1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa 5 * Copyright (c) 2014 Yandex LLC 6 * Copyright (c) 2014 Alexander V. Chernikov 7 * 8 * Supported by: Valeria Paoli 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 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 /* 36 * Control socket and rule management routines for ipfw. 37 * Control is currently implemented via IP_FW3 setsockopt() code. 38 */ 39 40 #include "opt_ipfw.h" 41 #include "opt_inet.h" 42 #ifndef INET 43 #error IPFIREWALL requires INET. 44 #endif /* INET */ 45 #include "opt_inet6.h" 46 47 #include <sys/param.h> 48 #include <sys/systm.h> 49 #include <sys/malloc.h> 50 #include <sys/mbuf.h> /* struct m_tag used by nested headers */ 51 #include <sys/kernel.h> 52 #include <sys/lock.h> 53 #include <sys/priv.h> 54 #include <sys/proc.h> 55 #include <sys/rwlock.h> 56 #include <sys/rmlock.h> 57 #include <sys/socket.h> 58 #include <sys/socketvar.h> 59 #include <sys/sysctl.h> 60 #include <sys/syslog.h> 61 #include <sys/fnv_hash.h> 62 #include <net/if.h> 63 #include <net/route.h> 64 #include <net/vnet.h> 65 #include <vm/vm.h> 66 #include <vm/vm_extern.h> 67 68 #include <netinet/in.h> 69 #include <netinet/ip_var.h> /* hooks */ 70 #include <netinet/ip_fw.h> 71 72 #include <netpfil/ipfw/ip_fw_private.h> 73 #include <netpfil/ipfw/ip_fw_table.h> 74 75 #ifdef MAC 76 #include <security/mac/mac_framework.h> 77 #endif 78 79 static int ipfw_ctl(struct sockopt *sopt); 80 static int check_ipfw_rule_body(ipfw_insn *cmd, int cmd_len, 81 struct rule_check_info *ci); 82 static int check_ipfw_rule1(struct ip_fw_rule *rule, int size, 83 struct rule_check_info *ci); 84 static int check_ipfw_rule0(struct ip_fw_rule0 *rule, int size, 85 struct rule_check_info *ci); 86 static int rewrite_rule_uidx(struct ip_fw_chain *chain, 87 struct rule_check_info *ci); 88 89 #define NAMEDOBJ_HASH_SIZE 32 90 91 struct namedobj_instance { 92 struct namedobjects_head *names; 93 struct namedobjects_head *values; 94 uint32_t nn_size; /* names hash size */ 95 uint32_t nv_size; /* number hash size */ 96 u_long *idx_mask; /* used items bitmask */ 97 uint32_t max_blocks; /* number of "long" blocks in bitmask */ 98 uint32_t count; /* number of items */ 99 uint16_t free_off[IPFW_MAX_SETS]; /* first possible free offset */ 100 objhash_hash_f *hash_f; 101 objhash_cmp_f *cmp_f; 102 }; 103 #define BLOCK_ITEMS (8 * sizeof(u_long)) /* Number of items for ffsl() */ 104 105 static uint32_t objhash_hash_name(struct namedobj_instance *ni, 106 const void *key, uint32_t kopt); 107 static uint32_t objhash_hash_idx(struct namedobj_instance *ni, uint32_t val); 108 static int objhash_cmp_name(struct named_object *no, const void *name, 109 uint32_t set); 110 111 MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 112 113 static int dump_config(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 114 struct sockopt_data *sd); 115 static int add_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 116 struct sockopt_data *sd); 117 static int del_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 118 struct sockopt_data *sd); 119 static int clear_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 120 struct sockopt_data *sd); 121 static int move_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 122 struct sockopt_data *sd); 123 static int manage_sets(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 124 struct sockopt_data *sd); 125 static int dump_soptcodes(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 126 struct sockopt_data *sd); 127 static int dump_srvobjects(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 128 struct sockopt_data *sd); 129 130 /* ctl3 handler data */ 131 struct mtx ctl3_lock; 132 #define CTL3_LOCK_INIT() mtx_init(&ctl3_lock, "ctl3_lock", NULL, MTX_DEF) 133 #define CTL3_LOCK_DESTROY() mtx_destroy(&ctl3_lock) 134 #define CTL3_LOCK() mtx_lock(&ctl3_lock) 135 #define CTL3_UNLOCK() mtx_unlock(&ctl3_lock) 136 137 static struct ipfw_sopt_handler *ctl3_handlers; 138 static size_t ctl3_hsize; 139 static uint64_t ctl3_refct, ctl3_gencnt; 140 #define CTL3_SMALLBUF 4096 /* small page-size write buffer */ 141 #define CTL3_LARGEBUF 16 * 1024 * 1024 /* handle large rulesets */ 142 143 static int ipfw_flush_sopt_data(struct sockopt_data *sd); 144 145 static struct ipfw_sopt_handler scodes[] = { 146 { IP_FW_XGET, 0, HDIR_GET, dump_config }, 147 { IP_FW_XADD, 0, HDIR_BOTH, add_rules }, 148 { IP_FW_XDEL, 0, HDIR_BOTH, del_rules }, 149 { IP_FW_XZERO, 0, HDIR_SET, clear_rules }, 150 { IP_FW_XRESETLOG, 0, HDIR_SET, clear_rules }, 151 { IP_FW_XMOVE, 0, HDIR_SET, move_rules }, 152 { IP_FW_SET_SWAP, 0, HDIR_SET, manage_sets }, 153 { IP_FW_SET_MOVE, 0, HDIR_SET, manage_sets }, 154 { IP_FW_SET_ENABLE, 0, HDIR_SET, manage_sets }, 155 { IP_FW_DUMP_SOPTCODES, 0, HDIR_GET, dump_soptcodes }, 156 { IP_FW_DUMP_SRVOBJECTS,0, HDIR_GET, dump_srvobjects }, 157 }; 158 159 static int 160 set_legacy_obj_kidx(struct ip_fw_chain *ch, struct ip_fw_rule0 *rule); 161 static struct opcode_obj_rewrite *find_op_rw(ipfw_insn *cmd, 162 uint16_t *puidx, uint8_t *ptype); 163 static int ref_rule_objects(struct ip_fw_chain *ch, struct ip_fw *rule, 164 struct rule_check_info *ci, struct obj_idx *oib, struct tid_info *ti); 165 static int ref_opcode_object(struct ip_fw_chain *ch, ipfw_insn *cmd, 166 struct tid_info *ti, struct obj_idx *pidx, int *unresolved); 167 static void unref_rule_objects(struct ip_fw_chain *chain, struct ip_fw *rule); 168 static void unref_oib_objects(struct ip_fw_chain *ch, ipfw_insn *cmd, 169 struct obj_idx *oib, struct obj_idx *end); 170 static int export_objhash_ntlv(struct namedobj_instance *ni, uint16_t kidx, 171 struct sockopt_data *sd); 172 173 /* 174 * Opcode object rewriter variables 175 */ 176 struct opcode_obj_rewrite *ctl3_rewriters; 177 static size_t ctl3_rsize; 178 179 /* 180 * static variables followed by global ones 181 */ 182 183 VNET_DEFINE_STATIC(uma_zone_t, ipfw_cntr_zone); 184 #define V_ipfw_cntr_zone VNET(ipfw_cntr_zone) 185 186 void 187 ipfw_init_counters(void) 188 { 189 190 V_ipfw_cntr_zone = uma_zcreate("IPFW counters", 191 IPFW_RULE_CNTR_SIZE, NULL, NULL, NULL, NULL, 192 UMA_ALIGN_PTR, UMA_ZONE_PCPU); 193 } 194 195 void 196 ipfw_destroy_counters(void) 197 { 198 199 uma_zdestroy(V_ipfw_cntr_zone); 200 } 201 202 struct ip_fw * 203 ipfw_alloc_rule(struct ip_fw_chain *chain, size_t rulesize) 204 { 205 struct ip_fw *rule; 206 207 rule = malloc(rulesize, M_IPFW, M_WAITOK | M_ZERO); 208 rule->cntr = uma_zalloc_pcpu(V_ipfw_cntr_zone, M_WAITOK | M_ZERO); 209 rule->refcnt = 1; 210 211 return (rule); 212 } 213 214 void 215 ipfw_free_rule(struct ip_fw *rule) 216 { 217 218 /* 219 * We don't release refcnt here, since this function 220 * can be called without any locks held. The caller 221 * must release reference under IPFW_UH_WLOCK, and then 222 * call this function if refcount becomes 1. 223 */ 224 if (rule->refcnt > 1) 225 return; 226 uma_zfree_pcpu(V_ipfw_cntr_zone, rule->cntr); 227 free(rule, M_IPFW); 228 } 229 230 /* 231 * Find the smallest rule >= key, id. 232 * We could use bsearch but it is so simple that we code it directly 233 */ 234 int 235 ipfw_find_rule(struct ip_fw_chain *chain, uint32_t key, uint32_t id) 236 { 237 int i, lo, hi; 238 struct ip_fw *r; 239 240 for (lo = 0, hi = chain->n_rules - 1; lo < hi;) { 241 i = (lo + hi) / 2; 242 r = chain->map[i]; 243 if (r->rulenum < key) 244 lo = i + 1; /* continue from the next one */ 245 else if (r->rulenum > key) 246 hi = i; /* this might be good */ 247 else if (r->id < id) 248 lo = i + 1; /* continue from the next one */ 249 else /* r->id >= id */ 250 hi = i; /* this might be good */ 251 } 252 return hi; 253 } 254 255 /* 256 * Builds skipto cache on rule set @map. 257 */ 258 static void 259 update_skipto_cache(struct ip_fw_chain *chain, struct ip_fw **map) 260 { 261 int *smap, rulenum; 262 int i, mi; 263 264 IPFW_UH_WLOCK_ASSERT(chain); 265 266 mi = 0; 267 rulenum = map[mi]->rulenum; 268 smap = chain->idxmap_back; 269 270 if (smap == NULL) 271 return; 272 273 for (i = 0; i < 65536; i++) { 274 smap[i] = mi; 275 /* Use the same rule index until i < rulenum */ 276 if (i != rulenum || i == 65535) 277 continue; 278 /* Find next rule with num > i */ 279 rulenum = map[++mi]->rulenum; 280 while (rulenum == i) 281 rulenum = map[++mi]->rulenum; 282 } 283 } 284 285 /* 286 * Swaps prepared (backup) index with current one. 287 */ 288 static void 289 swap_skipto_cache(struct ip_fw_chain *chain) 290 { 291 int *map; 292 293 IPFW_UH_WLOCK_ASSERT(chain); 294 IPFW_WLOCK_ASSERT(chain); 295 296 map = chain->idxmap; 297 chain->idxmap = chain->idxmap_back; 298 chain->idxmap_back = map; 299 } 300 301 /* 302 * Allocate and initialize skipto cache. 303 */ 304 void 305 ipfw_init_skipto_cache(struct ip_fw_chain *chain) 306 { 307 int *idxmap, *idxmap_back; 308 309 idxmap = malloc(65536 * sizeof(int), M_IPFW, M_WAITOK | M_ZERO); 310 idxmap_back = malloc(65536 * sizeof(int), M_IPFW, M_WAITOK); 311 312 /* 313 * Note we may be called at any time after initialization, 314 * for example, on first skipto rule, so we need to 315 * provide valid chain->idxmap on return 316 */ 317 318 IPFW_UH_WLOCK(chain); 319 if (chain->idxmap != NULL) { 320 IPFW_UH_WUNLOCK(chain); 321 free(idxmap, M_IPFW); 322 free(idxmap_back, M_IPFW); 323 return; 324 } 325 326 /* Set backup pointer first to permit building cache */ 327 chain->idxmap_back = idxmap_back; 328 update_skipto_cache(chain, chain->map); 329 IPFW_WLOCK(chain); 330 /* It is now safe to set chain->idxmap ptr */ 331 chain->idxmap = idxmap; 332 swap_skipto_cache(chain); 333 IPFW_WUNLOCK(chain); 334 IPFW_UH_WUNLOCK(chain); 335 } 336 337 /* 338 * Destroys skipto cache. 339 */ 340 void 341 ipfw_destroy_skipto_cache(struct ip_fw_chain *chain) 342 { 343 344 if (chain->idxmap != NULL) 345 free(chain->idxmap, M_IPFW); 346 if (chain->idxmap != NULL) 347 free(chain->idxmap_back, M_IPFW); 348 } 349 350 /* 351 * allocate a new map, returns the chain locked. extra is the number 352 * of entries to add or delete. 353 */ 354 static struct ip_fw ** 355 get_map(struct ip_fw_chain *chain, int extra, int locked) 356 { 357 358 for (;;) { 359 struct ip_fw **map; 360 u_int i, mflags; 361 362 mflags = M_ZERO | ((locked != 0) ? M_NOWAIT : M_WAITOK); 363 364 i = chain->n_rules + extra; 365 map = malloc(i * sizeof(struct ip_fw *), M_IPFW, mflags); 366 if (map == NULL) { 367 printf("%s: cannot allocate map\n", __FUNCTION__); 368 return NULL; 369 } 370 if (!locked) 371 IPFW_UH_WLOCK(chain); 372 if (i >= chain->n_rules + extra) /* good */ 373 return map; 374 /* otherwise we lost the race, free and retry */ 375 if (!locked) 376 IPFW_UH_WUNLOCK(chain); 377 free(map, M_IPFW); 378 } 379 } 380 381 /* 382 * swap the maps. It is supposed to be called with IPFW_UH_WLOCK 383 */ 384 static struct ip_fw ** 385 swap_map(struct ip_fw_chain *chain, struct ip_fw **new_map, int new_len) 386 { 387 struct ip_fw **old_map; 388 389 IPFW_WLOCK(chain); 390 chain->id++; 391 chain->n_rules = new_len; 392 old_map = chain->map; 393 chain->map = new_map; 394 swap_skipto_cache(chain); 395 IPFW_WUNLOCK(chain); 396 return old_map; 397 } 398 399 static void 400 export_cntr1_base(struct ip_fw *krule, struct ip_fw_bcounter *cntr) 401 { 402 struct timeval boottime; 403 404 cntr->size = sizeof(*cntr); 405 406 if (krule->cntr != NULL) { 407 cntr->pcnt = counter_u64_fetch(krule->cntr); 408 cntr->bcnt = counter_u64_fetch(krule->cntr + 1); 409 cntr->timestamp = krule->timestamp; 410 } 411 if (cntr->timestamp > 0) { 412 getboottime(&boottime); 413 cntr->timestamp += boottime.tv_sec; 414 } 415 } 416 417 static void 418 export_cntr0_base(struct ip_fw *krule, struct ip_fw_bcounter0 *cntr) 419 { 420 struct timeval boottime; 421 422 if (krule->cntr != NULL) { 423 cntr->pcnt = counter_u64_fetch(krule->cntr); 424 cntr->bcnt = counter_u64_fetch(krule->cntr + 1); 425 cntr->timestamp = krule->timestamp; 426 } 427 if (cntr->timestamp > 0) { 428 getboottime(&boottime); 429 cntr->timestamp += boottime.tv_sec; 430 } 431 } 432 433 /* 434 * Copies rule @urule from v1 userland format (current). 435 * to kernel @krule. 436 * Assume @krule is zeroed. 437 */ 438 static void 439 import_rule1(struct rule_check_info *ci) 440 { 441 struct ip_fw_rule *urule; 442 struct ip_fw *krule; 443 444 urule = (struct ip_fw_rule *)ci->urule; 445 krule = (struct ip_fw *)ci->krule; 446 447 /* copy header */ 448 krule->act_ofs = urule->act_ofs; 449 krule->cmd_len = urule->cmd_len; 450 krule->rulenum = urule->rulenum; 451 krule->set = urule->set; 452 krule->flags = urule->flags; 453 454 /* Save rulenum offset */ 455 ci->urule_numoff = offsetof(struct ip_fw_rule, rulenum); 456 457 /* Copy opcodes */ 458 memcpy(krule->cmd, urule->cmd, krule->cmd_len * sizeof(uint32_t)); 459 } 460 461 /* 462 * Export rule into v1 format (Current). 463 * Layout: 464 * [ ipfw_obj_tlv(IPFW_TLV_RULE_ENT) 465 * [ ip_fw_rule ] OR 466 * [ ip_fw_bcounter ip_fw_rule] (depends on rcntrs). 467 * ] 468 * Assume @data is zeroed. 469 */ 470 static void 471 export_rule1(struct ip_fw *krule, caddr_t data, int len, int rcntrs) 472 { 473 struct ip_fw_bcounter *cntr; 474 struct ip_fw_rule *urule; 475 ipfw_obj_tlv *tlv; 476 477 /* Fill in TLV header */ 478 tlv = (ipfw_obj_tlv *)data; 479 tlv->type = IPFW_TLV_RULE_ENT; 480 tlv->length = len; 481 482 if (rcntrs != 0) { 483 /* Copy counters */ 484 cntr = (struct ip_fw_bcounter *)(tlv + 1); 485 urule = (struct ip_fw_rule *)(cntr + 1); 486 export_cntr1_base(krule, cntr); 487 } else 488 urule = (struct ip_fw_rule *)(tlv + 1); 489 490 /* copy header */ 491 urule->act_ofs = krule->act_ofs; 492 urule->cmd_len = krule->cmd_len; 493 urule->rulenum = krule->rulenum; 494 urule->set = krule->set; 495 urule->flags = krule->flags; 496 urule->id = krule->id; 497 498 /* Copy opcodes */ 499 memcpy(urule->cmd, krule->cmd, krule->cmd_len * sizeof(uint32_t)); 500 } 501 502 /* 503 * Copies rule @urule from FreeBSD8 userland format (v0) 504 * to kernel @krule. 505 * Assume @krule is zeroed. 506 */ 507 static void 508 import_rule0(struct rule_check_info *ci) 509 { 510 struct ip_fw_rule0 *urule; 511 struct ip_fw *krule; 512 int cmdlen, l; 513 ipfw_insn *cmd; 514 ipfw_insn_limit *lcmd; 515 ipfw_insn_if *cmdif; 516 517 urule = (struct ip_fw_rule0 *)ci->urule; 518 krule = (struct ip_fw *)ci->krule; 519 520 /* copy header */ 521 krule->act_ofs = urule->act_ofs; 522 krule->cmd_len = urule->cmd_len; 523 krule->rulenum = urule->rulenum; 524 krule->set = urule->set; 525 if ((urule->_pad & 1) != 0) 526 krule->flags |= IPFW_RULE_NOOPT; 527 528 /* Save rulenum offset */ 529 ci->urule_numoff = offsetof(struct ip_fw_rule0, rulenum); 530 531 /* Copy opcodes */ 532 memcpy(krule->cmd, urule->cmd, krule->cmd_len * sizeof(uint32_t)); 533 534 /* 535 * Alter opcodes: 536 * 1) convert tablearg value from 65535 to 0 537 * 2) Add high bit to O_SETFIB/O_SETDSCP values (to make room 538 * for targ). 539 * 3) convert table number in iface opcodes to u16 540 * 4) convert old `nat global` into new 65535 541 */ 542 l = krule->cmd_len; 543 cmd = krule->cmd; 544 cmdlen = 0; 545 546 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 547 cmdlen = F_LEN(cmd); 548 549 switch (cmd->opcode) { 550 /* Opcodes supporting tablearg */ 551 case O_TAG: 552 case O_TAGGED: 553 case O_PIPE: 554 case O_QUEUE: 555 case O_DIVERT: 556 case O_TEE: 557 case O_SKIPTO: 558 case O_CALLRETURN: 559 case O_NETGRAPH: 560 case O_NGTEE: 561 case O_NAT: 562 if (cmd->arg1 == IP_FW_TABLEARG) 563 cmd->arg1 = IP_FW_TARG; 564 else if (cmd->arg1 == 0) 565 cmd->arg1 = IP_FW_NAT44_GLOBAL; 566 break; 567 case O_SETFIB: 568 case O_SETDSCP: 569 if (cmd->arg1 == IP_FW_TABLEARG) 570 cmd->arg1 = IP_FW_TARG; 571 else 572 cmd->arg1 |= 0x8000; 573 break; 574 case O_LIMIT: 575 lcmd = (ipfw_insn_limit *)cmd; 576 if (lcmd->conn_limit == IP_FW_TABLEARG) 577 lcmd->conn_limit = IP_FW_TARG; 578 break; 579 /* Interface tables */ 580 case O_XMIT: 581 case O_RECV: 582 case O_VIA: 583 /* Interface table, possibly */ 584 cmdif = (ipfw_insn_if *)cmd; 585 if (cmdif->name[0] != '\1') 586 break; 587 588 cmdif->p.kidx = (uint16_t)cmdif->p.glob; 589 break; 590 } 591 } 592 } 593 594 /* 595 * Copies rule @krule from kernel to FreeBSD8 userland format (v0) 596 */ 597 static void 598 export_rule0(struct ip_fw *krule, struct ip_fw_rule0 *urule, int len) 599 { 600 int cmdlen, l; 601 ipfw_insn *cmd; 602 ipfw_insn_limit *lcmd; 603 ipfw_insn_if *cmdif; 604 605 /* copy header */ 606 memset(urule, 0, len); 607 urule->act_ofs = krule->act_ofs; 608 urule->cmd_len = krule->cmd_len; 609 urule->rulenum = krule->rulenum; 610 urule->set = krule->set; 611 if ((krule->flags & IPFW_RULE_NOOPT) != 0) 612 urule->_pad |= 1; 613 614 /* Copy opcodes */ 615 memcpy(urule->cmd, krule->cmd, krule->cmd_len * sizeof(uint32_t)); 616 617 /* Export counters */ 618 export_cntr0_base(krule, (struct ip_fw_bcounter0 *)&urule->pcnt); 619 620 /* 621 * Alter opcodes: 622 * 1) convert tablearg value from 0 to 65535 623 * 2) Remove highest bit from O_SETFIB/O_SETDSCP values. 624 * 3) convert table number in iface opcodes to int 625 */ 626 l = urule->cmd_len; 627 cmd = urule->cmd; 628 cmdlen = 0; 629 630 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 631 cmdlen = F_LEN(cmd); 632 633 switch (cmd->opcode) { 634 /* Opcodes supporting tablearg */ 635 case O_TAG: 636 case O_TAGGED: 637 case O_PIPE: 638 case O_QUEUE: 639 case O_DIVERT: 640 case O_TEE: 641 case O_SKIPTO: 642 case O_CALLRETURN: 643 case O_NETGRAPH: 644 case O_NGTEE: 645 case O_NAT: 646 if (cmd->arg1 == IP_FW_TARG) 647 cmd->arg1 = IP_FW_TABLEARG; 648 else if (cmd->arg1 == IP_FW_NAT44_GLOBAL) 649 cmd->arg1 = 0; 650 break; 651 case O_SETFIB: 652 case O_SETDSCP: 653 if (cmd->arg1 == IP_FW_TARG) 654 cmd->arg1 = IP_FW_TABLEARG; 655 else 656 cmd->arg1 &= ~0x8000; 657 break; 658 case O_LIMIT: 659 lcmd = (ipfw_insn_limit *)cmd; 660 if (lcmd->conn_limit == IP_FW_TARG) 661 lcmd->conn_limit = IP_FW_TABLEARG; 662 break; 663 /* Interface tables */ 664 case O_XMIT: 665 case O_RECV: 666 case O_VIA: 667 /* Interface table, possibly */ 668 cmdif = (ipfw_insn_if *)cmd; 669 if (cmdif->name[0] != '\1') 670 break; 671 672 cmdif->p.glob = cmdif->p.kidx; 673 break; 674 } 675 } 676 } 677 678 /* 679 * Add new rule(s) to the list possibly creating rule number for each. 680 * Update the rule_number in the input struct so the caller knows it as well. 681 * Must be called without IPFW_UH held 682 */ 683 static int 684 commit_rules(struct ip_fw_chain *chain, struct rule_check_info *rci, int count) 685 { 686 int error, i, insert_before, tcount; 687 uint16_t rulenum, *pnum; 688 struct rule_check_info *ci; 689 struct ip_fw *krule; 690 struct ip_fw **map; /* the new array of pointers */ 691 692 /* Check if we need to do table/obj index remap */ 693 tcount = 0; 694 for (ci = rci, i = 0; i < count; ci++, i++) { 695 if (ci->object_opcodes == 0) 696 continue; 697 698 /* 699 * Rule has some object opcodes. 700 * We need to find (and create non-existing) 701 * kernel objects, and reference existing ones. 702 */ 703 error = rewrite_rule_uidx(chain, ci); 704 if (error != 0) { 705 /* 706 * rewrite failed, state for current rule 707 * has been reverted. Check if we need to 708 * revert more. 709 */ 710 if (tcount > 0) { 711 /* 712 * We have some more table rules 713 * we need to rollback. 714 */ 715 716 IPFW_UH_WLOCK(chain); 717 while (ci != rci) { 718 ci--; 719 if (ci->object_opcodes == 0) 720 continue; 721 unref_rule_objects(chain,ci->krule); 722 } 723 IPFW_UH_WUNLOCK(chain); 724 } 725 726 return (error); 727 } 728 729 tcount++; 730 } 731 732 /* get_map returns with IPFW_UH_WLOCK if successful */ 733 map = get_map(chain, count, 0 /* not locked */); 734 if (map == NULL) { 735 if (tcount > 0) { 736 /* Unbind tables */ 737 IPFW_UH_WLOCK(chain); 738 for (ci = rci, i = 0; i < count; ci++, i++) { 739 if (ci->object_opcodes == 0) 740 continue; 741 742 unref_rule_objects(chain, ci->krule); 743 } 744 IPFW_UH_WUNLOCK(chain); 745 } 746 747 return (ENOSPC); 748 } 749 750 if (V_autoinc_step < 1) 751 V_autoinc_step = 1; 752 else if (V_autoinc_step > 1000) 753 V_autoinc_step = 1000; 754 755 /* FIXME: Handle count > 1 */ 756 ci = rci; 757 krule = ci->krule; 758 rulenum = krule->rulenum; 759 760 /* find the insertion point, we will insert before */ 761 insert_before = rulenum ? rulenum + 1 : IPFW_DEFAULT_RULE; 762 i = ipfw_find_rule(chain, insert_before, 0); 763 /* duplicate first part */ 764 if (i > 0) 765 bcopy(chain->map, map, i * sizeof(struct ip_fw *)); 766 map[i] = krule; 767 /* duplicate remaining part, we always have the default rule */ 768 bcopy(chain->map + i, map + i + 1, 769 sizeof(struct ip_fw *) *(chain->n_rules - i)); 770 if (rulenum == 0) { 771 /* Compute rule number and write it back */ 772 rulenum = i > 0 ? map[i-1]->rulenum : 0; 773 if (rulenum < IPFW_DEFAULT_RULE - V_autoinc_step) 774 rulenum += V_autoinc_step; 775 krule->rulenum = rulenum; 776 /* Save number to userland rule */ 777 pnum = (uint16_t *)((caddr_t)ci->urule + ci->urule_numoff); 778 *pnum = rulenum; 779 } 780 781 krule->id = chain->id + 1; 782 update_skipto_cache(chain, map); 783 map = swap_map(chain, map, chain->n_rules + 1); 784 chain->static_len += RULEUSIZE0(krule); 785 IPFW_UH_WUNLOCK(chain); 786 if (map) 787 free(map, M_IPFW); 788 return (0); 789 } 790 791 int 792 ipfw_add_protected_rule(struct ip_fw_chain *chain, struct ip_fw *rule, 793 int locked) 794 { 795 struct ip_fw **map; 796 797 map = get_map(chain, 1, locked); 798 if (map == NULL) 799 return (ENOMEM); 800 if (chain->n_rules > 0) 801 bcopy(chain->map, map, 802 chain->n_rules * sizeof(struct ip_fw *)); 803 map[chain->n_rules] = rule; 804 rule->rulenum = IPFW_DEFAULT_RULE; 805 rule->set = RESVD_SET; 806 rule->id = chain->id + 1; 807 /* We add rule in the end of chain, no need to update skipto cache */ 808 map = swap_map(chain, map, chain->n_rules + 1); 809 chain->static_len += RULEUSIZE0(rule); 810 IPFW_UH_WUNLOCK(chain); 811 free(map, M_IPFW); 812 return (0); 813 } 814 815 /* 816 * Adds @rule to the list of rules to reap 817 */ 818 void 819 ipfw_reap_add(struct ip_fw_chain *chain, struct ip_fw **head, 820 struct ip_fw *rule) 821 { 822 823 IPFW_UH_WLOCK_ASSERT(chain); 824 825 /* Unlink rule from everywhere */ 826 unref_rule_objects(chain, rule); 827 828 rule->next = *head; 829 *head = rule; 830 } 831 832 /* 833 * Reclaim storage associated with a list of rules. This is 834 * typically the list created using remove_rule. 835 * A NULL pointer on input is handled correctly. 836 */ 837 void 838 ipfw_reap_rules(struct ip_fw *head) 839 { 840 struct ip_fw *rule; 841 842 while ((rule = head) != NULL) { 843 head = head->next; 844 ipfw_free_rule(rule); 845 } 846 } 847 848 /* 849 * Rules to keep are 850 * (default || reserved || !match_set || !match_number) 851 * where 852 * default ::= (rule->rulenum == IPFW_DEFAULT_RULE) 853 * // the default rule is always protected 854 * 855 * reserved ::= (cmd == 0 && n == 0 && rule->set == RESVD_SET) 856 * // RESVD_SET is protected only if cmd == 0 and n == 0 ("ipfw flush") 857 * 858 * match_set ::= (cmd == 0 || rule->set == set) 859 * // set number is ignored for cmd == 0 860 * 861 * match_number ::= (cmd == 1 || n == 0 || n == rule->rulenum) 862 * // number is ignored for cmd == 1 or n == 0 863 * 864 */ 865 int 866 ipfw_match_range(struct ip_fw *rule, ipfw_range_tlv *rt) 867 { 868 869 /* Don't match default rule for modification queries */ 870 if (rule->rulenum == IPFW_DEFAULT_RULE && 871 (rt->flags & IPFW_RCFLAG_DEFAULT) == 0) 872 return (0); 873 874 /* Don't match rules in reserved set for flush requests */ 875 if ((rt->flags & IPFW_RCFLAG_ALL) != 0 && rule->set == RESVD_SET) 876 return (0); 877 878 /* If we're filtering by set, don't match other sets */ 879 if ((rt->flags & IPFW_RCFLAG_SET) != 0 && rule->set != rt->set) 880 return (0); 881 882 if ((rt->flags & IPFW_RCFLAG_RANGE) != 0 && 883 (rule->rulenum < rt->start_rule || rule->rulenum > rt->end_rule)) 884 return (0); 885 886 return (1); 887 } 888 889 struct manage_sets_args { 890 uint16_t set; 891 uint8_t new_set; 892 }; 893 894 static int 895 swap_sets_cb(struct namedobj_instance *ni, struct named_object *no, 896 void *arg) 897 { 898 struct manage_sets_args *args; 899 900 args = (struct manage_sets_args *)arg; 901 if (no->set == (uint8_t)args->set) 902 no->set = args->new_set; 903 else if (no->set == args->new_set) 904 no->set = (uint8_t)args->set; 905 return (0); 906 } 907 908 static int 909 move_sets_cb(struct namedobj_instance *ni, struct named_object *no, 910 void *arg) 911 { 912 struct manage_sets_args *args; 913 914 args = (struct manage_sets_args *)arg; 915 if (no->set == (uint8_t)args->set) 916 no->set = args->new_set; 917 return (0); 918 } 919 920 static int 921 test_sets_cb(struct namedobj_instance *ni, struct named_object *no, 922 void *arg) 923 { 924 struct manage_sets_args *args; 925 926 args = (struct manage_sets_args *)arg; 927 if (no->set != (uint8_t)args->set) 928 return (0); 929 if (ipfw_objhash_lookup_name_type(ni, args->new_set, 930 no->etlv, no->name) != NULL) 931 return (EEXIST); 932 return (0); 933 } 934 935 /* 936 * Generic function to handler moving and swapping sets. 937 */ 938 int 939 ipfw_obj_manage_sets(struct namedobj_instance *ni, uint16_t type, 940 uint16_t set, uint8_t new_set, enum ipfw_sets_cmd cmd) 941 { 942 struct manage_sets_args args; 943 struct named_object *no; 944 945 args.set = set; 946 args.new_set = new_set; 947 switch (cmd) { 948 case SWAP_ALL: 949 return (ipfw_objhash_foreach_type(ni, swap_sets_cb, 950 &args, type)); 951 case TEST_ALL: 952 return (ipfw_objhash_foreach_type(ni, test_sets_cb, 953 &args, type)); 954 case MOVE_ALL: 955 return (ipfw_objhash_foreach_type(ni, move_sets_cb, 956 &args, type)); 957 case COUNT_ONE: 958 /* 959 * @set used to pass kidx. 960 * When @new_set is zero - reset object counter, 961 * otherwise increment it. 962 */ 963 no = ipfw_objhash_lookup_kidx(ni, set); 964 if (new_set != 0) 965 no->ocnt++; 966 else 967 no->ocnt = 0; 968 return (0); 969 case TEST_ONE: 970 /* @set used to pass kidx */ 971 no = ipfw_objhash_lookup_kidx(ni, set); 972 /* 973 * First check number of references: 974 * when it differs, this mean other rules are holding 975 * reference to given object, so it is not possible to 976 * change its set. Note that refcnt may account references 977 * to some going-to-be-added rules. Since we don't know 978 * their numbers (and even if they will be added) it is 979 * perfectly OK to return error here. 980 */ 981 if (no->ocnt != no->refcnt) 982 return (EBUSY); 983 if (ipfw_objhash_lookup_name_type(ni, new_set, type, 984 no->name) != NULL) 985 return (EEXIST); 986 return (0); 987 case MOVE_ONE: 988 /* @set used to pass kidx */ 989 no = ipfw_objhash_lookup_kidx(ni, set); 990 no->set = new_set; 991 return (0); 992 } 993 return (EINVAL); 994 } 995 996 /* 997 * Delete rules matching range @rt. 998 * Saves number of deleted rules in @ndel. 999 * 1000 * Returns 0 on success. 1001 */ 1002 static int 1003 delete_range(struct ip_fw_chain *chain, ipfw_range_tlv *rt, int *ndel) 1004 { 1005 struct ip_fw *reap, *rule, **map; 1006 int end, start; 1007 int i, n, ndyn, ofs; 1008 1009 reap = NULL; 1010 IPFW_UH_WLOCK(chain); /* arbitrate writers */ 1011 1012 /* 1013 * Stage 1: Determine range to inspect. 1014 * Range is half-inclusive, e.g [start, end). 1015 */ 1016 start = 0; 1017 end = chain->n_rules - 1; 1018 1019 if ((rt->flags & IPFW_RCFLAG_RANGE) != 0) { 1020 start = ipfw_find_rule(chain, rt->start_rule, 0); 1021 1022 if (rt->end_rule >= IPFW_DEFAULT_RULE) 1023 rt->end_rule = IPFW_DEFAULT_RULE - 1; 1024 end = ipfw_find_rule(chain, rt->end_rule, UINT32_MAX); 1025 } 1026 1027 if (rt->flags & IPFW_RCFLAG_DYNAMIC) { 1028 /* 1029 * Requested deleting only for dynamic states. 1030 */ 1031 *ndel = 0; 1032 ipfw_expire_dyn_states(chain, rt); 1033 IPFW_UH_WUNLOCK(chain); 1034 return (0); 1035 } 1036 1037 /* Allocate new map of the same size */ 1038 map = get_map(chain, 0, 1 /* locked */); 1039 if (map == NULL) { 1040 IPFW_UH_WUNLOCK(chain); 1041 return (ENOMEM); 1042 } 1043 1044 n = 0; 1045 ndyn = 0; 1046 ofs = start; 1047 /* 1. bcopy the initial part of the map */ 1048 if (start > 0) 1049 bcopy(chain->map, map, start * sizeof(struct ip_fw *)); 1050 /* 2. copy active rules between start and end */ 1051 for (i = start; i < end; i++) { 1052 rule = chain->map[i]; 1053 if (ipfw_match_range(rule, rt) == 0) { 1054 map[ofs++] = rule; 1055 continue; 1056 } 1057 1058 n++; 1059 if (ipfw_is_dyn_rule(rule) != 0) 1060 ndyn++; 1061 } 1062 /* 3. copy the final part of the map */ 1063 bcopy(chain->map + end, map + ofs, 1064 (chain->n_rules - end) * sizeof(struct ip_fw *)); 1065 /* 4. recalculate skipto cache */ 1066 update_skipto_cache(chain, map); 1067 /* 5. swap the maps (under UH_WLOCK + WHLOCK) */ 1068 map = swap_map(chain, map, chain->n_rules - n); 1069 /* 6. Remove all dynamic states originated by deleted rules */ 1070 if (ndyn > 0) 1071 ipfw_expire_dyn_states(chain, rt); 1072 /* 7. now remove the rules deleted from the old map */ 1073 for (i = start; i < end; i++) { 1074 rule = map[i]; 1075 if (ipfw_match_range(rule, rt) == 0) 1076 continue; 1077 chain->static_len -= RULEUSIZE0(rule); 1078 ipfw_reap_add(chain, &reap, rule); 1079 } 1080 IPFW_UH_WUNLOCK(chain); 1081 1082 ipfw_reap_rules(reap); 1083 if (map != NULL) 1084 free(map, M_IPFW); 1085 *ndel = n; 1086 return (0); 1087 } 1088 1089 static int 1090 move_objects(struct ip_fw_chain *ch, ipfw_range_tlv *rt) 1091 { 1092 struct opcode_obj_rewrite *rw; 1093 struct ip_fw *rule; 1094 ipfw_insn *cmd; 1095 int cmdlen, i, l, c; 1096 uint16_t kidx; 1097 1098 IPFW_UH_WLOCK_ASSERT(ch); 1099 1100 /* Stage 1: count number of references by given rules */ 1101 for (c = 0, i = 0; i < ch->n_rules - 1; i++) { 1102 rule = ch->map[i]; 1103 if (ipfw_match_range(rule, rt) == 0) 1104 continue; 1105 if (rule->set == rt->new_set) /* nothing to do */ 1106 continue; 1107 /* Search opcodes with named objects */ 1108 for (l = rule->cmd_len, cmdlen = 0, cmd = rule->cmd; 1109 l > 0; l -= cmdlen, cmd += cmdlen) { 1110 cmdlen = F_LEN(cmd); 1111 rw = find_op_rw(cmd, &kidx, NULL); 1112 if (rw == NULL || rw->manage_sets == NULL) 1113 continue; 1114 /* 1115 * When manage_sets() returns non-zero value to 1116 * COUNT_ONE command, consider this as an object 1117 * doesn't support sets (e.g. disabled with sysctl). 1118 * So, skip checks for this object. 1119 */ 1120 if (rw->manage_sets(ch, kidx, 1, COUNT_ONE) != 0) 1121 continue; 1122 c++; 1123 } 1124 } 1125 if (c == 0) /* No objects found */ 1126 return (0); 1127 /* Stage 2: verify "ownership" */ 1128 for (c = 0, i = 0; (i < ch->n_rules - 1) && c == 0; i++) { 1129 rule = ch->map[i]; 1130 if (ipfw_match_range(rule, rt) == 0) 1131 continue; 1132 if (rule->set == rt->new_set) /* nothing to do */ 1133 continue; 1134 /* Search opcodes with named objects */ 1135 for (l = rule->cmd_len, cmdlen = 0, cmd = rule->cmd; 1136 l > 0 && c == 0; l -= cmdlen, cmd += cmdlen) { 1137 cmdlen = F_LEN(cmd); 1138 rw = find_op_rw(cmd, &kidx, NULL); 1139 if (rw == NULL || rw->manage_sets == NULL) 1140 continue; 1141 /* Test for ownership and conflicting names */ 1142 c = rw->manage_sets(ch, kidx, 1143 (uint8_t)rt->new_set, TEST_ONE); 1144 } 1145 } 1146 /* Stage 3: change set and cleanup */ 1147 for (i = 0; i < ch->n_rules - 1; i++) { 1148 rule = ch->map[i]; 1149 if (ipfw_match_range(rule, rt) == 0) 1150 continue; 1151 if (rule->set == rt->new_set) /* nothing to do */ 1152 continue; 1153 /* Search opcodes with named objects */ 1154 for (l = rule->cmd_len, cmdlen = 0, cmd = rule->cmd; 1155 l > 0; l -= cmdlen, cmd += cmdlen) { 1156 cmdlen = F_LEN(cmd); 1157 rw = find_op_rw(cmd, &kidx, NULL); 1158 if (rw == NULL || rw->manage_sets == NULL) 1159 continue; 1160 /* cleanup object counter */ 1161 rw->manage_sets(ch, kidx, 1162 0 /* reset counter */, COUNT_ONE); 1163 if (c != 0) 1164 continue; 1165 /* change set */ 1166 rw->manage_sets(ch, kidx, 1167 (uint8_t)rt->new_set, MOVE_ONE); 1168 } 1169 } 1170 return (c); 1171 } 1172 1173 /* 1174 * Changes set of given rule rannge @rt 1175 * with each other. 1176 * 1177 * Returns 0 on success. 1178 */ 1179 static int 1180 move_range(struct ip_fw_chain *chain, ipfw_range_tlv *rt) 1181 { 1182 struct ip_fw *rule; 1183 int i; 1184 1185 IPFW_UH_WLOCK(chain); 1186 1187 /* 1188 * Move rules with matching paramenerts to a new set. 1189 * This one is much more complex. We have to ensure 1190 * that all referenced tables (if any) are referenced 1191 * by given rule subset only. Otherwise, we can't move 1192 * them to new set and have to return error. 1193 */ 1194 if ((i = move_objects(chain, rt)) != 0) { 1195 IPFW_UH_WUNLOCK(chain); 1196 return (i); 1197 } 1198 1199 /* XXX: We have to do swap holding WLOCK */ 1200 for (i = 0; i < chain->n_rules; i++) { 1201 rule = chain->map[i]; 1202 if (ipfw_match_range(rule, rt) == 0) 1203 continue; 1204 rule->set = rt->new_set; 1205 } 1206 1207 IPFW_UH_WUNLOCK(chain); 1208 1209 return (0); 1210 } 1211 1212 /* 1213 * Returns pointer to action instruction, skips all possible rule 1214 * modifiers like O_LOG, O_TAG, O_ALTQ. 1215 */ 1216 ipfw_insn * 1217 ipfw_get_action(struct ip_fw *rule) 1218 { 1219 ipfw_insn *cmd; 1220 int l, cmdlen; 1221 1222 cmd = ACTION_PTR(rule); 1223 l = rule->cmd_len - rule->act_ofs; 1224 while (l > 0) { 1225 switch (cmd->opcode) { 1226 case O_ALTQ: 1227 case O_LOG: 1228 case O_TAG: 1229 break; 1230 default: 1231 return (cmd); 1232 } 1233 cmdlen = F_LEN(cmd); 1234 l -= cmdlen; 1235 cmd += cmdlen; 1236 } 1237 panic("%s: rule (%p) has not action opcode", __func__, rule); 1238 return (NULL); 1239 } 1240 1241 /* 1242 * Clear counters for a specific rule. 1243 * Normally run under IPFW_UH_RLOCK, but these are idempotent ops 1244 * so we only care that rules do not disappear. 1245 */ 1246 static void 1247 clear_counters(struct ip_fw *rule, int log_only) 1248 { 1249 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 1250 1251 if (log_only == 0) 1252 IPFW_ZERO_RULE_COUNTER(rule); 1253 if (l->o.opcode == O_LOG) 1254 l->log_left = l->max_log; 1255 } 1256 1257 /* 1258 * Flushes rules counters and/or log values on matching range. 1259 * 1260 * Returns number of items cleared. 1261 */ 1262 static int 1263 clear_range(struct ip_fw_chain *chain, ipfw_range_tlv *rt, int log_only) 1264 { 1265 struct ip_fw *rule; 1266 int num; 1267 int i; 1268 1269 num = 0; 1270 rt->flags |= IPFW_RCFLAG_DEFAULT; 1271 1272 IPFW_UH_WLOCK(chain); /* arbitrate writers */ 1273 for (i = 0; i < chain->n_rules; i++) { 1274 rule = chain->map[i]; 1275 if (ipfw_match_range(rule, rt) == 0) 1276 continue; 1277 clear_counters(rule, log_only); 1278 num++; 1279 } 1280 IPFW_UH_WUNLOCK(chain); 1281 1282 return (num); 1283 } 1284 1285 static int 1286 check_range_tlv(ipfw_range_tlv *rt) 1287 { 1288 1289 if (rt->head.length != sizeof(*rt)) 1290 return (1); 1291 if (rt->start_rule > rt->end_rule) 1292 return (1); 1293 if (rt->set >= IPFW_MAX_SETS || rt->new_set >= IPFW_MAX_SETS) 1294 return (1); 1295 1296 if ((rt->flags & IPFW_RCFLAG_USER) != rt->flags) 1297 return (1); 1298 1299 return (0); 1300 } 1301 1302 /* 1303 * Delete rules matching specified parameters 1304 * Data layout (v0)(current): 1305 * Request: [ ipfw_obj_header ipfw_range_tlv ] 1306 * Reply: [ ipfw_obj_header ipfw_range_tlv ] 1307 * 1308 * Saves number of deleted rules in ipfw_range_tlv->new_set. 1309 * 1310 * Returns 0 on success. 1311 */ 1312 static int 1313 del_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 1314 struct sockopt_data *sd) 1315 { 1316 ipfw_range_header *rh; 1317 int error, ndel; 1318 1319 if (sd->valsize != sizeof(*rh)) 1320 return (EINVAL); 1321 1322 rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize); 1323 1324 if (check_range_tlv(&rh->range) != 0) 1325 return (EINVAL); 1326 1327 ndel = 0; 1328 if ((error = delete_range(chain, &rh->range, &ndel)) != 0) 1329 return (error); 1330 1331 /* Save number of rules deleted */ 1332 rh->range.new_set = ndel; 1333 return (0); 1334 } 1335 1336 /* 1337 * Move rules/sets matching specified parameters 1338 * Data layout (v0)(current): 1339 * Request: [ ipfw_obj_header ipfw_range_tlv ] 1340 * 1341 * Returns 0 on success. 1342 */ 1343 static int 1344 move_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 1345 struct sockopt_data *sd) 1346 { 1347 ipfw_range_header *rh; 1348 1349 if (sd->valsize != sizeof(*rh)) 1350 return (EINVAL); 1351 1352 rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize); 1353 1354 if (check_range_tlv(&rh->range) != 0) 1355 return (EINVAL); 1356 1357 return (move_range(chain, &rh->range)); 1358 } 1359 1360 /* 1361 * Clear rule accounting data matching specified parameters 1362 * Data layout (v0)(current): 1363 * Request: [ ipfw_obj_header ipfw_range_tlv ] 1364 * Reply: [ ipfw_obj_header ipfw_range_tlv ] 1365 * 1366 * Saves number of cleared rules in ipfw_range_tlv->new_set. 1367 * 1368 * Returns 0 on success. 1369 */ 1370 static int 1371 clear_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 1372 struct sockopt_data *sd) 1373 { 1374 ipfw_range_header *rh; 1375 int log_only, num; 1376 char *msg; 1377 1378 if (sd->valsize != sizeof(*rh)) 1379 return (EINVAL); 1380 1381 rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize); 1382 1383 if (check_range_tlv(&rh->range) != 0) 1384 return (EINVAL); 1385 1386 log_only = (op3->opcode == IP_FW_XRESETLOG); 1387 1388 num = clear_range(chain, &rh->range, log_only); 1389 1390 if (rh->range.flags & IPFW_RCFLAG_ALL) 1391 msg = log_only ? "All logging counts reset" : 1392 "Accounting cleared"; 1393 else 1394 msg = log_only ? "logging count reset" : "cleared"; 1395 1396 if (V_fw_verbose) { 1397 int lev = LOG_SECURITY | LOG_NOTICE; 1398 log(lev, "ipfw: %s.\n", msg); 1399 } 1400 1401 /* Save number of rules cleared */ 1402 rh->range.new_set = num; 1403 return (0); 1404 } 1405 1406 static void 1407 enable_sets(struct ip_fw_chain *chain, ipfw_range_tlv *rt) 1408 { 1409 uint32_t v_set; 1410 1411 IPFW_UH_WLOCK_ASSERT(chain); 1412 1413 /* Change enabled/disabled sets mask */ 1414 v_set = (V_set_disable | rt->set) & ~rt->new_set; 1415 v_set &= ~(1 << RESVD_SET); /* set RESVD_SET always enabled */ 1416 IPFW_WLOCK(chain); 1417 V_set_disable = v_set; 1418 IPFW_WUNLOCK(chain); 1419 } 1420 1421 static int 1422 swap_sets(struct ip_fw_chain *chain, ipfw_range_tlv *rt, int mv) 1423 { 1424 struct opcode_obj_rewrite *rw; 1425 struct ip_fw *rule; 1426 int i; 1427 1428 IPFW_UH_WLOCK_ASSERT(chain); 1429 1430 if (rt->set == rt->new_set) /* nothing to do */ 1431 return (0); 1432 1433 if (mv != 0) { 1434 /* 1435 * Berfore moving the rules we need to check that 1436 * there aren't any conflicting named objects. 1437 */ 1438 for (rw = ctl3_rewriters; 1439 rw < ctl3_rewriters + ctl3_rsize; rw++) { 1440 if (rw->manage_sets == NULL) 1441 continue; 1442 i = rw->manage_sets(chain, (uint8_t)rt->set, 1443 (uint8_t)rt->new_set, TEST_ALL); 1444 if (i != 0) 1445 return (EEXIST); 1446 } 1447 } 1448 /* Swap or move two sets */ 1449 for (i = 0; i < chain->n_rules - 1; i++) { 1450 rule = chain->map[i]; 1451 if (rule->set == (uint8_t)rt->set) 1452 rule->set = (uint8_t)rt->new_set; 1453 else if (rule->set == (uint8_t)rt->new_set && mv == 0) 1454 rule->set = (uint8_t)rt->set; 1455 } 1456 for (rw = ctl3_rewriters; rw < ctl3_rewriters + ctl3_rsize; rw++) { 1457 if (rw->manage_sets == NULL) 1458 continue; 1459 rw->manage_sets(chain, (uint8_t)rt->set, 1460 (uint8_t)rt->new_set, mv != 0 ? MOVE_ALL: SWAP_ALL); 1461 } 1462 return (0); 1463 } 1464 1465 /* 1466 * Swaps or moves set 1467 * Data layout (v0)(current): 1468 * Request: [ ipfw_obj_header ipfw_range_tlv ] 1469 * 1470 * Returns 0 on success. 1471 */ 1472 static int 1473 manage_sets(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 1474 struct sockopt_data *sd) 1475 { 1476 ipfw_range_header *rh; 1477 int ret; 1478 1479 if (sd->valsize != sizeof(*rh)) 1480 return (EINVAL); 1481 1482 rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize); 1483 1484 if (rh->range.head.length != sizeof(ipfw_range_tlv)) 1485 return (1); 1486 /* enable_sets() expects bitmasks. */ 1487 if (op3->opcode != IP_FW_SET_ENABLE && 1488 (rh->range.set >= IPFW_MAX_SETS || 1489 rh->range.new_set >= IPFW_MAX_SETS)) 1490 return (EINVAL); 1491 1492 ret = 0; 1493 IPFW_UH_WLOCK(chain); 1494 switch (op3->opcode) { 1495 case IP_FW_SET_SWAP: 1496 case IP_FW_SET_MOVE: 1497 ret = swap_sets(chain, &rh->range, 1498 op3->opcode == IP_FW_SET_MOVE); 1499 break; 1500 case IP_FW_SET_ENABLE: 1501 enable_sets(chain, &rh->range); 1502 break; 1503 } 1504 IPFW_UH_WUNLOCK(chain); 1505 1506 return (ret); 1507 } 1508 1509 /** 1510 * Remove all rules with given number, or do set manipulation. 1511 * Assumes chain != NULL && *chain != NULL. 1512 * 1513 * The argument is an uint32_t. The low 16 bit are the rule or set number; 1514 * the next 8 bits are the new set; the top 8 bits indicate the command: 1515 * 1516 * 0 delete rules numbered "rulenum" 1517 * 1 delete rules in set "rulenum" 1518 * 2 move rules "rulenum" to set "new_set" 1519 * 3 move rules from set "rulenum" to set "new_set" 1520 * 4 swap sets "rulenum" and "new_set" 1521 * 5 delete rules "rulenum" and set "new_set" 1522 */ 1523 static int 1524 del_entry(struct ip_fw_chain *chain, uint32_t arg) 1525 { 1526 uint32_t num; /* rule number or old_set */ 1527 uint8_t cmd, new_set; 1528 int do_del, ndel; 1529 int error = 0; 1530 ipfw_range_tlv rt; 1531 1532 num = arg & 0xffff; 1533 cmd = (arg >> 24) & 0xff; 1534 new_set = (arg >> 16) & 0xff; 1535 1536 if (cmd > 5 || new_set > RESVD_SET) 1537 return EINVAL; 1538 if (cmd == 0 || cmd == 2 || cmd == 5) { 1539 if (num >= IPFW_DEFAULT_RULE) 1540 return EINVAL; 1541 } else { 1542 if (num > RESVD_SET) /* old_set */ 1543 return EINVAL; 1544 } 1545 1546 /* Convert old requests into new representation */ 1547 memset(&rt, 0, sizeof(rt)); 1548 rt.start_rule = num; 1549 rt.end_rule = num; 1550 rt.set = num; 1551 rt.new_set = new_set; 1552 do_del = 0; 1553 1554 switch (cmd) { 1555 case 0: /* delete rules numbered "rulenum" */ 1556 if (num == 0) 1557 rt.flags |= IPFW_RCFLAG_ALL; 1558 else 1559 rt.flags |= IPFW_RCFLAG_RANGE; 1560 do_del = 1; 1561 break; 1562 case 1: /* delete rules in set "rulenum" */ 1563 rt.flags |= IPFW_RCFLAG_SET; 1564 do_del = 1; 1565 break; 1566 case 5: /* delete rules "rulenum" and set "new_set" */ 1567 rt.flags |= IPFW_RCFLAG_RANGE | IPFW_RCFLAG_SET; 1568 rt.set = new_set; 1569 rt.new_set = 0; 1570 do_del = 1; 1571 break; 1572 case 2: /* move rules "rulenum" to set "new_set" */ 1573 rt.flags |= IPFW_RCFLAG_RANGE; 1574 break; 1575 case 3: /* move rules from set "rulenum" to set "new_set" */ 1576 IPFW_UH_WLOCK(chain); 1577 error = swap_sets(chain, &rt, 1); 1578 IPFW_UH_WUNLOCK(chain); 1579 return (error); 1580 case 4: /* swap sets "rulenum" and "new_set" */ 1581 IPFW_UH_WLOCK(chain); 1582 error = swap_sets(chain, &rt, 0); 1583 IPFW_UH_WUNLOCK(chain); 1584 return (error); 1585 default: 1586 return (ENOTSUP); 1587 } 1588 1589 if (do_del != 0) { 1590 if ((error = delete_range(chain, &rt, &ndel)) != 0) 1591 return (error); 1592 1593 if (ndel == 0 && (cmd != 1 && num != 0)) 1594 return (EINVAL); 1595 1596 return (0); 1597 } 1598 1599 return (move_range(chain, &rt)); 1600 } 1601 1602 /** 1603 * Reset some or all counters on firewall rules. 1604 * The argument `arg' is an u_int32_t. The low 16 bit are the rule number, 1605 * the next 8 bits are the set number, the top 8 bits are the command: 1606 * 0 work with rules from all set's; 1607 * 1 work with rules only from specified set. 1608 * Specified rule number is zero if we want to clear all entries. 1609 * log_only is 1 if we only want to reset logs, zero otherwise. 1610 */ 1611 static int 1612 zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only) 1613 { 1614 struct ip_fw *rule; 1615 char *msg; 1616 int i; 1617 1618 uint16_t rulenum = arg & 0xffff; 1619 uint8_t set = (arg >> 16) & 0xff; 1620 uint8_t cmd = (arg >> 24) & 0xff; 1621 1622 if (cmd > 1) 1623 return (EINVAL); 1624 if (cmd == 1 && set > RESVD_SET) 1625 return (EINVAL); 1626 1627 IPFW_UH_RLOCK(chain); 1628 if (rulenum == 0) { 1629 V_norule_counter = 0; 1630 for (i = 0; i < chain->n_rules; i++) { 1631 rule = chain->map[i]; 1632 /* Skip rules not in our set. */ 1633 if (cmd == 1 && rule->set != set) 1634 continue; 1635 clear_counters(rule, log_only); 1636 } 1637 msg = log_only ? "All logging counts reset" : 1638 "Accounting cleared"; 1639 } else { 1640 int cleared = 0; 1641 for (i = 0; i < chain->n_rules; i++) { 1642 rule = chain->map[i]; 1643 if (rule->rulenum == rulenum) { 1644 if (cmd == 0 || rule->set == set) 1645 clear_counters(rule, log_only); 1646 cleared = 1; 1647 } 1648 if (rule->rulenum > rulenum) 1649 break; 1650 } 1651 if (!cleared) { /* we did not find any matching rules */ 1652 IPFW_UH_RUNLOCK(chain); 1653 return (EINVAL); 1654 } 1655 msg = log_only ? "logging count reset" : "cleared"; 1656 } 1657 IPFW_UH_RUNLOCK(chain); 1658 1659 if (V_fw_verbose) { 1660 int lev = LOG_SECURITY | LOG_NOTICE; 1661 1662 if (rulenum) 1663 log(lev, "ipfw: Entry %d %s.\n", rulenum, msg); 1664 else 1665 log(lev, "ipfw: %s.\n", msg); 1666 } 1667 return (0); 1668 } 1669 1670 /* 1671 * Check rule head in FreeBSD11 format 1672 * 1673 */ 1674 static int 1675 check_ipfw_rule1(struct ip_fw_rule *rule, int size, 1676 struct rule_check_info *ci) 1677 { 1678 int l; 1679 1680 if (size < sizeof(*rule)) { 1681 printf("ipfw: rule too short\n"); 1682 return (EINVAL); 1683 } 1684 1685 /* Check for valid cmd_len */ 1686 l = roundup2(RULESIZE(rule), sizeof(uint64_t)); 1687 if (l != size) { 1688 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 1689 return (EINVAL); 1690 } 1691 if (rule->act_ofs >= rule->cmd_len) { 1692 printf("ipfw: bogus action offset (%u > %u)\n", 1693 rule->act_ofs, rule->cmd_len - 1); 1694 return (EINVAL); 1695 } 1696 1697 if (rule->rulenum > IPFW_DEFAULT_RULE - 1) 1698 return (EINVAL); 1699 1700 return (check_ipfw_rule_body(rule->cmd, rule->cmd_len, ci)); 1701 } 1702 1703 /* 1704 * Check rule head in FreeBSD8 format 1705 * 1706 */ 1707 static int 1708 check_ipfw_rule0(struct ip_fw_rule0 *rule, int size, 1709 struct rule_check_info *ci) 1710 { 1711 int l; 1712 1713 if (size < sizeof(*rule)) { 1714 printf("ipfw: rule too short\n"); 1715 return (EINVAL); 1716 } 1717 1718 /* Check for valid cmd_len */ 1719 l = sizeof(*rule) + rule->cmd_len * 4 - 4; 1720 if (l != size) { 1721 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 1722 return (EINVAL); 1723 } 1724 if (rule->act_ofs >= rule->cmd_len) { 1725 printf("ipfw: bogus action offset (%u > %u)\n", 1726 rule->act_ofs, rule->cmd_len - 1); 1727 return (EINVAL); 1728 } 1729 1730 if (rule->rulenum > IPFW_DEFAULT_RULE - 1) 1731 return (EINVAL); 1732 1733 return (check_ipfw_rule_body(rule->cmd, rule->cmd_len, ci)); 1734 } 1735 1736 static int 1737 check_ipfw_rule_body(ipfw_insn *cmd, int cmd_len, struct rule_check_info *ci) 1738 { 1739 int cmdlen, l; 1740 int have_action; 1741 1742 have_action = 0; 1743 1744 /* 1745 * Now go for the individual checks. Very simple ones, basically only 1746 * instruction sizes. 1747 */ 1748 for (l = cmd_len; l > 0 ; l -= cmdlen, cmd += cmdlen) { 1749 cmdlen = F_LEN(cmd); 1750 if (cmdlen > l) { 1751 printf("ipfw: opcode %d size truncated\n", 1752 cmd->opcode); 1753 return EINVAL; 1754 } 1755 switch (cmd->opcode) { 1756 case O_PROBE_STATE: 1757 case O_KEEP_STATE: 1758 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 1759 goto bad_size; 1760 ci->object_opcodes++; 1761 break; 1762 case O_PROTO: 1763 case O_IP_SRC_ME: 1764 case O_IP_DST_ME: 1765 case O_LAYER2: 1766 case O_IN: 1767 case O_FRAG: 1768 case O_DIVERTED: 1769 case O_IPOPT: 1770 case O_IPTOS: 1771 case O_IPPRECEDENCE: 1772 case O_IPVER: 1773 case O_SOCKARG: 1774 case O_TCPFLAGS: 1775 case O_TCPOPTS: 1776 case O_ESTAB: 1777 case O_VERREVPATH: 1778 case O_VERSRCREACH: 1779 case O_ANTISPOOF: 1780 case O_IPSEC: 1781 #ifdef INET6 1782 case O_IP6_SRC_ME: 1783 case O_IP6_DST_ME: 1784 case O_EXT_HDR: 1785 case O_IP6: 1786 #endif 1787 case O_IP4: 1788 case O_TAG: 1789 case O_SKIP_ACTION: 1790 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 1791 goto bad_size; 1792 break; 1793 1794 case O_EXTERNAL_ACTION: 1795 if (cmd->arg1 == 0 || 1796 cmdlen != F_INSN_SIZE(ipfw_insn)) { 1797 printf("ipfw: invalid external " 1798 "action opcode\n"); 1799 return (EINVAL); 1800 } 1801 ci->object_opcodes++; 1802 /* 1803 * Do we have O_EXTERNAL_INSTANCE or O_EXTERNAL_DATA 1804 * opcode? 1805 */ 1806 if (l != cmdlen) { 1807 l -= cmdlen; 1808 cmd += cmdlen; 1809 cmdlen = F_LEN(cmd); 1810 if (cmd->opcode == O_EXTERNAL_DATA) 1811 goto check_action; 1812 if (cmd->opcode != O_EXTERNAL_INSTANCE) { 1813 printf("ipfw: invalid opcode " 1814 "next to external action %u\n", 1815 cmd->opcode); 1816 return (EINVAL); 1817 } 1818 if (cmd->arg1 == 0 || 1819 cmdlen != F_INSN_SIZE(ipfw_insn)) { 1820 printf("ipfw: invalid external " 1821 "action instance opcode\n"); 1822 return (EINVAL); 1823 } 1824 ci->object_opcodes++; 1825 } 1826 goto check_action; 1827 1828 case O_FIB: 1829 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 1830 goto bad_size; 1831 if (cmd->arg1 >= rt_numfibs) { 1832 printf("ipfw: invalid fib number %d\n", 1833 cmd->arg1); 1834 return EINVAL; 1835 } 1836 break; 1837 1838 case O_SETFIB: 1839 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 1840 goto bad_size; 1841 if ((cmd->arg1 != IP_FW_TARG) && 1842 ((cmd->arg1 & 0x7FFF) >= rt_numfibs)) { 1843 printf("ipfw: invalid fib number %d\n", 1844 cmd->arg1 & 0x7FFF); 1845 return EINVAL; 1846 } 1847 goto check_action; 1848 1849 case O_UID: 1850 case O_GID: 1851 case O_JAIL: 1852 case O_IP_SRC: 1853 case O_IP_DST: 1854 case O_TCPSEQ: 1855 case O_TCPACK: 1856 case O_PROB: 1857 case O_ICMPTYPE: 1858 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 1859 goto bad_size; 1860 break; 1861 1862 case O_LIMIT: 1863 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 1864 goto bad_size; 1865 ci->object_opcodes++; 1866 break; 1867 1868 case O_LOG: 1869 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 1870 goto bad_size; 1871 1872 ((ipfw_insn_log *)cmd)->log_left = 1873 ((ipfw_insn_log *)cmd)->max_log; 1874 1875 break; 1876 1877 case O_IP_SRC_MASK: 1878 case O_IP_DST_MASK: 1879 /* only odd command lengths */ 1880 if ((cmdlen & 1) == 0) 1881 goto bad_size; 1882 break; 1883 1884 case O_IP_SRC_SET: 1885 case O_IP_DST_SET: 1886 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 1887 printf("ipfw: invalid set size %d\n", 1888 cmd->arg1); 1889 return EINVAL; 1890 } 1891 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1892 (cmd->arg1+31)/32 ) 1893 goto bad_size; 1894 break; 1895 1896 case O_IP_SRC_LOOKUP: 1897 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) 1898 goto bad_size; 1899 case O_IP_DST_LOOKUP: 1900 if (cmd->arg1 >= V_fw_tables_max) { 1901 printf("ipfw: invalid table number %d\n", 1902 cmd->arg1); 1903 return (EINVAL); 1904 } 1905 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 1906 cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1 && 1907 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 1908 goto bad_size; 1909 ci->object_opcodes++; 1910 break; 1911 case O_IP_FLOW_LOOKUP: 1912 case O_MAC_DST_LOOKUP: 1913 case O_MAC_SRC_LOOKUP: 1914 if (cmd->arg1 >= V_fw_tables_max) { 1915 printf("ipfw: invalid table number %d\n", 1916 cmd->arg1); 1917 return (EINVAL); 1918 } 1919 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 1920 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 1921 goto bad_size; 1922 ci->object_opcodes++; 1923 break; 1924 case O_MACADDR2: 1925 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 1926 goto bad_size; 1927 break; 1928 1929 case O_NOP: 1930 case O_IPID: 1931 case O_IPTTL: 1932 case O_IPLEN: 1933 case O_TCPDATALEN: 1934 case O_TCPMSS: 1935 case O_TCPWIN: 1936 case O_TAGGED: 1937 if (cmdlen < 1 || cmdlen > 31) 1938 goto bad_size; 1939 break; 1940 1941 case O_DSCP: 1942 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1) 1943 goto bad_size; 1944 break; 1945 1946 case O_MAC_TYPE: 1947 case O_IP_SRCPORT: 1948 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 1949 if (cmdlen < 2 || cmdlen > 31) 1950 goto bad_size; 1951 break; 1952 1953 case O_RECV: 1954 case O_XMIT: 1955 case O_VIA: 1956 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 1957 goto bad_size; 1958 ci->object_opcodes++; 1959 break; 1960 1961 case O_ALTQ: 1962 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 1963 goto bad_size; 1964 break; 1965 1966 case O_PIPE: 1967 case O_QUEUE: 1968 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 1969 goto bad_size; 1970 goto check_action; 1971 1972 case O_FORWARD_IP: 1973 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 1974 goto bad_size; 1975 goto check_action; 1976 #ifdef INET6 1977 case O_FORWARD_IP6: 1978 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa6)) 1979 goto bad_size; 1980 goto check_action; 1981 #endif /* INET6 */ 1982 1983 case O_DIVERT: 1984 case O_TEE: 1985 if (ip_divert_ptr == NULL) 1986 return EINVAL; 1987 else 1988 goto check_size; 1989 case O_NETGRAPH: 1990 case O_NGTEE: 1991 if (ng_ipfw_input_p == NULL) 1992 return EINVAL; 1993 else 1994 goto check_size; 1995 case O_NAT: 1996 if (!IPFW_NAT_LOADED) 1997 return EINVAL; 1998 if (cmdlen != F_INSN_SIZE(ipfw_insn_nat)) 1999 goto bad_size; 2000 goto check_action; 2001 case O_CHECK_STATE: 2002 ci->object_opcodes++; 2003 /* FALLTHROUGH */ 2004 case O_FORWARD_MAC: /* XXX not implemented yet */ 2005 case O_COUNT: 2006 case O_ACCEPT: 2007 case O_DENY: 2008 case O_REJECT: 2009 case O_SETDSCP: 2010 #ifdef INET6 2011 case O_UNREACH6: 2012 #endif 2013 case O_SKIPTO: 2014 case O_REASS: 2015 case O_CALLRETURN: 2016 check_size: 2017 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 2018 goto bad_size; 2019 check_action: 2020 if (have_action) { 2021 printf("ipfw: opcode %d, multiple actions" 2022 " not allowed\n", 2023 cmd->opcode); 2024 return (EINVAL); 2025 } 2026 have_action = 1; 2027 if (l != cmdlen) { 2028 printf("ipfw: opcode %d, action must be" 2029 " last opcode\n", 2030 cmd->opcode); 2031 return (EINVAL); 2032 } 2033 break; 2034 #ifdef INET6 2035 case O_IP6_SRC: 2036 case O_IP6_DST: 2037 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 2038 F_INSN_SIZE(ipfw_insn)) 2039 goto bad_size; 2040 break; 2041 2042 case O_FLOW6ID: 2043 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 2044 ((ipfw_insn_u32 *)cmd)->o.arg1) 2045 goto bad_size; 2046 break; 2047 2048 case O_IP6_SRC_MASK: 2049 case O_IP6_DST_MASK: 2050 if ( !(cmdlen & 1) || cmdlen > 127) 2051 goto bad_size; 2052 break; 2053 case O_ICMP6TYPE: 2054 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 2055 goto bad_size; 2056 break; 2057 #endif 2058 2059 default: 2060 switch (cmd->opcode) { 2061 #ifndef INET6 2062 case O_IP6_SRC_ME: 2063 case O_IP6_DST_ME: 2064 case O_EXT_HDR: 2065 case O_IP6: 2066 case O_UNREACH6: 2067 case O_IP6_SRC: 2068 case O_IP6_DST: 2069 case O_FLOW6ID: 2070 case O_IP6_SRC_MASK: 2071 case O_IP6_DST_MASK: 2072 case O_ICMP6TYPE: 2073 printf("ipfw: no IPv6 support in kernel\n"); 2074 return (EPROTONOSUPPORT); 2075 #endif 2076 default: 2077 printf("ipfw: opcode %d, unknown opcode\n", 2078 cmd->opcode); 2079 return (EINVAL); 2080 } 2081 } 2082 } 2083 if (have_action == 0) { 2084 printf("ipfw: missing action\n"); 2085 return (EINVAL); 2086 } 2087 return 0; 2088 2089 bad_size: 2090 printf("ipfw: opcode %d size %d wrong\n", 2091 cmd->opcode, cmdlen); 2092 return (EINVAL); 2093 } 2094 2095 /* 2096 * Translation of requests for compatibility with FreeBSD 7.2/8. 2097 * a static variable tells us if we have an old client from userland, 2098 * and if necessary we translate requests and responses between the 2099 * two formats. 2100 */ 2101 static int is7 = 0; 2102 2103 struct ip_fw7 { 2104 struct ip_fw7 *next; /* linked list of rules */ 2105 struct ip_fw7 *next_rule; /* ptr to next [skipto] rule */ 2106 /* 'next_rule' is used to pass up 'set_disable' status */ 2107 2108 uint16_t act_ofs; /* offset of action in 32-bit units */ 2109 uint16_t cmd_len; /* # of 32-bit words in cmd */ 2110 uint16_t rulenum; /* rule number */ 2111 uint8_t set; /* rule set (0..31) */ 2112 // #define RESVD_SET 31 /* set for default and persistent rules */ 2113 uint8_t _pad; /* padding */ 2114 // uint32_t id; /* rule id, only in v.8 */ 2115 /* These fields are present in all rules. */ 2116 uint64_t pcnt; /* Packet counter */ 2117 uint64_t bcnt; /* Byte counter */ 2118 uint32_t timestamp; /* tv_sec of last match */ 2119 2120 ipfw_insn cmd[1]; /* storage for commands */ 2121 }; 2122 2123 static int convert_rule_to_7(struct ip_fw_rule0 *rule); 2124 static int convert_rule_to_8(struct ip_fw_rule0 *rule); 2125 2126 #ifndef RULESIZE7 2127 #define RULESIZE7(rule) (sizeof(struct ip_fw7) + \ 2128 ((struct ip_fw7 *)(rule))->cmd_len * 4 - 4) 2129 #endif 2130 2131 /* 2132 * Copy the static and dynamic rules to the supplied buffer 2133 * and return the amount of space actually used. 2134 * Must be run under IPFW_UH_RLOCK 2135 */ 2136 static size_t 2137 ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 2138 { 2139 char *bp = buf; 2140 char *ep = bp + space; 2141 struct ip_fw *rule; 2142 struct ip_fw_rule0 *dst; 2143 struct timeval boottime; 2144 int error, i, l, warnflag; 2145 time_t boot_seconds; 2146 2147 warnflag = 0; 2148 2149 getboottime(&boottime); 2150 boot_seconds = boottime.tv_sec; 2151 for (i = 0; i < chain->n_rules; i++) { 2152 rule = chain->map[i]; 2153 2154 if (is7) { 2155 /* Convert rule to FreeBSd 7.2 format */ 2156 l = RULESIZE7(rule); 2157 if (bp + l + sizeof(uint32_t) <= ep) { 2158 bcopy(rule, bp, l + sizeof(uint32_t)); 2159 error = set_legacy_obj_kidx(chain, 2160 (struct ip_fw_rule0 *)bp); 2161 if (error != 0) 2162 return (0); 2163 error = convert_rule_to_7((struct ip_fw_rule0 *) bp); 2164 if (error) 2165 return 0; /*XXX correct? */ 2166 /* 2167 * XXX HACK. Store the disable mask in the "next" 2168 * pointer in a wild attempt to keep the ABI the same. 2169 * Why do we do this on EVERY rule? 2170 */ 2171 bcopy(&V_set_disable, 2172 &(((struct ip_fw7 *)bp)->next_rule), 2173 sizeof(V_set_disable)); 2174 if (((struct ip_fw7 *)bp)->timestamp) 2175 ((struct ip_fw7 *)bp)->timestamp += boot_seconds; 2176 bp += l; 2177 } 2178 continue; /* go to next rule */ 2179 } 2180 2181 l = RULEUSIZE0(rule); 2182 if (bp + l > ep) { /* should not happen */ 2183 printf("overflow dumping static rules\n"); 2184 break; 2185 } 2186 dst = (struct ip_fw_rule0 *)bp; 2187 export_rule0(rule, dst, l); 2188 error = set_legacy_obj_kidx(chain, dst); 2189 2190 /* 2191 * XXX HACK. Store the disable mask in the "next" 2192 * pointer in a wild attempt to keep the ABI the same. 2193 * Why do we do this on EVERY rule? 2194 * 2195 * XXX: "ipfw set show" (ab)uses IP_FW_GET to read disabled mask 2196 * so we need to fail _after_ saving at least one mask. 2197 */ 2198 bcopy(&V_set_disable, &dst->next_rule, sizeof(V_set_disable)); 2199 if (dst->timestamp) 2200 dst->timestamp += boot_seconds; 2201 bp += l; 2202 2203 if (error != 0) { 2204 if (error == 2) { 2205 /* Non-fatal table rewrite error. */ 2206 warnflag = 1; 2207 continue; 2208 } 2209 printf("Stop on rule %d. Fail to convert table\n", 2210 rule->rulenum); 2211 break; 2212 } 2213 } 2214 if (warnflag != 0) 2215 printf("ipfw: process %s is using legacy interfaces," 2216 " consider rebuilding\n", ""); 2217 ipfw_get_dynamic(chain, &bp, ep); /* protected by the dynamic lock */ 2218 return (bp - (char *)buf); 2219 } 2220 2221 struct dump_args { 2222 uint32_t b; /* start rule */ 2223 uint32_t e; /* end rule */ 2224 uint32_t rcount; /* number of rules */ 2225 uint32_t rsize; /* rules size */ 2226 uint32_t tcount; /* number of tables */ 2227 int rcounters; /* counters */ 2228 uint32_t *bmask; /* index bitmask of used named objects */ 2229 }; 2230 2231 void 2232 ipfw_export_obj_ntlv(struct named_object *no, ipfw_obj_ntlv *ntlv) 2233 { 2234 2235 ntlv->head.type = no->etlv; 2236 ntlv->head.length = sizeof(*ntlv); 2237 ntlv->idx = no->kidx; 2238 strlcpy(ntlv->name, no->name, sizeof(ntlv->name)); 2239 } 2240 2241 /* 2242 * Export named object info in instance @ni, identified by @kidx 2243 * to ipfw_obj_ntlv. TLV is allocated from @sd space. 2244 * 2245 * Returns 0 on success. 2246 */ 2247 static int 2248 export_objhash_ntlv(struct namedobj_instance *ni, uint16_t kidx, 2249 struct sockopt_data *sd) 2250 { 2251 struct named_object *no; 2252 ipfw_obj_ntlv *ntlv; 2253 2254 no = ipfw_objhash_lookup_kidx(ni, kidx); 2255 KASSERT(no != NULL, ("invalid object kernel index passed")); 2256 2257 ntlv = (ipfw_obj_ntlv *)ipfw_get_sopt_space(sd, sizeof(*ntlv)); 2258 if (ntlv == NULL) 2259 return (ENOMEM); 2260 2261 ipfw_export_obj_ntlv(no, ntlv); 2262 return (0); 2263 } 2264 2265 static int 2266 export_named_objects(struct namedobj_instance *ni, struct dump_args *da, 2267 struct sockopt_data *sd) 2268 { 2269 int error, i; 2270 2271 for (i = 0; i < IPFW_TABLES_MAX && da->tcount > 0; i++) { 2272 if ((da->bmask[i / 32] & (1 << (i % 32))) == 0) 2273 continue; 2274 if ((error = export_objhash_ntlv(ni, i, sd)) != 0) 2275 return (error); 2276 da->tcount--; 2277 } 2278 return (0); 2279 } 2280 2281 static int 2282 dump_named_objects(struct ip_fw_chain *ch, struct dump_args *da, 2283 struct sockopt_data *sd) 2284 { 2285 ipfw_obj_ctlv *ctlv; 2286 int error; 2287 2288 MPASS(da->tcount > 0); 2289 /* Header first */ 2290 ctlv = (ipfw_obj_ctlv *)ipfw_get_sopt_space(sd, sizeof(*ctlv)); 2291 if (ctlv == NULL) 2292 return (ENOMEM); 2293 ctlv->head.type = IPFW_TLV_TBLNAME_LIST; 2294 ctlv->head.length = da->tcount * sizeof(ipfw_obj_ntlv) + 2295 sizeof(*ctlv); 2296 ctlv->count = da->tcount; 2297 ctlv->objsize = sizeof(ipfw_obj_ntlv); 2298 2299 /* Dump table names first (if any) */ 2300 error = export_named_objects(ipfw_get_table_objhash(ch), da, sd); 2301 if (error != 0) 2302 return (error); 2303 /* Then dump another named objects */ 2304 da->bmask += IPFW_TABLES_MAX / 32; 2305 return (export_named_objects(CHAIN_TO_SRV(ch), da, sd)); 2306 } 2307 2308 /* 2309 * Dumps static rules with table TLVs in buffer @sd. 2310 * 2311 * Returns 0 on success. 2312 */ 2313 static int 2314 dump_static_rules(struct ip_fw_chain *chain, struct dump_args *da, 2315 struct sockopt_data *sd) 2316 { 2317 ipfw_obj_ctlv *ctlv; 2318 struct ip_fw *krule; 2319 caddr_t dst; 2320 int i, l; 2321 2322 /* Dump rules */ 2323 ctlv = (ipfw_obj_ctlv *)ipfw_get_sopt_space(sd, sizeof(*ctlv)); 2324 if (ctlv == NULL) 2325 return (ENOMEM); 2326 ctlv->head.type = IPFW_TLV_RULE_LIST; 2327 ctlv->head.length = da->rsize + sizeof(*ctlv); 2328 ctlv->count = da->rcount; 2329 2330 for (i = da->b; i < da->e; i++) { 2331 krule = chain->map[i]; 2332 2333 l = RULEUSIZE1(krule) + sizeof(ipfw_obj_tlv); 2334 if (da->rcounters != 0) 2335 l += sizeof(struct ip_fw_bcounter); 2336 dst = (caddr_t)ipfw_get_sopt_space(sd, l); 2337 if (dst == NULL) 2338 return (ENOMEM); 2339 2340 export_rule1(krule, dst, l, da->rcounters); 2341 } 2342 2343 return (0); 2344 } 2345 2346 int 2347 ipfw_mark_object_kidx(uint32_t *bmask, uint16_t etlv, uint16_t kidx) 2348 { 2349 uint32_t bidx; 2350 2351 /* 2352 * Maintain separate bitmasks for table and non-table objects. 2353 */ 2354 bidx = (etlv == IPFW_TLV_TBL_NAME) ? 0: IPFW_TABLES_MAX / 32; 2355 bidx += kidx / 32; 2356 if ((bmask[bidx] & (1 << (kidx % 32))) != 0) 2357 return (0); 2358 2359 bmask[bidx] |= 1 << (kidx % 32); 2360 return (1); 2361 } 2362 2363 /* 2364 * Marks every object index used in @rule with bit in @bmask. 2365 * Used to generate bitmask of referenced tables/objects for given ruleset 2366 * or its part. 2367 */ 2368 static void 2369 mark_rule_objects(struct ip_fw_chain *ch, struct ip_fw *rule, 2370 struct dump_args *da) 2371 { 2372 struct opcode_obj_rewrite *rw; 2373 ipfw_insn *cmd; 2374 int cmdlen, l; 2375 uint16_t kidx; 2376 uint8_t subtype; 2377 2378 l = rule->cmd_len; 2379 cmd = rule->cmd; 2380 cmdlen = 0; 2381 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 2382 cmdlen = F_LEN(cmd); 2383 2384 rw = find_op_rw(cmd, &kidx, &subtype); 2385 if (rw == NULL) 2386 continue; 2387 2388 if (ipfw_mark_object_kidx(da->bmask, rw->etlv, kidx)) 2389 da->tcount++; 2390 } 2391 } 2392 2393 /* 2394 * Dumps requested objects data 2395 * Data layout (version 0)(current): 2396 * Request: [ ipfw_cfg_lheader ] + IPFW_CFG_GET_* flags 2397 * size = ipfw_cfg_lheader.size 2398 * Reply: [ ipfw_cfg_lheader 2399 * [ ipfw_obj_ctlv(IPFW_TLV_TBL_LIST) ipfw_obj_ntlv x N ] (optional) 2400 * [ ipfw_obj_ctlv(IPFW_TLV_RULE_LIST) 2401 * ipfw_obj_tlv(IPFW_TLV_RULE_ENT) [ ip_fw_bcounter (optional) ip_fw_rule ] 2402 * ] (optional) 2403 * [ ipfw_obj_ctlv(IPFW_TLV_STATE_LIST) ipfw_obj_dyntlv x N ] (optional) 2404 * ] 2405 * * NOTE IPFW_TLV_STATE_LIST has the single valid field: objsize. 2406 * The rest (size, count) are set to zero and needs to be ignored. 2407 * 2408 * Returns 0 on success. 2409 */ 2410 static int 2411 dump_config(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 2412 struct sockopt_data *sd) 2413 { 2414 struct dump_args da; 2415 ipfw_cfg_lheader *hdr; 2416 struct ip_fw *rule; 2417 size_t sz, rnum; 2418 uint32_t hdr_flags, *bmask; 2419 int error, i; 2420 2421 hdr = (ipfw_cfg_lheader *)ipfw_get_sopt_header(sd, sizeof(*hdr)); 2422 if (hdr == NULL) 2423 return (EINVAL); 2424 2425 error = 0; 2426 bmask = NULL; 2427 memset(&da, 0, sizeof(da)); 2428 /* 2429 * Allocate needed state. 2430 * Note we allocate 2xspace mask, for table & srv 2431 */ 2432 if (hdr->flags & (IPFW_CFG_GET_STATIC | IPFW_CFG_GET_STATES)) 2433 da.bmask = bmask = malloc( 2434 sizeof(uint32_t) * IPFW_TABLES_MAX * 2 / 32, M_TEMP, 2435 M_WAITOK | M_ZERO); 2436 IPFW_UH_RLOCK(chain); 2437 2438 /* 2439 * STAGE 1: Determine size/count for objects in range. 2440 * Prepare used tables bitmask. 2441 */ 2442 sz = sizeof(ipfw_cfg_lheader); 2443 da.e = chain->n_rules; 2444 2445 if (hdr->end_rule != 0) { 2446 /* Handle custom range */ 2447 if ((rnum = hdr->start_rule) > IPFW_DEFAULT_RULE) 2448 rnum = IPFW_DEFAULT_RULE; 2449 da.b = ipfw_find_rule(chain, rnum, 0); 2450 rnum = (hdr->end_rule < IPFW_DEFAULT_RULE) ? 2451 hdr->end_rule + 1: IPFW_DEFAULT_RULE; 2452 da.e = ipfw_find_rule(chain, rnum, UINT32_MAX) + 1; 2453 } 2454 2455 if (hdr->flags & IPFW_CFG_GET_STATIC) { 2456 for (i = da.b; i < da.e; i++) { 2457 rule = chain->map[i]; 2458 da.rsize += RULEUSIZE1(rule) + sizeof(ipfw_obj_tlv); 2459 da.rcount++; 2460 /* Update bitmask of used objects for given range */ 2461 mark_rule_objects(chain, rule, &da); 2462 } 2463 /* Add counters if requested */ 2464 if (hdr->flags & IPFW_CFG_GET_COUNTERS) { 2465 da.rsize += sizeof(struct ip_fw_bcounter) * da.rcount; 2466 da.rcounters = 1; 2467 } 2468 sz += da.rsize + sizeof(ipfw_obj_ctlv); 2469 } 2470 2471 if (hdr->flags & IPFW_CFG_GET_STATES) { 2472 sz += sizeof(ipfw_obj_ctlv) + 2473 ipfw_dyn_get_count(bmask, &i) * sizeof(ipfw_obj_dyntlv); 2474 da.tcount += i; 2475 } 2476 2477 if (da.tcount > 0) 2478 sz += da.tcount * sizeof(ipfw_obj_ntlv) + 2479 sizeof(ipfw_obj_ctlv); 2480 2481 /* 2482 * Fill header anyway. 2483 * Note we have to save header fields to stable storage 2484 * buffer inside @sd can be flushed after dumping rules 2485 */ 2486 hdr->size = sz; 2487 hdr->set_mask = ~V_set_disable; 2488 hdr_flags = hdr->flags; 2489 hdr = NULL; 2490 2491 if (sd->valsize < sz) { 2492 error = ENOMEM; 2493 goto cleanup; 2494 } 2495 2496 /* STAGE2: Store actual data */ 2497 if (da.tcount > 0) { 2498 error = dump_named_objects(chain, &da, sd); 2499 if (error != 0) 2500 goto cleanup; 2501 } 2502 2503 if (hdr_flags & IPFW_CFG_GET_STATIC) { 2504 error = dump_static_rules(chain, &da, sd); 2505 if (error != 0) 2506 goto cleanup; 2507 } 2508 2509 if (hdr_flags & IPFW_CFG_GET_STATES) 2510 error = ipfw_dump_states(chain, sd); 2511 2512 cleanup: 2513 IPFW_UH_RUNLOCK(chain); 2514 2515 if (bmask != NULL) 2516 free(bmask, M_TEMP); 2517 2518 return (error); 2519 } 2520 2521 int 2522 ipfw_check_object_name_generic(const char *name) 2523 { 2524 int nsize; 2525 2526 nsize = sizeof(((ipfw_obj_ntlv *)0)->name); 2527 if (strnlen(name, nsize) == nsize) 2528 return (EINVAL); 2529 if (name[0] == '\0') 2530 return (EINVAL); 2531 return (0); 2532 } 2533 2534 /* 2535 * Creates non-existent objects referenced by rule. 2536 * 2537 * Return 0 on success. 2538 */ 2539 int 2540 create_objects_compat(struct ip_fw_chain *ch, ipfw_insn *cmd, 2541 struct obj_idx *oib, struct obj_idx *pidx, struct tid_info *ti) 2542 { 2543 struct opcode_obj_rewrite *rw; 2544 struct obj_idx *p; 2545 uint16_t kidx; 2546 int error; 2547 2548 /* 2549 * Compatibility stuff: do actual creation for non-existing, 2550 * but referenced objects. 2551 */ 2552 for (p = oib; p < pidx; p++) { 2553 if (p->kidx != 0) 2554 continue; 2555 2556 ti->uidx = p->uidx; 2557 ti->type = p->type; 2558 ti->atype = 0; 2559 2560 rw = find_op_rw(cmd + p->off, NULL, NULL); 2561 KASSERT(rw != NULL, ("Unable to find handler for op %d", 2562 (cmd + p->off)->opcode)); 2563 2564 if (rw->create_object == NULL) 2565 error = EOPNOTSUPP; 2566 else 2567 error = rw->create_object(ch, ti, &kidx); 2568 if (error == 0) { 2569 p->kidx = kidx; 2570 continue; 2571 } 2572 2573 /* 2574 * Error happened. We have to rollback everything. 2575 * Drop all already acquired references. 2576 */ 2577 IPFW_UH_WLOCK(ch); 2578 unref_oib_objects(ch, cmd, oib, pidx); 2579 IPFW_UH_WUNLOCK(ch); 2580 2581 return (error); 2582 } 2583 2584 return (0); 2585 } 2586 2587 /* 2588 * Compatibility function for old ipfw(8) binaries. 2589 * Rewrites table/nat kernel indices with userland ones. 2590 * Convert tables matching '/^\d+$/' to their atoi() value. 2591 * Use number 65535 for other tables. 2592 * 2593 * Returns 0 on success. 2594 */ 2595 static int 2596 set_legacy_obj_kidx(struct ip_fw_chain *ch, struct ip_fw_rule0 *rule) 2597 { 2598 struct opcode_obj_rewrite *rw; 2599 struct named_object *no; 2600 ipfw_insn *cmd; 2601 char *end; 2602 long val; 2603 int cmdlen, error, l; 2604 uint16_t kidx, uidx; 2605 uint8_t subtype; 2606 2607 error = 0; 2608 2609 l = rule->cmd_len; 2610 cmd = rule->cmd; 2611 cmdlen = 0; 2612 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 2613 cmdlen = F_LEN(cmd); 2614 2615 /* Check if is index in given opcode */ 2616 rw = find_op_rw(cmd, &kidx, &subtype); 2617 if (rw == NULL) 2618 continue; 2619 2620 /* Try to find referenced kernel object */ 2621 no = rw->find_bykidx(ch, kidx); 2622 if (no == NULL) 2623 continue; 2624 2625 val = strtol(no->name, &end, 10); 2626 if (*end == '\0' && val < 65535) { 2627 uidx = val; 2628 } else { 2629 /* 2630 * We are called via legacy opcode. 2631 * Save error and show table as fake number 2632 * not to make ipfw(8) hang. 2633 */ 2634 uidx = 65535; 2635 error = 2; 2636 } 2637 2638 rw->update(cmd, uidx); 2639 } 2640 2641 return (error); 2642 } 2643 2644 /* 2645 * Unreferences all already-referenced objects in given @cmd rule, 2646 * using information in @oib. 2647 * 2648 * Used to rollback partially converted rule on error. 2649 */ 2650 static void 2651 unref_oib_objects(struct ip_fw_chain *ch, ipfw_insn *cmd, struct obj_idx *oib, 2652 struct obj_idx *end) 2653 { 2654 struct opcode_obj_rewrite *rw; 2655 struct named_object *no; 2656 struct obj_idx *p; 2657 2658 IPFW_UH_WLOCK_ASSERT(ch); 2659 2660 for (p = oib; p < end; p++) { 2661 if (p->kidx == 0) 2662 continue; 2663 2664 rw = find_op_rw(cmd + p->off, NULL, NULL); 2665 KASSERT(rw != NULL, ("Unable to find handler for op %d", 2666 (cmd + p->off)->opcode)); 2667 2668 /* Find & unref by existing idx */ 2669 no = rw->find_bykidx(ch, p->kidx); 2670 KASSERT(no != NULL, ("Ref'd object %d disappeared", p->kidx)); 2671 no->refcnt--; 2672 } 2673 } 2674 2675 /* 2676 * Remove references from every object used in @rule. 2677 * Used at rule removal code. 2678 */ 2679 static void 2680 unref_rule_objects(struct ip_fw_chain *ch, struct ip_fw *rule) 2681 { 2682 struct opcode_obj_rewrite *rw; 2683 struct named_object *no; 2684 ipfw_insn *cmd; 2685 int cmdlen, l; 2686 uint16_t kidx; 2687 uint8_t subtype; 2688 2689 IPFW_UH_WLOCK_ASSERT(ch); 2690 2691 l = rule->cmd_len; 2692 cmd = rule->cmd; 2693 cmdlen = 0; 2694 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 2695 cmdlen = F_LEN(cmd); 2696 2697 rw = find_op_rw(cmd, &kidx, &subtype); 2698 if (rw == NULL) 2699 continue; 2700 no = rw->find_bykidx(ch, kidx); 2701 2702 KASSERT(no != NULL, ("object id %d not found", kidx)); 2703 KASSERT(no->subtype == subtype, 2704 ("wrong type %d (%d) for object id %d", 2705 no->subtype, subtype, kidx)); 2706 KASSERT(no->refcnt > 0, ("refcount for object %d is %d", 2707 kidx, no->refcnt)); 2708 2709 if (no->refcnt == 1 && rw->destroy_object != NULL) 2710 rw->destroy_object(ch, no); 2711 else 2712 no->refcnt--; 2713 } 2714 } 2715 2716 /* 2717 * Find and reference object (if any) stored in instruction @cmd. 2718 * 2719 * Saves object info in @pidx, sets 2720 * - @unresolved to 1 if object should exists but not found 2721 * 2722 * Returns non-zero value in case of error. 2723 */ 2724 static int 2725 ref_opcode_object(struct ip_fw_chain *ch, ipfw_insn *cmd, struct tid_info *ti, 2726 struct obj_idx *pidx, int *unresolved) 2727 { 2728 struct named_object *no; 2729 struct opcode_obj_rewrite *rw; 2730 int error; 2731 2732 /* Check if this opcode is candidate for rewrite */ 2733 rw = find_op_rw(cmd, &ti->uidx, &ti->type); 2734 if (rw == NULL) 2735 return (0); 2736 2737 /* Need to rewrite. Save necessary fields */ 2738 pidx->uidx = ti->uidx; 2739 pidx->type = ti->type; 2740 2741 /* Try to find referenced kernel object */ 2742 error = rw->find_byname(ch, ti, &no); 2743 if (error != 0) 2744 return (error); 2745 if (no == NULL) { 2746 /* 2747 * Report about unresolved object for automaic 2748 * creation. 2749 */ 2750 *unresolved = 1; 2751 return (0); 2752 } 2753 2754 /* 2755 * Object is already exist. 2756 * Its subtype should match with expected value. 2757 */ 2758 if (ti->type != no->subtype) 2759 return (EINVAL); 2760 2761 /* Bump refcount and update kidx. */ 2762 no->refcnt++; 2763 rw->update(cmd, no->kidx); 2764 return (0); 2765 } 2766 2767 /* 2768 * Finds and bumps refcount for objects referenced by given @rule. 2769 * Auto-creates non-existing tables. 2770 * Fills in @oib array with userland/kernel indexes. 2771 * 2772 * Returns 0 on success. 2773 */ 2774 static int 2775 ref_rule_objects(struct ip_fw_chain *ch, struct ip_fw *rule, 2776 struct rule_check_info *ci, struct obj_idx *oib, struct tid_info *ti) 2777 { 2778 struct obj_idx *pidx; 2779 ipfw_insn *cmd; 2780 int cmdlen, error, l, unresolved; 2781 2782 pidx = oib; 2783 l = rule->cmd_len; 2784 cmd = rule->cmd; 2785 cmdlen = 0; 2786 error = 0; 2787 2788 IPFW_UH_WLOCK(ch); 2789 2790 /* Increase refcount on each existing referenced table. */ 2791 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 2792 cmdlen = F_LEN(cmd); 2793 unresolved = 0; 2794 2795 error = ref_opcode_object(ch, cmd, ti, pidx, &unresolved); 2796 if (error != 0) 2797 break; 2798 /* 2799 * Compatibility stuff for old clients: 2800 * prepare to automaitcally create non-existing objects. 2801 */ 2802 if (unresolved != 0) { 2803 pidx->off = rule->cmd_len - l; 2804 pidx++; 2805 } 2806 } 2807 2808 if (error != 0) { 2809 /* Unref everything we have already done */ 2810 unref_oib_objects(ch, rule->cmd, oib, pidx); 2811 IPFW_UH_WUNLOCK(ch); 2812 return (error); 2813 } 2814 IPFW_UH_WUNLOCK(ch); 2815 2816 /* Perform auto-creation for non-existing objects */ 2817 if (pidx != oib) 2818 error = create_objects_compat(ch, rule->cmd, oib, pidx, ti); 2819 2820 /* Calculate real number of dynamic objects */ 2821 ci->object_opcodes = (uint16_t)(pidx - oib); 2822 2823 return (error); 2824 } 2825 2826 /* 2827 * Checks is opcode is referencing table of appropriate type. 2828 * Adds reference count for found table if true. 2829 * Rewrites user-supplied opcode values with kernel ones. 2830 * 2831 * Returns 0 on success and appropriate error code otherwise. 2832 */ 2833 static int 2834 rewrite_rule_uidx(struct ip_fw_chain *chain, struct rule_check_info *ci) 2835 { 2836 int error; 2837 ipfw_insn *cmd; 2838 struct obj_idx *p, *pidx_first, *pidx_last; 2839 struct tid_info ti; 2840 2841 /* 2842 * Prepare an array for storing opcode indices. 2843 * Use stack allocation by default. 2844 */ 2845 if (ci->object_opcodes <= (sizeof(ci->obuf)/sizeof(ci->obuf[0]))) { 2846 /* Stack */ 2847 pidx_first = ci->obuf; 2848 } else 2849 pidx_first = malloc( 2850 ci->object_opcodes * sizeof(struct obj_idx), 2851 M_IPFW, M_WAITOK | M_ZERO); 2852 2853 error = 0; 2854 memset(&ti, 0, sizeof(ti)); 2855 2856 /* Use set rule is assigned to. */ 2857 ti.set = ci->krule->set; 2858 if (ci->ctlv != NULL) { 2859 ti.tlvs = (void *)(ci->ctlv + 1); 2860 ti.tlen = ci->ctlv->head.length - sizeof(ipfw_obj_ctlv); 2861 } 2862 2863 /* Reference all used tables and other objects */ 2864 error = ref_rule_objects(chain, ci->krule, ci, pidx_first, &ti); 2865 if (error != 0) 2866 goto free; 2867 /* 2868 * Note that ref_rule_objects() might have updated ci->object_opcodes 2869 * to reflect actual number of object opcodes. 2870 */ 2871 2872 /* Perform rewrite of remaining opcodes */ 2873 p = pidx_first; 2874 pidx_last = pidx_first + ci->object_opcodes; 2875 for (p = pidx_first; p < pidx_last; p++) { 2876 cmd = ci->krule->cmd + p->off; 2877 update_opcode_kidx(cmd, p->kidx); 2878 } 2879 2880 free: 2881 if (pidx_first != ci->obuf) 2882 free(pidx_first, M_IPFW); 2883 2884 return (error); 2885 } 2886 2887 /* 2888 * Adds one or more rules to ipfw @chain. 2889 * Data layout (version 0)(current): 2890 * Request: 2891 * [ 2892 * ip_fw3_opheader 2893 * [ ipfw_obj_ctlv(IPFW_TLV_TBL_LIST) ipfw_obj_ntlv x N ] (optional *1) 2894 * [ ipfw_obj_ctlv(IPFW_TLV_RULE_LIST) ip_fw x N ] (*2) (*3) 2895 * ] 2896 * Reply: 2897 * [ 2898 * ip_fw3_opheader 2899 * [ ipfw_obj_ctlv(IPFW_TLV_TBL_LIST) ipfw_obj_ntlv x N ] (optional) 2900 * [ ipfw_obj_ctlv(IPFW_TLV_RULE_LIST) ip_fw x N ] 2901 * ] 2902 * 2903 * Rules in reply are modified to store their actual ruleset number. 2904 * 2905 * (*1) TLVs inside IPFW_TLV_TBL_LIST needs to be sorted ascending 2906 * according to their idx field and there has to be no duplicates. 2907 * (*2) Numbered rules inside IPFW_TLV_RULE_LIST needs to be sorted ascending. 2908 * (*3) Each ip_fw structure needs to be aligned to u64 boundary. 2909 * 2910 * Returns 0 on success. 2911 */ 2912 static int 2913 add_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 2914 struct sockopt_data *sd) 2915 { 2916 ipfw_obj_ctlv *ctlv, *rtlv, *tstate; 2917 ipfw_obj_ntlv *ntlv; 2918 int clen, error, idx; 2919 uint32_t count, read; 2920 struct ip_fw_rule *r; 2921 struct rule_check_info rci, *ci, *cbuf; 2922 int i, rsize; 2923 2924 op3 = (ip_fw3_opheader *)ipfw_get_sopt_space(sd, sd->valsize); 2925 ctlv = (ipfw_obj_ctlv *)(op3 + 1); 2926 2927 read = sizeof(ip_fw3_opheader); 2928 rtlv = NULL; 2929 tstate = NULL; 2930 cbuf = NULL; 2931 memset(&rci, 0, sizeof(struct rule_check_info)); 2932 2933 if (read + sizeof(*ctlv) > sd->valsize) 2934 return (EINVAL); 2935 2936 if (ctlv->head.type == IPFW_TLV_TBLNAME_LIST) { 2937 clen = ctlv->head.length; 2938 /* Check size and alignment */ 2939 if (clen > sd->valsize || clen < sizeof(*ctlv)) 2940 return (EINVAL); 2941 if ((clen % sizeof(uint64_t)) != 0) 2942 return (EINVAL); 2943 2944 /* 2945 * Some table names or other named objects. 2946 * Check for validness. 2947 */ 2948 count = (ctlv->head.length - sizeof(*ctlv)) / sizeof(*ntlv); 2949 if (ctlv->count != count || ctlv->objsize != sizeof(*ntlv)) 2950 return (EINVAL); 2951 2952 /* 2953 * Check each TLV. 2954 * Ensure TLVs are sorted ascending and 2955 * there are no duplicates. 2956 */ 2957 idx = -1; 2958 ntlv = (ipfw_obj_ntlv *)(ctlv + 1); 2959 while (count > 0) { 2960 if (ntlv->head.length != sizeof(ipfw_obj_ntlv)) 2961 return (EINVAL); 2962 2963 error = ipfw_check_object_name_generic(ntlv->name); 2964 if (error != 0) 2965 return (error); 2966 2967 if (ntlv->idx <= idx) 2968 return (EINVAL); 2969 2970 idx = ntlv->idx; 2971 count--; 2972 ntlv++; 2973 } 2974 2975 tstate = ctlv; 2976 read += ctlv->head.length; 2977 ctlv = (ipfw_obj_ctlv *)((caddr_t)ctlv + ctlv->head.length); 2978 } 2979 2980 if (read + sizeof(*ctlv) > sd->valsize) 2981 return (EINVAL); 2982 2983 if (ctlv->head.type == IPFW_TLV_RULE_LIST) { 2984 clen = ctlv->head.length; 2985 if (clen + read > sd->valsize || clen < sizeof(*ctlv)) 2986 return (EINVAL); 2987 if ((clen % sizeof(uint64_t)) != 0) 2988 return (EINVAL); 2989 2990 /* 2991 * TODO: Permit adding multiple rules at once 2992 */ 2993 if (ctlv->count != 1) 2994 return (ENOTSUP); 2995 2996 clen -= sizeof(*ctlv); 2997 2998 if (ctlv->count > clen / sizeof(struct ip_fw_rule)) 2999 return (EINVAL); 3000 3001 /* Allocate state for each rule or use stack */ 3002 if (ctlv->count == 1) { 3003 memset(&rci, 0, sizeof(struct rule_check_info)); 3004 cbuf = &rci; 3005 } else 3006 cbuf = malloc(ctlv->count * sizeof(*ci), M_TEMP, 3007 M_WAITOK | M_ZERO); 3008 ci = cbuf; 3009 3010 /* 3011 * Check each rule for validness. 3012 * Ensure numbered rules are sorted ascending 3013 * and properly aligned 3014 */ 3015 idx = 0; 3016 r = (struct ip_fw_rule *)(ctlv + 1); 3017 count = 0; 3018 error = 0; 3019 while (clen > 0) { 3020 rsize = roundup2(RULESIZE(r), sizeof(uint64_t)); 3021 if (rsize > clen || ctlv->count <= count) { 3022 error = EINVAL; 3023 break; 3024 } 3025 3026 ci->ctlv = tstate; 3027 error = check_ipfw_rule1(r, rsize, ci); 3028 if (error != 0) 3029 break; 3030 3031 /* Check sorting */ 3032 if (r->rulenum != 0 && r->rulenum < idx) { 3033 printf("rulenum %d idx %d\n", r->rulenum, idx); 3034 error = EINVAL; 3035 break; 3036 } 3037 idx = r->rulenum; 3038 3039 ci->urule = (caddr_t)r; 3040 3041 rsize = roundup2(rsize, sizeof(uint64_t)); 3042 clen -= rsize; 3043 r = (struct ip_fw_rule *)((caddr_t)r + rsize); 3044 count++; 3045 ci++; 3046 } 3047 3048 if (ctlv->count != count || error != 0) { 3049 if (cbuf != &rci) 3050 free(cbuf, M_TEMP); 3051 return (EINVAL); 3052 } 3053 3054 rtlv = ctlv; 3055 read += ctlv->head.length; 3056 ctlv = (ipfw_obj_ctlv *)((caddr_t)ctlv + ctlv->head.length); 3057 } 3058 3059 if (read != sd->valsize || rtlv == NULL || rtlv->count == 0) { 3060 if (cbuf != NULL && cbuf != &rci) 3061 free(cbuf, M_TEMP); 3062 return (EINVAL); 3063 } 3064 3065 /* 3066 * Passed rules seems to be valid. 3067 * Allocate storage and try to add them to chain. 3068 */ 3069 for (i = 0, ci = cbuf; i < rtlv->count; i++, ci++) { 3070 clen = RULEKSIZE1((struct ip_fw_rule *)ci->urule); 3071 ci->krule = ipfw_alloc_rule(chain, clen); 3072 import_rule1(ci); 3073 } 3074 3075 if ((error = commit_rules(chain, cbuf, rtlv->count)) != 0) { 3076 /* Free allocate krules */ 3077 for (i = 0, ci = cbuf; i < rtlv->count; i++, ci++) 3078 ipfw_free_rule(ci->krule); 3079 } 3080 3081 if (cbuf != NULL && cbuf != &rci) 3082 free(cbuf, M_TEMP); 3083 3084 return (error); 3085 } 3086 3087 /* 3088 * Lists all sopts currently registered. 3089 * Data layout (v0)(current): 3090 * Request: [ ipfw_obj_lheader ], size = ipfw_obj_lheader.size 3091 * Reply: [ ipfw_obj_lheader ipfw_sopt_info x N ] 3092 * 3093 * Returns 0 on success 3094 */ 3095 static int 3096 dump_soptcodes(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 3097 struct sockopt_data *sd) 3098 { 3099 struct _ipfw_obj_lheader *olh; 3100 ipfw_sopt_info *i; 3101 struct ipfw_sopt_handler *sh; 3102 uint32_t count, n, size; 3103 3104 olh = (struct _ipfw_obj_lheader *)ipfw_get_sopt_header(sd,sizeof(*olh)); 3105 if (olh == NULL) 3106 return (EINVAL); 3107 if (sd->valsize < olh->size) 3108 return (EINVAL); 3109 3110 CTL3_LOCK(); 3111 count = ctl3_hsize; 3112 size = count * sizeof(ipfw_sopt_info) + sizeof(ipfw_obj_lheader); 3113 3114 /* Fill in header regadless of buffer size */ 3115 olh->count = count; 3116 olh->objsize = sizeof(ipfw_sopt_info); 3117 3118 if (size > olh->size) { 3119 olh->size = size; 3120 CTL3_UNLOCK(); 3121 return (ENOMEM); 3122 } 3123 olh->size = size; 3124 3125 for (n = 1; n <= count; n++) { 3126 i = (ipfw_sopt_info *)ipfw_get_sopt_space(sd, sizeof(*i)); 3127 KASSERT(i != NULL, ("previously checked buffer is not enough")); 3128 sh = &ctl3_handlers[n]; 3129 i->opcode = sh->opcode; 3130 i->version = sh->version; 3131 i->refcnt = sh->refcnt; 3132 } 3133 CTL3_UNLOCK(); 3134 3135 return (0); 3136 } 3137 3138 /* 3139 * Compares two opcodes. 3140 * Used both in qsort() and bsearch(). 3141 * 3142 * Returns 0 if match is found. 3143 */ 3144 static int 3145 compare_opcodes(const void *_a, const void *_b) 3146 { 3147 const struct opcode_obj_rewrite *a, *b; 3148 3149 a = (const struct opcode_obj_rewrite *)_a; 3150 b = (const struct opcode_obj_rewrite *)_b; 3151 3152 if (a->opcode < b->opcode) 3153 return (-1); 3154 else if (a->opcode > b->opcode) 3155 return (1); 3156 3157 return (0); 3158 } 3159 3160 /* 3161 * XXX: Rewrite bsearch() 3162 */ 3163 static int 3164 find_op_rw_range(uint16_t op, struct opcode_obj_rewrite **plo, 3165 struct opcode_obj_rewrite **phi) 3166 { 3167 struct opcode_obj_rewrite *ctl3_max, *lo, *hi, h, *rw; 3168 3169 memset(&h, 0, sizeof(h)); 3170 h.opcode = op; 3171 3172 rw = (struct opcode_obj_rewrite *)bsearch(&h, ctl3_rewriters, 3173 ctl3_rsize, sizeof(h), compare_opcodes); 3174 if (rw == NULL) 3175 return (1); 3176 3177 /* Find the first element matching the same opcode */ 3178 lo = rw; 3179 for ( ; lo > ctl3_rewriters && (lo - 1)->opcode == op; lo--) 3180 ; 3181 3182 /* Find the last element matching the same opcode */ 3183 hi = rw; 3184 ctl3_max = ctl3_rewriters + ctl3_rsize; 3185 for ( ; (hi + 1) < ctl3_max && (hi + 1)->opcode == op; hi++) 3186 ; 3187 3188 *plo = lo; 3189 *phi = hi; 3190 3191 return (0); 3192 } 3193 3194 /* 3195 * Finds opcode object rewriter based on @code. 3196 * 3197 * Returns pointer to handler or NULL. 3198 */ 3199 static struct opcode_obj_rewrite * 3200 find_op_rw(ipfw_insn *cmd, uint16_t *puidx, uint8_t *ptype) 3201 { 3202 struct opcode_obj_rewrite *rw, *lo, *hi; 3203 uint16_t uidx; 3204 uint8_t subtype; 3205 3206 if (find_op_rw_range(cmd->opcode, &lo, &hi) != 0) 3207 return (NULL); 3208 3209 for (rw = lo; rw <= hi; rw++) { 3210 if (rw->classifier(cmd, &uidx, &subtype) == 0) { 3211 if (puidx != NULL) 3212 *puidx = uidx; 3213 if (ptype != NULL) 3214 *ptype = subtype; 3215 return (rw); 3216 } 3217 } 3218 3219 return (NULL); 3220 } 3221 int 3222 classify_opcode_kidx(ipfw_insn *cmd, uint16_t *puidx) 3223 { 3224 3225 if (find_op_rw(cmd, puidx, NULL) == NULL) 3226 return (1); 3227 return (0); 3228 } 3229 3230 void 3231 update_opcode_kidx(ipfw_insn *cmd, uint16_t idx) 3232 { 3233 struct opcode_obj_rewrite *rw; 3234 3235 rw = find_op_rw(cmd, NULL, NULL); 3236 KASSERT(rw != NULL, ("No handler to update opcode %d", cmd->opcode)); 3237 rw->update(cmd, idx); 3238 } 3239 3240 void 3241 ipfw_init_obj_rewriter(void) 3242 { 3243 3244 ctl3_rewriters = NULL; 3245 ctl3_rsize = 0; 3246 } 3247 3248 void 3249 ipfw_destroy_obj_rewriter(void) 3250 { 3251 3252 if (ctl3_rewriters != NULL) 3253 free(ctl3_rewriters, M_IPFW); 3254 ctl3_rewriters = NULL; 3255 ctl3_rsize = 0; 3256 } 3257 3258 /* 3259 * Adds one or more opcode object rewrite handlers to the global array. 3260 * Function may sleep. 3261 */ 3262 void 3263 ipfw_add_obj_rewriter(struct opcode_obj_rewrite *rw, size_t count) 3264 { 3265 size_t sz; 3266 struct opcode_obj_rewrite *tmp; 3267 3268 CTL3_LOCK(); 3269 3270 for (;;) { 3271 sz = ctl3_rsize + count; 3272 CTL3_UNLOCK(); 3273 tmp = malloc(sizeof(*rw) * sz, M_IPFW, M_WAITOK | M_ZERO); 3274 CTL3_LOCK(); 3275 if (ctl3_rsize + count <= sz) 3276 break; 3277 3278 /* Retry */ 3279 free(tmp, M_IPFW); 3280 } 3281 3282 /* Merge old & new arrays */ 3283 sz = ctl3_rsize + count; 3284 memcpy(tmp, ctl3_rewriters, ctl3_rsize * sizeof(*rw)); 3285 memcpy(&tmp[ctl3_rsize], rw, count * sizeof(*rw)); 3286 qsort(tmp, sz, sizeof(*rw), compare_opcodes); 3287 /* Switch new and free old */ 3288 if (ctl3_rewriters != NULL) 3289 free(ctl3_rewriters, M_IPFW); 3290 ctl3_rewriters = tmp; 3291 ctl3_rsize = sz; 3292 3293 CTL3_UNLOCK(); 3294 } 3295 3296 /* 3297 * Removes one or more object rewrite handlers from the global array. 3298 */ 3299 int 3300 ipfw_del_obj_rewriter(struct opcode_obj_rewrite *rw, size_t count) 3301 { 3302 size_t sz; 3303 struct opcode_obj_rewrite *ctl3_max, *ktmp, *lo, *hi; 3304 int i; 3305 3306 CTL3_LOCK(); 3307 3308 for (i = 0; i < count; i++) { 3309 if (find_op_rw_range(rw[i].opcode, &lo, &hi) != 0) 3310 continue; 3311 3312 for (ktmp = lo; ktmp <= hi; ktmp++) { 3313 if (ktmp->classifier != rw[i].classifier) 3314 continue; 3315 3316 ctl3_max = ctl3_rewriters + ctl3_rsize; 3317 sz = (ctl3_max - (ktmp + 1)) * sizeof(*ktmp); 3318 memmove(ktmp, ktmp + 1, sz); 3319 ctl3_rsize--; 3320 break; 3321 } 3322 } 3323 3324 if (ctl3_rsize == 0) { 3325 if (ctl3_rewriters != NULL) 3326 free(ctl3_rewriters, M_IPFW); 3327 ctl3_rewriters = NULL; 3328 } 3329 3330 CTL3_UNLOCK(); 3331 3332 return (0); 3333 } 3334 3335 static int 3336 export_objhash_ntlv_internal(struct namedobj_instance *ni, 3337 struct named_object *no, void *arg) 3338 { 3339 struct sockopt_data *sd; 3340 ipfw_obj_ntlv *ntlv; 3341 3342 sd = (struct sockopt_data *)arg; 3343 ntlv = (ipfw_obj_ntlv *)ipfw_get_sopt_space(sd, sizeof(*ntlv)); 3344 if (ntlv == NULL) 3345 return (ENOMEM); 3346 ipfw_export_obj_ntlv(no, ntlv); 3347 return (0); 3348 } 3349 3350 /* 3351 * Lists all service objects. 3352 * Data layout (v0)(current): 3353 * Request: [ ipfw_obj_lheader ] size = ipfw_obj_lheader.size 3354 * Reply: [ ipfw_obj_lheader [ ipfw_obj_ntlv x N ] (optional) ] 3355 * Returns 0 on success 3356 */ 3357 static int 3358 dump_srvobjects(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 3359 struct sockopt_data *sd) 3360 { 3361 ipfw_obj_lheader *hdr; 3362 int count; 3363 3364 hdr = (ipfw_obj_lheader *)ipfw_get_sopt_header(sd, sizeof(*hdr)); 3365 if (hdr == NULL) 3366 return (EINVAL); 3367 3368 IPFW_UH_RLOCK(chain); 3369 count = ipfw_objhash_count(CHAIN_TO_SRV(chain)); 3370 hdr->size = sizeof(ipfw_obj_lheader) + count * sizeof(ipfw_obj_ntlv); 3371 if (sd->valsize < hdr->size) { 3372 IPFW_UH_RUNLOCK(chain); 3373 return (ENOMEM); 3374 } 3375 hdr->count = count; 3376 hdr->objsize = sizeof(ipfw_obj_ntlv); 3377 if (count > 0) 3378 ipfw_objhash_foreach(CHAIN_TO_SRV(chain), 3379 export_objhash_ntlv_internal, sd); 3380 IPFW_UH_RUNLOCK(chain); 3381 return (0); 3382 } 3383 3384 /* 3385 * Compares two sopt handlers (code, version and handler ptr). 3386 * Used both as qsort() and bsearch(). 3387 * Does not compare handler for latter case. 3388 * 3389 * Returns 0 if match is found. 3390 */ 3391 static int 3392 compare_sh(const void *_a, const void *_b) 3393 { 3394 const struct ipfw_sopt_handler *a, *b; 3395 3396 a = (const struct ipfw_sopt_handler *)_a; 3397 b = (const struct ipfw_sopt_handler *)_b; 3398 3399 if (a->opcode < b->opcode) 3400 return (-1); 3401 else if (a->opcode > b->opcode) 3402 return (1); 3403 3404 if (a->version < b->version) 3405 return (-1); 3406 else if (a->version > b->version) 3407 return (1); 3408 3409 /* bsearch helper */ 3410 if (a->handler == NULL) 3411 return (0); 3412 3413 if ((uintptr_t)a->handler < (uintptr_t)b->handler) 3414 return (-1); 3415 else if ((uintptr_t)a->handler > (uintptr_t)b->handler) 3416 return (1); 3417 3418 return (0); 3419 } 3420 3421 /* 3422 * Finds sopt handler based on @code and @version. 3423 * 3424 * Returns pointer to handler or NULL. 3425 */ 3426 static struct ipfw_sopt_handler * 3427 find_sh(uint16_t code, uint8_t version, sopt_handler_f *handler) 3428 { 3429 struct ipfw_sopt_handler *sh, h; 3430 3431 memset(&h, 0, sizeof(h)); 3432 h.opcode = code; 3433 h.version = version; 3434 h.handler = handler; 3435 3436 sh = (struct ipfw_sopt_handler *)bsearch(&h, ctl3_handlers, 3437 ctl3_hsize, sizeof(h), compare_sh); 3438 3439 return (sh); 3440 } 3441 3442 static int 3443 find_ref_sh(uint16_t opcode, uint8_t version, struct ipfw_sopt_handler *psh) 3444 { 3445 struct ipfw_sopt_handler *sh; 3446 3447 CTL3_LOCK(); 3448 if ((sh = find_sh(opcode, version, NULL)) == NULL) { 3449 CTL3_UNLOCK(); 3450 printf("ipfw: ipfw_ctl3 invalid option %d""v""%d\n", 3451 opcode, version); 3452 return (EINVAL); 3453 } 3454 sh->refcnt++; 3455 ctl3_refct++; 3456 /* Copy handler data to requested buffer */ 3457 *psh = *sh; 3458 CTL3_UNLOCK(); 3459 3460 return (0); 3461 } 3462 3463 static void 3464 find_unref_sh(struct ipfw_sopt_handler *psh) 3465 { 3466 struct ipfw_sopt_handler *sh; 3467 3468 CTL3_LOCK(); 3469 sh = find_sh(psh->opcode, psh->version, NULL); 3470 KASSERT(sh != NULL, ("ctl3 handler disappeared")); 3471 sh->refcnt--; 3472 ctl3_refct--; 3473 CTL3_UNLOCK(); 3474 } 3475 3476 void 3477 ipfw_init_sopt_handler(void) 3478 { 3479 3480 CTL3_LOCK_INIT(); 3481 IPFW_ADD_SOPT_HANDLER(1, scodes); 3482 } 3483 3484 void 3485 ipfw_destroy_sopt_handler(void) 3486 { 3487 3488 IPFW_DEL_SOPT_HANDLER(1, scodes); 3489 CTL3_LOCK_DESTROY(); 3490 } 3491 3492 /* 3493 * Adds one or more sockopt handlers to the global array. 3494 * Function may sleep. 3495 */ 3496 void 3497 ipfw_add_sopt_handler(struct ipfw_sopt_handler *sh, size_t count) 3498 { 3499 size_t sz; 3500 struct ipfw_sopt_handler *tmp; 3501 3502 CTL3_LOCK(); 3503 3504 for (;;) { 3505 sz = ctl3_hsize + count; 3506 CTL3_UNLOCK(); 3507 tmp = malloc(sizeof(*sh) * sz, M_IPFW, M_WAITOK | M_ZERO); 3508 CTL3_LOCK(); 3509 if (ctl3_hsize + count <= sz) 3510 break; 3511 3512 /* Retry */ 3513 free(tmp, M_IPFW); 3514 } 3515 3516 /* Merge old & new arrays */ 3517 sz = ctl3_hsize + count; 3518 memcpy(tmp, ctl3_handlers, ctl3_hsize * sizeof(*sh)); 3519 memcpy(&tmp[ctl3_hsize], sh, count * sizeof(*sh)); 3520 qsort(tmp, sz, sizeof(*sh), compare_sh); 3521 /* Switch new and free old */ 3522 if (ctl3_handlers != NULL) 3523 free(ctl3_handlers, M_IPFW); 3524 ctl3_handlers = tmp; 3525 ctl3_hsize = sz; 3526 ctl3_gencnt++; 3527 3528 CTL3_UNLOCK(); 3529 } 3530 3531 /* 3532 * Removes one or more sockopt handlers from the global array. 3533 */ 3534 int 3535 ipfw_del_sopt_handler(struct ipfw_sopt_handler *sh, size_t count) 3536 { 3537 size_t sz; 3538 struct ipfw_sopt_handler *tmp, *h; 3539 int i; 3540 3541 CTL3_LOCK(); 3542 3543 for (i = 0; i < count; i++) { 3544 tmp = &sh[i]; 3545 h = find_sh(tmp->opcode, tmp->version, tmp->handler); 3546 if (h == NULL) 3547 continue; 3548 3549 sz = (ctl3_handlers + ctl3_hsize - (h + 1)) * sizeof(*h); 3550 memmove(h, h + 1, sz); 3551 ctl3_hsize--; 3552 } 3553 3554 if (ctl3_hsize == 0) { 3555 if (ctl3_handlers != NULL) 3556 free(ctl3_handlers, M_IPFW); 3557 ctl3_handlers = NULL; 3558 } 3559 3560 ctl3_gencnt++; 3561 3562 CTL3_UNLOCK(); 3563 3564 return (0); 3565 } 3566 3567 /* 3568 * Writes data accumulated in @sd to sockopt buffer. 3569 * Zeroes internal @sd buffer. 3570 */ 3571 static int 3572 ipfw_flush_sopt_data(struct sockopt_data *sd) 3573 { 3574 struct sockopt *sopt; 3575 int error; 3576 size_t sz; 3577 3578 sz = sd->koff; 3579 if (sz == 0) 3580 return (0); 3581 3582 sopt = sd->sopt; 3583 3584 if (sopt->sopt_dir == SOPT_GET) { 3585 error = copyout(sd->kbuf, sopt->sopt_val, sz); 3586 if (error != 0) 3587 return (error); 3588 } 3589 3590 memset(sd->kbuf, 0, sd->ksize); 3591 sd->ktotal += sz; 3592 sd->koff = 0; 3593 if (sd->ktotal + sd->ksize < sd->valsize) 3594 sd->kavail = sd->ksize; 3595 else 3596 sd->kavail = sd->valsize - sd->ktotal; 3597 3598 /* Update sopt buffer data */ 3599 sopt->sopt_valsize = sd->ktotal; 3600 sopt->sopt_val = sd->sopt_val + sd->ktotal; 3601 3602 return (0); 3603 } 3604 3605 /* 3606 * Ensures that @sd buffer has contiguous @neeeded number of 3607 * bytes. 3608 * 3609 * Returns pointer to requested space or NULL. 3610 */ 3611 caddr_t 3612 ipfw_get_sopt_space(struct sockopt_data *sd, size_t needed) 3613 { 3614 int error; 3615 caddr_t addr; 3616 3617 if (sd->kavail < needed) { 3618 /* 3619 * Flush data and try another time. 3620 */ 3621 error = ipfw_flush_sopt_data(sd); 3622 3623 if (sd->kavail < needed || error != 0) 3624 return (NULL); 3625 } 3626 3627 addr = sd->kbuf + sd->koff; 3628 sd->koff += needed; 3629 sd->kavail -= needed; 3630 return (addr); 3631 } 3632 3633 /* 3634 * Requests @needed contiguous bytes from @sd buffer. 3635 * Function is used to notify subsystem that we are 3636 * interesed in first @needed bytes (request header) 3637 * and the rest buffer can be safely zeroed. 3638 * 3639 * Returns pointer to requested space or NULL. 3640 */ 3641 caddr_t 3642 ipfw_get_sopt_header(struct sockopt_data *sd, size_t needed) 3643 { 3644 caddr_t addr; 3645 3646 if ((addr = ipfw_get_sopt_space(sd, needed)) == NULL) 3647 return (NULL); 3648 3649 if (sd->kavail > 0) 3650 memset(sd->kbuf + sd->koff, 0, sd->kavail); 3651 3652 return (addr); 3653 } 3654 3655 /* 3656 * New sockopt handler. 3657 */ 3658 int 3659 ipfw_ctl3(struct sockopt *sopt) 3660 { 3661 int error, locked; 3662 size_t size, valsize; 3663 struct ip_fw_chain *chain; 3664 char xbuf[256]; 3665 struct sockopt_data sdata; 3666 struct ipfw_sopt_handler h; 3667 ip_fw3_opheader *op3 = NULL; 3668 3669 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW); 3670 if (error != 0) 3671 return (error); 3672 3673 if (sopt->sopt_name != IP_FW3) 3674 return (ipfw_ctl(sopt)); 3675 3676 chain = &V_layer3_chain; 3677 error = 0; 3678 3679 /* Save original valsize before it is altered via sooptcopyin() */ 3680 valsize = sopt->sopt_valsize; 3681 memset(&sdata, 0, sizeof(sdata)); 3682 /* Read op3 header first to determine actual operation */ 3683 op3 = (ip_fw3_opheader *)xbuf; 3684 error = sooptcopyin(sopt, op3, sizeof(*op3), sizeof(*op3)); 3685 if (error != 0) 3686 return (error); 3687 sopt->sopt_valsize = valsize; 3688 3689 /* 3690 * Find and reference command. 3691 */ 3692 error = find_ref_sh(op3->opcode, op3->version, &h); 3693 if (error != 0) 3694 return (error); 3695 3696 /* 3697 * Disallow modifications in really-really secure mode, but still allow 3698 * the logging counters to be reset. 3699 */ 3700 if ((h.dir & HDIR_SET) != 0 && h.opcode != IP_FW_XRESETLOG) { 3701 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3702 if (error != 0) { 3703 find_unref_sh(&h); 3704 return (error); 3705 } 3706 } 3707 3708 /* 3709 * Fill in sockopt_data structure that may be useful for 3710 * IP_FW3 get requests. 3711 */ 3712 locked = 0; 3713 if (valsize <= sizeof(xbuf)) { 3714 /* use on-stack buffer */ 3715 sdata.kbuf = xbuf; 3716 sdata.ksize = sizeof(xbuf); 3717 sdata.kavail = valsize; 3718 } else { 3719 /* 3720 * Determine opcode type/buffer size: 3721 * allocate sliding-window buf for data export or 3722 * contiguous buffer for special ops. 3723 */ 3724 if ((h.dir & HDIR_SET) != 0) { 3725 /* Set request. Allocate contigous buffer. */ 3726 if (valsize > CTL3_LARGEBUF) { 3727 find_unref_sh(&h); 3728 return (EFBIG); 3729 } 3730 3731 size = valsize; 3732 } else { 3733 /* Get request. Allocate sliding window buffer */ 3734 size = (valsize<CTL3_SMALLBUF) ? valsize:CTL3_SMALLBUF; 3735 3736 if (size < valsize) { 3737 /* We have to wire user buffer */ 3738 error = vslock(sopt->sopt_val, valsize); 3739 if (error != 0) 3740 return (error); 3741 locked = 1; 3742 } 3743 } 3744 3745 sdata.kbuf = malloc(size, M_TEMP, M_WAITOK | M_ZERO); 3746 sdata.ksize = size; 3747 sdata.kavail = size; 3748 } 3749 3750 sdata.sopt = sopt; 3751 sdata.sopt_val = sopt->sopt_val; 3752 sdata.valsize = valsize; 3753 3754 /* 3755 * Copy either all request (if valsize < bsize_max) 3756 * or first bsize_max bytes to guarantee most consumers 3757 * that all necessary data has been copied). 3758 * Anyway, copy not less than sizeof(ip_fw3_opheader). 3759 */ 3760 if ((error = sooptcopyin(sopt, sdata.kbuf, sdata.ksize, 3761 sizeof(ip_fw3_opheader))) != 0) 3762 return (error); 3763 op3 = (ip_fw3_opheader *)sdata.kbuf; 3764 3765 /* Finally, run handler */ 3766 error = h.handler(chain, op3, &sdata); 3767 find_unref_sh(&h); 3768 3769 /* Flush state and free buffers */ 3770 if (error == 0) 3771 error = ipfw_flush_sopt_data(&sdata); 3772 else 3773 ipfw_flush_sopt_data(&sdata); 3774 3775 if (locked != 0) 3776 vsunlock(sdata.sopt_val, valsize); 3777 3778 /* Restore original pointer and set number of bytes written */ 3779 sopt->sopt_val = sdata.sopt_val; 3780 sopt->sopt_valsize = sdata.ktotal; 3781 if (sdata.kbuf != xbuf) 3782 free(sdata.kbuf, M_TEMP); 3783 3784 return (error); 3785 } 3786 3787 /** 3788 * {set|get}sockopt parser. 3789 */ 3790 int 3791 ipfw_ctl(struct sockopt *sopt) 3792 { 3793 #define RULE_MAXSIZE (512*sizeof(u_int32_t)) 3794 int error; 3795 size_t size; 3796 struct ip_fw *buf; 3797 struct ip_fw_rule0 *rule; 3798 struct ip_fw_chain *chain; 3799 u_int32_t rulenum[2]; 3800 uint32_t opt; 3801 struct rule_check_info ci; 3802 IPFW_RLOCK_TRACKER; 3803 3804 chain = &V_layer3_chain; 3805 error = 0; 3806 3807 opt = sopt->sopt_name; 3808 3809 /* 3810 * Disallow modifications in really-really secure mode, but still allow 3811 * the logging counters to be reset. 3812 */ 3813 if (opt == IP_FW_ADD || 3814 (sopt->sopt_dir == SOPT_SET && opt != IP_FW_RESETLOG)) { 3815 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3816 if (error != 0) 3817 return (error); 3818 } 3819 3820 switch (opt) { 3821 case IP_FW_GET: 3822 /* 3823 * pass up a copy of the current rules. Static rules 3824 * come first (the last of which has number IPFW_DEFAULT_RULE), 3825 * followed by a possibly empty list of dynamic rule. 3826 * The last dynamic rule has NULL in the "next" field. 3827 * 3828 * Note that the calculated size is used to bound the 3829 * amount of data returned to the user. The rule set may 3830 * change between calculating the size and returning the 3831 * data in which case we'll just return what fits. 3832 */ 3833 for (;;) { 3834 int len = 0, want; 3835 3836 size = chain->static_len; 3837 size += ipfw_dyn_len(); 3838 if (size >= sopt->sopt_valsize) 3839 break; 3840 buf = malloc(size, M_TEMP, M_WAITOK | M_ZERO); 3841 IPFW_UH_RLOCK(chain); 3842 /* check again how much space we need */ 3843 want = chain->static_len + ipfw_dyn_len(); 3844 if (size >= want) 3845 len = ipfw_getrules(chain, buf, size); 3846 IPFW_UH_RUNLOCK(chain); 3847 if (size >= want) 3848 error = sooptcopyout(sopt, buf, len); 3849 free(buf, M_TEMP); 3850 if (size >= want) 3851 break; 3852 } 3853 break; 3854 3855 case IP_FW_FLUSH: 3856 /* locking is done within del_entry() */ 3857 error = del_entry(chain, 0); /* special case, rule=0, cmd=0 means all */ 3858 break; 3859 3860 case IP_FW_ADD: 3861 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 3862 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 3863 sizeof(struct ip_fw7) ); 3864 3865 memset(&ci, 0, sizeof(struct rule_check_info)); 3866 3867 /* 3868 * If the size of commands equals RULESIZE7 then we assume 3869 * a FreeBSD7.2 binary is talking to us (set is7=1). 3870 * is7 is persistent so the next 'ipfw list' command 3871 * will use this format. 3872 * NOTE: If wrong version is guessed (this can happen if 3873 * the first ipfw command is 'ipfw [pipe] list') 3874 * the ipfw binary may crash or loop infinitly... 3875 */ 3876 size = sopt->sopt_valsize; 3877 if (size == RULESIZE7(rule)) { 3878 is7 = 1; 3879 error = convert_rule_to_8(rule); 3880 if (error) { 3881 free(rule, M_TEMP); 3882 return error; 3883 } 3884 size = RULESIZE(rule); 3885 } else 3886 is7 = 0; 3887 if (error == 0) 3888 error = check_ipfw_rule0(rule, size, &ci); 3889 if (error == 0) { 3890 /* locking is done within add_rule() */ 3891 struct ip_fw *krule; 3892 krule = ipfw_alloc_rule(chain, RULEKSIZE0(rule)); 3893 ci.urule = (caddr_t)rule; 3894 ci.krule = krule; 3895 import_rule0(&ci); 3896 error = commit_rules(chain, &ci, 1); 3897 if (error != 0) 3898 ipfw_free_rule(ci.krule); 3899 else if (sopt->sopt_dir == SOPT_GET) { 3900 if (is7) { 3901 error = convert_rule_to_7(rule); 3902 size = RULESIZE7(rule); 3903 if (error) { 3904 free(rule, M_TEMP); 3905 return error; 3906 } 3907 } 3908 error = sooptcopyout(sopt, rule, size); 3909 } 3910 } 3911 free(rule, M_TEMP); 3912 break; 3913 3914 case IP_FW_DEL: 3915 /* 3916 * IP_FW_DEL is used for deleting single rules or sets, 3917 * and (ab)used to atomically manipulate sets. Argument size 3918 * is used to distinguish between the two: 3919 * sizeof(u_int32_t) 3920 * delete single rule or set of rules, 3921 * or reassign rules (or sets) to a different set. 3922 * 2*sizeof(u_int32_t) 3923 * atomic disable/enable sets. 3924 * first u_int32_t contains sets to be disabled, 3925 * second u_int32_t contains sets to be enabled. 3926 */ 3927 error = sooptcopyin(sopt, rulenum, 3928 2*sizeof(u_int32_t), sizeof(u_int32_t)); 3929 if (error) 3930 break; 3931 size = sopt->sopt_valsize; 3932 if (size == sizeof(u_int32_t) && rulenum[0] != 0) { 3933 /* delete or reassign, locking done in del_entry() */ 3934 error = del_entry(chain, rulenum[0]); 3935 } else if (size == 2*sizeof(u_int32_t)) { /* set enable/disable */ 3936 IPFW_UH_WLOCK(chain); 3937 V_set_disable = 3938 (V_set_disable | rulenum[0]) & ~rulenum[1] & 3939 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 3940 IPFW_UH_WUNLOCK(chain); 3941 } else 3942 error = EINVAL; 3943 break; 3944 3945 case IP_FW_ZERO: 3946 case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */ 3947 rulenum[0] = 0; 3948 if (sopt->sopt_val != 0) { 3949 error = sooptcopyin(sopt, rulenum, 3950 sizeof(u_int32_t), sizeof(u_int32_t)); 3951 if (error) 3952 break; 3953 } 3954 error = zero_entry(chain, rulenum[0], 3955 sopt->sopt_name == IP_FW_RESETLOG); 3956 break; 3957 3958 /*--- TABLE opcodes ---*/ 3959 case IP_FW_TABLE_ADD: 3960 case IP_FW_TABLE_DEL: 3961 { 3962 ipfw_table_entry ent; 3963 struct tentry_info tei; 3964 struct tid_info ti; 3965 struct table_value v; 3966 3967 error = sooptcopyin(sopt, &ent, 3968 sizeof(ent), sizeof(ent)); 3969 if (error) 3970 break; 3971 3972 memset(&tei, 0, sizeof(tei)); 3973 tei.paddr = &ent.addr; 3974 tei.subtype = AF_INET; 3975 tei.masklen = ent.masklen; 3976 ipfw_import_table_value_legacy(ent.value, &v); 3977 tei.pvalue = &v; 3978 memset(&ti, 0, sizeof(ti)); 3979 ti.uidx = ent.tbl; 3980 ti.type = IPFW_TABLE_CIDR; 3981 3982 error = (opt == IP_FW_TABLE_ADD) ? 3983 add_table_entry(chain, &ti, &tei, 0, 1) : 3984 del_table_entry(chain, &ti, &tei, 0, 1); 3985 } 3986 break; 3987 3988 case IP_FW_TABLE_FLUSH: 3989 { 3990 u_int16_t tbl; 3991 struct tid_info ti; 3992 3993 error = sooptcopyin(sopt, &tbl, 3994 sizeof(tbl), sizeof(tbl)); 3995 if (error) 3996 break; 3997 memset(&ti, 0, sizeof(ti)); 3998 ti.uidx = tbl; 3999 error = flush_table(chain, &ti); 4000 } 4001 break; 4002 4003 case IP_FW_TABLE_GETSIZE: 4004 { 4005 u_int32_t tbl, cnt; 4006 struct tid_info ti; 4007 4008 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4009 sizeof(tbl)))) 4010 break; 4011 memset(&ti, 0, sizeof(ti)); 4012 ti.uidx = tbl; 4013 IPFW_RLOCK(chain); 4014 error = ipfw_count_table(chain, &ti, &cnt); 4015 IPFW_RUNLOCK(chain); 4016 if (error) 4017 break; 4018 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4019 } 4020 break; 4021 4022 case IP_FW_TABLE_LIST: 4023 { 4024 ipfw_table *tbl; 4025 struct tid_info ti; 4026 4027 if (sopt->sopt_valsize < sizeof(*tbl)) { 4028 error = EINVAL; 4029 break; 4030 } 4031 size = sopt->sopt_valsize; 4032 tbl = malloc(size, M_TEMP, M_WAITOK); 4033 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4034 if (error) { 4035 free(tbl, M_TEMP); 4036 break; 4037 } 4038 tbl->size = (size - sizeof(*tbl)) / 4039 sizeof(ipfw_table_entry); 4040 memset(&ti, 0, sizeof(ti)); 4041 ti.uidx = tbl->tbl; 4042 IPFW_RLOCK(chain); 4043 error = ipfw_dump_table_legacy(chain, &ti, tbl); 4044 IPFW_RUNLOCK(chain); 4045 if (error) { 4046 free(tbl, M_TEMP); 4047 break; 4048 } 4049 error = sooptcopyout(sopt, tbl, size); 4050 free(tbl, M_TEMP); 4051 } 4052 break; 4053 4054 /*--- NAT operations are protected by the IPFW_LOCK ---*/ 4055 case IP_FW_NAT_CFG: 4056 if (IPFW_NAT_LOADED) 4057 error = ipfw_nat_cfg_ptr(sopt); 4058 else { 4059 printf("IP_FW_NAT_CFG: %s\n", 4060 "ipfw_nat not present, please load it"); 4061 error = EINVAL; 4062 } 4063 break; 4064 4065 case IP_FW_NAT_DEL: 4066 if (IPFW_NAT_LOADED) 4067 error = ipfw_nat_del_ptr(sopt); 4068 else { 4069 printf("IP_FW_NAT_DEL: %s\n", 4070 "ipfw_nat not present, please load it"); 4071 error = EINVAL; 4072 } 4073 break; 4074 4075 case IP_FW_NAT_GET_CONFIG: 4076 if (IPFW_NAT_LOADED) 4077 error = ipfw_nat_get_cfg_ptr(sopt); 4078 else { 4079 printf("IP_FW_NAT_GET_CFG: %s\n", 4080 "ipfw_nat not present, please load it"); 4081 error = EINVAL; 4082 } 4083 break; 4084 4085 case IP_FW_NAT_GET_LOG: 4086 if (IPFW_NAT_LOADED) 4087 error = ipfw_nat_get_log_ptr(sopt); 4088 else { 4089 printf("IP_FW_NAT_GET_LOG: %s\n", 4090 "ipfw_nat not present, please load it"); 4091 error = EINVAL; 4092 } 4093 break; 4094 4095 default: 4096 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4097 error = EINVAL; 4098 } 4099 4100 return (error); 4101 #undef RULE_MAXSIZE 4102 } 4103 #define RULE_MAXSIZE (256*sizeof(u_int32_t)) 4104 4105 /* Functions to convert rules 7.2 <==> 8.0 */ 4106 static int 4107 convert_rule_to_7(struct ip_fw_rule0 *rule) 4108 { 4109 /* Used to modify original rule */ 4110 struct ip_fw7 *rule7 = (struct ip_fw7 *)rule; 4111 /* copy of original rule, version 8 */ 4112 struct ip_fw_rule0 *tmp; 4113 4114 /* Used to copy commands */ 4115 ipfw_insn *ccmd, *dst; 4116 int ll = 0, ccmdlen = 0; 4117 4118 tmp = malloc(RULE_MAXSIZE, M_TEMP, M_NOWAIT | M_ZERO); 4119 if (tmp == NULL) { 4120 return 1; //XXX error 4121 } 4122 bcopy(rule, tmp, RULE_MAXSIZE); 4123 4124 /* Copy fields */ 4125 //rule7->_pad = tmp->_pad; 4126 rule7->set = tmp->set; 4127 rule7->rulenum = tmp->rulenum; 4128 rule7->cmd_len = tmp->cmd_len; 4129 rule7->act_ofs = tmp->act_ofs; 4130 rule7->next_rule = (struct ip_fw7 *)tmp->next_rule; 4131 rule7->cmd_len = tmp->cmd_len; 4132 rule7->pcnt = tmp->pcnt; 4133 rule7->bcnt = tmp->bcnt; 4134 rule7->timestamp = tmp->timestamp; 4135 4136 /* Copy commands */ 4137 for (ll = tmp->cmd_len, ccmd = tmp->cmd, dst = rule7->cmd ; 4138 ll > 0 ; ll -= ccmdlen, ccmd += ccmdlen, dst += ccmdlen) { 4139 ccmdlen = F_LEN(ccmd); 4140 4141 bcopy(ccmd, dst, F_LEN(ccmd)*sizeof(uint32_t)); 4142 4143 if (dst->opcode > O_NAT) 4144 /* O_REASS doesn't exists in 7.2 version, so 4145 * decrement opcode if it is after O_REASS 4146 */ 4147 dst->opcode--; 4148 4149 if (ccmdlen > ll) { 4150 printf("ipfw: opcode %d size truncated\n", 4151 ccmd->opcode); 4152 return EINVAL; 4153 } 4154 } 4155 free(tmp, M_TEMP); 4156 4157 return 0; 4158 } 4159 4160 static int 4161 convert_rule_to_8(struct ip_fw_rule0 *rule) 4162 { 4163 /* Used to modify original rule */ 4164 struct ip_fw7 *rule7 = (struct ip_fw7 *) rule; 4165 4166 /* Used to copy commands */ 4167 ipfw_insn *ccmd, *dst; 4168 int ll = 0, ccmdlen = 0; 4169 4170 /* Copy of original rule */ 4171 struct ip_fw7 *tmp = malloc(RULE_MAXSIZE, M_TEMP, M_NOWAIT | M_ZERO); 4172 if (tmp == NULL) { 4173 return 1; //XXX error 4174 } 4175 4176 bcopy(rule7, tmp, RULE_MAXSIZE); 4177 4178 for (ll = tmp->cmd_len, ccmd = tmp->cmd, dst = rule->cmd ; 4179 ll > 0 ; ll -= ccmdlen, ccmd += ccmdlen, dst += ccmdlen) { 4180 ccmdlen = F_LEN(ccmd); 4181 4182 bcopy(ccmd, dst, F_LEN(ccmd)*sizeof(uint32_t)); 4183 4184 if (dst->opcode > O_NAT) 4185 /* O_REASS doesn't exists in 7.2 version, so 4186 * increment opcode if it is after O_REASS 4187 */ 4188 dst->opcode++; 4189 4190 if (ccmdlen > ll) { 4191 printf("ipfw: opcode %d size truncated\n", 4192 ccmd->opcode); 4193 return EINVAL; 4194 } 4195 } 4196 4197 rule->_pad = tmp->_pad; 4198 rule->set = tmp->set; 4199 rule->rulenum = tmp->rulenum; 4200 rule->cmd_len = tmp->cmd_len; 4201 rule->act_ofs = tmp->act_ofs; 4202 rule->next_rule = (struct ip_fw *)tmp->next_rule; 4203 rule->cmd_len = tmp->cmd_len; 4204 rule->id = 0; /* XXX see if is ok = 0 */ 4205 rule->pcnt = tmp->pcnt; 4206 rule->bcnt = tmp->bcnt; 4207 rule->timestamp = tmp->timestamp; 4208 4209 free (tmp, M_TEMP); 4210 return 0; 4211 } 4212 4213 /* 4214 * Named object api 4215 * 4216 */ 4217 4218 void 4219 ipfw_init_srv(struct ip_fw_chain *ch) 4220 { 4221 4222 ch->srvmap = ipfw_objhash_create(IPFW_OBJECTS_DEFAULT); 4223 ch->srvstate = malloc(sizeof(void *) * IPFW_OBJECTS_DEFAULT, 4224 M_IPFW, M_WAITOK | M_ZERO); 4225 } 4226 4227 void 4228 ipfw_destroy_srv(struct ip_fw_chain *ch) 4229 { 4230 4231 free(ch->srvstate, M_IPFW); 4232 ipfw_objhash_destroy(ch->srvmap); 4233 } 4234 4235 /* 4236 * Allocate new bitmask which can be used to enlarge/shrink 4237 * named instance index. 4238 */ 4239 void 4240 ipfw_objhash_bitmap_alloc(uint32_t items, void **idx, int *pblocks) 4241 { 4242 size_t size; 4243 int max_blocks; 4244 u_long *idx_mask; 4245 4246 KASSERT((items % BLOCK_ITEMS) == 0, 4247 ("bitmask size needs to power of 2 and greater or equal to %zu", 4248 BLOCK_ITEMS)); 4249 4250 max_blocks = items / BLOCK_ITEMS; 4251 size = items / 8; 4252 idx_mask = malloc(size * IPFW_MAX_SETS, M_IPFW, M_WAITOK); 4253 /* Mark all as free */ 4254 memset(idx_mask, 0xFF, size * IPFW_MAX_SETS); 4255 *idx_mask &= ~(u_long)1; /* Skip index 0 */ 4256 4257 *idx = idx_mask; 4258 *pblocks = max_blocks; 4259 } 4260 4261 /* 4262 * Copy current bitmask index to new one. 4263 */ 4264 void 4265 ipfw_objhash_bitmap_merge(struct namedobj_instance *ni, void **idx, int *blocks) 4266 { 4267 int old_blocks, new_blocks; 4268 u_long *old_idx, *new_idx; 4269 int i; 4270 4271 old_idx = ni->idx_mask; 4272 old_blocks = ni->max_blocks; 4273 new_idx = *idx; 4274 new_blocks = *blocks; 4275 4276 for (i = 0; i < IPFW_MAX_SETS; i++) { 4277 memcpy(&new_idx[new_blocks * i], &old_idx[old_blocks * i], 4278 old_blocks * sizeof(u_long)); 4279 } 4280 } 4281 4282 /* 4283 * Swaps current @ni index with new one. 4284 */ 4285 void 4286 ipfw_objhash_bitmap_swap(struct namedobj_instance *ni, void **idx, int *blocks) 4287 { 4288 int old_blocks; 4289 u_long *old_idx; 4290 4291 old_idx = ni->idx_mask; 4292 old_blocks = ni->max_blocks; 4293 4294 ni->idx_mask = *idx; 4295 ni->max_blocks = *blocks; 4296 4297 /* Save old values */ 4298 *idx = old_idx; 4299 *blocks = old_blocks; 4300 } 4301 4302 void 4303 ipfw_objhash_bitmap_free(void *idx, int blocks) 4304 { 4305 4306 free(idx, M_IPFW); 4307 } 4308 4309 /* 4310 * Creates named hash instance. 4311 * Must be called without holding any locks. 4312 * Return pointer to new instance. 4313 */ 4314 struct namedobj_instance * 4315 ipfw_objhash_create(uint32_t items) 4316 { 4317 struct namedobj_instance *ni; 4318 int i; 4319 size_t size; 4320 4321 size = sizeof(struct namedobj_instance) + 4322 sizeof(struct namedobjects_head) * NAMEDOBJ_HASH_SIZE + 4323 sizeof(struct namedobjects_head) * NAMEDOBJ_HASH_SIZE; 4324 4325 ni = malloc(size, M_IPFW, M_WAITOK | M_ZERO); 4326 ni->nn_size = NAMEDOBJ_HASH_SIZE; 4327 ni->nv_size = NAMEDOBJ_HASH_SIZE; 4328 4329 ni->names = (struct namedobjects_head *)(ni +1); 4330 ni->values = &ni->names[ni->nn_size]; 4331 4332 for (i = 0; i < ni->nn_size; i++) 4333 TAILQ_INIT(&ni->names[i]); 4334 4335 for (i = 0; i < ni->nv_size; i++) 4336 TAILQ_INIT(&ni->values[i]); 4337 4338 /* Set default hashing/comparison functions */ 4339 ni->hash_f = objhash_hash_name; 4340 ni->cmp_f = objhash_cmp_name; 4341 4342 /* Allocate bitmask separately due to possible resize */ 4343 ipfw_objhash_bitmap_alloc(items, (void*)&ni->idx_mask, &ni->max_blocks); 4344 4345 return (ni); 4346 } 4347 4348 void 4349 ipfw_objhash_destroy(struct namedobj_instance *ni) 4350 { 4351 4352 free(ni->idx_mask, M_IPFW); 4353 free(ni, M_IPFW); 4354 } 4355 4356 void 4357 ipfw_objhash_set_funcs(struct namedobj_instance *ni, objhash_hash_f *hash_f, 4358 objhash_cmp_f *cmp_f) 4359 { 4360 4361 ni->hash_f = hash_f; 4362 ni->cmp_f = cmp_f; 4363 } 4364 4365 static uint32_t 4366 objhash_hash_name(struct namedobj_instance *ni, const void *name, uint32_t set) 4367 { 4368 4369 return (fnv_32_str((const char *)name, FNV1_32_INIT)); 4370 } 4371 4372 static int 4373 objhash_cmp_name(struct named_object *no, const void *name, uint32_t set) 4374 { 4375 4376 if ((strcmp(no->name, (const char *)name) == 0) && (no->set == set)) 4377 return (0); 4378 4379 return (1); 4380 } 4381 4382 static uint32_t 4383 objhash_hash_idx(struct namedobj_instance *ni, uint32_t val) 4384 { 4385 uint32_t v; 4386 4387 v = val % (ni->nv_size - 1); 4388 4389 return (v); 4390 } 4391 4392 struct named_object * 4393 ipfw_objhash_lookup_name(struct namedobj_instance *ni, uint32_t set, char *name) 4394 { 4395 struct named_object *no; 4396 uint32_t hash; 4397 4398 hash = ni->hash_f(ni, name, set) % ni->nn_size; 4399 4400 TAILQ_FOREACH(no, &ni->names[hash], nn_next) { 4401 if (ni->cmp_f(no, name, set) == 0) 4402 return (no); 4403 } 4404 4405 return (NULL); 4406 } 4407 4408 /* 4409 * Find named object by @uid. 4410 * Check @tlvs for valid data inside. 4411 * 4412 * Returns pointer to found TLV or NULL. 4413 */ 4414 ipfw_obj_ntlv * 4415 ipfw_find_name_tlv_type(void *tlvs, int len, uint16_t uidx, uint32_t etlv) 4416 { 4417 ipfw_obj_ntlv *ntlv; 4418 uintptr_t pa, pe; 4419 int l; 4420 4421 pa = (uintptr_t)tlvs; 4422 pe = pa + len; 4423 l = 0; 4424 for (; pa < pe; pa += l) { 4425 ntlv = (ipfw_obj_ntlv *)pa; 4426 l = ntlv->head.length; 4427 4428 if (l != sizeof(*ntlv)) 4429 return (NULL); 4430 4431 if (ntlv->idx != uidx) 4432 continue; 4433 /* 4434 * When userland has specified zero TLV type, do 4435 * not compare it with eltv. In some cases userland 4436 * doesn't know what type should it have. Use only 4437 * uidx and name for search named_object. 4438 */ 4439 if (ntlv->head.type != 0 && 4440 ntlv->head.type != (uint16_t)etlv) 4441 continue; 4442 4443 if (ipfw_check_object_name_generic(ntlv->name) != 0) 4444 return (NULL); 4445 4446 return (ntlv); 4447 } 4448 4449 return (NULL); 4450 } 4451 4452 /* 4453 * Finds object config based on either legacy index 4454 * or name in ntlv. 4455 * Note @ti structure contains unchecked data from userland. 4456 * 4457 * Returns 0 in success and fills in @pno with found config 4458 */ 4459 int 4460 ipfw_objhash_find_type(struct namedobj_instance *ni, struct tid_info *ti, 4461 uint32_t etlv, struct named_object **pno) 4462 { 4463 char *name; 4464 ipfw_obj_ntlv *ntlv; 4465 uint32_t set; 4466 4467 if (ti->tlvs == NULL) 4468 return (EINVAL); 4469 4470 ntlv = ipfw_find_name_tlv_type(ti->tlvs, ti->tlen, ti->uidx, etlv); 4471 if (ntlv == NULL) 4472 return (EINVAL); 4473 name = ntlv->name; 4474 4475 /* 4476 * Use set provided by @ti instead of @ntlv one. 4477 * This is needed due to different sets behavior 4478 * controlled by V_fw_tables_sets. 4479 */ 4480 set = ti->set; 4481 *pno = ipfw_objhash_lookup_name(ni, set, name); 4482 if (*pno == NULL) 4483 return (ESRCH); 4484 return (0); 4485 } 4486 4487 /* 4488 * Find named object by name, considering also its TLV type. 4489 */ 4490 struct named_object * 4491 ipfw_objhash_lookup_name_type(struct namedobj_instance *ni, uint32_t set, 4492 uint32_t type, const char *name) 4493 { 4494 struct named_object *no; 4495 uint32_t hash; 4496 4497 hash = ni->hash_f(ni, name, set) % ni->nn_size; 4498 4499 TAILQ_FOREACH(no, &ni->names[hash], nn_next) { 4500 if (ni->cmp_f(no, name, set) == 0 && 4501 no->etlv == (uint16_t)type) 4502 return (no); 4503 } 4504 4505 return (NULL); 4506 } 4507 4508 struct named_object * 4509 ipfw_objhash_lookup_kidx(struct namedobj_instance *ni, uint16_t kidx) 4510 { 4511 struct named_object *no; 4512 uint32_t hash; 4513 4514 hash = objhash_hash_idx(ni, kidx); 4515 4516 TAILQ_FOREACH(no, &ni->values[hash], nv_next) { 4517 if (no->kidx == kidx) 4518 return (no); 4519 } 4520 4521 return (NULL); 4522 } 4523 4524 int 4525 ipfw_objhash_same_name(struct namedobj_instance *ni, struct named_object *a, 4526 struct named_object *b) 4527 { 4528 4529 if ((strcmp(a->name, b->name) == 0) && a->set == b->set) 4530 return (1); 4531 4532 return (0); 4533 } 4534 4535 void 4536 ipfw_objhash_add(struct namedobj_instance *ni, struct named_object *no) 4537 { 4538 uint32_t hash; 4539 4540 hash = ni->hash_f(ni, no->name, no->set) % ni->nn_size; 4541 TAILQ_INSERT_HEAD(&ni->names[hash], no, nn_next); 4542 4543 hash = objhash_hash_idx(ni, no->kidx); 4544 TAILQ_INSERT_HEAD(&ni->values[hash], no, nv_next); 4545 4546 ni->count++; 4547 } 4548 4549 void 4550 ipfw_objhash_del(struct namedobj_instance *ni, struct named_object *no) 4551 { 4552 uint32_t hash; 4553 4554 hash = ni->hash_f(ni, no->name, no->set) % ni->nn_size; 4555 TAILQ_REMOVE(&ni->names[hash], no, nn_next); 4556 4557 hash = objhash_hash_idx(ni, no->kidx); 4558 TAILQ_REMOVE(&ni->values[hash], no, nv_next); 4559 4560 ni->count--; 4561 } 4562 4563 uint32_t 4564 ipfw_objhash_count(struct namedobj_instance *ni) 4565 { 4566 4567 return (ni->count); 4568 } 4569 4570 uint32_t 4571 ipfw_objhash_count_type(struct namedobj_instance *ni, uint16_t type) 4572 { 4573 struct named_object *no; 4574 uint32_t count; 4575 int i; 4576 4577 count = 0; 4578 for (i = 0; i < ni->nn_size; i++) { 4579 TAILQ_FOREACH(no, &ni->names[i], nn_next) { 4580 if (no->etlv == type) 4581 count++; 4582 } 4583 } 4584 return (count); 4585 } 4586 4587 /* 4588 * Runs @func for each found named object. 4589 * It is safe to delete objects from callback 4590 */ 4591 int 4592 ipfw_objhash_foreach(struct namedobj_instance *ni, objhash_cb_t *f, void *arg) 4593 { 4594 struct named_object *no, *no_tmp; 4595 int i, ret; 4596 4597 for (i = 0; i < ni->nn_size; i++) { 4598 TAILQ_FOREACH_SAFE(no, &ni->names[i], nn_next, no_tmp) { 4599 ret = f(ni, no, arg); 4600 if (ret != 0) 4601 return (ret); 4602 } 4603 } 4604 return (0); 4605 } 4606 4607 /* 4608 * Runs @f for each found named object with type @type. 4609 * It is safe to delete objects from callback 4610 */ 4611 int 4612 ipfw_objhash_foreach_type(struct namedobj_instance *ni, objhash_cb_t *f, 4613 void *arg, uint16_t type) 4614 { 4615 struct named_object *no, *no_tmp; 4616 int i, ret; 4617 4618 for (i = 0; i < ni->nn_size; i++) { 4619 TAILQ_FOREACH_SAFE(no, &ni->names[i], nn_next, no_tmp) { 4620 if (no->etlv != type) 4621 continue; 4622 ret = f(ni, no, arg); 4623 if (ret != 0) 4624 return (ret); 4625 } 4626 } 4627 return (0); 4628 } 4629 4630 /* 4631 * Removes index from given set. 4632 * Returns 0 on success. 4633 */ 4634 int 4635 ipfw_objhash_free_idx(struct namedobj_instance *ni, uint16_t idx) 4636 { 4637 u_long *mask; 4638 int i, v; 4639 4640 i = idx / BLOCK_ITEMS; 4641 v = idx % BLOCK_ITEMS; 4642 4643 if (i >= ni->max_blocks) 4644 return (1); 4645 4646 mask = &ni->idx_mask[i]; 4647 4648 if ((*mask & ((u_long)1 << v)) != 0) 4649 return (1); 4650 4651 /* Mark as free */ 4652 *mask |= (u_long)1 << v; 4653 4654 /* Update free offset */ 4655 if (ni->free_off[0] > i) 4656 ni->free_off[0] = i; 4657 4658 return (0); 4659 } 4660 4661 /* 4662 * Allocate new index in given instance and stores in in @pidx. 4663 * Returns 0 on success. 4664 */ 4665 int 4666 ipfw_objhash_alloc_idx(void *n, uint16_t *pidx) 4667 { 4668 struct namedobj_instance *ni; 4669 u_long *mask; 4670 int i, off, v; 4671 4672 ni = (struct namedobj_instance *)n; 4673 4674 off = ni->free_off[0]; 4675 mask = &ni->idx_mask[off]; 4676 4677 for (i = off; i < ni->max_blocks; i++, mask++) { 4678 if ((v = ffsl(*mask)) == 0) 4679 continue; 4680 4681 /* Mark as busy */ 4682 *mask &= ~ ((u_long)1 << (v - 1)); 4683 4684 ni->free_off[0] = i; 4685 4686 v = BLOCK_ITEMS * i + v - 1; 4687 4688 *pidx = v; 4689 return (0); 4690 } 4691 4692 return (1); 4693 } 4694 4695 /* end of file */ 4696