ip_fw_table.c revision 300021
1/*- 2 * Copyright (c) 2004 Ruslan Ermilov and Vsevolod Lobko. 3 * Copyright (c) 2014 Yandex LLC 4 * Copyright (c) 2014 Alexander V. Chernikov 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 25 * SUCH DAMAGE. 26 */ 27 28#include <sys/cdefs.h> 29__FBSDID("$FreeBSD: head/sys/netpfil/ipfw/ip_fw_table.c 300021 2016-05-17 07:47:23Z ae $"); 30 31/* 32 * Lookup table support for ipfw. 33 * 34 * This file contains handlers for all generic tables' operations: 35 * add/del/flush entries, list/dump tables etc.. 36 * 37 * Table data modification is protected by both UH and runtime lock 38 * while reading configuration/data is protected by UH lock. 39 * 40 * Lookup algorithms for all table types are located in ip_fw_table_algo.c 41 */ 42 43#include "opt_ipfw.h" 44 45#include <sys/param.h> 46#include <sys/systm.h> 47#include <sys/malloc.h> 48#include <sys/kernel.h> 49#include <sys/lock.h> 50#include <sys/rwlock.h> 51#include <sys/rmlock.h> 52#include <sys/socket.h> 53#include <sys/socketvar.h> 54#include <sys/queue.h> 55#include <net/if.h> /* ip_fw.h requires IFNAMSIZ */ 56 57#include <netinet/in.h> 58#include <netinet/ip_var.h> /* struct ipfw_rule_ref */ 59#include <netinet/ip_fw.h> 60 61#include <netpfil/ipfw/ip_fw_private.h> 62#include <netpfil/ipfw/ip_fw_table.h> 63 64 /* 65 * Table has the following `type` concepts: 66 * 67 * `no.type` represents lookup key type (addr, ifp, uid, etc..) 68 * vmask represents bitmask of table values which are present at the moment. 69 * Special IPFW_VTYPE_LEGACY ( (uint32_t)-1 ) represents old 70 * single-value-for-all approach. 71 */ 72struct table_config { 73 struct named_object no; 74 uint8_t tflags; /* type flags */ 75 uint8_t locked; /* 1 if locked from changes */ 76 uint8_t linked; /* 1 if already linked */ 77 uint8_t ochanged; /* used by set swapping */ 78 uint8_t vshared; /* 1 if using shared value array */ 79 uint8_t spare[3]; 80 uint32_t count; /* Number of records */ 81 uint32_t limit; /* Max number of records */ 82 uint32_t vmask; /* bitmask with supported values */ 83 uint32_t ocount; /* used by set swapping */ 84 uint64_t gencnt; /* generation count */ 85 char tablename[64]; /* table name */ 86 struct table_algo *ta; /* Callbacks for given algo */ 87 void *astate; /* algorithm state */ 88 struct table_info ti_copy; /* data to put to table_info */ 89 struct namedobj_instance *vi; 90}; 91 92static int find_table_err(struct namedobj_instance *ni, struct tid_info *ti, 93 struct table_config **tc); 94static struct table_config *find_table(struct namedobj_instance *ni, 95 struct tid_info *ti); 96static struct table_config *alloc_table_config(struct ip_fw_chain *ch, 97 struct tid_info *ti, struct table_algo *ta, char *adata, uint8_t tflags); 98static void free_table_config(struct namedobj_instance *ni, 99 struct table_config *tc); 100static int create_table_internal(struct ip_fw_chain *ch, struct tid_info *ti, 101 char *aname, ipfw_xtable_info *i, uint16_t *pkidx, int ref); 102static void link_table(struct ip_fw_chain *ch, struct table_config *tc); 103static void unlink_table(struct ip_fw_chain *ch, struct table_config *tc); 104static int find_ref_table(struct ip_fw_chain *ch, struct tid_info *ti, 105 struct tentry_info *tei, uint32_t count, int op, struct table_config **ptc); 106#define OP_ADD 1 107#define OP_DEL 0 108static int export_tables(struct ip_fw_chain *ch, ipfw_obj_lheader *olh, 109 struct sockopt_data *sd); 110static void export_table_info(struct ip_fw_chain *ch, struct table_config *tc, 111 ipfw_xtable_info *i); 112static int dump_table_tentry(void *e, void *arg); 113static int dump_table_xentry(void *e, void *arg); 114 115static int swap_tables(struct ip_fw_chain *ch, struct tid_info *a, 116 struct tid_info *b); 117 118static int check_table_name(const char *name); 119static int check_table_space(struct ip_fw_chain *ch, struct tableop_state *ts, 120 struct table_config *tc, struct table_info *ti, uint32_t count); 121static int destroy_table(struct ip_fw_chain *ch, struct tid_info *ti); 122 123static struct table_algo *find_table_algo(struct tables_config *tableconf, 124 struct tid_info *ti, char *name); 125 126static void objheader_to_ti(struct _ipfw_obj_header *oh, struct tid_info *ti); 127static void ntlv_to_ti(struct _ipfw_obj_ntlv *ntlv, struct tid_info *ti); 128 129#define CHAIN_TO_NI(chain) (CHAIN_TO_TCFG(chain)->namehash) 130#define KIDX_TO_TI(ch, k) (&(((struct table_info *)(ch)->tablestate)[k])) 131 132#define TA_BUF_SZ 128 /* On-stack buffer for add/delete state */ 133 134void 135rollback_toperation_state(struct ip_fw_chain *ch, void *object) 136{ 137 struct tables_config *tcfg; 138 struct op_state *os; 139 140 tcfg = CHAIN_TO_TCFG(ch); 141 TAILQ_FOREACH(os, &tcfg->state_list, next) 142 os->func(object, os); 143} 144 145void 146add_toperation_state(struct ip_fw_chain *ch, struct tableop_state *ts) 147{ 148 struct tables_config *tcfg; 149 150 tcfg = CHAIN_TO_TCFG(ch); 151 TAILQ_INSERT_HEAD(&tcfg->state_list, &ts->opstate, next); 152} 153 154void 155del_toperation_state(struct ip_fw_chain *ch, struct tableop_state *ts) 156{ 157 struct tables_config *tcfg; 158 159 tcfg = CHAIN_TO_TCFG(ch); 160 TAILQ_REMOVE(&tcfg->state_list, &ts->opstate, next); 161} 162 163void 164tc_ref(struct table_config *tc) 165{ 166 167 tc->no.refcnt++; 168} 169 170void 171tc_unref(struct table_config *tc) 172{ 173 174 tc->no.refcnt--; 175} 176 177static struct table_value * 178get_table_value(struct ip_fw_chain *ch, struct table_config *tc, uint32_t kidx) 179{ 180 struct table_value *pval; 181 182 pval = (struct table_value *)ch->valuestate; 183 184 return (&pval[kidx]); 185} 186 187 188/* 189 * Checks if we're able to insert/update entry @tei into table 190 * w.r.t @tc limits. 191 * May alter @tei to indicate insertion error / insert 192 * options. 193 * 194 * Returns 0 if operation can be performed/ 195 */ 196static int 197check_table_limit(struct table_config *tc, struct tentry_info *tei) 198{ 199 200 if (tc->limit == 0 || tc->count < tc->limit) 201 return (0); 202 203 if ((tei->flags & TEI_FLAGS_UPDATE) == 0) { 204 /* Notify userland on error cause */ 205 tei->flags |= TEI_FLAGS_LIMIT; 206 return (EFBIG); 207 } 208 209 /* 210 * We have UPDATE flag set. 211 * Permit updating record (if found), 212 * but restrict adding new one since we've 213 * already hit the limit. 214 */ 215 tei->flags |= TEI_FLAGS_DONTADD; 216 217 return (0); 218} 219 220/* 221 * Convert algorithm callback return code into 222 * one of pre-defined states known by userland. 223 */ 224static void 225store_tei_result(struct tentry_info *tei, int op, int error, uint32_t num) 226{ 227 int flag; 228 229 flag = 0; 230 231 switch (error) { 232 case 0: 233 if (op == OP_ADD && num != 0) 234 flag = TEI_FLAGS_ADDED; 235 if (op == OP_DEL) 236 flag = TEI_FLAGS_DELETED; 237 break; 238 case ENOENT: 239 flag = TEI_FLAGS_NOTFOUND; 240 break; 241 case EEXIST: 242 flag = TEI_FLAGS_EXISTS; 243 break; 244 default: 245 flag = TEI_FLAGS_ERROR; 246 } 247 248 tei->flags |= flag; 249} 250 251/* 252 * Creates and references table with default parameters. 253 * Saves table config, algo and allocated kidx info @ptc, @pta and 254 * @pkidx if non-zero. 255 * Used for table auto-creation to support old binaries. 256 * 257 * Returns 0 on success. 258 */ 259static int 260create_table_compat(struct ip_fw_chain *ch, struct tid_info *ti, 261 uint16_t *pkidx) 262{ 263 ipfw_xtable_info xi; 264 int error; 265 266 memset(&xi, 0, sizeof(xi)); 267 /* Set default value mask for legacy clients */ 268 xi.vmask = IPFW_VTYPE_LEGACY; 269 270 error = create_table_internal(ch, ti, NULL, &xi, pkidx, 1); 271 if (error != 0) 272 return (error); 273 274 return (0); 275} 276 277/* 278 * Find and reference existing table optionally 279 * creating new one. 280 * 281 * Saves found table config into @ptc. 282 * Note function may drop/acquire UH_WLOCK. 283 * Returns 0 if table was found/created and referenced 284 * or non-zero return code. 285 */ 286static int 287find_ref_table(struct ip_fw_chain *ch, struct tid_info *ti, 288 struct tentry_info *tei, uint32_t count, int op, 289 struct table_config **ptc) 290{ 291 struct namedobj_instance *ni; 292 struct table_config *tc; 293 uint16_t kidx; 294 int error; 295 296 IPFW_UH_WLOCK_ASSERT(ch); 297 298 ni = CHAIN_TO_NI(ch); 299 tc = NULL; 300 if ((tc = find_table(ni, ti)) != NULL) { 301 /* check table type */ 302 if (tc->no.subtype != ti->type) 303 return (EINVAL); 304 305 if (tc->locked != 0) 306 return (EACCES); 307 308 /* Try to exit early on limit hit */ 309 if (op == OP_ADD && count == 1 && 310 check_table_limit(tc, tei) != 0) 311 return (EFBIG); 312 313 /* Reference and return */ 314 tc->no.refcnt++; 315 *ptc = tc; 316 return (0); 317 } 318 319 if (op == OP_DEL) 320 return (ESRCH); 321 322 /* Compatibility mode: create new table for old clients */ 323 if ((tei->flags & TEI_FLAGS_COMPAT) == 0) 324 return (ESRCH); 325 326 IPFW_UH_WUNLOCK(ch); 327 error = create_table_compat(ch, ti, &kidx); 328 IPFW_UH_WLOCK(ch); 329 330 if (error != 0) 331 return (error); 332 333 tc = (struct table_config *)ipfw_objhash_lookup_kidx(ni, kidx); 334 KASSERT(tc != NULL, ("create_table_compat returned bad idx %d", kidx)); 335 336 /* OK, now we've got referenced table. */ 337 *ptc = tc; 338 return (0); 339} 340 341/* 342 * Rolls back already @added to @tc entries using state array @ta_buf_m. 343 * Assume the following layout: 344 * 1) ADD state (ta_buf_m[0] ... t_buf_m[added - 1]) for handling update cases 345 * 2) DEL state (ta_buf_m[count[ ... t_buf_m[count + added - 1]) 346 * for storing deleted state 347 */ 348static void 349rollback_added_entries(struct ip_fw_chain *ch, struct table_config *tc, 350 struct table_info *tinfo, struct tentry_info *tei, caddr_t ta_buf_m, 351 uint32_t count, uint32_t added) 352{ 353 struct table_algo *ta; 354 struct tentry_info *ptei; 355 caddr_t v, vv; 356 size_t ta_buf_sz; 357 int error, i; 358 uint32_t num; 359 360 IPFW_UH_WLOCK_ASSERT(ch); 361 362 ta = tc->ta; 363 ta_buf_sz = ta->ta_buf_size; 364 v = ta_buf_m; 365 vv = v + count * ta_buf_sz; 366 for (i = 0; i < added; i++, v += ta_buf_sz, vv += ta_buf_sz) { 367 ptei = &tei[i]; 368 if ((ptei->flags & TEI_FLAGS_UPDATED) != 0) { 369 370 /* 371 * We have old value stored by previous 372 * call in @ptei->value. Do add once again 373 * to restore it. 374 */ 375 error = ta->add(tc->astate, tinfo, ptei, v, &num); 376 KASSERT(error == 0, ("rollback UPDATE fail")); 377 KASSERT(num == 0, ("rollback UPDATE fail2")); 378 continue; 379 } 380 381 error = ta->prepare_del(ch, ptei, vv); 382 KASSERT(error == 0, ("pre-rollback INSERT failed")); 383 error = ta->del(tc->astate, tinfo, ptei, vv, &num); 384 KASSERT(error == 0, ("rollback INSERT failed")); 385 tc->count -= num; 386 } 387} 388 389/* 390 * Prepares add/del state for all @count entries in @tei. 391 * Uses either stack buffer (@ta_buf) or allocates a new one. 392 * Stores pointer to allocated buffer back to @ta_buf. 393 * 394 * Returns 0 on success. 395 */ 396static int 397prepare_batch_buffer(struct ip_fw_chain *ch, struct table_algo *ta, 398 struct tentry_info *tei, uint32_t count, int op, caddr_t *ta_buf) 399{ 400 caddr_t ta_buf_m, v; 401 size_t ta_buf_sz, sz; 402 struct tentry_info *ptei; 403 int error, i; 404 405 error = 0; 406 ta_buf_sz = ta->ta_buf_size; 407 if (count == 1) { 408 /* Sigle add/delete, use on-stack buffer */ 409 memset(*ta_buf, 0, TA_BUF_SZ); 410 ta_buf_m = *ta_buf; 411 } else { 412 413 /* 414 * Multiple adds/deletes, allocate larger buffer 415 * 416 * Note we need 2xcount buffer for add case: 417 * we have hold both ADD state 418 * and DELETE state (this may be needed 419 * if we need to rollback all changes) 420 */ 421 sz = count * ta_buf_sz; 422 ta_buf_m = malloc((op == OP_ADD) ? sz * 2 : sz, M_TEMP, 423 M_WAITOK | M_ZERO); 424 } 425 426 v = ta_buf_m; 427 for (i = 0; i < count; i++, v += ta_buf_sz) { 428 ptei = &tei[i]; 429 error = (op == OP_ADD) ? 430 ta->prepare_add(ch, ptei, v) : ta->prepare_del(ch, ptei, v); 431 432 /* 433 * Some syntax error (incorrect mask, or address, or 434 * anything). Return error regardless of atomicity 435 * settings. 436 */ 437 if (error != 0) 438 break; 439 } 440 441 *ta_buf = ta_buf_m; 442 return (error); 443} 444 445/* 446 * Flushes allocated state for each @count entries in @tei. 447 * Frees @ta_buf_m if differs from stack buffer @ta_buf. 448 */ 449static void 450flush_batch_buffer(struct ip_fw_chain *ch, struct table_algo *ta, 451 struct tentry_info *tei, uint32_t count, int rollback, 452 caddr_t ta_buf_m, caddr_t ta_buf) 453{ 454 caddr_t v; 455 struct tentry_info *ptei; 456 size_t ta_buf_sz; 457 int i; 458 459 ta_buf_sz = ta->ta_buf_size; 460 461 /* Run cleaning callback anyway */ 462 v = ta_buf_m; 463 for (i = 0; i < count; i++, v += ta_buf_sz) { 464 ptei = &tei[i]; 465 ta->flush_entry(ch, ptei, v); 466 if (ptei->ptv != NULL) { 467 free(ptei->ptv, M_IPFW); 468 ptei->ptv = NULL; 469 } 470 } 471 472 /* Clean up "deleted" state in case of rollback */ 473 if (rollback != 0) { 474 v = ta_buf_m + count * ta_buf_sz; 475 for (i = 0; i < count; i++, v += ta_buf_sz) 476 ta->flush_entry(ch, &tei[i], v); 477 } 478 479 if (ta_buf_m != ta_buf) 480 free(ta_buf_m, M_TEMP); 481} 482 483 484static void 485rollback_add_entry(void *object, struct op_state *_state) 486{ 487 struct ip_fw_chain *ch; 488 struct tableop_state *ts; 489 490 ts = (struct tableop_state *)_state; 491 492 if (ts->tc != object && ts->ch != object) 493 return; 494 495 ch = ts->ch; 496 497 IPFW_UH_WLOCK_ASSERT(ch); 498 499 /* Call specifid unlockers */ 500 rollback_table_values(ts); 501 502 /* Indicate we've called */ 503 ts->modified = 1; 504} 505 506/* 507 * Adds/updates one or more entries in table @ti. 508 * 509 * Function may drop/reacquire UH wlock multiple times due to 510 * items alloc, algorithm callbacks (check_space), value linkage 511 * (new values, value storage realloc), etc.. 512 * Other processes like other adds (which may involve storage resize), 513 * table swaps (which changes table data and may change algo type), 514 * table modify (which may change value mask) may be executed 515 * simultaneously so we need to deal with it. 516 * 517 * The following approach was implemented: 518 * we have per-chain linked list, protected with UH lock. 519 * add_table_entry prepares special on-stack structure wthich is passed 520 * to its descendants. Users add this structure to this list before unlock. 521 * After performing needed operations and acquiring UH lock back, each user 522 * checks if structure has changed. If true, it rolls local state back and 523 * returns without error to the caller. 524 * add_table_entry() on its own checks if structure has changed and restarts 525 * its operation from the beginning (goto restart). 526 * 527 * Functions which are modifying fields of interest (currently 528 * resize_shared_value_storage() and swap_tables() ) 529 * traverses given list while holding UH lock immediately before 530 * performing their operations calling function provided be list entry 531 * ( currently rollback_add_entry ) which performs rollback for all necessary 532 * state and sets appropriate values in structure indicating rollback 533 * has happened. 534 * 535 * Algo interaction: 536 * Function references @ti first to ensure table won't 537 * disappear or change its type. 538 * After that, prepare_add callback is called for each @tei entry. 539 * Next, we try to add each entry under UH+WHLOCK 540 * using add() callback. 541 * Finally, we free all state by calling flush_entry callback 542 * for each @tei. 543 * 544 * Returns 0 on success. 545 */ 546int 547add_table_entry(struct ip_fw_chain *ch, struct tid_info *ti, 548 struct tentry_info *tei, uint8_t flags, uint32_t count) 549{ 550 struct table_config *tc; 551 struct table_algo *ta; 552 uint16_t kidx; 553 int error, first_error, i, rollback; 554 uint32_t num, numadd; 555 struct tentry_info *ptei; 556 struct tableop_state ts; 557 char ta_buf[TA_BUF_SZ]; 558 caddr_t ta_buf_m, v; 559 560 memset(&ts, 0, sizeof(ts)); 561 ta = NULL; 562 IPFW_UH_WLOCK(ch); 563 564 /* 565 * Find and reference existing table. 566 */ 567restart: 568 if (ts.modified != 0) { 569 IPFW_UH_WUNLOCK(ch); 570 flush_batch_buffer(ch, ta, tei, count, rollback, 571 ta_buf_m, ta_buf); 572 memset(&ts, 0, sizeof(ts)); 573 ta = NULL; 574 IPFW_UH_WLOCK(ch); 575 } 576 577 error = find_ref_table(ch, ti, tei, count, OP_ADD, &tc); 578 if (error != 0) { 579 IPFW_UH_WUNLOCK(ch); 580 return (error); 581 } 582 ta = tc->ta; 583 584 /* Fill in tablestate */ 585 ts.ch = ch; 586 ts.opstate.func = rollback_add_entry; 587 ts.tc = tc; 588 ts.vshared = tc->vshared; 589 ts.vmask = tc->vmask; 590 ts.ta = ta; 591 ts.tei = tei; 592 ts.count = count; 593 rollback = 0; 594 add_toperation_state(ch, &ts); 595 IPFW_UH_WUNLOCK(ch); 596 597 /* Allocate memory and prepare record(s) */ 598 /* Pass stack buffer by default */ 599 ta_buf_m = ta_buf; 600 error = prepare_batch_buffer(ch, ta, tei, count, OP_ADD, &ta_buf_m); 601 602 IPFW_UH_WLOCK(ch); 603 del_toperation_state(ch, &ts); 604 /* Drop reference we've used in first search */ 605 tc->no.refcnt--; 606 607 /* Check prepare_batch_buffer() error */ 608 if (error != 0) 609 goto cleanup; 610 611 /* 612 * Check if table swap has happened. 613 * (so table algo might be changed). 614 * Restart operation to achieve consistent behavior. 615 */ 616 if (ts.modified != 0) 617 goto restart; 618 619 /* 620 * Link all values values to shared/per-table value array. 621 * 622 * May release/reacquire UH_WLOCK. 623 */ 624 error = ipfw_link_table_values(ch, &ts); 625 if (error != 0) 626 goto cleanup; 627 if (ts.modified != 0) 628 goto restart; 629 630 /* 631 * Ensure we are able to add all entries without additional 632 * memory allocations. May release/reacquire UH_WLOCK. 633 */ 634 kidx = tc->no.kidx; 635 error = check_table_space(ch, &ts, tc, KIDX_TO_TI(ch, kidx), count); 636 if (error != 0) 637 goto cleanup; 638 if (ts.modified != 0) 639 goto restart; 640 641 /* We've got valid table in @tc. Let's try to add data */ 642 kidx = tc->no.kidx; 643 ta = tc->ta; 644 numadd = 0; 645 first_error = 0; 646 647 IPFW_WLOCK(ch); 648 649 v = ta_buf_m; 650 for (i = 0; i < count; i++, v += ta->ta_buf_size) { 651 ptei = &tei[i]; 652 num = 0; 653 /* check limit before adding */ 654 if ((error = check_table_limit(tc, ptei)) == 0) { 655 error = ta->add(tc->astate, KIDX_TO_TI(ch, kidx), 656 ptei, v, &num); 657 /* Set status flag to inform userland */ 658 store_tei_result(ptei, OP_ADD, error, num); 659 } 660 if (error == 0) { 661 /* Update number of records to ease limit checking */ 662 tc->count += num; 663 numadd += num; 664 continue; 665 } 666 667 if (first_error == 0) 668 first_error = error; 669 670 /* 671 * Some error have happened. Check our atomicity 672 * settings: continue if atomicity is not required, 673 * rollback changes otherwise. 674 */ 675 if ((flags & IPFW_CTF_ATOMIC) == 0) 676 continue; 677 678 rollback_added_entries(ch, tc, KIDX_TO_TI(ch, kidx), 679 tei, ta_buf_m, count, i); 680 681 rollback = 1; 682 break; 683 } 684 685 IPFW_WUNLOCK(ch); 686 687 ipfw_garbage_table_values(ch, tc, tei, count, rollback); 688 689 /* Permit post-add algorithm grow/rehash. */ 690 if (numadd != 0) 691 check_table_space(ch, NULL, tc, KIDX_TO_TI(ch, kidx), 0); 692 693 /* Return first error to user, if any */ 694 error = first_error; 695 696cleanup: 697 IPFW_UH_WUNLOCK(ch); 698 699 flush_batch_buffer(ch, ta, tei, count, rollback, ta_buf_m, ta_buf); 700 701 return (error); 702} 703 704/* 705 * Deletes one or more entries in table @ti. 706 * 707 * Returns 0 on success. 708 */ 709int 710del_table_entry(struct ip_fw_chain *ch, struct tid_info *ti, 711 struct tentry_info *tei, uint8_t flags, uint32_t count) 712{ 713 struct table_config *tc; 714 struct table_algo *ta; 715 struct tentry_info *ptei; 716 uint16_t kidx; 717 int error, first_error, i; 718 uint32_t num, numdel; 719 char ta_buf[TA_BUF_SZ]; 720 caddr_t ta_buf_m, v; 721 722 /* 723 * Find and reference existing table. 724 */ 725 IPFW_UH_WLOCK(ch); 726 error = find_ref_table(ch, ti, tei, count, OP_DEL, &tc); 727 if (error != 0) { 728 IPFW_UH_WUNLOCK(ch); 729 return (error); 730 } 731 ta = tc->ta; 732 IPFW_UH_WUNLOCK(ch); 733 734 /* Allocate memory and prepare record(s) */ 735 /* Pass stack buffer by default */ 736 ta_buf_m = ta_buf; 737 error = prepare_batch_buffer(ch, ta, tei, count, OP_DEL, &ta_buf_m); 738 if (error != 0) 739 goto cleanup; 740 741 IPFW_UH_WLOCK(ch); 742 743 /* Drop reference we've used in first search */ 744 tc->no.refcnt--; 745 746 /* 747 * Check if table algo is still the same. 748 * (changed ta may be the result of table swap). 749 */ 750 if (ta != tc->ta) { 751 IPFW_UH_WUNLOCK(ch); 752 error = EINVAL; 753 goto cleanup; 754 } 755 756 kidx = tc->no.kidx; 757 numdel = 0; 758 first_error = 0; 759 760 IPFW_WLOCK(ch); 761 v = ta_buf_m; 762 for (i = 0; i < count; i++, v += ta->ta_buf_size) { 763 ptei = &tei[i]; 764 num = 0; 765 error = ta->del(tc->astate, KIDX_TO_TI(ch, kidx), ptei, v, 766 &num); 767 /* Save state for userland */ 768 store_tei_result(ptei, OP_DEL, error, num); 769 if (error != 0 && first_error == 0) 770 first_error = error; 771 tc->count -= num; 772 numdel += num; 773 } 774 IPFW_WUNLOCK(ch); 775 776 /* Unlink non-used values */ 777 ipfw_garbage_table_values(ch, tc, tei, count, 0); 778 779 if (numdel != 0) { 780 /* Run post-del hook to permit shrinking */ 781 check_table_space(ch, NULL, tc, KIDX_TO_TI(ch, kidx), 0); 782 } 783 784 IPFW_UH_WUNLOCK(ch); 785 786 /* Return first error to user, if any */ 787 error = first_error; 788 789cleanup: 790 flush_batch_buffer(ch, ta, tei, count, 0, ta_buf_m, ta_buf); 791 792 return (error); 793} 794 795/* 796 * Ensure that table @tc has enough space to add @count entries without 797 * need for reallocation. 798 * 799 * Callbacks order: 800 * 0) need_modify() (UH_WLOCK) - checks if @count items can be added w/o resize. 801 * 802 * 1) alloc_modify (no locks, M_WAITOK) - alloc new state based on @pflags. 803 * 2) prepare_modifyt (UH_WLOCK) - copy old data into new storage 804 * 3) modify (UH_WLOCK + WLOCK) - switch pointers 805 * 4) flush_modify (UH_WLOCK) - free state, if needed 806 * 807 * Returns 0 on success. 808 */ 809static int 810check_table_space(struct ip_fw_chain *ch, struct tableop_state *ts, 811 struct table_config *tc, struct table_info *ti, uint32_t count) 812{ 813 struct table_algo *ta; 814 uint64_t pflags; 815 char ta_buf[TA_BUF_SZ]; 816 int error; 817 818 IPFW_UH_WLOCK_ASSERT(ch); 819 820 error = 0; 821 ta = tc->ta; 822 if (ta->need_modify == NULL) 823 return (0); 824 825 /* Acquire reference not to loose @tc between locks/unlocks */ 826 tc->no.refcnt++; 827 828 /* 829 * TODO: think about avoiding race between large add/large delete 830 * operation on algorithm which implements shrinking along with 831 * growing. 832 */ 833 while (true) { 834 pflags = 0; 835 if (ta->need_modify(tc->astate, ti, count, &pflags) == 0) { 836 error = 0; 837 break; 838 } 839 840 /* We have to shrink/grow table */ 841 if (ts != NULL) 842 add_toperation_state(ch, ts); 843 IPFW_UH_WUNLOCK(ch); 844 845 memset(&ta_buf, 0, sizeof(ta_buf)); 846 error = ta->prepare_mod(ta_buf, &pflags); 847 848 IPFW_UH_WLOCK(ch); 849 if (ts != NULL) 850 del_toperation_state(ch, ts); 851 852 if (error != 0) 853 break; 854 855 if (ts != NULL && ts->modified != 0) { 856 857 /* 858 * Swap operation has happened 859 * so we're currently operating on other 860 * table data. Stop doing this. 861 */ 862 ta->flush_mod(ta_buf); 863 break; 864 } 865 866 /* Check if we still need to alter table */ 867 ti = KIDX_TO_TI(ch, tc->no.kidx); 868 if (ta->need_modify(tc->astate, ti, count, &pflags) == 0) { 869 IPFW_UH_WUNLOCK(ch); 870 871 /* 872 * Other thread has already performed resize. 873 * Flush our state and return. 874 */ 875 ta->flush_mod(ta_buf); 876 break; 877 } 878 879 error = ta->fill_mod(tc->astate, ti, ta_buf, &pflags); 880 if (error == 0) { 881 /* Do actual modification */ 882 IPFW_WLOCK(ch); 883 ta->modify(tc->astate, ti, ta_buf, pflags); 884 IPFW_WUNLOCK(ch); 885 } 886 887 /* Anyway, flush data and retry */ 888 ta->flush_mod(ta_buf); 889 } 890 891 tc->no.refcnt--; 892 return (error); 893} 894 895/* 896 * Adds or deletes record in table. 897 * Data layout (v0): 898 * Request: [ ip_fw3_opheader ipfw_table_xentry ] 899 * 900 * Returns 0 on success 901 */ 902static int 903manage_table_ent_v0(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 904 struct sockopt_data *sd) 905{ 906 ipfw_table_xentry *xent; 907 struct tentry_info tei; 908 struct tid_info ti; 909 struct table_value v; 910 int error, hdrlen, read; 911 912 hdrlen = offsetof(ipfw_table_xentry, k); 913 914 /* Check minimum header size */ 915 if (sd->valsize < (sizeof(*op3) + hdrlen)) 916 return (EINVAL); 917 918 read = sizeof(ip_fw3_opheader); 919 920 /* Check if xentry len field is valid */ 921 xent = (ipfw_table_xentry *)(op3 + 1); 922 if (xent->len < hdrlen || xent->len + read > sd->valsize) 923 return (EINVAL); 924 925 memset(&tei, 0, sizeof(tei)); 926 tei.paddr = &xent->k; 927 tei.masklen = xent->masklen; 928 ipfw_import_table_value_legacy(xent->value, &v); 929 tei.pvalue = &v; 930 /* Old requests compatibility */ 931 tei.flags = TEI_FLAGS_COMPAT; 932 if (xent->type == IPFW_TABLE_ADDR) { 933 if (xent->len - hdrlen == sizeof(in_addr_t)) 934 tei.subtype = AF_INET; 935 else 936 tei.subtype = AF_INET6; 937 } 938 939 memset(&ti, 0, sizeof(ti)); 940 ti.uidx = xent->tbl; 941 ti.type = xent->type; 942 943 error = (op3->opcode == IP_FW_TABLE_XADD) ? 944 add_table_entry(ch, &ti, &tei, 0, 1) : 945 del_table_entry(ch, &ti, &tei, 0, 1); 946 947 return (error); 948} 949 950/* 951 * Adds or deletes record in table. 952 * Data layout (v1)(current): 953 * Request: [ ipfw_obj_header 954 * ipfw_obj_ctlv(IPFW_TLV_TBLENT_LIST) [ ipfw_obj_tentry x N ] 955 * ] 956 * 957 * Returns 0 on success 958 */ 959static int 960manage_table_ent_v1(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 961 struct sockopt_data *sd) 962{ 963 ipfw_obj_tentry *tent, *ptent; 964 ipfw_obj_ctlv *ctlv; 965 ipfw_obj_header *oh; 966 struct tentry_info *ptei, tei, *tei_buf; 967 struct tid_info ti; 968 int error, i, kidx, read; 969 970 /* Check minimum header size */ 971 if (sd->valsize < (sizeof(*oh) + sizeof(*ctlv))) 972 return (EINVAL); 973 974 /* Check if passed data is too long */ 975 if (sd->valsize != sd->kavail) 976 return (EINVAL); 977 978 oh = (ipfw_obj_header *)sd->kbuf; 979 980 /* Basic length checks for TLVs */ 981 if (oh->ntlv.head.length != sizeof(oh->ntlv)) 982 return (EINVAL); 983 984 read = sizeof(*oh); 985 986 ctlv = (ipfw_obj_ctlv *)(oh + 1); 987 if (ctlv->head.length + read != sd->valsize) 988 return (EINVAL); 989 990 read += sizeof(*ctlv); 991 tent = (ipfw_obj_tentry *)(ctlv + 1); 992 if (ctlv->count * sizeof(*tent) + read != sd->valsize) 993 return (EINVAL); 994 995 if (ctlv->count == 0) 996 return (0); 997 998 /* 999 * Mark entire buffer as "read". 1000 * This instructs sopt api write it back 1001 * after function return. 1002 */ 1003 ipfw_get_sopt_header(sd, sd->valsize); 1004 1005 /* Perform basic checks for each entry */ 1006 ptent = tent; 1007 kidx = tent->idx; 1008 for (i = 0; i < ctlv->count; i++, ptent++) { 1009 if (ptent->head.length != sizeof(*ptent)) 1010 return (EINVAL); 1011 if (ptent->idx != kidx) 1012 return (ENOTSUP); 1013 } 1014 1015 /* Convert data into kernel request objects */ 1016 objheader_to_ti(oh, &ti); 1017 ti.type = oh->ntlv.type; 1018 ti.uidx = kidx; 1019 1020 /* Use on-stack buffer for single add/del */ 1021 if (ctlv->count == 1) { 1022 memset(&tei, 0, sizeof(tei)); 1023 tei_buf = &tei; 1024 } else 1025 tei_buf = malloc(ctlv->count * sizeof(tei), M_TEMP, 1026 M_WAITOK | M_ZERO); 1027 1028 ptei = tei_buf; 1029 ptent = tent; 1030 for (i = 0; i < ctlv->count; i++, ptent++, ptei++) { 1031 ptei->paddr = &ptent->k; 1032 ptei->subtype = ptent->subtype; 1033 ptei->masklen = ptent->masklen; 1034 if (ptent->head.flags & IPFW_TF_UPDATE) 1035 ptei->flags |= TEI_FLAGS_UPDATE; 1036 1037 ipfw_import_table_value_v1(&ptent->v.value); 1038 ptei->pvalue = (struct table_value *)&ptent->v.value; 1039 } 1040 1041 error = (oh->opheader.opcode == IP_FW_TABLE_XADD) ? 1042 add_table_entry(ch, &ti, tei_buf, ctlv->flags, ctlv->count) : 1043 del_table_entry(ch, &ti, tei_buf, ctlv->flags, ctlv->count); 1044 1045 /* Translate result back to userland */ 1046 ptei = tei_buf; 1047 ptent = tent; 1048 for (i = 0; i < ctlv->count; i++, ptent++, ptei++) { 1049 if (ptei->flags & TEI_FLAGS_ADDED) 1050 ptent->result = IPFW_TR_ADDED; 1051 else if (ptei->flags & TEI_FLAGS_DELETED) 1052 ptent->result = IPFW_TR_DELETED; 1053 else if (ptei->flags & TEI_FLAGS_UPDATED) 1054 ptent->result = IPFW_TR_UPDATED; 1055 else if (ptei->flags & TEI_FLAGS_LIMIT) 1056 ptent->result = IPFW_TR_LIMIT; 1057 else if (ptei->flags & TEI_FLAGS_ERROR) 1058 ptent->result = IPFW_TR_ERROR; 1059 else if (ptei->flags & TEI_FLAGS_NOTFOUND) 1060 ptent->result = IPFW_TR_NOTFOUND; 1061 else if (ptei->flags & TEI_FLAGS_EXISTS) 1062 ptent->result = IPFW_TR_EXISTS; 1063 ipfw_export_table_value_v1(ptei->pvalue, &ptent->v.value); 1064 } 1065 1066 if (tei_buf != &tei) 1067 free(tei_buf, M_TEMP); 1068 1069 return (error); 1070} 1071 1072/* 1073 * Looks up an entry in given table. 1074 * Data layout (v0)(current): 1075 * Request: [ ipfw_obj_header ipfw_obj_tentry ] 1076 * Reply: [ ipfw_obj_header ipfw_obj_tentry ] 1077 * 1078 * Returns 0 on success 1079 */ 1080static int 1081find_table_entry(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 1082 struct sockopt_data *sd) 1083{ 1084 ipfw_obj_tentry *tent; 1085 ipfw_obj_header *oh; 1086 struct tid_info ti; 1087 struct table_config *tc; 1088 struct table_algo *ta; 1089 struct table_info *kti; 1090 struct namedobj_instance *ni; 1091 int error; 1092 size_t sz; 1093 1094 /* Check minimum header size */ 1095 sz = sizeof(*oh) + sizeof(*tent); 1096 if (sd->valsize != sz) 1097 return (EINVAL); 1098 1099 oh = (struct _ipfw_obj_header *)ipfw_get_sopt_header(sd, sz); 1100 tent = (ipfw_obj_tentry *)(oh + 1); 1101 1102 /* Basic length checks for TLVs */ 1103 if (oh->ntlv.head.length != sizeof(oh->ntlv)) 1104 return (EINVAL); 1105 1106 objheader_to_ti(oh, &ti); 1107 ti.type = oh->ntlv.type; 1108 ti.uidx = tent->idx; 1109 1110 IPFW_UH_RLOCK(ch); 1111 ni = CHAIN_TO_NI(ch); 1112 1113 /* 1114 * Find existing table and check its type . 1115 */ 1116 ta = NULL; 1117 if ((tc = find_table(ni, &ti)) == NULL) { 1118 IPFW_UH_RUNLOCK(ch); 1119 return (ESRCH); 1120 } 1121 1122 /* check table type */ 1123 if (tc->no.subtype != ti.type) { 1124 IPFW_UH_RUNLOCK(ch); 1125 return (EINVAL); 1126 } 1127 1128 kti = KIDX_TO_TI(ch, tc->no.kidx); 1129 ta = tc->ta; 1130 1131 if (ta->find_tentry == NULL) 1132 return (ENOTSUP); 1133 1134 error = ta->find_tentry(tc->astate, kti, tent); 1135 1136 IPFW_UH_RUNLOCK(ch); 1137 1138 return (error); 1139} 1140 1141/* 1142 * Flushes all entries or destroys given table. 1143 * Data layout (v0)(current): 1144 * Request: [ ipfw_obj_header ] 1145 * 1146 * Returns 0 on success 1147 */ 1148static int 1149flush_table_v0(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 1150 struct sockopt_data *sd) 1151{ 1152 int error; 1153 struct _ipfw_obj_header *oh; 1154 struct tid_info ti; 1155 1156 if (sd->valsize != sizeof(*oh)) 1157 return (EINVAL); 1158 1159 oh = (struct _ipfw_obj_header *)op3; 1160 objheader_to_ti(oh, &ti); 1161 1162 if (op3->opcode == IP_FW_TABLE_XDESTROY) 1163 error = destroy_table(ch, &ti); 1164 else if (op3->opcode == IP_FW_TABLE_XFLUSH) 1165 error = flush_table(ch, &ti); 1166 else 1167 return (ENOTSUP); 1168 1169 return (error); 1170} 1171 1172static void 1173restart_flush(void *object, struct op_state *_state) 1174{ 1175 struct tableop_state *ts; 1176 1177 ts = (struct tableop_state *)_state; 1178 1179 if (ts->tc != object) 1180 return; 1181 1182 /* Indicate we've called */ 1183 ts->modified = 1; 1184} 1185 1186/* 1187 * Flushes given table. 1188 * 1189 * Function create new table instance with the same 1190 * parameters, swaps it with old one and 1191 * flushes state without holding runtime WLOCK. 1192 * 1193 * Returns 0 on success. 1194 */ 1195int 1196flush_table(struct ip_fw_chain *ch, struct tid_info *ti) 1197{ 1198 struct namedobj_instance *ni; 1199 struct table_config *tc; 1200 struct table_algo *ta; 1201 struct table_info ti_old, ti_new, *tablestate; 1202 void *astate_old, *astate_new; 1203 char algostate[64], *pstate; 1204 struct tableop_state ts; 1205 int error, need_gc; 1206 uint16_t kidx; 1207 uint8_t tflags; 1208 1209 /* 1210 * Stage 1: save table algorithm. 1211 * Reference found table to ensure it won't disappear. 1212 */ 1213 IPFW_UH_WLOCK(ch); 1214 ni = CHAIN_TO_NI(ch); 1215 if ((tc = find_table(ni, ti)) == NULL) { 1216 IPFW_UH_WUNLOCK(ch); 1217 return (ESRCH); 1218 } 1219 need_gc = 0; 1220 astate_new = NULL; 1221 memset(&ti_new, 0, sizeof(ti_new)); 1222restart: 1223 /* Set up swap handler */ 1224 memset(&ts, 0, sizeof(ts)); 1225 ts.opstate.func = restart_flush; 1226 ts.tc = tc; 1227 1228 ta = tc->ta; 1229 /* Do not flush readonly tables */ 1230 if ((ta->flags & TA_FLAG_READONLY) != 0) { 1231 IPFW_UH_WUNLOCK(ch); 1232 return (EACCES); 1233 } 1234 /* Save startup algo parameters */ 1235 if (ta->print_config != NULL) { 1236 ta->print_config(tc->astate, KIDX_TO_TI(ch, tc->no.kidx), 1237 algostate, sizeof(algostate)); 1238 pstate = algostate; 1239 } else 1240 pstate = NULL; 1241 tflags = tc->tflags; 1242 tc->no.refcnt++; 1243 add_toperation_state(ch, &ts); 1244 IPFW_UH_WUNLOCK(ch); 1245 1246 /* 1247 * Stage 1.5: if this is not the first attempt, destroy previous state 1248 */ 1249 if (need_gc != 0) { 1250 ta->destroy(astate_new, &ti_new); 1251 need_gc = 0; 1252 } 1253 1254 /* 1255 * Stage 2: allocate new table instance using same algo. 1256 */ 1257 memset(&ti_new, 0, sizeof(struct table_info)); 1258 error = ta->init(ch, &astate_new, &ti_new, pstate, tflags); 1259 1260 /* 1261 * Stage 3: swap old state pointers with newly-allocated ones. 1262 * Decrease refcount. 1263 */ 1264 IPFW_UH_WLOCK(ch); 1265 tc->no.refcnt--; 1266 del_toperation_state(ch, &ts); 1267 1268 if (error != 0) { 1269 IPFW_UH_WUNLOCK(ch); 1270 return (error); 1271 } 1272 1273 /* 1274 * Restart operation if table swap has happened: 1275 * even if algo may be the same, algo init parameters 1276 * may change. Restart operation instead of doing 1277 * complex checks. 1278 */ 1279 if (ts.modified != 0) { 1280 /* Delay destroying data since we're holding UH lock */ 1281 need_gc = 1; 1282 goto restart; 1283 } 1284 1285 ni = CHAIN_TO_NI(ch); 1286 kidx = tc->no.kidx; 1287 tablestate = (struct table_info *)ch->tablestate; 1288 1289 IPFW_WLOCK(ch); 1290 ti_old = tablestate[kidx]; 1291 tablestate[kidx] = ti_new; 1292 IPFW_WUNLOCK(ch); 1293 1294 astate_old = tc->astate; 1295 tc->astate = astate_new; 1296 tc->ti_copy = ti_new; 1297 tc->count = 0; 1298 1299 /* Notify algo on real @ti address */ 1300 if (ta->change_ti != NULL) 1301 ta->change_ti(tc->astate, &tablestate[kidx]); 1302 1303 /* 1304 * Stage 4: unref values. 1305 */ 1306 ipfw_unref_table_values(ch, tc, ta, astate_old, &ti_old); 1307 IPFW_UH_WUNLOCK(ch); 1308 1309 /* 1310 * Stage 5: perform real flush/destroy. 1311 */ 1312 ta->destroy(astate_old, &ti_old); 1313 1314 return (0); 1315} 1316 1317/* 1318 * Swaps two tables. 1319 * Data layout (v0)(current): 1320 * Request: [ ipfw_obj_header ipfw_obj_ntlv ] 1321 * 1322 * Returns 0 on success 1323 */ 1324static int 1325swap_table(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 1326 struct sockopt_data *sd) 1327{ 1328 int error; 1329 struct _ipfw_obj_header *oh; 1330 struct tid_info ti_a, ti_b; 1331 1332 if (sd->valsize != sizeof(*oh) + sizeof(ipfw_obj_ntlv)) 1333 return (EINVAL); 1334 1335 oh = (struct _ipfw_obj_header *)op3; 1336 ntlv_to_ti(&oh->ntlv, &ti_a); 1337 ntlv_to_ti((ipfw_obj_ntlv *)(oh + 1), &ti_b); 1338 1339 error = swap_tables(ch, &ti_a, &ti_b); 1340 1341 return (error); 1342} 1343 1344/* 1345 * Swaps two tables of the same type/valtype. 1346 * 1347 * Checks if tables are compatible and limits 1348 * permits swap, than actually perform swap. 1349 * 1350 * Each table consists of 2 different parts: 1351 * config: 1352 * @tc (with name, set, kidx) and rule bindings, which is "stable". 1353 * number of items 1354 * table algo 1355 * runtime: 1356 * runtime data @ti (ch->tablestate) 1357 * runtime cache in @tc 1358 * algo-specific data (@tc->astate) 1359 * 1360 * So we switch: 1361 * all runtime data 1362 * number of items 1363 * table algo 1364 * 1365 * After that we call @ti change handler for each table. 1366 * 1367 * Note that referencing @tc won't protect tc->ta from change. 1368 * XXX: Do we need to restrict swap between locked tables? 1369 * XXX: Do we need to exchange ftype? 1370 * 1371 * Returns 0 on success. 1372 */ 1373static int 1374swap_tables(struct ip_fw_chain *ch, struct tid_info *a, 1375 struct tid_info *b) 1376{ 1377 struct namedobj_instance *ni; 1378 struct table_config *tc_a, *tc_b; 1379 struct table_algo *ta; 1380 struct table_info ti, *tablestate; 1381 void *astate; 1382 uint32_t count; 1383 1384 /* 1385 * Stage 1: find both tables and ensure they are of 1386 * the same type. 1387 */ 1388 IPFW_UH_WLOCK(ch); 1389 ni = CHAIN_TO_NI(ch); 1390 if ((tc_a = find_table(ni, a)) == NULL) { 1391 IPFW_UH_WUNLOCK(ch); 1392 return (ESRCH); 1393 } 1394 if ((tc_b = find_table(ni, b)) == NULL) { 1395 IPFW_UH_WUNLOCK(ch); 1396 return (ESRCH); 1397 } 1398 1399 /* It is very easy to swap between the same table */ 1400 if (tc_a == tc_b) { 1401 IPFW_UH_WUNLOCK(ch); 1402 return (0); 1403 } 1404 1405 /* Check type and value are the same */ 1406 if (tc_a->no.subtype!=tc_b->no.subtype || tc_a->tflags!=tc_b->tflags) { 1407 IPFW_UH_WUNLOCK(ch); 1408 return (EINVAL); 1409 } 1410 1411 /* Check limits before swap */ 1412 if ((tc_a->limit != 0 && tc_b->count > tc_a->limit) || 1413 (tc_b->limit != 0 && tc_a->count > tc_b->limit)) { 1414 IPFW_UH_WUNLOCK(ch); 1415 return (EFBIG); 1416 } 1417 1418 /* Check if one of the tables is readonly */ 1419 if (((tc_a->ta->flags | tc_b->ta->flags) & TA_FLAG_READONLY) != 0) { 1420 IPFW_UH_WUNLOCK(ch); 1421 return (EACCES); 1422 } 1423 1424 /* Notify we're going to swap */ 1425 rollback_toperation_state(ch, tc_a); 1426 rollback_toperation_state(ch, tc_b); 1427 1428 /* Everything is fine, prepare to swap */ 1429 tablestate = (struct table_info *)ch->tablestate; 1430 ti = tablestate[tc_a->no.kidx]; 1431 ta = tc_a->ta; 1432 astate = tc_a->astate; 1433 count = tc_a->count; 1434 1435 IPFW_WLOCK(ch); 1436 /* a <- b */ 1437 tablestate[tc_a->no.kidx] = tablestate[tc_b->no.kidx]; 1438 tc_a->ta = tc_b->ta; 1439 tc_a->astate = tc_b->astate; 1440 tc_a->count = tc_b->count; 1441 /* b <- a */ 1442 tablestate[tc_b->no.kidx] = ti; 1443 tc_b->ta = ta; 1444 tc_b->astate = astate; 1445 tc_b->count = count; 1446 IPFW_WUNLOCK(ch); 1447 1448 /* Ensure tc.ti copies are in sync */ 1449 tc_a->ti_copy = tablestate[tc_a->no.kidx]; 1450 tc_b->ti_copy = tablestate[tc_b->no.kidx]; 1451 1452 /* Notify both tables on @ti change */ 1453 if (tc_a->ta->change_ti != NULL) 1454 tc_a->ta->change_ti(tc_a->astate, &tablestate[tc_a->no.kidx]); 1455 if (tc_b->ta->change_ti != NULL) 1456 tc_b->ta->change_ti(tc_b->astate, &tablestate[tc_b->no.kidx]); 1457 1458 IPFW_UH_WUNLOCK(ch); 1459 1460 return (0); 1461} 1462 1463/* 1464 * Destroys table specified by @ti. 1465 * Data layout (v0)(current): 1466 * Request: [ ip_fw3_opheader ] 1467 * 1468 * Returns 0 on success 1469 */ 1470static int 1471destroy_table(struct ip_fw_chain *ch, struct tid_info *ti) 1472{ 1473 struct namedobj_instance *ni; 1474 struct table_config *tc; 1475 1476 IPFW_UH_WLOCK(ch); 1477 1478 ni = CHAIN_TO_NI(ch); 1479 if ((tc = find_table(ni, ti)) == NULL) { 1480 IPFW_UH_WUNLOCK(ch); 1481 return (ESRCH); 1482 } 1483 1484 /* Do not permit destroying referenced tables */ 1485 if (tc->no.refcnt > 0) { 1486 IPFW_UH_WUNLOCK(ch); 1487 return (EBUSY); 1488 } 1489 1490 IPFW_WLOCK(ch); 1491 unlink_table(ch, tc); 1492 IPFW_WUNLOCK(ch); 1493 1494 /* Free obj index */ 1495 if (ipfw_objhash_free_idx(ni, tc->no.kidx) != 0) 1496 printf("Error unlinking kidx %d from table %s\n", 1497 tc->no.kidx, tc->tablename); 1498 1499 /* Unref values used in tables while holding UH lock */ 1500 ipfw_unref_table_values(ch, tc, tc->ta, tc->astate, &tc->ti_copy); 1501 IPFW_UH_WUNLOCK(ch); 1502 1503 free_table_config(ni, tc); 1504 1505 return (0); 1506} 1507 1508static uint32_t 1509roundup2p(uint32_t v) 1510{ 1511 1512 v--; 1513 v |= v >> 1; 1514 v |= v >> 2; 1515 v |= v >> 4; 1516 v |= v >> 8; 1517 v |= v >> 16; 1518 v++; 1519 1520 return (v); 1521} 1522 1523/* 1524 * Grow tables index. 1525 * 1526 * Returns 0 on success. 1527 */ 1528int 1529ipfw_resize_tables(struct ip_fw_chain *ch, unsigned int ntables) 1530{ 1531 unsigned int ntables_old, tbl; 1532 struct namedobj_instance *ni; 1533 void *new_idx, *old_tablestate, *tablestate; 1534 struct table_info *ti; 1535 struct table_config *tc; 1536 int i, new_blocks; 1537 1538 /* Check new value for validity */ 1539 if (ntables == 0) 1540 return (EINVAL); 1541 if (ntables > IPFW_TABLES_MAX) 1542 ntables = IPFW_TABLES_MAX; 1543 /* Alight to nearest power of 2 */ 1544 ntables = (unsigned int)roundup2p(ntables); 1545 1546 /* Allocate new pointers */ 1547 tablestate = malloc(ntables * sizeof(struct table_info), 1548 M_IPFW, M_WAITOK | M_ZERO); 1549 1550 ipfw_objhash_bitmap_alloc(ntables, (void *)&new_idx, &new_blocks); 1551 1552 IPFW_UH_WLOCK(ch); 1553 1554 tbl = (ntables >= V_fw_tables_max) ? V_fw_tables_max : ntables; 1555 ni = CHAIN_TO_NI(ch); 1556 1557 /* Temporary restrict decreasing max_tables */ 1558 if (ntables < V_fw_tables_max) { 1559 1560 /* 1561 * FIXME: Check if we really can shrink 1562 */ 1563 IPFW_UH_WUNLOCK(ch); 1564 return (EINVAL); 1565 } 1566 1567 /* Copy table info/indices */ 1568 memcpy(tablestate, ch->tablestate, sizeof(struct table_info) * tbl); 1569 ipfw_objhash_bitmap_merge(ni, &new_idx, &new_blocks); 1570 1571 IPFW_WLOCK(ch); 1572 1573 /* Change pointers */ 1574 old_tablestate = ch->tablestate; 1575 ch->tablestate = tablestate; 1576 ipfw_objhash_bitmap_swap(ni, &new_idx, &new_blocks); 1577 1578 ntables_old = V_fw_tables_max; 1579 V_fw_tables_max = ntables; 1580 1581 IPFW_WUNLOCK(ch); 1582 1583 /* Notify all consumers that their @ti pointer has changed */ 1584 ti = (struct table_info *)ch->tablestate; 1585 for (i = 0; i < tbl; i++, ti++) { 1586 if (ti->lookup == NULL) 1587 continue; 1588 tc = (struct table_config *)ipfw_objhash_lookup_kidx(ni, i); 1589 if (tc == NULL || tc->ta->change_ti == NULL) 1590 continue; 1591 1592 tc->ta->change_ti(tc->astate, ti); 1593 } 1594 1595 IPFW_UH_WUNLOCK(ch); 1596 1597 /* Free old pointers */ 1598 free(old_tablestate, M_IPFW); 1599 ipfw_objhash_bitmap_free(new_idx, new_blocks); 1600 1601 return (0); 1602} 1603 1604/* 1605 * Lookup an IP @addr in table @tbl. 1606 * Stores found value in @val. 1607 * 1608 * Returns 1 if @addr was found. 1609 */ 1610int 1611ipfw_lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1612 uint32_t *val) 1613{ 1614 struct table_info *ti; 1615 1616 ti = KIDX_TO_TI(ch, tbl); 1617 1618 return (ti->lookup(ti, &addr, sizeof(in_addr_t), val)); 1619} 1620 1621/* 1622 * Lookup an arbtrary key @paddr of legth @plen in table @tbl. 1623 * Stores found value in @val. 1624 * 1625 * Returns 1 if key was found. 1626 */ 1627int 1628ipfw_lookup_table_extended(struct ip_fw_chain *ch, uint16_t tbl, uint16_t plen, 1629 void *paddr, uint32_t *val) 1630{ 1631 struct table_info *ti; 1632 1633 ti = KIDX_TO_TI(ch, tbl); 1634 1635 return (ti->lookup(ti, paddr, plen, val)); 1636} 1637 1638/* 1639 * Info/List/dump support for tables. 1640 * 1641 */ 1642 1643/* 1644 * High-level 'get' cmds sysctl handlers 1645 */ 1646 1647/* 1648 * Lists all tables currently available in kernel. 1649 * Data layout (v0)(current): 1650 * Request: [ ipfw_obj_lheader ], size = ipfw_obj_lheader.size 1651 * Reply: [ ipfw_obj_lheader ipfw_xtable_info x N ] 1652 * 1653 * Returns 0 on success 1654 */ 1655static int 1656list_tables(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 1657 struct sockopt_data *sd) 1658{ 1659 struct _ipfw_obj_lheader *olh; 1660 int error; 1661 1662 olh = (struct _ipfw_obj_lheader *)ipfw_get_sopt_header(sd,sizeof(*olh)); 1663 if (olh == NULL) 1664 return (EINVAL); 1665 if (sd->valsize < olh->size) 1666 return (EINVAL); 1667 1668 IPFW_UH_RLOCK(ch); 1669 error = export_tables(ch, olh, sd); 1670 IPFW_UH_RUNLOCK(ch); 1671 1672 return (error); 1673} 1674 1675/* 1676 * Store table info to buffer provided by @sd. 1677 * Data layout (v0)(current): 1678 * Request: [ ipfw_obj_header ipfw_xtable_info(empty)] 1679 * Reply: [ ipfw_obj_header ipfw_xtable_info ] 1680 * 1681 * Returns 0 on success. 1682 */ 1683static int 1684describe_table(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 1685 struct sockopt_data *sd) 1686{ 1687 struct _ipfw_obj_header *oh; 1688 struct table_config *tc; 1689 struct tid_info ti; 1690 size_t sz; 1691 1692 sz = sizeof(*oh) + sizeof(ipfw_xtable_info); 1693 oh = (struct _ipfw_obj_header *)ipfw_get_sopt_header(sd, sz); 1694 if (oh == NULL) 1695 return (EINVAL); 1696 1697 objheader_to_ti(oh, &ti); 1698 1699 IPFW_UH_RLOCK(ch); 1700 if ((tc = find_table(CHAIN_TO_NI(ch), &ti)) == NULL) { 1701 IPFW_UH_RUNLOCK(ch); 1702 return (ESRCH); 1703 } 1704 1705 export_table_info(ch, tc, (ipfw_xtable_info *)(oh + 1)); 1706 IPFW_UH_RUNLOCK(ch); 1707 1708 return (0); 1709} 1710 1711/* 1712 * Modifies existing table. 1713 * Data layout (v0)(current): 1714 * Request: [ ipfw_obj_header ipfw_xtable_info ] 1715 * 1716 * Returns 0 on success 1717 */ 1718static int 1719modify_table(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 1720 struct sockopt_data *sd) 1721{ 1722 struct _ipfw_obj_header *oh; 1723 ipfw_xtable_info *i; 1724 char *tname; 1725 struct tid_info ti; 1726 struct namedobj_instance *ni; 1727 struct table_config *tc; 1728 1729 if (sd->valsize != sizeof(*oh) + sizeof(ipfw_xtable_info)) 1730 return (EINVAL); 1731 1732 oh = (struct _ipfw_obj_header *)sd->kbuf; 1733 i = (ipfw_xtable_info *)(oh + 1); 1734 1735 /* 1736 * Verify user-supplied strings. 1737 * Check for null-terminated/zero-length strings/ 1738 */ 1739 tname = oh->ntlv.name; 1740 if (check_table_name(tname) != 0) 1741 return (EINVAL); 1742 1743 objheader_to_ti(oh, &ti); 1744 ti.type = i->type; 1745 1746 IPFW_UH_WLOCK(ch); 1747 ni = CHAIN_TO_NI(ch); 1748 if ((tc = find_table(ni, &ti)) == NULL) { 1749 IPFW_UH_WUNLOCK(ch); 1750 return (ESRCH); 1751 } 1752 1753 /* Do not support any modifications for readonly tables */ 1754 if ((tc->ta->flags & TA_FLAG_READONLY) != 0) { 1755 IPFW_UH_WUNLOCK(ch); 1756 return (EACCES); 1757 } 1758 1759 if ((i->mflags & IPFW_TMFLAGS_LIMIT) != 0) 1760 tc->limit = i->limit; 1761 if ((i->mflags & IPFW_TMFLAGS_LOCK) != 0) 1762 tc->locked = ((i->flags & IPFW_TGFLAGS_LOCKED) != 0); 1763 IPFW_UH_WUNLOCK(ch); 1764 1765 return (0); 1766} 1767 1768/* 1769 * Creates new table. 1770 * Data layout (v0)(current): 1771 * Request: [ ipfw_obj_header ipfw_xtable_info ] 1772 * 1773 * Returns 0 on success 1774 */ 1775static int 1776create_table(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 1777 struct sockopt_data *sd) 1778{ 1779 struct _ipfw_obj_header *oh; 1780 ipfw_xtable_info *i; 1781 char *tname, *aname; 1782 struct tid_info ti; 1783 struct namedobj_instance *ni; 1784 1785 if (sd->valsize != sizeof(*oh) + sizeof(ipfw_xtable_info)) 1786 return (EINVAL); 1787 1788 oh = (struct _ipfw_obj_header *)sd->kbuf; 1789 i = (ipfw_xtable_info *)(oh + 1); 1790 1791 /* 1792 * Verify user-supplied strings. 1793 * Check for null-terminated/zero-length strings/ 1794 */ 1795 tname = oh->ntlv.name; 1796 aname = i->algoname; 1797 if (check_table_name(tname) != 0 || 1798 strnlen(aname, sizeof(i->algoname)) == sizeof(i->algoname)) 1799 return (EINVAL); 1800 1801 if (aname[0] == '\0') { 1802 /* Use default algorithm */ 1803 aname = NULL; 1804 } 1805 1806 objheader_to_ti(oh, &ti); 1807 ti.type = i->type; 1808 1809 ni = CHAIN_TO_NI(ch); 1810 1811 IPFW_UH_RLOCK(ch); 1812 if (find_table(ni, &ti) != NULL) { 1813 IPFW_UH_RUNLOCK(ch); 1814 return (EEXIST); 1815 } 1816 IPFW_UH_RUNLOCK(ch); 1817 1818 return (create_table_internal(ch, &ti, aname, i, NULL, 0)); 1819} 1820 1821/* 1822 * Creates new table based on @ti and @aname. 1823 * 1824 * Assume @aname to be checked and valid. 1825 * Stores allocated table kidx inside @pkidx (if non-NULL). 1826 * Reference created table if @compat is non-zero. 1827 * 1828 * Returns 0 on success. 1829 */ 1830static int 1831create_table_internal(struct ip_fw_chain *ch, struct tid_info *ti, 1832 char *aname, ipfw_xtable_info *i, uint16_t *pkidx, int compat) 1833{ 1834 struct namedobj_instance *ni; 1835 struct table_config *tc, *tc_new, *tmp; 1836 struct table_algo *ta; 1837 uint16_t kidx; 1838 1839 ni = CHAIN_TO_NI(ch); 1840 1841 ta = find_table_algo(CHAIN_TO_TCFG(ch), ti, aname); 1842 if (ta == NULL) 1843 return (ENOTSUP); 1844 1845 tc = alloc_table_config(ch, ti, ta, aname, i->tflags); 1846 if (tc == NULL) 1847 return (ENOMEM); 1848 1849 tc->vmask = i->vmask; 1850 tc->limit = i->limit; 1851 if (ta->flags & TA_FLAG_READONLY) 1852 tc->locked = 1; 1853 else 1854 tc->locked = (i->flags & IPFW_TGFLAGS_LOCKED) != 0; 1855 1856 IPFW_UH_WLOCK(ch); 1857 1858 /* Check if table has been already created */ 1859 tc_new = find_table(ni, ti); 1860 if (tc_new != NULL) { 1861 1862 /* 1863 * Compat: do not fail if we're 1864 * requesting to create existing table 1865 * which has the same type 1866 */ 1867 if (compat == 0 || tc_new->no.subtype != tc->no.subtype) { 1868 IPFW_UH_WUNLOCK(ch); 1869 free_table_config(ni, tc); 1870 return (EEXIST); 1871 } 1872 1873 /* Exchange tc and tc_new for proper refcounting & freeing */ 1874 tmp = tc; 1875 tc = tc_new; 1876 tc_new = tmp; 1877 } else { 1878 /* New table */ 1879 if (ipfw_objhash_alloc_idx(ni, &kidx) != 0) { 1880 IPFW_UH_WUNLOCK(ch); 1881 printf("Unable to allocate table index." 1882 " Consider increasing net.inet.ip.fw.tables_max"); 1883 free_table_config(ni, tc); 1884 return (EBUSY); 1885 } 1886 tc->no.kidx = kidx; 1887 tc->no.etlv = IPFW_TLV_TBL_NAME; 1888 1889 IPFW_WLOCK(ch); 1890 link_table(ch, tc); 1891 IPFW_WUNLOCK(ch); 1892 } 1893 1894 if (compat != 0) 1895 tc->no.refcnt++; 1896 if (pkidx != NULL) 1897 *pkidx = tc->no.kidx; 1898 1899 IPFW_UH_WUNLOCK(ch); 1900 1901 if (tc_new != NULL) 1902 free_table_config(ni, tc_new); 1903 1904 return (0); 1905} 1906 1907static void 1908ntlv_to_ti(ipfw_obj_ntlv *ntlv, struct tid_info *ti) 1909{ 1910 1911 memset(ti, 0, sizeof(struct tid_info)); 1912 ti->set = ntlv->set; 1913 ti->uidx = ntlv->idx; 1914 ti->tlvs = ntlv; 1915 ti->tlen = ntlv->head.length; 1916} 1917 1918static void 1919objheader_to_ti(struct _ipfw_obj_header *oh, struct tid_info *ti) 1920{ 1921 1922 ntlv_to_ti(&oh->ntlv, ti); 1923} 1924 1925struct namedobj_instance * 1926ipfw_get_table_objhash(struct ip_fw_chain *ch) 1927{ 1928 1929 return (CHAIN_TO_NI(ch)); 1930} 1931 1932/* 1933 * Exports basic table info as name TLV. 1934 * Used inside dump_static_rules() to provide info 1935 * about all tables referenced by current ruleset. 1936 * 1937 * Returns 0 on success. 1938 */ 1939int 1940ipfw_export_table_ntlv(struct ip_fw_chain *ch, uint16_t kidx, 1941 struct sockopt_data *sd) 1942{ 1943 struct namedobj_instance *ni; 1944 struct named_object *no; 1945 ipfw_obj_ntlv *ntlv; 1946 1947 ni = CHAIN_TO_NI(ch); 1948 1949 no = ipfw_objhash_lookup_kidx(ni, kidx); 1950 KASSERT(no != NULL, ("invalid table kidx passed")); 1951 1952 ntlv = (ipfw_obj_ntlv *)ipfw_get_sopt_space(sd, sizeof(*ntlv)); 1953 if (ntlv == NULL) 1954 return (ENOMEM); 1955 1956 ntlv->head.type = IPFW_TLV_TBL_NAME; 1957 ntlv->head.length = sizeof(*ntlv); 1958 ntlv->idx = no->kidx; 1959 strlcpy(ntlv->name, no->name, sizeof(ntlv->name)); 1960 1961 return (0); 1962} 1963 1964struct dump_args { 1965 struct ip_fw_chain *ch; 1966 struct table_info *ti; 1967 struct table_config *tc; 1968 struct sockopt_data *sd; 1969 uint32_t cnt; 1970 uint16_t uidx; 1971 int error; 1972 uint32_t size; 1973 ipfw_table_entry *ent; 1974 ta_foreach_f *f; 1975 void *farg; 1976 ipfw_obj_tentry tent; 1977}; 1978 1979static int 1980count_ext_entries(void *e, void *arg) 1981{ 1982 struct dump_args *da; 1983 1984 da = (struct dump_args *)arg; 1985 da->cnt++; 1986 1987 return (0); 1988} 1989 1990/* 1991 * Gets number of items from table either using 1992 * internal counter or calling algo callback for 1993 * externally-managed tables. 1994 * 1995 * Returns number of records. 1996 */ 1997static uint32_t 1998table_get_count(struct ip_fw_chain *ch, struct table_config *tc) 1999{ 2000 struct table_info *ti; 2001 struct table_algo *ta; 2002 struct dump_args da; 2003 2004 ti = KIDX_TO_TI(ch, tc->no.kidx); 2005 ta = tc->ta; 2006 2007 /* Use internal counter for self-managed tables */ 2008 if ((ta->flags & TA_FLAG_READONLY) == 0) 2009 return (tc->count); 2010 2011 /* Use callback to quickly get number of items */ 2012 if ((ta->flags & TA_FLAG_EXTCOUNTER) != 0) 2013 return (ta->get_count(tc->astate, ti)); 2014 2015 /* Count number of iterms ourselves */ 2016 memset(&da, 0, sizeof(da)); 2017 ta->foreach(tc->astate, ti, count_ext_entries, &da); 2018 2019 return (da.cnt); 2020} 2021 2022/* 2023 * Exports table @tc info into standard ipfw_xtable_info format. 2024 */ 2025static void 2026export_table_info(struct ip_fw_chain *ch, struct table_config *tc, 2027 ipfw_xtable_info *i) 2028{ 2029 struct table_info *ti; 2030 struct table_algo *ta; 2031 2032 i->type = tc->no.subtype; 2033 i->tflags = tc->tflags; 2034 i->vmask = tc->vmask; 2035 i->set = tc->no.set; 2036 i->kidx = tc->no.kidx; 2037 i->refcnt = tc->no.refcnt; 2038 i->count = table_get_count(ch, tc); 2039 i->limit = tc->limit; 2040 i->flags |= (tc->locked != 0) ? IPFW_TGFLAGS_LOCKED : 0; 2041 i->size = i->count * sizeof(ipfw_obj_tentry); 2042 i->size += sizeof(ipfw_obj_header) + sizeof(ipfw_xtable_info); 2043 strlcpy(i->tablename, tc->tablename, sizeof(i->tablename)); 2044 ti = KIDX_TO_TI(ch, tc->no.kidx); 2045 ta = tc->ta; 2046 if (ta->print_config != NULL) { 2047 /* Use algo function to print table config to string */ 2048 ta->print_config(tc->astate, ti, i->algoname, 2049 sizeof(i->algoname)); 2050 } else 2051 strlcpy(i->algoname, ta->name, sizeof(i->algoname)); 2052 /* Dump algo-specific data, if possible */ 2053 if (ta->dump_tinfo != NULL) { 2054 ta->dump_tinfo(tc->astate, ti, &i->ta_info); 2055 i->ta_info.flags |= IPFW_TATFLAGS_DATA; 2056 } 2057} 2058 2059struct dump_table_args { 2060 struct ip_fw_chain *ch; 2061 struct sockopt_data *sd; 2062}; 2063 2064static int 2065export_table_internal(struct namedobj_instance *ni, struct named_object *no, 2066 void *arg) 2067{ 2068 ipfw_xtable_info *i; 2069 struct dump_table_args *dta; 2070 2071 dta = (struct dump_table_args *)arg; 2072 2073 i = (ipfw_xtable_info *)ipfw_get_sopt_space(dta->sd, sizeof(*i)); 2074 KASSERT(i != NULL, ("previously checked buffer is not enough")); 2075 2076 export_table_info(dta->ch, (struct table_config *)no, i); 2077 return (0); 2078} 2079 2080/* 2081 * Export all tables as ipfw_xtable_info structures to 2082 * storage provided by @sd. 2083 * 2084 * If supplied buffer is too small, fills in required size 2085 * and returns ENOMEM. 2086 * Returns 0 on success. 2087 */ 2088static int 2089export_tables(struct ip_fw_chain *ch, ipfw_obj_lheader *olh, 2090 struct sockopt_data *sd) 2091{ 2092 uint32_t size; 2093 uint32_t count; 2094 struct dump_table_args dta; 2095 2096 count = ipfw_objhash_count(CHAIN_TO_NI(ch)); 2097 size = count * sizeof(ipfw_xtable_info) + sizeof(ipfw_obj_lheader); 2098 2099 /* Fill in header regadless of buffer size */ 2100 olh->count = count; 2101 olh->objsize = sizeof(ipfw_xtable_info); 2102 2103 if (size > olh->size) { 2104 olh->size = size; 2105 return (ENOMEM); 2106 } 2107 2108 olh->size = size; 2109 2110 dta.ch = ch; 2111 dta.sd = sd; 2112 2113 ipfw_objhash_foreach(CHAIN_TO_NI(ch), export_table_internal, &dta); 2114 2115 return (0); 2116} 2117 2118/* 2119 * Dumps all table data 2120 * Data layout (v1)(current): 2121 * Request: [ ipfw_obj_header ], size = ipfw_xtable_info.size 2122 * Reply: [ ipfw_obj_header ipfw_xtable_info ipfw_obj_tentry x N ] 2123 * 2124 * Returns 0 on success 2125 */ 2126static int 2127dump_table_v1(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 2128 struct sockopt_data *sd) 2129{ 2130 struct _ipfw_obj_header *oh; 2131 ipfw_xtable_info *i; 2132 struct tid_info ti; 2133 struct table_config *tc; 2134 struct table_algo *ta; 2135 struct dump_args da; 2136 uint32_t sz; 2137 2138 sz = sizeof(ipfw_obj_header) + sizeof(ipfw_xtable_info); 2139 oh = (struct _ipfw_obj_header *)ipfw_get_sopt_header(sd, sz); 2140 if (oh == NULL) 2141 return (EINVAL); 2142 2143 i = (ipfw_xtable_info *)(oh + 1); 2144 objheader_to_ti(oh, &ti); 2145 2146 IPFW_UH_RLOCK(ch); 2147 if ((tc = find_table(CHAIN_TO_NI(ch), &ti)) == NULL) { 2148 IPFW_UH_RUNLOCK(ch); 2149 return (ESRCH); 2150 } 2151 export_table_info(ch, tc, i); 2152 2153 if (sd->valsize < i->size) { 2154 2155 /* 2156 * Submitted buffer size is not enough. 2157 * WE've already filled in @i structure with 2158 * relevant table info including size, so we 2159 * can return. Buffer will be flushed automatically. 2160 */ 2161 IPFW_UH_RUNLOCK(ch); 2162 return (ENOMEM); 2163 } 2164 2165 /* 2166 * Do the actual dump in eXtended format 2167 */ 2168 memset(&da, 0, sizeof(da)); 2169 da.ch = ch; 2170 da.ti = KIDX_TO_TI(ch, tc->no.kidx); 2171 da.tc = tc; 2172 da.sd = sd; 2173 2174 ta = tc->ta; 2175 2176 ta->foreach(tc->astate, da.ti, dump_table_tentry, &da); 2177 IPFW_UH_RUNLOCK(ch); 2178 2179 return (da.error); 2180} 2181 2182/* 2183 * Dumps all table data 2184 * Data layout (version 0)(legacy): 2185 * Request: [ ipfw_xtable ], size = IP_FW_TABLE_XGETSIZE() 2186 * Reply: [ ipfw_xtable ipfw_table_xentry x N ] 2187 * 2188 * Returns 0 on success 2189 */ 2190static int 2191dump_table_v0(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 2192 struct sockopt_data *sd) 2193{ 2194 ipfw_xtable *xtbl; 2195 struct tid_info ti; 2196 struct table_config *tc; 2197 struct table_algo *ta; 2198 struct dump_args da; 2199 size_t sz, count; 2200 2201 xtbl = (ipfw_xtable *)ipfw_get_sopt_header(sd, sizeof(ipfw_xtable)); 2202 if (xtbl == NULL) 2203 return (EINVAL); 2204 2205 memset(&ti, 0, sizeof(ti)); 2206 ti.uidx = xtbl->tbl; 2207 2208 IPFW_UH_RLOCK(ch); 2209 if ((tc = find_table(CHAIN_TO_NI(ch), &ti)) == NULL) { 2210 IPFW_UH_RUNLOCK(ch); 2211 return (0); 2212 } 2213 count = table_get_count(ch, tc); 2214 sz = count * sizeof(ipfw_table_xentry) + sizeof(ipfw_xtable); 2215 2216 xtbl->cnt = count; 2217 xtbl->size = sz; 2218 xtbl->type = tc->no.subtype; 2219 xtbl->tbl = ti.uidx; 2220 2221 if (sd->valsize < sz) { 2222 2223 /* 2224 * Submitted buffer size is not enough. 2225 * WE've already filled in @i structure with 2226 * relevant table info including size, so we 2227 * can return. Buffer will be flushed automatically. 2228 */ 2229 IPFW_UH_RUNLOCK(ch); 2230 return (ENOMEM); 2231 } 2232 2233 /* Do the actual dump in eXtended format */ 2234 memset(&da, 0, sizeof(da)); 2235 da.ch = ch; 2236 da.ti = KIDX_TO_TI(ch, tc->no.kidx); 2237 da.tc = tc; 2238 da.sd = sd; 2239 2240 ta = tc->ta; 2241 2242 ta->foreach(tc->astate, da.ti, dump_table_xentry, &da); 2243 IPFW_UH_RUNLOCK(ch); 2244 2245 return (0); 2246} 2247 2248/* 2249 * Legacy function to retrieve number of items in table. 2250 */ 2251static int 2252get_table_size(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 2253 struct sockopt_data *sd) 2254{ 2255 uint32_t *tbl; 2256 struct tid_info ti; 2257 size_t sz; 2258 int error; 2259 2260 sz = sizeof(*op3) + sizeof(uint32_t); 2261 op3 = (ip_fw3_opheader *)ipfw_get_sopt_header(sd, sz); 2262 if (op3 == NULL) 2263 return (EINVAL); 2264 2265 tbl = (uint32_t *)(op3 + 1); 2266 memset(&ti, 0, sizeof(ti)); 2267 ti.uidx = *tbl; 2268 IPFW_UH_RLOCK(ch); 2269 error = ipfw_count_xtable(ch, &ti, tbl); 2270 IPFW_UH_RUNLOCK(ch); 2271 return (error); 2272} 2273 2274/* 2275 * Legacy IP_FW_TABLE_GETSIZE handler 2276 */ 2277int 2278ipfw_count_table(struct ip_fw_chain *ch, struct tid_info *ti, uint32_t *cnt) 2279{ 2280 struct table_config *tc; 2281 2282 if ((tc = find_table(CHAIN_TO_NI(ch), ti)) == NULL) 2283 return (ESRCH); 2284 *cnt = table_get_count(ch, tc); 2285 return (0); 2286} 2287 2288/* 2289 * Legacy IP_FW_TABLE_XGETSIZE handler 2290 */ 2291int 2292ipfw_count_xtable(struct ip_fw_chain *ch, struct tid_info *ti, uint32_t *cnt) 2293{ 2294 struct table_config *tc; 2295 uint32_t count; 2296 2297 if ((tc = find_table(CHAIN_TO_NI(ch), ti)) == NULL) { 2298 *cnt = 0; 2299 return (0); /* 'table all list' requires success */ 2300 } 2301 2302 count = table_get_count(ch, tc); 2303 *cnt = count * sizeof(ipfw_table_xentry); 2304 if (count > 0) 2305 *cnt += sizeof(ipfw_xtable); 2306 return (0); 2307} 2308 2309static int 2310dump_table_entry(void *e, void *arg) 2311{ 2312 struct dump_args *da; 2313 struct table_config *tc; 2314 struct table_algo *ta; 2315 ipfw_table_entry *ent; 2316 struct table_value *pval; 2317 int error; 2318 2319 da = (struct dump_args *)arg; 2320 2321 tc = da->tc; 2322 ta = tc->ta; 2323 2324 /* Out of memory, returning */ 2325 if (da->cnt == da->size) 2326 return (1); 2327 ent = da->ent++; 2328 ent->tbl = da->uidx; 2329 da->cnt++; 2330 2331 error = ta->dump_tentry(tc->astate, da->ti, e, &da->tent); 2332 if (error != 0) 2333 return (error); 2334 2335 ent->addr = da->tent.k.addr.s_addr; 2336 ent->masklen = da->tent.masklen; 2337 pval = get_table_value(da->ch, da->tc, da->tent.v.kidx); 2338 ent->value = ipfw_export_table_value_legacy(pval); 2339 2340 return (0); 2341} 2342 2343/* 2344 * Dumps table in pre-8.1 legacy format. 2345 */ 2346int 2347ipfw_dump_table_legacy(struct ip_fw_chain *ch, struct tid_info *ti, 2348 ipfw_table *tbl) 2349{ 2350 struct table_config *tc; 2351 struct table_algo *ta; 2352 struct dump_args da; 2353 2354 tbl->cnt = 0; 2355 2356 if ((tc = find_table(CHAIN_TO_NI(ch), ti)) == NULL) 2357 return (0); /* XXX: We should return ESRCH */ 2358 2359 ta = tc->ta; 2360 2361 /* This dump format supports IPv4 only */ 2362 if (tc->no.subtype != IPFW_TABLE_ADDR) 2363 return (0); 2364 2365 memset(&da, 0, sizeof(da)); 2366 da.ch = ch; 2367 da.ti = KIDX_TO_TI(ch, tc->no.kidx); 2368 da.tc = tc; 2369 da.ent = &tbl->ent[0]; 2370 da.size = tbl->size; 2371 2372 tbl->cnt = 0; 2373 ta->foreach(tc->astate, da.ti, dump_table_entry, &da); 2374 tbl->cnt = da.cnt; 2375 2376 return (0); 2377} 2378 2379/* 2380 * Dumps table entry in eXtended format (v1)(current). 2381 */ 2382static int 2383dump_table_tentry(void *e, void *arg) 2384{ 2385 struct dump_args *da; 2386 struct table_config *tc; 2387 struct table_algo *ta; 2388 struct table_value *pval; 2389 ipfw_obj_tentry *tent; 2390 int error; 2391 2392 da = (struct dump_args *)arg; 2393 2394 tc = da->tc; 2395 ta = tc->ta; 2396 2397 tent = (ipfw_obj_tentry *)ipfw_get_sopt_space(da->sd, sizeof(*tent)); 2398 /* Out of memory, returning */ 2399 if (tent == NULL) { 2400 da->error = ENOMEM; 2401 return (1); 2402 } 2403 tent->head.length = sizeof(ipfw_obj_tentry); 2404 tent->idx = da->uidx; 2405 2406 error = ta->dump_tentry(tc->astate, da->ti, e, tent); 2407 if (error != 0) 2408 return (error); 2409 2410 pval = get_table_value(da->ch, da->tc, tent->v.kidx); 2411 ipfw_export_table_value_v1(pval, &tent->v.value); 2412 2413 return (0); 2414} 2415 2416/* 2417 * Dumps table entry in eXtended format (v0). 2418 */ 2419static int 2420dump_table_xentry(void *e, void *arg) 2421{ 2422 struct dump_args *da; 2423 struct table_config *tc; 2424 struct table_algo *ta; 2425 ipfw_table_xentry *xent; 2426 ipfw_obj_tentry *tent; 2427 struct table_value *pval; 2428 int error; 2429 2430 da = (struct dump_args *)arg; 2431 2432 tc = da->tc; 2433 ta = tc->ta; 2434 2435 xent = (ipfw_table_xentry *)ipfw_get_sopt_space(da->sd, sizeof(*xent)); 2436 /* Out of memory, returning */ 2437 if (xent == NULL) 2438 return (1); 2439 xent->len = sizeof(ipfw_table_xentry); 2440 xent->tbl = da->uidx; 2441 2442 memset(&da->tent, 0, sizeof(da->tent)); 2443 tent = &da->tent; 2444 error = ta->dump_tentry(tc->astate, da->ti, e, tent); 2445 if (error != 0) 2446 return (error); 2447 2448 /* Convert current format to previous one */ 2449 xent->masklen = tent->masklen; 2450 pval = get_table_value(da->ch, da->tc, da->tent.v.kidx); 2451 xent->value = ipfw_export_table_value_legacy(pval); 2452 /* Apply some hacks */ 2453 if (tc->no.subtype == IPFW_TABLE_ADDR && tent->subtype == AF_INET) { 2454 xent->k.addr6.s6_addr32[3] = tent->k.addr.s_addr; 2455 xent->flags = IPFW_TCF_INET; 2456 } else 2457 memcpy(&xent->k, &tent->k, sizeof(xent->k)); 2458 2459 return (0); 2460} 2461 2462/* 2463 * Helper function to export table algo data 2464 * to tentry format before calling user function. 2465 * 2466 * Returns 0 on success. 2467 */ 2468static int 2469prepare_table_tentry(void *e, void *arg) 2470{ 2471 struct dump_args *da; 2472 struct table_config *tc; 2473 struct table_algo *ta; 2474 int error; 2475 2476 da = (struct dump_args *)arg; 2477 2478 tc = da->tc; 2479 ta = tc->ta; 2480 2481 error = ta->dump_tentry(tc->astate, da->ti, e, &da->tent); 2482 if (error != 0) 2483 return (error); 2484 2485 da->f(&da->tent, da->farg); 2486 2487 return (0); 2488} 2489 2490/* 2491 * Allow external consumers to read table entries in standard format. 2492 */ 2493int 2494ipfw_foreach_table_tentry(struct ip_fw_chain *ch, uint16_t kidx, 2495 ta_foreach_f *f, void *arg) 2496{ 2497 struct namedobj_instance *ni; 2498 struct table_config *tc; 2499 struct table_algo *ta; 2500 struct dump_args da; 2501 2502 ni = CHAIN_TO_NI(ch); 2503 2504 tc = (struct table_config *)ipfw_objhash_lookup_kidx(ni, kidx); 2505 if (tc == NULL) 2506 return (ESRCH); 2507 2508 ta = tc->ta; 2509 2510 memset(&da, 0, sizeof(da)); 2511 da.ch = ch; 2512 da.ti = KIDX_TO_TI(ch, tc->no.kidx); 2513 da.tc = tc; 2514 da.f = f; 2515 da.farg = arg; 2516 2517 ta->foreach(tc->astate, da.ti, prepare_table_tentry, &da); 2518 2519 return (0); 2520} 2521 2522/* 2523 * Table algorithms 2524 */ 2525 2526/* 2527 * Finds algorithm by index, table type or supplied name. 2528 * 2529 * Returns pointer to algo or NULL. 2530 */ 2531static struct table_algo * 2532find_table_algo(struct tables_config *tcfg, struct tid_info *ti, char *name) 2533{ 2534 int i, l; 2535 struct table_algo *ta; 2536 2537 if (ti->type > IPFW_TABLE_MAXTYPE) 2538 return (NULL); 2539 2540 /* Search by index */ 2541 if (ti->atype != 0) { 2542 if (ti->atype > tcfg->algo_count) 2543 return (NULL); 2544 return (tcfg->algo[ti->atype]); 2545 } 2546 2547 if (name == NULL) { 2548 /* Return default algorithm for given type if set */ 2549 return (tcfg->def_algo[ti->type]); 2550 } 2551 2552 /* Search by name */ 2553 /* TODO: better search */ 2554 for (i = 1; i <= tcfg->algo_count; i++) { 2555 ta = tcfg->algo[i]; 2556 2557 /* 2558 * One can supply additional algorithm 2559 * parameters so we compare only the first word 2560 * of supplied name: 2561 * 'addr:chash hsize=32' 2562 * '^^^^^^^^^' 2563 * 2564 */ 2565 l = strlen(ta->name); 2566 if (strncmp(name, ta->name, l) != 0) 2567 continue; 2568 if (name[l] != '\0' && name[l] != ' ') 2569 continue; 2570 /* Check if we're requesting proper table type */ 2571 if (ti->type != 0 && ti->type != ta->type) 2572 return (NULL); 2573 return (ta); 2574 } 2575 2576 return (NULL); 2577} 2578 2579/* 2580 * Register new table algo @ta. 2581 * Stores algo id inside @idx. 2582 * 2583 * Returns 0 on success. 2584 */ 2585int 2586ipfw_add_table_algo(struct ip_fw_chain *ch, struct table_algo *ta, size_t size, 2587 int *idx) 2588{ 2589 struct tables_config *tcfg; 2590 struct table_algo *ta_new; 2591 size_t sz; 2592 2593 if (size > sizeof(struct table_algo)) 2594 return (EINVAL); 2595 2596 /* Check for the required on-stack size for add/del */ 2597 sz = roundup2(ta->ta_buf_size, sizeof(void *)); 2598 if (sz > TA_BUF_SZ) 2599 return (EINVAL); 2600 2601 KASSERT(ta->type <= IPFW_TABLE_MAXTYPE,("Increase IPFW_TABLE_MAXTYPE")); 2602 2603 /* Copy algorithm data to stable storage. */ 2604 ta_new = malloc(sizeof(struct table_algo), M_IPFW, M_WAITOK | M_ZERO); 2605 memcpy(ta_new, ta, size); 2606 2607 tcfg = CHAIN_TO_TCFG(ch); 2608 2609 KASSERT(tcfg->algo_count < 255, ("Increase algo array size")); 2610 2611 tcfg->algo[++tcfg->algo_count] = ta_new; 2612 ta_new->idx = tcfg->algo_count; 2613 2614 /* Set algorithm as default one for given type */ 2615 if ((ta_new->flags & TA_FLAG_DEFAULT) != 0 && 2616 tcfg->def_algo[ta_new->type] == NULL) 2617 tcfg->def_algo[ta_new->type] = ta_new; 2618 2619 *idx = ta_new->idx; 2620 2621 return (0); 2622} 2623 2624/* 2625 * Unregisters table algo using @idx as id. 2626 * XXX: It is NOT safe to call this function in any place 2627 * other than ipfw instance destroy handler. 2628 */ 2629void 2630ipfw_del_table_algo(struct ip_fw_chain *ch, int idx) 2631{ 2632 struct tables_config *tcfg; 2633 struct table_algo *ta; 2634 2635 tcfg = CHAIN_TO_TCFG(ch); 2636 2637 KASSERT(idx <= tcfg->algo_count, ("algo idx %d out of range 1..%d", 2638 idx, tcfg->algo_count)); 2639 2640 ta = tcfg->algo[idx]; 2641 KASSERT(ta != NULL, ("algo idx %d is NULL", idx)); 2642 2643 if (tcfg->def_algo[ta->type] == ta) 2644 tcfg->def_algo[ta->type] = NULL; 2645 2646 free(ta, M_IPFW); 2647} 2648 2649/* 2650 * Lists all table algorithms currently available. 2651 * Data layout (v0)(current): 2652 * Request: [ ipfw_obj_lheader ], size = ipfw_obj_lheader.size 2653 * Reply: [ ipfw_obj_lheader ipfw_ta_info x N ] 2654 * 2655 * Returns 0 on success 2656 */ 2657static int 2658list_table_algo(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 2659 struct sockopt_data *sd) 2660{ 2661 struct _ipfw_obj_lheader *olh; 2662 struct tables_config *tcfg; 2663 ipfw_ta_info *i; 2664 struct table_algo *ta; 2665 uint32_t count, n, size; 2666 2667 olh = (struct _ipfw_obj_lheader *)ipfw_get_sopt_header(sd,sizeof(*olh)); 2668 if (olh == NULL) 2669 return (EINVAL); 2670 if (sd->valsize < olh->size) 2671 return (EINVAL); 2672 2673 IPFW_UH_RLOCK(ch); 2674 tcfg = CHAIN_TO_TCFG(ch); 2675 count = tcfg->algo_count; 2676 size = count * sizeof(ipfw_ta_info) + sizeof(ipfw_obj_lheader); 2677 2678 /* Fill in header regadless of buffer size */ 2679 olh->count = count; 2680 olh->objsize = sizeof(ipfw_ta_info); 2681 2682 if (size > olh->size) { 2683 olh->size = size; 2684 IPFW_UH_RUNLOCK(ch); 2685 return (ENOMEM); 2686 } 2687 olh->size = size; 2688 2689 for (n = 1; n <= count; n++) { 2690 i = (ipfw_ta_info *)ipfw_get_sopt_space(sd, sizeof(*i)); 2691 KASSERT(i != NULL, ("previously checked buffer is not enough")); 2692 ta = tcfg->algo[n]; 2693 strlcpy(i->algoname, ta->name, sizeof(i->algoname)); 2694 i->type = ta->type; 2695 i->refcnt = ta->refcnt; 2696 } 2697 2698 IPFW_UH_RUNLOCK(ch); 2699 2700 return (0); 2701} 2702 2703static int 2704classify_srcdst(ipfw_insn *cmd, uint16_t *puidx, uint8_t *ptype) 2705{ 2706 /* Basic IPv4/IPv6 or u32 lookups */ 2707 *puidx = cmd->arg1; 2708 /* Assume ADDR by default */ 2709 *ptype = IPFW_TABLE_ADDR; 2710 int v; 2711 2712 if (F_LEN(cmd) > F_INSN_SIZE(ipfw_insn_u32)) { 2713 /* 2714 * generic lookup. The key must be 2715 * in 32bit big-endian format. 2716 */ 2717 v = ((ipfw_insn_u32 *)cmd)->d[1]; 2718 switch (v) { 2719 case 0: 2720 case 1: 2721 /* IPv4 src/dst */ 2722 break; 2723 case 2: 2724 case 3: 2725 /* src/dst port */ 2726 *ptype = IPFW_TABLE_NUMBER; 2727 break; 2728 case 4: 2729 /* uid/gid */ 2730 *ptype = IPFW_TABLE_NUMBER; 2731 break; 2732 case 5: 2733 /* jid */ 2734 *ptype = IPFW_TABLE_NUMBER; 2735 break; 2736 case 6: 2737 /* dscp */ 2738 *ptype = IPFW_TABLE_NUMBER; 2739 break; 2740 } 2741 } 2742 2743 return (0); 2744} 2745 2746static int 2747classify_via(ipfw_insn *cmd, uint16_t *puidx, uint8_t *ptype) 2748{ 2749 ipfw_insn_if *cmdif; 2750 2751 /* Interface table, possibly */ 2752 cmdif = (ipfw_insn_if *)cmd; 2753 if (cmdif->name[0] != '\1') 2754 return (1); 2755 2756 *ptype = IPFW_TABLE_INTERFACE; 2757 *puidx = cmdif->p.kidx; 2758 2759 return (0); 2760} 2761 2762static int 2763classify_flow(ipfw_insn *cmd, uint16_t *puidx, uint8_t *ptype) 2764{ 2765 2766 *puidx = cmd->arg1; 2767 *ptype = IPFW_TABLE_FLOW; 2768 2769 return (0); 2770} 2771 2772static void 2773update_arg1(ipfw_insn *cmd, uint16_t idx) 2774{ 2775 2776 cmd->arg1 = idx; 2777} 2778 2779static void 2780update_via(ipfw_insn *cmd, uint16_t idx) 2781{ 2782 ipfw_insn_if *cmdif; 2783 2784 cmdif = (ipfw_insn_if *)cmd; 2785 cmdif->p.kidx = idx; 2786} 2787 2788static int 2789table_findbyname(struct ip_fw_chain *ch, struct tid_info *ti, 2790 struct named_object **pno) 2791{ 2792 struct table_config *tc; 2793 int error; 2794 2795 IPFW_UH_WLOCK_ASSERT(ch); 2796 2797 error = find_table_err(CHAIN_TO_NI(ch), ti, &tc); 2798 if (error != 0) 2799 return (error); 2800 2801 *pno = &tc->no; 2802 return (0); 2803} 2804 2805/* XXX: sets-sets! */ 2806static struct named_object * 2807table_findbykidx(struct ip_fw_chain *ch, uint16_t idx) 2808{ 2809 struct namedobj_instance *ni; 2810 struct table_config *tc; 2811 2812 IPFW_UH_WLOCK_ASSERT(ch); 2813 ni = CHAIN_TO_NI(ch); 2814 tc = (struct table_config *)ipfw_objhash_lookup_kidx(ni, idx); 2815 KASSERT(tc != NULL, ("Table with index %d not found", idx)); 2816 2817 return (&tc->no); 2818} 2819 2820static int 2821table_manage_sets(struct ip_fw_chain *ch, uint16_t set, uint8_t new_set, 2822 enum ipfw_sets_cmd cmd) 2823{ 2824 2825 switch (cmd) { 2826 case SWAP_ALL: 2827 case TEST_ALL: 2828 /* 2829 * Return success for TEST_ALL, since nothing prevents 2830 * move rules from one set to another. All tables are 2831 * accessible from all sets when per-set tables sysctl 2832 * is disabled. 2833 */ 2834 case MOVE_ALL: 2835 case TEST_ONE: 2836 case MOVE_ONE: 2837 /* 2838 * NOTE: we need to use ipfw_objhash_del/ipfw_objhash_add 2839 * if set number will be used in hash function. Currently 2840 * we can just use generic handler that replaces set value. 2841 */ 2842 if (V_fw_tables_sets == 0) 2843 return (0); 2844 break; 2845 case COUNT_ONE: 2846 /* 2847 * Return EOPNOTSUPP for COUNT_ONE when per-set sysctl is 2848 * disabled. This allow skip table's opcodes from additional 2849 * checks when specific rules moved to another set. 2850 */ 2851 if (V_fw_tables_sets == 0) 2852 return (EOPNOTSUPP); 2853 } 2854 /* Use generic sets handler when per-set sysctl is enabled. */ 2855 return (ipfw_obj_manage_sets(CHAIN_TO_NI(ch), IPFW_TLV_TBL_NAME, 2856 set, new_set, cmd)); 2857} 2858 2859static struct opcode_obj_rewrite opcodes[] = { 2860 { 2861 .opcode = O_IP_SRC_LOOKUP, 2862 .etlv = IPFW_TLV_TBL_NAME, 2863 .classifier = classify_srcdst, 2864 .update = update_arg1, 2865 .find_byname = table_findbyname, 2866 .find_bykidx = table_findbykidx, 2867 .create_object = create_table_compat, 2868 .manage_sets = table_manage_sets, 2869 }, 2870 { 2871 .opcode = O_IP_DST_LOOKUP, 2872 .etlv = IPFW_TLV_TBL_NAME, 2873 .classifier = classify_srcdst, 2874 .update = update_arg1, 2875 .find_byname = table_findbyname, 2876 .find_bykidx = table_findbykidx, 2877 .create_object = create_table_compat, 2878 .manage_sets = table_manage_sets, 2879 }, 2880 { 2881 .opcode = O_IP_FLOW_LOOKUP, 2882 .etlv = IPFW_TLV_TBL_NAME, 2883 .classifier = classify_flow, 2884 .update = update_arg1, 2885 .find_byname = table_findbyname, 2886 .find_bykidx = table_findbykidx, 2887 .create_object = create_table_compat, 2888 .manage_sets = table_manage_sets, 2889 }, 2890 { 2891 .opcode = O_XMIT, 2892 .etlv = IPFW_TLV_TBL_NAME, 2893 .classifier = classify_via, 2894 .update = update_via, 2895 .find_byname = table_findbyname, 2896 .find_bykidx = table_findbykidx, 2897 .create_object = create_table_compat, 2898 .manage_sets = table_manage_sets, 2899 }, 2900 { 2901 .opcode = O_RECV, 2902 .etlv = IPFW_TLV_TBL_NAME, 2903 .classifier = classify_via, 2904 .update = update_via, 2905 .find_byname = table_findbyname, 2906 .find_bykidx = table_findbykidx, 2907 .create_object = create_table_compat, 2908 .manage_sets = table_manage_sets, 2909 }, 2910 { 2911 .opcode = O_VIA, 2912 .etlv = IPFW_TLV_TBL_NAME, 2913 .classifier = classify_via, 2914 .update = update_via, 2915 .find_byname = table_findbyname, 2916 .find_bykidx = table_findbykidx, 2917 .create_object = create_table_compat, 2918 .manage_sets = table_manage_sets, 2919 }, 2920}; 2921 2922static int 2923test_sets_cb(struct namedobj_instance *ni __unused, struct named_object *no, 2924 void *arg __unused) 2925{ 2926 2927 /* Check that there aren't any tables in not default set */ 2928 if (no->set != 0) 2929 return (EBUSY); 2930 return (0); 2931} 2932 2933/* 2934 * Switch between "set 0" and "rule's set" table binding, 2935 * Check all ruleset bindings and permits changing 2936 * IFF each binding has both rule AND table in default set (set 0). 2937 * 2938 * Returns 0 on success. 2939 */ 2940int 2941ipfw_switch_tables_namespace(struct ip_fw_chain *ch, unsigned int sets) 2942{ 2943 struct opcode_obj_rewrite *rw; 2944 struct namedobj_instance *ni; 2945 struct named_object *no; 2946 struct ip_fw *rule; 2947 ipfw_insn *cmd; 2948 int cmdlen, i, l; 2949 uint16_t kidx; 2950 uint8_t subtype; 2951 2952 IPFW_UH_WLOCK(ch); 2953 2954 if (V_fw_tables_sets == sets) { 2955 IPFW_UH_WUNLOCK(ch); 2956 return (0); 2957 } 2958 ni = CHAIN_TO_NI(ch); 2959 if (sets == 0) { 2960 /* 2961 * Prevent disabling sets support if we have some tables 2962 * in not default sets. 2963 */ 2964 if (ipfw_objhash_foreach_type(ni, test_sets_cb, 2965 NULL, IPFW_TLV_TBL_NAME) != 0) { 2966 IPFW_UH_WUNLOCK(ch); 2967 return (EBUSY); 2968 } 2969 } 2970 /* 2971 * Scan all rules and examine tables opcodes. 2972 */ 2973 for (i = 0; i < ch->n_rules; i++) { 2974 rule = ch->map[i]; 2975 2976 l = rule->cmd_len; 2977 cmd = rule->cmd; 2978 cmdlen = 0; 2979 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 2980 cmdlen = F_LEN(cmd); 2981 /* Check only tables opcodes */ 2982 for (kidx = 0, rw = opcodes; 2983 rw < opcodes + nitems(opcodes); rw++) { 2984 if (rw->opcode != cmd->opcode) 2985 continue; 2986 if (rw->classifier(cmd, &kidx, &subtype) == 0) 2987 break; 2988 } 2989 if (kidx == 0) 2990 continue; 2991 no = ipfw_objhash_lookup_kidx(ni, kidx); 2992 /* Check if both table object and rule has the set 0 */ 2993 if (no->set != 0 || rule->set != 0) { 2994 IPFW_UH_WUNLOCK(ch); 2995 return (EBUSY); 2996 } 2997 2998 } 2999 } 3000 V_fw_tables_sets = sets; 3001 IPFW_UH_WUNLOCK(ch); 3002 return (0); 3003} 3004 3005/* 3006 * Checks table name for validity. 3007 * Enforce basic length checks, the rest 3008 * should be done in userland. 3009 * 3010 * Returns 0 if name is considered valid. 3011 */ 3012static int 3013check_table_name(const char *name) 3014{ 3015 3016 /* 3017 * TODO: do some more complicated checks 3018 */ 3019 return (ipfw_check_object_name_generic(name)); 3020} 3021 3022/* 3023 * Finds table config based on either legacy index 3024 * or name in ntlv. 3025 * Note @ti structure contains unchecked data from userland. 3026 * 3027 * Returns 0 in success and fills in @tc with found config 3028 */ 3029static int 3030find_table_err(struct namedobj_instance *ni, struct tid_info *ti, 3031 struct table_config **tc) 3032{ 3033 char *name, bname[16]; 3034 struct named_object *no; 3035 ipfw_obj_ntlv *ntlv; 3036 uint32_t set; 3037 3038 if (ti->tlvs != NULL) { 3039 ntlv = ipfw_find_name_tlv_type(ti->tlvs, ti->tlen, ti->uidx, 3040 IPFW_TLV_TBL_NAME); 3041 if (ntlv == NULL) 3042 return (EINVAL); 3043 name = ntlv->name; 3044 3045 /* 3046 * Use set provided by @ti instead of @ntlv one. 3047 * This is needed due to different sets behavior 3048 * controlled by V_fw_tables_sets. 3049 */ 3050 set = (V_fw_tables_sets != 0) ? ti->set : 0; 3051 } else { 3052 snprintf(bname, sizeof(bname), "%d", ti->uidx); 3053 name = bname; 3054 set = 0; 3055 } 3056 3057 no = ipfw_objhash_lookup_name(ni, set, name); 3058 *tc = (struct table_config *)no; 3059 3060 return (0); 3061} 3062 3063/* 3064 * Finds table config based on either legacy index 3065 * or name in ntlv. 3066 * Note @ti structure contains unchecked data from userland. 3067 * 3068 * Returns pointer to table_config or NULL. 3069 */ 3070static struct table_config * 3071find_table(struct namedobj_instance *ni, struct tid_info *ti) 3072{ 3073 struct table_config *tc; 3074 3075 if (find_table_err(ni, ti, &tc) != 0) 3076 return (NULL); 3077 3078 return (tc); 3079} 3080 3081/* 3082 * Allocate new table config structure using 3083 * specified @algo and @aname. 3084 * 3085 * Returns pointer to config or NULL. 3086 */ 3087static struct table_config * 3088alloc_table_config(struct ip_fw_chain *ch, struct tid_info *ti, 3089 struct table_algo *ta, char *aname, uint8_t tflags) 3090{ 3091 char *name, bname[16]; 3092 struct table_config *tc; 3093 int error; 3094 ipfw_obj_ntlv *ntlv; 3095 uint32_t set; 3096 3097 if (ti->tlvs != NULL) { 3098 ntlv = ipfw_find_name_tlv_type(ti->tlvs, ti->tlen, ti->uidx, 3099 IPFW_TLV_TBL_NAME); 3100 if (ntlv == NULL) 3101 return (NULL); 3102 name = ntlv->name; 3103 set = ntlv->set; 3104 } else { 3105 /* Compat part: convert number to string representation */ 3106 snprintf(bname, sizeof(bname), "%d", ti->uidx); 3107 name = bname; 3108 set = 0; 3109 } 3110 3111 tc = malloc(sizeof(struct table_config), M_IPFW, M_WAITOK | M_ZERO); 3112 tc->no.name = tc->tablename; 3113 tc->no.subtype = ta->type; 3114 tc->no.set = set; 3115 tc->tflags = tflags; 3116 tc->ta = ta; 3117 strlcpy(tc->tablename, name, sizeof(tc->tablename)); 3118 /* Set "shared" value type by default */ 3119 tc->vshared = 1; 3120 3121 /* Preallocate data structures for new tables */ 3122 error = ta->init(ch, &tc->astate, &tc->ti_copy, aname, tflags); 3123 if (error != 0) { 3124 free(tc, M_IPFW); 3125 return (NULL); 3126 } 3127 3128 return (tc); 3129} 3130 3131/* 3132 * Destroys table state and config. 3133 */ 3134static void 3135free_table_config(struct namedobj_instance *ni, struct table_config *tc) 3136{ 3137 3138 KASSERT(tc->linked == 0, ("free() on linked config")); 3139 /* UH lock MUST NOT be held */ 3140 3141 /* 3142 * We're using ta without any locking/referencing. 3143 * TODO: fix this if we're going to use unloadable algos. 3144 */ 3145 tc->ta->destroy(tc->astate, &tc->ti_copy); 3146 free(tc, M_IPFW); 3147} 3148 3149/* 3150 * Links @tc to @chain table named instance. 3151 * Sets appropriate type/states in @chain table info. 3152 */ 3153static void 3154link_table(struct ip_fw_chain *ch, struct table_config *tc) 3155{ 3156 struct namedobj_instance *ni; 3157 struct table_info *ti; 3158 uint16_t kidx; 3159 3160 IPFW_UH_WLOCK_ASSERT(ch); 3161 IPFW_WLOCK_ASSERT(ch); 3162 3163 ni = CHAIN_TO_NI(ch); 3164 kidx = tc->no.kidx; 3165 3166 ipfw_objhash_add(ni, &tc->no); 3167 3168 ti = KIDX_TO_TI(ch, kidx); 3169 *ti = tc->ti_copy; 3170 3171 /* Notify algo on real @ti address */ 3172 if (tc->ta->change_ti != NULL) 3173 tc->ta->change_ti(tc->astate, ti); 3174 3175 tc->linked = 1; 3176 tc->ta->refcnt++; 3177} 3178 3179/* 3180 * Unlinks @tc from @chain table named instance. 3181 * Zeroes states in @chain and stores them in @tc. 3182 */ 3183static void 3184unlink_table(struct ip_fw_chain *ch, struct table_config *tc) 3185{ 3186 struct namedobj_instance *ni; 3187 struct table_info *ti; 3188 uint16_t kidx; 3189 3190 IPFW_UH_WLOCK_ASSERT(ch); 3191 IPFW_WLOCK_ASSERT(ch); 3192 3193 ni = CHAIN_TO_NI(ch); 3194 kidx = tc->no.kidx; 3195 3196 /* Clear state. @ti copy is already saved inside @tc */ 3197 ipfw_objhash_del(ni, &tc->no); 3198 ti = KIDX_TO_TI(ch, kidx); 3199 memset(ti, 0, sizeof(struct table_info)); 3200 tc->linked = 0; 3201 tc->ta->refcnt--; 3202 3203 /* Notify algo on real @ti address */ 3204 if (tc->ta->change_ti != NULL) 3205 tc->ta->change_ti(tc->astate, NULL); 3206} 3207 3208static struct ipfw_sopt_handler scodes[] = { 3209 { IP_FW_TABLE_XCREATE, 0, HDIR_SET, create_table }, 3210 { IP_FW_TABLE_XDESTROY, 0, HDIR_SET, flush_table_v0 }, 3211 { IP_FW_TABLE_XFLUSH, 0, HDIR_SET, flush_table_v0 }, 3212 { IP_FW_TABLE_XMODIFY, 0, HDIR_BOTH, modify_table }, 3213 { IP_FW_TABLE_XINFO, 0, HDIR_GET, describe_table }, 3214 { IP_FW_TABLES_XLIST, 0, HDIR_GET, list_tables }, 3215 { IP_FW_TABLE_XLIST, 0, HDIR_GET, dump_table_v0 }, 3216 { IP_FW_TABLE_XLIST, 1, HDIR_GET, dump_table_v1 }, 3217 { IP_FW_TABLE_XADD, 0, HDIR_BOTH, manage_table_ent_v0 }, 3218 { IP_FW_TABLE_XADD, 1, HDIR_BOTH, manage_table_ent_v1 }, 3219 { IP_FW_TABLE_XDEL, 0, HDIR_BOTH, manage_table_ent_v0 }, 3220 { IP_FW_TABLE_XDEL, 1, HDIR_BOTH, manage_table_ent_v1 }, 3221 { IP_FW_TABLE_XFIND, 0, HDIR_GET, find_table_entry }, 3222 { IP_FW_TABLE_XSWAP, 0, HDIR_SET, swap_table }, 3223 { IP_FW_TABLES_ALIST, 0, HDIR_GET, list_table_algo }, 3224 { IP_FW_TABLE_XGETSIZE, 0, HDIR_GET, get_table_size }, 3225}; 3226 3227static int 3228destroy_table_locked(struct namedobj_instance *ni, struct named_object *no, 3229 void *arg) 3230{ 3231 3232 unlink_table((struct ip_fw_chain *)arg, (struct table_config *)no); 3233 if (ipfw_objhash_free_idx(ni, no->kidx) != 0) 3234 printf("Error unlinking kidx %d from table %s\n", 3235 no->kidx, no->name); 3236 free_table_config(ni, (struct table_config *)no); 3237 return (0); 3238} 3239 3240/* 3241 * Shuts tables module down. 3242 */ 3243void 3244ipfw_destroy_tables(struct ip_fw_chain *ch, int last) 3245{ 3246 3247 IPFW_DEL_SOPT_HANDLER(last, scodes); 3248 IPFW_DEL_OBJ_REWRITER(last, opcodes); 3249 3250 /* Remove all tables from working set */ 3251 IPFW_UH_WLOCK(ch); 3252 IPFW_WLOCK(ch); 3253 ipfw_objhash_foreach(CHAIN_TO_NI(ch), destroy_table_locked, ch); 3254 IPFW_WUNLOCK(ch); 3255 IPFW_UH_WUNLOCK(ch); 3256 3257 /* Free pointers itself */ 3258 free(ch->tablestate, M_IPFW); 3259 3260 ipfw_table_value_destroy(ch, last); 3261 ipfw_table_algo_destroy(ch); 3262 3263 ipfw_objhash_destroy(CHAIN_TO_NI(ch)); 3264 free(CHAIN_TO_TCFG(ch), M_IPFW); 3265} 3266 3267/* 3268 * Starts tables module. 3269 */ 3270int 3271ipfw_init_tables(struct ip_fw_chain *ch, int first) 3272{ 3273 struct tables_config *tcfg; 3274 3275 /* Allocate pointers */ 3276 ch->tablestate = malloc(V_fw_tables_max * sizeof(struct table_info), 3277 M_IPFW, M_WAITOK | M_ZERO); 3278 3279 tcfg = malloc(sizeof(struct tables_config), M_IPFW, M_WAITOK | M_ZERO); 3280 tcfg->namehash = ipfw_objhash_create(V_fw_tables_max); 3281 ch->tblcfg = tcfg; 3282 3283 ipfw_table_value_init(ch, first); 3284 ipfw_table_algo_init(ch); 3285 3286 IPFW_ADD_OBJ_REWRITER(first, opcodes); 3287 IPFW_ADD_SOPT_HANDLER(first, scodes); 3288 return (0); 3289} 3290 3291 3292 3293