1 /* $NetBSD: pf.c,v 1.27 2006/10/12 19:59:08 peter Exp $ */ 2 /* $OpenBSD: pf.c,v 1.487 2005/04/22 09:53:18 dhartmei 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 "bpfilter.h" 40 #include "pflog.h" 41 #ifdef __OpenBSD__ 42 #include "pfsync.h" 43 #else 44 #define NPFSYNC 0 45 #endif 46 47 #include <sys/param.h> 48 #include <sys/systm.h> 49 #include <sys/mbuf.h> 50 #include <sys/filio.h> 51 #include <sys/socket.h> 52 #include <sys/socketvar.h> 53 #include <sys/kernel.h> 54 #include <sys/time.h> 55 #include <sys/pool.h> 56 #ifdef __NetBSD__ 57 #include <sys/endian.h> 58 #endif 59 60 #include <net/if.h> 61 #include <net/if_types.h> 62 #include <net/bpf.h> 63 #include <net/route.h> 64 65 #include <netinet/in.h> 66 #include <netinet/in_var.h> 67 #include <netinet/in_systm.h> 68 #include <netinet/ip.h> 69 #include <netinet/ip_var.h> 70 #include <netinet/tcp.h> 71 #include <netinet/tcp_seq.h> 72 #include <netinet/udp.h> 73 #include <netinet/ip_icmp.h> 74 #include <netinet/in_pcb.h> 75 #include <netinet/tcp_timer.h> 76 #include <netinet/tcp_var.h> 77 #include <netinet/udp_var.h> 78 #include <netinet/icmp_var.h> 79 #ifdef __OpenBSD__ 80 #include <netinet/if_ether.h> 81 #else 82 #include <net/if_ether.h> 83 #endif 84 85 #ifdef __OpenBSD__ 86 #include <dev/rndvar.h> 87 #else 88 #include <sys/rnd.h> 89 #endif 90 #include <net/pfvar.h> 91 #include <net/if_pflog.h> 92 93 #if NPFSYNC > 0 94 #include <net/if_pfsync.h> 95 #endif /* NPFSYNC > 0 */ 96 97 #ifdef INET6 98 #include <netinet/ip6.h> 99 #include <netinet6/ip6_var.h> 100 #ifdef __NetBSD__ 101 #include <netinet6/in6_pcb.h> 102 #endif 103 #include <netinet/icmp6.h> 104 #include <netinet6/nd6.h> 105 #endif /* INET6 */ 106 107 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x 108 109 /* 110 * Global variables 111 */ 112 113 struct pf_anchor_global pf_anchors; 114 struct pf_ruleset pf_main_ruleset; 115 struct pf_altqqueue pf_altqs[2]; 116 struct pf_palist pf_pabuf; 117 struct pf_altqqueue *pf_altqs_active; 118 struct pf_altqqueue *pf_altqs_inactive; 119 struct pf_status pf_status; 120 121 u_int32_t ticket_altqs_active; 122 u_int32_t ticket_altqs_inactive; 123 int altqs_inactive_open; 124 u_int32_t ticket_pabuf; 125 126 #ifdef __OpenBSD__ 127 struct timeout pf_expire_to; /* expire timeout */ 128 #else 129 struct callout pf_expire_to; /* expire timeout */ 130 #endif 131 132 struct pf_anchor_stackframe { 133 struct pf_ruleset *rs; 134 struct pf_rule *r; 135 struct pf_anchor_node *parent; 136 struct pf_anchor *child; 137 } pf_anchor_stack[64]; 138 139 struct pool pf_src_tree_pl, pf_rule_pl; 140 struct pool pf_state_pl, pf_altq_pl, pf_pooladdr_pl; 141 142 void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t); 143 144 void pf_init_threshold(struct pf_threshold *, u_int32_t, 145 u_int32_t); 146 void pf_add_threshold(struct pf_threshold *); 147 int pf_check_threshold(struct pf_threshold *); 148 149 void pf_change_ap(struct pf_addr *, u_int16_t *, 150 u_int16_t *, u_int16_t *, struct pf_addr *, 151 u_int16_t, u_int8_t, sa_family_t); 152 #ifdef INET6 153 void pf_change_a6(struct pf_addr *, u_int16_t *, 154 struct pf_addr *, u_int8_t); 155 #endif /* INET6 */ 156 void pf_change_icmp(struct pf_addr *, u_int16_t *, 157 struct pf_addr *, struct pf_addr *, u_int16_t, 158 u_int16_t *, u_int16_t *, u_int16_t *, 159 u_int16_t *, u_int8_t, sa_family_t); 160 void pf_send_tcp(const struct pf_rule *, sa_family_t, 161 const struct pf_addr *, const struct pf_addr *, 162 u_int16_t, u_int16_t, u_int32_t, u_int32_t, 163 u_int8_t, u_int16_t, u_int16_t, u_int8_t, int, 164 u_int16_t, struct ether_header *, struct ifnet *); 165 void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t, 166 sa_family_t, struct pf_rule *); 167 struct pf_rule *pf_match_translation(struct pf_pdesc *, struct mbuf *, 168 int, int, struct pfi_kif *, 169 struct pf_addr *, u_int16_t, struct pf_addr *, 170 u_int16_t, int); 171 struct pf_rule *pf_get_translation(struct pf_pdesc *, struct mbuf *, 172 int, int, struct pfi_kif *, struct pf_src_node **, 173 struct pf_addr *, u_int16_t, 174 struct pf_addr *, u_int16_t, 175 struct pf_addr *, u_int16_t *); 176 int pf_test_tcp(struct pf_rule **, struct pf_state **, 177 int, struct pfi_kif *, struct mbuf *, int, 178 void *, struct pf_pdesc *, struct pf_rule **, 179 struct pf_ruleset **, struct ifqueue *); 180 int pf_test_udp(struct pf_rule **, struct pf_state **, 181 int, struct pfi_kif *, struct mbuf *, int, 182 void *, struct pf_pdesc *, struct pf_rule **, 183 struct pf_ruleset **, struct ifqueue *); 184 int pf_test_icmp(struct pf_rule **, struct pf_state **, 185 int, struct pfi_kif *, struct mbuf *, int, 186 void *, struct pf_pdesc *, struct pf_rule **, 187 struct pf_ruleset **, struct ifqueue *); 188 int pf_test_other(struct pf_rule **, struct pf_state **, 189 int, struct pfi_kif *, struct mbuf *, int, void *, 190 struct pf_pdesc *, struct pf_rule **, 191 struct pf_ruleset **, struct ifqueue *); 192 int pf_test_fragment(struct pf_rule **, int, 193 struct pfi_kif *, struct mbuf *, void *, 194 struct pf_pdesc *, struct pf_rule **, 195 struct pf_ruleset **); 196 int pf_test_state_tcp(struct pf_state **, int, 197 struct pfi_kif *, struct mbuf *, int, 198 void *, struct pf_pdesc *, u_short *); 199 int pf_test_state_udp(struct pf_state **, int, 200 struct pfi_kif *, struct mbuf *, int, 201 void *, struct pf_pdesc *); 202 int pf_test_state_icmp(struct pf_state **, int, 203 struct pfi_kif *, struct mbuf *, int, 204 void *, struct pf_pdesc *, u_short *); 205 int pf_test_state_other(struct pf_state **, int, 206 struct pfi_kif *, struct pf_pdesc *); 207 struct pf_tag *pf_get_tag(struct mbuf *); 208 int pf_match_tag(struct mbuf *, struct pf_rule *, 209 struct pf_tag **, int *); 210 void pf_hash(struct pf_addr *, struct pf_addr *, 211 struct pf_poolhashkey *, sa_family_t); 212 int pf_map_addr(u_int8_t, struct pf_rule *, 213 struct pf_addr *, struct pf_addr *, 214 struct pf_addr *, struct pf_src_node **); 215 int pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *, 216 struct pf_addr *, struct pf_addr *, u_int16_t, 217 struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t, 218 struct pf_src_node **); 219 void pf_route(struct mbuf **, struct pf_rule *, int, 220 struct ifnet *, struct pf_state *); 221 void pf_route6(struct mbuf **, struct pf_rule *, int, 222 struct ifnet *, struct pf_state *); 223 int pf_socket_lookup(uid_t *, gid_t *, 224 int, struct pf_pdesc *); 225 u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t, 226 sa_family_t); 227 u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t, 228 sa_family_t); 229 u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t, 230 u_int16_t); 231 void pf_set_rt_ifp(struct pf_state *, 232 struct pf_addr *); 233 int pf_check_proto_cksum(struct mbuf *, int, int, 234 u_int8_t, sa_family_t); 235 int pf_addr_wrap_neq(struct pf_addr_wrap *, 236 struct pf_addr_wrap *); 237 static int pf_add_mbuf_tag(struct mbuf *, u_int); 238 struct pf_state *pf_find_state_recurse(struct pfi_kif *, 239 struct pf_state *, u_int8_t); 240 int pf_src_connlimit(struct pf_state **); 241 int pf_check_congestion(struct ifqueue *); 242 243 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = { 244 { &pf_state_pl, PFSTATE_HIWAT }, 245 { &pf_src_tree_pl, PFSNODE_HIWAT }, 246 { &pf_frent_pl, PFFRAG_FRENT_HIWAT } 247 }; 248 249 #define STATE_LOOKUP() \ 250 do { \ 251 if (direction == PF_IN) \ 252 *state = pf_find_state_recurse( \ 253 kif, &key, PF_EXT_GWY); \ 254 else \ 255 *state = pf_find_state_recurse( \ 256 kif, &key, PF_LAN_EXT); \ 257 if (*state == NULL || (*state)->timeout == PFTM_PURGE) \ 258 return (PF_DROP); \ 259 if (direction == PF_OUT && \ 260 (((*state)->rule.ptr->rt == PF_ROUTETO && \ 261 (*state)->rule.ptr->direction == PF_OUT) || \ 262 ((*state)->rule.ptr->rt == PF_REPLYTO && \ 263 (*state)->rule.ptr->direction == PF_IN)) && \ 264 (*state)->rt_kif != NULL && \ 265 (*state)->rt_kif != kif) \ 266 return (PF_PASS); \ 267 } while (0) 268 269 #define STATE_TRANSLATE(s) \ 270 (s)->lan.addr.addr32[0] != (s)->gwy.addr.addr32[0] || \ 271 ((s)->af == AF_INET6 && \ 272 ((s)->lan.addr.addr32[1] != (s)->gwy.addr.addr32[1] || \ 273 (s)->lan.addr.addr32[2] != (s)->gwy.addr.addr32[2] || \ 274 (s)->lan.addr.addr32[3] != (s)->gwy.addr.addr32[3])) || \ 275 (s)->lan.port != (s)->gwy.port 276 277 #define BOUND_IFACE(r, k) (((r)->rule_flag & PFRULE_IFBOUND) ? (k) : \ 278 ((r)->rule_flag & PFRULE_GRBOUND) ? (k)->pfik_parent : \ 279 (k)->pfik_parent->pfik_parent) 280 281 #define STATE_INC_COUNTERS(s) \ 282 do { \ 283 s->rule.ptr->states++; \ 284 if (s->anchor.ptr != NULL) \ 285 s->anchor.ptr->states++; \ 286 if (s->nat_rule.ptr != NULL) \ 287 s->nat_rule.ptr->states++; \ 288 } while (0) 289 290 #define STATE_DEC_COUNTERS(s) \ 291 do { \ 292 if (s->nat_rule.ptr != NULL) \ 293 s->nat_rule.ptr->states--; \ 294 if (s->anchor.ptr != NULL) \ 295 s->anchor.ptr->states--; \ 296 s->rule.ptr->states--; \ 297 } while (0) 298 299 static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *); 300 static __inline int pf_state_compare_lan_ext(struct pf_state *, 301 struct pf_state *); 302 static __inline int pf_state_compare_ext_gwy(struct pf_state *, 303 struct pf_state *); 304 static __inline int pf_state_compare_id(struct pf_state *, 305 struct pf_state *); 306 static __inline int pf_anchor_compare(struct pf_anchor *, struct pf_anchor *); 307 308 struct pf_src_tree tree_src_tracking; 309 310 struct pf_state_tree_id tree_id; 311 struct pf_state_queue state_updates; 312 313 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare); 314 RB_GENERATE(pf_state_tree_lan_ext, pf_state, 315 u.s.entry_lan_ext, pf_state_compare_lan_ext); 316 RB_GENERATE(pf_state_tree_ext_gwy, pf_state, 317 u.s.entry_ext_gwy, pf_state_compare_ext_gwy); 318 RB_GENERATE(pf_state_tree_id, pf_state, 319 u.s.entry_id, pf_state_compare_id); 320 RB_GENERATE(pf_anchor_global, pf_anchor, entry_global, pf_anchor_compare); 321 RB_GENERATE(pf_anchor_node, pf_anchor, entry_node, pf_anchor_compare); 322 323 static __inline int 324 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b) 325 { 326 int diff; 327 328 if (a->rule.ptr > b->rule.ptr) 329 return (1); 330 if (a->rule.ptr < b->rule.ptr) 331 return (-1); 332 if ((diff = a->af - b->af) != 0) 333 return (diff); 334 switch (a->af) { 335 #ifdef INET 336 case AF_INET: 337 if (a->addr.addr32[0] > b->addr.addr32[0]) 338 return (1); 339 if (a->addr.addr32[0] < b->addr.addr32[0]) 340 return (-1); 341 break; 342 #endif /* INET */ 343 #ifdef INET6 344 case AF_INET6: 345 if (a->addr.addr32[3] > b->addr.addr32[3]) 346 return (1); 347 if (a->addr.addr32[3] < b->addr.addr32[3]) 348 return (-1); 349 if (a->addr.addr32[2] > b->addr.addr32[2]) 350 return (1); 351 if (a->addr.addr32[2] < b->addr.addr32[2]) 352 return (-1); 353 if (a->addr.addr32[1] > b->addr.addr32[1]) 354 return (1); 355 if (a->addr.addr32[1] < b->addr.addr32[1]) 356 return (-1); 357 if (a->addr.addr32[0] > b->addr.addr32[0]) 358 return (1); 359 if (a->addr.addr32[0] < b->addr.addr32[0]) 360 return (-1); 361 break; 362 #endif /* INET6 */ 363 } 364 return (0); 365 } 366 367 static __inline int 368 pf_state_compare_lan_ext(struct pf_state *a, struct pf_state *b) 369 { 370 int diff; 371 372 if ((diff = a->proto - b->proto) != 0) 373 return (diff); 374 if ((diff = a->af - b->af) != 0) 375 return (diff); 376 switch (a->af) { 377 #ifdef INET 378 case AF_INET: 379 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0]) 380 return (1); 381 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0]) 382 return (-1); 383 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 384 return (1); 385 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 386 return (-1); 387 break; 388 #endif /* INET */ 389 #ifdef INET6 390 case AF_INET6: 391 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3]) 392 return (1); 393 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3]) 394 return (-1); 395 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3]) 396 return (1); 397 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3]) 398 return (-1); 399 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2]) 400 return (1); 401 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2]) 402 return (-1); 403 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2]) 404 return (1); 405 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2]) 406 return (-1); 407 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1]) 408 return (1); 409 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1]) 410 return (-1); 411 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1]) 412 return (1); 413 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1]) 414 return (-1); 415 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0]) 416 return (1); 417 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0]) 418 return (-1); 419 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 420 return (1); 421 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 422 return (-1); 423 break; 424 #endif /* INET6 */ 425 } 426 427 if ((diff = a->lan.port - b->lan.port) != 0) 428 return (diff); 429 if ((diff = a->ext.port - b->ext.port) != 0) 430 return (diff); 431 432 return (0); 433 } 434 435 static __inline int 436 pf_state_compare_ext_gwy(struct pf_state *a, struct pf_state *b) 437 { 438 int diff; 439 440 if ((diff = a->proto - b->proto) != 0) 441 return (diff); 442 if ((diff = a->af - b->af) != 0) 443 return (diff); 444 switch (a->af) { 445 #ifdef INET 446 case AF_INET: 447 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 448 return (1); 449 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 450 return (-1); 451 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0]) 452 return (1); 453 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0]) 454 return (-1); 455 break; 456 #endif /* INET */ 457 #ifdef INET6 458 case AF_INET6: 459 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3]) 460 return (1); 461 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3]) 462 return (-1); 463 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3]) 464 return (1); 465 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3]) 466 return (-1); 467 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2]) 468 return (1); 469 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2]) 470 return (-1); 471 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2]) 472 return (1); 473 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2]) 474 return (-1); 475 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1]) 476 return (1); 477 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1]) 478 return (-1); 479 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1]) 480 return (1); 481 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1]) 482 return (-1); 483 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0]) 484 return (1); 485 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0]) 486 return (-1); 487 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0]) 488 return (1); 489 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0]) 490 return (-1); 491 break; 492 #endif /* INET6 */ 493 } 494 495 if ((diff = a->ext.port - b->ext.port) != 0) 496 return (diff); 497 if ((diff = a->gwy.port - b->gwy.port) != 0) 498 return (diff); 499 500 return (0); 501 } 502 503 static __inline int 504 pf_state_compare_id(struct pf_state *a, struct pf_state *b) 505 { 506 if (a->id > b->id) 507 return (1); 508 if (a->id < b->id) 509 return (-1); 510 if (a->creatorid > b->creatorid) 511 return (1); 512 if (a->creatorid < b->creatorid) 513 return (-1); 514 515 return (0); 516 } 517 518 static __inline int 519 pf_anchor_compare(struct pf_anchor *a, struct pf_anchor *b) 520 { 521 int c = strcmp(a->path, b->path); 522 523 return (c ? (c < 0 ? -1 : 1) : 0); 524 } 525 526 #ifdef INET6 527 void 528 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af) 529 { 530 switch (af) { 531 #ifdef INET 532 case AF_INET: 533 dst->addr32[0] = src->addr32[0]; 534 break; 535 #endif /* INET */ 536 case AF_INET6: 537 dst->addr32[0] = src->addr32[0]; 538 dst->addr32[1] = src->addr32[1]; 539 dst->addr32[2] = src->addr32[2]; 540 dst->addr32[3] = src->addr32[3]; 541 break; 542 } 543 } 544 #endif /* INET6 */ 545 546 struct pf_state * 547 pf_find_state_byid(struct pf_state *key) 548 { 549 pf_status.fcounters[FCNT_STATE_SEARCH]++; 550 return (RB_FIND(pf_state_tree_id, &tree_id, key)); 551 } 552 553 struct pf_state * 554 pf_find_state_recurse(struct pfi_kif *kif, struct pf_state *key, u_int8_t tree) 555 { 556 struct pf_state *s; 557 558 pf_status.fcounters[FCNT_STATE_SEARCH]++; 559 560 switch (tree) { 561 case PF_LAN_EXT: 562 for (; kif != NULL; kif = kif->pfik_parent) { 563 s = RB_FIND(pf_state_tree_lan_ext, 564 &kif->pfik_lan_ext, key); 565 if (s != NULL) 566 return (s); 567 } 568 return (NULL); 569 case PF_EXT_GWY: 570 for (; kif != NULL; kif = kif->pfik_parent) { 571 s = RB_FIND(pf_state_tree_ext_gwy, 572 &kif->pfik_ext_gwy, key); 573 if (s != NULL) 574 return (s); 575 } 576 return (NULL); 577 default: 578 panic("pf_find_state_recurse"); 579 } 580 } 581 582 struct pf_state * 583 pf_find_state_all(struct pf_state *key, u_int8_t tree, int *more) 584 { 585 struct pf_state *s, *ss = NULL; 586 struct pfi_kif *kif; 587 588 pf_status.fcounters[FCNT_STATE_SEARCH]++; 589 590 switch (tree) { 591 case PF_LAN_EXT: 592 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) { 593 s = RB_FIND(pf_state_tree_lan_ext, 594 &kif->pfik_lan_ext, key); 595 if (s == NULL) 596 continue; 597 if (more == NULL) 598 return (s); 599 ss = s; 600 (*more)++; 601 } 602 return (ss); 603 case PF_EXT_GWY: 604 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) { 605 s = RB_FIND(pf_state_tree_ext_gwy, 606 &kif->pfik_ext_gwy, key); 607 if (s == NULL) 608 continue; 609 if (more == NULL) 610 return (s); 611 ss = s; 612 (*more)++; 613 } 614 return (ss); 615 default: 616 panic("pf_find_state_all"); 617 } 618 } 619 620 void 621 pf_init_threshold(struct pf_threshold *threshold, 622 u_int32_t limit, u_int32_t seconds) 623 { 624 threshold->limit = limit * PF_THRESHOLD_MULT; 625 threshold->seconds = seconds; 626 threshold->count = 0; 627 threshold->last = time_second; 628 } 629 630 void 631 pf_add_threshold(struct pf_threshold *threshold) 632 { 633 u_int32_t t = time_second, diff = t - threshold->last; 634 635 if (diff >= threshold->seconds) 636 threshold->count = 0; 637 else 638 threshold->count -= threshold->count * diff / 639 threshold->seconds; 640 threshold->count += PF_THRESHOLD_MULT; 641 threshold->last = t; 642 } 643 644 int 645 pf_check_threshold(struct pf_threshold *threshold) 646 { 647 return (threshold->count > threshold->limit); 648 } 649 650 int 651 pf_src_connlimit(struct pf_state **state) 652 { 653 struct pf_state *s; 654 int bad = 0; 655 656 (*state)->src_node->conn++; 657 pf_add_threshold(&(*state)->src_node->conn_rate); 658 659 if ((*state)->rule.ptr->max_src_conn && 660 (*state)->rule.ptr->max_src_conn < 661 (*state)->src_node->conn) { 662 pf_status.lcounters[LCNT_SRCCONN]++; 663 bad++; 664 } 665 666 if ((*state)->rule.ptr->max_src_conn_rate.limit && 667 pf_check_threshold(&(*state)->src_node->conn_rate)) { 668 pf_status.lcounters[LCNT_SRCCONNRATE]++; 669 bad++; 670 } 671 672 if (!bad) 673 return (0); 674 675 if ((*state)->rule.ptr->overload_tbl) { 676 struct pfr_addr p; 677 u_int32_t killed = 0; 678 679 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++; 680 if (pf_status.debug >= PF_DEBUG_MISC) { 681 printf("pf_src_connlimit: blocking address "); 682 pf_print_host(&(*state)->src_node->addr, 0, 683 (*state)->af); 684 } 685 686 bzero(&p, sizeof(p)); 687 p.pfra_af = (*state)->af; 688 switch ((*state)->af) { 689 #ifdef INET 690 case AF_INET: 691 p.pfra_net = 32; 692 p.pfra_ip4addr = (*state)->src_node->addr.v4; 693 break; 694 #endif /* INET */ 695 #ifdef INET6 696 case AF_INET6: 697 p.pfra_net = 128; 698 p.pfra_ip6addr = (*state)->src_node->addr.v6; 699 break; 700 #endif /* INET6 */ 701 } 702 703 pfr_insert_kentry((*state)->rule.ptr->overload_tbl, 704 &p, time_second); 705 706 /* kill existing states if that's required. */ 707 if ((*state)->rule.ptr->flush) { 708 pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++; 709 710 RB_FOREACH(s, pf_state_tree_id, &tree_id) { 711 /* 712 * Kill states from this source. (Only those 713 * from the same rule if PF_FLUSH_GLOBAL is not 714 * set) 715 */ 716 if (s->af == (*state)->af && 717 (((*state)->direction == PF_OUT && 718 PF_AEQ(&(*state)->src_node->addr, 719 &s->lan.addr, s->af)) || 720 ((*state)->direction == PF_IN && 721 PF_AEQ(&(*state)->src_node->addr, 722 &s->ext.addr, s->af))) && 723 ((*state)->rule.ptr->flush & 724 PF_FLUSH_GLOBAL || 725 (*state)->rule.ptr == s->rule.ptr)) { 726 s->timeout = PFTM_PURGE; 727 s->src.state = s->dst.state = 728 TCPS_CLOSED; 729 killed++; 730 } 731 } 732 if (pf_status.debug >= PF_DEBUG_MISC) 733 printf(", %u states killed", killed); 734 } 735 if (pf_status.debug >= PF_DEBUG_MISC) 736 printf("\n"); 737 } 738 739 /* kill this state */ 740 (*state)->timeout = PFTM_PURGE; 741 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED; 742 return (1); 743 } 744 745 int 746 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule, 747 struct pf_addr *src, sa_family_t af) 748 { 749 struct pf_src_node k; 750 751 if (*sn == NULL) { 752 k.af = af; 753 PF_ACPY(&k.addr, src, af); 754 if (rule->rule_flag & PFRULE_RULESRCTRACK || 755 rule->rpool.opts & PF_POOL_STICKYADDR) 756 k.rule.ptr = rule; 757 else 758 k.rule.ptr = NULL; 759 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++; 760 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k); 761 } 762 if (*sn == NULL) { 763 if (!rule->max_src_nodes || 764 rule->src_nodes < rule->max_src_nodes) 765 (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT); 766 else 767 pf_status.lcounters[LCNT_SRCNODES]++; 768 if ((*sn) == NULL) 769 return (-1); 770 bzero(*sn, sizeof(struct pf_src_node)); 771 772 pf_init_threshold(&(*sn)->conn_rate, 773 rule->max_src_conn_rate.limit, 774 rule->max_src_conn_rate.seconds); 775 776 (*sn)->af = af; 777 if (rule->rule_flag & PFRULE_RULESRCTRACK || 778 rule->rpool.opts & PF_POOL_STICKYADDR) 779 (*sn)->rule.ptr = rule; 780 else 781 (*sn)->rule.ptr = NULL; 782 PF_ACPY(&(*sn)->addr, src, af); 783 if (RB_INSERT(pf_src_tree, 784 &tree_src_tracking, *sn) != NULL) { 785 if (pf_status.debug >= PF_DEBUG_MISC) { 786 printf("pf: src_tree insert failed: "); 787 pf_print_host(&(*sn)->addr, 0, af); 788 printf("\n"); 789 } 790 pool_put(&pf_src_tree_pl, *sn); 791 return (-1); 792 } 793 (*sn)->creation = time_second; 794 (*sn)->ruletype = rule->action; 795 if ((*sn)->rule.ptr != NULL) 796 (*sn)->rule.ptr->src_nodes++; 797 pf_status.scounters[SCNT_SRC_NODE_INSERT]++; 798 pf_status.src_nodes++; 799 } else { 800 if (rule->max_src_states && 801 (*sn)->states >= rule->max_src_states) { 802 pf_status.lcounters[LCNT_SRCSTATES]++; 803 return (-1); 804 } 805 } 806 return (0); 807 } 808 809 int 810 pf_insert_state(struct pfi_kif *kif, struct pf_state *state) 811 { 812 /* Thou MUST NOT insert multiple duplicate keys */ 813 state->u.s.kif = kif; 814 if (RB_INSERT(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state)) { 815 if (pf_status.debug >= PF_DEBUG_MISC) { 816 printf("pf: state insert failed: tree_lan_ext"); 817 printf(" lan: "); 818 pf_print_host(&state->lan.addr, state->lan.port, 819 state->af); 820 printf(" gwy: "); 821 pf_print_host(&state->gwy.addr, state->gwy.port, 822 state->af); 823 printf(" ext: "); 824 pf_print_host(&state->ext.addr, state->ext.port, 825 state->af); 826 if (state->sync_flags & PFSTATE_FROMSYNC) 827 printf(" (from sync)"); 828 printf("\n"); 829 } 830 return (-1); 831 } 832 833 if (RB_INSERT(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state)) { 834 if (pf_status.debug >= PF_DEBUG_MISC) { 835 printf("pf: state insert failed: tree_ext_gwy"); 836 printf(" lan: "); 837 pf_print_host(&state->lan.addr, state->lan.port, 838 state->af); 839 printf(" gwy: "); 840 pf_print_host(&state->gwy.addr, state->gwy.port, 841 state->af); 842 printf(" ext: "); 843 pf_print_host(&state->ext.addr, state->ext.port, 844 state->af); 845 if (state->sync_flags & PFSTATE_FROMSYNC) 846 printf(" (from sync)"); 847 printf("\n"); 848 } 849 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state); 850 return (-1); 851 } 852 853 if (state->id == 0 && state->creatorid == 0) { 854 state->id = htobe64(pf_status.stateid++); 855 state->creatorid = pf_status.hostid; 856 } 857 if (RB_INSERT(pf_state_tree_id, &tree_id, state) != NULL) { 858 if (pf_status.debug >= PF_DEBUG_MISC) { 859 #ifdef __OpenBSD__ 860 printf("pf: state insert failed: " 861 "id: %016llx creatorid: %08x", 862 betoh64(state->id), ntohl(state->creatorid)); 863 #else 864 printf("pf: state insert failed: " 865 "id: %016" PRIx64 " creatorid: %08x", 866 be64toh(state->id), ntohl(state->creatorid)); 867 #endif 868 if (state->sync_flags & PFSTATE_FROMSYNC) 869 printf(" (from sync)"); 870 printf("\n"); 871 } 872 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state); 873 RB_REMOVE(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state); 874 return (-1); 875 } 876 TAILQ_INSERT_HEAD(&state_updates, state, u.s.entry_updates); 877 878 pf_status.fcounters[FCNT_STATE_INSERT]++; 879 pf_status.states++; 880 pfi_attach_state(kif); 881 #if NPFSYNC 882 pfsync_insert_state(state); 883 #endif 884 return (0); 885 } 886 887 void 888 pf_purge_timeout(void *arg) 889 { 890 #ifdef __OpenBSD__ 891 struct timeout *to = arg; 892 #else 893 struct callout *to = arg; 894 #endif 895 int s; 896 897 s = splsoftnet(); 898 pf_purge_expired_states(); 899 pf_purge_expired_fragments(); 900 pf_purge_expired_src_nodes(); 901 splx(s); 902 903 #ifdef __OpenBSD__ 904 timeout_add(to, pf_default_rule.timeout[PFTM_INTERVAL] * hz); 905 #else 906 callout_schedule(to, pf_default_rule.timeout[PFTM_INTERVAL] * hz); 907 #endif 908 } 909 910 u_int32_t 911 pf_state_expires(const struct pf_state *state) 912 { 913 u_int32_t timeout; 914 u_int32_t start; 915 u_int32_t end; 916 u_int32_t states; 917 918 /* handle all PFTM_* > PFTM_MAX here */ 919 if (state->timeout == PFTM_PURGE) 920 return (time_second); 921 if (state->timeout == PFTM_UNTIL_PACKET) 922 return (0); 923 KASSERT(state->timeout < PFTM_MAX); 924 timeout = state->rule.ptr->timeout[state->timeout]; 925 if (!timeout) 926 timeout = pf_default_rule.timeout[state->timeout]; 927 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START]; 928 if (start) { 929 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END]; 930 states = state->rule.ptr->states; 931 } else { 932 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START]; 933 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END]; 934 states = pf_status.states; 935 } 936 if (end && states > start && start < end) { 937 if (states < end) 938 return (state->expire + timeout * (end - states) / 939 (end - start)); 940 else 941 return (time_second); 942 } 943 return (state->expire + timeout); 944 } 945 946 void 947 pf_purge_expired_src_nodes(void) 948 { 949 struct pf_src_node *cur, *next; 950 951 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) { 952 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur); 953 954 if (cur->states <= 0 && cur->expire <= time_second) { 955 if (cur->rule.ptr != NULL) { 956 cur->rule.ptr->src_nodes--; 957 if (cur->rule.ptr->states <= 0 && 958 cur->rule.ptr->max_src_nodes <= 0) 959 pf_rm_rule(NULL, cur->rule.ptr); 960 } 961 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur); 962 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 963 pf_status.src_nodes--; 964 pool_put(&pf_src_tree_pl, cur); 965 } 966 } 967 } 968 969 void 970 pf_src_tree_remove_state(struct pf_state *s) 971 { 972 u_int32_t timeout; 973 974 if (s->src_node != NULL) { 975 if (s->proto == IPPROTO_TCP) { 976 if (s->src.state == PF_TCPS_PROXY_DST || 977 s->timeout >= PFTM_TCP_ESTABLISHED) 978 --s->src_node->conn; 979 } 980 if (--s->src_node->states <= 0) { 981 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE]; 982 if (!timeout) 983 timeout = 984 pf_default_rule.timeout[PFTM_SRC_NODE]; 985 s->src_node->expire = time_second + timeout; 986 } 987 } 988 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) { 989 if (--s->nat_src_node->states <= 0) { 990 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE]; 991 if (!timeout) 992 timeout = 993 pf_default_rule.timeout[PFTM_SRC_NODE]; 994 s->nat_src_node->expire = time_second + timeout; 995 } 996 } 997 s->src_node = s->nat_src_node = NULL; 998 } 999 1000 void 1001 pf_purge_expired_state(struct pf_state *cur) 1002 { 1003 if (cur->src.state == PF_TCPS_PROXY_DST) 1004 pf_send_tcp(cur->rule.ptr, cur->af, 1005 &cur->ext.addr, &cur->lan.addr, 1006 cur->ext.port, cur->lan.port, 1007 cur->src.seqhi, cur->src.seqlo + 1, 1008 TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL); 1009 RB_REMOVE(pf_state_tree_ext_gwy, 1010 &cur->u.s.kif->pfik_ext_gwy, cur); 1011 RB_REMOVE(pf_state_tree_lan_ext, 1012 &cur->u.s.kif->pfik_lan_ext, cur); 1013 RB_REMOVE(pf_state_tree_id, &tree_id, cur); 1014 #if NPFSYNC 1015 pfsync_delete_state(cur); 1016 #endif 1017 pf_src_tree_remove_state(cur); 1018 if (--cur->rule.ptr->states <= 0 && 1019 cur->rule.ptr->src_nodes <= 0) 1020 pf_rm_rule(NULL, cur->rule.ptr); 1021 if (cur->nat_rule.ptr != NULL) 1022 if (--cur->nat_rule.ptr->states <= 0 && 1023 cur->nat_rule.ptr->src_nodes <= 0) 1024 pf_rm_rule(NULL, cur->nat_rule.ptr); 1025 if (cur->anchor.ptr != NULL) 1026 if (--cur->anchor.ptr->states <= 0) 1027 pf_rm_rule(NULL, cur->anchor.ptr); 1028 pf_normalize_tcp_cleanup(cur); 1029 pfi_detach_state(cur->u.s.kif); 1030 TAILQ_REMOVE(&state_updates, cur, u.s.entry_updates); 1031 if (cur->tag) 1032 pf_tag_unref(cur->tag); 1033 pool_put(&pf_state_pl, cur); 1034 pf_status.fcounters[FCNT_STATE_REMOVALS]++; 1035 pf_status.states--; 1036 } 1037 1038 void 1039 pf_purge_expired_states(void) 1040 { 1041 struct pf_state *cur, *next; 1042 1043 for (cur = RB_MIN(pf_state_tree_id, &tree_id); 1044 cur; cur = next) { 1045 next = RB_NEXT(pf_state_tree_id, &tree_id, cur); 1046 if (pf_state_expires(cur) <= time_second) 1047 pf_purge_expired_state(cur); 1048 } 1049 } 1050 1051 int 1052 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw) 1053 { 1054 if (aw->type != PF_ADDR_TABLE) 1055 return (0); 1056 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL) 1057 return (1); 1058 return (0); 1059 } 1060 1061 void 1062 pf_tbladdr_remove(struct pf_addr_wrap *aw) 1063 { 1064 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL) 1065 return; 1066 pfr_detach_table(aw->p.tbl); 1067 aw->p.tbl = NULL; 1068 } 1069 1070 void 1071 pf_tbladdr_copyout(struct pf_addr_wrap *aw) 1072 { 1073 struct pfr_ktable *kt = aw->p.tbl; 1074 1075 if (aw->type != PF_ADDR_TABLE || kt == NULL) 1076 return; 1077 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) 1078 kt = kt->pfrkt_root; 1079 aw->p.tbl = NULL; 1080 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ? 1081 kt->pfrkt_cnt : -1; 1082 } 1083 1084 void 1085 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af) 1086 { 1087 switch (af) { 1088 #ifdef INET 1089 case AF_INET: { 1090 u_int32_t a = ntohl(addr->addr32[0]); 1091 printf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255, 1092 (a>>8)&255, a&255); 1093 if (p) { 1094 p = ntohs(p); 1095 printf(":%u", p); 1096 } 1097 break; 1098 } 1099 #endif /* INET */ 1100 #ifdef INET6 1101 case AF_INET6: { 1102 u_int16_t b; 1103 u_int8_t i, curstart = 255, curend = 0, 1104 maxstart = 255, maxend = 255; 1105 for (i = 0; i < 8; i++) { 1106 if (!addr->addr16[i]) { 1107 if (curstart == 255) 1108 curstart = i; 1109 else 1110 curend = i; 1111 } else { 1112 if (curstart != 255) { 1113 if ((curend - curstart) > 1114 (maxend - maxstart)) { 1115 maxstart = curstart; 1116 maxend = curend; 1117 } 1118 curstart = 255; 1119 } 1120 } 1121 } 1122 for (i = 0; i < 8; i++) { 1123 if (i >= maxstart && i <= maxend) { 1124 if (maxstart == 0) { 1125 if (i < 2) 1126 printf(":"); 1127 } else if (maxend != 7) { 1128 if (i == maxstart) 1129 printf(":"); 1130 } else { 1131 if (i == maxend) 1132 printf(":"); 1133 } 1134 } else { 1135 b = ntohs(addr->addr16[i]); 1136 printf("%x", b); 1137 if (i < 7) 1138 printf(":"); 1139 } 1140 } 1141 if (p) { 1142 p = ntohs(p); 1143 printf("[%u]", p); 1144 } 1145 break; 1146 } 1147 #endif /* INET6 */ 1148 } 1149 } 1150 1151 void 1152 pf_print_state(struct pf_state *s) 1153 { 1154 switch (s->proto) { 1155 case IPPROTO_TCP: 1156 printf("TCP "); 1157 break; 1158 case IPPROTO_UDP: 1159 printf("UDP "); 1160 break; 1161 case IPPROTO_ICMP: 1162 printf("ICMP "); 1163 break; 1164 case IPPROTO_ICMPV6: 1165 printf("ICMPV6 "); 1166 break; 1167 default: 1168 printf("%u ", s->proto); 1169 break; 1170 } 1171 pf_print_host(&s->lan.addr, s->lan.port, s->af); 1172 printf(" "); 1173 pf_print_host(&s->gwy.addr, s->gwy.port, s->af); 1174 printf(" "); 1175 pf_print_host(&s->ext.addr, s->ext.port, s->af); 1176 printf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo, 1177 s->src.seqhi, s->src.max_win, s->src.seqdiff); 1178 if (s->src.wscale && s->dst.wscale) 1179 printf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK); 1180 printf("]"); 1181 printf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo, 1182 s->dst.seqhi, s->dst.max_win, s->dst.seqdiff); 1183 if (s->src.wscale && s->dst.wscale) 1184 printf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK); 1185 printf("]"); 1186 printf(" %u:%u", s->src.state, s->dst.state); 1187 } 1188 1189 void 1190 pf_print_flags(u_int8_t f) 1191 { 1192 if (f) 1193 printf(" "); 1194 if (f & TH_FIN) 1195 printf("F"); 1196 if (f & TH_SYN) 1197 printf("S"); 1198 if (f & TH_RST) 1199 printf("R"); 1200 if (f & TH_PUSH) 1201 printf("P"); 1202 if (f & TH_ACK) 1203 printf("A"); 1204 if (f & TH_URG) 1205 printf("U"); 1206 if (f & TH_ECE) 1207 printf("E"); 1208 if (f & TH_CWR) 1209 printf("W"); 1210 } 1211 1212 #define PF_SET_SKIP_STEPS(i) \ 1213 do { \ 1214 while (head[i] != cur) { \ 1215 head[i]->skip[i].ptr = cur; \ 1216 head[i] = TAILQ_NEXT(head[i], entries); \ 1217 } \ 1218 } while (0) 1219 1220 void 1221 pf_calc_skip_steps(struct pf_rulequeue *rules) 1222 { 1223 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT]; 1224 int i; 1225 1226 cur = TAILQ_FIRST(rules); 1227 prev = cur; 1228 for (i = 0; i < PF_SKIP_COUNT; ++i) 1229 head[i] = cur; 1230 while (cur != NULL) { 1231 1232 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot) 1233 PF_SET_SKIP_STEPS(PF_SKIP_IFP); 1234 if (cur->direction != prev->direction) 1235 PF_SET_SKIP_STEPS(PF_SKIP_DIR); 1236 if (cur->af != prev->af) 1237 PF_SET_SKIP_STEPS(PF_SKIP_AF); 1238 if (cur->proto != prev->proto) 1239 PF_SET_SKIP_STEPS(PF_SKIP_PROTO); 1240 if (cur->src.neg != prev->src.neg || 1241 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr)) 1242 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR); 1243 if (cur->src.port[0] != prev->src.port[0] || 1244 cur->src.port[1] != prev->src.port[1] || 1245 cur->src.port_op != prev->src.port_op) 1246 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT); 1247 if (cur->dst.neg != prev->dst.neg || 1248 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr)) 1249 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR); 1250 if (cur->dst.port[0] != prev->dst.port[0] || 1251 cur->dst.port[1] != prev->dst.port[1] || 1252 cur->dst.port_op != prev->dst.port_op) 1253 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT); 1254 1255 prev = cur; 1256 cur = TAILQ_NEXT(cur, entries); 1257 } 1258 for (i = 0; i < PF_SKIP_COUNT; ++i) 1259 PF_SET_SKIP_STEPS(i); 1260 } 1261 1262 int 1263 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2) 1264 { 1265 if (aw1->type != aw2->type) 1266 return (1); 1267 switch (aw1->type) { 1268 case PF_ADDR_ADDRMASK: 1269 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0)) 1270 return (1); 1271 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0)) 1272 return (1); 1273 return (0); 1274 case PF_ADDR_DYNIFTL: 1275 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt); 1276 case PF_ADDR_NOROUTE: 1277 return (0); 1278 case PF_ADDR_TABLE: 1279 return (aw1->p.tbl != aw2->p.tbl); 1280 default: 1281 printf("invalid address type: %d\n", aw1->type); 1282 return (1); 1283 } 1284 } 1285 1286 u_int16_t 1287 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp) 1288 { 1289 u_int32_t l; 1290 1291 if (udp && !cksum) 1292 return (0x0000); 1293 l = cksum + old - new; 1294 l = (l >> 16) + (l & 65535); 1295 l = l & 65535; 1296 if (udp && !l) 1297 return (0xFFFF); 1298 return (l); 1299 } 1300 1301 void 1302 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc, 1303 struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af) 1304 { 1305 struct pf_addr ao; 1306 u_int16_t po = *p; 1307 1308 PF_ACPY(&ao, a, af); 1309 PF_ACPY(a, an, af); 1310 1311 *p = pn; 1312 1313 switch (af) { 1314 #ifdef INET 1315 case AF_INET: 1316 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic, 1317 ao.addr16[0], an->addr16[0], 0), 1318 ao.addr16[1], an->addr16[1], 0); 1319 *p = pn; 1320 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc, 1321 ao.addr16[0], an->addr16[0], u), 1322 ao.addr16[1], an->addr16[1], u), 1323 po, pn, u); 1324 break; 1325 #endif /* INET */ 1326 #ifdef INET6 1327 case AF_INET6: 1328 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1329 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1330 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc, 1331 ao.addr16[0], an->addr16[0], u), 1332 ao.addr16[1], an->addr16[1], u), 1333 ao.addr16[2], an->addr16[2], u), 1334 ao.addr16[3], an->addr16[3], u), 1335 ao.addr16[4], an->addr16[4], u), 1336 ao.addr16[5], an->addr16[5], u), 1337 ao.addr16[6], an->addr16[6], u), 1338 ao.addr16[7], an->addr16[7], u), 1339 po, pn, u); 1340 break; 1341 #endif /* INET6 */ 1342 } 1343 } 1344 1345 1346 /* Changes a u_int32_t. Uses a void * so there are no align restrictions */ 1347 void 1348 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u) 1349 { 1350 u_int32_t ao; 1351 1352 memcpy(&ao, a, sizeof(ao)); 1353 memcpy(a, &an, sizeof(u_int32_t)); 1354 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u), 1355 ao % 65536, an % 65536, u); 1356 } 1357 1358 #ifdef INET6 1359 void 1360 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u) 1361 { 1362 struct pf_addr ao; 1363 1364 PF_ACPY(&ao, a, AF_INET6); 1365 PF_ACPY(a, an, AF_INET6); 1366 1367 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1368 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1369 pf_cksum_fixup(pf_cksum_fixup(*c, 1370 ao.addr16[0], an->addr16[0], u), 1371 ao.addr16[1], an->addr16[1], u), 1372 ao.addr16[2], an->addr16[2], u), 1373 ao.addr16[3], an->addr16[3], u), 1374 ao.addr16[4], an->addr16[4], u), 1375 ao.addr16[5], an->addr16[5], u), 1376 ao.addr16[6], an->addr16[6], u), 1377 ao.addr16[7], an->addr16[7], u); 1378 } 1379 #endif /* INET6 */ 1380 1381 void 1382 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa, 1383 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c, 1384 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af) 1385 { 1386 struct pf_addr oia, ooa; 1387 1388 PF_ACPY(&oia, ia, af); 1389 PF_ACPY(&ooa, oa, af); 1390 1391 /* Change inner protocol port, fix inner protocol checksum. */ 1392 if (ip != NULL) { 1393 u_int16_t oip = *ip; 1394 u_int32_t opc = 0; 1395 1396 if (pc != NULL) 1397 opc = *pc; 1398 *ip = np; 1399 if (pc != NULL) 1400 *pc = pf_cksum_fixup(*pc, oip, *ip, u); 1401 *ic = pf_cksum_fixup(*ic, oip, *ip, 0); 1402 if (pc != NULL) 1403 *ic = pf_cksum_fixup(*ic, opc, *pc, 0); 1404 } 1405 /* Change inner ip address, fix inner ip and icmp checksums. */ 1406 PF_ACPY(ia, na, af); 1407 switch (af) { 1408 #ifdef INET 1409 case AF_INET: { 1410 u_int32_t oh2c = *h2c; 1411 1412 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c, 1413 oia.addr16[0], ia->addr16[0], 0), 1414 oia.addr16[1], ia->addr16[1], 0); 1415 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic, 1416 oia.addr16[0], ia->addr16[0], 0), 1417 oia.addr16[1], ia->addr16[1], 0); 1418 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0); 1419 break; 1420 } 1421 #endif /* INET */ 1422 #ifdef INET6 1423 case AF_INET6: 1424 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1425 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1426 pf_cksum_fixup(pf_cksum_fixup(*ic, 1427 oia.addr16[0], ia->addr16[0], u), 1428 oia.addr16[1], ia->addr16[1], u), 1429 oia.addr16[2], ia->addr16[2], u), 1430 oia.addr16[3], ia->addr16[3], u), 1431 oia.addr16[4], ia->addr16[4], u), 1432 oia.addr16[5], ia->addr16[5], u), 1433 oia.addr16[6], ia->addr16[6], u), 1434 oia.addr16[7], ia->addr16[7], u); 1435 break; 1436 #endif /* INET6 */ 1437 } 1438 /* Change outer ip address, fix outer ip or icmpv6 checksum. */ 1439 PF_ACPY(oa, na, af); 1440 switch (af) { 1441 #ifdef INET 1442 case AF_INET: 1443 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc, 1444 ooa.addr16[0], oa->addr16[0], 0), 1445 ooa.addr16[1], oa->addr16[1], 0); 1446 break; 1447 #endif /* INET */ 1448 #ifdef INET6 1449 case AF_INET6: 1450 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1451 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1452 pf_cksum_fixup(pf_cksum_fixup(*ic, 1453 ooa.addr16[0], oa->addr16[0], u), 1454 ooa.addr16[1], oa->addr16[1], u), 1455 ooa.addr16[2], oa->addr16[2], u), 1456 ooa.addr16[3], oa->addr16[3], u), 1457 ooa.addr16[4], oa->addr16[4], u), 1458 ooa.addr16[5], oa->addr16[5], u), 1459 ooa.addr16[6], oa->addr16[6], u), 1460 ooa.addr16[7], oa->addr16[7], u); 1461 break; 1462 #endif /* INET6 */ 1463 } 1464 } 1465 1466 void 1467 pf_send_tcp(const struct pf_rule *r __unused, sa_family_t af, 1468 const struct pf_addr *saddr, const struct pf_addr *daddr, 1469 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack, 1470 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag, 1471 u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp __unused) 1472 { 1473 struct mbuf *m; 1474 int len, tlen; 1475 #ifdef INET 1476 struct ip *h = NULL; 1477 #endif /* INET */ 1478 #ifdef INET6 1479 struct ip6_hdr *h6 = NULL; 1480 #endif /* INET6 */ 1481 struct tcphdr *th; 1482 char *opt; 1483 1484 /* maximum segment size tcp option */ 1485 tlen = sizeof(struct tcphdr); 1486 if (mss) 1487 tlen += 4; 1488 1489 switch (af) { 1490 #ifdef INET 1491 case AF_INET: 1492 len = sizeof(struct ip) + tlen; 1493 break; 1494 #endif /* INET */ 1495 #ifdef INET6 1496 case AF_INET6: 1497 len = sizeof(struct ip6_hdr) + tlen; 1498 break; 1499 #endif /* INET6 */ 1500 default: 1501 return; 1502 } 1503 1504 /* create outgoing mbuf */ 1505 m = m_gethdr(M_DONTWAIT, MT_HEADER); 1506 if (m == NULL) 1507 return; 1508 if (tag) { 1509 struct m_tag *mtag; 1510 1511 mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT); 1512 if (mtag == NULL) { 1513 m_freem(m); 1514 return; 1515 } 1516 m_tag_prepend(m, mtag); 1517 } 1518 if (rtag) 1519 if (pf_tag_packet(m, NULL, rtag)) { 1520 m_freem(m); 1521 return; 1522 } 1523 #ifdef ALTQ 1524 if (r != NULL && r->qid) { 1525 struct m_tag *mtag; 1526 struct altq_tag *atag; 1527 1528 mtag = m_tag_get(PACKET_TAG_PF_QID, sizeof(*atag), M_NOWAIT); 1529 if (mtag != NULL) { 1530 atag = (struct altq_tag *)(mtag + 1); 1531 atag->qid = r->qid; 1532 /* add hints for ecn */ 1533 atag->af = af; 1534 atag->hdr = mtod(m, struct ip *); 1535 m_tag_prepend(m, mtag); 1536 } 1537 } 1538 #endif /* ALTQ */ 1539 m->m_data += max_linkhdr; 1540 m->m_pkthdr.len = m->m_len = len; 1541 m->m_pkthdr.rcvif = NULL; 1542 bzero(m->m_data, len); 1543 switch (af) { 1544 #ifdef INET 1545 case AF_INET: 1546 h = mtod(m, struct ip *); 1547 1548 /* IP header fields included in the TCP checksum */ 1549 h->ip_p = IPPROTO_TCP; 1550 h->ip_len = htons(tlen); 1551 h->ip_src.s_addr = saddr->v4.s_addr; 1552 h->ip_dst.s_addr = daddr->v4.s_addr; 1553 1554 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip)); 1555 break; 1556 #endif /* INET */ 1557 #ifdef INET6 1558 case AF_INET6: 1559 h6 = mtod(m, struct ip6_hdr *); 1560 1561 /* IP header fields included in the TCP checksum */ 1562 h6->ip6_nxt = IPPROTO_TCP; 1563 h6->ip6_plen = htons(tlen); 1564 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr)); 1565 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr)); 1566 1567 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr)); 1568 break; 1569 #endif /* INET6 */ 1570 default: 1571 m_freem(m); 1572 return; 1573 } 1574 1575 /* TCP header */ 1576 th->th_sport = sport; 1577 th->th_dport = dport; 1578 th->th_seq = htonl(seq); 1579 th->th_ack = htonl(ack); 1580 th->th_off = tlen >> 2; 1581 th->th_flags = flags; 1582 th->th_win = htons(win); 1583 1584 if (mss) { 1585 opt = (char *)(th + 1); 1586 opt[0] = TCPOPT_MAXSEG; 1587 opt[1] = 4; 1588 HTONS(mss); 1589 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2); 1590 } 1591 1592 switch (af) { 1593 #ifdef INET 1594 case AF_INET: 1595 /* TCP checksum */ 1596 th->th_sum = in_cksum(m, len); 1597 1598 /* Finish the IP header */ 1599 h->ip_v = 4; 1600 h->ip_hl = sizeof(*h) >> 2; 1601 h->ip_tos = IPTOS_LOWDELAY; 1602 h->ip_len = htons(len); 1603 h->ip_off = htons(ip_mtudisc ? IP_DF : 0); 1604 h->ip_ttl = ttl ? ttl : ip_defttl; 1605 h->ip_sum = 0; 1606 if (eh == NULL) { 1607 ip_output(m, (void *)NULL, (void *)NULL, 0, 1608 (void *)NULL, (void *)NULL); 1609 } else { 1610 #ifdef __OpenBSD__ 1611 struct route ro; 1612 struct rtentry rt; 1613 struct ether_header *e = (void *)ro.ro_dst.sa_data; 1614 1615 if (ifp == NULL) { 1616 m_freem(m); 1617 return; 1618 } 1619 rt.rt_ifp = ifp; 1620 ro.ro_rt = &rt; 1621 ro.ro_dst.sa_len = sizeof(ro.ro_dst); 1622 ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT; 1623 bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN); 1624 bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN); 1625 e->ether_type = eh->ether_type; 1626 ip_output(m, (void *)NULL, &ro, IP_ROUTETOETHER, 1627 (void *)NULL, (void *)NULL); 1628 #else 1629 /* 1630 * on netbsd, pf_test and pf_test6 are always called 1631 * with eh == NULL. 1632 */ 1633 panic("pf_send_tcp: eh != NULL"); 1634 #endif 1635 } 1636 break; 1637 #endif /* INET */ 1638 #ifdef INET6 1639 case AF_INET6: 1640 /* TCP checksum */ 1641 th->th_sum = in6_cksum(m, IPPROTO_TCP, 1642 sizeof(struct ip6_hdr), tlen); 1643 1644 h6->ip6_vfc |= IPV6_VERSION; 1645 h6->ip6_hlim = IPV6_DEFHLIM; 1646 1647 #ifdef __OpenBSD__ 1648 ip6_output(m, NULL, NULL, 0, NULL, NULL); 1649 #else 1650 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL); 1651 #endif 1652 break; 1653 #endif /* INET6 */ 1654 } 1655 } 1656 1657 void 1658 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af, 1659 struct pf_rule *r __unused) 1660 { 1661 struct m_tag *mtag; 1662 struct mbuf *m0; 1663 1664 mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT); 1665 if (mtag == NULL) 1666 return; 1667 m0 = m_copy(m, 0, M_COPYALL); 1668 if (m0 == NULL) { 1669 m_tag_free(mtag); 1670 return; 1671 } 1672 m_tag_prepend(m0, mtag); 1673 1674 #ifdef ALTQ 1675 if (r->qid) { 1676 struct altq_tag *atag; 1677 1678 mtag = m_tag_get(PACKET_TAG_PF_QID, sizeof(*atag), M_NOWAIT); 1679 if (mtag != NULL) { 1680 atag = (struct altq_tag *)(mtag + 1); 1681 atag->qid = r->qid; 1682 /* add hints for ecn */ 1683 atag->af = af; 1684 atag->hdr = mtod(m0, struct ip *); 1685 m_tag_prepend(m0, mtag); 1686 } 1687 } 1688 #endif /* ALTQ */ 1689 1690 switch (af) { 1691 #ifdef INET 1692 case AF_INET: 1693 icmp_error(m0, type, code, 0, 0); 1694 break; 1695 #endif /* INET */ 1696 #ifdef INET6 1697 case AF_INET6: 1698 icmp6_error(m0, type, code, 0); 1699 break; 1700 #endif /* INET6 */ 1701 } 1702 } 1703 1704 /* 1705 * Return 1 if the addresses a and b match (with mask m), otherwise return 0. 1706 * If n is 0, they match if they are equal. If n is != 0, they match if they 1707 * are different. 1708 */ 1709 int 1710 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m, 1711 struct pf_addr *b, sa_family_t af) 1712 { 1713 int match = 0; 1714 1715 switch (af) { 1716 #ifdef INET 1717 case AF_INET: 1718 if ((a->addr32[0] & m->addr32[0]) == 1719 (b->addr32[0] & m->addr32[0])) 1720 match++; 1721 break; 1722 #endif /* INET */ 1723 #ifdef INET6 1724 case AF_INET6: 1725 if (((a->addr32[0] & m->addr32[0]) == 1726 (b->addr32[0] & m->addr32[0])) && 1727 ((a->addr32[1] & m->addr32[1]) == 1728 (b->addr32[1] & m->addr32[1])) && 1729 ((a->addr32[2] & m->addr32[2]) == 1730 (b->addr32[2] & m->addr32[2])) && 1731 ((a->addr32[3] & m->addr32[3]) == 1732 (b->addr32[3] & m->addr32[3]))) 1733 match++; 1734 break; 1735 #endif /* INET6 */ 1736 } 1737 if (match) { 1738 if (n) 1739 return (0); 1740 else 1741 return (1); 1742 } else { 1743 if (n) 1744 return (1); 1745 else 1746 return (0); 1747 } 1748 } 1749 1750 int 1751 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p) 1752 { 1753 switch (op) { 1754 case PF_OP_IRG: 1755 return ((p > a1) && (p < a2)); 1756 case PF_OP_XRG: 1757 return ((p < a1) || (p > a2)); 1758 case PF_OP_RRG: 1759 return ((p >= a1) && (p <= a2)); 1760 case PF_OP_EQ: 1761 return (p == a1); 1762 case PF_OP_NE: 1763 return (p != a1); 1764 case PF_OP_LT: 1765 return (p < a1); 1766 case PF_OP_LE: 1767 return (p <= a1); 1768 case PF_OP_GT: 1769 return (p > a1); 1770 case PF_OP_GE: 1771 return (p >= a1); 1772 } 1773 return (0); /* never reached */ 1774 } 1775 1776 int 1777 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p) 1778 { 1779 NTOHS(a1); 1780 NTOHS(a2); 1781 NTOHS(p); 1782 return (pf_match(op, a1, a2, p)); 1783 } 1784 1785 int 1786 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u) 1787 { 1788 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE) 1789 return (0); 1790 return (pf_match(op, a1, a2, u)); 1791 } 1792 1793 int 1794 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g) 1795 { 1796 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE) 1797 return (0); 1798 return (pf_match(op, a1, a2, g)); 1799 } 1800 1801 struct pf_tag * 1802 pf_get_tag(struct mbuf *m) 1803 { 1804 struct m_tag *mtag; 1805 1806 if ((mtag = m_tag_find(m, PACKET_TAG_PF_TAG, NULL)) != NULL) 1807 return ((struct pf_tag *)(mtag + 1)); 1808 else 1809 return (NULL); 1810 } 1811 1812 int 1813 pf_match_tag(struct mbuf *m, struct pf_rule *r, struct pf_tag **pftag, int *tag) 1814 { 1815 if (*tag == -1) { /* find mbuf tag */ 1816 *pftag = pf_get_tag(m); 1817 if (*pftag != NULL) 1818 *tag = (*pftag)->tag; 1819 else 1820 *tag = 0; 1821 } 1822 1823 return ((!r->match_tag_not && r->match_tag == *tag) || 1824 (r->match_tag_not && r->match_tag != *tag)); 1825 } 1826 1827 int 1828 pf_tag_packet(struct mbuf *m, struct pf_tag *pftag, int tag) 1829 { 1830 struct m_tag *mtag; 1831 1832 if (tag <= 0) 1833 return (0); 1834 1835 if (pftag == NULL) { 1836 mtag = m_tag_get(PACKET_TAG_PF_TAG, sizeof(*pftag), M_NOWAIT); 1837 if (mtag == NULL) 1838 return (1); 1839 ((struct pf_tag *)(mtag + 1))->tag = tag; 1840 m_tag_prepend(m, mtag); 1841 } else 1842 pftag->tag = tag; 1843 1844 return (0); 1845 } 1846 1847 static void 1848 pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n, 1849 struct pf_rule **r, struct pf_rule **a) 1850 { 1851 struct pf_anchor_stackframe *f; 1852 1853 if (*depth >= sizeof(pf_anchor_stack) / 1854 sizeof(pf_anchor_stack[0])) { 1855 printf("pf_step_into_anchor: stack overflow\n"); 1856 *r = TAILQ_NEXT(*r, entries); 1857 return; 1858 } else if (*depth == 0 && a != NULL) 1859 *a = *r; 1860 f = pf_anchor_stack + (*depth)++; 1861 f->rs = *rs; 1862 f->r = *r; 1863 if ((*r)->anchor_wildcard) { 1864 f->parent = &(*r)->anchor->children; 1865 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) == 1866 NULL) { 1867 *r = NULL; 1868 return; 1869 } 1870 *rs = &f->child->ruleset; 1871 } else { 1872 f->parent = NULL; 1873 f->child = NULL; 1874 *rs = &(*r)->anchor->ruleset; 1875 } 1876 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr); 1877 } 1878 1879 static void 1880 pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n, 1881 struct pf_rule **r, struct pf_rule **a) 1882 { 1883 struct pf_anchor_stackframe *f; 1884 1885 do { 1886 if (*depth <= 0) 1887 break; 1888 f = pf_anchor_stack + *depth - 1; 1889 if (f->parent != NULL && f->child != NULL) { 1890 f->child = RB_NEXT(pf_anchor_node, f->parent, f->child); 1891 if (f->child != NULL) { 1892 *rs = &f->child->ruleset; 1893 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr); 1894 if (*r == NULL) 1895 continue; 1896 else 1897 break; 1898 } 1899 } 1900 (*depth)--; 1901 if (*depth == 0 && a != NULL) 1902 *a = NULL; 1903 *rs = f->rs; 1904 *r = TAILQ_NEXT(f->r, entries); 1905 } while (*r == NULL); 1906 } 1907 1908 #ifdef INET6 1909 void 1910 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr, 1911 struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af) 1912 { 1913 switch (af) { 1914 #ifdef INET 1915 case AF_INET: 1916 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) | 1917 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]); 1918 break; 1919 #endif /* INET */ 1920 case AF_INET6: 1921 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) | 1922 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]); 1923 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) | 1924 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]); 1925 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) | 1926 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]); 1927 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) | 1928 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]); 1929 break; 1930 } 1931 } 1932 1933 void 1934 pf_addr_inc(struct pf_addr *addr, sa_family_t af) 1935 { 1936 switch (af) { 1937 #ifdef INET 1938 case AF_INET: 1939 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1); 1940 break; 1941 #endif /* INET */ 1942 case AF_INET6: 1943 if (addr->addr32[3] == 0xffffffff) { 1944 addr->addr32[3] = 0; 1945 if (addr->addr32[2] == 0xffffffff) { 1946 addr->addr32[2] = 0; 1947 if (addr->addr32[1] == 0xffffffff) { 1948 addr->addr32[1] = 0; 1949 addr->addr32[0] = 1950 htonl(ntohl(addr->addr32[0]) + 1); 1951 } else 1952 addr->addr32[1] = 1953 htonl(ntohl(addr->addr32[1]) + 1); 1954 } else 1955 addr->addr32[2] = 1956 htonl(ntohl(addr->addr32[2]) + 1); 1957 } else 1958 addr->addr32[3] = 1959 htonl(ntohl(addr->addr32[3]) + 1); 1960 break; 1961 } 1962 } 1963 #endif /* INET6 */ 1964 1965 #define mix(a,b,c) \ 1966 do { \ 1967 a -= b; a -= c; a ^= (c >> 13); \ 1968 b -= c; b -= a; b ^= (a << 8); \ 1969 c -= a; c -= b; c ^= (b >> 13); \ 1970 a -= b; a -= c; a ^= (c >> 12); \ 1971 b -= c; b -= a; b ^= (a << 16); \ 1972 c -= a; c -= b; c ^= (b >> 5); \ 1973 a -= b; a -= c; a ^= (c >> 3); \ 1974 b -= c; b -= a; b ^= (a << 10); \ 1975 c -= a; c -= b; c ^= (b >> 15); \ 1976 } while (0) 1977 1978 /* 1979 * hash function based on bridge_hash in if_bridge.c 1980 */ 1981 void 1982 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash, 1983 struct pf_poolhashkey *key, sa_family_t af) 1984 { 1985 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0]; 1986 1987 switch (af) { 1988 #ifdef INET 1989 case AF_INET: 1990 a += inaddr->addr32[0]; 1991 b += key->key32[1]; 1992 mix(a, b, c); 1993 hash->addr32[0] = c + key->key32[2]; 1994 break; 1995 #endif /* INET */ 1996 #ifdef INET6 1997 case AF_INET6: 1998 a += inaddr->addr32[0]; 1999 b += inaddr->addr32[2]; 2000 mix(a, b, c); 2001 hash->addr32[0] = c; 2002 a += inaddr->addr32[1]; 2003 b += inaddr->addr32[3]; 2004 c += key->key32[1]; 2005 mix(a, b, c); 2006 hash->addr32[1] = c; 2007 a += inaddr->addr32[2]; 2008 b += inaddr->addr32[1]; 2009 c += key->key32[2]; 2010 mix(a, b, c); 2011 hash->addr32[2] = c; 2012 a += inaddr->addr32[3]; 2013 b += inaddr->addr32[0]; 2014 c += key->key32[3]; 2015 mix(a, b, c); 2016 hash->addr32[3] = c; 2017 break; 2018 #endif /* INET6 */ 2019 } 2020 } 2021 2022 int 2023 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr, 2024 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn) 2025 { 2026 unsigned char hash[16]; 2027 struct pf_pool *rpool = &r->rpool; 2028 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr; 2029 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask; 2030 struct pf_pooladdr *acur = rpool->cur; 2031 struct pf_src_node k; 2032 2033 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR && 2034 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { 2035 k.af = af; 2036 PF_ACPY(&k.addr, saddr, af); 2037 if (r->rule_flag & PFRULE_RULESRCTRACK || 2038 r->rpool.opts & PF_POOL_STICKYADDR) 2039 k.rule.ptr = r; 2040 else 2041 k.rule.ptr = NULL; 2042 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++; 2043 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k); 2044 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) { 2045 PF_ACPY(naddr, &(*sn)->raddr, af); 2046 if (pf_status.debug >= PF_DEBUG_MISC) { 2047 printf("pf_map_addr: src tracking maps "); 2048 pf_print_host(&k.addr, 0, af); 2049 printf(" to "); 2050 pf_print_host(naddr, 0, af); 2051 printf("\n"); 2052 } 2053 return (0); 2054 } 2055 } 2056 2057 if (rpool->cur->addr.type == PF_ADDR_NOROUTE) 2058 return (1); 2059 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2060 switch (af) { 2061 #ifdef INET 2062 case AF_INET: 2063 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 && 2064 (rpool->opts & PF_POOL_TYPEMASK) != 2065 PF_POOL_ROUNDROBIN) 2066 return (1); 2067 raddr = &rpool->cur->addr.p.dyn->pfid_addr4; 2068 rmask = &rpool->cur->addr.p.dyn->pfid_mask4; 2069 break; 2070 #endif /* INET */ 2071 #ifdef INET6 2072 case AF_INET6: 2073 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 && 2074 (rpool->opts & PF_POOL_TYPEMASK) != 2075 PF_POOL_ROUNDROBIN) 2076 return (1); 2077 raddr = &rpool->cur->addr.p.dyn->pfid_addr6; 2078 rmask = &rpool->cur->addr.p.dyn->pfid_mask6; 2079 break; 2080 #endif /* INET6 */ 2081 } 2082 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2083 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN) 2084 return (1); /* unsupported */ 2085 } else { 2086 raddr = &rpool->cur->addr.v.a.addr; 2087 rmask = &rpool->cur->addr.v.a.mask; 2088 } 2089 2090 switch (rpool->opts & PF_POOL_TYPEMASK) { 2091 case PF_POOL_NONE: 2092 PF_ACPY(naddr, raddr, af); 2093 break; 2094 case PF_POOL_BITMASK: 2095 PF_POOLMASK(naddr, raddr, rmask, saddr, af); 2096 break; 2097 case PF_POOL_RANDOM: 2098 if (init_addr != NULL && PF_AZERO(init_addr, af)) { 2099 switch (af) { 2100 #ifdef INET 2101 case AF_INET: 2102 rpool->counter.addr32[0] = htonl(arc4random()); 2103 break; 2104 #endif /* INET */ 2105 #ifdef INET6 2106 case AF_INET6: 2107 if (rmask->addr32[3] != 0xffffffff) 2108 rpool->counter.addr32[3] = 2109 htonl(arc4random()); 2110 else 2111 break; 2112 if (rmask->addr32[2] != 0xffffffff) 2113 rpool->counter.addr32[2] = 2114 htonl(arc4random()); 2115 else 2116 break; 2117 if (rmask->addr32[1] != 0xffffffff) 2118 rpool->counter.addr32[1] = 2119 htonl(arc4random()); 2120 else 2121 break; 2122 if (rmask->addr32[0] != 0xffffffff) 2123 rpool->counter.addr32[0] = 2124 htonl(arc4random()); 2125 break; 2126 #endif /* INET6 */ 2127 } 2128 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af); 2129 PF_ACPY(init_addr, naddr, af); 2130 2131 } else { 2132 PF_AINC(&rpool->counter, af); 2133 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af); 2134 } 2135 break; 2136 case PF_POOL_SRCHASH: 2137 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af); 2138 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af); 2139 break; 2140 case PF_POOL_ROUNDROBIN: 2141 if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2142 if (!pfr_pool_get(rpool->cur->addr.p.tbl, 2143 &rpool->tblidx, &rpool->counter, 2144 &raddr, &rmask, af)) 2145 goto get_addr; 2146 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2147 if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, 2148 &rpool->tblidx, &rpool->counter, 2149 &raddr, &rmask, af)) 2150 goto get_addr; 2151 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af)) 2152 goto get_addr; 2153 2154 try_next: 2155 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL) 2156 rpool->cur = TAILQ_FIRST(&rpool->list); 2157 if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2158 rpool->tblidx = -1; 2159 if (pfr_pool_get(rpool->cur->addr.p.tbl, 2160 &rpool->tblidx, &rpool->counter, 2161 &raddr, &rmask, af)) { 2162 /* table contains no address of type 'af' */ 2163 if (rpool->cur != acur) 2164 goto try_next; 2165 return (1); 2166 } 2167 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2168 rpool->tblidx = -1; 2169 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, 2170 &rpool->tblidx, &rpool->counter, 2171 &raddr, &rmask, af)) { 2172 /* table contains no address of type 'af' */ 2173 if (rpool->cur != acur) 2174 goto try_next; 2175 return (1); 2176 } 2177 } else { 2178 raddr = &rpool->cur->addr.v.a.addr; 2179 rmask = &rpool->cur->addr.v.a.mask; 2180 PF_ACPY(&rpool->counter, raddr, af); 2181 } 2182 2183 get_addr: 2184 PF_ACPY(naddr, &rpool->counter, af); 2185 if (init_addr != NULL && PF_AZERO(init_addr, af)) 2186 PF_ACPY(init_addr, naddr, af); 2187 PF_AINC(&rpool->counter, af); 2188 break; 2189 } 2190 if (*sn != NULL) 2191 PF_ACPY(&(*sn)->raddr, naddr, af); 2192 2193 if (pf_status.debug >= PF_DEBUG_MISC && 2194 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { 2195 printf("pf_map_addr: selected address "); 2196 pf_print_host(naddr, 0, af); 2197 printf("\n"); 2198 } 2199 2200 return (0); 2201 } 2202 2203 int 2204 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r, 2205 struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport, 2206 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high, 2207 struct pf_src_node **sn) 2208 { 2209 struct pf_state key; 2210 struct pf_addr init_addr; 2211 u_int16_t cut; 2212 2213 bzero(&init_addr, sizeof(init_addr)); 2214 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn)) 2215 return (1); 2216 2217 do { 2218 key.af = af; 2219 key.proto = proto; 2220 PF_ACPY(&key.ext.addr, daddr, key.af); 2221 PF_ACPY(&key.gwy.addr, naddr, key.af); 2222 key.ext.port = dport; 2223 2224 /* 2225 * port search; start random, step; 2226 * similar 2 portloop in in_pcbbind 2227 */ 2228 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP)) { 2229 key.gwy.port = dport; 2230 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) 2231 return (0); 2232 } else if (low == 0 && high == 0) { 2233 key.gwy.port = *nport; 2234 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) 2235 return (0); 2236 } else if (low == high) { 2237 key.gwy.port = htons(low); 2238 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) { 2239 *nport = htons(low); 2240 return (0); 2241 } 2242 } else { 2243 u_int16_t tmp; 2244 2245 if (low > high) { 2246 tmp = low; 2247 low = high; 2248 high = tmp; 2249 } 2250 /* low < high */ 2251 cut = htonl(arc4random()) % (1 + high - low) + low; 2252 /* low <= cut <= high */ 2253 for (tmp = cut; tmp <= high; ++(tmp)) { 2254 key.gwy.port = htons(tmp); 2255 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == 2256 NULL) { 2257 *nport = htons(tmp); 2258 return (0); 2259 } 2260 } 2261 for (tmp = cut - 1; tmp >= low; --(tmp)) { 2262 key.gwy.port = htons(tmp); 2263 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == 2264 NULL) { 2265 *nport = htons(tmp); 2266 return (0); 2267 } 2268 } 2269 } 2270 2271 switch (r->rpool.opts & PF_POOL_TYPEMASK) { 2272 case PF_POOL_RANDOM: 2273 case PF_POOL_ROUNDROBIN: 2274 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn)) 2275 return (1); 2276 break; 2277 case PF_POOL_NONE: 2278 case PF_POOL_SRCHASH: 2279 case PF_POOL_BITMASK: 2280 default: 2281 return (1); 2282 } 2283 } while (! PF_AEQ(&init_addr, naddr, af) ); 2284 2285 return (1); /* none available */ 2286 } 2287 2288 struct pf_rule * 2289 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off, 2290 int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport, 2291 struct pf_addr *daddr, u_int16_t dport, int rs_num) 2292 { 2293 struct pf_rule *r, *rm = NULL; 2294 struct pf_ruleset *ruleset = NULL; 2295 struct pf_tag *pftag = NULL; 2296 int tag = -1; 2297 int asd = 0; 2298 2299 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr); 2300 while (r && rm == NULL) { 2301 struct pf_rule_addr *src = NULL, *dst = NULL; 2302 struct pf_addr_wrap *xdst = NULL; 2303 2304 if (r->action == PF_BINAT && direction == PF_IN) { 2305 src = &r->dst; 2306 if (r->rpool.cur != NULL) 2307 xdst = &r->rpool.cur->addr; 2308 } else { 2309 src = &r->src; 2310 dst = &r->dst; 2311 } 2312 2313 r->evaluations++; 2314 if (r->kif != NULL && 2315 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot) 2316 r = r->skip[PF_SKIP_IFP].ptr; 2317 else if (r->direction && r->direction != direction) 2318 r = r->skip[PF_SKIP_DIR].ptr; 2319 else if (r->af && r->af != pd->af) 2320 r = r->skip[PF_SKIP_AF].ptr; 2321 else if (r->proto && r->proto != pd->proto) 2322 r = r->skip[PF_SKIP_PROTO].ptr; 2323 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af, src->neg)) 2324 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR : 2325 PF_SKIP_DST_ADDR].ptr; 2326 else if (src->port_op && !pf_match_port(src->port_op, 2327 src->port[0], src->port[1], sport)) 2328 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT : 2329 PF_SKIP_DST_PORT].ptr; 2330 else if (dst != NULL && 2331 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg)) 2332 r = r->skip[PF_SKIP_DST_ADDR].ptr; 2333 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af, 0)) 2334 r = TAILQ_NEXT(r, entries); 2335 else if (dst != NULL && dst->port_op && 2336 !pf_match_port(dst->port_op, dst->port[0], 2337 dst->port[1], dport)) 2338 r = r->skip[PF_SKIP_DST_PORT].ptr; 2339 else if (r->match_tag && !pf_match_tag(m, r, &pftag, &tag)) 2340 r = TAILQ_NEXT(r, entries); 2341 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto != 2342 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m, 2343 off, pd->hdr.tcp), r->os_fingerprint))) 2344 r = TAILQ_NEXT(r, entries); 2345 else { 2346 if (r->tag) 2347 tag = r->tag; 2348 if (r->anchor == NULL) { 2349 rm = r; 2350 } else 2351 pf_step_into_anchor(&asd, &ruleset, rs_num, &r, NULL); 2352 } 2353 if (r == NULL) 2354 pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r, NULL); 2355 } 2356 if (pf_tag_packet(m, pftag, tag)) 2357 return (NULL); 2358 if (rm != NULL && (rm->action == PF_NONAT || 2359 rm->action == PF_NORDR || rm->action == PF_NOBINAT)) 2360 return (NULL); 2361 return (rm); 2362 } 2363 2364 struct pf_rule * 2365 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction, 2366 struct pfi_kif *kif, struct pf_src_node **sn, 2367 struct pf_addr *saddr, u_int16_t sport, 2368 struct pf_addr *daddr, u_int16_t dport, 2369 struct pf_addr *naddr, u_int16_t *nport) 2370 { 2371 struct pf_rule *r = NULL; 2372 2373 if (direction == PF_OUT) { 2374 r = pf_match_translation(pd, m, off, direction, kif, saddr, 2375 sport, daddr, dport, PF_RULESET_BINAT); 2376 if (r == NULL) 2377 r = pf_match_translation(pd, m, off, direction, kif, 2378 saddr, sport, daddr, dport, PF_RULESET_NAT); 2379 } else { 2380 r = pf_match_translation(pd, m, off, direction, kif, saddr, 2381 sport, daddr, dport, PF_RULESET_RDR); 2382 if (r == NULL) 2383 r = pf_match_translation(pd, m, off, direction, kif, 2384 saddr, sport, daddr, dport, PF_RULESET_BINAT); 2385 } 2386 2387 if (r != NULL) { 2388 switch (r->action) { 2389 case PF_NONAT: 2390 case PF_NOBINAT: 2391 case PF_NORDR: 2392 return (NULL); 2393 case PF_NAT: 2394 if (pf_get_sport(pd->af, pd->proto, r, saddr, 2395 daddr, dport, naddr, nport, r->rpool.proxy_port[0], 2396 r->rpool.proxy_port[1], sn)) { 2397 DPFPRINTF(PF_DEBUG_MISC, 2398 ("pf: NAT proxy port allocation " 2399 "(%u-%u) failed\n", 2400 r->rpool.proxy_port[0], 2401 r->rpool.proxy_port[1])); 2402 return (NULL); 2403 } 2404 break; 2405 case PF_BINAT: 2406 switch (direction) { 2407 case PF_OUT: 2408 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){ 2409 switch (pd->af) { 2410 #ifdef INET 2411 case AF_INET: 2412 if (r->rpool.cur->addr.p.dyn-> 2413 pfid_acnt4 < 1) 2414 return (NULL); 2415 PF_POOLMASK(naddr, 2416 &r->rpool.cur->addr.p.dyn-> 2417 pfid_addr4, 2418 &r->rpool.cur->addr.p.dyn-> 2419 pfid_mask4, 2420 saddr, AF_INET); 2421 break; 2422 #endif /* INET */ 2423 #ifdef INET6 2424 case AF_INET6: 2425 if (r->rpool.cur->addr.p.dyn-> 2426 pfid_acnt6 < 1) 2427 return (NULL); 2428 PF_POOLMASK(naddr, 2429 &r->rpool.cur->addr.p.dyn-> 2430 pfid_addr6, 2431 &r->rpool.cur->addr.p.dyn-> 2432 pfid_mask6, 2433 saddr, AF_INET6); 2434 break; 2435 #endif /* INET6 */ 2436 } 2437 } else 2438 PF_POOLMASK(naddr, 2439 &r->rpool.cur->addr.v.a.addr, 2440 &r->rpool.cur->addr.v.a.mask, 2441 saddr, pd->af); 2442 break; 2443 case PF_IN: 2444 if (r->src.addr.type == PF_ADDR_DYNIFTL) { 2445 switch (pd->af) { 2446 #ifdef INET 2447 case AF_INET: 2448 if (r->src.addr.p.dyn-> 2449 pfid_acnt4 < 1) 2450 return (NULL); 2451 PF_POOLMASK(naddr, 2452 &r->src.addr.p.dyn-> 2453 pfid_addr4, 2454 &r->src.addr.p.dyn-> 2455 pfid_mask4, 2456 daddr, AF_INET); 2457 break; 2458 #endif /* INET */ 2459 #ifdef INET6 2460 case AF_INET6: 2461 if (r->src.addr.p.dyn-> 2462 pfid_acnt6 < 1) 2463 return (NULL); 2464 PF_POOLMASK(naddr, 2465 &r->src.addr.p.dyn-> 2466 pfid_addr6, 2467 &r->src.addr.p.dyn-> 2468 pfid_mask6, 2469 daddr, AF_INET6); 2470 break; 2471 #endif /* INET6 */ 2472 } 2473 } else 2474 PF_POOLMASK(naddr, 2475 &r->src.addr.v.a.addr, 2476 &r->src.addr.v.a.mask, daddr, 2477 pd->af); 2478 break; 2479 } 2480 break; 2481 case PF_RDR: { 2482 if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn)) 2483 return (NULL); 2484 2485 if (r->rpool.proxy_port[1]) { 2486 u_int32_t tmp_nport; 2487 2488 tmp_nport = ((ntohs(dport) - 2489 ntohs(r->dst.port[0])) % 2490 (r->rpool.proxy_port[1] - 2491 r->rpool.proxy_port[0] + 1)) + 2492 r->rpool.proxy_port[0]; 2493 2494 /* wrap around if necessary */ 2495 if (tmp_nport > 65535) 2496 tmp_nport -= 65535; 2497 *nport = htons((u_int16_t)tmp_nport); 2498 } else if (r->rpool.proxy_port[0]) 2499 *nport = htons(r->rpool.proxy_port[0]); 2500 break; 2501 } 2502 default: 2503 return (NULL); 2504 } 2505 } 2506 2507 return (r); 2508 } 2509 2510 int 2511 pf_socket_lookup(uid_t *uid, gid_t *gid, int direction, struct pf_pdesc *pd) 2512 { 2513 struct pf_addr *saddr, *daddr; 2514 u_int16_t sport, dport; 2515 struct inpcbtable *tb; 2516 struct inpcb *inp = NULL; 2517 #if defined(__NetBSD__) && defined(INET6) 2518 struct in6pcb *in6p = NULL; 2519 #endif 2520 2521 *uid = UID_MAX; 2522 *gid = GID_MAX; 2523 switch (pd->proto) { 2524 case IPPROTO_TCP: 2525 sport = pd->hdr.tcp->th_sport; 2526 dport = pd->hdr.tcp->th_dport; 2527 tb = &tcbtable; 2528 break; 2529 case IPPROTO_UDP: 2530 sport = pd->hdr.udp->uh_sport; 2531 dport = pd->hdr.udp->uh_dport; 2532 tb = &udbtable; 2533 break; 2534 default: 2535 return (0); 2536 } 2537 if (direction == PF_IN) { 2538 saddr = pd->src; 2539 daddr = pd->dst; 2540 } else { 2541 u_int16_t p; 2542 2543 p = sport; 2544 sport = dport; 2545 dport = p; 2546 saddr = pd->dst; 2547 daddr = pd->src; 2548 } 2549 switch (pd->af) { 2550 #ifdef INET 2551 case AF_INET: 2552 #ifdef __OpenBSD__ 2553 inp = in_pcbhashlookup(tb, saddr->v4, sport, daddr->v4, dport); 2554 if (inp == NULL) { 2555 inp = in_pcblookup_listen(tb, daddr->v4, dport, 0); 2556 if (inp == NULL) 2557 return (0); 2558 } 2559 #else 2560 inp = in_pcblookup_connect(tb, saddr->v4, sport, daddr->v4, 2561 dport); 2562 if (inp == NULL) { 2563 inp = in_pcblookup_bind(tb, daddr->v4, dport); 2564 if (inp == NULL) 2565 return (0); 2566 } 2567 #endif 2568 break; 2569 #endif /* INET */ 2570 #ifdef INET6 2571 case AF_INET6: 2572 #ifdef __OpenBSD__ 2573 inp = in6_pcbhashlookup(tb, &saddr->v6, sport, &daddr->v6, 2574 dport); 2575 if (inp == NULL) { 2576 inp = in6_pcblookup_listen(tb, &daddr->v6, dport, 0); 2577 if (inp == NULL) 2578 return (0); 2579 } 2580 #else 2581 in6p = in6_pcblookup_connect(tb, &saddr->v6, sport, &daddr->v6, 2582 dport, 0); 2583 if (in6p == NULL) { 2584 in6p = in6_pcblookup_bind(tb, &daddr->v6, dport, 0); 2585 if (in6p == NULL) 2586 return (0); 2587 } 2588 #endif 2589 break; 2590 #endif /* INET6 */ 2591 2592 default: 2593 return (0); 2594 } 2595 #ifdef __OpenBSD__ 2596 *uid = inp->inp_socket->so_euid; 2597 *gid = inp->inp_socket->so_egid; 2598 #else 2599 switch (pd->af) { 2600 case AF_INET: 2601 *uid = inp->inp_socket->so_uidinfo->ui_uid; 2602 /* XXX gid */ 2603 break; 2604 #ifdef INET6 2605 case AF_INET6: 2606 *uid = in6p->in6p_socket->so_uidinfo->ui_uid; 2607 /* XXX gid */ 2608 break; 2609 #endif 2610 } 2611 #endif 2612 return (1); 2613 } 2614 2615 u_int8_t 2616 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af) 2617 { 2618 int hlen; 2619 u_int8_t hdr[60]; 2620 u_int8_t *opt, optlen; 2621 u_int8_t wscale = 0; 2622 2623 hlen = th_off << 2; /* hlen <= sizeof(hdr) */ 2624 if (hlen <= sizeof(struct tcphdr)) 2625 return (0); 2626 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af)) 2627 return (0); 2628 opt = hdr + sizeof(struct tcphdr); 2629 hlen -= sizeof(struct tcphdr); 2630 while (hlen >= 3) { 2631 switch (*opt) { 2632 case TCPOPT_EOL: 2633 case TCPOPT_NOP: 2634 ++opt; 2635 --hlen; 2636 break; 2637 case TCPOPT_WINDOW: 2638 wscale = opt[2]; 2639 if (wscale > TCP_MAX_WINSHIFT) 2640 wscale = TCP_MAX_WINSHIFT; 2641 wscale |= PF_WSCALE_FLAG; 2642 /* FALLTHROUGH */ 2643 default: 2644 optlen = opt[1]; 2645 if (optlen < 2) 2646 optlen = 2; 2647 hlen -= optlen; 2648 opt += optlen; 2649 break; 2650 } 2651 } 2652 return (wscale); 2653 } 2654 2655 u_int16_t 2656 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af) 2657 { 2658 int hlen; 2659 u_int8_t hdr[60]; 2660 u_int8_t *opt, optlen; 2661 u_int16_t mss = tcp_mssdflt; 2662 2663 hlen = th_off << 2; /* hlen <= sizeof(hdr) */ 2664 if (hlen <= sizeof(struct tcphdr)) 2665 return (0); 2666 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af)) 2667 return (0); 2668 opt = hdr + sizeof(struct tcphdr); 2669 hlen -= sizeof(struct tcphdr); 2670 while (hlen >= TCPOLEN_MAXSEG) { 2671 switch (*opt) { 2672 case TCPOPT_EOL: 2673 case TCPOPT_NOP: 2674 ++opt; 2675 --hlen; 2676 break; 2677 case TCPOPT_MAXSEG: 2678 bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2); 2679 NTOHS(mss); 2680 /* FALLTHROUGH */ 2681 default: 2682 optlen = opt[1]; 2683 if (optlen < 2) 2684 optlen = 2; 2685 hlen -= optlen; 2686 opt += optlen; 2687 break; 2688 } 2689 } 2690 return (mss); 2691 } 2692 2693 u_int16_t 2694 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer) 2695 { 2696 #ifdef INET 2697 struct sockaddr_in *dst; 2698 struct route ro; 2699 #endif /* INET */ 2700 #ifdef INET6 2701 struct sockaddr_in6 *dst6; 2702 struct route_in6 ro6; 2703 #endif /* INET6 */ 2704 struct rtentry *rt = NULL; 2705 int hlen; 2706 u_int16_t mss = tcp_mssdflt; 2707 2708 hlen = 0; /* XXXGCC - -Wunitialized m68k */ 2709 2710 switch (af) { 2711 #ifdef INET 2712 case AF_INET: 2713 hlen = sizeof(struct ip); 2714 bzero(&ro, sizeof(ro)); 2715 dst = (struct sockaddr_in *)&ro.ro_dst; 2716 dst->sin_family = AF_INET; 2717 dst->sin_len = sizeof(*dst); 2718 dst->sin_addr = addr->v4; 2719 #ifdef __OpenBSD__ 2720 rtalloc_noclone(&ro, NO_CLONING); 2721 #else 2722 rtalloc(&ro); 2723 #endif 2724 rt = ro.ro_rt; 2725 break; 2726 #endif /* INET */ 2727 #ifdef INET6 2728 case AF_INET6: 2729 hlen = sizeof(struct ip6_hdr); 2730 bzero(&ro6, sizeof(ro6)); 2731 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst; 2732 dst6->sin6_family = AF_INET6; 2733 dst6->sin6_len = sizeof(*dst6); 2734 dst6->sin6_addr = addr->v6; 2735 #ifdef __OpenBSD__ 2736 rtalloc_noclone((struct route *)&ro6, NO_CLONING); 2737 #else 2738 rtalloc((struct route *)&ro6); 2739 #endif 2740 rt = ro6.ro_rt; 2741 break; 2742 #endif /* INET6 */ 2743 } 2744 2745 if (rt && rt->rt_ifp) { 2746 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr); 2747 mss = max(tcp_mssdflt, mss); 2748 RTFREE(rt); 2749 } 2750 mss = min(mss, offer); 2751 mss = max(mss, 64); /* sanity - at least max opt space */ 2752 return (mss); 2753 } 2754 2755 void 2756 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr) 2757 { 2758 struct pf_rule *r = s->rule.ptr; 2759 2760 s->rt_kif = NULL; 2761 if (!r->rt || r->rt == PF_FASTROUTE) 2762 return; 2763 switch (s->af) { 2764 #ifdef INET 2765 case AF_INET: 2766 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL, 2767 &s->nat_src_node); 2768 s->rt_kif = r->rpool.cur->kif; 2769 break; 2770 #endif /* INET */ 2771 #ifdef INET6 2772 case AF_INET6: 2773 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL, 2774 &s->nat_src_node); 2775 s->rt_kif = r->rpool.cur->kif; 2776 break; 2777 #endif /* INET6 */ 2778 } 2779 } 2780 2781 int 2782 pf_test_tcp(struct pf_rule **rm, struct pf_state **sm, int direction, 2783 struct pfi_kif *kif, struct mbuf *m, int off, void *h __unused, 2784 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 2785 struct ifqueue *ifq) 2786 { 2787 struct pf_rule *nr = NULL; 2788 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 2789 struct tcphdr *th = pd->hdr.tcp; 2790 u_int16_t bport, nport = 0; 2791 sa_family_t af = pd->af; 2792 int lookup = -1; 2793 uid_t uid = 0; /* XXX: GCC */ 2794 gid_t gid = 0; /* XXX: GCC */ 2795 struct pf_rule *r, *a = NULL; 2796 struct pf_ruleset *ruleset = NULL; 2797 struct pf_src_node *nsn = NULL; 2798 u_short reason; 2799 int rewrite = 0; 2800 struct pf_tag *pftag = NULL; 2801 int tag = -1; 2802 u_int16_t mss = tcp_mssdflt; 2803 int asd = 0; 2804 2805 if (pf_check_congestion(ifq)) { 2806 REASON_SET(&reason, PFRES_CONGEST); 2807 return (PF_DROP); 2808 } 2809 2810 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 2811 2812 if (direction == PF_OUT) { 2813 bport = nport = th->th_sport; 2814 /* check outgoing packet for BINAT/NAT */ 2815 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, 2816 saddr, th->th_sport, daddr, th->th_dport, 2817 &pd->naddr, &nport)) != NULL) { 2818 PF_ACPY(&pd->baddr, saddr, af); 2819 pf_change_ap(saddr, &th->th_sport, pd->ip_sum, 2820 &th->th_sum, &pd->naddr, nport, 0, af); 2821 rewrite++; 2822 if (nr->natpass) 2823 r = NULL; 2824 pd->nat_rule = nr; 2825 } 2826 } else { 2827 bport = nport = th->th_dport; 2828 /* check incoming packet for BINAT/RDR */ 2829 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, 2830 saddr, th->th_sport, daddr, th->th_dport, 2831 &pd->naddr, &nport)) != NULL) { 2832 PF_ACPY(&pd->baddr, daddr, af); 2833 pf_change_ap(daddr, &th->th_dport, pd->ip_sum, 2834 &th->th_sum, &pd->naddr, nport, 0, af); 2835 rewrite++; 2836 if (nr->natpass) 2837 r = NULL; 2838 pd->nat_rule = nr; 2839 } 2840 } 2841 2842 while (r != NULL) { 2843 r->evaluations++; 2844 if (r->kif != NULL && 2845 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot) 2846 r = r->skip[PF_SKIP_IFP].ptr; 2847 else if (r->direction && r->direction != direction) 2848 r = r->skip[PF_SKIP_DIR].ptr; 2849 else if (r->af && r->af != af) 2850 r = r->skip[PF_SKIP_AF].ptr; 2851 else if (r->proto && r->proto != IPPROTO_TCP) 2852 r = r->skip[PF_SKIP_PROTO].ptr; 2853 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.neg)) 2854 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 2855 else if (r->src.port_op && !pf_match_port(r->src.port_op, 2856 r->src.port[0], r->src.port[1], th->th_sport)) 2857 r = r->skip[PF_SKIP_SRC_PORT].ptr; 2858 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.neg)) 2859 r = r->skip[PF_SKIP_DST_ADDR].ptr; 2860 else if (r->dst.port_op && !pf_match_port(r->dst.port_op, 2861 r->dst.port[0], r->dst.port[1], th->th_dport)) 2862 r = r->skip[PF_SKIP_DST_PORT].ptr; 2863 else if (r->tos && !(r->tos & pd->tos)) 2864 r = TAILQ_NEXT(r, entries); 2865 else if (r->rule_flag & PFRULE_FRAGMENT) 2866 r = TAILQ_NEXT(r, entries); 2867 else if ((r->flagset & th->th_flags) != r->flags) 2868 r = TAILQ_NEXT(r, entries); 2869 else if (r->uid.op && (lookup != -1 || (lookup = 2870 pf_socket_lookup(&uid, &gid, direction, pd), 1)) && 2871 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1], 2872 uid)) 2873 r = TAILQ_NEXT(r, entries); 2874 else if (r->gid.op && (lookup != -1 || (lookup = 2875 pf_socket_lookup(&uid, &gid, direction, pd), 1)) && 2876 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1], 2877 gid)) 2878 r = TAILQ_NEXT(r, entries); 2879 else if (r->prob && r->prob <= arc4random()) 2880 r = TAILQ_NEXT(r, entries); 2881 else if (r->match_tag && !pf_match_tag(m, r, &pftag, &tag)) 2882 r = TAILQ_NEXT(r, entries); 2883 else if (r->os_fingerprint != PF_OSFP_ANY && !pf_osfp_match( 2884 pf_osfp_fingerprint(pd, m, off, th), r->os_fingerprint)) 2885 r = TAILQ_NEXT(r, entries); 2886 else { 2887 if (r->tag) 2888 tag = r->tag; 2889 if (r->anchor == NULL) { 2890 *rm = r; 2891 *am = a; 2892 *rsm = ruleset; 2893 if ((*rm)->quick) 2894 break; 2895 r = TAILQ_NEXT(r, entries); 2896 } else 2897 pf_step_into_anchor(&asd, &ruleset, 2898 PF_RULESET_FILTER, &r, &a); 2899 } 2900 if (r == NULL) 2901 pf_step_out_of_anchor(&asd, &ruleset, 2902 PF_RULESET_FILTER, &r, &a); 2903 } 2904 r = *rm; 2905 a = *am; 2906 ruleset = *rsm; 2907 2908 REASON_SET(&reason, PFRES_MATCH); 2909 2910 if (r->log) { 2911 if (rewrite) 2912 m_copyback(m, off, sizeof(*th), th); 2913 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset); 2914 } 2915 2916 if ((r->action == PF_DROP) && 2917 ((r->rule_flag & PFRULE_RETURNRST) || 2918 (r->rule_flag & PFRULE_RETURNICMP) || 2919 (r->rule_flag & PFRULE_RETURN))) { 2920 /* undo NAT changes, if they have taken place */ 2921 if (nr != NULL) { 2922 if (direction == PF_OUT) { 2923 pf_change_ap(saddr, &th->th_sport, pd->ip_sum, 2924 &th->th_sum, &pd->baddr, bport, 0, af); 2925 rewrite++; 2926 } else { 2927 pf_change_ap(daddr, &th->th_dport, pd->ip_sum, 2928 &th->th_sum, &pd->baddr, bport, 0, af); 2929 rewrite++; 2930 } 2931 } 2932 if (((r->rule_flag & PFRULE_RETURNRST) || 2933 (r->rule_flag & PFRULE_RETURN)) && 2934 !(th->th_flags & TH_RST)) { 2935 u_int32_t ack = ntohl(th->th_seq) + pd->p_len; 2936 2937 if (th->th_flags & TH_SYN) 2938 ack++; 2939 if (th->th_flags & TH_FIN) 2940 ack++; 2941 pf_send_tcp(r, af, pd->dst, 2942 pd->src, th->th_dport, th->th_sport, 2943 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0, 2944 r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp); 2945 } else if ((af == AF_INET) && r->return_icmp) 2946 pf_send_icmp(m, r->return_icmp >> 8, 2947 r->return_icmp & 255, af, r); 2948 else if ((af == AF_INET6) && r->return_icmp6) 2949 pf_send_icmp(m, r->return_icmp6 >> 8, 2950 r->return_icmp6 & 255, af, r); 2951 } 2952 2953 if (r->action == PF_DROP) 2954 return (PF_DROP); 2955 2956 if (pf_tag_packet(m, pftag, tag)) { 2957 REASON_SET(&reason, PFRES_MEMORY); 2958 return (PF_DROP); 2959 } 2960 2961 if (r->keep_state || nr != NULL || 2962 (pd->flags & PFDESC_TCP_NORM)) { 2963 /* create new state */ 2964 u_int16_t len; 2965 struct pf_state *s = NULL; 2966 struct pf_src_node *sn = NULL; 2967 2968 len = pd->tot_len - off - (th->th_off << 2); 2969 2970 /* check maximums */ 2971 if (r->max_states && (r->states >= r->max_states)) { 2972 pf_status.lcounters[LCNT_STATES]++; 2973 REASON_SET(&reason, PFRES_MAXSTATES); 2974 goto cleanup; 2975 } 2976 /* src node for flter rule */ 2977 if ((r->rule_flag & PFRULE_SRCTRACK || 2978 r->rpool.opts & PF_POOL_STICKYADDR) && 2979 pf_insert_src_node(&sn, r, saddr, af) != 0) { 2980 REASON_SET(&reason, PFRES_SRCLIMIT); 2981 goto cleanup; 2982 } 2983 /* src node for translation rule */ 2984 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 2985 ((direction == PF_OUT && 2986 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || 2987 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { 2988 REASON_SET(&reason, PFRES_SRCLIMIT); 2989 goto cleanup; 2990 } 2991 s = pool_get(&pf_state_pl, PR_NOWAIT); 2992 if (s == NULL) { 2993 REASON_SET(&reason, PFRES_MEMORY); 2994 cleanup: 2995 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 2996 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 2997 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 2998 pf_status.src_nodes--; 2999 pool_put(&pf_src_tree_pl, sn); 3000 } 3001 if (nsn != sn && nsn != NULL && nsn->states == 0 && 3002 nsn->expire == 0) { 3003 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 3004 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3005 pf_status.src_nodes--; 3006 pool_put(&pf_src_tree_pl, nsn); 3007 } 3008 return (PF_DROP); 3009 } 3010 bzero(s, sizeof(*s)); 3011 s->rule.ptr = r; 3012 s->nat_rule.ptr = nr; 3013 s->anchor.ptr = a; 3014 STATE_INC_COUNTERS(s); 3015 s->allow_opts = r->allow_opts; 3016 s->log = r->log & 2; 3017 s->proto = IPPROTO_TCP; 3018 s->direction = direction; 3019 s->af = af; 3020 if (direction == PF_OUT) { 3021 PF_ACPY(&s->gwy.addr, saddr, af); 3022 s->gwy.port = th->th_sport; /* sport */ 3023 PF_ACPY(&s->ext.addr, daddr, af); 3024 s->ext.port = th->th_dport; 3025 if (nr != NULL) { 3026 PF_ACPY(&s->lan.addr, &pd->baddr, af); 3027 s->lan.port = bport; 3028 } else { 3029 PF_ACPY(&s->lan.addr, &s->gwy.addr, af); 3030 s->lan.port = s->gwy.port; 3031 } 3032 } else { 3033 PF_ACPY(&s->lan.addr, daddr, af); 3034 s->lan.port = th->th_dport; 3035 PF_ACPY(&s->ext.addr, saddr, af); 3036 s->ext.port = th->th_sport; 3037 if (nr != NULL) { 3038 PF_ACPY(&s->gwy.addr, &pd->baddr, af); 3039 s->gwy.port = bport; 3040 } else { 3041 PF_ACPY(&s->gwy.addr, &s->lan.addr, af); 3042 s->gwy.port = s->lan.port; 3043 } 3044 } 3045 3046 s->src.seqlo = ntohl(th->th_seq); 3047 s->src.seqhi = s->src.seqlo + len + 1; 3048 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN && 3049 r->keep_state == PF_STATE_MODULATE) { 3050 /* Generate sequence number modulator */ 3051 while ((s->src.seqdiff = htonl(arc4random())) == 0) 3052 ; 3053 pf_change_a(&th->th_seq, &th->th_sum, 3054 htonl(s->src.seqlo + s->src.seqdiff), 0); 3055 rewrite = 1; 3056 } else 3057 s->src.seqdiff = 0; 3058 if (th->th_flags & TH_SYN) { 3059 s->src.seqhi++; 3060 s->src.wscale = pf_get_wscale(m, off, th->th_off, af); 3061 } 3062 s->src.max_win = MAX(ntohs(th->th_win), 1); 3063 if (s->src.wscale & PF_WSCALE_MASK) { 3064 /* Remove scale factor from initial window */ 3065 int win = s->src.max_win; 3066 win += 1 << (s->src.wscale & PF_WSCALE_MASK); 3067 s->src.max_win = (win - 1) >> 3068 (s->src.wscale & PF_WSCALE_MASK); 3069 } 3070 if (th->th_flags & TH_FIN) 3071 s->src.seqhi++; 3072 s->dst.seqhi = 1; 3073 s->dst.max_win = 1; 3074 s->src.state = TCPS_SYN_SENT; 3075 s->dst.state = TCPS_CLOSED; 3076 s->creation = time_second; 3077 s->expire = time_second; 3078 s->timeout = PFTM_TCP_FIRST_PACKET; 3079 pf_set_rt_ifp(s, saddr); 3080 if (sn != NULL) { 3081 s->src_node = sn; 3082 s->src_node->states++; 3083 } 3084 if (nsn != NULL) { 3085 PF_ACPY(&nsn->raddr, &pd->naddr, af); 3086 s->nat_src_node = nsn; 3087 s->nat_src_node->states++; 3088 } 3089 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m, 3090 off, pd, th, &s->src, &s->dst)) { 3091 REASON_SET(&reason, PFRES_MEMORY); 3092 pf_src_tree_remove_state(s); 3093 STATE_DEC_COUNTERS(s); 3094 pool_put(&pf_state_pl, s); 3095 return (PF_DROP); 3096 } 3097 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub && 3098 pf_normalize_tcp_stateful(m, off, pd, &reason, th, s, 3099 &s->src, &s->dst, &rewrite)) { 3100 /* This really shouldn't happen!!! */ 3101 DPFPRINTF(PF_DEBUG_URGENT, 3102 ("pf_normalize_tcp_stateful failed on first pkt")); 3103 pf_normalize_tcp_cleanup(s); 3104 pf_src_tree_remove_state(s); 3105 STATE_DEC_COUNTERS(s); 3106 pool_put(&pf_state_pl, s); 3107 return (PF_DROP); 3108 } 3109 if (pf_insert_state(BOUND_IFACE(r, kif), s)) { 3110 pf_normalize_tcp_cleanup(s); 3111 REASON_SET(&reason, PFRES_STATEINS); 3112 pf_src_tree_remove_state(s); 3113 STATE_DEC_COUNTERS(s); 3114 pool_put(&pf_state_pl, s); 3115 return (PF_DROP); 3116 } else 3117 *sm = s; 3118 if (tag > 0) { 3119 pf_tag_ref(tag); 3120 s->tag = tag; 3121 } 3122 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN && 3123 r->keep_state == PF_STATE_SYNPROXY) { 3124 s->src.state = PF_TCPS_PROXY_SRC; 3125 if (nr != NULL) { 3126 if (direction == PF_OUT) { 3127 pf_change_ap(saddr, &th->th_sport, 3128 pd->ip_sum, &th->th_sum, &pd->baddr, 3129 bport, 0, af); 3130 } else { 3131 pf_change_ap(daddr, &th->th_dport, 3132 pd->ip_sum, &th->th_sum, &pd->baddr, 3133 bport, 0, af); 3134 } 3135 } 3136 s->src.seqhi = htonl(arc4random()); 3137 /* Find mss option */ 3138 mss = pf_get_mss(m, off, th->th_off, af); 3139 mss = pf_calc_mss(saddr, af, mss); 3140 mss = pf_calc_mss(daddr, af, mss); 3141 s->src.mss = mss; 3142 pf_send_tcp(r, af, daddr, saddr, th->th_dport, 3143 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1, 3144 TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL); 3145 REASON_SET(&reason, PFRES_SYNPROXY); 3146 return (PF_SYNPROXY_DROP); 3147 } 3148 } 3149 3150 /* copy back packet headers if we performed NAT operations */ 3151 if (rewrite) 3152 m_copyback(m, off, sizeof(*th), th); 3153 3154 return (PF_PASS); 3155 } 3156 3157 int 3158 pf_test_udp(struct pf_rule **rm, struct pf_state **sm, int direction, 3159 struct pfi_kif *kif, struct mbuf *m, int off, void *h __unused, 3160 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3161 struct ifqueue *ifq) 3162 { 3163 struct pf_rule *nr = NULL; 3164 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 3165 struct udphdr *uh = pd->hdr.udp; 3166 u_int16_t bport, nport = 0; 3167 sa_family_t af = pd->af; 3168 int lookup = -1; 3169 uid_t uid = 0; /* XXX: GCC */ 3170 gid_t gid = 0; /* XXX: GCC */ 3171 struct pf_rule *r, *a = NULL; 3172 struct pf_ruleset *ruleset = NULL; 3173 struct pf_src_node *nsn = NULL; 3174 u_short reason; 3175 int rewrite = 0; 3176 struct pf_tag *pftag = NULL; 3177 int tag = -1; 3178 int asd = 0; 3179 3180 if (pf_check_congestion(ifq)) { 3181 REASON_SET(&reason, PFRES_CONGEST); 3182 return (PF_DROP); 3183 } 3184 3185 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 3186 3187 if (direction == PF_OUT) { 3188 bport = nport = uh->uh_sport; 3189 /* check outgoing packet for BINAT/NAT */ 3190 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, 3191 saddr, uh->uh_sport, daddr, uh->uh_dport, 3192 &pd->naddr, &nport)) != NULL) { 3193 PF_ACPY(&pd->baddr, saddr, af); 3194 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum, 3195 &uh->uh_sum, &pd->naddr, nport, 1, af); 3196 rewrite++; 3197 if (nr->natpass) 3198 r = NULL; 3199 pd->nat_rule = nr; 3200 } 3201 } else { 3202 bport = nport = uh->uh_dport; 3203 /* check incoming packet for BINAT/RDR */ 3204 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, 3205 saddr, uh->uh_sport, daddr, uh->uh_dport, &pd->naddr, 3206 &nport)) != NULL) { 3207 PF_ACPY(&pd->baddr, daddr, af); 3208 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum, 3209 &uh->uh_sum, &pd->naddr, nport, 1, af); 3210 rewrite++; 3211 if (nr->natpass) 3212 r = NULL; 3213 pd->nat_rule = nr; 3214 } 3215 } 3216 3217 while (r != NULL) { 3218 r->evaluations++; 3219 if (r->kif != NULL && 3220 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot) 3221 r = r->skip[PF_SKIP_IFP].ptr; 3222 else if (r->direction && r->direction != direction) 3223 r = r->skip[PF_SKIP_DIR].ptr; 3224 else if (r->af && r->af != af) 3225 r = r->skip[PF_SKIP_AF].ptr; 3226 else if (r->proto && r->proto != IPPROTO_UDP) 3227 r = r->skip[PF_SKIP_PROTO].ptr; 3228 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.neg)) 3229 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 3230 else if (r->src.port_op && !pf_match_port(r->src.port_op, 3231 r->src.port[0], r->src.port[1], uh->uh_sport)) 3232 r = r->skip[PF_SKIP_SRC_PORT].ptr; 3233 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.neg)) 3234 r = r->skip[PF_SKIP_DST_ADDR].ptr; 3235 else if (r->dst.port_op && !pf_match_port(r->dst.port_op, 3236 r->dst.port[0], r->dst.port[1], uh->uh_dport)) 3237 r = r->skip[PF_SKIP_DST_PORT].ptr; 3238 else if (r->tos && !(r->tos & pd->tos)) 3239 r = TAILQ_NEXT(r, entries); 3240 else if (r->rule_flag & PFRULE_FRAGMENT) 3241 r = TAILQ_NEXT(r, entries); 3242 else if (r->uid.op && (lookup != -1 || (lookup = 3243 pf_socket_lookup(&uid, &gid, direction, pd), 1)) && 3244 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1], 3245 uid)) 3246 r = TAILQ_NEXT(r, entries); 3247 else if (r->gid.op && (lookup != -1 || (lookup = 3248 pf_socket_lookup(&uid, &gid, direction, pd), 1)) && 3249 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1], 3250 gid)) 3251 r = TAILQ_NEXT(r, entries); 3252 else if (r->prob && r->prob <= arc4random()) 3253 r = TAILQ_NEXT(r, entries); 3254 else if (r->match_tag && !pf_match_tag(m, r, &pftag, &tag)) 3255 r = TAILQ_NEXT(r, entries); 3256 else if (r->os_fingerprint != PF_OSFP_ANY) 3257 r = TAILQ_NEXT(r, entries); 3258 else { 3259 if (r->tag) 3260 tag = r->tag; 3261 if (r->anchor == NULL) { 3262 *rm = r; 3263 *am = a; 3264 *rsm = ruleset; 3265 if ((*rm)->quick) 3266 break; 3267 r = TAILQ_NEXT(r, entries); 3268 } else 3269 pf_step_into_anchor(&asd, &ruleset, 3270 PF_RULESET_FILTER, &r, &a); 3271 } 3272 if (r == NULL) 3273 pf_step_out_of_anchor(&asd, &ruleset, 3274 PF_RULESET_FILTER, &r, &a); 3275 } 3276 r = *rm; 3277 a = *am; 3278 ruleset = *rsm; 3279 3280 REASON_SET(&reason, PFRES_MATCH); 3281 3282 if (r->log) { 3283 if (rewrite) 3284 m_copyback(m, off, sizeof(*uh), uh); 3285 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset); 3286 } 3287 3288 if ((r->action == PF_DROP) && 3289 ((r->rule_flag & PFRULE_RETURNICMP) || 3290 (r->rule_flag & PFRULE_RETURN))) { 3291 /* undo NAT changes, if they have taken place */ 3292 if (nr != NULL) { 3293 if (direction == PF_OUT) { 3294 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum, 3295 &uh->uh_sum, &pd->baddr, bport, 1, af); 3296 rewrite++; 3297 } else { 3298 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum, 3299 &uh->uh_sum, &pd->baddr, bport, 1, af); 3300 rewrite++; 3301 } 3302 } 3303 if ((af == AF_INET) && r->return_icmp) 3304 pf_send_icmp(m, r->return_icmp >> 8, 3305 r->return_icmp & 255, af, r); 3306 else if ((af == AF_INET6) && r->return_icmp6) 3307 pf_send_icmp(m, r->return_icmp6 >> 8, 3308 r->return_icmp6 & 255, af, r); 3309 } 3310 3311 if (r->action == PF_DROP) 3312 return (PF_DROP); 3313 3314 if (pf_tag_packet(m, pftag, tag)) { 3315 REASON_SET(&reason, PFRES_MEMORY); 3316 return (PF_DROP); 3317 } 3318 3319 if (r->keep_state || nr != NULL) { 3320 /* create new state */ 3321 struct pf_state *s = NULL; 3322 struct pf_src_node *sn = NULL; 3323 3324 /* check maximums */ 3325 if (r->max_states && (r->states >= r->max_states)) { 3326 pf_status.lcounters[LCNT_STATES]++; 3327 REASON_SET(&reason, PFRES_MAXSTATES); 3328 goto cleanup; 3329 } 3330 /* src node for flter rule */ 3331 if ((r->rule_flag & PFRULE_SRCTRACK || 3332 r->rpool.opts & PF_POOL_STICKYADDR) && 3333 pf_insert_src_node(&sn, r, saddr, af) != 0) { 3334 REASON_SET(&reason, PFRES_SRCLIMIT); 3335 goto cleanup; 3336 } 3337 /* src node for translation rule */ 3338 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 3339 ((direction == PF_OUT && 3340 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || 3341 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { 3342 REASON_SET(&reason, PFRES_SRCLIMIT); 3343 goto cleanup; 3344 } 3345 s = pool_get(&pf_state_pl, PR_NOWAIT); 3346 if (s == NULL) { 3347 REASON_SET(&reason, PFRES_MEMORY); 3348 cleanup: 3349 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 3350 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 3351 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3352 pf_status.src_nodes--; 3353 pool_put(&pf_src_tree_pl, sn); 3354 } 3355 if (nsn != sn && nsn != NULL && nsn->states == 0 && 3356 nsn->expire == 0) { 3357 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 3358 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3359 pf_status.src_nodes--; 3360 pool_put(&pf_src_tree_pl, nsn); 3361 } 3362 return (PF_DROP); 3363 } 3364 bzero(s, sizeof(*s)); 3365 s->rule.ptr = r; 3366 s->nat_rule.ptr = nr; 3367 s->anchor.ptr = a; 3368 STATE_INC_COUNTERS(s); 3369 s->allow_opts = r->allow_opts; 3370 s->log = r->log & 2; 3371 s->proto = IPPROTO_UDP; 3372 s->direction = direction; 3373 s->af = af; 3374 if (direction == PF_OUT) { 3375 PF_ACPY(&s->gwy.addr, saddr, af); 3376 s->gwy.port = uh->uh_sport; 3377 PF_ACPY(&s->ext.addr, daddr, af); 3378 s->ext.port = uh->uh_dport; 3379 if (nr != NULL) { 3380 PF_ACPY(&s->lan.addr, &pd->baddr, af); 3381 s->lan.port = bport; 3382 } else { 3383 PF_ACPY(&s->lan.addr, &s->gwy.addr, af); 3384 s->lan.port = s->gwy.port; 3385 } 3386 } else { 3387 PF_ACPY(&s->lan.addr, daddr, af); 3388 s->lan.port = uh->uh_dport; 3389 PF_ACPY(&s->ext.addr, saddr, af); 3390 s->ext.port = uh->uh_sport; 3391 if (nr != NULL) { 3392 PF_ACPY(&s->gwy.addr, &pd->baddr, af); 3393 s->gwy.port = bport; 3394 } else { 3395 PF_ACPY(&s->gwy.addr, &s->lan.addr, af); 3396 s->gwy.port = s->lan.port; 3397 } 3398 } 3399 s->src.state = PFUDPS_SINGLE; 3400 s->dst.state = PFUDPS_NO_TRAFFIC; 3401 s->creation = time_second; 3402 s->expire = time_second; 3403 s->timeout = PFTM_UDP_FIRST_PACKET; 3404 pf_set_rt_ifp(s, saddr); 3405 if (sn != NULL) { 3406 s->src_node = sn; 3407 s->src_node->states++; 3408 } 3409 if (nsn != NULL) { 3410 PF_ACPY(&nsn->raddr, &pd->naddr, af); 3411 s->nat_src_node = nsn; 3412 s->nat_src_node->states++; 3413 } 3414 if (pf_insert_state(BOUND_IFACE(r, kif), s)) { 3415 REASON_SET(&reason, PFRES_STATEINS); 3416 pf_src_tree_remove_state(s); 3417 STATE_DEC_COUNTERS(s); 3418 pool_put(&pf_state_pl, s); 3419 return (PF_DROP); 3420 } else 3421 *sm = s; 3422 if (tag > 0) { 3423 pf_tag_ref(tag); 3424 s->tag = tag; 3425 } 3426 } 3427 3428 /* copy back packet headers if we performed NAT operations */ 3429 if (rewrite) 3430 m_copyback(m, off, sizeof(*uh), uh); 3431 3432 return (PF_PASS); 3433 } 3434 3435 int 3436 pf_test_icmp(struct pf_rule **rm, struct pf_state **sm, int direction, 3437 struct pfi_kif *kif, struct mbuf *m, int off, void *h __unused, 3438 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3439 struct ifqueue *ifq) 3440 { 3441 struct pf_rule *nr = NULL; 3442 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 3443 struct pf_rule *r, *a = NULL; 3444 struct pf_ruleset *ruleset = NULL; 3445 struct pf_src_node *nsn = NULL; 3446 u_short reason; 3447 u_int16_t icmpid = 0; 3448 sa_family_t af = pd->af; 3449 u_int8_t icmptype = 0, icmpcode = 0; 3450 int state_icmp = 0; 3451 struct pf_tag *pftag = NULL; 3452 int tag = -1; 3453 #ifdef INET6 3454 int rewrite = 0; 3455 #endif /* INET6 */ 3456 int asd = 0; 3457 3458 if (pf_check_congestion(ifq)) { 3459 REASON_SET(&reason, PFRES_CONGEST); 3460 return (PF_DROP); 3461 } 3462 3463 switch (pd->proto) { 3464 #ifdef INET 3465 case IPPROTO_ICMP: 3466 icmptype = pd->hdr.icmp->icmp_type; 3467 icmpcode = pd->hdr.icmp->icmp_code; 3468 icmpid = pd->hdr.icmp->icmp_id; 3469 3470 if (icmptype == ICMP_UNREACH || 3471 icmptype == ICMP_SOURCEQUENCH || 3472 icmptype == ICMP_REDIRECT || 3473 icmptype == ICMP_TIMXCEED || 3474 icmptype == ICMP_PARAMPROB) 3475 state_icmp++; 3476 break; 3477 #endif /* INET */ 3478 #ifdef INET6 3479 case IPPROTO_ICMPV6: 3480 icmptype = pd->hdr.icmp6->icmp6_type; 3481 icmpcode = pd->hdr.icmp6->icmp6_code; 3482 icmpid = pd->hdr.icmp6->icmp6_id; 3483 3484 if (icmptype == ICMP6_DST_UNREACH || 3485 icmptype == ICMP6_PACKET_TOO_BIG || 3486 icmptype == ICMP6_TIME_EXCEEDED || 3487 icmptype == ICMP6_PARAM_PROB) 3488 state_icmp++; 3489 break; 3490 #endif /* INET6 */ 3491 } 3492 3493 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 3494 3495 if (direction == PF_OUT) { 3496 /* check outgoing packet for BINAT/NAT */ 3497 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, 3498 saddr, icmpid, daddr, icmpid, &pd->naddr, NULL)) != NULL) { 3499 PF_ACPY(&pd->baddr, saddr, af); 3500 switch (af) { 3501 #ifdef INET 3502 case AF_INET: 3503 pf_change_a(&saddr->v4.s_addr, pd->ip_sum, 3504 pd->naddr.v4.s_addr, 0); 3505 break; 3506 #endif /* INET */ 3507 #ifdef INET6 3508 case AF_INET6: 3509 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum, 3510 &pd->naddr, 0); 3511 rewrite++; 3512 break; 3513 #endif /* INET6 */ 3514 } 3515 if (nr->natpass) 3516 r = NULL; 3517 pd->nat_rule = nr; 3518 } 3519 } else { 3520 /* check incoming packet for BINAT/RDR */ 3521 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, 3522 saddr, icmpid, daddr, icmpid, &pd->naddr, NULL)) != NULL) { 3523 PF_ACPY(&pd->baddr, daddr, af); 3524 switch (af) { 3525 #ifdef INET 3526 case AF_INET: 3527 pf_change_a(&daddr->v4.s_addr, 3528 pd->ip_sum, pd->naddr.v4.s_addr, 0); 3529 break; 3530 #endif /* INET */ 3531 #ifdef INET6 3532 case AF_INET6: 3533 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum, 3534 &pd->naddr, 0); 3535 rewrite++; 3536 break; 3537 #endif /* INET6 */ 3538 } 3539 if (nr->natpass) 3540 r = NULL; 3541 pd->nat_rule = nr; 3542 } 3543 } 3544 3545 while (r != NULL) { 3546 r->evaluations++; 3547 if (r->kif != NULL && 3548 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot) 3549 r = r->skip[PF_SKIP_IFP].ptr; 3550 else if (r->direction && r->direction != direction) 3551 r = r->skip[PF_SKIP_DIR].ptr; 3552 else if (r->af && r->af != af) 3553 r = r->skip[PF_SKIP_AF].ptr; 3554 else if (r->proto && r->proto != pd->proto) 3555 r = r->skip[PF_SKIP_PROTO].ptr; 3556 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.neg)) 3557 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 3558 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.neg)) 3559 r = r->skip[PF_SKIP_DST_ADDR].ptr; 3560 else if (r->type && r->type != icmptype + 1) 3561 r = TAILQ_NEXT(r, entries); 3562 else if (r->code && r->code != icmpcode + 1) 3563 r = TAILQ_NEXT(r, entries); 3564 else if (r->tos && !(r->tos & pd->tos)) 3565 r = TAILQ_NEXT(r, entries); 3566 else if (r->rule_flag & PFRULE_FRAGMENT) 3567 r = TAILQ_NEXT(r, entries); 3568 else if (r->prob && r->prob <= arc4random()) 3569 r = TAILQ_NEXT(r, entries); 3570 else if (r->match_tag && !pf_match_tag(m, r, &pftag, &tag)) 3571 r = TAILQ_NEXT(r, entries); 3572 else if (r->os_fingerprint != PF_OSFP_ANY) 3573 r = TAILQ_NEXT(r, entries); 3574 else { 3575 if (r->tag) 3576 tag = r->tag; 3577 if (r->anchor == NULL) { 3578 *rm = r; 3579 *am = a; 3580 *rsm = ruleset; 3581 if ((*rm)->quick) 3582 break; 3583 r = TAILQ_NEXT(r, entries); 3584 } else 3585 pf_step_into_anchor(&asd, &ruleset, 3586 PF_RULESET_FILTER, &r, &a); 3587 } 3588 if (r == NULL) 3589 pf_step_out_of_anchor(&asd, &ruleset, 3590 PF_RULESET_FILTER, &r, &a); 3591 } 3592 r = *rm; 3593 a = *am; 3594 ruleset = *rsm; 3595 3596 REASON_SET(&reason, PFRES_MATCH); 3597 3598 if (r->log) { 3599 #ifdef INET6 3600 if (rewrite) 3601 m_copyback(m, off, sizeof(struct icmp6_hdr), 3602 pd->hdr.icmp6); 3603 #endif /* INET6 */ 3604 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset); 3605 } 3606 3607 if (r->action != PF_PASS) 3608 return (PF_DROP); 3609 3610 if (pf_tag_packet(m, pftag, tag)) { 3611 REASON_SET(&reason, PFRES_MEMORY); 3612 return (PF_DROP); 3613 } 3614 3615 if (!state_icmp && (r->keep_state || nr != NULL)) { 3616 /* create new state */ 3617 struct pf_state *s = NULL; 3618 struct pf_src_node *sn = NULL; 3619 3620 /* check maximums */ 3621 if (r->max_states && (r->states >= r->max_states)) { 3622 pf_status.lcounters[LCNT_STATES]++; 3623 REASON_SET(&reason, PFRES_MAXSTATES); 3624 goto cleanup; 3625 } 3626 /* src node for flter rule */ 3627 if ((r->rule_flag & PFRULE_SRCTRACK || 3628 r->rpool.opts & PF_POOL_STICKYADDR) && 3629 pf_insert_src_node(&sn, r, saddr, af) != 0) { 3630 REASON_SET(&reason, PFRES_SRCLIMIT); 3631 goto cleanup; 3632 } 3633 /* src node for translation rule */ 3634 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 3635 ((direction == PF_OUT && 3636 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || 3637 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { 3638 REASON_SET(&reason, PFRES_SRCLIMIT); 3639 goto cleanup; 3640 } 3641 s = pool_get(&pf_state_pl, PR_NOWAIT); 3642 if (s == NULL) { 3643 REASON_SET(&reason, PFRES_MEMORY); 3644 cleanup: 3645 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 3646 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 3647 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3648 pf_status.src_nodes--; 3649 pool_put(&pf_src_tree_pl, sn); 3650 } 3651 if (nsn != sn && nsn != NULL && nsn->states == 0 && 3652 nsn->expire == 0) { 3653 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 3654 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3655 pf_status.src_nodes--; 3656 pool_put(&pf_src_tree_pl, nsn); 3657 } 3658 return (PF_DROP); 3659 } 3660 bzero(s, sizeof(*s)); 3661 s->rule.ptr = r; 3662 s->nat_rule.ptr = nr; 3663 s->anchor.ptr = a; 3664 STATE_INC_COUNTERS(s); 3665 s->allow_opts = r->allow_opts; 3666 s->log = r->log & 2; 3667 s->proto = pd->proto; 3668 s->direction = direction; 3669 s->af = af; 3670 if (direction == PF_OUT) { 3671 PF_ACPY(&s->gwy.addr, saddr, af); 3672 s->gwy.port = icmpid; 3673 PF_ACPY(&s->ext.addr, daddr, af); 3674 s->ext.port = icmpid; 3675 if (nr != NULL) 3676 PF_ACPY(&s->lan.addr, &pd->baddr, af); 3677 else 3678 PF_ACPY(&s->lan.addr, &s->gwy.addr, af); 3679 s->lan.port = icmpid; 3680 } else { 3681 PF_ACPY(&s->lan.addr, daddr, af); 3682 s->lan.port = icmpid; 3683 PF_ACPY(&s->ext.addr, saddr, af); 3684 s->ext.port = icmpid; 3685 if (nr != NULL) 3686 PF_ACPY(&s->gwy.addr, &pd->baddr, af); 3687 else 3688 PF_ACPY(&s->gwy.addr, &s->lan.addr, af); 3689 s->gwy.port = icmpid; 3690 } 3691 s->creation = time_second; 3692 s->expire = time_second; 3693 s->timeout = PFTM_ICMP_FIRST_PACKET; 3694 pf_set_rt_ifp(s, saddr); 3695 if (sn != NULL) { 3696 s->src_node = sn; 3697 s->src_node->states++; 3698 } 3699 if (nsn != NULL) { 3700 PF_ACPY(&nsn->raddr, &pd->naddr, af); 3701 s->nat_src_node = nsn; 3702 s->nat_src_node->states++; 3703 } 3704 if (pf_insert_state(BOUND_IFACE(r, kif), s)) { 3705 REASON_SET(&reason, PFRES_STATEINS); 3706 pf_src_tree_remove_state(s); 3707 STATE_DEC_COUNTERS(s); 3708 pool_put(&pf_state_pl, s); 3709 return (PF_DROP); 3710 } else 3711 *sm = s; 3712 if (tag > 0) { 3713 pf_tag_ref(tag); 3714 s->tag = tag; 3715 } 3716 } 3717 3718 #ifdef INET6 3719 /* copy back packet headers if we performed IPv6 NAT operations */ 3720 if (rewrite) 3721 m_copyback(m, off, sizeof(struct icmp6_hdr), 3722 pd->hdr.icmp6); 3723 #endif /* INET6 */ 3724 3725 return (PF_PASS); 3726 } 3727 3728 int 3729 pf_test_other(struct pf_rule **rm, struct pf_state **sm, int direction, 3730 struct pfi_kif *kif, struct mbuf *m, int off, void *h __unused, 3731 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3732 struct ifqueue *ifq) 3733 { 3734 struct pf_rule *nr = NULL; 3735 struct pf_rule *r, *a = NULL; 3736 struct pf_ruleset *ruleset = NULL; 3737 struct pf_src_node *nsn = NULL; 3738 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 3739 sa_family_t af = pd->af; 3740 u_short reason; 3741 struct pf_tag *pftag = NULL; 3742 int tag = -1; 3743 int asd = 0; 3744 3745 if (pf_check_congestion(ifq)) { 3746 REASON_SET(&reason, PFRES_CONGEST); 3747 return (PF_DROP); 3748 } 3749 3750 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 3751 3752 if (direction == PF_OUT) { 3753 /* check outgoing packet for BINAT/NAT */ 3754 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn, 3755 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) { 3756 PF_ACPY(&pd->baddr, saddr, af); 3757 switch (af) { 3758 #ifdef INET 3759 case AF_INET: 3760 pf_change_a(&saddr->v4.s_addr, pd->ip_sum, 3761 pd->naddr.v4.s_addr, 0); 3762 break; 3763 #endif /* INET */ 3764 #ifdef INET6 3765 case AF_INET6: 3766 PF_ACPY(saddr, &pd->naddr, af); 3767 break; 3768 #endif /* INET6 */ 3769 } 3770 if (nr->natpass) 3771 r = NULL; 3772 pd->nat_rule = nr; 3773 } 3774 } else { 3775 /* check incoming packet for BINAT/RDR */ 3776 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn, 3777 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) { 3778 PF_ACPY(&pd->baddr, daddr, af); 3779 switch (af) { 3780 #ifdef INET 3781 case AF_INET: 3782 pf_change_a(&daddr->v4.s_addr, 3783 pd->ip_sum, pd->naddr.v4.s_addr, 0); 3784 break; 3785 #endif /* INET */ 3786 #ifdef INET6 3787 case AF_INET6: 3788 PF_ACPY(daddr, &pd->naddr, af); 3789 break; 3790 #endif /* INET6 */ 3791 } 3792 if (nr->natpass) 3793 r = NULL; 3794 pd->nat_rule = nr; 3795 } 3796 } 3797 3798 while (r != NULL) { 3799 r->evaluations++; 3800 if (r->kif != NULL && 3801 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot) 3802 r = r->skip[PF_SKIP_IFP].ptr; 3803 else if (r->direction && r->direction != direction) 3804 r = r->skip[PF_SKIP_DIR].ptr; 3805 else if (r->af && r->af != af) 3806 r = r->skip[PF_SKIP_AF].ptr; 3807 else if (r->proto && r->proto != pd->proto) 3808 r = r->skip[PF_SKIP_PROTO].ptr; 3809 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.neg)) 3810 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 3811 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.neg)) 3812 r = r->skip[PF_SKIP_DST_ADDR].ptr; 3813 else if (r->tos && !(r->tos & pd->tos)) 3814 r = TAILQ_NEXT(r, entries); 3815 else if (r->rule_flag & PFRULE_FRAGMENT) 3816 r = TAILQ_NEXT(r, entries); 3817 else if (r->prob && r->prob <= arc4random()) 3818 r = TAILQ_NEXT(r, entries); 3819 else if (r->match_tag && !pf_match_tag(m, r, &pftag, &tag)) 3820 r = TAILQ_NEXT(r, entries); 3821 else if (r->os_fingerprint != PF_OSFP_ANY) 3822 r = TAILQ_NEXT(r, entries); 3823 else { 3824 if (r->tag) 3825 tag = r->tag; 3826 if (r->anchor == NULL) { 3827 *rm = r; 3828 *am = a; 3829 *rsm = ruleset; 3830 if ((*rm)->quick) 3831 break; 3832 r = TAILQ_NEXT(r, entries); 3833 } else 3834 pf_step_into_anchor(&asd, &ruleset, 3835 PF_RULESET_FILTER, &r, &a); 3836 } 3837 if (r == NULL) 3838 pf_step_out_of_anchor(&asd, &ruleset, 3839 PF_RULESET_FILTER, &r, &a); 3840 } 3841 r = *rm; 3842 a = *am; 3843 ruleset = *rsm; 3844 3845 REASON_SET(&reason, PFRES_MATCH); 3846 3847 if (r->log) 3848 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset); 3849 3850 if ((r->action == PF_DROP) && 3851 ((r->rule_flag & PFRULE_RETURNICMP) || 3852 (r->rule_flag & PFRULE_RETURN))) { 3853 struct pf_addr *a = NULL; 3854 3855 if (nr != NULL) { 3856 if (direction == PF_OUT) 3857 a = saddr; 3858 else 3859 a = daddr; 3860 } 3861 if (a != NULL) { 3862 switch (af) { 3863 #ifdef INET 3864 case AF_INET: 3865 pf_change_a(&a->v4.s_addr, pd->ip_sum, 3866 pd->baddr.v4.s_addr, 0); 3867 break; 3868 #endif /* INET */ 3869 #ifdef INET6 3870 case AF_INET6: 3871 PF_ACPY(a, &pd->baddr, af); 3872 break; 3873 #endif /* INET6 */ 3874 } 3875 } 3876 if ((af == AF_INET) && r->return_icmp) 3877 pf_send_icmp(m, r->return_icmp >> 8, 3878 r->return_icmp & 255, af, r); 3879 else if ((af == AF_INET6) && r->return_icmp6) 3880 pf_send_icmp(m, r->return_icmp6 >> 8, 3881 r->return_icmp6 & 255, af, r); 3882 } 3883 3884 if (r->action != PF_PASS) 3885 return (PF_DROP); 3886 3887 if (pf_tag_packet(m, pftag, tag)) { 3888 REASON_SET(&reason, PFRES_MEMORY); 3889 return (PF_DROP); 3890 } 3891 3892 if (r->keep_state || nr != NULL) { 3893 /* create new state */ 3894 struct pf_state *s = NULL; 3895 struct pf_src_node *sn = NULL; 3896 3897 /* check maximums */ 3898 if (r->max_states && (r->states >= r->max_states)) { 3899 pf_status.lcounters[LCNT_STATES]++; 3900 REASON_SET(&reason, PFRES_MAXSTATES); 3901 goto cleanup; 3902 } 3903 /* src node for flter rule */ 3904 if ((r->rule_flag & PFRULE_SRCTRACK || 3905 r->rpool.opts & PF_POOL_STICKYADDR) && 3906 pf_insert_src_node(&sn, r, saddr, af) != 0) { 3907 REASON_SET(&reason, PFRES_SRCLIMIT); 3908 goto cleanup; 3909 } 3910 /* src node for translation rule */ 3911 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 3912 ((direction == PF_OUT && 3913 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) || 3914 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) { 3915 REASON_SET(&reason, PFRES_SRCLIMIT); 3916 goto cleanup; 3917 } 3918 s = pool_get(&pf_state_pl, PR_NOWAIT); 3919 if (s == NULL) { 3920 REASON_SET(&reason, PFRES_MEMORY); 3921 cleanup: 3922 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 3923 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 3924 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3925 pf_status.src_nodes--; 3926 pool_put(&pf_src_tree_pl, sn); 3927 } 3928 if (nsn != sn && nsn != NULL && nsn->states == 0 && 3929 nsn->expire == 0) { 3930 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 3931 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3932 pf_status.src_nodes--; 3933 pool_put(&pf_src_tree_pl, nsn); 3934 } 3935 return (PF_DROP); 3936 } 3937 bzero(s, sizeof(*s)); 3938 s->rule.ptr = r; 3939 s->nat_rule.ptr = nr; 3940 s->anchor.ptr = a; 3941 STATE_INC_COUNTERS(s); 3942 s->allow_opts = r->allow_opts; 3943 s->log = r->log & 2; 3944 s->proto = pd->proto; 3945 s->direction = direction; 3946 s->af = af; 3947 if (direction == PF_OUT) { 3948 PF_ACPY(&s->gwy.addr, saddr, af); 3949 PF_ACPY(&s->ext.addr, daddr, af); 3950 if (nr != NULL) 3951 PF_ACPY(&s->lan.addr, &pd->baddr, af); 3952 else 3953 PF_ACPY(&s->lan.addr, &s->gwy.addr, af); 3954 } else { 3955 PF_ACPY(&s->lan.addr, daddr, af); 3956 PF_ACPY(&s->ext.addr, saddr, af); 3957 if (nr != NULL) 3958 PF_ACPY(&s->gwy.addr, &pd->baddr, af); 3959 else 3960 PF_ACPY(&s->gwy.addr, &s->lan.addr, af); 3961 } 3962 s->src.state = PFOTHERS_SINGLE; 3963 s->dst.state = PFOTHERS_NO_TRAFFIC; 3964 s->creation = time_second; 3965 s->expire = time_second; 3966 s->timeout = PFTM_OTHER_FIRST_PACKET; 3967 pf_set_rt_ifp(s, saddr); 3968 if (sn != NULL) { 3969 s->src_node = sn; 3970 s->src_node->states++; 3971 } 3972 if (nsn != NULL) { 3973 PF_ACPY(&nsn->raddr, &pd->naddr, af); 3974 s->nat_src_node = nsn; 3975 s->nat_src_node->states++; 3976 } 3977 if (pf_insert_state(BOUND_IFACE(r, kif), s)) { 3978 REASON_SET(&reason, PFRES_STATEINS); 3979 pf_src_tree_remove_state(s); 3980 STATE_DEC_COUNTERS(s); 3981 pool_put(&pf_state_pl, s); 3982 return (PF_DROP); 3983 } else 3984 *sm = s; 3985 if (tag > 0) { 3986 pf_tag_ref(tag); 3987 s->tag = tag; 3988 } 3989 } 3990 3991 return (PF_PASS); 3992 } 3993 3994 int 3995 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif, 3996 struct mbuf *m, void *h __unused, struct pf_pdesc *pd, struct pf_rule **am, 3997 struct pf_ruleset **rsm) 3998 { 3999 struct pf_rule *r, *a = NULL; 4000 struct pf_ruleset *ruleset = NULL; 4001 sa_family_t af = pd->af; 4002 u_short reason; 4003 struct pf_tag *pftag = NULL; 4004 int tag = -1; 4005 int asd = 0; 4006 4007 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 4008 while (r != NULL) { 4009 r->evaluations++; 4010 if (r->kif != NULL && 4011 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot) 4012 r = r->skip[PF_SKIP_IFP].ptr; 4013 else if (r->direction && r->direction != direction) 4014 r = r->skip[PF_SKIP_DIR].ptr; 4015 else if (r->af && r->af != af) 4016 r = r->skip[PF_SKIP_AF].ptr; 4017 else if (r->proto && r->proto != pd->proto) 4018 r = r->skip[PF_SKIP_PROTO].ptr; 4019 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.neg)) 4020 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 4021 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.neg)) 4022 r = r->skip[PF_SKIP_DST_ADDR].ptr; 4023 else if (r->tos && !(r->tos & pd->tos)) 4024 r = TAILQ_NEXT(r, entries); 4025 else if (r->src.port_op || r->dst.port_op || 4026 r->flagset || r->type || r->code || 4027 r->os_fingerprint != PF_OSFP_ANY) 4028 r = TAILQ_NEXT(r, entries); 4029 else if (r->prob && r->prob <= arc4random()) 4030 r = TAILQ_NEXT(r, entries); 4031 else if (r->match_tag && !pf_match_tag(m, r, &pftag, &tag)) 4032 r = TAILQ_NEXT(r, entries); 4033 else { 4034 if (r->anchor == NULL) { 4035 *rm = r; 4036 *am = a; 4037 *rsm = ruleset; 4038 if ((*rm)->quick) 4039 break; 4040 r = TAILQ_NEXT(r, entries); 4041 } else 4042 pf_step_into_anchor(&asd, &ruleset, 4043 PF_RULESET_FILTER, &r, &a); 4044 } 4045 if (r == NULL) 4046 pf_step_out_of_anchor(&asd, &ruleset, 4047 PF_RULESET_FILTER, &r, &a); 4048 } 4049 r = *rm; 4050 a = *am; 4051 ruleset = *rsm; 4052 4053 REASON_SET(&reason, PFRES_MATCH); 4054 4055 if (r->log) 4056 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset); 4057 4058 if (r->action != PF_PASS) 4059 return (PF_DROP); 4060 4061 if (pf_tag_packet(m, pftag, tag)) { 4062 REASON_SET(&reason, PFRES_MEMORY); 4063 return (PF_DROP); 4064 } 4065 4066 return (PF_PASS); 4067 } 4068 4069 int 4070 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif, 4071 struct mbuf *m, int off, void *h __unused, struct pf_pdesc *pd, 4072 u_short *reason) 4073 { 4074 struct pf_state key; 4075 struct tcphdr *th = pd->hdr.tcp; 4076 u_int16_t win = ntohs(th->th_win); 4077 u_int32_t ack, end, seq, orig_seq; 4078 u_int8_t sws, dws; 4079 int ackskew; 4080 int copyback = 0; 4081 struct pf_state_peer *src, *dst; 4082 4083 key.af = pd->af; 4084 key.proto = IPPROTO_TCP; 4085 if (direction == PF_IN) { 4086 PF_ACPY(&key.ext.addr, pd->src, key.af); 4087 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 4088 key.ext.port = th->th_sport; 4089 key.gwy.port = th->th_dport; 4090 } else { 4091 PF_ACPY(&key.lan.addr, pd->src, key.af); 4092 PF_ACPY(&key.ext.addr, pd->dst, key.af); 4093 key.lan.port = th->th_sport; 4094 key.ext.port = th->th_dport; 4095 } 4096 4097 STATE_LOOKUP(); 4098 4099 if (direction == (*state)->direction) { 4100 src = &(*state)->src; 4101 dst = &(*state)->dst; 4102 } else { 4103 src = &(*state)->dst; 4104 dst = &(*state)->src; 4105 } 4106 4107 if ((*state)->src.state == PF_TCPS_PROXY_SRC) { 4108 if (direction != (*state)->direction) { 4109 REASON_SET(reason, PFRES_SYNPROXY); 4110 return (PF_SYNPROXY_DROP); 4111 } 4112 if (th->th_flags & TH_SYN) { 4113 if (ntohl(th->th_seq) != (*state)->src.seqlo) { 4114 REASON_SET(reason, PFRES_SYNPROXY); 4115 return (PF_DROP); 4116 } 4117 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, 4118 pd->src, th->th_dport, th->th_sport, 4119 (*state)->src.seqhi, ntohl(th->th_seq) + 1, 4120 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1, 4121 0, NULL, NULL); 4122 REASON_SET(reason, PFRES_SYNPROXY); 4123 return (PF_SYNPROXY_DROP); 4124 } else if (!(th->th_flags & TH_ACK) || 4125 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) || 4126 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) { 4127 REASON_SET(reason, PFRES_SYNPROXY); 4128 return (PF_DROP); 4129 } else if ((*state)->src_node != NULL && 4130 pf_src_connlimit(state)) { 4131 REASON_SET(reason, PFRES_SRCLIMIT); 4132 return (PF_DROP); 4133 } else 4134 (*state)->src.state = PF_TCPS_PROXY_DST; 4135 } 4136 if ((*state)->src.state == PF_TCPS_PROXY_DST) { 4137 struct pf_state_host *src, *dst; 4138 4139 if (direction == PF_OUT) { 4140 src = &(*state)->gwy; 4141 dst = &(*state)->ext; 4142 } else { 4143 src = &(*state)->ext; 4144 dst = &(*state)->lan; 4145 } 4146 if (direction == (*state)->direction) { 4147 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) || 4148 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) || 4149 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) { 4150 REASON_SET(reason, PFRES_SYNPROXY); 4151 return (PF_DROP); 4152 } 4153 (*state)->src.max_win = MAX(ntohs(th->th_win), 1); 4154 if ((*state)->dst.seqhi == 1) 4155 (*state)->dst.seqhi = htonl(arc4random()); 4156 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr, 4157 &dst->addr, src->port, dst->port, 4158 (*state)->dst.seqhi, 0, TH_SYN, 0, 4159 (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL); 4160 REASON_SET(reason, PFRES_SYNPROXY); 4161 return (PF_SYNPROXY_DROP); 4162 } else if (((th->th_flags & (TH_SYN|TH_ACK)) != 4163 (TH_SYN|TH_ACK)) || 4164 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) { 4165 REASON_SET(reason, PFRES_SYNPROXY); 4166 return (PF_DROP); 4167 } else { 4168 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1); 4169 (*state)->dst.seqlo = ntohl(th->th_seq); 4170 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, 4171 pd->src, th->th_dport, th->th_sport, 4172 ntohl(th->th_ack), ntohl(th->th_seq) + 1, 4173 TH_ACK, (*state)->src.max_win, 0, 0, 0, 4174 (*state)->tag, NULL, NULL); 4175 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr, 4176 &dst->addr, src->port, dst->port, 4177 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1, 4178 TH_ACK, (*state)->dst.max_win, 0, 0, 1, 4179 0, NULL, NULL); 4180 (*state)->src.seqdiff = (*state)->dst.seqhi - 4181 (*state)->src.seqlo; 4182 (*state)->dst.seqdiff = (*state)->src.seqhi - 4183 (*state)->dst.seqlo; 4184 (*state)->src.seqhi = (*state)->src.seqlo + 4185 (*state)->dst.max_win; 4186 (*state)->dst.seqhi = (*state)->dst.seqlo + 4187 (*state)->src.max_win; 4188 (*state)->src.wscale = (*state)->dst.wscale = 0; 4189 (*state)->src.state = (*state)->dst.state = 4190 TCPS_ESTABLISHED; 4191 REASON_SET(reason, PFRES_SYNPROXY); 4192 return (PF_SYNPROXY_DROP); 4193 } 4194 } 4195 4196 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) { 4197 sws = src->wscale & PF_WSCALE_MASK; 4198 dws = dst->wscale & PF_WSCALE_MASK; 4199 } else 4200 sws = dws = 0; 4201 4202 /* 4203 * Sequence tracking algorithm from Guido van Rooij's paper: 4204 * http://www.madison-gurkha.com/publications/tcp_filtering/ 4205 * tcp_filtering.ps 4206 */ 4207 4208 orig_seq = seq = ntohl(th->th_seq); 4209 if (src->seqlo == 0) { 4210 /* First packet from this end. Set its state */ 4211 4212 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) && 4213 src->scrub == NULL) { 4214 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) { 4215 REASON_SET(reason, PFRES_MEMORY); 4216 return (PF_DROP); 4217 } 4218 } 4219 4220 /* Deferred generation of sequence number modulator */ 4221 if (dst->seqdiff && !src->seqdiff) { 4222 while ((src->seqdiff = htonl(arc4random())) == 0) 4223 ; 4224 ack = ntohl(th->th_ack) - dst->seqdiff; 4225 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq + 4226 src->seqdiff), 0); 4227 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0); 4228 copyback = 1; 4229 } else { 4230 ack = ntohl(th->th_ack); 4231 } 4232 4233 end = seq + pd->p_len; 4234 if (th->th_flags & TH_SYN) { 4235 end++; 4236 if (dst->wscale & PF_WSCALE_FLAG) { 4237 src->wscale = pf_get_wscale(m, off, th->th_off, 4238 pd->af); 4239 if (src->wscale & PF_WSCALE_FLAG) { 4240 /* Remove scale factor from initial 4241 * window */ 4242 sws = src->wscale & PF_WSCALE_MASK; 4243 win = ((u_int32_t)win + (1 << sws) - 1) 4244 >> sws; 4245 dws = dst->wscale & PF_WSCALE_MASK; 4246 } else { 4247 /* fixup other window */ 4248 dst->max_win <<= dst->wscale & 4249 PF_WSCALE_MASK; 4250 /* in case of a retrans SYN|ACK */ 4251 dst->wscale = 0; 4252 } 4253 } 4254 } 4255 if (th->th_flags & TH_FIN) 4256 end++; 4257 4258 src->seqlo = seq; 4259 if (src->state < TCPS_SYN_SENT) 4260 src->state = TCPS_SYN_SENT; 4261 4262 /* 4263 * May need to slide the window (seqhi may have been set by 4264 * the crappy stack check or if we picked up the connection 4265 * after establishment) 4266 */ 4267 if (src->seqhi == 1 || 4268 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi)) 4269 src->seqhi = end + MAX(1, dst->max_win << dws); 4270 if (win > src->max_win) 4271 src->max_win = win; 4272 4273 } else { 4274 ack = ntohl(th->th_ack) - dst->seqdiff; 4275 if (src->seqdiff) { 4276 /* Modulate sequence numbers */ 4277 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq + 4278 src->seqdiff), 0); 4279 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0); 4280 copyback = 1; 4281 } 4282 end = seq + pd->p_len; 4283 if (th->th_flags & TH_SYN) 4284 end++; 4285 if (th->th_flags & TH_FIN) 4286 end++; 4287 } 4288 4289 if ((th->th_flags & TH_ACK) == 0) { 4290 /* Let it pass through the ack skew check */ 4291 ack = dst->seqlo; 4292 } else if ((ack == 0 && 4293 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) || 4294 /* broken tcp stacks do not set ack */ 4295 (dst->state < TCPS_SYN_SENT)) { 4296 /* 4297 * Many stacks (ours included) will set the ACK number in an 4298 * FIN|ACK if the SYN times out -- no sequence to ACK. 4299 */ 4300 ack = dst->seqlo; 4301 } 4302 4303 if (seq == end) { 4304 /* Ease sequencing restrictions on no data packets */ 4305 seq = src->seqlo; 4306 end = seq; 4307 } 4308 4309 ackskew = dst->seqlo - ack; 4310 4311 #define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */ 4312 if (SEQ_GEQ(src->seqhi, end) && 4313 /* Last octet inside other's window space */ 4314 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) && 4315 /* Retrans: not more than one window back */ 4316 (ackskew >= -MAXACKWINDOW) && 4317 /* Acking not more than one reassembled fragment backwards */ 4318 (ackskew <= (MAXACKWINDOW << sws)) && 4319 /* Acking not more than one window forward */ 4320 ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo || 4321 (pd->flags & PFDESC_IP_REAS) == 0)) { 4322 /* Require an exact sequence match on resets when possible */ 4323 4324 if (dst->scrub || src->scrub) { 4325 if (pf_normalize_tcp_stateful(m, off, pd, reason, th, 4326 *state, src, dst, ©back)) 4327 return (PF_DROP); 4328 } 4329 4330 /* update max window */ 4331 if (src->max_win < win) 4332 src->max_win = win; 4333 /* synchronize sequencing */ 4334 if (SEQ_GT(end, src->seqlo)) 4335 src->seqlo = end; 4336 /* slide the window of what the other end can send */ 4337 if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) 4338 dst->seqhi = ack + MAX((win << sws), 1); 4339 4340 4341 /* update states */ 4342 if (th->th_flags & TH_SYN) 4343 if (src->state < TCPS_SYN_SENT) 4344 src->state = TCPS_SYN_SENT; 4345 if (th->th_flags & TH_FIN) 4346 if (src->state < TCPS_CLOSING) 4347 src->state = TCPS_CLOSING; 4348 if (th->th_flags & TH_ACK) { 4349 if (dst->state == TCPS_SYN_SENT) { 4350 dst->state = TCPS_ESTABLISHED; 4351 if (src->state == TCPS_ESTABLISHED && 4352 (*state)->src_node != NULL && 4353 pf_src_connlimit(state)) { 4354 REASON_SET(reason, PFRES_SRCLIMIT); 4355 return (PF_DROP); 4356 } 4357 } else if (dst->state == TCPS_CLOSING) 4358 dst->state = TCPS_FIN_WAIT_2; 4359 } 4360 if (th->th_flags & TH_RST) 4361 src->state = dst->state = TCPS_TIME_WAIT; 4362 4363 /* update expire time */ 4364 (*state)->expire = time_second; 4365 if (src->state >= TCPS_FIN_WAIT_2 && 4366 dst->state >= TCPS_FIN_WAIT_2) 4367 (*state)->timeout = PFTM_TCP_CLOSED; 4368 else if (src->state >= TCPS_FIN_WAIT_2 || 4369 dst->state >= TCPS_FIN_WAIT_2) 4370 (*state)->timeout = PFTM_TCP_FIN_WAIT; 4371 else if (src->state < TCPS_ESTABLISHED || 4372 dst->state < TCPS_ESTABLISHED) 4373 (*state)->timeout = PFTM_TCP_OPENING; 4374 else if (src->state >= TCPS_CLOSING || 4375 dst->state >= TCPS_CLOSING) 4376 (*state)->timeout = PFTM_TCP_CLOSING; 4377 else 4378 (*state)->timeout = PFTM_TCP_ESTABLISHED; 4379 4380 /* Fall through to PASS packet */ 4381 4382 } else if ((dst->state < TCPS_SYN_SENT || 4383 dst->state >= TCPS_FIN_WAIT_2 || 4384 src->state >= TCPS_FIN_WAIT_2) && 4385 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) && 4386 /* Within a window forward of the originating packet */ 4387 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) { 4388 /* Within a window backward of the originating packet */ 4389 4390 /* 4391 * This currently handles three situations: 4392 * 1) Stupid stacks will shotgun SYNs before their peer 4393 * replies. 4394 * 2) When PF catches an already established stream (the 4395 * firewall rebooted, the state table was flushed, routes 4396 * changed...) 4397 * 3) Packets get funky immediately after the connection 4398 * closes (this should catch Solaris spurious ACK|FINs 4399 * that web servers like to spew after a close) 4400 * 4401 * This must be a little more careful than the above code 4402 * since packet floods will also be caught here. We don't 4403 * update the TTL here to mitigate the damage of a packet 4404 * flood and so the same code can handle awkward establishment 4405 * and a loosened connection close. 4406 * In the establishment case, a correct peer response will 4407 * validate the connection, go through the normal state code 4408 * and keep updating the state TTL. 4409 */ 4410 4411 if (pf_status.debug >= PF_DEBUG_MISC) { 4412 printf("pf: loose state match: "); 4413 pf_print_state(*state); 4414 pf_print_flags(th->th_flags); 4415 printf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d\n", 4416 seq, ack, pd->p_len, ackskew, 4417 (*state)->packets[0], (*state)->packets[1]); 4418 } 4419 4420 if (dst->scrub || src->scrub) { 4421 if (pf_normalize_tcp_stateful(m, off, pd, reason, th, 4422 *state, src, dst, ©back)) 4423 return (PF_DROP); 4424 } 4425 4426 /* update max window */ 4427 if (src->max_win < win) 4428 src->max_win = win; 4429 /* synchronize sequencing */ 4430 if (SEQ_GT(end, src->seqlo)) 4431 src->seqlo = end; 4432 /* slide the window of what the other end can send */ 4433 if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) 4434 dst->seqhi = ack + MAX((win << sws), 1); 4435 4436 /* 4437 * Cannot set dst->seqhi here since this could be a shotgunned 4438 * SYN and not an already established connection. 4439 */ 4440 4441 if (th->th_flags & TH_FIN) 4442 if (src->state < TCPS_CLOSING) 4443 src->state = TCPS_CLOSING; 4444 if (th->th_flags & TH_RST) 4445 src->state = dst->state = TCPS_TIME_WAIT; 4446 4447 /* Fall through to PASS packet */ 4448 4449 } else { 4450 if ((*state)->dst.state == TCPS_SYN_SENT && 4451 (*state)->src.state == TCPS_SYN_SENT) { 4452 /* Send RST for state mismatches during handshake */ 4453 if (!(th->th_flags & TH_RST)) 4454 pf_send_tcp((*state)->rule.ptr, pd->af, 4455 pd->dst, pd->src, th->th_dport, 4456 th->th_sport, ntohl(th->th_ack), 0, 4457 TH_RST, 0, 0, 4458 (*state)->rule.ptr->return_ttl, 1, 0, 4459 pd->eh, kif->pfik_ifp); 4460 src->seqlo = 0; 4461 src->seqhi = 1; 4462 src->max_win = 1; 4463 } else if (pf_status.debug >= PF_DEBUG_MISC) { 4464 printf("pf: BAD state: "); 4465 pf_print_state(*state); 4466 pf_print_flags(th->th_flags); 4467 printf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d " 4468 "dir=%s,%s\n", seq, ack, pd->p_len, ackskew, 4469 (*state)->packets[0], (*state)->packets[1], 4470 direction == PF_IN ? "in" : "out", 4471 direction == (*state)->direction ? "fwd" : "rev"); 4472 printf("pf: State failure on: %c %c %c %c | %c %c\n", 4473 SEQ_GEQ(src->seqhi, end) ? ' ' : '1', 4474 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ? 4475 ' ': '2', 4476 (ackskew >= -MAXACKWINDOW) ? ' ' : '3', 4477 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4', 4478 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5', 4479 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6'); 4480 } 4481 REASON_SET(reason, PFRES_BADSTATE); 4482 return (PF_DROP); 4483 } 4484 4485 /* Any packets which have gotten here are to be passed */ 4486 4487 /* translate source/destination address, if necessary */ 4488 if (STATE_TRANSLATE(*state)) { 4489 if (direction == PF_OUT) 4490 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum, 4491 &th->th_sum, &(*state)->gwy.addr, 4492 (*state)->gwy.port, 0, pd->af); 4493 else 4494 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum, 4495 &th->th_sum, &(*state)->lan.addr, 4496 (*state)->lan.port, 0, pd->af); 4497 m_copyback(m, off, sizeof(*th), th); 4498 } else if (copyback) { 4499 /* Copyback sequence modulation or stateful scrub changes */ 4500 m_copyback(m, off, sizeof(*th), th); 4501 } 4502 4503 return (PF_PASS); 4504 } 4505 4506 int 4507 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif, 4508 struct mbuf *m, int off, void *h __unused, struct pf_pdesc *pd) 4509 { 4510 struct pf_state_peer *src, *dst; 4511 struct pf_state key; 4512 struct udphdr *uh = pd->hdr.udp; 4513 4514 key.af = pd->af; 4515 key.proto = IPPROTO_UDP; 4516 if (direction == PF_IN) { 4517 PF_ACPY(&key.ext.addr, pd->src, key.af); 4518 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 4519 key.ext.port = uh->uh_sport; 4520 key.gwy.port = uh->uh_dport; 4521 } else { 4522 PF_ACPY(&key.lan.addr, pd->src, key.af); 4523 PF_ACPY(&key.ext.addr, pd->dst, key.af); 4524 key.lan.port = uh->uh_sport; 4525 key.ext.port = uh->uh_dport; 4526 } 4527 4528 STATE_LOOKUP(); 4529 4530 if (direction == (*state)->direction) { 4531 src = &(*state)->src; 4532 dst = &(*state)->dst; 4533 } else { 4534 src = &(*state)->dst; 4535 dst = &(*state)->src; 4536 } 4537 4538 /* update states */ 4539 if (src->state < PFUDPS_SINGLE) 4540 src->state = PFUDPS_SINGLE; 4541 if (dst->state == PFUDPS_SINGLE) 4542 dst->state = PFUDPS_MULTIPLE; 4543 4544 /* update expire time */ 4545 (*state)->expire = time_second; 4546 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE) 4547 (*state)->timeout = PFTM_UDP_MULTIPLE; 4548 else 4549 (*state)->timeout = PFTM_UDP_SINGLE; 4550 4551 /* translate source/destination address, if necessary */ 4552 if (STATE_TRANSLATE(*state)) { 4553 if (direction == PF_OUT) 4554 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum, 4555 &uh->uh_sum, &(*state)->gwy.addr, 4556 (*state)->gwy.port, 1, pd->af); 4557 else 4558 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum, 4559 &uh->uh_sum, &(*state)->lan.addr, 4560 (*state)->lan.port, 1, pd->af); 4561 m_copyback(m, off, sizeof(*uh), uh); 4562 } 4563 4564 return (PF_PASS); 4565 } 4566 4567 int 4568 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif, 4569 struct mbuf *m, int off, void *h __unused, struct pf_pdesc *pd, 4570 u_short *reason) 4571 { 4572 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 4573 u_int16_t icmpid = 0, *icmpsum; 4574 u_int8_t icmptype; 4575 int state_icmp = 0; 4576 4577 icmpsum = NULL; /* XXXGCC -Wunitialized m68k */ 4578 icmptype = 0; /* XXXGCC -Wunitialized m68k */ 4579 4580 switch (pd->proto) { 4581 #ifdef INET 4582 case IPPROTO_ICMP: 4583 icmptype = pd->hdr.icmp->icmp_type; 4584 icmpid = pd->hdr.icmp->icmp_id; 4585 icmpsum = &pd->hdr.icmp->icmp_cksum; 4586 4587 if (icmptype == ICMP_UNREACH || 4588 icmptype == ICMP_SOURCEQUENCH || 4589 icmptype == ICMP_REDIRECT || 4590 icmptype == ICMP_TIMXCEED || 4591 icmptype == ICMP_PARAMPROB) 4592 state_icmp++; 4593 break; 4594 #endif /* INET */ 4595 #ifdef INET6 4596 case IPPROTO_ICMPV6: 4597 icmptype = pd->hdr.icmp6->icmp6_type; 4598 icmpid = pd->hdr.icmp6->icmp6_id; 4599 icmpsum = &pd->hdr.icmp6->icmp6_cksum; 4600 4601 if (icmptype == ICMP6_DST_UNREACH || 4602 icmptype == ICMP6_PACKET_TOO_BIG || 4603 icmptype == ICMP6_TIME_EXCEEDED || 4604 icmptype == ICMP6_PARAM_PROB) 4605 state_icmp++; 4606 break; 4607 #endif /* INET6 */ 4608 } 4609 4610 if (!state_icmp) { 4611 4612 /* 4613 * ICMP query/reply message not related to a TCP/UDP packet. 4614 * Search for an ICMP state. 4615 */ 4616 struct pf_state key; 4617 4618 key.af = pd->af; 4619 key.proto = pd->proto; 4620 if (direction == PF_IN) { 4621 PF_ACPY(&key.ext.addr, pd->src, key.af); 4622 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 4623 key.ext.port = icmpid; 4624 key.gwy.port = icmpid; 4625 } else { 4626 PF_ACPY(&key.lan.addr, pd->src, key.af); 4627 PF_ACPY(&key.ext.addr, pd->dst, key.af); 4628 key.lan.port = icmpid; 4629 key.ext.port = icmpid; 4630 } 4631 4632 STATE_LOOKUP(); 4633 4634 (*state)->expire = time_second; 4635 (*state)->timeout = PFTM_ICMP_ERROR_REPLY; 4636 4637 /* translate source/destination address, if necessary */ 4638 if (PF_ANEQ(&(*state)->lan.addr, &(*state)->gwy.addr, pd->af)) { 4639 if (direction == PF_OUT) { 4640 switch (pd->af) { 4641 #ifdef INET 4642 case AF_INET: 4643 pf_change_a(&saddr->v4.s_addr, 4644 pd->ip_sum, 4645 (*state)->gwy.addr.v4.s_addr, 0); 4646 break; 4647 #endif /* INET */ 4648 #ifdef INET6 4649 case AF_INET6: 4650 pf_change_a6(saddr, 4651 &pd->hdr.icmp6->icmp6_cksum, 4652 &(*state)->gwy.addr, 0); 4653 m_copyback(m, off, 4654 sizeof(struct icmp6_hdr), 4655 pd->hdr.icmp6); 4656 break; 4657 #endif /* INET6 */ 4658 } 4659 } else { 4660 switch (pd->af) { 4661 #ifdef INET 4662 case AF_INET: 4663 pf_change_a(&daddr->v4.s_addr, 4664 pd->ip_sum, 4665 (*state)->lan.addr.v4.s_addr, 0); 4666 break; 4667 #endif /* INET */ 4668 #ifdef INET6 4669 case AF_INET6: 4670 pf_change_a6(daddr, 4671 &pd->hdr.icmp6->icmp6_cksum, 4672 &(*state)->lan.addr, 0); 4673 m_copyback(m, off, 4674 sizeof(struct icmp6_hdr), 4675 pd->hdr.icmp6); 4676 break; 4677 #endif /* INET6 */ 4678 } 4679 } 4680 } 4681 4682 return (PF_PASS); 4683 4684 } else { 4685 /* 4686 * ICMP error message in response to a TCP/UDP packet. 4687 * Extract the inner TCP/UDP header and search for that state. 4688 */ 4689 4690 struct pf_pdesc pd2; 4691 #ifdef INET 4692 struct ip h2; 4693 #endif /* INET */ 4694 #ifdef INET6 4695 struct ip6_hdr h2_6; 4696 int terminal = 0; 4697 #endif /* INET6 */ 4698 int ipoff2 = 0; 4699 int off2 = 0; 4700 4701 memset(&pd2, 0, sizeof pd2); /* XXX gcc */ 4702 4703 pd2.af = pd->af; 4704 switch (pd->af) { 4705 #ifdef INET 4706 case AF_INET: 4707 /* offset of h2 in mbuf chain */ 4708 ipoff2 = off + ICMP_MINLEN; 4709 4710 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2), 4711 NULL, reason, pd2.af)) { 4712 DPFPRINTF(PF_DEBUG_MISC, 4713 ("pf: ICMP error message too short " 4714 "(ip)\n")); 4715 return (PF_DROP); 4716 } 4717 /* 4718 * ICMP error messages don't refer to non-first 4719 * fragments 4720 */ 4721 if (h2.ip_off & htons(IP_OFFMASK)) { 4722 REASON_SET(reason, PFRES_FRAG); 4723 return (PF_DROP); 4724 } 4725 4726 /* offset of protocol header that follows h2 */ 4727 off2 = ipoff2 + (h2.ip_hl << 2); 4728 4729 pd2.proto = h2.ip_p; 4730 pd2.src = (struct pf_addr *)&h2.ip_src; 4731 pd2.dst = (struct pf_addr *)&h2.ip_dst; 4732 pd2.ip_sum = &h2.ip_sum; 4733 break; 4734 #endif /* INET */ 4735 #ifdef INET6 4736 case AF_INET6: 4737 ipoff2 = off + sizeof(struct icmp6_hdr); 4738 4739 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6), 4740 NULL, reason, pd2.af)) { 4741 DPFPRINTF(PF_DEBUG_MISC, 4742 ("pf: ICMP error message too short " 4743 "(ip6)\n")); 4744 return (PF_DROP); 4745 } 4746 pd2.proto = h2_6.ip6_nxt; 4747 pd2.src = (struct pf_addr *)&h2_6.ip6_src; 4748 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst; 4749 pd2.ip_sum = NULL; 4750 off2 = ipoff2 + sizeof(h2_6); 4751 do { 4752 switch (pd2.proto) { 4753 case IPPROTO_FRAGMENT: 4754 /* 4755 * ICMPv6 error messages for 4756 * non-first fragments 4757 */ 4758 REASON_SET(reason, PFRES_FRAG); 4759 return (PF_DROP); 4760 case IPPROTO_AH: 4761 case IPPROTO_HOPOPTS: 4762 case IPPROTO_ROUTING: 4763 case IPPROTO_DSTOPTS: { 4764 /* get next header and header length */ 4765 struct ip6_ext opt6; 4766 4767 if (!pf_pull_hdr(m, off2, &opt6, 4768 sizeof(opt6), NULL, reason, 4769 pd2.af)) { 4770 DPFPRINTF(PF_DEBUG_MISC, 4771 ("pf: ICMPv6 short opt\n")); 4772 return (PF_DROP); 4773 } 4774 if (pd2.proto == IPPROTO_AH) 4775 off2 += (opt6.ip6e_len + 2) * 4; 4776 else 4777 off2 += (opt6.ip6e_len + 1) * 8; 4778 pd2.proto = opt6.ip6e_nxt; 4779 /* goto the next header */ 4780 break; 4781 } 4782 default: 4783 terminal++; 4784 break; 4785 } 4786 } while (!terminal); 4787 break; 4788 #endif /* INET6 */ 4789 } 4790 4791 switch (pd2.proto) { 4792 case IPPROTO_TCP: { 4793 struct tcphdr th; 4794 u_int32_t seq; 4795 struct pf_state key; 4796 struct pf_state_peer *src, *dst; 4797 u_int8_t dws; 4798 int copyback = 0; 4799 4800 /* 4801 * Only the first 8 bytes of the TCP header can be 4802 * expected. Don't access any TCP header fields after 4803 * th_seq, an ackskew test is not possible. 4804 */ 4805 if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason, 4806 pd2.af)) { 4807 DPFPRINTF(PF_DEBUG_MISC, 4808 ("pf: ICMP error message too short " 4809 "(tcp)\n")); 4810 return (PF_DROP); 4811 } 4812 4813 key.af = pd2.af; 4814 key.proto = IPPROTO_TCP; 4815 if (direction == PF_IN) { 4816 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 4817 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 4818 key.ext.port = th.th_dport; 4819 key.gwy.port = th.th_sport; 4820 } else { 4821 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 4822 PF_ACPY(&key.ext.addr, pd2.src, key.af); 4823 key.lan.port = th.th_dport; 4824 key.ext.port = th.th_sport; 4825 } 4826 4827 STATE_LOOKUP(); 4828 4829 if (direction == (*state)->direction) { 4830 src = &(*state)->dst; 4831 dst = &(*state)->src; 4832 } else { 4833 src = &(*state)->src; 4834 dst = &(*state)->dst; 4835 } 4836 4837 if (src->wscale && dst->wscale && 4838 !(th.th_flags & TH_SYN)) 4839 dws = dst->wscale & PF_WSCALE_MASK; 4840 else 4841 dws = 0; 4842 4843 /* Demodulate sequence number */ 4844 seq = ntohl(th.th_seq) - src->seqdiff; 4845 if (src->seqdiff) { 4846 pf_change_a(&th.th_seq, icmpsum, 4847 htonl(seq), 0); 4848 copyback = 1; 4849 } 4850 4851 if (!SEQ_GEQ(src->seqhi, seq) || 4852 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) { 4853 if (pf_status.debug >= PF_DEBUG_MISC) { 4854 printf("pf: BAD ICMP %d:%d ", 4855 icmptype, pd->hdr.icmp->icmp_code); 4856 pf_print_host(pd->src, 0, pd->af); 4857 printf(" -> "); 4858 pf_print_host(pd->dst, 0, pd->af); 4859 printf(" state: "); 4860 pf_print_state(*state); 4861 printf(" seq=%u\n", seq); 4862 } 4863 REASON_SET(reason, PFRES_BADSTATE); 4864 return (PF_DROP); 4865 } 4866 4867 if (STATE_TRANSLATE(*state)) { 4868 if (direction == PF_IN) { 4869 pf_change_icmp(pd2.src, &th.th_sport, 4870 daddr, &(*state)->lan.addr, 4871 (*state)->lan.port, NULL, 4872 pd2.ip_sum, icmpsum, 4873 pd->ip_sum, 0, pd2.af); 4874 } else { 4875 pf_change_icmp(pd2.dst, &th.th_dport, 4876 saddr, &(*state)->gwy.addr, 4877 (*state)->gwy.port, NULL, 4878 pd2.ip_sum, icmpsum, 4879 pd->ip_sum, 0, pd2.af); 4880 } 4881 copyback = 1; 4882 } 4883 4884 if (copyback) { 4885 switch (pd2.af) { 4886 #ifdef INET 4887 case AF_INET: 4888 m_copyback(m, off, ICMP_MINLEN, 4889 pd->hdr.icmp); 4890 m_copyback(m, ipoff2, sizeof(h2), 4891 &h2); 4892 break; 4893 #endif /* INET */ 4894 #ifdef INET6 4895 case AF_INET6: 4896 m_copyback(m, off, 4897 sizeof(struct icmp6_hdr), 4898 pd->hdr.icmp6); 4899 m_copyback(m, ipoff2, sizeof(h2_6), 4900 &h2_6); 4901 break; 4902 #endif /* INET6 */ 4903 } 4904 m_copyback(m, off2, 8, &th); 4905 } 4906 4907 return (PF_PASS); 4908 break; 4909 } 4910 case IPPROTO_UDP: { 4911 struct udphdr uh; 4912 struct pf_state key; 4913 4914 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh), 4915 NULL, reason, pd2.af)) { 4916 DPFPRINTF(PF_DEBUG_MISC, 4917 ("pf: ICMP error message too short " 4918 "(udp)\n")); 4919 return (PF_DROP); 4920 } 4921 4922 key.af = pd2.af; 4923 key.proto = IPPROTO_UDP; 4924 if (direction == PF_IN) { 4925 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 4926 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 4927 key.ext.port = uh.uh_dport; 4928 key.gwy.port = uh.uh_sport; 4929 } else { 4930 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 4931 PF_ACPY(&key.ext.addr, pd2.src, key.af); 4932 key.lan.port = uh.uh_dport; 4933 key.ext.port = uh.uh_sport; 4934 } 4935 4936 STATE_LOOKUP(); 4937 4938 if (STATE_TRANSLATE(*state)) { 4939 if (direction == PF_IN) { 4940 pf_change_icmp(pd2.src, &uh.uh_sport, 4941 daddr, &(*state)->lan.addr, 4942 (*state)->lan.port, &uh.uh_sum, 4943 pd2.ip_sum, icmpsum, 4944 pd->ip_sum, 1, pd2.af); 4945 } else { 4946 pf_change_icmp(pd2.dst, &uh.uh_dport, 4947 saddr, &(*state)->gwy.addr, 4948 (*state)->gwy.port, &uh.uh_sum, 4949 pd2.ip_sum, icmpsum, 4950 pd->ip_sum, 1, pd2.af); 4951 } 4952 switch (pd2.af) { 4953 #ifdef INET 4954 case AF_INET: 4955 m_copyback(m, off, ICMP_MINLEN, 4956 pd->hdr.icmp); 4957 m_copyback(m, ipoff2, sizeof(h2), &h2); 4958 break; 4959 #endif /* INET */ 4960 #ifdef INET6 4961 case AF_INET6: 4962 m_copyback(m, off, 4963 sizeof(struct icmp6_hdr), 4964 pd->hdr.icmp6); 4965 m_copyback(m, ipoff2, sizeof(h2_6), 4966 &h2_6); 4967 break; 4968 #endif /* INET6 */ 4969 } 4970 m_copyback(m, off2, sizeof(uh), &uh); 4971 } 4972 4973 return (PF_PASS); 4974 break; 4975 } 4976 #ifdef INET 4977 case IPPROTO_ICMP: { 4978 struct icmp iih; 4979 struct pf_state key; 4980 4981 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN, 4982 NULL, reason, pd2.af)) { 4983 DPFPRINTF(PF_DEBUG_MISC, 4984 ("pf: ICMP error message too short i" 4985 "(icmp)\n")); 4986 return (PF_DROP); 4987 } 4988 4989 key.af = pd2.af; 4990 key.proto = IPPROTO_ICMP; 4991 if (direction == PF_IN) { 4992 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 4993 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 4994 key.ext.port = iih.icmp_id; 4995 key.gwy.port = iih.icmp_id; 4996 } else { 4997 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 4998 PF_ACPY(&key.ext.addr, pd2.src, key.af); 4999 key.lan.port = iih.icmp_id; 5000 key.ext.port = iih.icmp_id; 5001 } 5002 5003 STATE_LOOKUP(); 5004 5005 if (STATE_TRANSLATE(*state)) { 5006 if (direction == PF_IN) { 5007 pf_change_icmp(pd2.src, &iih.icmp_id, 5008 daddr, &(*state)->lan.addr, 5009 (*state)->lan.port, NULL, 5010 pd2.ip_sum, icmpsum, 5011 pd->ip_sum, 0, AF_INET); 5012 } else { 5013 pf_change_icmp(pd2.dst, &iih.icmp_id, 5014 saddr, &(*state)->gwy.addr, 5015 (*state)->gwy.port, NULL, 5016 pd2.ip_sum, icmpsum, 5017 pd->ip_sum, 0, AF_INET); 5018 } 5019 m_copyback(m, off, ICMP_MINLEN, pd->hdr.icmp); 5020 m_copyback(m, ipoff2, sizeof(h2), &h2); 5021 m_copyback(m, off2, ICMP_MINLEN, &iih); 5022 } 5023 5024 return (PF_PASS); 5025 break; 5026 } 5027 #endif /* INET */ 5028 #ifdef INET6 5029 case IPPROTO_ICMPV6: { 5030 struct icmp6_hdr iih; 5031 struct pf_state key; 5032 5033 if (!pf_pull_hdr(m, off2, &iih, 5034 sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) { 5035 DPFPRINTF(PF_DEBUG_MISC, 5036 ("pf: ICMP error message too short " 5037 "(icmp6)\n")); 5038 return (PF_DROP); 5039 } 5040 5041 key.af = pd2.af; 5042 key.proto = IPPROTO_ICMPV6; 5043 if (direction == PF_IN) { 5044 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 5045 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 5046 key.ext.port = iih.icmp6_id; 5047 key.gwy.port = iih.icmp6_id; 5048 } else { 5049 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 5050 PF_ACPY(&key.ext.addr, pd2.src, key.af); 5051 key.lan.port = iih.icmp6_id; 5052 key.ext.port = iih.icmp6_id; 5053 } 5054 5055 STATE_LOOKUP(); 5056 5057 if (STATE_TRANSLATE(*state)) { 5058 if (direction == PF_IN) { 5059 pf_change_icmp(pd2.src, &iih.icmp6_id, 5060 daddr, &(*state)->lan.addr, 5061 (*state)->lan.port, NULL, 5062 pd2.ip_sum, icmpsum, 5063 pd->ip_sum, 0, AF_INET6); 5064 } else { 5065 pf_change_icmp(pd2.dst, &iih.icmp6_id, 5066 saddr, &(*state)->gwy.addr, 5067 (*state)->gwy.port, NULL, 5068 pd2.ip_sum, icmpsum, 5069 pd->ip_sum, 0, AF_INET6); 5070 } 5071 m_copyback(m, off, sizeof(struct icmp6_hdr), 5072 pd->hdr.icmp6); 5073 m_copyback(m, ipoff2, sizeof(h2_6), &h2_6); 5074 m_copyback(m, off2, sizeof(struct icmp6_hdr), 5075 &iih); 5076 } 5077 5078 return (PF_PASS); 5079 break; 5080 } 5081 #endif /* INET6 */ 5082 default: { 5083 struct pf_state key; 5084 5085 key.af = pd2.af; 5086 key.proto = pd2.proto; 5087 if (direction == PF_IN) { 5088 PF_ACPY(&key.ext.addr, pd2.dst, key.af); 5089 PF_ACPY(&key.gwy.addr, pd2.src, key.af); 5090 key.ext.port = 0; 5091 key.gwy.port = 0; 5092 } else { 5093 PF_ACPY(&key.lan.addr, pd2.dst, key.af); 5094 PF_ACPY(&key.ext.addr, pd2.src, key.af); 5095 key.lan.port = 0; 5096 key.ext.port = 0; 5097 } 5098 5099 STATE_LOOKUP(); 5100 5101 if (STATE_TRANSLATE(*state)) { 5102 if (direction == PF_IN) { 5103 pf_change_icmp(pd2.src, NULL, 5104 daddr, &(*state)->lan.addr, 5105 0, NULL, 5106 pd2.ip_sum, icmpsum, 5107 pd->ip_sum, 0, pd2.af); 5108 } else { 5109 pf_change_icmp(pd2.dst, NULL, 5110 saddr, &(*state)->gwy.addr, 5111 0, NULL, 5112 pd2.ip_sum, icmpsum, 5113 pd->ip_sum, 0, pd2.af); 5114 } 5115 switch (pd2.af) { 5116 #ifdef INET 5117 case AF_INET: 5118 m_copyback(m, off, ICMP_MINLEN, 5119 pd->hdr.icmp); 5120 m_copyback(m, ipoff2, sizeof(h2), &h2); 5121 break; 5122 #endif /* INET */ 5123 #ifdef INET6 5124 case AF_INET6: 5125 m_copyback(m, off, 5126 sizeof(struct icmp6_hdr), 5127 pd->hdr.icmp6); 5128 m_copyback(m, ipoff2, sizeof(h2_6), 5129 &h2_6); 5130 break; 5131 #endif /* INET6 */ 5132 } 5133 } 5134 5135 return (PF_PASS); 5136 break; 5137 } 5138 } 5139 } 5140 } 5141 5142 int 5143 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif, 5144 struct pf_pdesc *pd) 5145 { 5146 struct pf_state_peer *src, *dst; 5147 struct pf_state key; 5148 5149 key.af = pd->af; 5150 key.proto = pd->proto; 5151 if (direction == PF_IN) { 5152 PF_ACPY(&key.ext.addr, pd->src, key.af); 5153 PF_ACPY(&key.gwy.addr, pd->dst, key.af); 5154 key.ext.port = 0; 5155 key.gwy.port = 0; 5156 } else { 5157 PF_ACPY(&key.lan.addr, pd->src, key.af); 5158 PF_ACPY(&key.ext.addr, pd->dst, key.af); 5159 key.lan.port = 0; 5160 key.ext.port = 0; 5161 } 5162 5163 STATE_LOOKUP(); 5164 5165 if (direction == (*state)->direction) { 5166 src = &(*state)->src; 5167 dst = &(*state)->dst; 5168 } else { 5169 src = &(*state)->dst; 5170 dst = &(*state)->src; 5171 } 5172 5173 /* update states */ 5174 if (src->state < PFOTHERS_SINGLE) 5175 src->state = PFOTHERS_SINGLE; 5176 if (dst->state == PFOTHERS_SINGLE) 5177 dst->state = PFOTHERS_MULTIPLE; 5178 5179 /* update expire time */ 5180 (*state)->expire = time_second; 5181 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE) 5182 (*state)->timeout = PFTM_OTHER_MULTIPLE; 5183 else 5184 (*state)->timeout = PFTM_OTHER_SINGLE; 5185 5186 /* translate source/destination address, if necessary */ 5187 if (STATE_TRANSLATE(*state)) { 5188 if (direction == PF_OUT) 5189 switch (pd->af) { 5190 #ifdef INET 5191 case AF_INET: 5192 pf_change_a(&pd->src->v4.s_addr, 5193 pd->ip_sum, (*state)->gwy.addr.v4.s_addr, 5194 0); 5195 break; 5196 #endif /* INET */ 5197 #ifdef INET6 5198 case AF_INET6: 5199 PF_ACPY(pd->src, &(*state)->gwy.addr, pd->af); 5200 break; 5201 #endif /* INET6 */ 5202 } 5203 else 5204 switch (pd->af) { 5205 #ifdef INET 5206 case AF_INET: 5207 pf_change_a(&pd->dst->v4.s_addr, 5208 pd->ip_sum, (*state)->lan.addr.v4.s_addr, 5209 0); 5210 break; 5211 #endif /* INET */ 5212 #ifdef INET6 5213 case AF_INET6: 5214 PF_ACPY(pd->dst, &(*state)->lan.addr, pd->af); 5215 break; 5216 #endif /* INET6 */ 5217 } 5218 } 5219 5220 return (PF_PASS); 5221 } 5222 5223 /* 5224 * ipoff and off are measured from the start of the mbuf chain. 5225 * h must be at "ipoff" on the mbuf chain. 5226 */ 5227 void * 5228 pf_pull_hdr(struct mbuf *m, int off, void *p, int len, 5229 u_short *actionp, u_short *reasonp, sa_family_t af) 5230 { 5231 switch (af) { 5232 #ifdef INET 5233 case AF_INET: { 5234 struct ip *h = mtod(m, struct ip *); 5235 u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3; 5236 5237 if (fragoff) { 5238 if (fragoff >= len) 5239 ACTION_SET(actionp, PF_PASS); 5240 else { 5241 ACTION_SET(actionp, PF_DROP); 5242 REASON_SET(reasonp, PFRES_FRAG); 5243 } 5244 return (NULL); 5245 } 5246 if (m->m_pkthdr.len < off + len || 5247 ntohs(h->ip_len) < off + len) { 5248 ACTION_SET(actionp, PF_DROP); 5249 REASON_SET(reasonp, PFRES_SHORT); 5250 return (NULL); 5251 } 5252 break; 5253 } 5254 #endif /* INET */ 5255 #ifdef INET6 5256 case AF_INET6: { 5257 struct ip6_hdr *h = mtod(m, struct ip6_hdr *); 5258 5259 if (m->m_pkthdr.len < off + len || 5260 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) < 5261 (unsigned)(off + len)) { 5262 ACTION_SET(actionp, PF_DROP); 5263 REASON_SET(reasonp, PFRES_SHORT); 5264 return (NULL); 5265 } 5266 break; 5267 } 5268 #endif /* INET6 */ 5269 } 5270 m_copydata(m, off, len, p); 5271 return (p); 5272 } 5273 5274 int 5275 pf_routable(struct pf_addr *addr, sa_family_t af) 5276 { 5277 struct sockaddr_in *dst; 5278 #ifdef INET6 5279 struct sockaddr_in6 *dst6; 5280 struct route_in6 ro; 5281 #else 5282 struct route ro; 5283 #endif 5284 5285 bzero(&ro, sizeof(ro)); 5286 switch (af) { 5287 case AF_INET: 5288 dst = satosin(&ro.ro_dst); 5289 dst->sin_family = AF_INET; 5290 dst->sin_len = sizeof(*dst); 5291 dst->sin_addr = addr->v4; 5292 break; 5293 #ifdef INET6 5294 case AF_INET6: 5295 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 5296 dst6->sin6_family = AF_INET6; 5297 dst6->sin6_len = sizeof(*dst6); 5298 dst6->sin6_addr = addr->v6; 5299 break; 5300 #endif /* INET6 */ 5301 default: 5302 return (0); 5303 } 5304 5305 #ifdef __OpenBSD__ 5306 rtalloc_noclone((struct route *)&ro, NO_CLONING); 5307 #else 5308 rtalloc((struct route *)&ro); 5309 #endif 5310 5311 if (ro.ro_rt != NULL) { 5312 RTFREE(ro.ro_rt); 5313 return (1); 5314 } 5315 5316 return (0); 5317 } 5318 5319 int 5320 pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, 5321 struct pf_addr_wrap *aw __unused) 5322 { 5323 struct sockaddr_in *dst; 5324 #ifdef INET6 5325 struct sockaddr_in6 *dst6; 5326 struct route_in6 ro; 5327 #else 5328 struct route ro; 5329 #endif 5330 int ret = 0; 5331 5332 bzero(&ro, sizeof(ro)); 5333 switch (af) { 5334 case AF_INET: 5335 dst = satosin(&ro.ro_dst); 5336 dst->sin_family = AF_INET; 5337 dst->sin_len = sizeof(*dst); 5338 dst->sin_addr = addr->v4; 5339 break; 5340 #ifdef INET6 5341 case AF_INET6: 5342 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 5343 dst6->sin6_family = AF_INET6; 5344 dst6->sin6_len = sizeof(*dst6); 5345 dst6->sin6_addr = addr->v6; 5346 break; 5347 #endif /* INET6 */ 5348 default: 5349 return (0); 5350 } 5351 5352 #ifdef __OpenBSD__ 5353 rtalloc_noclone((struct route *)&ro, NO_CLONING); 5354 #else 5355 rtalloc((struct route *)&ro); 5356 #endif 5357 5358 if (ro.ro_rt != NULL) { 5359 #ifdef __OpenBSD__ 5360 if (ro.ro_rt->rt_labelid == aw->v.rtlabel) 5361 ret = 1; 5362 #endif 5363 RTFREE(ro.ro_rt); 5364 } 5365 5366 return (ret); 5367 } 5368 5369 #ifdef INET 5370 void 5371 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp, 5372 struct pf_state *s) 5373 { 5374 struct mbuf *m0, *m1; 5375 struct m_tag *mtag; 5376 struct route iproute; 5377 struct route *ro = NULL; 5378 struct sockaddr_in *dst; 5379 struct ip *ip; 5380 struct ifnet *ifp = NULL; 5381 struct pf_addr naddr; 5382 struct pf_src_node *sn = NULL; 5383 int error = 0; 5384 5385 if (m == NULL || *m == NULL || r == NULL || 5386 (dir != PF_IN && dir != PF_OUT) || oifp == NULL) 5387 panic("pf_route: invalid parameters"); 5388 5389 if ((mtag = m_tag_find(*m, PACKET_TAG_PF_ROUTED, NULL)) == NULL) { 5390 if ((mtag = m_tag_get(PACKET_TAG_PF_ROUTED, 1, M_NOWAIT)) == 5391 NULL) { 5392 m0 = *m; 5393 *m = NULL; 5394 goto bad; 5395 } 5396 *(char *)(mtag + 1) = 1; 5397 m_tag_prepend(*m, mtag); 5398 } else { 5399 if (*(char *)(mtag + 1) > 3) { 5400 m0 = *m; 5401 *m = NULL; 5402 goto bad; 5403 } 5404 (*(char *)(mtag + 1))++; 5405 } 5406 5407 if (r->rt == PF_DUPTO) { 5408 if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL) 5409 return; 5410 } else { 5411 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) 5412 return; 5413 m0 = *m; 5414 } 5415 5416 if (m0->m_len < sizeof(struct ip)) { 5417 DPFPRINTF(PF_DEBUG_URGENT, 5418 ("pf_route: m0->m_len < sizeof(struct ip)\n")); 5419 goto bad; 5420 } 5421 5422 ip = mtod(m0, struct ip *); 5423 5424 ro = &iproute; 5425 bzero((caddr_t)ro, sizeof(*ro)); 5426 dst = satosin(&ro->ro_dst); 5427 dst->sin_family = AF_INET; 5428 dst->sin_len = sizeof(*dst); 5429 dst->sin_addr = ip->ip_dst; 5430 5431 if (r->rt == PF_FASTROUTE) { 5432 rtalloc(ro); 5433 if (ro->ro_rt == 0) { 5434 ipstat.ips_noroute++; 5435 goto bad; 5436 } 5437 5438 ifp = ro->ro_rt->rt_ifp; 5439 ro->ro_rt->rt_use++; 5440 5441 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 5442 dst = satosin(ro->ro_rt->rt_gateway); 5443 } else { 5444 if (TAILQ_EMPTY(&r->rpool.list)) { 5445 DPFPRINTF(PF_DEBUG_URGENT, 5446 ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n")); 5447 goto bad; 5448 } 5449 if (s == NULL) { 5450 pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src, 5451 &naddr, NULL, &sn); 5452 if (!PF_AZERO(&naddr, AF_INET)) 5453 dst->sin_addr.s_addr = naddr.v4.s_addr; 5454 ifp = r->rpool.cur->kif ? 5455 r->rpool.cur->kif->pfik_ifp : NULL; 5456 } else { 5457 if (!PF_AZERO(&s->rt_addr, AF_INET)) 5458 dst->sin_addr.s_addr = 5459 s->rt_addr.v4.s_addr; 5460 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; 5461 } 5462 } 5463 if (ifp == NULL) 5464 goto bad; 5465 5466 if (oifp != ifp) { 5467 if (pf_test(PF_OUT, ifp, &m0, NULL) != PF_PASS) 5468 goto bad; 5469 else if (m0 == NULL) 5470 goto done; 5471 if (m0->m_len < sizeof(struct ip)) { 5472 DPFPRINTF(PF_DEBUG_URGENT, 5473 ("pf_route: m0->m_len < sizeof(struct ip)\n")); 5474 goto bad; 5475 } 5476 ip = mtod(m0, struct ip *); 5477 } 5478 5479 /* Copied from ip_output. */ 5480 #ifdef IPSEC 5481 /* 5482 * If deferred crypto processing is needed, check that the 5483 * interface supports it. 5484 */ 5485 if ((mtag = m_tag_find(m0, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL)) 5486 != NULL && (ifp->if_capabilities & IFCAP_IPSEC) == 0) { 5487 /* Notify IPsec to do its own crypto. */ 5488 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 5489 goto bad; 5490 } 5491 #endif /* IPSEC */ 5492 5493 /* Catch routing changes wrt. hardware checksumming for TCP or UDP. */ 5494 #ifdef __OpenBSD__ 5495 if (m0->m_pkthdr.csum & M_TCPV4_CSUM_OUT) { 5496 if (!(ifp->if_capabilities & IFCAP_CSUM_TCPv4) || 5497 ifp->if_bridge != NULL) { 5498 in_delayed_cksum(m0); 5499 m0->m_pkthdr.csum &= ~M_TCPV4_CSUM_OUT; /* Clear */ 5500 } 5501 } else if (m0->m_pkthdr.csum & M_UDPV4_CSUM_OUT) { 5502 if (!(ifp->if_capabilities & IFCAP_CSUM_UDPv4) || 5503 ifp->if_bridge != NULL) { 5504 in_delayed_cksum(m0); 5505 m0->m_pkthdr.csum &= ~M_UDPV4_CSUM_OUT; /* Clear */ 5506 } 5507 } 5508 #else 5509 if (m0->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 5510 in_delayed_cksum(m0); 5511 m0->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 5512 } 5513 #endif 5514 5515 if (ntohs(ip->ip_len) <= ifp->if_mtu) { 5516 #ifdef __OpenBSD__ 5517 if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) && 5518 ifp->if_bridge == NULL) { 5519 m0->m_pkthdr.csum |= M_IPV4_CSUM_OUT; 5520 ipstat.ips_outhwcsum++; 5521 } else { 5522 ip->ip_sum = 0; 5523 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2); 5524 } 5525 #else 5526 ip->ip_sum = 0; 5527 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2); 5528 5529 m0->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 5530 #endif 5531 #ifdef __OpenBSD__ 5532 /* Update relevant hardware checksum stats for TCP/UDP */ 5533 if (m0->m_pkthdr.csum & M_TCPV4_CSUM_OUT) 5534 tcpstat.tcps_outhwcsum++; 5535 else if (m0->m_pkthdr.csum & M_UDPV4_CSUM_OUT) 5536 udpstat.udps_outhwcsum++; 5537 #endif 5538 error = (*ifp->if_output)(ifp, m0, sintosa(dst), NULL); 5539 goto done; 5540 } 5541 5542 /* 5543 * Too large for interface; fragment if possible. 5544 * Must be able to put at least 8 bytes per fragment. 5545 */ 5546 if (ip->ip_off & htons(IP_DF)) { 5547 ipstat.ips_cantfrag++; 5548 if (r->rt != PF_DUPTO) { 5549 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0, 5550 ifp->if_mtu); 5551 goto done; 5552 } else 5553 goto bad; 5554 } 5555 5556 m1 = m0; 5557 error = ip_fragment(m0, ifp, ifp->if_mtu); 5558 if (error) { 5559 m0 = NULL; 5560 goto bad; 5561 } 5562 5563 for (m0 = m1; m0; m0 = m1) { 5564 m1 = m0->m_nextpkt; 5565 m0->m_nextpkt = 0; 5566 if (error == 0) 5567 error = (*ifp->if_output)(ifp, m0, sintosa(dst), 5568 NULL); 5569 else 5570 m_freem(m0); 5571 } 5572 5573 if (error == 0) 5574 ipstat.ips_fragmented++; 5575 5576 done: 5577 if (r->rt != PF_DUPTO) 5578 *m = NULL; 5579 if (ro == &iproute && ro->ro_rt) 5580 RTFREE(ro->ro_rt); 5581 return; 5582 5583 bad: 5584 m_freem(m0); 5585 goto done; 5586 } 5587 #endif /* INET */ 5588 5589 #ifdef INET6 5590 void 5591 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp, 5592 struct pf_state *s) 5593 { 5594 struct mbuf *m0; 5595 struct m_tag *mtag; 5596 struct route_in6 ip6route; 5597 struct route_in6 *ro; 5598 struct sockaddr_in6 *dst; 5599 struct ip6_hdr *ip6; 5600 struct ifnet *ifp = NULL; 5601 struct pf_addr naddr; 5602 struct pf_src_node *sn = NULL; 5603 int error = 0; 5604 5605 if (m == NULL || *m == NULL || r == NULL || 5606 (dir != PF_IN && dir != PF_OUT) || oifp == NULL) 5607 panic("pf_route6: invalid parameters"); 5608 5609 if ((mtag = m_tag_find(*m, PACKET_TAG_PF_ROUTED, NULL)) == NULL) { 5610 if ((mtag = m_tag_get(PACKET_TAG_PF_ROUTED, 1, M_NOWAIT)) == 5611 NULL) { 5612 m0 = *m; 5613 *m = NULL; 5614 goto bad; 5615 } 5616 *(char *)(mtag + 1) = 1; 5617 m_tag_prepend(*m, mtag); 5618 } else { 5619 if (*(char *)(mtag + 1) > 3) { 5620 m0 = *m; 5621 *m = NULL; 5622 goto bad; 5623 } 5624 (*(char *)(mtag + 1))++; 5625 } 5626 5627 if (r->rt == PF_DUPTO) { 5628 if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL) 5629 return; 5630 } else { 5631 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) 5632 return; 5633 m0 = *m; 5634 } 5635 5636 if (m0->m_len < sizeof(struct ip6_hdr)) { 5637 DPFPRINTF(PF_DEBUG_URGENT, 5638 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n")); 5639 goto bad; 5640 } 5641 ip6 = mtod(m0, struct ip6_hdr *); 5642 5643 ro = &ip6route; 5644 bzero((caddr_t)ro, sizeof(*ro)); 5645 dst = (struct sockaddr_in6 *)&ro->ro_dst; 5646 dst->sin6_family = AF_INET6; 5647 dst->sin6_len = sizeof(*dst); 5648 dst->sin6_addr = ip6->ip6_dst; 5649 5650 /* Cheat. */ 5651 if (r->rt == PF_FASTROUTE) { 5652 mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT); 5653 if (mtag == NULL) 5654 goto bad; 5655 m_tag_prepend(m0, mtag); 5656 #ifdef __OpenBSD__ 5657 ip6_output(m0, NULL, NULL, 0, NULL, NULL); 5658 #else 5659 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL); 5660 #endif 5661 return; 5662 } 5663 5664 if (TAILQ_EMPTY(&r->rpool.list)) { 5665 DPFPRINTF(PF_DEBUG_URGENT, 5666 ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n")); 5667 goto bad; 5668 } 5669 if (s == NULL) { 5670 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src, 5671 &naddr, NULL, &sn); 5672 if (!PF_AZERO(&naddr, AF_INET6)) 5673 PF_ACPY((struct pf_addr *)&dst->sin6_addr, 5674 &naddr, AF_INET6); 5675 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL; 5676 } else { 5677 if (!PF_AZERO(&s->rt_addr, AF_INET6)) 5678 PF_ACPY((struct pf_addr *)&dst->sin6_addr, 5679 &s->rt_addr, AF_INET6); 5680 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; 5681 } 5682 if (ifp == NULL) 5683 goto bad; 5684 5685 if (oifp != ifp) { 5686 if (pf_test6(PF_OUT, ifp, &m0, NULL) != PF_PASS) 5687 goto bad; 5688 else if (m0 == NULL) 5689 goto done; 5690 if (m0->m_len < sizeof(struct ip6_hdr)) { 5691 DPFPRINTF(PF_DEBUG_URGENT, 5692 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n")); 5693 goto bad; 5694 } 5695 ip6 = mtod(m0, struct ip6_hdr *); 5696 } 5697 5698 /* 5699 * If the packet is too large for the outgoing interface, 5700 * send back an icmp6 error. 5701 */ 5702 if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr)) 5703 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index); 5704 if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) { 5705 error = nd6_output(ifp, ifp, m0, dst, NULL); 5706 } else { 5707 in6_ifstat_inc(ifp, ifs6_in_toobig); 5708 if (r->rt != PF_DUPTO) 5709 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu); 5710 else 5711 goto bad; 5712 } 5713 5714 done: 5715 if (r->rt != PF_DUPTO) 5716 *m = NULL; 5717 return; 5718 5719 bad: 5720 m_freem(m0); 5721 goto done; 5722 } 5723 #endif /* INET6 */ 5724 5725 5726 /* 5727 * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag 5728 * off is the offset where the protocol header starts 5729 * len is the total length of protocol header plus payload 5730 * returns 0 when the checksum is valid, otherwise returns 1. 5731 */ 5732 int 5733 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p, 5734 sa_family_t af) 5735 { 5736 #ifdef __OpenBSD__ 5737 u_int16_t flag_ok, flag_bad; 5738 #endif 5739 u_int16_t sum; 5740 5741 #ifdef __OpenBSD__ 5742 switch (p) { 5743 case IPPROTO_TCP: 5744 flag_ok = M_TCP_CSUM_IN_OK; 5745 flag_bad = M_TCP_CSUM_IN_BAD; 5746 break; 5747 case IPPROTO_UDP: 5748 flag_ok = M_UDP_CSUM_IN_OK; 5749 flag_bad = M_UDP_CSUM_IN_BAD; 5750 break; 5751 case IPPROTO_ICMP: 5752 #ifdef INET6 5753 case IPPROTO_ICMPV6: 5754 #endif /* INET6 */ 5755 flag_ok = flag_bad = 0; 5756 break; 5757 default: 5758 return (1); 5759 } 5760 if (m->m_pkthdr.csum & flag_ok) 5761 return (0); 5762 if (m->m_pkthdr.csum & flag_bad) 5763 return (1); 5764 #endif 5765 if (off < sizeof(struct ip) || len < sizeof(struct udphdr)) 5766 return (1); 5767 if (m->m_pkthdr.len < off + len) 5768 return (1); 5769 #ifdef __NetBSD__ 5770 switch (p) { 5771 case IPPROTO_TCP: { 5772 struct tcphdr th; /* XXX */ 5773 int thlen; 5774 5775 m_copydata(m, off, sizeof(th), &th); /* XXX */ 5776 thlen = th.th_off << 2; 5777 return tcp_input_checksum(af, m, &th, off, 5778 thlen, len - thlen) != 0; 5779 } 5780 5781 case IPPROTO_UDP: { 5782 struct udphdr uh; /* XXX */ 5783 5784 m_copydata(m, off, sizeof(uh), &uh); /* XXX */ 5785 return udp_input_checksum(af, m, &uh, off, len) != 0; 5786 } 5787 break; 5788 } 5789 #endif /* __NetBSD__ */ 5790 switch (af) { 5791 #ifdef INET 5792 case AF_INET: 5793 if (p == IPPROTO_ICMP) { 5794 if (m->m_len < off) 5795 return (1); 5796 m->m_data += off; 5797 m->m_len -= off; 5798 sum = in_cksum(m, len); 5799 m->m_data -= off; 5800 m->m_len += off; 5801 } else { 5802 if (m->m_len < sizeof(struct ip)) 5803 return (1); 5804 sum = in4_cksum(m, p, off, len); 5805 } 5806 break; 5807 #endif /* INET */ 5808 #ifdef INET6 5809 case AF_INET6: 5810 if (m->m_len < sizeof(struct ip6_hdr)) 5811 return (1); 5812 sum = in6_cksum(m, p, off, len); 5813 break; 5814 #endif /* INET6 */ 5815 default: 5816 return (1); 5817 } 5818 if (sum) { 5819 #ifdef __OpenBSD__ 5820 m->m_pkthdr.csum |= flag_bad; 5821 #endif 5822 switch (p) { 5823 case IPPROTO_TCP: 5824 tcpstat.tcps_rcvbadsum++; 5825 break; 5826 case IPPROTO_UDP: 5827 udpstat.udps_badsum++; 5828 break; 5829 case IPPROTO_ICMP: 5830 icmpstat.icps_checksum++; 5831 break; 5832 #ifdef INET6 5833 case IPPROTO_ICMPV6: 5834 icmp6stat.icp6s_checksum++; 5835 break; 5836 #endif /* INET6 */ 5837 } 5838 return (1); 5839 } 5840 #ifdef __OpenBSD__ 5841 m->m_pkthdr.csum |= flag_ok; 5842 #endif 5843 return (0); 5844 } 5845 5846 static int 5847 pf_add_mbuf_tag(struct mbuf *m, u_int tag) 5848 { 5849 struct m_tag *mtag; 5850 5851 if (m_tag_find(m, tag, NULL) != NULL) 5852 return (0); 5853 mtag = m_tag_get(tag, 0, M_NOWAIT); 5854 if (mtag == NULL) 5855 return (1); 5856 m_tag_prepend(m, mtag); 5857 return (0); 5858 } 5859 5860 #ifdef INET 5861 int 5862 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0, 5863 struct ether_header *eh) 5864 { 5865 struct pfi_kif *kif; 5866 u_short action, reason = 0, log = 0; 5867 struct mbuf *m = *m0; 5868 struct ip *h = NULL; 5869 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr; 5870 struct pf_state *s = NULL; 5871 struct pf_ruleset *ruleset = NULL; 5872 struct pf_pdesc pd; 5873 int off, dirndx, pqid = 0; 5874 5875 if (!pf_status.running || 5876 (m_tag_find(m, PACKET_TAG_PF_GENERATED, NULL) != NULL)) 5877 return (PF_PASS); 5878 5879 #if NCARP > 0 5880 if (ifp->if_type == IFT_CARP && ifp->if_carpdev) 5881 ifp = ifp->if_carpdev; 5882 #endif 5883 5884 kif = pfi_index2kif[ifp->if_index]; 5885 if (kif == NULL) { 5886 DPFPRINTF(PF_DEBUG_URGENT, 5887 ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname)); 5888 return (PF_DROP); 5889 } 5890 if (kif->pfik_flags & PFI_IFLAG_SKIP) 5891 return (PF_PASS); 5892 5893 #ifdef DIAGNOSTIC 5894 if ((m->m_flags & M_PKTHDR) == 0) 5895 panic("non-M_PKTHDR is passed to pf_test"); 5896 #endif /* DIAGNOSTIC */ 5897 5898 memset(&pd, 0, sizeof(pd)); 5899 if (m->m_pkthdr.len < (int)sizeof(*h)) { 5900 action = PF_DROP; 5901 REASON_SET(&reason, PFRES_SHORT); 5902 log = 1; 5903 goto done; 5904 } 5905 5906 /* We do IP header normalization and packet reassembly here */ 5907 if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) { 5908 action = PF_DROP; 5909 goto done; 5910 } 5911 m = *m0; 5912 h = mtod(m, struct ip *); 5913 5914 off = h->ip_hl << 2; 5915 if (off < (int)sizeof(*h)) { 5916 action = PF_DROP; 5917 REASON_SET(&reason, PFRES_SHORT); 5918 log = 1; 5919 goto done; 5920 } 5921 5922 pd.src = (struct pf_addr *)&h->ip_src; 5923 pd.dst = (struct pf_addr *)&h->ip_dst; 5924 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET); 5925 pd.ip_sum = &h->ip_sum; 5926 pd.proto = h->ip_p; 5927 pd.af = AF_INET; 5928 pd.tos = h->ip_tos; 5929 pd.tot_len = ntohs(h->ip_len); 5930 pd.eh = eh; 5931 5932 /* handle fragments that didn't get reassembled by normalization */ 5933 if (h->ip_off & htons(IP_MF | IP_OFFMASK)) { 5934 action = pf_test_fragment(&r, dir, kif, m, h, 5935 &pd, &a, &ruleset); 5936 goto done; 5937 } 5938 5939 switch (h->ip_p) { 5940 5941 case IPPROTO_TCP: { 5942 struct tcphdr th; 5943 5944 pd.hdr.tcp = &th; 5945 if (!pf_pull_hdr(m, off, &th, sizeof(th), 5946 &action, &reason, AF_INET)) { 5947 log = action != PF_PASS; 5948 goto done; 5949 } 5950 if (dir == PF_IN && pf_check_proto_cksum(m, off, 5951 ntohs(h->ip_len) - off, IPPROTO_TCP, AF_INET)) { 5952 action = PF_DROP; 5953 goto done; 5954 } 5955 pd.p_len = pd.tot_len - off - (th.th_off << 2); 5956 if ((th.th_flags & TH_ACK) && pd.p_len == 0) 5957 pqid = 1; 5958 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd); 5959 if (action == PF_DROP) 5960 goto done; 5961 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd, 5962 &reason); 5963 if (action == PF_PASS) { 5964 #if NPFSYNC 5965 pfsync_update_state(s); 5966 #endif /* NPFSYNC */ 5967 r = s->rule.ptr; 5968 a = s->anchor.ptr; 5969 log = s->log; 5970 } else if (s == NULL) 5971 action = pf_test_tcp(&r, &s, dir, kif, 5972 m, off, h, &pd, &a, &ruleset, &ipintrq); 5973 break; 5974 } 5975 5976 case IPPROTO_UDP: { 5977 struct udphdr uh; 5978 5979 pd.hdr.udp = &uh; 5980 if (!pf_pull_hdr(m, off, &uh, sizeof(uh), 5981 &action, &reason, AF_INET)) { 5982 log = action != PF_PASS; 5983 goto done; 5984 } 5985 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m, 5986 off, ntohs(h->ip_len) - off, IPPROTO_UDP, AF_INET)) { 5987 action = PF_DROP; 5988 goto done; 5989 } 5990 if (uh.uh_dport == 0 || 5991 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off || 5992 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) { 5993 action = PF_DROP; 5994 goto done; 5995 } 5996 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd); 5997 if (action == PF_PASS) { 5998 #if NPFSYNC 5999 pfsync_update_state(s); 6000 #endif /* NPFSYNC */ 6001 r = s->rule.ptr; 6002 a = s->anchor.ptr; 6003 log = s->log; 6004 } else if (s == NULL) 6005 action = pf_test_udp(&r, &s, dir, kif, 6006 m, off, h, &pd, &a, &ruleset, &ipintrq); 6007 break; 6008 } 6009 6010 case IPPROTO_ICMP: { 6011 struct icmp ih; 6012 6013 pd.hdr.icmp = &ih; 6014 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN, 6015 &action, &reason, AF_INET)) { 6016 log = action != PF_PASS; 6017 goto done; 6018 } 6019 if (dir == PF_IN && pf_check_proto_cksum(m, off, 6020 ntohs(h->ip_len) - off, IPPROTO_ICMP, AF_INET)) { 6021 action = PF_DROP; 6022 goto done; 6023 } 6024 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd, 6025 &reason); 6026 if (action == PF_PASS) { 6027 #if NPFSYNC 6028 pfsync_update_state(s); 6029 #endif /* NPFSYNC */ 6030 r = s->rule.ptr; 6031 a = s->anchor.ptr; 6032 log = s->log; 6033 } else if (s == NULL) 6034 action = pf_test_icmp(&r, &s, dir, kif, 6035 m, off, h, &pd, &a, &ruleset, &ipintrq); 6036 break; 6037 } 6038 6039 default: 6040 action = pf_test_state_other(&s, dir, kif, &pd); 6041 if (action == PF_PASS) { 6042 #if NPFSYNC 6043 pfsync_update_state(s); 6044 #endif /* NPFSYNC */ 6045 r = s->rule.ptr; 6046 a = s->anchor.ptr; 6047 log = s->log; 6048 } else if (s == NULL) 6049 action = pf_test_other(&r, &s, dir, kif, m, off, h, 6050 &pd, &a, &ruleset, &ipintrq); 6051 break; 6052 } 6053 6054 done: 6055 if (action == PF_PASS && h->ip_hl > 5 && 6056 !((s && s->allow_opts) || r->allow_opts)) { 6057 action = PF_DROP; 6058 REASON_SET(&reason, PFRES_IPOPTIONS); 6059 log = 1; 6060 DPFPRINTF(PF_DEBUG_MISC, 6061 ("pf: dropping packet with ip options\n")); 6062 } 6063 6064 if (s && s->tag) 6065 pf_tag_packet(m, pf_get_tag(m), s->tag); 6066 6067 #ifdef ALTQ 6068 if (action == PF_PASS && r->qid) { 6069 struct m_tag *mtag; 6070 struct altq_tag *atag; 6071 6072 mtag = m_tag_get(PACKET_TAG_PF_QID, sizeof(*atag), M_NOWAIT); 6073 if (mtag != NULL) { 6074 atag = (struct altq_tag *)(mtag + 1); 6075 if (pqid || pd.tos == IPTOS_LOWDELAY) 6076 atag->qid = r->pqid; 6077 else 6078 atag->qid = r->qid; 6079 /* add hints for ecn */ 6080 atag->af = AF_INET; 6081 atag->hdr = h; 6082 m_tag_prepend(m, mtag); 6083 } 6084 } 6085 #endif /* ALTQ */ 6086 6087 /* 6088 * connections redirected to loopback should not match sockets 6089 * bound specifically to loopback due to security implications, 6090 * see tcp_input() and in_pcblookup_listen(). 6091 */ 6092 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP || 6093 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL && 6094 (s->nat_rule.ptr->action == PF_RDR || 6095 s->nat_rule.ptr->action == PF_BINAT) && 6096 (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET && 6097 pf_add_mbuf_tag(m, PACKET_TAG_PF_TRANSLATE_LOCALHOST)) { 6098 action = PF_DROP; 6099 REASON_SET(&reason, PFRES_MEMORY); 6100 } 6101 6102 if (log) 6103 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, r, a, ruleset); 6104 6105 kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len; 6106 kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++; 6107 6108 if (action == PF_PASS || r->action == PF_DROP) { 6109 r->packets++; 6110 r->bytes += pd.tot_len; 6111 if (a != NULL) { 6112 a->packets++; 6113 a->bytes += pd.tot_len; 6114 } 6115 if (s != NULL) { 6116 dirndx = (dir == s->direction) ? 0 : 1; 6117 s->packets[dirndx]++; 6118 s->bytes[dirndx] += pd.tot_len; 6119 if (s->nat_rule.ptr != NULL) { 6120 s->nat_rule.ptr->packets++; 6121 s->nat_rule.ptr->bytes += pd.tot_len; 6122 } 6123 if (s->src_node != NULL) { 6124 s->src_node->packets++; 6125 s->src_node->bytes += pd.tot_len; 6126 } 6127 if (s->nat_src_node != NULL) { 6128 s->nat_src_node->packets++; 6129 s->nat_src_node->bytes += pd.tot_len; 6130 } 6131 } 6132 tr = r; 6133 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule; 6134 if (nr != NULL) { 6135 struct pf_addr *x; 6136 /* 6137 * XXX: we need to make sure that the addresses 6138 * passed to pfr_update_stats() are the same than 6139 * the addresses used during matching (pfr_match) 6140 */ 6141 if (r == &pf_default_rule) { 6142 tr = nr; 6143 x = (s == NULL || s->direction == dir) ? 6144 &pd.baddr : &pd.naddr; 6145 } else 6146 x = (s == NULL || s->direction == dir) ? 6147 &pd.naddr : &pd.baddr; 6148 if (x == &pd.baddr || s == NULL) { 6149 /* we need to change the address */ 6150 if (dir == PF_OUT) 6151 pd.src = x; 6152 else 6153 pd.dst = x; 6154 } 6155 } 6156 if (tr->src.addr.type == PF_ADDR_TABLE) 6157 pfr_update_stats(tr->src.addr.p.tbl, (s == NULL || 6158 s->direction == dir) ? pd.src : pd.dst, pd.af, 6159 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 6160 tr->src.neg); 6161 if (tr->dst.addr.type == PF_ADDR_TABLE) 6162 pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL || 6163 s->direction == dir) ? pd.dst : pd.src, pd.af, 6164 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 6165 tr->dst.neg); 6166 } 6167 6168 6169 if (action == PF_SYNPROXY_DROP) { 6170 m_freem(*m0); 6171 *m0 = NULL; 6172 action = PF_PASS; 6173 } else if (r->rt) 6174 /* pf_route can free the mbuf causing *m0 to become NULL */ 6175 pf_route(m0, r, dir, ifp, s); 6176 6177 return (action); 6178 } 6179 #endif /* INET */ 6180 6181 #ifdef INET6 6182 int 6183 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0, 6184 struct ether_header *eh) 6185 { 6186 struct pfi_kif *kif; 6187 u_short action, reason = 0, log = 0; 6188 struct mbuf *m = *m0; 6189 struct ip6_hdr *h = NULL; 6190 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr; 6191 struct pf_state *s = NULL; 6192 struct pf_ruleset *ruleset = NULL; 6193 struct pf_pdesc pd; 6194 int off, terminal = 0, dirndx; 6195 6196 if (!pf_status.running || 6197 (m_tag_find(m, PACKET_TAG_PF_GENERATED, NULL) != NULL)) 6198 return (PF_PASS); 6199 6200 #if NCARP > 0 6201 if (ifp->if_type == IFT_CARP && ifp->if_carpdev) 6202 ifp = ifp->if_carpdev; 6203 #endif 6204 6205 kif = pfi_index2kif[ifp->if_index]; 6206 if (kif == NULL) { 6207 DPFPRINTF(PF_DEBUG_URGENT, 6208 ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname)); 6209 return (PF_DROP); 6210 } 6211 if (kif->pfik_flags & PFI_IFLAG_SKIP) 6212 return (PF_PASS); 6213 6214 #ifdef DIAGNOSTIC 6215 if ((m->m_flags & M_PKTHDR) == 0) 6216 panic("non-M_PKTHDR is passed to pf_test6"); 6217 #endif /* DIAGNOSTIC */ 6218 6219 memset(&pd, 0, sizeof(pd)); 6220 if (m->m_pkthdr.len < (int)sizeof(*h)) { 6221 action = PF_DROP; 6222 REASON_SET(&reason, PFRES_SHORT); 6223 log = 1; 6224 goto done; 6225 } 6226 6227 /* We do IP header normalization and packet reassembly here */ 6228 if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) { 6229 action = PF_DROP; 6230 goto done; 6231 } 6232 m = *m0; 6233 h = mtod(m, struct ip6_hdr *); 6234 6235 pd.src = (struct pf_addr *)&h->ip6_src; 6236 pd.dst = (struct pf_addr *)&h->ip6_dst; 6237 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6); 6238 pd.ip_sum = NULL; 6239 pd.af = AF_INET6; 6240 pd.tos = 0; 6241 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr); 6242 pd.eh = eh; 6243 6244 off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr); 6245 pd.proto = h->ip6_nxt; 6246 do { 6247 switch (pd.proto) { 6248 case IPPROTO_FRAGMENT: 6249 action = pf_test_fragment(&r, dir, kif, m, h, 6250 &pd, &a, &ruleset); 6251 if (action == PF_DROP) 6252 REASON_SET(&reason, PFRES_FRAG); 6253 goto done; 6254 case IPPROTO_AH: 6255 case IPPROTO_HOPOPTS: 6256 case IPPROTO_ROUTING: 6257 case IPPROTO_DSTOPTS: { 6258 /* get next header and header length */ 6259 struct ip6_ext opt6; 6260 6261 if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6), 6262 NULL, &reason, pd.af)) { 6263 DPFPRINTF(PF_DEBUG_MISC, 6264 ("pf: IPv6 short opt\n")); 6265 action = PF_DROP; 6266 log = 1; 6267 goto done; 6268 } 6269 if (pd.proto == IPPROTO_AH) 6270 off += (opt6.ip6e_len + 2) * 4; 6271 else 6272 off += (opt6.ip6e_len + 1) * 8; 6273 pd.proto = opt6.ip6e_nxt; 6274 /* goto the next header */ 6275 break; 6276 } 6277 default: 6278 terminal++; 6279 break; 6280 } 6281 } while (!terminal); 6282 6283 switch (pd.proto) { 6284 6285 case IPPROTO_TCP: { 6286 struct tcphdr th; 6287 6288 pd.hdr.tcp = &th; 6289 if (!pf_pull_hdr(m, off, &th, sizeof(th), 6290 &action, &reason, AF_INET6)) { 6291 log = action != PF_PASS; 6292 goto done; 6293 } 6294 if (dir == PF_IN && pf_check_proto_cksum(m, off, 6295 ntohs(h->ip6_plen) - (off - sizeof(struct ip6_hdr)), 6296 IPPROTO_TCP, AF_INET6)) { 6297 action = PF_DROP; 6298 REASON_SET(&reason, PFRES_PROTCKSUM); 6299 goto done; 6300 } 6301 pd.p_len = pd.tot_len - off - (th.th_off << 2); 6302 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd); 6303 if (action == PF_DROP) 6304 goto done; 6305 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd, 6306 &reason); 6307 if (action == PF_PASS) { 6308 #if NPFSYNC 6309 pfsync_update_state(s); 6310 #endif /* NPFSYNC */ 6311 r = s->rule.ptr; 6312 a = s->anchor.ptr; 6313 log = s->log; 6314 } else if (s == NULL) 6315 action = pf_test_tcp(&r, &s, dir, kif, 6316 m, off, h, &pd, &a, &ruleset, &ip6intrq); 6317 break; 6318 } 6319 6320 case IPPROTO_UDP: { 6321 struct udphdr uh; 6322 6323 pd.hdr.udp = &uh; 6324 if (!pf_pull_hdr(m, off, &uh, sizeof(uh), 6325 &action, &reason, AF_INET6)) { 6326 log = action != PF_PASS; 6327 goto done; 6328 } 6329 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m, 6330 off, ntohs(h->ip6_plen) - (off - sizeof(struct ip6_hdr)), 6331 IPPROTO_UDP, AF_INET6)) { 6332 action = PF_DROP; 6333 REASON_SET(&reason, PFRES_PROTCKSUM); 6334 goto done; 6335 } 6336 if (uh.uh_dport == 0 || 6337 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off || 6338 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) { 6339 action = PF_DROP; 6340 goto done; 6341 } 6342 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd); 6343 if (action == PF_PASS) { 6344 #if NPFSYNC 6345 pfsync_update_state(s); 6346 #endif /* NPFSYNC */ 6347 r = s->rule.ptr; 6348 a = s->anchor.ptr; 6349 log = s->log; 6350 } else if (s == NULL) 6351 action = pf_test_udp(&r, &s, dir, kif, 6352 m, off, h, &pd, &a, &ruleset, &ip6intrq); 6353 break; 6354 } 6355 6356 case IPPROTO_ICMPV6: { 6357 struct icmp6_hdr ih; 6358 6359 pd.hdr.icmp6 = &ih; 6360 if (!pf_pull_hdr(m, off, &ih, sizeof(ih), 6361 &action, &reason, AF_INET6)) { 6362 log = action != PF_PASS; 6363 goto done; 6364 } 6365 if (dir == PF_IN && pf_check_proto_cksum(m, off, 6366 ntohs(h->ip6_plen) - (off - sizeof(struct ip6_hdr)), 6367 IPPROTO_ICMPV6, AF_INET6)) { 6368 action = PF_DROP; 6369 REASON_SET(&reason, PFRES_PROTCKSUM); 6370 goto done; 6371 } 6372 action = pf_test_state_icmp(&s, dir, kif, 6373 m, off, h, &pd, &reason); 6374 if (action == PF_PASS) { 6375 #if NPFSYNC 6376 pfsync_update_state(s); 6377 #endif /* NPFSYNC */ 6378 r = s->rule.ptr; 6379 a = s->anchor.ptr; 6380 log = s->log; 6381 } else if (s == NULL) 6382 action = pf_test_icmp(&r, &s, dir, kif, 6383 m, off, h, &pd, &a, &ruleset, &ip6intrq); 6384 break; 6385 } 6386 6387 default: 6388 action = pf_test_state_other(&s, dir, kif, &pd); 6389 if (action == PF_PASS) { 6390 #if NPFSYNC 6391 pfsync_update_state(s); 6392 #endif /* NPFSYNC */ 6393 r = s->rule.ptr; 6394 a = s->anchor.ptr; 6395 log = s->log; 6396 } else if (s == NULL) 6397 action = pf_test_other(&r, &s, dir, kif, m, off, h, 6398 &pd, &a, &ruleset, &ip6intrq); 6399 break; 6400 } 6401 6402 done: 6403 /* XXX handle IPv6 options, if not allowed. not implemented. */ 6404 6405 if (s && s->tag) 6406 pf_tag_packet(m, pf_get_tag(m), s->tag); 6407 6408 #ifdef ALTQ 6409 if (action == PF_PASS && r->qid) { 6410 struct m_tag *mtag; 6411 struct altq_tag *atag; 6412 6413 mtag = m_tag_get(PACKET_TAG_PF_QID, sizeof(*atag), M_NOWAIT); 6414 if (mtag != NULL) { 6415 atag = (struct altq_tag *)(mtag + 1); 6416 if (pd.tos == IPTOS_LOWDELAY) 6417 atag->qid = r->pqid; 6418 else 6419 atag->qid = r->qid; 6420 /* add hints for ecn */ 6421 atag->af = AF_INET6; 6422 atag->hdr = h; 6423 m_tag_prepend(m, mtag); 6424 } 6425 } 6426 #endif /* ALTQ */ 6427 6428 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP || 6429 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL && 6430 (s->nat_rule.ptr->action == PF_RDR || 6431 s->nat_rule.ptr->action == PF_BINAT) && 6432 IN6_IS_ADDR_LOOPBACK(&pd.dst->v6) && 6433 pf_add_mbuf_tag(m, PACKET_TAG_PF_TRANSLATE_LOCALHOST)) { 6434 action = PF_DROP; 6435 REASON_SET(&reason, PFRES_MEMORY); 6436 } 6437 6438 if (log) 6439 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, r, a, ruleset); 6440 6441 kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len; 6442 kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++; 6443 6444 if (action == PF_PASS || r->action == PF_DROP) { 6445 r->packets++; 6446 r->bytes += pd.tot_len; 6447 if (a != NULL) { 6448 a->packets++; 6449 a->bytes += pd.tot_len; 6450 } 6451 if (s != NULL) { 6452 dirndx = (dir == s->direction) ? 0 : 1; 6453 s->packets[dirndx]++; 6454 s->bytes[dirndx] += pd.tot_len; 6455 if (s->nat_rule.ptr != NULL) { 6456 s->nat_rule.ptr->packets++; 6457 s->nat_rule.ptr->bytes += pd.tot_len; 6458 } 6459 if (s->src_node != NULL) { 6460 s->src_node->packets++; 6461 s->src_node->bytes += pd.tot_len; 6462 } 6463 if (s->nat_src_node != NULL) { 6464 s->nat_src_node->packets++; 6465 s->nat_src_node->bytes += pd.tot_len; 6466 } 6467 } 6468 tr = r; 6469 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule; 6470 if (nr != NULL) { 6471 struct pf_addr *x; 6472 /* 6473 * XXX: we need to make sure that the addresses 6474 * passed to pfr_update_stats() are the same than 6475 * the addresses used during matching (pfr_match) 6476 */ 6477 if (r == &pf_default_rule) { 6478 tr = nr; 6479 x = (s == NULL || s->direction == dir) ? 6480 &pd.baddr : &pd.naddr; 6481 } else { 6482 x = (s == NULL || s->direction == dir) ? 6483 &pd.naddr : &pd.baddr; 6484 } 6485 if (x == &pd.baddr || s == NULL) { 6486 if (dir == PF_OUT) 6487 pd.src = x; 6488 else 6489 pd.dst = x; 6490 } 6491 } 6492 if (tr->src.addr.type == PF_ADDR_TABLE) 6493 pfr_update_stats(tr->src.addr.p.tbl, (s == NULL || 6494 s->direction == dir) ? pd.src : pd.dst, pd.af, 6495 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 6496 tr->src.neg); 6497 if (tr->dst.addr.type == PF_ADDR_TABLE) 6498 pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL || 6499 s->direction == dir) ? pd.dst : pd.src, pd.af, 6500 pd.tot_len, dir == PF_OUT, r->action == PF_PASS, 6501 tr->dst.neg); 6502 } 6503 6504 6505 if (action == PF_SYNPROXY_DROP) { 6506 m_freem(*m0); 6507 *m0 = NULL; 6508 action = PF_PASS; 6509 } else if (r->rt) 6510 /* pf_route6 can free the mbuf causing *m0 to become NULL */ 6511 pf_route6(m0, r, dir, ifp, s); 6512 6513 return (action); 6514 } 6515 #endif /* INET6 */ 6516 6517 int 6518 pf_check_congestion(struct ifqueue *ifq __unused) 6519 { 6520 #ifdef __OpenBSD__ 6521 if (ifq->ifq_congestion) 6522 return (1); 6523 else 6524 return (0); 6525 #else 6526 return 0; 6527 #endif 6528 } 6529