1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause 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 * $OpenBSD: pf_lb.c,v 1.2 2009/02/12 02:13:15 sthen Exp $ 37 */ 38 39 #include <sys/cdefs.h> 40 #include "opt_pf.h" 41 #include "opt_inet.h" 42 #include "opt_inet6.h" 43 44 #include <sys/param.h> 45 #include <sys/lock.h> 46 #include <sys/mbuf.h> 47 #include <sys/socket.h> 48 #include <sys/sysctl.h> 49 50 #include <net/if.h> 51 #include <net/vnet.h> 52 #include <net/pfvar.h> 53 #include <net/if_pflog.h> 54 55 /* 56 * Limit the amount of work we do to find a free source port for redirects that 57 * introduce a state conflict. 58 */ 59 #define V_pf_rdr_srcport_rewrite_tries VNET(pf_rdr_srcport_rewrite_tries) 60 VNET_DEFINE_STATIC(int, pf_rdr_srcport_rewrite_tries) = 16; 61 62 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x 63 64 static void pf_hash(struct pf_addr *, struct pf_addr *, 65 struct pf_poolhashkey *, sa_family_t); 66 static struct pf_krule *pf_match_translation(struct pf_pdesc *, struct mbuf *, 67 int, struct pfi_kkif *, 68 struct pf_addr *, u_int16_t, struct pf_addr *, 69 uint16_t, int, struct pf_kanchor_stackframe *); 70 static int pf_get_sport(sa_family_t, uint8_t, struct pf_krule *, 71 struct pf_addr *, uint16_t, struct pf_addr *, uint16_t, struct pf_addr *, 72 uint16_t *, uint16_t, uint16_t, struct pf_ksrc_node **, 73 struct pf_udp_mapping **); 74 static bool pf_islinklocal(const sa_family_t, const struct pf_addr *); 75 76 #define mix(a,b,c) \ 77 do { \ 78 a -= b; a -= c; a ^= (c >> 13); \ 79 b -= c; b -= a; b ^= (a << 8); \ 80 c -= a; c -= b; c ^= (b >> 13); \ 81 a -= b; a -= c; a ^= (c >> 12); \ 82 b -= c; b -= a; b ^= (a << 16); \ 83 c -= a; c -= b; c ^= (b >> 5); \ 84 a -= b; a -= c; a ^= (c >> 3); \ 85 b -= c; b -= a; b ^= (a << 10); \ 86 c -= a; c -= b; c ^= (b >> 15); \ 87 } while (0) 88 89 /* 90 * hash function based on bridge_hash in if_bridge.c 91 */ 92 static void 93 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash, 94 struct pf_poolhashkey *key, sa_family_t af) 95 { 96 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0]; 97 98 switch (af) { 99 #ifdef INET 100 case AF_INET: 101 a += inaddr->addr32[0]; 102 b += key->key32[1]; 103 mix(a, b, c); 104 hash->addr32[0] = c + key->key32[2]; 105 break; 106 #endif /* INET */ 107 #ifdef INET6 108 case AF_INET6: 109 a += inaddr->addr32[0]; 110 b += inaddr->addr32[2]; 111 mix(a, b, c); 112 hash->addr32[0] = c; 113 a += inaddr->addr32[1]; 114 b += inaddr->addr32[3]; 115 c += key->key32[1]; 116 mix(a, b, c); 117 hash->addr32[1] = c; 118 a += inaddr->addr32[2]; 119 b += inaddr->addr32[1]; 120 c += key->key32[2]; 121 mix(a, b, c); 122 hash->addr32[2] = c; 123 a += inaddr->addr32[3]; 124 b += inaddr->addr32[0]; 125 c += key->key32[3]; 126 mix(a, b, c); 127 hash->addr32[3] = c; 128 break; 129 #endif /* INET6 */ 130 } 131 } 132 133 static struct pf_krule * 134 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off, 135 struct pfi_kkif *kif, struct pf_addr *saddr, u_int16_t sport, 136 struct pf_addr *daddr, uint16_t dport, int rs_num, 137 struct pf_kanchor_stackframe *anchor_stack) 138 { 139 struct pf_krule *r, *rm = NULL; 140 struct pf_kruleset *ruleset = NULL; 141 int tag = -1; 142 int rtableid = -1; 143 int asd = 0; 144 145 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr); 146 while (r != NULL) { 147 struct pf_rule_addr *src = NULL, *dst = NULL; 148 struct pf_addr_wrap *xdst = NULL; 149 150 if (r->action == PF_BINAT && pd->dir == PF_IN) { 151 src = &r->dst; 152 if (r->rpool.cur != NULL) 153 xdst = &r->rpool.cur->addr; 154 } else { 155 src = &r->src; 156 dst = &r->dst; 157 } 158 159 pf_counter_u64_add(&r->evaluations, 1); 160 if (pfi_kkif_match(r->kif, kif) == r->ifnot) 161 r = r->skip[PF_SKIP_IFP].ptr; 162 else if (r->direction && r->direction != pd->dir) 163 r = r->skip[PF_SKIP_DIR].ptr; 164 else if (r->af && r->af != pd->af) 165 r = r->skip[PF_SKIP_AF].ptr; 166 else if (r->proto && r->proto != pd->proto) 167 r = r->skip[PF_SKIP_PROTO].ptr; 168 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af, 169 src->neg, kif, M_GETFIB(m))) 170 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR : 171 PF_SKIP_DST_ADDR].ptr; 172 else if (src->port_op && !pf_match_port(src->port_op, 173 src->port[0], src->port[1], sport)) 174 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT : 175 PF_SKIP_DST_PORT].ptr; 176 else if (dst != NULL && 177 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL, 178 M_GETFIB(m))) 179 r = r->skip[PF_SKIP_DST_ADDR].ptr; 180 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af, 181 0, NULL, M_GETFIB(m))) 182 r = TAILQ_NEXT(r, entries); 183 else if (dst != NULL && dst->port_op && 184 !pf_match_port(dst->port_op, dst->port[0], 185 dst->port[1], dport)) 186 r = r->skip[PF_SKIP_DST_PORT].ptr; 187 else if (r->match_tag && !pf_match_tag(m, r, &tag, 188 pd->pf_mtag ? pd->pf_mtag->tag : 0)) 189 r = TAILQ_NEXT(r, entries); 190 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto != 191 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m, 192 off, &pd->hdr.tcp), r->os_fingerprint))) 193 r = TAILQ_NEXT(r, entries); 194 else { 195 if (r->tag) 196 tag = r->tag; 197 if (r->rtableid >= 0) 198 rtableid = r->rtableid; 199 if (r->anchor == NULL) { 200 rm = r; 201 if (rm->action == PF_NONAT || 202 rm->action == PF_NORDR || 203 rm->action == PF_NOBINAT) { 204 rm = NULL; 205 } 206 break; 207 } else 208 pf_step_into_anchor(anchor_stack, &asd, 209 &ruleset, rs_num, &r, NULL, NULL); 210 } 211 if (r == NULL) 212 pf_step_out_of_anchor(anchor_stack, &asd, &ruleset, 213 rs_num, &r, NULL, NULL); 214 } 215 216 if (tag > 0 && pf_tag_packet(m, pd, tag)) 217 return (NULL); 218 if (rtableid >= 0) 219 M_SETFIB(m, rtableid); 220 221 return (rm); 222 } 223 224 static int 225 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_krule *r, 226 struct pf_addr *saddr, uint16_t sport, struct pf_addr *daddr, 227 uint16_t dport, struct pf_addr *naddr, uint16_t *nport, uint16_t low, 228 uint16_t high, struct pf_ksrc_node **sn, 229 struct pf_udp_mapping **udp_mapping) 230 { 231 struct pf_state_key_cmp key; 232 struct pf_addr init_addr; 233 struct pf_srchash *sh = NULL; 234 235 bzero(&init_addr, sizeof(init_addr)); 236 237 MPASS(*udp_mapping == NULL); 238 239 /* 240 * If we are UDP and have an existing mapping we can get source port 241 * from the mapping. In this case we have to look up the src_node as 242 * pf_map_addr would. 243 */ 244 if (proto == IPPROTO_UDP && (r->rpool.opts & PF_POOL_ENDPI)) { 245 struct pf_udp_endpoint_cmp udp_source; 246 247 bzero(&udp_source, sizeof(udp_source)); 248 udp_source.af = af; 249 PF_ACPY(&udp_source.addr, saddr, af); 250 udp_source.port = sport; 251 *udp_mapping = pf_udp_mapping_find(&udp_source); 252 if (*udp_mapping) { 253 PF_ACPY(naddr, &(*udp_mapping)->endpoints[1].addr, af); 254 *nport = (*udp_mapping)->endpoints[1].port; 255 /* Try to find a src_node as per pf_map_addr(). */ 256 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR && 257 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) 258 *sn = pf_find_src_node(saddr, r, af, &sh, 0); 259 return (0); 260 } else { 261 *udp_mapping = pf_udp_mapping_create(af, saddr, sport, &init_addr, 0); 262 if (*udp_mapping == NULL) 263 return (1); 264 } 265 } 266 267 if (pf_map_addr(af, r, saddr, naddr, NULL, &init_addr, sn)) 268 goto failed; 269 270 if (proto == IPPROTO_ICMP) { 271 if (*nport == htons(ICMP_ECHO)) { 272 low = 1; 273 high = 65535; 274 } else 275 return (0); /* Don't try to modify non-echo ICMP */ 276 } 277 #ifdef INET6 278 if (proto == IPPROTO_ICMPV6) { 279 if (*nport == htons(ICMP6_ECHO_REQUEST)) { 280 low = 1; 281 high = 65535; 282 } else 283 return (0); /* Don't try to modify non-echo ICMP */ 284 } 285 #endif /* INET6 */ 286 287 bzero(&key, sizeof(key)); 288 key.af = af; 289 key.proto = proto; 290 key.port[0] = dport; 291 PF_ACPY(&key.addr[0], daddr, key.af); 292 293 do { 294 PF_ACPY(&key.addr[1], naddr, key.af); 295 if (*udp_mapping) 296 PF_ACPY(&(*udp_mapping)->endpoints[1].addr, naddr, af); 297 298 /* 299 * port search; start random, step; 300 * similar 2 portloop in in_pcbbind 301 */ 302 if (proto == IPPROTO_SCTP) { 303 key.port[1] = sport; 304 if (!pf_find_state_all_exists(&key, PF_IN)) { 305 *nport = sport; 306 return (0); 307 } else { 308 return (1); /* Fail mapping. */ 309 } 310 } else if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP || 311 proto == IPPROTO_ICMP) || (low == 0 && high == 0)) { 312 /* 313 * XXX bug: icmp states don't use the id on both sides. 314 * (traceroute -I through nat) 315 */ 316 key.port[1] = sport; 317 if (!pf_find_state_all_exists(&key, PF_IN)) { 318 *nport = sport; 319 return (0); 320 } 321 } else if (low == high) { 322 key.port[1] = htons(low); 323 if (!pf_find_state_all_exists(&key, PF_IN)) { 324 if (*udp_mapping != NULL) { 325 (*udp_mapping)->endpoints[1].port = htons(low); 326 if (pf_udp_mapping_insert(*udp_mapping) == 0) { 327 *nport = htons(low); 328 return (0); 329 } 330 } else { 331 *nport = htons(low); 332 return (0); 333 } 334 } 335 } else { 336 uint32_t tmp; 337 uint16_t cut; 338 339 if (low > high) { 340 tmp = low; 341 low = high; 342 high = tmp; 343 } 344 /* low < high */ 345 cut = arc4random() % (1 + high - low) + low; 346 /* low <= cut <= high */ 347 for (tmp = cut; tmp <= high && tmp <= 0xffff; ++tmp) { 348 if (*udp_mapping != NULL) { 349 (*udp_mapping)->endpoints[1].port = htons(tmp); 350 if (pf_udp_mapping_insert(*udp_mapping) == 0) { 351 *nport = htons(tmp); 352 return (0); 353 } 354 } else { 355 key.port[1] = htons(tmp); 356 if (!pf_find_state_all_exists(&key, PF_IN)) { 357 *nport = htons(tmp); 358 return (0); 359 } 360 } 361 } 362 tmp = cut; 363 for (tmp -= 1; tmp >= low && tmp <= 0xffff; --tmp) { 364 if (proto == IPPROTO_UDP && 365 (r->rpool.opts & PF_POOL_ENDPI)) { 366 (*udp_mapping)->endpoints[1].port = htons(tmp); 367 if (pf_udp_mapping_insert(*udp_mapping) == 0) { 368 *nport = htons(tmp); 369 return (0); 370 } 371 } else { 372 key.port[1] = htons(tmp); 373 if (!pf_find_state_all_exists(&key, PF_IN)) { 374 *nport = htons(tmp); 375 return (0); 376 } 377 } 378 } 379 } 380 381 switch (r->rpool.opts & PF_POOL_TYPEMASK) { 382 case PF_POOL_RANDOM: 383 case PF_POOL_ROUNDROBIN: 384 /* 385 * pick a different source address since we're out 386 * of free port choices for the current one. 387 */ 388 if (pf_map_addr(af, r, saddr, naddr, NULL, &init_addr, sn)) 389 return (1); 390 break; 391 case PF_POOL_NONE: 392 case PF_POOL_SRCHASH: 393 case PF_POOL_BITMASK: 394 default: 395 return (1); 396 } 397 } while (! PF_AEQ(&init_addr, naddr, af) ); 398 399 failed: 400 uma_zfree(V_pf_udp_mapping_z, *udp_mapping); 401 *udp_mapping = NULL; 402 return (1); /* none available */ 403 } 404 405 static bool 406 pf_islinklocal(const sa_family_t af, const struct pf_addr *addr) 407 { 408 if (af == AF_INET6 && IN6_IS_ADDR_LINKLOCAL(&addr->v6)) 409 return (true); 410 return (false); 411 } 412 413 static int 414 pf_get_mape_sport(sa_family_t af, u_int8_t proto, struct pf_krule *r, 415 struct pf_addr *saddr, uint16_t sport, struct pf_addr *daddr, 416 uint16_t dport, struct pf_addr *naddr, uint16_t *nport, 417 struct pf_ksrc_node **sn, struct pf_udp_mapping **udp_mapping) 418 { 419 uint16_t psmask, low, highmask; 420 uint16_t i, ahigh, cut; 421 int ashift, psidshift; 422 423 ashift = 16 - r->rpool.mape.offset; 424 psidshift = ashift - r->rpool.mape.psidlen; 425 psmask = r->rpool.mape.psid & ((1U << r->rpool.mape.psidlen) - 1); 426 psmask = psmask << psidshift; 427 highmask = (1U << psidshift) - 1; 428 429 ahigh = (1U << r->rpool.mape.offset) - 1; 430 cut = arc4random() & ahigh; 431 if (cut == 0) 432 cut = 1; 433 434 for (i = cut; i <= ahigh; i++) { 435 low = (i << ashift) | psmask; 436 if (!pf_get_sport(af, proto, r, saddr, sport, daddr, dport, 437 naddr, nport, low, low | highmask, sn, udp_mapping)) 438 return (0); 439 } 440 for (i = cut - 1; i > 0; i--) { 441 low = (i << ashift) | psmask; 442 if (!pf_get_sport(af, proto, r, saddr, sport, daddr, dport, 443 naddr, nport, low, low | highmask, sn, udp_mapping)) 444 return (0); 445 } 446 return (1); 447 } 448 449 u_short 450 pf_map_addr(sa_family_t af, struct pf_krule *r, struct pf_addr *saddr, 451 struct pf_addr *naddr, struct pfi_kkif **nkif, struct pf_addr *init_addr, 452 struct pf_ksrc_node **sn) 453 { 454 u_short reason = PFRES_MATCH; 455 struct pf_kpool *rpool = &r->rpool; 456 struct pf_addr *raddr = NULL, *rmask = NULL; 457 struct pf_srchash *sh = NULL; 458 459 /* Try to find a src_node if none was given and this 460 is a sticky-address rule. */ 461 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR && 462 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) 463 *sn = pf_find_src_node(saddr, r, af, &sh, false); 464 465 /* If a src_node was found or explicitly given and it has a non-zero 466 route address, use this address. A zeroed address is found if the 467 src node was created just a moment ago in pf_create_state and it 468 needs to be filled in with routing decision calculated here. */ 469 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) { 470 /* If the supplied address is the same as the current one we've 471 * been asked before, so tell the caller that there's no other 472 * address to be had. */ 473 if (PF_AEQ(naddr, &(*sn)->raddr, af)) { 474 reason = PFRES_MAPFAILED; 475 goto done; 476 } 477 478 PF_ACPY(naddr, &(*sn)->raddr, af); 479 if (nkif) 480 *nkif = (*sn)->rkif; 481 if (V_pf_status.debug >= PF_DEBUG_NOISY) { 482 printf("pf_map_addr: src tracking maps "); 483 pf_print_host(saddr, 0, af); 484 printf(" to "); 485 pf_print_host(naddr, 0, af); 486 if (nkif) 487 printf("@%s", (*nkif)->pfik_name); 488 printf("\n"); 489 } 490 goto done; 491 } 492 493 mtx_lock(&rpool->mtx); 494 /* Find the route using chosen algorithm. Store the found route 495 in src_node if it was given or found. */ 496 if (rpool->cur->addr.type == PF_ADDR_NOROUTE) { 497 reason = PFRES_MAPFAILED; 498 goto done_pool_mtx; 499 } 500 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 501 switch (af) { 502 #ifdef INET 503 case AF_INET: 504 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 && 505 (rpool->opts & PF_POOL_TYPEMASK) != 506 PF_POOL_ROUNDROBIN) { 507 reason = PFRES_MAPFAILED; 508 goto done_pool_mtx; 509 } 510 raddr = &rpool->cur->addr.p.dyn->pfid_addr4; 511 rmask = &rpool->cur->addr.p.dyn->pfid_mask4; 512 break; 513 #endif /* INET */ 514 #ifdef INET6 515 case AF_INET6: 516 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 && 517 (rpool->opts & PF_POOL_TYPEMASK) != 518 PF_POOL_ROUNDROBIN) { 519 reason = PFRES_MAPFAILED; 520 goto done_pool_mtx; 521 } 522 raddr = &rpool->cur->addr.p.dyn->pfid_addr6; 523 rmask = &rpool->cur->addr.p.dyn->pfid_mask6; 524 break; 525 #endif /* INET6 */ 526 } 527 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) { 528 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN) { 529 reason = PFRES_MAPFAILED; 530 goto done_pool_mtx; /* unsupported */ 531 } 532 } else { 533 raddr = &rpool->cur->addr.v.a.addr; 534 rmask = &rpool->cur->addr.v.a.mask; 535 } 536 537 switch (rpool->opts & PF_POOL_TYPEMASK) { 538 case PF_POOL_NONE: 539 PF_ACPY(naddr, raddr, af); 540 break; 541 case PF_POOL_BITMASK: 542 PF_POOLMASK(naddr, raddr, rmask, saddr, af); 543 break; 544 case PF_POOL_RANDOM: 545 if (init_addr != NULL && PF_AZERO(init_addr, af)) { 546 switch (af) { 547 #ifdef INET 548 case AF_INET: 549 rpool->counter.addr32[0] = htonl(arc4random()); 550 break; 551 #endif /* INET */ 552 #ifdef INET6 553 case AF_INET6: 554 if (rmask->addr32[3] != 0xffffffff) 555 rpool->counter.addr32[3] = 556 htonl(arc4random()); 557 else 558 break; 559 if (rmask->addr32[2] != 0xffffffff) 560 rpool->counter.addr32[2] = 561 htonl(arc4random()); 562 else 563 break; 564 if (rmask->addr32[1] != 0xffffffff) 565 rpool->counter.addr32[1] = 566 htonl(arc4random()); 567 else 568 break; 569 if (rmask->addr32[0] != 0xffffffff) 570 rpool->counter.addr32[0] = 571 htonl(arc4random()); 572 break; 573 #endif /* INET6 */ 574 } 575 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af); 576 PF_ACPY(init_addr, naddr, af); 577 578 } else { 579 PF_AINC(&rpool->counter, af); 580 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af); 581 } 582 break; 583 case PF_POOL_SRCHASH: 584 { 585 unsigned char hash[16]; 586 587 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af); 588 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af); 589 break; 590 } 591 case PF_POOL_ROUNDROBIN: 592 { 593 struct pf_kpooladdr *acur = rpool->cur; 594 595 if (rpool->cur->addr.type == PF_ADDR_TABLE) { 596 if (!pfr_pool_get(rpool->cur->addr.p.tbl, 597 &rpool->tblidx, &rpool->counter, af, NULL)) 598 goto get_addr; 599 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 600 if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, 601 &rpool->tblidx, &rpool->counter, af, pf_islinklocal)) 602 goto get_addr; 603 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af)) 604 goto get_addr; 605 606 try_next: 607 if (TAILQ_NEXT(rpool->cur, entries) == NULL) 608 rpool->cur = TAILQ_FIRST(&rpool->list); 609 else 610 rpool->cur = TAILQ_NEXT(rpool->cur, entries); 611 if (rpool->cur->addr.type == PF_ADDR_TABLE) { 612 rpool->tblidx = -1; 613 if (pfr_pool_get(rpool->cur->addr.p.tbl, 614 &rpool->tblidx, &rpool->counter, af, NULL)) { 615 /* table contains no address of type 'af' */ 616 if (rpool->cur != acur) 617 goto try_next; 618 reason = PFRES_MAPFAILED; 619 goto done_pool_mtx; 620 } 621 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { 622 rpool->tblidx = -1; 623 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, 624 &rpool->tblidx, &rpool->counter, af, pf_islinklocal)) { 625 /* table contains no address of type 'af' */ 626 if (rpool->cur != acur) 627 goto try_next; 628 reason = PFRES_MAPFAILED; 629 goto done_pool_mtx; 630 } 631 } else { 632 raddr = &rpool->cur->addr.v.a.addr; 633 rmask = &rpool->cur->addr.v.a.mask; 634 PF_ACPY(&rpool->counter, raddr, af); 635 } 636 637 get_addr: 638 PF_ACPY(naddr, &rpool->counter, af); 639 if (init_addr != NULL && PF_AZERO(init_addr, af)) 640 PF_ACPY(init_addr, naddr, af); 641 PF_AINC(&rpool->counter, af); 642 break; 643 } 644 } 645 646 if (nkif) 647 *nkif = rpool->cur->kif; 648 649 if (*sn != NULL) { 650 PF_ACPY(&(*sn)->raddr, naddr, af); 651 if (nkif) 652 (*sn)->rkif = *nkif; 653 } 654 655 if (V_pf_status.debug >= PF_DEBUG_NOISY && 656 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { 657 printf("pf_map_addr: selected address "); 658 pf_print_host(naddr, 0, af); 659 if (nkif) 660 printf("@%s", (*nkif)->pfik_name); 661 printf("\n"); 662 } 663 664 done_pool_mtx: 665 mtx_unlock(&rpool->mtx); 666 667 done: 668 if (reason) { 669 counter_u64_add(V_pf_status.counters[reason], 1); 670 } 671 672 return (reason); 673 } 674 675 u_short 676 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, 677 struct pfi_kkif *kif, struct pf_ksrc_node **sn, 678 struct pf_state_key **skp, struct pf_state_key **nkp, 679 struct pf_addr *saddr, struct pf_addr *daddr, 680 uint16_t sport, uint16_t dport, struct pf_kanchor_stackframe *anchor_stack, 681 struct pf_krule **rp, 682 struct pf_udp_mapping **udp_mapping) 683 { 684 struct pf_krule *r = NULL; 685 struct pf_addr *naddr; 686 uint16_t *nportp; 687 uint16_t low, high; 688 u_short reason; 689 690 PF_RULES_RASSERT(); 691 KASSERT(*skp == NULL, ("*skp not NULL")); 692 KASSERT(*nkp == NULL, ("*nkp not NULL")); 693 694 *rp = NULL; 695 696 if (pd->dir == PF_OUT) { 697 r = pf_match_translation(pd, m, off, kif, saddr, 698 sport, daddr, dport, PF_RULESET_BINAT, anchor_stack); 699 if (r == NULL) 700 r = pf_match_translation(pd, m, off, kif, 701 saddr, sport, daddr, dport, PF_RULESET_NAT, 702 anchor_stack); 703 } else { 704 r = pf_match_translation(pd, m, off, kif, saddr, 705 sport, daddr, dport, PF_RULESET_RDR, anchor_stack); 706 if (r == NULL) 707 r = pf_match_translation(pd, m, off, kif, 708 saddr, sport, daddr, dport, PF_RULESET_BINAT, 709 anchor_stack); 710 } 711 712 if (r == NULL) 713 return (PFRES_MAX); 714 715 switch (r->action) { 716 case PF_NONAT: 717 case PF_NOBINAT: 718 case PF_NORDR: 719 return (PFRES_MAX); 720 } 721 722 *skp = pf_state_key_setup(pd, m, off, saddr, daddr, sport, dport); 723 if (*skp == NULL) 724 return (PFRES_MEMORY); 725 *nkp = pf_state_key_clone(*skp); 726 if (*nkp == NULL) { 727 uma_zfree(V_pf_state_key_z, *skp); 728 *skp = NULL; 729 return (PFRES_MEMORY); 730 } 731 732 naddr = &(*nkp)->addr[1]; 733 nportp = &(*nkp)->port[1]; 734 735 switch (r->action) { 736 case PF_NAT: 737 if (pd->proto == IPPROTO_ICMP) { 738 low = 1; 739 high = 65535; 740 } else { 741 low = r->rpool.proxy_port[0]; 742 high = r->rpool.proxy_port[1]; 743 } 744 if (r->rpool.mape.offset > 0) { 745 if (pf_get_mape_sport(pd->af, pd->proto, r, saddr, 746 sport, daddr, dport, naddr, nportp, sn, udp_mapping)) { 747 DPFPRINTF(PF_DEBUG_MISC, 748 ("pf: MAP-E port allocation (%u/%u/%u)" 749 " failed\n", 750 r->rpool.mape.offset, 751 r->rpool.mape.psidlen, 752 r->rpool.mape.psid)); 753 reason = PFRES_MAPFAILED; 754 goto notrans; 755 } 756 } else if (pf_get_sport(pd->af, pd->proto, r, saddr, sport, 757 daddr, dport, naddr, nportp, low, high, sn, udp_mapping)) { 758 DPFPRINTF(PF_DEBUG_MISC, 759 ("pf: NAT proxy port allocation (%u-%u) failed\n", 760 r->rpool.proxy_port[0], r->rpool.proxy_port[1])); 761 reason = PFRES_MAPFAILED; 762 goto notrans; 763 } 764 break; 765 case PF_BINAT: 766 switch (pd->dir) { 767 case PF_OUT: 768 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){ 769 switch (pd->af) { 770 #ifdef INET 771 case AF_INET: 772 if (r->rpool.cur->addr.p.dyn-> 773 pfid_acnt4 < 1) { 774 reason = PFRES_MAPFAILED; 775 goto notrans; 776 } 777 PF_POOLMASK(naddr, 778 &r->rpool.cur->addr.p.dyn-> 779 pfid_addr4, 780 &r->rpool.cur->addr.p.dyn-> 781 pfid_mask4, saddr, AF_INET); 782 break; 783 #endif /* INET */ 784 #ifdef INET6 785 case AF_INET6: 786 if (r->rpool.cur->addr.p.dyn-> 787 pfid_acnt6 < 1) { 788 reason = PFRES_MAPFAILED; 789 goto notrans; 790 } 791 PF_POOLMASK(naddr, 792 &r->rpool.cur->addr.p.dyn-> 793 pfid_addr6, 794 &r->rpool.cur->addr.p.dyn-> 795 pfid_mask6, saddr, AF_INET6); 796 break; 797 #endif /* INET6 */ 798 } 799 } else 800 PF_POOLMASK(naddr, 801 &r->rpool.cur->addr.v.a.addr, 802 &r->rpool.cur->addr.v.a.mask, saddr, 803 pd->af); 804 break; 805 case PF_IN: 806 if (r->src.addr.type == PF_ADDR_DYNIFTL) { 807 switch (pd->af) { 808 #ifdef INET 809 case AF_INET: 810 if (r->src.addr.p.dyn->pfid_acnt4 < 1) { 811 reason = PFRES_MAPFAILED; 812 goto notrans; 813 } 814 PF_POOLMASK(naddr, 815 &r->src.addr.p.dyn->pfid_addr4, 816 &r->src.addr.p.dyn->pfid_mask4, 817 daddr, AF_INET); 818 break; 819 #endif /* INET */ 820 #ifdef INET6 821 case AF_INET6: 822 if (r->src.addr.p.dyn->pfid_acnt6 < 1) { 823 reason = PFRES_MAPFAILED; 824 goto notrans; 825 } 826 PF_POOLMASK(naddr, 827 &r->src.addr.p.dyn->pfid_addr6, 828 &r->src.addr.p.dyn->pfid_mask6, 829 daddr, AF_INET6); 830 break; 831 #endif /* INET6 */ 832 } 833 } else 834 PF_POOLMASK(naddr, &r->src.addr.v.a.addr, 835 &r->src.addr.v.a.mask, daddr, pd->af); 836 break; 837 } 838 break; 839 case PF_RDR: { 840 struct pf_state_key_cmp key; 841 int tries; 842 uint16_t cut, low, high, nport; 843 844 reason = pf_map_addr(pd->af, r, saddr, naddr, NULL, NULL, sn); 845 if (reason != 0) 846 goto notrans; 847 if ((r->rpool.opts & PF_POOL_TYPEMASK) == PF_POOL_BITMASK) 848 PF_POOLMASK(naddr, naddr, &r->rpool.cur->addr.v.a.mask, 849 daddr, pd->af); 850 851 /* Do not change SCTP ports. */ 852 if (pd->proto == IPPROTO_SCTP) 853 break; 854 855 if (r->rpool.proxy_port[1]) { 856 uint32_t tmp_nport; 857 858 tmp_nport = ((ntohs(dport) - ntohs(r->dst.port[0])) % 859 (r->rpool.proxy_port[1] - r->rpool.proxy_port[0] + 860 1)) + r->rpool.proxy_port[0]; 861 862 /* Wrap around if necessary. */ 863 if (tmp_nport > 65535) 864 tmp_nport -= 65535; 865 nport = htons((uint16_t)tmp_nport); 866 } else if (r->rpool.proxy_port[0]) 867 nport = htons(r->rpool.proxy_port[0]); 868 else 869 nport = dport; 870 871 /* 872 * Update the destination port. 873 */ 874 *nportp = nport; 875 876 /* 877 * Do we have a source port conflict in the stack state? Try to 878 * modulate the source port if so. Note that this is racy since 879 * the state lookup may not find any matches here but will once 880 * pf_create_state() actually instantiates the state. 881 */ 882 bzero(&key, sizeof(key)); 883 key.af = pd->af; 884 key.proto = pd->proto; 885 key.port[0] = sport; 886 PF_ACPY(&key.addr[0], saddr, key.af); 887 key.port[1] = nport; 888 PF_ACPY(&key.addr[1], naddr, key.af); 889 890 if (!pf_find_state_all_exists(&key, PF_OUT)) 891 break; 892 893 tries = 0; 894 895 low = 50001; /* XXX-MJ PF_NAT_PROXY_PORT_LOW/HIGH */ 896 high = 65535; 897 cut = arc4random() % (1 + high - low) + low; 898 for (uint32_t tmp = cut; 899 tmp <= high && tmp <= UINT16_MAX && 900 tries < V_pf_rdr_srcport_rewrite_tries; 901 tmp++, tries++) { 902 key.port[0] = htons(tmp); 903 if (!pf_find_state_all_exists(&key, PF_OUT)) { 904 /* Update the source port. */ 905 (*nkp)->port[0] = htons(tmp); 906 goto out; 907 } 908 } 909 for (uint32_t tmp = cut - 1; 910 tmp >= low && tries < V_pf_rdr_srcport_rewrite_tries; 911 tmp--, tries++) { 912 key.port[0] = htons(tmp); 913 if (!pf_find_state_all_exists(&key, PF_OUT)) { 914 /* Update the source port. */ 915 (*nkp)->port[0] = htons(tmp); 916 goto out; 917 } 918 } 919 920 /* 921 * We failed to find a match. Push on ahead anyway, let 922 * pf_state_insert() be the arbiter of whether the state 923 * conflict is tolerable. In particular, with TCP connections 924 * the state may be reused if the TCP state is terminal. 925 */ 926 DPFPRINTF(PF_DEBUG_MISC, 927 ("pf: RDR source port allocation failed\n")); 928 break; 929 930 out: 931 DPFPRINTF(PF_DEBUG_MISC, 932 ("pf: RDR source port allocation %u->%u\n", 933 ntohs(sport), ntohs((*nkp)->port[0]))); 934 break; 935 } 936 default: 937 panic("%s: unknown action %u", __func__, r->action); 938 } 939 940 /* Return success only if translation really happened. */ 941 if (bcmp(*skp, *nkp, sizeof(struct pf_state_key_cmp))) { 942 *rp = r; 943 return (PFRES_MATCH); 944 } 945 946 reason = PFRES_MAX; 947 notrans: 948 uma_zfree(V_pf_state_key_z, *nkp); 949 uma_zfree(V_pf_state_key_z, *skp); 950 *skp = *nkp = NULL; 951 *sn = NULL; 952 953 return (reason); 954 } 955