1 /* $OpenBSD: pf_lb.c,v 1.70 2022/02/16 08:46:11 sashan Exp $ */ 2 3 /* 4 * Copyright (c) 2001 Daniel Hartmeier 5 * Copyright (c) 2002 - 2008 Henning Brauer 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 12 * - Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * - Redistributions in binary form must reproduce the above 15 * copyright notice, this list of conditions and the following 16 * disclaimer in the documentation and/or other materials provided 17 * with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 20 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 21 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 22 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 23 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 24 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, 25 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 26 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 27 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN 29 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 * POSSIBILITY OF SUCH DAMAGE. 31 * 32 * Effort sponsored in part by the Defense Advanced Research Projects 33 * Agency (DARPA) and Air Force Research Laboratory, Air Force 34 * Materiel Command, USAF, under agreement number F30602-01-2-0537. 35 * 36 */ 37 38 #include "bpfilter.h" 39 #include "pflog.h" 40 #include "pfsync.h" 41 #include "pflow.h" 42 43 #include <sys/param.h> 44 #include <sys/systm.h> 45 #include <sys/mbuf.h> 46 #include <sys/filio.h> 47 #include <sys/socket.h> 48 #include <sys/socketvar.h> 49 #include <sys/kernel.h> 50 #include <sys/time.h> 51 #include <sys/pool.h> 52 #include <sys/rwlock.h> 53 #include <sys/syslog.h> 54 #include <sys/stdint.h> 55 56 #include <crypto/siphash.h> 57 58 #include <net/if.h> 59 #include <net/bpf.h> 60 #include <net/route.h> 61 62 #include <netinet/in.h> 63 #include <netinet/ip.h> 64 #include <netinet/in_pcb.h> 65 #include <netinet/ip_var.h> 66 #include <netinet/ip_icmp.h> 67 #include <netinet/icmp_var.h> 68 #include <netinet/tcp.h> 69 #include <netinet/tcp_seq.h> 70 #include <netinet/tcp_timer.h> 71 #include <netinet/udp.h> 72 #include <netinet/udp_var.h> 73 #include <netinet/if_ether.h> 74 75 #ifdef INET6 76 #include <netinet/ip6.h> 77 #include <netinet/icmp6.h> 78 #endif /* INET6 */ 79 80 #include <net/pfvar.h> 81 #include <net/pfvar_priv.h> 82 83 #if NPFLOG > 0 84 #include <net/if_pflog.h> 85 #endif /* NPFLOG > 0 */ 86 87 #if NPFLOW > 0 88 #include <net/if_pflow.h> 89 #endif /* NPFLOW > 0 */ 90 91 #if NPFSYNC > 0 92 #include <net/if_pfsync.h> 93 #endif /* NPFSYNC > 0 */ 94 95 u_int64_t pf_hash(struct pf_addr *, struct pf_addr *, 96 struct pf_poolhashkey *, sa_family_t); 97 int pf_get_sport(struct pf_pdesc *, struct pf_rule *, 98 struct pf_addr *, u_int16_t *, u_int16_t, 99 u_int16_t, struct pf_src_node **); 100 int pf_map_addr_states_increase(sa_family_t, 101 struct pf_pool *, struct pf_addr *); 102 int pf_get_transaddr_af(struct pf_rule *, 103 struct pf_pdesc *, struct pf_src_node **); 104 int pf_map_addr_sticky(sa_family_t, struct pf_rule *, 105 struct pf_addr *, struct pf_addr *, 106 struct pf_src_node **, struct pf_pool *, 107 enum pf_sn_types); 108 109 u_int64_t 110 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash, 111 struct pf_poolhashkey *key, sa_family_t af) 112 { 113 uint64_t res = 0; 114 #ifdef INET6 115 union { 116 uint64_t hash64; 117 uint32_t hash32[2]; 118 } h; 119 #endif /* INET6 */ 120 121 switch (af) { 122 case AF_INET: 123 res = SipHash24((SIPHASH_KEY *)key, 124 &inaddr->addr32[0], sizeof(inaddr->addr32[0])); 125 hash->addr32[0] = res; 126 break; 127 #ifdef INET6 128 case AF_INET6: 129 res = SipHash24((SIPHASH_KEY *)key, &inaddr->addr32[0], 130 4 * sizeof(inaddr->addr32[0])); 131 h.hash64 = res; 132 hash->addr32[0] = h.hash32[0]; 133 hash->addr32[1] = h.hash32[1]; 134 /* 135 * siphash isn't big enough, but flipping it around is 136 * good enough here. 137 */ 138 hash->addr32[2] = ~h.hash32[1]; 139 hash->addr32[3] = ~h.hash32[0]; 140 break; 141 #endif /* INET6 */ 142 default: 143 unhandled_af(af); 144 } 145 return (res); 146 } 147 148 int 149 pf_get_sport(struct pf_pdesc *pd, struct pf_rule *r, 150 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high, 151 struct pf_src_node **sn) 152 { 153 struct pf_state_key_cmp key; 154 struct pf_addr init_addr; 155 u_int16_t cut; 156 int dir = (pd->dir == PF_IN) ? PF_OUT : PF_IN; 157 int sidx = pd->sidx; 158 int didx = pd->didx; 159 160 memset(&init_addr, 0, sizeof(init_addr)); 161 if (pf_map_addr(pd->naf, r, &pd->nsaddr, naddr, &init_addr, sn, &r->nat, 162 PF_SN_NAT)) 163 return (1); 164 165 if (pd->proto == IPPROTO_ICMP) { 166 if (pd->ndport == htons(ICMP_ECHO)) { 167 low = 1; 168 high = 65535; 169 } else 170 return (0); /* Don't try to modify non-echo ICMP */ 171 } 172 #ifdef INET6 173 if (pd->proto == IPPROTO_ICMPV6) { 174 if (pd->ndport == htons(ICMP6_ECHO_REQUEST)) { 175 low = 1; 176 high = 65535; 177 } else 178 return (0); /* Don't try to modify non-echo ICMP */ 179 } 180 #endif /* INET6 */ 181 182 do { 183 key.af = pd->naf; 184 key.proto = pd->proto; 185 key.rdomain = pd->rdomain; 186 pf_addrcpy(&key.addr[didx], &pd->ndaddr, key.af); 187 pf_addrcpy(&key.addr[sidx], naddr, key.af); 188 key.port[didx] = pd->ndport; 189 190 /* 191 * port search; start random, step; 192 * similar 2 portloop in in_pcbbind 193 */ 194 if (!(pd->proto == IPPROTO_TCP || pd->proto == IPPROTO_UDP || 195 pd->proto == IPPROTO_ICMP || pd->proto == IPPROTO_ICMPV6)) { 196 /* XXX bug: icmp states dont use the id on both 197 * XXX sides (traceroute -I through nat) */ 198 key.port[sidx] = pd->nsport; 199 if (pf_find_state_all(&key, dir, NULL) == NULL) { 200 *nport = pd->nsport; 201 return (0); 202 } 203 } else if (low == 0 && high == 0) { 204 key.port[sidx] = pd->nsport; 205 if (pf_find_state_all(&key, dir, NULL) == NULL) { 206 *nport = pd->nsport; 207 return (0); 208 } 209 } else if (low == high) { 210 key.port[sidx] = htons(low); 211 if (pf_find_state_all(&key, dir, NULL) == NULL) { 212 *nport = htons(low); 213 return (0); 214 } 215 } else { 216 u_int32_t tmp; 217 218 if (low > high) { 219 tmp = low; 220 low = high; 221 high = tmp; 222 } 223 /* low < high */ 224 cut = arc4random_uniform(1 + high - low) + low; 225 /* low <= cut <= high */ 226 for (tmp = cut; tmp <= high && tmp <= 0xffff; ++tmp) { 227 key.port[sidx] = htons(tmp); 228 if (pf_find_state_all(&key, dir, NULL) == 229 NULL && !in_baddynamic(tmp, pd->proto)) { 230 *nport = htons(tmp); 231 return (0); 232 } 233 } 234 tmp = cut; 235 for (tmp -= 1; tmp >= low && tmp <= 0xffff; --tmp) { 236 key.port[sidx] = htons(tmp); 237 if (pf_find_state_all(&key, dir, NULL) == 238 NULL && !in_baddynamic(tmp, pd->proto)) { 239 *nport = htons(tmp); 240 return (0); 241 } 242 } 243 } 244 245 switch (r->nat.opts & PF_POOL_TYPEMASK) { 246 case PF_POOL_RANDOM: 247 case PF_POOL_ROUNDROBIN: 248 case PF_POOL_LEASTSTATES: 249 /* 250 * pick a different source address since we're out 251 * of free port choices for the current one. 252 */ 253 if (pf_map_addr(pd->naf, r, &pd->nsaddr, naddr, 254 &init_addr, sn, &r->nat, PF_SN_NAT)) 255 return (1); 256 break; 257 case PF_POOL_NONE: 258 case PF_POOL_SRCHASH: 259 case PF_POOL_BITMASK: 260 default: 261 return (1); 262 } 263 } while (! PF_AEQ(&init_addr, naddr, pd->naf) ); 264 return (1); /* none available */ 265 } 266 267 int 268 pf_map_addr_sticky(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr, 269 struct pf_addr *naddr, struct pf_src_node **sns, struct pf_pool *rpool, 270 enum pf_sn_types type) 271 { 272 struct pf_addr *raddr, *rmask, *cached; 273 struct pf_state *s; 274 struct pf_src_node k; 275 int valid; 276 277 k.af = af; 278 k.type = type; 279 pf_addrcpy(&k.addr, saddr, af); 280 k.rule.ptr = r; 281 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++; 282 sns[type] = RB_FIND(pf_src_tree, &tree_src_tracking, &k); 283 if (sns[type] == NULL) 284 return (-1); 285 286 /* check if the cached entry is still valid */ 287 cached = &(sns[type])->raddr; 288 valid = 0; 289 if (PF_AZERO(cached, af)) { 290 valid = 1; 291 } else if (rpool->addr.type == PF_ADDR_DYNIFTL) { 292 if (pfr_kentry_byaddr(rpool->addr.p.dyn->pfid_kt, cached, 293 af, 0)) 294 valid = 1; 295 } else if (rpool->addr.type == PF_ADDR_TABLE) { 296 if (pfr_kentry_byaddr(rpool->addr.p.tbl, cached, af, 0)) 297 valid = 1; 298 } else if (rpool->addr.type != PF_ADDR_NOROUTE) { 299 raddr = &rpool->addr.v.a.addr; 300 rmask = &rpool->addr.v.a.mask; 301 valid = pf_match_addr(0, raddr, rmask, cached, af); 302 } 303 if (!valid) { 304 if (pf_status.debug >= LOG_DEBUG) { 305 log(LOG_DEBUG, "pf: pf_map_addr: " 306 "stale src tracking (%u) ", type); 307 pf_print_host(&k.addr, 0, af); 308 addlog(" to "); 309 pf_print_host(cached, 0, af); 310 addlog("\n"); 311 } 312 if (sns[type]->states != 0) { 313 /* XXX expensive */ 314 RB_FOREACH(s, pf_state_tree_id, 315 &tree_id) 316 pf_state_rm_src_node(s, 317 sns[type]); 318 } 319 sns[type]->expire = 1; 320 pf_remove_src_node(sns[type]); 321 sns[type] = NULL; 322 return (-1); 323 } 324 325 326 if (!PF_AZERO(cached, af)) { 327 pf_addrcpy(naddr, cached, af); 328 if ((rpool->opts & PF_POOL_TYPEMASK) == PF_POOL_LEASTSTATES && 329 pf_map_addr_states_increase(af, rpool, cached) == -1) 330 return (-1); 331 } 332 if (pf_status.debug >= LOG_DEBUG) { 333 log(LOG_DEBUG, "pf: pf_map_addr: " 334 "src tracking (%u) maps ", type); 335 pf_print_host(&k.addr, 0, af); 336 addlog(" to "); 337 pf_print_host(naddr, 0, af); 338 addlog("\n"); 339 } 340 341 if (sns[type]->kif != NULL) 342 rpool->kif = sns[type]->kif; 343 344 return (0); 345 } 346 347 int 348 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr, 349 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sns, 350 struct pf_pool *rpool, enum pf_sn_types type) 351 { 352 struct pf_addr hash; 353 struct pf_addr faddr; 354 struct pf_addr *raddr = &rpool->addr.v.a.addr; 355 struct pf_addr *rmask = &rpool->addr.v.a.mask; 356 struct pfr_ktable *kt; 357 struct pfi_kif *kif; 358 u_int64_t states; 359 u_int16_t weight; 360 u_int64_t load; 361 u_int64_t cload; 362 u_int64_t hashidx; 363 int cnt; 364 365 if (sns[type] == NULL && rpool->opts & PF_POOL_STICKYADDR && 366 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE && 367 pf_map_addr_sticky(af, r, saddr, naddr, sns, rpool, type) == 0) 368 return (0); 369 370 if (rpool->addr.type == PF_ADDR_NOROUTE) 371 return (1); 372 if (rpool->addr.type == PF_ADDR_DYNIFTL) { 373 switch (af) { 374 case AF_INET: 375 if (rpool->addr.p.dyn->pfid_acnt4 < 1 && 376 !PF_POOL_DYNTYPE(rpool->opts)) 377 return (1); 378 raddr = &rpool->addr.p.dyn->pfid_addr4; 379 rmask = &rpool->addr.p.dyn->pfid_mask4; 380 break; 381 #ifdef INET6 382 case AF_INET6: 383 if (rpool->addr.p.dyn->pfid_acnt6 < 1 && 384 !PF_POOL_DYNTYPE(rpool->opts)) 385 return (1); 386 raddr = &rpool->addr.p.dyn->pfid_addr6; 387 rmask = &rpool->addr.p.dyn->pfid_mask6; 388 break; 389 #endif /* INET6 */ 390 default: 391 unhandled_af(af); 392 } 393 } else if (rpool->addr.type == PF_ADDR_TABLE) { 394 if (!PF_POOL_DYNTYPE(rpool->opts)) 395 return (1); /* unsupported */ 396 } else { 397 raddr = &rpool->addr.v.a.addr; 398 rmask = &rpool->addr.v.a.mask; 399 } 400 401 switch (rpool->opts & PF_POOL_TYPEMASK) { 402 case PF_POOL_NONE: 403 pf_addrcpy(naddr, raddr, af); 404 break; 405 case PF_POOL_BITMASK: 406 pf_poolmask(naddr, raddr, rmask, saddr, af); 407 break; 408 case PF_POOL_RANDOM: 409 if (rpool->addr.type == PF_ADDR_TABLE || 410 rpool->addr.type == PF_ADDR_DYNIFTL) { 411 if (rpool->addr.type == PF_ADDR_TABLE) 412 kt = rpool->addr.p.tbl; 413 else 414 kt = rpool->addr.p.dyn->pfid_kt; 415 kt = pfr_ktable_select_active(kt); 416 if (kt == NULL) 417 return (1); 418 419 cnt = kt->pfrkt_cnt; 420 if (cnt == 0) 421 rpool->tblidx = 0; 422 else 423 rpool->tblidx = (int)arc4random_uniform(cnt); 424 memset(&rpool->counter, 0, sizeof(rpool->counter)); 425 if (pfr_pool_get(rpool, &raddr, &rmask, af)) 426 return (1); 427 pf_addrcpy(naddr, &rpool->counter, af); 428 } else if (init_addr != NULL && PF_AZERO(init_addr, af)) { 429 switch (af) { 430 case AF_INET: 431 rpool->counter.addr32[0] = arc4random(); 432 break; 433 #ifdef INET6 434 case AF_INET6: 435 if (rmask->addr32[3] != 0xffffffff) 436 rpool->counter.addr32[3] = arc4random(); 437 else 438 break; 439 if (rmask->addr32[2] != 0xffffffff) 440 rpool->counter.addr32[2] = arc4random(); 441 else 442 break; 443 if (rmask->addr32[1] != 0xffffffff) 444 rpool->counter.addr32[1] = arc4random(); 445 else 446 break; 447 if (rmask->addr32[0] != 0xffffffff) 448 rpool->counter.addr32[0] = arc4random(); 449 break; 450 #endif /* INET6 */ 451 default: 452 unhandled_af(af); 453 } 454 pf_poolmask(naddr, raddr, rmask, &rpool->counter, af); 455 pf_addrcpy(init_addr, naddr, af); 456 457 } else { 458 pf_addr_inc(&rpool->counter, af); 459 pf_poolmask(naddr, raddr, rmask, &rpool->counter, af); 460 } 461 break; 462 case PF_POOL_SRCHASH: 463 hashidx = pf_hash(saddr, &hash, &rpool->key, af); 464 465 if (rpool->addr.type == PF_ADDR_TABLE || 466 rpool->addr.type == PF_ADDR_DYNIFTL) { 467 if (rpool->addr.type == PF_ADDR_TABLE) 468 kt = rpool->addr.p.tbl; 469 else 470 kt = rpool->addr.p.dyn->pfid_kt; 471 kt = pfr_ktable_select_active(kt); 472 if (kt == NULL) 473 return (1); 474 475 cnt = kt->pfrkt_cnt; 476 if (cnt == 0) 477 rpool->tblidx = 0; 478 else 479 rpool->tblidx = (int)(hashidx % cnt); 480 memset(&rpool->counter, 0, sizeof(rpool->counter)); 481 if (pfr_pool_get(rpool, &raddr, &rmask, af)) 482 return (1); 483 pf_addrcpy(naddr, &rpool->counter, af); 484 } else { 485 pf_poolmask(naddr, raddr, rmask, &hash, af); 486 } 487 break; 488 case PF_POOL_ROUNDROBIN: 489 if (rpool->addr.type == PF_ADDR_TABLE || 490 rpool->addr.type == PF_ADDR_DYNIFTL) { 491 if (pfr_pool_get(rpool, &raddr, &rmask, af)) { 492 /* 493 * reset counter in case its value 494 * has been removed from the pool. 495 */ 496 memset(&rpool->counter, 0, 497 sizeof(rpool->counter)); 498 if (pfr_pool_get(rpool, &raddr, &rmask, af)) 499 return (1); 500 } 501 } else if (PF_AZERO(&rpool->counter, af)) { 502 /* 503 * fall back to POOL_NONE if there are no addresses in 504 * pool 505 */ 506 pf_addrcpy(naddr, raddr, af); 507 break; 508 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af)) 509 return (1); 510 511 /* iterate over table if it contains entries which are weighted */ 512 if ((rpool->addr.type == PF_ADDR_TABLE && 513 rpool->addr.p.tbl->pfrkt_refcntcost > 0) || 514 (rpool->addr.type == PF_ADDR_DYNIFTL && 515 rpool->addr.p.dyn->pfid_kt->pfrkt_refcntcost > 0)) { 516 do { 517 if (rpool->addr.type == PF_ADDR_TABLE || 518 rpool->addr.type == PF_ADDR_DYNIFTL) { 519 if (pfr_pool_get(rpool, 520 &raddr, &rmask, af)) 521 return (1); 522 } else { 523 log(LOG_ERR, "pf: pf_map_addr: " 524 "weighted RR failure"); 525 return (1); 526 } 527 if (rpool->weight >= rpool->curweight) 528 break; 529 pf_addr_inc(&rpool->counter, af); 530 } while (1); 531 532 weight = rpool->weight; 533 } 534 535 pf_addrcpy(naddr, &rpool->counter, af); 536 if (init_addr != NULL && PF_AZERO(init_addr, af)) 537 pf_addrcpy(init_addr, naddr, af); 538 pf_addr_inc(&rpool->counter, af); 539 break; 540 case PF_POOL_LEASTSTATES: 541 /* retrieve an address first */ 542 if (rpool->addr.type == PF_ADDR_TABLE || 543 rpool->addr.type == PF_ADDR_DYNIFTL) { 544 if (pfr_pool_get(rpool, &raddr, &rmask, af)) { 545 /* see PF_POOL_ROUNDROBIN */ 546 memset(&rpool->counter, 0, 547 sizeof(rpool->counter)); 548 if (pfr_pool_get(rpool, &raddr, &rmask, af)) 549 return (1); 550 } 551 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af)) 552 return (1); 553 554 states = rpool->states; 555 weight = rpool->weight; 556 kif = rpool->kif; 557 558 if ((rpool->addr.type == PF_ADDR_TABLE && 559 rpool->addr.p.tbl->pfrkt_refcntcost > 0) || 560 (rpool->addr.type == PF_ADDR_DYNIFTL && 561 rpool->addr.p.dyn->pfid_kt->pfrkt_refcntcost > 0)) 562 load = ((UINT16_MAX * rpool->states) / rpool->weight); 563 else 564 load = states; 565 566 pf_addrcpy(&faddr, &rpool->counter, af); 567 568 pf_addrcpy(naddr, &rpool->counter, af); 569 if (init_addr != NULL && PF_AZERO(init_addr, af)) 570 pf_addrcpy(init_addr, naddr, af); 571 572 /* 573 * iterate *once* over whole table and find destination with 574 * least connection 575 */ 576 do { 577 pf_addr_inc(&rpool->counter, af); 578 if (rpool->addr.type == PF_ADDR_TABLE || 579 rpool->addr.type == PF_ADDR_DYNIFTL) { 580 if (pfr_pool_get(rpool, &raddr, &rmask, af)) 581 return (1); 582 } else if (pf_match_addr(0, raddr, rmask, 583 &rpool->counter, af)) 584 return (1); 585 586 if ((rpool->addr.type == PF_ADDR_TABLE && 587 rpool->addr.p.tbl->pfrkt_refcntcost > 0) || 588 (rpool->addr.type == PF_ADDR_DYNIFTL && 589 rpool->addr.p.dyn->pfid_kt->pfrkt_refcntcost > 0)) 590 cload = ((UINT16_MAX * rpool->states) 591 / rpool->weight); 592 else 593 cload = rpool->states; 594 595 /* find lc minimum */ 596 if (cload < load) { 597 states = rpool->states; 598 weight = rpool->weight; 599 kif = rpool->kif; 600 load = cload; 601 602 pf_addrcpy(naddr, &rpool->counter, af); 603 if (init_addr != NULL && 604 PF_AZERO(init_addr, af)) 605 pf_addrcpy(init_addr, naddr, af); 606 } 607 } while (pf_match_addr(1, &faddr, rmask, &rpool->counter, af) && 608 (states > 0)); 609 610 if (pf_map_addr_states_increase(af, rpool, naddr) == -1) 611 return (1); 612 /* revert the kif which was set by pfr_pool_get() */ 613 rpool->kif = kif; 614 break; 615 } 616 617 if (rpool->opts & PF_POOL_STICKYADDR) { 618 if (sns[type] != NULL) { 619 pf_remove_src_node(sns[type]); 620 sns[type] = NULL; 621 } 622 if (pf_insert_src_node(&sns[type], r, type, af, saddr, naddr, 623 rpool->kif)) 624 return (1); 625 } 626 627 if (pf_status.debug >= LOG_INFO && 628 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { 629 log(LOG_INFO, "pf: pf_map_addr: selected address "); 630 pf_print_host(naddr, 0, af); 631 if ((rpool->opts & PF_POOL_TYPEMASK) == 632 PF_POOL_LEASTSTATES) 633 addlog(" with state count %llu", states); 634 if ((rpool->addr.type == PF_ADDR_TABLE && 635 rpool->addr.p.tbl->pfrkt_refcntcost > 0) || 636 (rpool->addr.type == PF_ADDR_DYNIFTL && 637 rpool->addr.p.dyn->pfid_kt->pfrkt_refcntcost > 0)) 638 addlog(" with weight %u", weight); 639 addlog("\n"); 640 } 641 642 return (0); 643 } 644 645 int 646 pf_map_addr_states_increase(sa_family_t af, struct pf_pool *rpool, 647 struct pf_addr *naddr) 648 { 649 if (rpool->addr.type == PF_ADDR_TABLE) { 650 if (pfr_states_increase(rpool->addr.p.tbl, 651 naddr, af) == -1) { 652 if (pf_status.debug >= LOG_DEBUG) { 653 log(LOG_DEBUG, 654 "pf: pf_map_addr_states_increase: " 655 "selected address "); 656 pf_print_host(naddr, 0, af); 657 addlog(". Failed to increase count!\n"); 658 } 659 return (-1); 660 } 661 } else if (rpool->addr.type == PF_ADDR_DYNIFTL) { 662 if (pfr_states_increase(rpool->addr.p.dyn->pfid_kt, 663 naddr, af) == -1) { 664 if (pf_status.debug >= LOG_DEBUG) { 665 log(LOG_DEBUG, 666 "pf: pf_map_addr_states_increase: " 667 "selected address "); 668 pf_print_host(naddr, 0, af); 669 addlog(". Failed to increase count!\n"); 670 } 671 return (-1); 672 } 673 } 674 return (0); 675 } 676 677 int 678 pf_get_transaddr(struct pf_rule *r, struct pf_pdesc *pd, 679 struct pf_src_node **sns, struct pf_rule **nr) 680 { 681 struct pf_addr naddr; 682 u_int16_t nport; 683 684 #ifdef INET6 685 if (pd->af != pd->naf) 686 return (pf_get_transaddr_af(r, pd, sns)); 687 #endif /* INET6 */ 688 689 if (r->nat.addr.type != PF_ADDR_NONE) { 690 /* XXX is this right? what if rtable is changed at the same 691 * XXX time? where do I need to figure out the sport? */ 692 nport = 0; 693 if (pf_get_sport(pd, r, &naddr, &nport, 694 r->nat.proxy_port[0], r->nat.proxy_port[1], sns)) { 695 DPFPRINTF(LOG_NOTICE, 696 "pf: NAT proxy port allocation (%u-%u) failed", 697 r->nat.proxy_port[0], 698 r->nat.proxy_port[1]); 699 return (-1); 700 } 701 *nr = r; 702 pf_addrcpy(&pd->nsaddr, &naddr, pd->af); 703 pd->nsport = nport; 704 } 705 if (r->rdr.addr.type != PF_ADDR_NONE) { 706 if (pf_map_addr(pd->af, r, &pd->nsaddr, &naddr, NULL, sns, 707 &r->rdr, PF_SN_RDR)) 708 return (-1); 709 if ((r->rdr.opts & PF_POOL_TYPEMASK) == PF_POOL_BITMASK) 710 pf_poolmask(&naddr, &naddr, &r->rdr.addr.v.a.mask, 711 &pd->ndaddr, pd->af); 712 713 nport = 0; 714 if (r->rdr.proxy_port[1]) { 715 u_int32_t tmp_nport; 716 u_int16_t div; 717 718 div = r->rdr.proxy_port[1] - r->rdr.proxy_port[0] + 1; 719 div = (div == 0) ? 1 : div; 720 721 tmp_nport = ((ntohs(pd->ndport) - ntohs(r->dst.port[0])) % div) + 722 r->rdr.proxy_port[0]; 723 724 /* wrap around if necessary */ 725 if (tmp_nport > 65535) 726 tmp_nport -= 65535; 727 nport = htons((u_int16_t)tmp_nport); 728 } else if (r->rdr.proxy_port[0]) 729 nport = htons(r->rdr.proxy_port[0]); 730 *nr = r; 731 pf_addrcpy(&pd->ndaddr, &naddr, pd->af); 732 if (nport) 733 pd->ndport = nport; 734 } 735 736 return (0); 737 } 738 739 #ifdef INET6 740 int 741 pf_get_transaddr_af(struct pf_rule *r, struct pf_pdesc *pd, 742 struct pf_src_node **sns) 743 { 744 struct pf_addr ndaddr, nsaddr, naddr; 745 u_int16_t nport; 746 int prefixlen = 96; 747 748 if (pf_status.debug >= LOG_INFO) { 749 log(LOG_INFO, "pf: af-to %s %s, ", 750 pd->naf == AF_INET ? "inet" : "inet6", 751 r->rdr.addr.type == PF_ADDR_NONE ? "nat" : "rdr"); 752 pf_print_host(&pd->nsaddr, pd->nsport, pd->af); 753 addlog(" -> "); 754 pf_print_host(&pd->ndaddr, pd->ndport, pd->af); 755 addlog("\n"); 756 } 757 758 if (r->nat.addr.type == PF_ADDR_NONE) 759 panic("pf_get_transaddr_af: no nat pool for source address"); 760 761 /* get source address and port */ 762 nport = 0; 763 if (pf_get_sport(pd, r, &nsaddr, &nport, 764 r->nat.proxy_port[0], r->nat.proxy_port[1], sns)) { 765 DPFPRINTF(LOG_NOTICE, 766 "pf: af-to NAT proxy port allocation (%u-%u) failed", 767 r->nat.proxy_port[0], 768 r->nat.proxy_port[1]); 769 return (-1); 770 } 771 pd->nsport = nport; 772 773 if (pd->proto == IPPROTO_ICMPV6 && pd->naf == AF_INET) { 774 if (pd->dir == PF_IN) { 775 pd->ndport = ntohs(pd->ndport); 776 if (pd->ndport == ICMP6_ECHO_REQUEST) 777 pd->ndport = ICMP_ECHO; 778 else if (pd->ndport == ICMP6_ECHO_REPLY) 779 pd->ndport = ICMP_ECHOREPLY; 780 pd->ndport = htons(pd->ndport); 781 } else { 782 pd->nsport = ntohs(pd->nsport); 783 if (pd->nsport == ICMP6_ECHO_REQUEST) 784 pd->nsport = ICMP_ECHO; 785 else if (pd->nsport == ICMP6_ECHO_REPLY) 786 pd->nsport = ICMP_ECHOREPLY; 787 pd->nsport = htons(pd->nsport); 788 } 789 } else if (pd->proto == IPPROTO_ICMP && pd->naf == AF_INET6) { 790 if (pd->dir == PF_IN) { 791 pd->ndport = ntohs(pd->ndport); 792 if (pd->ndport == ICMP_ECHO) 793 pd->ndport = ICMP6_ECHO_REQUEST; 794 else if (pd->ndport == ICMP_ECHOREPLY) 795 pd->ndport = ICMP6_ECHO_REPLY; 796 pd->ndport = htons(pd->ndport); 797 } else { 798 pd->nsport = ntohs(pd->nsport); 799 if (pd->nsport == ICMP_ECHO) 800 pd->nsport = ICMP6_ECHO_REQUEST; 801 else if (pd->nsport == ICMP_ECHOREPLY) 802 pd->nsport = ICMP6_ECHO_REPLY; 803 pd->nsport = htons(pd->nsport); 804 } 805 } 806 807 /* get the destination address and port */ 808 if (r->rdr.addr.type != PF_ADDR_NONE) { 809 if (pf_map_addr(pd->naf, r, &nsaddr, &naddr, NULL, sns, 810 &r->rdr, PF_SN_RDR)) 811 return (-1); 812 if (r->rdr.proxy_port[0]) 813 pd->ndport = htons(r->rdr.proxy_port[0]); 814 815 if (pd->naf == AF_INET) { 816 /* The prefix is the IPv4 rdr address */ 817 prefixlen = in_mask2len((struct in_addr *) 818 &r->rdr.addr.v.a.mask); 819 inet_nat46(pd->naf, &pd->ndaddr, 820 &ndaddr, &naddr, prefixlen); 821 } else { 822 /* The prefix is the IPv6 rdr address */ 823 prefixlen = 824 in6_mask2len((struct in6_addr *) 825 &r->rdr.addr.v.a.mask, NULL); 826 inet_nat64(pd->naf, &pd->ndaddr, 827 &ndaddr, &naddr, prefixlen); 828 } 829 } else { 830 if (pd->naf == AF_INET) { 831 /* The prefix is the IPv6 dst address */ 832 prefixlen = 833 in6_mask2len((struct in6_addr *) 834 &r->dst.addr.v.a.mask, NULL); 835 if (prefixlen < 32) 836 prefixlen = 96; 837 inet_nat64(pd->naf, &pd->ndaddr, 838 &ndaddr, &pd->ndaddr, prefixlen); 839 } else { 840 /* 841 * The prefix is the IPv6 nat address 842 * (that was stored in pd->nsaddr) 843 */ 844 prefixlen = in6_mask2len((struct in6_addr *) 845 &r->nat.addr.v.a.mask, NULL); 846 if (prefixlen > 96) 847 prefixlen = 96; 848 inet_nat64(pd->naf, &pd->ndaddr, 849 &ndaddr, &nsaddr, prefixlen); 850 } 851 } 852 853 pf_addrcpy(&pd->nsaddr, &nsaddr, pd->naf); 854 pf_addrcpy(&pd->ndaddr, &ndaddr, pd->naf); 855 856 if (pf_status.debug >= LOG_INFO) { 857 log(LOG_INFO, "pf: af-to %s %s done, prefixlen %d, ", 858 pd->naf == AF_INET ? "inet" : "inet6", 859 r->rdr.addr.type == PF_ADDR_NONE ? "nat" : "rdr", 860 prefixlen); 861 pf_print_host(&pd->nsaddr, pd->nsport, pd->naf); 862 addlog(" -> "); 863 pf_print_host(&pd->ndaddr, pd->ndport, pd->naf); 864 addlog("\n"); 865 } 866 867 return (0); 868 } 869 #endif /* INET6 */ 870 871 int 872 pf_postprocess_addr(struct pf_state *cur) 873 { 874 struct pf_rule *nr; 875 struct pf_state_key *sks; 876 struct pf_pool rpool; 877 struct pf_addr lookup_addr; 878 int slbcount = -1; 879 880 nr = cur->natrule.ptr; 881 882 if (nr == NULL) 883 return (0); 884 885 /* decrease counter */ 886 887 sks = cur->key[PF_SK_STACK]; 888 889 /* check for outgoing or ingoing balancing */ 890 if (nr->rt == PF_ROUTETO) 891 lookup_addr = cur->rt_addr; 892 else if (sks != NULL) 893 lookup_addr = sks->addr[1]; 894 else { 895 if (pf_status.debug >= LOG_DEBUG) { 896 log(LOG_DEBUG, "pf: %s: unable to obtain address", 897 __func__); 898 } 899 return (1); 900 } 901 902 /* check for appropriate pool */ 903 if (nr->rdr.addr.type != PF_ADDR_NONE) 904 rpool = nr->rdr; 905 else if (nr->nat.addr.type != PF_ADDR_NONE) 906 rpool = nr->nat; 907 else if (nr->route.addr.type != PF_ADDR_NONE) 908 rpool = nr->route; 909 else 910 return (0); 911 912 if (((rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_LEASTSTATES)) 913 return (0); 914 915 if (rpool.addr.type == PF_ADDR_TABLE) { 916 if ((slbcount = pfr_states_decrease( 917 rpool.addr.p.tbl, 918 &lookup_addr, sks->af)) == -1) { 919 if (pf_status.debug >= LOG_DEBUG) { 920 log(LOG_DEBUG, "pf: %s: selected address ", 921 __func__); 922 pf_print_host(&lookup_addr, 923 sks->port[0], sks->af); 924 addlog(". Failed to " 925 "decrease count!\n"); 926 } 927 return (1); 928 } 929 } else if (rpool.addr.type == PF_ADDR_DYNIFTL) { 930 if ((slbcount = pfr_states_decrease( 931 rpool.addr.p.dyn->pfid_kt, 932 &lookup_addr, sks->af)) == -1) { 933 if (pf_status.debug >= LOG_DEBUG) { 934 log(LOG_DEBUG, "pf: %s: selected address ", 935 __func__); 936 pf_print_host(&lookup_addr, 937 sks->port[0], sks->af); 938 addlog(". Failed to " 939 "decrease count!\n"); 940 } 941 return (1); 942 } 943 } 944 if (slbcount > -1) { 945 if (pf_status.debug >= LOG_INFO) { 946 log(LOG_INFO, "pf: %s: selected address ", __func__); 947 pf_print_host(&lookup_addr, sks->port[0], 948 sks->af); 949 addlog(" decreased state count to %u\n", 950 slbcount); 951 } 952 } 953 return (0); 954 } 955