ip_fw_sockopt.c revision 304843
1/*- 2 * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa 3 * Copyright (c) 2014 Yandex LLC 4 * Copyright (c) 2014 Alexander V. Chernikov 5 * 6 * Supported by: Valeria Paoli 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 */ 29 30#include <sys/cdefs.h> 31__FBSDID("$FreeBSD: stable/11/sys/netpfil/ipfw/ip_fw_sockopt.c 304843 2016-08-26 10:04:10Z kib $"); 32 33/* 34 * Control socket and rule management routines for ipfw. 35 * Control is currently implemented via IP_FW3 setsockopt() code. 36 */ 37 38#include "opt_ipfw.h" 39#include "opt_inet.h" 40#ifndef INET 41#error IPFIREWALL requires INET. 42#endif /* INET */ 43#include "opt_inet6.h" 44 45#include <sys/param.h> 46#include <sys/systm.h> 47#include <sys/malloc.h> 48#include <sys/mbuf.h> /* struct m_tag used by nested headers */ 49#include <sys/kernel.h> 50#include <sys/lock.h> 51#include <sys/priv.h> 52#include <sys/proc.h> 53#include <sys/rwlock.h> 54#include <sys/rmlock.h> 55#include <sys/socket.h> 56#include <sys/socketvar.h> 57#include <sys/sysctl.h> 58#include <sys/syslog.h> 59#include <sys/fnv_hash.h> 60#include <net/if.h> 61#include <net/route.h> 62#include <net/vnet.h> 63#include <vm/vm.h> 64#include <vm/vm_extern.h> 65 66#include <netinet/in.h> 67#include <netinet/ip_var.h> /* hooks */ 68#include <netinet/ip_fw.h> 69 70#include <netpfil/ipfw/ip_fw_private.h> 71#include <netpfil/ipfw/ip_fw_table.h> 72 73#ifdef MAC 74#include <security/mac/mac_framework.h> 75#endif 76 77static int ipfw_ctl(struct sockopt *sopt); 78static int check_ipfw_rule_body(ipfw_insn *cmd, int cmd_len, 79 struct rule_check_info *ci); 80static int check_ipfw_rule1(struct ip_fw_rule *rule, int size, 81 struct rule_check_info *ci); 82static int check_ipfw_rule0(struct ip_fw_rule0 *rule, int size, 83 struct rule_check_info *ci); 84static int rewrite_rule_uidx(struct ip_fw_chain *chain, 85 struct rule_check_info *ci); 86 87#define NAMEDOBJ_HASH_SIZE 32 88 89struct namedobj_instance { 90 struct namedobjects_head *names; 91 struct namedobjects_head *values; 92 uint32_t nn_size; /* names hash size */ 93 uint32_t nv_size; /* number hash size */ 94 u_long *idx_mask; /* used items bitmask */ 95 uint32_t max_blocks; /* number of "long" blocks in bitmask */ 96 uint32_t count; /* number of items */ 97 uint16_t free_off[IPFW_MAX_SETS]; /* first possible free offset */ 98 objhash_hash_f *hash_f; 99 objhash_cmp_f *cmp_f; 100}; 101#define BLOCK_ITEMS (8 * sizeof(u_long)) /* Number of items for ffsl() */ 102 103static uint32_t objhash_hash_name(struct namedobj_instance *ni, 104 const void *key, uint32_t kopt); 105static uint32_t objhash_hash_idx(struct namedobj_instance *ni, uint32_t val); 106static int objhash_cmp_name(struct named_object *no, const void *name, 107 uint32_t set); 108 109MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 110 111static int dump_config(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 112 struct sockopt_data *sd); 113static int add_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 114 struct sockopt_data *sd); 115static int del_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 116 struct sockopt_data *sd); 117static int clear_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 118 struct sockopt_data *sd); 119static int move_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 120 struct sockopt_data *sd); 121static int manage_sets(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 122 struct sockopt_data *sd); 123static int dump_soptcodes(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 124 struct sockopt_data *sd); 125static int dump_srvobjects(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 126 struct sockopt_data *sd); 127 128/* ctl3 handler data */ 129struct mtx ctl3_lock; 130#define CTL3_LOCK_INIT() mtx_init(&ctl3_lock, "ctl3_lock", NULL, MTX_DEF) 131#define CTL3_LOCK_DESTROY() mtx_destroy(&ctl3_lock) 132#define CTL3_LOCK() mtx_lock(&ctl3_lock) 133#define CTL3_UNLOCK() mtx_unlock(&ctl3_lock) 134 135static struct ipfw_sopt_handler *ctl3_handlers; 136static size_t ctl3_hsize; 137static uint64_t ctl3_refct, ctl3_gencnt; 138#define CTL3_SMALLBUF 4096 /* small page-size write buffer */ 139#define CTL3_LARGEBUF 16 * 1024 * 1024 /* handle large rulesets */ 140 141static int ipfw_flush_sopt_data(struct sockopt_data *sd); 142 143static struct ipfw_sopt_handler scodes[] = { 144 { IP_FW_XGET, 0, HDIR_GET, dump_config }, 145 { IP_FW_XADD, 0, HDIR_BOTH, add_rules }, 146 { IP_FW_XDEL, 0, HDIR_BOTH, del_rules }, 147 { IP_FW_XZERO, 0, HDIR_SET, clear_rules }, 148 { IP_FW_XRESETLOG, 0, HDIR_SET, clear_rules }, 149 { IP_FW_XMOVE, 0, HDIR_SET, move_rules }, 150 { IP_FW_SET_SWAP, 0, HDIR_SET, manage_sets }, 151 { IP_FW_SET_MOVE, 0, HDIR_SET, manage_sets }, 152 { IP_FW_SET_ENABLE, 0, HDIR_SET, manage_sets }, 153 { IP_FW_DUMP_SOPTCODES, 0, HDIR_GET, dump_soptcodes }, 154 { IP_FW_DUMP_SRVOBJECTS,0, HDIR_GET, dump_srvobjects }, 155}; 156 157static int 158set_legacy_obj_kidx(struct ip_fw_chain *ch, struct ip_fw_rule0 *rule); 159static struct opcode_obj_rewrite *find_op_rw(ipfw_insn *cmd, 160 uint16_t *puidx, uint8_t *ptype); 161static int mark_object_kidx(struct ip_fw_chain *ch, struct ip_fw *rule, 162 uint32_t *bmask); 163static 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); 165static int ref_opcode_object(struct ip_fw_chain *ch, ipfw_insn *cmd, 166 struct tid_info *ti, struct obj_idx *pidx, int *unresolved); 167static void unref_rule_objects(struct ip_fw_chain *chain, struct ip_fw *rule); 168static void unref_oib_objects(struct ip_fw_chain *ch, ipfw_insn *cmd, 169 struct obj_idx *oib, struct obj_idx *end); 170static int export_objhash_ntlv(struct namedobj_instance *ni, uint16_t kidx, 171 struct sockopt_data *sd); 172 173/* 174 * Opcode object rewriter variables 175 */ 176struct opcode_obj_rewrite *ctl3_rewriters; 177static size_t ctl3_rsize; 178 179/* 180 * static variables followed by global ones 181 */ 182 183static VNET_DEFINE(uma_zone_t, ipfw_cntr_zone); 184#define V_ipfw_cntr_zone VNET(ipfw_cntr_zone) 185 186void 187ipfw_init_counters() 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 195void 196ipfw_destroy_counters() 197{ 198 199 uma_zdestroy(V_ipfw_cntr_zone); 200} 201 202struct ip_fw * 203ipfw_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(V_ipfw_cntr_zone, M_WAITOK | M_ZERO); 209 210 return (rule); 211} 212 213static void 214free_rule(struct ip_fw *rule) 215{ 216 217 uma_zfree(V_ipfw_cntr_zone, rule->cntr); 218 free(rule, M_IPFW); 219} 220 221 222/* 223 * Find the smallest rule >= key, id. 224 * We could use bsearch but it is so simple that we code it directly 225 */ 226int 227ipfw_find_rule(struct ip_fw_chain *chain, uint32_t key, uint32_t id) 228{ 229 int i, lo, hi; 230 struct ip_fw *r; 231 232 for (lo = 0, hi = chain->n_rules - 1; lo < hi;) { 233 i = (lo + hi) / 2; 234 r = chain->map[i]; 235 if (r->rulenum < key) 236 lo = i + 1; /* continue from the next one */ 237 else if (r->rulenum > key) 238 hi = i; /* this might be good */ 239 else if (r->id < id) 240 lo = i + 1; /* continue from the next one */ 241 else /* r->id >= id */ 242 hi = i; /* this might be good */ 243 } 244 return hi; 245} 246 247/* 248 * Builds skipto cache on rule set @map. 249 */ 250static void 251update_skipto_cache(struct ip_fw_chain *chain, struct ip_fw **map) 252{ 253 int *smap, rulenum; 254 int i, mi; 255 256 IPFW_UH_WLOCK_ASSERT(chain); 257 258 mi = 0; 259 rulenum = map[mi]->rulenum; 260 smap = chain->idxmap_back; 261 262 if (smap == NULL) 263 return; 264 265 for (i = 0; i < 65536; i++) { 266 smap[i] = mi; 267 /* Use the same rule index until i < rulenum */ 268 if (i != rulenum || i == 65535) 269 continue; 270 /* Find next rule with num > i */ 271 rulenum = map[++mi]->rulenum; 272 while (rulenum == i) 273 rulenum = map[++mi]->rulenum; 274 } 275} 276 277/* 278 * Swaps prepared (backup) index with current one. 279 */ 280static void 281swap_skipto_cache(struct ip_fw_chain *chain) 282{ 283 int *map; 284 285 IPFW_UH_WLOCK_ASSERT(chain); 286 IPFW_WLOCK_ASSERT(chain); 287 288 map = chain->idxmap; 289 chain->idxmap = chain->idxmap_back; 290 chain->idxmap_back = map; 291} 292 293/* 294 * Allocate and initialize skipto cache. 295 */ 296void 297ipfw_init_skipto_cache(struct ip_fw_chain *chain) 298{ 299 int *idxmap, *idxmap_back; 300 301 idxmap = malloc(65536 * sizeof(uint32_t *), M_IPFW, 302 M_WAITOK | M_ZERO); 303 idxmap_back = malloc(65536 * sizeof(uint32_t *), M_IPFW, 304 M_WAITOK | M_ZERO); 305 306 /* 307 * Note we may be called at any time after initialization, 308 * for example, on first skipto rule, so we need to 309 * provide valid chain->idxmap on return 310 */ 311 312 IPFW_UH_WLOCK(chain); 313 if (chain->idxmap != NULL) { 314 IPFW_UH_WUNLOCK(chain); 315 free(idxmap, M_IPFW); 316 free(idxmap_back, M_IPFW); 317 return; 318 } 319 320 /* Set backup pointer first to permit building cache */ 321 chain->idxmap_back = idxmap_back; 322 update_skipto_cache(chain, chain->map); 323 IPFW_WLOCK(chain); 324 /* It is now safe to set chain->idxmap ptr */ 325 chain->idxmap = idxmap; 326 swap_skipto_cache(chain); 327 IPFW_WUNLOCK(chain); 328 IPFW_UH_WUNLOCK(chain); 329} 330 331/* 332 * Destroys skipto cache. 333 */ 334void 335ipfw_destroy_skipto_cache(struct ip_fw_chain *chain) 336{ 337 338 if (chain->idxmap != NULL) 339 free(chain->idxmap, M_IPFW); 340 if (chain->idxmap != NULL) 341 free(chain->idxmap_back, M_IPFW); 342} 343 344 345/* 346 * allocate a new map, returns the chain locked. extra is the number 347 * of entries to add or delete. 348 */ 349static struct ip_fw ** 350get_map(struct ip_fw_chain *chain, int extra, int locked) 351{ 352 353 for (;;) { 354 struct ip_fw **map; 355 int i, mflags; 356 357 mflags = M_ZERO | ((locked != 0) ? M_NOWAIT : M_WAITOK); 358 359 i = chain->n_rules + extra; 360 map = malloc(i * sizeof(struct ip_fw *), M_IPFW, mflags); 361 if (map == NULL) { 362 printf("%s: cannot allocate map\n", __FUNCTION__); 363 return NULL; 364 } 365 if (!locked) 366 IPFW_UH_WLOCK(chain); 367 if (i >= chain->n_rules + extra) /* good */ 368 return map; 369 /* otherwise we lost the race, free and retry */ 370 if (!locked) 371 IPFW_UH_WUNLOCK(chain); 372 free(map, M_IPFW); 373 } 374} 375 376/* 377 * swap the maps. It is supposed to be called with IPFW_UH_WLOCK 378 */ 379static struct ip_fw ** 380swap_map(struct ip_fw_chain *chain, struct ip_fw **new_map, int new_len) 381{ 382 struct ip_fw **old_map; 383 384 IPFW_WLOCK(chain); 385 chain->id++; 386 chain->n_rules = new_len; 387 old_map = chain->map; 388 chain->map = new_map; 389 swap_skipto_cache(chain); 390 IPFW_WUNLOCK(chain); 391 return old_map; 392} 393 394 395static void 396export_cntr1_base(struct ip_fw *krule, struct ip_fw_bcounter *cntr) 397{ 398 struct timeval boottime; 399 400 cntr->size = sizeof(*cntr); 401 402 if (krule->cntr != NULL) { 403 cntr->pcnt = counter_u64_fetch(krule->cntr); 404 cntr->bcnt = counter_u64_fetch(krule->cntr + 1); 405 cntr->timestamp = krule->timestamp; 406 } 407 if (cntr->timestamp > 0) { 408 getboottime(&boottime); 409 cntr->timestamp += boottime.tv_sec; 410 } 411} 412 413static void 414export_cntr0_base(struct ip_fw *krule, struct ip_fw_bcounter0 *cntr) 415{ 416 struct timeval boottime; 417 418 if (krule->cntr != NULL) { 419 cntr->pcnt = counter_u64_fetch(krule->cntr); 420 cntr->bcnt = counter_u64_fetch(krule->cntr + 1); 421 cntr->timestamp = krule->timestamp; 422 } 423 if (cntr->timestamp > 0) { 424 getboottime(&boottime); 425 cntr->timestamp += boottime.tv_sec; 426 } 427} 428 429/* 430 * Copies rule @urule from v1 userland format (current). 431 * to kernel @krule. 432 * Assume @krule is zeroed. 433 */ 434static void 435import_rule1(struct rule_check_info *ci) 436{ 437 struct ip_fw_rule *urule; 438 struct ip_fw *krule; 439 440 urule = (struct ip_fw_rule *)ci->urule; 441 krule = (struct ip_fw *)ci->krule; 442 443 /* copy header */ 444 krule->act_ofs = urule->act_ofs; 445 krule->cmd_len = urule->cmd_len; 446 krule->rulenum = urule->rulenum; 447 krule->set = urule->set; 448 krule->flags = urule->flags; 449 450 /* Save rulenum offset */ 451 ci->urule_numoff = offsetof(struct ip_fw_rule, rulenum); 452 453 /* Copy opcodes */ 454 memcpy(krule->cmd, urule->cmd, krule->cmd_len * sizeof(uint32_t)); 455} 456 457/* 458 * Export rule into v1 format (Current). 459 * Layout: 460 * [ ipfw_obj_tlv(IPFW_TLV_RULE_ENT) 461 * [ ip_fw_rule ] OR 462 * [ ip_fw_bcounter ip_fw_rule] (depends on rcntrs). 463 * ] 464 * Assume @data is zeroed. 465 */ 466static void 467export_rule1(struct ip_fw *krule, caddr_t data, int len, int rcntrs) 468{ 469 struct ip_fw_bcounter *cntr; 470 struct ip_fw_rule *urule; 471 ipfw_obj_tlv *tlv; 472 473 /* Fill in TLV header */ 474 tlv = (ipfw_obj_tlv *)data; 475 tlv->type = IPFW_TLV_RULE_ENT; 476 tlv->length = len; 477 478 if (rcntrs != 0) { 479 /* Copy counters */ 480 cntr = (struct ip_fw_bcounter *)(tlv + 1); 481 urule = (struct ip_fw_rule *)(cntr + 1); 482 export_cntr1_base(krule, cntr); 483 } else 484 urule = (struct ip_fw_rule *)(tlv + 1); 485 486 /* copy header */ 487 urule->act_ofs = krule->act_ofs; 488 urule->cmd_len = krule->cmd_len; 489 urule->rulenum = krule->rulenum; 490 urule->set = krule->set; 491 urule->flags = krule->flags; 492 urule->id = krule->id; 493 494 /* Copy opcodes */ 495 memcpy(urule->cmd, krule->cmd, krule->cmd_len * sizeof(uint32_t)); 496} 497 498 499/* 500 * Copies rule @urule from FreeBSD8 userland format (v0) 501 * to kernel @krule. 502 * Assume @krule is zeroed. 503 */ 504static void 505import_rule0(struct rule_check_info *ci) 506{ 507 struct ip_fw_rule0 *urule; 508 struct ip_fw *krule; 509 int cmdlen, l; 510 ipfw_insn *cmd; 511 ipfw_insn_limit *lcmd; 512 ipfw_insn_if *cmdif; 513 514 urule = (struct ip_fw_rule0 *)ci->urule; 515 krule = (struct ip_fw *)ci->krule; 516 517 /* copy header */ 518 krule->act_ofs = urule->act_ofs; 519 krule->cmd_len = urule->cmd_len; 520 krule->rulenum = urule->rulenum; 521 krule->set = urule->set; 522 if ((urule->_pad & 1) != 0) 523 krule->flags |= IPFW_RULE_NOOPT; 524 525 /* Save rulenum offset */ 526 ci->urule_numoff = offsetof(struct ip_fw_rule0, rulenum); 527 528 /* Copy opcodes */ 529 memcpy(krule->cmd, urule->cmd, krule->cmd_len * sizeof(uint32_t)); 530 531 /* 532 * Alter opcodes: 533 * 1) convert tablearg value from 65535 to 0 534 * 2) Add high bit to O_SETFIB/O_SETDSCP values (to make room 535 * for targ). 536 * 3) convert table number in iface opcodes to u16 537 * 4) convert old `nat global` into new 65535 538 */ 539 l = krule->cmd_len; 540 cmd = krule->cmd; 541 cmdlen = 0; 542 543 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 544 cmdlen = F_LEN(cmd); 545 546 switch (cmd->opcode) { 547 /* Opcodes supporting tablearg */ 548 case O_TAG: 549 case O_TAGGED: 550 case O_PIPE: 551 case O_QUEUE: 552 case O_DIVERT: 553 case O_TEE: 554 case O_SKIPTO: 555 case O_CALLRETURN: 556 case O_NETGRAPH: 557 case O_NGTEE: 558 case O_NAT: 559 if (cmd->arg1 == IP_FW_TABLEARG) 560 cmd->arg1 = IP_FW_TARG; 561 else if (cmd->arg1 == 0) 562 cmd->arg1 = IP_FW_NAT44_GLOBAL; 563 break; 564 case O_SETFIB: 565 case O_SETDSCP: 566 if (cmd->arg1 == IP_FW_TABLEARG) 567 cmd->arg1 = IP_FW_TARG; 568 else 569 cmd->arg1 |= 0x8000; 570 break; 571 case O_LIMIT: 572 lcmd = (ipfw_insn_limit *)cmd; 573 if (lcmd->conn_limit == IP_FW_TABLEARG) 574 lcmd->conn_limit = IP_FW_TARG; 575 break; 576 /* Interface tables */ 577 case O_XMIT: 578 case O_RECV: 579 case O_VIA: 580 /* Interface table, possibly */ 581 cmdif = (ipfw_insn_if *)cmd; 582 if (cmdif->name[0] != '\1') 583 break; 584 585 cmdif->p.kidx = (uint16_t)cmdif->p.glob; 586 break; 587 } 588 } 589} 590 591/* 592 * Copies rule @krule from kernel to FreeBSD8 userland format (v0) 593 */ 594static void 595export_rule0(struct ip_fw *krule, struct ip_fw_rule0 *urule, int len) 596{ 597 int cmdlen, l; 598 ipfw_insn *cmd; 599 ipfw_insn_limit *lcmd; 600 ipfw_insn_if *cmdif; 601 602 /* copy header */ 603 memset(urule, 0, len); 604 urule->act_ofs = krule->act_ofs; 605 urule->cmd_len = krule->cmd_len; 606 urule->rulenum = krule->rulenum; 607 urule->set = krule->set; 608 if ((krule->flags & IPFW_RULE_NOOPT) != 0) 609 urule->_pad |= 1; 610 611 /* Copy opcodes */ 612 memcpy(urule->cmd, krule->cmd, krule->cmd_len * sizeof(uint32_t)); 613 614 /* Export counters */ 615 export_cntr0_base(krule, (struct ip_fw_bcounter0 *)&urule->pcnt); 616 617 /* 618 * Alter opcodes: 619 * 1) convert tablearg value from 0 to 65535 620 * 2) Remove highest bit from O_SETFIB/O_SETDSCP values. 621 * 3) convert table number in iface opcodes to int 622 */ 623 l = urule->cmd_len; 624 cmd = urule->cmd; 625 cmdlen = 0; 626 627 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 628 cmdlen = F_LEN(cmd); 629 630 switch (cmd->opcode) { 631 /* Opcodes supporting tablearg */ 632 case O_TAG: 633 case O_TAGGED: 634 case O_PIPE: 635 case O_QUEUE: 636 case O_DIVERT: 637 case O_TEE: 638 case O_SKIPTO: 639 case O_CALLRETURN: 640 case O_NETGRAPH: 641 case O_NGTEE: 642 case O_NAT: 643 if (cmd->arg1 == IP_FW_TARG) 644 cmd->arg1 = IP_FW_TABLEARG; 645 else if (cmd->arg1 == IP_FW_NAT44_GLOBAL) 646 cmd->arg1 = 0; 647 break; 648 case O_SETFIB: 649 case O_SETDSCP: 650 if (cmd->arg1 == IP_FW_TARG) 651 cmd->arg1 = IP_FW_TABLEARG; 652 else 653 cmd->arg1 &= ~0x8000; 654 break; 655 case O_LIMIT: 656 lcmd = (ipfw_insn_limit *)cmd; 657 if (lcmd->conn_limit == IP_FW_TARG) 658 lcmd->conn_limit = IP_FW_TABLEARG; 659 break; 660 /* Interface tables */ 661 case O_XMIT: 662 case O_RECV: 663 case O_VIA: 664 /* Interface table, possibly */ 665 cmdif = (ipfw_insn_if *)cmd; 666 if (cmdif->name[0] != '\1') 667 break; 668 669 cmdif->p.glob = cmdif->p.kidx; 670 break; 671 } 672 } 673} 674 675/* 676 * Add new rule(s) to the list possibly creating rule number for each. 677 * Update the rule_number in the input struct so the caller knows it as well. 678 * Must be called without IPFW_UH held 679 */ 680static int 681commit_rules(struct ip_fw_chain *chain, struct rule_check_info *rci, int count) 682{ 683 int error, i, insert_before, tcount; 684 uint16_t rulenum, *pnum; 685 struct rule_check_info *ci; 686 struct ip_fw *krule; 687 struct ip_fw **map; /* the new array of pointers */ 688 689 /* Check if we need to do table/obj index remap */ 690 tcount = 0; 691 for (ci = rci, i = 0; i < count; ci++, i++) { 692 if (ci->object_opcodes == 0) 693 continue; 694 695 /* 696 * Rule has some object opcodes. 697 * We need to find (and create non-existing) 698 * kernel objects, and reference existing ones. 699 */ 700 error = rewrite_rule_uidx(chain, ci); 701 if (error != 0) { 702 703 /* 704 * rewrite failed, state for current rule 705 * has been reverted. Check if we need to 706 * revert more. 707 */ 708 if (tcount > 0) { 709 710 /* 711 * We have some more table rules 712 * we need to rollback. 713 */ 714 715 IPFW_UH_WLOCK(chain); 716 while (ci != rci) { 717 ci--; 718 if (ci->object_opcodes == 0) 719 continue; 720 unref_rule_objects(chain,ci->krule); 721 722 } 723 IPFW_UH_WUNLOCK(chain); 724 725 } 726 727 return (error); 728 } 729 730 tcount++; 731 } 732 733 /* get_map returns with IPFW_UH_WLOCK if successful */ 734 map = get_map(chain, count, 0 /* not locked */); 735 if (map == NULL) { 736 if (tcount > 0) { 737 /* Unbind tables */ 738 IPFW_UH_WLOCK(chain); 739 for (ci = rci, i = 0; i < count; ci++, i++) { 740 if (ci->object_opcodes == 0) 741 continue; 742 743 unref_rule_objects(chain, ci->krule); 744 } 745 IPFW_UH_WUNLOCK(chain); 746 } 747 748 return (ENOSPC); 749 } 750 751 if (V_autoinc_step < 1) 752 V_autoinc_step = 1; 753 else if (V_autoinc_step > 1000) 754 V_autoinc_step = 1000; 755 756 /* FIXME: Handle count > 1 */ 757 ci = rci; 758 krule = ci->krule; 759 rulenum = krule->rulenum; 760 761 /* find the insertion point, we will insert before */ 762 insert_before = rulenum ? rulenum + 1 : IPFW_DEFAULT_RULE; 763 i = ipfw_find_rule(chain, insert_before, 0); 764 /* duplicate first part */ 765 if (i > 0) 766 bcopy(chain->map, map, i * sizeof(struct ip_fw *)); 767 map[i] = krule; 768 /* duplicate remaining part, we always have the default rule */ 769 bcopy(chain->map + i, map + i + 1, 770 sizeof(struct ip_fw *) *(chain->n_rules - i)); 771 if (rulenum == 0) { 772 /* Compute rule number and write it back */ 773 rulenum = i > 0 ? map[i-1]->rulenum : 0; 774 if (rulenum < IPFW_DEFAULT_RULE - V_autoinc_step) 775 rulenum += V_autoinc_step; 776 krule->rulenum = rulenum; 777 /* Save number to userland rule */ 778 pnum = (uint16_t *)((caddr_t)ci->urule + ci->urule_numoff); 779 *pnum = rulenum; 780 } 781 782 krule->id = chain->id + 1; 783 update_skipto_cache(chain, map); 784 map = swap_map(chain, map, chain->n_rules + 1); 785 chain->static_len += RULEUSIZE0(krule); 786 IPFW_UH_WUNLOCK(chain); 787 if (map) 788 free(map, M_IPFW); 789 return (0); 790} 791 792/* 793 * Adds @rule to the list of rules to reap 794 */ 795void 796ipfw_reap_add(struct ip_fw_chain *chain, struct ip_fw **head, 797 struct ip_fw *rule) 798{ 799 800 IPFW_UH_WLOCK_ASSERT(chain); 801 802 /* Unlink rule from everywhere */ 803 unref_rule_objects(chain, rule); 804 805 *((struct ip_fw **)rule) = *head; 806 *head = rule; 807} 808 809/* 810 * Reclaim storage associated with a list of rules. This is 811 * typically the list created using remove_rule. 812 * A NULL pointer on input is handled correctly. 813 */ 814void 815ipfw_reap_rules(struct ip_fw *head) 816{ 817 struct ip_fw *rule; 818 819 while ((rule = head) != NULL) { 820 head = *((struct ip_fw **)head); 821 free_rule(rule); 822 } 823} 824 825/* 826 * Rules to keep are 827 * (default || reserved || !match_set || !match_number) 828 * where 829 * default ::= (rule->rulenum == IPFW_DEFAULT_RULE) 830 * // the default rule is always protected 831 * 832 * reserved ::= (cmd == 0 && n == 0 && rule->set == RESVD_SET) 833 * // RESVD_SET is protected only if cmd == 0 and n == 0 ("ipfw flush") 834 * 835 * match_set ::= (cmd == 0 || rule->set == set) 836 * // set number is ignored for cmd == 0 837 * 838 * match_number ::= (cmd == 1 || n == 0 || n == rule->rulenum) 839 * // number is ignored for cmd == 1 or n == 0 840 * 841 */ 842int 843ipfw_match_range(struct ip_fw *rule, ipfw_range_tlv *rt) 844{ 845 846 /* Don't match default rule for modification queries */ 847 if (rule->rulenum == IPFW_DEFAULT_RULE && 848 (rt->flags & IPFW_RCFLAG_DEFAULT) == 0) 849 return (0); 850 851 /* Don't match rules in reserved set for flush requests */ 852 if ((rt->flags & IPFW_RCFLAG_ALL) != 0 && rule->set == RESVD_SET) 853 return (0); 854 855 /* If we're filtering by set, don't match other sets */ 856 if ((rt->flags & IPFW_RCFLAG_SET) != 0 && rule->set != rt->set) 857 return (0); 858 859 if ((rt->flags & IPFW_RCFLAG_RANGE) != 0 && 860 (rule->rulenum < rt->start_rule || rule->rulenum > rt->end_rule)) 861 return (0); 862 863 return (1); 864} 865 866struct manage_sets_args { 867 uint16_t set; 868 uint8_t new_set; 869}; 870 871static int 872swap_sets_cb(struct namedobj_instance *ni, struct named_object *no, 873 void *arg) 874{ 875 struct manage_sets_args *args; 876 877 args = (struct manage_sets_args *)arg; 878 if (no->set == (uint8_t)args->set) 879 no->set = args->new_set; 880 else if (no->set == args->new_set) 881 no->set = (uint8_t)args->set; 882 return (0); 883} 884 885static int 886move_sets_cb(struct namedobj_instance *ni, struct named_object *no, 887 void *arg) 888{ 889 struct manage_sets_args *args; 890 891 args = (struct manage_sets_args *)arg; 892 if (no->set == (uint8_t)args->set) 893 no->set = args->new_set; 894 return (0); 895} 896 897static int 898test_sets_cb(struct namedobj_instance *ni, struct named_object *no, 899 void *arg) 900{ 901 struct manage_sets_args *args; 902 903 args = (struct manage_sets_args *)arg; 904 if (no->set != (uint8_t)args->set) 905 return (0); 906 if (ipfw_objhash_lookup_name_type(ni, args->new_set, 907 no->etlv, no->name) != NULL) 908 return (EEXIST); 909 return (0); 910} 911 912/* 913 * Generic function to handler moving and swapping sets. 914 */ 915int 916ipfw_obj_manage_sets(struct namedobj_instance *ni, uint16_t type, 917 uint16_t set, uint8_t new_set, enum ipfw_sets_cmd cmd) 918{ 919 struct manage_sets_args args; 920 struct named_object *no; 921 922 args.set = set; 923 args.new_set = new_set; 924 switch (cmd) { 925 case SWAP_ALL: 926 return (ipfw_objhash_foreach_type(ni, swap_sets_cb, 927 &args, type)); 928 case TEST_ALL: 929 return (ipfw_objhash_foreach_type(ni, test_sets_cb, 930 &args, type)); 931 case MOVE_ALL: 932 return (ipfw_objhash_foreach_type(ni, move_sets_cb, 933 &args, type)); 934 case COUNT_ONE: 935 /* 936 * @set used to pass kidx. 937 * When @new_set is zero - reset object counter, 938 * otherwise increment it. 939 */ 940 no = ipfw_objhash_lookup_kidx(ni, set); 941 if (new_set != 0) 942 no->ocnt++; 943 else 944 no->ocnt = 0; 945 return (0); 946 case TEST_ONE: 947 /* @set used to pass kidx */ 948 no = ipfw_objhash_lookup_kidx(ni, set); 949 /* 950 * First check number of references: 951 * when it differs, this mean other rules are holding 952 * reference to given object, so it is not possible to 953 * change its set. Note that refcnt may account references 954 * to some going-to-be-added rules. Since we don't know 955 * their numbers (and even if they will be added) it is 956 * perfectly OK to return error here. 957 */ 958 if (no->ocnt != no->refcnt) 959 return (EBUSY); 960 if (ipfw_objhash_lookup_name_type(ni, new_set, type, 961 no->name) != NULL) 962 return (EEXIST); 963 return (0); 964 case MOVE_ONE: 965 /* @set used to pass kidx */ 966 no = ipfw_objhash_lookup_kidx(ni, set); 967 no->set = new_set; 968 return (0); 969 } 970 return (EINVAL); 971} 972 973/* 974 * Delete rules matching range @rt. 975 * Saves number of deleted rules in @ndel. 976 * 977 * Returns 0 on success. 978 */ 979static int 980delete_range(struct ip_fw_chain *chain, ipfw_range_tlv *rt, int *ndel) 981{ 982 struct ip_fw *reap, *rule, **map; 983 int end, start; 984 int i, n, ndyn, ofs; 985 986 reap = NULL; 987 IPFW_UH_WLOCK(chain); /* arbitrate writers */ 988 989 /* 990 * Stage 1: Determine range to inspect. 991 * Range is half-inclusive, e.g [start, end). 992 */ 993 start = 0; 994 end = chain->n_rules - 1; 995 996 if ((rt->flags & IPFW_RCFLAG_RANGE) != 0) { 997 start = ipfw_find_rule(chain, rt->start_rule, 0); 998 999 end = ipfw_find_rule(chain, rt->end_rule, 0); 1000 if (rt->end_rule != IPFW_DEFAULT_RULE) 1001 while (chain->map[end]->rulenum == rt->end_rule) 1002 end++; 1003 } 1004 1005 /* Allocate new map of the same size */ 1006 map = get_map(chain, 0, 1 /* locked */); 1007 if (map == NULL) { 1008 IPFW_UH_WUNLOCK(chain); 1009 return (ENOMEM); 1010 } 1011 1012 n = 0; 1013 ndyn = 0; 1014 ofs = start; 1015 /* 1. bcopy the initial part of the map */ 1016 if (start > 0) 1017 bcopy(chain->map, map, start * sizeof(struct ip_fw *)); 1018 /* 2. copy active rules between start and end */ 1019 for (i = start; i < end; i++) { 1020 rule = chain->map[i]; 1021 if (ipfw_match_range(rule, rt) == 0) { 1022 map[ofs++] = rule; 1023 continue; 1024 } 1025 1026 n++; 1027 if (ipfw_is_dyn_rule(rule) != 0) 1028 ndyn++; 1029 } 1030 /* 3. copy the final part of the map */ 1031 bcopy(chain->map + end, map + ofs, 1032 (chain->n_rules - end) * sizeof(struct ip_fw *)); 1033 /* 4. recalculate skipto cache */ 1034 update_skipto_cache(chain, map); 1035 /* 5. swap the maps (under UH_WLOCK + WHLOCK) */ 1036 map = swap_map(chain, map, chain->n_rules - n); 1037 /* 6. Remove all dynamic states originated by deleted rules */ 1038 if (ndyn > 0) 1039 ipfw_expire_dyn_rules(chain, rt); 1040 /* 7. now remove the rules deleted from the old map */ 1041 for (i = start; i < end; i++) { 1042 rule = map[i]; 1043 if (ipfw_match_range(rule, rt) == 0) 1044 continue; 1045 chain->static_len -= RULEUSIZE0(rule); 1046 ipfw_reap_add(chain, &reap, rule); 1047 } 1048 IPFW_UH_WUNLOCK(chain); 1049 1050 ipfw_reap_rules(reap); 1051 if (map != NULL) 1052 free(map, M_IPFW); 1053 *ndel = n; 1054 return (0); 1055} 1056 1057static int 1058move_objects(struct ip_fw_chain *ch, ipfw_range_tlv *rt) 1059{ 1060 struct opcode_obj_rewrite *rw; 1061 struct ip_fw *rule; 1062 ipfw_insn *cmd; 1063 int cmdlen, i, l, c; 1064 uint16_t kidx; 1065 1066 IPFW_UH_WLOCK_ASSERT(ch); 1067 1068 /* Stage 1: count number of references by given rules */ 1069 for (c = 0, i = 0; i < ch->n_rules - 1; i++) { 1070 rule = ch->map[i]; 1071 if (ipfw_match_range(rule, rt) == 0) 1072 continue; 1073 if (rule->set == rt->new_set) /* nothing to do */ 1074 continue; 1075 /* Search opcodes with named objects */ 1076 for (l = rule->cmd_len, cmdlen = 0, cmd = rule->cmd; 1077 l > 0; l -= cmdlen, cmd += cmdlen) { 1078 cmdlen = F_LEN(cmd); 1079 rw = find_op_rw(cmd, &kidx, NULL); 1080 if (rw == NULL || rw->manage_sets == NULL) 1081 continue; 1082 /* 1083 * When manage_sets() returns non-zero value to 1084 * COUNT_ONE command, consider this as an object 1085 * doesn't support sets (e.g. disabled with sysctl). 1086 * So, skip checks for this object. 1087 */ 1088 if (rw->manage_sets(ch, kidx, 1, COUNT_ONE) != 0) 1089 continue; 1090 c++; 1091 } 1092 } 1093 if (c == 0) /* No objects found */ 1094 return (0); 1095 /* Stage 2: verify "ownership" */ 1096 for (c = 0, i = 0; (i < ch->n_rules - 1) && c == 0; i++) { 1097 rule = ch->map[i]; 1098 if (ipfw_match_range(rule, rt) == 0) 1099 continue; 1100 if (rule->set == rt->new_set) /* nothing to do */ 1101 continue; 1102 /* Search opcodes with named objects */ 1103 for (l = rule->cmd_len, cmdlen = 0, cmd = rule->cmd; 1104 l > 0 && c == 0; l -= cmdlen, cmd += cmdlen) { 1105 cmdlen = F_LEN(cmd); 1106 rw = find_op_rw(cmd, &kidx, NULL); 1107 if (rw == NULL || rw->manage_sets == NULL) 1108 continue; 1109 /* Test for ownership and conflicting names */ 1110 c = rw->manage_sets(ch, kidx, 1111 (uint8_t)rt->new_set, TEST_ONE); 1112 } 1113 } 1114 /* Stage 3: change set and cleanup */ 1115 for (i = 0; i < ch->n_rules - 1; i++) { 1116 rule = ch->map[i]; 1117 if (ipfw_match_range(rule, rt) == 0) 1118 continue; 1119 if (rule->set == rt->new_set) /* nothing to do */ 1120 continue; 1121 /* Search opcodes with named objects */ 1122 for (l = rule->cmd_len, cmdlen = 0, cmd = rule->cmd; 1123 l > 0; l -= cmdlen, cmd += cmdlen) { 1124 cmdlen = F_LEN(cmd); 1125 rw = find_op_rw(cmd, &kidx, NULL); 1126 if (rw == NULL || rw->manage_sets == NULL) 1127 continue; 1128 /* cleanup object counter */ 1129 rw->manage_sets(ch, kidx, 1130 0 /* reset counter */, COUNT_ONE); 1131 if (c != 0) 1132 continue; 1133 /* change set */ 1134 rw->manage_sets(ch, kidx, 1135 (uint8_t)rt->new_set, MOVE_ONE); 1136 } 1137 } 1138 return (c); 1139}/* 1140 * Changes set of given rule rannge @rt 1141 * with each other. 1142 * 1143 * Returns 0 on success. 1144 */ 1145static int 1146move_range(struct ip_fw_chain *chain, ipfw_range_tlv *rt) 1147{ 1148 struct ip_fw *rule; 1149 int i; 1150 1151 IPFW_UH_WLOCK(chain); 1152 1153 /* 1154 * Move rules with matching paramenerts to a new set. 1155 * This one is much more complex. We have to ensure 1156 * that all referenced tables (if any) are referenced 1157 * by given rule subset only. Otherwise, we can't move 1158 * them to new set and have to return error. 1159 */ 1160 if ((i = move_objects(chain, rt)) != 0) { 1161 IPFW_UH_WUNLOCK(chain); 1162 return (i); 1163 } 1164 1165 /* XXX: We have to do swap holding WLOCK */ 1166 for (i = 0; i < chain->n_rules; i++) { 1167 rule = chain->map[i]; 1168 if (ipfw_match_range(rule, rt) == 0) 1169 continue; 1170 rule->set = rt->new_set; 1171 } 1172 1173 IPFW_UH_WUNLOCK(chain); 1174 1175 return (0); 1176} 1177 1178/* 1179 * Clear counters for a specific rule. 1180 * Normally run under IPFW_UH_RLOCK, but these are idempotent ops 1181 * so we only care that rules do not disappear. 1182 */ 1183static void 1184clear_counters(struct ip_fw *rule, int log_only) 1185{ 1186 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 1187 1188 if (log_only == 0) 1189 IPFW_ZERO_RULE_COUNTER(rule); 1190 if (l->o.opcode == O_LOG) 1191 l->log_left = l->max_log; 1192} 1193 1194/* 1195 * Flushes rules counters and/or log values on matching range. 1196 * 1197 * Returns number of items cleared. 1198 */ 1199static int 1200clear_range(struct ip_fw_chain *chain, ipfw_range_tlv *rt, int log_only) 1201{ 1202 struct ip_fw *rule; 1203 int num; 1204 int i; 1205 1206 num = 0; 1207 rt->flags |= IPFW_RCFLAG_DEFAULT; 1208 1209 IPFW_UH_WLOCK(chain); /* arbitrate writers */ 1210 for (i = 0; i < chain->n_rules; i++) { 1211 rule = chain->map[i]; 1212 if (ipfw_match_range(rule, rt) == 0) 1213 continue; 1214 clear_counters(rule, log_only); 1215 num++; 1216 } 1217 IPFW_UH_WUNLOCK(chain); 1218 1219 return (num); 1220} 1221 1222static int 1223check_range_tlv(ipfw_range_tlv *rt) 1224{ 1225 1226 if (rt->head.length != sizeof(*rt)) 1227 return (1); 1228 if (rt->start_rule > rt->end_rule) 1229 return (1); 1230 if (rt->set >= IPFW_MAX_SETS || rt->new_set >= IPFW_MAX_SETS) 1231 return (1); 1232 1233 if ((rt->flags & IPFW_RCFLAG_USER) != rt->flags) 1234 return (1); 1235 1236 return (0); 1237} 1238 1239/* 1240 * Delete rules matching specified parameters 1241 * Data layout (v0)(current): 1242 * Request: [ ipfw_obj_header ipfw_range_tlv ] 1243 * Reply: [ ipfw_obj_header ipfw_range_tlv ] 1244 * 1245 * Saves number of deleted rules in ipfw_range_tlv->new_set. 1246 * 1247 * Returns 0 on success. 1248 */ 1249static int 1250del_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 1251 struct sockopt_data *sd) 1252{ 1253 ipfw_range_header *rh; 1254 int error, ndel; 1255 1256 if (sd->valsize != sizeof(*rh)) 1257 return (EINVAL); 1258 1259 rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize); 1260 1261 if (check_range_tlv(&rh->range) != 0) 1262 return (EINVAL); 1263 1264 ndel = 0; 1265 if ((error = delete_range(chain, &rh->range, &ndel)) != 0) 1266 return (error); 1267 1268 /* Save number of rules deleted */ 1269 rh->range.new_set = ndel; 1270 return (0); 1271} 1272 1273/* 1274 * Move rules/sets matching specified parameters 1275 * Data layout (v0)(current): 1276 * Request: [ ipfw_obj_header ipfw_range_tlv ] 1277 * 1278 * Returns 0 on success. 1279 */ 1280static int 1281move_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 1282 struct sockopt_data *sd) 1283{ 1284 ipfw_range_header *rh; 1285 1286 if (sd->valsize != sizeof(*rh)) 1287 return (EINVAL); 1288 1289 rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize); 1290 1291 if (check_range_tlv(&rh->range) != 0) 1292 return (EINVAL); 1293 1294 return (move_range(chain, &rh->range)); 1295} 1296 1297/* 1298 * Clear rule accounting data matching specified parameters 1299 * Data layout (v0)(current): 1300 * Request: [ ipfw_obj_header ipfw_range_tlv ] 1301 * Reply: [ ipfw_obj_header ipfw_range_tlv ] 1302 * 1303 * Saves number of cleared rules in ipfw_range_tlv->new_set. 1304 * 1305 * Returns 0 on success. 1306 */ 1307static int 1308clear_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 1309 struct sockopt_data *sd) 1310{ 1311 ipfw_range_header *rh; 1312 int log_only, num; 1313 char *msg; 1314 1315 if (sd->valsize != sizeof(*rh)) 1316 return (EINVAL); 1317 1318 rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize); 1319 1320 if (check_range_tlv(&rh->range) != 0) 1321 return (EINVAL); 1322 1323 log_only = (op3->opcode == IP_FW_XRESETLOG); 1324 1325 num = clear_range(chain, &rh->range, log_only); 1326 1327 if (rh->range.flags & IPFW_RCFLAG_ALL) 1328 msg = log_only ? "All logging counts reset" : 1329 "Accounting cleared"; 1330 else 1331 msg = log_only ? "logging count reset" : "cleared"; 1332 1333 if (V_fw_verbose) { 1334 int lev = LOG_SECURITY | LOG_NOTICE; 1335 log(lev, "ipfw: %s.\n", msg); 1336 } 1337 1338 /* Save number of rules cleared */ 1339 rh->range.new_set = num; 1340 return (0); 1341} 1342 1343static void 1344enable_sets(struct ip_fw_chain *chain, ipfw_range_tlv *rt) 1345{ 1346 uint32_t v_set; 1347 1348 IPFW_UH_WLOCK_ASSERT(chain); 1349 1350 /* Change enabled/disabled sets mask */ 1351 v_set = (V_set_disable | rt->set) & ~rt->new_set; 1352 v_set &= ~(1 << RESVD_SET); /* set RESVD_SET always enabled */ 1353 IPFW_WLOCK(chain); 1354 V_set_disable = v_set; 1355 IPFW_WUNLOCK(chain); 1356} 1357 1358static int 1359swap_sets(struct ip_fw_chain *chain, ipfw_range_tlv *rt, int mv) 1360{ 1361 struct opcode_obj_rewrite *rw; 1362 struct ip_fw *rule; 1363 int i; 1364 1365 IPFW_UH_WLOCK_ASSERT(chain); 1366 1367 if (rt->set == rt->new_set) /* nothing to do */ 1368 return (0); 1369 1370 if (mv != 0) { 1371 /* 1372 * Berfore moving the rules we need to check that 1373 * there aren't any conflicting named objects. 1374 */ 1375 for (rw = ctl3_rewriters; 1376 rw < ctl3_rewriters + ctl3_rsize; rw++) { 1377 if (rw->manage_sets == NULL) 1378 continue; 1379 i = rw->manage_sets(chain, (uint8_t)rt->set, 1380 (uint8_t)rt->new_set, TEST_ALL); 1381 if (i != 0) 1382 return (EEXIST); 1383 } 1384 } 1385 /* Swap or move two sets */ 1386 for (i = 0; i < chain->n_rules - 1; i++) { 1387 rule = chain->map[i]; 1388 if (rule->set == (uint8_t)rt->set) 1389 rule->set = (uint8_t)rt->new_set; 1390 else if (rule->set == (uint8_t)rt->new_set && mv == 0) 1391 rule->set = (uint8_t)rt->set; 1392 } 1393 for (rw = ctl3_rewriters; rw < ctl3_rewriters + ctl3_rsize; rw++) { 1394 if (rw->manage_sets == NULL) 1395 continue; 1396 rw->manage_sets(chain, (uint8_t)rt->set, 1397 (uint8_t)rt->new_set, mv != 0 ? MOVE_ALL: SWAP_ALL); 1398 } 1399 return (0); 1400} 1401 1402/* 1403 * Swaps or moves set 1404 * Data layout (v0)(current): 1405 * Request: [ ipfw_obj_header ipfw_range_tlv ] 1406 * 1407 * Returns 0 on success. 1408 */ 1409static int 1410manage_sets(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 1411 struct sockopt_data *sd) 1412{ 1413 ipfw_range_header *rh; 1414 int ret; 1415 1416 if (sd->valsize != sizeof(*rh)) 1417 return (EINVAL); 1418 1419 rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize); 1420 1421 if (rh->range.head.length != sizeof(ipfw_range_tlv)) 1422 return (1); 1423 /* enable_sets() expects bitmasks. */ 1424 if (op3->opcode != IP_FW_SET_ENABLE && 1425 (rh->range.set >= IPFW_MAX_SETS || 1426 rh->range.new_set >= IPFW_MAX_SETS)) 1427 return (EINVAL); 1428 1429 ret = 0; 1430 IPFW_UH_WLOCK(chain); 1431 switch (op3->opcode) { 1432 case IP_FW_SET_SWAP: 1433 case IP_FW_SET_MOVE: 1434 ret = swap_sets(chain, &rh->range, 1435 op3->opcode == IP_FW_SET_MOVE); 1436 break; 1437 case IP_FW_SET_ENABLE: 1438 enable_sets(chain, &rh->range); 1439 break; 1440 } 1441 IPFW_UH_WUNLOCK(chain); 1442 1443 return (ret); 1444} 1445 1446/** 1447 * Remove all rules with given number, or do set manipulation. 1448 * Assumes chain != NULL && *chain != NULL. 1449 * 1450 * The argument is an uint32_t. The low 16 bit are the rule or set number; 1451 * the next 8 bits are the new set; the top 8 bits indicate the command: 1452 * 1453 * 0 delete rules numbered "rulenum" 1454 * 1 delete rules in set "rulenum" 1455 * 2 move rules "rulenum" to set "new_set" 1456 * 3 move rules from set "rulenum" to set "new_set" 1457 * 4 swap sets "rulenum" and "new_set" 1458 * 5 delete rules "rulenum" and set "new_set" 1459 */ 1460static int 1461del_entry(struct ip_fw_chain *chain, uint32_t arg) 1462{ 1463 uint32_t num; /* rule number or old_set */ 1464 uint8_t cmd, new_set; 1465 int do_del, ndel; 1466 int error = 0; 1467 ipfw_range_tlv rt; 1468 1469 num = arg & 0xffff; 1470 cmd = (arg >> 24) & 0xff; 1471 new_set = (arg >> 16) & 0xff; 1472 1473 if (cmd > 5 || new_set > RESVD_SET) 1474 return EINVAL; 1475 if (cmd == 0 || cmd == 2 || cmd == 5) { 1476 if (num >= IPFW_DEFAULT_RULE) 1477 return EINVAL; 1478 } else { 1479 if (num > RESVD_SET) /* old_set */ 1480 return EINVAL; 1481 } 1482 1483 /* Convert old requests into new representation */ 1484 memset(&rt, 0, sizeof(rt)); 1485 rt.start_rule = num; 1486 rt.end_rule = num; 1487 rt.set = num; 1488 rt.new_set = new_set; 1489 do_del = 0; 1490 1491 switch (cmd) { 1492 case 0: /* delete rules numbered "rulenum" */ 1493 if (num == 0) 1494 rt.flags |= IPFW_RCFLAG_ALL; 1495 else 1496 rt.flags |= IPFW_RCFLAG_RANGE; 1497 do_del = 1; 1498 break; 1499 case 1: /* delete rules in set "rulenum" */ 1500 rt.flags |= IPFW_RCFLAG_SET; 1501 do_del = 1; 1502 break; 1503 case 5: /* delete rules "rulenum" and set "new_set" */ 1504 rt.flags |= IPFW_RCFLAG_RANGE | IPFW_RCFLAG_SET; 1505 rt.set = new_set; 1506 rt.new_set = 0; 1507 do_del = 1; 1508 break; 1509 case 2: /* move rules "rulenum" to set "new_set" */ 1510 rt.flags |= IPFW_RCFLAG_RANGE; 1511 break; 1512 case 3: /* move rules from set "rulenum" to set "new_set" */ 1513 IPFW_UH_WLOCK(chain); 1514 error = swap_sets(chain, &rt, 1); 1515 IPFW_UH_WUNLOCK(chain); 1516 return (error); 1517 case 4: /* swap sets "rulenum" and "new_set" */ 1518 IPFW_UH_WLOCK(chain); 1519 error = swap_sets(chain, &rt, 0); 1520 IPFW_UH_WUNLOCK(chain); 1521 return (error); 1522 default: 1523 return (ENOTSUP); 1524 } 1525 1526 if (do_del != 0) { 1527 if ((error = delete_range(chain, &rt, &ndel)) != 0) 1528 return (error); 1529 1530 if (ndel == 0 && (cmd != 1 && num != 0)) 1531 return (EINVAL); 1532 1533 return (0); 1534 } 1535 1536 return (move_range(chain, &rt)); 1537} 1538 1539/** 1540 * Reset some or all counters on firewall rules. 1541 * The argument `arg' is an u_int32_t. The low 16 bit are the rule number, 1542 * the next 8 bits are the set number, the top 8 bits are the command: 1543 * 0 work with rules from all set's; 1544 * 1 work with rules only from specified set. 1545 * Specified rule number is zero if we want to clear all entries. 1546 * log_only is 1 if we only want to reset logs, zero otherwise. 1547 */ 1548static int 1549zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only) 1550{ 1551 struct ip_fw *rule; 1552 char *msg; 1553 int i; 1554 1555 uint16_t rulenum = arg & 0xffff; 1556 uint8_t set = (arg >> 16) & 0xff; 1557 uint8_t cmd = (arg >> 24) & 0xff; 1558 1559 if (cmd > 1) 1560 return (EINVAL); 1561 if (cmd == 1 && set > RESVD_SET) 1562 return (EINVAL); 1563 1564 IPFW_UH_RLOCK(chain); 1565 if (rulenum == 0) { 1566 V_norule_counter = 0; 1567 for (i = 0; i < chain->n_rules; i++) { 1568 rule = chain->map[i]; 1569 /* Skip rules not in our set. */ 1570 if (cmd == 1 && rule->set != set) 1571 continue; 1572 clear_counters(rule, log_only); 1573 } 1574 msg = log_only ? "All logging counts reset" : 1575 "Accounting cleared"; 1576 } else { 1577 int cleared = 0; 1578 for (i = 0; i < chain->n_rules; i++) { 1579 rule = chain->map[i]; 1580 if (rule->rulenum == rulenum) { 1581 if (cmd == 0 || rule->set == set) 1582 clear_counters(rule, log_only); 1583 cleared = 1; 1584 } 1585 if (rule->rulenum > rulenum) 1586 break; 1587 } 1588 if (!cleared) { /* we did not find any matching rules */ 1589 IPFW_UH_RUNLOCK(chain); 1590 return (EINVAL); 1591 } 1592 msg = log_only ? "logging count reset" : "cleared"; 1593 } 1594 IPFW_UH_RUNLOCK(chain); 1595 1596 if (V_fw_verbose) { 1597 int lev = LOG_SECURITY | LOG_NOTICE; 1598 1599 if (rulenum) 1600 log(lev, "ipfw: Entry %d %s.\n", rulenum, msg); 1601 else 1602 log(lev, "ipfw: %s.\n", msg); 1603 } 1604 return (0); 1605} 1606 1607 1608/* 1609 * Check rule head in FreeBSD11 format 1610 * 1611 */ 1612static int 1613check_ipfw_rule1(struct ip_fw_rule *rule, int size, 1614 struct rule_check_info *ci) 1615{ 1616 int l; 1617 1618 if (size < sizeof(*rule)) { 1619 printf("ipfw: rule too short\n"); 1620 return (EINVAL); 1621 } 1622 1623 /* Check for valid cmd_len */ 1624 l = roundup2(RULESIZE(rule), sizeof(uint64_t)); 1625 if (l != size) { 1626 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 1627 return (EINVAL); 1628 } 1629 if (rule->act_ofs >= rule->cmd_len) { 1630 printf("ipfw: bogus action offset (%u > %u)\n", 1631 rule->act_ofs, rule->cmd_len - 1); 1632 return (EINVAL); 1633 } 1634 1635 if (rule->rulenum > IPFW_DEFAULT_RULE - 1) 1636 return (EINVAL); 1637 1638 return (check_ipfw_rule_body(rule->cmd, rule->cmd_len, ci)); 1639} 1640 1641/* 1642 * Check rule head in FreeBSD8 format 1643 * 1644 */ 1645static int 1646check_ipfw_rule0(struct ip_fw_rule0 *rule, int size, 1647 struct rule_check_info *ci) 1648{ 1649 int l; 1650 1651 if (size < sizeof(*rule)) { 1652 printf("ipfw: rule too short\n"); 1653 return (EINVAL); 1654 } 1655 1656 /* Check for valid cmd_len */ 1657 l = sizeof(*rule) + rule->cmd_len * 4 - 4; 1658 if (l != size) { 1659 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 1660 return (EINVAL); 1661 } 1662 if (rule->act_ofs >= rule->cmd_len) { 1663 printf("ipfw: bogus action offset (%u > %u)\n", 1664 rule->act_ofs, rule->cmd_len - 1); 1665 return (EINVAL); 1666 } 1667 1668 if (rule->rulenum > IPFW_DEFAULT_RULE - 1) 1669 return (EINVAL); 1670 1671 return (check_ipfw_rule_body(rule->cmd, rule->cmd_len, ci)); 1672} 1673 1674static int 1675check_ipfw_rule_body(ipfw_insn *cmd, int cmd_len, struct rule_check_info *ci) 1676{ 1677 int cmdlen, l; 1678 int have_action; 1679 1680 have_action = 0; 1681 1682 /* 1683 * Now go for the individual checks. Very simple ones, basically only 1684 * instruction sizes. 1685 */ 1686 for (l = cmd_len; l > 0 ; l -= cmdlen, cmd += cmdlen) { 1687 cmdlen = F_LEN(cmd); 1688 if (cmdlen > l) { 1689 printf("ipfw: opcode %d size truncated\n", 1690 cmd->opcode); 1691 return EINVAL; 1692 } 1693 switch (cmd->opcode) { 1694 case O_PROBE_STATE: 1695 case O_KEEP_STATE: 1696 case O_PROTO: 1697 case O_IP_SRC_ME: 1698 case O_IP_DST_ME: 1699 case O_LAYER2: 1700 case O_IN: 1701 case O_FRAG: 1702 case O_DIVERTED: 1703 case O_IPOPT: 1704 case O_IPTOS: 1705 case O_IPPRECEDENCE: 1706 case O_IPVER: 1707 case O_SOCKARG: 1708 case O_TCPFLAGS: 1709 case O_TCPOPTS: 1710 case O_ESTAB: 1711 case O_VERREVPATH: 1712 case O_VERSRCREACH: 1713 case O_ANTISPOOF: 1714 case O_IPSEC: 1715#ifdef INET6 1716 case O_IP6_SRC_ME: 1717 case O_IP6_DST_ME: 1718 case O_EXT_HDR: 1719 case O_IP6: 1720#endif 1721 case O_IP4: 1722 case O_TAG: 1723 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 1724 goto bad_size; 1725 break; 1726 1727 case O_EXTERNAL_ACTION: 1728 if (cmd->arg1 == 0 || 1729 cmdlen != F_INSN_SIZE(ipfw_insn)) { 1730 printf("ipfw: invalid external " 1731 "action opcode\n"); 1732 return (EINVAL); 1733 } 1734 ci->object_opcodes++; 1735 /* Do we have O_EXTERNAL_INSTANCE opcode? */ 1736 if (l != cmdlen) { 1737 l -= cmdlen; 1738 cmd += cmdlen; 1739 cmdlen = F_LEN(cmd); 1740 if (cmd->opcode != O_EXTERNAL_INSTANCE) { 1741 printf("ipfw: invalid opcode " 1742 "next to external action %u\n", 1743 cmd->opcode); 1744 return (EINVAL); 1745 } 1746 if (cmd->arg1 == 0 || 1747 cmdlen != F_INSN_SIZE(ipfw_insn)) { 1748 printf("ipfw: invalid external " 1749 "action instance opcode\n"); 1750 return (EINVAL); 1751 } 1752 ci->object_opcodes++; 1753 } 1754 goto check_action; 1755 1756 case O_FIB: 1757 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 1758 goto bad_size; 1759 if (cmd->arg1 >= rt_numfibs) { 1760 printf("ipfw: invalid fib number %d\n", 1761 cmd->arg1); 1762 return EINVAL; 1763 } 1764 break; 1765 1766 case O_SETFIB: 1767 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 1768 goto bad_size; 1769 if ((cmd->arg1 != IP_FW_TARG) && 1770 ((cmd->arg1 & 0x7FFF) >= rt_numfibs)) { 1771 printf("ipfw: invalid fib number %d\n", 1772 cmd->arg1 & 0x7FFF); 1773 return EINVAL; 1774 } 1775 goto check_action; 1776 1777 case O_UID: 1778 case O_GID: 1779 case O_JAIL: 1780 case O_IP_SRC: 1781 case O_IP_DST: 1782 case O_TCPSEQ: 1783 case O_TCPACK: 1784 case O_PROB: 1785 case O_ICMPTYPE: 1786 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 1787 goto bad_size; 1788 break; 1789 1790 case O_LIMIT: 1791 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 1792 goto bad_size; 1793 break; 1794 1795 case O_LOG: 1796 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 1797 goto bad_size; 1798 1799 ((ipfw_insn_log *)cmd)->log_left = 1800 ((ipfw_insn_log *)cmd)->max_log; 1801 1802 break; 1803 1804 case O_IP_SRC_MASK: 1805 case O_IP_DST_MASK: 1806 /* only odd command lengths */ 1807 if ((cmdlen & 1) == 0) 1808 goto bad_size; 1809 break; 1810 1811 case O_IP_SRC_SET: 1812 case O_IP_DST_SET: 1813 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 1814 printf("ipfw: invalid set size %d\n", 1815 cmd->arg1); 1816 return EINVAL; 1817 } 1818 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1819 (cmd->arg1+31)/32 ) 1820 goto bad_size; 1821 break; 1822 1823 case O_IP_SRC_LOOKUP: 1824 case O_IP_DST_LOOKUP: 1825 if (cmd->arg1 >= V_fw_tables_max) { 1826 printf("ipfw: invalid table number %d\n", 1827 cmd->arg1); 1828 return (EINVAL); 1829 } 1830 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 1831 cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1 && 1832 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 1833 goto bad_size; 1834 ci->object_opcodes++; 1835 break; 1836 case O_IP_FLOW_LOOKUP: 1837 if (cmd->arg1 >= V_fw_tables_max) { 1838 printf("ipfw: invalid table number %d\n", 1839 cmd->arg1); 1840 return (EINVAL); 1841 } 1842 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 1843 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 1844 goto bad_size; 1845 ci->object_opcodes++; 1846 break; 1847 case O_MACADDR2: 1848 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 1849 goto bad_size; 1850 break; 1851 1852 case O_NOP: 1853 case O_IPID: 1854 case O_IPTTL: 1855 case O_IPLEN: 1856 case O_TCPDATALEN: 1857 case O_TCPWIN: 1858 case O_TAGGED: 1859 if (cmdlen < 1 || cmdlen > 31) 1860 goto bad_size; 1861 break; 1862 1863 case O_DSCP: 1864 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1) 1865 goto bad_size; 1866 break; 1867 1868 case O_MAC_TYPE: 1869 case O_IP_SRCPORT: 1870 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 1871 if (cmdlen < 2 || cmdlen > 31) 1872 goto bad_size; 1873 break; 1874 1875 case O_RECV: 1876 case O_XMIT: 1877 case O_VIA: 1878 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 1879 goto bad_size; 1880 ci->object_opcodes++; 1881 break; 1882 1883 case O_ALTQ: 1884 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 1885 goto bad_size; 1886 break; 1887 1888 case O_PIPE: 1889 case O_QUEUE: 1890 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 1891 goto bad_size; 1892 goto check_action; 1893 1894 case O_FORWARD_IP: 1895 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 1896 goto bad_size; 1897 goto check_action; 1898#ifdef INET6 1899 case O_FORWARD_IP6: 1900 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa6)) 1901 goto bad_size; 1902 goto check_action; 1903#endif /* INET6 */ 1904 1905 case O_DIVERT: 1906 case O_TEE: 1907 if (ip_divert_ptr == NULL) 1908 return EINVAL; 1909 else 1910 goto check_size; 1911 case O_NETGRAPH: 1912 case O_NGTEE: 1913 if (ng_ipfw_input_p == NULL) 1914 return EINVAL; 1915 else 1916 goto check_size; 1917 case O_NAT: 1918 if (!IPFW_NAT_LOADED) 1919 return EINVAL; 1920 if (cmdlen != F_INSN_SIZE(ipfw_insn_nat)) 1921 goto bad_size; 1922 goto check_action; 1923 case O_FORWARD_MAC: /* XXX not implemented yet */ 1924 case O_CHECK_STATE: 1925 case O_COUNT: 1926 case O_ACCEPT: 1927 case O_DENY: 1928 case O_REJECT: 1929 case O_SETDSCP: 1930#ifdef INET6 1931 case O_UNREACH6: 1932#endif 1933 case O_SKIPTO: 1934 case O_REASS: 1935 case O_CALLRETURN: 1936check_size: 1937 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 1938 goto bad_size; 1939check_action: 1940 if (have_action) { 1941 printf("ipfw: opcode %d, multiple actions" 1942 " not allowed\n", 1943 cmd->opcode); 1944 return (EINVAL); 1945 } 1946 have_action = 1; 1947 if (l != cmdlen) { 1948 printf("ipfw: opcode %d, action must be" 1949 " last opcode\n", 1950 cmd->opcode); 1951 return (EINVAL); 1952 } 1953 break; 1954#ifdef INET6 1955 case O_IP6_SRC: 1956 case O_IP6_DST: 1957 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 1958 F_INSN_SIZE(ipfw_insn)) 1959 goto bad_size; 1960 break; 1961 1962 case O_FLOW6ID: 1963 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1964 ((ipfw_insn_u32 *)cmd)->o.arg1) 1965 goto bad_size; 1966 break; 1967 1968 case O_IP6_SRC_MASK: 1969 case O_IP6_DST_MASK: 1970 if ( !(cmdlen & 1) || cmdlen > 127) 1971 goto bad_size; 1972 break; 1973 case O_ICMP6TYPE: 1974 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 1975 goto bad_size; 1976 break; 1977#endif 1978 1979 default: 1980 switch (cmd->opcode) { 1981#ifndef INET6 1982 case O_IP6_SRC_ME: 1983 case O_IP6_DST_ME: 1984 case O_EXT_HDR: 1985 case O_IP6: 1986 case O_UNREACH6: 1987 case O_IP6_SRC: 1988 case O_IP6_DST: 1989 case O_FLOW6ID: 1990 case O_IP6_SRC_MASK: 1991 case O_IP6_DST_MASK: 1992 case O_ICMP6TYPE: 1993 printf("ipfw: no IPv6 support in kernel\n"); 1994 return (EPROTONOSUPPORT); 1995#endif 1996 default: 1997 printf("ipfw: opcode %d, unknown opcode\n", 1998 cmd->opcode); 1999 return (EINVAL); 2000 } 2001 } 2002 } 2003 if (have_action == 0) { 2004 printf("ipfw: missing action\n"); 2005 return (EINVAL); 2006 } 2007 return 0; 2008 2009bad_size: 2010 printf("ipfw: opcode %d size %d wrong\n", 2011 cmd->opcode, cmdlen); 2012 return (EINVAL); 2013} 2014 2015 2016/* 2017 * Translation of requests for compatibility with FreeBSD 7.2/8. 2018 * a static variable tells us if we have an old client from userland, 2019 * and if necessary we translate requests and responses between the 2020 * two formats. 2021 */ 2022static int is7 = 0; 2023 2024struct ip_fw7 { 2025 struct ip_fw7 *next; /* linked list of rules */ 2026 struct ip_fw7 *next_rule; /* ptr to next [skipto] rule */ 2027 /* 'next_rule' is used to pass up 'set_disable' status */ 2028 2029 uint16_t act_ofs; /* offset of action in 32-bit units */ 2030 uint16_t cmd_len; /* # of 32-bit words in cmd */ 2031 uint16_t rulenum; /* rule number */ 2032 uint8_t set; /* rule set (0..31) */ 2033 // #define RESVD_SET 31 /* set for default and persistent rules */ 2034 uint8_t _pad; /* padding */ 2035 // uint32_t id; /* rule id, only in v.8 */ 2036 /* These fields are present in all rules. */ 2037 uint64_t pcnt; /* Packet counter */ 2038 uint64_t bcnt; /* Byte counter */ 2039 uint32_t timestamp; /* tv_sec of last match */ 2040 2041 ipfw_insn cmd[1]; /* storage for commands */ 2042}; 2043 2044static int convert_rule_to_7(struct ip_fw_rule0 *rule); 2045static int convert_rule_to_8(struct ip_fw_rule0 *rule); 2046 2047#ifndef RULESIZE7 2048#define RULESIZE7(rule) (sizeof(struct ip_fw7) + \ 2049 ((struct ip_fw7 *)(rule))->cmd_len * 4 - 4) 2050#endif 2051 2052 2053/* 2054 * Copy the static and dynamic rules to the supplied buffer 2055 * and return the amount of space actually used. 2056 * Must be run under IPFW_UH_RLOCK 2057 */ 2058static size_t 2059ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 2060{ 2061 char *bp = buf; 2062 char *ep = bp + space; 2063 struct ip_fw *rule; 2064 struct ip_fw_rule0 *dst; 2065 struct timeval boottime; 2066 int error, i, l, warnflag; 2067 time_t boot_seconds; 2068 2069 warnflag = 0; 2070 2071 getboottime(&boottime); 2072 boot_seconds = boottime.tv_sec; 2073 for (i = 0; i < chain->n_rules; i++) { 2074 rule = chain->map[i]; 2075 2076 if (is7) { 2077 /* Convert rule to FreeBSd 7.2 format */ 2078 l = RULESIZE7(rule); 2079 if (bp + l + sizeof(uint32_t) <= ep) { 2080 bcopy(rule, bp, l + sizeof(uint32_t)); 2081 error = set_legacy_obj_kidx(chain, 2082 (struct ip_fw_rule0 *)bp); 2083 if (error != 0) 2084 return (0); 2085 error = convert_rule_to_7((struct ip_fw_rule0 *) bp); 2086 if (error) 2087 return 0; /*XXX correct? */ 2088 /* 2089 * XXX HACK. Store the disable mask in the "next" 2090 * pointer in a wild attempt to keep the ABI the same. 2091 * Why do we do this on EVERY rule? 2092 */ 2093 bcopy(&V_set_disable, 2094 &(((struct ip_fw7 *)bp)->next_rule), 2095 sizeof(V_set_disable)); 2096 if (((struct ip_fw7 *)bp)->timestamp) 2097 ((struct ip_fw7 *)bp)->timestamp += boot_seconds; 2098 bp += l; 2099 } 2100 continue; /* go to next rule */ 2101 } 2102 2103 l = RULEUSIZE0(rule); 2104 if (bp + l > ep) { /* should not happen */ 2105 printf("overflow dumping static rules\n"); 2106 break; 2107 } 2108 dst = (struct ip_fw_rule0 *)bp; 2109 export_rule0(rule, dst, l); 2110 error = set_legacy_obj_kidx(chain, dst); 2111 2112 /* 2113 * XXX HACK. Store the disable mask in the "next" 2114 * pointer in a wild attempt to keep the ABI the same. 2115 * Why do we do this on EVERY rule? 2116 * 2117 * XXX: "ipfw set show" (ab)uses IP_FW_GET to read disabled mask 2118 * so we need to fail _after_ saving at least one mask. 2119 */ 2120 bcopy(&V_set_disable, &dst->next_rule, sizeof(V_set_disable)); 2121 if (dst->timestamp) 2122 dst->timestamp += boot_seconds; 2123 bp += l; 2124 2125 if (error != 0) { 2126 if (error == 2) { 2127 /* Non-fatal table rewrite error. */ 2128 warnflag = 1; 2129 continue; 2130 } 2131 printf("Stop on rule %d. Fail to convert table\n", 2132 rule->rulenum); 2133 break; 2134 } 2135 } 2136 if (warnflag != 0) 2137 printf("ipfw: process %s is using legacy interfaces," 2138 " consider rebuilding\n", ""); 2139 ipfw_get_dynamic(chain, &bp, ep); /* protected by the dynamic lock */ 2140 return (bp - (char *)buf); 2141} 2142 2143 2144struct dump_args { 2145 uint32_t b; /* start rule */ 2146 uint32_t e; /* end rule */ 2147 uint32_t rcount; /* number of rules */ 2148 uint32_t rsize; /* rules size */ 2149 uint32_t tcount; /* number of tables */ 2150 int rcounters; /* counters */ 2151}; 2152 2153void 2154ipfw_export_obj_ntlv(struct named_object *no, ipfw_obj_ntlv *ntlv) 2155{ 2156 2157 ntlv->head.type = no->etlv; 2158 ntlv->head.length = sizeof(*ntlv); 2159 ntlv->idx = no->kidx; 2160 strlcpy(ntlv->name, no->name, sizeof(ntlv->name)); 2161} 2162 2163/* 2164 * Export named object info in instance @ni, identified by @kidx 2165 * to ipfw_obj_ntlv. TLV is allocated from @sd space. 2166 * 2167 * Returns 0 on success. 2168 */ 2169static int 2170export_objhash_ntlv(struct namedobj_instance *ni, uint16_t kidx, 2171 struct sockopt_data *sd) 2172{ 2173 struct named_object *no; 2174 ipfw_obj_ntlv *ntlv; 2175 2176 no = ipfw_objhash_lookup_kidx(ni, kidx); 2177 KASSERT(no != NULL, ("invalid object kernel index passed")); 2178 2179 ntlv = (ipfw_obj_ntlv *)ipfw_get_sopt_space(sd, sizeof(*ntlv)); 2180 if (ntlv == NULL) 2181 return (ENOMEM); 2182 2183 ipfw_export_obj_ntlv(no, ntlv); 2184 return (0); 2185} 2186 2187/* 2188 * Dumps static rules with table TLVs in buffer @sd. 2189 * 2190 * Returns 0 on success. 2191 */ 2192static int 2193dump_static_rules(struct ip_fw_chain *chain, struct dump_args *da, 2194 uint32_t *bmask, struct sockopt_data *sd) 2195{ 2196 int error; 2197 int i, l; 2198 uint32_t tcount; 2199 ipfw_obj_ctlv *ctlv; 2200 struct ip_fw *krule; 2201 struct namedobj_instance *ni; 2202 caddr_t dst; 2203 2204 /* Dump table names first (if any) */ 2205 if (da->tcount > 0) { 2206 /* Header first */ 2207 ctlv = (ipfw_obj_ctlv *)ipfw_get_sopt_space(sd, sizeof(*ctlv)); 2208 if (ctlv == NULL) 2209 return (ENOMEM); 2210 ctlv->head.type = IPFW_TLV_TBLNAME_LIST; 2211 ctlv->head.length = da->tcount * sizeof(ipfw_obj_ntlv) + 2212 sizeof(*ctlv); 2213 ctlv->count = da->tcount; 2214 ctlv->objsize = sizeof(ipfw_obj_ntlv); 2215 } 2216 2217 i = 0; 2218 tcount = da->tcount; 2219 ni = ipfw_get_table_objhash(chain); 2220 while (tcount > 0) { 2221 if ((bmask[i / 32] & (1 << (i % 32))) == 0) { 2222 i++; 2223 continue; 2224 } 2225 2226 /* Jump to shared named object bitmask */ 2227 if (i >= IPFW_TABLES_MAX) { 2228 ni = CHAIN_TO_SRV(chain); 2229 i -= IPFW_TABLES_MAX; 2230 bmask += IPFW_TABLES_MAX / 32; 2231 } 2232 2233 if ((error = export_objhash_ntlv(ni, i, sd)) != 0) 2234 return (error); 2235 2236 i++; 2237 tcount--; 2238 } 2239 2240 /* Dump rules */ 2241 ctlv = (ipfw_obj_ctlv *)ipfw_get_sopt_space(sd, sizeof(*ctlv)); 2242 if (ctlv == NULL) 2243 return (ENOMEM); 2244 ctlv->head.type = IPFW_TLV_RULE_LIST; 2245 ctlv->head.length = da->rsize + sizeof(*ctlv); 2246 ctlv->count = da->rcount; 2247 2248 for (i = da->b; i < da->e; i++) { 2249 krule = chain->map[i]; 2250 2251 l = RULEUSIZE1(krule) + sizeof(ipfw_obj_tlv); 2252 if (da->rcounters != 0) 2253 l += sizeof(struct ip_fw_bcounter); 2254 dst = (caddr_t)ipfw_get_sopt_space(sd, l); 2255 if (dst == NULL) 2256 return (ENOMEM); 2257 2258 export_rule1(krule, dst, l, da->rcounters); 2259 } 2260 2261 return (0); 2262} 2263 2264/* 2265 * Marks every object index used in @rule with bit in @bmask. 2266 * Used to generate bitmask of referenced tables/objects for given ruleset 2267 * or its part. 2268 * 2269 * Returns number of newly-referenced objects. 2270 */ 2271static int 2272mark_object_kidx(struct ip_fw_chain *ch, struct ip_fw *rule, 2273 uint32_t *bmask) 2274{ 2275 struct opcode_obj_rewrite *rw; 2276 ipfw_insn *cmd; 2277 int bidx, cmdlen, l, count; 2278 uint16_t kidx; 2279 uint8_t subtype; 2280 2281 l = rule->cmd_len; 2282 cmd = rule->cmd; 2283 cmdlen = 0; 2284 count = 0; 2285 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 2286 cmdlen = F_LEN(cmd); 2287 2288 rw = find_op_rw(cmd, &kidx, &subtype); 2289 if (rw == NULL) 2290 continue; 2291 2292 bidx = kidx / 32; 2293 /* 2294 * Maintain separate bitmasks for table and 2295 * non-table objects. 2296 */ 2297 if (rw->etlv != IPFW_TLV_TBL_NAME) 2298 bidx += IPFW_TABLES_MAX / 32; 2299 2300 if ((bmask[bidx] & (1 << (kidx % 32))) == 0) 2301 count++; 2302 2303 bmask[bidx] |= 1 << (kidx % 32); 2304 } 2305 2306 return (count); 2307} 2308 2309/* 2310 * Dumps requested objects data 2311 * Data layout (version 0)(current): 2312 * Request: [ ipfw_cfg_lheader ] + IPFW_CFG_GET_* flags 2313 * size = ipfw_cfg_lheader.size 2314 * Reply: [ ipfw_cfg_lheader 2315 * [ ipfw_obj_ctlv(IPFW_TLV_TBL_LIST) ipfw_obj_ntlv x N ] (optional) 2316 * [ ipfw_obj_ctlv(IPFW_TLV_RULE_LIST) 2317 * ipfw_obj_tlv(IPFW_TLV_RULE_ENT) [ ip_fw_bcounter (optional) ip_fw_rule ] 2318 * ] (optional) 2319 * [ ipfw_obj_ctlv(IPFW_TLV_STATE_LIST) ipfw_obj_dyntlv x N ] (optional) 2320 * ] 2321 * * NOTE IPFW_TLV_STATE_LIST has the single valid field: objsize. 2322 * The rest (size, count) are set to zero and needs to be ignored. 2323 * 2324 * Returns 0 on success. 2325 */ 2326static int 2327dump_config(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 2328 struct sockopt_data *sd) 2329{ 2330 ipfw_cfg_lheader *hdr; 2331 struct ip_fw *rule; 2332 size_t sz, rnum; 2333 uint32_t hdr_flags; 2334 int error, i; 2335 struct dump_args da; 2336 uint32_t *bmask; 2337 2338 hdr = (ipfw_cfg_lheader *)ipfw_get_sopt_header(sd, sizeof(*hdr)); 2339 if (hdr == NULL) 2340 return (EINVAL); 2341 2342 error = 0; 2343 bmask = NULL; 2344 /* Allocate needed state. Note we allocate 2xspace mask, for table&srv */ 2345 if (hdr->flags & IPFW_CFG_GET_STATIC) 2346 bmask = malloc(IPFW_TABLES_MAX / 4, M_TEMP, M_WAITOK | M_ZERO); 2347 2348 IPFW_UH_RLOCK(chain); 2349 2350 /* 2351 * STAGE 1: Determine size/count for objects in range. 2352 * Prepare used tables bitmask. 2353 */ 2354 sz = sizeof(ipfw_cfg_lheader); 2355 memset(&da, 0, sizeof(da)); 2356 2357 da.b = 0; 2358 da.e = chain->n_rules; 2359 2360 if (hdr->end_rule != 0) { 2361 /* Handle custom range */ 2362 if ((rnum = hdr->start_rule) > IPFW_DEFAULT_RULE) 2363 rnum = IPFW_DEFAULT_RULE; 2364 da.b = ipfw_find_rule(chain, rnum, 0); 2365 rnum = hdr->end_rule; 2366 rnum = (rnum < IPFW_DEFAULT_RULE) ? rnum+1 : IPFW_DEFAULT_RULE; 2367 da.e = ipfw_find_rule(chain, rnum, 0) + 1; 2368 } 2369 2370 if (hdr->flags & IPFW_CFG_GET_STATIC) { 2371 for (i = da.b; i < da.e; i++) { 2372 rule = chain->map[i]; 2373 da.rsize += RULEUSIZE1(rule) + sizeof(ipfw_obj_tlv); 2374 da.rcount++; 2375 /* Update bitmask of used objects for given range */ 2376 da.tcount += mark_object_kidx(chain, rule, bmask); 2377 } 2378 /* Add counters if requested */ 2379 if (hdr->flags & IPFW_CFG_GET_COUNTERS) { 2380 da.rsize += sizeof(struct ip_fw_bcounter) * da.rcount; 2381 da.rcounters = 1; 2382 } 2383 2384 if (da.tcount > 0) 2385 sz += da.tcount * sizeof(ipfw_obj_ntlv) + 2386 sizeof(ipfw_obj_ctlv); 2387 sz += da.rsize + sizeof(ipfw_obj_ctlv); 2388 } 2389 2390 if (hdr->flags & IPFW_CFG_GET_STATES) 2391 sz += ipfw_dyn_get_count() * sizeof(ipfw_obj_dyntlv) + 2392 sizeof(ipfw_obj_ctlv); 2393 2394 2395 /* 2396 * Fill header anyway. 2397 * Note we have to save header fields to stable storage 2398 * buffer inside @sd can be flushed after dumping rules 2399 */ 2400 hdr->size = sz; 2401 hdr->set_mask = ~V_set_disable; 2402 hdr_flags = hdr->flags; 2403 hdr = NULL; 2404 2405 if (sd->valsize < sz) { 2406 error = ENOMEM; 2407 goto cleanup; 2408 } 2409 2410 /* STAGE2: Store actual data */ 2411 if (hdr_flags & IPFW_CFG_GET_STATIC) { 2412 error = dump_static_rules(chain, &da, bmask, sd); 2413 if (error != 0) 2414 goto cleanup; 2415 } 2416 2417 if (hdr_flags & IPFW_CFG_GET_STATES) 2418 error = ipfw_dump_states(chain, sd); 2419 2420cleanup: 2421 IPFW_UH_RUNLOCK(chain); 2422 2423 if (bmask != NULL) 2424 free(bmask, M_TEMP); 2425 2426 return (error); 2427} 2428 2429int 2430ipfw_check_object_name_generic(const char *name) 2431{ 2432 int nsize; 2433 2434 nsize = sizeof(((ipfw_obj_ntlv *)0)->name); 2435 if (strnlen(name, nsize) == nsize) 2436 return (EINVAL); 2437 if (name[0] == '\0') 2438 return (EINVAL); 2439 return (0); 2440} 2441 2442/* 2443 * Creates non-existent objects referenced by rule. 2444 * 2445 * Return 0 on success. 2446 */ 2447int 2448create_objects_compat(struct ip_fw_chain *ch, ipfw_insn *cmd, 2449 struct obj_idx *oib, struct obj_idx *pidx, struct tid_info *ti) 2450{ 2451 struct opcode_obj_rewrite *rw; 2452 struct obj_idx *p; 2453 uint16_t kidx; 2454 int error; 2455 2456 /* 2457 * Compatibility stuff: do actual creation for non-existing, 2458 * but referenced objects. 2459 */ 2460 for (p = oib; p < pidx; p++) { 2461 if (p->kidx != 0) 2462 continue; 2463 2464 ti->uidx = p->uidx; 2465 ti->type = p->type; 2466 ti->atype = 0; 2467 2468 rw = find_op_rw(cmd + p->off, NULL, NULL); 2469 KASSERT(rw != NULL, ("Unable to find handler for op %d", 2470 (cmd + p->off)->opcode)); 2471 2472 if (rw->create_object == NULL) 2473 error = EOPNOTSUPP; 2474 else 2475 error = rw->create_object(ch, ti, &kidx); 2476 if (error == 0) { 2477 p->kidx = kidx; 2478 continue; 2479 } 2480 2481 /* 2482 * Error happened. We have to rollback everything. 2483 * Drop all already acquired references. 2484 */ 2485 IPFW_UH_WLOCK(ch); 2486 unref_oib_objects(ch, cmd, oib, pidx); 2487 IPFW_UH_WUNLOCK(ch); 2488 2489 return (error); 2490 } 2491 2492 return (0); 2493} 2494 2495/* 2496 * Compatibility function for old ipfw(8) binaries. 2497 * Rewrites table/nat kernel indices with userland ones. 2498 * Convert tables matching '/^\d+$/' to their atoi() value. 2499 * Use number 65535 for other tables. 2500 * 2501 * Returns 0 on success. 2502 */ 2503static int 2504set_legacy_obj_kidx(struct ip_fw_chain *ch, struct ip_fw_rule0 *rule) 2505{ 2506 struct opcode_obj_rewrite *rw; 2507 struct named_object *no; 2508 ipfw_insn *cmd; 2509 char *end; 2510 long val; 2511 int cmdlen, error, l; 2512 uint16_t kidx, uidx; 2513 uint8_t subtype; 2514 2515 error = 0; 2516 2517 l = rule->cmd_len; 2518 cmd = rule->cmd; 2519 cmdlen = 0; 2520 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 2521 cmdlen = F_LEN(cmd); 2522 2523 /* Check if is index in given opcode */ 2524 rw = find_op_rw(cmd, &kidx, &subtype); 2525 if (rw == NULL) 2526 continue; 2527 2528 /* Try to find referenced kernel object */ 2529 no = rw->find_bykidx(ch, kidx); 2530 if (no == NULL) 2531 continue; 2532 2533 val = strtol(no->name, &end, 10); 2534 if (*end == '\0' && val < 65535) { 2535 uidx = val; 2536 } else { 2537 2538 /* 2539 * We are called via legacy opcode. 2540 * Save error and show table as fake number 2541 * not to make ipfw(8) hang. 2542 */ 2543 uidx = 65535; 2544 error = 2; 2545 } 2546 2547 rw->update(cmd, uidx); 2548 } 2549 2550 return (error); 2551} 2552 2553 2554/* 2555 * Unreferences all already-referenced objects in given @cmd rule, 2556 * using information in @oib. 2557 * 2558 * Used to rollback partially converted rule on error. 2559 */ 2560static void 2561unref_oib_objects(struct ip_fw_chain *ch, ipfw_insn *cmd, struct obj_idx *oib, 2562 struct obj_idx *end) 2563{ 2564 struct opcode_obj_rewrite *rw; 2565 struct named_object *no; 2566 struct obj_idx *p; 2567 2568 IPFW_UH_WLOCK_ASSERT(ch); 2569 2570 for (p = oib; p < end; p++) { 2571 if (p->kidx == 0) 2572 continue; 2573 2574 rw = find_op_rw(cmd + p->off, NULL, NULL); 2575 KASSERT(rw != NULL, ("Unable to find handler for op %d", 2576 (cmd + p->off)->opcode)); 2577 2578 /* Find & unref by existing idx */ 2579 no = rw->find_bykidx(ch, p->kidx); 2580 KASSERT(no != NULL, ("Ref'd object %d disappeared", p->kidx)); 2581 no->refcnt--; 2582 } 2583} 2584 2585/* 2586 * Remove references from every object used in @rule. 2587 * Used at rule removal code. 2588 */ 2589static void 2590unref_rule_objects(struct ip_fw_chain *ch, struct ip_fw *rule) 2591{ 2592 struct opcode_obj_rewrite *rw; 2593 struct named_object *no; 2594 ipfw_insn *cmd; 2595 int cmdlen, l; 2596 uint16_t kidx; 2597 uint8_t subtype; 2598 2599 IPFW_UH_WLOCK_ASSERT(ch); 2600 2601 l = rule->cmd_len; 2602 cmd = rule->cmd; 2603 cmdlen = 0; 2604 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 2605 cmdlen = F_LEN(cmd); 2606 2607 rw = find_op_rw(cmd, &kidx, &subtype); 2608 if (rw == NULL) 2609 continue; 2610 no = rw->find_bykidx(ch, kidx); 2611 2612 KASSERT(no != NULL, ("table id %d not found", kidx)); 2613 KASSERT(no->subtype == subtype, 2614 ("wrong type %d (%d) for table id %d", 2615 no->subtype, subtype, kidx)); 2616 KASSERT(no->refcnt > 0, ("refcount for table %d is %d", 2617 kidx, no->refcnt)); 2618 2619 if (no->refcnt == 1 && rw->destroy_object != NULL) 2620 rw->destroy_object(ch, no); 2621 else 2622 no->refcnt--; 2623 } 2624} 2625 2626 2627/* 2628 * Find and reference object (if any) stored in instruction @cmd. 2629 * 2630 * Saves object info in @pidx, sets 2631 * - @unresolved to 1 if object should exists but not found 2632 * 2633 * Returns non-zero value in case of error. 2634 */ 2635static int 2636ref_opcode_object(struct ip_fw_chain *ch, ipfw_insn *cmd, struct tid_info *ti, 2637 struct obj_idx *pidx, int *unresolved) 2638{ 2639 struct named_object *no; 2640 struct opcode_obj_rewrite *rw; 2641 int error; 2642 2643 /* Check if this opcode is candidate for rewrite */ 2644 rw = find_op_rw(cmd, &ti->uidx, &ti->type); 2645 if (rw == NULL) 2646 return (0); 2647 2648 /* Need to rewrite. Save necessary fields */ 2649 pidx->uidx = ti->uidx; 2650 pidx->type = ti->type; 2651 2652 /* Try to find referenced kernel object */ 2653 error = rw->find_byname(ch, ti, &no); 2654 if (error != 0) 2655 return (error); 2656 if (no == NULL) { 2657 /* 2658 * Report about unresolved object for automaic 2659 * creation. 2660 */ 2661 *unresolved = 1; 2662 return (0); 2663 } 2664 2665 /* Found. Bump refcount and update kidx. */ 2666 no->refcnt++; 2667 rw->update(cmd, no->kidx); 2668 return (0); 2669} 2670 2671/* 2672 * Finds and bumps refcount for objects referenced by given @rule. 2673 * Auto-creates non-existing tables. 2674 * Fills in @oib array with userland/kernel indexes. 2675 * 2676 * Returns 0 on success. 2677 */ 2678static int 2679ref_rule_objects(struct ip_fw_chain *ch, struct ip_fw *rule, 2680 struct rule_check_info *ci, struct obj_idx *oib, struct tid_info *ti) 2681{ 2682 struct obj_idx *pidx; 2683 ipfw_insn *cmd; 2684 int cmdlen, error, l, unresolved; 2685 2686 pidx = oib; 2687 l = rule->cmd_len; 2688 cmd = rule->cmd; 2689 cmdlen = 0; 2690 error = 0; 2691 2692 IPFW_UH_WLOCK(ch); 2693 2694 /* Increase refcount on each existing referenced table. */ 2695 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 2696 cmdlen = F_LEN(cmd); 2697 unresolved = 0; 2698 2699 error = ref_opcode_object(ch, cmd, ti, pidx, &unresolved); 2700 if (error != 0) 2701 break; 2702 /* 2703 * Compatibility stuff for old clients: 2704 * prepare to automaitcally create non-existing objects. 2705 */ 2706 if (unresolved != 0) { 2707 pidx->off = rule->cmd_len - l; 2708 pidx++; 2709 } 2710 } 2711 2712 if (error != 0) { 2713 /* Unref everything we have already done */ 2714 unref_oib_objects(ch, rule->cmd, oib, pidx); 2715 IPFW_UH_WUNLOCK(ch); 2716 return (error); 2717 } 2718 IPFW_UH_WUNLOCK(ch); 2719 2720 /* Perform auto-creation for non-existing objects */ 2721 if (pidx != oib) 2722 error = create_objects_compat(ch, rule->cmd, oib, pidx, ti); 2723 2724 /* Calculate real number of dynamic objects */ 2725 ci->object_opcodes = (uint16_t)(pidx - oib); 2726 2727 return (error); 2728} 2729 2730/* 2731 * Checks is opcode is referencing table of appropriate type. 2732 * Adds reference count for found table if true. 2733 * Rewrites user-supplied opcode values with kernel ones. 2734 * 2735 * Returns 0 on success and appropriate error code otherwise. 2736 */ 2737static int 2738rewrite_rule_uidx(struct ip_fw_chain *chain, struct rule_check_info *ci) 2739{ 2740 int error; 2741 ipfw_insn *cmd; 2742 uint8_t type; 2743 struct obj_idx *p, *pidx_first, *pidx_last; 2744 struct tid_info ti; 2745 2746 /* 2747 * Prepare an array for storing opcode indices. 2748 * Use stack allocation by default. 2749 */ 2750 if (ci->object_opcodes <= (sizeof(ci->obuf)/sizeof(ci->obuf[0]))) { 2751 /* Stack */ 2752 pidx_first = ci->obuf; 2753 } else 2754 pidx_first = malloc( 2755 ci->object_opcodes * sizeof(struct obj_idx), 2756 M_IPFW, M_WAITOK | M_ZERO); 2757 2758 error = 0; 2759 type = 0; 2760 memset(&ti, 0, sizeof(ti)); 2761 2762 /* Use set rule is assigned to. */ 2763 ti.set = ci->krule->set; 2764 if (ci->ctlv != NULL) { 2765 ti.tlvs = (void *)(ci->ctlv + 1); 2766 ti.tlen = ci->ctlv->head.length - sizeof(ipfw_obj_ctlv); 2767 } 2768 2769 /* Reference all used tables and other objects */ 2770 error = ref_rule_objects(chain, ci->krule, ci, pidx_first, &ti); 2771 if (error != 0) 2772 goto free; 2773 /* 2774 * Note that ref_rule_objects() might have updated ci->object_opcodes 2775 * to reflect actual number of object opcodes. 2776 */ 2777 2778 /* Perform rewrite of remaining opcodes */ 2779 p = pidx_first; 2780 pidx_last = pidx_first + ci->object_opcodes; 2781 for (p = pidx_first; p < pidx_last; p++) { 2782 cmd = ci->krule->cmd + p->off; 2783 update_opcode_kidx(cmd, p->kidx); 2784 } 2785 2786free: 2787 if (pidx_first != ci->obuf) 2788 free(pidx_first, M_IPFW); 2789 2790 return (error); 2791} 2792 2793/* 2794 * Adds one or more rules to ipfw @chain. 2795 * Data layout (version 0)(current): 2796 * Request: 2797 * [ 2798 * ip_fw3_opheader 2799 * [ ipfw_obj_ctlv(IPFW_TLV_TBL_LIST) ipfw_obj_ntlv x N ] (optional *1) 2800 * [ ipfw_obj_ctlv(IPFW_TLV_RULE_LIST) ip_fw x N ] (*2) (*3) 2801 * ] 2802 * Reply: 2803 * [ 2804 * ip_fw3_opheader 2805 * [ ipfw_obj_ctlv(IPFW_TLV_TBL_LIST) ipfw_obj_ntlv x N ] (optional) 2806 * [ ipfw_obj_ctlv(IPFW_TLV_RULE_LIST) ip_fw x N ] 2807 * ] 2808 * 2809 * Rules in reply are modified to store their actual ruleset number. 2810 * 2811 * (*1) TLVs inside IPFW_TLV_TBL_LIST needs to be sorted ascending 2812 * according to their idx field and there has to be no duplicates. 2813 * (*2) Numbered rules inside IPFW_TLV_RULE_LIST needs to be sorted ascending. 2814 * (*3) Each ip_fw structure needs to be aligned to u64 boundary. 2815 * 2816 * Returns 0 on success. 2817 */ 2818static int 2819add_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 2820 struct sockopt_data *sd) 2821{ 2822 ipfw_obj_ctlv *ctlv, *rtlv, *tstate; 2823 ipfw_obj_ntlv *ntlv; 2824 int clen, error, idx; 2825 uint32_t count, read; 2826 struct ip_fw_rule *r; 2827 struct rule_check_info rci, *ci, *cbuf; 2828 int i, rsize; 2829 2830 op3 = (ip_fw3_opheader *)ipfw_get_sopt_space(sd, sd->valsize); 2831 ctlv = (ipfw_obj_ctlv *)(op3 + 1); 2832 2833 read = sizeof(ip_fw3_opheader); 2834 rtlv = NULL; 2835 tstate = NULL; 2836 cbuf = NULL; 2837 memset(&rci, 0, sizeof(struct rule_check_info)); 2838 2839 if (read + sizeof(*ctlv) > sd->valsize) 2840 return (EINVAL); 2841 2842 if (ctlv->head.type == IPFW_TLV_TBLNAME_LIST) { 2843 clen = ctlv->head.length; 2844 /* Check size and alignment */ 2845 if (clen > sd->valsize || clen < sizeof(*ctlv)) 2846 return (EINVAL); 2847 if ((clen % sizeof(uint64_t)) != 0) 2848 return (EINVAL); 2849 2850 /* 2851 * Some table names or other named objects. 2852 * Check for validness. 2853 */ 2854 count = (ctlv->head.length - sizeof(*ctlv)) / sizeof(*ntlv); 2855 if (ctlv->count != count || ctlv->objsize != sizeof(*ntlv)) 2856 return (EINVAL); 2857 2858 /* 2859 * Check each TLV. 2860 * Ensure TLVs are sorted ascending and 2861 * there are no duplicates. 2862 */ 2863 idx = -1; 2864 ntlv = (ipfw_obj_ntlv *)(ctlv + 1); 2865 while (count > 0) { 2866 if (ntlv->head.length != sizeof(ipfw_obj_ntlv)) 2867 return (EINVAL); 2868 2869 error = ipfw_check_object_name_generic(ntlv->name); 2870 if (error != 0) 2871 return (error); 2872 2873 if (ntlv->idx <= idx) 2874 return (EINVAL); 2875 2876 idx = ntlv->idx; 2877 count--; 2878 ntlv++; 2879 } 2880 2881 tstate = ctlv; 2882 read += ctlv->head.length; 2883 ctlv = (ipfw_obj_ctlv *)((caddr_t)ctlv + ctlv->head.length); 2884 } 2885 2886 if (read + sizeof(*ctlv) > sd->valsize) 2887 return (EINVAL); 2888 2889 if (ctlv->head.type == IPFW_TLV_RULE_LIST) { 2890 clen = ctlv->head.length; 2891 if (clen + read > sd->valsize || clen < sizeof(*ctlv)) 2892 return (EINVAL); 2893 if ((clen % sizeof(uint64_t)) != 0) 2894 return (EINVAL); 2895 2896 /* 2897 * TODO: Permit adding multiple rules at once 2898 */ 2899 if (ctlv->count != 1) 2900 return (ENOTSUP); 2901 2902 clen -= sizeof(*ctlv); 2903 2904 if (ctlv->count > clen / sizeof(struct ip_fw_rule)) 2905 return (EINVAL); 2906 2907 /* Allocate state for each rule or use stack */ 2908 if (ctlv->count == 1) { 2909 memset(&rci, 0, sizeof(struct rule_check_info)); 2910 cbuf = &rci; 2911 } else 2912 cbuf = malloc(ctlv->count * sizeof(*ci), M_TEMP, 2913 M_WAITOK | M_ZERO); 2914 ci = cbuf; 2915 2916 /* 2917 * Check each rule for validness. 2918 * Ensure numbered rules are sorted ascending 2919 * and properly aligned 2920 */ 2921 idx = 0; 2922 r = (struct ip_fw_rule *)(ctlv + 1); 2923 count = 0; 2924 error = 0; 2925 while (clen > 0) { 2926 rsize = roundup2(RULESIZE(r), sizeof(uint64_t)); 2927 if (rsize > clen || ctlv->count <= count) { 2928 error = EINVAL; 2929 break; 2930 } 2931 2932 ci->ctlv = tstate; 2933 error = check_ipfw_rule1(r, rsize, ci); 2934 if (error != 0) 2935 break; 2936 2937 /* Check sorting */ 2938 if (r->rulenum != 0 && r->rulenum < idx) { 2939 printf("rulenum %d idx %d\n", r->rulenum, idx); 2940 error = EINVAL; 2941 break; 2942 } 2943 idx = r->rulenum; 2944 2945 ci->urule = (caddr_t)r; 2946 2947 rsize = roundup2(rsize, sizeof(uint64_t)); 2948 clen -= rsize; 2949 r = (struct ip_fw_rule *)((caddr_t)r + rsize); 2950 count++; 2951 ci++; 2952 } 2953 2954 if (ctlv->count != count || error != 0) { 2955 if (cbuf != &rci) 2956 free(cbuf, M_TEMP); 2957 return (EINVAL); 2958 } 2959 2960 rtlv = ctlv; 2961 read += ctlv->head.length; 2962 ctlv = (ipfw_obj_ctlv *)((caddr_t)ctlv + ctlv->head.length); 2963 } 2964 2965 if (read != sd->valsize || rtlv == NULL || rtlv->count == 0) { 2966 if (cbuf != NULL && cbuf != &rci) 2967 free(cbuf, M_TEMP); 2968 return (EINVAL); 2969 } 2970 2971 /* 2972 * Passed rules seems to be valid. 2973 * Allocate storage and try to add them to chain. 2974 */ 2975 for (i = 0, ci = cbuf; i < rtlv->count; i++, ci++) { 2976 clen = RULEKSIZE1((struct ip_fw_rule *)ci->urule); 2977 ci->krule = ipfw_alloc_rule(chain, clen); 2978 import_rule1(ci); 2979 } 2980 2981 if ((error = commit_rules(chain, cbuf, rtlv->count)) != 0) { 2982 /* Free allocate krules */ 2983 for (i = 0, ci = cbuf; i < rtlv->count; i++, ci++) 2984 free_rule(ci->krule); 2985 } 2986 2987 if (cbuf != NULL && cbuf != &rci) 2988 free(cbuf, M_TEMP); 2989 2990 return (error); 2991} 2992 2993/* 2994 * Lists all sopts currently registered. 2995 * Data layout (v0)(current): 2996 * Request: [ ipfw_obj_lheader ], size = ipfw_obj_lheader.size 2997 * Reply: [ ipfw_obj_lheader ipfw_sopt_info x N ] 2998 * 2999 * Returns 0 on success 3000 */ 3001static int 3002dump_soptcodes(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 3003 struct sockopt_data *sd) 3004{ 3005 struct _ipfw_obj_lheader *olh; 3006 ipfw_sopt_info *i; 3007 struct ipfw_sopt_handler *sh; 3008 uint32_t count, n, size; 3009 3010 olh = (struct _ipfw_obj_lheader *)ipfw_get_sopt_header(sd,sizeof(*olh)); 3011 if (olh == NULL) 3012 return (EINVAL); 3013 if (sd->valsize < olh->size) 3014 return (EINVAL); 3015 3016 CTL3_LOCK(); 3017 count = ctl3_hsize; 3018 size = count * sizeof(ipfw_sopt_info) + sizeof(ipfw_obj_lheader); 3019 3020 /* Fill in header regadless of buffer size */ 3021 olh->count = count; 3022 olh->objsize = sizeof(ipfw_sopt_info); 3023 3024 if (size > olh->size) { 3025 olh->size = size; 3026 CTL3_UNLOCK(); 3027 return (ENOMEM); 3028 } 3029 olh->size = size; 3030 3031 for (n = 1; n <= count; n++) { 3032 i = (ipfw_sopt_info *)ipfw_get_sopt_space(sd, sizeof(*i)); 3033 KASSERT(i != NULL, ("previously checked buffer is not enough")); 3034 sh = &ctl3_handlers[n]; 3035 i->opcode = sh->opcode; 3036 i->version = sh->version; 3037 i->refcnt = sh->refcnt; 3038 } 3039 CTL3_UNLOCK(); 3040 3041 return (0); 3042} 3043 3044/* 3045 * Compares two opcodes. 3046 * Used both in qsort() and bsearch(). 3047 * 3048 * Returns 0 if match is found. 3049 */ 3050static int 3051compare_opcodes(const void *_a, const void *_b) 3052{ 3053 const struct opcode_obj_rewrite *a, *b; 3054 3055 a = (const struct opcode_obj_rewrite *)_a; 3056 b = (const struct opcode_obj_rewrite *)_b; 3057 3058 if (a->opcode < b->opcode) 3059 return (-1); 3060 else if (a->opcode > b->opcode) 3061 return (1); 3062 3063 return (0); 3064} 3065 3066/* 3067 * XXX: Rewrite bsearch() 3068 */ 3069static int 3070find_op_rw_range(uint16_t op, struct opcode_obj_rewrite **plo, 3071 struct opcode_obj_rewrite **phi) 3072{ 3073 struct opcode_obj_rewrite *ctl3_max, *lo, *hi, h, *rw; 3074 3075 memset(&h, 0, sizeof(h)); 3076 h.opcode = op; 3077 3078 rw = (struct opcode_obj_rewrite *)bsearch(&h, ctl3_rewriters, 3079 ctl3_rsize, sizeof(h), compare_opcodes); 3080 if (rw == NULL) 3081 return (1); 3082 3083 /* Find the first element matching the same opcode */ 3084 lo = rw; 3085 for ( ; lo > ctl3_rewriters && (lo - 1)->opcode == op; lo--) 3086 ; 3087 3088 /* Find the last element matching the same opcode */ 3089 hi = rw; 3090 ctl3_max = ctl3_rewriters + ctl3_rsize; 3091 for ( ; (hi + 1) < ctl3_max && (hi + 1)->opcode == op; hi++) 3092 ; 3093 3094 *plo = lo; 3095 *phi = hi; 3096 3097 return (0); 3098} 3099 3100/* 3101 * Finds opcode object rewriter based on @code. 3102 * 3103 * Returns pointer to handler or NULL. 3104 */ 3105static struct opcode_obj_rewrite * 3106find_op_rw(ipfw_insn *cmd, uint16_t *puidx, uint8_t *ptype) 3107{ 3108 struct opcode_obj_rewrite *rw, *lo, *hi; 3109 uint16_t uidx; 3110 uint8_t subtype; 3111 3112 if (find_op_rw_range(cmd->opcode, &lo, &hi) != 0) 3113 return (NULL); 3114 3115 for (rw = lo; rw <= hi; rw++) { 3116 if (rw->classifier(cmd, &uidx, &subtype) == 0) { 3117 if (puidx != NULL) 3118 *puidx = uidx; 3119 if (ptype != NULL) 3120 *ptype = subtype; 3121 return (rw); 3122 } 3123 } 3124 3125 return (NULL); 3126} 3127int 3128classify_opcode_kidx(ipfw_insn *cmd, uint16_t *puidx) 3129{ 3130 3131 if (find_op_rw(cmd, puidx, NULL) == 0) 3132 return (1); 3133 return (0); 3134} 3135 3136void 3137update_opcode_kidx(ipfw_insn *cmd, uint16_t idx) 3138{ 3139 struct opcode_obj_rewrite *rw; 3140 3141 rw = find_op_rw(cmd, NULL, NULL); 3142 KASSERT(rw != NULL, ("No handler to update opcode %d", cmd->opcode)); 3143 rw->update(cmd, idx); 3144} 3145 3146void 3147ipfw_init_obj_rewriter() 3148{ 3149 3150 ctl3_rewriters = NULL; 3151 ctl3_rsize = 0; 3152} 3153 3154void 3155ipfw_destroy_obj_rewriter() 3156{ 3157 3158 if (ctl3_rewriters != NULL) 3159 free(ctl3_rewriters, M_IPFW); 3160 ctl3_rewriters = NULL; 3161 ctl3_rsize = 0; 3162} 3163 3164/* 3165 * Adds one or more opcode object rewrite handlers to the global array. 3166 * Function may sleep. 3167 */ 3168void 3169ipfw_add_obj_rewriter(struct opcode_obj_rewrite *rw, size_t count) 3170{ 3171 size_t sz; 3172 struct opcode_obj_rewrite *tmp; 3173 3174 CTL3_LOCK(); 3175 3176 for (;;) { 3177 sz = ctl3_rsize + count; 3178 CTL3_UNLOCK(); 3179 tmp = malloc(sizeof(*rw) * sz, M_IPFW, M_WAITOK | M_ZERO); 3180 CTL3_LOCK(); 3181 if (ctl3_rsize + count <= sz) 3182 break; 3183 3184 /* Retry */ 3185 free(tmp, M_IPFW); 3186 } 3187 3188 /* Merge old & new arrays */ 3189 sz = ctl3_rsize + count; 3190 memcpy(tmp, ctl3_rewriters, ctl3_rsize * sizeof(*rw)); 3191 memcpy(&tmp[ctl3_rsize], rw, count * sizeof(*rw)); 3192 qsort(tmp, sz, sizeof(*rw), compare_opcodes); 3193 /* Switch new and free old */ 3194 if (ctl3_rewriters != NULL) 3195 free(ctl3_rewriters, M_IPFW); 3196 ctl3_rewriters = tmp; 3197 ctl3_rsize = sz; 3198 3199 CTL3_UNLOCK(); 3200} 3201 3202/* 3203 * Removes one or more object rewrite handlers from the global array. 3204 */ 3205int 3206ipfw_del_obj_rewriter(struct opcode_obj_rewrite *rw, size_t count) 3207{ 3208 size_t sz; 3209 struct opcode_obj_rewrite *ctl3_max, *ktmp, *lo, *hi; 3210 int i; 3211 3212 CTL3_LOCK(); 3213 3214 for (i = 0; i < count; i++) { 3215 if (find_op_rw_range(rw[i].opcode, &lo, &hi) != 0) 3216 continue; 3217 3218 for (ktmp = lo; ktmp <= hi; ktmp++) { 3219 if (ktmp->classifier != rw[i].classifier) 3220 continue; 3221 3222 ctl3_max = ctl3_rewriters + ctl3_rsize; 3223 sz = (ctl3_max - (ktmp + 1)) * sizeof(*ktmp); 3224 memmove(ktmp, ktmp + 1, sz); 3225 ctl3_rsize--; 3226 break; 3227 } 3228 3229 } 3230 3231 if (ctl3_rsize == 0) { 3232 if (ctl3_rewriters != NULL) 3233 free(ctl3_rewriters, M_IPFW); 3234 ctl3_rewriters = NULL; 3235 } 3236 3237 CTL3_UNLOCK(); 3238 3239 return (0); 3240} 3241 3242static int 3243export_objhash_ntlv_internal(struct namedobj_instance *ni, 3244 struct named_object *no, void *arg) 3245{ 3246 struct sockopt_data *sd; 3247 ipfw_obj_ntlv *ntlv; 3248 3249 sd = (struct sockopt_data *)arg; 3250 ntlv = (ipfw_obj_ntlv *)ipfw_get_sopt_space(sd, sizeof(*ntlv)); 3251 if (ntlv == NULL) 3252 return (ENOMEM); 3253 ipfw_export_obj_ntlv(no, ntlv); 3254 return (0); 3255} 3256 3257/* 3258 * Lists all service objects. 3259 * Data layout (v0)(current): 3260 * Request: [ ipfw_obj_lheader ] size = ipfw_obj_lheader.size 3261 * Reply: [ ipfw_obj_lheader [ ipfw_obj_ntlv x N ] (optional) ] 3262 * Returns 0 on success 3263 */ 3264static int 3265dump_srvobjects(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 3266 struct sockopt_data *sd) 3267{ 3268 ipfw_obj_lheader *hdr; 3269 int count; 3270 3271 hdr = (ipfw_obj_lheader *)ipfw_get_sopt_header(sd, sizeof(*hdr)); 3272 if (hdr == NULL) 3273 return (EINVAL); 3274 3275 IPFW_UH_RLOCK(chain); 3276 count = ipfw_objhash_count(CHAIN_TO_SRV(chain)); 3277 hdr->size = sizeof(ipfw_obj_lheader) + count * sizeof(ipfw_obj_ntlv); 3278 if (sd->valsize < hdr->size) { 3279 IPFW_UH_RUNLOCK(chain); 3280 return (ENOMEM); 3281 } 3282 hdr->count = count; 3283 hdr->objsize = sizeof(ipfw_obj_ntlv); 3284 if (count > 0) 3285 ipfw_objhash_foreach(CHAIN_TO_SRV(chain), 3286 export_objhash_ntlv_internal, sd); 3287 IPFW_UH_RUNLOCK(chain); 3288 return (0); 3289} 3290 3291/* 3292 * Compares two sopt handlers (code, version and handler ptr). 3293 * Used both as qsort() and bsearch(). 3294 * Does not compare handler for latter case. 3295 * 3296 * Returns 0 if match is found. 3297 */ 3298static int 3299compare_sh(const void *_a, const void *_b) 3300{ 3301 const struct ipfw_sopt_handler *a, *b; 3302 3303 a = (const struct ipfw_sopt_handler *)_a; 3304 b = (const struct ipfw_sopt_handler *)_b; 3305 3306 if (a->opcode < b->opcode) 3307 return (-1); 3308 else if (a->opcode > b->opcode) 3309 return (1); 3310 3311 if (a->version < b->version) 3312 return (-1); 3313 else if (a->version > b->version) 3314 return (1); 3315 3316 /* bsearch helper */ 3317 if (a->handler == NULL) 3318 return (0); 3319 3320 if ((uintptr_t)a->handler < (uintptr_t)b->handler) 3321 return (-1); 3322 else if ((uintptr_t)a->handler > (uintptr_t)b->handler) 3323 return (1); 3324 3325 return (0); 3326} 3327 3328/* 3329 * Finds sopt handler based on @code and @version. 3330 * 3331 * Returns pointer to handler or NULL. 3332 */ 3333static struct ipfw_sopt_handler * 3334find_sh(uint16_t code, uint8_t version, sopt_handler_f *handler) 3335{ 3336 struct ipfw_sopt_handler *sh, h; 3337 3338 memset(&h, 0, sizeof(h)); 3339 h.opcode = code; 3340 h.version = version; 3341 h.handler = handler; 3342 3343 sh = (struct ipfw_sopt_handler *)bsearch(&h, ctl3_handlers, 3344 ctl3_hsize, sizeof(h), compare_sh); 3345 3346 return (sh); 3347} 3348 3349static int 3350find_ref_sh(uint16_t opcode, uint8_t version, struct ipfw_sopt_handler *psh) 3351{ 3352 struct ipfw_sopt_handler *sh; 3353 3354 CTL3_LOCK(); 3355 if ((sh = find_sh(opcode, version, NULL)) == NULL) { 3356 CTL3_UNLOCK(); 3357 printf("ipfw: ipfw_ctl3 invalid option %d""v""%d\n", 3358 opcode, version); 3359 return (EINVAL); 3360 } 3361 sh->refcnt++; 3362 ctl3_refct++; 3363 /* Copy handler data to requested buffer */ 3364 *psh = *sh; 3365 CTL3_UNLOCK(); 3366 3367 return (0); 3368} 3369 3370static void 3371find_unref_sh(struct ipfw_sopt_handler *psh) 3372{ 3373 struct ipfw_sopt_handler *sh; 3374 3375 CTL3_LOCK(); 3376 sh = find_sh(psh->opcode, psh->version, NULL); 3377 KASSERT(sh != NULL, ("ctl3 handler disappeared")); 3378 sh->refcnt--; 3379 ctl3_refct--; 3380 CTL3_UNLOCK(); 3381} 3382 3383void 3384ipfw_init_sopt_handler() 3385{ 3386 3387 CTL3_LOCK_INIT(); 3388 IPFW_ADD_SOPT_HANDLER(1, scodes); 3389} 3390 3391void 3392ipfw_destroy_sopt_handler() 3393{ 3394 3395 IPFW_DEL_SOPT_HANDLER(1, scodes); 3396 CTL3_LOCK_DESTROY(); 3397} 3398 3399/* 3400 * Adds one or more sockopt handlers to the global array. 3401 * Function may sleep. 3402 */ 3403void 3404ipfw_add_sopt_handler(struct ipfw_sopt_handler *sh, size_t count) 3405{ 3406 size_t sz; 3407 struct ipfw_sopt_handler *tmp; 3408 3409 CTL3_LOCK(); 3410 3411 for (;;) { 3412 sz = ctl3_hsize + count; 3413 CTL3_UNLOCK(); 3414 tmp = malloc(sizeof(*sh) * sz, M_IPFW, M_WAITOK | M_ZERO); 3415 CTL3_LOCK(); 3416 if (ctl3_hsize + count <= sz) 3417 break; 3418 3419 /* Retry */ 3420 free(tmp, M_IPFW); 3421 } 3422 3423 /* Merge old & new arrays */ 3424 sz = ctl3_hsize + count; 3425 memcpy(tmp, ctl3_handlers, ctl3_hsize * sizeof(*sh)); 3426 memcpy(&tmp[ctl3_hsize], sh, count * sizeof(*sh)); 3427 qsort(tmp, sz, sizeof(*sh), compare_sh); 3428 /* Switch new and free old */ 3429 if (ctl3_handlers != NULL) 3430 free(ctl3_handlers, M_IPFW); 3431 ctl3_handlers = tmp; 3432 ctl3_hsize = sz; 3433 ctl3_gencnt++; 3434 3435 CTL3_UNLOCK(); 3436} 3437 3438/* 3439 * Removes one or more sockopt handlers from the global array. 3440 */ 3441int 3442ipfw_del_sopt_handler(struct ipfw_sopt_handler *sh, size_t count) 3443{ 3444 size_t sz; 3445 struct ipfw_sopt_handler *tmp, *h; 3446 int i; 3447 3448 CTL3_LOCK(); 3449 3450 for (i = 0; i < count; i++) { 3451 tmp = &sh[i]; 3452 h = find_sh(tmp->opcode, tmp->version, tmp->handler); 3453 if (h == NULL) 3454 continue; 3455 3456 sz = (ctl3_handlers + ctl3_hsize - (h + 1)) * sizeof(*h); 3457 memmove(h, h + 1, sz); 3458 ctl3_hsize--; 3459 } 3460 3461 if (ctl3_hsize == 0) { 3462 if (ctl3_handlers != NULL) 3463 free(ctl3_handlers, M_IPFW); 3464 ctl3_handlers = NULL; 3465 } 3466 3467 ctl3_gencnt++; 3468 3469 CTL3_UNLOCK(); 3470 3471 return (0); 3472} 3473 3474/* 3475 * Writes data accumulated in @sd to sockopt buffer. 3476 * Zeroes internal @sd buffer. 3477 */ 3478static int 3479ipfw_flush_sopt_data(struct sockopt_data *sd) 3480{ 3481 struct sockopt *sopt; 3482 int error; 3483 size_t sz; 3484 3485 sz = sd->koff; 3486 if (sz == 0) 3487 return (0); 3488 3489 sopt = sd->sopt; 3490 3491 if (sopt->sopt_dir == SOPT_GET) { 3492 error = copyout(sd->kbuf, sopt->sopt_val, sz); 3493 if (error != 0) 3494 return (error); 3495 } 3496 3497 memset(sd->kbuf, 0, sd->ksize); 3498 sd->ktotal += sz; 3499 sd->koff = 0; 3500 if (sd->ktotal + sd->ksize < sd->valsize) 3501 sd->kavail = sd->ksize; 3502 else 3503 sd->kavail = sd->valsize - sd->ktotal; 3504 3505 /* Update sopt buffer data */ 3506 sopt->sopt_valsize = sd->ktotal; 3507 sopt->sopt_val = sd->sopt_val + sd->ktotal; 3508 3509 return (0); 3510} 3511 3512/* 3513 * Ensures that @sd buffer has contiguous @neeeded number of 3514 * bytes. 3515 * 3516 * Returns pointer to requested space or NULL. 3517 */ 3518caddr_t 3519ipfw_get_sopt_space(struct sockopt_data *sd, size_t needed) 3520{ 3521 int error; 3522 caddr_t addr; 3523 3524 if (sd->kavail < needed) { 3525 /* 3526 * Flush data and try another time. 3527 */ 3528 error = ipfw_flush_sopt_data(sd); 3529 3530 if (sd->kavail < needed || error != 0) 3531 return (NULL); 3532 } 3533 3534 addr = sd->kbuf + sd->koff; 3535 sd->koff += needed; 3536 sd->kavail -= needed; 3537 return (addr); 3538} 3539 3540/* 3541 * Requests @needed contiguous bytes from @sd buffer. 3542 * Function is used to notify subsystem that we are 3543 * interesed in first @needed bytes (request header) 3544 * and the rest buffer can be safely zeroed. 3545 * 3546 * Returns pointer to requested space or NULL. 3547 */ 3548caddr_t 3549ipfw_get_sopt_header(struct sockopt_data *sd, size_t needed) 3550{ 3551 caddr_t addr; 3552 3553 if ((addr = ipfw_get_sopt_space(sd, needed)) == NULL) 3554 return (NULL); 3555 3556 if (sd->kavail > 0) 3557 memset(sd->kbuf + sd->koff, 0, sd->kavail); 3558 3559 return (addr); 3560} 3561 3562/* 3563 * New sockopt handler. 3564 */ 3565int 3566ipfw_ctl3(struct sockopt *sopt) 3567{ 3568 int error, locked; 3569 size_t size, valsize; 3570 struct ip_fw_chain *chain; 3571 char xbuf[256]; 3572 struct sockopt_data sdata; 3573 struct ipfw_sopt_handler h; 3574 ip_fw3_opheader *op3 = NULL; 3575 3576 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW); 3577 if (error != 0) 3578 return (error); 3579 3580 if (sopt->sopt_name != IP_FW3) 3581 return (ipfw_ctl(sopt)); 3582 3583 chain = &V_layer3_chain; 3584 error = 0; 3585 3586 /* Save original valsize before it is altered via sooptcopyin() */ 3587 valsize = sopt->sopt_valsize; 3588 memset(&sdata, 0, sizeof(sdata)); 3589 /* Read op3 header first to determine actual operation */ 3590 op3 = (ip_fw3_opheader *)xbuf; 3591 error = sooptcopyin(sopt, op3, sizeof(*op3), sizeof(*op3)); 3592 if (error != 0) 3593 return (error); 3594 sopt->sopt_valsize = valsize; 3595 3596 /* 3597 * Find and reference command. 3598 */ 3599 error = find_ref_sh(op3->opcode, op3->version, &h); 3600 if (error != 0) 3601 return (error); 3602 3603 /* 3604 * Disallow modifications in really-really secure mode, but still allow 3605 * the logging counters to be reset. 3606 */ 3607 if ((h.dir & HDIR_SET) != 0 && h.opcode != IP_FW_XRESETLOG) { 3608 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3609 if (error != 0) { 3610 find_unref_sh(&h); 3611 return (error); 3612 } 3613 } 3614 3615 /* 3616 * Fill in sockopt_data structure that may be useful for 3617 * IP_FW3 get requests. 3618 */ 3619 locked = 0; 3620 if (valsize <= sizeof(xbuf)) { 3621 /* use on-stack buffer */ 3622 sdata.kbuf = xbuf; 3623 sdata.ksize = sizeof(xbuf); 3624 sdata.kavail = valsize; 3625 } else { 3626 3627 /* 3628 * Determine opcode type/buffer size: 3629 * allocate sliding-window buf for data export or 3630 * contiguous buffer for special ops. 3631 */ 3632 if ((h.dir & HDIR_SET) != 0) { 3633 /* Set request. Allocate contigous buffer. */ 3634 if (valsize > CTL3_LARGEBUF) { 3635 find_unref_sh(&h); 3636 return (EFBIG); 3637 } 3638 3639 size = valsize; 3640 } else { 3641 /* Get request. Allocate sliding window buffer */ 3642 size = (valsize<CTL3_SMALLBUF) ? valsize:CTL3_SMALLBUF; 3643 3644 if (size < valsize) { 3645 /* We have to wire user buffer */ 3646 error = vslock(sopt->sopt_val, valsize); 3647 if (error != 0) 3648 return (error); 3649 locked = 1; 3650 } 3651 } 3652 3653 sdata.kbuf = malloc(size, M_TEMP, M_WAITOK | M_ZERO); 3654 sdata.ksize = size; 3655 sdata.kavail = size; 3656 } 3657 3658 sdata.sopt = sopt; 3659 sdata.sopt_val = sopt->sopt_val; 3660 sdata.valsize = valsize; 3661 3662 /* 3663 * Copy either all request (if valsize < bsize_max) 3664 * or first bsize_max bytes to guarantee most consumers 3665 * that all necessary data has been copied). 3666 * Anyway, copy not less than sizeof(ip_fw3_opheader). 3667 */ 3668 if ((error = sooptcopyin(sopt, sdata.kbuf, sdata.ksize, 3669 sizeof(ip_fw3_opheader))) != 0) 3670 return (error); 3671 op3 = (ip_fw3_opheader *)sdata.kbuf; 3672 3673 /* Finally, run handler */ 3674 error = h.handler(chain, op3, &sdata); 3675 find_unref_sh(&h); 3676 3677 /* Flush state and free buffers */ 3678 if (error == 0) 3679 error = ipfw_flush_sopt_data(&sdata); 3680 else 3681 ipfw_flush_sopt_data(&sdata); 3682 3683 if (locked != 0) 3684 vsunlock(sdata.sopt_val, valsize); 3685 3686 /* Restore original pointer and set number of bytes written */ 3687 sopt->sopt_val = sdata.sopt_val; 3688 sopt->sopt_valsize = sdata.ktotal; 3689 if (sdata.kbuf != xbuf) 3690 free(sdata.kbuf, M_TEMP); 3691 3692 return (error); 3693} 3694 3695/** 3696 * {set|get}sockopt parser. 3697 */ 3698int 3699ipfw_ctl(struct sockopt *sopt) 3700{ 3701#define RULE_MAXSIZE (512*sizeof(u_int32_t)) 3702 int error; 3703 size_t size, valsize; 3704 struct ip_fw *buf; 3705 struct ip_fw_rule0 *rule; 3706 struct ip_fw_chain *chain; 3707 u_int32_t rulenum[2]; 3708 uint32_t opt; 3709 struct rule_check_info ci; 3710 IPFW_RLOCK_TRACKER; 3711 3712 chain = &V_layer3_chain; 3713 error = 0; 3714 3715 /* Save original valsize before it is altered via sooptcopyin() */ 3716 valsize = sopt->sopt_valsize; 3717 opt = sopt->sopt_name; 3718 3719 /* 3720 * Disallow modifications in really-really secure mode, but still allow 3721 * the logging counters to be reset. 3722 */ 3723 if (opt == IP_FW_ADD || 3724 (sopt->sopt_dir == SOPT_SET && opt != IP_FW_RESETLOG)) { 3725 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3726 if (error != 0) 3727 return (error); 3728 } 3729 3730 switch (opt) { 3731 case IP_FW_GET: 3732 /* 3733 * pass up a copy of the current rules. Static rules 3734 * come first (the last of which has number IPFW_DEFAULT_RULE), 3735 * followed by a possibly empty list of dynamic rule. 3736 * The last dynamic rule has NULL in the "next" field. 3737 * 3738 * Note that the calculated size is used to bound the 3739 * amount of data returned to the user. The rule set may 3740 * change between calculating the size and returning the 3741 * data in which case we'll just return what fits. 3742 */ 3743 for (;;) { 3744 int len = 0, want; 3745 3746 size = chain->static_len; 3747 size += ipfw_dyn_len(); 3748 if (size >= sopt->sopt_valsize) 3749 break; 3750 buf = malloc(size, M_TEMP, M_WAITOK | M_ZERO); 3751 IPFW_UH_RLOCK(chain); 3752 /* check again how much space we need */ 3753 want = chain->static_len + ipfw_dyn_len(); 3754 if (size >= want) 3755 len = ipfw_getrules(chain, buf, size); 3756 IPFW_UH_RUNLOCK(chain); 3757 if (size >= want) 3758 error = sooptcopyout(sopt, buf, len); 3759 free(buf, M_TEMP); 3760 if (size >= want) 3761 break; 3762 } 3763 break; 3764 3765 case IP_FW_FLUSH: 3766 /* locking is done within del_entry() */ 3767 error = del_entry(chain, 0); /* special case, rule=0, cmd=0 means all */ 3768 break; 3769 3770 case IP_FW_ADD: 3771 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 3772 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 3773 sizeof(struct ip_fw7) ); 3774 3775 memset(&ci, 0, sizeof(struct rule_check_info)); 3776 3777 /* 3778 * If the size of commands equals RULESIZE7 then we assume 3779 * a FreeBSD7.2 binary is talking to us (set is7=1). 3780 * is7 is persistent so the next 'ipfw list' command 3781 * will use this format. 3782 * NOTE: If wrong version is guessed (this can happen if 3783 * the first ipfw command is 'ipfw [pipe] list') 3784 * the ipfw binary may crash or loop infinitly... 3785 */ 3786 size = sopt->sopt_valsize; 3787 if (size == RULESIZE7(rule)) { 3788 is7 = 1; 3789 error = convert_rule_to_8(rule); 3790 if (error) { 3791 free(rule, M_TEMP); 3792 return error; 3793 } 3794 size = RULESIZE(rule); 3795 } else 3796 is7 = 0; 3797 if (error == 0) 3798 error = check_ipfw_rule0(rule, size, &ci); 3799 if (error == 0) { 3800 /* locking is done within add_rule() */ 3801 struct ip_fw *krule; 3802 krule = ipfw_alloc_rule(chain, RULEKSIZE0(rule)); 3803 ci.urule = (caddr_t)rule; 3804 ci.krule = krule; 3805 import_rule0(&ci); 3806 error = commit_rules(chain, &ci, 1); 3807 if (error != 0) 3808 free_rule(ci.krule); 3809 else if (sopt->sopt_dir == SOPT_GET) { 3810 if (is7) { 3811 error = convert_rule_to_7(rule); 3812 size = RULESIZE7(rule); 3813 if (error) { 3814 free(rule, M_TEMP); 3815 return error; 3816 } 3817 } 3818 error = sooptcopyout(sopt, rule, size); 3819 } 3820 } 3821 free(rule, M_TEMP); 3822 break; 3823 3824 case IP_FW_DEL: 3825 /* 3826 * IP_FW_DEL is used for deleting single rules or sets, 3827 * and (ab)used to atomically manipulate sets. Argument size 3828 * is used to distinguish between the two: 3829 * sizeof(u_int32_t) 3830 * delete single rule or set of rules, 3831 * or reassign rules (or sets) to a different set. 3832 * 2*sizeof(u_int32_t) 3833 * atomic disable/enable sets. 3834 * first u_int32_t contains sets to be disabled, 3835 * second u_int32_t contains sets to be enabled. 3836 */ 3837 error = sooptcopyin(sopt, rulenum, 3838 2*sizeof(u_int32_t), sizeof(u_int32_t)); 3839 if (error) 3840 break; 3841 size = sopt->sopt_valsize; 3842 if (size == sizeof(u_int32_t) && rulenum[0] != 0) { 3843 /* delete or reassign, locking done in del_entry() */ 3844 error = del_entry(chain, rulenum[0]); 3845 } else if (size == 2*sizeof(u_int32_t)) { /* set enable/disable */ 3846 IPFW_UH_WLOCK(chain); 3847 V_set_disable = 3848 (V_set_disable | rulenum[0]) & ~rulenum[1] & 3849 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 3850 IPFW_UH_WUNLOCK(chain); 3851 } else 3852 error = EINVAL; 3853 break; 3854 3855 case IP_FW_ZERO: 3856 case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */ 3857 rulenum[0] = 0; 3858 if (sopt->sopt_val != 0) { 3859 error = sooptcopyin(sopt, rulenum, 3860 sizeof(u_int32_t), sizeof(u_int32_t)); 3861 if (error) 3862 break; 3863 } 3864 error = zero_entry(chain, rulenum[0], 3865 sopt->sopt_name == IP_FW_RESETLOG); 3866 break; 3867 3868 /*--- TABLE opcodes ---*/ 3869 case IP_FW_TABLE_ADD: 3870 case IP_FW_TABLE_DEL: 3871 { 3872 ipfw_table_entry ent; 3873 struct tentry_info tei; 3874 struct tid_info ti; 3875 struct table_value v; 3876 3877 error = sooptcopyin(sopt, &ent, 3878 sizeof(ent), sizeof(ent)); 3879 if (error) 3880 break; 3881 3882 memset(&tei, 0, sizeof(tei)); 3883 tei.paddr = &ent.addr; 3884 tei.subtype = AF_INET; 3885 tei.masklen = ent.masklen; 3886 ipfw_import_table_value_legacy(ent.value, &v); 3887 tei.pvalue = &v; 3888 memset(&ti, 0, sizeof(ti)); 3889 ti.uidx = ent.tbl; 3890 ti.type = IPFW_TABLE_CIDR; 3891 3892 error = (opt == IP_FW_TABLE_ADD) ? 3893 add_table_entry(chain, &ti, &tei, 0, 1) : 3894 del_table_entry(chain, &ti, &tei, 0, 1); 3895 } 3896 break; 3897 3898 3899 case IP_FW_TABLE_FLUSH: 3900 { 3901 u_int16_t tbl; 3902 struct tid_info ti; 3903 3904 error = sooptcopyin(sopt, &tbl, 3905 sizeof(tbl), sizeof(tbl)); 3906 if (error) 3907 break; 3908 memset(&ti, 0, sizeof(ti)); 3909 ti.uidx = tbl; 3910 error = flush_table(chain, &ti); 3911 } 3912 break; 3913 3914 case IP_FW_TABLE_GETSIZE: 3915 { 3916 u_int32_t tbl, cnt; 3917 struct tid_info ti; 3918 3919 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 3920 sizeof(tbl)))) 3921 break; 3922 memset(&ti, 0, sizeof(ti)); 3923 ti.uidx = tbl; 3924 IPFW_RLOCK(chain); 3925 error = ipfw_count_table(chain, &ti, &cnt); 3926 IPFW_RUNLOCK(chain); 3927 if (error) 3928 break; 3929 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 3930 } 3931 break; 3932 3933 case IP_FW_TABLE_LIST: 3934 { 3935 ipfw_table *tbl; 3936 struct tid_info ti; 3937 3938 if (sopt->sopt_valsize < sizeof(*tbl)) { 3939 error = EINVAL; 3940 break; 3941 } 3942 size = sopt->sopt_valsize; 3943 tbl = malloc(size, M_TEMP, M_WAITOK); 3944 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 3945 if (error) { 3946 free(tbl, M_TEMP); 3947 break; 3948 } 3949 tbl->size = (size - sizeof(*tbl)) / 3950 sizeof(ipfw_table_entry); 3951 memset(&ti, 0, sizeof(ti)); 3952 ti.uidx = tbl->tbl; 3953 IPFW_RLOCK(chain); 3954 error = ipfw_dump_table_legacy(chain, &ti, tbl); 3955 IPFW_RUNLOCK(chain); 3956 if (error) { 3957 free(tbl, M_TEMP); 3958 break; 3959 } 3960 error = sooptcopyout(sopt, tbl, size); 3961 free(tbl, M_TEMP); 3962 } 3963 break; 3964 3965 /*--- NAT operations are protected by the IPFW_LOCK ---*/ 3966 case IP_FW_NAT_CFG: 3967 if (IPFW_NAT_LOADED) 3968 error = ipfw_nat_cfg_ptr(sopt); 3969 else { 3970 printf("IP_FW_NAT_CFG: %s\n", 3971 "ipfw_nat not present, please load it"); 3972 error = EINVAL; 3973 } 3974 break; 3975 3976 case IP_FW_NAT_DEL: 3977 if (IPFW_NAT_LOADED) 3978 error = ipfw_nat_del_ptr(sopt); 3979 else { 3980 printf("IP_FW_NAT_DEL: %s\n", 3981 "ipfw_nat not present, please load it"); 3982 error = EINVAL; 3983 } 3984 break; 3985 3986 case IP_FW_NAT_GET_CONFIG: 3987 if (IPFW_NAT_LOADED) 3988 error = ipfw_nat_get_cfg_ptr(sopt); 3989 else { 3990 printf("IP_FW_NAT_GET_CFG: %s\n", 3991 "ipfw_nat not present, please load it"); 3992 error = EINVAL; 3993 } 3994 break; 3995 3996 case IP_FW_NAT_GET_LOG: 3997 if (IPFW_NAT_LOADED) 3998 error = ipfw_nat_get_log_ptr(sopt); 3999 else { 4000 printf("IP_FW_NAT_GET_LOG: %s\n", 4001 "ipfw_nat not present, please load it"); 4002 error = EINVAL; 4003 } 4004 break; 4005 4006 default: 4007 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4008 error = EINVAL; 4009 } 4010 4011 return (error); 4012#undef RULE_MAXSIZE 4013} 4014#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 4015 4016/* Functions to convert rules 7.2 <==> 8.0 */ 4017static int 4018convert_rule_to_7(struct ip_fw_rule0 *rule) 4019{ 4020 /* Used to modify original rule */ 4021 struct ip_fw7 *rule7 = (struct ip_fw7 *)rule; 4022 /* copy of original rule, version 8 */ 4023 struct ip_fw_rule0 *tmp; 4024 4025 /* Used to copy commands */ 4026 ipfw_insn *ccmd, *dst; 4027 int ll = 0, ccmdlen = 0; 4028 4029 tmp = malloc(RULE_MAXSIZE, M_TEMP, M_NOWAIT | M_ZERO); 4030 if (tmp == NULL) { 4031 return 1; //XXX error 4032 } 4033 bcopy(rule, tmp, RULE_MAXSIZE); 4034 4035 /* Copy fields */ 4036 //rule7->_pad = tmp->_pad; 4037 rule7->set = tmp->set; 4038 rule7->rulenum = tmp->rulenum; 4039 rule7->cmd_len = tmp->cmd_len; 4040 rule7->act_ofs = tmp->act_ofs; 4041 rule7->next_rule = (struct ip_fw7 *)tmp->next_rule; 4042 rule7->cmd_len = tmp->cmd_len; 4043 rule7->pcnt = tmp->pcnt; 4044 rule7->bcnt = tmp->bcnt; 4045 rule7->timestamp = tmp->timestamp; 4046 4047 /* Copy commands */ 4048 for (ll = tmp->cmd_len, ccmd = tmp->cmd, dst = rule7->cmd ; 4049 ll > 0 ; ll -= ccmdlen, ccmd += ccmdlen, dst += ccmdlen) { 4050 ccmdlen = F_LEN(ccmd); 4051 4052 bcopy(ccmd, dst, F_LEN(ccmd)*sizeof(uint32_t)); 4053 4054 if (dst->opcode > O_NAT) 4055 /* O_REASS doesn't exists in 7.2 version, so 4056 * decrement opcode if it is after O_REASS 4057 */ 4058 dst->opcode--; 4059 4060 if (ccmdlen > ll) { 4061 printf("ipfw: opcode %d size truncated\n", 4062 ccmd->opcode); 4063 return EINVAL; 4064 } 4065 } 4066 free(tmp, M_TEMP); 4067 4068 return 0; 4069} 4070 4071static int 4072convert_rule_to_8(struct ip_fw_rule0 *rule) 4073{ 4074 /* Used to modify original rule */ 4075 struct ip_fw7 *rule7 = (struct ip_fw7 *) rule; 4076 4077 /* Used to copy commands */ 4078 ipfw_insn *ccmd, *dst; 4079 int ll = 0, ccmdlen = 0; 4080 4081 /* Copy of original rule */ 4082 struct ip_fw7 *tmp = malloc(RULE_MAXSIZE, M_TEMP, M_NOWAIT | M_ZERO); 4083 if (tmp == NULL) { 4084 return 1; //XXX error 4085 } 4086 4087 bcopy(rule7, tmp, RULE_MAXSIZE); 4088 4089 for (ll = tmp->cmd_len, ccmd = tmp->cmd, dst = rule->cmd ; 4090 ll > 0 ; ll -= ccmdlen, ccmd += ccmdlen, dst += ccmdlen) { 4091 ccmdlen = F_LEN(ccmd); 4092 4093 bcopy(ccmd, dst, F_LEN(ccmd)*sizeof(uint32_t)); 4094 4095 if (dst->opcode > O_NAT) 4096 /* O_REASS doesn't exists in 7.2 version, so 4097 * increment opcode if it is after O_REASS 4098 */ 4099 dst->opcode++; 4100 4101 if (ccmdlen > ll) { 4102 printf("ipfw: opcode %d size truncated\n", 4103 ccmd->opcode); 4104 return EINVAL; 4105 } 4106 } 4107 4108 rule->_pad = tmp->_pad; 4109 rule->set = tmp->set; 4110 rule->rulenum = tmp->rulenum; 4111 rule->cmd_len = tmp->cmd_len; 4112 rule->act_ofs = tmp->act_ofs; 4113 rule->next_rule = (struct ip_fw *)tmp->next_rule; 4114 rule->cmd_len = tmp->cmd_len; 4115 rule->id = 0; /* XXX see if is ok = 0 */ 4116 rule->pcnt = tmp->pcnt; 4117 rule->bcnt = tmp->bcnt; 4118 rule->timestamp = tmp->timestamp; 4119 4120 free (tmp, M_TEMP); 4121 return 0; 4122} 4123 4124/* 4125 * Named object api 4126 * 4127 */ 4128 4129void 4130ipfw_init_srv(struct ip_fw_chain *ch) 4131{ 4132 4133 ch->srvmap = ipfw_objhash_create(IPFW_OBJECTS_DEFAULT); 4134 ch->srvstate = malloc(sizeof(void *) * IPFW_OBJECTS_DEFAULT, 4135 M_IPFW, M_WAITOK | M_ZERO); 4136} 4137 4138void 4139ipfw_destroy_srv(struct ip_fw_chain *ch) 4140{ 4141 4142 free(ch->srvstate, M_IPFW); 4143 ipfw_objhash_destroy(ch->srvmap); 4144} 4145 4146/* 4147 * Allocate new bitmask which can be used to enlarge/shrink 4148 * named instance index. 4149 */ 4150void 4151ipfw_objhash_bitmap_alloc(uint32_t items, void **idx, int *pblocks) 4152{ 4153 size_t size; 4154 int max_blocks; 4155 u_long *idx_mask; 4156 4157 KASSERT((items % BLOCK_ITEMS) == 0, 4158 ("bitmask size needs to power of 2 and greater or equal to %zu", 4159 BLOCK_ITEMS)); 4160 4161 max_blocks = items / BLOCK_ITEMS; 4162 size = items / 8; 4163 idx_mask = malloc(size * IPFW_MAX_SETS, M_IPFW, M_WAITOK); 4164 /* Mark all as free */ 4165 memset(idx_mask, 0xFF, size * IPFW_MAX_SETS); 4166 *idx_mask &= ~(u_long)1; /* Skip index 0 */ 4167 4168 *idx = idx_mask; 4169 *pblocks = max_blocks; 4170} 4171 4172/* 4173 * Copy current bitmask index to new one. 4174 */ 4175void 4176ipfw_objhash_bitmap_merge(struct namedobj_instance *ni, void **idx, int *blocks) 4177{ 4178 int old_blocks, new_blocks; 4179 u_long *old_idx, *new_idx; 4180 int i; 4181 4182 old_idx = ni->idx_mask; 4183 old_blocks = ni->max_blocks; 4184 new_idx = *idx; 4185 new_blocks = *blocks; 4186 4187 for (i = 0; i < IPFW_MAX_SETS; i++) { 4188 memcpy(&new_idx[new_blocks * i], &old_idx[old_blocks * i], 4189 old_blocks * sizeof(u_long)); 4190 } 4191} 4192 4193/* 4194 * Swaps current @ni index with new one. 4195 */ 4196void 4197ipfw_objhash_bitmap_swap(struct namedobj_instance *ni, void **idx, int *blocks) 4198{ 4199 int old_blocks; 4200 u_long *old_idx; 4201 4202 old_idx = ni->idx_mask; 4203 old_blocks = ni->max_blocks; 4204 4205 ni->idx_mask = *idx; 4206 ni->max_blocks = *blocks; 4207 4208 /* Save old values */ 4209 *idx = old_idx; 4210 *blocks = old_blocks; 4211} 4212 4213void 4214ipfw_objhash_bitmap_free(void *idx, int blocks) 4215{ 4216 4217 free(idx, M_IPFW); 4218} 4219 4220/* 4221 * Creates named hash instance. 4222 * Must be called without holding any locks. 4223 * Return pointer to new instance. 4224 */ 4225struct namedobj_instance * 4226ipfw_objhash_create(uint32_t items) 4227{ 4228 struct namedobj_instance *ni; 4229 int i; 4230 size_t size; 4231 4232 size = sizeof(struct namedobj_instance) + 4233 sizeof(struct namedobjects_head) * NAMEDOBJ_HASH_SIZE + 4234 sizeof(struct namedobjects_head) * NAMEDOBJ_HASH_SIZE; 4235 4236 ni = malloc(size, M_IPFW, M_WAITOK | M_ZERO); 4237 ni->nn_size = NAMEDOBJ_HASH_SIZE; 4238 ni->nv_size = NAMEDOBJ_HASH_SIZE; 4239 4240 ni->names = (struct namedobjects_head *)(ni +1); 4241 ni->values = &ni->names[ni->nn_size]; 4242 4243 for (i = 0; i < ni->nn_size; i++) 4244 TAILQ_INIT(&ni->names[i]); 4245 4246 for (i = 0; i < ni->nv_size; i++) 4247 TAILQ_INIT(&ni->values[i]); 4248 4249 /* Set default hashing/comparison functions */ 4250 ni->hash_f = objhash_hash_name; 4251 ni->cmp_f = objhash_cmp_name; 4252 4253 /* Allocate bitmask separately due to possible resize */ 4254 ipfw_objhash_bitmap_alloc(items, (void*)&ni->idx_mask, &ni->max_blocks); 4255 4256 return (ni); 4257} 4258 4259void 4260ipfw_objhash_destroy(struct namedobj_instance *ni) 4261{ 4262 4263 free(ni->idx_mask, M_IPFW); 4264 free(ni, M_IPFW); 4265} 4266 4267void 4268ipfw_objhash_set_funcs(struct namedobj_instance *ni, objhash_hash_f *hash_f, 4269 objhash_cmp_f *cmp_f) 4270{ 4271 4272 ni->hash_f = hash_f; 4273 ni->cmp_f = cmp_f; 4274} 4275 4276static uint32_t 4277objhash_hash_name(struct namedobj_instance *ni, const void *name, uint32_t set) 4278{ 4279 4280 return (fnv_32_str((const char *)name, FNV1_32_INIT)); 4281} 4282 4283static int 4284objhash_cmp_name(struct named_object *no, const void *name, uint32_t set) 4285{ 4286 4287 if ((strcmp(no->name, (const char *)name) == 0) && (no->set == set)) 4288 return (0); 4289 4290 return (1); 4291} 4292 4293static uint32_t 4294objhash_hash_idx(struct namedobj_instance *ni, uint32_t val) 4295{ 4296 uint32_t v; 4297 4298 v = val % (ni->nv_size - 1); 4299 4300 return (v); 4301} 4302 4303struct named_object * 4304ipfw_objhash_lookup_name(struct namedobj_instance *ni, uint32_t set, char *name) 4305{ 4306 struct named_object *no; 4307 uint32_t hash; 4308 4309 hash = ni->hash_f(ni, name, set) % ni->nn_size; 4310 4311 TAILQ_FOREACH(no, &ni->names[hash], nn_next) { 4312 if (ni->cmp_f(no, name, set) == 0) 4313 return (no); 4314 } 4315 4316 return (NULL); 4317} 4318 4319/* 4320 * Find named object by @uid. 4321 * Check @tlvs for valid data inside. 4322 * 4323 * Returns pointer to found TLV or NULL. 4324 */ 4325ipfw_obj_ntlv * 4326ipfw_find_name_tlv_type(void *tlvs, int len, uint16_t uidx, uint32_t etlv) 4327{ 4328 ipfw_obj_ntlv *ntlv; 4329 uintptr_t pa, pe; 4330 int l; 4331 4332 pa = (uintptr_t)tlvs; 4333 pe = pa + len; 4334 l = 0; 4335 for (; pa < pe; pa += l) { 4336 ntlv = (ipfw_obj_ntlv *)pa; 4337 l = ntlv->head.length; 4338 4339 if (l != sizeof(*ntlv)) 4340 return (NULL); 4341 4342 if (ntlv->idx != uidx) 4343 continue; 4344 /* 4345 * When userland has specified zero TLV type, do 4346 * not compare it with eltv. In some cases userland 4347 * doesn't know what type should it have. Use only 4348 * uidx and name for search named_object. 4349 */ 4350 if (ntlv->head.type != 0 && 4351 ntlv->head.type != (uint16_t)etlv) 4352 continue; 4353 4354 if (ipfw_check_object_name_generic(ntlv->name) != 0) 4355 return (NULL); 4356 4357 return (ntlv); 4358 } 4359 4360 return (NULL); 4361} 4362 4363/* 4364 * Finds object config based on either legacy index 4365 * or name in ntlv. 4366 * Note @ti structure contains unchecked data from userland. 4367 * 4368 * Returns 0 in success and fills in @pno with found config 4369 */ 4370int 4371ipfw_objhash_find_type(struct namedobj_instance *ni, struct tid_info *ti, 4372 uint32_t etlv, struct named_object **pno) 4373{ 4374 char *name; 4375 ipfw_obj_ntlv *ntlv; 4376 uint32_t set; 4377 4378 if (ti->tlvs == NULL) 4379 return (EINVAL); 4380 4381 ntlv = ipfw_find_name_tlv_type(ti->tlvs, ti->tlen, ti->uidx, etlv); 4382 if (ntlv == NULL) 4383 return (EINVAL); 4384 name = ntlv->name; 4385 4386 /* 4387 * Use set provided by @ti instead of @ntlv one. 4388 * This is needed due to different sets behavior 4389 * controlled by V_fw_tables_sets. 4390 */ 4391 set = ti->set; 4392 *pno = ipfw_objhash_lookup_name(ni, set, name); 4393 if (*pno == NULL) 4394 return (ESRCH); 4395 return (0); 4396} 4397 4398/* 4399 * Find named object by name, considering also its TLV type. 4400 */ 4401struct named_object * 4402ipfw_objhash_lookup_name_type(struct namedobj_instance *ni, uint32_t set, 4403 uint32_t type, const char *name) 4404{ 4405 struct named_object *no; 4406 uint32_t hash; 4407 4408 hash = ni->hash_f(ni, name, set) % ni->nn_size; 4409 4410 TAILQ_FOREACH(no, &ni->names[hash], nn_next) { 4411 if (ni->cmp_f(no, name, set) == 0 && 4412 no->etlv == (uint16_t)type) 4413 return (no); 4414 } 4415 4416 return (NULL); 4417} 4418 4419struct named_object * 4420ipfw_objhash_lookup_kidx(struct namedobj_instance *ni, uint16_t kidx) 4421{ 4422 struct named_object *no; 4423 uint32_t hash; 4424 4425 hash = objhash_hash_idx(ni, kidx); 4426 4427 TAILQ_FOREACH(no, &ni->values[hash], nv_next) { 4428 if (no->kidx == kidx) 4429 return (no); 4430 } 4431 4432 return (NULL); 4433} 4434 4435int 4436ipfw_objhash_same_name(struct namedobj_instance *ni, struct named_object *a, 4437 struct named_object *b) 4438{ 4439 4440 if ((strcmp(a->name, b->name) == 0) && a->set == b->set) 4441 return (1); 4442 4443 return (0); 4444} 4445 4446void 4447ipfw_objhash_add(struct namedobj_instance *ni, struct named_object *no) 4448{ 4449 uint32_t hash; 4450 4451 hash = ni->hash_f(ni, no->name, no->set) % ni->nn_size; 4452 TAILQ_INSERT_HEAD(&ni->names[hash], no, nn_next); 4453 4454 hash = objhash_hash_idx(ni, no->kidx); 4455 TAILQ_INSERT_HEAD(&ni->values[hash], no, nv_next); 4456 4457 ni->count++; 4458} 4459 4460void 4461ipfw_objhash_del(struct namedobj_instance *ni, struct named_object *no) 4462{ 4463 uint32_t hash; 4464 4465 hash = ni->hash_f(ni, no->name, no->set) % ni->nn_size; 4466 TAILQ_REMOVE(&ni->names[hash], no, nn_next); 4467 4468 hash = objhash_hash_idx(ni, no->kidx); 4469 TAILQ_REMOVE(&ni->values[hash], no, nv_next); 4470 4471 ni->count--; 4472} 4473 4474uint32_t 4475ipfw_objhash_count(struct namedobj_instance *ni) 4476{ 4477 4478 return (ni->count); 4479} 4480 4481uint32_t 4482ipfw_objhash_count_type(struct namedobj_instance *ni, uint16_t type) 4483{ 4484 struct named_object *no; 4485 uint32_t count; 4486 int i; 4487 4488 count = 0; 4489 for (i = 0; i < ni->nn_size; i++) { 4490 TAILQ_FOREACH(no, &ni->names[i], nn_next) { 4491 if (no->etlv == type) 4492 count++; 4493 } 4494 } 4495 return (count); 4496} 4497 4498/* 4499 * Runs @func for each found named object. 4500 * It is safe to delete objects from callback 4501 */ 4502int 4503ipfw_objhash_foreach(struct namedobj_instance *ni, objhash_cb_t *f, void *arg) 4504{ 4505 struct named_object *no, *no_tmp; 4506 int i, ret; 4507 4508 for (i = 0; i < ni->nn_size; i++) { 4509 TAILQ_FOREACH_SAFE(no, &ni->names[i], nn_next, no_tmp) { 4510 ret = f(ni, no, arg); 4511 if (ret != 0) 4512 return (ret); 4513 } 4514 } 4515 return (0); 4516} 4517 4518/* 4519 * Runs @f for each found named object with type @type. 4520 * It is safe to delete objects from callback 4521 */ 4522int 4523ipfw_objhash_foreach_type(struct namedobj_instance *ni, objhash_cb_t *f, 4524 void *arg, uint16_t type) 4525{ 4526 struct named_object *no, *no_tmp; 4527 int i, ret; 4528 4529 for (i = 0; i < ni->nn_size; i++) { 4530 TAILQ_FOREACH_SAFE(no, &ni->names[i], nn_next, no_tmp) { 4531 if (no->etlv != type) 4532 continue; 4533 ret = f(ni, no, arg); 4534 if (ret != 0) 4535 return (ret); 4536 } 4537 } 4538 return (0); 4539} 4540 4541/* 4542 * Removes index from given set. 4543 * Returns 0 on success. 4544 */ 4545int 4546ipfw_objhash_free_idx(struct namedobj_instance *ni, uint16_t idx) 4547{ 4548 u_long *mask; 4549 int i, v; 4550 4551 i = idx / BLOCK_ITEMS; 4552 v = idx % BLOCK_ITEMS; 4553 4554 if (i >= ni->max_blocks) 4555 return (1); 4556 4557 mask = &ni->idx_mask[i]; 4558 4559 if ((*mask & ((u_long)1 << v)) != 0) 4560 return (1); 4561 4562 /* Mark as free */ 4563 *mask |= (u_long)1 << v; 4564 4565 /* Update free offset */ 4566 if (ni->free_off[0] > i) 4567 ni->free_off[0] = i; 4568 4569 return (0); 4570} 4571 4572/* 4573 * Allocate new index in given instance and stores in in @pidx. 4574 * Returns 0 on success. 4575 */ 4576int 4577ipfw_objhash_alloc_idx(void *n, uint16_t *pidx) 4578{ 4579 struct namedobj_instance *ni; 4580 u_long *mask; 4581 int i, off, v; 4582 4583 ni = (struct namedobj_instance *)n; 4584 4585 off = ni->free_off[0]; 4586 mask = &ni->idx_mask[off]; 4587 4588 for (i = off; i < ni->max_blocks; i++, mask++) { 4589 if ((v = ffsl(*mask)) == 0) 4590 continue; 4591 4592 /* Mark as busy */ 4593 *mask &= ~ ((u_long)1 << (v - 1)); 4594 4595 ni->free_off[0] = i; 4596 4597 v = BLOCK_ITEMS * i + v - 1; 4598 4599 *pidx = v; 4600 return (0); 4601 } 4602 4603 return (1); 4604} 4605 4606/* end of file */ 4607