); 4619 copyback = 1; 4620 } 4621 4622 if (!SEQ_GEQ(src->seqhi, seq) || 4623 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) { 4624 if (pf_status.debug >= PF_DEBUG_MISC) { 4625 printf("pf: BAD ICMP %d:%d ", 4626 icmptype, pd->hdr.icmp->icmp_code); 4627 pf_print_host(pd->src, 0, pd->af); 4628 printf(" -> "); 4629 pf_print_host(pd->dst, 0, pd->af); 4630 printf(" state: "); 4631 pf_print_state(*state); 4632 printf(" seq=%u\n", seq); 4633 } 4634 REASON_SET(reason, PFRES_BADSTATE); 4635 return (PF_DROP); 4636 } 4637 4638 if (STATE_TRANSLATE((*state)->state_key)) { 4639 if (direction == PF_IN) { 4640 pf_change_icmp(pd2.src, &th.th_sport, 4641 daddr, &(*state)->state_key->lan.addr, 4642 (*state)->state_key->lan.port, NULL, 4643 pd2.ip_sum, icmpsum, 4644 pd->ip_sum, 0, pd2.af); 4645 } else { 4646 pf_change_icmp(pd2.dst, &th.th_dport, 4647 saddr, &(*state)->state_key->gwy.addr, 4648 (*state)->state_key->gwy.port, NULL, 4649 pd2.ip_sum, icmpsum, 4650 pd->ip_sum, 0, pd2.af); 4651 } 4652 copyback = 1; 4653 } 4654 4655 if (copyback) { 4656 switch (pd2.af) { 4657 #ifdef INET 4658 case AF_INET: 4659 m_copyback(m, off, ICMP_MINLEN, 4660 pd->hdr.icmp); 4661 m_copyback(m, ipoff2, sizeof(h2), 4662 &h2); 4663 break; 4664 #endif /* INET */ 4665 #ifdef INET6 4666 case AF_INET6: 4667 m_copyback(m, off, 4668 sizeof(struct icmp6_hdr), 4669 pd->hdr.icmp6); 4670 m_copyback(m, ipoff2, sizeof(h2_6), 4671 &h2_6); 4672 break; 4673 #endif /* INET6 */ 4674 } 4675 m_copyback(m, off2, 8, &th); 4676 } 4677 4678 return (PF_PASS); 4679 break; 4680 } 4681 case IPPROTO_UDP: { 4682 struct udphdr uh; 4683 4684 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh), 4685 NULL, reason, pd2.af)) { 4686 DPFPRINTF(PF_DEBUG_MISC, 4687 ("pf: ICMP error message too short " 4688 "(udp)\n")); 4689 return (PF_DROP); 4690 } 4691 4692 key.af = pd2.af; 4693 key.proto = IPPROTO_UDP; 4694 if (direction == PF_IN) { 4695 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 4696 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 4697 key.ext.port = uh.uh_dport; 4698 key.gwy.port = uh.uh_sport; 4699 } else { 4700 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 4701 PF_ACPY(&key.ext.addr, pd2.src, key.af); 4702 key.lan.port = uh.uh_dport; 4703 key.ext.port = uh.uh_sport; 4704 } 4705 4706 STATE_LOOKUP(); 4707 4708 if (STATE_TRANSLATE((*state)->state_key)) { 4709 if (direction == PF_IN) { 4710 pf_change_icmp(pd2.src, &uh.uh_sport, 4711 daddr, 4712 &(*state)->state_key->lan.addr, 4713 (*state)->state_key->lan.port, 4714 &uh.uh_sum, 4715 pd2.ip_sum, icmpsum, 4716 pd->ip_sum, 1, pd2.af); 4717 } else { 4718 pf_change_icmp(pd2.dst, &uh.uh_dport, 4719 saddr, 4720 &(*state)->state_key->gwy.addr, 4721 (*state)->state_key->gwy.port, &uh.uh_sum, 4722 pd2.ip_sum, icmpsum, 4723 pd->ip_sum, 1, pd2.af); 4724 } 4725 switch (pd2.af) { 4726 #ifdef INET 4727 case AF_INET: 4728 m_copyback(m, off, ICMP_MINLEN, 4729 pd->hdr.icmp); 4730 m_copyback(m, ipoff2, sizeof(h2), &h2); 4731 break; 4732 #endif /* INET */ 4733 #ifdef INET6 4734 case AF_INET6: 4735 m_copyback(m, off, 4736 sizeof(struct icmp6_hdr), 4737 pd->hdr.icmp6); 4738 m_copyback(m, ipoff2, sizeof(h2_6), 4739 &h2_6); 4740 break; 4741 #endif /* INET6 */ 4742 } 4743 m_copyback(m, off2, sizeof(uh), &uh); 4744 } 4745 4746 return (PF_PASS); 4747 break; 4748 } 4749 #ifdef INET 4750 case IPPROTO_ICMP: { 4751 struct icmp iih; 4752 4753 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN, 4754 NULL, reason, pd2.af)) { 4755 DPFPRINTF(PF_DEBUG_MISC, 4756 ("pf: ICMP error message too short i" 4757 "(icmp)\n")); 4758 return (PF_DROP); 4759 } 4760 4761 key.af = pd2.af; 4762 key.proto = IPPROTO_ICMP; 4763 if (direction == PF_IN) { 4764 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 4765 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 4766 key.ext.port = 0; 4767 key.gwy.port = iih.icmp_id; 4768 } else { 4769 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 4770 PF_ACPY(&key.ext.addr, pd2.src, key.af); 4771 key.lan.port = iih.icmp_id; 4772 key.ext.port = 0; 4773 } 4774 4775 STATE_LOOKUP(); 4776 4777 if (STATE_TRANSLATE((*state)->state_key)) { 4778 if (direction == PF_IN) { 4779 pf_change_icmp(pd2.src, &iih.icmp_id, 4780 daddr, 4781 &(*state)->state_key->lan.addr, 4782 (*state)->state_key->lan.port, NULL, 4783 pd2.ip_sum, icmpsum, 4784 pd->ip_sum, 0, AF_INET); 4785 } else { 4786 pf_change_icmp(pd2.dst, &iih.icmp_id, 4787 saddr, 4788 &(*state)->state_key->gwy.addr, 4789 (*state)->state_key->gwy.port, NULL, 4790 pd2.ip_sum, icmpsum, 4791 pd->ip_sum, 0, AF_INET); 4792 } 4793 m_copyback(m, off, ICMP_MINLEN, pd->hdr.icmp); 4794 m_copyback(m, ipoff2, sizeof(h2), &h2); 4795 m_copyback(m, off2, ICMP_MINLEN, &iih); 4796 } 4797 4798 return (PF_PASS); 4799 break; 4800 } 4801 #endif /* INET */ 4802 #ifdef INET6 4803 case IPPROTO_ICMPV6: { 4804 struct icmp6_hdr iih; 4805 4806 if (!pf_pull_hdr(m, off2, &iih, 4807 sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) { 4808 DPFPRINTF(PF_DEBUG_MISC, 4809 ("pf: ICMP error message too short " 4810 "(icmp6)\n")); 4811 return (PF_DROP); 4812 } 4813 4814 key.af = pd2.af; 4815 key.proto = IPPROTO_ICMPV6; 4816 if (direction == PF_IN) { 4817 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 4818 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 4819 key.ext.port = 0; 4820 key.gwy.port = iih.icmp6_id; 4821 } else { 4822 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 4823 PF_ACPY(&key.ext.addr, pd2.src, key.af); 4824 key.lan.port = iih.icmp6_id; 4825 key.ext.port = 0; 4826 } 4827 4828 STATE_LOOKUP(); 4829 4830 if (STATE_TRANSLATE((*state)->state_key)) { 4831 if (direction == PF_IN) { 4832 pf_change_icmp(pd2.src, &iih.icmp6_id, 4833 daddr, 4834 &(*state)->state_key->lan.addr, 4835 (*state)->state_key->lan.port, NULL, 4836 pd2.ip_sum, icmpsum, 4837 pd->ip_sum, 0, AF_INET6); 4838 } else { 4839 pf_change_icmp(pd2.dst, &iih.icmp6_id, 4840 saddr, &(*state)->state_key->gwy.addr, 4841 (*state)->state_key->gwy.port, NULL, 4842 pd2.ip_sum, icmpsum, 4843 pd->ip_sum, 0, AF_INET6); 4844 } 4845 m_copyback(m, off, sizeof(struct icmp6_hdr), 4846 pd->hdr.icmp6); 4847 m_copyback(m, ipoff2, sizeof(h2_6), &h2_6); 4848 m_copyback(m, off2, sizeof(struct icmp6_hdr), 4849 &iih); 4850 } 4851 4852 return (PF_PASS); 4853 break; 4854 } 4855 #endif /* INET6 */ 4856 default: { 4857 key.af = pd2.af; 4858 key.proto = pd2.proto; 4859 if (direction == PF_IN) { 4860 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 4861 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 4862 key.ext.port = 0; 4863 key.gwy.port = 0; 4864 } else { 4865 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 4866 PF_ACPY(&key.ext.addr, pd2.src, key.af); 4867 key.lan.port = 0; 4868 key.ext.port = 0; 4869 } 4870 4871 STATE_LOOKUP(); 4872 4873 if (STATE_TRANSLATE((*state)->state_key)) { 4874 if (direction == PF_IN) { 4875 pf_change_icmp(pd2.src, NULL, 4876 daddr, 4877 &(*state)->state_key->lan.addr, 4878 0, NULL, 4879 pd2.ip_sum, icmpsum, 4880 pd->ip_sum, 0, pd2.af); 4881 } else { 4882 pf_change_icmp(pd2.dst, NULL, 4883 saddr, 4884 &(*state)->state_key->gwy.addr, 4885 0, NULL, 4886 pd2.ip_sum, icmpsum, 4887 pd->ip_sum, 0, pd2.af); 4888 } 4889 switch (pd2.af) { 4890 #ifdef INET 4891 case AF_INET: 4892 m_copyback(m, off, ICMP_MINLEN, 4893 pd->hdr.icmp); 4894 m_copyback(m, ipoff2, sizeof(h2), &h2); 4895 break; 4896 #endif /* INET */ 4897 #ifdef INET6 4898 case AF_INET6: 4899 m_copyback(m, off, 4900 sizeof(struct icmp6_hdr), 4901 pd->hdr.icmp6); 4902 m_copyback(m, ipoff2, sizeof(h2_6), 4903 &h2_6); 4904 break; 4905 #endif /* INET6 */ 4906 } 4907 } 4908 4909 return (PF_PASS); 4910 break; 4911 } 4912 } 4913 } 4914 } 4915 4916 int 4917 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif, 4918 struct pf_pdesc *pd) 4919 { 4920 struct pf_state_peer *src, *dst; 4921 struct pf_state_key_cmp key; 4922 4923 key.af = pd->af; 4924 key.proto = pd->proto; 4925 if (direction == PF_IN) { 4926 PF_ACPY(&key.ext.addr, pd->src, key.af); 4927 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 4928 key.ext.port = 0; 4929 key.gwy.port = 0; 4930 } else { 4931 PF_ACPY(&key.lan.addr, pd->src, key.af); 4932 PF_ACPY(&key.ext.addr, pd->dst, key.af); 4933 key.lan.port = 0; 4934 key.ext.port = 0; 4935 } 4936 4937 STATE_LOOKUP(); 4938 4939 if (direction == (*state)->state_key->direction) { 4940 src = &(*state)->src; 4941 dst = &(*state)->dst; 4942 } else { 4943 src = &(*state)->dst; 4944 dst = &(*state)->src; 4945 } 4946 4947 /* update states */ 4948 if (src->state < PFOTHERS_SINGLE) 4949 src->state = PFOTHERS_SINGLE; 4950 if (dst->state == PFOTHERS_SINGLE) 4951 dst->state = PFOTHERS_MULTIPLE; 4952 4953 /* update expire time */ 4954 (*state)->expire = time_second; 4955 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE) 4956 (*state)->timeout = PFTM_OTHER_MULTIPLE; 4957 else 4958 (*state)->timeout = PFTM_OTHER_SINGLE; 4959 4960 /* translate source/destination address, if necessary */ 4961 if (STATE_TRANSLATE((*state)->state_key)) { 4962 if (direction == PF_OUT) 4963 switch (pd->af) { 4964 #ifdef INET 4965 case AF_INET: 4966 pf_change_a(&pd->src->v4.s_addr, 4967 pd->ip_sum, 4968 (*state)->state_key->gwy.addr.v4.s_addr, 4969 0); 4970 break; 4971 #endif /* INET */ 4972 #ifdef INET6 4973 case AF_INET6: 4974 PF_ACPY(pd->src, 4975 &(*state)->state_key->gwy.addr, pd->af); 4976 break; 4977 #endif /* INET6 */ 4978 } 4979 else 4980 switch (pd->af) { 4981 #ifdef INET 4982 case AF_INET: 4983 pf_change_a(&pd->dst->v4.s_addr, 4984 pd->ip_sum, 4985 (*state)->state_key->lan.addr.v4.s_addr, 4986 0); 4987 break; 4988 #endif /* INET */ 4989 #ifdef INET6 4990 case AF_INET6: 4991 PF_ACPY(pd->dst, 4992 &(*state)->state_key->lan.addr, pd->af); 4993 break; 4994 #endif /* INET6 */ 4995 } 4996 } 4997 4998 return (PF_PASS); 4999 } 5000 5001 /* 5002 * ipoff and off are measured from the start of the mbuf chain. 5003 * h must be at "ipoff" on the mbuf chain. 5004 */ 5005 void * 5006 pf_pull_hdr(struct mbuf *m, int off, void *p, int len, 5007 u_short *actionp, u_short *reasonp, sa_family_t af) 5008 { 5009 switch (af) { 5010 #ifdef INET 5011 case AF_INET: { 5012 struct ip *h = mtod(m, struct ip *); 5013 u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3; 5014 5015 if (fragoff) { 5016 if (fragoff >= len) 5017 ACTION_SET(actionp, PF_PASS); 5018 else { 5019 ACTION_SET(actionp, PF_DROP); 5020 REASON_SET(reasonp, PFRES_FRAG); 5021 } 5022 return (NULL); 5023 } 5024 if (m->m_pkthdr.len < off + len || 5025 ntohs(h->ip_len) < off + len) { 5026 ACTION_SET(actionp, PF_DROP); 5027 REASON_SET(reasonp, PFRES_SHORT); 5028 return (NULL); 5029 } 5030 break; 5031 } 5032 #endif /* INET */ 5033 #ifdef INET6 5034 case AF_INET6: { 5035 struct ip6_hdr *h = mtod(m, struct ip6_hdr *); 5036 5037 if (m->m_pkthdr.len < off + len || 5038 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) < 5039 (unsigned)(off + len)) { 5040 ACTION_SET(actionp, PF_DROP); 5041 REASON_SET(reasonp, PFRES_SHORT); 5042 return (NULL); 5043 } 5044 break; 5045 } 5046 #endif /* INET6 */ 5047 } 5048 m_copydata(m, off, len, p); 5049 return (p); 5050 } 5051 5052 int 5053 pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif) 5054 { 5055 #ifdef __NetBSD__ 5056 union { 5057 struct sockaddr dst; 5058 struct sockaddr_in dst4; 5059 struct sockaddr_in6 dst6; 5060 } u; 5061 struct route ro; 5062 int ret = 1; 5063 5064 bzero(&ro, sizeof(ro)); 5065 switch (af) { 5066 case AF_INET: 5067 sockaddr_in_init(&u.dst4, &addr->v4, 0); 5068 break; 5069 #ifdef INET6 5070 case AF_INET6: 5071 sockaddr_in6_init(&u.dst6, &addr->v6, 0, 0, 0); 5072 break; 5073 #endif /* INET6 */ 5074 default: 5075 return (0); 5076 } 5077 rtcache_setdst(&ro, &u.dst); 5078 5079 ret = rtcache_init(&ro) != NULL ? 1 : 0; 5080 rtcache_free(&ro); 5081 5082 return (ret); 5083 #else /* !__NetBSD__ */ 5084 struct sockaddr_in *dst; 5085 int ret = 1; 5086 int check_mpath; 5087 extern int ipmultipath; 5088 #ifdef INET6 5089 extern int ip6_multipath; 5090 struct sockaddr_in6 *dst6; 5091 struct route_in6 ro; 5092 #else 5093 struct route ro; 5094 #endif 5095 struct radix_node *rn; 5096 struct rtentry *rt; 5097 struct ifnet *ifp; 5098 5099 check_mpath = 0; 5100 bzero(&ro, sizeof(ro)); 5101 switch (af) { 5102 case AF_INET: 5103 dst = satosin(&ro.ro_dst); 5104 dst->sin_family = AF_INET; 5105 dst->sin_len = sizeof(*dst); 5106 dst->sin_addr = addr->v4; 5107 if (ipmultipath) 5108 check_mpath = 1; 5109 break; 5110 #ifdef INET6 5111 case AF_INET6: 5112 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 5113 dst6->sin6_family = AF_INET6; 5114 dst6->sin6_len = sizeof(*dst6); 5115 dst6->sin6_addr = addr->v6; 5116 if (ip6_multipath) 5117 check_mpath = 1; 5118 break; 5119 #endif /* INET6 */ 5120 default: 5121 return (0); 5122 } 5123 5124 /* Skip checks for ipsec interfaces */ 5125 if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC) 5126 goto out; 5127 5128 rtalloc_noclone((struct route *)&ro, NO_CLONING); 5129 5130 if (ro.ro_rt != NULL) { 5131 /* No interface given, this is a no-route check */ 5132 if (kif == NULL) 5133 goto out; 5134 5135 if (kif->pfik_ifp == NULL) { 5136 ret = 0; 5137 goto out; 5138 } 5139 5140 /* Perform uRPF check if passed input interface */ 5141 ret = 0; 5142 rn = (struct radix_node *)ro.ro_rt; 5143 do { 5144 rt = (struct rtentry *)rn; 5145 if (rt->rt_ifp->if_type == IFT_CARP) 5146 ifp = rt->rt_ifp->if_carpdev; 5147 else 5148 ifp = rt->rt_ifp; 5149 5150 if (kif->pfik_ifp == ifp) 5151 ret = 1; 5152 rn = rn_mpath_next(rn); 5153 } while (check_mpath == 1 && rn != NULL && ret == 0); 5154 } else 5155 ret = 0; 5156 out: 5157 if (ro.ro_rt != NULL) 5158 RTFREE(ro.ro_rt); 5159 return (ret); 5160 #endif /* !__NetBSD__ */ 5161 } 5162 5163 int 5164 pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw) 5165 { 5166 #ifdef __NetBSD__ 5167 /* NetBSD doesn't have route labels. */ 5168 5169 return (0); 5170 #else 5171 struct sockaddr_in *dst; 5172 #ifdef INET6 5173 struct sockaddr_in6 *dst6; 5174 struct route_in6 ro; 5175 #else 5176 struct route ro; 5177 #endif 5178 int ret = 0; 5179 5180 bzero(&ro, sizeof(ro)); 5181 switch (af) { 5182 case AF_INET: 5183 dst = satosin(&ro.ro_dst); 5184 dst->sin_family = AF_INET; 5185 dst->sin_len = sizeof(*dst); 5186 dst->sin_addr = addr->v4; 5187 break; 5188 #ifdef INET6 5189 case AF_INET6: 5190 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 5191 dst6->sin6_family = AF_INET6; 5192 dst6->sin6_len = sizeof(*dst6); 5193 dst6->sin6_addr = addr->v6; 5194 break; 5195 #endif /* INET6 */ 5196 default: 5197 return (0); 5198 } 5199 5200 rtalloc_noclone((struct route *)&ro, NO_CLONING); 5201 5202 if (ro.ro_rt != NULL) { 5203 if (ro.ro_rt->rt_labelid == aw->v.rtlabel) 5204 ret = 1; 5205 RTFREE(ro.ro_rt); 5206 } 5207 5208 return (ret); 5209 #endif /* !__NetBSD__ */ 5210 } 5211 5212 #ifdef INET 5213 void 5214 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp, 5215 struct pf_state *s, struct pf_pdesc *pd) 5216 { 5217 struct mbuf *m0, *m1; 5218 struct route iproute; 5219 struct route *ro = NULL; 5220 const struct sockaddr *dst; 5221 union { 5222 struct sockaddr dst; 5223 struct sockaddr_in dst4; 5224 } u; 5225 struct ip *ip; 5226 struct ifnet *ifp = NULL; 5227 struct pf_addr naddr; 5228 struct pf_src_node *sn = NULL; 5229 int error = 0; 5230 #ifdef IPSEC 5231 struct m_tag *mtag; 5232 #endif /* IPSEC */ 5233 #ifdef __NetBSD__ 5234 struct pf_mtag *pf_mtag; 5235 #endif /* __NetBSD__ */ 5236 5237 if (m == NULL || *m == NULL || r == NULL || 5238 (dir != PF_IN && dir != PF_OUT) || oifp == NULL) 5239 panic("pf_route: invalid parameters"); 5240 5241 #ifdef __NetBSD__ 5242 if ((pf_mtag = pf_get_mtag(*m)) == NULL) { 5243 m0 = *m; 5244 *m = NULL; 5245 goto bad; 5246 } 5247 if (pf_mtag->routed++ > 3) { 5248 m0 = *m; 5249 *m = NULL; 5250 goto bad; 5251 } 5252 #else 5253 if ((*m)->m_pkthdr.pf.routed++ > 3) { 5254 m0 = *m; 5255 *m = NULL; 5256 goto bad; 5257 } 5258 #endif /* !__NetBSD__ */ 5259 5260 if (r->rt == PF_DUPTO) { 5261 if ((m0 = m_dup(*m, 0, M_COPYALL, M_NOWAIT)) == NULL) 5262 return; 5263 } else { 5264 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) 5265 return; 5266 m0 = *m; 5267 } 5268 5269 if (m0->m_len < sizeof(struct ip)) { 5270 DPFPRINTF(PF_DEBUG_URGENT, 5271 ("pf_route: m0->m_len < sizeof(struct ip)\n")); 5272 goto bad; 5273 } 5274 5275 ip = mtod(m0, struct ip *); 5276 5277 ro = &iproute; 5278 memset(ro, 0, sizeof(*ro)); 5279 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 5280 dst = &u.dst; 5281 rtcache_setdst(ro, dst); 5282 5283 if (r->rt == PF_FASTROUTE) { 5284 struct rtentry *rt; 5285 5286 rt = rtcache_init(ro); 5287 5288 if (rt == NULL) { 5289 ip_statinc(IP_STAT_NOROUTE); 5290 goto bad; 5291 } 5292 5293 ifp = rt->rt_ifp; 5294 rt->rt_use++; 5295 5296 if (rt->rt_flags & RTF_GATEWAY) 5297 dst = rt->rt_gateway; 5298 } else { 5299 if (TAILQ_EMPTY(&r->rpool.list)) { 5300 DPFPRINTF(PF_DEBUG_URGENT, 5301 ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n")); 5302 goto bad; 5303 } 5304 if (s == NULL) { 5305 pf_map_addr(AF_INET, r, 5306 (const struct pf_addr *)&ip->ip_src, 5307 &naddr, NULL, &sn); 5308 if (!PF_AZERO(&naddr, AF_INET)) 5309 u.dst4.sin_addr.s_addr = naddr.v4.s_addr; 5310 ifp = r->rpool.cur->kif ? 5311 r->rpool.cur->kif->pfik_ifp : NULL; 5312 } else { 5313 if (!PF_AZERO(&s->rt_addr, AF_INET)) 5314 u.dst4.sin_addr.s_addr = s->rt_addr.v4.s_addr; 5315 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; 5316 } 5317 } 5318 if (ifp == NULL) 5319 goto bad; 5320 5321 if (oifp != ifp) { 5322 if (pf_test(PF_OUT, ifp, &m0, NULL) != PF_PASS) 5323 goto bad; 5324 else if (m0 == NULL) 5325 goto done; 5326 if (m0->m_len < sizeof(struct ip)) { 5327 DPFPRINTF(PF_DEBUG_URGENT, 5328 ("pf_route: m0->m_len < sizeof(struct ip)\n")); 5329 goto bad; 5330 } 5331 ip = mtod(m0, struct ip *); 5332 } 5333 5334 /* Copied from ip_output. */ 5335 #ifdef IPSEC 5336 /* 5337 * If deferred crypto processing is needed, check that the 5338 * interface supports it. 5339 */ 5340 if ((mtag = m_tag_find(m0, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL)) 5341 != NULL && (ifp->if_capabilities & IFCAP_IPSEC) == 0) { 5342 /* Notify IPsec to do its own crypto. */ 5343 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 5344 goto bad; 5345 } 5346 #endif /* IPSEC */ 5347 5348 /* Catch routing changes wrt. hardware checksumming for TCP or UDP. */ 5349 #ifdef __NetBSD__ 5350 if (m0->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 5351 in_delayed_cksum(m0); 5352 m0->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 5353 } 5354 #else 5355 if (m0->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT) { 5356 if (!(ifp->if_capabilities & IFCAP_CSUM_TCPv4) || 5357 ifp->if_bridge != NULL) { 5358 in_delayed_cksum(m0); 5359 m0->m_pkthdr.csum_flags &= ~M_TCPV4_CSUM_OUT; /* Clear */ 5360 } 5361 } else if (m0->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT) { 5362 if (!(ifp->if_capabilities & IFCAP_CSUM_UDPv4) || 5363 ifp->if_bridge != NULL) { 5364 in_delayed_cksum(m0); 5365 m0->m_pkthdr.csum_flags &= ~M_UDPV4_CSUM_OUT; /* Clear */ 5366 } 5367 } 5368 #endif /* !__NetBSD__ */ 5369 5370 if (ntohs(ip->ip_len) <= ifp->if_mtu) { 5371 #ifdef __NetBSD__ 5372 ip->ip_sum = 0; 5373 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2); 5374 5375 m0->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 5376 #else 5377 if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) && 5378 ifp->if_bridge == NULL) { 5379 m0->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT; 5380 ipstat.ips_outhwcsum++; 5381 } else { 5382 ip->ip_sum = 0; 5383 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2); 5384 } 5385 /* Update relevant hardware checksum stats for TCP/UDP */ 5386 if (m0->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT) 5387 tcpstat.tcps_outhwcsum++; 5388 else if (m0->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT) 5389 udpstat.udps_outhwcsum++; 5390 #endif /* !__NetBSD__ */ 5391 error = (*ifp->if_output)(ifp, m0, dst, NULL); 5392 goto done; 5393 } 5394 5395 /* 5396 * Too large for interface; fragment if possible. 5397 * Must be able to put at least 8 bytes per fragment. 5398 */ 5399 if (ip->ip_off & htons(IP_DF)) { 5400 ip_statinc(IP_STAT_CANTFRAG); 5401 if (r->rt != PF_DUPTO) { 5402 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0, 5403 ifp->if_mtu); 5404 goto done; 5405 } else 5406 goto bad; 5407 } 5408 5409 #ifdef __NetBSD__ 5410 /* Make ip_fragment re-compute checksums. */ 5411 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) { 5412 m0->m_pkthdr.csum_flags |= M_CSUM_IPv4; 5413 } 5414 #endif /* __NetBSD__ */ 5415 m1 = m0; 5416 error = ip_fragment(m0, ifp, ifp->if_mtu); 5417 if (error) { 5418 m0 = NULL; 5419 goto bad; 5420 } 5421 5422 for (m0 = m1; m0; m0 = m1) { 5423 m1 = m0->m_nextpkt; 5424 m0->m_nextpkt = 0; 5425 if (error == 0) 5426 error = (*ifp->if_output)(ifp, m0, dst, NULL); 5427 else 5428 m_freem(m0); 5429 } 5430 5431 if (error == 0) 5432 ip_statinc(IP_STAT_FRAGMENTED); 5433 5434 done: 5435 if (r->rt != PF_DUPTO) 5436 *m = NULL; 5437 if (ro == &iproute) 5438 rtcache_free(ro); 5439 return; 5440 5441 bad: 5442 m_freem(m0); 5443 goto done; 5444 } 5445 #endif /* INET */ 5446 5447 #ifdef INET6 5448 void 5449 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp, 5450 struct pf_state *s, struct pf_pdesc *pd) 5451 { 5452 struct mbuf *m0; 5453 struct sockaddr_in6 dst; 5454 struct ip6_hdr *ip6; 5455 struct ifnet *ifp = NULL; 5456 struct pf_addr naddr; 5457 struct pf_src_node *sn = NULL; 5458 int error = 0; 5459 #ifdef __NetBSD__ 5460 struct pf_mtag *pf_mtag; 5461 #endif /* __NetBSD__ */ 5462 5463 if (m == NULL || *m == NULL || r == NULL || 5464 (dir != PF_IN && dir != PF_OUT) || oifp == NULL) 5465 panic("pf_route6: invalid parameters"); 5466 5467 #ifdef __NetBSD__ 5468 if ((pf_mtag = pf_get_mtag(*m)) == NULL) { 5469 m0 = *m; 5470 *m = NULL; 5471 goto bad; 5472 } 5473 if (pf_mtag->routed++ > 3) { 5474 m0 = *m; 5475 *m = NULL; 5476 goto bad; 5477 } 5478 #else 5479 if ((*m)->m_pkthdr.pf.routed++ > 3) { 5480 m0 = *m; 5481 *m = NULL; 5482 goto bad; 5483 } 5484 #endif /* !__NetBSD__ */ 5485 5486 if (r->rt == PF_DUPTO) { 5487 if ((m0 = m_dup(*m, 0, M_COPYALL, M_NOWAIT)) == NULL) 5488 return; 5489 } else { 5490 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) 5491 return; 5492 m0 = *m; 5493 } 5494 5495 if (m0->m_len < sizeof(struct ip6_hdr)) { 5496 DPFPRINTF(PF_DEBUG_URGENT, 5497 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n")); 5498 goto bad; 5499 } 5500 ip6 = mtod(m0, struct ip6_hdr *); 5501 5502 dst.sin6_family = AF_INET6; 5503 dst.sin6_len = sizeof(dst); 5504 dst.sin6_addr = ip6->ip6_dst; 5505 5506 /* Cheat. XXX why only in the v6 case??? */ 5507 if (r->rt == PF_FASTROUTE) { 5508 #ifdef __NetBSD__ 5509 pf_mtag->flags |= PF_TAG_GENERATED; 5510 #else 5511 m0->m_pkthdr.pf.flags |= PF_TAG_GENERATED; 5512 #endif /* !__NetBSD__ */ 5513 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL); 5514 return; 5515 } 5516 5517 if (TAILQ_EMPTY(&r->rpool.list)) { 5518 DPFPRINTF(PF_DEBUG_URGENT, 5519 ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n")); 5520 goto bad; 5521 } 5522 if (s == NULL) { 5523 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src, 5524 &naddr, NULL, &sn); 5525 if (!PF_AZERO(&naddr, AF_INET6)) 5526 PF_ACPY((struct pf_addr *)&dst.sin6_addr, 5527 &naddr, AF_INET6); 5528 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL; 5529 } else { 5530 if (!PF_AZERO(&s->rt_addr, AF_INET6)) 5531 PF_ACPY((struct pf_addr *)&dst.sin6_addr, 5532 &s->rt_addr, AF_INET6); 5533 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; 5534 } 5535 if (ifp == NULL) 5536 goto bad; 5537 5538 if (oifp != ifp) { 5539 if (pf_test6(PF_OUT, ifp, &m0, NULL) != PF_PASS) 5540 goto bad; 5541 else if (m0 == NULL) 5542 goto done; 5543 if (m0->m_len < sizeof(struct ip6_hdr)) { 5544 DPFPRINTF(PF_DEBUG_URGENT, 5545 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n")); 5546 goto bad; 5547 } 5548 ip6 = mtod(m0, struct ip6_hdr *); 5549 } 5550 5551 /* 5552 * If the packet is too large for the outgoing interface, 5553 * send back an icmp6 error. 5554 */ 5555 if (IN6_IS_SCOPE_EMBEDDABLE(&dst.sin6_addr)) 5556 dst.sin6_addr.s6_addr16[1] = htons(ifp->if_index); 5557 if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) { 5558 error = nd6_output(ifp, ifp, m0, &dst, NULL); 5559 } else { 5560 in6_ifstat_inc(ifp, ifs6_in_toobig); 5561 if (r->rt != PF_DUPTO) 5562 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu); 5563 else 5564 goto bad; 5565 } 5566 5567 done: 5568 if (r->rt != PF_DUPTO) 5569 *m = NULL; 5570 return; 5571 5572 bad: 5573 m_freem(m0); 5574 goto done; 5575 } 5576 #endif /* INET6 */ 5577 5578 5579 /* 5580 * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag 5581 * off is the offset where the protocol header starts 5582 * len is the total length of protocol header plus payload 5583 * returns 0 when the checksum is valid, otherwise returns 1. 5584 */ 5585 #ifdef __NetBSD__ 5586 int 5587 pf_check_proto_cksum(struct mbuf *m, int direction, int off, int len, 5588 u_int8_t p, sa_family_t af) 5589 #else 5590 int 5591 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p, 5592 sa_family_t af) 5593 #endif /* !__NetBSD__ */ 5594 { 5595 #ifndef __NetBSD__ 5596 u_int16_t flag_ok, flag_bad; 5597 #endif /* !__NetBSD__ */ 5598 u_int16_t sum; 5599 5600 #ifndef __NetBSD__ 5601 switch (p) { 5602 case IPPROTO_TCP: 5603 flag_ok = M_TCP_CSUM_IN_OK; 5604 flag_bad = M_TCP_CSUM_IN_BAD; 5605 break; 5606 case IPPROTO_UDP: 5607 flag_ok = M_UDP_CSUM_IN_OK; 5608 flag_bad = M_UDP_CSUM_IN_BAD; 5609 break; 5610 case IPPROTO_ICMP: 5611 #ifdef INET6 5612 case IPPROTO_ICMPV6: 5613 #endif /* INET6 */ 5614 flag_ok = flag_bad = 0; 5615 break; 5616 default: 5617 return (1); 5618 } 5619 if (m->m_pkthdr.csum_flags & flag_ok) 5620 return (0); 5621 if (m->m_pkthdr.csum_flags & flag_bad) 5622 return (1); 5623 #endif /* !__NetBSD__ */ 5624 if (off < sizeof(struct ip) || len < sizeof(struct udphdr)) 5625 return (1); 5626 if (m->m_pkthdr.len < off + len) 5627 return (1); 5628 #ifdef __NetBSD__ 5629 if (direction == PF_IN) { 5630 switch (p) { 5631 case IPPROTO_TCP: { 5632 struct tcphdr th; /* XXX */ 5633 int thlen; 5634 5635 m_copydata(m, off, sizeof(th), &th); /* XXX */ 5636 thlen = th.th_off << 2; 5637 return tcp_input_checksum(af, m, &th, off, 5638 thlen, len - thlen) != 0; 5639 } 5640 5641 case IPPROTO_UDP: { 5642 struct udphdr uh; /* XXX */ 5643 5644 m_copydata(m, off, sizeof(uh), &uh); /* XXX */ 5645 return udp_input_checksum(af, m, &uh, off, len) != 0; 5646 } 5647 } 5648 } 5649 #endif /* __NetBSD__ */ 5650 switch (af) { 5651 #ifdef INET 5652 case AF_INET: 5653 if (p == IPPROTO_ICMP) { 5654 if (m->m_len < off) 5655 return (1); 5656 m->m_data += off; 5657 m->m_len -= off; 5658 sum = in_cksum(m, len); 5659 m->m_data -= off; 5660 m->m_len += off; 5661 } else { 5662 if (m->m_len < sizeof(struct ip)) 5663 return (1); 5664 sum = in4_cksum(m, p, off, len); 5665 } 5666 break; 5667 #endif /* INET */ 5668 #ifdef INET6 5669 case AF_INET6: 5670 if (m->m_len < sizeof(struct ip6_hdr)) 5671 return (1); 5672 sum = in6_cksum(m, p, off, len); 5673 break; 5674 #endif /* INET6 */ 5675 default: 5676 return (1); 5677 } 5678 if (sum) { 5679 #ifndef __NetBSD__ 5680 m->m_pkthdr.csum_flags |= flag_bad; 5681 #endif /* !__NetBSD__ */ 5682 switch (p) { 5683 case IPPROTO_TCP: 5684 tcp_statinc(TCP_STAT_RCVBADSUM); 5685 break; 5686 case IPPROTO_UDP: 5687 udp_statinc(UDP_STAT_BADSUM); 5688 break; 5689 case IPPROTO_ICMP: 5690 icmp_statinc(ICMP_STAT_CHECKSUM); 5691 break; 5692 #ifdef INET6 5693 case IPPROTO_ICMPV6: 5694 icmp6_statinc(ICMP6_STAT_CHECKSUM); 5695 break; 5696 #endif /* INET6 */ 5697 } 5698 return (1); 5699 } 5700 #ifndef __NetBSD__ 5701 m->m_pkthdr.csum_flags |= flag_ok; 5702 #endif /* !__NetBSD__ */ 5703 return (0); 5704 } 5705 5706 #ifdef INET 5707 int 5708 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0, 5709 struct ether_header *eh) 5710 { 5711 struct pfi_kif *kif; 5712 u_short action, reason = 0, log = 0; 5713 struct mbuf *m = *m0; 5714 struct ip *h = NULL; 5715 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr; 5716 struct pf_state *s = NULL; 5717 struct pf_state_key *sk = NULL; 5718 struct pf_ruleset *ruleset = NULL; 5719 struct pf_pdesc pd; 5720 int off, dirndx, pqid = 0; 5721 #ifdef __NetBSD__ 5722 struct pf_mtag *pf_mtag = NULL; /* XXX gcc */ 5723 #endif /* __NetBSD__ */ 5724 5725 if (!pf_status.running) 5726 return (PF_PASS); 5727 5728 memset(&pd, 0, sizeof(pd)); 5729 if (ifp->if_type == IFT_CARP && ifp->if_carpdev) 5730 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif; 5731 else 5732 kif = (struct pfi_kif *)ifp->if_pf_kif; 5733 5734 if (kif == NULL) { 5735 DPFPRINTF(PF_DEBUG_URGENT, 5736 ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname)); 5737 return (PF_DROP); 5738 } 5739 if (kif->pfik_flags & PFI_IFLAG_SKIP) 5740 return (PF_PASS); 5741 5742 #ifdef DIAGNOSTIC 5743 if ((m->m_flags & M_PKTHDR) == 0) 5744 panic("non-M_PKTHDR is passed to pf_test"); 5745 #endif /* DIAGNOSTIC */ 5746 5747 if (m->m_pkthdr.len < (int)sizeof(*h)) { 5748 action = PF_DROP; 5749 REASON_SET(&reason, PFRES_SHORT); 5750 log = 1; 5751 goto done; 5752 } 5753 5754 #ifdef __NetBSD__ 5755 if ((pf_mtag = pf_get_mtag(m)) == NULL) { 5756 DPFPRINTF(PF_DEBUG_URGENT, 5757 ("pf_test: pf_get_mtag returned NULL\n")); 5758 return (PF_DROP); 5759 } 5760 if (pf_mtag->flags & PF_TAG_GENERATED) 5761 return (PF_PASS); 5762 #else 5763 if (m->m_pkthdr.pf.flags & PF_TAG_GENERATED) 5764 return (PF_PASS); 5765 #endif /* !__NetBSD__ */ 5766 5767 /* We do IP header normalization and packet reassembly here */ 5768 if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) { 5769 action = PF_DROP; 5770 goto done; 5771 } 5772 m = *m0; /* pf_normalize messes with m0 */ 5773 h = mtod(m, struct ip *); 5774 5775 off = h->ip_hl << 2; 5776 if (off < (int)sizeof(*h)) { 5777 action = PF_DROP; 5778 REASON_SET(&reason, PFRES_SHORT); 5779 log = 1; 5780 goto done; 5781 } 5782 5783 pd.src = (struct pf_addr *)&h->ip_src; 5784 pd.dst = (struct pf_addr *)&h->ip_dst; 5785 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET); 5786 pd.ip_sum = &h->ip_sum; 5787 pd.proto = h->ip_p; 5788 pd.af = AF_INET; 5789 pd.tos = h->ip_tos; 5790 pd.tot_len = ntohs(h->ip_len); 5791 pd.eh = eh; 5792 5793 /* handle fragments that didn't get reassembled by normalization */ 5794 if (h->ip_off & htons(IP_MF | IP_OFFMASK)) { 5795 action = pf_test_fragment(&r, dir, kif, m, h, 5796 &pd, &a, &ruleset); 5797 goto done; 5798 } 5799 5800 switch (h->ip_p) { 5801 5802 case IPPROTO_TCP: { 5803 struct tcphdr th; 5804 5805 pd.hdr.tcp = &th; 5806 if (!pf_pull_hdr(m, off, &th, sizeof(th), 5807 &action, &reason, AF_INET)) { 5808 log = action != PF_PASS; 5809 goto done; 5810 } 5811 pd.p_len = pd.tot_len - off - (th.th_off << 2); 5812 if ((th.th_flags & TH_ACK) && pd.p_len == 0) 5813 pqid = 1; 5814 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd); 5815 if (action == PF_DROP) 5816 goto done; 5817 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd, 5818 &reason); 5819 if (action == PF_PASS) { 5820 #if NPFSYNC 5821 pfsync_update_state(s); 5822 #endif /* NPFSYNC */ 5823 r = s->rule.ptr; 5824 a = s->anchor.ptr; 5825 log = s->log; 5826 } else if (s == NULL) 5827 action = pf_test_rule(&r, &s, dir, kif, 5828 m, off, h, &pd, &a, &ruleset, &ipintrq); 5829 break; 5830 } 5831 5832 case IPPROTO_UDP: { 5833 struct udphdr uh; 5834 5835 pd.hdr.udp = &uh; 5836 if (!pf_pull_hdr(m, off, &uh, sizeof(uh), 5837 &action, &reason, AF_INET)) { 5838 log = action != PF_PASS; 5839 goto done; 5840 } 5841 if (uh.uh_dport == 0 || 5842 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off || 5843 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) { 5844 action = PF_DROP; 5845 REASON_SET(&reason, PFRES_SHORT); 5846 goto done; 5847 } 5848 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd); 5849 if (action == PF_PASS) { 5850 #if NPFSYNC 5851 pfsync_update_state(s); 5852 #endif /* NPFSYNC */ 5853 r = s->rule.ptr; 5854 a = s->anchor.ptr; 5855 log = s->log; 5856 } else if (s == NULL) 5857 action = pf_test_rule(&r, &s, dir, kif, 5858 m, off, h, &pd, &a, &ruleset, &ipintrq); 5859 break; 5860 } 5861 5862 case IPPROTO_ICMP: { 5863 struct icmp ih; 5864 5865 pd.hdr.icmp = &ih; 5866 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN, 5867 &action, &reason, AF_INET)) { 5868 log = action != PF_PASS; 5869 goto done; 5870 } 5871 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd, 5872 &reason); 5873 if (action == PF_PASS) { 5874 #if NPFSYNC 5875 pfsync_update_state(s); 5876 #endif /* NPFSYNC */ 5877 r = s->rule.ptr; 5878 a = s->anchor.ptr; 5879 log = s->log; 5880 } else if (s == NULL) 5881 action = pf_test_rule(&r, &s, dir, kif, 5882 m, off, h, &pd, &a, &ruleset, &ipintrq); 5883 break; 5884 } 5885 5886 default: 5887 action = pf_test_state_other(&s, dir, kif, &pd); 5888 if (action == PF_PASS) { 5889 #if NPFSYNC 5890 pfsync_update_state(s); 5891 #endif /* NPFSYNC */ 5892 r = s->rule.ptr; 5893 a = s->anchor.ptr; 5894 log = s->log; 5895 } else if (s == NULL) 5896 action = pf_test_rule(&r, &s, dir, kif, m, off, h, 5897 &pd, &a, &ruleset, &ipintrq); 5898 break; 5899 } 5900 5901 done: 5902 if (action == PF_PASS && h->ip_hl > 5 && 5903 !((s && s->allow_opts) || r->allow_opts)) { 5904 action = PF_DROP; 5905 REASON_SET(&reason, PFRES_IPOPTIONS); 5906 log = 1; 5907 DPFPRINTF(PF_DEBUG_MISC, 5908 ("pf: dropping packet with ip options\n")); 5909 } 5910 5911 if ((s && s->tag) || r->rtableid) 5912 pf_tag_packet(m, s ? s->tag : 0, r->rtableid); 5913 5914 #ifdef ALTQ 5915 if (action == PF_PASS && r->qid) { 5916 #ifdef __NetBSD__ 5917 struct m_tag *mtag; 5918 struct altq_tag *atag; 5919 5920 mtag = m_tag_get(PACKET_TAG_ALTQ_QID, sizeof(*atag), M_NOWAIT); 5921 if (mtag != NULL) { 5922 atag = (struct altq_tag *)(mtag + 1); 5923 if (pqid || (pd.tos & IPTOS_LOWDELAY)) 5924 atag->qid = r->pqid; 5925 else 5926 atag->qid = r->qid; 5927 /* add hints for ecn */ 5928 atag->af = AF_INET; 5929 atag->hdr = h; 5930 m_tag_prepend(m, mtag); 5931 } 5932 #else 5933 if (pqid || (pd.tos & IPTOS_LOWDELAY)) 5934 m->m_pkthdr.pf.qid = r->pqid; 5935 else 5936 m->m_pkthdr.pf.qid = r->qid; 5937 /* add hints for ecn */ 5938 m->m_pkthdr.pf.hdr = h; 5939 #endif /* !__NetBSD__ */ 5940 } 5941 #endif /* ALTQ */ 5942 5943 /* 5944 * connections redirected to loopback should not match sockets 5945 * bound specifically to loopback due to security implications, 5946 * see tcp_input() and in_pcblookup_listen(). 5947 */ 5948 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP || 5949 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL && 5950 (s->nat_rule.ptr->action == PF_RDR || 5951 s->nat_rule.ptr->action == PF_BINAT) && 5952 (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) 5953 #ifdef __NetBSD__ 5954 pf_mtag->flags |= PF_TAG_TRANSLATE_LOCALHOST; 5955 #else 5956 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST; 5957 #endif /* !__NetBSD__ */ 5958 5959 if (log) { 5960 struct pf_rule *lr; 5961 5962 if (s != NULL && s->nat_rule.ptr != NULL && 5963 s->nat_rule.ptr->log & PF_LOG_ALL) 5964 lr = s->nat_rule.ptr; 5965 else 5966 lr = r; 5967 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset, 5968 &pd); 5969 } 5970 5971 kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len; 5972 kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++; 5973 5974 if (action == PF_PASS || r->action == PF_DROP) { 5975 dirndx = (dir == PF_OUT); 5976 r->packets[dirndx]++; 5977 r->bytes[dirndx] += pd.tot_len; 5978 if (a != NULL) { 5979 a->packets[dirndx]++; 5980 a->bytes[dirndx] += pd.tot_len; 5981 } 5982 if (s != NULL) { 5983 sk = s->state_key; 5984 if (s->nat_rule.ptr != NULL) { 5985 s->nat_rule.ptr->packets[dirndx]++; 5986 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len; 5987 } 5988 if (s->src_node != NULL) { 5989 s->src_node->packets[dirndx]++; 5990 s->src_node->bytes[dirndx] += pd.tot_len; 5991 } 5992 if (s->nat_src_node != NULL) { 5993 s->nat_src_node->packets[dirndx]++; 5994 s->nat_src_node->bytes[dirndx] += pd.tot_len; 5995 } 5996 dirndx = (dir == sk->direction) ? 0 : 1; 5997 s->packets[dirndx]++; 5998 s->bytes[dirndx] += pd.tot_len; 5999 } 6000 tr = r; 6001 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule; 6002 if (nr != NULL) { 6003 struct pf_addr *x; 6004 /* 6005 * XXX: we need to make sure that the addresses 6006 * passed to pfr_update_stats() are the same than 6007 * the addresses used during matching (pfr_match) 6008 */ 6009 if (r == &pf_default_rule) { 6010 tr = nr; 6011 x = (sk == NULL || sk->direction == dir) ? 6012 &pd.baddr : &pd.naddr; 6013 } else 6014 x = (sk == NULL || sk->direction == dir) ? 6015 &pd.naddr : &pd.baddr; 6016 if (x == &pd.baddr || s == NULL) { 6017 /* we need to change the address */ 6018 if (dir == PF_OUT) 6019 pd.src = x; 6020 else 6021 pd.dst = x; 6022 } 6023 } 6024 if (tr->src.addr.type == PF_ADDR_TABLE) 6025 pfr_update_stats(tr->src.addr.p.tbl, (sk == NULL || 6026 sk->direction == dir) ? 6027 pd.src : pd.dst, pd.af, 6028 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 6029 tr->src.neg); 6030 if (tr->dst.addr.type == PF_ADDR_TABLE) 6031 pfr_update_stats(tr->dst.addr.p.tbl, (sk == NULL || 6032 sk->direction == dir) ? pd.dst : pd.src, pd.af, 6033 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 6034 tr->dst.neg); 6035 } 6036 6037 6038 if (action == PF_SYNPROXY_DROP) { 6039 m_freem(*m0); 6040 *m0 = NULL; 6041 action = PF_PASS; 6042 } else if (r->rt) 6043 /* pf_route can free the mbuf causing *m0 to become NULL */ 6044 pf_route(m0, r, dir, kif->pfik_ifp, s, &pd); 6045 6046 return (action); 6047 } 6048 #endif /* INET */ 6049 6050 #ifdef INET6 6051 int 6052 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0, 6053 struct ether_header *eh) 6054 { 6055 struct pfi_kif *kif; 6056 u_short action, reason = 0, log = 0; 6057 struct mbuf *m = *m0, *n = NULL; 6058 struct ip6_hdr *h = NULL; /* XXX gcc */ 6059 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr; 6060 struct pf_state *s = NULL; 6061 struct pf_state_key *sk = NULL; 6062 struct pf_ruleset *ruleset = NULL; 6063 struct pf_pdesc pd; 6064 int off, terminal = 0, dirndx, rh_cnt = 0; 6065 #ifdef __NetBSD__ 6066 struct pf_mtag *pf_mtag = NULL; /* XXX gcc */ 6067 #endif /* __NetBSD__ */ 6068 6069 if (!pf_status.running) 6070 return (PF_PASS); 6071 6072 memset(&pd, 0, sizeof(pd)); 6073 if (ifp->if_type == IFT_CARP && ifp->if_carpdev) 6074 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif; 6075 else 6076 kif = (struct pfi_kif *)ifp->if_pf_kif; 6077 6078 if (kif == NULL) { 6079 DPFPRINTF(PF_DEBUG_URGENT, 6080 ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname)); 6081 return (PF_DROP); 6082 } 6083 if (kif->pfik_flags & PFI_IFLAG_SKIP) 6084 return (PF_PASS); 6085 6086 #ifdef DIAGNOSTIC 6087 if ((m->m_flags & M_PKTHDR) == 0) 6088 panic("non-M_PKTHDR is passed to pf_test6"); 6089 #endif /* DIAGNOSTIC */ 6090 6091 if (m->m_pkthdr.len < (int)sizeof(*h)) { 6092 action = PF_DROP; 6093 REASON_SET(&reason, PFRES_SHORT); 6094 log = 1; 6095 goto done; 6096 } 6097 6098 #ifdef __NetBSD__ 6099 if ((pf_mtag = pf_get_mtag(m)) == NULL) { 6100 DPFPRINTF(PF_DEBUG_URGENT, 6101 ("pf_test6: pf_get_mtag returned NULL\n")); 6102 return (PF_DROP); 6103 } 6104 if (pf_mtag->flags & PF_TAG_GENERATED) 6105 return (PF_PASS); 6106 #else 6107 if (m->m_pkthdr.pf.flags & PF_TAG_GENERATED) 6108 return (PF_PASS); 6109 #endif /* !__NetBSD__ */ 6110 6111 /* We do IP header normalization and packet reassembly here */ 6112 if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) { 6113 action = PF_DROP; 6114 goto done; 6115 } 6116 m = *m0; /* pf_normalize messes with m0 */ 6117 h = mtod(m, struct ip6_hdr *); 6118 6119 #if 1 6120 /* 6121 * we do not support jumbogram yet. if we keep going, zero ip6_plen 6122 * will do something bad, so drop the packet for now. 6123 */ 6124 if (htons(h->ip6_plen) == 0) { 6125 action = PF_DROP; 6126 REASON_SET(&reason, PFRES_NORM); /*XXX*/ 6127 goto done; 6128 } 6129 #endif 6130 6131 pd.src = (struct pf_addr *)&h->ip6_src; 6132 pd.dst = (struct pf_addr *)&h->ip6_dst; 6133 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6); 6134 pd.ip_sum = NULL; 6135 pd.af = AF_INET6; 6136 pd.tos = 0; 6137 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr); 6138 pd.eh = eh; 6139 6140 off = ((char *)h - m->m_data) + sizeof(struct ip6_hdr); 6141 pd.proto = h->ip6_nxt; 6142 do { 6143 switch (pd.proto) { 6144 case IPPROTO_FRAGMENT: 6145 action = pf_test_fragment(&r, dir, kif, m, h, 6146 &pd, &a, &ruleset); 6147 if (action == PF_DROP) 6148 REASON_SET(&reason, PFRES_FRAG); 6149 goto done; 6150 case IPPROTO_ROUTING: { 6151 struct ip6_rthdr rthdr; 6152 6153 if (rh_cnt++) { 6154 DPFPRINTF(PF_DEBUG_MISC, 6155 ("pf: IPv6 more than one rthdr\n")); 6156 action = PF_DROP; 6157 REASON_SET(&reason, PFRES_IPOPTIONS); 6158 log = 1; 6159 goto done; 6160 } 6161 if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL, 6162 &reason, pd.af)) { 6163 DPFPRINTF(PF_DEBUG_MISC, 6164 ("pf: IPv6 short rthdr\n")); 6165 action = PF_DROP; 6166 REASON_SET(&reason, PFRES_SHORT); 6167 log = 1; 6168 goto done; 6169 } 6170 if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) { 6171 DPFPRINTF(PF_DEBUG_MISC, 6172 ("pf: IPv6 rthdr0\n")); 6173 action = PF_DROP; 6174 REASON_SET(&reason, PFRES_IPOPTIONS); 6175 log = 1; 6176 goto done; 6177 } 6178 /* FALLTHROUGH */ 6179 } 6180 case IPPROTO_AH: 6181 case IPPROTO_HOPOPTS: 6182 case IPPROTO_DSTOPTS: { 6183 /* get next header and header length */ 6184 struct ip6_ext opt6; 6185 6186 if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6), 6187 NULL, &reason, pd.af)) { 6188 DPFPRINTF(PF_DEBUG_MISC, 6189 ("pf: IPv6 short opt\n")); 6190 action = PF_DROP; 6191 log = 1; 6192 goto done; 6193 } 6194 if (pd.proto == IPPROTO_AH) 6195 off += (opt6.ip6e_len + 2) * 4; 6196 else 6197 off += (opt6.ip6e_len + 1) * 8; 6198 pd.proto = opt6.ip6e_nxt; 6199 /* goto the next header */ 6200 break; 6201 } 6202 default: 6203 terminal++; 6204 break; 6205 } 6206 } while (!terminal); 6207 6208 /* if there's no routing header, use unmodified mbuf for checksumming */ 6209 if (!n) 6210 n = m; 6211 6212 switch (pd.proto) { 6213 6214 case IPPROTO_TCP: { 6215 struct tcphdr th; 6216 6217 pd.hdr.tcp = &th; 6218 if (!pf_pull_hdr(m, off, &th, sizeof(th), 6219 &action, &reason, AF_INET6)) { 6220 log = action != PF_PASS; 6221 goto done; 6222 } 6223 pd.p_len = pd.tot_len - off - (th.th_off << 2); 6224 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd); 6225 if (action == PF_DROP) 6226 goto done; 6227 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd, 6228 &reason); 6229 if (action == PF_PASS) { 6230 #if NPFSYNC 6231 pfsync_update_state(s); 6232 #endif /* NPFSYNC */ 6233 r = s->rule.ptr; 6234 a = s->anchor.ptr; 6235 log = s->log; 6236 } else if (s == NULL) 6237 action = pf_test_rule(&r, &s, dir, kif, 6238 m, off, h, &pd, &a, &ruleset, &ip6intrq); 6239 break; 6240 } 6241 6242 case IPPROTO_UDP: { 6243 struct udphdr uh; 6244 6245 pd.hdr.udp = &uh; 6246 if (!pf_pull_hdr(m, off, &uh, sizeof(uh), 6247 &action, &reason, AF_INET6)) { 6248 log = action != PF_PASS; 6249 goto done; 6250 } 6251 if (uh.uh_dport == 0 || 6252 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off || 6253 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) { 6254 action = PF_DROP; 6255 REASON_SET(&reason, PFRES_SHORT); 6256 goto done; 6257 } 6258 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd); 6259 if (action == PF_PASS) { 6260 #if NPFSYNC 6261 pfsync_update_state(s); 6262 #endif /* NPFSYNC */ 6263 r = s->rule.ptr; 6264 a = s->anchor.ptr; 6265 log = s->log; 6266 } else if (s == NULL) 6267 action = pf_test_rule(&r, &s, dir, kif, 6268 m, off, h, &pd, &a, &ruleset, &ip6intrq); 6269 break; 6270 } 6271 6272 case IPPROTO_ICMPV6: { 6273 struct icmp6_hdr ih; 6274 6275 pd.hdr.icmp6 = &ih; 6276 if (!pf_pull_hdr(m, off, &ih, sizeof(ih), 6277 &action, &reason, AF_INET6)) { 6278 log = action != PF_PASS; 6279 goto done; 6280 } 6281 action = pf_test_state_icmp(&s, dir, kif, 6282 m, off, h, &pd, &reason); 6283 if (action == PF_PASS) { 6284 #if NPFSYNC 6285 pfsync_update_state(s); 6286 #endif /* NPFSYNC */ 6287 r = s->rule.ptr; 6288 a = s->anchor.ptr; 6289 log = s->log; 6290 } else if (s == NULL) 6291 action = pf_test_rule(&r, &s, dir, kif, 6292 m, off, h, &pd, &a, &ruleset, &ip6intrq); 6293 break; 6294 } 6295 6296 default: 6297 action = pf_test_state_other(&s, dir, kif, &pd); 6298 if (action == PF_PASS) { 6299 #if NPFSYNC 6300 pfsync_update_state(s); 6301 #endif /* NPFSYNC */ 6302 r = s->rule.ptr; 6303 a = s->anchor.ptr; 6304 log = s->log; 6305 } else if (s == NULL) 6306 action = pf_test_rule(&r, &s, dir, kif, m, off, h, 6307 &pd, &a, &ruleset, &ip6intrq); 6308 break; 6309 } 6310 6311 done: 6312 if (n != m) { 6313 m_freem(n); 6314 n = NULL; 6315 } 6316 6317 /* handle dangerous IPv6 extension headers. */ 6318 if (action == PF_PASS && rh_cnt && 6319 !((s && s->allow_opts) || r->allow_opts)) { 6320 action = PF_DROP; 6321 REASON_SET(&reason, PFRES_IPOPTIONS); 6322 log = 1; 6323 DPFPRINTF(PF_DEBUG_MISC, 6324 ("pf: dropping packet with dangerous v6 headers\n")); 6325 } 6326 6327 if ((s && s->tag) || r->rtableid) 6328 pf_tag_packet(m, s ? s->tag : 0, r->rtableid); 6329 6330 #ifdef ALTQ 6331 if (action == PF_PASS && r->qid) { 6332 #ifdef __NetBSD__ 6333 struct m_tag *mtag; 6334 struct altq_tag *atag; 6335 6336 mtag = m_tag_get(PACKET_TAG_ALTQ_QID, sizeof(*atag), M_NOWAIT); 6337 if (mtag != NULL) { 6338 atag = (struct altq_tag *)(mtag + 1); 6339 if (pd.tos & IPTOS_LOWDELAY) 6340 atag->qid = r->pqid; 6341 else 6342 atag->qid = r->qid; 6343 /* add hints for ecn */ 6344 atag->af = AF_INET6; 6345 atag->hdr = h; 6346 m_tag_prepend(m, mtag); 6347 } 6348 #else 6349 if (pd.tos & IPTOS_LOWDELAY) 6350 m->m_pkthdr.pf.qid = r->pqid; 6351 else 6352 m->m_pkthdr.pf.qid = r->qid; 6353 /* add hints for ecn */ 6354 m->m_pkthdr.pf.hdr = h; 6355 #endif /* !__NetBSD__ */ 6356 } 6357 #endif /* ALTQ */ 6358 6359 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP || 6360 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL && 6361 (s->nat_rule.ptr->action == PF_RDR || 6362 s->nat_rule.ptr->action == PF_BINAT) && 6363 IN6_IS_ADDR_LOOPBACK(&pd.dst->v6)) 6364 #ifdef __NetBSD__ 6365 pf_mtag->flags |= PF_TAG_TRANSLATE_LOCALHOST; 6366 #else 6367 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST; 6368 #endif /* !__NetBSD__ */ 6369 6370 if (log) { 6371 struct pf_rule *lr; 6372 6373 if (s != NULL && s->nat_rule.ptr != NULL && 6374 s->nat_rule.ptr->log & PF_LOG_ALL) 6375 lr = s->nat_rule.ptr; 6376 else 6377 lr = r; 6378 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset, 6379 &pd); 6380 } 6381 6382 kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len; 6383 kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++; 6384 6385 if (action == PF_PASS || r->action == PF_DROP) { 6386 dirndx = (dir == PF_OUT); 6387 r->packets[dirndx]++; 6388 r->bytes[dirndx] += pd.tot_len; 6389 if (a != NULL) { 6390 a->packets[dirndx]++; 6391 a->bytes[dirndx] += pd.tot_len; 6392 } 6393 if (s != NULL) { 6394 sk = s->state_key; 6395 if (s->nat_rule.ptr != NULL) { 6396 s->nat_rule.ptr->packets[dirndx]++; 6397 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len; 6398 } 6399 if (s->src_node != NULL) { 6400 s->src_node->packets[dirndx]++; 6401 s->src_node->bytes[dirndx] += pd.tot_len; 6402 } 6403 if (s->nat_src_node != NULL) { 6404 s->nat_src_node->packets[dirndx]++; 6405 s->nat_src_node->bytes[dirndx] += pd.tot_len; 6406 } 6407 dirndx = (dir == sk->direction) ? 0 : 1; 6408 s->packets[dirndx]++; 6409 s->bytes[dirndx] += pd.tot_len; 6410 } 6411 tr = r; 6412 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule; 6413 if (nr != NULL) { 6414 struct pf_addr *x; 6415 /* 6416 * XXX: we need to make sure that the addresses 6417 * passed to pfr_update_stats() are the same than 6418 * the addresses used during matching (pfr_match) 6419 */ 6420 if (r == &pf_default_rule) { 6421 tr = nr; 6422 x = (s == NULL || sk->direction == dir) ? 6423 &pd.baddr : &pd.naddr; 6424 } else { 6425 x = (s == NULL || sk->direction == dir) ? 6426 &pd.naddr : &pd.baddr; 6427 } 6428 if (x == &pd.baddr || s == NULL) { 6429 if (dir == PF_OUT) 6430 pd.src = x; 6431 else 6432 pd.dst = x; 6433 } 6434 } 6435 if (tr->src.addr.type == PF_ADDR_TABLE) 6436 pfr_update_stats(tr->src.addr.p.tbl, (sk == NULL || 6437 sk->direction == dir) ? pd.src : pd.dst, pd.af, 6438 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 6439 tr->src.neg); 6440 if (tr->dst.addr.type == PF_ADDR_TABLE) 6441 pfr_update_stats(tr->dst.addr.p.tbl, (sk == NULL || 6442 sk->direction == dir) ? pd.dst : pd.src, pd.af, 6443 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 6444 tr->dst.neg); 6445 } 6446 6447 6448 if (action == PF_SYNPROXY_DROP) { 6449 m_freem(*m0); 6450 *m0 = NULL; 6451 action = PF_PASS; 6452 } else if (r->rt) 6453 /* pf_route6 can free the mbuf causing *m0 to become NULL */ 6454 pf_route6(m0, r, dir, kif->pfik_ifp, s, &pd); 6455 6456 return (action); 6457 } 6458 #endif /* INET6 */ 6459 6460 int 6461 pf_check_congestion(struct ifqueue *ifq) 6462 { 6463 #ifdef __NetBSD__ 6464 return (0); 6465 #else 6466 if (ifq->ifq_congestion) 6467 return (1); 6468 else 6469 return (0); 6470 #endif /* !__NetBSD__ */ 6471 } 6472