ip_fw_sockopt.c revision 315532
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 315532 2017-03-19 07:34:19Z ae $"); 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 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) 1825 goto bad_size; 1826 case O_IP_DST_LOOKUP: 1827 if (cmd->arg1 >= V_fw_tables_max) { 1828 printf("ipfw: invalid table number %d\n", 1829 cmd->arg1); 1830 return (EINVAL); 1831 } 1832 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 1833 cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1 && 1834 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 1835 goto bad_size; 1836 ci->object_opcodes++; 1837 break; 1838 case O_IP_FLOW_LOOKUP: 1839 if (cmd->arg1 >= V_fw_tables_max) { 1840 printf("ipfw: invalid table number %d\n", 1841 cmd->arg1); 1842 return (EINVAL); 1843 } 1844 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 1845 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 1846 goto bad_size; 1847 ci->object_opcodes++; 1848 break; 1849 case O_MACADDR2: 1850 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 1851 goto bad_size; 1852 break; 1853 1854 case O_NOP: 1855 case O_IPID: 1856 case O_IPTTL: 1857 case O_IPLEN: 1858 case O_TCPDATALEN: 1859 case O_TCPWIN: 1860 case O_TAGGED: 1861 if (cmdlen < 1 || cmdlen > 31) 1862 goto bad_size; 1863 break; 1864 1865 case O_DSCP: 1866 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1) 1867 goto bad_size; 1868 break; 1869 1870 case O_MAC_TYPE: 1871 case O_IP_SRCPORT: 1872 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 1873 if (cmdlen < 2 || cmdlen > 31) 1874 goto bad_size; 1875 break; 1876 1877 case O_RECV: 1878 case O_XMIT: 1879 case O_VIA: 1880 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 1881 goto bad_size; 1882 ci->object_opcodes++; 1883 break; 1884 1885 case O_ALTQ: 1886 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 1887 goto bad_size; 1888 break; 1889 1890 case O_PIPE: 1891 case O_QUEUE: 1892 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 1893 goto bad_size; 1894 goto check_action; 1895 1896 case O_FORWARD_IP: 1897 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 1898 goto bad_size; 1899 goto check_action; 1900#ifdef INET6 1901 case O_FORWARD_IP6: 1902 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa6)) 1903 goto bad_size; 1904 goto check_action; 1905#endif /* INET6 */ 1906 1907 case O_DIVERT: 1908 case O_TEE: 1909 if (ip_divert_ptr == NULL) 1910 return EINVAL; 1911 else 1912 goto check_size; 1913 case O_NETGRAPH: 1914 case O_NGTEE: 1915 if (ng_ipfw_input_p == NULL) 1916 return EINVAL; 1917 else 1918 goto check_size; 1919 case O_NAT: 1920 if (!IPFW_NAT_LOADED) 1921 return EINVAL; 1922 if (cmdlen != F_INSN_SIZE(ipfw_insn_nat)) 1923 goto bad_size; 1924 goto check_action; 1925 case O_FORWARD_MAC: /* XXX not implemented yet */ 1926 case O_CHECK_STATE: 1927 case O_COUNT: 1928 case O_ACCEPT: 1929 case O_DENY: 1930 case O_REJECT: 1931 case O_SETDSCP: 1932#ifdef INET6 1933 case O_UNREACH6: 1934#endif 1935 case O_SKIPTO: 1936 case O_REASS: 1937 case O_CALLRETURN: 1938check_size: 1939 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 1940 goto bad_size; 1941check_action: 1942 if (have_action) { 1943 printf("ipfw: opcode %d, multiple actions" 1944 " not allowed\n", 1945 cmd->opcode); 1946 return (EINVAL); 1947 } 1948 have_action = 1; 1949 if (l != cmdlen) { 1950 printf("ipfw: opcode %d, action must be" 1951 " last opcode\n", 1952 cmd->opcode); 1953 return (EINVAL); 1954 } 1955 break; 1956#ifdef INET6 1957 case O_IP6_SRC: 1958 case O_IP6_DST: 1959 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 1960 F_INSN_SIZE(ipfw_insn)) 1961 goto bad_size; 1962 break; 1963 1964 case O_FLOW6ID: 1965 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1966 ((ipfw_insn_u32 *)cmd)->o.arg1) 1967 goto bad_size; 1968 break; 1969 1970 case O_IP6_SRC_MASK: 1971 case O_IP6_DST_MASK: 1972 if ( !(cmdlen & 1) || cmdlen > 127) 1973 goto bad_size; 1974 break; 1975 case O_ICMP6TYPE: 1976 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 1977 goto bad_size; 1978 break; 1979#endif 1980 1981 default: 1982 switch (cmd->opcode) { 1983#ifndef INET6 1984 case O_IP6_SRC_ME: 1985 case O_IP6_DST_ME: 1986 case O_EXT_HDR: 1987 case O_IP6: 1988 case O_UNREACH6: 1989 case O_IP6_SRC: 1990 case O_IP6_DST: 1991 case O_FLOW6ID: 1992 case O_IP6_SRC_MASK: 1993 case O_IP6_DST_MASK: 1994 case O_ICMP6TYPE: 1995 printf("ipfw: no IPv6 support in kernel\n"); 1996 return (EPROTONOSUPPORT); 1997#endif 1998 default: 1999 printf("ipfw: opcode %d, unknown opcode\n", 2000 cmd->opcode); 2001 return (EINVAL); 2002 } 2003 } 2004 } 2005 if (have_action == 0) { 2006 printf("ipfw: missing action\n"); 2007 return (EINVAL); 2008 } 2009 return 0; 2010 2011bad_size: 2012 printf("ipfw: opcode %d size %d wrong\n", 2013 cmd->opcode, cmdlen); 2014 return (EINVAL); 2015} 2016 2017 2018/* 2019 * Translation of requests for compatibility with FreeBSD 7.2/8. 2020 * a static variable tells us if we have an old client from userland, 2021 * and if necessary we translate requests and responses between the 2022 * two formats. 2023 */ 2024static int is7 = 0; 2025 2026struct ip_fw7 { 2027 struct ip_fw7 *next; /* linked list of rules */ 2028 struct ip_fw7 *next_rule; /* ptr to next [skipto] rule */ 2029 /* 'next_rule' is used to pass up 'set_disable' status */ 2030 2031 uint16_t act_ofs; /* offset of action in 32-bit units */ 2032 uint16_t cmd_len; /* # of 32-bit words in cmd */ 2033 uint16_t rulenum; /* rule number */ 2034 uint8_t set; /* rule set (0..31) */ 2035 // #define RESVD_SET 31 /* set for default and persistent rules */ 2036 uint8_t _pad; /* padding */ 2037 // uint32_t id; /* rule id, only in v.8 */ 2038 /* These fields are present in all rules. */ 2039 uint64_t pcnt; /* Packet counter */ 2040 uint64_t bcnt; /* Byte counter */ 2041 uint32_t timestamp; /* tv_sec of last match */ 2042 2043 ipfw_insn cmd[1]; /* storage for commands */ 2044}; 2045 2046static int convert_rule_to_7(struct ip_fw_rule0 *rule); 2047static int convert_rule_to_8(struct ip_fw_rule0 *rule); 2048 2049#ifndef RULESIZE7 2050#define RULESIZE7(rule) (sizeof(struct ip_fw7) + \ 2051 ((struct ip_fw7 *)(rule))->cmd_len * 4 - 4) 2052#endif 2053 2054 2055/* 2056 * Copy the static and dynamic rules to the supplied buffer 2057 * and return the amount of space actually used. 2058 * Must be run under IPFW_UH_RLOCK 2059 */ 2060static size_t 2061ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 2062{ 2063 char *bp = buf; 2064 char *ep = bp + space; 2065 struct ip_fw *rule; 2066 struct ip_fw_rule0 *dst; 2067 struct timeval boottime; 2068 int error, i, l, warnflag; 2069 time_t boot_seconds; 2070 2071 warnflag = 0; 2072 2073 getboottime(&boottime); 2074 boot_seconds = boottime.tv_sec; 2075 for (i = 0; i < chain->n_rules; i++) { 2076 rule = chain->map[i]; 2077 2078 if (is7) { 2079 /* Convert rule to FreeBSd 7.2 format */ 2080 l = RULESIZE7(rule); 2081 if (bp + l + sizeof(uint32_t) <= ep) { 2082 bcopy(rule, bp, l + sizeof(uint32_t)); 2083 error = set_legacy_obj_kidx(chain, 2084 (struct ip_fw_rule0 *)bp); 2085 if (error != 0) 2086 return (0); 2087 error = convert_rule_to_7((struct ip_fw_rule0 *) bp); 2088 if (error) 2089 return 0; /*XXX correct? */ 2090 /* 2091 * XXX HACK. Store the disable mask in the "next" 2092 * pointer in a wild attempt to keep the ABI the same. 2093 * Why do we do this on EVERY rule? 2094 */ 2095 bcopy(&V_set_disable, 2096 &(((struct ip_fw7 *)bp)->next_rule), 2097 sizeof(V_set_disable)); 2098 if (((struct ip_fw7 *)bp)->timestamp) 2099 ((struct ip_fw7 *)bp)->timestamp += boot_seconds; 2100 bp += l; 2101 } 2102 continue; /* go to next rule */ 2103 } 2104 2105 l = RULEUSIZE0(rule); 2106 if (bp + l > ep) { /* should not happen */ 2107 printf("overflow dumping static rules\n"); 2108 break; 2109 } 2110 dst = (struct ip_fw_rule0 *)bp; 2111 export_rule0(rule, dst, l); 2112 error = set_legacy_obj_kidx(chain, dst); 2113 2114 /* 2115 * XXX HACK. Store the disable mask in the "next" 2116 * pointer in a wild attempt to keep the ABI the same. 2117 * Why do we do this on EVERY rule? 2118 * 2119 * XXX: "ipfw set show" (ab)uses IP_FW_GET to read disabled mask 2120 * so we need to fail _after_ saving at least one mask. 2121 */ 2122 bcopy(&V_set_disable, &dst->next_rule, sizeof(V_set_disable)); 2123 if (dst->timestamp) 2124 dst->timestamp += boot_seconds; 2125 bp += l; 2126 2127 if (error != 0) { 2128 if (error == 2) { 2129 /* Non-fatal table rewrite error. */ 2130 warnflag = 1; 2131 continue; 2132 } 2133 printf("Stop on rule %d. Fail to convert table\n", 2134 rule->rulenum); 2135 break; 2136 } 2137 } 2138 if (warnflag != 0) 2139 printf("ipfw: process %s is using legacy interfaces," 2140 " consider rebuilding\n", ""); 2141 ipfw_get_dynamic(chain, &bp, ep); /* protected by the dynamic lock */ 2142 return (bp - (char *)buf); 2143} 2144 2145 2146struct dump_args { 2147 uint32_t b; /* start rule */ 2148 uint32_t e; /* end rule */ 2149 uint32_t rcount; /* number of rules */ 2150 uint32_t rsize; /* rules size */ 2151 uint32_t tcount; /* number of tables */ 2152 int rcounters; /* counters */ 2153}; 2154 2155void 2156ipfw_export_obj_ntlv(struct named_object *no, ipfw_obj_ntlv *ntlv) 2157{ 2158 2159 ntlv->head.type = no->etlv; 2160 ntlv->head.length = sizeof(*ntlv); 2161 ntlv->idx = no->kidx; 2162 strlcpy(ntlv->name, no->name, sizeof(ntlv->name)); 2163} 2164 2165/* 2166 * Export named object info in instance @ni, identified by @kidx 2167 * to ipfw_obj_ntlv. TLV is allocated from @sd space. 2168 * 2169 * Returns 0 on success. 2170 */ 2171static int 2172export_objhash_ntlv(struct namedobj_instance *ni, uint16_t kidx, 2173 struct sockopt_data *sd) 2174{ 2175 struct named_object *no; 2176 ipfw_obj_ntlv *ntlv; 2177 2178 no = ipfw_objhash_lookup_kidx(ni, kidx); 2179 KASSERT(no != NULL, ("invalid object kernel index passed")); 2180 2181 ntlv = (ipfw_obj_ntlv *)ipfw_get_sopt_space(sd, sizeof(*ntlv)); 2182 if (ntlv == NULL) 2183 return (ENOMEM); 2184 2185 ipfw_export_obj_ntlv(no, ntlv); 2186 return (0); 2187} 2188 2189/* 2190 * Dumps static rules with table TLVs in buffer @sd. 2191 * 2192 * Returns 0 on success. 2193 */ 2194static int 2195dump_static_rules(struct ip_fw_chain *chain, struct dump_args *da, 2196 uint32_t *bmask, struct sockopt_data *sd) 2197{ 2198 int error; 2199 int i, l; 2200 uint32_t tcount; 2201 ipfw_obj_ctlv *ctlv; 2202 struct ip_fw *krule; 2203 struct namedobj_instance *ni; 2204 caddr_t dst; 2205 2206 /* Dump table names first (if any) */ 2207 if (da->tcount > 0) { 2208 /* Header first */ 2209 ctlv = (ipfw_obj_ctlv *)ipfw_get_sopt_space(sd, sizeof(*ctlv)); 2210 if (ctlv == NULL) 2211 return (ENOMEM); 2212 ctlv->head.type = IPFW_TLV_TBLNAME_LIST; 2213 ctlv->head.length = da->tcount * sizeof(ipfw_obj_ntlv) + 2214 sizeof(*ctlv); 2215 ctlv->count = da->tcount; 2216 ctlv->objsize = sizeof(ipfw_obj_ntlv); 2217 } 2218 2219 i = 0; 2220 tcount = da->tcount; 2221 ni = ipfw_get_table_objhash(chain); 2222 while (tcount > 0) { 2223 if ((bmask[i / 32] & (1 << (i % 32))) == 0) { 2224 i++; 2225 continue; 2226 } 2227 2228 /* Jump to shared named object bitmask */ 2229 if (i >= IPFW_TABLES_MAX) { 2230 ni = CHAIN_TO_SRV(chain); 2231 i -= IPFW_TABLES_MAX; 2232 bmask += IPFW_TABLES_MAX / 32; 2233 } 2234 2235 if ((error = export_objhash_ntlv(ni, i, sd)) != 0) 2236 return (error); 2237 2238 i++; 2239 tcount--; 2240 } 2241 2242 /* Dump rules */ 2243 ctlv = (ipfw_obj_ctlv *)ipfw_get_sopt_space(sd, sizeof(*ctlv)); 2244 if (ctlv == NULL) 2245 return (ENOMEM); 2246 ctlv->head.type = IPFW_TLV_RULE_LIST; 2247 ctlv->head.length = da->rsize + sizeof(*ctlv); 2248 ctlv->count = da->rcount; 2249 2250 for (i = da->b; i < da->e; i++) { 2251 krule = chain->map[i]; 2252 2253 l = RULEUSIZE1(krule) + sizeof(ipfw_obj_tlv); 2254 if (da->rcounters != 0) 2255 l += sizeof(struct ip_fw_bcounter); 2256 dst = (caddr_t)ipfw_get_sopt_space(sd, l); 2257 if (dst == NULL) 2258 return (ENOMEM); 2259 2260 export_rule1(krule, dst, l, da->rcounters); 2261 } 2262 2263 return (0); 2264} 2265 2266/* 2267 * Marks every object index used in @rule with bit in @bmask. 2268 * Used to generate bitmask of referenced tables/objects for given ruleset 2269 * or its part. 2270 * 2271 * Returns number of newly-referenced objects. 2272 */ 2273static int 2274mark_object_kidx(struct ip_fw_chain *ch, struct ip_fw *rule, 2275 uint32_t *bmask) 2276{ 2277 struct opcode_obj_rewrite *rw; 2278 ipfw_insn *cmd; 2279 int bidx, cmdlen, l, count; 2280 uint16_t kidx; 2281 uint8_t subtype; 2282 2283 l = rule->cmd_len; 2284 cmd = rule->cmd; 2285 cmdlen = 0; 2286 count = 0; 2287 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 2288 cmdlen = F_LEN(cmd); 2289 2290 rw = find_op_rw(cmd, &kidx, &subtype); 2291 if (rw == NULL) 2292 continue; 2293 2294 bidx = kidx / 32; 2295 /* 2296 * Maintain separate bitmasks for table and 2297 * non-table objects. 2298 */ 2299 if (rw->etlv != IPFW_TLV_TBL_NAME) 2300 bidx += IPFW_TABLES_MAX / 32; 2301 2302 if ((bmask[bidx] & (1 << (kidx % 32))) == 0) 2303 count++; 2304 2305 bmask[bidx] |= 1 << (kidx % 32); 2306 } 2307 2308 return (count); 2309} 2310 2311/* 2312 * Dumps requested objects data 2313 * Data layout (version 0)(current): 2314 * Request: [ ipfw_cfg_lheader ] + IPFW_CFG_GET_* flags 2315 * size = ipfw_cfg_lheader.size 2316 * Reply: [ ipfw_cfg_lheader 2317 * [ ipfw_obj_ctlv(IPFW_TLV_TBL_LIST) ipfw_obj_ntlv x N ] (optional) 2318 * [ ipfw_obj_ctlv(IPFW_TLV_RULE_LIST) 2319 * ipfw_obj_tlv(IPFW_TLV_RULE_ENT) [ ip_fw_bcounter (optional) ip_fw_rule ] 2320 * ] (optional) 2321 * [ ipfw_obj_ctlv(IPFW_TLV_STATE_LIST) ipfw_obj_dyntlv x N ] (optional) 2322 * ] 2323 * * NOTE IPFW_TLV_STATE_LIST has the single valid field: objsize. 2324 * The rest (size, count) are set to zero and needs to be ignored. 2325 * 2326 * Returns 0 on success. 2327 */ 2328static int 2329dump_config(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 2330 struct sockopt_data *sd) 2331{ 2332 ipfw_cfg_lheader *hdr; 2333 struct ip_fw *rule; 2334 size_t sz, rnum; 2335 uint32_t hdr_flags; 2336 int error, i; 2337 struct dump_args da; 2338 uint32_t *bmask; 2339 2340 hdr = (ipfw_cfg_lheader *)ipfw_get_sopt_header(sd, sizeof(*hdr)); 2341 if (hdr == NULL) 2342 return (EINVAL); 2343 2344 error = 0; 2345 bmask = NULL; 2346 /* Allocate needed state. Note we allocate 2xspace mask, for table&srv */ 2347 if (hdr->flags & IPFW_CFG_GET_STATIC) 2348 bmask = malloc(IPFW_TABLES_MAX / 4, M_TEMP, M_WAITOK | M_ZERO); 2349 2350 IPFW_UH_RLOCK(chain); 2351 2352 /* 2353 * STAGE 1: Determine size/count for objects in range. 2354 * Prepare used tables bitmask. 2355 */ 2356 sz = sizeof(ipfw_cfg_lheader); 2357 memset(&da, 0, sizeof(da)); 2358 2359 da.b = 0; 2360 da.e = chain->n_rules; 2361 2362 if (hdr->end_rule != 0) { 2363 /* Handle custom range */ 2364 if ((rnum = hdr->start_rule) > IPFW_DEFAULT_RULE) 2365 rnum = IPFW_DEFAULT_RULE; 2366 da.b = ipfw_find_rule(chain, rnum, 0); 2367 rnum = hdr->end_rule; 2368 rnum = (rnum < IPFW_DEFAULT_RULE) ? rnum+1 : IPFW_DEFAULT_RULE; 2369 da.e = ipfw_find_rule(chain, rnum, 0) + 1; 2370 } 2371 2372 if (hdr->flags & IPFW_CFG_GET_STATIC) { 2373 for (i = da.b; i < da.e; i++) { 2374 rule = chain->map[i]; 2375 da.rsize += RULEUSIZE1(rule) + sizeof(ipfw_obj_tlv); 2376 da.rcount++; 2377 /* Update bitmask of used objects for given range */ 2378 da.tcount += mark_object_kidx(chain, rule, bmask); 2379 } 2380 /* Add counters if requested */ 2381 if (hdr->flags & IPFW_CFG_GET_COUNTERS) { 2382 da.rsize += sizeof(struct ip_fw_bcounter) * da.rcount; 2383 da.rcounters = 1; 2384 } 2385 2386 if (da.tcount > 0) 2387 sz += da.tcount * sizeof(ipfw_obj_ntlv) + 2388 sizeof(ipfw_obj_ctlv); 2389 sz += da.rsize + sizeof(ipfw_obj_ctlv); 2390 } 2391 2392 if (hdr->flags & IPFW_CFG_GET_STATES) 2393 sz += ipfw_dyn_get_count() * sizeof(ipfw_obj_dyntlv) + 2394 sizeof(ipfw_obj_ctlv); 2395 2396 2397 /* 2398 * Fill header anyway. 2399 * Note we have to save header fields to stable storage 2400 * buffer inside @sd can be flushed after dumping rules 2401 */ 2402 hdr->size = sz; 2403 hdr->set_mask = ~V_set_disable; 2404 hdr_flags = hdr->flags; 2405 hdr = NULL; 2406 2407 if (sd->valsize < sz) { 2408 error = ENOMEM; 2409 goto cleanup; 2410 } 2411 2412 /* STAGE2: Store actual data */ 2413 if (hdr_flags & IPFW_CFG_GET_STATIC) { 2414 error = dump_static_rules(chain, &da, bmask, sd); 2415 if (error != 0) 2416 goto cleanup; 2417 } 2418 2419 if (hdr_flags & IPFW_CFG_GET_STATES) 2420 error = ipfw_dump_states(chain, sd); 2421 2422cleanup: 2423 IPFW_UH_RUNLOCK(chain); 2424 2425 if (bmask != NULL) 2426 free(bmask, M_TEMP); 2427 2428 return (error); 2429} 2430 2431int 2432ipfw_check_object_name_generic(const char *name) 2433{ 2434 int nsize; 2435 2436 nsize = sizeof(((ipfw_obj_ntlv *)0)->name); 2437 if (strnlen(name, nsize) == nsize) 2438 return (EINVAL); 2439 if (name[0] == '\0') 2440 return (EINVAL); 2441 return (0); 2442} 2443 2444/* 2445 * Creates non-existent objects referenced by rule. 2446 * 2447 * Return 0 on success. 2448 */ 2449int 2450create_objects_compat(struct ip_fw_chain *ch, ipfw_insn *cmd, 2451 struct obj_idx *oib, struct obj_idx *pidx, struct tid_info *ti) 2452{ 2453 struct opcode_obj_rewrite *rw; 2454 struct obj_idx *p; 2455 uint16_t kidx; 2456 int error; 2457 2458 /* 2459 * Compatibility stuff: do actual creation for non-existing, 2460 * but referenced objects. 2461 */ 2462 for (p = oib; p < pidx; p++) { 2463 if (p->kidx != 0) 2464 continue; 2465 2466 ti->uidx = p->uidx; 2467 ti->type = p->type; 2468 ti->atype = 0; 2469 2470 rw = find_op_rw(cmd + p->off, NULL, NULL); 2471 KASSERT(rw != NULL, ("Unable to find handler for op %d", 2472 (cmd + p->off)->opcode)); 2473 2474 if (rw->create_object == NULL) 2475 error = EOPNOTSUPP; 2476 else 2477 error = rw->create_object(ch, ti, &kidx); 2478 if (error == 0) { 2479 p->kidx = kidx; 2480 continue; 2481 } 2482 2483 /* 2484 * Error happened. We have to rollback everything. 2485 * Drop all already acquired references. 2486 */ 2487 IPFW_UH_WLOCK(ch); 2488 unref_oib_objects(ch, cmd, oib, pidx); 2489 IPFW_UH_WUNLOCK(ch); 2490 2491 return (error); 2492 } 2493 2494 return (0); 2495} 2496 2497/* 2498 * Compatibility function for old ipfw(8) binaries. 2499 * Rewrites table/nat kernel indices with userland ones. 2500 * Convert tables matching '/^\d+$/' to their atoi() value. 2501 * Use number 65535 for other tables. 2502 * 2503 * Returns 0 on success. 2504 */ 2505static int 2506set_legacy_obj_kidx(struct ip_fw_chain *ch, struct ip_fw_rule0 *rule) 2507{ 2508 struct opcode_obj_rewrite *rw; 2509 struct named_object *no; 2510 ipfw_insn *cmd; 2511 char *end; 2512 long val; 2513 int cmdlen, error, l; 2514 uint16_t kidx, uidx; 2515 uint8_t subtype; 2516 2517 error = 0; 2518 2519 l = rule->cmd_len; 2520 cmd = rule->cmd; 2521 cmdlen = 0; 2522 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 2523 cmdlen = F_LEN(cmd); 2524 2525 /* Check if is index in given opcode */ 2526 rw = find_op_rw(cmd, &kidx, &subtype); 2527 if (rw == NULL) 2528 continue; 2529 2530 /* Try to find referenced kernel object */ 2531 no = rw->find_bykidx(ch, kidx); 2532 if (no == NULL) 2533 continue; 2534 2535 val = strtol(no->name, &end, 10); 2536 if (*end == '\0' && val < 65535) { 2537 uidx = val; 2538 } else { 2539 2540 /* 2541 * We are called via legacy opcode. 2542 * Save error and show table as fake number 2543 * not to make ipfw(8) hang. 2544 */ 2545 uidx = 65535; 2546 error = 2; 2547 } 2548 2549 rw->update(cmd, uidx); 2550 } 2551 2552 return (error); 2553} 2554 2555 2556/* 2557 * Unreferences all already-referenced objects in given @cmd rule, 2558 * using information in @oib. 2559 * 2560 * Used to rollback partially converted rule on error. 2561 */ 2562static void 2563unref_oib_objects(struct ip_fw_chain *ch, ipfw_insn *cmd, struct obj_idx *oib, 2564 struct obj_idx *end) 2565{ 2566 struct opcode_obj_rewrite *rw; 2567 struct named_object *no; 2568 struct obj_idx *p; 2569 2570 IPFW_UH_WLOCK_ASSERT(ch); 2571 2572 for (p = oib; p < end; p++) { 2573 if (p->kidx == 0) 2574 continue; 2575 2576 rw = find_op_rw(cmd + p->off, NULL, NULL); 2577 KASSERT(rw != NULL, ("Unable to find handler for op %d", 2578 (cmd + p->off)->opcode)); 2579 2580 /* Find & unref by existing idx */ 2581 no = rw->find_bykidx(ch, p->kidx); 2582 KASSERT(no != NULL, ("Ref'd object %d disappeared", p->kidx)); 2583 no->refcnt--; 2584 } 2585} 2586 2587/* 2588 * Remove references from every object used in @rule. 2589 * Used at rule removal code. 2590 */ 2591static void 2592unref_rule_objects(struct ip_fw_chain *ch, struct ip_fw *rule) 2593{ 2594 struct opcode_obj_rewrite *rw; 2595 struct named_object *no; 2596 ipfw_insn *cmd; 2597 int cmdlen, l; 2598 uint16_t kidx; 2599 uint8_t subtype; 2600 2601 IPFW_UH_WLOCK_ASSERT(ch); 2602 2603 l = rule->cmd_len; 2604 cmd = rule->cmd; 2605 cmdlen = 0; 2606 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 2607 cmdlen = F_LEN(cmd); 2608 2609 rw = find_op_rw(cmd, &kidx, &subtype); 2610 if (rw == NULL) 2611 continue; 2612 no = rw->find_bykidx(ch, kidx); 2613 2614 KASSERT(no != NULL, ("table id %d not found", kidx)); 2615 KASSERT(no->subtype == subtype, 2616 ("wrong type %d (%d) for table id %d", 2617 no->subtype, subtype, kidx)); 2618 KASSERT(no->refcnt > 0, ("refcount for table %d is %d", 2619 kidx, no->refcnt)); 2620 2621 if (no->refcnt == 1 && rw->destroy_object != NULL) 2622 rw->destroy_object(ch, no); 2623 else 2624 no->refcnt--; 2625 } 2626} 2627 2628 2629/* 2630 * Find and reference object (if any) stored in instruction @cmd. 2631 * 2632 * Saves object info in @pidx, sets 2633 * - @unresolved to 1 if object should exists but not found 2634 * 2635 * Returns non-zero value in case of error. 2636 */ 2637static int 2638ref_opcode_object(struct ip_fw_chain *ch, ipfw_insn *cmd, struct tid_info *ti, 2639 struct obj_idx *pidx, int *unresolved) 2640{ 2641 struct named_object *no; 2642 struct opcode_obj_rewrite *rw; 2643 int error; 2644 2645 /* Check if this opcode is candidate for rewrite */ 2646 rw = find_op_rw(cmd, &ti->uidx, &ti->type); 2647 if (rw == NULL) 2648 return (0); 2649 2650 /* Need to rewrite. Save necessary fields */ 2651 pidx->uidx = ti->uidx; 2652 pidx->type = ti->type; 2653 2654 /* Try to find referenced kernel object */ 2655 error = rw->find_byname(ch, ti, &no); 2656 if (error != 0) 2657 return (error); 2658 if (no == NULL) { 2659 /* 2660 * Report about unresolved object for automaic 2661 * creation. 2662 */ 2663 *unresolved = 1; 2664 return (0); 2665 } 2666 2667 /* 2668 * Object is already exist. 2669 * Its subtype should match with expected value. 2670 */ 2671 if (ti->type != no->subtype) 2672 return (EINVAL); 2673 2674 /* Bump refcount and update kidx. */ 2675 no->refcnt++; 2676 rw->update(cmd, no->kidx); 2677 return (0); 2678} 2679 2680/* 2681 * Finds and bumps refcount for objects referenced by given @rule. 2682 * Auto-creates non-existing tables. 2683 * Fills in @oib array with userland/kernel indexes. 2684 * 2685 * Returns 0 on success. 2686 */ 2687static int 2688ref_rule_objects(struct ip_fw_chain *ch, struct ip_fw *rule, 2689 struct rule_check_info *ci, struct obj_idx *oib, struct tid_info *ti) 2690{ 2691 struct obj_idx *pidx; 2692 ipfw_insn *cmd; 2693 int cmdlen, error, l, unresolved; 2694 2695 pidx = oib; 2696 l = rule->cmd_len; 2697 cmd = rule->cmd; 2698 cmdlen = 0; 2699 error = 0; 2700 2701 IPFW_UH_WLOCK(ch); 2702 2703 /* Increase refcount on each existing referenced table. */ 2704 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 2705 cmdlen = F_LEN(cmd); 2706 unresolved = 0; 2707 2708 error = ref_opcode_object(ch, cmd, ti, pidx, &unresolved); 2709 if (error != 0) 2710 break; 2711 /* 2712 * Compatibility stuff for old clients: 2713 * prepare to automaitcally create non-existing objects. 2714 */ 2715 if (unresolved != 0) { 2716 pidx->off = rule->cmd_len - l; 2717 pidx++; 2718 } 2719 } 2720 2721 if (error != 0) { 2722 /* Unref everything we have already done */ 2723 unref_oib_objects(ch, rule->cmd, oib, pidx); 2724 IPFW_UH_WUNLOCK(ch); 2725 return (error); 2726 } 2727 IPFW_UH_WUNLOCK(ch); 2728 2729 /* Perform auto-creation for non-existing objects */ 2730 if (pidx != oib) 2731 error = create_objects_compat(ch, rule->cmd, oib, pidx, ti); 2732 2733 /* Calculate real number of dynamic objects */ 2734 ci->object_opcodes = (uint16_t)(pidx - oib); 2735 2736 return (error); 2737} 2738 2739/* 2740 * Checks is opcode is referencing table of appropriate type. 2741 * Adds reference count for found table if true. 2742 * Rewrites user-supplied opcode values with kernel ones. 2743 * 2744 * Returns 0 on success and appropriate error code otherwise. 2745 */ 2746static int 2747rewrite_rule_uidx(struct ip_fw_chain *chain, struct rule_check_info *ci) 2748{ 2749 int error; 2750 ipfw_insn *cmd; 2751 uint8_t type; 2752 struct obj_idx *p, *pidx_first, *pidx_last; 2753 struct tid_info ti; 2754 2755 /* 2756 * Prepare an array for storing opcode indices. 2757 * Use stack allocation by default. 2758 */ 2759 if (ci->object_opcodes <= (sizeof(ci->obuf)/sizeof(ci->obuf[0]))) { 2760 /* Stack */ 2761 pidx_first = ci->obuf; 2762 } else 2763 pidx_first = malloc( 2764 ci->object_opcodes * sizeof(struct obj_idx), 2765 M_IPFW, M_WAITOK | M_ZERO); 2766 2767 error = 0; 2768 type = 0; 2769 memset(&ti, 0, sizeof(ti)); 2770 2771 /* Use set rule is assigned to. */ 2772 ti.set = ci->krule->set; 2773 if (ci->ctlv != NULL) { 2774 ti.tlvs = (void *)(ci->ctlv + 1); 2775 ti.tlen = ci->ctlv->head.length - sizeof(ipfw_obj_ctlv); 2776 } 2777 2778 /* Reference all used tables and other objects */ 2779 error = ref_rule_objects(chain, ci->krule, ci, pidx_first, &ti); 2780 if (error != 0) 2781 goto free; 2782 /* 2783 * Note that ref_rule_objects() might have updated ci->object_opcodes 2784 * to reflect actual number of object opcodes. 2785 */ 2786 2787 /* Perform rewrite of remaining opcodes */ 2788 p = pidx_first; 2789 pidx_last = pidx_first + ci->object_opcodes; 2790 for (p = pidx_first; p < pidx_last; p++) { 2791 cmd = ci->krule->cmd + p->off; 2792 update_opcode_kidx(cmd, p->kidx); 2793 } 2794 2795free: 2796 if (pidx_first != ci->obuf) 2797 free(pidx_first, M_IPFW); 2798 2799 return (error); 2800} 2801 2802/* 2803 * Adds one or more rules to ipfw @chain. 2804 * Data layout (version 0)(current): 2805 * Request: 2806 * [ 2807 * ip_fw3_opheader 2808 * [ ipfw_obj_ctlv(IPFW_TLV_TBL_LIST) ipfw_obj_ntlv x N ] (optional *1) 2809 * [ ipfw_obj_ctlv(IPFW_TLV_RULE_LIST) ip_fw x N ] (*2) (*3) 2810 * ] 2811 * Reply: 2812 * [ 2813 * ip_fw3_opheader 2814 * [ ipfw_obj_ctlv(IPFW_TLV_TBL_LIST) ipfw_obj_ntlv x N ] (optional) 2815 * [ ipfw_obj_ctlv(IPFW_TLV_RULE_LIST) ip_fw x N ] 2816 * ] 2817 * 2818 * Rules in reply are modified to store their actual ruleset number. 2819 * 2820 * (*1) TLVs inside IPFW_TLV_TBL_LIST needs to be sorted ascending 2821 * according to their idx field and there has to be no duplicates. 2822 * (*2) Numbered rules inside IPFW_TLV_RULE_LIST needs to be sorted ascending. 2823 * (*3) Each ip_fw structure needs to be aligned to u64 boundary. 2824 * 2825 * Returns 0 on success. 2826 */ 2827static int 2828add_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 2829 struct sockopt_data *sd) 2830{ 2831 ipfw_obj_ctlv *ctlv, *rtlv, *tstate; 2832 ipfw_obj_ntlv *ntlv; 2833 int clen, error, idx; 2834 uint32_t count, read; 2835 struct ip_fw_rule *r; 2836 struct rule_check_info rci, *ci, *cbuf; 2837 int i, rsize; 2838 2839 op3 = (ip_fw3_opheader *)ipfw_get_sopt_space(sd, sd->valsize); 2840 ctlv = (ipfw_obj_ctlv *)(op3 + 1); 2841 2842 read = sizeof(ip_fw3_opheader); 2843 rtlv = NULL; 2844 tstate = NULL; 2845 cbuf = NULL; 2846 memset(&rci, 0, sizeof(struct rule_check_info)); 2847 2848 if (read + sizeof(*ctlv) > sd->valsize) 2849 return (EINVAL); 2850 2851 if (ctlv->head.type == IPFW_TLV_TBLNAME_LIST) { 2852 clen = ctlv->head.length; 2853 /* Check size and alignment */ 2854 if (clen > sd->valsize || clen < sizeof(*ctlv)) 2855 return (EINVAL); 2856 if ((clen % sizeof(uint64_t)) != 0) 2857 return (EINVAL); 2858 2859 /* 2860 * Some table names or other named objects. 2861 * Check for validness. 2862 */ 2863 count = (ctlv->head.length - sizeof(*ctlv)) / sizeof(*ntlv); 2864 if (ctlv->count != count || ctlv->objsize != sizeof(*ntlv)) 2865 return (EINVAL); 2866 2867 /* 2868 * Check each TLV. 2869 * Ensure TLVs are sorted ascending and 2870 * there are no duplicates. 2871 */ 2872 idx = -1; 2873 ntlv = (ipfw_obj_ntlv *)(ctlv + 1); 2874 while (count > 0) { 2875 if (ntlv->head.length != sizeof(ipfw_obj_ntlv)) 2876 return (EINVAL); 2877 2878 error = ipfw_check_object_name_generic(ntlv->name); 2879 if (error != 0) 2880 return (error); 2881 2882 if (ntlv->idx <= idx) 2883 return (EINVAL); 2884 2885 idx = ntlv->idx; 2886 count--; 2887 ntlv++; 2888 } 2889 2890 tstate = ctlv; 2891 read += ctlv->head.length; 2892 ctlv = (ipfw_obj_ctlv *)((caddr_t)ctlv + ctlv->head.length); 2893 } 2894 2895 if (read + sizeof(*ctlv) > sd->valsize) 2896 return (EINVAL); 2897 2898 if (ctlv->head.type == IPFW_TLV_RULE_LIST) { 2899 clen = ctlv->head.length; 2900 if (clen + read > sd->valsize || clen < sizeof(*ctlv)) 2901 return (EINVAL); 2902 if ((clen % sizeof(uint64_t)) != 0) 2903 return (EINVAL); 2904 2905 /* 2906 * TODO: Permit adding multiple rules at once 2907 */ 2908 if (ctlv->count != 1) 2909 return (ENOTSUP); 2910 2911 clen -= sizeof(*ctlv); 2912 2913 if (ctlv->count > clen / sizeof(struct ip_fw_rule)) 2914 return (EINVAL); 2915 2916 /* Allocate state for each rule or use stack */ 2917 if (ctlv->count == 1) { 2918 memset(&rci, 0, sizeof(struct rule_check_info)); 2919 cbuf = &rci; 2920 } else 2921 cbuf = malloc(ctlv->count * sizeof(*ci), M_TEMP, 2922 M_WAITOK | M_ZERO); 2923 ci = cbuf; 2924 2925 /* 2926 * Check each rule for validness. 2927 * Ensure numbered rules are sorted ascending 2928 * and properly aligned 2929 */ 2930 idx = 0; 2931 r = (struct ip_fw_rule *)(ctlv + 1); 2932 count = 0; 2933 error = 0; 2934 while (clen > 0) { 2935 rsize = roundup2(RULESIZE(r), sizeof(uint64_t)); 2936 if (rsize > clen || ctlv->count <= count) { 2937 error = EINVAL; 2938 break; 2939 } 2940 2941 ci->ctlv = tstate; 2942 error = check_ipfw_rule1(r, rsize, ci); 2943 if (error != 0) 2944 break; 2945 2946 /* Check sorting */ 2947 if (r->rulenum != 0 && r->rulenum < idx) { 2948 printf("rulenum %d idx %d\n", r->rulenum, idx); 2949 error = EINVAL; 2950 break; 2951 } 2952 idx = r->rulenum; 2953 2954 ci->urule = (caddr_t)r; 2955 2956 rsize = roundup2(rsize, sizeof(uint64_t)); 2957 clen -= rsize; 2958 r = (struct ip_fw_rule *)((caddr_t)r + rsize); 2959 count++; 2960 ci++; 2961 } 2962 2963 if (ctlv->count != count || error != 0) { 2964 if (cbuf != &rci) 2965 free(cbuf, M_TEMP); 2966 return (EINVAL); 2967 } 2968 2969 rtlv = ctlv; 2970 read += ctlv->head.length; 2971 ctlv = (ipfw_obj_ctlv *)((caddr_t)ctlv + ctlv->head.length); 2972 } 2973 2974 if (read != sd->valsize || rtlv == NULL || rtlv->count == 0) { 2975 if (cbuf != NULL && cbuf != &rci) 2976 free(cbuf, M_TEMP); 2977 return (EINVAL); 2978 } 2979 2980 /* 2981 * Passed rules seems to be valid. 2982 * Allocate storage and try to add them to chain. 2983 */ 2984 for (i = 0, ci = cbuf; i < rtlv->count; i++, ci++) { 2985 clen = RULEKSIZE1((struct ip_fw_rule *)ci->urule); 2986 ci->krule = ipfw_alloc_rule(chain, clen); 2987 import_rule1(ci); 2988 } 2989 2990 if ((error = commit_rules(chain, cbuf, rtlv->count)) != 0) { 2991 /* Free allocate krules */ 2992 for (i = 0, ci = cbuf; i < rtlv->count; i++, ci++) 2993 free_rule(ci->krule); 2994 } 2995 2996 if (cbuf != NULL && cbuf != &rci) 2997 free(cbuf, M_TEMP); 2998 2999 return (error); 3000} 3001 3002/* 3003 * Lists all sopts currently registered. 3004 * Data layout (v0)(current): 3005 * Request: [ ipfw_obj_lheader ], size = ipfw_obj_lheader.size 3006 * Reply: [ ipfw_obj_lheader ipfw_sopt_info x N ] 3007 * 3008 * Returns 0 on success 3009 */ 3010static int 3011dump_soptcodes(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 3012 struct sockopt_data *sd) 3013{ 3014 struct _ipfw_obj_lheader *olh; 3015 ipfw_sopt_info *i; 3016 struct ipfw_sopt_handler *sh; 3017 uint32_t count, n, size; 3018 3019 olh = (struct _ipfw_obj_lheader *)ipfw_get_sopt_header(sd,sizeof(*olh)); 3020 if (olh == NULL) 3021 return (EINVAL); 3022 if (sd->valsize < olh->size) 3023 return (EINVAL); 3024 3025 CTL3_LOCK(); 3026 count = ctl3_hsize; 3027 size = count * sizeof(ipfw_sopt_info) + sizeof(ipfw_obj_lheader); 3028 3029 /* Fill in header regadless of buffer size */ 3030 olh->count = count; 3031 olh->objsize = sizeof(ipfw_sopt_info); 3032 3033 if (size > olh->size) { 3034 olh->size = size; 3035 CTL3_UNLOCK(); 3036 return (ENOMEM); 3037 } 3038 olh->size = size; 3039 3040 for (n = 1; n <= count; n++) { 3041 i = (ipfw_sopt_info *)ipfw_get_sopt_space(sd, sizeof(*i)); 3042 KASSERT(i != NULL, ("previously checked buffer is not enough")); 3043 sh = &ctl3_handlers[n]; 3044 i->opcode = sh->opcode; 3045 i->version = sh->version; 3046 i->refcnt = sh->refcnt; 3047 } 3048 CTL3_UNLOCK(); 3049 3050 return (0); 3051} 3052 3053/* 3054 * Compares two opcodes. 3055 * Used both in qsort() and bsearch(). 3056 * 3057 * Returns 0 if match is found. 3058 */ 3059static int 3060compare_opcodes(const void *_a, const void *_b) 3061{ 3062 const struct opcode_obj_rewrite *a, *b; 3063 3064 a = (const struct opcode_obj_rewrite *)_a; 3065 b = (const struct opcode_obj_rewrite *)_b; 3066 3067 if (a->opcode < b->opcode) 3068 return (-1); 3069 else if (a->opcode > b->opcode) 3070 return (1); 3071 3072 return (0); 3073} 3074 3075/* 3076 * XXX: Rewrite bsearch() 3077 */ 3078static int 3079find_op_rw_range(uint16_t op, struct opcode_obj_rewrite **plo, 3080 struct opcode_obj_rewrite **phi) 3081{ 3082 struct opcode_obj_rewrite *ctl3_max, *lo, *hi, h, *rw; 3083 3084 memset(&h, 0, sizeof(h)); 3085 h.opcode = op; 3086 3087 rw = (struct opcode_obj_rewrite *)bsearch(&h, ctl3_rewriters, 3088 ctl3_rsize, sizeof(h), compare_opcodes); 3089 if (rw == NULL) 3090 return (1); 3091 3092 /* Find the first element matching the same opcode */ 3093 lo = rw; 3094 for ( ; lo > ctl3_rewriters && (lo - 1)->opcode == op; lo--) 3095 ; 3096 3097 /* Find the last element matching the same opcode */ 3098 hi = rw; 3099 ctl3_max = ctl3_rewriters + ctl3_rsize; 3100 for ( ; (hi + 1) < ctl3_max && (hi + 1)->opcode == op; hi++) 3101 ; 3102 3103 *plo = lo; 3104 *phi = hi; 3105 3106 return (0); 3107} 3108 3109/* 3110 * Finds opcode object rewriter based on @code. 3111 * 3112 * Returns pointer to handler or NULL. 3113 */ 3114static struct opcode_obj_rewrite * 3115find_op_rw(ipfw_insn *cmd, uint16_t *puidx, uint8_t *ptype) 3116{ 3117 struct opcode_obj_rewrite *rw, *lo, *hi; 3118 uint16_t uidx; 3119 uint8_t subtype; 3120 3121 if (find_op_rw_range(cmd->opcode, &lo, &hi) != 0) 3122 return (NULL); 3123 3124 for (rw = lo; rw <= hi; rw++) { 3125 if (rw->classifier(cmd, &uidx, &subtype) == 0) { 3126 if (puidx != NULL) 3127 *puidx = uidx; 3128 if (ptype != NULL) 3129 *ptype = subtype; 3130 return (rw); 3131 } 3132 } 3133 3134 return (NULL); 3135} 3136int 3137classify_opcode_kidx(ipfw_insn *cmd, uint16_t *puidx) 3138{ 3139 3140 if (find_op_rw(cmd, puidx, NULL) == NULL) 3141 return (1); 3142 return (0); 3143} 3144 3145void 3146update_opcode_kidx(ipfw_insn *cmd, uint16_t idx) 3147{ 3148 struct opcode_obj_rewrite *rw; 3149 3150 rw = find_op_rw(cmd, NULL, NULL); 3151 KASSERT(rw != NULL, ("No handler to update opcode %d", cmd->opcode)); 3152 rw->update(cmd, idx); 3153} 3154 3155void 3156ipfw_init_obj_rewriter() 3157{ 3158 3159 ctl3_rewriters = NULL; 3160 ctl3_rsize = 0; 3161} 3162 3163void 3164ipfw_destroy_obj_rewriter() 3165{ 3166 3167 if (ctl3_rewriters != NULL) 3168 free(ctl3_rewriters, M_IPFW); 3169 ctl3_rewriters = NULL; 3170 ctl3_rsize = 0; 3171} 3172 3173/* 3174 * Adds one or more opcode object rewrite handlers to the global array. 3175 * Function may sleep. 3176 */ 3177void 3178ipfw_add_obj_rewriter(struct opcode_obj_rewrite *rw, size_t count) 3179{ 3180 size_t sz; 3181 struct opcode_obj_rewrite *tmp; 3182 3183 CTL3_LOCK(); 3184 3185 for (;;) { 3186 sz = ctl3_rsize + count; 3187 CTL3_UNLOCK(); 3188 tmp = malloc(sizeof(*rw) * sz, M_IPFW, M_WAITOK | M_ZERO); 3189 CTL3_LOCK(); 3190 if (ctl3_rsize + count <= sz) 3191 break; 3192 3193 /* Retry */ 3194 free(tmp, M_IPFW); 3195 } 3196 3197 /* Merge old & new arrays */ 3198 sz = ctl3_rsize + count; 3199 memcpy(tmp, ctl3_rewriters, ctl3_rsize * sizeof(*rw)); 3200 memcpy(&tmp[ctl3_rsize], rw, count * sizeof(*rw)); 3201 qsort(tmp, sz, sizeof(*rw), compare_opcodes); 3202 /* Switch new and free old */ 3203 if (ctl3_rewriters != NULL) 3204 free(ctl3_rewriters, M_IPFW); 3205 ctl3_rewriters = tmp; 3206 ctl3_rsize = sz; 3207 3208 CTL3_UNLOCK(); 3209} 3210 3211/* 3212 * Removes one or more object rewrite handlers from the global array. 3213 */ 3214int 3215ipfw_del_obj_rewriter(struct opcode_obj_rewrite *rw, size_t count) 3216{ 3217 size_t sz; 3218 struct opcode_obj_rewrite *ctl3_max, *ktmp, *lo, *hi; 3219 int i; 3220 3221 CTL3_LOCK(); 3222 3223 for (i = 0; i < count; i++) { 3224 if (find_op_rw_range(rw[i].opcode, &lo, &hi) != 0) 3225 continue; 3226 3227 for (ktmp = lo; ktmp <= hi; ktmp++) { 3228 if (ktmp->classifier != rw[i].classifier) 3229 continue; 3230 3231 ctl3_max = ctl3_rewriters + ctl3_rsize; 3232 sz = (ctl3_max - (ktmp + 1)) * sizeof(*ktmp); 3233 memmove(ktmp, ktmp + 1, sz); 3234 ctl3_rsize--; 3235 break; 3236 } 3237 3238 } 3239 3240 if (ctl3_rsize == 0) { 3241 if (ctl3_rewriters != NULL) 3242 free(ctl3_rewriters, M_IPFW); 3243 ctl3_rewriters = NULL; 3244 } 3245 3246 CTL3_UNLOCK(); 3247 3248 return (0); 3249} 3250 3251static int 3252export_objhash_ntlv_internal(struct namedobj_instance *ni, 3253 struct named_object *no, void *arg) 3254{ 3255 struct sockopt_data *sd; 3256 ipfw_obj_ntlv *ntlv; 3257 3258 sd = (struct sockopt_data *)arg; 3259 ntlv = (ipfw_obj_ntlv *)ipfw_get_sopt_space(sd, sizeof(*ntlv)); 3260 if (ntlv == NULL) 3261 return (ENOMEM); 3262 ipfw_export_obj_ntlv(no, ntlv); 3263 return (0); 3264} 3265 3266/* 3267 * Lists all service objects. 3268 * Data layout (v0)(current): 3269 * Request: [ ipfw_obj_lheader ] size = ipfw_obj_lheader.size 3270 * Reply: [ ipfw_obj_lheader [ ipfw_obj_ntlv x N ] (optional) ] 3271 * Returns 0 on success 3272 */ 3273static int 3274dump_srvobjects(struct ip_fw_chain *chain, ip_fw3_opheader *op3, 3275 struct sockopt_data *sd) 3276{ 3277 ipfw_obj_lheader *hdr; 3278 int count; 3279 3280 hdr = (ipfw_obj_lheader *)ipfw_get_sopt_header(sd, sizeof(*hdr)); 3281 if (hdr == NULL) 3282 return (EINVAL); 3283 3284 IPFW_UH_RLOCK(chain); 3285 count = ipfw_objhash_count(CHAIN_TO_SRV(chain)); 3286 hdr->size = sizeof(ipfw_obj_lheader) + count * sizeof(ipfw_obj_ntlv); 3287 if (sd->valsize < hdr->size) { 3288 IPFW_UH_RUNLOCK(chain); 3289 return (ENOMEM); 3290 } 3291 hdr->count = count; 3292 hdr->objsize = sizeof(ipfw_obj_ntlv); 3293 if (count > 0) 3294 ipfw_objhash_foreach(CHAIN_TO_SRV(chain), 3295 export_objhash_ntlv_internal, sd); 3296 IPFW_UH_RUNLOCK(chain); 3297 return (0); 3298} 3299 3300/* 3301 * Compares two sopt handlers (code, version and handler ptr). 3302 * Used both as qsort() and bsearch(). 3303 * Does not compare handler for latter case. 3304 * 3305 * Returns 0 if match is found. 3306 */ 3307static int 3308compare_sh(const void *_a, const void *_b) 3309{ 3310 const struct ipfw_sopt_handler *a, *b; 3311 3312 a = (const struct ipfw_sopt_handler *)_a; 3313 b = (const struct ipfw_sopt_handler *)_b; 3314 3315 if (a->opcode < b->opcode) 3316 return (-1); 3317 else if (a->opcode > b->opcode) 3318 return (1); 3319 3320 if (a->version < b->version) 3321 return (-1); 3322 else if (a->version > b->version) 3323 return (1); 3324 3325 /* bsearch helper */ 3326 if (a->handler == NULL) 3327 return (0); 3328 3329 if ((uintptr_t)a->handler < (uintptr_t)b->handler) 3330 return (-1); 3331 else if ((uintptr_t)a->handler > (uintptr_t)b->handler) 3332 return (1); 3333 3334 return (0); 3335} 3336 3337/* 3338 * Finds sopt handler based on @code and @version. 3339 * 3340 * Returns pointer to handler or NULL. 3341 */ 3342static struct ipfw_sopt_handler * 3343find_sh(uint16_t code, uint8_t version, sopt_handler_f *handler) 3344{ 3345 struct ipfw_sopt_handler *sh, h; 3346 3347 memset(&h, 0, sizeof(h)); 3348 h.opcode = code; 3349 h.version = version; 3350 h.handler = handler; 3351 3352 sh = (struct ipfw_sopt_handler *)bsearch(&h, ctl3_handlers, 3353 ctl3_hsize, sizeof(h), compare_sh); 3354 3355 return (sh); 3356} 3357 3358static int 3359find_ref_sh(uint16_t opcode, uint8_t version, struct ipfw_sopt_handler *psh) 3360{ 3361 struct ipfw_sopt_handler *sh; 3362 3363 CTL3_LOCK(); 3364 if ((sh = find_sh(opcode, version, NULL)) == NULL) { 3365 CTL3_UNLOCK(); 3366 printf("ipfw: ipfw_ctl3 invalid option %d""v""%d\n", 3367 opcode, version); 3368 return (EINVAL); 3369 } 3370 sh->refcnt++; 3371 ctl3_refct++; 3372 /* Copy handler data to requested buffer */ 3373 *psh = *sh; 3374 CTL3_UNLOCK(); 3375 3376 return (0); 3377} 3378 3379static void 3380find_unref_sh(struct ipfw_sopt_handler *psh) 3381{ 3382 struct ipfw_sopt_handler *sh; 3383 3384 CTL3_LOCK(); 3385 sh = find_sh(psh->opcode, psh->version, NULL); 3386 KASSERT(sh != NULL, ("ctl3 handler disappeared")); 3387 sh->refcnt--; 3388 ctl3_refct--; 3389 CTL3_UNLOCK(); 3390} 3391 3392void 3393ipfw_init_sopt_handler() 3394{ 3395 3396 CTL3_LOCK_INIT(); 3397 IPFW_ADD_SOPT_HANDLER(1, scodes); 3398} 3399 3400void 3401ipfw_destroy_sopt_handler() 3402{ 3403 3404 IPFW_DEL_SOPT_HANDLER(1, scodes); 3405 CTL3_LOCK_DESTROY(); 3406} 3407 3408/* 3409 * Adds one or more sockopt handlers to the global array. 3410 * Function may sleep. 3411 */ 3412void 3413ipfw_add_sopt_handler(struct ipfw_sopt_handler *sh, size_t count) 3414{ 3415 size_t sz; 3416 struct ipfw_sopt_handler *tmp; 3417 3418 CTL3_LOCK(); 3419 3420 for (;;) { 3421 sz = ctl3_hsize + count; 3422 CTL3_UNLOCK(); 3423 tmp = malloc(sizeof(*sh) * sz, M_IPFW, M_WAITOK | M_ZERO); 3424 CTL3_LOCK(); 3425 if (ctl3_hsize + count <= sz) 3426 break; 3427 3428 /* Retry */ 3429 free(tmp, M_IPFW); 3430 } 3431 3432 /* Merge old & new arrays */ 3433 sz = ctl3_hsize + count; 3434 memcpy(tmp, ctl3_handlers, ctl3_hsize * sizeof(*sh)); 3435 memcpy(&tmp[ctl3_hsize], sh, count * sizeof(*sh)); 3436 qsort(tmp, sz, sizeof(*sh), compare_sh); 3437 /* Switch new and free old */ 3438 if (ctl3_handlers != NULL) 3439 free(ctl3_handlers, M_IPFW); 3440 ctl3_handlers = tmp; 3441 ctl3_hsize = sz; 3442 ctl3_gencnt++; 3443 3444 CTL3_UNLOCK(); 3445} 3446 3447/* 3448 * Removes one or more sockopt handlers from the global array. 3449 */ 3450int 3451ipfw_del_sopt_handler(struct ipfw_sopt_handler *sh, size_t count) 3452{ 3453 size_t sz; 3454 struct ipfw_sopt_handler *tmp, *h; 3455 int i; 3456 3457 CTL3_LOCK(); 3458 3459 for (i = 0; i < count; i++) { 3460 tmp = &sh[i]; 3461 h = find_sh(tmp->opcode, tmp->version, tmp->handler); 3462 if (h == NULL) 3463 continue; 3464 3465 sz = (ctl3_handlers + ctl3_hsize - (h + 1)) * sizeof(*h); 3466 memmove(h, h + 1, sz); 3467 ctl3_hsize--; 3468 } 3469 3470 if (ctl3_hsize == 0) { 3471 if (ctl3_handlers != NULL) 3472 free(ctl3_handlers, M_IPFW); 3473 ctl3_handlers = NULL; 3474 } 3475 3476 ctl3_gencnt++; 3477 3478 CTL3_UNLOCK(); 3479 3480 return (0); 3481} 3482 3483/* 3484 * Writes data accumulated in @sd to sockopt buffer. 3485 * Zeroes internal @sd buffer. 3486 */ 3487static int 3488ipfw_flush_sopt_data(struct sockopt_data *sd) 3489{ 3490 struct sockopt *sopt; 3491 int error; 3492 size_t sz; 3493 3494 sz = sd->koff; 3495 if (sz == 0) 3496 return (0); 3497 3498 sopt = sd->sopt; 3499 3500 if (sopt->sopt_dir == SOPT_GET) { 3501 error = copyout(sd->kbuf, sopt->sopt_val, sz); 3502 if (error != 0) 3503 return (error); 3504 } 3505 3506 memset(sd->kbuf, 0, sd->ksize); 3507 sd->ktotal += sz; 3508 sd->koff = 0; 3509 if (sd->ktotal + sd->ksize < sd->valsize) 3510 sd->kavail = sd->ksize; 3511 else 3512 sd->kavail = sd->valsize - sd->ktotal; 3513 3514 /* Update sopt buffer data */ 3515 sopt->sopt_valsize = sd->ktotal; 3516 sopt->sopt_val = sd->sopt_val + sd->ktotal; 3517 3518 return (0); 3519} 3520 3521/* 3522 * Ensures that @sd buffer has contiguous @neeeded number of 3523 * bytes. 3524 * 3525 * Returns pointer to requested space or NULL. 3526 */ 3527caddr_t 3528ipfw_get_sopt_space(struct sockopt_data *sd, size_t needed) 3529{ 3530 int error; 3531 caddr_t addr; 3532 3533 if (sd->kavail < needed) { 3534 /* 3535 * Flush data and try another time. 3536 */ 3537 error = ipfw_flush_sopt_data(sd); 3538 3539 if (sd->kavail < needed || error != 0) 3540 return (NULL); 3541 } 3542 3543 addr = sd->kbuf + sd->koff; 3544 sd->koff += needed; 3545 sd->kavail -= needed; 3546 return (addr); 3547} 3548 3549/* 3550 * Requests @needed contiguous bytes from @sd buffer. 3551 * Function is used to notify subsystem that we are 3552 * interesed in first @needed bytes (request header) 3553 * and the rest buffer can be safely zeroed. 3554 * 3555 * Returns pointer to requested space or NULL. 3556 */ 3557caddr_t 3558ipfw_get_sopt_header(struct sockopt_data *sd, size_t needed) 3559{ 3560 caddr_t addr; 3561 3562 if ((addr = ipfw_get_sopt_space(sd, needed)) == NULL) 3563 return (NULL); 3564 3565 if (sd->kavail > 0) 3566 memset(sd->kbuf + sd->koff, 0, sd->kavail); 3567 3568 return (addr); 3569} 3570 3571/* 3572 * New sockopt handler. 3573 */ 3574int 3575ipfw_ctl3(struct sockopt *sopt) 3576{ 3577 int error, locked; 3578 size_t size, valsize; 3579 struct ip_fw_chain *chain; 3580 char xbuf[256]; 3581 struct sockopt_data sdata; 3582 struct ipfw_sopt_handler h; 3583 ip_fw3_opheader *op3 = NULL; 3584 3585 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW); 3586 if (error != 0) 3587 return (error); 3588 3589 if (sopt->sopt_name != IP_FW3) 3590 return (ipfw_ctl(sopt)); 3591 3592 chain = &V_layer3_chain; 3593 error = 0; 3594 3595 /* Save original valsize before it is altered via sooptcopyin() */ 3596 valsize = sopt->sopt_valsize; 3597 memset(&sdata, 0, sizeof(sdata)); 3598 /* Read op3 header first to determine actual operation */ 3599 op3 = (ip_fw3_opheader *)xbuf; 3600 error = sooptcopyin(sopt, op3, sizeof(*op3), sizeof(*op3)); 3601 if (error != 0) 3602 return (error); 3603 sopt->sopt_valsize = valsize; 3604 3605 /* 3606 * Find and reference command. 3607 */ 3608 error = find_ref_sh(op3->opcode, op3->version, &h); 3609 if (error != 0) 3610 return (error); 3611 3612 /* 3613 * Disallow modifications in really-really secure mode, but still allow 3614 * the logging counters to be reset. 3615 */ 3616 if ((h.dir & HDIR_SET) != 0 && h.opcode != IP_FW_XRESETLOG) { 3617 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3618 if (error != 0) { 3619 find_unref_sh(&h); 3620 return (error); 3621 } 3622 } 3623 3624 /* 3625 * Fill in sockopt_data structure that may be useful for 3626 * IP_FW3 get requests. 3627 */ 3628 locked = 0; 3629 if (valsize <= sizeof(xbuf)) { 3630 /* use on-stack buffer */ 3631 sdata.kbuf = xbuf; 3632 sdata.ksize = sizeof(xbuf); 3633 sdata.kavail = valsize; 3634 } else { 3635 3636 /* 3637 * Determine opcode type/buffer size: 3638 * allocate sliding-window buf for data export or 3639 * contiguous buffer for special ops. 3640 */ 3641 if ((h.dir & HDIR_SET) != 0) { 3642 /* Set request. Allocate contigous buffer. */ 3643 if (valsize > CTL3_LARGEBUF) { 3644 find_unref_sh(&h); 3645 return (EFBIG); 3646 } 3647 3648 size = valsize; 3649 } else { 3650 /* Get request. Allocate sliding window buffer */ 3651 size = (valsize<CTL3_SMALLBUF) ? valsize:CTL3_SMALLBUF; 3652 3653 if (size < valsize) { 3654 /* We have to wire user buffer */ 3655 error = vslock(sopt->sopt_val, valsize); 3656 if (error != 0) 3657 return (error); 3658 locked = 1; 3659 } 3660 } 3661 3662 sdata.kbuf = malloc(size, M_TEMP, M_WAITOK | M_ZERO); 3663 sdata.ksize = size; 3664 sdata.kavail = size; 3665 } 3666 3667 sdata.sopt = sopt; 3668 sdata.sopt_val = sopt->sopt_val; 3669 sdata.valsize = valsize; 3670 3671 /* 3672 * Copy either all request (if valsize < bsize_max) 3673 * or first bsize_max bytes to guarantee most consumers 3674 * that all necessary data has been copied). 3675 * Anyway, copy not less than sizeof(ip_fw3_opheader). 3676 */ 3677 if ((error = sooptcopyin(sopt, sdata.kbuf, sdata.ksize, 3678 sizeof(ip_fw3_opheader))) != 0) 3679 return (error); 3680 op3 = (ip_fw3_opheader *)sdata.kbuf; 3681 3682 /* Finally, run handler */ 3683 error = h.handler(chain, op3, &sdata); 3684 find_unref_sh(&h); 3685 3686 /* Flush state and free buffers */ 3687 if (error == 0) 3688 error = ipfw_flush_sopt_data(&sdata); 3689 else 3690 ipfw_flush_sopt_data(&sdata); 3691 3692 if (locked != 0) 3693 vsunlock(sdata.sopt_val, valsize); 3694 3695 /* Restore original pointer and set number of bytes written */ 3696 sopt->sopt_val = sdata.sopt_val; 3697 sopt->sopt_valsize = sdata.ktotal; 3698 if (sdata.kbuf != xbuf) 3699 free(sdata.kbuf, M_TEMP); 3700 3701 return (error); 3702} 3703 3704/** 3705 * {set|get}sockopt parser. 3706 */ 3707int 3708ipfw_ctl(struct sockopt *sopt) 3709{ 3710#define RULE_MAXSIZE (512*sizeof(u_int32_t)) 3711 int error; 3712 size_t size, valsize; 3713 struct ip_fw *buf; 3714 struct ip_fw_rule0 *rule; 3715 struct ip_fw_chain *chain; 3716 u_int32_t rulenum[2]; 3717 uint32_t opt; 3718 struct rule_check_info ci; 3719 IPFW_RLOCK_TRACKER; 3720 3721 chain = &V_layer3_chain; 3722 error = 0; 3723 3724 /* Save original valsize before it is altered via sooptcopyin() */ 3725 valsize = sopt->sopt_valsize; 3726 opt = sopt->sopt_name; 3727 3728 /* 3729 * Disallow modifications in really-really secure mode, but still allow 3730 * the logging counters to be reset. 3731 */ 3732 if (opt == IP_FW_ADD || 3733 (sopt->sopt_dir == SOPT_SET && opt != IP_FW_RESETLOG)) { 3734 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3735 if (error != 0) 3736 return (error); 3737 } 3738 3739 switch (opt) { 3740 case IP_FW_GET: 3741 /* 3742 * pass up a copy of the current rules. Static rules 3743 * come first (the last of which has number IPFW_DEFAULT_RULE), 3744 * followed by a possibly empty list of dynamic rule. 3745 * The last dynamic rule has NULL in the "next" field. 3746 * 3747 * Note that the calculated size is used to bound the 3748 * amount of data returned to the user. The rule set may 3749 * change between calculating the size and returning the 3750 * data in which case we'll just return what fits. 3751 */ 3752 for (;;) { 3753 int len = 0, want; 3754 3755 size = chain->static_len; 3756 size += ipfw_dyn_len(); 3757 if (size >= sopt->sopt_valsize) 3758 break; 3759 buf = malloc(size, M_TEMP, M_WAITOK | M_ZERO); 3760 IPFW_UH_RLOCK(chain); 3761 /* check again how much space we need */ 3762 want = chain->static_len + ipfw_dyn_len(); 3763 if (size >= want) 3764 len = ipfw_getrules(chain, buf, size); 3765 IPFW_UH_RUNLOCK(chain); 3766 if (size >= want) 3767 error = sooptcopyout(sopt, buf, len); 3768 free(buf, M_TEMP); 3769 if (size >= want) 3770 break; 3771 } 3772 break; 3773 3774 case IP_FW_FLUSH: 3775 /* locking is done within del_entry() */ 3776 error = del_entry(chain, 0); /* special case, rule=0, cmd=0 means all */ 3777 break; 3778 3779 case IP_FW_ADD: 3780 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 3781 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 3782 sizeof(struct ip_fw7) ); 3783 3784 memset(&ci, 0, sizeof(struct rule_check_info)); 3785 3786 /* 3787 * If the size of commands equals RULESIZE7 then we assume 3788 * a FreeBSD7.2 binary is talking to us (set is7=1). 3789 * is7 is persistent so the next 'ipfw list' command 3790 * will use this format. 3791 * NOTE: If wrong version is guessed (this can happen if 3792 * the first ipfw command is 'ipfw [pipe] list') 3793 * the ipfw binary may crash or loop infinitly... 3794 */ 3795 size = sopt->sopt_valsize; 3796 if (size == RULESIZE7(rule)) { 3797 is7 = 1; 3798 error = convert_rule_to_8(rule); 3799 if (error) { 3800 free(rule, M_TEMP); 3801 return error; 3802 } 3803 size = RULESIZE(rule); 3804 } else 3805 is7 = 0; 3806 if (error == 0) 3807 error = check_ipfw_rule0(rule, size, &ci); 3808 if (error == 0) { 3809 /* locking is done within add_rule() */ 3810 struct ip_fw *krule; 3811 krule = ipfw_alloc_rule(chain, RULEKSIZE0(rule)); 3812 ci.urule = (caddr_t)rule; 3813 ci.krule = krule; 3814 import_rule0(&ci); 3815 error = commit_rules(chain, &ci, 1); 3816 if (error != 0) 3817 free_rule(ci.krule); 3818 else if (sopt->sopt_dir == SOPT_GET) { 3819 if (is7) { 3820 error = convert_rule_to_7(rule); 3821 size = RULESIZE7(rule); 3822 if (error) { 3823 free(rule, M_TEMP); 3824 return error; 3825 } 3826 } 3827 error = sooptcopyout(sopt, rule, size); 3828 } 3829 } 3830 free(rule, M_TEMP); 3831 break; 3832 3833 case IP_FW_DEL: 3834 /* 3835 * IP_FW_DEL is used for deleting single rules or sets, 3836 * and (ab)used to atomically manipulate sets. Argument size 3837 * is used to distinguish between the two: 3838 * sizeof(u_int32_t) 3839 * delete single rule or set of rules, 3840 * or reassign rules (or sets) to a different set. 3841 * 2*sizeof(u_int32_t) 3842 * atomic disable/enable sets. 3843 * first u_int32_t contains sets to be disabled, 3844 * second u_int32_t contains sets to be enabled. 3845 */ 3846 error = sooptcopyin(sopt, rulenum, 3847 2*sizeof(u_int32_t), sizeof(u_int32_t)); 3848 if (error) 3849 break; 3850 size = sopt->sopt_valsize; 3851 if (size == sizeof(u_int32_t) && rulenum[0] != 0) { 3852 /* delete or reassign, locking done in del_entry() */ 3853 error = del_entry(chain, rulenum[0]); 3854 } else if (size == 2*sizeof(u_int32_t)) { /* set enable/disable */ 3855 IPFW_UH_WLOCK(chain); 3856 V_set_disable = 3857 (V_set_disable | rulenum[0]) & ~rulenum[1] & 3858 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 3859 IPFW_UH_WUNLOCK(chain); 3860 } else 3861 error = EINVAL; 3862 break; 3863 3864 case IP_FW_ZERO: 3865 case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */ 3866 rulenum[0] = 0; 3867 if (sopt->sopt_val != 0) { 3868 error = sooptcopyin(sopt, rulenum, 3869 sizeof(u_int32_t), sizeof(u_int32_t)); 3870 if (error) 3871 break; 3872 } 3873 error = zero_entry(chain, rulenum[0], 3874 sopt->sopt_name == IP_FW_RESETLOG); 3875 break; 3876 3877 /*--- TABLE opcodes ---*/ 3878 case IP_FW_TABLE_ADD: 3879 case IP_FW_TABLE_DEL: 3880 { 3881 ipfw_table_entry ent; 3882 struct tentry_info tei; 3883 struct tid_info ti; 3884 struct table_value v; 3885 3886 error = sooptcopyin(sopt, &ent, 3887 sizeof(ent), sizeof(ent)); 3888 if (error) 3889 break; 3890 3891 memset(&tei, 0, sizeof(tei)); 3892 tei.paddr = &ent.addr; 3893 tei.subtype = AF_INET; 3894 tei.masklen = ent.masklen; 3895 ipfw_import_table_value_legacy(ent.value, &v); 3896 tei.pvalue = &v; 3897 memset(&ti, 0, sizeof(ti)); 3898 ti.uidx = ent.tbl; 3899 ti.type = IPFW_TABLE_CIDR; 3900 3901 error = (opt == IP_FW_TABLE_ADD) ? 3902 add_table_entry(chain, &ti, &tei, 0, 1) : 3903 del_table_entry(chain, &ti, &tei, 0, 1); 3904 } 3905 break; 3906 3907 3908 case IP_FW_TABLE_FLUSH: 3909 { 3910 u_int16_t tbl; 3911 struct tid_info ti; 3912 3913 error = sooptcopyin(sopt, &tbl, 3914 sizeof(tbl), sizeof(tbl)); 3915 if (error) 3916 break; 3917 memset(&ti, 0, sizeof(ti)); 3918 ti.uidx = tbl; 3919 error = flush_table(chain, &ti); 3920 } 3921 break; 3922 3923 case IP_FW_TABLE_GETSIZE: 3924 { 3925 u_int32_t tbl, cnt; 3926 struct tid_info ti; 3927 3928 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 3929 sizeof(tbl)))) 3930 break; 3931 memset(&ti, 0, sizeof(ti)); 3932 ti.uidx = tbl; 3933 IPFW_RLOCK(chain); 3934 error = ipfw_count_table(chain, &ti, &cnt); 3935 IPFW_RUNLOCK(chain); 3936 if (error) 3937 break; 3938 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 3939 } 3940 break; 3941 3942 case IP_FW_TABLE_LIST: 3943 { 3944 ipfw_table *tbl; 3945 struct tid_info ti; 3946 3947 if (sopt->sopt_valsize < sizeof(*tbl)) { 3948 error = EINVAL; 3949 break; 3950 } 3951 size = sopt->sopt_valsize; 3952 tbl = malloc(size, M_TEMP, M_WAITOK); 3953 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 3954 if (error) { 3955 free(tbl, M_TEMP); 3956 break; 3957 } 3958 tbl->size = (size - sizeof(*tbl)) / 3959 sizeof(ipfw_table_entry); 3960 memset(&ti, 0, sizeof(ti)); 3961 ti.uidx = tbl->tbl; 3962 IPFW_RLOCK(chain); 3963 error = ipfw_dump_table_legacy(chain, &ti, tbl); 3964 IPFW_RUNLOCK(chain); 3965 if (error) { 3966 free(tbl, M_TEMP); 3967 break; 3968 } 3969 error = sooptcopyout(sopt, tbl, size); 3970 free(tbl, M_TEMP); 3971 } 3972 break; 3973 3974 /*--- NAT operations are protected by the IPFW_LOCK ---*/ 3975 case IP_FW_NAT_CFG: 3976 if (IPFW_NAT_LOADED) 3977 error = ipfw_nat_cfg_ptr(sopt); 3978 else { 3979 printf("IP_FW_NAT_CFG: %s\n", 3980 "ipfw_nat not present, please load it"); 3981 error = EINVAL; 3982 } 3983 break; 3984 3985 case IP_FW_NAT_DEL: 3986 if (IPFW_NAT_LOADED) 3987 error = ipfw_nat_del_ptr(sopt); 3988 else { 3989 printf("IP_FW_NAT_DEL: %s\n", 3990 "ipfw_nat not present, please load it"); 3991 error = EINVAL; 3992 } 3993 break; 3994 3995 case IP_FW_NAT_GET_CONFIG: 3996 if (IPFW_NAT_LOADED) 3997 error = ipfw_nat_get_cfg_ptr(sopt); 3998 else { 3999 printf("IP_FW_NAT_GET_CFG: %s\n", 4000 "ipfw_nat not present, please load it"); 4001 error = EINVAL; 4002 } 4003 break; 4004 4005 case IP_FW_NAT_GET_LOG: 4006 if (IPFW_NAT_LOADED) 4007 error = ipfw_nat_get_log_ptr(sopt); 4008 else { 4009 printf("IP_FW_NAT_GET_LOG: %s\n", 4010 "ipfw_nat not present, please load it"); 4011 error = EINVAL; 4012 } 4013 break; 4014 4015 default: 4016 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4017 error = EINVAL; 4018 } 4019 4020 return (error); 4021#undef RULE_MAXSIZE 4022} 4023#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 4024 4025/* Functions to convert rules 7.2 <==> 8.0 */ 4026static int 4027convert_rule_to_7(struct ip_fw_rule0 *rule) 4028{ 4029 /* Used to modify original rule */ 4030 struct ip_fw7 *rule7 = (struct ip_fw7 *)rule; 4031 /* copy of original rule, version 8 */ 4032 struct ip_fw_rule0 *tmp; 4033 4034 /* Used to copy commands */ 4035 ipfw_insn *ccmd, *dst; 4036 int ll = 0, ccmdlen = 0; 4037 4038 tmp = malloc(RULE_MAXSIZE, M_TEMP, M_NOWAIT | M_ZERO); 4039 if (tmp == NULL) { 4040 return 1; //XXX error 4041 } 4042 bcopy(rule, tmp, RULE_MAXSIZE); 4043 4044 /* Copy fields */ 4045 //rule7->_pad = tmp->_pad; 4046 rule7->set = tmp->set; 4047 rule7->rulenum = tmp->rulenum; 4048 rule7->cmd_len = tmp->cmd_len; 4049 rule7->act_ofs = tmp->act_ofs; 4050 rule7->next_rule = (struct ip_fw7 *)tmp->next_rule; 4051 rule7->cmd_len = tmp->cmd_len; 4052 rule7->pcnt = tmp->pcnt; 4053 rule7->bcnt = tmp->bcnt; 4054 rule7->timestamp = tmp->timestamp; 4055 4056 /* Copy commands */ 4057 for (ll = tmp->cmd_len, ccmd = tmp->cmd, dst = rule7->cmd ; 4058 ll > 0 ; ll -= ccmdlen, ccmd += ccmdlen, dst += ccmdlen) { 4059 ccmdlen = F_LEN(ccmd); 4060 4061 bcopy(ccmd, dst, F_LEN(ccmd)*sizeof(uint32_t)); 4062 4063 if (dst->opcode > O_NAT) 4064 /* O_REASS doesn't exists in 7.2 version, so 4065 * decrement opcode if it is after O_REASS 4066 */ 4067 dst->opcode--; 4068 4069 if (ccmdlen > ll) { 4070 printf("ipfw: opcode %d size truncated\n", 4071 ccmd->opcode); 4072 return EINVAL; 4073 } 4074 } 4075 free(tmp, M_TEMP); 4076 4077 return 0; 4078} 4079 4080static int 4081convert_rule_to_8(struct ip_fw_rule0 *rule) 4082{ 4083 /* Used to modify original rule */ 4084 struct ip_fw7 *rule7 = (struct ip_fw7 *) rule; 4085 4086 /* Used to copy commands */ 4087 ipfw_insn *ccmd, *dst; 4088 int ll = 0, ccmdlen = 0; 4089 4090 /* Copy of original rule */ 4091 struct ip_fw7 *tmp = malloc(RULE_MAXSIZE, M_TEMP, M_NOWAIT | M_ZERO); 4092 if (tmp == NULL) { 4093 return 1; //XXX error 4094 } 4095 4096 bcopy(rule7, tmp, RULE_MAXSIZE); 4097 4098 for (ll = tmp->cmd_len, ccmd = tmp->cmd, dst = rule->cmd ; 4099 ll > 0 ; ll -= ccmdlen, ccmd += ccmdlen, dst += ccmdlen) { 4100 ccmdlen = F_LEN(ccmd); 4101 4102 bcopy(ccmd, dst, F_LEN(ccmd)*sizeof(uint32_t)); 4103 4104 if (dst->opcode > O_NAT) 4105 /* O_REASS doesn't exists in 7.2 version, so 4106 * increment opcode if it is after O_REASS 4107 */ 4108 dst->opcode++; 4109 4110 if (ccmdlen > ll) { 4111 printf("ipfw: opcode %d size truncated\n", 4112 ccmd->opcode); 4113 return EINVAL; 4114 } 4115 } 4116 4117 rule->_pad = tmp->_pad; 4118 rule->set = tmp->set; 4119 rule->rulenum = tmp->rulenum; 4120 rule->cmd_len = tmp->cmd_len; 4121 rule->act_ofs = tmp->act_ofs; 4122 rule->next_rule = (struct ip_fw *)tmp->next_rule; 4123 rule->cmd_len = tmp->cmd_len; 4124 rule->id = 0; /* XXX see if is ok = 0 */ 4125 rule->pcnt = tmp->pcnt; 4126 rule->bcnt = tmp->bcnt; 4127 rule->timestamp = tmp->timestamp; 4128 4129 free (tmp, M_TEMP); 4130 return 0; 4131} 4132 4133/* 4134 * Named object api 4135 * 4136 */ 4137 4138void 4139ipfw_init_srv(struct ip_fw_chain *ch) 4140{ 4141 4142 ch->srvmap = ipfw_objhash_create(IPFW_OBJECTS_DEFAULT); 4143 ch->srvstate = malloc(sizeof(void *) * IPFW_OBJECTS_DEFAULT, 4144 M_IPFW, M_WAITOK | M_ZERO); 4145} 4146 4147void 4148ipfw_destroy_srv(struct ip_fw_chain *ch) 4149{ 4150 4151 free(ch->srvstate, M_IPFW); 4152 ipfw_objhash_destroy(ch->srvmap); 4153} 4154 4155/* 4156 * Allocate new bitmask which can be used to enlarge/shrink 4157 * named instance index. 4158 */ 4159void 4160ipfw_objhash_bitmap_alloc(uint32_t items, void **idx, int *pblocks) 4161{ 4162 size_t size; 4163 int max_blocks; 4164 u_long *idx_mask; 4165 4166 KASSERT((items % BLOCK_ITEMS) == 0, 4167 ("bitmask size needs to power of 2 and greater or equal to %zu", 4168 BLOCK_ITEMS)); 4169 4170 max_blocks = items / BLOCK_ITEMS; 4171 size = items / 8; 4172 idx_mask = malloc(size * IPFW_MAX_SETS, M_IPFW, M_WAITOK); 4173 /* Mark all as free */ 4174 memset(idx_mask, 0xFF, size * IPFW_MAX_SETS); 4175 *idx_mask &= ~(u_long)1; /* Skip index 0 */ 4176 4177 *idx = idx_mask; 4178 *pblocks = max_blocks; 4179} 4180 4181/* 4182 * Copy current bitmask index to new one. 4183 */ 4184void 4185ipfw_objhash_bitmap_merge(struct namedobj_instance *ni, void **idx, int *blocks) 4186{ 4187 int old_blocks, new_blocks; 4188 u_long *old_idx, *new_idx; 4189 int i; 4190 4191 old_idx = ni->idx_mask; 4192 old_blocks = ni->max_blocks; 4193 new_idx = *idx; 4194 new_blocks = *blocks; 4195 4196 for (i = 0; i < IPFW_MAX_SETS; i++) { 4197 memcpy(&new_idx[new_blocks * i], &old_idx[old_blocks * i], 4198 old_blocks * sizeof(u_long)); 4199 } 4200} 4201 4202/* 4203 * Swaps current @ni index with new one. 4204 */ 4205void 4206ipfw_objhash_bitmap_swap(struct namedobj_instance *ni, void **idx, int *blocks) 4207{ 4208 int old_blocks; 4209 u_long *old_idx; 4210 4211 old_idx = ni->idx_mask; 4212 old_blocks = ni->max_blocks; 4213 4214 ni->idx_mask = *idx; 4215 ni->max_blocks = *blocks; 4216 4217 /* Save old values */ 4218 *idx = old_idx; 4219 *blocks = old_blocks; 4220} 4221 4222void 4223ipfw_objhash_bitmap_free(void *idx, int blocks) 4224{ 4225 4226 free(idx, M_IPFW); 4227} 4228 4229/* 4230 * Creates named hash instance. 4231 * Must be called without holding any locks. 4232 * Return pointer to new instance. 4233 */ 4234struct namedobj_instance * 4235ipfw_objhash_create(uint32_t items) 4236{ 4237 struct namedobj_instance *ni; 4238 int i; 4239 size_t size; 4240 4241 size = sizeof(struct namedobj_instance) + 4242 sizeof(struct namedobjects_head) * NAMEDOBJ_HASH_SIZE + 4243 sizeof(struct namedobjects_head) * NAMEDOBJ_HASH_SIZE; 4244 4245 ni = malloc(size, M_IPFW, M_WAITOK | M_ZERO); 4246 ni->nn_size = NAMEDOBJ_HASH_SIZE; 4247 ni->nv_size = NAMEDOBJ_HASH_SIZE; 4248 4249 ni->names = (struct namedobjects_head *)(ni +1); 4250 ni->values = &ni->names[ni->nn_size]; 4251 4252 for (i = 0; i < ni->nn_size; i++) 4253 TAILQ_INIT(&ni->names[i]); 4254 4255 for (i = 0; i < ni->nv_size; i++) 4256 TAILQ_INIT(&ni->values[i]); 4257 4258 /* Set default hashing/comparison functions */ 4259 ni->hash_f = objhash_hash_name; 4260 ni->cmp_f = objhash_cmp_name; 4261 4262 /* Allocate bitmask separately due to possible resize */ 4263 ipfw_objhash_bitmap_alloc(items, (void*)&ni->idx_mask, &ni->max_blocks); 4264 4265 return (ni); 4266} 4267 4268void 4269ipfw_objhash_destroy(struct namedobj_instance *ni) 4270{ 4271 4272 free(ni->idx_mask, M_IPFW); 4273 free(ni, M_IPFW); 4274} 4275 4276void 4277ipfw_objhash_set_funcs(struct namedobj_instance *ni, objhash_hash_f *hash_f, 4278 objhash_cmp_f *cmp_f) 4279{ 4280 4281 ni->hash_f = hash_f; 4282 ni->cmp_f = cmp_f; 4283} 4284 4285static uint32_t 4286objhash_hash_name(struct namedobj_instance *ni, const void *name, uint32_t set) 4287{ 4288 4289 return (fnv_32_str((const char *)name, FNV1_32_INIT)); 4290} 4291 4292static int 4293objhash_cmp_name(struct named_object *no, const void *name, uint32_t set) 4294{ 4295 4296 if ((strcmp(no->name, (const char *)name) == 0) && (no->set == set)) 4297 return (0); 4298 4299 return (1); 4300} 4301 4302static uint32_t 4303objhash_hash_idx(struct namedobj_instance *ni, uint32_t val) 4304{ 4305 uint32_t v; 4306 4307 v = val % (ni->nv_size - 1); 4308 4309 return (v); 4310} 4311 4312struct named_object * 4313ipfw_objhash_lookup_name(struct namedobj_instance *ni, uint32_t set, char *name) 4314{ 4315 struct named_object *no; 4316 uint32_t hash; 4317 4318 hash = ni->hash_f(ni, name, set) % ni->nn_size; 4319 4320 TAILQ_FOREACH(no, &ni->names[hash], nn_next) { 4321 if (ni->cmp_f(no, name, set) == 0) 4322 return (no); 4323 } 4324 4325 return (NULL); 4326} 4327 4328/* 4329 * Find named object by @uid. 4330 * Check @tlvs for valid data inside. 4331 * 4332 * Returns pointer to found TLV or NULL. 4333 */ 4334ipfw_obj_ntlv * 4335ipfw_find_name_tlv_type(void *tlvs, int len, uint16_t uidx, uint32_t etlv) 4336{ 4337 ipfw_obj_ntlv *ntlv; 4338 uintptr_t pa, pe; 4339 int l; 4340 4341 pa = (uintptr_t)tlvs; 4342 pe = pa + len; 4343 l = 0; 4344 for (; pa < pe; pa += l) { 4345 ntlv = (ipfw_obj_ntlv *)pa; 4346 l = ntlv->head.length; 4347 4348 if (l != sizeof(*ntlv)) 4349 return (NULL); 4350 4351 if (ntlv->idx != uidx) 4352 continue; 4353 /* 4354 * When userland has specified zero TLV type, do 4355 * not compare it with eltv. In some cases userland 4356 * doesn't know what type should it have. Use only 4357 * uidx and name for search named_object. 4358 */ 4359 if (ntlv->head.type != 0 && 4360 ntlv->head.type != (uint16_t)etlv) 4361 continue; 4362 4363 if (ipfw_check_object_name_generic(ntlv->name) != 0) 4364 return (NULL); 4365 4366 return (ntlv); 4367 } 4368 4369 return (NULL); 4370} 4371 4372/* 4373 * Finds object config based on either legacy index 4374 * or name in ntlv. 4375 * Note @ti structure contains unchecked data from userland. 4376 * 4377 * Returns 0 in success and fills in @pno with found config 4378 */ 4379int 4380ipfw_objhash_find_type(struct namedobj_instance *ni, struct tid_info *ti, 4381 uint32_t etlv, struct named_object **pno) 4382{ 4383 char *name; 4384 ipfw_obj_ntlv *ntlv; 4385 uint32_t set; 4386 4387 if (ti->tlvs == NULL) 4388 return (EINVAL); 4389 4390 ntlv = ipfw_find_name_tlv_type(ti->tlvs, ti->tlen, ti->uidx, etlv); 4391 if (ntlv == NULL) 4392 return (EINVAL); 4393 name = ntlv->name; 4394 4395 /* 4396 * Use set provided by @ti instead of @ntlv one. 4397 * This is needed due to different sets behavior 4398 * controlled by V_fw_tables_sets. 4399 */ 4400 set = ti->set; 4401 *pno = ipfw_objhash_lookup_name(ni, set, name); 4402 if (*pno == NULL) 4403 return (ESRCH); 4404 return (0); 4405} 4406 4407/* 4408 * Find named object by name, considering also its TLV type. 4409 */ 4410struct named_object * 4411ipfw_objhash_lookup_name_type(struct namedobj_instance *ni, uint32_t set, 4412 uint32_t type, const char *name) 4413{ 4414 struct named_object *no; 4415 uint32_t hash; 4416 4417 hash = ni->hash_f(ni, name, set) % ni->nn_size; 4418 4419 TAILQ_FOREACH(no, &ni->names[hash], nn_next) { 4420 if (ni->cmp_f(no, name, set) == 0 && 4421 no->etlv == (uint16_t)type) 4422 return (no); 4423 } 4424 4425 return (NULL); 4426} 4427 4428struct named_object * 4429ipfw_objhash_lookup_kidx(struct namedobj_instance *ni, uint16_t kidx) 4430{ 4431 struct named_object *no; 4432 uint32_t hash; 4433 4434 hash = objhash_hash_idx(ni, kidx); 4435 4436 TAILQ_FOREACH(no, &ni->values[hash], nv_next) { 4437 if (no->kidx == kidx) 4438 return (no); 4439 } 4440 4441 return (NULL); 4442} 4443 4444int 4445ipfw_objhash_same_name(struct namedobj_instance *ni, struct named_object *a, 4446 struct named_object *b) 4447{ 4448 4449 if ((strcmp(a->name, b->name) == 0) && a->set == b->set) 4450 return (1); 4451 4452 return (0); 4453} 4454 4455void 4456ipfw_objhash_add(struct namedobj_instance *ni, struct named_object *no) 4457{ 4458 uint32_t hash; 4459 4460 hash = ni->hash_f(ni, no->name, no->set) % ni->nn_size; 4461 TAILQ_INSERT_HEAD(&ni->names[hash], no, nn_next); 4462 4463 hash = objhash_hash_idx(ni, no->kidx); 4464 TAILQ_INSERT_HEAD(&ni->values[hash], no, nv_next); 4465 4466 ni->count++; 4467} 4468 4469void 4470ipfw_objhash_del(struct namedobj_instance *ni, struct named_object *no) 4471{ 4472 uint32_t hash; 4473 4474 hash = ni->hash_f(ni, no->name, no->set) % ni->nn_size; 4475 TAILQ_REMOVE(&ni->names[hash], no, nn_next); 4476 4477 hash = objhash_hash_idx(ni, no->kidx); 4478 TAILQ_REMOVE(&ni->values[hash], no, nv_next); 4479 4480 ni->count--; 4481} 4482 4483uint32_t 4484ipfw_objhash_count(struct namedobj_instance *ni) 4485{ 4486 4487 return (ni->count); 4488} 4489 4490uint32_t 4491ipfw_objhash_count_type(struct namedobj_instance *ni, uint16_t type) 4492{ 4493 struct named_object *no; 4494 uint32_t count; 4495 int i; 4496 4497 count = 0; 4498 for (i = 0; i < ni->nn_size; i++) { 4499 TAILQ_FOREACH(no, &ni->names[i], nn_next) { 4500 if (no->etlv == type) 4501 count++; 4502 } 4503 } 4504 return (count); 4505} 4506 4507/* 4508 * Runs @func for each found named object. 4509 * It is safe to delete objects from callback 4510 */ 4511int 4512ipfw_objhash_foreach(struct namedobj_instance *ni, objhash_cb_t *f, void *arg) 4513{ 4514 struct named_object *no, *no_tmp; 4515 int i, ret; 4516 4517 for (i = 0; i < ni->nn_size; i++) { 4518 TAILQ_FOREACH_SAFE(no, &ni->names[i], nn_next, no_tmp) { 4519 ret = f(ni, no, arg); 4520 if (ret != 0) 4521 return (ret); 4522 } 4523 } 4524 return (0); 4525} 4526 4527/* 4528 * Runs @f for each found named object with type @type. 4529 * It is safe to delete objects from callback 4530 */ 4531int 4532ipfw_objhash_foreach_type(struct namedobj_instance *ni, objhash_cb_t *f, 4533 void *arg, uint16_t type) 4534{ 4535 struct named_object *no, *no_tmp; 4536 int i, ret; 4537 4538 for (i = 0; i < ni->nn_size; i++) { 4539 TAILQ_FOREACH_SAFE(no, &ni->names[i], nn_next, no_tmp) { 4540 if (no->etlv != type) 4541 continue; 4542 ret = f(ni, no, arg); 4543 if (ret != 0) 4544 return (ret); 4545 } 4546 } 4547 return (0); 4548} 4549 4550/* 4551 * Removes index from given set. 4552 * Returns 0 on success. 4553 */ 4554int 4555ipfw_objhash_free_idx(struct namedobj_instance *ni, uint16_t idx) 4556{ 4557 u_long *mask; 4558 int i, v; 4559 4560 i = idx / BLOCK_ITEMS; 4561 v = idx % BLOCK_ITEMS; 4562 4563 if (i >= ni->max_blocks) 4564 return (1); 4565 4566 mask = &ni->idx_mask[i]; 4567 4568 if ((*mask & ((u_long)1 << v)) != 0) 4569 return (1); 4570 4571 /* Mark as free */ 4572 *mask |= (u_long)1 << v; 4573 4574 /* Update free offset */ 4575 if (ni->free_off[0] > i) 4576 ni->free_off[0] = i; 4577 4578 return (0); 4579} 4580 4581/* 4582 * Allocate new index in given instance and stores in in @pidx. 4583 * Returns 0 on success. 4584 */ 4585int 4586ipfw_objhash_alloc_idx(void *n, uint16_t *pidx) 4587{ 4588 struct namedobj_instance *ni; 4589 u_long *mask; 4590 int i, off, v; 4591 4592 ni = (struct namedobj_instance *)n; 4593 4594 off = ni->free_off[0]; 4595 mask = &ni->idx_mask[off]; 4596 4597 for (i = off; i < ni->max_blocks; i++, mask++) { 4598 if ((v = ffsl(*mask)) == 0) 4599 continue; 4600 4601 /* Mark as busy */ 4602 *mask &= ~ ((u_long)1 << (v - 1)); 4603 4604 ni->free_off[0] = i; 4605 4606 v = BLOCK_ITEMS * i + v - 1; 4607 4608 *pidx = v; 4609 return (0); 4610 } 4611 4612 return (1); 4613} 4614 4615/* end of file */ 4616