1 /* $OpenBSD: pf.c,v 1.614 2008/08/02 12:34:37 henning Exp $ */ 2 3 /* 4 * Copyright (c) 2004 The DragonFly Project. All rights reserved. 5 * 6 * Copyright (c) 2001 Daniel Hartmeier 7 * Copyright (c) 2002 - 2008 Henning Brauer 8 * All rights reserved. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 14 * - Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * - Redistributions in binary form must reproduce the above 17 * copyright notice, this list of conditions and the following 18 * disclaimer in the documentation and/or other materials provided 19 * with the distribution. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 26 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, 27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 29 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN 31 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 32 * POSSIBILITY OF SUCH DAMAGE. 33 * 34 * Effort sponsored in part by the Defense Advanced Research Projects 35 * Agency (DARPA) and Air Force Research Laboratory, Air Force 36 * Materiel Command, USAF, under agreement number F30602-01-2-0537. 37 * 38 */ 39 40 #include "opt_inet.h" 41 #include "opt_inet6.h" 42 43 #include <sys/param.h> 44 #include <sys/systm.h> 45 #include <sys/malloc.h> 46 #include <sys/mbuf.h> 47 #include <sys/filio.h> 48 #include <sys/socket.h> 49 #include <sys/socketvar.h> 50 #include <sys/kernel.h> 51 #include <sys/time.h> 52 #include <sys/sysctl.h> 53 #include <sys/endian.h> 54 #include <sys/proc.h> 55 #include <sys/kthread.h> 56 57 #include <machine/inttypes.h> 58 59 #include <sys/md5.h> 60 61 #include <net/if.h> 62 #include <net/if_types.h> 63 #include <net/bpf.h> 64 #include <net/netisr.h> 65 #include <net/route.h> 66 67 #include <netinet/in.h> 68 #include <netinet/in_var.h> 69 #include <netinet/in_systm.h> 70 #include <netinet/ip.h> 71 #include <netinet/ip_var.h> 72 #include <netinet/tcp.h> 73 #include <netinet/tcp_seq.h> 74 #include <netinet/udp.h> 75 #include <netinet/ip_icmp.h> 76 #include <netinet/in_pcb.h> 77 #include <netinet/tcp_timer.h> 78 #include <netinet/tcp_var.h> 79 #include <netinet/udp_var.h> 80 #include <netinet/icmp_var.h> 81 #include <netinet/if_ether.h> 82 83 #include <net/pf/pfvar.h> 84 #include <net/pf/if_pflog.h> 85 86 #include <net/pf/if_pfsync.h> 87 88 #ifdef INET6 89 #include <netinet/ip6.h> 90 #include <netinet/in_pcb.h> 91 #include <netinet/icmp6.h> 92 #include <netinet6/nd6.h> 93 #include <netinet6/ip6_var.h> 94 #include <netinet6/in6_pcb.h> 95 #endif /* INET6 */ 96 97 #include <sys/in_cksum.h> 98 #include <sys/ucred.h> 99 #include <machine/limits.h> 100 #include <sys/msgport2.h> 101 #include <net/netmsg2.h> 102 103 extern int ip_optcopy(struct ip *, struct ip *); 104 extern int debug_pfugidhack; 105 106 struct lwkt_token pf_token = LWKT_TOKEN_INITIALIZER(pf_token); 107 108 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x 109 110 /* 111 * Global variables 112 */ 113 114 /* mask radix tree */ 115 struct radix_node_head *pf_maskhead; 116 117 /* state tables */ 118 struct pf_state_tree pf_statetbl; 119 120 struct pf_altqqueue pf_altqs[2]; 121 struct pf_palist pf_pabuf; 122 struct pf_altqqueue *pf_altqs_active; 123 struct pf_altqqueue *pf_altqs_inactive; 124 struct pf_status pf_status; 125 126 u_int32_t ticket_altqs_active; 127 u_int32_t ticket_altqs_inactive; 128 int altqs_inactive_open; 129 u_int32_t ticket_pabuf; 130 131 MD5_CTX pf_tcp_secret_ctx; 132 u_char pf_tcp_secret[16]; 133 int pf_tcp_secret_init; 134 int pf_tcp_iss_off; 135 136 struct pf_anchor_stackframe { 137 struct pf_ruleset *rs; 138 struct pf_rule *r; 139 struct pf_anchor_node *parent; 140 struct pf_anchor *child; 141 } pf_anchor_stack[64]; 142 143 struct malloc_type *pf_src_tree_pl, *pf_rule_pl, *pf_pooladdr_pl; 144 struct malloc_type *pf_state_pl, *pf_state_key_pl, *pf_state_item_pl; 145 struct malloc_type *pf_altq_pl; 146 147 void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t); 148 149 void pf_init_threshold(struct pf_threshold *, u_int32_t, 150 u_int32_t); 151 void pf_add_threshold(struct pf_threshold *); 152 int pf_check_threshold(struct pf_threshold *); 153 154 void pf_change_ap(struct pf_addr *, u_int16_t *, 155 u_int16_t *, u_int16_t *, struct pf_addr *, 156 u_int16_t, u_int8_t, sa_family_t); 157 int pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *, 158 struct tcphdr *, struct pf_state_peer *); 159 #ifdef INET6 160 void pf_change_a6(struct pf_addr *, u_int16_t *, 161 struct pf_addr *, u_int8_t); 162 #endif /* INET6 */ 163 void pf_change_icmp(struct pf_addr *, u_int16_t *, 164 struct pf_addr *, struct pf_addr *, u_int16_t, 165 u_int16_t *, u_int16_t *, u_int16_t *, 166 u_int16_t *, u_int8_t, sa_family_t); 167 void pf_send_tcp(const struct pf_rule *, sa_family_t, 168 const struct pf_addr *, const struct pf_addr *, 169 u_int16_t, u_int16_t, u_int32_t, u_int32_t, 170 u_int8_t, u_int16_t, u_int16_t, u_int8_t, int, 171 u_int16_t, struct ether_header *, struct ifnet *); 172 void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t, 173 sa_family_t, struct pf_rule *); 174 struct pf_rule *pf_match_translation(struct pf_pdesc *, struct mbuf *, 175 int, int, struct pfi_kif *, 176 struct pf_addr *, u_int16_t, struct pf_addr *, 177 u_int16_t, int); 178 struct pf_rule *pf_get_translation(struct pf_pdesc *, struct mbuf *, 179 int, int, struct pfi_kif *, struct pf_src_node **, 180 struct pf_state_key **, struct pf_state_key **, 181 struct pf_state_key **, struct pf_state_key **, 182 struct pf_addr *, struct pf_addr *, 183 u_int16_t, u_int16_t); 184 void pf_detach_state(struct pf_state *); 185 int pf_state_key_setup(struct pf_pdesc *, struct pf_rule *, 186 struct pf_state_key **, struct pf_state_key **, 187 struct pf_state_key **, struct pf_state_key **, 188 struct pf_addr *, struct pf_addr *, 189 u_int16_t, u_int16_t); 190 void pf_state_key_detach(struct pf_state *, int); 191 u_int32_t pf_tcp_iss(struct pf_pdesc *); 192 int pf_test_rule(struct pf_rule **, struct pf_state **, 193 int, struct pfi_kif *, struct mbuf *, int, 194 void *, struct pf_pdesc *, struct pf_rule **, 195 struct pf_ruleset **, struct ifqueue *, struct inpcb *); 196 static __inline int pf_create_state(struct pf_rule *, struct pf_rule *, 197 struct pf_rule *, struct pf_pdesc *, 198 struct pf_src_node *, struct pf_state_key *, 199 struct pf_state_key *, struct pf_state_key *, 200 struct pf_state_key *, struct mbuf *, int, 201 u_int16_t, u_int16_t, int *, struct pfi_kif *, 202 struct pf_state **, int, u_int16_t, u_int16_t, 203 int); 204 int pf_test_fragment(struct pf_rule **, int, 205 struct pfi_kif *, struct mbuf *, void *, 206 struct pf_pdesc *, struct pf_rule **, 207 struct pf_ruleset **); 208 int pf_tcp_track_full(struct pf_state_peer *, 209 struct pf_state_peer *, struct pf_state **, 210 struct pfi_kif *, struct mbuf *, int, 211 struct pf_pdesc *, u_short *, int *); 212 int pf_tcp_track_sloppy(struct pf_state_peer *, 213 struct pf_state_peer *, struct pf_state **, 214 struct pf_pdesc *, u_short *); 215 int pf_test_state_tcp(struct pf_state **, int, 216 struct pfi_kif *, struct mbuf *, int, 217 void *, struct pf_pdesc *, u_short *); 218 int pf_test_state_udp(struct pf_state **, int, 219 struct pfi_kif *, struct mbuf *, int, 220 void *, struct pf_pdesc *); 221 int pf_test_state_icmp(struct pf_state **, int, 222 struct pfi_kif *, struct mbuf *, int, 223 void *, struct pf_pdesc *, u_short *); 224 int pf_test_state_other(struct pf_state **, int, 225 struct pfi_kif *, struct mbuf *, struct pf_pdesc *); 226 void pf_step_into_anchor(int *, struct pf_ruleset **, int, 227 struct pf_rule **, struct pf_rule **, int *); 228 int pf_step_out_of_anchor(int *, struct pf_ruleset **, 229 int, struct pf_rule **, struct pf_rule **, 230 int *); 231 void pf_hash(struct pf_addr *, struct pf_addr *, 232 struct pf_poolhashkey *, sa_family_t); 233 int pf_map_addr(u_int8_t, struct pf_rule *, 234 struct pf_addr *, struct pf_addr *, 235 struct pf_addr *, struct pf_src_node **); 236 int pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *, 237 struct pf_addr *, struct pf_addr *, u_int16_t, 238 struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t, 239 struct pf_src_node **); 240 void pf_route(struct mbuf **, struct pf_rule *, int, 241 struct ifnet *, struct pf_state *, 242 struct pf_pdesc *); 243 void pf_route6(struct mbuf **, struct pf_rule *, int, 244 struct ifnet *, struct pf_state *, 245 struct pf_pdesc *); 246 u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t, 247 sa_family_t); 248 u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t, 249 sa_family_t); 250 u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t, 251 u_int16_t); 252 void pf_set_rt_ifp(struct pf_state *, 253 struct pf_addr *); 254 int pf_check_proto_cksum(struct mbuf *, int, int, 255 u_int8_t, sa_family_t); 256 struct pf_divert *pf_get_divert(struct mbuf *); 257 void pf_print_state_parts(struct pf_state *, 258 struct pf_state_key *, struct pf_state_key *); 259 int pf_addr_wrap_neq(struct pf_addr_wrap *, 260 struct pf_addr_wrap *); 261 struct pf_state *pf_find_state(struct pfi_kif *, 262 struct pf_state_key_cmp *, u_int, struct mbuf *); 263 int pf_src_connlimit(struct pf_state **); 264 int pf_check_congestion(struct ifqueue *); 265 266 extern int pf_end_threads; 267 268 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = { 269 { &pf_state_pl, PFSTATE_HIWAT }, 270 { &pf_src_tree_pl, PFSNODE_HIWAT }, 271 { &pf_frent_pl, PFFRAG_FRENT_HIWAT }, 272 { &pfr_ktable_pl, PFR_KTABLE_HIWAT }, 273 { &pfr_kentry_pl, PFR_KENTRY_HIWAT } 274 }; 275 276 #define STATE_LOOKUP(i, k, d, s, m) \ 277 do { \ 278 s = pf_find_state(i, k, d, m); \ 279 if (s == NULL || (s)->timeout == PFTM_PURGE) \ 280 return (PF_DROP); \ 281 if (d == PF_OUT && \ 282 (((s)->rule.ptr->rt == PF_ROUTETO && \ 283 (s)->rule.ptr->direction == PF_OUT) || \ 284 ((s)->rule.ptr->rt == PF_REPLYTO && \ 285 (s)->rule.ptr->direction == PF_IN)) && \ 286 (s)->rt_kif != NULL && \ 287 (s)->rt_kif != i) \ 288 return (PF_PASS); \ 289 } while (0) 290 291 #define BOUND_IFACE(r, k) \ 292 ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all 293 294 #define STATE_INC_COUNTERS(s) \ 295 do { \ 296 s->rule.ptr->states_cur++; \ 297 s->rule.ptr->states_tot++; \ 298 if (s->anchor.ptr != NULL) { \ 299 s->anchor.ptr->states_cur++; \ 300 s->anchor.ptr->states_tot++; \ 301 } \ 302 if (s->nat_rule.ptr != NULL) { \ 303 s->nat_rule.ptr->states_cur++; \ 304 s->nat_rule.ptr->states_tot++; \ 305 } \ 306 } while (0) 307 308 #define STATE_DEC_COUNTERS(s) \ 309 do { \ 310 if (s->nat_rule.ptr != NULL) \ 311 s->nat_rule.ptr->states_cur--; \ 312 if (s->anchor.ptr != NULL) \ 313 s->anchor.ptr->states_cur--; \ 314 s->rule.ptr->states_cur--; \ 315 } while (0) 316 317 static MALLOC_DEFINE(M_PFSTATEPL, "pfstatepl", "pf state pool list"); 318 static MALLOC_DEFINE(M_PFSRCTREEPL, "pfsrctpl", "pf source tree pool list"); 319 static MALLOC_DEFINE(M_PFSTATEKEYPL, "pfstatekeypl", "pf state key pool list"); 320 static MALLOC_DEFINE(M_PFSTATEITEMPL, "pfstateitempl", "pf state item pool list"); 321 322 static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *); 323 static __inline int pf_state_compare_key(struct pf_state_key *, 324 struct pf_state_key *); 325 static __inline int pf_state_compare_id(struct pf_state *, 326 struct pf_state *); 327 328 struct pf_src_tree tree_src_tracking; 329 330 struct pf_state_tree_id tree_id; 331 struct pf_state_queue state_list; 332 333 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare); 334 RB_GENERATE(pf_state_tree, pf_state_key, entry, pf_state_compare_key); 335 RB_GENERATE(pf_state_tree_id, pf_state, 336 entry_id, pf_state_compare_id); 337 338 static __inline int 339 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b) 340 { 341 int diff; 342 343 if (a->rule.ptr > b->rule.ptr) 344 return (1); 345 if (a->rule.ptr < b->rule.ptr) 346 return (-1); 347 if ((diff = a->af - b->af) != 0) 348 return (diff); 349 switch (a->af) { 350 #ifdef INET 351 case AF_INET: 352 if (a->addr.addr32[0] > b->addr.addr32[0]) 353 return (1); 354 if (a->addr.addr32[0] < b->addr.addr32[0]) 355 return (-1); 356 break; 357 #endif /* INET */ 358 #ifdef INET6 359 case AF_INET6: 360 if (a->addr.addr32[3] > b->addr.addr32[3]) 361 return (1); 362 if (a->addr.addr32[3] < b->addr.addr32[3]) 363 return (-1); 364 if (a->addr.addr32[2] > b->addr.addr32[2]) 365 return (1); 366 if (a->addr.addr32[2] < b->addr.addr32[2]) 367 return (-1); 368 if (a->addr.addr32[1] > b->addr.addr32[1]) 369 return (1); 370 if (a->addr.addr32[1] < b->addr.addr32[1]) 371 return (-1); 372 if (a->addr.addr32[0] > b->addr.addr32[0]) 373 return (1); 374 if (a->addr.addr32[0] < b->addr.addr32[0]) 375 return (-1); 376 break; 377 #endif /* INET6 */ 378 } 379 return (0); 380 } 381 382 u_int32_t 383 pf_state_hash(struct pf_state_key *sk) 384 { 385 u_int32_t hv = (u_int32_t)(((intptr_t)sk >> 6) ^ ((intptr_t)sk >> 15)); 386 if (hv == 0) /* disallow 0 */ 387 hv = 1; 388 return(hv); 389 } 390 391 #ifdef INET6 392 void 393 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af) 394 { 395 switch (af) { 396 #ifdef INET 397 case AF_INET: 398 dst->addr32[0] = src->addr32[0]; 399 break; 400 #endif /* INET */ 401 case AF_INET6: 402 dst->addr32[0] = src->addr32[0]; 403 dst->addr32[1] = src->addr32[1]; 404 dst->addr32[2] = src->addr32[2]; 405 dst->addr32[3] = src->addr32[3]; 406 break; 407 } 408 } 409 #endif /* INET6 */ 410 411 void 412 pf_init_threshold(struct pf_threshold *threshold, 413 u_int32_t limit, u_int32_t seconds) 414 { 415 threshold->limit = limit * PF_THRESHOLD_MULT; 416 threshold->seconds = seconds; 417 threshold->count = 0; 418 threshold->last = time_second; 419 } 420 421 void 422 pf_add_threshold(struct pf_threshold *threshold) 423 { 424 u_int32_t t = time_second, diff = t - threshold->last; 425 426 if (diff >= threshold->seconds) 427 threshold->count = 0; 428 else 429 threshold->count -= threshold->count * diff / 430 threshold->seconds; 431 threshold->count += PF_THRESHOLD_MULT; 432 threshold->last = t; 433 } 434 435 int 436 pf_check_threshold(struct pf_threshold *threshold) 437 { 438 return (threshold->count > threshold->limit); 439 } 440 441 int 442 pf_src_connlimit(struct pf_state **state) 443 { 444 int bad = 0; 445 446 (*state)->src_node->conn++; 447 (*state)->src.tcp_est = 1; 448 pf_add_threshold(&(*state)->src_node->conn_rate); 449 450 if ((*state)->rule.ptr->max_src_conn && 451 (*state)->rule.ptr->max_src_conn < 452 (*state)->src_node->conn) { 453 pf_status.lcounters[LCNT_SRCCONN]++; 454 bad++; 455 } 456 457 if ((*state)->rule.ptr->max_src_conn_rate.limit && 458 pf_check_threshold(&(*state)->src_node->conn_rate)) { 459 pf_status.lcounters[LCNT_SRCCONNRATE]++; 460 bad++; 461 } 462 463 if (!bad) 464 return (0); 465 466 if ((*state)->rule.ptr->overload_tbl) { 467 struct pfr_addr p; 468 u_int32_t killed = 0; 469 470 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++; 471 if (pf_status.debug >= PF_DEBUG_MISC) { 472 kprintf("pf_src_connlimit: blocking address "); 473 pf_print_host(&(*state)->src_node->addr, 0, 474 (*state)->key[PF_SK_WIRE]->af); 475 } 476 477 bzero(&p, sizeof(p)); 478 p.pfra_af = (*state)->key[PF_SK_WIRE]->af; 479 switch ((*state)->key[PF_SK_WIRE]->af) { 480 #ifdef INET 481 case AF_INET: 482 p.pfra_net = 32; 483 p.pfra_ip4addr = (*state)->src_node->addr.v4; 484 break; 485 #endif /* INET */ 486 #ifdef INET6 487 case AF_INET6: 488 p.pfra_net = 128; 489 p.pfra_ip6addr = (*state)->src_node->addr.v6; 490 break; 491 #endif /* INET6 */ 492 } 493 494 pfr_insert_kentry((*state)->rule.ptr->overload_tbl, 495 &p, time_second); 496 497 /* kill existing states if that's required. */ 498 if ((*state)->rule.ptr->flush) { 499 struct pf_state_key *sk; 500 struct pf_state *st; 501 502 pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++; 503 RB_FOREACH(st, pf_state_tree_id, &tree_id) { 504 sk = st->key[PF_SK_WIRE]; 505 /* 506 * Kill states from this source. (Only those 507 * from the same rule if PF_FLUSH_GLOBAL is not 508 * set) 509 */ 510 if (sk->af == 511 (*state)->key[PF_SK_WIRE]->af && 512 (((*state)->direction == PF_OUT && 513 PF_AEQ(&(*state)->src_node->addr, 514 &sk->addr[0], sk->af)) || 515 ((*state)->direction == PF_IN && 516 PF_AEQ(&(*state)->src_node->addr, 517 &sk->addr[1], sk->af))) && 518 ((*state)->rule.ptr->flush & 519 PF_FLUSH_GLOBAL || 520 (*state)->rule.ptr == st->rule.ptr)) { 521 st->timeout = PFTM_PURGE; 522 st->src.state = st->dst.state = 523 TCPS_CLOSED; 524 killed++; 525 } 526 } 527 if (pf_status.debug >= PF_DEBUG_MISC) 528 kprintf(", %u states killed", killed); 529 } 530 if (pf_status.debug >= PF_DEBUG_MISC) 531 kprintf("\n"); 532 } 533 534 /* kill this state */ 535 (*state)->timeout = PFTM_PURGE; 536 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED; 537 return (1); 538 } 539 540 int 541 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule, 542 struct pf_addr *src, sa_family_t af) 543 { 544 struct pf_src_node k; 545 546 if (*sn == NULL) { 547 k.af = af; 548 PF_ACPY(&k.addr, src, af); 549 if (rule->rule_flag & PFRULE_RULESRCTRACK || 550 rule->rpool.opts & PF_POOL_STICKYADDR) 551 k.rule.ptr = rule; 552 else 553 k.rule.ptr = NULL; 554 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++; 555 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k); 556 } 557 if (*sn == NULL) { 558 if (!rule->max_src_nodes || 559 rule->src_nodes < rule->max_src_nodes) 560 (*sn) = kmalloc(sizeof(struct pf_src_node), M_PFSRCTREEPL, M_NOWAIT|M_ZERO); 561 else 562 pf_status.lcounters[LCNT_SRCNODES]++; 563 if ((*sn) == NULL) 564 return (-1); 565 566 pf_init_threshold(&(*sn)->conn_rate, 567 rule->max_src_conn_rate.limit, 568 rule->max_src_conn_rate.seconds); 569 570 (*sn)->af = af; 571 if (rule->rule_flag & PFRULE_RULESRCTRACK || 572 rule->rpool.opts & PF_POOL_STICKYADDR) 573 (*sn)->rule.ptr = rule; 574 else 575 (*sn)->rule.ptr = NULL; 576 PF_ACPY(&(*sn)->addr, src, af); 577 if (RB_INSERT(pf_src_tree, 578 &tree_src_tracking, *sn) != NULL) { 579 if (pf_status.debug >= PF_DEBUG_MISC) { 580 kprintf("pf: src_tree insert failed: "); 581 pf_print_host(&(*sn)->addr, 0, af); 582 kprintf("\n"); 583 } 584 kfree(*sn, M_PFSRCTREEPL); 585 return (-1); 586 } 587 (*sn)->creation = time_second; 588 (*sn)->ruletype = rule->action; 589 if ((*sn)->rule.ptr != NULL) 590 (*sn)->rule.ptr->src_nodes++; 591 pf_status.scounters[SCNT_SRC_NODE_INSERT]++; 592 pf_status.src_nodes++; 593 } else { 594 if (rule->max_src_states && 595 (*sn)->states >= rule->max_src_states) { 596 pf_status.lcounters[LCNT_SRCSTATES]++; 597 return (-1); 598 } 599 } 600 return (0); 601 } 602 603 /* state table stuff */ 604 605 static __inline int 606 pf_state_compare_key(struct pf_state_key *a, struct pf_state_key *b) 607 { 608 int diff; 609 610 if ((diff = a->proto - b->proto) != 0) 611 return (diff); 612 if ((diff = a->af - b->af) != 0) 613 return (diff); 614 switch (a->af) { 615 #ifdef INET 616 case AF_INET: 617 if (a->addr[0].addr32[0] > b->addr[0].addr32[0]) 618 return (1); 619 if (a->addr[0].addr32[0] < b->addr[0].addr32[0]) 620 return (-1); 621 if (a->addr[1].addr32[0] > b->addr[1].addr32[0]) 622 return (1); 623 if (a->addr[1].addr32[0] < b->addr[1].addr32[0]) 624 return (-1); 625 break; 626 #endif /* INET */ 627 #ifdef INET6 628 case AF_INET6: 629 if (a->addr[0].addr32[3] > b->addr[0].addr32[3]) 630 return (1); 631 if (a->addr[0].addr32[3] < b->addr[0].addr32[3]) 632 return (-1); 633 if (a->addr[1].addr32[3] > b->addr[1].addr32[3]) 634 return (1); 635 if (a->addr[1].addr32[3] < b->addr[1].addr32[3]) 636 return (-1); 637 if (a->addr[0].addr32[2] > b->addr[0].addr32[2]) 638 return (1); 639 if (a->addr[0].addr32[2] < b->addr[0].addr32[2]) 640 return (-1); 641 if (a->addr[1].addr32[2] > b->addr[1].addr32[2]) 642 return (1); 643 if (a->addr[1].addr32[2] < b->addr[1].addr32[2]) 644 return (-1); 645 if (a->addr[0].addr32[1] > b->addr[0].addr32[1]) 646 return (1); 647 if (a->addr[0].addr32[1] < b->addr[0].addr32[1]) 648 return (-1); 649 if (a->addr[1].addr32[1] > b->addr[1].addr32[1]) 650 return (1); 651 if (a->addr[1].addr32[1] < b->addr[1].addr32[1]) 652 return (-1); 653 if (a->addr[0].addr32[0] > b->addr[0].addr32[0]) 654 return (1); 655 if (a->addr[0].addr32[0] < b->addr[0].addr32[0]) 656 return (-1); 657 if (a->addr[1].addr32[0] > b->addr[1].addr32[0]) 658 return (1); 659 if (a->addr[1].addr32[0] < b->addr[1].addr32[0]) 660 return (-1); 661 break; 662 #endif /* INET6 */ 663 } 664 665 if ((diff = a->port[0] - b->port[0]) != 0) 666 return (diff); 667 if ((diff = a->port[1] - b->port[1]) != 0) 668 return (diff); 669 670 return (0); 671 } 672 673 static __inline int 674 pf_state_compare_id(struct pf_state *a, struct pf_state *b) 675 { 676 if (a->id > b->id) 677 return (1); 678 if (a->id < b->id) 679 return (-1); 680 if (a->creatorid > b->creatorid) 681 return (1); 682 if (a->creatorid < b->creatorid) 683 return (-1); 684 685 return (0); 686 } 687 688 int 689 pf_state_key_attach(struct pf_state_key *sk, struct pf_state *s, int idx) 690 { 691 struct pf_state_item *si; 692 struct pf_state_key *cur; 693 694 KKASSERT(s->key[idx] == NULL); /* XXX handle this? */ 695 696 if ((cur = RB_INSERT(pf_state_tree, &pf_statetbl, sk)) != NULL) { 697 /* key exists. check for same kif, if none, add to key */ 698 TAILQ_FOREACH(si, &cur->states, entry) 699 if (si->s->kif == s->kif && 700 si->s->direction == s->direction) { 701 if (pf_status.debug >= PF_DEBUG_MISC) { 702 kprintf( 703 "pf: %s key attach failed on %s: ", 704 (idx == PF_SK_WIRE) ? 705 "wire" : "stack", 706 s->kif->pfik_name); 707 pf_print_state_parts(s, 708 (idx == PF_SK_WIRE) ? sk : NULL, 709 (idx == PF_SK_STACK) ? sk : NULL); 710 kprintf("\n"); 711 } 712 kfree(sk, M_PFSTATEKEYPL); 713 return (-1); /* collision! */ 714 } 715 kfree(sk, M_PFSTATEKEYPL); 716 717 s->key[idx] = cur; 718 } else 719 s->key[idx] = sk; 720 721 if ((si = kmalloc(sizeof(struct pf_state_item), M_PFSTATEITEMPL, M_NOWAIT)) == NULL) { 722 pf_state_key_detach(s, idx); 723 return (-1); 724 } 725 si->s = s; 726 727 /* list is sorted, if-bound states before floating */ 728 if (s->kif == pfi_all) 729 TAILQ_INSERT_TAIL(&s->key[idx]->states, si, entry); 730 else 731 TAILQ_INSERT_HEAD(&s->key[idx]->states, si, entry); 732 return (0); 733 } 734 735 void 736 pf_detach_state(struct pf_state *s) 737 { 738 if (s->key[PF_SK_WIRE] == s->key[PF_SK_STACK]) 739 s->key[PF_SK_WIRE] = NULL; 740 741 if (s->key[PF_SK_STACK] != NULL) 742 pf_state_key_detach(s, PF_SK_STACK); 743 744 if (s->key[PF_SK_WIRE] != NULL) 745 pf_state_key_detach(s, PF_SK_WIRE); 746 } 747 748 void 749 pf_state_key_detach(struct pf_state *s, int idx) 750 { 751 struct pf_state_item *si; 752 si = TAILQ_FIRST(&s->key[idx]->states); 753 while (si && si->s != s) 754 si = TAILQ_NEXT(si, entry); 755 756 if (si) { 757 TAILQ_REMOVE(&s->key[idx]->states, si, entry); 758 kfree(si, M_PFSTATEITEMPL); 759 } 760 761 if (TAILQ_EMPTY(&s->key[idx]->states)) { 762 RB_REMOVE(pf_state_tree, &pf_statetbl, s->key[idx]); 763 if (s->key[idx]->reverse) 764 s->key[idx]->reverse->reverse = NULL; 765 if (s->key[idx]->inp) 766 s->key[idx]->inp->inp_pf_sk = NULL; 767 kfree(s->key[idx], M_PFSTATEKEYPL); 768 } 769 s->key[idx] = NULL; 770 } 771 772 struct pf_state_key * 773 pf_alloc_state_key(int pool_flags) 774 { 775 struct pf_state_key *sk; 776 777 if ((sk = kmalloc(sizeof(struct pf_state_key), M_PFSTATEKEYPL, pool_flags)) == NULL) 778 return (NULL); 779 TAILQ_INIT(&sk->states); 780 781 return (sk); 782 } 783 784 int 785 pf_state_key_setup(struct pf_pdesc *pd, struct pf_rule *nr, 786 struct pf_state_key **skw, struct pf_state_key **sks, 787 struct pf_state_key **skp, struct pf_state_key **nkp, 788 struct pf_addr *saddr, struct pf_addr *daddr, 789 u_int16_t sport, u_int16_t dport) 790 { 791 KKASSERT((*skp == NULL && *nkp == NULL)); 792 793 if ((*skp = pf_alloc_state_key(M_NOWAIT | M_ZERO)) == NULL) 794 return (ENOMEM); 795 796 PF_ACPY(&(*skp)->addr[pd->sidx], saddr, pd->af); 797 PF_ACPY(&(*skp)->addr[pd->didx], daddr, pd->af); 798 (*skp)->port[pd->sidx] = sport; 799 (*skp)->port[pd->didx] = dport; 800 (*skp)->proto = pd->proto; 801 (*skp)->af = pd->af; 802 803 if (nr != NULL) { 804 if ((*nkp = pf_alloc_state_key(M_NOWAIT | M_ZERO)) == NULL) 805 return (ENOMEM); /* caller must handle cleanup */ 806 807 /* XXX maybe just bcopy and TAILQ_INIT(&(*nkp)->states) */ 808 PF_ACPY(&(*nkp)->addr[0], &(*skp)->addr[0], pd->af); 809 PF_ACPY(&(*nkp)->addr[1], &(*skp)->addr[1], pd->af); 810 (*nkp)->port[0] = (*skp)->port[0]; 811 (*nkp)->port[1] = (*skp)->port[1]; 812 (*nkp)->proto = pd->proto; 813 (*nkp)->af = pd->af; 814 } else 815 *nkp = *skp; 816 817 if (pd->dir == PF_IN) { 818 *skw = *skp; 819 *sks = *nkp; 820 } else { 821 *sks = *skp; 822 *skw = *nkp; 823 } 824 return (0); 825 } 826 827 828 int 829 pf_state_insert(struct pfi_kif *kif, struct pf_state_key *skw, 830 struct pf_state_key *sks, struct pf_state *s) 831 { 832 s->kif = kif; 833 834 if (skw == sks) { 835 if (pf_state_key_attach(skw, s, PF_SK_WIRE)) 836 return (-1); 837 s->key[PF_SK_STACK] = s->key[PF_SK_WIRE]; 838 } else { 839 if (pf_state_key_attach(skw, s, PF_SK_WIRE)) { 840 kfree(sks, M_PFSTATEKEYPL); 841 return (-1); 842 } 843 if (pf_state_key_attach(sks, s, PF_SK_STACK)) { 844 pf_state_key_detach(s, PF_SK_WIRE); 845 return (-1); 846 } 847 } 848 849 if (s->id == 0 && s->creatorid == 0) { 850 s->id = htobe64(pf_status.stateid++); 851 s->creatorid = pf_status.hostid; 852 } 853 854 /* 855 * Calculate hash code for altq 856 */ 857 s->hash = crc32(s->key[PF_SK_WIRE], sizeof(*sks)); 858 859 if (RB_INSERT(pf_state_tree_id, &tree_id, s) != NULL) { 860 if (pf_status.debug >= PF_DEBUG_MISC) { 861 kprintf("pf: state insert failed: " 862 "id: %016jx creatorid: %08x", 863 (uintmax_t)be64toh(s->id), ntohl(s->creatorid)); 864 if (s->sync_flags & PFSTATE_FROMSYNC) 865 kprintf(" (from sync)"); 866 kprintf("\n"); 867 } 868 pf_detach_state(s); 869 return (-1); 870 } 871 TAILQ_INSERT_TAIL(&state_list, s, entry_list); 872 pf_status.fcounters[FCNT_STATE_INSERT]++; 873 pf_status.states++; 874 pfi_kif_ref(kif, PFI_KIF_REF_STATE); 875 pfsync_insert_state(s); 876 return (0); 877 } 878 879 struct pf_state * 880 pf_find_state_byid(struct pf_state_cmp *key) 881 { 882 pf_status.fcounters[FCNT_STATE_SEARCH]++; 883 884 return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key)); 885 } 886 887 struct pf_state * 888 pf_find_state(struct pfi_kif *kif, struct pf_state_key_cmp *key, u_int dir, 889 struct mbuf *m) 890 { 891 struct pf_state_key *sk; 892 struct pf_state_item *si; 893 894 pf_status.fcounters[FCNT_STATE_SEARCH]++; 895 896 if (dir == PF_OUT && m->m_pkthdr.pf.statekey && 897 ((struct pf_state_key *)m->m_pkthdr.pf.statekey)->reverse) 898 sk = ((struct pf_state_key *)m->m_pkthdr.pf.statekey)->reverse; 899 else { 900 if ((sk = RB_FIND(pf_state_tree, &pf_statetbl, 901 (struct pf_state_key *)key)) == NULL) 902 return (NULL); 903 if (dir == PF_OUT && m->m_pkthdr.pf.statekey) { 904 ((struct pf_state_key *) 905 m->m_pkthdr.pf.statekey)->reverse = sk; 906 sk->reverse = m->m_pkthdr.pf.statekey; 907 } 908 } 909 910 if (dir == PF_OUT) 911 m->m_pkthdr.pf.statekey = NULL; 912 913 /* list is sorted, if-bound states before floating ones */ 914 TAILQ_FOREACH(si, &sk->states, entry) 915 if ((si->s->kif == pfi_all || si->s->kif == kif) && 916 sk == (dir == PF_IN ? si->s->key[PF_SK_WIRE] : 917 si->s->key[PF_SK_STACK])) 918 return (si->s); 919 920 return (NULL); 921 } 922 923 struct pf_state * 924 pf_find_state_all(struct pf_state_key_cmp *key, u_int dir, int *more) 925 { 926 struct pf_state_key *sk; 927 struct pf_state_item *si, *ret = NULL; 928 929 pf_status.fcounters[FCNT_STATE_SEARCH]++; 930 931 sk = RB_FIND(pf_state_tree, &pf_statetbl, (struct pf_state_key *)key); 932 933 if (sk != NULL) { 934 TAILQ_FOREACH(si, &sk->states, entry) 935 if (dir == PF_INOUT || 936 (sk == (dir == PF_IN ? si->s->key[PF_SK_WIRE] : 937 si->s->key[PF_SK_STACK]))) { 938 if (more == NULL) 939 return (si->s); 940 941 if (ret) 942 (*more)++; 943 else 944 ret = si; 945 } 946 } 947 return (ret ? ret->s : NULL); 948 } 949 950 /* END state table stuff */ 951 952 953 void 954 pf_purge_thread(void *v) 955 { 956 int nloops = 0; 957 int locked = 0; 958 959 lwkt_gettoken(&pf_token); 960 for (;;) { 961 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz); 962 963 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE); 964 965 if (pf_end_threads) { 966 pf_purge_expired_states(pf_status.states, 0); 967 pf_purge_expired_fragments(); 968 pf_purge_expired_src_nodes(1); 969 pf_end_threads++; 970 971 lockmgr(&pf_consistency_lock, LK_RELEASE); 972 wakeup(pf_purge_thread); 973 kthread_exit(); 974 } 975 crit_enter(); 976 977 /* process a fraction of the state table every second */ 978 if(!pf_purge_expired_states(1 + (pf_status.states 979 / pf_default_rule.timeout[PFTM_INTERVAL]), 0)) { 980 981 pf_purge_expired_states(1 + (pf_status.states 982 / pf_default_rule.timeout[PFTM_INTERVAL]), 1); 983 } 984 985 /* purge other expired types every PFTM_INTERVAL seconds */ 986 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) { 987 pf_purge_expired_fragments(); 988 if (!pf_purge_expired_src_nodes(locked)) { 989 pf_purge_expired_src_nodes(1); 990 } 991 nloops = 0; 992 } 993 crit_exit(); 994 lockmgr(&pf_consistency_lock, LK_RELEASE); 995 } 996 lwkt_reltoken(&pf_token); 997 } 998 999 u_int32_t 1000 pf_state_expires(const struct pf_state *state) 1001 { 1002 u_int32_t timeout; 1003 u_int32_t start; 1004 u_int32_t end; 1005 u_int32_t states; 1006 1007 /* handle all PFTM_* > PFTM_MAX here */ 1008 if (state->timeout == PFTM_PURGE) 1009 return (time_second); 1010 if (state->timeout == PFTM_UNTIL_PACKET) 1011 return (0); 1012 KKASSERT(state->timeout != PFTM_UNLINKED); 1013 KKASSERT(state->timeout < PFTM_MAX); 1014 timeout = state->rule.ptr->timeout[state->timeout]; 1015 if (!timeout) 1016 timeout = pf_default_rule.timeout[state->timeout]; 1017 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START]; 1018 if (start) { 1019 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END]; 1020 states = state->rule.ptr->states_cur; 1021 } else { 1022 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START]; 1023 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END]; 1024 states = pf_status.states; 1025 } 1026 if (end && states > start && start < end) { 1027 if (states < end) 1028 return (state->expire + timeout * (end - states) / 1029 (end - start)); 1030 else 1031 return (time_second); 1032 } 1033 return (state->expire + timeout); 1034 } 1035 1036 int 1037 pf_purge_expired_src_nodes(int waslocked) 1038 { 1039 struct pf_src_node *cur, *next; 1040 int locked = waslocked; 1041 1042 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) { 1043 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur); 1044 1045 if (cur->states <= 0 && cur->expire <= time_second) { 1046 if (! locked) { 1047 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE); 1048 next = RB_NEXT(pf_src_tree, 1049 &tree_src_tracking, cur); 1050 locked = 1; 1051 } 1052 if (cur->rule.ptr != NULL) { 1053 cur->rule.ptr->src_nodes--; 1054 if (cur->rule.ptr->states_cur <= 0 && 1055 cur->rule.ptr->max_src_nodes <= 0) 1056 pf_rm_rule(NULL, cur->rule.ptr); 1057 } 1058 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur); 1059 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 1060 pf_status.src_nodes--; 1061 kfree(cur, M_PFSRCTREEPL); 1062 } 1063 } 1064 1065 if (locked && !waslocked) 1066 lockmgr(&pf_consistency_lock, LK_RELEASE); 1067 return(1); 1068 } 1069 1070 void 1071 pf_src_tree_remove_state(struct pf_state *s) 1072 { 1073 u_int32_t timeout; 1074 1075 if (s->src_node != NULL) { 1076 if (s->src.tcp_est) 1077 --s->src_node->conn; 1078 if (--s->src_node->states <= 0) { 1079 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE]; 1080 if (!timeout) 1081 timeout = 1082 pf_default_rule.timeout[PFTM_SRC_NODE]; 1083 s->src_node->expire = time_second + timeout; 1084 } 1085 } 1086 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) { 1087 if (--s->nat_src_node->states <= 0) { 1088 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE]; 1089 if (!timeout) 1090 timeout = 1091 pf_default_rule.timeout[PFTM_SRC_NODE]; 1092 s->nat_src_node->expire = time_second + timeout; 1093 } 1094 } 1095 s->src_node = s->nat_src_node = NULL; 1096 } 1097 1098 /* callers should be at crit_enter() */ 1099 void 1100 pf_unlink_state(struct pf_state *cur) 1101 { 1102 if (cur->src.state == PF_TCPS_PROXY_DST) { 1103 /* XXX wire key the right one? */ 1104 pf_send_tcp(cur->rule.ptr, cur->key[PF_SK_WIRE]->af, 1105 &cur->key[PF_SK_WIRE]->addr[1], 1106 &cur->key[PF_SK_WIRE]->addr[0], 1107 cur->key[PF_SK_WIRE]->port[1], 1108 cur->key[PF_SK_WIRE]->port[0], 1109 cur->src.seqhi, cur->src.seqlo + 1, 1110 TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL); 1111 } 1112 RB_REMOVE(pf_state_tree_id, &tree_id, cur); 1113 if (cur->creatorid == pf_status.hostid) 1114 pfsync_delete_state(cur); 1115 cur->timeout = PFTM_UNLINKED; 1116 pf_src_tree_remove_state(cur); 1117 pf_detach_state(cur); 1118 } 1119 1120 static struct pf_state *purge_cur; 1121 1122 /* callers should be at crit_enter() and hold the 1123 * write_lock on pf_consistency_lock */ 1124 void 1125 pf_free_state(struct pf_state *cur) 1126 { 1127 if (pfsyncif != NULL && 1128 (pfsyncif->sc_bulk_send_next == cur || 1129 pfsyncif->sc_bulk_terminator == cur)) 1130 return; 1131 KKASSERT(cur->timeout == PFTM_UNLINKED); 1132 if (--cur->rule.ptr->states_cur <= 0 && 1133 cur->rule.ptr->src_nodes <= 0) 1134 pf_rm_rule(NULL, cur->rule.ptr); 1135 if (cur->nat_rule.ptr != NULL) 1136 if (--cur->nat_rule.ptr->states_cur <= 0 && 1137 cur->nat_rule.ptr->src_nodes <= 0) 1138 pf_rm_rule(NULL, cur->nat_rule.ptr); 1139 if (cur->anchor.ptr != NULL) 1140 if (--cur->anchor.ptr->states_cur <= 0) 1141 pf_rm_rule(NULL, cur->anchor.ptr); 1142 pf_normalize_tcp_cleanup(cur); 1143 pfi_kif_unref(cur->kif, PFI_KIF_REF_STATE); 1144 1145 /* 1146 * We may be freeing pf_purge_expired_states()'s saved scan entry, 1147 * adjust it if necessary. 1148 */ 1149 if (purge_cur == cur) { 1150 kprintf("PURGE CONFLICT\n"); 1151 purge_cur = TAILQ_NEXT(purge_cur, entry_list); 1152 } 1153 TAILQ_REMOVE(&state_list, cur, entry_list); 1154 if (cur->tag) 1155 pf_tag_unref(cur->tag); 1156 kfree(cur, M_PFSTATEPL); 1157 pf_status.fcounters[FCNT_STATE_REMOVALS]++; 1158 pf_status.states--; 1159 } 1160 1161 int 1162 pf_purge_expired_states(u_int32_t maxcheck, int waslocked) 1163 { 1164 struct pf_state *cur; 1165 int locked = waslocked; 1166 1167 while (maxcheck--) { 1168 /* 1169 * Wrap to start of list when we hit the end 1170 */ 1171 cur = purge_cur; 1172 if (cur == NULL) { 1173 cur = TAILQ_FIRST(&state_list); 1174 if (cur == NULL) 1175 break; /* list empty */ 1176 } 1177 1178 /* 1179 * Setup next (purge_cur) while we process this one. If we block and 1180 * something else deletes purge_cur, pf_free_state() will adjust it further 1181 * ahead. 1182 */ 1183 purge_cur = TAILQ_NEXT(cur, entry_list); 1184 1185 if (cur->timeout == PFTM_UNLINKED) { 1186 /* free unlinked state */ 1187 if (! locked) { 1188 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE); 1189 locked = 1; 1190 } 1191 pf_free_state(cur); 1192 } else if (pf_state_expires(cur) <= time_second) { 1193 /* unlink and free expired state */ 1194 pf_unlink_state(cur); 1195 if (! locked) { 1196 if (!lockmgr(&pf_consistency_lock, LK_EXCLUSIVE)) 1197 return (0); 1198 locked = 1; 1199 } 1200 pf_free_state(cur); 1201 } 1202 } 1203 1204 if (locked) 1205 lockmgr(&pf_consistency_lock, LK_RELEASE); 1206 return (1); 1207 } 1208 1209 int 1210 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw) 1211 { 1212 if (aw->type != PF_ADDR_TABLE) 1213 return (0); 1214 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL) 1215 return (1); 1216 return (0); 1217 } 1218 1219 void 1220 pf_tbladdr_remove(struct pf_addr_wrap *aw) 1221 { 1222 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL) 1223 return; 1224 pfr_detach_table(aw->p.tbl); 1225 aw->p.tbl = NULL; 1226 } 1227 1228 void 1229 pf_tbladdr_copyout(struct pf_addr_wrap *aw) 1230 { 1231 struct pfr_ktable *kt = aw->p.tbl; 1232 1233 if (aw->type != PF_ADDR_TABLE || kt == NULL) 1234 return; 1235 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) 1236 kt = kt->pfrkt_root; 1237 aw->p.tbl = NULL; 1238 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ? 1239 kt->pfrkt_cnt : -1; 1240 } 1241 1242 void 1243 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af) 1244 { 1245 switch (af) { 1246 #ifdef INET 1247 case AF_INET: { 1248 u_int32_t a = ntohl(addr->addr32[0]); 1249 kprintf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255, 1250 (a>>8)&255, a&255); 1251 if (p) { 1252 p = ntohs(p); 1253 kprintf(":%u", p); 1254 } 1255 break; 1256 } 1257 #endif /* INET */ 1258 #ifdef INET6 1259 case AF_INET6: { 1260 u_int16_t b; 1261 u_int8_t i, curstart = 255, curend = 0, 1262 maxstart = 0, maxend = 0; 1263 for (i = 0; i < 8; i++) { 1264 if (!addr->addr16[i]) { 1265 if (curstart == 255) 1266 curstart = i; 1267 else 1268 curend = i; 1269 } else { 1270 if (curstart) { 1271 if ((curend - curstart) > 1272 (maxend - maxstart)) { 1273 maxstart = curstart; 1274 maxend = curend; 1275 curstart = 255; 1276 } 1277 } 1278 } 1279 } 1280 for (i = 0; i < 8; i++) { 1281 if (i >= maxstart && i <= maxend) { 1282 if (maxend != 7) { 1283 if (i == maxstart) 1284 kprintf(":"); 1285 } else { 1286 if (i == maxend) 1287 kprintf(":"); 1288 } 1289 } else { 1290 b = ntohs(addr->addr16[i]); 1291 kprintf("%x", b); 1292 if (i < 7) 1293 kprintf(":"); 1294 } 1295 } 1296 if (p) { 1297 p = ntohs(p); 1298 kprintf("[%u]", p); 1299 } 1300 break; 1301 } 1302 #endif /* INET6 */ 1303 } 1304 } 1305 1306 void 1307 pf_print_state(struct pf_state *s) 1308 { 1309 pf_print_state_parts(s, NULL, NULL); 1310 } 1311 1312 void 1313 pf_print_state_parts(struct pf_state *s, 1314 struct pf_state_key *skwp, struct pf_state_key *sksp) 1315 { 1316 struct pf_state_key *skw, *sks; 1317 u_int8_t proto, dir; 1318 1319 /* Do our best to fill these, but they're skipped if NULL */ 1320 skw = skwp ? skwp : (s ? s->key[PF_SK_WIRE] : NULL); 1321 sks = sksp ? sksp : (s ? s->key[PF_SK_STACK] : NULL); 1322 proto = skw ? skw->proto : (sks ? sks->proto : 0); 1323 dir = s ? s->direction : 0; 1324 1325 switch (proto) { 1326 case IPPROTO_TCP: 1327 kprintf("TCP "); 1328 break; 1329 case IPPROTO_UDP: 1330 kprintf("UDP "); 1331 break; 1332 case IPPROTO_ICMP: 1333 kprintf("ICMP "); 1334 break; 1335 case IPPROTO_ICMPV6: 1336 kprintf("ICMPV6 "); 1337 break; 1338 default: 1339 kprintf("%u ", skw->proto); 1340 break; 1341 } 1342 switch (dir) { 1343 case PF_IN: 1344 kprintf(" in"); 1345 break; 1346 case PF_OUT: 1347 kprintf(" out"); 1348 break; 1349 } 1350 if (skw) { 1351 kprintf(" wire: "); 1352 pf_print_host(&skw->addr[0], skw->port[0], skw->af); 1353 kprintf(" "); 1354 pf_print_host(&skw->addr[1], skw->port[1], skw->af); 1355 } 1356 if (sks) { 1357 kprintf(" stack: "); 1358 if (sks != skw) { 1359 pf_print_host(&sks->addr[0], sks->port[0], sks->af); 1360 kprintf(" "); 1361 pf_print_host(&sks->addr[1], sks->port[1], sks->af); 1362 } else 1363 kprintf("-"); 1364 } 1365 if (s) { 1366 if (proto == IPPROTO_TCP) { 1367 kprintf(" [lo=%u high=%u win=%u modulator=%u", 1368 s->src.seqlo, s->src.seqhi, 1369 s->src.max_win, s->src.seqdiff); 1370 if (s->src.wscale && s->dst.wscale) 1371 kprintf(" wscale=%u", 1372 s->src.wscale & PF_WSCALE_MASK); 1373 kprintf("]"); 1374 kprintf(" [lo=%u high=%u win=%u modulator=%u", 1375 s->dst.seqlo, s->dst.seqhi, 1376 s->dst.max_win, s->dst.seqdiff); 1377 if (s->src.wscale && s->dst.wscale) 1378 kprintf(" wscale=%u", 1379 s->dst.wscale & PF_WSCALE_MASK); 1380 kprintf("]"); 1381 } 1382 kprintf(" %u:%u", s->src.state, s->dst.state); 1383 } 1384 } 1385 1386 void 1387 pf_print_flags(u_int8_t f) 1388 { 1389 if (f) 1390 kprintf(" "); 1391 if (f & TH_FIN) 1392 kprintf("F"); 1393 if (f & TH_SYN) 1394 kprintf("S"); 1395 if (f & TH_RST) 1396 kprintf("R"); 1397 if (f & TH_PUSH) 1398 kprintf("P"); 1399 if (f & TH_ACK) 1400 kprintf("A"); 1401 if (f & TH_URG) 1402 kprintf("U"); 1403 if (f & TH_ECE) 1404 kprintf("E"); 1405 if (f & TH_CWR) 1406 kprintf("W"); 1407 } 1408 1409 #define PF_SET_SKIP_STEPS(i) \ 1410 do { \ 1411 while (head[i] != cur) { \ 1412 head[i]->skip[i].ptr = cur; \ 1413 head[i] = TAILQ_NEXT(head[i], entries); \ 1414 } \ 1415 } while (0) 1416 1417 void 1418 pf_calc_skip_steps(struct pf_rulequeue *rules) 1419 { 1420 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT]; 1421 int i; 1422 1423 cur = TAILQ_FIRST(rules); 1424 prev = cur; 1425 for (i = 0; i < PF_SKIP_COUNT; ++i) 1426 head[i] = cur; 1427 while (cur != NULL) { 1428 1429 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot) 1430 PF_SET_SKIP_STEPS(PF_SKIP_IFP); 1431 if (cur->direction != prev->direction) 1432 PF_SET_SKIP_STEPS(PF_SKIP_DIR); 1433 if (cur->af != prev->af) 1434 PF_SET_SKIP_STEPS(PF_SKIP_AF); 1435 if (cur->proto != prev->proto) 1436 PF_SET_SKIP_STEPS(PF_SKIP_PROTO); 1437 if (cur->src.neg != prev->src.neg || 1438 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr)) 1439 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR); 1440 if (cur->src.port[0] != prev->src.port[0] || 1441 cur->src.port[1] != prev->src.port[1] || 1442 cur->src.port_op != prev->src.port_op) 1443 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT); 1444 if (cur->dst.neg != prev->dst.neg || 1445 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr)) 1446 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR); 1447 if (cur->dst.port[0] != prev->dst.port[0] || 1448 cur->dst.port[1] != prev->dst.port[1] || 1449 cur->dst.port_op != prev->dst.port_op) 1450 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT); 1451 1452 prev = cur; 1453 cur = TAILQ_NEXT(cur, entries); 1454 } 1455 for (i = 0; i < PF_SKIP_COUNT; ++i) 1456 PF_SET_SKIP_STEPS(i); 1457 } 1458 1459 int 1460 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2) 1461 { 1462 if (aw1->type != aw2->type) 1463 return (1); 1464 switch (aw1->type) { 1465 case PF_ADDR_ADDRMASK: 1466 case PF_ADDR_RANGE: 1467 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0)) 1468 return (1); 1469 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0)) 1470 return (1); 1471 return (0); 1472 case PF_ADDR_DYNIFTL: 1473 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt); 1474 case PF_ADDR_NOROUTE: 1475 case PF_ADDR_URPFFAILED: 1476 return (0); 1477 case PF_ADDR_TABLE: 1478 return (aw1->p.tbl != aw2->p.tbl); 1479 case PF_ADDR_RTLABEL: 1480 return (aw1->v.rtlabel != aw2->v.rtlabel); 1481 default: 1482 kprintf("invalid address type: %d\n", aw1->type); 1483 return (1); 1484 } 1485 } 1486 1487 u_int16_t 1488 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp) 1489 { 1490 u_int32_t l; 1491 1492 if (udp && !cksum) 1493 return (0x0000); 1494 l = cksum + old - new; 1495 l = (l >> 16) + (l & 65535); 1496 l = l & 65535; 1497 if (udp && !l) 1498 return (0xFFFF); 1499 return (l); 1500 } 1501 1502 void 1503 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc, 1504 struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af) 1505 { 1506 struct pf_addr ao; 1507 u_int16_t po = *p; 1508 1509 PF_ACPY(&ao, a, af); 1510 PF_ACPY(a, an, af); 1511 1512 *p = pn; 1513 1514 switch (af) { 1515 #ifdef INET 1516 case AF_INET: 1517 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic, 1518 ao.addr16[0], an->addr16[0], 0), 1519 ao.addr16[1], an->addr16[1], 0); 1520 *p = pn; 1521 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc, 1522 ao.addr16[0], an->addr16[0], u), 1523 ao.addr16[1], an->addr16[1], u), 1524 po, pn, u); 1525 break; 1526 #endif /* INET */ 1527 #ifdef INET6 1528 case AF_INET6: 1529 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1530 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1531 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc, 1532 ao.addr16[0], an->addr16[0], u), 1533 ao.addr16[1], an->addr16[1], u), 1534 ao.addr16[2], an->addr16[2], u), 1535 ao.addr16[3], an->addr16[3], u), 1536 ao.addr16[4], an->addr16[4], u), 1537 ao.addr16[5], an->addr16[5], u), 1538 ao.addr16[6], an->addr16[6], u), 1539 ao.addr16[7], an->addr16[7], u), 1540 po, pn, u); 1541 break; 1542 #endif /* INET6 */ 1543 } 1544 } 1545 1546 1547 /* Changes a u_int32_t. Uses a void * so there are no align restrictions */ 1548 void 1549 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u) 1550 { 1551 u_int32_t ao; 1552 1553 memcpy(&ao, a, sizeof(ao)); 1554 memcpy(a, &an, sizeof(u_int32_t)); 1555 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u), 1556 ao % 65536, an % 65536, u); 1557 } 1558 1559 #ifdef INET6 1560 void 1561 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u) 1562 { 1563 struct pf_addr ao; 1564 1565 PF_ACPY(&ao, a, AF_INET6); 1566 PF_ACPY(a, an, AF_INET6); 1567 1568 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1569 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1570 pf_cksum_fixup(pf_cksum_fixup(*c, 1571 ao.addr16[0], an->addr16[0], u), 1572 ao.addr16[1], an->addr16[1], u), 1573 ao.addr16[2], an->addr16[2], u), 1574 ao.addr16[3], an->addr16[3], u), 1575 ao.addr16[4], an->addr16[4], u), 1576 ao.addr16[5], an->addr16[5], u), 1577 ao.addr16[6], an->addr16[6], u), 1578 ao.addr16[7], an->addr16[7], u); 1579 } 1580 #endif /* INET6 */ 1581 1582 void 1583 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa, 1584 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c, 1585 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af) 1586 { 1587 struct pf_addr oia, ooa; 1588 1589 PF_ACPY(&oia, ia, af); 1590 if (oa) 1591 PF_ACPY(&ooa, oa, af); 1592 1593 /* Change inner protocol port, fix inner protocol checksum. */ 1594 if (ip != NULL) { 1595 u_int16_t oip = *ip; 1596 u_int32_t opc = 0; 1597 1598 if (pc != NULL) 1599 opc = *pc; 1600 *ip = np; 1601 if (pc != NULL) 1602 *pc = pf_cksum_fixup(*pc, oip, *ip, u); 1603 *ic = pf_cksum_fixup(*ic, oip, *ip, 0); 1604 if (pc != NULL) 1605 *ic = pf_cksum_fixup(*ic, opc, *pc, 0); 1606 } 1607 /* Change inner ip address, fix inner ip and icmp checksums. */ 1608 PF_ACPY(ia, na, af); 1609 switch (af) { 1610 #ifdef INET 1611 case AF_INET: { 1612 u_int32_t oh2c = *h2c; 1613 1614 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c, 1615 oia.addr16[0], ia->addr16[0], 0), 1616 oia.addr16[1], ia->addr16[1], 0); 1617 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic, 1618 oia.addr16[0], ia->addr16[0], 0), 1619 oia.addr16[1], ia->addr16[1], 0); 1620 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0); 1621 break; 1622 } 1623 #endif /* INET */ 1624 #ifdef INET6 1625 case AF_INET6: 1626 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1627 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1628 pf_cksum_fixup(pf_cksum_fixup(*ic, 1629 oia.addr16[0], ia->addr16[0], u), 1630 oia.addr16[1], ia->addr16[1], u), 1631 oia.addr16[2], ia->addr16[2], u), 1632 oia.addr16[3], ia->addr16[3], u), 1633 oia.addr16[4], ia->addr16[4], u), 1634 oia.addr16[5], ia->addr16[5], u), 1635 oia.addr16[6], ia->addr16[6], u), 1636 oia.addr16[7], ia->addr16[7], u); 1637 break; 1638 #endif /* INET6 */ 1639 } 1640 /* Outer ip address, fix outer ip or icmpv6 checksum, if necessary. */ 1641 if (oa) { 1642 PF_ACPY(oa, na, af); 1643 switch (af) { 1644 #ifdef INET 1645 case AF_INET: 1646 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc, 1647 ooa.addr16[0], oa->addr16[0], 0), 1648 ooa.addr16[1], oa->addr16[1], 0); 1649 break; 1650 #endif /* INET */ 1651 #ifdef INET6 1652 case AF_INET6: 1653 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1654 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( 1655 pf_cksum_fixup(pf_cksum_fixup(*ic, 1656 ooa.addr16[0], oa->addr16[0], u), 1657 ooa.addr16[1], oa->addr16[1], u), 1658 ooa.addr16[2], oa->addr16[2], u), 1659 ooa.addr16[3], oa->addr16[3], u), 1660 ooa.addr16[4], oa->addr16[4], u), 1661 ooa.addr16[5], oa->addr16[5], u), 1662 ooa.addr16[6], oa->addr16[6], u), 1663 ooa.addr16[7], oa->addr16[7], u); 1664 break; 1665 #endif /* INET6 */ 1666 } 1667 } 1668 } 1669 1670 1671 /* 1672 * Need to modulate the sequence numbers in the TCP SACK option 1673 * (credits to Krzysztof Pfaff for report and patch) 1674 */ 1675 int 1676 pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd, 1677 struct tcphdr *th, struct pf_state_peer *dst) 1678 { 1679 int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen; 1680 u_int8_t opts[TCP_MAXOLEN], *opt = opts; 1681 int copyback = 0, i, olen; 1682 struct raw_sackblock sack; 1683 1684 #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2) 1685 if (hlen < TCPOLEN_SACKLEN || 1686 !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af)) 1687 return 0; 1688 1689 while (hlen >= TCPOLEN_SACKLEN) { 1690 olen = opt[1]; 1691 switch (*opt) { 1692 case TCPOPT_EOL: /* FALLTHROUGH */ 1693 case TCPOPT_NOP: 1694 opt++; 1695 hlen--; 1696 break; 1697 case TCPOPT_SACK: 1698 if (olen > hlen) 1699 olen = hlen; 1700 if (olen >= TCPOLEN_SACKLEN) { 1701 for (i = 2; i + TCPOLEN_SACK <= olen; 1702 i += TCPOLEN_SACK) { 1703 memcpy(&sack, &opt[i], sizeof(sack)); 1704 pf_change_a(&sack.rblk_start, &th->th_sum, 1705 htonl(ntohl(sack.rblk_start) - 1706 dst->seqdiff), 0); 1707 pf_change_a(&sack.rblk_end, &th->th_sum, 1708 htonl(ntohl(sack.rblk_end) - 1709 dst->seqdiff), 0); 1710 memcpy(&opt[i], &sack, sizeof(sack)); 1711 } 1712 copyback = 1; 1713 } 1714 /* FALLTHROUGH */ 1715 default: 1716 if (olen < 2) 1717 olen = 2; 1718 hlen -= olen; 1719 opt += olen; 1720 } 1721 } 1722 1723 if (copyback) 1724 m_copyback(m, off + sizeof(*th), thoptlen, opts); 1725 return (copyback); 1726 } 1727 1728 void 1729 pf_send_tcp(const struct pf_rule *r, sa_family_t af, 1730 const struct pf_addr *saddr, const struct pf_addr *daddr, 1731 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack, 1732 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag, 1733 u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp) 1734 { 1735 struct mbuf *m; 1736 int len = 0, tlen; 1737 #ifdef INET 1738 struct ip *h = NULL; 1739 #endif /* INET */ 1740 #ifdef INET6 1741 struct ip6_hdr *h6 = NULL; 1742 #endif /* INET6 */ 1743 struct tcphdr *th = NULL; 1744 char *opt; 1745 1746 ASSERT_LWKT_TOKEN_HELD(&pf_token); 1747 1748 /* maximum segment size tcp option */ 1749 tlen = sizeof(struct tcphdr); 1750 if (mss) 1751 tlen += 4; 1752 1753 switch (af) { 1754 #ifdef INET 1755 case AF_INET: 1756 len = sizeof(struct ip) + tlen; 1757 break; 1758 #endif /* INET */ 1759 #ifdef INET6 1760 case AF_INET6: 1761 len = sizeof(struct ip6_hdr) + tlen; 1762 break; 1763 #endif /* INET6 */ 1764 } 1765 1766 /* 1767 * Create outgoing mbuf. 1768 * 1769 * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags, 1770 * so make sure pf.flags is clear. 1771 */ 1772 m = m_gethdr(MB_DONTWAIT, MT_HEADER); 1773 if (m == NULL) { 1774 return; 1775 } 1776 if (tag) 1777 m->m_pkthdr.fw_flags |= PF_MBUF_TAGGED; 1778 m->m_pkthdr.pf.flags = 0; 1779 m->m_pkthdr.pf.tag = rtag; 1780 /* XXX Recheck when upgrading to > 4.4 */ 1781 m->m_pkthdr.pf.statekey = NULL; 1782 if (r != NULL && r->rtableid >= 0) 1783 m->m_pkthdr.pf.rtableid = r->rtableid; 1784 1785 #ifdef ALTQ 1786 if (r != NULL && r->qid) { 1787 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE; 1788 m->m_pkthdr.pf.qid = r->qid; 1789 m->m_pkthdr.pf.ecn_af = af; 1790 m->m_pkthdr.pf.hdr = mtod(m, struct ip *); 1791 } 1792 #endif /* ALTQ */ 1793 m->m_data += max_linkhdr; 1794 m->m_pkthdr.len = m->m_len = len; 1795 m->m_pkthdr.rcvif = NULL; 1796 bzero(m->m_data, len); 1797 switch (af) { 1798 #ifdef INET 1799 case AF_INET: 1800 h = mtod(m, struct ip *); 1801 1802 /* IP header fields included in the TCP checksum */ 1803 h->ip_p = IPPROTO_TCP; 1804 h->ip_len = tlen; 1805 h->ip_src.s_addr = saddr->v4.s_addr; 1806 h->ip_dst.s_addr = daddr->v4.s_addr; 1807 1808 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip)); 1809 break; 1810 #endif /* INET */ 1811 #ifdef INET6 1812 case AF_INET6: 1813 h6 = mtod(m, struct ip6_hdr *); 1814 1815 /* IP header fields included in the TCP checksum */ 1816 h6->ip6_nxt = IPPROTO_TCP; 1817 h6->ip6_plen = htons(tlen); 1818 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr)); 1819 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr)); 1820 1821 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr)); 1822 break; 1823 #endif /* INET6 */ 1824 } 1825 1826 /* TCP header */ 1827 th->th_sport = sport; 1828 th->th_dport = dport; 1829 th->th_seq = htonl(seq); 1830 th->th_ack = htonl(ack); 1831 th->th_off = tlen >> 2; 1832 th->th_flags = flags; 1833 th->th_win = htons(win); 1834 1835 if (mss) { 1836 opt = (char *)(th + 1); 1837 opt[0] = TCPOPT_MAXSEG; 1838 opt[1] = 4; 1839 mss = htons(mss); 1840 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2); 1841 } 1842 1843 switch (af) { 1844 #ifdef INET 1845 case AF_INET: 1846 /* TCP checksum */ 1847 th->th_sum = in_cksum(m, len); 1848 1849 /* Finish the IP header */ 1850 h->ip_v = 4; 1851 h->ip_hl = sizeof(*h) >> 2; 1852 h->ip_tos = IPTOS_LOWDELAY; 1853 h->ip_len = len; 1854 h->ip_off = path_mtu_discovery ? IP_DF : 0; 1855 h->ip_ttl = ttl ? ttl : ip_defttl; 1856 h->ip_sum = 0; 1857 if (eh == NULL) { 1858 lwkt_reltoken(&pf_token); 1859 ip_output(m, NULL, NULL, 0, NULL, NULL); 1860 lwkt_gettoken(&pf_token); 1861 } else { 1862 struct route ro; 1863 struct rtentry rt; 1864 struct ether_header *e = (void *)ro.ro_dst.sa_data; 1865 1866 if (ifp == NULL) { 1867 m_freem(m); 1868 return; 1869 } 1870 rt.rt_ifp = ifp; 1871 ro.ro_rt = &rt; 1872 ro.ro_dst.sa_len = sizeof(ro.ro_dst); 1873 ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT; 1874 bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN); 1875 bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN); 1876 e->ether_type = eh->ether_type; 1877 /* XXX_IMPORT: later */ 1878 lwkt_reltoken(&pf_token); 1879 ip_output(m, NULL, &ro, 0, NULL, NULL); 1880 lwkt_gettoken(&pf_token); 1881 } 1882 break; 1883 #endif /* INET */ 1884 #ifdef INET6 1885 case AF_INET6: 1886 /* TCP checksum */ 1887 th->th_sum = in6_cksum(m, IPPROTO_TCP, 1888 sizeof(struct ip6_hdr), tlen); 1889 1890 h6->ip6_vfc |= IPV6_VERSION; 1891 h6->ip6_hlim = IPV6_DEFHLIM; 1892 1893 lwkt_reltoken(&pf_token); 1894 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL); 1895 lwkt_gettoken(&pf_token); 1896 break; 1897 #endif /* INET6 */ 1898 } 1899 } 1900 1901 void 1902 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af, 1903 struct pf_rule *r) 1904 { 1905 struct mbuf *m0; 1906 1907 /* 1908 * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags, 1909 * so make sure pf.flags is clear. 1910 */ 1911 if ((m0 = m_copy(m, 0, M_COPYALL)) == NULL) 1912 return; 1913 1914 m0->m_pkthdr.fw_flags |= PF_MBUF_TAGGED; 1915 m0->m_pkthdr.pf.flags = 0; 1916 /* XXX Re-Check when Upgrading to > 4.4 */ 1917 m0->m_pkthdr.pf.statekey = NULL; 1918 1919 if (r->rtableid >= 0) 1920 m0->m_pkthdr.pf.rtableid = r->rtableid; 1921 1922 #ifdef ALTQ 1923 if (r->qid) { 1924 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE; 1925 m0->m_pkthdr.pf.qid = r->qid; 1926 m0->m_pkthdr.pf.ecn_af = af; 1927 m0->m_pkthdr.pf.hdr = mtod(m0, struct ip *); 1928 } 1929 #endif /* ALTQ */ 1930 1931 switch (af) { 1932 #ifdef INET 1933 case AF_INET: 1934 icmp_error(m0, type, code, 0, 0); 1935 break; 1936 #endif /* INET */ 1937 #ifdef INET6 1938 case AF_INET6: 1939 icmp6_error(m0, type, code, 0); 1940 break; 1941 #endif /* INET6 */ 1942 } 1943 } 1944 1945 /* 1946 * Return 1 if the addresses a and b match (with mask m), otherwise return 0. 1947 * If n is 0, they match if they are equal. If n is != 0, they match if they 1948 * are different. 1949 */ 1950 int 1951 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m, 1952 struct pf_addr *b, sa_family_t af) 1953 { 1954 int match = 0; 1955 1956 switch (af) { 1957 #ifdef INET 1958 case AF_INET: 1959 if ((a->addr32[0] & m->addr32[0]) == 1960 (b->addr32[0] & m->addr32[0])) 1961 match++; 1962 break; 1963 #endif /* INET */ 1964 #ifdef INET6 1965 case AF_INET6: 1966 if (((a->addr32[0] & m->addr32[0]) == 1967 (b->addr32[0] & m->addr32[0])) && 1968 ((a->addr32[1] & m->addr32[1]) == 1969 (b->addr32[1] & m->addr32[1])) && 1970 ((a->addr32[2] & m->addr32[2]) == 1971 (b->addr32[2] & m->addr32[2])) && 1972 ((a->addr32[3] & m->addr32[3]) == 1973 (b->addr32[3] & m->addr32[3]))) 1974 match++; 1975 break; 1976 #endif /* INET6 */ 1977 } 1978 if (match) { 1979 if (n) 1980 return (0); 1981 else 1982 return (1); 1983 } else { 1984 if (n) 1985 return (1); 1986 else 1987 return (0); 1988 } 1989 } 1990 1991 /* 1992 * Return 1 if b <= a <= e, otherwise return 0. 1993 */ 1994 int 1995 pf_match_addr_range(struct pf_addr *b, struct pf_addr *e, 1996 struct pf_addr *a, sa_family_t af) 1997 { 1998 switch (af) { 1999 #ifdef INET 2000 case AF_INET: 2001 if ((a->addr32[0] < b->addr32[0]) || 2002 (a->addr32[0] > e->addr32[0])) 2003 return (0); 2004 break; 2005 #endif /* INET */ 2006 #ifdef INET6 2007 case AF_INET6: { 2008 int i; 2009 2010 /* check a >= b */ 2011 for (i = 0; i < 4; ++i) 2012 if (a->addr32[i] > b->addr32[i]) 2013 break; 2014 else if (a->addr32[i] < b->addr32[i]) 2015 return (0); 2016 /* check a <= e */ 2017 for (i = 0; i < 4; ++i) 2018 if (a->addr32[i] < e->addr32[i]) 2019 break; 2020 else if (a->addr32[i] > e->addr32[i]) 2021 return (0); 2022 break; 2023 } 2024 #endif /* INET6 */ 2025 } 2026 return (1); 2027 } 2028 2029 int 2030 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p) 2031 { 2032 switch (op) { 2033 case PF_OP_IRG: 2034 return ((p > a1) && (p < a2)); 2035 case PF_OP_XRG: 2036 return ((p < a1) || (p > a2)); 2037 case PF_OP_RRG: 2038 return ((p >= a1) && (p <= a2)); 2039 case PF_OP_EQ: 2040 return (p == a1); 2041 case PF_OP_NE: 2042 return (p != a1); 2043 case PF_OP_LT: 2044 return (p < a1); 2045 case PF_OP_LE: 2046 return (p <= a1); 2047 case PF_OP_GT: 2048 return (p > a1); 2049 case PF_OP_GE: 2050 return (p >= a1); 2051 } 2052 return (0); /* never reached */ 2053 } 2054 2055 int 2056 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p) 2057 { 2058 a1 = ntohs(a1); 2059 a2 = ntohs(a2); 2060 p = ntohs(p); 2061 return (pf_match(op, a1, a2, p)); 2062 } 2063 2064 int 2065 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u) 2066 { 2067 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE) 2068 return (0); 2069 return (pf_match(op, a1, a2, u)); 2070 } 2071 2072 int 2073 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g) 2074 { 2075 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE) 2076 return (0); 2077 return (pf_match(op, a1, a2, g)); 2078 } 2079 2080 int 2081 pf_match_tag(struct mbuf *m, struct pf_rule *r, int *tag) 2082 { 2083 if (*tag == -1) 2084 *tag = m->m_pkthdr.pf.tag; 2085 2086 return ((!r->match_tag_not && r->match_tag == *tag) || 2087 (r->match_tag_not && r->match_tag != *tag)); 2088 } 2089 2090 int 2091 pf_tag_packet(struct mbuf *m, int tag, int rtableid) 2092 { 2093 if (tag <= 0 && rtableid < 0) 2094 return (0); 2095 2096 if (tag > 0) 2097 m->m_pkthdr.pf.tag = tag; 2098 if (rtableid >= 0) 2099 m->m_pkthdr.pf.rtableid = rtableid; 2100 2101 return (0); 2102 } 2103 2104 void 2105 pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n, 2106 struct pf_rule **r, struct pf_rule **a, int *match) 2107 { 2108 struct pf_anchor_stackframe *f; 2109 2110 (*r)->anchor->match = 0; 2111 if (match) 2112 *match = 0; 2113 if (*depth >= NELEM(pf_anchor_stack)) { 2114 kprintf("pf_step_into_anchor: stack overflow\n"); 2115 *r = TAILQ_NEXT(*r, entries); 2116 return; 2117 } else if (*depth == 0 && a != NULL) 2118 *a = *r; 2119 f = pf_anchor_stack + (*depth)++; 2120 f->rs = *rs; 2121 f->r = *r; 2122 if ((*r)->anchor_wildcard) { 2123 f->parent = &(*r)->anchor->children; 2124 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) == 2125 NULL) { 2126 *r = NULL; 2127 return; 2128 } 2129 *rs = &f->child->ruleset; 2130 } else { 2131 f->parent = NULL; 2132 f->child = NULL; 2133 *rs = &(*r)->anchor->ruleset; 2134 } 2135 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr); 2136 } 2137 2138 int 2139 pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n, 2140 struct pf_rule **r, struct pf_rule **a, int *match) 2141 { 2142 struct pf_anchor_stackframe *f; 2143 int quick = 0; 2144 2145 do { 2146 if (*depth <= 0) 2147 break; 2148 f = pf_anchor_stack + *depth - 1; 2149 if (f->parent != NULL && f->child != NULL) { 2150 if (f->child->match || 2151 (match != NULL && *match)) { 2152 f->r->anchor->match = 1; 2153 *match = 0; 2154 } 2155 f->child = RB_NEXT(pf_anchor_node, f->parent, f->child); 2156 if (f->child != NULL) { 2157 *rs = &f->child->ruleset; 2158 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr); 2159 if (*r == NULL) 2160 continue; 2161 else 2162 break; 2163 } 2164 } 2165 (*depth)--; 2166 if (*depth == 0 && a != NULL) 2167 *a = NULL; 2168 *rs = f->rs; 2169 if (f->r->anchor->match || (match != NULL && *match)) 2170 quick = f->r->quick; 2171 *r = TAILQ_NEXT(f->r, entries); 2172 } while (*r == NULL); 2173 2174 return (quick); 2175 } 2176 2177 #ifdef INET6 2178 void 2179 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr, 2180 struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af) 2181 { 2182 switch (af) { 2183 #ifdef INET 2184 case AF_INET: 2185 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) | 2186 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]); 2187 break; 2188 #endif /* INET */ 2189 case AF_INET6: 2190 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) | 2191 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]); 2192 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) | 2193 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]); 2194 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) | 2195 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]); 2196 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) | 2197 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]); 2198 break; 2199 } 2200 } 2201 2202 void 2203 pf_addr_inc(struct pf_addr *addr, sa_family_t af) 2204 { 2205 switch (af) { 2206 #ifdef INET 2207 case AF_INET: 2208 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1); 2209 break; 2210 #endif /* INET */ 2211 case AF_INET6: 2212 if (addr->addr32[3] == 0xffffffff) { 2213 addr->addr32[3] = 0; 2214 if (addr->addr32[2] == 0xffffffff) { 2215 addr->addr32[2] = 0; 2216 if (addr->addr32[1] == 0xffffffff) { 2217 addr->addr32[1] = 0; 2218 addr->addr32[0] = 2219 htonl(ntohl(addr->addr32[0]) + 1); 2220 } else 2221 addr->addr32[1] = 2222 htonl(ntohl(addr->addr32[1]) + 1); 2223 } else 2224 addr->addr32[2] = 2225 htonl(ntohl(addr->addr32[2]) + 1); 2226 } else 2227 addr->addr32[3] = 2228 htonl(ntohl(addr->addr32[3]) + 1); 2229 break; 2230 } 2231 } 2232 #endif /* INET6 */ 2233 2234 #define mix(a,b,c) \ 2235 do { \ 2236 a -= b; a -= c; a ^= (c >> 13); \ 2237 b -= c; b -= a; b ^= (a << 8); \ 2238 c -= a; c -= b; c ^= (b >> 13); \ 2239 a -= b; a -= c; a ^= (c >> 12); \ 2240 b -= c; b -= a; b ^= (a << 16); \ 2241 c -= a; c -= b; c ^= (b >> 5); \ 2242 a -= b; a -= c; a ^= (c >> 3); \ 2243 b -= c; b -= a; b ^= (a << 10); \ 2244 c -= a; c -= b; c ^= (b >> 15); \ 2245 } while (0) 2246 2247 /* 2248 * hash function based on bridge_hash in if_bridge.c 2249 */ 2250 void 2251 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash, 2252 struct pf_poolhashkey *key, sa_family_t af) 2253 { 2254 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0]; 2255 2256 switch (af) { 2257 #ifdef INET 2258 case AF_INET: 2259 a += inaddr->addr32[0]; 2260 b += key->key32[1]; 2261 mix(a, b, c); 2262 hash->addr32[0] = c + key->key32[2]; 2263 break; 2264 #endif /* INET */ 2265 #ifdef INET6 2266 case AF_INET6: 2267 a += inaddr->addr32[0]; 2268 b += inaddr->addr32[2]; 2269 mix(a, b, c); 2270 hash->addr32[0] = c; 2271 a += inaddr->addr32[1]; 2272 b += inaddr->addr32[3]; 2273 c += key->key32[1]; 2274 mix(a, b, c); 2275 hash->addr32[1] = c; 2276 a += inaddr->addr32[2]; 2277 b += inaddr->addr32[1]; 2278 c += key->key32[2]; 2279 mix(a, b, c); 2280 hash->addr32[2] = c; 2281 a += inaddr->addr32[3]; 2282 b += inaddr->addr32[0]; 2283 c += key->key32[3]; 2284 mix(a, b, c); 2285 hash->addr32[3] = c; 2286 break; 2287 #endif /* INET6 */ 2288 } 2289 } 2290 2291 int 2292 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr, 2293 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn) 2294 { 2295 unsigned char hash[16]; 2296 struct pf_pool *rpool = &r->rpool; 2297 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr; 2298 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask; 2299 struct pf_pooladdr *acur = rpool->cur; 2300 struct pf_src_node k; 2301 2302 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR && 2303 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { 2304 k.af = af; 2305 PF_ACPY(&k.addr, saddr, af); 2306 if (r->rule_flag & PFRULE_RULESRCTRACK || 2307 r->rpool.opts & PF_POOL_STICKYADDR) 2308 k.rule.ptr = r; 2309 else 2310 k.rule.ptr = NULL; 2311 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++; 2312 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k); 2313 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) { 2314 PF_ACPY(naddr, &(*sn)->raddr, af); 2315 if (pf_status.debug >= PF_DEBUG_MISC) { 2316 kprintf("pf_map_addr: src tracking maps "); 2317 pf_print_host(&k.addr, 0, af); 2318 kprintf(" to "); 2319 pf_print_host(naddr, 0, af); 2320 kprintf("\n"); 2321 } 2322 return (0); 2323 } 2324 } 2325 2326 if (rpool->cur->addr.type == PF_ADDR_NOROUTE) 2327 return (1); 2328 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2329 switch (af) { 2330 #ifdef INET 2331 case AF_INET: 2332 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 && 2333 (rpool->opts & PF_POOL_TYPEMASK) != 2334 PF_POOL_ROUNDROBIN) 2335 return (1); 2336 raddr = &rpool->cur->addr.p.dyn->pfid_addr4; 2337 rmask = &rpool->cur->addr.p.dyn->pfid_mask4; 2338 break; 2339 #endif /* INET */ 2340 #ifdef INET6 2341 case AF_INET6: 2342 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 && 2343 (rpool->opts & PF_POOL_TYPEMASK) != 2344 PF_POOL_ROUNDROBIN) 2345 return (1); 2346 raddr = &rpool->cur->addr.p.dyn->pfid_addr6; 2347 rmask = &rpool->cur->addr.p.dyn->pfid_mask6; 2348 break; 2349 #endif /* INET6 */ 2350 } 2351 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2352 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN) 2353 return (1); /* unsupported */ 2354 } else { 2355 raddr = &rpool->cur->addr.v.a.addr; 2356 rmask = &rpool->cur->addr.v.a.mask; 2357 } 2358 2359 switch (rpool->opts & PF_POOL_TYPEMASK) { 2360 case PF_POOL_NONE: 2361 PF_ACPY(naddr, raddr, af); 2362 break; 2363 case PF_POOL_BITMASK: 2364 PF_POOLMASK(naddr, raddr, rmask, saddr, af); 2365 break; 2366 case PF_POOL_RANDOM: 2367 if (init_addr != NULL && PF_AZERO(init_addr, af)) { 2368 switch (af) { 2369 #ifdef INET 2370 case AF_INET: 2371 rpool->counter.addr32[0] = htonl(karc4random()); 2372 break; 2373 #endif /* INET */ 2374 #ifdef INET6 2375 case AF_INET6: 2376 if (rmask->addr32[3] != 0xffffffff) 2377 rpool->counter.addr32[3] = 2378 htonl(karc4random()); 2379 else 2380 break; 2381 if (rmask->addr32[2] != 0xffffffff) 2382 rpool->counter.addr32[2] = 2383 htonl(karc4random()); 2384 else 2385 break; 2386 if (rmask->addr32[1] != 0xffffffff) 2387 rpool->counter.addr32[1] = 2388 htonl(karc4random()); 2389 else 2390 break; 2391 if (rmask->addr32[0] != 0xffffffff) 2392 rpool->counter.addr32[0] = 2393 htonl(karc4random()); 2394 break; 2395 #endif /* INET6 */ 2396 } 2397 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af); 2398 PF_ACPY(init_addr, naddr, af); 2399 2400 } else { 2401 PF_AINC(&rpool->counter, af); 2402 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af); 2403 } 2404 break; 2405 case PF_POOL_SRCHASH: 2406 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af); 2407 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af); 2408 break; 2409 case PF_POOL_ROUNDROBIN: 2410 if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2411 if (!pfr_pool_get(rpool->cur->addr.p.tbl, 2412 &rpool->tblidx, &rpool->counter, 2413 &raddr, &rmask, af)) 2414 goto get_addr; 2415 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2416 if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, 2417 &rpool->tblidx, &rpool->counter, 2418 &raddr, &rmask, af)) 2419 goto get_addr; 2420 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af)) 2421 goto get_addr; 2422 2423 try_next: 2424 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL) 2425 rpool->cur = TAILQ_FIRST(&rpool->list); 2426 if (rpool->cur->addr.type == PF_ADDR_TABLE) { 2427 rpool->tblidx = -1; 2428 if (pfr_pool_get(rpool->cur->addr.p.tbl, 2429 &rpool->tblidx, &rpool->counter, 2430 &raddr, &rmask, af)) { 2431 /* table contains no address of type 'af' */ 2432 if (rpool->cur != acur) 2433 goto try_next; 2434 return (1); 2435 } 2436 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 2437 rpool->tblidx = -1; 2438 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, 2439 &rpool->tblidx, &rpool->counter, 2440 &raddr, &rmask, af)) { 2441 /* table contains no address of type 'af' */ 2442 if (rpool->cur != acur) 2443 goto try_next; 2444 return (1); 2445 } 2446 } else { 2447 raddr = &rpool->cur->addr.v.a.addr; 2448 rmask = &rpool->cur->addr.v.a.mask; 2449 PF_ACPY(&rpool->counter, raddr, af); 2450 } 2451 2452 get_addr: 2453 PF_ACPY(naddr, &rpool->counter, af); 2454 if (init_addr != NULL && PF_AZERO(init_addr, af)) 2455 PF_ACPY(init_addr, naddr, af); 2456 PF_AINC(&rpool->counter, af); 2457 break; 2458 } 2459 if (*sn != NULL) 2460 PF_ACPY(&(*sn)->raddr, naddr, af); 2461 2462 if (pf_status.debug >= PF_DEBUG_MISC && 2463 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { 2464 kprintf("pf_map_addr: selected address "); 2465 pf_print_host(naddr, 0, af); 2466 kprintf("\n"); 2467 } 2468 2469 return (0); 2470 } 2471 2472 int 2473 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r, 2474 struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport, 2475 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high, 2476 struct pf_src_node **sn) 2477 { 2478 struct pf_state_key_cmp key; 2479 struct pf_addr init_addr; 2480 u_int16_t cut; 2481 2482 bzero(&init_addr, sizeof(init_addr)); 2483 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn)) 2484 return (1); 2485 2486 if (proto == IPPROTO_ICMP) { 2487 low = 1; 2488 high = 65535; 2489 } 2490 2491 do { 2492 key.af = af; 2493 key.proto = proto; 2494 PF_ACPY(&key.addr[1], daddr, key.af); 2495 PF_ACPY(&key.addr[0], naddr, key.af); 2496 key.port[1] = dport; 2497 2498 /* 2499 * port search; start random, step; 2500 * similar 2 portloop in in_pcbbind 2501 */ 2502 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP || 2503 proto == IPPROTO_ICMP)) { 2504 key.port[0] = dport; 2505 if (pf_find_state_all(&key, PF_IN, NULL) == NULL) 2506 return (0); 2507 } else if (low == 0 && high == 0) { 2508 key.port[0] = *nport; 2509 if (pf_find_state_all(&key, PF_IN, NULL) == NULL) 2510 return (0); 2511 } else if (low == high) { 2512 key.port[0] = htons(low); 2513 if (pf_find_state_all(&key, PF_IN, NULL) == NULL) { 2514 *nport = htons(low); 2515 return (0); 2516 } 2517 } else { 2518 u_int16_t tmp; 2519 2520 if (low > high) { 2521 tmp = low; 2522 low = high; 2523 high = tmp; 2524 } 2525 /* low < high */ 2526 cut = htonl(karc4random()) % (1 + high - low) + low; 2527 /* low <= cut <= high */ 2528 for (tmp = cut; tmp <= high; ++(tmp)) { 2529 key.port[0] = htons(tmp); 2530 if (pf_find_state_all(&key, PF_IN, NULL) == 2531 NULL && !in_baddynamic(tmp, proto)) { 2532 *nport = htons(tmp); 2533 return (0); 2534 } 2535 } 2536 for (tmp = cut - 1; tmp >= low; --(tmp)) { 2537 key.port[0] = htons(tmp); 2538 if (pf_find_state_all(&key, PF_IN, NULL) == 2539 NULL && !in_baddynamic(tmp, proto)) { 2540 *nport = htons(tmp); 2541 return (0); 2542 } 2543 } 2544 } 2545 2546 switch (r->rpool.opts & PF_POOL_TYPEMASK) { 2547 case PF_POOL_RANDOM: 2548 case PF_POOL_ROUNDROBIN: 2549 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn)) 2550 return (1); 2551 break; 2552 case PF_POOL_NONE: 2553 case PF_POOL_SRCHASH: 2554 case PF_POOL_BITMASK: 2555 default: 2556 return (1); 2557 } 2558 } while (! PF_AEQ(&init_addr, naddr, af) ); 2559 return (1); /* none available */ 2560 } 2561 2562 struct pf_rule * 2563 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off, 2564 int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport, 2565 struct pf_addr *daddr, u_int16_t dport, int rs_num) 2566 { 2567 struct pf_rule *r, *rm = NULL; 2568 struct pf_ruleset *ruleset = NULL; 2569 int tag = -1; 2570 int rtableid = -1; 2571 int asd = 0; 2572 2573 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr); 2574 while (r && rm == NULL) { 2575 struct pf_rule_addr *src = NULL, *dst = NULL; 2576 struct pf_addr_wrap *xdst = NULL; 2577 2578 if (r->action == PF_BINAT && direction == PF_IN) { 2579 src = &r->dst; 2580 if (r->rpool.cur != NULL) 2581 xdst = &r->rpool.cur->addr; 2582 } else { 2583 src = &r->src; 2584 dst = &r->dst; 2585 } 2586 2587 r->evaluations++; 2588 if (pfi_kif_match(r->kif, kif) == r->ifnot) 2589 r = r->skip[PF_SKIP_IFP].ptr; 2590 else if (r->direction && r->direction != direction) 2591 r = r->skip[PF_SKIP_DIR].ptr; 2592 else if (r->af && r->af != pd->af) 2593 r = r->skip[PF_SKIP_AF].ptr; 2594 else if (r->proto && r->proto != pd->proto) 2595 r = r->skip[PF_SKIP_PROTO].ptr; 2596 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af, 2597 src->neg, kif)) 2598 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR : 2599 PF_SKIP_DST_ADDR].ptr; 2600 else if (src->port_op && !pf_match_port(src->port_op, 2601 src->port[0], src->port[1], sport)) 2602 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT : 2603 PF_SKIP_DST_PORT].ptr; 2604 else if (dst != NULL && 2605 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL)) 2606 r = r->skip[PF_SKIP_DST_ADDR].ptr; 2607 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af, 2608 0, NULL)) 2609 r = TAILQ_NEXT(r, entries); 2610 else if (dst != NULL && dst->port_op && 2611 !pf_match_port(dst->port_op, dst->port[0], 2612 dst->port[1], dport)) 2613 r = r->skip[PF_SKIP_DST_PORT].ptr; 2614 else if (r->match_tag && !pf_match_tag(m, r, &tag)) 2615 r = TAILQ_NEXT(r, entries); 2616 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto != 2617 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m, 2618 off, pd->hdr.tcp), r->os_fingerprint))) 2619 r = TAILQ_NEXT(r, entries); 2620 else { 2621 if (r->tag) 2622 tag = r->tag; 2623 if (r->rtableid >= 0) 2624 rtableid = r->rtableid; 2625 if (r->anchor == NULL) { 2626 rm = r; 2627 } else 2628 pf_step_into_anchor(&asd, &ruleset, rs_num, 2629 &r, NULL, NULL); 2630 } 2631 if (r == NULL) 2632 pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r, 2633 NULL, NULL); 2634 } 2635 if (pf_tag_packet(m, tag, rtableid)) 2636 return (NULL); 2637 if (rm != NULL && (rm->action == PF_NONAT || 2638 rm->action == PF_NORDR || rm->action == PF_NOBINAT)) 2639 return (NULL); 2640 return (rm); 2641 } 2642 2643 struct pf_rule * 2644 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction, 2645 struct pfi_kif *kif, struct pf_src_node **sn, 2646 struct pf_state_key **skw, struct pf_state_key **sks, 2647 struct pf_state_key **skp, struct pf_state_key **nkp, 2648 struct pf_addr *saddr, struct pf_addr *daddr, 2649 u_int16_t sport, u_int16_t dport) 2650 { 2651 struct pf_rule *r = NULL; 2652 2653 2654 if (direction == PF_OUT) { 2655 r = pf_match_translation(pd, m, off, direction, kif, saddr, 2656 sport, daddr, dport, PF_RULESET_BINAT); 2657 if (r == NULL) 2658 r = pf_match_translation(pd, m, off, direction, kif, 2659 saddr, sport, daddr, dport, PF_RULESET_NAT); 2660 } else { 2661 r = pf_match_translation(pd, m, off, direction, kif, saddr, 2662 sport, daddr, dport, PF_RULESET_RDR); 2663 if (r == NULL) 2664 r = pf_match_translation(pd, m, off, direction, kif, 2665 saddr, sport, daddr, dport, PF_RULESET_BINAT); 2666 } 2667 2668 if (r != NULL) { 2669 struct pf_addr *naddr; 2670 u_int16_t *nport; 2671 2672 if (pf_state_key_setup(pd, r, skw, sks, skp, nkp, 2673 saddr, daddr, sport, dport)) 2674 return r; 2675 2676 /* XXX We only modify one side for now. */ 2677 naddr = &(*nkp)->addr[1]; 2678 nport = &(*nkp)->port[1]; 2679 2680 /* 2681 * NOTE: Currently all translations will clear 2682 * BRIDGE_MBUF_TAGGED, telling the bridge to 2683 * ignore the original input encapsulation. 2684 */ 2685 switch (r->action) { 2686 case PF_NONAT: 2687 case PF_NOBINAT: 2688 case PF_NORDR: 2689 return (NULL); 2690 case PF_NAT: 2691 m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED; 2692 if (pf_get_sport(pd->af, pd->proto, r, saddr, 2693 daddr, dport, naddr, nport, r->rpool.proxy_port[0], 2694 r->rpool.proxy_port[1], sn)) { 2695 DPFPRINTF(PF_DEBUG_MISC, 2696 ("pf: NAT proxy port allocation " 2697 "(%u-%u) failed\n", 2698 r->rpool.proxy_port[0], 2699 r->rpool.proxy_port[1])); 2700 return (NULL); 2701 } 2702 break; 2703 case PF_BINAT: 2704 m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED; 2705 switch (direction) { 2706 case PF_OUT: 2707 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){ 2708 switch (pd->af) { 2709 #ifdef INET 2710 case AF_INET: 2711 if (r->rpool.cur->addr.p.dyn-> 2712 pfid_acnt4 < 1) 2713 return (NULL); 2714 PF_POOLMASK(naddr, 2715 &r->rpool.cur->addr.p.dyn-> 2716 pfid_addr4, 2717 &r->rpool.cur->addr.p.dyn-> 2718 pfid_mask4, 2719 saddr, AF_INET); 2720 break; 2721 #endif /* INET */ 2722 #ifdef INET6 2723 case AF_INET6: 2724 if (r->rpool.cur->addr.p.dyn-> 2725 pfid_acnt6 < 1) 2726 return (NULL); 2727 PF_POOLMASK(naddr, 2728 &r->rpool.cur->addr.p.dyn-> 2729 pfid_addr6, 2730 &r->rpool.cur->addr.p.dyn-> 2731 pfid_mask6, 2732 saddr, AF_INET6); 2733 break; 2734 #endif /* INET6 */ 2735 } 2736 } else 2737 PF_POOLMASK(naddr, 2738 &r->rpool.cur->addr.v.a.addr, 2739 &r->rpool.cur->addr.v.a.mask, 2740 saddr, pd->af); 2741 break; 2742 case PF_IN: 2743 if (r->src.addr.type == PF_ADDR_DYNIFTL) { 2744 switch (pd->af) { 2745 #ifdef INET 2746 case AF_INET: 2747 if (r->src.addr.p.dyn-> 2748 pfid_acnt4 < 1) 2749 return (NULL); 2750 PF_POOLMASK(naddr, 2751 &r->src.addr.p.dyn-> 2752 pfid_addr4, 2753 &r->src.addr.p.dyn-> 2754 pfid_mask4, 2755 daddr, AF_INET); 2756 break; 2757 #endif /* INET */ 2758 #ifdef INET6 2759 case AF_INET6: 2760 if (r->src.addr.p.dyn-> 2761 pfid_acnt6 < 1) 2762 return (NULL); 2763 PF_POOLMASK(naddr, 2764 &r->src.addr.p.dyn-> 2765 pfid_addr6, 2766 &r->src.addr.p.dyn-> 2767 pfid_mask6, 2768 daddr, AF_INET6); 2769 break; 2770 #endif /* INET6 */ 2771 } 2772 } else 2773 PF_POOLMASK(naddr, 2774 &r->src.addr.v.a.addr, 2775 &r->src.addr.v.a.mask, daddr, 2776 pd->af); 2777 break; 2778 } 2779 break; 2780 case PF_RDR: { 2781 m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED; 2782 if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn)) 2783 return (NULL); 2784 if ((r->rpool.opts & PF_POOL_TYPEMASK) == 2785 PF_POOL_BITMASK) 2786 PF_POOLMASK(naddr, naddr, 2787 &r->rpool.cur->addr.v.a.mask, daddr, 2788 pd->af); 2789 2790 if (r->rpool.proxy_port[1]) { 2791 u_int32_t tmp_nport; 2792 2793 tmp_nport = ((ntohs(dport) - 2794 ntohs(r->dst.port[0])) % 2795 (r->rpool.proxy_port[1] - 2796 r->rpool.proxy_port[0] + 1)) + 2797 r->rpool.proxy_port[0]; 2798 2799 /* wrap around if necessary */ 2800 if (tmp_nport > 65535) 2801 tmp_nport -= 65535; 2802 *nport = htons((u_int16_t)tmp_nport); 2803 } else if (r->rpool.proxy_port[0]) 2804 *nport = htons(r->rpool.proxy_port[0]); 2805 break; 2806 } 2807 default: 2808 return (NULL); 2809 } 2810 } 2811 2812 return (r); 2813 } 2814 2815 struct netmsg_hashlookup { 2816 struct netmsg_base base; 2817 struct inpcb **nm_pinp; 2818 struct inpcbinfo *nm_pcbinfo; 2819 struct pf_addr *nm_saddr; 2820 struct pf_addr *nm_daddr; 2821 uint16_t nm_sport; 2822 uint16_t nm_dport; 2823 sa_family_t nm_af; 2824 }; 2825 2826 #ifdef PF_SOCKET_LOOKUP_DOMSG 2827 static void 2828 in_pcblookup_hash_handler(netmsg_t msg) 2829 { 2830 struct netmsg_hashlookup *rmsg = (struct netmsg_hashlookup *)msg; 2831 2832 if (rmsg->nm_af == AF_INET) 2833 *rmsg->nm_pinp = in_pcblookup_hash(rmsg->nm_pcbinfo, 2834 rmsg->nm_saddr->v4, rmsg->nm_sport, rmsg->nm_daddr->v4, 2835 rmsg->nm_dport, INPLOOKUP_WILDCARD, NULL); 2836 #ifdef INET6 2837 else 2838 *rmsg->nm_pinp = in6_pcblookup_hash(rmsg->nm_pcbinfo, 2839 &rmsg->nm_saddr->v6, rmsg->nm_sport, &rmsg->nm_daddr->v6, 2840 rmsg->nm_dport, INPLOOKUP_WILDCARD, NULL); 2841 #endif /* INET6 */ 2842 lwkt_replymsg(&rmsg->base.lmsg, 0); 2843 } 2844 #endif /* PF_SOCKET_LOOKUP_DOMSG */ 2845 2846 int 2847 pf_socket_lookup(int direction, struct pf_pdesc *pd) 2848 { 2849 struct pf_addr *saddr, *daddr; 2850 u_int16_t sport, dport; 2851 struct inpcbinfo *pi; 2852 struct inpcb *inp; 2853 struct netmsg_hashlookup *msg = NULL; 2854 #ifdef PF_SOCKET_LOOKUP_DOMSG 2855 struct netmsg_hashlookup msg0; 2856 #endif 2857 int pi_cpu = 0; 2858 2859 if (pd == NULL) 2860 return (-1); 2861 pd->lookup.uid = UID_MAX; 2862 pd->lookup.gid = GID_MAX; 2863 pd->lookup.pid = NO_PID; 2864 if (direction == PF_IN) { 2865 saddr = pd->src; 2866 daddr = pd->dst; 2867 } else { 2868 saddr = pd->dst; 2869 daddr = pd->src; 2870 } 2871 switch (pd->proto) { 2872 case IPPROTO_TCP: 2873 if (pd->hdr.tcp == NULL) 2874 return (-1); 2875 sport = pd->hdr.tcp->th_sport; 2876 dport = pd->hdr.tcp->th_dport; 2877 2878 pi_cpu = tcp_addrcpu(saddr->v4.s_addr, sport, daddr->v4.s_addr, dport); 2879 pi = &tcbinfo[pi_cpu]; 2880 /* 2881 * Our netstack runs lockless on MP systems 2882 * (only for TCP connections at the moment). 2883 * 2884 * As we are not allowed to read another CPU's tcbinfo, 2885 * we have to ask that CPU via remote call to search the 2886 * table for us. 2887 * 2888 * Prepare a msg iff data belongs to another CPU. 2889 */ 2890 if (pi_cpu != mycpu->gd_cpuid) { 2891 #ifdef PF_SOCKET_LOOKUP_DOMSG 2892 /* 2893 * NOTE: 2894 * 2895 * Following lwkt_domsg() is dangerous and could 2896 * lockup the network system, e.g. 2897 * 2898 * On 2 CPU system: 2899 * netisr0 domsg to netisr1 (due to lookup) 2900 * netisr1 domsg to netisr0 (due to lookup) 2901 * 2902 * We simply return -1 here, since we are probably 2903 * called before NAT, so the TCP packet should 2904 * already be on the correct CPU. 2905 */ 2906 msg = &msg0; 2907 netmsg_init(&msg->base, NULL, &curthread->td_msgport, 2908 0, in_pcblookup_hash_handler); 2909 msg->nm_pinp = &inp; 2910 msg->nm_pcbinfo = pi; 2911 msg->nm_saddr = saddr; 2912 msg->nm_sport = sport; 2913 msg->nm_daddr = daddr; 2914 msg->nm_dport = dport; 2915 msg->nm_af = pd->af; 2916 #else /* !PF_SOCKET_LOOKUP_DOMSG */ 2917 kprintf("pf_socket_lookup: tcp packet not on the " 2918 "correct cpu %d, cur cpu %d\n", 2919 pi_cpu, mycpuid); 2920 print_backtrace(-1); 2921 return -1; 2922 #endif /* PF_SOCKET_LOOKUP_DOMSG */ 2923 } 2924 break; 2925 case IPPROTO_UDP: 2926 if (pd->hdr.udp == NULL) 2927 return (-1); 2928 sport = pd->hdr.udp->uh_sport; 2929 dport = pd->hdr.udp->uh_dport; 2930 pi = &udbinfo; 2931 break; 2932 default: 2933 return (-1); 2934 } 2935 if (direction != PF_IN) { 2936 u_int16_t p; 2937 2938 p = sport; 2939 sport = dport; 2940 dport = p; 2941 } 2942 switch (pd->af) { 2943 #ifdef INET6 2944 case AF_INET6: 2945 /* 2946 * Query other CPU, second part 2947 * 2948 * msg only gets initialized when: 2949 * 1) packet is TCP 2950 * 2) the info belongs to another CPU 2951 * 2952 * Use some switch/case magic to avoid code duplication. 2953 */ 2954 if (msg == NULL) { 2955 inp = in6_pcblookup_hash(pi, &saddr->v6, sport, 2956 &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL); 2957 2958 if (inp == NULL) 2959 return (-1); 2960 break; 2961 } 2962 /* FALLTHROUGH if SMP and on other CPU */ 2963 #endif /* INET6 */ 2964 case AF_INET: 2965 if (msg != NULL) { 2966 lwkt_domsg(netisr_portfn(pi_cpu), 2967 &msg->base.lmsg, 0); 2968 } else 2969 { 2970 inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4, 2971 dport, INPLOOKUP_WILDCARD, NULL); 2972 } 2973 if (inp == NULL) 2974 return (-1); 2975 break; 2976 2977 default: 2978 return (-1); 2979 } 2980 pd->lookup.uid = inp->inp_socket->so_cred->cr_uid; 2981 pd->lookup.gid = inp->inp_socket->so_cred->cr_groups[0]; 2982 return (1); 2983 } 2984 2985 u_int8_t 2986 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af) 2987 { 2988 int hlen; 2989 u_int8_t hdr[60]; 2990 u_int8_t *opt, optlen; 2991 u_int8_t wscale = 0; 2992 2993 hlen = th_off << 2; /* hlen <= sizeof(hdr) */ 2994 if (hlen <= sizeof(struct tcphdr)) 2995 return (0); 2996 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af)) 2997 return (0); 2998 opt = hdr + sizeof(struct tcphdr); 2999 hlen -= sizeof(struct tcphdr); 3000 while (hlen >= 3) { 3001 switch (*opt) { 3002 case TCPOPT_EOL: 3003 case TCPOPT_NOP: 3004 ++opt; 3005 --hlen; 3006 break; 3007 case TCPOPT_WINDOW: 3008 wscale = opt[2]; 3009 if (wscale > TCP_MAX_WINSHIFT) 3010 wscale = TCP_MAX_WINSHIFT; 3011 wscale |= PF_WSCALE_FLAG; 3012 /* FALLTHROUGH */ 3013 default: 3014 optlen = opt[1]; 3015 if (optlen < 2) 3016 optlen = 2; 3017 hlen -= optlen; 3018 opt += optlen; 3019 break; 3020 } 3021 } 3022 return (wscale); 3023 } 3024 3025 u_int16_t 3026 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af) 3027 { 3028 int hlen; 3029 u_int8_t hdr[60]; 3030 u_int8_t *opt, optlen; 3031 u_int16_t mss = tcp_mssdflt; 3032 3033 hlen = th_off << 2; /* hlen <= sizeof(hdr) */ 3034 if (hlen <= sizeof(struct tcphdr)) 3035 return (0); 3036 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af)) 3037 return (0); 3038 opt = hdr + sizeof(struct tcphdr); 3039 hlen -= sizeof(struct tcphdr); 3040 while (hlen >= TCPOLEN_MAXSEG) { 3041 switch (*opt) { 3042 case TCPOPT_EOL: 3043 case TCPOPT_NOP: 3044 ++opt; 3045 --hlen; 3046 break; 3047 case TCPOPT_MAXSEG: 3048 bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2); 3049 /* FALLTHROUGH */ 3050 default: 3051 optlen = opt[1]; 3052 if (optlen < 2) 3053 optlen = 2; 3054 hlen -= optlen; 3055 opt += optlen; 3056 break; 3057 } 3058 } 3059 return (mss); 3060 } 3061 3062 u_int16_t 3063 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer) 3064 { 3065 #ifdef INET 3066 struct sockaddr_in *dst; 3067 struct route ro; 3068 #endif /* INET */ 3069 #ifdef INET6 3070 struct sockaddr_in6 *dst6; 3071 struct route_in6 ro6; 3072 #endif /* INET6 */ 3073 struct rtentry *rt = NULL; 3074 int hlen = 0; 3075 u_int16_t mss = tcp_mssdflt; 3076 3077 switch (af) { 3078 #ifdef INET 3079 case AF_INET: 3080 hlen = sizeof(struct ip); 3081 bzero(&ro, sizeof(ro)); 3082 dst = (struct sockaddr_in *)&ro.ro_dst; 3083 dst->sin_family = AF_INET; 3084 dst->sin_len = sizeof(*dst); 3085 dst->sin_addr = addr->v4; 3086 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING)); 3087 rt = ro.ro_rt; 3088 break; 3089 #endif /* INET */ 3090 #ifdef INET6 3091 case AF_INET6: 3092 hlen = sizeof(struct ip6_hdr); 3093 bzero(&ro6, sizeof(ro6)); 3094 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst; 3095 dst6->sin6_family = AF_INET6; 3096 dst6->sin6_len = sizeof(*dst6); 3097 dst6->sin6_addr = addr->v6; 3098 rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING)); 3099 rt = ro6.ro_rt; 3100 break; 3101 #endif /* INET6 */ 3102 } 3103 3104 if (rt && rt->rt_ifp) { 3105 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr); 3106 mss = max(tcp_mssdflt, mss); 3107 RTFREE(rt); 3108 } 3109 mss = min(mss, offer); 3110 mss = max(mss, 64); /* sanity - at least max opt space */ 3111 return (mss); 3112 } 3113 3114 void 3115 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr) 3116 { 3117 struct pf_rule *r = s->rule.ptr; 3118 3119 s->rt_kif = NULL; 3120 if (!r->rt || r->rt == PF_FASTROUTE) 3121 return; 3122 switch (s->key[PF_SK_WIRE]->af) { 3123 #ifdef INET 3124 case AF_INET: 3125 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL, 3126 &s->nat_src_node); 3127 s->rt_kif = r->rpool.cur->kif; 3128 break; 3129 #endif /* INET */ 3130 #ifdef INET6 3131 case AF_INET6: 3132 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL, 3133 &s->nat_src_node); 3134 s->rt_kif = r->rpool.cur->kif; 3135 break; 3136 #endif /* INET6 */ 3137 } 3138 } 3139 3140 u_int32_t 3141 pf_tcp_iss(struct pf_pdesc *pd) 3142 { 3143 MD5_CTX ctx; 3144 u_int32_t digest[4]; 3145 3146 if (pf_tcp_secret_init == 0) { 3147 karc4rand(pf_tcp_secret, sizeof(pf_tcp_secret)); 3148 MD5Init(&pf_tcp_secret_ctx); 3149 MD5Update(&pf_tcp_secret_ctx, pf_tcp_secret, 3150 sizeof(pf_tcp_secret)); 3151 pf_tcp_secret_init = 1; 3152 } 3153 ctx = pf_tcp_secret_ctx; 3154 3155 MD5Update(&ctx, (char *)&pd->hdr.tcp->th_sport, sizeof(u_short)); 3156 MD5Update(&ctx, (char *)&pd->hdr.tcp->th_dport, sizeof(u_short)); 3157 if (pd->af == AF_INET6) { 3158 MD5Update(&ctx, (char *)&pd->src->v6, sizeof(struct in6_addr)); 3159 MD5Update(&ctx, (char *)&pd->dst->v6, sizeof(struct in6_addr)); 3160 } else { 3161 MD5Update(&ctx, (char *)&pd->src->v4, sizeof(struct in_addr)); 3162 MD5Update(&ctx, (char *)&pd->dst->v4, sizeof(struct in_addr)); 3163 } 3164 MD5Final((u_char *)digest, &ctx); 3165 pf_tcp_iss_off += 4096; 3166 return (digest[0] + pd->hdr.tcp->th_seq + pf_tcp_iss_off); 3167 } 3168 3169 int 3170 pf_test_rule(struct pf_rule **rm, struct pf_state **sm, int direction, 3171 struct pfi_kif *kif, struct mbuf *m, int off, void *h, 3172 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm, 3173 struct ifqueue *ifq, struct inpcb *inp) 3174 { 3175 struct pf_rule *nr = NULL; 3176 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 3177 sa_family_t af = pd->af; 3178 struct pf_rule *r, *a = NULL; 3179 struct pf_ruleset *ruleset = NULL; 3180 struct pf_src_node *nsn = NULL; 3181 struct tcphdr *th = pd->hdr.tcp; 3182 struct pf_state_key *skw = NULL, *sks = NULL; 3183 struct pf_state_key *sk = NULL, *nk = NULL; 3184 u_short reason; 3185 int rewrite = 0, hdrlen = 0; 3186 int tag = -1, rtableid = -1; 3187 int asd = 0; 3188 int match = 0; 3189 int state_icmp = 0; 3190 u_int16_t sport = 0, dport = 0; 3191 u_int16_t nport = 0, bport = 0; 3192 u_int16_t bproto_sum = 0, bip_sum = 0; 3193 u_int8_t icmptype = 0, icmpcode = 0; 3194 3195 3196 if (direction == PF_IN && pf_check_congestion(ifq)) { 3197 REASON_SET(&reason, PFRES_CONGEST); 3198 return (PF_DROP); 3199 } 3200 3201 if (inp != NULL) 3202 pd->lookup.done = pf_socket_lookup(direction, pd); 3203 else if (debug_pfugidhack) { 3204 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n")); 3205 pd->lookup.done = pf_socket_lookup(direction, pd); 3206 } 3207 3208 switch (pd->proto) { 3209 case IPPROTO_TCP: 3210 sport = th->th_sport; 3211 dport = th->th_dport; 3212 hdrlen = sizeof(*th); 3213 break; 3214 case IPPROTO_UDP: 3215 sport = pd->hdr.udp->uh_sport; 3216 dport = pd->hdr.udp->uh_dport; 3217 hdrlen = sizeof(*pd->hdr.udp); 3218 break; 3219 #ifdef INET 3220 case IPPROTO_ICMP: 3221 if (pd->af != AF_INET) 3222 break; 3223 sport = dport = pd->hdr.icmp->icmp_id; 3224 hdrlen = sizeof(*pd->hdr.icmp); 3225 icmptype = pd->hdr.icmp->icmp_type; 3226 icmpcode = pd->hdr.icmp->icmp_code; 3227 3228 if (icmptype == ICMP_UNREACH || 3229 icmptype == ICMP_SOURCEQUENCH || 3230 icmptype == ICMP_REDIRECT || 3231 icmptype == ICMP_TIMXCEED || 3232 icmptype == ICMP_PARAMPROB) 3233 state_icmp++; 3234 break; 3235 #endif /* INET */ 3236 #ifdef INET6 3237 case IPPROTO_ICMPV6: 3238 if (af != AF_INET6) 3239 break; 3240 sport = dport = pd->hdr.icmp6->icmp6_id; 3241 hdrlen = sizeof(*pd->hdr.icmp6); 3242 icmptype = pd->hdr.icmp6->icmp6_type; 3243 icmpcode = pd->hdr.icmp6->icmp6_code; 3244 3245 if (icmptype == ICMP6_DST_UNREACH || 3246 icmptype == ICMP6_PACKET_TOO_BIG || 3247 icmptype == ICMP6_TIME_EXCEEDED || 3248 icmptype == ICMP6_PARAM_PROB) 3249 state_icmp++; 3250 break; 3251 #endif /* INET6 */ 3252 default: 3253 sport = dport = hdrlen = 0; 3254 break; 3255 } 3256 3257 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 3258 3259 bport = nport = sport; 3260 /* check packet for BINAT/NAT/RDR */ 3261 if ((nr = pf_get_translation(pd, m, off, direction, kif, &nsn, 3262 &skw, &sks, &sk, &nk, saddr, daddr, sport, dport)) != NULL) { 3263 if (nk == NULL || sk == NULL) { 3264 REASON_SET(&reason, PFRES_MEMORY); 3265 goto cleanup; 3266 } 3267 3268 if (pd->ip_sum) 3269 bip_sum = *pd->ip_sum; 3270 3271 switch (pd->proto) { 3272 case IPPROTO_TCP: 3273 bproto_sum = th->th_sum; 3274 pd->proto_sum = &th->th_sum; 3275 3276 if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) || 3277 nk->port[pd->sidx] != sport) { 3278 pf_change_ap(saddr, &th->th_sport, pd->ip_sum, 3279 &th->th_sum, &nk->addr[pd->sidx], 3280 nk->port[pd->sidx], 0, af); 3281 pd->sport = &th->th_sport; 3282 sport = th->th_sport; 3283 } 3284 3285 if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) || 3286 nk->port[pd->didx] != dport) { 3287 pf_change_ap(daddr, &th->th_dport, pd->ip_sum, 3288 &th->th_sum, &nk->addr[pd->didx], 3289 nk->port[pd->didx], 0, af); 3290 dport = th->th_dport; 3291 pd->dport = &th->th_dport; 3292 } 3293 rewrite++; 3294 break; 3295 case IPPROTO_UDP: 3296 bproto_sum = pd->hdr.udp->uh_sum; 3297 pd->proto_sum = &pd->hdr.udp->uh_sum; 3298 3299 if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) || 3300 nk->port[pd->sidx] != sport) { 3301 pf_change_ap(saddr, &pd->hdr.udp->uh_sport, 3302 pd->ip_sum, &pd->hdr.udp->uh_sum, 3303 &nk->addr[pd->sidx], 3304 nk->port[pd->sidx], 1, af); 3305 sport = pd->hdr.udp->uh_sport; 3306 pd->sport = &pd->hdr.udp->uh_sport; 3307 } 3308 3309 if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) || 3310 nk->port[pd->didx] != dport) { 3311 pf_change_ap(daddr, &pd->hdr.udp->uh_dport, 3312 pd->ip_sum, &pd->hdr.udp->uh_sum, 3313 &nk->addr[pd->didx], 3314 nk->port[pd->didx], 1, af); 3315 dport = pd->hdr.udp->uh_dport; 3316 pd->dport = &pd->hdr.udp->uh_dport; 3317 } 3318 rewrite++; 3319 break; 3320 #ifdef INET 3321 case IPPROTO_ICMP: 3322 nk->port[0] = nk->port[1]; 3323 if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET)) 3324 pf_change_a(&saddr->v4.s_addr, pd->ip_sum, 3325 nk->addr[pd->sidx].v4.s_addr, 0); 3326 3327 if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET)) 3328 pf_change_a(&daddr->v4.s_addr, pd->ip_sum, 3329 nk->addr[pd->didx].v4.s_addr, 0); 3330 3331 if (nk->port[1] != pd->hdr.icmp->icmp_id) { 3332 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup( 3333 pd->hdr.icmp->icmp_cksum, sport, 3334 nk->port[1], 0); 3335 pd->hdr.icmp->icmp_id = nk->port[1]; 3336 pd->sport = &pd->hdr.icmp->icmp_id; 3337 } 3338 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp); 3339 break; 3340 #endif /* INET */ 3341 #ifdef INET6 3342 case IPPROTO_ICMPV6: 3343 nk->port[0] = nk->port[1]; 3344 if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET6)) 3345 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum, 3346 &nk->addr[pd->sidx], 0); 3347 3348 if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET6)) 3349 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum, 3350 &nk->addr[pd->didx], 0); 3351 rewrite++; 3352 break; 3353 #endif /* INET */ 3354 default: 3355 switch (af) { 3356 #ifdef INET 3357 case AF_INET: 3358 if (PF_ANEQ(saddr, 3359 &nk->addr[pd->sidx], AF_INET)) 3360 pf_change_a(&saddr->v4.s_addr, 3361 pd->ip_sum, 3362 nk->addr[pd->sidx].v4.s_addr, 0); 3363 3364 if (PF_ANEQ(daddr, 3365 &nk->addr[pd->didx], AF_INET)) 3366 pf_change_a(&daddr->v4.s_addr, 3367 pd->ip_sum, 3368 nk->addr[pd->didx].v4.s_addr, 0); 3369 break; 3370 #endif /* INET */ 3371 #ifdef INET6 3372 case AF_INET6: 3373 if (PF_ANEQ(saddr, 3374 &nk->addr[pd->sidx], AF_INET6)) 3375 PF_ACPY(saddr, &nk->addr[pd->sidx], af); 3376 3377 if (PF_ANEQ(daddr, 3378 &nk->addr[pd->didx], AF_INET6)) 3379 PF_ACPY(saddr, &nk->addr[pd->didx], af); 3380 break; 3381 #endif /* INET */ 3382 } 3383 break; 3384 } 3385 if (nr->natpass) 3386 r = NULL; 3387 pd->nat_rule = nr; 3388 } 3389 3390 while (r != NULL) { 3391 r->evaluations++; 3392 if (pfi_kif_match(r->kif, kif) == r->ifnot) 3393 r = r->skip[PF_SKIP_IFP].ptr; 3394 else if (r->direction && r->direction != direction) 3395 r = r->skip[PF_SKIP_DIR].ptr; 3396 else if (r->af && r->af != af) 3397 r = r->skip[PF_SKIP_AF].ptr; 3398 else if (r->proto && r->proto != pd->proto) 3399 r = r->skip[PF_SKIP_PROTO].ptr; 3400 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, 3401 r->src.neg, kif)) 3402 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 3403 /* tcp/udp only. port_op always 0 in other cases */ 3404 else if (r->src.port_op && !pf_match_port(r->src.port_op, 3405 r->src.port[0], r->src.port[1], sport)) 3406 r = r->skip[PF_SKIP_SRC_PORT].ptr; 3407 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, 3408 r->dst.neg, NULL)) 3409 r = r->skip[PF_SKIP_DST_ADDR].ptr; 3410 /* tcp/udp only. port_op always 0 in other cases */ 3411 else if (r->dst.port_op && !pf_match_port(r->dst.port_op, 3412 r->dst.port[0], r->dst.port[1], dport)) 3413 r = r->skip[PF_SKIP_DST_PORT].ptr; 3414 /* icmp only. type always 0 in other cases */ 3415 else if (r->type && r->type != icmptype + 1) 3416 r = TAILQ_NEXT(r, entries); 3417 /* icmp only. type always 0 in other cases */ 3418 else if (r->code && r->code != icmpcode + 1) 3419 r = TAILQ_NEXT(r, entries); 3420 else if (r->tos && !(r->tos == pd->tos)) 3421 r = TAILQ_NEXT(r, entries); 3422 else if (r->rule_flag & PFRULE_FRAGMENT) 3423 r = TAILQ_NEXT(r, entries); 3424 else if (pd->proto == IPPROTO_TCP && 3425 (r->flagset & th->th_flags) != r->flags) 3426 r = TAILQ_NEXT(r, entries); 3427 /* tcp/udp only. uid.op always 0 in other cases */ 3428 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done = 3429 pf_socket_lookup(direction, pd), 1)) && 3430 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1], 3431 pd->lookup.uid)) 3432 r = TAILQ_NEXT(r, entries); 3433 /* tcp/udp only. gid.op always 0 in other cases */ 3434 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done = 3435 pf_socket_lookup(direction, pd), 1)) && 3436 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1], 3437 pd->lookup.gid)) 3438 r = TAILQ_NEXT(r, entries); 3439 else if (r->prob && 3440 r->prob <= karc4random()) 3441 r = TAILQ_NEXT(r, entries); 3442 else if (r->match_tag && !pf_match_tag(m, r, &tag)) 3443 r = TAILQ_NEXT(r, entries); 3444 else if (r->os_fingerprint != PF_OSFP_ANY && 3445 (pd->proto != IPPROTO_TCP || !pf_osfp_match( 3446 pf_osfp_fingerprint(pd, m, off, th), 3447 r->os_fingerprint))) 3448 r = TAILQ_NEXT(r, entries); 3449 else { 3450 if (r->tag) 3451 tag = r->tag; 3452 if (r->rtableid >= 0) 3453 rtableid = r->rtableid; 3454 if (r->anchor == NULL) { 3455 match = 1; 3456 *rm = r; 3457 *am = a; 3458 *rsm = ruleset; 3459 if ((*rm)->quick) 3460 break; 3461 r = TAILQ_NEXT(r, entries); 3462 } else 3463 pf_step_into_anchor(&asd, &ruleset, 3464 PF_RULESET_FILTER, &r, &a, &match); 3465 } 3466 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 3467 PF_RULESET_FILTER, &r, &a, &match)) 3468 break; 3469 } 3470 r = *rm; 3471 a = *am; 3472 ruleset = *rsm; 3473 3474 REASON_SET(&reason, PFRES_MATCH); 3475 3476 if (r->log || (nr != NULL && nr->log)) { 3477 if (rewrite) 3478 m_copyback(m, off, hdrlen, pd->hdr.any); 3479 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr, 3480 a, ruleset, pd); 3481 } 3482 3483 if ((r->action == PF_DROP) && 3484 ((r->rule_flag & PFRULE_RETURNRST) || 3485 (r->rule_flag & PFRULE_RETURNICMP) || 3486 (r->rule_flag & PFRULE_RETURN))) { 3487 /* undo NAT changes, if they have taken place */ 3488 if (nr != NULL) { 3489 PF_ACPY(saddr, &sk->addr[pd->sidx], af); 3490 PF_ACPY(daddr, &sk->addr[pd->didx], af); 3491 if (pd->sport) 3492 *pd->sport = sk->port[pd->sidx]; 3493 if (pd->dport) 3494 *pd->dport = sk->port[pd->didx]; 3495 if (pd->proto_sum) 3496 *pd->proto_sum = bproto_sum; 3497 if (pd->ip_sum) 3498 *pd->ip_sum = bip_sum; 3499 m_copyback(m, off, hdrlen, pd->hdr.any); 3500 } 3501 if (pd->proto == IPPROTO_TCP && 3502 ((r->rule_flag & PFRULE_RETURNRST) || 3503 (r->rule_flag & PFRULE_RETURN)) && 3504 !(th->th_flags & TH_RST)) { 3505 u_int32_t ack = ntohl(th->th_seq) + pd->p_len; 3506 int len = 0; 3507 struct ip *h4; 3508 #ifdef INET6 3509 struct ip6_hdr *h6; 3510 #endif 3511 switch (af) { 3512 case AF_INET: 3513 h4 = mtod(m, struct ip *); 3514 len = h4->ip_len - off; 3515 break; 3516 #ifdef INET6 3517 case AF_INET6: 3518 h6 = mtod(m, struct ip6_hdr *); 3519 len = h6->ip6_plen - (off - sizeof(*h6)); 3520 break; 3521 #endif 3522 } 3523 3524 if (pf_check_proto_cksum(m, off, len, IPPROTO_TCP, af)) 3525 REASON_SET(&reason, PFRES_PROTCKSUM); 3526 else { 3527 if (th->th_flags & TH_SYN) 3528 ack++; 3529 if (th->th_flags & TH_FIN) 3530 ack++; 3531 pf_send_tcp(r, af, pd->dst, 3532 pd->src, th->th_dport, th->th_sport, 3533 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0, 3534 r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp); 3535 } 3536 } else if (pd->proto != IPPROTO_ICMP && af == AF_INET && 3537 r->return_icmp) 3538 pf_send_icmp(m, r->return_icmp >> 8, 3539 r->return_icmp & 255, af, r); 3540 else if (pd->proto != IPPROTO_ICMPV6 && af == AF_INET6 && 3541 r->return_icmp6) 3542 pf_send_icmp(m, r->return_icmp6 >> 8, 3543 r->return_icmp6 & 255, af, r); 3544 } 3545 3546 if (r->action == PF_DROP) 3547 goto cleanup; 3548 3549 if (pf_tag_packet(m, tag, rtableid)) { 3550 REASON_SET(&reason, PFRES_MEMORY); 3551 goto cleanup; 3552 } 3553 3554 if (!state_icmp && (r->keep_state || nr != NULL || 3555 (pd->flags & PFDESC_TCP_NORM))) { 3556 int action; 3557 action = pf_create_state(r, nr, a, pd, nsn, skw, sks, nk, sk, m, 3558 off, sport, dport, &rewrite, kif, sm, tag, bproto_sum, 3559 bip_sum, hdrlen); 3560 if (action != PF_PASS) 3561 return (action); 3562 } 3563 3564 /* copy back packet headers if we performed NAT operations */ 3565 if (rewrite) 3566 m_copyback(m, off, hdrlen, pd->hdr.any); 3567 3568 return (PF_PASS); 3569 3570 cleanup: 3571 if (sk != NULL) 3572 kfree(sk, M_PFSTATEKEYPL); 3573 if (nk != NULL) 3574 kfree(nk, M_PFSTATEKEYPL); 3575 return (PF_DROP); 3576 } 3577 3578 static __inline int 3579 pf_create_state(struct pf_rule *r, struct pf_rule *nr, struct pf_rule *a, 3580 struct pf_pdesc *pd, struct pf_src_node *nsn, struct pf_state_key *skw, 3581 struct pf_state_key *sks, struct pf_state_key *nk, struct pf_state_key *sk, 3582 struct mbuf *m, int off, u_int16_t sport, u_int16_t dport, int *rewrite, 3583 struct pfi_kif *kif, struct pf_state **sm, int tag, u_int16_t bproto_sum, 3584 u_int16_t bip_sum, int hdrlen) 3585 { 3586 struct pf_state *s = NULL; 3587 struct pf_src_node *sn = NULL; 3588 struct tcphdr *th = pd->hdr.tcp; 3589 u_int16_t mss = tcp_mssdflt; 3590 u_short reason; 3591 3592 /* check maximums */ 3593 if (r->max_states && (r->states_cur >= r->max_states)) { 3594 pf_status.lcounters[LCNT_STATES]++; 3595 REASON_SET(&reason, PFRES_MAXSTATES); 3596 return (PF_DROP); 3597 } 3598 /* src node for filter rule */ 3599 if ((r->rule_flag & PFRULE_SRCTRACK || 3600 r->rpool.opts & PF_POOL_STICKYADDR) && 3601 pf_insert_src_node(&sn, r, pd->src, pd->af) != 0) { 3602 REASON_SET(&reason, PFRES_SRCLIMIT); 3603 goto csfailed; 3604 } 3605 /* src node for translation rule */ 3606 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && 3607 pf_insert_src_node(&nsn, nr, &sk->addr[pd->sidx], pd->af)) { 3608 REASON_SET(&reason, PFRES_SRCLIMIT); 3609 goto csfailed; 3610 } 3611 s = kmalloc(sizeof(struct pf_state), M_PFSTATEPL, M_NOWAIT|M_ZERO); 3612 if (s == NULL) { 3613 REASON_SET(&reason, PFRES_MEMORY); 3614 goto csfailed; 3615 } 3616 s->id = 0; /* XXX Do we really need that? not in OpenBSD */ 3617 s->creatorid = 0; 3618 s->rule.ptr = r; 3619 s->nat_rule.ptr = nr; 3620 s->anchor.ptr = a; 3621 STATE_INC_COUNTERS(s); 3622 if (r->allow_opts) 3623 s->state_flags |= PFSTATE_ALLOWOPTS; 3624 if (r->rule_flag & PFRULE_STATESLOPPY) 3625 s->state_flags |= PFSTATE_SLOPPY; 3626 s->log = r->log & PF_LOG_ALL; 3627 if (nr != NULL) 3628 s->log |= nr->log & PF_LOG_ALL; 3629 switch (pd->proto) { 3630 case IPPROTO_TCP: 3631 s->src.seqlo = ntohl(th->th_seq); 3632 s->src.seqhi = s->src.seqlo + pd->p_len + 1; 3633 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN && 3634 r->keep_state == PF_STATE_MODULATE) { 3635 /* Generate sequence number modulator */ 3636 if ((s->src.seqdiff = pf_tcp_iss(pd) - s->src.seqlo) == 3637 0) 3638 s->src.seqdiff = 1; 3639 pf_change_a(&th->th_seq, &th->th_sum, 3640 htonl(s->src.seqlo + s->src.seqdiff), 0); 3641 *rewrite = 1; 3642 } else 3643 s->src.seqdiff = 0; 3644 if (th->th_flags & TH_SYN) { 3645 s->src.seqhi++; 3646 s->src.wscale = pf_get_wscale(m, off, 3647 th->th_off, pd->af); 3648 } 3649 s->src.max_win = MAX(ntohs(th->th_win), 1); 3650 if (s->src.wscale & PF_WSCALE_MASK) { 3651 /* Remove scale factor from initial window */ 3652 int win = s->src.max_win; 3653 win += 1 << (s->src.wscale & PF_WSCALE_MASK); 3654 s->src.max_win = (win - 1) >> 3655 (s->src.wscale & PF_WSCALE_MASK); 3656 } 3657 if (th->th_flags & TH_FIN) 3658 s->src.seqhi++; 3659 s->dst.seqhi = 1; 3660 s->dst.max_win = 1; 3661 s->src.state = TCPS_SYN_SENT; 3662 s->dst.state = TCPS_CLOSED; 3663 s->timeout = PFTM_TCP_FIRST_PACKET; 3664 break; 3665 case IPPROTO_UDP: 3666 s->src.state = PFUDPS_SINGLE; 3667 s->dst.state = PFUDPS_NO_TRAFFIC; 3668 s->timeout = PFTM_UDP_FIRST_PACKET; 3669 break; 3670 case IPPROTO_ICMP: 3671 #ifdef INET6 3672 case IPPROTO_ICMPV6: 3673 #endif 3674 s->timeout = PFTM_ICMP_FIRST_PACKET; 3675 break; 3676 default: 3677 s->src.state = PFOTHERS_SINGLE; 3678 s->dst.state = PFOTHERS_NO_TRAFFIC; 3679 s->timeout = PFTM_OTHER_FIRST_PACKET; 3680 } 3681 3682 s->creation = time_second; 3683 s->expire = time_second; 3684 3685 if (sn != NULL) { 3686 s->src_node = sn; 3687 s->src_node->states++; 3688 } 3689 if (nsn != NULL) { 3690 /* XXX We only modify one side for now. */ 3691 PF_ACPY(&nsn->raddr, &nk->addr[1], pd->af); 3692 s->nat_src_node = nsn; 3693 s->nat_src_node->states++; 3694 } 3695 if (pd->proto == IPPROTO_TCP) { 3696 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m, 3697 off, pd, th, &s->src, &s->dst)) { 3698 REASON_SET(&reason, PFRES_MEMORY); 3699 pf_src_tree_remove_state(s); 3700 STATE_DEC_COUNTERS(s); 3701 kfree(s, M_PFSTATEPL); 3702 return (PF_DROP); 3703 } 3704 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub && 3705 pf_normalize_tcp_stateful(m, off, pd, &reason, th, s, 3706 &s->src, &s->dst, rewrite)) { 3707 /* This really shouldn't happen!!! */ 3708 DPFPRINTF(PF_DEBUG_URGENT, 3709 ("pf_normalize_tcp_stateful failed on first pkt")); 3710 pf_normalize_tcp_cleanup(s); 3711 pf_src_tree_remove_state(s); 3712 STATE_DEC_COUNTERS(s); 3713 kfree(s, M_PFSTATEPL); 3714 return (PF_DROP); 3715 } 3716 } 3717 s->direction = pd->dir; 3718 3719 if (sk == NULL && pf_state_key_setup(pd, nr, &skw, &sks, &sk, &nk, 3720 pd->src, pd->dst, sport, dport)) 3721 goto csfailed; 3722 3723 if (pf_state_insert(BOUND_IFACE(r, kif), skw, sks, s)) { 3724 if (pd->proto == IPPROTO_TCP) 3725 pf_normalize_tcp_cleanup(s); 3726 REASON_SET(&reason, PFRES_STATEINS); 3727 pf_src_tree_remove_state(s); 3728 STATE_DEC_COUNTERS(s); 3729 kfree(s, M_PFSTATEPL); 3730 return (PF_DROP); 3731 } else 3732 *sm = s; 3733 3734 pf_set_rt_ifp(s, pd->src); /* needs s->state_key set */ 3735 if (tag > 0) { 3736 pf_tag_ref(tag); 3737 s->tag = tag; 3738 } 3739 if (pd->proto == IPPROTO_TCP && (th->th_flags & (TH_SYN|TH_ACK)) == 3740 TH_SYN && r->keep_state == PF_STATE_SYNPROXY) { 3741 s->src.state = PF_TCPS_PROXY_SRC; 3742 /* undo NAT changes, if they have taken place */ 3743 if (nr != NULL) { 3744 struct pf_state_key *skt = s->key[PF_SK_WIRE]; 3745 if (pd->dir == PF_OUT) 3746 skt = s->key[PF_SK_STACK]; 3747 PF_ACPY(pd->src, &skt->addr[pd->sidx], pd->af); 3748 PF_ACPY(pd->dst, &skt->addr[pd->didx], pd->af); 3749 if (pd->sport) 3750 *pd->sport = skt->port[pd->sidx]; 3751 if (pd->dport) 3752 *pd->dport = skt->port[pd->didx]; 3753 if (pd->proto_sum) 3754 *pd->proto_sum = bproto_sum; 3755 if (pd->ip_sum) 3756 *pd->ip_sum = bip_sum; 3757 m_copyback(m, off, hdrlen, pd->hdr.any); 3758 } 3759 s->src.seqhi = htonl(karc4random()); 3760 /* Find mss option */ 3761 mss = pf_get_mss(m, off, th->th_off, pd->af); 3762 mss = pf_calc_mss(pd->src, pd->af, mss); 3763 mss = pf_calc_mss(pd->dst, pd->af, mss); 3764 s->src.mss = mss; 3765 pf_send_tcp(r, pd->af, pd->dst, pd->src, th->th_dport, 3766 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1, 3767 TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL); 3768 REASON_SET(&reason, PFRES_SYNPROXY); 3769 return (PF_SYNPROXY_DROP); 3770 } 3771 3772 return (PF_PASS); 3773 3774 csfailed: 3775 if (sk != NULL) 3776 kfree(sk, M_PFSTATEKEYPL); 3777 if (nk != NULL) 3778 kfree(nk, M_PFSTATEKEYPL); 3779 3780 if (sn != NULL && sn->states == 0 && sn->expire == 0) { 3781 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn); 3782 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3783 pf_status.src_nodes--; 3784 kfree(sn, M_PFSRCTREEPL); 3785 } 3786 if (nsn != sn && nsn != NULL && nsn->states == 0 && nsn->expire == 0) { 3787 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn); 3788 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++; 3789 pf_status.src_nodes--; 3790 kfree(nsn, M_PFSRCTREEPL); 3791 } 3792 return (PF_DROP); 3793 } 3794 3795 int 3796 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif, 3797 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am, 3798 struct pf_ruleset **rsm) 3799 { 3800 struct pf_rule *r, *a = NULL; 3801 struct pf_ruleset *ruleset = NULL; 3802 sa_family_t af = pd->af; 3803 u_short reason; 3804 int tag = -1; 3805 int asd = 0; 3806 int match = 0; 3807 3808 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); 3809 while (r != NULL) { 3810 r->evaluations++; 3811 if (pfi_kif_match(r->kif, kif) == r->ifnot) 3812 r = r->skip[PF_SKIP_IFP].ptr; 3813 else if (r->direction && r->direction != direction) 3814 r = r->skip[PF_SKIP_DIR].ptr; 3815 else if (r->af && r->af != af) 3816 r = r->skip[PF_SKIP_AF].ptr; 3817 else if (r->proto && r->proto != pd->proto) 3818 r = r->skip[PF_SKIP_PROTO].ptr; 3819 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, 3820 r->src.neg, kif)) 3821 r = r->skip[PF_SKIP_SRC_ADDR].ptr; 3822 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, 3823 r->dst.neg, NULL)) 3824 r = r->skip[PF_SKIP_DST_ADDR].ptr; 3825 else if (r->tos && !(r->tos == pd->tos)) 3826 r = TAILQ_NEXT(r, entries); 3827 else if (r->os_fingerprint != PF_OSFP_ANY) 3828 r = TAILQ_NEXT(r, entries); 3829 else if (pd->proto == IPPROTO_UDP && 3830 (r->src.port_op || r->dst.port_op)) 3831 r = TAILQ_NEXT(r, entries); 3832 else if (pd->proto == IPPROTO_TCP && 3833 (r->src.port_op || r->dst.port_op || r->flagset)) 3834 r = TAILQ_NEXT(r, entries); 3835 else if ((pd->proto == IPPROTO_ICMP || 3836 pd->proto == IPPROTO_ICMPV6) && 3837 (r->type || r->code)) 3838 r = TAILQ_NEXT(r, entries); 3839 else if (r->prob && r->prob <= karc4random()) 3840 r = TAILQ_NEXT(r, entries); 3841 else if (r->match_tag && !pf_match_tag(m, r, &tag)) 3842 r = TAILQ_NEXT(r, entries); 3843 else { 3844 if (r->anchor == NULL) { 3845 match = 1; 3846 *rm = r; 3847 *am = a; 3848 *rsm = ruleset; 3849 if ((*rm)->quick) 3850 break; 3851 r = TAILQ_NEXT(r, entries); 3852 } else 3853 pf_step_into_anchor(&asd, &ruleset, 3854 PF_RULESET_FILTER, &r, &a, &match); 3855 } 3856 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset, 3857 PF_RULESET_FILTER, &r, &a, &match)) 3858 break; 3859 } 3860 r = *rm; 3861 a = *am; 3862 ruleset = *rsm; 3863 3864 REASON_SET(&reason, PFRES_MATCH); 3865 3866 if (r->log) 3867 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset, 3868 pd); 3869 3870 if (r->action != PF_PASS) 3871 return (PF_DROP); 3872 3873 if (pf_tag_packet(m, tag, -1)) { 3874 REASON_SET(&reason, PFRES_MEMORY); 3875 return (PF_DROP); 3876 } 3877 3878 return (PF_PASS); 3879 } 3880 3881 int 3882 pf_tcp_track_full(struct pf_state_peer *src, struct pf_state_peer *dst, 3883 struct pf_state **state, struct pfi_kif *kif, struct mbuf *m, int off, 3884 struct pf_pdesc *pd, u_short *reason, int *copyback) 3885 { 3886 struct tcphdr *th = pd->hdr.tcp; 3887 u_int16_t win = ntohs(th->th_win); 3888 u_int32_t ack, end, seq, orig_seq; 3889 u_int8_t sws, dws; 3890 int ackskew; 3891 3892 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) { 3893 sws = src->wscale & PF_WSCALE_MASK; 3894 dws = dst->wscale & PF_WSCALE_MASK; 3895 } else 3896 sws = dws = 0; 3897 3898 /* 3899 * Sequence tracking algorithm from Guido van Rooij's paper: 3900 * http://www.madison-gurkha.com/publications/tcp_filtering/ 3901 * tcp_filtering.ps 3902 */ 3903 3904 orig_seq = seq = ntohl(th->th_seq); 3905 if (src->seqlo == 0) { 3906 /* First packet from this end. Set its state */ 3907 3908 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) && 3909 src->scrub == NULL) { 3910 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) { 3911 REASON_SET(reason, PFRES_MEMORY); 3912 return (PF_DROP); 3913 } 3914 } 3915 3916 /* Deferred generation of sequence number modulator */ 3917 if (dst->seqdiff && !src->seqdiff) { 3918 /* use random iss for the TCP server */ 3919 while ((src->seqdiff = karc4random() - seq) == 0) 3920 ; 3921 ack = ntohl(th->th_ack) - dst->seqdiff; 3922 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq + 3923 src->seqdiff), 0); 3924 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0); 3925 *copyback = 1; 3926 } else { 3927 ack = ntohl(th->th_ack); 3928 } 3929 3930 end = seq + pd->p_len; 3931 if (th->th_flags & TH_SYN) { 3932 end++; 3933 (*state)->sync_flags |= PFSTATE_GOT_SYN2; 3934 if (dst->wscale & PF_WSCALE_FLAG) { 3935 src->wscale = pf_get_wscale(m, off, th->th_off, 3936 pd->af); 3937 if (src->wscale & PF_WSCALE_FLAG) { 3938 /* Remove scale factor from initial 3939 * window */ 3940 sws = src->wscale & PF_WSCALE_MASK; 3941 win = ((u_int32_t)win + (1 << sws) - 1) 3942 >> sws; 3943 dws = dst->wscale & PF_WSCALE_MASK; 3944 } else { 3945 /* fixup other window */ 3946 dst->max_win <<= dst->wscale & 3947 PF_WSCALE_MASK; 3948 /* in case of a retrans SYN|ACK */ 3949 dst->wscale = 0; 3950 } 3951 } 3952 } 3953 if (th->th_flags & TH_FIN) 3954 end++; 3955 3956 src->seqlo = seq; 3957 if (src->state < TCPS_SYN_SENT) 3958 src->state = TCPS_SYN_SENT; 3959 3960 /* 3961 * May need to slide the window (seqhi may have been set by 3962 * the crappy stack check or if we picked up the connection 3963 * after establishment) 3964 */ 3965 if (src->seqhi == 1 || 3966 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi)) 3967 src->seqhi = end + MAX(1, dst->max_win << dws); 3968 if (win > src->max_win) 3969 src->max_win = win; 3970 3971 } else { 3972 ack = ntohl(th->th_ack) - dst->seqdiff; 3973 if (src->seqdiff) { 3974 /* Modulate sequence numbers */ 3975 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq + 3976 src->seqdiff), 0); 3977 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0); 3978 *copyback = 1; 3979 } 3980 end = seq + pd->p_len; 3981 if (th->th_flags & TH_SYN) 3982 end++; 3983 if (th->th_flags & TH_FIN) 3984 end++; 3985 } 3986 3987 if ((th->th_flags & TH_ACK) == 0) { 3988 /* Let it pass through the ack skew check */ 3989 ack = dst->seqlo; 3990 } else if ((ack == 0 && 3991 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) || 3992 /* broken tcp stacks do not set ack */ 3993 (dst->state < TCPS_SYN_SENT)) { 3994 /* 3995 * Many stacks (ours included) will set the ACK number in an 3996 * FIN|ACK if the SYN times out -- no sequence to ACK. 3997 */ 3998 ack = dst->seqlo; 3999 } 4000 4001 if (seq == end) { 4002 /* Ease sequencing restrictions on no data packets */ 4003 seq = src->seqlo; 4004 end = seq; 4005 } 4006 4007 ackskew = dst->seqlo - ack; 4008 4009 4010 /* 4011 * Need to demodulate the sequence numbers in any TCP SACK options 4012 * (Selective ACK). We could optionally validate the SACK values 4013 * against the current ACK window, either forwards or backwards, but 4014 * I'm not confident that SACK has been implemented properly 4015 * everywhere. It wouldn't surprise me if several stacks accidently 4016 * SACK too far backwards of previously ACKed data. There really aren't 4017 * any security implications of bad SACKing unless the target stack 4018 * doesn't validate the option length correctly. Someone trying to 4019 * spoof into a TCP connection won't bother blindly sending SACK 4020 * options anyway. 4021 */ 4022 if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) { 4023 if (pf_modulate_sack(m, off, pd, th, dst)) 4024 *copyback = 1; 4025 } 4026 4027 4028 #define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */ 4029 if (SEQ_GEQ(src->seqhi, end) && 4030 /* Last octet inside other's window space */ 4031 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) && 4032 /* Retrans: not more than one window back */ 4033 (ackskew >= -MAXACKWINDOW) && 4034 /* Acking not more than one reassembled fragment backwards */ 4035 (ackskew <= (MAXACKWINDOW << sws)) && 4036 /* Acking not more than one window forward */ 4037 ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo || 4038 (orig_seq == src->seqlo + 1) || (orig_seq + 1 == src->seqlo) || 4039 (pd->flags & PFDESC_IP_REAS) == 0)) { 4040 /* Require an exact/+1 sequence match on resets when possible */ 4041 4042 if (dst->scrub || src->scrub) { 4043 if (pf_normalize_tcp_stateful(m, off, pd, reason, th, 4044 *state, src, dst, copyback)) 4045 return (PF_DROP); 4046 } 4047 4048 /* update max window */ 4049 if (src->max_win < win) 4050 src->max_win = win; 4051 /* synchronize sequencing */ 4052 if (SEQ_GT(end, src->seqlo)) 4053 src->seqlo = end; 4054 /* slide the window of what the other end can send */ 4055 if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) 4056 dst->seqhi = ack + MAX((win << sws), 1); 4057 4058 4059 /* update states */ 4060 if (th->th_flags & TH_SYN) 4061 if (src->state < TCPS_SYN_SENT) 4062 src->state = TCPS_SYN_SENT; 4063 if (th->th_flags & TH_FIN) 4064 if (src->state < TCPS_CLOSING) 4065 src->state = TCPS_CLOSING; 4066 if (th->th_flags & TH_ACK) { 4067 if (dst->state == TCPS_SYN_SENT) { 4068 dst->state = TCPS_ESTABLISHED; 4069 if (src->state == TCPS_ESTABLISHED && 4070 (*state)->src_node != NULL && 4071 pf_src_connlimit(state)) { 4072 REASON_SET(reason, PFRES_SRCLIMIT); 4073 return (PF_DROP); 4074 } 4075 } else if (dst->state == TCPS_CLOSING) 4076 dst->state = TCPS_FIN_WAIT_2; 4077 } 4078 if (th->th_flags & TH_RST) 4079 src->state = dst->state = TCPS_TIME_WAIT; 4080 4081 /* update expire time */ 4082 (*state)->expire = time_second; 4083 if (src->state >= TCPS_FIN_WAIT_2 && 4084 dst->state >= TCPS_FIN_WAIT_2) 4085 (*state)->timeout = PFTM_TCP_CLOSED; 4086 else if (src->state >= TCPS_CLOSING && 4087 dst->state >= TCPS_CLOSING) 4088 (*state)->timeout = PFTM_TCP_FIN_WAIT; 4089 else if (src->state < TCPS_ESTABLISHED || 4090 dst->state < TCPS_ESTABLISHED) 4091 (*state)->timeout = PFTM_TCP_OPENING; 4092 else if (src->state >= TCPS_CLOSING || 4093 dst->state >= TCPS_CLOSING) 4094 (*state)->timeout = PFTM_TCP_CLOSING; 4095 else 4096 (*state)->timeout = PFTM_TCP_ESTABLISHED; 4097 4098 /* Fall through to PASS packet */ 4099 4100 } else if ((dst->state < TCPS_SYN_SENT || 4101 dst->state >= TCPS_FIN_WAIT_2 || 4102 src->state >= TCPS_FIN_WAIT_2) && 4103 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) && 4104 /* Within a window forward of the originating packet */ 4105 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) { 4106 /* Within a window backward of the originating packet */ 4107 4108 /* 4109 * This currently handles three situations: 4110 * 1) Stupid stacks will shotgun SYNs before their peer 4111 * replies. 4112 * 2) When PF catches an already established stream (the 4113 * firewall rebooted, the state table was flushed, routes 4114 * changed...) 4115 * 3) Packets get funky immediately after the connection 4116 * closes (this should catch Solaris spurious ACK|FINs 4117 * that web servers like to spew after a close) 4118 * 4119 * This must be a little more careful than the above code 4120 * since packet floods will also be caught here. We don't 4121 * update the TTL here to mitigate the damage of a packet 4122 * flood and so the same code can handle awkward establishment 4123 * and a loosened connection close. 4124 * In the establishment case, a correct peer response will 4125 * validate the connection, go through the normal state code 4126 * and keep updating the state TTL. 4127 */ 4128 4129 if (pf_status.debug >= PF_DEBUG_MISC) { 4130 kprintf("pf: loose state match: "); 4131 pf_print_state(*state); 4132 pf_print_flags(th->th_flags); 4133 kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d " 4134 "pkts=%llu:%llu dir=%s,%s\n", seq, orig_seq, ack, pd->p_len, 4135 ackskew, (unsigned long long)(*state)->packets[0], 4136 (unsigned long long)(*state)->packets[1], 4137 pd->dir == PF_IN ? "in" : "out", 4138 pd->dir == (*state)->direction ? "fwd" : "rev"); 4139 } 4140 4141 if (dst->scrub || src->scrub) { 4142 if (pf_normalize_tcp_stateful(m, off, pd, reason, th, 4143 *state, src, dst, copyback)) 4144 return (PF_DROP); 4145 } 4146 4147 /* update max window */ 4148 if (src->max_win < win) 4149 src->max_win = win; 4150 /* synchronize sequencing */ 4151 if (SEQ_GT(end, src->seqlo)) 4152 src->seqlo = end; 4153 /* slide the window of what the other end can send */ 4154 if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) 4155 dst->seqhi = ack + MAX((win << sws), 1); 4156 4157 /* 4158 * Cannot set dst->seqhi here since this could be a shotgunned 4159 * SYN and not an already established connection. 4160 */ 4161 4162 if (th->th_flags & TH_FIN) 4163 if (src->state < TCPS_CLOSING) 4164 src->state = TCPS_CLOSING; 4165 if (th->th_flags & TH_RST) 4166 src->state = dst->state = TCPS_TIME_WAIT; 4167 4168 /* Fall through to PASS packet */ 4169 4170 } else if ((*state)->pickup_mode == PF_PICKUPS_HASHONLY || 4171 ((*state)->pickup_mode == PF_PICKUPS_ENABLED && 4172 ((*state)->sync_flags & PFSTATE_GOT_SYN_MASK) != 4173 PFSTATE_GOT_SYN_MASK)) { 4174 /* 4175 * If pickup mode is hash only, do not fail on sequence checks. 4176 * 4177 * If pickup mode is enabled and we did not see the SYN in 4178 * both direction, do not fail on sequence checks because 4179 * we do not have complete information on window scale. 4180 * 4181 * Adjust expiration and fall through to PASS packet. 4182 * XXX Add a FIN check to reduce timeout? 4183 */ 4184 (*state)->expire = time_second; 4185 } else { 4186 /* 4187 * Failure processing 4188 */ 4189 if ((*state)->dst.state == TCPS_SYN_SENT && 4190 (*state)->src.state == TCPS_SYN_SENT) { 4191 /* Send RST for state mismatches during handshake */ 4192 if (!(th->th_flags & TH_RST)) 4193 pf_send_tcp((*state)->rule.ptr, pd->af, 4194 pd->dst, pd->src, th->th_dport, 4195 th->th_sport, ntohl(th->th_ack), 0, 4196 TH_RST, 0, 0, 4197 (*state)->rule.ptr->return_ttl, 1, 0, 4198 pd->eh, kif->pfik_ifp); 4199 src->seqlo = 0; 4200 src->seqhi = 1; 4201 src->max_win = 1; 4202 } else if (pf_status.debug >= PF_DEBUG_MISC) { 4203 kprintf("pf: BAD state: "); 4204 pf_print_state(*state); 4205 pf_print_flags(th->th_flags); 4206 kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d " 4207 "pkts=%llu:%llu dir=%s,%s\n", 4208 seq, orig_seq, ack, pd->p_len, ackskew, 4209 (unsigned long long)(*state)->packets[0], 4210 (unsigned long long)(*state)->packets[1], 4211 pd->dir == PF_IN ? "in" : "out", 4212 pd->dir == (*state)->direction ? "fwd" : "rev"); 4213 kprintf("pf: State failure on: %c %c %c %c | %c %c\n", 4214 SEQ_GEQ(src->seqhi, end) ? ' ' : '1', 4215 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ? 4216 ' ': '2', 4217 (ackskew >= -MAXACKWINDOW) ? ' ' : '3', 4218 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4', 4219 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5', 4220 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6'); 4221 } 4222 REASON_SET(reason, PFRES_BADSTATE); 4223 return (PF_DROP); 4224 } 4225 4226 return (PF_PASS); 4227 } 4228 4229 int 4230 pf_tcp_track_sloppy(struct pf_state_peer *src, struct pf_state_peer *dst, 4231 struct pf_state **state, struct pf_pdesc *pd, u_short *reason) 4232 { 4233 struct tcphdr *th = pd->hdr.tcp; 4234 4235 if (th->th_flags & TH_SYN) 4236 if (src->state < TCPS_SYN_SENT) 4237 src->state = TCPS_SYN_SENT; 4238 if (th->th_flags & TH_FIN) 4239 if (src->state < TCPS_CLOSING) 4240 src->state = TCPS_CLOSING; 4241 if (th->th_flags & TH_ACK) { 4242 if (dst->state == TCPS_SYN_SENT) { 4243 dst->state = TCPS_ESTABLISHED; 4244 if (src->state == TCPS_ESTABLISHED && 4245 (*state)->src_node != NULL && 4246 pf_src_connlimit(state)) { 4247 REASON_SET(reason, PFRES_SRCLIMIT); 4248 return (PF_DROP); 4249 } 4250 } else if (dst->state == TCPS_CLOSING) { 4251 dst->state = TCPS_FIN_WAIT_2; 4252 } else if (src->state == TCPS_SYN_SENT && 4253 dst->state < TCPS_SYN_SENT) { 4254 /* 4255 * Handle a special sloppy case where we only see one 4256 * half of the connection. If there is a ACK after 4257 * the initial SYN without ever seeing a packet from 4258 * the destination, set the connection to established. 4259 */ 4260 dst->state = src->state = TCPS_ESTABLISHED; 4261 if ((*state)->src_node != NULL && 4262 pf_src_connlimit(state)) { 4263 REASON_SET(reason, PFRES_SRCLIMIT); 4264 return (PF_DROP); 4265 } 4266 } else if (src->state == TCPS_CLOSING && 4267 dst->state == TCPS_ESTABLISHED && 4268 dst->seqlo == 0) { 4269 /* 4270 * Handle the closing of half connections where we 4271 * don't see the full bidirectional FIN/ACK+ACK 4272 * handshake. 4273 */ 4274 dst->state = TCPS_CLOSING; 4275 } 4276 } 4277 if (th->th_flags & TH_RST) 4278 src->state = dst->state = TCPS_TIME_WAIT; 4279 4280 /* update expire time */ 4281 (*state)->expire = time_second; 4282 if (src->state >= TCPS_FIN_WAIT_2 && 4283 dst->state >= TCPS_FIN_WAIT_2) 4284 (*state)->timeout = PFTM_TCP_CLOSED; 4285 else if (src->state >= TCPS_CLOSING && 4286 dst->state >= TCPS_CLOSING) 4287 (*state)->timeout = PFTM_TCP_FIN_WAIT; 4288 else if (src->state < TCPS_ESTABLISHED || 4289 dst->state < TCPS_ESTABLISHED) 4290 (*state)->timeout = PFTM_TCP_OPENING; 4291 else if (src->state >= TCPS_CLOSING || 4292 dst->state >= TCPS_CLOSING) 4293 (*state)->timeout = PFTM_TCP_CLOSING; 4294 else 4295 (*state)->timeout = PFTM_TCP_ESTABLISHED; 4296 4297 return (PF_PASS); 4298 } 4299 4300 int 4301 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif, 4302 struct mbuf *m, int off, void *h, struct pf_pdesc *pd, 4303 u_short *reason) 4304 { 4305 struct pf_state_key_cmp key; 4306 struct tcphdr *th = pd->hdr.tcp; 4307 int copyback = 0; 4308 struct pf_state_peer *src, *dst; 4309 struct pf_state_key *sk; 4310 4311 key.af = pd->af; 4312 key.proto = IPPROTO_TCP; 4313 if (direction == PF_IN) { /* wire side, straight */ 4314 PF_ACPY(&key.addr[0], pd->src, key.af); 4315 PF_ACPY(&key.addr[1], pd->dst, key.af); 4316 key.port[0] = th->th_sport; 4317 key.port[1] = th->th_dport; 4318 } else { /* stack side, reverse */ 4319 PF_ACPY(&key.addr[1], pd->src, key.af); 4320 PF_ACPY(&key.addr[0], pd->dst, key.af); 4321 key.port[1] = th->th_sport; 4322 key.port[0] = th->th_dport; 4323 } 4324 4325 STATE_LOOKUP(kif, &key, direction, *state, m); 4326 4327 if (direction == (*state)->direction) { 4328 src = &(*state)->src; 4329 dst = &(*state)->dst; 4330 } else { 4331 src = &(*state)->dst; 4332 dst = &(*state)->src; 4333 } 4334 4335 sk = (*state)->key[pd->didx]; 4336 4337 if ((*state)->src.state == PF_TCPS_PROXY_SRC) { 4338 if (direction != (*state)->direction) { 4339 REASON_SET(reason, PFRES_SYNPROXY); 4340 return (PF_SYNPROXY_DROP); 4341 } 4342 if (th->th_flags & TH_SYN) { 4343 if (ntohl(th->th_seq) != (*state)->src.seqlo) { 4344 REASON_SET(reason, PFRES_SYNPROXY); 4345 return (PF_DROP); 4346 } 4347 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, 4348 pd->src, th->th_dport, th->th_sport, 4349 (*state)->src.seqhi, ntohl(th->th_seq) + 1, 4350 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1, 4351 0, NULL, NULL); 4352 REASON_SET(reason, PFRES_SYNPROXY); 4353 return (PF_SYNPROXY_DROP); 4354 } else if (!(th->th_flags & TH_ACK) || 4355 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) || 4356 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) { 4357 REASON_SET(reason, PFRES_SYNPROXY); 4358 return (PF_DROP); 4359 } else if ((*state)->src_node != NULL && 4360 pf_src_connlimit(state)) { 4361 REASON_SET(reason, PFRES_SRCLIMIT); 4362 return (PF_DROP); 4363 } else 4364 (*state)->src.state = PF_TCPS_PROXY_DST; 4365 } 4366 if ((*state)->src.state == PF_TCPS_PROXY_DST) { 4367 if (direction == (*state)->direction) { 4368 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) || 4369 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) || 4370 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) { 4371 REASON_SET(reason, PFRES_SYNPROXY); 4372 return (PF_DROP); 4373 } 4374 (*state)->src.max_win = MAX(ntohs(th->th_win), 1); 4375 if ((*state)->dst.seqhi == 1) 4376 (*state)->dst.seqhi = htonl(karc4random()); 4377 pf_send_tcp((*state)->rule.ptr, pd->af, 4378 &sk->addr[pd->sidx], &sk->addr[pd->didx], 4379 sk->port[pd->sidx], sk->port[pd->didx], 4380 (*state)->dst.seqhi, 0, TH_SYN, 0, 4381 (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL); 4382 REASON_SET(reason, PFRES_SYNPROXY); 4383 return (PF_SYNPROXY_DROP); 4384 } else if (((th->th_flags & (TH_SYN|TH_ACK)) != 4385 (TH_SYN|TH_ACK)) || 4386 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) { 4387 REASON_SET(reason, PFRES_SYNPROXY); 4388 return (PF_DROP); 4389 } else { 4390 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1); 4391 (*state)->dst.seqlo = ntohl(th->th_seq); 4392 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, 4393 pd->src, th->th_dport, th->th_sport, 4394 ntohl(th->th_ack), ntohl(th->th_seq) + 1, 4395 TH_ACK, (*state)->src.max_win, 0, 0, 0, 4396 (*state)->tag, NULL, NULL); 4397 pf_send_tcp((*state)->rule.ptr, pd->af, 4398 &sk->addr[pd->sidx], &sk->addr[pd->didx], 4399 sk->port[pd->sidx], sk->port[pd->didx], 4400 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1, 4401 TH_ACK, (*state)->dst.max_win, 0, 0, 1, 4402 0, NULL, NULL); 4403 (*state)->src.seqdiff = (*state)->dst.seqhi - 4404 (*state)->src.seqlo; 4405 (*state)->dst.seqdiff = (*state)->src.seqhi - 4406 (*state)->dst.seqlo; 4407 (*state)->src.seqhi = (*state)->src.seqlo + 4408 (*state)->dst.max_win; 4409 (*state)->dst.seqhi = (*state)->dst.seqlo + 4410 (*state)->src.max_win; 4411 (*state)->src.wscale = (*state)->dst.wscale = 0; 4412 (*state)->src.state = (*state)->dst.state = 4413 TCPS_ESTABLISHED; 4414 REASON_SET(reason, PFRES_SYNPROXY); 4415 return (PF_SYNPROXY_DROP); 4416 } 4417 } 4418 4419 if (((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN) && 4420 dst->state >= TCPS_FIN_WAIT_2 && 4421 src->state >= TCPS_FIN_WAIT_2) { 4422 if (pf_status.debug >= PF_DEBUG_MISC) { 4423 kprintf("pf: state reuse "); 4424 pf_print_state(*state); 4425 pf_print_flags(th->th_flags); 4426 kprintf("\n"); 4427 } 4428 /* XXX make sure it's the same direction ?? */ 4429 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED; 4430 pf_unlink_state(*state); 4431 *state = NULL; 4432 return (PF_DROP); 4433 } 4434 4435 if ((*state)->state_flags & PFSTATE_SLOPPY) { 4436 if (pf_tcp_track_sloppy(src, dst, state, pd, reason) == PF_DROP) 4437 return (PF_DROP); 4438 } else { 4439 if (pf_tcp_track_full(src, dst, state, kif, m, off, pd, reason, 4440 ©back) == PF_DROP) 4441 return (PF_DROP); 4442 } 4443 4444 /* translate source/destination address, if necessary */ 4445 if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) { 4446 struct pf_state_key *nk = (*state)->key[pd->didx]; 4447 4448 if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af) || 4449 nk->port[pd->sidx] != th->th_sport) { 4450 /* 4451 * The translated source address may be completely 4452 * unrelated to the saved link header, make sure 4453 * a bridge doesn't try to use it. 4454 */ 4455 m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED; 4456 m->m_flags &= ~M_HASH; 4457 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum, 4458 &th->th_sum, &nk->addr[pd->sidx], 4459 nk->port[pd->sidx], 0, pd->af); 4460 } 4461 4462 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af) || 4463 nk->port[pd->didx] != th->th_dport) { 4464 /* 4465 * If we don't redispatch the packet will go into 4466 * the protocol stack on the wrong cpu for the 4467 * post-translated address. 4468 */ 4469 m->m_flags &= ~M_HASH; 4470 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum, 4471 &th->th_sum, &nk->addr[pd->didx], 4472 nk->port[pd->didx], 0, pd->af); 4473 } 4474 copyback = 1; 4475 } 4476 4477 /* Copyback sequence modulation or stateful scrub changes if needed */ 4478 if (copyback) 4479 m_copyback(m, off, sizeof(*th), (caddr_t)th); 4480 4481 return (PF_PASS); 4482 } 4483 4484 int 4485 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif, 4486 struct mbuf *m, int off, void *h, struct pf_pdesc *pd) 4487 { 4488 struct pf_state_peer *src, *dst; 4489 struct pf_state_key_cmp key; 4490 struct udphdr *uh = pd->hdr.udp; 4491 4492 key.af = pd->af; 4493 key.proto = IPPROTO_UDP; 4494 if (direction == PF_IN) { /* wire side, straight */ 4495 PF_ACPY(&key.addr[0], pd->src, key.af); 4496 PF_ACPY(&key.addr[1], pd->dst, key.af); 4497 key.port[0] = uh->uh_sport; 4498 key.port[1] = uh->uh_dport; 4499 } else { /* stack side, reverse */ 4500 PF_ACPY(&key.addr[1], pd->src, key.af); 4501 PF_ACPY(&key.addr[0], pd->dst, key.af); 4502 key.port[1] = uh->uh_sport; 4503 key.port[0] = uh->uh_dport; 4504 } 4505 4506 STATE_LOOKUP(kif, &key, direction, *state, m); 4507 4508 if (direction == (*state)->direction) { 4509 src = &(*state)->src; 4510 dst = &(*state)->dst; 4511 } else { 4512 src = &(*state)->dst; 4513 dst = &(*state)->src; 4514 } 4515 4516 /* update states */ 4517 if (src->state < PFUDPS_SINGLE) 4518 src->state = PFUDPS_SINGLE; 4519 if (dst->state == PFUDPS_SINGLE) 4520 dst->state = PFUDPS_MULTIPLE; 4521 4522 /* update expire time */ 4523 (*state)->expire = time_second; 4524 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE) 4525 (*state)->timeout = PFTM_UDP_MULTIPLE; 4526 else 4527 (*state)->timeout = PFTM_UDP_SINGLE; 4528 4529 /* translate source/destination address, if necessary */ 4530 if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) { 4531 struct pf_state_key *nk = (*state)->key[pd->didx]; 4532 4533 if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af) || 4534 nk->port[pd->sidx] != uh->uh_sport) { 4535 /* 4536 * The translated source address may be completely 4537 * unrelated to the saved link header, make sure 4538 * a bridge doesn't try to use it. 4539 */ 4540 m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED; 4541 m->m_flags &= ~M_HASH; 4542 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum, 4543 &uh->uh_sum, &nk->addr[pd->sidx], 4544 nk->port[pd->sidx], 1, pd->af); 4545 } 4546 4547 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af) || 4548 nk->port[pd->didx] != uh->uh_dport) { 4549 /* 4550 * If we don't redispatch the packet will go into 4551 * the protocol stack on the wrong cpu for the 4552 * post-translated address. 4553 */ 4554 m->m_flags &= ~M_HASH; 4555 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum, 4556 &uh->uh_sum, &nk->addr[pd->didx], 4557 nk->port[pd->didx], 1, pd->af); 4558 } 4559 m_copyback(m, off, sizeof(*uh), (caddr_t)uh); 4560 } 4561 4562 return (PF_PASS); 4563 } 4564 4565 int 4566 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif, 4567 struct mbuf *m, int off, void *h, struct pf_pdesc *pd, u_short *reason) 4568 { 4569 struct pf_addr *saddr = pd->src, *daddr = pd->dst; 4570 u_int16_t icmpid = 0, *icmpsum; 4571 u_int8_t icmptype; 4572 int state_icmp = 0; 4573 struct pf_state_key_cmp key; 4574 4575 switch (pd->proto) { 4576 #ifdef INET 4577 case IPPROTO_ICMP: 4578 icmptype = pd->hdr.icmp->icmp_type; 4579 icmpid = pd->hdr.icmp->icmp_id; 4580 icmpsum = &pd->hdr.icmp->icmp_cksum; 4581 4582 if (icmptype == ICMP_UNREACH || 4583 icmptype == ICMP_SOURCEQUENCH || 4584 icmptype == ICMP_REDIRECT || 4585 icmptype == ICMP_TIMXCEED || 4586 icmptype == ICMP_PARAMPROB) 4587 state_icmp++; 4588 break; 4589 #endif /* INET */ 4590 #ifdef INET6 4591 case IPPROTO_ICMPV6: 4592 icmptype = pd->hdr.icmp6->icmp6_type; 4593 icmpid = pd->hdr.icmp6->icmp6_id; 4594 icmpsum = &pd->hdr.icmp6->icmp6_cksum; 4595 4596 if (icmptype == ICMP6_DST_UNREACH || 4597 icmptype == ICMP6_PACKET_TOO_BIG || 4598 icmptype == ICMP6_TIME_EXCEEDED || 4599 icmptype == ICMP6_PARAM_PROB) 4600 state_icmp++; 4601 break; 4602 #endif /* INET6 */ 4603 } 4604 4605 if (!state_icmp) { 4606 4607 /* 4608 * ICMP query/reply message not related to a TCP/UDP packet. 4609 * Search for an ICMP state. 4610 */ 4611 key.af = pd->af; 4612 key.proto = pd->proto; 4613 key.port[0] = key.port[1] = icmpid; 4614 if (direction == PF_IN) { /* wire side, straight */ 4615 PF_ACPY(&key.addr[0], pd->src, key.af); 4616 PF_ACPY(&key.addr[1], pd->dst, key.af); 4617 } else { /* stack side, reverse */ 4618 PF_ACPY(&key.addr[1], pd->src, key.af); 4619 PF_ACPY(&key.addr[0], pd->dst, key.af); 4620 } 4621 4622 STATE_LOOKUP(kif, &key, direction, *state, m); 4623 4624 (*state)->expire = time_second; 4625 (*state)->timeout = PFTM_ICMP_ERROR_REPLY; 4626 4627 /* translate source/destination address, if necessary */ 4628 if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) { 4629 struct pf_state_key *nk = (*state)->key[pd->didx]; 4630 4631 switch (pd->af) { 4632 #ifdef INET 4633 case AF_INET: 4634 if (PF_ANEQ(pd->src, 4635 &nk->addr[pd->sidx], AF_INET)) 4636 pf_change_a(&saddr->v4.s_addr, 4637 pd->ip_sum, 4638 nk->addr[pd->sidx].v4.s_addr, 0); 4639 4640 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], 4641 AF_INET)) 4642 pf_change_a(&daddr->v4.s_addr, 4643 pd->ip_sum, 4644 nk->addr[pd->didx].v4.s_addr, 0); 4645 4646 if (nk->port[0] != 4647 pd->hdr.icmp->icmp_id) { 4648 pd->hdr.icmp->icmp_cksum = 4649 pf_cksum_fixup( 4650 pd->hdr.icmp->icmp_cksum, icmpid, 4651 nk->port[pd->sidx], 0); 4652 pd->hdr.icmp->icmp_id = 4653 nk->port[pd->sidx]; 4654 } 4655 4656 m_copyback(m, off, ICMP_MINLEN, 4657 (caddr_t)pd->hdr.icmp); 4658 break; 4659 #endif /* INET */ 4660 #ifdef INET6 4661 case AF_INET6: 4662 if (PF_ANEQ(pd->src, 4663 &nk->addr[pd->sidx], AF_INET6)) 4664 pf_change_a6(saddr, 4665 &pd->hdr.icmp6->icmp6_cksum, 4666 &nk->addr[pd->sidx], 0); 4667 4668 if (PF_ANEQ(pd->dst, 4669 &nk->addr[pd->didx], AF_INET6)) 4670 pf_change_a6(daddr, 4671 &pd->hdr.icmp6->icmp6_cksum, 4672 &nk->addr[pd->didx], 0); 4673 4674 m_copyback(m, off, 4675 sizeof(struct icmp6_hdr), 4676 (caddr_t)pd->hdr.icmp6); 4677 break; 4678 #endif /* INET6 */ 4679 } 4680 } 4681 return (PF_PASS); 4682 4683 } else { 4684 /* 4685 * ICMP error message in response to a TCP/UDP packet. 4686 * Extract the inner TCP/UDP header and search for that state. 4687 */ 4688 4689 struct pf_pdesc pd2; 4690 #ifdef INET 4691 struct ip h2; 4692 #endif /* INET */ 4693 #ifdef INET6 4694 struct ip6_hdr h2_6; 4695 int terminal = 0; 4696 #endif /* INET6 */ 4697 int ipoff2; 4698 int off2; 4699 4700 pd2.af = pd->af; 4701 /* Payload packet is from the opposite direction. */ 4702 pd2.sidx = (direction == PF_IN) ? 1 : 0; 4703 pd2.didx = (direction == PF_IN) ? 0 : 1; 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 default: 4790 DPFPRINTF(PF_DEBUG_MISC, 4791 ("pf: ICMP AF %d unknown (ip6)\n", pd->af)); 4792 return (PF_DROP); 4793 break; 4794 } 4795 4796 switch (pd2.proto) { 4797 case IPPROTO_TCP: { 4798 struct tcphdr th; 4799 u_int32_t seq; 4800 struct pf_state_peer *src, *dst; 4801 u_int8_t dws; 4802 int copyback = 0; 4803 4804 /* 4805 * Only the first 8 bytes of the TCP header can be 4806 * expected. Don't access any TCP header fields after 4807 * th_seq, an ackskew test is not possible. 4808 */ 4809 if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason, 4810 pd2.af)) { 4811 DPFPRINTF(PF_DEBUG_MISC, 4812 ("pf: ICMP error message too short " 4813 "(tcp)\n")); 4814 return (PF_DROP); 4815 } 4816 4817 key.af = pd2.af; 4818 key.proto = IPPROTO_TCP; 4819 PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af); 4820 PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af); 4821 key.port[pd2.sidx] = th.th_sport; 4822 key.port[pd2.didx] = th.th_dport; 4823 4824 STATE_LOOKUP(kif, &key, direction, *state, m); 4825 4826 if (direction == (*state)->direction) { 4827 src = &(*state)->dst; 4828 dst = &(*state)->src; 4829 } else { 4830 src = &(*state)->src; 4831 dst = &(*state)->dst; 4832 } 4833 4834 if (src->wscale && dst->wscale) 4835 dws = dst->wscale & PF_WSCALE_MASK; 4836 else 4837 dws = 0; 4838 4839 /* Demodulate sequence number */ 4840 seq = ntohl(th.th_seq) - src->seqdiff; 4841 if (src->seqdiff) { 4842 pf_change_a(&th.th_seq, icmpsum, 4843 htonl(seq), 0); 4844 copyback = 1; 4845 } 4846 4847 if (!((*state)->state_flags & PFSTATE_SLOPPY) && 4848 (!SEQ_GEQ(src->seqhi, seq) || 4849 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)))) { 4850 if (pf_status.debug >= PF_DEBUG_MISC) { 4851 kprintf("pf: BAD ICMP %d:%d ", 4852 icmptype, pd->hdr.icmp->icmp_code); 4853 pf_print_host(pd->src, 0, pd->af); 4854 kprintf(" -> "); 4855 pf_print_host(pd->dst, 0, pd->af); 4856 kprintf(" state: "); 4857 pf_print_state(*state); 4858 kprintf(" seq=%u\n", seq); 4859 } 4860 REASON_SET(reason, PFRES_BADSTATE); 4861 return (PF_DROP); 4862 } else { 4863 if (pf_status.debug >= PF_DEBUG_MISC) { 4864 kprintf("pf: OK ICMP %d:%d ", 4865 icmptype, pd->hdr.icmp->icmp_code); 4866 pf_print_host(pd->src, 0, pd->af); 4867 kprintf(" -> "); 4868 pf_print_host(pd->dst, 0, pd->af); 4869 kprintf(" state: "); 4870 pf_print_state(*state); 4871 kprintf(" seq=%u\n", seq); 4872 } 4873 } 4874 4875 /* translate source/destination address, if necessary */ 4876 if ((*state)->key[PF_SK_WIRE] != 4877 (*state)->key[PF_SK_STACK]) { 4878 struct pf_state_key *nk = 4879 (*state)->key[pd->didx]; 4880 4881 if (PF_ANEQ(pd2.src, 4882 &nk->addr[pd2.sidx], pd2.af) || 4883 nk->port[pd2.sidx] != th.th_sport) 4884 pf_change_icmp(pd2.src, &th.th_sport, 4885 daddr, &nk->addr[pd2.sidx], 4886 nk->port[pd2.sidx], NULL, 4887 pd2.ip_sum, icmpsum, 4888 pd->ip_sum, 0, pd2.af); 4889 4890 if (PF_ANEQ(pd2.dst, 4891 &nk->addr[pd2.didx], pd2.af) || 4892 nk->port[pd2.didx] != th.th_dport) 4893 pf_change_icmp(pd2.dst, &th.th_dport, 4894 NULL, /* XXX Inbound NAT? */ 4895 &nk->addr[pd2.didx], 4896 nk->port[pd2.didx], NULL, 4897 pd2.ip_sum, icmpsum, 4898 pd->ip_sum, 0, pd2.af); 4899 copyback = 1; 4900 } 4901 4902 if (copyback) { 4903 switch (pd2.af) { 4904 #ifdef INET 4905 case AF_INET: 4906 m_copyback(m, off, ICMP_MINLEN, 4907 (caddr_t)pd->hdr.icmp); 4908 m_copyback(m, ipoff2, sizeof(h2), 4909 (caddr_t)&h2); 4910 break; 4911 #endif /* INET */ 4912 #ifdef INET6 4913 case AF_INET6: 4914 m_copyback(m, off, 4915 sizeof(struct icmp6_hdr), 4916 (caddr_t)pd->hdr.icmp6); 4917 m_copyback(m, ipoff2, sizeof(h2_6), 4918 (caddr_t)&h2_6); 4919 break; 4920 #endif /* INET6 */ 4921 } 4922 m_copyback(m, off2, 8, (caddr_t)&th); 4923 } 4924 4925 return (PF_PASS); 4926 break; 4927 } 4928 case IPPROTO_UDP: { 4929 struct udphdr uh; 4930 4931 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh), 4932 NULL, reason, pd2.af)) { 4933 DPFPRINTF(PF_DEBUG_MISC, 4934 ("pf: ICMP error message too short " 4935 "(udp)\n")); 4936 return (PF_DROP); 4937 } 4938 4939 key.af = pd2.af; 4940 key.proto = IPPROTO_UDP; 4941 PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af); 4942 PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af); 4943 key.port[pd2.sidx] = uh.uh_sport; 4944 key.port[pd2.didx] = uh.uh_dport; 4945 4946 STATE_LOOKUP(kif, &key, direction, *state, m); 4947 4948 /* translate source/destination address, if necessary */ 4949 if ((*state)->key[PF_SK_WIRE] != 4950 (*state)->key[PF_SK_STACK]) { 4951 struct pf_state_key *nk = 4952 (*state)->key[pd->didx]; 4953 4954 if (PF_ANEQ(pd2.src, 4955 &nk->addr[pd2.sidx], pd2.af) || 4956 nk->port[pd2.sidx] != uh.uh_sport) 4957 pf_change_icmp(pd2.src, &uh.uh_sport, 4958 daddr, &nk->addr[pd2.sidx], 4959 nk->port[pd2.sidx], &uh.uh_sum, 4960 pd2.ip_sum, icmpsum, 4961 pd->ip_sum, 1, pd2.af); 4962 4963 if (PF_ANEQ(pd2.dst, 4964 &nk->addr[pd2.didx], pd2.af) || 4965 nk->port[pd2.didx] != uh.uh_dport) 4966 pf_change_icmp(pd2.dst, &uh.uh_dport, 4967 NULL, /* XXX Inbound NAT? */ 4968 &nk->addr[pd2.didx], 4969 nk->port[pd2.didx], &uh.uh_sum, 4970 pd2.ip_sum, icmpsum, 4971 pd->ip_sum, 1, pd2.af); 4972 4973 switch (pd2.af) { 4974 #ifdef INET 4975 case AF_INET: 4976 m_copyback(m, off, ICMP_MINLEN, 4977 (caddr_t)pd->hdr.icmp); 4978 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2); 4979 break; 4980 #endif /* INET */ 4981 #ifdef INET6 4982 case AF_INET6: 4983 m_copyback(m, off, 4984 sizeof(struct icmp6_hdr), 4985 (caddr_t)pd->hdr.icmp6); 4986 m_copyback(m, ipoff2, sizeof(h2_6), 4987 (caddr_t)&h2_6); 4988 break; 4989 #endif /* INET6 */ 4990 } 4991 m_copyback(m, off2, sizeof(uh), (caddr_t)&uh); 4992 } 4993 4994 return (PF_PASS); 4995 break; 4996 } 4997 #ifdef INET 4998 case IPPROTO_ICMP: { 4999 struct icmp iih; 5000 5001 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN, 5002 NULL, reason, pd2.af)) { 5003 DPFPRINTF(PF_DEBUG_MISC, 5004 ("pf: ICMP error message too short i" 5005 "(icmp)\n")); 5006 return (PF_DROP); 5007 } 5008 5009 key.af = pd2.af; 5010 key.proto = IPPROTO_ICMP; 5011 PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af); 5012 PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af); 5013 key.port[0] = key.port[1] = iih.icmp_id; 5014 5015 STATE_LOOKUP(kif, &key, direction, *state, m); 5016 5017 /* translate source/destination address, if necessary */ 5018 if ((*state)->key[PF_SK_WIRE] != 5019 (*state)->key[PF_SK_STACK]) { 5020 struct pf_state_key *nk = 5021 (*state)->key[pd->didx]; 5022 5023 if (PF_ANEQ(pd2.src, 5024 &nk->addr[pd2.sidx], pd2.af) || 5025 nk->port[pd2.sidx] != iih.icmp_id) 5026 pf_change_icmp(pd2.src, &iih.icmp_id, 5027 daddr, &nk->addr[pd2.sidx], 5028 nk->port[pd2.sidx], NULL, 5029 pd2.ip_sum, icmpsum, 5030 pd->ip_sum, 0, AF_INET); 5031 5032 if (PF_ANEQ(pd2.dst, 5033 &nk->addr[pd2.didx], pd2.af) || 5034 nk->port[pd2.didx] != iih.icmp_id) 5035 pf_change_icmp(pd2.dst, &iih.icmp_id, 5036 NULL, /* XXX Inbound NAT? */ 5037 &nk->addr[pd2.didx], 5038 nk->port[pd2.didx], NULL, 5039 pd2.ip_sum, icmpsum, 5040 pd->ip_sum, 0, AF_INET); 5041 5042 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp); 5043 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2); 5044 m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih); 5045 } 5046 return (PF_PASS); 5047 break; 5048 } 5049 #endif /* INET */ 5050 #ifdef INET6 5051 case IPPROTO_ICMPV6: { 5052 struct icmp6_hdr iih; 5053 5054 if (!pf_pull_hdr(m, off2, &iih, 5055 sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) { 5056 DPFPRINTF(PF_DEBUG_MISC, 5057 ("pf: ICMP error message too short " 5058 "(icmp6)\n")); 5059 return (PF_DROP); 5060 } 5061 5062 key.af = pd2.af; 5063 key.proto = IPPROTO_ICMPV6; 5064 PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af); 5065 PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af); 5066 key.port[0] = key.port[1] = iih.icmp6_id; 5067 5068 STATE_LOOKUP(kif, &key, direction, *state, m); 5069 5070 /* translate source/destination address, if necessary */ 5071 if ((*state)->key[PF_SK_WIRE] != 5072 (*state)->key[PF_SK_STACK]) { 5073 struct pf_state_key *nk = 5074 (*state)->key[pd->didx]; 5075 5076 if (PF_ANEQ(pd2.src, 5077 &nk->addr[pd2.sidx], pd2.af) || 5078 nk->port[pd2.sidx] != iih.icmp6_id) 5079 pf_change_icmp(pd2.src, &iih.icmp6_id, 5080 daddr, &nk->addr[pd2.sidx], 5081 nk->port[pd2.sidx], NULL, 5082 pd2.ip_sum, icmpsum, 5083 pd->ip_sum, 0, AF_INET6); 5084 5085 if (PF_ANEQ(pd2.dst, 5086 &nk->addr[pd2.didx], pd2.af) || 5087 nk->port[pd2.didx] != iih.icmp6_id) 5088 pf_change_icmp(pd2.dst, &iih.icmp6_id, 5089 NULL, /* XXX Inbound NAT? */ 5090 &nk->addr[pd2.didx], 5091 nk->port[pd2.didx], NULL, 5092 pd2.ip_sum, icmpsum, 5093 pd->ip_sum, 0, AF_INET6); 5094 5095 m_copyback(m, off, sizeof(struct icmp6_hdr), 5096 (caddr_t)pd->hdr.icmp6); 5097 m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6); 5098 m_copyback(m, off2, sizeof(struct icmp6_hdr), 5099 (caddr_t)&iih); 5100 } 5101 5102 return (PF_PASS); 5103 break; 5104 } 5105 #endif /* INET6 */ 5106 default: { 5107 key.af = pd2.af; 5108 key.proto = pd2.proto; 5109 PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af); 5110 PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af); 5111 key.port[0] = key.port[1] = 0; 5112 5113 STATE_LOOKUP(kif, &key, direction, *state, m); 5114 5115 /* translate source/destination address, if necessary */ 5116 if ((*state)->key[PF_SK_WIRE] != 5117 (*state)->key[PF_SK_STACK]) { 5118 struct pf_state_key *nk = 5119 (*state)->key[pd->didx]; 5120 5121 if (PF_ANEQ(pd2.src, 5122 &nk->addr[pd2.sidx], pd2.af)) 5123 pf_change_icmp(pd2.src, NULL, daddr, 5124 &nk->addr[pd2.sidx], 0, NULL, 5125 pd2.ip_sum, icmpsum, 5126 pd->ip_sum, 0, pd2.af); 5127 5128 if (PF_ANEQ(pd2.dst, 5129 &nk->addr[pd2.didx], pd2.af)) 5130 pf_change_icmp(pd2.src, NULL, 5131 NULL, /* XXX Inbound NAT? */ 5132 &nk->addr[pd2.didx], 0, NULL, 5133 pd2.ip_sum, icmpsum, 5134 pd->ip_sum, 0, pd2.af); 5135 5136 switch (pd2.af) { 5137 #ifdef INET 5138 case AF_INET: 5139 m_copyback(m, off, ICMP_MINLEN, 5140 (caddr_t)pd->hdr.icmp); 5141 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2); 5142 break; 5143 #endif /* INET */ 5144 #ifdef INET6 5145 case AF_INET6: 5146 m_copyback(m, off, 5147 sizeof(struct icmp6_hdr), 5148 (caddr_t)pd->hdr.icmp6); 5149 m_copyback(m, ipoff2, sizeof(h2_6), 5150 (caddr_t)&h2_6); 5151 break; 5152 #endif /* INET6 */ 5153 } 5154 } 5155 return (PF_PASS); 5156 break; 5157 } 5158 } 5159 } 5160 } 5161 5162 int 5163 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif, 5164 struct mbuf *m, struct pf_pdesc *pd) 5165 { 5166 struct pf_state_peer *src, *dst; 5167 struct pf_state_key_cmp key; 5168 5169 key.af = pd->af; 5170 key.proto = pd->proto; 5171 if (direction == PF_IN) { 5172 PF_ACPY(&key.addr[0], pd->src, key.af); 5173 PF_ACPY(&key.addr[1], pd->dst, key.af); 5174 key.port[0] = key.port[1] = 0; 5175 } else { 5176 PF_ACPY(&key.addr[1], pd->src, key.af); 5177 PF_ACPY(&key.addr[0], pd->dst, key.af); 5178 key.port[1] = key.port[0] = 0; 5179 } 5180 5181 STATE_LOOKUP(kif, &key, direction, *state, m); 5182 5183 if (direction == (*state)->direction) { 5184 src = &(*state)->src; 5185 dst = &(*state)->dst; 5186 } else { 5187 src = &(*state)->dst; 5188 dst = &(*state)->src; 5189 } 5190 5191 /* update states */ 5192 if (src->state < PFOTHERS_SINGLE) 5193 src->state = PFOTHERS_SINGLE; 5194 if (dst->state == PFOTHERS_SINGLE) 5195 dst->state = PFOTHERS_MULTIPLE; 5196 5197 /* update expire time */ 5198 (*state)->expire = time_second; 5199 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE) 5200 (*state)->timeout = PFTM_OTHER_MULTIPLE; 5201 else 5202 (*state)->timeout = PFTM_OTHER_SINGLE; 5203 5204 /* translate source/destination address, if necessary */ 5205 if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) { 5206 struct pf_state_key *nk = (*state)->key[pd->didx]; 5207 5208 KKASSERT(nk); 5209 KKASSERT(pd); 5210 KKASSERT(pd->src); 5211 KKASSERT(pd->dst); 5212 switch (pd->af) { 5213 #ifdef INET 5214 case AF_INET: 5215 if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], AF_INET)) 5216 pf_change_a(&pd->src->v4.s_addr, 5217 pd->ip_sum, 5218 nk->addr[pd->sidx].v4.s_addr, 5219 0); 5220 5221 5222 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], AF_INET)) 5223 pf_change_a(&pd->dst->v4.s_addr, 5224 pd->ip_sum, 5225 nk->addr[pd->didx].v4.s_addr, 5226 0); 5227 5228 break; 5229 #endif /* INET */ 5230 #ifdef INET6 5231 case AF_INET6: 5232 if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], AF_INET)) 5233 PF_ACPY(pd->src, &nk->addr[pd->sidx], pd->af); 5234 5235 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], AF_INET)) 5236 PF_ACPY(pd->dst, &nk->addr[pd->didx], pd->af); 5237 #endif /* INET6 */ 5238 } 5239 } 5240 return (PF_PASS); 5241 } 5242 5243 /* 5244 * ipoff and off are measured from the start of the mbuf chain. 5245 * h must be at "ipoff" on the mbuf chain. 5246 */ 5247 void * 5248 pf_pull_hdr(struct mbuf *m, int off, void *p, int len, 5249 u_short *actionp, u_short *reasonp, sa_family_t af) 5250 { 5251 switch (af) { 5252 #ifdef INET 5253 case AF_INET: { 5254 struct ip *h = mtod(m, struct ip *); 5255 u_int16_t fragoff = (h->ip_off & IP_OFFMASK) << 3; 5256 5257 if (fragoff) { 5258 if (fragoff >= len) 5259 ACTION_SET(actionp, PF_PASS); 5260 else { 5261 ACTION_SET(actionp, PF_DROP); 5262 REASON_SET(reasonp, PFRES_FRAG); 5263 } 5264 return (NULL); 5265 } 5266 if (m->m_pkthdr.len < off + len || 5267 h->ip_len < off + len) { 5268 ACTION_SET(actionp, PF_DROP); 5269 REASON_SET(reasonp, PFRES_SHORT); 5270 return (NULL); 5271 } 5272 break; 5273 } 5274 #endif /* INET */ 5275 #ifdef INET6 5276 case AF_INET6: { 5277 struct ip6_hdr *h = mtod(m, struct ip6_hdr *); 5278 5279 if (m->m_pkthdr.len < off + len || 5280 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) < 5281 (unsigned)(off + len)) { 5282 ACTION_SET(actionp, PF_DROP); 5283 REASON_SET(reasonp, PFRES_SHORT); 5284 return (NULL); 5285 } 5286 break; 5287 } 5288 #endif /* INET6 */ 5289 } 5290 m_copydata(m, off, len, p); 5291 return (p); 5292 } 5293 5294 int 5295 pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif) 5296 { 5297 struct sockaddr_in *dst; 5298 int ret = 1; 5299 int check_mpath; 5300 #ifdef INET6 5301 struct sockaddr_in6 *dst6; 5302 struct route_in6 ro; 5303 #else 5304 struct route ro; 5305 #endif 5306 struct radix_node *rn; 5307 struct rtentry *rt; 5308 struct ifnet *ifp; 5309 5310 check_mpath = 0; 5311 bzero(&ro, sizeof(ro)); 5312 switch (af) { 5313 case AF_INET: 5314 dst = satosin(&ro.ro_dst); 5315 dst->sin_family = AF_INET; 5316 dst->sin_len = sizeof(*dst); 5317 dst->sin_addr = addr->v4; 5318 break; 5319 #ifdef INET6 5320 case AF_INET6: 5321 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 5322 dst6->sin6_family = AF_INET6; 5323 dst6->sin6_len = sizeof(*dst6); 5324 dst6->sin6_addr = addr->v6; 5325 break; 5326 #endif /* INET6 */ 5327 default: 5328 return (0); 5329 } 5330 5331 /* Skip checks for ipsec interfaces */ 5332 if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC) 5333 goto out; 5334 5335 rtalloc_ign((struct route *)&ro, 0); 5336 5337 if (ro.ro_rt != NULL) { 5338 /* No interface given, this is a no-route check */ 5339 if (kif == NULL) 5340 goto out; 5341 5342 if (kif->pfik_ifp == NULL) { 5343 ret = 0; 5344 goto out; 5345 } 5346 5347 /* Perform uRPF check if passed input interface */ 5348 ret = 0; 5349 rn = (struct radix_node *)ro.ro_rt; 5350 do { 5351 rt = (struct rtentry *)rn; 5352 ifp = rt->rt_ifp; 5353 5354 if (kif->pfik_ifp == ifp) 5355 ret = 1; 5356 rn = NULL; 5357 } while (check_mpath == 1 && rn != NULL && ret == 0); 5358 } else 5359 ret = 0; 5360 out: 5361 if (ro.ro_rt != NULL) 5362 RTFREE(ro.ro_rt); 5363 return (ret); 5364 } 5365 5366 int 5367 pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw) 5368 { 5369 struct sockaddr_in *dst; 5370 #ifdef INET6 5371 struct sockaddr_in6 *dst6; 5372 struct route_in6 ro; 5373 #else 5374 struct route ro; 5375 #endif 5376 int ret = 0; 5377 5378 ASSERT_LWKT_TOKEN_HELD(&pf_token); 5379 5380 bzero(&ro, sizeof(ro)); 5381 switch (af) { 5382 case AF_INET: 5383 dst = satosin(&ro.ro_dst); 5384 dst->sin_family = AF_INET; 5385 dst->sin_len = sizeof(*dst); 5386 dst->sin_addr = addr->v4; 5387 break; 5388 #ifdef INET6 5389 case AF_INET6: 5390 dst6 = (struct sockaddr_in6 *)&ro.ro_dst; 5391 dst6->sin6_family = AF_INET6; 5392 dst6->sin6_len = sizeof(*dst6); 5393 dst6->sin6_addr = addr->v6; 5394 break; 5395 #endif /* INET6 */ 5396 default: 5397 return (0); 5398 } 5399 5400 rtalloc_ign((struct route *)&ro, (RTF_CLONING | RTF_PRCLONING)); 5401 5402 if (ro.ro_rt != NULL) { 5403 RTFREE(ro.ro_rt); 5404 } 5405 5406 return (ret); 5407 } 5408 5409 #ifdef INET 5410 void 5411 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp, 5412 struct pf_state *s, struct pf_pdesc *pd) 5413 { 5414 struct mbuf *m0, *m1; 5415 struct route iproute; 5416 struct route *ro = NULL; 5417 struct sockaddr_in *dst; 5418 struct ip *ip; 5419 struct ifnet *ifp = NULL; 5420 struct pf_addr naddr; 5421 struct pf_src_node *sn = NULL; 5422 int error = 0; 5423 int sw_csum; 5424 #ifdef IPSEC 5425 struct m_tag *mtag; 5426 #endif /* IPSEC */ 5427 5428 ASSERT_LWKT_TOKEN_HELD(&pf_token); 5429 5430 if (m == NULL || *m == NULL || r == NULL || 5431 (dir != PF_IN && dir != PF_OUT) || oifp == NULL) 5432 panic("pf_route: invalid parameters"); 5433 5434 if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) { 5435 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED; 5436 (*m)->m_pkthdr.pf.routed = 1; 5437 } else { 5438 if ((*m)->m_pkthdr.pf.routed++ > 3) { 5439 m0 = *m; 5440 *m = NULL; 5441 goto bad; 5442 } 5443 } 5444 5445 if (r->rt == PF_DUPTO) { 5446 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL) { 5447 return; 5448 } 5449 } else { 5450 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) { 5451 return; 5452 } 5453 m0 = *m; 5454 } 5455 5456 if (m0->m_len < sizeof(struct ip)) { 5457 DPFPRINTF(PF_DEBUG_URGENT, 5458 ("pf_route: m0->m_len < sizeof(struct ip)\n")); 5459 goto bad; 5460 } 5461 5462 ip = mtod(m0, struct ip *); 5463 5464 ro = &iproute; 5465 bzero((caddr_t)ro, sizeof(*ro)); 5466 dst = satosin(&ro->ro_dst); 5467 dst->sin_family = AF_INET; 5468 dst->sin_len = sizeof(*dst); 5469 dst->sin_addr = ip->ip_dst; 5470 5471 if (r->rt == PF_FASTROUTE) { 5472 rtalloc(ro); 5473 if (ro->ro_rt == 0) { 5474 ipstat.ips_noroute++; 5475 goto bad; 5476 } 5477 5478 ifp = ro->ro_rt->rt_ifp; 5479 ro->ro_rt->rt_use++; 5480 5481 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 5482 dst = satosin(ro->ro_rt->rt_gateway); 5483 } else { 5484 if (TAILQ_EMPTY(&r->rpool.list)) { 5485 DPFPRINTF(PF_DEBUG_URGENT, 5486 ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n")); 5487 goto bad; 5488 } 5489 if (s == NULL) { 5490 pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src, 5491 &naddr, NULL, &sn); 5492 if (!PF_AZERO(&naddr, AF_INET)) 5493 dst->sin_addr.s_addr = naddr.v4.s_addr; 5494 ifp = r->rpool.cur->kif ? 5495 r->rpool.cur->kif->pfik_ifp : NULL; 5496 } else { 5497 if (!PF_AZERO(&s->rt_addr, AF_INET)) 5498 dst->sin_addr.s_addr = 5499 s->rt_addr.v4.s_addr; 5500 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; 5501 } 5502 } 5503 if (ifp == NULL) 5504 goto bad; 5505 5506 if (oifp != ifp) { 5507 if (pf_test(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) { 5508 goto bad; 5509 } else if (m0 == NULL) { 5510 goto done; 5511 } 5512 if (m0->m_len < sizeof(struct ip)) { 5513 DPFPRINTF(PF_DEBUG_URGENT, 5514 ("pf_route: m0->m_len < sizeof(struct ip)\n")); 5515 goto bad; 5516 } 5517 ip = mtod(m0, struct ip *); 5518 } 5519 5520 /* Copied from FreeBSD 5.1-CURRENT ip_output. */ 5521 m0->m_pkthdr.csum_flags |= CSUM_IP; 5522 sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist; 5523 if (sw_csum & CSUM_DELAY_DATA) { 5524 in_delayed_cksum(m0); 5525 sw_csum &= ~CSUM_DELAY_DATA; 5526 } 5527 m0->m_pkthdr.csum_flags &= ifp->if_hwassist; 5528 m0->m_pkthdr.csum_iphlen = (ip->ip_hl << 2); 5529 5530 if (ip->ip_len <= ifp->if_mtu || 5531 (ifp->if_hwassist & CSUM_FRAGMENT && 5532 (ip->ip_off & IP_DF) == 0)) { 5533 ip->ip_len = htons(ip->ip_len); 5534 ip->ip_off = htons(ip->ip_off); 5535 ip->ip_sum = 0; 5536 if (sw_csum & CSUM_DELAY_IP) { 5537 /* From KAME */ 5538 if (ip->ip_v == IPVERSION && 5539 (ip->ip_hl << 2) == sizeof(*ip)) { 5540 ip->ip_sum = in_cksum_hdr(ip); 5541 } else { 5542 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2); 5543 } 5544 } 5545 lwkt_reltoken(&pf_token); 5546 error = ifp->if_output(ifp, m0, sintosa(dst), ro->ro_rt); 5547 lwkt_gettoken(&pf_token); 5548 goto done; 5549 } 5550 5551 /* 5552 * Too large for interface; fragment if possible. 5553 * Must be able to put at least 8 bytes per fragment. 5554 */ 5555 if (ip->ip_off & IP_DF) { 5556 ipstat.ips_cantfrag++; 5557 if (r->rt != PF_DUPTO) { 5558 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0, 5559 ifp->if_mtu); 5560 goto done; 5561 } else 5562 goto bad; 5563 } 5564 5565 m1 = m0; 5566 error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum); 5567 if (error) { 5568 goto bad; 5569 } 5570 5571 for (m0 = m1; m0; m0 = m1) { 5572 m1 = m0->m_nextpkt; 5573 m0->m_nextpkt = 0; 5574 if (error == 0) { 5575 lwkt_reltoken(&pf_token); 5576 error = (*ifp->if_output)(ifp, m0, sintosa(dst), 5577 NULL); 5578 lwkt_gettoken(&pf_token); 5579 } else 5580 m_freem(m0); 5581 } 5582 5583 if (error == 0) 5584 ipstat.ips_fragmented++; 5585 5586 done: 5587 if (r->rt != PF_DUPTO) 5588 *m = NULL; 5589 if (ro == &iproute && ro->ro_rt) 5590 RTFREE(ro->ro_rt); 5591 return; 5592 5593 bad: 5594 m_freem(m0); 5595 goto done; 5596 } 5597 #endif /* INET */ 5598 5599 #ifdef INET6 5600 void 5601 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp, 5602 struct pf_state *s, struct pf_pdesc *pd) 5603 { 5604 struct mbuf *m0; 5605 struct route_in6 ip6route; 5606 struct route_in6 *ro; 5607 struct sockaddr_in6 *dst; 5608 struct ip6_hdr *ip6; 5609 struct ifnet *ifp = NULL; 5610 struct pf_addr naddr; 5611 struct pf_src_node *sn = NULL; 5612 int error = 0; 5613 5614 if (m == NULL || *m == NULL || r == NULL || 5615 (dir != PF_IN && dir != PF_OUT) || oifp == NULL) 5616 panic("pf_route6: invalid parameters"); 5617 5618 if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) { 5619 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED; 5620 (*m)->m_pkthdr.pf.routed = 1; 5621 } else { 5622 if ((*m)->m_pkthdr.pf.routed++ > 3) { 5623 m0 = *m; 5624 *m = NULL; 5625 goto bad; 5626 } 5627 } 5628 5629 if (r->rt == PF_DUPTO) { 5630 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL) 5631 return; 5632 } else { 5633 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) 5634 return; 5635 m0 = *m; 5636 } 5637 5638 if (m0->m_len < sizeof(struct ip6_hdr)) { 5639 DPFPRINTF(PF_DEBUG_URGENT, 5640 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n")); 5641 goto bad; 5642 } 5643 ip6 = mtod(m0, struct ip6_hdr *); 5644 5645 ro = &ip6route; 5646 bzero((caddr_t)ro, sizeof(*ro)); 5647 dst = (struct sockaddr_in6 *)&ro->ro_dst; 5648 dst->sin6_family = AF_INET6; 5649 dst->sin6_len = sizeof(*dst); 5650 dst->sin6_addr = ip6->ip6_dst; 5651 5652 /* 5653 * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags, 5654 * so make sure pf.flags is clear. 5655 * 5656 * Cheat. XXX why only in the v6 case??? 5657 */ 5658 if (r->rt == PF_FASTROUTE) { 5659 m0->m_pkthdr.fw_flags |= PF_MBUF_TAGGED; 5660 m0->m_pkthdr.pf.flags = 0; 5661 /* XXX Re-Check when Upgrading to > 4.4 */ 5662 m0->m_pkthdr.pf.statekey = NULL; 5663 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL); 5664 return; 5665 } 5666 5667 if (TAILQ_EMPTY(&r->rpool.list)) { 5668 DPFPRINTF(PF_DEBUG_URGENT, 5669 ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n")); 5670 goto bad; 5671 } 5672 if (s == NULL) { 5673 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src, 5674 &naddr, NULL, &sn); 5675 if (!PF_AZERO(&naddr, AF_INET6)) 5676 PF_ACPY((struct pf_addr *)&dst->sin6_addr, 5677 &naddr, AF_INET6); 5678 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL; 5679 } else { 5680 if (!PF_AZERO(&s->rt_addr, AF_INET6)) 5681 PF_ACPY((struct pf_addr *)&dst->sin6_addr, 5682 &s->rt_addr, AF_INET6); 5683 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; 5684 } 5685 if (ifp == NULL) 5686 goto bad; 5687 5688 if (oifp != ifp) { 5689 if (pf_test6(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) { 5690 goto bad; 5691 } else if (m0 == NULL) { 5692 goto done; 5693 } 5694 if (m0->m_len < sizeof(struct ip6_hdr)) { 5695 DPFPRINTF(PF_DEBUG_URGENT, 5696 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n")); 5697 goto bad; 5698 } 5699 ip6 = mtod(m0, struct ip6_hdr *); 5700 } 5701 5702 /* 5703 * If the packet is too large for the outgoing interface, 5704 * send back an icmp6 error. 5705 */ 5706 if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr)) 5707 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index); 5708 if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) { 5709 error = nd6_output(ifp, ifp, m0, dst, NULL); 5710 } else { 5711 in6_ifstat_inc(ifp, ifs6_in_toobig); 5712 if (r->rt != PF_DUPTO) 5713 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu); 5714 else 5715 goto bad; 5716 } 5717 5718 done: 5719 if (r->rt != PF_DUPTO) 5720 *m = NULL; 5721 return; 5722 5723 bad: 5724 m_freem(m0); 5725 goto done; 5726 } 5727 #endif /* INET6 */ 5728 5729 5730 /* 5731 * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag 5732 * off is the offset where the protocol header starts 5733 * len is the total length of protocol header plus payload 5734 * returns 0 when the checksum is valid, otherwise returns 1. 5735 */ 5736 /* 5737 * XXX 5738 * FreeBSD supports cksum offload for the following drivers. 5739 * em(4), gx(4), lge(4), nge(4), ti(4), xl(4) 5740 * If we can make full use of it we would outperform ipfw/ipfilter in 5741 * very heavy traffic. 5742 * I have not tested 'cause I don't have NICs that supports cksum offload. 5743 * (There might be problems. Typical phenomena would be 5744 * 1. No route message for UDP packet. 5745 * 2. No connection acceptance from external hosts regardless of rule set.) 5746 */ 5747 int 5748 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p, 5749 sa_family_t af) 5750 { 5751 u_int16_t sum = 0; 5752 int hw_assist = 0; 5753 struct ip *ip; 5754 5755 if (off < sizeof(struct ip) || len < sizeof(struct udphdr)) 5756 return (1); 5757 if (m->m_pkthdr.len < off + len) 5758 return (1); 5759 5760 switch (p) { 5761 case IPPROTO_TCP: 5762 case IPPROTO_UDP: 5763 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { 5764 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) { 5765 sum = m->m_pkthdr.csum_data; 5766 } else { 5767 ip = mtod(m, struct ip *); 5768 sum = in_pseudo(ip->ip_src.s_addr, 5769 ip->ip_dst.s_addr, htonl((u_short)len + 5770 m->m_pkthdr.csum_data + p)); 5771 } 5772 sum ^= 0xffff; 5773 ++hw_assist; 5774 } 5775 break; 5776 case IPPROTO_ICMP: 5777 #ifdef INET6 5778 case IPPROTO_ICMPV6: 5779 #endif /* INET6 */ 5780 break; 5781 default: 5782 return (1); 5783 } 5784 5785 if (!hw_assist) { 5786 switch (af) { 5787 case AF_INET: 5788 if (p == IPPROTO_ICMP) { 5789 if (m->m_len < off) 5790 return (1); 5791 m->m_data += off; 5792 m->m_len -= off; 5793 sum = in_cksum(m, len); 5794 m->m_data -= off; 5795 m->m_len += off; 5796 } else { 5797 if (m->m_len < sizeof(struct ip)) 5798 return (1); 5799 sum = in_cksum_range(m, p, off, len); 5800 if (sum == 0) { 5801 m->m_pkthdr.csum_flags |= 5802 (CSUM_DATA_VALID | 5803 CSUM_PSEUDO_HDR); 5804 m->m_pkthdr.csum_data = 0xffff; 5805 } 5806 } 5807 break; 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 /* 5814 * XXX 5815 * IPv6 H/W cksum off-load not supported yet! 5816 * 5817 * if (sum == 0) { 5818 * m->m_pkthdr.csum_flags |= 5819 * (CSUM_DATA_VALID|CSUM_PSEUDO_HDR); 5820 * m->m_pkthdr.csum_data = 0xffff; 5821 *} 5822 */ 5823 break; 5824 #endif /* INET6 */ 5825 default: 5826 return (1); 5827 } 5828 } 5829 if (sum) { 5830 switch (p) { 5831 case IPPROTO_TCP: 5832 tcpstat.tcps_rcvbadsum++; 5833 break; 5834 case IPPROTO_UDP: 5835 udpstat.udps_badsum++; 5836 break; 5837 case IPPROTO_ICMP: 5838 icmpstat.icps_checksum++; 5839 break; 5840 #ifdef INET6 5841 case IPPROTO_ICMPV6: 5842 icmp6stat.icp6s_checksum++; 5843 break; 5844 #endif /* INET6 */ 5845 } 5846 return (1); 5847 } 5848 return (0); 5849 } 5850 5851 struct pf_divert * 5852 pf_find_divert(struct mbuf *m) 5853 { 5854 struct m_tag *mtag; 5855 5856 if ((mtag = m_tag_find(m, PACKET_TAG_PF_DIVERT, NULL)) == NULL) 5857 return (NULL); 5858 5859 return ((struct pf_divert *)(mtag + 1)); 5860 } 5861 5862 struct pf_divert * 5863 pf_get_divert(struct mbuf *m) 5864 { 5865 struct m_tag *mtag; 5866 5867 if ((mtag = m_tag_find(m, PACKET_TAG_PF_DIVERT, NULL)) == NULL) { 5868 mtag = m_tag_get(PACKET_TAG_PF_DIVERT, sizeof(struct pf_divert), 5869 M_NOWAIT); 5870 if (mtag == NULL) 5871 return (NULL); 5872 bzero(mtag + 1, sizeof(struct pf_divert)); 5873 m_tag_prepend(m, mtag); 5874 } 5875 5876 return ((struct pf_divert *)(mtag + 1)); 5877 } 5878 5879 #ifdef INET 5880 int 5881 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0, 5882 struct ether_header *eh, struct inpcb *inp) 5883 { 5884 struct pfi_kif *kif; 5885 u_short action, reason = 0, log = 0; 5886 struct mbuf *m = *m0; 5887 struct ip *h = NULL; 5888 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr; 5889 struct pf_state *s = NULL; 5890 struct pf_ruleset *ruleset = NULL; 5891 struct pf_pdesc pd; 5892 int off, dirndx, pqid = 0; 5893 5894 if (!pf_status.running) 5895 return (PF_PASS); 5896 5897 memset(&pd, 0, sizeof(pd)); 5898 #ifdef foo 5899 if (ifp->if_type == IFT_CARP && ifp->if_carpdev) 5900 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif; 5901 else 5902 #endif 5903 kif = (struct pfi_kif *)ifp->if_pf_kif; 5904 5905 if (kif == NULL) { 5906 DPFPRINTF(PF_DEBUG_URGENT, 5907 ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname)); 5908 return (PF_DROP); 5909 } 5910 if (kif->pfik_flags & PFI_IFLAG_SKIP) 5911 return (PF_PASS); 5912 5913 #ifdef DIAGNOSTIC 5914 if ((m->m_flags & M_PKTHDR) == 0) 5915 panic("non-M_PKTHDR is passed to pf_test"); 5916 #endif /* DIAGNOSTIC */ 5917 5918 if (m->m_pkthdr.len < (int)sizeof(*h)) { 5919 action = PF_DROP; 5920 REASON_SET(&reason, PFRES_SHORT); 5921 log = 1; 5922 goto done; 5923 } 5924 5925 /* 5926 * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags, 5927 * so make sure pf.flags is clear. 5928 */ 5929 if (m->m_pkthdr.fw_flags & PF_MBUF_TAGGED) 5930 return (PF_PASS); 5931 m->m_pkthdr.pf.flags = 0; 5932 /* Re-Check when updating to > 4.4 */ 5933 m->m_pkthdr.pf.statekey = NULL; 5934 5935 /* We do IP header normalization and packet reassembly here */ 5936 if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) { 5937 action = PF_DROP; 5938 goto done; 5939 } 5940 m = *m0; /* pf_normalize messes with m0 */ 5941 h = mtod(m, struct ip *); 5942 5943 off = h->ip_hl << 2; 5944 if (off < (int)sizeof(*h)) { 5945 action = PF_DROP; 5946 REASON_SET(&reason, PFRES_SHORT); 5947 log = 1; 5948 goto done; 5949 } 5950 5951 pd.src = (struct pf_addr *)&h->ip_src; 5952 pd.dst = (struct pf_addr *)&h->ip_dst; 5953 pd.sport = pd.dport = NULL; 5954 pd.ip_sum = &h->ip_sum; 5955 pd.proto_sum = NULL; 5956 pd.proto = h->ip_p; 5957 pd.dir = dir; 5958 pd.sidx = (dir == PF_IN) ? 0 : 1; 5959 pd.didx = (dir == PF_IN) ? 1 : 0; 5960 pd.af = AF_INET; 5961 pd.tos = h->ip_tos; 5962 pd.tot_len = h->ip_len; 5963 pd.eh = eh; 5964 5965 /* handle fragments that didn't get reassembled by normalization */ 5966 if (h->ip_off & (IP_MF | IP_OFFMASK)) { 5967 action = pf_test_fragment(&r, dir, kif, m, h, 5968 &pd, &a, &ruleset); 5969 goto done; 5970 } 5971 5972 switch (h->ip_p) { 5973 5974 case IPPROTO_TCP: { 5975 struct tcphdr th; 5976 5977 pd.hdr.tcp = &th; 5978 if (!pf_pull_hdr(m, off, &th, sizeof(th), 5979 &action, &reason, AF_INET)) { 5980 log = action != PF_PASS; 5981 goto done; 5982 } 5983 pd.p_len = pd.tot_len - off - (th.th_off << 2); 5984 if ((th.th_flags & TH_ACK) && pd.p_len == 0) 5985 pqid = 1; 5986 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd); 5987 if (action == PF_DROP) 5988 goto done; 5989 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd, 5990 &reason); 5991 if (action == PF_PASS) { 5992 pfsync_update_state(s); 5993 r = s->rule.ptr; 5994 a = s->anchor.ptr; 5995 log = s->log; 5996 } else if (s == NULL) 5997 action = pf_test_rule(&r, &s, dir, kif, 5998 m, off, h, &pd, &a, &ruleset, NULL, inp); 5999 break; 6000 } 6001 6002 case IPPROTO_UDP: { 6003 struct udphdr uh; 6004 6005 pd.hdr.udp = &uh; 6006 if (!pf_pull_hdr(m, off, &uh, sizeof(uh), 6007 &action, &reason, AF_INET)) { 6008 log = action != PF_PASS; 6009 goto done; 6010 } 6011 if (uh.uh_dport == 0 || 6012 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off || 6013 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) { 6014 action = PF_DROP; 6015 REASON_SET(&reason, PFRES_SHORT); 6016 goto done; 6017 } 6018 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd); 6019 if (action == PF_PASS) { 6020 pfsync_update_state(s); 6021 r = s->rule.ptr; 6022 a = s->anchor.ptr; 6023 log = s->log; 6024 } else if (s == NULL) 6025 action = pf_test_rule(&r, &s, dir, kif, 6026 m, off, h, &pd, &a, &ruleset, NULL, inp); 6027 break; 6028 } 6029 6030 case IPPROTO_ICMP: { 6031 struct icmp ih; 6032 6033 pd.hdr.icmp = &ih; 6034 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN, 6035 &action, &reason, AF_INET)) { 6036 log = action != PF_PASS; 6037 goto done; 6038 } 6039 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd, 6040 &reason); 6041 if (action == PF_PASS) { 6042 pfsync_update_state(s); 6043 r = s->rule.ptr; 6044 a = s->anchor.ptr; 6045 log = s->log; 6046 } else if (s == NULL) 6047 action = pf_test_rule(&r, &s, dir, kif, 6048 m, off, h, &pd, &a, &ruleset, NULL, inp); 6049 break; 6050 } 6051 6052 default: 6053 action = pf_test_state_other(&s, dir, kif, m, &pd); 6054 if (action == PF_PASS) { 6055 pfsync_update_state(s); 6056 r = s->rule.ptr; 6057 a = s->anchor.ptr; 6058 log = s->log; 6059 } else if (s == NULL) 6060 action = pf_test_rule(&r, &s, dir, kif, m, off, h, 6061 &pd, &a, &ruleset, NULL, inp); 6062 break; 6063 } 6064 6065 done: 6066 if (action == PF_PASS && h->ip_hl > 5 && 6067 !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) { 6068 action = PF_DROP; 6069 REASON_SET(&reason, PFRES_IPOPTIONS); 6070 log = 1; 6071 DPFPRINTF(PF_DEBUG_MISC, 6072 ("pf: dropping packet with ip options\n")); 6073 } 6074 6075 if ((s && s->tag) || r->rtableid) 6076 pf_tag_packet(m, s ? s->tag : 0, r->rtableid); 6077 6078 #if 0 6079 if (dir == PF_IN && s && s->key[PF_SK_STACK]) 6080 m->m_pkthdr.pf.statekey = s->key[PF_SK_STACK]; 6081 #endif 6082 6083 #ifdef ALTQ 6084 if (action == PF_PASS && r->qid) { 6085 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE; 6086 if (pqid || (pd.tos & IPTOS_LOWDELAY)) 6087 m->m_pkthdr.pf.qid = r->pqid; 6088 else 6089 m->m_pkthdr.pf.qid = r->qid; 6090 m->m_pkthdr.pf.ecn_af = AF_INET; 6091 m->m_pkthdr.pf.hdr = h; 6092 /* add connection hash for fairq */ 6093 if (s) { 6094 /* for fairq */ 6095 m->m_pkthdr.pf.state_hash = s->hash; 6096 m->m_pkthdr.pf.flags |= PF_TAG_STATE_HASHED; 6097 } 6098 } 6099 #endif /* ALTQ */ 6100 6101 /* 6102 * connections redirected to loopback should not match sockets 6103 * bound specifically to loopback due to security implications, 6104 * see tcp_input() and in_pcblookup_listen(). 6105 */ 6106 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP || 6107 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL && 6108 (s->nat_rule.ptr->action == PF_RDR || 6109 s->nat_rule.ptr->action == PF_BINAT) && 6110 (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) 6111 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST; 6112 6113 if (dir == PF_IN && action == PF_PASS && r->divert.port) { 6114 struct pf_divert *divert; 6115 6116 if ((divert = pf_get_divert(m))) { 6117 m->m_pkthdr.pf.flags |= PF_TAG_DIVERTED; 6118 divert->port = r->divert.port; 6119 divert->addr.ipv4 = r->divert.addr.v4; 6120 } 6121 } 6122 6123 if (log) { 6124 struct pf_rule *lr; 6125 6126 if (s != NULL && s->nat_rule.ptr != NULL && 6127 s->nat_rule.ptr->log & PF_LOG_ALL) 6128 lr = s->nat_rule.ptr; 6129 else 6130 lr = r; 6131 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset, 6132 &pd); 6133 } 6134 6135 kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len; 6136 kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++; 6137 6138 if (action == PF_PASS || r->action == PF_DROP) { 6139 dirndx = (dir == PF_OUT); 6140 r->packets[dirndx]++; 6141 r->bytes[dirndx] += pd.tot_len; 6142 if (a != NULL) { 6143 a->packets[dirndx]++; 6144 a->bytes[dirndx] += pd.tot_len; 6145 } 6146 if (s != NULL) { 6147 if (s->nat_rule.ptr != NULL) { 6148 s->nat_rule.ptr->packets[dirndx]++; 6149 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len; 6150 } 6151 if (s->src_node != NULL) { 6152 s->src_node->packets[dirndx]++; 6153 s->src_node->bytes[dirndx] += pd.tot_len; 6154 } 6155 if (s->nat_src_node != NULL) { 6156 s->nat_src_node->packets[dirndx]++; 6157 s->nat_src_node->bytes[dirndx] += pd.tot_len; 6158 } 6159 dirndx = (dir == s->direction) ? 0 : 1; 6160 s->packets[dirndx]++; 6161 s->bytes[dirndx] += pd.tot_len; 6162 } 6163 tr = r; 6164 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule; 6165 if (nr != NULL && r == &pf_default_rule) 6166 tr = nr; 6167 if (tr->src.addr.type == PF_ADDR_TABLE) 6168 pfr_update_stats(tr->src.addr.p.tbl, 6169 (s == NULL) ? pd.src : 6170 &s->key[(s->direction == PF_IN)]-> 6171 addr[(s->direction == PF_OUT)], 6172 pd.af, pd.tot_len, dir == PF_OUT, 6173 r->action == PF_PASS, tr->src.neg); 6174 if (tr->dst.addr.type == PF_ADDR_TABLE) 6175 pfr_update_stats(tr->dst.addr.p.tbl, 6176 (s == NULL) ? pd.dst : 6177 &s->key[(s->direction == PF_IN)]-> 6178 addr[(s->direction == PF_IN)], 6179 pd.af, pd.tot_len, dir == PF_OUT, 6180 r->action == PF_PASS, tr->dst.neg); 6181 } 6182 6183 6184 if (action == PF_SYNPROXY_DROP) { 6185 m_freem(*m0); 6186 *m0 = NULL; 6187 action = PF_PASS; 6188 } else if (r->rt) 6189 /* pf_route can free the mbuf causing *m0 to become NULL */ 6190 pf_route(m0, r, dir, kif->pfik_ifp, s, &pd); 6191 6192 return (action); 6193 } 6194 #endif /* INET */ 6195 6196 #ifdef INET6 6197 int 6198 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0, 6199 struct ether_header *eh, struct inpcb *inp) 6200 { 6201 struct pfi_kif *kif; 6202 u_short action, reason = 0, log = 0; 6203 struct mbuf *m = *m0, *n = NULL; 6204 struct ip6_hdr *h = NULL; 6205 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr; 6206 struct pf_state *s = NULL; 6207 struct pf_ruleset *ruleset = NULL; 6208 struct pf_pdesc pd; 6209 int off, terminal = 0, dirndx, rh_cnt = 0; 6210 6211 if (!pf_status.running) 6212 return (PF_PASS); 6213 6214 memset(&pd, 0, sizeof(pd)); 6215 #ifdef foo 6216 if (ifp->if_type == IFT_CARP && ifp->if_carpdev) 6217 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif; 6218 else 6219 #endif 6220 kif = (struct pfi_kif *)ifp->if_pf_kif; 6221 6222 if (kif == NULL) { 6223 DPFPRINTF(PF_DEBUG_URGENT, 6224 ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname)); 6225 return (PF_DROP); 6226 } 6227 if (kif->pfik_flags & PFI_IFLAG_SKIP) 6228 return (PF_PASS); 6229 6230 #ifdef DIAGNOSTIC 6231 if ((m->m_flags & M_PKTHDR) == 0) 6232 panic("non-M_PKTHDR is passed to pf_test6"); 6233 #endif /* DIAGNOSTIC */ 6234 6235 if (m->m_pkthdr.len < (int)sizeof(*h)) { 6236 action = PF_DROP; 6237 REASON_SET(&reason, PFRES_SHORT); 6238 log = 1; 6239 goto done; 6240 } 6241 6242 /* 6243 * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags, 6244 * so make sure pf.flags is clear. 6245 */ 6246 if (m->m_pkthdr.fw_flags & PF_MBUF_TAGGED) 6247 return (PF_PASS); 6248 m->m_pkthdr.pf.flags = 0; 6249 /* Re-Check when updating to > 4.4 */ 6250 m->m_pkthdr.pf.statekey = NULL; 6251 6252 /* We do IP header normalization and packet reassembly here */ 6253 if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) { 6254 action = PF_DROP; 6255 goto done; 6256 } 6257 m = *m0; /* pf_normalize messes with m0 */ 6258 h = mtod(m, struct ip6_hdr *); 6259 6260 #if 1 6261 /* 6262 * we do not support jumbogram yet. if we keep going, zero ip6_plen 6263 * will do something bad, so drop the packet for now. 6264 */ 6265 if (htons(h->ip6_plen) == 0) { 6266 action = PF_DROP; 6267 REASON_SET(&reason, PFRES_NORM); /*XXX*/ 6268 goto done; 6269 } 6270 #endif 6271 6272 pd.src = (struct pf_addr *)&h->ip6_src; 6273 pd.dst = (struct pf_addr *)&h->ip6_dst; 6274 pd.sport = pd.dport = NULL; 6275 pd.ip_sum = NULL; 6276 pd.proto_sum = NULL; 6277 pd.dir = dir; 6278 pd.sidx = (dir == PF_IN) ? 0 : 1; 6279 pd.didx = (dir == PF_IN) ? 1 : 0; 6280 pd.af = AF_INET6; 6281 pd.tos = 0; 6282 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr); 6283 pd.eh = eh; 6284 6285 off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr); 6286 pd.proto = h->ip6_nxt; 6287 do { 6288 switch (pd.proto) { 6289 case IPPROTO_FRAGMENT: 6290 action = pf_test_fragment(&r, dir, kif, m, h, 6291 &pd, &a, &ruleset); 6292 if (action == PF_DROP) 6293 REASON_SET(&reason, PFRES_FRAG); 6294 goto done; 6295 case IPPROTO_ROUTING: { 6296 struct ip6_rthdr rthdr; 6297 6298 if (rh_cnt++) { 6299 DPFPRINTF(PF_DEBUG_MISC, 6300 ("pf: IPv6 more than one rthdr\n")); 6301 action = PF_DROP; 6302 REASON_SET(&reason, PFRES_IPOPTIONS); 6303 log = 1; 6304 goto done; 6305 } 6306 if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL, 6307 &reason, pd.af)) { 6308 DPFPRINTF(PF_DEBUG_MISC, 6309 ("pf: IPv6 short rthdr\n")); 6310 action = PF_DROP; 6311 REASON_SET(&reason, PFRES_SHORT); 6312 log = 1; 6313 goto done; 6314 } 6315 if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) { 6316 DPFPRINTF(PF_DEBUG_MISC, 6317 ("pf: IPv6 rthdr0\n")); 6318 action = PF_DROP; 6319 REASON_SET(&reason, PFRES_IPOPTIONS); 6320 log = 1; 6321 goto done; 6322 } 6323 /* FALLTHROUGH */ 6324 } 6325 case IPPROTO_AH: 6326 case IPPROTO_HOPOPTS: 6327 case IPPROTO_DSTOPTS: { 6328 /* get next header and header length */ 6329 struct ip6_ext opt6; 6330 6331 if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6), 6332 NULL, &reason, pd.af)) { 6333 DPFPRINTF(PF_DEBUG_MISC, 6334 ("pf: IPv6 short opt\n")); 6335 action = PF_DROP; 6336 log = 1; 6337 goto done; 6338 } 6339 if (pd.proto == IPPROTO_AH) 6340 off += (opt6.ip6e_len + 2) * 4; 6341 else 6342 off += (opt6.ip6e_len + 1) * 8; 6343 pd.proto = opt6.ip6e_nxt; 6344 /* goto the next header */ 6345 break; 6346 } 6347 default: 6348 terminal++; 6349 break; 6350 } 6351 } while (!terminal); 6352 6353 /* if there's no routing header, use unmodified mbuf for checksumming */ 6354 if (!n) 6355 n = m; 6356 6357 switch (pd.proto) { 6358 6359 case IPPROTO_TCP: { 6360 struct tcphdr th; 6361 6362 pd.hdr.tcp = &th; 6363 if (!pf_pull_hdr(m, off, &th, sizeof(th), 6364 &action, &reason, AF_INET6)) { 6365 log = action != PF_PASS; 6366 goto done; 6367 } 6368 pd.p_len = pd.tot_len - off - (th.th_off << 2); 6369 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd); 6370 if (action == PF_DROP) 6371 goto done; 6372 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd, 6373 &reason); 6374 if (action == PF_PASS) { 6375 pfsync_update_state(s); 6376 r = s->rule.ptr; 6377 a = s->anchor.ptr; 6378 log = s->log; 6379 } else if (s == NULL) 6380 action = pf_test_rule(&r, &s, dir, kif, 6381 m, off, h, &pd, &a, &ruleset, NULL, inp); 6382 break; 6383 } 6384 6385 case IPPROTO_UDP: { 6386 struct udphdr uh; 6387 6388 pd.hdr.udp = &uh; 6389 if (!pf_pull_hdr(m, off, &uh, sizeof(uh), 6390 &action, &reason, AF_INET6)) { 6391 log = action != PF_PASS; 6392 goto done; 6393 } 6394 if (uh.uh_dport == 0 || 6395 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off || 6396 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) { 6397 action = PF_DROP; 6398 REASON_SET(&reason, PFRES_SHORT); 6399 goto done; 6400 } 6401 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd); 6402 if (action == PF_PASS) { 6403 pfsync_update_state(s); 6404 r = s->rule.ptr; 6405 a = s->anchor.ptr; 6406 log = s->log; 6407 } else if (s == NULL) 6408 action = pf_test_rule(&r, &s, dir, kif, 6409 m, off, h, &pd, &a, &ruleset, NULL, inp); 6410 break; 6411 } 6412 6413 case IPPROTO_ICMPV6: { 6414 struct icmp6_hdr ih; 6415 6416 pd.hdr.icmp6 = &ih; 6417 if (!pf_pull_hdr(m, off, &ih, sizeof(ih), 6418 &action, &reason, AF_INET6)) { 6419 log = action != PF_PASS; 6420 goto done; 6421 } 6422 action = pf_test_state_icmp(&s, dir, kif, 6423 m, off, h, &pd, &reason); 6424 if (action == PF_PASS) { 6425 pfsync_update_state(s); 6426 r = s->rule.ptr; 6427 a = s->anchor.ptr; 6428 log = s->log; 6429 } else if (s == NULL) 6430 action = pf_test_rule(&r, &s, dir, kif, 6431 m, off, h, &pd, &a, &ruleset, NULL, inp); 6432 break; 6433 } 6434 6435 default: 6436 action = pf_test_state_other(&s, dir, kif, m, &pd); 6437 if (action == PF_PASS) { 6438 pfsync_update_state(s); 6439 r = s->rule.ptr; 6440 a = s->anchor.ptr; 6441 log = s->log; 6442 } else if (s == NULL) 6443 action = pf_test_rule(&r, &s, dir, kif, m, off, h, 6444 &pd, &a, &ruleset, NULL, inp); 6445 break; 6446 } 6447 6448 done: 6449 if (n != m) { 6450 m_freem(n); 6451 n = NULL; 6452 } 6453 6454 /* handle dangerous IPv6 extension headers. */ 6455 if (action == PF_PASS && rh_cnt && 6456 !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) { 6457 action = PF_DROP; 6458 REASON_SET(&reason, PFRES_IPOPTIONS); 6459 log = 1; 6460 DPFPRINTF(PF_DEBUG_MISC, 6461 ("pf: dropping packet with dangerous v6 headers\n")); 6462 } 6463 6464 if ((s && s->tag) || r->rtableid) 6465 pf_tag_packet(m, s ? s->tag : 0, r->rtableid); 6466 6467 #if 0 6468 if (dir == PF_IN && s && s->key[PF_SK_STACK]) 6469 m->m_pkthdr.pf.statekey = s->key[PF_SK_STACK]; 6470 #endif 6471 6472 #ifdef ALTQ 6473 if (action == PF_PASS && r->qid) { 6474 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE; 6475 if (pd.tos & IPTOS_LOWDELAY) 6476 m->m_pkthdr.pf.qid = r->pqid; 6477 else 6478 m->m_pkthdr.pf.qid = r->qid; 6479 m->m_pkthdr.pf.ecn_af = AF_INET6; 6480 m->m_pkthdr.pf.hdr = h; 6481 if (s) { 6482 /* for fairq */ 6483 m->m_pkthdr.pf.state_hash = s->hash; 6484 m->m_pkthdr.pf.flags |= PF_TAG_STATE_HASHED; 6485 } 6486 } 6487 #endif /* ALTQ */ 6488 6489 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP || 6490 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL && 6491 (s->nat_rule.ptr->action == PF_RDR || 6492 s->nat_rule.ptr->action == PF_BINAT) && 6493 IN6_IS_ADDR_LOOPBACK(&pd.dst->v6)) 6494 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST; 6495 6496 if (dir == PF_IN && action == PF_PASS && r->divert.port) { 6497 struct pf_divert *divert; 6498 6499 if ((divert = pf_get_divert(m))) { 6500 m->m_pkthdr.pf.flags |= PF_TAG_DIVERTED; 6501 divert->port = r->divert.port; 6502 divert->addr.ipv6 = r->divert.addr.v6; 6503 } 6504 } 6505 6506 if (log) { 6507 struct pf_rule *lr; 6508 6509 if (s != NULL && s->nat_rule.ptr != NULL && 6510 s->nat_rule.ptr->log & PF_LOG_ALL) 6511 lr = s->nat_rule.ptr; 6512 else 6513 lr = r; 6514 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset, 6515 &pd); 6516 } 6517 6518 kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len; 6519 kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++; 6520 6521 if (action == PF_PASS || r->action == PF_DROP) { 6522 dirndx = (dir == PF_OUT); 6523 r->packets[dirndx]++; 6524 r->bytes[dirndx] += pd.tot_len; 6525 if (a != NULL) { 6526 a->packets[dirndx]++; 6527 a->bytes[dirndx] += pd.tot_len; 6528 } 6529 if (s != NULL) { 6530 if (s->nat_rule.ptr != NULL) { 6531 s->nat_rule.ptr->packets[dirndx]++; 6532 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len; 6533 } 6534 if (s->src_node != NULL) { 6535 s->src_node->packets[dirndx]++; 6536 s->src_node->bytes[dirndx] += pd.tot_len; 6537 } 6538 if (s->nat_src_node != NULL) { 6539 s->nat_src_node->packets[dirndx]++; 6540 s->nat_src_node->bytes[dirndx] += pd.tot_len; 6541 } 6542 dirndx = (dir == s->direction) ? 0 : 1; 6543 s->packets[dirndx]++; 6544 s->bytes[dirndx] += pd.tot_len; 6545 } 6546 tr = r; 6547 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule; 6548 if (nr != NULL && r == &pf_default_rule) 6549 tr = nr; 6550 if (tr->src.addr.type == PF_ADDR_TABLE) 6551 pfr_update_stats(tr->src.addr.p.tbl, 6552 (s == NULL) ? pd.src : 6553 &s->key[(s->direction == PF_IN)]->addr[0], 6554 pd.af, pd.tot_len, dir == PF_OUT, 6555 r->action == PF_PASS, tr->src.neg); 6556 if (tr->dst.addr.type == PF_ADDR_TABLE) 6557 pfr_update_stats(tr->dst.addr.p.tbl, 6558 (s == NULL) ? pd.dst : 6559 &s->key[(s->direction == PF_IN)]->addr[1], 6560 pd.af, pd.tot_len, dir == PF_OUT, 6561 r->action == PF_PASS, tr->dst.neg); 6562 } 6563 6564 6565 if (action == PF_SYNPROXY_DROP) { 6566 m_freem(*m0); 6567 *m0 = NULL; 6568 action = PF_PASS; 6569 } else if (r->rt) 6570 /* pf_route6 can free the mbuf causing *m0 to become NULL */ 6571 pf_route6(m0, r, dir, kif->pfik_ifp, s, &pd); 6572 6573 return (action); 6574 } 6575 #endif /* INET6 */ 6576 6577 int 6578 pf_check_congestion(struct ifqueue *ifq) 6579 { 6580 return (0); 6581 } 6582