1 /* $NetBSD: pf.c,v 1.61 2010/01/19 22:08:00 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 1.61 2010/01/19 22:08:00 pooka Exp $"); 41 42 #include "pflog.h" 43 44 #include "pfsync.h" 45 46 #include <sys/param.h> 47 #include <sys/systm.h> 48 #include <sys/mbuf.h> 49 #include <sys/filio.h> 50 #include <sys/socket.h> 51 #include <sys/socketvar.h> 52 #include <sys/kernel.h> 53 #include <sys/time.h> 54 #include <sys/pool.h> 55 #include <sys/proc.h> 56 #include <sys/rwlock.h> 57 #ifdef __NetBSD__ 58 #include <sys/kthread.h> 59 #include <sys/kauth.h> 60 #endif /* __NetBSD__ */ 61 62 #include <net/if.h> 63 #include <net/if_types.h> 64 #include <net/bpf.h> 65 #include <net/route.h> 66 #ifndef __NetBSD__ 67 #include <net/radix_mpath.h> 68 #endif /* !__NetBSD__ */ 69 70 #include <netinet/in.h> 71 #ifdef __NetBSD__ 72 #include <netinet/in_offload.h> 73 #endif /* __NetBSD__ */ 74 #include <netinet/in_var.h> 75 #include <netinet/in_systm.h> 76 #include <netinet/ip.h> 77 #include <netinet/ip_var.h> 78 #include <netinet/tcp.h> 79 #include <netinet/tcp_seq.h> 80 #include <netinet/udp.h> 81 #include <netinet/ip_icmp.h> 82 #include <netinet/in_pcb.h> 83 #include <netinet/tcp_timer.h> 84 #include <netinet/tcp_var.h> 85 #include <netinet/udp_var.h> 86 #include <netinet/icmp_var.h> 87 #ifndef __NetBSD__ 88 #include <netinet/if_ether.h> 89 #else 90 #include <net/if_ether.h> 91 #endif /* __NetBSD__ */ 92 93 #ifndef __NetBSD__ 94 #include <dev/rndvar.h> 95 #else 96 #include <sys/rnd.h> 97 #endif /* __NetBSD__ */ 98 99 #include <net/pfvar.h> 100 #include <net/if_pflog.h> 101 102 #if NPFSYNC > 0 103 #include <net/if_pfsync.h> 104 #endif /* NPFSYNC > 0 */ 105 106 #ifdef INET6 107 #include <netinet/ip6.h> 108 #include <netinet6/ip6_var.h> 109 #ifdef __NetBSD__ 110 #include <netinet6/in6_pcb.h> 111 #endif /* __NetBSD__ */ 112 #include <netinet/icmp6.h> 113 #include <netinet6/nd6.h> 114 #endif /* INET6 */ 115 116 #ifdef __NetBSD__ 117 #include <netinet/tcp_rndiss.h> 118 #endif /* __NetBSD__ */ 119 120 121 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x 122 123 /* 124 * Global variables 125 */ 126 127 /* state tables */ 128 struct pf_state_tree_lan_ext pf_statetbl_lan_ext; 129 struct pf_state_tree_ext_gwy pf_statetbl_ext_gwy; 130 131 struct pf_altqqueue pf_altqs[2]; 132 struct pf_palist pf_pabuf; 133 struct pf_altqqueue *pf_altqs_active; 134 struct pf_altqqueue *pf_altqs_inactive; 135 struct pf_status pf_status; 136 137 u_int32_t ticket_altqs_active; 138 u_int32_t ticket_altqs_inactive; 139 int altqs_inactive_open; 140 u_int32_t ticket_pabuf; 141 142 struct pf_anchor_stackframe { 143 struct pf_ruleset *rs; 144 struct pf_rule *r; 145 struct pf_anchor_node *parent; 146 struct pf_anchor *child; 147 } pf_anchor_stack[64]; 148 149 struct pool pf_src_tree_pl, pf_rule_pl, pf_pooladdr_pl; 150 struct pool pf_state_pl, pf_state_key_pl; 151 struct pool pf_altq_pl; 152 153 void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t); 154 155 void pf_init_threshold(struct pf_threshold *, u_int32_t, 156 u_int32_t); 157 void pf_add_threshold(struct pf_threshold *); 158 int pf_check_threshold(struct pf_threshold *); 159 160 void pf_change_ap(struct pf_addr *, u_int16_t *, 161 u_int16_t *, u_int16_t *, struct pf_addr *, 162 u_int16_t, u_int8_t, sa_family_t); 163 int pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *, 164 struct tcphdr *, struct pf_state_peer *); 165 #ifdef INET6 166 void pf_change_a6(struct pf_addr *, u_int16_t *, 167 struct pf_addr *, u_int8_t); 168 #endif /* INET6 */ 169 void pf_change_icmp(struct pf_addr *, u_int16_t *, 170 struct pf_addr *, struct pf_addr *, u_int16_t, 171 u_int16_t *, u_int16_t *, u_int16_t *, 172 u_int16_t *, u_int8_t, sa_family_t); 173 void pf_send_tcp(const struct pf_rule *, sa_family_t, 174 const struct pf_addr *, const struct pf_addr *, 175 u_int16_t, u_int16_t, u_int32_t, u_int32_t, 176 u_int8_t, u_int16_t, u_int16_t, u_int8_t, int, 177 u_int16_t, struct ether_header *, struct ifnet *); 178 void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t, 179 sa_family_t, struct pf_rule *); 180 struct pf_rule *pf_match_translation(struct pf_pdesc *, struct mbuf *, 181 int, int, struct pfi_kif *, 182 struct pf_addr *, u_int16_t, struct pf_addr *, 183 u_int16_t, int); 184 struct pf_rule *pf_get_translation(struct pf_pdesc *, struct mbuf *, 185 int, int, struct pfi_kif *, struct pf_src_node **, 186 struct pf_addr *, u_int16_t, 187 struct pf_addr *, u_int16_t, 188 struct pf_addr *, u_int16_t *); 189 void pf_attach_state(struct pf_state_key *, 190 struct pf_state *, int); 191 void pf_detach_state(struct pf_state *, int); 192 int pf_test_rule(struct pf_rule **, struct pf_state **, 193 int, struct pfi_kif *, struct mbuf *, int, 194 void *, struct pf_pdesc *, struct pf_rule **, 195 struct pf_ruleset **, struct ifqueue *); 196 int pf_test_fragment(struct pf_rule **, int, 197 struct pfi_kif *, struct mbuf *, void *, 198 struct pf_pdesc *, struct pf_rule **, 199 struct pf_ruleset **); 200 int pf_test_state_tcp(struct pf_state **, int, 201 struct pfi_kif *, struct mbuf *, int, 202 void *, struct pf_pdesc *, u_short *); 203 int pf_test_state_udp(struct pf_state **, int, 204 struct pfi_kif *, struct mbuf *, int, 205 void *, struct pf_pdesc *); 206 int pf_test_state_icmp(struct pf_state **, int, 207 struct pfi_kif *, struct mbuf *, int, 208 void *, struct pf_pdesc *, u_short *); 209 int pf_test_state_other(struct pf_state **, int, 210 struct pfi_kif *, struct pf_pdesc *); 211 int pf_match_tag(struct mbuf *, struct pf_rule *, int *); 212 void pf_step_into_anchor(int *, struct pf_ruleset **, int, 213 struct pf_rule **, struct pf_rule **, int *); 214 int pf_step_out_of_anchor(int *, struct pf_ruleset **, 215 int, struct pf_rule **, struct pf_rule **, 216 int *); 217 void pf_hash(const struct pf_addr *, struct pf_addr *, 218 struct pf_poolhashkey *, sa_family_t); 219 int pf_map_addr(u_int8_t, struct pf_rule *, 220 const struct pf_addr *, struct pf_addr *, 221 struct pf_addr *, struct pf_src_node **); 222 int pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *, 223 struct pf_addr *, struct pf_addr *, u_int16_t, 224 struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t, 225 struct pf_src_node **); 226 void pf_route(struct mbuf **, struct pf_rule *, int, 227 struct ifnet *, struct pf_state *, 228 struct pf_pdesc *); 229 void pf_route6(struct mbuf **, struct pf_rule *, int, 230 struct ifnet *, struct pf_state *, 231 struct pf_pdesc *); 232 int pf_socket_lookup(int, struct pf_pdesc *); 233 u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t, 234 sa_family_t); 235 u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t, 236 sa_family_t); 237 u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t, 238 u_int16_t); 239 void pf_set_rt_ifp(struct pf_state *, 240 struct pf_addr *); 241 #ifdef __NetBSD__ 242 int pf_check_proto_cksum(struct mbuf *, int, int, int, 243 u_int8_t, sa_family_t); 244 #else 245 int pf_check_proto_cksum(struct mbuf *, int, int, 246 u_int8_t, sa_family_t); 247 #endif /* !__NetBSD__ */ 248 int pf_addr_wrap_neq(struct pf_addr_wrap *, 249 struct pf_addr_wrap *); 250 struct pf_state *pf_find_state(struct pfi_kif *, 251 struct pf_state_key_cmp *, u_int8_t); 252 int pf_src_connlimit(struct pf_state **); 253 void pf_stateins_err(const char *, struct pf_state *, 254 struct pfi_kif *); 255 int pf_check_congestion(struct ifqueue *); 256 257 extern struct pool pfr_ktable_pl; 258 extern struct pool pfr_kentry_pl; 259 260 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = { 261 { &pf_state_pl, PFSTATE_HIWAT }, 262 { &pf_src_tree_pl, PFSNODE_HIWAT }, 263 { &pf_frent_pl, PFFRAG_FRENT_HIWAT }, 264 { &pfr_ktable_pl, PFR_KTABLE_HIWAT }, 265 { &pfr_kentry_pl, PFR_KENTRY_HIWAT } 266 }; 267 268 #define STATE_LOOKUP() \ 269 do { \ 270 if (direction == PF_IN) \ 271 *state = pf_find_state(kif, &key, PF_EXT_GWY); \ 272 else \ 273 *state = pf_find_state(kif, &key, PF_LAN_EXT); \ 274 if (*state == NULL || (*state)->timeout == PFTM_PURGE) \ 275 return (PF_DROP); \ 276 if (direction == PF_OUT && \ 277 (((*state)->rule.ptr->rt == PF_ROUTETO && \ 278 (*state)->rule.ptr->direction == PF_OUT) || \ 279 ((*state)->rule.ptr->rt == PF_REPLYTO && \ 280 (*state)->rule.ptr->direction == PF_IN)) && \ 281 (*state)->rt_kif != NULL && \ 282 (*state)->rt_kif != kif) \ 283 return (PF_PASS); \ 284 } while (0) 285 286 #define STATE_TRANSLATE(sk) \ 287 (sk)->lan.addr.addr32[0] != (sk)->gwy.addr.addr32[0] || \ 288 ((sk)->af == AF_INET6 && \ 289 ((sk)->lan.addr.addr32[1] != (sk)->gwy.addr.addr32[1] || \ 290 (sk)->lan.addr.addr32[2] != (sk)->gwy.addr.addr32[2] || \ 291 (sk)->lan.addr.addr32[3] != (sk)->gwy.addr.addr32[3])) || \ 292 (sk)->lan.port != (sk)->gwy.port 293 294 #define BOUND_IFACE(r, k) \ 295 ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all 296 297 #define STATE_INC_COUNTERS(s) \ 298 do { \ 299 s->rule.ptr->states++; \ 300 if (s->anchor.ptr != NULL) \ 301 s->anchor.ptr->states++; \ 302 if (s->nat_rule.ptr != NULL) \ 303 s->nat_rule.ptr->states++; \ 304 } while (0) 305 306 #define STATE_DEC_COUNTERS(s) \ 307 do { \ 308 if (s->nat_rule.ptr != NULL) \ 309 s->nat_rule.ptr->states--; \ 310 if (s->anchor.ptr != NULL) \ 311 s->anchor.ptr->states--; \ 312 s->rule.ptr->states--; \ 313 } while (0) 314 315 static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *); 316 static __inline int pf_state_compare_lan_ext(struct pf_state_key *, 317 struct pf_state_key *); 318 static __inline int pf_state_compare_ext_gwy(struct pf_state_key *, 319 struct pf_state_key *); 320 static __inline int pf_state_compare_id(struct pf_state *, 321 struct pf_state *); 322 323 struct pf_src_tree tree_src_tracking; 324 325 struct pf_state_tree_id tree_id; 326 struct pf_state_queue state_list; 327 328 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare); 329 RB_GENERATE(pf_state_tree_lan_ext, pf_state_key, 330 entry_lan_ext, pf_state_compare_lan_ext); 331 RB_GENERATE(pf_state_tree_ext_gwy, pf_state_key, 332 entry_ext_gwy, pf_state_compare_ext_gwy); 333 RB_GENERATE(pf_state_tree_id, pf_state, 334 entry_id, pf_state_compare_id); 335 336 #define PF_DT_SKIP_LANEXT 0x01 337 #define PF_DT_SKIP_EXTGWY 0x02 338 339 #ifdef __NetBSD__ 340 static __inline struct pfi_kif * 341 bound_iface(const struct pf_rule *r, const struct pf_rule *nr, 342 struct pfi_kif *k) 343 { 344 uint32_t rule_flag; 345 346 rule_flag = r->rule_flag; 347 if (nr != NULL) 348 rule_flag |= nr->rule_flag; 349 350 return ((rule_flag & PFRULE_IFBOUND) != 0) ? k : pfi_all; 351 } 352 #endif /* __NetBSD__ */ 353 354 static __inline int 355 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b) 356 { 357 int diff; 358 359 if (a->rule.ptr > b->rule.ptr) 360 return (1); 361 if (a->rule.ptr < b->rule.ptr) 362 return (-1); 363 if ((diff = a->af - b->af) != 0) 364 return (diff); 365 switch (a->af) { 366 #ifdef INET 367 case AF_INET: 368 if (a->addr.addr32[0] > b->addr.addr32[0]) 369 return (1); 370 if (a->addr.addr32[0] < b->addr.addr32[0]) 371 return (-1); 372 break; 373 #endif /* INET */ 374 #ifdef INET6 375 case AF_INET6: 376 if (a->addr.addr32[3] > b->addr.addr32[3]) 377 return (1); 378 if (a->addr.addr32[3] < b->addr.addr32[3]) 379 return (-1); 380 if (a->addr.addr32[2] > b->addr.addr32[2]) 381 return (1); 382 if (a->addr.addr32[2] < b->addr.addr32[2]) 383 return (-1); 384 if (a->addr.addr32[1] > b->addr.addr32[1]) 385 return (1); 386 if (a->addr.addr32[1] < b->addr.addr32[1]) 387 return (-1); 388 if (a->addr.addr32[0] > b->addr.addr32[0]) 389 return (1); 390 if (a->addr.addr32[0] < b->addr.addr32[0]) 391 return (-1); 392 break; 393 #endif /* INET6 */ 394 } 395 return (0); 396 } 397 398 static __inline int 399 pf_state_compare_lan_ext(struct pf_state_key *a, struct pf_state_key *b) 400 { 401 int diff; 402 403 if ((diff = a->proto - b->proto) != 0) 404 return (diff); 405 if ((diff = a->af - b->af) != 0) 406 return (diff); 407 switch (a->af) { 408 #ifdef INET 409 case AF_INET: 410 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0]) 411 return (1); 412 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0]) 413 return (-1); 414 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 415 return (1); 416 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 417 return (-1); 418 break; 419 #endif /* INET */ 420 #ifdef INET6 421 case AF_INET6: 422 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3]) 423 return (1); 424 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3]) 425 return (-1); 426 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3]) 427 return (1); 428 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3]) 429 return (-1); 430 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2]) 431 return (1); 432 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2]) 433 return (-1); 434 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2]) 435 return (1); 436 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2]) 437 return (-1); 438 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1]) 439 return (1); 440 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1]) 441 return (-1); 442 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1]) 443 return (1); 444 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1]) 445 return (-1); 446 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0]) 447 return (1); 448 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0]) 449 return (-1); 450 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 451 return (1); 452 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 453 return (-1); 454 break; 455 #endif /* INET6 */ 456 } 457 458 if ((diff = a->lan.port - b->lan.port) != 0) 459 return (diff); 460 if ((diff = a->ext.port - b->ext.port) != 0) 461 return (diff); 462 463 return (0); 464 } 465 466 static __inline int 467 pf_state_compare_ext_gwy(struct pf_state_key *a, struct pf_state_key *b) 468 { 469 int diff; 470 471 if ((diff = a->proto - b->proto) != 0) 472 return (diff); 473 if ((diff = a->af - b->af) != 0) 474 return (diff); 475 switch (a->af) { 476 #ifdef INET 477 case AF_INET: 478 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 479 return (1); 480 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 481 return (-1); 482 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0]) 483 return (1); 484 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0]) 485 return (-1); 486 break; 487 #endif /* INET */ 488 #ifdef INET6 489 case AF_INET6: 490 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3]) 491 return (1); 492 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3]) 493 return (-1); 494 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3]) 495 return (1); 496 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3]) 497 return (-1); 498 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2]) 499 return (1); 500 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2]) 501 return (-1); 502 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2]) 503 return (1); 504 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2]) 505 return (-1); 506 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1]) 507 return (1); 508 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1]) 509 return (-1); 510 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1]) 511 return (1); 512 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1]) 513 return (-1); 514 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 515 return (1); 516 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 517 return (-1); 518 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0]) 519 return (1); 520 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0]) 521 return (-1); 522 break; 523 #endif /* INET6 */ 524 } 525 526 if ((diff = a->ext.port - b->ext.port) != 0) 527 return (diff); 528 if ((diff = a->gwy.port - b->gwy.port) != 0) 529 return (diff); 530 531 return (0); 532 } 533 534 static __inline int 535 pf_state_compare_id(struct pf_state *a, struct pf_state *b) 536 { 537 if (a->id > b->id) 538 return (1); 539 if (a->id < b->id) 540 return (-1); 541 if (a->creatorid > b->creatorid) 542 return (1); 543 if (a->creatorid < b->creatorid) 544 return (-1); 545 546 return (0); 547 } 548 549 #ifdef INET6 550 void 551 pf_addrcpy(struct pf_addr *dst, const struct pf_addr *src, sa_family_t af) 552 { 553 switch (af) { 554 #ifdef INET 555 case AF_INET: 556 dst->addr32[0] = src->addr32[0]; 557 break; 558 #endif /* INET */ 559 case AF_INET6: 560 dst->addr32[0] = src->addr32[0]; 561 dst->addr32[1] = src->addr32[1]; 562 dst->addr32[2] = src->addr32[2]; 563 dst->addr32[3] = src->addr32[3]; 564 break; 565 } 566 } 567 #endif /* INET6 */ 568 569 struct pf_state * 570 pf_find_state_byid(struct pf_state_cmp *key) 571 { 572 pf_status.fcounters[FCNT_STATE_SEARCH]++; 573 574 return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key)); 575 } 576 577 struct pf_state * 578 pf_find_state(struct pfi_kif *kif, struct pf_state_key_cmp *key, u_int8_t tree) 579 { 580 struct pf_state_key *sk; 581 struct pf_state *s; 582 583 pf_status.fcounters[FCNT_STATE_SEARCH]++; 584 585 switch (tree) { 586 case PF_LAN_EXT: 587 sk = RB_FIND(pf_state_tree_lan_ext, &pf_statetbl_lan_ext, 588 (struct pf_state_key *)key); 589 break; 590 case PF_EXT_GWY: 591 sk = RB_FIND(pf_state_tree_ext_gwy, &pf_statetbl_ext_gwy, 592 (struct pf_state_key *)key); 593 break; 594 default: 595 panic("pf_find_state"); 596 } 597 598 /* list is sorted, if-bound states before floating ones */ 599 if (sk != NULL) 600 TAILQ_FOREACH(s, &sk->states, next) 601 if (s->kif == pfi_all || s->kif == kif) 602 return (s); 603 604 return (NULL); 605 } 606 607 struct pf_state * 608 pf_find_state_all(struct pf_state_key_cmp *key, u_int8_t tree, int *more) 609 { 610 struct pf_state_key *sk; 611 struct pf_state *s, *ret = NULL; 612 613 pf_status.fcounters[FCNT_STATE_SEARCH]++; 614 615 switch (tree) { 616 case PF_LAN_EXT: 617 sk = RB_FIND(pf_state_tree_lan_ext, 618 &pf_statetbl_lan_ext, (struct pf_state_key *)key); 619 break; 620 case PF_EXT_GWY: 621 sk = RB_FIND(pf_state_tree_ext_gwy, 622 &pf_statetbl_ext_gwy, (struct pf_state_key *)key); 623 break; 624 default: 625 panic("pf_find_state_all"); 626 } 627 628 if (sk != NULL) { 629 ret = TAILQ_FIRST(&sk->states); 630 if (more == NULL) 631 return (ret); 632 633 TAILQ_FOREACH(s, &sk->states, next) 634 (*more)++; 635 } 636 637 return (ret); 638 } 639 640 void 641 pf_init_threshold(struct pf_threshold *threshold, 642 u_int32_t limit, u_int32_t seconds) 643 { 644 threshold->limit = limit * PF_THRESHOLD_MULT; 645 threshold->seconds = seconds; 646 threshold->count = 0; 647 threshold->last = time_second; 648 } 649 650 void 651 pf_add_threshold(struct pf_threshold *threshold) 652 { 653 u_int32_t t = time_second, diff = t - threshold->last; 654 655 if (diff >= threshold->seconds) 656 threshold->count = 0; 657 else 658 threshold->count -= threshold->count * diff / 659 threshold->seconds; 660 threshold->count += PF_THRESHOLD_MULT; 661 threshold->last = t; 662 } 663 664 int 665 pf_check_threshold(struct pf_threshold *threshold) 666 { 667 return (threshold->count > threshold->limit); 668 } 669 670 int 671 pf_src_connlimit(struct pf_state **state) 672 { 673 int bad = 0; 674 675 (*state)->src_node->conn++; 676 (*state)->src.tcp_est = 1; 677 pf_add_threshold(&(*state)->src_node->conn_rate); 678 679 if ((*state)->rule.ptr->max_src_conn && 680 (*state)->rule.ptr->max_src_conn < 681 (*state)->src_node->conn) { 682 pf_status.lcounters[LCNT_SRCCONN]++; 683 bad++; 684 } 685 686 if ((*state)->rule.ptr->max_src_conn_rate.limit && 687 pf_check_threshold(&(*state)->src_node->conn_rate)) { 688 pf_status.lcounters[LCNT_SRCCONNRATE]++; 689 bad++; 690 } 691 692 if (!bad) 693 return (0); 694 695 if ((*state)->rule.ptr->overload_tbl) { 696 struct pfr_addr p; 697 u_int32_t killed = 0; 698 699 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++; 700 if (pf_status.debug >= PF_DEBUG_MISC) { 701 printf("pf_src_connlimit: blocking address "); 702 pf_print_host(&(*state)->src_node->addr, 0, 703 (*state)->state_key->af); 704 } 705 706 bzero(&p, sizeof(p)); 707 p.pfra_af = (*state)->state_key->af; 708 switch ((*state)->state_key->af) { 709 #ifdef INET 710 case AF_INET: 711 p.pfra_net = 32; 712 p.pfra_ip4addr = (*state)->src_node->addr.v4; 713 break; 714 #endif /* INET */ 715 #ifdef INET6 716 case AF_INET6: 717 p.pfra_net = 128; 718 p.pfra_ip6addr = (*state)->src_node->addr.v6; 719 break; 720 #endif /* INET6 */ 721 } 722 723 pfr_insert_kentry((*state)->rule.ptr->overload_tbl, 724 &p, time_second); 725 726 /* kill existing states if that's required. */ 727 if ((*state)->rule.ptr->flush) { 728 struct pf_state_key *sk; 729 struct pf_state *st; 730 731 pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++; 732 RB_FOREACH(st, pf_state_tree_id, &tree_id) { 733 sk = st->state_key; 734 /* 735 * Kill states from this source. (Only those 736 * from the same rule if PF_FLUSH_GLOBAL is not 737 * set) 738 */ 739 if (sk->af == 740 (*state)->state_key->af && 741 (((*state)->state_key->direction == 742 PF_OUT && 743 PF_AEQ(&(*state)->src_node->addr, 744 &sk->lan.addr, sk->af)) || 745 ((*state)->state_key->direction == PF_IN && 746 PF_AEQ(&(*state)->src_node->addr, 747 &sk->ext.addr, sk->af))) && 748 ((*state)->rule.ptr->flush & 749 PF_FLUSH_GLOBAL || 750 (*state)->rule.ptr == st->rule.ptr)) { 751 st->timeout = PFTM_PURGE; 752 st->src.state = st->dst.state = 753 TCPS_CLOSED; 754 killed++; 755 } 756 } 757 if (pf_status.debug >= PF_DEBUG_MISC) 758 printf(", %u states killed", killed); 759 } 760 if (pf_status.debug >= PF_DEBUG_MISC) 761 printf("\n"); 762 } 763 764 /* kill this state */ 765 (*state)->timeout = PFTM_PURGE; 766 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED; 767 return (1); 768 } 769 770 int 771 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule, 772 struct pf_addr *src, sa_family_t af) 773 { 774 struct pf_src_node k; 775 776 if (*sn == NULL) { 777 k.af = af; 778 PF_ACPY(&k.addr, src, af); 779 if (rule->rule_flag & PFRULE_RULESRCTRACK || 780 rule->rpool.opts & PF_POOL_STICKYADDR) 781 k.rule.ptr = rule; 782 else 783 k.rule.ptr = NULL; 784 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++; 785 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k); 786 } 787 if (*sn == NULL) { 788 if (!rule->max_src_nodes || 789 rule->src_nodes < rule->max_src_nodes) 790 (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT); 791 else 792 pf_status.lcounters[LCNT_SRCNODES]++; 793 if ((*sn) == NULL) 794 return (-1); 795 bzero(*sn, sizeof(struct pf_src_node)); 796 797 pf_init_threshold(&(*sn)->conn_rate, 798 rule->max_src_conn_rate.limit, 799 rule->max_src_conn_rate.seconds); 800 801 (*sn)->af = af; 802 if (rule->rule_flag & PFRULE_RULESRCTRACK || 803 rule->rpool.opts & PF_POOL_STICKYADDR) 804 (*sn)->rule.ptr = rule; 805 else 806 (*sn)->rule.ptr = NULL; 807 PF_ACPY(&(*sn)->addr, src, af); 808 if (RB_INSERT(pf_src_tree, 809 &tree_src_tracking, *sn) != NULL) { 810 if (pf_status.debug >= PF_DEBUG_MISC) { 811 printf("pf: src_tree insert failed: "); 812 pf_print_host(&(*sn)->addr, 0, af); 813 printf("\n"); 814 } 815 pool_put(&pf_src_tree_pl, *sn); 816 return (-1); 817 } 818 (*sn)->creation = time_second; 819 (*sn)->ruletype = rule->action; 820 if ((*sn)->rule.ptr != NULL) 821 (*sn)->rule.ptr->src_nodes++; 822 pf_status.scounters[SCNT_SRC_NODE_INSERT]++; 823 pf_status.src_nodes++; 824 } else { 825 if (rule->max_src_states && 826 (*sn)->states >= rule->max_src_states) { 827 pf_status.lcounters[LCNT_SRCSTATES]++; 828 return (-1); 829 } 830 } 831 return (0); 832 } 833 834 void 835 pf_stateins_err(const char *tree, struct pf_state *s, struct pfi_kif *kif) 836 { 837 struct pf_state_key *sk = s->state_key; 838 839 if (pf_status.debug >= PF_DEBUG_MISC) { 840 printf("pf: state insert failed: %s %s", tree, kif->pfik_name); 841 printf(" lan: "); 842 pf_print_host(&sk->lan.addr, sk->lan.port, 843 sk->af); 844 printf(" gwy: "); 845 pf_print_host(&sk->gwy.addr, sk->gwy.port, 846 sk->af); 847 printf(" ext: "); 848 pf_print_host(&sk->ext.addr, sk->ext.port, 849 sk->af); 850 if (s->sync_flags & PFSTATE_FROMSYNC) 851 printf(" (from sync)"); 852 printf("\n"); 853 } 854 } 855 856 int 857 pf_insert_state(struct pfi_kif *kif, struct pf_state *s) 858 { 859 struct pf_state_key *cur; 860 struct pf_state *sp; 861 862 KASSERT(s->state_key != NULL); 863 s->kif = kif; 864 865 if ((cur = RB_INSERT(pf_state_tree_lan_ext, &pf_statetbl_lan_ext, 866 s->state_key)) != NULL) { 867 /* key exists. check for same kif, if none, add to key */ 868 TAILQ_FOREACH(sp, &cur->states, next) 869 if (sp->kif == kif) { /* collision! */ 870 pf_stateins_err("tree_lan_ext", s, kif); 871 pf_detach_state(s, 872 PF_DT_SKIP_LANEXT|PF_DT_SKIP_EXTGWY); 873 return (-1); 874 } 875 pf_detach_state(s, PF_DT_SKIP_LANEXT|PF_DT_SKIP_EXTGWY); 876 pf_attach_state(cur, s, kif == pfi_all ? 1 : 0); 877 } 878 879 /* if cur != NULL, we already found a state key and attached to it */ 880 if (cur == NULL && (cur = RB_INSERT(pf_state_tree_ext_gwy, 881 &pf_statetbl_ext_gwy, s->state_key)) != NULL) { 882 /* must not happen. we must have found the sk above! */ 883 pf_stateins_err("tree_ext_gwy", s, kif); 884 pf_detach_state(s, PF_DT_SKIP_EXTGWY); 885 return (-1); 886 } 887 888 if (s->id == 0 && s->creatorid == 0) { 889 s->id = htobe64(pf_status.stateid++); 890 s->creatorid = pf_status.hostid; 891 } 892 if (RB_INSERT(pf_state_tree_id, &tree_id, s) != NULL) { 893 if (pf_status.debug >= PF_DEBUG_MISC) { 894 #ifdef __NetBSD__ 895 printf("pf: state insert failed: " 896 "id: %016" PRIx64 " creatorid: %08x", 897 be64toh(s->id), ntohl(s->creatorid)); 898 #else 899 printf("pf: state insert failed: " 900 "id: %016llx creatorid: %08x", 901 betoh64(s->id), ntohl(s->creatorid)); 902 #endif /* !__NetBSD__ */ 903 if (s->sync_flags & PFSTATE_FROMSYNC) 904 printf(" (from sync)"); 905 printf("\n"); 906 } 907 pf_detach_state(s, 0); 908 return (-1); 909 } 910 TAILQ_INSERT_TAIL(&state_list, s, entry_list); 911 pf_status.fcounters[FCNT_STATE_INSERT]++; 912 pf_status.states++; 913 pfi_kif_ref(kif, PFI_KIF_REF_STATE); 914 #if NPFSYNC 915 pfsync_insert_state(s); 916 #endif 917 return (0); 918 } 919 920 void 921 pf_purge_thread(void *v) 922 { 923 int nloops = 0, s; 924 925 for (;;) { 926 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz); 927 928 s = splsoftnet(); 929 930 /* process a fraction of the state table every second */ 931 pf_purge_expired_states(1 + (pf_status.states 932 / pf_default_rule.timeout[PFTM_INTERVAL])); 933 934 /* purge other expired types every PFTM_INTERVAL seconds */ 935 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) { 936 pf_purge_expired_fragments(); 937 pf_purge_expired_src_nodes(0); 938 nloops = 0; 939 } 940 941 splx(s); 942 } 943 } 944 945 u_int32_t 946 pf_state_expires(const struct pf_state *state) 947 { 948 u_int32_t timeout; 949 u_int32_t start; 950 u_int32_t end; 951 u_int32_t states; 952 953 /* handle all PFTM_* > PFTM_MAX here */ 954 if (state->timeout == PFTM_PURGE) 955 return (time_second); 956 if (state->timeout == PFTM_UNTIL_PACKET) 957 return (0); 958 KASSERT(state->timeout != PFTM_UNLINKED); 959 KASSERT(state->timeout < PFTM_MAX); 960 timeout = state->rule.ptr->timeout[state->timeout]; 961 if (!timeout) 962 timeout = pf_default_rule.timeout[state->timeout]; 963 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START]; 964 if (start) { 965 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END]; 966 states = state->rule.ptr->states; 967 } else { 968 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START]; 969 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END]; 970 states = pf_status.states; 971 } 972 if (end && states > start && start < end) { 973 if (states < end) 974 return (state->expire + timeout * (end - states) / 975 (end - start)); 976 else 977 return (time_second); 978 } 979 return (state->expire + timeout); 980 } 981 982 void 983 pf_purge_expired_src_nodes(int waslocked) 984 { 985 struct pf_src_node *cur, *next; 986 int locked = waslocked; 987 988 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) { 989 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur); 990 991 if (cur->states <= 0 && cur->expire <= time_second) { 992 if (! locked) { 993 rw_enter_write(&pf_consistency_lock); 994 next = RB_NEXT(pf_src_tree, 995 &tree_src_tracking, cur); 996 locked = 1; 997 } 998 if (cur->rule.ptr != NULL) { 999 cur->rule.ptr->src_nodes--; 1000 if (cur->rule.ptr->states <= 0 && 1001 cur->rule.ptr->max_src_nodes <= 0) 1002 pf_rm_rule(NULL, cur->rule.ptr); 1003 } 1004 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur); 1005 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 1006 pf_status.src_nodes--; 1007 pool_put(&pf_src_tree_pl, cur); 1008 } 1009 } 1010 1011 if (locked && !waslocked) 1012 rw_exit_write(&pf_consistency_lock); 1013 } 1014 1015 void 1016 pf_src_tree_remove_state(struct pf_state *s) 1017 { 1018 u_int32_t timeout; 1019 1020 if (s->src_node != NULL) { 1021 if (s->src.tcp_est) 1022 --s->src_node->conn; 1023 if (--s->src_node->states <= 0) { 1024 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE]; 1025 if (!timeout) 1026 timeout = 1027 pf_default_rule.timeout[PFTM_SRC_NODE]; 1028 s->src_node->expire = time_second + timeout; 1029 } 1030 } 1031 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) { 1032 if (--s->nat_src_node->states <= 0) { 1033 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE]; 1034 if (!timeout) 1035 timeout = 1036 pf_default_rule.timeout[PFTM_SRC_NODE]; 1037 s->nat_src_node->expire = time_second + timeout; 1038 } 1039 } 1040 s->src_node = s->nat_src_node = NULL; 1041 } 1042 1043 /* callers should be at splsoftnet */ 1044 void 1045 pf_unlink_state(struct pf_state *cur) 1046 { 1047 if (cur->src.state == PF_TCPS_PROXY_DST) { 1048 pf_send_tcp(cur->rule.ptr, cur->state_key->af, 1049 &cur->state_key->ext.addr, &cur->state_key->lan.addr, 1050 cur->state_key->ext.port, cur->state_key->lan.port, 1051 cur->src.seqhi, cur->src.seqlo + 1, 1052 TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL); 1053 } 1054 RB_REMOVE(pf_state_tree_id, &tree_id, cur); 1055 #if NPFSYNC 1056 if (cur->creatorid == pf_status.hostid) 1057 pfsync_delete_state(cur); 1058 #endif 1059 cur->timeout = PFTM_UNLINKED; 1060 pf_src_tree_remove_state(cur); 1061 pf_detach_state(cur, 0); 1062 } 1063 1064 /* callers should be at splsoftnet and hold the 1065 * write_lock on pf_consistency_lock */ 1066 void 1067 pf_free_state(struct pf_state *cur) 1068 { 1069 #if NPFSYNC 1070 if (pfsyncif != NULL && 1071 (pfsyncif->sc_bulk_send_next == cur || 1072 pfsyncif->sc_bulk_terminator == cur)) 1073 return; 1074 #endif 1075 KASSERT(cur->timeout == PFTM_UNLINKED); 1076 if (--cur->rule.ptr->states <= 0 && 1077 cur->rule.ptr->src_nodes <= 0) 1078 pf_rm_rule(NULL, cur->rule.ptr); 1079 if (cur->nat_rule.ptr != NULL) 1080 if (--cur->nat_rule.ptr->states <= 0 && 1081 cur->nat_rule.ptr->src_nodes <= 0) 1082 pf_rm_rule(NULL, cur->nat_rule.ptr); 1083 if (cur->anchor.ptr != NULL) 1084 if (--cur->anchor.ptr->states <= 0) 1085 pf_rm_rule(NULL, cur->anchor.ptr); 1086 pf_normalize_tcp_cleanup(cur); 1087 pfi_kif_unref(cur->kif, PFI_KIF_REF_STATE); 1088 TAILQ_REMOVE(&state_list, cur, entry_list); 1089 if (cur->tag) 1090 pf_tag_unref(cur->tag); 1091 pool_put(&pf_state_pl, cur); 1092 pf_status.fcounters[FCNT_STATE_REMOVALS]++; 1093 pf_status.states--; 1094 } 1095 1096 void 1097 pf_purge_expired_states(u_int32_t maxcheck) 1098 { 1099 static struct pf_state *cur = NULL; 1100 struct pf_state *next; 1101 int locked = 0; 1102 1103 while (maxcheck--) { 1104 /* wrap to start of list when we hit the end */ 1105 if (cur == NULL) { 1106 cur = TAILQ_FIRST(&state_list); 1107 if (cur == NULL) 1108 break; /* list empty */ 1109 } 1110 1111 /* get next state, as cur may get deleted */ 1112 next = TAILQ_NEXT(cur, entry_list); 1113 1114 if (cur->timeout == PFTM_UNLINKED) { 1115 /* free unlinked state */ 1116 if (! locked) { 1117 rw_enter_write(&pf_consistency_lock); 1118 locked = 1; 1119 } 1120 pf_free_state(cur); 1121 } else if (pf_state_expires(cur) <= time_second) { 1122 /* unlink and free expired state */ 1123 pf_unlink_state(cur); 1124 if (! locked) { 1125 rw_enter_write(&pf_consistency_lock); 1126 locked = 1; 1127 } 1128 pf_free_state(cur); 1129 } 1130 cur = next; 1131 } 1132 1133 if (locked) 1134 rw_exit_write(&pf_consistency_lock); 1135 } 1136 1137 int 1138 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw) 1139 { 1140 if (aw->type != PF_ADDR_TABLE) 1141 return (0); 1142 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL) 1143 return (1); 1144 return (0); 1145 } 1146 1147 void 1148 pf_tbladdr_remove(struct pf_addr_wrap *aw) 1149 { 1150 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL) 1151 return; 1152 pfr_detach_table(aw->p.tbl); 1153 aw->p.tbl = NULL; 1154 } 1155 1156 void 1157 pf_tbladdr_copyout(struct pf_addr_wrap *aw) 1158 { 1159 struct pfr_ktable *kt = aw->p.tbl; 1160 1161 if (aw->type != PF_ADDR_TABLE || kt == NULL) 1162 return; 1163 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) 1164 kt = kt->pfrkt_root; 1165 aw->p.tbl = NULL; 1166 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ? 1167 kt->pfrkt_cnt : -1; 1168 } 1169 1170 void 1171 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af) 1172 { 1173 switch (af) { 1174 #ifdef INET 1175 case AF_INET: { 1176 u_int32_t a = ntohl(addr->addr32[0]); 1177 printf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255, 1178 (a>>8)&255, a&255); 1179 if (p) { 1180 p = ntohs(p); 1181 printf(":%u", p); 1182 } 1183 break; 1184 } 1185 #endif /* INET */ 1186 #ifdef INET6 1187 case AF_INET6: { 1188 u_int16_t b; 1189 u_int8_t i, curstart = 255, curend = 0, 1190 maxstart = 0, maxend = 0; 1191 for (i = 0; i < 8; i++) { 1192 if (!addr->addr16[i]) { 1193 if (curstart == 255) 1194 curstart = i; 1195 else 1196 curend = i; 1197 } else { 1198 if (curstart) { 1199 if ((curend - curstart) > 1200 (maxend - maxstart)) { 1201 maxstart = curstart; 1202 maxend = curend; 1203 curstart = 255; 1204 } 1205 } 1206 } 1207 } 1208 for (i = 0; i < 8; i++) { 1209 if (i >= maxstart && i <= maxend) { 1210 if (maxend != 7) { 1211 if (i == maxstart) 1212 printf(":"); 1213 } else { 1214 if (i == maxend) 1215 printf(":"); 1216 } 1217 } else { 1218 b = ntohs(addr->addr16[i]); 1219 printf("%x", b); 1220 if (i < 7) 1221 printf(":"); 1222 } 1223 } 1224 if (p) { 1225 p = ntohs(p); 1226 printf("[%u]", p); 1227 } 1228 break; 1229 } 1230 #endif /* INET6 */ 1231 } 1232 } 1233 1234 void 1235 pf_print_state(struct pf_state *s) 1236 { 1237 struct pf_state_key *sk = s->state_key; 1238 switch (sk->proto) { 1239 case IPPROTO_TCP: 1240 printf("TCP "); 1241 break; 1242 case IPPROTO_UDP: 1243 printf("UDP "); 1244 break; 1245 case IPPROTO_ICMP: 1246 printf("ICMP "); 1247 break; 1248 case IPPROTO_ICMPV6: 1249 printf("ICMPV6 "); 1250 break; 1251 default: 1252 printf("%u ", sk->proto); 1253 break; 1254 } 1255 pf_print_host(&sk->lan.addr, sk->lan.port, sk->af); 1256 printf(" "); 1257 pf_print_host(&sk->gwy.addr, sk->gwy.port, sk->af); 1258 printf(" "); 1259 pf_print_host(&sk->ext.addr, sk->ext.port, sk->af); 1260 printf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo, 1261 s->src.seqhi, s->src.max_win, s->src.seqdiff); 1262 if (s->src.wscale && s->dst.wscale) 1263 printf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK); 1264 printf("]"); 1265 printf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo, 1266 s->dst.seqhi, s->dst.max_win, s->dst.seqdiff); 1267 if (s->src.wscale && s->dst.wscale) 1268 printf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK); 1269 printf("]"); 1270 printf(" %u:%u", s->src.state, s->dst.state); 1271 } 1272 1273 void 1274 pf_print_flags(u_int8_t f) 1275 { 1276 if (f) 1277 printf(" "); 1278 if (f & TH_FIN) 1279 printf("F"); 1280 if (f & TH_SYN) 1281 printf("S"); 1282 if (f & TH_RST) 1283 printf("R"); 1284 if (f & TH_PUSH) 1285 printf("P"); 1286 if (f & TH_ACK) 1287 printf("A"); 1288 if (f & TH_URG) 1289 printf("U"); 1290 if (f & TH_ECE) 1291 printf("E"); 1292 if (f & TH_CWR) 1293 printf("W"); 1294 } 1295 1296 #define PF_SET_SKIP_STEPS(i) \ 1297 do { \ 1298 while (head[i] != cur) { \ 1299 head[i]->skip[i].ptr = cur; \ 1300 head[i] = TAILQ_NEXT(head[i], entries); \ 1301 } \ 1302 } while (0) 1303 1304 void 1305 pf_calc_skip_steps(struct pf_rulequeue *rules) 1306 { 1307 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT]; 1308 int i; 1309 1310 cur = TAILQ_FIRST(rules); 1311 prev = cur; 1312 for (i = 0; i < PF_SKIP_COUNT; ++i) 1313 head[i] = cur; 1314 while (cur != NULL) { 1315 1316 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot) 1317 PF_SET_SKIP_STEPS(PF_SKIP_IFP); 1318 if (cur->direction != prev->direction) 1319 PF_SET_SKIP_STEPS(PF_SKIP_DIR); 1320 if (cur->af != prev->af) 1321 PF_SET_SKIP_STEPS(PF_SKIP_AF); 1322 if (cur->proto != prev->proto) 1323 PF_SET_SKIP_STEPS(PF_SKIP_PROTO); 1324 if (cur->src.neg != prev->src.neg || 1325 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr)) 1326 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR); 1327 if (cur->src.port[0] != prev->src.port[0] || 1328 cur->src.port[1] != prev->src.port[1] || 1329 cur->src.port_op != prev->src.port_op) 1330 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT); 1331 if (cur->dst.neg != prev->dst.neg || 1332 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr)) 1333 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR); 1334 if (cur->dst.port[0] != prev->dst.port[0] || 1335 cur->dst.port[1] != prev->dst.port[1] || 1336 cur->dst.port_op != prev->dst.port_op) 1337 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT); 1338 1339 prev = cur; 1340 cur = TAILQ_NEXT(cur, entries); 1341 } 1342 for (i = 0; i < PF_SKIP_COUNT; ++i) 1343 PF_SET_SKIP_STEPS(i); 1344 } 1345 1346 int 1347 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2) 1348 { 1349 if (aw1->type != aw2->type) 1350 return (1); 1351 switch (aw1->type) { 1352 case PF_ADDR_ADDRMASK: 1353 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0)) 1354 return (1); 1355 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0)) 1356 return (1); 1357 return (0); 1358 case PF_ADDR_DYNIFTL: 1359 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt); 1360 case PF_ADDR_NOROUTE: 1361 case PF_ADDR_URPFFAILED: 1362 return (0); 1363 case PF_ADDR_TABLE: 1364 return (aw1->p.tbl != aw2->p.tbl); 1365 case PF_ADDR_RTLABEL: 1366 return (aw1->v.rtlabel != aw2->v.rtlabel); 1367 default: 1368 printf("invalid address type: %d\n", aw1->type); 1369 return (1); 1370 } 1371 } 1372 1373 u_int16_t 1374 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp) 1375 { 1376 u_int32_t l; 1377 1378 if (udp && !cksum) 1379 return (0x0000); 1380 l = cksum + old - new; 1381 l = (l >> 16) + (l & 65535); 1382 l = l & 65535; 1383 if (udp && !l) 1384 return (0xFFFF); 1385 return (l); 1386 } 1387 1388 void 1389 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc, 1390 struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af) 1391 { 1392 struct pf_addr ao; 1393 u_int16_t po = *p; 1394 1395 PF_ACPY(&ao, a, af); 1396 PF_ACPY(a, an, af); 1397 1398 *p = pn; 1399 1400 switch (af) { 1401 #ifdef INET 1402 case AF_INET: 1403 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic, 1404 ao.addr16[0], an->addr16[0], 0), 1405 ao.addr16[1], an->addr16[1], 0); 1406 *p = pn; 1407 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc, 1408 ao.addr16[0], an->addr16[0], u), 1409 ao.addr16[1], an->addr16[1], u), 1410 po, pn, u); 1411 break; 1412 #endif /* INET */ 1413 #ifdef INET6 1414 case AF_INET6: 1415 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1416 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1417 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc, 1418 ao.addr16[0], an->addr16[0], u), 1419 ao.addr16[1], an->addr16[1], u), 1420 ao.addr16[2], an->addr16[2], u), 1421 ao.addr16[3], an->addr16[3], u), 1422 ao.addr16[4], an->addr16[4], u), 1423 ao.addr16[5], an->addr16[5], u), 1424 ao.addr16[6], an->addr16[6], u), 1425 ao.addr16[7], an->addr16[7], u), 1426 po, pn, u); 1427 break; 1428 #endif /* INET6 */ 1429 } 1430 } 1431 1432 1433 /* Changes a u_int32_t. Uses a void * so there are no align restrictions */ 1434 void 1435 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u) 1436 { 1437 u_int32_t ao; 1438 1439 memcpy(&ao, a, sizeof(ao)); 1440 memcpy(a, &an, sizeof(u_int32_t)); 1441 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u), 1442 ao % 65536, an % 65536, u); 1443 } 1444 1445 #ifdef INET6 1446 void 1447 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u) 1448 { 1449 struct pf_addr ao; 1450 1451 PF_ACPY(&ao, a, AF_INET6); 1452 PF_ACPY(a, an, AF_INET6); 1453 1454 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1455 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1456 pf_cksum_fixup(pf_cksum_fixup(*c, 1457 ao.addr16[0], an->addr16[0], u), 1458 ao.addr16[1], an->addr16[1], u), 1459 ao.addr16[2], an->addr16[2], u), 1460 ao.addr16[3], an->addr16[3], u), 1461 ao.addr16[4], an->addr16[4], u), 1462 ao.addr16[5], an->addr16[5], u), 1463 ao.addr16[6], an->addr16[6], u), 1464 ao.addr16[7], an->addr16[7], u); 1465 } 1466 #endif /* INET6 */ 1467 1468 void 1469 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa, 1470 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c, 1471 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af) 1472 { 1473 struct pf_addr oia, ooa; 1474 1475 PF_ACPY(&oia, ia, af); 1476 PF_ACPY(&ooa, oa, af); 1477 1478 /* Change inner protocol port, fix inner protocol checksum. */ 1479 if (ip != NULL) { 1480 u_int16_t oip = *ip; 1481 u_int32_t opc = 0; 1482 1483 if (pc != NULL) 1484 opc = *pc; 1485 *ip = np; 1486 if (pc != NULL) 1487 *pc = pf_cksum_fixup(*pc, oip, *ip, u); 1488 *ic = pf_cksum_fixup(*ic, oip, *ip, 0); 1489 if (pc != NULL) 1490 *ic = pf_cksum_fixup(*ic, opc, *pc, 0); 1491 } 1492 /* Change inner ip address, fix inner ip and icmp checksums. */ 1493 PF_ACPY(ia, na, af); 1494 switch (af) { 1495 #ifdef INET 1496 case AF_INET: { 1497 u_int32_t oh2c = *h2c; 1498 1499 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c, 1500 oia.addr16[0], ia->addr16[0], 0), 1501 oia.addr16[1], ia->addr16[1], 0); 1502 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic, 1503 oia.addr16[0], ia->addr16[0], 0), 1504 oia.addr16[1], ia->addr16[1], 0); 1505 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0); 1506 break; 1507 } 1508 #endif /* INET */ 1509 #ifdef INET6 1510 case AF_INET6: 1511 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1512 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1513 pf_cksum_fixup(pf_cksum_fixup(*ic, 1514 oia.addr16[0], ia->addr16[0], u), 1515 oia.addr16[1], ia->addr16[1], u), 1516 oia.addr16[2], ia->addr16[2], u), 1517 oia.addr16[3], ia->addr16[3], u), 1518 oia.addr16[4], ia->addr16[4], u), 1519 oia.addr16[5], ia->addr16[5], u), 1520 oia.addr16[6], ia->addr16[6], u), 1521 oia.addr16[7], ia->addr16[7], u); 1522 break; 1523 #endif /* INET6 */ 1524 } 1525 /* Change outer ip address, fix outer ip or icmpv6 checksum. */ 1526 PF_ACPY(oa, na, af); 1527 switch (af) { 1528 #ifdef INET 1529 case AF_INET: 1530 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc, 1531 ooa.addr16[0], oa->addr16[0], 0), 1532 ooa.addr16[1], oa->addr16[1], 0); 1533 break; 1534 #endif /* INET */ 1535 #ifdef INET6 1536 case AF_INET6: 1537 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1538 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1539 pf_cksum_fixup(pf_cksum_fixup(*ic, 1540 ooa.addr16[0], oa->addr16[0], u), 1541 ooa.addr16[1], oa->addr16[1], u), 1542 ooa.addr16[2], oa->addr16[2], u), 1543 ooa.addr16[3], oa->addr16[3], u), 1544 ooa.addr16[4], oa->addr16[4], u), 1545 ooa.addr16[5], oa->addr16[5], u), 1546 ooa.addr16[6], oa->addr16[6], u), 1547 ooa.addr16[7], oa->addr16[7], u); 1548 break; 1549 #endif /* INET6 */ 1550 } 1551 } 1552 1553 1554 /* 1555 * Need to modulate the sequence numbers in the TCP SACK option 1556 * (credits to Krzysztof Pfaff for report and patch) 1557 */ 1558 int 1559 pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd, 1560 struct tcphdr *th, struct pf_state_peer *dst) 1561 { 1562 int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen; 1563 u_int8_t opts[MAX_TCPOPTLEN], *opt = opts; 1564 int copyback = 0, i, olen; 1565 struct sackblk sack; 1566 1567 #ifdef __NetBSD__ 1568 #define TCPOLEN_SACK (2 * sizeof(uint32_t)) 1569 #endif 1570 1571 #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2) 1572 if (hlen < TCPOLEN_SACKLEN || 1573 !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af)) 1574 return 0; 1575 1576 while (hlen >= TCPOLEN_SACKLEN) { 1577 olen = opt[1]; 1578 switch (*opt) { 1579 case TCPOPT_EOL: /* FALLTHROUGH */ 1580 case TCPOPT_NOP: 1581 opt++; 1582 hlen--; 1583 break; 1584 case TCPOPT_SACK: 1585 if (olen > hlen) 1586 olen = hlen; 1587 if (olen >= TCPOLEN_SACKLEN) { 1588 for (i = 2; i + TCPOLEN_SACK <= olen; 1589 i += TCPOLEN_SACK) { 1590 memcpy(&sack, &opt[i], sizeof(sack)); 1591 #ifdef __NetBSD__ 1592 #define SACK_START sack.left 1593 #define SACK_END sack.right 1594 #else 1595 #define SACK_START sack.start 1596 #define SACK_END sack.end 1597 #endif 1598 pf_change_a(&SACK_START, &th->th_sum, 1599 htonl(ntohl(SACK_START) - 1600 dst->seqdiff), 0); 1601 pf_change_a(&SACK_END, &th->th_sum, 1602 htonl(ntohl(SACK_END) - 1603 dst->seqdiff), 0); 1604 #undef SACK_START 1605 #undef SACK_END 1606 memcpy(&opt[i], &sack, sizeof(sack)); 1607 } 1608 copyback = 1; 1609 } 1610 /* FALLTHROUGH */ 1611 default: 1612 if (olen < 2) 1613 olen = 2; 1614 hlen -= olen; 1615 opt += olen; 1616 } 1617 } 1618 1619 if (copyback) 1620 m_copyback(m, off + sizeof(*th), thoptlen, opts); 1621 return (copyback); 1622 } 1623 1624 void 1625 pf_send_tcp(const struct pf_rule *r, sa_family_t af, 1626 const struct pf_addr *saddr, const struct pf_addr *daddr, 1627 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack, 1628 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag, 1629 u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp) 1630 { 1631 struct mbuf *m; 1632 int len, tlen; 1633 #ifdef INET 1634 struct ip *h = NULL; 1635 #endif /* INET */ 1636 #ifdef INET6 1637 struct ip6_hdr *h6 = NULL; 1638 #endif /* INET6 */ 1639 struct tcphdr *th; 1640 char *opt; 1641 #ifdef __NetBSD__ 1642 struct pf_mtag *pf_mtag; 1643 #endif /* __NetBSD__ */ 1644 1645 /* maximum segment size tcp option */ 1646 tlen = sizeof(struct tcphdr); 1647 if (mss) 1648 tlen += 4; 1649 1650 switch (af) { 1651 #ifdef INET 1652 case AF_INET: 1653 len = sizeof(struct ip) + tlen; 1654 break; 1655 #endif /* INET */ 1656 #ifdef INET6 1657 case AF_INET6: 1658 len = sizeof(struct ip6_hdr) + tlen; 1659 break; 1660 #endif /* INET6 */ 1661 default: 1662 return; 1663 } 1664 1665 /* create outgoing mbuf */ 1666 m = m_gethdr(M_DONTWAIT, MT_HEADER); 1667 if (m == NULL) 1668 return; 1669 #ifdef __NetBSD__ 1670 if ((pf_mtag = pf_get_mtag(m)) == NULL) { 1671 m_freem(m); 1672 return; 1673 } 1674 if (tag) 1675 pf_mtag->flags |= PF_TAG_GENERATED; 1676 pf_mtag->tag = rtag; 1677 1678 if (r != NULL && r->rtableid >= 0) 1679 pf_mtag->rtableid = r->rtableid; 1680 #else 1681 if (tag) 1682 m->m_pkthdr.pf.flags |= PF_TAG_GENERATED; 1683 m->m_pkthdr.pf.tag = rtag; 1684 1685 if (r != NULL && r->rtableid >= 0) 1686 m->m_pkthdr.pf.rtableid = m->m_pkthdr.pf.rtableid; 1687 #endif /* !__NetBSD__ */ 1688 1689 #ifdef ALTQ 1690 if (r != NULL && r->qid) { 1691 #ifdef __NetBSD__ 1692 struct m_tag *mtag; 1693 struct altq_tag *atag; 1694 1695 mtag = m_tag_get(PACKET_TAG_ALTQ_QID, sizeof(*atag), M_NOWAIT); 1696 if (mtag != NULL) { 1697 atag = (struct altq_tag *)(mtag + 1); 1698 atag->qid = r->qid; 1699 /* add hints for ecn */ 1700 atag->af = af; 1701 atag->hdr = mtod(m, struct ip *); 1702 m_tag_prepend(m, mtag); 1703 } 1704 #else 1705 m->m_pkthdr.pf.qid = r->qid; 1706 /* add hints for ecn */ 1707 m->m_pkthdr.pf.hdr = mtod(m, struct ip *); 1708 #endif /* !__NetBSD__ */ 1709 } 1710 #endif /* ALTQ */ 1711 m->m_data += max_linkhdr; 1712 m->m_pkthdr.len = m->m_len = len; 1713 m->m_pkthdr.rcvif = NULL; 1714 bzero(m->m_data, len); 1715 switch (af) { 1716 #ifdef INET 1717 case AF_INET: 1718 h = mtod(m, struct ip *); 1719 1720 /* IP header fields included in the TCP checksum */ 1721 h->ip_p = IPPROTO_TCP; 1722 h->ip_len = htons(tlen); 1723 h->ip_src.s_addr = saddr->v4.s_addr; 1724 h->ip_dst.s_addr = daddr->v4.s_addr; 1725 1726 th = (struct tcphdr *)((char *)h + sizeof(struct ip)); 1727 break; 1728 #endif /* INET */ 1729 #ifdef INET6 1730 case AF_INET6: 1731 h6 = mtod(m, struct ip6_hdr *); 1732 1733 /* IP header fields included in the TCP checksum */ 1734 h6->ip6_nxt = IPPROTO_TCP; 1735 h6->ip6_plen = htons(tlen); 1736 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr)); 1737 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr)); 1738 1739 th = (struct tcphdr *)((char *)h6 + sizeof(struct ip6_hdr)); 1740 break; 1741 #endif /* INET6 */ 1742 default: 1743 m_freem(m); 1744 return; 1745 } 1746 1747 /* TCP header */ 1748 th->th_sport = sport; 1749 th->th_dport = dport; 1750 th->th_seq = htonl(seq); 1751 th->th_ack = htonl(ack); 1752 th->th_off = tlen >> 2; 1753 th->th_flags = flags; 1754 th->th_win = htons(win); 1755 1756 if (mss) { 1757 opt = (char *)(th + 1); 1758 opt[0] = TCPOPT_MAXSEG; 1759 opt[1] = 4; 1760 HTONS(mss); 1761 bcopy((void *)&mss, (void *)(opt + 2), 2); 1762 } 1763 1764 switch (af) { 1765 #ifdef INET 1766 case AF_INET: 1767 /* TCP checksum */ 1768 th->th_sum = in_cksum(m, len); 1769 1770 /* Finish the IP header */ 1771 h->ip_v = 4; 1772 h->ip_hl = sizeof(*h) >> 2; 1773 h->ip_tos = IPTOS_LOWDELAY; 1774 h->ip_len = htons(len); 1775 h->ip_off = htons(ip_mtudisc ? IP_DF : 0); 1776 h->ip_ttl = ttl ? ttl : ip_defttl; 1777 h->ip_sum = 0; 1778 if (eh == NULL) { 1779 ip_output(m, (void *)NULL, (void *)NULL, 0, 1780 (void *)NULL, (void *)NULL); 1781 } else { 1782 #ifdef __NetBSD__ 1783 /* 1784 * On netbsd, pf_test and pf_test6 are always called 1785 * with eh == NULL. 1786 */ 1787 panic("pf_send_tcp: eh != NULL"); 1788 #else 1789 struct route ro; 1790 struct rtentry rt; 1791 struct ether_header *e = (void *)ro.ro_dst.sa_data; 1792 1793 if (ifp == NULL) { 1794 m_freem(m); 1795 return; 1796 } 1797 rt.rt_ifp = ifp; 1798 ro.ro_rt = &rt; 1799 ro.ro_dst.sa_len = sizeof(ro.ro_dst); 1800 ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT; 1801 bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN); 1802 bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN); 1803 e->ether_type = eh->ether_type; 1804 ip_output(m, (void *)NULL, &ro, IP_ROUTETOETHER, 1805 (void *)NULL, (void *)NULL); 1806 #endif /* !__NetBSD__ */ 1807 } 1808 break; 1809 #endif /* INET */ 1810 #ifdef INET6 1811 case AF_INET6: 1812 /* TCP checksum */ 1813 th->th_sum = in6_cksum(m, IPPROTO_TCP, 1814 sizeof(struct ip6_hdr), tlen); 1815 1816 h6->ip6_vfc |= IPV6_VERSION; 1817 h6->ip6_hlim = IPV6_DEFHLIM; 1818 1819 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL); 1820 break; 1821 #endif /* INET6 */ 1822 } 1823 } 1824 1825 void 1826 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af, 1827 struct pf_rule *r) 1828 { 1829 struct mbuf *m0; 1830 #ifdef __NetBSD__ 1831 struct pf_mtag *pf_mtag; 1832 #endif /* __NetBSD__ */ 1833 1834 m0 = m_copy(m, 0, M_COPYALL); 1835 1836 #ifdef __NetBSD__ 1837 if ((pf_mtag = pf_get_mtag(m0)) == NULL) 1838 return; 1839 pf_mtag->flags |= PF_TAG_GENERATED; 1840 1841 if (r->rtableid >= 0) 1842 pf_mtag->rtableid = r->rtableid; 1843 #else 1844 m0->m_pkthdr.pf.flags |= PF_TAG_GENERATED; 1845 1846 if (r->rtableid >= 0) 1847 m0->m_pkthdr.pf.rtableid = r->rtableid; 1848 #endif /* !__NetBSD__ */ 1849 1850 #ifdef ALTQ 1851 if (r->qid) { 1852 #ifdef __NetBSD__ 1853 struct m_tag *mtag; 1854 struct altq_tag *atag; 1855 1856 mtag = m_tag_get(PACKET_TAG_ALTQ_QID, sizeof(*atag), M_NOWAIT); 1857 if (mtag != NULL) { 1858 atag = (struct altq_tag *)(mtag + 1); 1859 atag->qid = r->qid; 1860 /* add hints for ecn */ 1861 atag->af = af; 1862 atag->hdr = mtod(m0, struct ip *); 1863 m_tag_prepend(m0, mtag); 1864 } 1865 #else 1866 m0->m_pkthdr.pf.qid = r->qid; 1867 /* add hints for ecn */ 1868 m0->m_pkthdr.pf.hdr = mtod(m0, struct ip *); 1869 #endif /* !__NetBSD__ */ 1870 } 1871 #endif /* ALTQ */ 1872 1873 switch (af) { 1874 #ifdef INET 1875 case AF_INET: 1876 icmp_error(m0, type, code, 0, 0); 1877 break; 1878 #endif /* INET */ 1879 #ifdef INET6 1880 case AF_INET6: 1881 icmp6_error(m0, type, code, 0); 1882 break; 1883 #endif /* INET6 */ 1884 } 1885 } 1886 1887 /* 1888 * Return 1 if the addresses a and b match (with mask m), otherwise return 0. 1889 * If n is 0, they match if they are equal. If n is != 0, they match if they 1890 * are different. 1891 */ 1892 int 1893 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m, 1894 struct pf_addr *b, sa_family_t af) 1895 { 1896 int match = 0; 1897 1898 switch (af) { 1899 #ifdef INET 1900 case AF_INET: 1901 if ((a->addr32[0] & m->addr32[0]) == 1902 (b->addr32[0] & m->addr32[0])) 1903 match++; 1904 break; 1905 #endif /* INET */ 1906 #ifdef INET6 1907 case AF_INET6: 1908 if (((a->addr32[0] & m->addr32[0]) == 1909 (b->addr32[0] & m->addr32[0])) && 1910 ((a->addr32[1] & m->addr32[1]) == 1911 (b->addr32[1] & m->addr32[1])) && 1912 ((a->addr32[2] & m->addr32[2]) == 1913 (b->addr32[2] & m->addr32[2])) && 1914 ((a->addr32[3] & m->addr32[3]) == 1915 (b->addr32[3] & m->addr32[3]))) 1916 match++; 1917 break; 1918 #endif /* INET6 */ 1919 } 1920 if (match) { 1921 if (n) 1922 return (0); 1923 else 1924 return (1); 1925 } else { 1926 if (n) 1927 return (1); 1928 else 1929 return (0); 1930 } 1931 } 1932 1933 int 1934 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p) 1935 { 1936 switch (op) { 1937 case PF_OP_IRG: 1938 return ((p > a1) && (p < a2)); 1939 case PF_OP_XRG: 1940 return ((p < a1) || (p > a2)); 1941 case PF_OP_RRG: 1942 return ((p >= a1) && (p <= a2)); 1943 case PF_OP_EQ: 1944 return (p == a1); 1945 case PF_OP_NE: 1946 return (p != a1); 1947 case PF_OP_LT: 1948 return (p < a1); 1949 case PF_OP_LE: 1950 return (p <= a1); 1951 case PF_OP_GT: 1952 return (p > a1); 1953 case PF_OP_GE: 1954 return (p >= a1); 1955 } 1956 return (0); /* never reached */ 1957 } 1958 1959 int 1960 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p) 1961 { 1962 NTOHS(a1); 1963 NTOHS(a2); 1964 NTOHS(p); 1965 return (pf_match(op, a1, a2, p)); 1966 } 1967 1968 int 1969 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u) 1970 { 1971 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE) 1972 return (0); 1973 return (pf_match(op, a1, a2, u)); 1974 } 1975 1976 int 1977 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g) 1978 { 1979 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE) 1980 return (0); 1981 return (pf_match(op, a1, a2, g)); 1982 } 1983 1984 int 1985 pf_match_tag(struct mbuf *m, struct pf_rule *r, int *tag) 1986 { 1987 #ifdef __NetBSD__ 1988 if (*tag == -1) { 1989 struct pf_mtag *pf_mtag = pf_get_mtag(m); 1990 if (pf_mtag == NULL) 1991 return (0); 1992 1993 *tag = pf_mtag->tag; 1994 } 1995 #else 1996 if (*tag == -1) 1997 *tag = m->m_pkthdr.pf.tag; 1998 #endif /* !__NetBSD__ */ 1999 2000 return ((!r->match_tag_not && r->match_tag == *tag) || 2001 (r->match_tag_not && r->match_tag != *tag)); 2002 } 2003 2004 int 2005 pf_tag_packet(struct mbuf *m, int tag, int rtableid) 2006 { 2007 if (tag <= 0 && rtableid < 0) 2008 return (0); 2009 2010 #ifdef __NetBSD__ 2011 if (tag > 0 || rtableid > 0) { 2012 struct pf_mtag *pf_mtag = pf_get_mtag(m); 2013 if (pf_mtag == NULL) 2014 return (1); 2015 2016 if (tag > 0) 2017 pf_mtag->tag = tag; 2018 if (rtableid > 0) 2019 pf_mtag->rtableid = rtableid; 2020 } 2021 #else 2022 if (tag > 0) 2023 m->m_pkthdr.pf.tag = tag; 2024 if (rtableid >= 0) 2025 m->m_pkthdr.pf.rtableid = rtableid; 2026 #endif /* !__NetBSD__ */ 2027 2028 return (0); 2029 } 2030 2031 void 2032 pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n, 2033 struct pf_rule **r, struct pf_rule **a, int *match) 2034 { 2035 struct pf_anchor_stackframe *f; 2036 2037 (*r)->anchor->match = 0; 2038 if (match) 2039 *match = 0; 2040 if (*depth >= sizeof(pf_anchor_stack) / 2041 sizeof(pf_anchor_stack[0])) { 2042 printf("pf_step_into_anchor: stack overflow\n"); 2043 *r = TAILQ_NEXT(*r, entries); 2044 return; 2045 } else if (*depth == 0 && a != NULL) 2046 *a = *r; 2047 f = pf_anchor_stack + (*depth)++; 2048 f->rs = *rs; 2049 f->r = *r; 2050 if ((*r)->anchor_wildcard) { 2051 f->parent = &(*r)->anchor->children; 2052 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) == 2053 NULL) { 2054 *r = NULL; 2055 return; 2056 } 2057 *rs = &f->child->ruleset; 2058 } else { 2059 f->parent = NULL; 2060 f->child = NULL; 2061 *rs = &(*r)->anchor->ruleset; 2062 } 2063 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr); 2064 } 2065 2066 int 2067 pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n, 2068 struct pf_rule **r, struct pf_rule **a, int *match) 2069 { 2070 struct pf_anchor_stackframe *f; 2071 int quick = 0; 2072 2073 do { 2074 if (*depth <= 0) 2075 break; 2076 f = pf_anchor_stack + *depth - 1; 2077 if (f->parent != NULL && f->child != NULL) { 2078 if (f->child->match || 2079 (match != NULL && *match)) { 2080 f->r->anchor->match = 1; 2081 *match = 0; 2082 } 2083 f->child = RB_NEXT(pf_anchor_node, f->parent, f->child); 2084 if (f->child != NULL) { 2085 *rs = &f->child->ruleset; 2086 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr); 2087 if (*r == NULL) 2088 continue; 2089 else 2090 break; 2091 } 2092 } 2093 (*depth)--; 2094 if (*depth == 0 && a != NULL) 2095 *a = NULL; 2096 *rs = f->rs; 2097 if (f->r->anchor->match || (match != NULL && *match)) 2098 quick = f->r->quick; 2099 *r = TAILQ_NEXT(f->r, entries); 2100 } while (*r == NULL); 2101 2102 return (quick); 2103 } 2104 2105 #ifdef INET6 2106 void 2107 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr, 2108 struct pf_addr *rmask, const struct pf_addr *saddr, sa_family_t af) 2109 { 2110 switch (af) { 2111 #ifdef INET 2112 case AF_INET: 2113 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) | 2114 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]); 2115 break; 2116 #endif /* INET */ 2117 case AF_INET6: 2118 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) | 2119 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]); 2120 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) | 2121 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]); 2122 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) | 2123 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]); 2124 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) | 2125 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]); 2126 break; 2127 } 2128 } 2129 2130 void 2131 pf_addr_inc(struct pf_addr *addr, sa_family_t af) 2132 { 2133 switch (af) { 2134 #ifdef INET 2135 case AF_INET: 2136 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1); 2137 break; 2138 #endif /* INET */ 2139 case AF_INET6: 2140 if (addr->addr32[3] == 0xffffffff) { 2141 addr->addr32[3] = 0; 2142 if (addr->addr32[2] == 0xffffffff) { 2143 addr->addr32[2] = 0; 2144 if (addr->addr32[1] == 0xffffffff) { 2145 addr->addr32[1] = 0; 2146 addr->addr32[0] = 2147 htonl(ntohl(addr->addr32[0]) + 1); 2148 } else 2149 addr->addr32[1] = 2150 htonl(ntohl(addr->addr32[1]) + 1); 2151 } else 2152 addr->addr32[2] = 2153 htonl(ntohl(addr->addr32[2]) + 1); 2154 } else 2155 addr->addr32[3] = 2156 htonl(ntohl(addr->addr32[3]) + 1); 2157 break; 2158 } 2159 } 2160 #endif /* INET6 */ 2161 2162 #define mix(a,b,c) \ 2163 do { \ 2164 a -= b; a -= c; a ^= (c >> 13); \ 2165 b -= c; b -= a; b ^= (a << 8); \ 2166 c -= a; c -= b; c ^= (b >> 13); \ 2167 a -= b; a -= c; a ^= (c >> 12); \ 2168 b -= c; b -= a; b ^= (a << 16); \ 2169 c -= a; c -= b; c ^= (b >> 5); \ 2170 a -= b; a -= c; a ^= (c >> 3); \ 2171 b -= c; b -= a; b ^= (a << 10); \ 2172 c -= a; c -= b; c ^= (b >> 15); \ 2173 } while (0) 2174 2175 /* 2176 * hash function based on bridge_hash in if_bridge.c 2177 */ 2178 void 2179 pf_hash(const struct pf_addr *inaddr, struct pf_addr *hash, 2180 struct pf_poolhashkey *key, sa_family_t af) 2181 { 2182 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0]; 2183 2184 switch (af) { 2185 #ifdef INET 2186 case AF_INET: 2187 a += inaddr->addr32[0]; 2188 b += key->key32[1]; 2189 mix(a, b, c); 2190 hash->addr32[0] = c + key->key32[2]; 2191 break; 2192 #endif /* INET */ 2193 #ifdef INET6 2194 case AF_INET6: 2195 a += inaddr->addr32[0]; 2196 b += inaddr->addr32[2]; 2197 mix(a, b, c); 2198 hash->addr32[0] = c; 2199 a += inaddr->addr32[1]; 2200 b += inaddr->addr32[3]; 2201 c += key->key32[1]; 2202 mix(a, b, c); 2203 hash->addr32[1] = c; 2204 a += inaddr->addr32[2]; 2205 b += inaddr->addr32[1]; 2206 c += key->key32[2]; 2207 mix(a, b, c); 2208 hash->addr32[2] = c; 2209 a += inaddr->addr32[3]; 2210 b += inaddr->addr32[0]; 2211 c += key->key32[3]; 2212 mix(a, b, c); 2213 hash->addr32[3] = c; 2214 break; 2215 #endif /* INET6 */ 2216 } 2217 } 2218 2219 int 2220 pf_map_addr(sa_family_t af, struct pf_rule *r, const struct pf_addr *saddr, 2221 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn) 2222 { 2223 unsigned char hash[16]; 2224 struct pf_pool *rpool = &r->rpool; 2225 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr; 2226 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask; 2227 struct pf_pooladdr *acur = rpool->cur; 2228 struct pf_src_node k; 2229 2230 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR && 2231 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { 2232 k.af = af; 2233 PF_ACPY(&k.addr, saddr, af); 2234 if (r->rule_flag & PFRULE_RULESRCTRACK || 2235 r->rpool.opts & PF_POOL_STICKYADDR) 2236 k.rule.ptr = r; 2237 else 2238 k.rule.ptr = NULL; 2239 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++; 2240 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k); 2241 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) { 2242 PF_ACPY(naddr, &(*sn)->raddr, af); 2243 if (pf_status.debug >= PF_DEBUG_MISC) { 2244 printf("pf_map_addr: src tracking maps "); 2245 pf_print_host(&k.addr, 0, af); 2246 printf(" to "); 2247 pf_print_host(naddr, 0, af); 2248 printf("\n"); 2249 } 2250 return (0); 2251 } 2252 } 2253 2254 if (rpool->cur->addr.type == PF_ADDR_NOROUTE) 2255 return (1); 2256 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2257 switch (af) { 2258 #ifdef INET 2259 case AF_INET: 2260 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 && 2261 (rpool->opts & PF_POOL_TYPEMASK) != 2262 PF_POOL_ROUNDROBIN) 2263 return (1); 2264 raddr = &rpool->cur->addr.p.dyn->pfid_addr4; 2265 rmask = &rpool->cur->addr.p.dyn->pfid_mask4; 2266 break; 2267 #endif /* INET */ 2268 #ifdef INET6 2269 case AF_INET6: 2270 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 && 2271 (rpool->opts & PF_POOL_TYPEMASK) != 2272 PF_POOL_ROUNDROBIN) 2273 return (1); 2274 raddr = &rpool->cur->addr.p.dyn->pfid_addr6; 2275 rmask = &rpool->cur->addr.p.dyn->pfid_mask6; 2276 break; 2277 #endif /* INET6 */ 2278 } 2279 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2280 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN) 2281 return (1); /* unsupported */ 2282 } else { 2283 raddr = &rpool->cur->addr.v.a.addr; 2284 rmask = &rpool->cur->addr.v.a.mask; 2285 } 2286 2287 switch (rpool->opts & PF_POOL_TYPEMASK) { 2288 case PF_POOL_NONE: 2289 PF_ACPY(naddr, raddr, af); 2290 break; 2291 case PF_POOL_BITMASK: 2292 PF_POOLMASK(naddr, raddr, rmask, saddr, af); 2293 break; 2294 case PF_POOL_RANDOM: 2295 if (init_addr != NULL && PF_AZERO(init_addr, af)) { 2296 switch (af) { 2297 #ifdef INET 2298 case AF_INET: 2299 rpool->counter.addr32[0] = htonl(arc4random()); 2300 break; 2301 #endif /* INET */ 2302 #ifdef INET6 2303 case AF_INET6: 2304 if (rmask->addr32[3] != 0xffffffff) 2305 rpool->counter.addr32[3] = 2306 htonl(arc4random()); 2307 else 2308 break; 2309 if (rmask->addr32[2] != 0xffffffff) 2310 rpool->counter.addr32[2] = 2311 htonl(arc4random()); 2312 else 2313 break; 2314 if (rmask->addr32[1] != 0xffffffff) 2315 rpool->counter.addr32[1] = 2316 htonl(arc4random()); 2317 else 2318 break; 2319 if (rmask->addr32[0] != 0xffffffff) 2320 rpool->counter.addr32[0] = 2321 htonl(arc4random()); 2322 break; 2323 #endif /* INET6 */ 2324 } 2325 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af); 2326 PF_ACPY(init_addr, naddr, af); 2327 2328 } else { 2329 PF_AINC(&rpool->counter, af); 2330 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af); 2331 } 2332 break; 2333 case PF_POOL_SRCHASH: 2334 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af); 2335 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af); 2336 break; 2337 case PF_POOL_ROUNDROBIN: 2338 if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2339 if (!pfr_pool_get(rpool->cur->addr.p.tbl, 2340 &rpool->tblidx, &rpool->counter, 2341 &raddr, &rmask, af)) 2342 goto get_addr; 2343 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2344 if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, 2345 &rpool->tblidx, &rpool->counter, 2346 &raddr, &rmask, af)) 2347 goto get_addr; 2348 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af)) 2349 goto get_addr; 2350 2351 try_next: 2352 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL) 2353 rpool->cur = TAILQ_FIRST(&rpool->list); 2354 if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2355 rpool->tblidx = -1; 2356 if (pfr_pool_get(rpool->cur->addr.p.tbl, 2357 &rpool->tblidx, &rpool->counter, 2358 &raddr, &rmask, af)) { 2359 /* table contains no address of type 'af' */ 2360 if (rpool->cur != acur) 2361 goto try_next; 2362 return (1); 2363 } 2364 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2365 rpool->tblidx = -1; 2366 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, 2367 &rpool->tblidx, &rpool->counter, 2368 &raddr, &rmask, af)) { 2369 /* table contains no address of type 'af' */ 2370 if (rpool->cur != acur) 2371 goto try_next; 2372 return (1); 2373 } 2374 } else { 2375 raddr = &rpool->cur->addr.v.a.addr; 2376 rmask = &rpool->cur->addr.v.a.mask; 2377 PF_ACPY(&rpool->counter, raddr, af); 2378 } 2379 2380 get_addr: 2381 PF_ACPY(naddr, &rpool->counter, af); 2382 if (init_addr != NULL && PF_AZERO(init_addr, af)) 2383 PF_ACPY(init_addr, naddr, af); 2384 PF_AINC(&rpool->counter, af); 2385 break; 2386 } 2387 if (*sn != NULL) 2388 PF_ACPY(&(*sn)->raddr, naddr, af); 2389 2390 if (pf_status.debug >= PF_DEBUG_MISC && 2391 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { 2392 printf("pf_map_addr: selected address "); 2393 pf_print_host(naddr, 0, af); 2394 printf("\n"); 2395 } 2396 2397 return (0); 2398 } 2399 2400 int 2401 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r, 2402 struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport, 2403 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high, 2404 struct pf_src_node **sn) 2405 { 2406 struct pf_state_key_cmp key; 2407 struct pf_addr init_addr; 2408 u_int16_t cut; 2409 2410 bzero(&init_addr, sizeof(init_addr)); 2411 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn)) 2412 return (1); 2413 2414 if (proto == IPPROTO_ICMP) { 2415 low = 1; 2416 high = 65535; 2417 } 2418 2419 do { 2420 key.af = af; 2421 key.proto = proto; 2422 PF_ACPY(&key.ext.addr, daddr, key.af); 2423 PF_ACPY(&key.gwy.addr, naddr, key.af); 2424 key.ext.port = dport; 2425 2426 /* 2427 * port search; start random, step; 2428 * similar 2 portloop in in_pcbbind 2429 */ 2430 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP || 2431 proto == IPPROTO_ICMP)) { 2432 key.gwy.port = dport; 2433 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) 2434 return (0); 2435 } else if (low == 0 && high == 0) { 2436 key.gwy.port = *nport; 2437 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) 2438 return (0); 2439 } else if (low == high) { 2440 key.gwy.port = htons(low); 2441 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) { 2442 *nport = htons(low); 2443 return (0); 2444 } 2445 } else { 2446 u_int16_t tmp; 2447 2448 if (low > high) { 2449 tmp = low; 2450 low = high; 2451 high = tmp; 2452 } 2453 /* low < high */ 2454 cut = htonl(arc4random()) % (1 + high - low) + low; 2455 /* low <= cut <= high */ 2456 for (tmp = cut; tmp <= high; ++(tmp)) { 2457 key.gwy.port = htons(tmp); 2458 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == 2459 NULL) { 2460 *nport = htons(tmp); 2461 return (0); 2462 } 2463 } 2464 for (tmp = cut - 1; tmp >= low; --(tmp)) { 2465 key.gwy.port = htons(tmp); 2466 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == 2467 NULL) { 2468 *nport = htons(tmp); 2469 return (0); 2470 } 2471 } 2472 } 2473 2474 switch (r->rpool.opts & PF_POOL_TYPEMASK) { 2475 case PF_POOL_RANDOM: 2476 case PF_POOL_ROUNDROBIN: 2477 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn)) 2478 return (1); 2479 break; 2480 case PF_POOL_NONE: 2481 case PF_POOL_SRCHASH: 2482 case PF_POOL_BITMASK: 2483 default: 2484 return (1); 2485 } 2486 } while (! PF_AEQ(&init_addr, naddr, af) ); 2487 2488 return (1); /* none available */ 2489 } 2490 2491 struct pf_rule * 2492 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off, 2493 int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport, 2494 struct pf_addr *daddr, u_int16_t dport, int rs_num) 2495 { 2496 struct pf_rule *r, *rm = NULL; 2497 struct pf_ruleset *ruleset = NULL; 2498 int tag = -1; 2499 int rtableid = -1; 2500 int asd = 0; 2501 2502 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr); 2503 while (r && rm == NULL) { 2504 struct pf_rule_addr *src = NULL, *dst = NULL; 2505 struct pf_addr_wrap *xdst = NULL; 2506 2507 if (r->action == PF_BINAT && direction == PF_IN) { 2508 src = &r->dst; 2509 if (r->rpool.cur != NULL) 2510 xdst = &r->rpool.cur->addr; 2511 } else { 2512 src = &r->src; 2513 dst = &r->dst; 2514 } 2515 2516 r->evaluations++; 2517 if (pfi_kif_match(r->kif, kif) == r->ifnot) 2518 r = r->skip[PF_SKIP_IFP].ptr; 2519 else if (r->direction && r->direction != direction) 2520 r = r->skip[PF_SKIP_DIR].ptr; 2521 else if (r->af && r->af != pd->af) 2522 r = r->skip[PF_SKIP_AF].ptr; 2523 else if (r->proto && r->proto != pd->proto) 2524 r = r->skip[PF_SKIP_PROTO].ptr; 2525 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af, 2526 src->neg, kif)) 2527 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR : 2528 PF_SKIP_DST_ADDR].ptr; 2529 else if (src->port_op && !pf_match_port(src->port_op, 2530 src->port[0], src->port[1], sport)) 2531 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT : 2532 PF_SKIP_DST_PORT].ptr; 2533 else if (dst != NULL && 2534 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL)) 2535 r = r->skip[PF_SKIP_DST_ADDR].ptr; 2536 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af, 2537 0, NULL)) 2538 r = TAILQ_NEXT(r, entries); 2539 else if (dst != NULL && dst->port_op && 2540 !pf_match_port(dst->port_op, dst->port[0], 2541 dst->port[1], dport)) 2542 r = r->skip[PF_SKIP_DST_PORT].ptr; 2543 else if (r->match_tag && !pf_match_tag(m, r, &tag)) 2544 r = TAILQ_NEXT(r, entries); 2545 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto != 2546 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m, 2547 off, pd->hdr.tcp), r->os_fingerprint))) 2548 r = TAILQ_NEXT(r, entries); 2549 else { 2550 if (r->tag) 2551 tag = r->tag; 2552 if (r->rtableid >= 0) 2553 rtableid = r->rtableid; 2554 if (r->anchor == NULL) { 2555 rm = r; 2556 } else 2557 pf_step_into_anchor(&asd, &ruleset, rs_num, 2558 &r, NULL, NULL); 2559 } 2560 if (r == NULL) 2561 pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r, 2562 NULL, NULL); 2563 } 2564 if (pf_tag_packet(m, tag, rtableid)) 2565 return (NULL); 2566 if (rm != NULL && (rm->action == PF_NONAT || 2567 rm->action == PF_NORDR || rm->action == PF_NOBINAT)) 2568 return (NULL); 2569 return (rm); 2570 } 2571 2572 struct pf_rule * 2573 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction, 2574 struct pfi_kif *kif, struct pf_src_node **sn, 2575 struct pf_addr *saddr, u_int16_t sport, 2576 struct pf_addr *daddr, u_int16_t dport, 2577 struct pf_addr *naddr, u_int16_t *nport) 2578 { 2579 struct pf_rule *r = NULL; 2580 2581 if (direction == PF_OUT) { 2582 r = pf_match_translation(pd, m, off, direction, kif, saddr, 2583 sport, daddr, dport, PF_RULESET_BINAT); 2584 if (r == NULL) 2585 r = pf_match_translation(pd, m, off, direction, kif, 2586 saddr, sport, daddr, dport, PF_RULESET_NAT); 2587 } else { 2588 r = pf_match_translation(pd, m, off, direction, kif, saddr, 2589 sport, daddr, dport, PF_RULESET_RDR); 2590 if (r == NULL) 2591 r = pf_match_translation(pd, m, off, direction, kif, 2592 saddr, sport, daddr, dport, PF_RULESET_BINAT); 2593 } 2594 2595 if (r != NULL) { 2596 switch (r->action) { 2597 case PF_NONAT: 2598 case PF_NOBINAT: 2599 case PF_NORDR: 2600 return (NULL); 2601 case PF_NAT: 2602 if (pf_get_sport(pd->af, pd->proto, r, saddr, 2603 daddr, dport, naddr, nport, r->rpool.proxy_port[0], 2604 r->rpool.proxy_port[1], sn)) { 2605 DPFPRINTF(PF_DEBUG_MISC, 2606 ("pf: NAT proxy port allocation " 2607 "(%u-%u) failed\n", 2608 r->rpool.proxy_port[0], 2609 r->rpool.proxy_port[1])); 2610 return (NULL); 2611 } 2612 break; 2613 case PF_BINAT: 2614 switch (direction) { 2615 case PF_OUT: 2616 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){ 2617 switch (pd->af) { 2618 #ifdef INET 2619 case AF_INET: 2620 if (r->rpool.cur->addr.p.dyn-> 2621 pfid_acnt4 < 1) 2622 return (NULL); 2623 PF_POOLMASK(naddr, 2624 &r->rpool.cur->addr.p.dyn-> 2625 pfid_addr4, 2626 &r->rpool.cur->addr.p.dyn-> 2627 pfid_mask4, 2628 saddr, AF_INET); 2629 break; 2630 #endif /* INET */ 2631 #ifdef INET6 2632 case AF_INET6: 2633 if (r->rpool.cur->addr.p.dyn-> 2634 pfid_acnt6 < 1) 2635 return (NULL); 2636 PF_POOLMASK(naddr, 2637 &r->rpool.cur->addr.p.dyn-> 2638 pfid_addr6, 2639 &r->rpool.cur->addr.p.dyn-> 2640 pfid_mask6, 2641 saddr, AF_INET6); 2642 break; 2643 #endif /* INET6 */ 2644 } 2645 } else 2646 PF_POOLMASK(naddr, 2647 &r->rpool.cur->addr.v.a.addr, 2648 &r->rpool.cur->addr.v.a.mask, 2649 saddr, pd->af); 2650 break; 2651 case PF_IN: 2652 if (r->src.addr.type == PF_ADDR_DYNIFTL) { 2653 switch (pd->af) { 2654 #ifdef INET 2655 case AF_INET: 2656 if (r->src.addr.p.dyn-> 2657 pfid_acnt4 < 1) 2658 return (NULL); 2659 PF_POOLMASK(naddr, 2660 &r->src.addr.p.dyn-> 2661 pfid_addr4, 2662 &r->src.addr.p.dyn-> 2663 pfid_mask4, 2664 daddr, AF_INET); 2665 break; 2666 #endif /* INET */ 2667 #ifdef INET6 2668 case AF_INET6: 2669 if (r->src.addr.p.dyn-> 2670 pfid_acnt6 < 1) 2671 return (NULL); 2672 PF_POOLMASK(naddr, 2673 &r->src.addr.p.dyn-> 2674 pfid_addr6, 2675 &r->src.addr.p.dyn-> 2676 pfid_mask6, 2677 daddr, AF_INET6); 2678 break; 2679 #endif /* INET6 */ 2680 } 2681 } else 2682 PF_POOLMASK(naddr, 2683 &r->src.addr.v.a.addr, 2684 &r->src.addr.v.a.mask, daddr, 2685 pd->af); 2686 break; 2687 } 2688 break; 2689 case PF_RDR: { 2690 if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn)) 2691 return (NULL); 2692 if ((r->rpool.opts & PF_POOL_TYPEMASK) == 2693 PF_POOL_BITMASK) 2694 PF_POOLMASK(naddr, naddr, 2695 &r->rpool.cur->addr.v.a.mask, daddr, 2696 pd->af); 2697 2698 if (r->rpool.proxy_port[1]) { 2699 u_int32_t tmp_nport; 2700 2701 tmp_nport = ((ntohs(dport) - 2702 ntohs(r->dst.port[0])) % 2703 (r->rpool.proxy_port[1] - 2704 r->rpool.proxy_port[0] + 1)) + 2705 r->rpool.proxy_port[0]; 2706 2707 /* wrap around if necessary */ 2708 if (tmp_nport > 65535) 2709 tmp_nport -= 65535; 2710 *nport = htons((u_int16_t)tmp_nport); 2711 } else if (r->rpool.proxy_port[0]) 2712 *nport = htons(r->rpool.proxy_port[0]); 2713 break; 2714 } 2715 default: 2716 return (NULL); 2717 } 2718 } 2719 2720 return (r); 2721 } 2722 2723 int 2724 pf_socket_lookup(int direction, struct pf_pdesc *pd) 2725 { 2726 struct pf_addr *saddr, *daddr; 2727 u_int16_t sport, dport; 2728 struct inpcbtable *tb; 2729 struct inpcb *inp = NULL; 2730 struct socket *so = NULL; 2731 #if defined(__NetBSD__) && defined(INET6) 2732 struct in6pcb *in6p = NULL; 2733 #else 2734 #define in6p inp 2735 #endif /* __NetBSD__ && INET6 */ 2736 2737 if (pd == NULL) 2738 return (-1); 2739 pd->lookup.uid = UID_MAX; 2740 pd->lookup.gid = GID_MAX; 2741 pd->lookup.pid = NO_PID; 2742 switch (pd->proto) { 2743 case IPPROTO_TCP: 2744 if (pd->hdr.tcp == NULL) 2745 return (-1); 2746 sport = pd->hdr.tcp->th_sport; 2747 dport = pd->hdr.tcp->th_dport; 2748 tb = &tcbtable; 2749 break; 2750 case IPPROTO_UDP: 2751 if (pd->hdr.udp == NULL) 2752 return (-1); 2753 sport = pd->hdr.udp->uh_sport; 2754 dport = pd->hdr.udp->uh_dport; 2755 tb = &udbtable; 2756 break; 2757 default: 2758 return (-1); 2759 } 2760 if (direction == PF_IN) { 2761 saddr = pd->src; 2762 daddr = pd->dst; 2763 } else { 2764 u_int16_t p; 2765 2766 p = sport; 2767 sport = dport; 2768 dport = p; 2769 saddr = pd->dst; 2770 daddr = pd->src; 2771 } 2772 switch (pd->af) { 2773 2774 #ifdef __NetBSD__ 2775 #define in_pcbhashlookup(tbl, saddr, sport, daddr, dport) \ 2776 in_pcblookup_connect(tbl, saddr, sport, daddr, dport) 2777 #define in6_pcbhashlookup(tbl, saddr, sport, daddr, dport) \ 2778 in6_pcblookup_connect(tbl, saddr, sport, daddr, dport, 0) 2779 #define in_pcblookup_listen(tbl, addr, port, zero) \ 2780 in_pcblookup_bind(tbl, addr, port) 2781 #define in6_pcblookup_listen(tbl, addr, port, zero) \ 2782 in6_pcblookup_bind(tbl, addr, port, zero) 2783 #endif 2784 2785 #ifdef INET 2786 case AF_INET: 2787 inp = in_pcbhashlookup(tb, saddr->v4, sport, daddr->v4, dport); 2788 if (inp == NULL) { 2789 inp = in_pcblookup_listen(tb, daddr->v4, dport, 0); 2790 if (inp == NULL) 2791 return (-1); 2792 } 2793 break; 2794 #endif /* INET */ 2795 #ifdef INET6 2796 case AF_INET6: 2797 /*###2817 [cc] warning: assignment from incompatible pointer type%%%*/ 2798 in6p = in6_pcbhashlookup(tb, &saddr->v6, sport, &daddr->v6, 2799 dport); 2800 if (inp == NULL) { 2801 in6p = in6_pcblookup_listen(tb, &daddr->v6, dport, 0); 2802 if (inp == NULL) 2803 return (-1); 2804 } 2805 break; 2806 #endif /* INET6 */ 2807 2808 default: 2809 return (-1); 2810 } 2811 2812 #ifdef __NetBSD__ 2813 switch (pd->af) { 2814 #ifdef INET 2815 case AF_INET: 2816 so = inp->inp_socket; 2817 break; 2818 #endif 2819 #ifdef INET6 2820 case AF_INET6: 2821 /*###2840 [cc] error: 'struct inpcb' has no member named 'in6p_head'%%%*/ 2822 so = in6p->in6p_socket; 2823 break; 2824 #endif /* INET6 */ 2825 } 2826 pd->lookup.uid = kauth_cred_geteuid(so->so_cred); 2827 pd->lookup.gid = kauth_cred_getegid(so->so_cred); 2828 #else 2829 so = inp->inp_socket; 2830 pd->lookup.uid = so->so_euid; 2831 pd->lookup.gid = so->so_egid; 2832 #endif /* !__NetBSD__ */ 2833 pd->lookup.pid = so->so_cpid; 2834 return (1); 2835 } 2836 2837 u_int8_t 2838 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af) 2839 { 2840 int hlen; 2841 u_int8_t hdr[60]; 2842 u_int8_t *opt, optlen; 2843 u_int8_t wscale = 0; 2844 2845 hlen = th_off << 2; /* hlen <= sizeof(hdr) */ 2846 if (hlen <= sizeof(struct tcphdr)) 2847 return (0); 2848 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af)) 2849 return (0); 2850 opt = hdr + sizeof(struct tcphdr); 2851 hlen -= sizeof(struct tcphdr); 2852 while (hlen >= 3) { 2853 switch (*opt) { 2854 case TCPOPT_EOL: 2855 case TCPOPT_NOP: 2856 ++opt; 2857 --hlen; 2858 break; 2859 case TCPOPT_WINDOW: 2860 wscale = opt[2]; 2861 if (wscale > TCP_MAX_WINSHIFT) 2862 wscale = TCP_MAX_WINSHIFT; 2863 wscale |= PF_WSCALE_FLAG; 2864 /* FALLTHROUGH */ 2865 default: 2866 optlen = opt[1]; 2867 if (optlen < 2) 2868 optlen = 2; 2869 hlen -= optlen; 2870 opt += optlen; 2871 break; 2872 } 2873 } 2874 return (wscale); 2875 } 2876 2877 u_int16_t 2878 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af) 2879 { 2880 int hlen; 2881 u_int8_t hdr[60]; 2882 u_int8_t *opt, optlen; 2883 u_int16_t mss = tcp_mssdflt; 2884 2885 hlen = th_off << 2; /* hlen <= sizeof(hdr) */ 2886 if (hlen <= sizeof(struct tcphdr)) 2887 return (0); 2888 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af)) 2889 return (0); 2890 opt = hdr + sizeof(struct tcphdr); 2891 hlen -= sizeof(struct tcphdr); 2892 while (hlen >= TCPOLEN_MAXSEG) { 2893 switch (*opt) { 2894 case TCPOPT_EOL: 2895 case TCPOPT_NOP: 2896 ++opt; 2897 --hlen; 2898 break; 2899 case TCPOPT_MAXSEG: 2900 bcopy((void *)(opt + 2), (void *)&mss, 2); 2901 NTOHS(mss); 2902 /* FALLTHROUGH */ 2903 default: 2904 optlen = opt[1]; 2905 if (optlen < 2) 2906 optlen = 2; 2907 hlen -= optlen; 2908 opt += optlen; 2909 break; 2910 } 2911 } 2912 return (mss); 2913 } 2914 2915 u_int16_t 2916 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer) 2917 { 2918 union { 2919 struct sockaddr dst; 2920 struct sockaddr_in dst4; 2921 struct sockaddr_in6 dst6; 2922 } u; 2923 struct route ro; 2924 struct route *rop = &ro; 2925 struct rtentry *rt; 2926 int hlen; 2927 u_int16_t mss = tcp_mssdflt; 2928 2929 hlen = 0; /* XXXGCC -Wunitialized m68k */ 2930 2931 memset(&ro, 0, sizeof(ro)); 2932 switch (af) { 2933 #ifdef INET 2934 case AF_INET: 2935 hlen = sizeof(struct ip); 2936 sockaddr_in_init(&u.dst4, &addr->v4, 0); 2937 rtcache_setdst(rop, &u.dst); 2938 break; 2939 #endif /* INET */ 2940 #ifdef INET6 2941 case AF_INET6: 2942 hlen = sizeof(struct ip6_hdr); 2943 sockaddr_in6_init(&u.dst6, &addr->v6, 0, 0, 0); 2944 rtcache_setdst(rop, &u.dst); 2945 break; 2946 #endif /* INET6 */ 2947 } 2948 2949 #ifndef __NetBSD__ 2950 rtalloc_noclone(rop, NO_CLONING); 2951 if ((rt = ro->ro_rt) != NULL) { 2952 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr); 2953 mss = max(tcp_mssdflt, mss); 2954 } 2955 #else 2956 if ((rt = rtcache_init_noclone(rop)) != NULL) { 2957 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr); 2958 mss = max(tcp_mssdflt, mss); 2959 } 2960 rtcache_free(rop); 2961 #endif 2962 mss = min(mss, offer); 2963 mss = max(mss, 64); /* sanity - at least max opt space */ 2964 return (mss); 2965 } 2966 2967 void 2968 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr) 2969 { 2970 struct pf_rule *r = s->rule.ptr; 2971 2972 s->rt_kif = NULL; 2973 if (!r->rt || r->rt == PF_FASTROUTE) 2974 return; 2975 switch (s->state_key->af) { 2976 #ifdef INET 2977 case AF_INET: 2978 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL, 2979 &s->nat_src_node); 2980 s->rt_kif = r->rpool.cur->kif; 2981 break; 2982 #endif /* INET */ 2983 #ifdef INET6 2984 case AF_INET6: 2985 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL, 2986 &s->nat_src_node); 2987 s->rt_kif = r->rpool.cur->kif; 2988 break; 2989 #endif /* INET6 */ 2990 } 2991 } 2992 2993 void 2994 pf_attach_state(struct pf_state_key *sk, struct pf_state *s, int tail) 2995 { 2996 s->state_key = sk; 2997 sk->refcnt++; 2998 2999 /* list is sorted, if-bound states before floating */ 3000 if (tail) 3001 TAILQ_INSERT_TAIL(&sk->states, s, next); 3002 else 3003 TAILQ_INSERT_HEAD(&sk->states, s, next); 3004 } 3005 3006 void 3007 pf_detach_state(struct pf_state *s, int flags) 3008 { 3009 struct pf_state_key *sk = s->state_key; 3010 3011 if (sk == NULL) 3012 return; 3013 3014 s->state_key = NULL; 3015 TAILQ_REMOVE(&sk->states, s, next); 3016 if (--sk->refcnt == 0) { 3017 if (!(flags & PF_DT_SKIP_EXTGWY)) 3018 RB_REMOVE(pf_state_tree_ext_gwy, 3019 &pf_statetbl_ext_gwy, sk); 3020 if (!(flags & PF_DT_SKIP_LANEXT)) 3021 RB_REMOVE(pf_state_tree_lan_ext, 3022 &pf_statetbl_lan_ext, sk); 3023 pool_put(&pf_state_key_pl, sk); 3024 } 3025 } 3026 3027 struct pf_state_key * 3028 pf_alloc_state_key(struct pf_state *s) 3029 { 3030 struct pf_state_key *sk; 3031 3032 if ((sk = pool_get(&pf_state_key_pl, PR_NOWAIT)) == NULL) 3033 return (NULL); 3034 bzero(sk, sizeof(*sk)); 3035 TAILQ_INIT(&sk->states); 3036 pf_attach_state(sk, s, 0); 3037 3038 return (sk); 3039 } 3040 3041 int 3042 pf_test_rule(struct pf_rule **rm, struct pf_state **sm, int direction, 3043 struct pfi_kif *kif, struct mbuf *m, int off, void *h, 3044 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3045 struct ifqueue *ifq) 3046 { 3047 struct pf_rule *nr = NULL; 3048 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 3049 u_int16_t bport, nport = 0; 3050 sa_family_t af = pd->af; 3051 struct pf_rule *r, *a = NULL; 3052 struct pf_ruleset *ruleset = NULL; 3053 struct pf_src_node *nsn = NULL; 3054 struct tcphdr *th = pd->hdr.tcp; 3055 u_short reason; 3056 int rewrite = 0, hdrlen = 0; 3057 int tag = -1, rtableid = -1; 3058 int asd = 0; 3059 int match = 0; 3060 int state_icmp = 0; 3061 u_int16_t mss = tcp_mssdflt; 3062 u_int16_t sport, dport; 3063 u_int8_t icmptype = 0, icmpcode = 0; 3064 3065 if (direction == PF_IN && pf_check_congestion(ifq)) { 3066 REASON_SET(&reason, PFRES_CONGEST); 3067 return (PF_DROP); 3068 } 3069 3070 sport = dport = hdrlen = 0; 3071 3072 switch (pd->proto) { 3073 case IPPROTO_TCP: 3074 sport = th->th_sport; 3075 dport = th->th_dport; 3076 hdrlen = sizeof(*th); 3077 break; 3078 case IPPROTO_UDP: 3079 sport = pd->hdr.udp->uh_sport; 3080 dport = pd->hdr.udp->uh_dport; 3081 hdrlen = sizeof(*pd->hdr.udp); 3082 break; 3083 #ifdef INET 3084 case IPPROTO_ICMP: 3085 if (pd->af != AF_INET) 3086 break; 3087 sport = dport = pd->hdr.icmp->icmp_id; 3088 icmptype = pd->hdr.icmp->icmp_type; 3089 icmpcode = pd->hdr.icmp->icmp_code; 3090 3091 if (icmptype == ICMP_UNREACH || 3092 icmptype == ICMP_SOURCEQUENCH || 3093 icmptype == ICMP_REDIRECT || 3094 icmptype == ICMP_TIMXCEED || 3095 icmptype == ICMP_PARAMPROB) 3096 state_icmp++; 3097 break; 3098 #endif /* INET */ 3099 3100 #ifdef INET6 3101 case IPPROTO_ICMPV6: 3102 if (pd->af != AF_INET6) 3103 break; 3104 sport = dport = pd->hdr.icmp6->icmp6_id; 3105 hdrlen = sizeof(*pd->hdr.icmp6); 3106 icmptype = pd->hdr.icmp6->icmp6_type; 3107 icmpcode = pd->hdr.icmp6->icmp6_code; 3108 3109 if (icmptype == ICMP6_DST_UNREACH || 3110 icmptype == ICMP6_PACKET_TOO_BIG || 3111 icmptype == ICMP6_TIME_EXCEEDED || 3112 icmptype == ICMP6_PARAM_PROB) 3113 state_icmp++; 3114 break; 3115 #endif /* INET6 */ 3116 } 3117 3118 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 3119 3120 if (direction == PF_OUT) { 3121 bport = nport = sport; 3122 /* check outgoing packet for BINAT/NAT */ 3123 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, 3124 saddr, sport, daddr, dport, &pd->naddr, &nport)) != NULL) { 3125 PF_ACPY(&pd->baddr, saddr, af); 3126 switch (pd->proto) { 3127 case IPPROTO_TCP: 3128 pf_change_ap(saddr, &th->th_sport, pd->ip_sum, 3129 &th->th_sum, &pd->naddr, nport, 0, af); 3130 sport = th->th_sport; 3131 rewrite++; 3132 break; 3133 case IPPROTO_UDP: 3134 pf_change_ap(saddr, &pd->hdr.udp->uh_sport, 3135 pd->ip_sum, &pd->hdr.udp->uh_sum, 3136 &pd->naddr, nport, 1, af); 3137 sport = pd->hdr.udp->uh_sport; 3138 rewrite++; 3139 break; 3140 #ifdef INET 3141 case IPPROTO_ICMP: 3142 pf_change_a(&saddr->v4.s_addr, pd->ip_sum, 3143 pd->naddr.v4.s_addr, 0); 3144 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup( 3145 pd->hdr.icmp->icmp_cksum, sport, nport, 0); 3146 pd->hdr.icmp->icmp_id = nport; 3147 m_copyback(m, off, ICMP_MINLEN, pd->hdr.icmp); 3148 break; 3149 #endif /* INET */ 3150 #ifdef INET6 3151 case IPPROTO_ICMPV6: 3152 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum, 3153 &pd->naddr, 0); 3154 rewrite++; 3155 break; 3156 #endif /* INET */ 3157 default: 3158 switch (af) { 3159 #ifdef INET 3160 case AF_INET: 3161 pf_change_a(&saddr->v4.s_addr, 3162 pd->ip_sum, pd->naddr.v4.s_addr, 0); 3163 break; 3164 #endif /* INET */ 3165 #ifdef INET6 3166 case AF_INET6: 3167 PF_ACPY(saddr, &pd->naddr, af); 3168 break; 3169 #endif /* INET */ 3170 } 3171 break; 3172 } 3173 3174 if (nr->natpass) 3175 r = NULL; 3176 pd->nat_rule = nr; 3177 } 3178 } else { 3179 bport = nport = dport; 3180 /* check incoming packet for BINAT/RDR */ 3181 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, 3182 saddr, sport, daddr, dport, &pd->naddr, &nport)) != NULL) { 3183 PF_ACPY(&pd->baddr, daddr, af); 3184 switch (pd->proto) { 3185 case IPPROTO_TCP: 3186 pf_change_ap(daddr, &th->th_dport, pd->ip_sum, 3187 &th->th_sum, &pd->naddr, nport, 0, af); 3188 dport = th->th_dport; 3189 rewrite++; 3190 break; 3191 case IPPROTO_UDP: 3192 pf_change_ap(daddr, &pd->hdr.udp->uh_dport, 3193 pd->ip_sum, &pd->hdr.udp->uh_sum, 3194 &pd->naddr, nport, 1, af); 3195 dport = pd->hdr.udp->uh_dport; 3196 rewrite++; 3197 break; 3198 #ifdef INET 3199 case IPPROTO_ICMP: 3200 pf_change_a(&daddr->v4.s_addr, pd->ip_sum, 3201 pd->naddr.v4.s_addr, 0); 3202 break; 3203 #endif /* INET */ 3204 #ifdef INET6 3205 case IPPROTO_ICMPV6: 3206 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum, 3207 &pd->naddr, 0); 3208 rewrite++; 3209 break; 3210 #endif /* INET6 */ 3211 default: 3212 switch (af) { 3213 #ifdef INET 3214 case AF_INET: 3215 pf_change_a(&daddr->v4.s_addr, 3216 pd->ip_sum, pd->naddr.v4.s_addr, 0); 3217 break; 3218 #endif /* INET */ 3219 #ifdef INET6 3220 case AF_INET6: 3221 PF_ACPY(daddr, &pd->naddr, af); 3222 break; 3223 #endif /* INET */ 3224 } 3225 break; 3226 } 3227 3228 if (nr->natpass) 3229 r = NULL; 3230 pd->nat_rule = nr; 3231 } 3232 } 3233 3234 while (r != NULL) { 3235 r->evaluations++; 3236 if (pfi_kif_match(r->kif, kif) == r->ifnot) 3237 r = r->skip[PF_SKIP_IFP].ptr; 3238 else if (r->direction && r->direction != direction) 3239 r = r->skip[PF_SKIP_DIR].ptr; 3240 else if (r->af && r->af != af) 3241 r = r->skip[PF_SKIP_AF].ptr; 3242 else if (r->proto && r->proto != pd->proto) 3243 r = r->skip[PF_SKIP_PROTO].ptr; 3244 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, 3245 r->src.neg, kif)) 3246 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 3247 /* tcp/udp only. port_op always 0 in other cases */ 3248 else if (r->src.port_op && !pf_match_port(r->src.port_op, 3249 r->src.port[0], r->src.port[1], sport)) 3250 r = r->skip[PF_SKIP_SRC_PORT].ptr; 3251 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, 3252 r->dst.neg, NULL)) 3253 r = r->skip[PF_SKIP_DST_ADDR].ptr; 3254 /* tcp/udp only. port_op always 0 in other cases */ 3255 else if (r->dst.port_op && !pf_match_port(r->dst.port_op, 3256 r->dst.port[0], r->dst.port[1], dport)) 3257 r = r->skip[PF_SKIP_DST_PORT].ptr; 3258 /* icmp only. type always 0 in other cases */ 3259 else if (r->type && r->type != icmptype + 1) 3260 r = TAILQ_NEXT(r, entries); 3261 /* icmp only. type always 0 in other cases */ 3262 else if (r->code && r->code != icmpcode + 1) 3263 r = TAILQ_NEXT(r, entries); 3264 else if (r->tos && !(r->tos == pd->tos)) 3265 r = TAILQ_NEXT(r, entries); 3266 else if (r->rule_flag & PFRULE_FRAGMENT) 3267 r = TAILQ_NEXT(r, entries); 3268 else if (pd->proto == IPPROTO_TCP && 3269 (r->flagset & th->th_flags) != r->flags) 3270 r = TAILQ_NEXT(r, entries); 3271 /* tcp/udp only. uid.op always 0 in other cases */ 3272 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done = 3273 pf_socket_lookup(direction, pd), 1)) && 3274 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1], 3275 pd->lookup.uid)) 3276 r = TAILQ_NEXT(r, entries); 3277 /* tcp/udp only. gid.op always 0 in other cases */ 3278 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done = 3279 pf_socket_lookup(direction, pd), 1)) && 3280 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1], 3281 pd->lookup.gid)) 3282 r = TAILQ_NEXT(r, entries); 3283 else if (r->prob && r->prob <= arc4random()) 3284 r = TAILQ_NEXT(r, entries); 3285 else if (r->match_tag && !pf_match_tag(m, r, &tag)) 3286 r = TAILQ_NEXT(r, entries); 3287 else if (r->os_fingerprint != PF_OSFP_ANY && 3288 (pd->proto != IPPROTO_TCP || !pf_osfp_match( 3289 pf_osfp_fingerprint(pd, m, off, th), 3290 r->os_fingerprint))) 3291 r = TAILQ_NEXT(r, entries); 3292 else { 3293 if (r->tag) 3294 tag = r->tag; 3295 if (r->rtableid >= 0) 3296 rtableid = r->rtableid; 3297 if (r->anchor == NULL) { 3298 match = 1; 3299 *rm = r; 3300 *am = a; 3301 *rsm = ruleset; 3302 if ((*rm)->quick) 3303 break; 3304 r = TAILQ_NEXT(r, entries); 3305 } else 3306 pf_step_into_anchor(&asd, &ruleset, 3307 PF_RULESET_FILTER, &r, &a, &match); 3308 } 3309 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 3310 PF_RULESET_FILTER, &r, &a, &match)) 3311 break; 3312 } 3313 r = *rm; 3314 a = *am; 3315 ruleset = *rsm; 3316 3317 REASON_SET(&reason, PFRES_MATCH); 3318 3319 if (r->log || (nr != NULL && nr->log)) { 3320 if (rewrite) 3321 m_copyback(m, off, hdrlen, pd->hdr.any); 3322 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, 3323 a, ruleset, pd); 3324 } 3325 3326 if ((r->action == PF_DROP) && 3327 ((r->rule_flag & PFRULE_RETURNRST) || 3328 (r->rule_flag & PFRULE_RETURNICMP) || 3329 (r->rule_flag & PFRULE_RETURN))) { 3330 /* undo NAT changes, if they have taken place */ 3331 if (nr != NULL) { 3332 if (direction == PF_OUT) { 3333 switch (pd->proto) { 3334 case IPPROTO_TCP: 3335 pf_change_ap(saddr, &th->th_sport, 3336 pd->ip_sum, &th->th_sum, 3337 &pd->baddr, bport, 0, af); 3338 sport = th->th_sport; 3339 rewrite++; 3340 break; 3341 case IPPROTO_UDP: 3342 pf_change_ap(saddr, 3343 &pd->hdr.udp->uh_sport, pd->ip_sum, 3344 &pd->hdr.udp->uh_sum, &pd->baddr, 3345 bport, 1, af); 3346 sport = pd->hdr.udp->uh_sport; 3347 rewrite++; 3348 break; 3349 case IPPROTO_ICMP: 3350 #ifdef INET6 3351 case IPPROTO_ICMPV6: 3352 #endif 3353 /* nothing! */ 3354 break; 3355 default: 3356 switch (af) { 3357 case AF_INET: 3358 pf_change_a(&saddr->v4.s_addr, 3359 pd->ip_sum, 3360 pd->baddr.v4.s_addr, 0); 3361 break; 3362 case AF_INET6: 3363 PF_ACPY(saddr, &pd->baddr, af); 3364 break; 3365 } 3366 } 3367 } else { 3368 switch (pd->proto) { 3369 case IPPROTO_TCP: 3370 pf_change_ap(daddr, &th->th_dport, 3371 pd->ip_sum, &th->th_sum, 3372 &pd->baddr, bport, 0, af); 3373 dport = th->th_dport; 3374 rewrite++; 3375 break; 3376 case IPPROTO_UDP: 3377 pf_change_ap(daddr, 3378 &pd->hdr.udp->uh_dport, pd->ip_sum, 3379 &pd->hdr.udp->uh_sum, &pd->baddr, 3380 bport, 1, af); 3381 dport = pd->hdr.udp->uh_dport; 3382 rewrite++; 3383 break; 3384 case IPPROTO_ICMP: 3385 #ifdef INET6 3386 case IPPROTO_ICMPV6: 3387 #endif 3388 /* nothing! */ 3389 break; 3390 default: 3391 switch (af) { 3392 case AF_INET: 3393 pf_change_a(&daddr->v4.s_addr, 3394 pd->ip_sum, 3395 pd->baddr.v4.s_addr, 0); 3396 break; 3397 case AF_INET6: 3398 PF_ACPY(daddr, &pd->baddr, af); 3399 break; 3400 } 3401 } 3402 } 3403 } 3404 if (pd->proto == IPPROTO_TCP && 3405 ((r->rule_flag & PFRULE_RETURNRST) || 3406 (r->rule_flag & PFRULE_RETURN)) && 3407 !(th->th_flags & TH_RST)) { 3408 u_int32_t ack = ntohl(th->th_seq) + pd->p_len; 3409 struct ip *h = mtod(m, struct ip *); 3410 3411 #ifdef __NetBSD__ 3412 if (pf_check_proto_cksum(m, direction, off, 3413 ntohs(h->ip_len) - off, IPPROTO_TCP, AF_INET)) 3414 #else 3415 if (pf_check_proto_cksum(m, off, 3416 ntohs(h->ip_len) - off, IPPROTO_TCP, AF_INET)) 3417 #endif /* !__NetBSD__ */ 3418 REASON_SET(&reason, PFRES_PROTCKSUM); 3419 else { 3420 if (th->th_flags & TH_SYN) 3421 ack++; 3422 if (th->th_flags & TH_FIN) 3423 ack++; 3424 pf_send_tcp(r, af, pd->dst, 3425 pd->src, th->th_dport, th->th_sport, 3426 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0, 3427 r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp); 3428 } 3429 } else if ((af == AF_INET) && r->return_icmp) 3430 pf_send_icmp(m, r->return_icmp >> 8, 3431 r->return_icmp & 255, af, r); 3432 else if ((af == AF_INET6) && r->return_icmp6) 3433 pf_send_icmp(m, r->return_icmp6 >> 8, 3434 r->return_icmp6 & 255, af, r); 3435 } 3436 3437 if (r->action == PF_DROP) 3438 return (PF_DROP); 3439 3440 if (pf_tag_packet(m, tag, rtableid)) { 3441 REASON_SET(&reason, PFRES_MEMORY); 3442 return (PF_DROP); 3443 } 3444 3445 if (!state_icmp && (r->keep_state || nr != NULL || 3446 (pd->flags & PFDESC_TCP_NORM))) { 3447 /* create new state */ 3448 u_int16_t len; 3449 struct pf_state *s = NULL; 3450 struct pf_state_key *sk = NULL; 3451 struct pf_src_node *sn = NULL; 3452 3453 /* check maximums */ 3454 if (r->max_states && (r->states >= r->max_states)) { 3455 pf_status.lcounters[LCNT_STATES]++; 3456 REASON_SET(&reason, PFRES_MAXSTATES); 3457 goto cleanup; 3458 } 3459 /* src node for filter rule */ 3460 if ((r->rule_flag & PFRULE_SRCTRACK || 3461 r->rpool.opts & PF_POOL_STICKYADDR) && 3462 pf_insert_src_node(&sn, r, saddr, af) != 0) { 3463 REASON_SET(&reason, PFRES_SRCLIMIT); 3464 goto cleanup; 3465 } 3466 /* src node for translation rule */ 3467 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 3468 ((direction == PF_OUT && 3469 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || 3470 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { 3471 REASON_SET(&reason, PFRES_SRCLIMIT); 3472 goto cleanup; 3473 } 3474 s = pool_get(&pf_state_pl, PR_NOWAIT); 3475 if (s == NULL) { 3476 REASON_SET(&reason, PFRES_MEMORY); 3477 cleanup: 3478 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 3479 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 3480 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3481 pf_status.src_nodes--; 3482 pool_put(&pf_src_tree_pl, sn); 3483 } 3484 if (nsn != sn && nsn != NULL && nsn->states == 0 && 3485 nsn->expire == 0) { 3486 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 3487 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3488 pf_status.src_nodes--; 3489 pool_put(&pf_src_tree_pl, nsn); 3490 } 3491 if (sk != NULL) { 3492 pool_put(&pf_state_key_pl, sk); 3493 } 3494 return (PF_DROP); 3495 } 3496 bzero(s, sizeof(*s)); 3497 s->rule.ptr = r; 3498 s->nat_rule.ptr = nr; 3499 s->anchor.ptr = a; 3500 STATE_INC_COUNTERS(s); 3501 s->allow_opts = r->allow_opts; 3502 s->log = r->log & PF_LOG_ALL; 3503 if (nr != NULL) 3504 s->log |= nr->log & PF_LOG_ALL; 3505 switch (pd->proto) { 3506 case IPPROTO_TCP: 3507 len = pd->tot_len - off - (th->th_off << 2); 3508 s->src.seqlo = ntohl(th->th_seq); 3509 s->src.seqhi = s->src.seqlo + len + 1; 3510 if ((th->th_flags & (TH_SYN|TH_ACK)) == 3511 TH_SYN && r->keep_state == PF_STATE_MODULATE) { 3512 /* Generate sequence number modulator */ 3513 while ((s->src.seqdiff = 3514 tcp_rndiss_next() - s->src.seqlo) == 0) 3515 ; 3516 pf_change_a(&th->th_seq, &th->th_sum, 3517 htonl(s->src.seqlo + s->src.seqdiff), 0); 3518 rewrite = 1; 3519 } else 3520 s->src.seqdiff = 0; 3521 if (th->th_flags & TH_SYN) { 3522 s->src.seqhi++; 3523 s->src.wscale = pf_get_wscale(m, off, 3524 th->th_off, af); 3525 } 3526 s->src.max_win = MAX(ntohs(th->th_win), 1); 3527 if (s->src.wscale & PF_WSCALE_MASK) { 3528 /* Remove scale factor from initial window */ 3529 int win = s->src.max_win; 3530 win += 1 << (s->src.wscale & PF_WSCALE_MASK); 3531 s->src.max_win = (win - 1) >> 3532 (s->src.wscale & PF_WSCALE_MASK); 3533 } 3534 if (th->th_flags & TH_FIN) 3535 s->src.seqhi++; 3536 s->dst.seqhi = 1; 3537 s->dst.max_win = 1; 3538 s->src.state = TCPS_SYN_SENT; 3539 s->dst.state = TCPS_CLOSED; 3540 s->timeout = PFTM_TCP_FIRST_PACKET; 3541 break; 3542 case IPPROTO_UDP: 3543 s->src.state = PFUDPS_SINGLE; 3544 s->dst.state = PFUDPS_NO_TRAFFIC; 3545 s->timeout = PFTM_UDP_FIRST_PACKET; 3546 break; 3547 case IPPROTO_ICMP: 3548 #ifdef INET6 3549 case IPPROTO_ICMPV6: 3550 #endif 3551 s->timeout = PFTM_ICMP_FIRST_PACKET; 3552 break; 3553 default: 3554 s->src.state = PFOTHERS_SINGLE; 3555 s->dst.state = PFOTHERS_NO_TRAFFIC; 3556 s->timeout = PFTM_OTHER_FIRST_PACKET; 3557 } 3558 3559 s->creation = time_second; 3560 s->expire = time_second; 3561 3562 if (sn != NULL) { 3563 s->src_node = sn; 3564 s->src_node->states++; 3565 } 3566 if (nsn != NULL) { 3567 PF_ACPY(&nsn->raddr, &pd->naddr, af); 3568 s->nat_src_node = nsn; 3569 s->nat_src_node->states++; 3570 } 3571 if (pd->proto == IPPROTO_TCP) { 3572 if ((pd->flags & PFDESC_TCP_NORM) && 3573 pf_normalize_tcp_init(m, off, pd, th, &s->src, 3574 &s->dst)) { 3575 REASON_SET(&reason, PFRES_MEMORY); 3576 pf_src_tree_remove_state(s); 3577 STATE_DEC_COUNTERS(s); 3578 pool_put(&pf_state_pl, s); 3579 return (PF_DROP); 3580 } 3581 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub && 3582 pf_normalize_tcp_stateful(m, off, pd, &reason, 3583 th, s, &s->src, &s->dst, &rewrite)) { 3584 /* This really shouldn't happen!!! */ 3585 DPFPRINTF(PF_DEBUG_URGENT, 3586 ("pf_normalize_tcp_stateful failed on " 3587 "first pkt")); 3588 pf_normalize_tcp_cleanup(s); 3589 pf_src_tree_remove_state(s); 3590 STATE_DEC_COUNTERS(s); 3591 pool_put(&pf_state_pl, s); 3592 return (PF_DROP); 3593 } 3594 } 3595 3596 if ((sk = pf_alloc_state_key(s)) == NULL) { 3597 REASON_SET(&reason, PFRES_MEMORY); 3598 goto cleanup; 3599 } 3600 3601 sk->proto = pd->proto; 3602 sk->direction = direction; 3603 sk->af = af; 3604 if (direction == PF_OUT) { 3605 PF_ACPY(&sk->gwy.addr, saddr, af); 3606 PF_ACPY(&sk->ext.addr, daddr, af); 3607 switch (pd->proto) { 3608 case IPPROTO_ICMP: 3609 #ifdef INET6 3610 case IPPROTO_ICMPV6: 3611 #endif 3612 sk->gwy.port = nport; 3613 sk->ext.port = 0; 3614 break; 3615 default: 3616 sk->gwy.port = sport; 3617 sk->ext.port = dport; 3618 } 3619 if (nr != NULL) { 3620 PF_ACPY(&sk->lan.addr, &pd->baddr, af); 3621 sk->lan.port = bport; 3622 } else { 3623 PF_ACPY(&sk->lan.addr, &sk->gwy.addr, af); 3624 sk->lan.port = sk->gwy.port; 3625 } 3626 } else { 3627 PF_ACPY(&sk->lan.addr, daddr, af); 3628 PF_ACPY(&sk->ext.addr, saddr, af); 3629 switch (pd->proto) { 3630 case IPPROTO_ICMP: 3631 #ifdef INET6 3632 case IPPROTO_ICMPV6: 3633 #endif 3634 sk->lan.port = nport; 3635 sk->ext.port = 0; 3636 break; 3637 default: 3638 sk->lan.port = dport; 3639 sk->ext.port = sport; 3640 } 3641 if (nr != NULL) { 3642 PF_ACPY(&sk->gwy.addr, &pd->baddr, af); 3643 sk->gwy.port = bport; 3644 } else { 3645 PF_ACPY(&sk->gwy.addr, &sk->lan.addr, af); 3646 sk->gwy.port = sk->lan.port; 3647 } 3648 } 3649 3650 pf_set_rt_ifp(s, saddr); /* needs s->state_key set */ 3651 3652 if (pf_insert_state(bound_iface(r, nr, kif), s)) { 3653 if (pd->proto == IPPROTO_TCP) 3654 pf_normalize_tcp_cleanup(s); 3655 REASON_SET(&reason, PFRES_STATEINS); 3656 pf_src_tree_remove_state(s); 3657 STATE_DEC_COUNTERS(s); 3658 pool_put(&pf_state_pl, s); 3659 return (PF_DROP); 3660 } else 3661 *sm = s; 3662 if (tag > 0) { 3663 pf_tag_ref(tag); 3664 s->tag = tag; 3665 } 3666 if (pd->proto == IPPROTO_TCP && 3667 (th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN && 3668 r->keep_state == PF_STATE_SYNPROXY) { 3669 s->src.state = PF_TCPS_PROXY_SRC; 3670 if (nr != NULL) { 3671 if (direction == PF_OUT) { 3672 pf_change_ap(saddr, &th->th_sport, 3673 pd->ip_sum, &th->th_sum, &pd->baddr, 3674 bport, 0, af); 3675 sport = th->th_sport; 3676 } else { 3677 pf_change_ap(daddr, &th->th_dport, 3678 pd->ip_sum, &th->th_sum, &pd->baddr, 3679 bport, 0, af); 3680 sport = th->th_dport; 3681 } 3682 } 3683 s->src.seqhi = htonl(arc4random()); 3684 /* Find mss option */ 3685 mss = pf_get_mss(m, off, th->th_off, af); 3686 mss = pf_calc_mss(saddr, af, mss); 3687 mss = pf_calc_mss(daddr, af, mss); 3688 s->src.mss = mss; 3689 pf_send_tcp(r, af, daddr, saddr, th->th_dport, 3690 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1, 3691 TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL); 3692 REASON_SET(&reason, PFRES_SYNPROXY); 3693 return (PF_SYNPROXY_DROP); 3694 } 3695 } 3696 3697 /* copy back packet headers if we performed NAT operations */ 3698 if (rewrite) 3699 m_copyback(m, off, hdrlen, pd->hdr.any); 3700 3701 return (PF_PASS); 3702 } 3703 3704 int 3705 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif, 3706 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am, 3707 struct pf_ruleset **rsm) 3708 { 3709 struct pf_rule *r, *a = NULL; 3710 struct pf_ruleset *ruleset = NULL; 3711 sa_family_t af = pd->af; 3712 u_short reason; 3713 int tag = -1; 3714 int asd = 0; 3715 int match = 0; 3716 3717 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 3718 while (r != NULL) { 3719 r->evaluations++; 3720 if (pfi_kif_match(r->kif, kif) == r->ifnot) 3721 r = r->skip[PF_SKIP_IFP].ptr; 3722 else if (r->direction && r->direction != direction) 3723 r = r->skip[PF_SKIP_DIR].ptr; 3724 else if (r->af && r->af != af) 3725 r = r->skip[PF_SKIP_AF].ptr; 3726 else if (r->proto && r->proto != pd->proto) 3727 r = r->skip[PF_SKIP_PROTO].ptr; 3728 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, 3729 r->src.neg, kif)) 3730 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 3731 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, 3732 r->dst.neg, NULL)) 3733 r = r->skip[PF_SKIP_DST_ADDR].ptr; 3734 else if (r->tos && !(r->tos == pd->tos)) 3735 r = TAILQ_NEXT(r, entries); 3736 else if (r->src.port_op || r->dst.port_op || 3737 r->flagset || r->type || r->code || 3738 r->os_fingerprint != PF_OSFP_ANY) 3739 r = TAILQ_NEXT(r, entries); 3740 else if (r->prob && r->prob <= arc4random()) 3741 r = TAILQ_NEXT(r, entries); 3742 else if (r->match_tag && !pf_match_tag(m, r, &tag)) 3743 r = TAILQ_NEXT(r, entries); 3744 else { 3745 if (r->anchor == NULL) { 3746 match = 1; 3747 *rm = r; 3748 *am = a; 3749 *rsm = ruleset; 3750 if ((*rm)->quick) 3751 break; 3752 r = TAILQ_NEXT(r, entries); 3753 } else 3754 pf_step_into_anchor(&asd, &ruleset, 3755 PF_RULESET_FILTER, &r, &a, &match); 3756 } 3757 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 3758 PF_RULESET_FILTER, &r, &a, &match)) 3759 break; 3760 } 3761 r = *rm; 3762 a = *am; 3763 ruleset = *rsm; 3764 3765 REASON_SET(&reason, PFRES_MATCH); 3766 3767 if (r->log) 3768 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset, 3769 pd); 3770 3771 if (r->action != PF_PASS) 3772 return (PF_DROP); 3773 3774 if (pf_tag_packet(m, tag, -1)) { 3775 REASON_SET(&reason, PFRES_MEMORY); 3776 return (PF_DROP); 3777 } 3778 3779 return (PF_PASS); 3780 } 3781 3782 int 3783 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif, 3784 struct mbuf *m, int off, void *h, struct pf_pdesc *pd, 3785 u_short *reason) 3786 { 3787 struct pf_state_key_cmp key; 3788 struct tcphdr *th = pd->hdr.tcp; 3789 u_int16_t win = ntohs(th->th_win); 3790 u_int32_t ack, end, seq, orig_seq; 3791 u_int8_t sws, dws; 3792 int ackskew; 3793 int copyback = 0; 3794 struct pf_state_peer *src, *dst; 3795 3796 key.af = pd->af; 3797 key.proto = IPPROTO_TCP; 3798 if (direction == PF_IN) { 3799 PF_ACPY(&key.ext.addr, pd->src, key.af); 3800 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 3801 key.ext.port = th->th_sport; 3802 key.gwy.port = th->th_dport; 3803 } else { 3804 PF_ACPY(&key.lan.addr, pd->src, key.af); 3805 PF_ACPY(&key.ext.addr, pd->dst, key.af); 3806 key.lan.port = th->th_sport; 3807 key.ext.port = th->th_dport; 3808 } 3809 3810 STATE_LOOKUP(); 3811 3812 if (direction == (*state)->state_key->direction) { 3813 src = &(*state)->src; 3814 dst = &(*state)->dst; 3815 } else { 3816 src = &(*state)->dst; 3817 dst = &(*state)->src; 3818 } 3819 3820 if ((*state)->src.state == PF_TCPS_PROXY_SRC) { 3821 if (direction != (*state)->state_key->direction) { 3822 REASON_SET(reason, PFRES_SYNPROXY); 3823 return (PF_SYNPROXY_DROP); 3824 } 3825 if (th->th_flags & TH_SYN) { 3826 if (ntohl(th->th_seq) != (*state)->src.seqlo) { 3827 REASON_SET(reason, PFRES_SYNPROXY); 3828 return (PF_DROP); 3829 } 3830 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, 3831 pd->src, th->th_dport, th->th_sport, 3832 (*state)->src.seqhi, ntohl(th->th_seq) + 1, 3833 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1, 3834 0, NULL, NULL); 3835 REASON_SET(reason, PFRES_SYNPROXY); 3836 return (PF_SYNPROXY_DROP); 3837 } else if (!(th->th_flags & TH_ACK) || 3838 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) || 3839 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) { 3840 REASON_SET(reason, PFRES_SYNPROXY); 3841 return (PF_DROP); 3842 } else if ((*state)->src_node != NULL && 3843 pf_src_connlimit(state)) { 3844 REASON_SET(reason, PFRES_SRCLIMIT); 3845 return (PF_DROP); 3846 } else 3847 (*state)->src.state = PF_TCPS_PROXY_DST; 3848 } 3849 if ((*state)->src.state == PF_TCPS_PROXY_DST) { 3850 struct pf_state_host *src, *dst; 3851 3852 if (direction == PF_OUT) { 3853 src = &(*state)->state_key->gwy; 3854 dst = &(*state)->state_key->ext; 3855 } else { 3856 src = &(*state)->state_key->ext; 3857 dst = &(*state)->state_key->lan; 3858 } 3859 if (direction == (*state)->state_key->direction) { 3860 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) || 3861 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) || 3862 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) { 3863 REASON_SET(reason, PFRES_SYNPROXY); 3864 return (PF_DROP); 3865 } 3866 (*state)->src.max_win = MAX(ntohs(th->th_win), 1); 3867 if ((*state)->dst.seqhi == 1) 3868 (*state)->dst.seqhi = htonl(arc4random()); 3869 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr, 3870 &dst->addr, src->port, dst->port, 3871 (*state)->dst.seqhi, 0, TH_SYN, 0, 3872 (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL); 3873 REASON_SET(reason, PFRES_SYNPROXY); 3874 return (PF_SYNPROXY_DROP); 3875 } else if (((th->th_flags & (TH_SYN|TH_ACK)) != 3876 (TH_SYN|TH_ACK)) || 3877 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) { 3878 REASON_SET(reason, PFRES_SYNPROXY); 3879 return (PF_DROP); 3880 } else { 3881 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1); 3882 (*state)->dst.seqlo = ntohl(th->th_seq); 3883 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, 3884 pd->src, th->th_dport, th->th_sport, 3885 ntohl(th->th_ack), ntohl(th->th_seq) + 1, 3886 TH_ACK, (*state)->src.max_win, 0, 0, 0, 3887 (*state)->tag, NULL, NULL); 3888 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr, 3889 &dst->addr, src->port, dst->port, 3890 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1, 3891 TH_ACK, (*state)->dst.max_win, 0, 0, 1, 3892 0, NULL, NULL); 3893 (*state)->src.seqdiff = (*state)->dst.seqhi - 3894 (*state)->src.seqlo; 3895 (*state)->dst.seqdiff = (*state)->src.seqhi - 3896 (*state)->dst.seqlo; 3897 (*state)->src.seqhi = (*state)->src.seqlo + 3898 (*state)->dst.max_win; 3899 (*state)->dst.seqhi = (*state)->dst.seqlo + 3900 (*state)->src.max_win; 3901 (*state)->src.wscale = (*state)->dst.wscale = 0; 3902 (*state)->src.state = (*state)->dst.state = 3903 TCPS_ESTABLISHED; 3904 REASON_SET(reason, PFRES_SYNPROXY); 3905 return (PF_SYNPROXY_DROP); 3906 } 3907 } 3908 3909 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) { 3910 sws = src->wscale & PF_WSCALE_MASK; 3911 dws = dst->wscale & PF_WSCALE_MASK; 3912 } else 3913 sws = dws = 0; 3914 3915 /* 3916 * Sequence tracking algorithm from Guido van Rooij's paper: 3917 * http://www.madison-gurkha.com/publications/tcp_filtering/ 3918 * tcp_filtering.ps 3919 */ 3920 3921 orig_seq = seq = ntohl(th->th_seq); 3922 if (src->seqlo == 0) { 3923 /* First packet from this end. Set its state */ 3924 3925 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) && 3926 src->scrub == NULL) { 3927 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) { 3928 REASON_SET(reason, PFRES_MEMORY); 3929 return (PF_DROP); 3930 } 3931 } 3932 3933 /* Deferred generation of sequence number modulator */ 3934 if (dst->seqdiff && !src->seqdiff) { 3935 while ((src->seqdiff = tcp_rndiss_next() - seq) == 0) 3936 ; 3937 ack = ntohl(th->th_ack) - dst->seqdiff; 3938 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq + 3939 src->seqdiff), 0); 3940 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0); 3941 copyback = 1; 3942 } else { 3943 ack = ntohl(th->th_ack); 3944 } 3945 3946 end = seq + pd->p_len; 3947 if (th->th_flags & TH_SYN) { 3948 end++; 3949 if (dst->wscale & PF_WSCALE_FLAG) { 3950 src->wscale = pf_get_wscale(m, off, th->th_off, 3951 pd->af); 3952 if (src->wscale & PF_WSCALE_FLAG) { 3953 /* Remove scale factor from initial 3954 * window */ 3955 sws = src->wscale & PF_WSCALE_MASK; 3956 win = ((u_int32_t)win + (1 << sws) - 1) 3957 >> sws; 3958 dws = dst->wscale & PF_WSCALE_MASK; 3959 } else { 3960 /* fixup other window */ 3961 dst->max_win <<= dst->wscale & 3962 PF_WSCALE_MASK; 3963 /* in case of a retrans SYN|ACK */ 3964 dst->wscale = 0; 3965 } 3966 } 3967 } 3968 if (th->th_flags & TH_FIN) 3969 end++; 3970 3971 src->seqlo = seq; 3972 if (src->state < TCPS_SYN_SENT) 3973 src->state = TCPS_SYN_SENT; 3974 3975 /* 3976 * May need to slide the window (seqhi may have been set by 3977 * the crappy stack check or if we picked up the connection 3978 * after establishment) 3979 */ 3980 if (src->seqhi == 1 || 3981 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi)) 3982 src->seqhi = end + MAX(1, dst->max_win << dws); 3983 if (win > src->max_win) 3984 src->max_win = win; 3985 3986 } else { 3987 ack = ntohl(th->th_ack) - dst->seqdiff; 3988 if (src->seqdiff) { 3989 /* Modulate sequence numbers */ 3990 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq + 3991 src->seqdiff), 0); 3992 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0); 3993 copyback = 1; 3994 } 3995 end = seq + pd->p_len; 3996 if (th->th_flags & TH_SYN) 3997 end++; 3998 if (th->th_flags & TH_FIN) 3999 end++; 4000 } 4001 4002 if ((th->th_flags & TH_ACK) == 0) { 4003 /* Let it pass through the ack skew check */ 4004 ack = dst->seqlo; 4005 } else if ((ack == 0 && 4006 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) || 4007 /* broken tcp stacks do not set ack */ 4008 (dst->state < TCPS_SYN_SENT)) { 4009 /* 4010 * Many stacks (ours included) will set the ACK number in an 4011 * FIN|ACK if the SYN times out -- no sequence to ACK. 4012 */ 4013 ack = dst->seqlo; 4014 } 4015 4016 if (seq == end) { 4017 /* Ease sequencing restrictions on no data packets */ 4018 seq = src->seqlo; 4019 end = seq; 4020 } 4021 4022 ackskew = dst->seqlo - ack; 4023 4024 4025 /* 4026 * Need to demodulate the sequence numbers in any TCP SACK options 4027 * (Selective ACK). We could optionally validate the SACK values 4028 * against the current ACK window, either forwards or backwards, but 4029 * I'm not confident that SACK has been implemented properly 4030 * everywhere. It wouldn't surprise me if several stacks accidently 4031 * SACK too far backwards of previously ACKed data. There really aren't 4032 * any security implications of bad SACKing unless the target stack 4033 * doesn't validate the option length correctly. Someone trying to 4034 * spoof into a TCP connection won't bother blindly sending SACK 4035 * options anyway. 4036 */ 4037 if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) { 4038 if (pf_modulate_sack(m, off, pd, th, dst)) 4039 copyback = 1; 4040 } 4041 4042 4043 #define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */ 4044 if (SEQ_GEQ(src->seqhi, end) && 4045 /* Last octet inside other's window space */ 4046 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) && 4047 /* Retrans: not more than one window back */ 4048 (ackskew >= -MAXACKWINDOW) && 4049 /* Acking not more than one reassembled fragment backwards */ 4050 (ackskew <= (MAXACKWINDOW << sws)) && 4051 /* Acking not more than one window forward */ 4052 ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo || 4053 (orig_seq == src->seqlo + 1) || (pd->flags & PFDESC_IP_REAS) == 0)) { 4054 /* Require an exact/+1 sequence match on resets when possible */ 4055 4056 if (dst->scrub || src->scrub) { 4057 if (pf_normalize_tcp_stateful(m, off, pd, reason, th, 4058 *state, src, dst, ©back)) 4059 return (PF_DROP); 4060 } 4061 4062 /* update max window */ 4063 if (src->max_win < win) 4064 src->max_win = win; 4065 /* synchronize sequencing */ 4066 if (SEQ_GT(end, src->seqlo)) 4067 src->seqlo = end; 4068 /* slide the window of what the other end can send */ 4069 if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) 4070 dst->seqhi = ack + MAX((win << sws), 1); 4071 4072 4073 /* update states */ 4074 if (th->th_flags & TH_SYN) 4075 if (src->state < TCPS_SYN_SENT) 4076 src->state = TCPS_SYN_SENT; 4077 if (th->th_flags & TH_FIN) 4078 if (src->state < TCPS_CLOSING) 4079 src->state = TCPS_CLOSING; 4080 if (th->th_flags & TH_ACK) { 4081 if (dst->state == TCPS_SYN_SENT) { 4082 dst->state = TCPS_ESTABLISHED; 4083 if (src->state == TCPS_ESTABLISHED && 4084 (*state)->src_node != NULL && 4085 pf_src_connlimit(state)) { 4086 REASON_SET(reason, PFRES_SRCLIMIT); 4087 return (PF_DROP); 4088 } 4089 } else if (dst->state == TCPS_CLOSING) 4090 dst->state = TCPS_FIN_WAIT_2; 4091 } 4092 if (th->th_flags & TH_RST) 4093 src->state = dst->state = TCPS_TIME_WAIT; 4094 4095 /* update expire time */ 4096 (*state)->expire = time_second; 4097 if (src->state >= TCPS_FIN_WAIT_2 && 4098 dst->state >= TCPS_FIN_WAIT_2) 4099 (*state)->timeout = PFTM_TCP_CLOSED; 4100 else if (src->state >= TCPS_CLOSING && 4101 dst->state >= TCPS_CLOSING) 4102 (*state)->timeout = PFTM_TCP_FIN_WAIT; 4103 else if (src->state < TCPS_ESTABLISHED || 4104 dst->state < TCPS_ESTABLISHED) 4105 (*state)->timeout = PFTM_TCP_OPENING; 4106 else if (src->state >= TCPS_CLOSING || 4107 dst->state >= TCPS_CLOSING) 4108 (*state)->timeout = PFTM_TCP_CLOSING; 4109 else 4110 (*state)->timeout = PFTM_TCP_ESTABLISHED; 4111 4112 /* Fall through to PASS packet */ 4113 4114 } else if ((dst->state < TCPS_SYN_SENT || 4115 dst->state >= TCPS_FIN_WAIT_2 || 4116 src->state >= TCPS_FIN_WAIT_2) && 4117 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) && 4118 /* Within a window forward of the originating packet */ 4119 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) { 4120 /* Within a window backward of the originating packet */ 4121 4122 /* 4123 * This currently handles three situations: 4124 * 1) Stupid stacks will shotgun SYNs before their peer 4125 * replies. 4126 * 2) When PF catches an already established stream (the 4127 * firewall rebooted, the state table was flushed, routes 4128 * changed...) 4129 * 3) Packets get funky immediately after the connection 4130 * closes (this should catch Solaris spurious ACK|FINs 4131 * that web servers like to spew after a close) 4132 * 4133 * This must be a little more careful than the above code 4134 * since packet floods will also be caught here. We don't 4135 * update the TTL here to mitigate the damage of a packet 4136 * flood and so the same code can handle awkward establishment 4137 * and a loosened connection close. 4138 * In the establishment case, a correct peer response will 4139 * validate the connection, go through the normal state code 4140 * and keep updating the state TTL. 4141 */ 4142 4143 if (pf_status.debug >= PF_DEBUG_MISC) { 4144 printf("pf: loose state match: "); 4145 pf_print_state(*state); 4146 pf_print_flags(th->th_flags); 4147 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d " 4148 "pkts=%llu:%llu\n", seq, orig_seq, ack, pd->p_len, 4149 ackskew, 4150 (unsigned long long int)(*state)->packets[0], 4151 (unsigned long long int)(*state)->packets[1]); 4152 } 4153 4154 if (dst->scrub || src->scrub) { 4155 if (pf_normalize_tcp_stateful(m, off, pd, reason, th, 4156 *state, src, dst, ©back)) 4157 return (PF_DROP); 4158 } 4159 4160 /* update max window */ 4161 if (src->max_win < win) 4162 src->max_win = win; 4163 /* synchronize sequencing */ 4164 if (SEQ_GT(end, src->seqlo)) 4165 src->seqlo = end; 4166 /* slide the window of what the other end can send */ 4167 if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) 4168 dst->seqhi = ack + MAX((win << sws), 1); 4169 4170 /* 4171 * Cannot set dst->seqhi here since this could be a shotgunned 4172 * SYN and not an already established connection. 4173 */ 4174 4175 if (th->th_flags & TH_FIN) 4176 if (src->state < TCPS_CLOSING) 4177 src->state = TCPS_CLOSING; 4178 if (th->th_flags & TH_RST) 4179 src->state = dst->state = TCPS_TIME_WAIT; 4180 4181 /* Fall through to PASS packet */ 4182 4183 } else { 4184 if ((*state)->dst.state == TCPS_SYN_SENT && 4185 (*state)->src.state == TCPS_SYN_SENT) { 4186 /* Send RST for state mismatches during handshake */ 4187 if (!(th->th_flags & TH_RST)) 4188 pf_send_tcp((*state)->rule.ptr, pd->af, 4189 pd->dst, pd->src, th->th_dport, 4190 th->th_sport, ntohl(th->th_ack), 0, 4191 TH_RST, 0, 0, 4192 (*state)->rule.ptr->return_ttl, 1, 0, 4193 pd->eh, kif->pfik_ifp); 4194 src->seqlo = 0; 4195 src->seqhi = 1; 4196 src->max_win = 1; 4197 } else if (pf_status.debug >= PF_DEBUG_MISC) { 4198 printf("pf: BAD state: "); 4199 pf_print_state(*state); 4200 pf_print_flags(th->th_flags); 4201 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d " 4202 "pkts=%llu:%llu dir=%s,%s\n", 4203 seq, orig_seq, ack, pd->p_len, ackskew, 4204 (unsigned long long int)(*state)->packets[0], 4205 (unsigned long long int)(*state)->packets[1], 4206 direction == PF_IN ? "in" : "out", 4207 direction == (*state)->state_key->direction ? 4208 "fwd" : "rev"); 4209 printf("pf: State failure on: %c %c %c %c | %c %c\n", 4210 SEQ_GEQ(src->seqhi, end) ? ' ' : '1', 4211 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ? 4212 ' ': '2', 4213 (ackskew >= -MAXACKWINDOW) ? ' ' : '3', 4214 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4', 4215 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5', 4216 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6'); 4217 } 4218 REASON_SET(reason, PFRES_BADSTATE); 4219 return (PF_DROP); 4220 } 4221 4222 /* Any packets which have gotten here are to be passed */ 4223 4224 /* translate source/destination address, if necessary */ 4225 if (STATE_TRANSLATE((*state)->state_key)) { 4226 if (direction == PF_OUT) 4227 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum, 4228 &th->th_sum, &(*state)->state_key->gwy.addr, 4229 (*state)->state_key->gwy.port, 0, pd->af); 4230 else 4231 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum, 4232 &th->th_sum, &(*state)->state_key->lan.addr, 4233 (*state)->state_key->lan.port, 0, pd->af); 4234 m_copyback(m, off, sizeof(*th), th); 4235 } else if (copyback) { 4236 /* Copyback sequence modulation or stateful scrub changes */ 4237 m_copyback(m, off, sizeof(*th), th); 4238 } 4239 4240 return (PF_PASS); 4241 } 4242 4243 int 4244 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif, 4245 struct mbuf *m, int off, void *h, struct pf_pdesc *pd) 4246 { 4247 struct pf_state_peer *src, *dst; 4248 struct pf_state_key_cmp key; 4249 struct udphdr *uh = pd->hdr.udp; 4250 4251 key.af = pd->af; 4252 key.proto = IPPROTO_UDP; 4253 if (direction == PF_IN) { 4254 PF_ACPY(&key.ext.addr, pd->src, key.af); 4255 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 4256 key.ext.port = uh->uh_sport; 4257 key.gwy.port = uh->uh_dport; 4258 } else { 4259 PF_ACPY(&key.lan.addr, pd->src, key.af); 4260 PF_ACPY(&key.ext.addr, pd->dst, key.af); 4261 key.lan.port = uh->uh_sport; 4262 key.ext.port = uh->uh_dport; 4263 } 4264 4265 STATE_LOOKUP(); 4266 4267 if (direction == (*state)->state_key->direction) { 4268 src = &(*state)->src; 4269 dst = &(*state)->dst; 4270 } else { 4271 src = &(*state)->dst; 4272 dst = &(*state)->src; 4273 } 4274 4275 /* update states */ 4276 if (src->state < PFUDPS_SINGLE) 4277 src->state = PFUDPS_SINGLE; 4278 if (dst->state == PFUDPS_SINGLE) 4279 dst->state = PFUDPS_MULTIPLE; 4280 4281 /* update expire time */ 4282 (*state)->expire = time_second; 4283 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE) 4284 (*state)->timeout = PFTM_UDP_MULTIPLE; 4285 else 4286 (*state)->timeout = PFTM_UDP_SINGLE; 4287 4288 /* translate source/destination address, if necessary */ 4289 if (STATE_TRANSLATE((*state)->state_key)) { 4290 if (direction == PF_OUT) 4291 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum, 4292 &uh->uh_sum, &(*state)->state_key->gwy.addr, 4293 (*state)->state_key->gwy.port, 1, pd->af); 4294 else 4295 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum, 4296 &uh->uh_sum, &(*state)->state_key->lan.addr, 4297 (*state)->state_key->lan.port, 1, pd->af); 4298 m_copyback(m, off, sizeof(*uh), uh); 4299 } 4300 4301 return (PF_PASS); 4302 } 4303 4304 int 4305 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif, 4306 struct mbuf *m, int off, void *h, struct pf_pdesc *pd, 4307 u_short *reason) 4308 { 4309 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 4310 u_int16_t icmpid = 0, *icmpsum; 4311 u_int8_t icmptype; 4312 int state_icmp = 0; 4313 struct pf_state_key_cmp key; 4314 4315 icmpsum = NULL; /* XXXGCC -Wunitialized m68k */ 4316 icmptype = 0; /* XXXGCC -Wunitialized m68k */ 4317 4318 switch (pd->proto) { 4319 #ifdef INET 4320 case IPPROTO_ICMP: 4321 icmptype = pd->hdr.icmp->icmp_type; 4322 icmpid = pd->hdr.icmp->icmp_id; 4323 icmpsum = &pd->hdr.icmp->icmp_cksum; 4324 4325 if (icmptype == ICMP_UNREACH || 4326 icmptype == ICMP_SOURCEQUENCH || 4327 icmptype == ICMP_REDIRECT || 4328 icmptype == ICMP_TIMXCEED || 4329 icmptype == ICMP_PARAMPROB) 4330 state_icmp++; 4331 break; 4332 #endif /* INET */ 4333 #ifdef INET6 4334 case IPPROTO_ICMPV6: 4335 icmptype = pd->hdr.icmp6->icmp6_type; 4336 icmpid = pd->hdr.icmp6->icmp6_id; 4337 icmpsum = &pd->hdr.icmp6->icmp6_cksum; 4338 4339 if (icmptype == ICMP6_DST_UNREACH || 4340 icmptype == ICMP6_PACKET_TOO_BIG || 4341 icmptype == ICMP6_TIME_EXCEEDED || 4342 icmptype == ICMP6_PARAM_PROB) 4343 state_icmp++; 4344 break; 4345 #endif /* INET6 */ 4346 } 4347 4348 if (!state_icmp) { 4349 4350 /* 4351 * ICMP query/reply message not related to a TCP/UDP packet. 4352 * Search for an ICMP state. 4353 */ 4354 key.af = pd->af; 4355 key.proto = pd->proto; 4356 if (direction == PF_IN) { 4357 PF_ACPY(&key.ext.addr, pd->src, key.af); 4358 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 4359 key.ext.port = 0; 4360 key.gwy.port = icmpid; 4361 } else { 4362 PF_ACPY(&key.lan.addr, pd->src, key.af); 4363 PF_ACPY(&key.ext.addr, pd->dst, key.af); 4364 key.lan.port = icmpid; 4365 key.ext.port = 0; 4366 } 4367 4368 STATE_LOOKUP(); 4369 4370 (*state)->expire = time_second; 4371 (*state)->timeout = PFTM_ICMP_ERROR_REPLY; 4372 4373 /* translate source/destination address, if necessary */ 4374 if (STATE_TRANSLATE((*state)->state_key)) { 4375 if (direction == PF_OUT) { 4376 switch (pd->af) { 4377 #ifdef INET 4378 case AF_INET: 4379 pf_change_a(&saddr->v4.s_addr, 4380 pd->ip_sum, 4381 (*state)->state_key->gwy.addr.v4.s_addr, 0); 4382 pd->hdr.icmp->icmp_cksum = 4383 pf_cksum_fixup( 4384 pd->hdr.icmp->icmp_cksum, icmpid, 4385 (*state)->state_key->gwy.port, 0); 4386 pd->hdr.icmp->icmp_id = 4387 (*state)->state_key->gwy.port; 4388 m_copyback(m, off, ICMP_MINLEN, 4389 pd->hdr.icmp); 4390 break; 4391 #endif /* INET */ 4392 #ifdef INET6 4393 case AF_INET6: 4394 pf_change_a6(saddr, 4395 &pd->hdr.icmp6->icmp6_cksum, 4396 &(*state)->state_key->gwy.addr, 0); 4397 m_copyback(m, off, 4398 sizeof(struct icmp6_hdr), 4399 pd->hdr.icmp6); 4400 break; 4401 #endif /* INET6 */ 4402 } 4403 } else { 4404 switch (pd->af) { 4405 #ifdef INET 4406 case AF_INET: 4407 pf_change_a(&daddr->v4.s_addr, 4408 pd->ip_sum, 4409 (*state)->state_key->lan.addr.v4.s_addr, 0); 4410 pd->hdr.icmp->icmp_cksum = 4411 pf_cksum_fixup( 4412 pd->hdr.icmp->icmp_cksum, icmpid, 4413 (*state)->state_key->lan.port, 0); 4414 pd->hdr.icmp->icmp_id = 4415 (*state)->state_key->lan.port; 4416 m_copyback(m, off, ICMP_MINLEN, 4417 pd->hdr.icmp); 4418 break; 4419 #endif /* INET */ 4420 #ifdef INET6 4421 case AF_INET6: 4422 pf_change_a6(daddr, 4423 &pd->hdr.icmp6->icmp6_cksum, 4424 &(*state)->state_key->lan.addr, 0); 4425 m_copyback(m, off, 4426 sizeof(struct icmp6_hdr), 4427 pd->hdr.icmp6); 4428 break; 4429 #endif /* INET6 */ 4430 } 4431 } 4432 } 4433 4434 return (PF_PASS); 4435 4436 } else { 4437 /* 4438 * ICMP error message in response to a TCP/UDP packet. 4439 * Extract the inner TCP/UDP header and search for that state. 4440 */ 4441 4442 struct pf_pdesc pd2; 4443 #ifdef INET 4444 struct ip h2; 4445 #endif /* INET */ 4446 #ifdef INET6 4447 struct ip6_hdr h2_6; 4448 int terminal = 0; 4449 #endif /* INET6 */ 4450 int ipoff2 = 0; 4451 int off2 = 0; 4452 4453 memset(&pd2, 0, sizeof pd2); /* XXX gcc */ 4454 4455 pd2.af = pd->af; 4456 switch (pd->af) { 4457 #ifdef INET 4458 case AF_INET: 4459 /* offset of h2 in mbuf chain */ 4460 ipoff2 = off + ICMP_MINLEN; 4461 4462 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2), 4463 NULL, reason, pd2.af)) { 4464 DPFPRINTF(PF_DEBUG_MISC, 4465 ("pf: ICMP error message too short " 4466 "(ip)\n")); 4467 return (PF_DROP); 4468 } 4469 /* 4470 * ICMP error messages don't refer to non-first 4471 * fragments 4472 */ 4473 if (h2.ip_off & htons(IP_OFFMASK)) { 4474 REASON_SET(reason, PFRES_FRAG); 4475 return (PF_DROP); 4476 } 4477 4478 /* offset of protocol header that follows h2 */ 4479 off2 = ipoff2 + (h2.ip_hl << 2); 4480 4481 pd2.proto = h2.ip_p; 4482 pd2.src = (struct pf_addr *)&h2.ip_src; 4483 pd2.dst = (struct pf_addr *)&h2.ip_dst; 4484 pd2.ip_sum = &h2.ip_sum; 4485 break; 4486 #endif /* INET */ 4487 #ifdef INET6 4488 case AF_INET6: 4489 ipoff2 = off + sizeof(struct icmp6_hdr); 4490 4491 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6), 4492 NULL, reason, pd2.af)) { 4493 DPFPRINTF(PF_DEBUG_MISC, 4494 ("pf: ICMP error message too short " 4495 "(ip6)\n")); 4496 return (PF_DROP); 4497 } 4498 pd2.proto = h2_6.ip6_nxt; 4499 pd2.src = (struct pf_addr *)&h2_6.ip6_src; 4500 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst; 4501 pd2.ip_sum = NULL; 4502 off2 = ipoff2 + sizeof(h2_6); 4503 do { 4504 switch (pd2.proto) { 4505 case IPPROTO_FRAGMENT: 4506 /* 4507 * ICMPv6 error messages for 4508 * non-first fragments 4509 */ 4510 REASON_SET(reason, PFRES_FRAG); 4511 return (PF_DROP); 4512 case IPPROTO_AH: 4513 case IPPROTO_HOPOPTS: 4514 case IPPROTO_ROUTING: 4515 case IPPROTO_DSTOPTS: { 4516 /* get next header and header length */ 4517 struct ip6_ext opt6; 4518 4519 if (!pf_pull_hdr(m, off2, &opt6, 4520 sizeof(opt6), NULL, reason, 4521 pd2.af)) { 4522 DPFPRINTF(PF_DEBUG_MISC, 4523 ("pf: ICMPv6 short opt\n")); 4524 return (PF_DROP); 4525 } 4526 if (pd2.proto == IPPROTO_AH) 4527 off2 += (opt6.ip6e_len + 2) * 4; 4528 else 4529 off2 += (opt6.ip6e_len + 1) * 8; 4530 pd2.proto = opt6.ip6e_nxt; 4531 /* goto the next header */ 4532 break; 4533 } 4534 default: 4535 terminal++; 4536 break; 4537 } 4538 } while (!terminal); 4539 break; 4540 #endif /* INET6 */ 4541 } 4542 4543 switch (pd2.proto) { 4544 case IPPROTO_TCP: { 4545 struct tcphdr th; 4546 u_int32_t seq; 4547 struct pf_state_peer *src, *dst; 4548 u_int8_t dws; 4549 int copyback = 0; 4550 4551 /* 4552 * Only the first 8 bytes of the TCP header can be 4553 * expected. Don't access any TCP header fields after 4554 * th_seq, an ackskew test is not possible. 4555 */ 4556 if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason, 4557 pd2.af)) { 4558 DPFPRINTF(PF_DEBUG_MISC, 4559 ("pf: ICMP error message too short " 4560 "(tcp)\n")); 4561 return (PF_DROP); 4562 } 4563 4564 key.af = pd2.af; 4565 key.proto = IPPROTO_TCP; 4566 if (direction == PF_IN) { 4567 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 4568 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 4569 key.ext.port = th.th_dport; 4570 key.gwy.port = th.th_sport; 4571 } else { 4572 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 4573 PF_ACPY(&key.ext.addr, pd2.src, key.af); 4574 key.lan.port = th.th_dport; 4575 key.ext.port = th.th_sport; 4576 } 4577 4578 STATE_LOOKUP(); 4579 4580 if (direction == (*state)->state_key->direction) { 4581 src = &(*state)->dst; 4582 dst = &(*state)->src; 4583 } else { 4584 src = &(*state)->src; 4585 dst = &(*state)->dst; 4586 } 4587 4588 if (src->wscale && dst->wscale) 4589 dws = dst->wscale & PF_WSCALE_MASK; 4590 else 4591 dws = 0; 4592 4593 /* Demodulate sequence number */ 4594 seq = ntohl(th.th_seq) - src->seqdiff; 4595 if (src->seqdiff) { 4596 pf_change_a(&th.th_seq, icmpsum, 4597 htonl(seq), 0); 4598 copyback = 1; 4599 } 4600 4601 if (!SEQ_GEQ(src->seqhi, seq) || 4602 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) { 4603 if (pf_status.debug >= PF_DEBUG_MISC) { 4604 printf("pf: BAD ICMP %d:%d ", 4605 icmptype, pd->hdr.icmp->icmp_code); 4606 pf_print_host(pd->src, 0, pd->af); 4607 printf(" -> "); 4608 pf_print_host(pd->dst, 0, pd->af); 4609 printf(" state: "); 4610 pf_print_state(*state); 4611 printf(" seq=%u\n", seq); 4612 } 4613 REASON_SET(reason, PFRES_BADSTATE); 4614 return (PF_DROP); 4615 } 4616 4617 if (STATE_TRANSLATE((*state)->state_key)) { 4618 if (direction == PF_IN) { 4619 pf_change_icmp(pd2.src, &th.th_sport, 4620 daddr, &(*state)->state_key->lan.addr, 4621 (*state)->state_key->lan.port, NULL, 4622 pd2.ip_sum, icmpsum, 4623 pd->ip_sum, 0, pd2.af); 4624 } else { 4625 pf_change_icmp(pd2.dst, &th.th_dport, 4626 saddr, &(*state)->state_key->gwy.addr, 4627 (*state)->state_key->gwy.port, NULL, 4628 pd2.ip_sum, icmpsum, 4629 pd->ip_sum, 0, pd2.af); 4630 } 4631 copyback = 1; 4632 } 4633 4634 if (copyback) { 4635 switch (pd2.af) { 4636 #ifdef INET 4637 case AF_INET: 4638 m_copyback(m, off, ICMP_MINLEN, 4639 pd->hdr.icmp); 4640 m_copyback(m, ipoff2, sizeof(h2), 4641 &h2); 4642 break; 4643 #endif /* INET */ 4644 #ifdef INET6 4645 case AF_INET6: 4646 m_copyback(m, off, 4647 sizeof(struct icmp6_hdr), 4648 pd->hdr.icmp6); 4649 m_copyback(m, ipoff2, sizeof(h2_6), 4650 &h2_6); 4651 break; 4652 #endif /* INET6 */ 4653 } 4654 m_copyback(m, off2, 8, &th); 4655 } 4656 4657 return (PF_PASS); 4658 break; 4659 } 4660 case IPPROTO_UDP: { 4661 struct udphdr uh; 4662 4663 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh), 4664 NULL, reason, pd2.af)) { 4665 DPFPRINTF(PF_DEBUG_MISC, 4666 ("pf: ICMP error message too short " 4667 "(udp)\n")); 4668 return (PF_DROP); 4669 } 4670 4671 key.af = pd2.af; 4672 key.proto = IPPROTO_UDP; 4673 if (direction == PF_IN) { 4674 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 4675 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 4676 key.ext.port = uh.uh_dport; 4677 key.gwy.port = uh.uh_sport; 4678 } else { 4679 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 4680 PF_ACPY(&key.ext.addr, pd2.src, key.af); 4681 key.lan.port = uh.uh_dport; 4682 key.ext.port = uh.uh_sport; 4683 } 4684 4685 STATE_LOOKUP(); 4686 4687 if (STATE_TRANSLATE((*state)->state_key)) { 4688 if (direction == PF_IN) { 4689 pf_change_icmp(pd2.src, &uh.uh_sport, 4690 daddr, 4691 &(*state)->state_key->lan.addr, 4692 (*state)->state_key->lan.port, 4693 &uh.uh_sum, 4694 pd2.ip_sum, icmpsum, 4695 pd->ip_sum, 1, pd2.af); 4696 } else { 4697 pf_change_icmp(pd2.dst, &uh.uh_dport, 4698 saddr, 4699 &(*state)->state_key->gwy.addr, 4700 (*state)->state_key->gwy.port, &uh.uh_sum, 4701 pd2.ip_sum, icmpsum, 4702 pd->ip_sum, 1, pd2.af); 4703 } 4704 switch (pd2.af) { 4705 #ifdef INET 4706 case AF_INET: 4707 m_copyback(m, off, ICMP_MINLEN, 4708 pd->hdr.icmp); 4709 m_copyback(m, ipoff2, sizeof(h2), &h2); 4710 break; 4711 #endif /* INET */ 4712 #ifdef INET6 4713 case AF_INET6: 4714 m_copyback(m, off, 4715 sizeof(struct icmp6_hdr), 4716 pd->hdr.icmp6); 4717 m_copyback(m, ipoff2, sizeof(h2_6), 4718 &h2_6); 4719 break; 4720 #endif /* INET6 */ 4721 } 4722 m_copyback(m, off2, sizeof(uh), &uh); 4723 } 4724 4725 return (PF_PASS); 4726 break; 4727 } 4728 #ifdef INET 4729 case IPPROTO_ICMP: { 4730 struct icmp iih; 4731 4732 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN, 4733 NULL, reason, pd2.af)) { 4734 DPFPRINTF(PF_DEBUG_MISC, 4735 ("pf: ICMP error message too short i" 4736 "(icmp)\n")); 4737 return (PF_DROP); 4738 } 4739 4740 key.af = pd2.af; 4741 key.proto = IPPROTO_ICMP; 4742 if (direction == PF_IN) { 4743 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 4744 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 4745 key.ext.port = 0; 4746 key.gwy.port = iih.icmp_id; 4747 } else { 4748 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 4749 PF_ACPY(&key.ext.addr, pd2.src, key.af); 4750 key.lan.port = iih.icmp_id; 4751 key.ext.port = 0; 4752 } 4753 4754 STATE_LOOKUP(); 4755 4756 if (STATE_TRANSLATE((*state)->state_key)) { 4757 if (direction == PF_IN) { 4758 pf_change_icmp(pd2.src, &iih.icmp_id, 4759 daddr, 4760 &(*state)->state_key->lan.addr, 4761 (*state)->state_key->lan.port, NULL, 4762 pd2.ip_sum, icmpsum, 4763 pd->ip_sum, 0, AF_INET); 4764 } else { 4765 pf_change_icmp(pd2.dst, &iih.icmp_id, 4766 saddr, 4767 &(*state)->state_key->gwy.addr, 4768 (*state)->state_key->gwy.port, NULL, 4769 pd2.ip_sum, icmpsum, 4770 pd->ip_sum, 0, AF_INET); 4771 } 4772 m_copyback(m, off, ICMP_MINLEN, pd->hdr.icmp); 4773 m_copyback(m, ipoff2, sizeof(h2), &h2); 4774 m_copyback(m, off2, ICMP_MINLEN, &iih); 4775 } 4776 4777 return (PF_PASS); 4778 break; 4779 } 4780 #endif /* INET */ 4781 #ifdef INET6 4782 case IPPROTO_ICMPV6: { 4783 struct icmp6_hdr iih; 4784 4785 if (!pf_pull_hdr(m, off2, &iih, 4786 sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) { 4787 DPFPRINTF(PF_DEBUG_MISC, 4788 ("pf: ICMP error message too short " 4789 "(icmp6)\n")); 4790 return (PF_DROP); 4791 } 4792 4793 key.af = pd2.af; 4794 key.proto = IPPROTO_ICMPV6; 4795 if (direction == PF_IN) { 4796 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 4797 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 4798 key.ext.port = 0; 4799 key.gwy.port = iih.icmp6_id; 4800 } else { 4801 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 4802 PF_ACPY(&key.ext.addr, pd2.src, key.af); 4803 key.lan.port = iih.icmp6_id; 4804 key.ext.port = 0; 4805 } 4806 4807 STATE_LOOKUP(); 4808 4809 if (STATE_TRANSLATE((*state)->state_key)) { 4810 if (direction == PF_IN) { 4811 pf_change_icmp(pd2.src, &iih.icmp6_id, 4812 daddr, 4813 &(*state)->state_key->lan.addr, 4814 (*state)->state_key->lan.port, NULL, 4815 pd2.ip_sum, icmpsum, 4816 pd->ip_sum, 0, AF_INET6); 4817 } else { 4818 pf_change_icmp(pd2.dst, &iih.icmp6_id, 4819 saddr, &(*state)->state_key->gwy.addr, 4820 (*state)->state_key->gwy.port, NULL, 4821 pd2.ip_sum, icmpsum, 4822 pd->ip_sum, 0, AF_INET6); 4823 } 4824 m_copyback(m, off, sizeof(struct icmp6_hdr), 4825 pd->hdr.icmp6); 4826 m_copyback(m, ipoff2, sizeof(h2_6), &h2_6); 4827 m_copyback(m, off2, sizeof(struct icmp6_hdr), 4828 &iih); 4829 } 4830 4831 return (PF_PASS); 4832 break; 4833 } 4834 #endif /* INET6 */ 4835 default: { 4836 key.af = pd2.af; 4837 key.proto = pd2.proto; 4838 if (direction == PF_IN) { 4839 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 4840 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 4841 key.ext.port = 0; 4842 key.gwy.port = 0; 4843 } else { 4844 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 4845 PF_ACPY(&key.ext.addr, pd2.src, key.af); 4846 key.lan.port = 0; 4847 key.ext.port = 0; 4848 } 4849 4850 STATE_LOOKUP(); 4851 4852 if (STATE_TRANSLATE((*state)->state_key)) { 4853 if (direction == PF_IN) { 4854 pf_change_icmp(pd2.src, NULL, 4855 daddr, 4856 &(*state)->state_key->lan.addr, 4857 0, NULL, 4858 pd2.ip_sum, icmpsum, 4859 pd->ip_sum, 0, pd2.af); 4860 } else { 4861 pf_change_icmp(pd2.dst, NULL, 4862 saddr, 4863 &(*state)->state_key->gwy.addr, 4864 0, NULL, 4865 pd2.ip_sum, icmpsum, 4866 pd->ip_sum, 0, pd2.af); 4867 } 4868 switch (pd2.af) { 4869 #ifdef INET 4870 case AF_INET: 4871 m_copyback(m, off, ICMP_MINLEN, 4872 pd->hdr.icmp); 4873 m_copyback(m, ipoff2, sizeof(h2), &h2); 4874 break; 4875 #endif /* INET */ 4876 #ifdef INET6 4877 case AF_INET6: 4878 m_copyback(m, off, 4879 sizeof(struct icmp6_hdr), 4880 pd->hdr.icmp6); 4881 m_copyback(m, ipoff2, sizeof(h2_6), 4882 &h2_6); 4883 break; 4884 #endif /* INET6 */ 4885 } 4886 } 4887 4888 return (PF_PASS); 4889 break; 4890 } 4891 } 4892 } 4893 } 4894 4895 int 4896 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif, 4897 struct pf_pdesc *pd) 4898 { 4899 struct pf_state_peer *src, *dst; 4900 struct pf_state_key_cmp key; 4901 4902 key.af = pd->af; 4903 key.proto = pd->proto; 4904 if (direction == PF_IN) { 4905 PF_ACPY(&key.ext.addr, pd->src, key.af); 4906 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 4907 key.ext.port = 0; 4908 key.gwy.port = 0; 4909 } else { 4910 PF_ACPY(&key.lan.addr, pd->src, key.af); 4911 PF_ACPY(&key.ext.addr, pd->dst, key.af); 4912 key.lan.port = 0; 4913 key.ext.port = 0; 4914 } 4915 4916 STATE_LOOKUP(); 4917 4918 if (direction == (*state)->state_key->direction) { 4919 src = &(*state)->src; 4920 dst = &(*state)->dst; 4921 } else { 4922 src = &(*state)->dst; 4923 dst = &(*state)->src; 4924 } 4925 4926 /* update states */ 4927 if (src->state < PFOTHERS_SINGLE) 4928 src->state = PFOTHERS_SINGLE; 4929 if (dst->state == PFOTHERS_SINGLE) 4930 dst->state = PFOTHERS_MULTIPLE; 4931 4932 /* update expire time */ 4933 (*state)->expire = time_second; 4934 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE) 4935 (*state)->timeout = PFTM_OTHER_MULTIPLE; 4936 else 4937 (*state)->timeout = PFTM_OTHER_SINGLE; 4938 4939 /* translate source/destination address, if necessary */ 4940 if (STATE_TRANSLATE((*state)->state_key)) { 4941 if (direction == PF_OUT) 4942 switch (pd->af) { 4943 #ifdef INET 4944 case AF_INET: 4945 pf_change_a(&pd->src->v4.s_addr, 4946 pd->ip_sum, 4947 (*state)->state_key->gwy.addr.v4.s_addr, 4948 0); 4949 break; 4950 #endif /* INET */ 4951 #ifdef INET6 4952 case AF_INET6: 4953 PF_ACPY(pd->src, 4954 &(*state)->state_key->gwy.addr, pd->af); 4955 break; 4956 #endif /* INET6 */ 4957 } 4958 else 4959 switch (pd->af) { 4960 #ifdef INET 4961 case AF_INET: 4962 pf_change_a(&pd->dst->v4.s_addr, 4963 pd->ip_sum, 4964 (*state)->state_key->lan.addr.v4.s_addr, 4965 0); 4966 break; 4967 #endif /* INET */ 4968 #ifdef INET6 4969 case AF_INET6: 4970 PF_ACPY(pd->dst, 4971 &(*state)->state_key->lan.addr, pd->af); 4972 break; 4973 #endif /* INET6 */ 4974 } 4975 } 4976 4977 return (PF_PASS); 4978 } 4979 4980 /* 4981 * ipoff and off are measured from the start of the mbuf chain. 4982 * h must be at "ipoff" on the mbuf chain. 4983 */ 4984 void * 4985 pf_pull_hdr(struct mbuf *m, int off, void *p, int len, 4986 u_short *actionp, u_short *reasonp, sa_family_t af) 4987 { 4988 switch (af) { 4989 #ifdef INET 4990 case AF_INET: { 4991 struct ip *h = mtod(m, struct ip *); 4992 u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3; 4993 4994 if (fragoff) { 4995 if (fragoff >= len) 4996 ACTION_SET(actionp, PF_PASS); 4997 else { 4998 ACTION_SET(actionp, PF_DROP); 4999 REASON_SET(reasonp, PFRES_FRAG); 5000 } 5001 return (NULL); 5002 } 5003 if (m->m_pkthdr.len < off + len || 5004 ntohs(h->ip_len) < off + len) { 5005 ACTION_SET(actionp, PF_DROP); 5006 REASON_SET(reasonp, PFRES_SHORT); 5007 return (NULL); 5008 } 5009 break; 5010 } 5011 #endif /* INET */ 5012 #ifdef INET6 5013 case AF_INET6: { 5014 struct ip6_hdr *h = mtod(m, struct ip6_hdr *); 5015 5016 if (m->m_pkthdr.len < off + len || 5017 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) < 5018 (unsigned)(off + len)) { 5019 ACTION_SET(actionp, PF_DROP); 5020 REASON_SET(reasonp, PFRES_SHORT); 5021 return (NULL); 5022 } 5023 break; 5024 } 5025 #endif /* INET6 */ 5026 } 5027 m_copydata(m, off, len, p); 5028 return (p); 5029 } 5030 5031 int 5032 pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif) 5033 { 5034 #ifdef __NetBSD__ 5035 union { 5036 struct sockaddr dst; 5037 struct sockaddr_in dst4; 5038 struct sockaddr_in6 dst6; 5039 } u; 5040 struct route ro; 5041 int ret = 1; 5042 5043 bzero(&ro, sizeof(ro)); 5044 switch (af) { 5045 case AF_INET: 5046 sockaddr_in_init(&u.dst4, &addr->v4, 0); 5047 break; 5048 #ifdef INET6 5049 case AF_INET6: 5050 sockaddr_in6_init(&u.dst6, &addr->v6, 0, 0, 0); 5051 break; 5052 #endif /* INET6 */ 5053 default: 5054 return (0); 5055 } 5056 rtcache_setdst(&ro, &u.dst); 5057 5058 ret = rtcache_init(&ro) != NULL ? 1 : 0; 5059 rtcache_free(&ro); 5060 5061 return (ret); 5062 #else /* !__NetBSD__ */ 5063 struct sockaddr_in *dst; 5064 int ret = 1; 5065 int check_mpath; 5066 extern int ipmultipath; 5067 #ifdef INET6 5068 extern int ip6_multipath; 5069 struct sockaddr_in6 *dst6; 5070 struct route_in6 ro; 5071 #else 5072 struct route ro; 5073 #endif 5074 struct radix_node *rn; 5075 struct rtentry *rt; 5076 struct ifnet *ifp; 5077 5078 check_mpath = 0; 5079 bzero(&ro, sizeof(ro)); 5080 switch (af) { 5081 case AF_INET: 5082 dst = satosin(&ro.ro_dst); 5083 dst->sin_family = AF_INET; 5084 dst->sin_len = sizeof(*dst); 5085 dst->sin_addr = addr->v4; 5086 if (ipmultipath) 5087 check_mpath = 1; 5088 break; 5089 #ifdef INET6 5090 case AF_INET6: 5091 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 5092 dst6->sin6_family = AF_INET6; 5093 dst6->sin6_len = sizeof(*dst6); 5094 dst6->sin6_addr = addr->v6; 5095 if (ip6_multipath) 5096 check_mpath = 1; 5097 break; 5098 #endif /* INET6 */ 5099 default: 5100 return (0); 5101 } 5102 5103 /* Skip checks for ipsec interfaces */ 5104 if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC) 5105 goto out; 5106 5107 rtalloc_noclone((struct route *)&ro, NO_CLONING); 5108 5109 if (ro.ro_rt != NULL) { 5110 /* No interface given, this is a no-route check */ 5111 if (kif == NULL) 5112 goto out; 5113 5114 if (kif->pfik_ifp == NULL) { 5115 ret = 0; 5116 goto out; 5117 } 5118 5119 /* Perform uRPF check if passed input interface */ 5120 ret = 0; 5121 rn = (struct radix_node *)ro.ro_rt; 5122 do { 5123 rt = (struct rtentry *)rn; 5124 if (rt->rt_ifp->if_type == IFT_CARP) 5125 ifp = rt->rt_ifp->if_carpdev; 5126 else 5127 ifp = rt->rt_ifp; 5128 5129 if (kif->pfik_ifp == ifp) 5130 ret = 1; 5131 rn = rn_mpath_next(rn); 5132 } while (check_mpath == 1 && rn != NULL && ret == 0); 5133 } else 5134 ret = 0; 5135 out: 5136 if (ro.ro_rt != NULL) 5137 RTFREE(ro.ro_rt); 5138 return (ret); 5139 #endif /* !__NetBSD__ */ 5140 } 5141 5142 int 5143 pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw) 5144 { 5145 #ifdef __NetBSD__ 5146 /* NetBSD doesn't have route labels. */ 5147 5148 return (0); 5149 #else 5150 struct sockaddr_in *dst; 5151 #ifdef INET6 5152 struct sockaddr_in6 *dst6; 5153 struct route_in6 ro; 5154 #else 5155 struct route ro; 5156 #endif 5157 int ret = 0; 5158 5159 bzero(&ro, sizeof(ro)); 5160 switch (af) { 5161 case AF_INET: 5162 dst = satosin(&ro.ro_dst); 5163 dst->sin_family = AF_INET; 5164 dst->sin_len = sizeof(*dst); 5165 dst->sin_addr = addr->v4; 5166 break; 5167 #ifdef INET6 5168 case AF_INET6: 5169 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 5170 dst6->sin6_family = AF_INET6; 5171 dst6->sin6_len = sizeof(*dst6); 5172 dst6->sin6_addr = addr->v6; 5173 break; 5174 #endif /* INET6 */ 5175 default: 5176 return (0); 5177 } 5178 5179 rtalloc_noclone((struct route *)&ro, NO_CLONING); 5180 5181 if (ro.ro_rt != NULL) { 5182 if (ro.ro_rt->rt_labelid == aw->v.rtlabel) 5183 ret = 1; 5184 RTFREE(ro.ro_rt); 5185 } 5186 5187 return (ret); 5188 #endif /* !__NetBSD__ */ 5189 } 5190 5191 #ifdef INET 5192 void 5193 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp, 5194 struct pf_state *s, struct pf_pdesc *pd) 5195 { 5196 struct mbuf *m0, *m1; 5197 struct route iproute; 5198 struct route *ro = NULL; 5199 const struct sockaddr *dst; 5200 union { 5201 struct sockaddr dst; 5202 struct sockaddr_in dst4; 5203 } u; 5204 struct ip *ip; 5205 struct ifnet *ifp = NULL; 5206 struct pf_addr naddr; 5207 struct pf_src_node *sn = NULL; 5208 int error = 0; 5209 #ifdef IPSEC 5210 struct m_tag *mtag; 5211 #endif /* IPSEC */ 5212 #ifdef __NetBSD__ 5213 struct pf_mtag *pf_mtag; 5214 #endif /* __NetBSD__ */ 5215 5216 if (m == NULL || *m == NULL || r == NULL || 5217 (dir != PF_IN && dir != PF_OUT) || oifp == NULL) 5218 panic("pf_route: invalid parameters"); 5219 5220 #ifdef __NetBSD__ 5221 if ((pf_mtag = pf_get_mtag(*m)) == NULL) { 5222 m0 = *m; 5223 *m = NULL; 5224 goto bad; 5225 } 5226 if (pf_mtag->routed++ > 3) { 5227 m0 = *m; 5228 *m = NULL; 5229 goto bad; 5230 } 5231 #else 5232 if ((*m)->m_pkthdr.pf.routed++ > 3) { 5233 m0 = *m; 5234 *m = NULL; 5235 goto bad; 5236 } 5237 #endif /* !__NetBSD__ */ 5238 5239 if (r->rt == PF_DUPTO) { 5240 if ((m0 = m_dup(*m, 0, M_COPYALL, M_NOWAIT)) == NULL) 5241 return; 5242 } else { 5243 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) 5244 return; 5245 m0 = *m; 5246 } 5247 5248 if (m0->m_len < sizeof(struct ip)) { 5249 DPFPRINTF(PF_DEBUG_URGENT, 5250 ("pf_route: m0->m_len < sizeof(struct ip)\n")); 5251 goto bad; 5252 } 5253 5254 ip = mtod(m0, struct ip *); 5255 5256 ro = &iproute; 5257 memset(ro, 0, sizeof(*ro)); 5258 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 5259 dst = &u.dst; 5260 rtcache_setdst(ro, dst); 5261 5262 if (r->rt == PF_FASTROUTE) { 5263 struct rtentry *rt; 5264 5265 rt = rtcache_init(ro); 5266 5267 if (rt == NULL) { 5268 ip_statinc(IP_STAT_NOROUTE); 5269 goto bad; 5270 } 5271 5272 ifp = rt->rt_ifp; 5273 rt->rt_use++; 5274 5275 if (rt->rt_flags & RTF_GATEWAY) 5276 dst = rt->rt_gateway; 5277 } else { 5278 if (TAILQ_EMPTY(&r->rpool.list)) { 5279 DPFPRINTF(PF_DEBUG_URGENT, 5280 ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n")); 5281 goto bad; 5282 } 5283 if (s == NULL) { 5284 pf_map_addr(AF_INET, r, 5285 (const struct pf_addr *)&ip->ip_src, 5286 &naddr, NULL, &sn); 5287 if (!PF_AZERO(&naddr, AF_INET)) 5288 u.dst4.sin_addr.s_addr = naddr.v4.s_addr; 5289 ifp = r->rpool.cur->kif ? 5290 r->rpool.cur->kif->pfik_ifp : NULL; 5291 } else { 5292 if (!PF_AZERO(&s->rt_addr, AF_INET)) 5293 u.dst4.sin_addr.s_addr = s->rt_addr.v4.s_addr; 5294 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; 5295 } 5296 } 5297 if (ifp == NULL) 5298 goto bad; 5299 5300 if (oifp != ifp) { 5301 if (pf_test(PF_OUT, ifp, &m0, NULL) != PF_PASS) 5302 goto bad; 5303 else if (m0 == NULL) 5304 goto done; 5305 if (m0->m_len < sizeof(struct ip)) { 5306 DPFPRINTF(PF_DEBUG_URGENT, 5307 ("pf_route: m0->m_len < sizeof(struct ip)\n")); 5308 goto bad; 5309 } 5310 ip = mtod(m0, struct ip *); 5311 } 5312 5313 /* Copied from ip_output. */ 5314 #ifdef IPSEC 5315 /* 5316 * If deferred crypto processing is needed, check that the 5317 * interface supports it. 5318 */ 5319 if ((mtag = m_tag_find(m0, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL)) 5320 != NULL && (ifp->if_capabilities & IFCAP_IPSEC) == 0) { 5321 /* Notify IPsec to do its own crypto. */ 5322 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 5323 goto bad; 5324 } 5325 #endif /* IPSEC */ 5326 5327 /* Catch routing changes wrt. hardware checksumming for TCP or UDP. */ 5328 #ifdef __NetBSD__ 5329 if (m0->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 5330 in_delayed_cksum(m0); 5331 m0->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 5332 } 5333 #else 5334 if (m0->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT) { 5335 if (!(ifp->if_capabilities & IFCAP_CSUM_TCPv4) || 5336 ifp->if_bridge != NULL) { 5337 in_delayed_cksum(m0); 5338 m0->m_pkthdr.csum_flags &= ~M_TCPV4_CSUM_OUT; /* Clear */ 5339 } 5340 } else if (m0->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT) { 5341 if (!(ifp->if_capabilities & IFCAP_CSUM_UDPv4) || 5342 ifp->if_bridge != NULL) { 5343 in_delayed_cksum(m0); 5344 m0->m_pkthdr.csum_flags &= ~M_UDPV4_CSUM_OUT; /* Clear */ 5345 } 5346 } 5347 #endif /* !__NetBSD__ */ 5348 5349 if (ntohs(ip->ip_len) <= ifp->if_mtu) { 5350 #ifdef __NetBSD__ 5351 ip->ip_sum = 0; 5352 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2); 5353 5354 m0->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 5355 #else 5356 if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) && 5357 ifp->if_bridge == NULL) { 5358 m0->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT; 5359 ipstat.ips_outhwcsum++; 5360 } else { 5361 ip->ip_sum = 0; 5362 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2); 5363 } 5364 /* Update relevant hardware checksum stats for TCP/UDP */ 5365 if (m0->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT) 5366 tcpstat.tcps_outhwcsum++; 5367 else if (m0->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT) 5368 udpstat.udps_outhwcsum++; 5369 #endif /* !__NetBSD__ */ 5370 error = (*ifp->if_output)(ifp, m0, dst, NULL); 5371 goto done; 5372 } 5373 5374 /* 5375 * Too large for interface; fragment if possible. 5376 * Must be able to put at least 8 bytes per fragment. 5377 */ 5378 if (ip->ip_off & htons(IP_DF)) { 5379 ip_statinc(IP_STAT_CANTFRAG); 5380 if (r->rt != PF_DUPTO) { 5381 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0, 5382 ifp->if_mtu); 5383 goto done; 5384 } else 5385 goto bad; 5386 } 5387 5388 #ifdef __NetBSD__ 5389 /* Make ip_fragment re-compute checksums. */ 5390 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) { 5391 m0->m_pkthdr.csum_flags |= M_CSUM_IPv4; 5392 } 5393 #endif /* __NetBSD__ */ 5394 m1 = m0; 5395 error = ip_fragment(m0, ifp, ifp->if_mtu); 5396 if (error) { 5397 m0 = NULL; 5398 goto bad; 5399 } 5400 5401 for (m0 = m1; m0; m0 = m1) { 5402 m1 = m0->m_nextpkt; 5403 m0->m_nextpkt = 0; 5404 if (error == 0) 5405 error = (*ifp->if_output)(ifp, m0, dst, NULL); 5406 else 5407 m_freem(m0); 5408 } 5409 5410 if (error == 0) 5411 ip_statinc(IP_STAT_FRAGMENTED); 5412 5413 done: 5414 if (r->rt != PF_DUPTO) 5415 *m = NULL; 5416 if (ro == &iproute) 5417 rtcache_free(ro); 5418 return; 5419 5420 bad: 5421 m_freem(m0); 5422 goto done; 5423 } 5424 #endif /* INET */ 5425 5426 #ifdef INET6 5427 void 5428 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp, 5429 struct pf_state *s, struct pf_pdesc *pd) 5430 { 5431 struct mbuf *m0; 5432 struct sockaddr_in6 dst; 5433 struct ip6_hdr *ip6; 5434 struct ifnet *ifp = NULL; 5435 struct pf_addr naddr; 5436 struct pf_src_node *sn = NULL; 5437 int error = 0; 5438 #ifdef __NetBSD__ 5439 struct pf_mtag *pf_mtag; 5440 #endif /* __NetBSD__ */ 5441 5442 if (m == NULL || *m == NULL || r == NULL || 5443 (dir != PF_IN && dir != PF_OUT) || oifp == NULL) 5444 panic("pf_route6: invalid parameters"); 5445 5446 #ifdef __NetBSD__ 5447 if ((pf_mtag = pf_get_mtag(*m)) == NULL) { 5448 m0 = *m; 5449 *m = NULL; 5450 goto bad; 5451 } 5452 if (pf_mtag->routed++ > 3) { 5453 m0 = *m; 5454 *m = NULL; 5455 goto bad; 5456 } 5457 #else 5458 if ((*m)->m_pkthdr.pf.routed++ > 3) { 5459 m0 = *m; 5460 *m = NULL; 5461 goto bad; 5462 } 5463 #endif /* !__NetBSD__ */ 5464 5465 if (r->rt == PF_DUPTO) { 5466 if ((m0 = m_dup(*m, 0, M_COPYALL, M_NOWAIT)) == NULL) 5467 return; 5468 } else { 5469 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) 5470 return; 5471 m0 = *m; 5472 } 5473 5474 if (m0->m_len < sizeof(struct ip6_hdr)) { 5475 DPFPRINTF(PF_DEBUG_URGENT, 5476 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n")); 5477 goto bad; 5478 } 5479 ip6 = mtod(m0, struct ip6_hdr *); 5480 5481 dst.sin6_family = AF_INET6; 5482 dst.sin6_len = sizeof(dst); 5483 dst.sin6_addr = ip6->ip6_dst; 5484 5485 /* Cheat. XXX why only in the v6 case??? */ 5486 if (r->rt == PF_FASTROUTE) { 5487 #ifdef __NetBSD__ 5488 pf_mtag->flags |= PF_TAG_GENERATED; 5489 #else 5490 m0->m_pkthdr.pf.flags |= PF_TAG_GENERATED; 5491 #endif /* !__NetBSD__ */ 5492 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL); 5493 return; 5494 } 5495 5496 if (TAILQ_EMPTY(&r->rpool.list)) { 5497 DPFPRINTF(PF_DEBUG_URGENT, 5498 ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n")); 5499 goto bad; 5500 } 5501 if (s == NULL) { 5502 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src, 5503 &naddr, NULL, &sn); 5504 if (!PF_AZERO(&naddr, AF_INET6)) 5505 PF_ACPY((struct pf_addr *)&dst.sin6_addr, 5506 &naddr, AF_INET6); 5507 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL; 5508 } else { 5509 if (!PF_AZERO(&s->rt_addr, AF_INET6)) 5510 PF_ACPY((struct pf_addr *)&dst.sin6_addr, 5511 &s->rt_addr, AF_INET6); 5512 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; 5513 } 5514 if (ifp == NULL) 5515 goto bad; 5516 5517 if (oifp != ifp) { 5518 if (pf_test6(PF_OUT, ifp, &m0, NULL) != PF_PASS) 5519 goto bad; 5520 else if (m0 == NULL) 5521 goto done; 5522 if (m0->m_len < sizeof(struct ip6_hdr)) { 5523 DPFPRINTF(PF_DEBUG_URGENT, 5524 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n")); 5525 goto bad; 5526 } 5527 ip6 = mtod(m0, struct ip6_hdr *); 5528 } 5529 5530 /* 5531 * If the packet is too large for the outgoing interface, 5532 * send back an icmp6 error. 5533 */ 5534 if (IN6_IS_SCOPE_EMBEDDABLE(&dst.sin6_addr)) 5535 dst.sin6_addr.s6_addr16[1] = htons(ifp->if_index); 5536 if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) { 5537 error = nd6_output(ifp, ifp, m0, &dst, NULL); 5538 } else { 5539 in6_ifstat_inc(ifp, ifs6_in_toobig); 5540 if (r->rt != PF_DUPTO) 5541 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu); 5542 else 5543 goto bad; 5544 } 5545 5546 done: 5547 if (r->rt != PF_DUPTO) 5548 *m = NULL; 5549 return; 5550 5551 bad: 5552 m_freem(m0); 5553 goto done; 5554 } 5555 #endif /* INET6 */ 5556 5557 5558 /* 5559 * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag 5560 * off is the offset where the protocol header starts 5561 * len is the total length of protocol header plus payload 5562 * returns 0 when the checksum is valid, otherwise returns 1. 5563 */ 5564 #ifdef __NetBSD__ 5565 int 5566 pf_check_proto_cksum(struct mbuf *m, int direction, int off, int len, 5567 u_int8_t p, sa_family_t af) 5568 #else 5569 int 5570 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p, 5571 sa_family_t af) 5572 #endif /* !__NetBSD__ */ 5573 { 5574 #ifndef __NetBSD__ 5575 u_int16_t flag_ok, flag_bad; 5576 #endif /* !__NetBSD__ */ 5577 u_int16_t sum; 5578 5579 #ifndef __NetBSD__ 5580 switch (p) { 5581 case IPPROTO_TCP: 5582 flag_ok = M_TCP_CSUM_IN_OK; 5583 flag_bad = M_TCP_CSUM_IN_BAD; 5584 break; 5585 case IPPROTO_UDP: 5586 flag_ok = M_UDP_CSUM_IN_OK; 5587 flag_bad = M_UDP_CSUM_IN_BAD; 5588 break; 5589 case IPPROTO_ICMP: 5590 #ifdef INET6 5591 case IPPROTO_ICMPV6: 5592 #endif /* INET6 */ 5593 flag_ok = flag_bad = 0; 5594 break; 5595 default: 5596 return (1); 5597 } 5598 if (m->m_pkthdr.csum_flags & flag_ok) 5599 return (0); 5600 if (m->m_pkthdr.csum_flags & flag_bad) 5601 return (1); 5602 #endif /* !__NetBSD__ */ 5603 if (off < sizeof(struct ip) || len < sizeof(struct udphdr)) 5604 return (1); 5605 if (m->m_pkthdr.len < off + len) 5606 return (1); 5607 #ifdef __NetBSD__ 5608 if (direction == PF_IN) { 5609 switch (p) { 5610 case IPPROTO_TCP: { 5611 struct tcphdr th; /* XXX */ 5612 int thlen; 5613 5614 m_copydata(m, off, sizeof(th), &th); /* XXX */ 5615 thlen = th.th_off << 2; 5616 return tcp_input_checksum(af, m, &th, off, 5617 thlen, len - thlen) != 0; 5618 } 5619 5620 case IPPROTO_UDP: { 5621 struct udphdr uh; /* XXX */ 5622 5623 m_copydata(m, off, sizeof(uh), &uh); /* XXX */ 5624 return udp_input_checksum(af, m, &uh, off, len) != 0; 5625 } 5626 } 5627 } 5628 #endif /* __NetBSD__ */ 5629 switch (af) { 5630 #ifdef INET 5631 case AF_INET: 5632 if (p == IPPROTO_ICMP) { 5633 if (m->m_len < off) 5634 return (1); 5635 m->m_data += off; 5636 m->m_len -= off; 5637 sum = in_cksum(m, len); 5638 m->m_data -= off; 5639 m->m_len += off; 5640 } else { 5641 if (m->m_len < sizeof(struct ip)) 5642 return (1); 5643 sum = in4_cksum(m, p, off, len); 5644 } 5645 break; 5646 #endif /* INET */ 5647 #ifdef INET6 5648 case AF_INET6: 5649 if (m->m_len < sizeof(struct ip6_hdr)) 5650 return (1); 5651 sum = in6_cksum(m, p, off, len); 5652 break; 5653 #endif /* INET6 */ 5654 default: 5655 return (1); 5656 } 5657 if (sum) { 5658 #ifndef __NetBSD__ 5659 m->m_pkthdr.csum_flags |= flag_bad; 5660 #endif /* !__NetBSD__ */ 5661 switch (p) { 5662 case IPPROTO_TCP: 5663 tcp_statinc(TCP_STAT_RCVBADSUM); 5664 break; 5665 case IPPROTO_UDP: 5666 udp_statinc(UDP_STAT_BADSUM); 5667 break; 5668 case IPPROTO_ICMP: 5669 icmp_statinc(ICMP_STAT_CHECKSUM); 5670 break; 5671 #ifdef INET6 5672 case IPPROTO_ICMPV6: 5673 icmp6_statinc(ICMP6_STAT_CHECKSUM); 5674 break; 5675 #endif /* INET6 */ 5676 } 5677 return (1); 5678 } 5679 #ifndef __NetBSD__ 5680 m->m_pkthdr.csum_flags |= flag_ok; 5681 #endif /* !__NetBSD__ */ 5682 return (0); 5683 } 5684 5685 #ifdef INET 5686 int 5687 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0, 5688 struct ether_header *eh) 5689 { 5690 struct pfi_kif *kif; 5691 u_short action, reason = 0, log = 0; 5692 struct mbuf *m = *m0; 5693 struct ip *h = NULL; 5694 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr; 5695 struct pf_state *s = NULL; 5696 struct pf_state_key *sk = NULL; 5697 struct pf_ruleset *ruleset = NULL; 5698 struct pf_pdesc pd; 5699 int off, dirndx, pqid = 0; 5700 #ifdef __NetBSD__ 5701 struct pf_mtag *pf_mtag = NULL; /* XXX gcc */ 5702 #endif /* __NetBSD__ */ 5703 5704 if (!pf_status.running) 5705 return (PF_PASS); 5706 5707 memset(&pd, 0, sizeof(pd)); 5708 if (ifp->if_type == IFT_CARP && ifp->if_carpdev) 5709 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif; 5710 else 5711 kif = (struct pfi_kif *)ifp->if_pf_kif; 5712 5713 if (kif == NULL) { 5714 DPFPRINTF(PF_DEBUG_URGENT, 5715 ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname)); 5716 return (PF_DROP); 5717 } 5718 if (kif->pfik_flags & PFI_IFLAG_SKIP) 5719 return (PF_PASS); 5720 5721 #ifdef DIAGNOSTIC 5722 if ((m->m_flags & M_PKTHDR) == 0) 5723 panic("non-M_PKTHDR is passed to pf_test"); 5724 #endif /* DIAGNOSTIC */ 5725 5726 if (m->m_pkthdr.len < (int)sizeof(*h)) { 5727 action = PF_DROP; 5728 REASON_SET(&reason, PFRES_SHORT); 5729 log = 1; 5730 goto done; 5731 } 5732 5733 #ifdef __NetBSD__ 5734 if ((pf_mtag = pf_get_mtag(m)) == NULL) { 5735 DPFPRINTF(PF_DEBUG_URGENT, 5736 ("pf_test: pf_get_mtag returned NULL\n")); 5737 return (PF_DROP); 5738 } 5739 if (pf_mtag->flags & PF_TAG_GENERATED) 5740 return (PF_PASS); 5741 #else 5742 if (m->m_pkthdr.pf.flags & PF_TAG_GENERATED) 5743 return (PF_PASS); 5744 #endif /* !__NetBSD__ */ 5745 5746 /* We do IP header normalization and packet reassembly here */ 5747 if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) { 5748 action = PF_DROP; 5749 goto done; 5750 } 5751 m = *m0; /* pf_normalize messes with m0 */ 5752 h = mtod(m, struct ip *); 5753 5754 off = h->ip_hl << 2; 5755 if (off < (int)sizeof(*h)) { 5756 action = PF_DROP; 5757 REASON_SET(&reason, PFRES_SHORT); 5758 log = 1; 5759 goto done; 5760 } 5761 5762 pd.src = (struct pf_addr *)&h->ip_src; 5763 pd.dst = (struct pf_addr *)&h->ip_dst; 5764 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET); 5765 pd.ip_sum = &h->ip_sum; 5766 pd.proto = h->ip_p; 5767 pd.af = AF_INET; 5768 pd.tos = h->ip_tos; 5769 pd.tot_len = ntohs(h->ip_len); 5770 pd.eh = eh; 5771 5772 /* handle fragments that didn't get reassembled by normalization */ 5773 if (h->ip_off & htons(IP_MF | IP_OFFMASK)) { 5774 action = pf_test_fragment(&r, dir, kif, m, h, 5775 &pd, &a, &ruleset); 5776 goto done; 5777 } 5778 5779 switch (h->ip_p) { 5780 5781 case IPPROTO_TCP: { 5782 struct tcphdr th; 5783 5784 pd.hdr.tcp = &th; 5785 if (!pf_pull_hdr(m, off, &th, sizeof(th), 5786 &action, &reason, AF_INET)) { 5787 log = action != PF_PASS; 5788 goto done; 5789 } 5790 pd.p_len = pd.tot_len - off - (th.th_off << 2); 5791 if ((th.th_flags & TH_ACK) && pd.p_len == 0) 5792 pqid = 1; 5793 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd); 5794 if (action == PF_DROP) 5795 goto done; 5796 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd, 5797 &reason); 5798 if (action == PF_PASS) { 5799 #if NPFSYNC 5800 pfsync_update_state(s); 5801 #endif /* NPFSYNC */ 5802 r = s->rule.ptr; 5803 a = s->anchor.ptr; 5804 log = s->log; 5805 } else if (s == NULL) 5806 action = pf_test_rule(&r, &s, dir, kif, 5807 m, off, h, &pd, &a, &ruleset, &ipintrq); 5808 break; 5809 } 5810 5811 case IPPROTO_UDP: { 5812 struct udphdr uh; 5813 5814 pd.hdr.udp = &uh; 5815 if (!pf_pull_hdr(m, off, &uh, sizeof(uh), 5816 &action, &reason, AF_INET)) { 5817 log = action != PF_PASS; 5818 goto done; 5819 } 5820 if (uh.uh_dport == 0 || 5821 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off || 5822 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) { 5823 action = PF_DROP; 5824 REASON_SET(&reason, PFRES_SHORT); 5825 goto done; 5826 } 5827 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd); 5828 if (action == PF_PASS) { 5829 #if NPFSYNC 5830 pfsync_update_state(s); 5831 #endif /* NPFSYNC */ 5832 r = s->rule.ptr; 5833 a = s->anchor.ptr; 5834 log = s->log; 5835 } else if (s == NULL) 5836 action = pf_test_rule(&r, &s, dir, kif, 5837 m, off, h, &pd, &a, &ruleset, &ipintrq); 5838 break; 5839 } 5840 5841 case IPPROTO_ICMP: { 5842 struct icmp ih; 5843 5844 pd.hdr.icmp = &ih; 5845 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN, 5846 &action, &reason, AF_INET)) { 5847 log = action != PF_PASS; 5848 goto done; 5849 } 5850 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd, 5851 &reason); 5852 if (action == PF_PASS) { 5853 #if NPFSYNC 5854 pfsync_update_state(s); 5855 #endif /* NPFSYNC */ 5856 r = s->rule.ptr; 5857 a = s->anchor.ptr; 5858 log = s->log; 5859 } else if (s == NULL) 5860 action = pf_test_rule(&r, &s, dir, kif, 5861 m, off, h, &pd, &a, &ruleset, &ipintrq); 5862 break; 5863 } 5864 5865 #ifdef INET6 5866 case IPPROTO_ICMPV6: { 5867 action = PF_DROP; 5868 DPFPRINTF(PF_DEBUG_MISC, 5869 ("pf: dropping IPv4 packet with ICMPv6 payload\n")); 5870 goto done; 5871 } 5872 #endif 5873 5874 default: 5875 action = pf_test_state_other(&s, dir, kif, &pd); 5876 if (action == PF_PASS) { 5877 #if NPFSYNC 5878 pfsync_update_state(s); 5879 #endif /* NPFSYNC */ 5880 r = s->rule.ptr; 5881 a = s->anchor.ptr; 5882 log = s->log; 5883 } else if (s == NULL) 5884 action = pf_test_rule(&r, &s, dir, kif, m, off, h, 5885 &pd, &a, &ruleset, &ipintrq); 5886 break; 5887 } 5888 5889 done: 5890 if (action == PF_PASS && h->ip_hl > 5 && 5891 !((s && s->allow_opts) || r->allow_opts)) { 5892 action = PF_DROP; 5893 REASON_SET(&reason, PFRES_IPOPTIONS); 5894 log = 1; 5895 DPFPRINTF(PF_DEBUG_MISC, 5896 ("pf: dropping packet with ip options\n")); 5897 } 5898 5899 if ((s && s->tag) || r->rtableid) 5900 pf_tag_packet(m, s ? s->tag : 0, r->rtableid); 5901 5902 #ifdef ALTQ 5903 if (action == PF_PASS && r->qid) { 5904 #ifdef __NetBSD__ 5905 struct m_tag *mtag; 5906 struct altq_tag *atag; 5907 5908 mtag = m_tag_get(PACKET_TAG_ALTQ_QID, sizeof(*atag), M_NOWAIT); 5909 if (mtag != NULL) { 5910 atag = (struct altq_tag *)(mtag + 1); 5911 if (pqid || (pd.tos & IPTOS_LOWDELAY)) 5912 atag->qid = r->pqid; 5913 else 5914 atag->qid = r->qid; 5915 /* add hints for ecn */ 5916 atag->af = AF_INET; 5917 atag->hdr = h; 5918 m_tag_prepend(m, mtag); 5919 } 5920 #else 5921 if (pqid || (pd.tos & IPTOS_LOWDELAY)) 5922 m->m_pkthdr.pf.qid = r->pqid; 5923 else 5924 m->m_pkthdr.pf.qid = r->qid; 5925 /* add hints for ecn */ 5926 m->m_pkthdr.pf.hdr = h; 5927 #endif /* !__NetBSD__ */ 5928 } 5929 #endif /* ALTQ */ 5930 5931 /* 5932 * connections redirected to loopback should not match sockets 5933 * bound specifically to loopback due to security implications, 5934 * see tcp_input() and in_pcblookup_listen(). 5935 */ 5936 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP || 5937 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL && 5938 (s->nat_rule.ptr->action == PF_RDR || 5939 s->nat_rule.ptr->action == PF_BINAT) && 5940 (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) 5941 #ifdef __NetBSD__ 5942 pf_mtag->flags |= PF_TAG_TRANSLATE_LOCALHOST; 5943 #else 5944 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST; 5945 #endif /* !__NetBSD__ */ 5946 5947 if (log) { 5948 struct pf_rule *lr; 5949 5950 if (s != NULL && s->nat_rule.ptr != NULL && 5951 s->nat_rule.ptr->log & PF_LOG_ALL) 5952 lr = s->nat_rule.ptr; 5953 else 5954 lr = r; 5955 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset, 5956 &pd); 5957 } 5958 5959 kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len; 5960 kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++; 5961 5962 if (action == PF_PASS || r->action == PF_DROP) { 5963 dirndx = (dir == PF_OUT); 5964 r->packets[dirndx]++; 5965 r->bytes[dirndx] += pd.tot_len; 5966 if (a != NULL) { 5967 a->packets[dirndx]++; 5968 a->bytes[dirndx] += pd.tot_len; 5969 } 5970 if (s != NULL) { 5971 sk = s->state_key; 5972 if (s->nat_rule.ptr != NULL) { 5973 s->nat_rule.ptr->packets[dirndx]++; 5974 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len; 5975 } 5976 if (s->src_node != NULL) { 5977 s->src_node->packets[dirndx]++; 5978 s->src_node->bytes[dirndx] += pd.tot_len; 5979 } 5980 if (s->nat_src_node != NULL) { 5981 s->nat_src_node->packets[dirndx]++; 5982 s->nat_src_node->bytes[dirndx] += pd.tot_len; 5983 } 5984 dirndx = (dir == sk->direction) ? 0 : 1; 5985 s->packets[dirndx]++; 5986 s->bytes[dirndx] += pd.tot_len; 5987 } 5988 tr = r; 5989 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule; 5990 if (nr != NULL) { 5991 struct pf_addr *x; 5992 /* 5993 * XXX: we need to make sure that the addresses 5994 * passed to pfr_update_stats() are the same than 5995 * the addresses used during matching (pfr_match) 5996 */ 5997 if (r == &pf_default_rule) { 5998 tr = nr; 5999 x = (sk == NULL || sk->direction == dir) ? 6000 &pd.baddr : &pd.naddr; 6001 } else 6002 x = (sk == NULL || sk->direction == dir) ? 6003 &pd.naddr : &pd.baddr; 6004 if (x == &pd.baddr || s == NULL) { 6005 /* we need to change the address */ 6006 if (dir == PF_OUT) 6007 pd.src = x; 6008 else 6009 pd.dst = x; 6010 } 6011 } 6012 if (tr->src.addr.type == PF_ADDR_TABLE) 6013 pfr_update_stats(tr->src.addr.p.tbl, (sk == NULL || 6014 sk->direction == dir) ? 6015 pd.src : pd.dst, pd.af, 6016 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 6017 tr->src.neg); 6018 if (tr->dst.addr.type == PF_ADDR_TABLE) 6019 pfr_update_stats(tr->dst.addr.p.tbl, (sk == NULL || 6020 sk->direction == dir) ? pd.dst : pd.src, pd.af, 6021 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 6022 tr->dst.neg); 6023 } 6024 6025 6026 if (action == PF_SYNPROXY_DROP) { 6027 m_freem(*m0); 6028 *m0 = NULL; 6029 action = PF_PASS; 6030 } else if (r->rt) 6031 /* pf_route can free the mbuf causing *m0 to become NULL */ 6032 pf_route(m0, r, dir, kif->pfik_ifp, s, &pd); 6033 6034 return (action); 6035 } 6036 #endif /* INET */ 6037 6038 #ifdef INET6 6039 int 6040 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0, 6041 struct ether_header *eh) 6042 { 6043 struct pfi_kif *kif; 6044 u_short action, reason = 0, log = 0; 6045 struct mbuf *m = *m0, *n = NULL; 6046 struct ip6_hdr *h = NULL; /* XXX gcc */ 6047 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr; 6048 struct pf_state *s = NULL; 6049 struct pf_state_key *sk = NULL; 6050 struct pf_ruleset *ruleset = NULL; 6051 struct pf_pdesc pd; 6052 int off, terminal = 0, dirndx, rh_cnt = 0; 6053 #ifdef __NetBSD__ 6054 struct pf_mtag *pf_mtag = NULL; /* XXX gcc */ 6055 #endif /* __NetBSD__ */ 6056 6057 if (!pf_status.running) 6058 return (PF_PASS); 6059 6060 memset(&pd, 0, sizeof(pd)); 6061 if (ifp->if_type == IFT_CARP && ifp->if_carpdev) 6062 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif; 6063 else 6064 kif = (struct pfi_kif *)ifp->if_pf_kif; 6065 6066 if (kif == NULL) { 6067 DPFPRINTF(PF_DEBUG_URGENT, 6068 ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname)); 6069 return (PF_DROP); 6070 } 6071 if (kif->pfik_flags & PFI_IFLAG_SKIP) 6072 return (PF_PASS); 6073 6074 #ifdef DIAGNOSTIC 6075 if ((m->m_flags & M_PKTHDR) == 0) 6076 panic("non-M_PKTHDR is passed to pf_test6"); 6077 #endif /* DIAGNOSTIC */ 6078 6079 if (m->m_pkthdr.len < (int)sizeof(*h)) { 6080 action = PF_DROP; 6081 REASON_SET(&reason, PFRES_SHORT); 6082 log = 1; 6083 goto done; 6084 } 6085 6086 #ifdef __NetBSD__ 6087 if ((pf_mtag = pf_get_mtag(m)) == NULL) { 6088 DPFPRINTF(PF_DEBUG_URGENT, 6089 ("pf_test6: pf_get_mtag returned NULL\n")); 6090 return (PF_DROP); 6091 } 6092 if (pf_mtag->flags & PF_TAG_GENERATED) 6093 return (PF_PASS); 6094 #else 6095 if (m->m_pkthdr.pf.flags & PF_TAG_GENERATED) 6096 return (PF_PASS); 6097 #endif /* !__NetBSD__ */ 6098 6099 /* We do IP header normalization and packet reassembly here */ 6100 if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) { 6101 action = PF_DROP; 6102 goto done; 6103 } 6104 m = *m0; /* pf_normalize messes with m0 */ 6105 h = mtod(m, struct ip6_hdr *); 6106 6107 #if 1 6108 /* 6109 * we do not support jumbogram yet. if we keep going, zero ip6_plen 6110 * will do something bad, so drop the packet for now. 6111 */ 6112 if (htons(h->ip6_plen) == 0) { 6113 action = PF_DROP; 6114 REASON_SET(&reason, PFRES_NORM); /*XXX*/ 6115 goto done; 6116 } 6117 #endif 6118 6119 pd.src = (struct pf_addr *)&h->ip6_src; 6120 pd.dst = (struct pf_addr *)&h->ip6_dst; 6121 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6); 6122 pd.ip_sum = NULL; 6123 pd.af = AF_INET6; 6124 pd.tos = 0; 6125 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr); 6126 pd.eh = eh; 6127 6128 off = ((char *)h - m->m_data) + sizeof(struct ip6_hdr); 6129 pd.proto = h->ip6_nxt; 6130 do { 6131 switch (pd.proto) { 6132 case IPPROTO_FRAGMENT: 6133 action = pf_test_fragment(&r, dir, kif, m, h, 6134 &pd, &a, &ruleset); 6135 if (action == PF_DROP) 6136 REASON_SET(&reason, PFRES_FRAG); 6137 goto done; 6138 case IPPROTO_ROUTING: { 6139 struct ip6_rthdr rthdr; 6140 6141 if (rh_cnt++) { 6142 DPFPRINTF(PF_DEBUG_MISC, 6143 ("pf: IPv6 more than one rthdr\n")); 6144 action = PF_DROP; 6145 REASON_SET(&reason, PFRES_IPOPTIONS); 6146 log = 1; 6147 goto done; 6148 } 6149 if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL, 6150 &reason, pd.af)) { 6151 DPFPRINTF(PF_DEBUG_MISC, 6152 ("pf: IPv6 short rthdr\n")); 6153 action = PF_DROP; 6154 REASON_SET(&reason, PFRES_SHORT); 6155 log = 1; 6156 goto done; 6157 } 6158 if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) { 6159 DPFPRINTF(PF_DEBUG_MISC, 6160 ("pf: IPv6 rthdr0\n")); 6161 action = PF_DROP; 6162 REASON_SET(&reason, PFRES_IPOPTIONS); 6163 log = 1; 6164 goto done; 6165 } 6166 /* FALLTHROUGH */ 6167 } 6168 case IPPROTO_AH: 6169 case IPPROTO_HOPOPTS: 6170 case IPPROTO_DSTOPTS: { 6171 /* get next header and header length */ 6172 struct ip6_ext opt6; 6173 6174 if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6), 6175 NULL, &reason, pd.af)) { 6176 DPFPRINTF(PF_DEBUG_MISC, 6177 ("pf: IPv6 short opt\n")); 6178 action = PF_DROP; 6179 log = 1; 6180 goto done; 6181 } 6182 if (pd.proto == IPPROTO_AH) 6183 off += (opt6.ip6e_len + 2) * 4; 6184 else 6185 off += (opt6.ip6e_len + 1) * 8; 6186 pd.proto = opt6.ip6e_nxt; 6187 /* goto the next header */ 6188 break; 6189 } 6190 default: 6191 terminal++; 6192 break; 6193 } 6194 } while (!terminal); 6195 6196 /* if there's no routing header, use unmodified mbuf for checksumming */ 6197 if (!n) 6198 n = m; 6199 6200 switch (pd.proto) { 6201 6202 case IPPROTO_TCP: { 6203 struct tcphdr th; 6204 6205 pd.hdr.tcp = &th; 6206 if (!pf_pull_hdr(m, off, &th, sizeof(th), 6207 &action, &reason, AF_INET6)) { 6208 log = action != PF_PASS; 6209 goto done; 6210 } 6211 pd.p_len = pd.tot_len - off - (th.th_off << 2); 6212 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd); 6213 if (action == PF_DROP) 6214 goto done; 6215 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd, 6216 &reason); 6217 if (action == PF_PASS) { 6218 #if NPFSYNC 6219 pfsync_update_state(s); 6220 #endif /* NPFSYNC */ 6221 r = s->rule.ptr; 6222 a = s->anchor.ptr; 6223 log = s->log; 6224 } else if (s == NULL) 6225 action = pf_test_rule(&r, &s, dir, kif, 6226 m, off, h, &pd, &a, &ruleset, &ip6intrq); 6227 break; 6228 } 6229 6230 case IPPROTO_UDP: { 6231 struct udphdr uh; 6232 6233 pd.hdr.udp = &uh; 6234 if (!pf_pull_hdr(m, off, &uh, sizeof(uh), 6235 &action, &reason, AF_INET6)) { 6236 log = action != PF_PASS; 6237 goto done; 6238 } 6239 if (uh.uh_dport == 0 || 6240 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off || 6241 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) { 6242 action = PF_DROP; 6243 REASON_SET(&reason, PFRES_SHORT); 6244 goto done; 6245 } 6246 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd); 6247 if (action == PF_PASS) { 6248 #if NPFSYNC 6249 pfsync_update_state(s); 6250 #endif /* NPFSYNC */ 6251 r = s->rule.ptr; 6252 a = s->anchor.ptr; 6253 log = s->log; 6254 } else if (s == NULL) 6255 action = pf_test_rule(&r, &s, dir, kif, 6256 m, off, h, &pd, &a, &ruleset, &ip6intrq); 6257 break; 6258 } 6259 6260 #ifdef INET 6261 case IPPROTO_ICMP: { 6262 action = PF_DROP; 6263 DPFPRINTF(PF_DEBUG_MISC, 6264 ("pf: dropping IPv6 packet with ICMPv4 payload\n")); 6265 goto done; 6266 } 6267 #endif 6268 6269 case IPPROTO_ICMPV6: { 6270 struct icmp6_hdr ih; 6271 6272 pd.hdr.icmp6 = &ih; 6273 if (!pf_pull_hdr(m, off, &ih, sizeof(ih), 6274 &action, &reason, AF_INET6)) { 6275 log = action != PF_PASS; 6276 goto done; 6277 } 6278 action = pf_test_state_icmp(&s, dir, kif, 6279 m, off, h, &pd, &reason); 6280 if (action == PF_PASS) { 6281 #if NPFSYNC 6282 pfsync_update_state(s); 6283 #endif /* NPFSYNC */ 6284 r = s->rule.ptr; 6285 a = s->anchor.ptr; 6286 log = s->log; 6287 } else if (s == NULL) 6288 action = pf_test_rule(&r, &s, dir, kif, 6289 m, off, h, &pd, &a, &ruleset, &ip6intrq); 6290 break; 6291 } 6292 6293 default: 6294 action = pf_test_state_other(&s, dir, kif, &pd); 6295 if (action == PF_PASS) { 6296 #if NPFSYNC 6297 pfsync_update_state(s); 6298 #endif /* NPFSYNC */ 6299 r = s->rule.ptr; 6300 a = s->anchor.ptr; 6301 log = s->log; 6302 } else if (s == NULL) 6303 action = pf_test_rule(&r, &s, dir, kif, m, off, h, 6304 &pd, &a, &ruleset, &ip6intrq); 6305 break; 6306 } 6307 6308 done: 6309 if (n != m) { 6310 m_freem(n); 6311 n = NULL; 6312 } 6313 6314 /* handle dangerous IPv6 extension headers. */ 6315 if (action == PF_PASS && rh_cnt && 6316 !((s && s->allow_opts) || r->allow_opts)) { 6317 action = PF_DROP; 6318 REASON_SET(&reason, PFRES_IPOPTIONS); 6319 log = 1; 6320 DPFPRINTF(PF_DEBUG_MISC, 6321 ("pf: dropping packet with dangerous v6 headers\n")); 6322 } 6323 6324 if ((s && s->tag) || r->rtableid) 6325 pf_tag_packet(m, s ? s->tag : 0, r->rtableid); 6326 6327 #ifdef ALTQ 6328 if (action == PF_PASS && r->qid) { 6329 #ifdef __NetBSD__ 6330 struct m_tag *mtag; 6331 struct altq_tag *atag; 6332 6333 mtag = m_tag_get(PACKET_TAG_ALTQ_QID, sizeof(*atag), M_NOWAIT); 6334 if (mtag != NULL) { 6335 atag = (struct altq_tag *)(mtag + 1); 6336 if (pd.tos & IPTOS_LOWDELAY) 6337 atag->qid = r->pqid; 6338 else 6339 atag->qid = r->qid; 6340 /* add hints for ecn */ 6341 atag->af = AF_INET6; 6342 atag->hdr = h; 6343 m_tag_prepend(m, mtag); 6344 } 6345 #else 6346 if (pd.tos & IPTOS_LOWDELAY) 6347 m->m_pkthdr.pf.qid = r->pqid; 6348 else 6349 m->m_pkthdr.pf.qid = r->qid; 6350 /* add hints for ecn */ 6351 m->m_pkthdr.pf.hdr = h; 6352 #endif /* !__NetBSD__ */ 6353 } 6354 #endif /* ALTQ */ 6355 6356 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP || 6357 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL && 6358 (s->nat_rule.ptr->action == PF_RDR || 6359 s->nat_rule.ptr->action == PF_BINAT) && 6360 IN6_IS_ADDR_LOOPBACK(&pd.dst->v6)) 6361 #ifdef __NetBSD__ 6362 pf_mtag->flags |= PF_TAG_TRANSLATE_LOCALHOST; 6363 #else 6364 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST; 6365 #endif /* !__NetBSD__ */ 6366 6367 if (log) { 6368 struct pf_rule *lr; 6369 6370 if (s != NULL && s->nat_rule.ptr != NULL && 6371 s->nat_rule.ptr->log & PF_LOG_ALL) 6372 lr = s->nat_rule.ptr; 6373 else 6374 lr = r; 6375 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset, 6376 &pd); 6377 } 6378 6379 kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len; 6380 kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++; 6381 6382 if (action == PF_PASS || r->action == PF_DROP) { 6383 dirndx = (dir == PF_OUT); 6384 r->packets[dirndx]++; 6385 r->bytes[dirndx] += pd.tot_len; 6386 if (a != NULL) { 6387 a->packets[dirndx]++; 6388 a->bytes[dirndx] += pd.tot_len; 6389 } 6390 if (s != NULL) { 6391 sk = s->state_key; 6392 if (s->nat_rule.ptr != NULL) { 6393 s->nat_rule.ptr->packets[dirndx]++; 6394 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len; 6395 } 6396 if (s->src_node != NULL) { 6397 s->src_node->packets[dirndx]++; 6398 s->src_node->bytes[dirndx] += pd.tot_len; 6399 } 6400 if (s->nat_src_node != NULL) { 6401 s->nat_src_node->packets[dirndx]++; 6402 s->nat_src_node->bytes[dirndx] += pd.tot_len; 6403 } 6404 dirndx = (dir == sk->direction) ? 0 : 1; 6405 s->packets[dirndx]++; 6406 s->bytes[dirndx] += pd.tot_len; 6407 } 6408 tr = r; 6409 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule; 6410 if (nr != NULL) { 6411 struct pf_addr *x; 6412 /* 6413 * XXX: we need to make sure that the addresses 6414 * passed to pfr_update_stats() are the same than 6415 * the addresses used during matching (pfr_match) 6416 */ 6417 if (r == &pf_default_rule) { 6418 tr = nr; 6419 x = (s == NULL || sk->direction == dir) ? 6420 &pd.baddr : &pd.naddr; 6421 } else { 6422 x = (s == NULL || sk->direction == dir) ? 6423 &pd.naddr : &pd.baddr; 6424 } 6425 if (x == &pd.baddr || s == NULL) { 6426 if (dir == PF_OUT) 6427 pd.src = x; 6428 else 6429 pd.dst = x; 6430 } 6431 } 6432 if (tr->src.addr.type == PF_ADDR_TABLE) 6433 pfr_update_stats(tr->src.addr.p.tbl, (sk == NULL || 6434 sk->direction == dir) ? pd.src : pd.dst, pd.af, 6435 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 6436 tr->src.neg); 6437 if (tr->dst.addr.type == PF_ADDR_TABLE) 6438 pfr_update_stats(tr->dst.addr.p.tbl, (sk == NULL || 6439 sk->direction == dir) ? pd.dst : pd.src, pd.af, 6440 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 6441 tr->dst.neg); 6442 } 6443 6444 6445 if (action == PF_SYNPROXY_DROP) { 6446 m_freem(*m0); 6447 *m0 = NULL; 6448 action = PF_PASS; 6449 } else if (r->rt) 6450 /* pf_route6 can free the mbuf causing *m0 to become NULL */ 6451 pf_route6(m0, r, dir, kif->pfik_ifp, s, &pd); 6452 6453 return (action); 6454 } 6455 #endif /* INET6 */ 6456 6457 int 6458 pf_check_congestion(struct ifqueue *ifq) 6459 { 6460 #ifdef __NetBSD__ 6461 return (0); 6462 #else 6463 if (ifq->ifq_congestion) 6464 return (1); 6465 else 6466 return (0); 6467 #endif /* !__NetBSD__ */ 6468 } 6469