ip_fw_table.c revision 299136
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 299136 2016-05-05 20:15:46Z 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 * Switch between "set 0" and "rule's set" table binding, 1606 * Check all ruleset bindings and permits changing 1607 * IFF each binding has both rule AND table in default set (set 0). 1608 * 1609 * Returns 0 on success. 1610 */ 1611int 1612ipfw_switch_tables_namespace(struct ip_fw_chain *ch, unsigned int sets) 1613{ 1614 struct namedobj_instance *ni; 1615 struct named_object *no; 1616 struct ip_fw *rule; 1617 ipfw_insn *cmd; 1618 int cmdlen, i, l; 1619 uint16_t kidx; 1620 1621 IPFW_UH_WLOCK(ch); 1622 1623 if (V_fw_tables_sets == sets) { 1624 IPFW_UH_WUNLOCK(ch); 1625 return (0); 1626 } 1627 1628 ni = CHAIN_TO_NI(ch); 1629 1630 /* 1631 * Scan all rules and examine tables opcodes. 1632 */ 1633 for (i = 0; i < ch->n_rules; i++) { 1634 rule = ch->map[i]; 1635 1636 l = rule->cmd_len; 1637 cmd = rule->cmd; 1638 cmdlen = 0; 1639 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 1640 cmdlen = F_LEN(cmd); 1641 1642 if (classify_opcode_kidx(cmd, &kidx) != 0) 1643 continue; 1644 1645 no = ipfw_objhash_lookup_kidx(ni, kidx); 1646 1647 /* Check if both table object and rule has the set 0 */ 1648 if (no->set != 0 || rule->set != 0) { 1649 IPFW_UH_WUNLOCK(ch); 1650 return (EBUSY); 1651 } 1652 1653 } 1654 } 1655 V_fw_tables_sets = sets; 1656 1657 IPFW_UH_WUNLOCK(ch); 1658 1659 return (0); 1660} 1661 1662/* 1663 * Lookup an IP @addr in table @tbl. 1664 * Stores found value in @val. 1665 * 1666 * Returns 1 if @addr was found. 1667 */ 1668int 1669ipfw_lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1670 uint32_t *val) 1671{ 1672 struct table_info *ti; 1673 1674 ti = KIDX_TO_TI(ch, tbl); 1675 1676 return (ti->lookup(ti, &addr, sizeof(in_addr_t), val)); 1677} 1678 1679/* 1680 * Lookup an arbtrary key @paddr of legth @plen in table @tbl. 1681 * Stores found value in @val. 1682 * 1683 * Returns 1 if key was found. 1684 */ 1685int 1686ipfw_lookup_table_extended(struct ip_fw_chain *ch, uint16_t tbl, uint16_t plen, 1687 void *paddr, uint32_t *val) 1688{ 1689 struct table_info *ti; 1690 1691 ti = KIDX_TO_TI(ch, tbl); 1692 1693 return (ti->lookup(ti, paddr, plen, val)); 1694} 1695 1696/* 1697 * Info/List/dump support for tables. 1698 * 1699 */ 1700 1701/* 1702 * High-level 'get' cmds sysctl handlers 1703 */ 1704 1705/* 1706 * Lists all tables currently available in kernel. 1707 * Data layout (v0)(current): 1708 * Request: [ ipfw_obj_lheader ], size = ipfw_obj_lheader.size 1709 * Reply: [ ipfw_obj_lheader ipfw_xtable_info x N ] 1710 * 1711 * Returns 0 on success 1712 */ 1713static int 1714list_tables(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 1715 struct sockopt_data *sd) 1716{ 1717 struct _ipfw_obj_lheader *olh; 1718 int error; 1719 1720 olh = (struct _ipfw_obj_lheader *)ipfw_get_sopt_header(sd,sizeof(*olh)); 1721 if (olh == NULL) 1722 return (EINVAL); 1723 if (sd->valsize < olh->size) 1724 return (EINVAL); 1725 1726 IPFW_UH_RLOCK(ch); 1727 error = export_tables(ch, olh, sd); 1728 IPFW_UH_RUNLOCK(ch); 1729 1730 return (error); 1731} 1732 1733/* 1734 * Store table info to buffer provided by @sd. 1735 * Data layout (v0)(current): 1736 * Request: [ ipfw_obj_header ipfw_xtable_info(empty)] 1737 * Reply: [ ipfw_obj_header ipfw_xtable_info ] 1738 * 1739 * Returns 0 on success. 1740 */ 1741static int 1742describe_table(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 1743 struct sockopt_data *sd) 1744{ 1745 struct _ipfw_obj_header *oh; 1746 struct table_config *tc; 1747 struct tid_info ti; 1748 size_t sz; 1749 1750 sz = sizeof(*oh) + sizeof(ipfw_xtable_info); 1751 oh = (struct _ipfw_obj_header *)ipfw_get_sopt_header(sd, sz); 1752 if (oh == NULL) 1753 return (EINVAL); 1754 1755 objheader_to_ti(oh, &ti); 1756 1757 IPFW_UH_RLOCK(ch); 1758 if ((tc = find_table(CHAIN_TO_NI(ch), &ti)) == NULL) { 1759 IPFW_UH_RUNLOCK(ch); 1760 return (ESRCH); 1761 } 1762 1763 export_table_info(ch, tc, (ipfw_xtable_info *)(oh + 1)); 1764 IPFW_UH_RUNLOCK(ch); 1765 1766 return (0); 1767} 1768 1769/* 1770 * Modifies existing table. 1771 * Data layout (v0)(current): 1772 * Request: [ ipfw_obj_header ipfw_xtable_info ] 1773 * 1774 * Returns 0 on success 1775 */ 1776static int 1777modify_table(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 1778 struct sockopt_data *sd) 1779{ 1780 struct _ipfw_obj_header *oh; 1781 ipfw_xtable_info *i; 1782 char *tname; 1783 struct tid_info ti; 1784 struct namedobj_instance *ni; 1785 struct table_config *tc; 1786 1787 if (sd->valsize != sizeof(*oh) + sizeof(ipfw_xtable_info)) 1788 return (EINVAL); 1789 1790 oh = (struct _ipfw_obj_header *)sd->kbuf; 1791 i = (ipfw_xtable_info *)(oh + 1); 1792 1793 /* 1794 * Verify user-supplied strings. 1795 * Check for null-terminated/zero-length strings/ 1796 */ 1797 tname = oh->ntlv.name; 1798 if (check_table_name(tname) != 0) 1799 return (EINVAL); 1800 1801 objheader_to_ti(oh, &ti); 1802 ti.type = i->type; 1803 1804 IPFW_UH_WLOCK(ch); 1805 ni = CHAIN_TO_NI(ch); 1806 if ((tc = find_table(ni, &ti)) == NULL) { 1807 IPFW_UH_WUNLOCK(ch); 1808 return (ESRCH); 1809 } 1810 1811 /* Do not support any modifications for readonly tables */ 1812 if ((tc->ta->flags & TA_FLAG_READONLY) != 0) { 1813 IPFW_UH_WUNLOCK(ch); 1814 return (EACCES); 1815 } 1816 1817 if ((i->mflags & IPFW_TMFLAGS_LIMIT) != 0) 1818 tc->limit = i->limit; 1819 if ((i->mflags & IPFW_TMFLAGS_LOCK) != 0) 1820 tc->locked = ((i->flags & IPFW_TGFLAGS_LOCKED) != 0); 1821 IPFW_UH_WUNLOCK(ch); 1822 1823 return (0); 1824} 1825 1826/* 1827 * Creates new table. 1828 * Data layout (v0)(current): 1829 * Request: [ ipfw_obj_header ipfw_xtable_info ] 1830 * 1831 * Returns 0 on success 1832 */ 1833static int 1834create_table(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 1835 struct sockopt_data *sd) 1836{ 1837 struct _ipfw_obj_header *oh; 1838 ipfw_xtable_info *i; 1839 char *tname, *aname; 1840 struct tid_info ti; 1841 struct namedobj_instance *ni; 1842 1843 if (sd->valsize != sizeof(*oh) + sizeof(ipfw_xtable_info)) 1844 return (EINVAL); 1845 1846 oh = (struct _ipfw_obj_header *)sd->kbuf; 1847 i = (ipfw_xtable_info *)(oh + 1); 1848 1849 /* 1850 * Verify user-supplied strings. 1851 * Check for null-terminated/zero-length strings/ 1852 */ 1853 tname = oh->ntlv.name; 1854 aname = i->algoname; 1855 if (check_table_name(tname) != 0 || 1856 strnlen(aname, sizeof(i->algoname)) == sizeof(i->algoname)) 1857 return (EINVAL); 1858 1859 if (aname[0] == '\0') { 1860 /* Use default algorithm */ 1861 aname = NULL; 1862 } 1863 1864 objheader_to_ti(oh, &ti); 1865 ti.type = i->type; 1866 1867 ni = CHAIN_TO_NI(ch); 1868 1869 IPFW_UH_RLOCK(ch); 1870 if (find_table(ni, &ti) != NULL) { 1871 IPFW_UH_RUNLOCK(ch); 1872 return (EEXIST); 1873 } 1874 IPFW_UH_RUNLOCK(ch); 1875 1876 return (create_table_internal(ch, &ti, aname, i, NULL, 0)); 1877} 1878 1879/* 1880 * Creates new table based on @ti and @aname. 1881 * 1882 * Assume @aname to be checked and valid. 1883 * Stores allocated table kidx inside @pkidx (if non-NULL). 1884 * Reference created table if @compat is non-zero. 1885 * 1886 * Returns 0 on success. 1887 */ 1888static int 1889create_table_internal(struct ip_fw_chain *ch, struct tid_info *ti, 1890 char *aname, ipfw_xtable_info *i, uint16_t *pkidx, int compat) 1891{ 1892 struct namedobj_instance *ni; 1893 struct table_config *tc, *tc_new, *tmp; 1894 struct table_algo *ta; 1895 uint16_t kidx; 1896 1897 ni = CHAIN_TO_NI(ch); 1898 1899 ta = find_table_algo(CHAIN_TO_TCFG(ch), ti, aname); 1900 if (ta == NULL) 1901 return (ENOTSUP); 1902 1903 tc = alloc_table_config(ch, ti, ta, aname, i->tflags); 1904 if (tc == NULL) 1905 return (ENOMEM); 1906 1907 tc->vmask = i->vmask; 1908 tc->limit = i->limit; 1909 if (ta->flags & TA_FLAG_READONLY) 1910 tc->locked = 1; 1911 else 1912 tc->locked = (i->flags & IPFW_TGFLAGS_LOCKED) != 0; 1913 1914 IPFW_UH_WLOCK(ch); 1915 1916 /* Check if table has been already created */ 1917 tc_new = find_table(ni, ti); 1918 if (tc_new != NULL) { 1919 1920 /* 1921 * Compat: do not fail if we're 1922 * requesting to create existing table 1923 * which has the same type 1924 */ 1925 if (compat == 0 || tc_new->no.subtype != tc->no.subtype) { 1926 IPFW_UH_WUNLOCK(ch); 1927 free_table_config(ni, tc); 1928 return (EEXIST); 1929 } 1930 1931 /* Exchange tc and tc_new for proper refcounting & freeing */ 1932 tmp = tc; 1933 tc = tc_new; 1934 tc_new = tmp; 1935 } else { 1936 /* New table */ 1937 if (ipfw_objhash_alloc_idx(ni, &kidx) != 0) { 1938 IPFW_UH_WUNLOCK(ch); 1939 printf("Unable to allocate table index." 1940 " Consider increasing net.inet.ip.fw.tables_max"); 1941 free_table_config(ni, tc); 1942 return (EBUSY); 1943 } 1944 tc->no.kidx = kidx; 1945 tc->no.etlv = IPFW_TLV_TBL_NAME; 1946 1947 IPFW_WLOCK(ch); 1948 link_table(ch, tc); 1949 IPFW_WUNLOCK(ch); 1950 } 1951 1952 if (compat != 0) 1953 tc->no.refcnt++; 1954 if (pkidx != NULL) 1955 *pkidx = tc->no.kidx; 1956 1957 IPFW_UH_WUNLOCK(ch); 1958 1959 if (tc_new != NULL) 1960 free_table_config(ni, tc_new); 1961 1962 return (0); 1963} 1964 1965static void 1966ntlv_to_ti(ipfw_obj_ntlv *ntlv, struct tid_info *ti) 1967{ 1968 1969 memset(ti, 0, sizeof(struct tid_info)); 1970 ti->set = ntlv->set; 1971 ti->uidx = ntlv->idx; 1972 ti->tlvs = ntlv; 1973 ti->tlen = ntlv->head.length; 1974} 1975 1976static void 1977objheader_to_ti(struct _ipfw_obj_header *oh, struct tid_info *ti) 1978{ 1979 1980 ntlv_to_ti(&oh->ntlv, ti); 1981} 1982 1983struct namedobj_instance * 1984ipfw_get_table_objhash(struct ip_fw_chain *ch) 1985{ 1986 1987 return (CHAIN_TO_NI(ch)); 1988} 1989 1990/* 1991 * Exports basic table info as name TLV. 1992 * Used inside dump_static_rules() to provide info 1993 * about all tables referenced by current ruleset. 1994 * 1995 * Returns 0 on success. 1996 */ 1997int 1998ipfw_export_table_ntlv(struct ip_fw_chain *ch, uint16_t kidx, 1999 struct sockopt_data *sd) 2000{ 2001 struct namedobj_instance *ni; 2002 struct named_object *no; 2003 ipfw_obj_ntlv *ntlv; 2004 2005 ni = CHAIN_TO_NI(ch); 2006 2007 no = ipfw_objhash_lookup_kidx(ni, kidx); 2008 KASSERT(no != NULL, ("invalid table kidx passed")); 2009 2010 ntlv = (ipfw_obj_ntlv *)ipfw_get_sopt_space(sd, sizeof(*ntlv)); 2011 if (ntlv == NULL) 2012 return (ENOMEM); 2013 2014 ntlv->head.type = IPFW_TLV_TBL_NAME; 2015 ntlv->head.length = sizeof(*ntlv); 2016 ntlv->idx = no->kidx; 2017 strlcpy(ntlv->name, no->name, sizeof(ntlv->name)); 2018 2019 return (0); 2020} 2021 2022struct dump_args { 2023 struct ip_fw_chain *ch; 2024 struct table_info *ti; 2025 struct table_config *tc; 2026 struct sockopt_data *sd; 2027 uint32_t cnt; 2028 uint16_t uidx; 2029 int error; 2030 uint32_t size; 2031 ipfw_table_entry *ent; 2032 ta_foreach_f *f; 2033 void *farg; 2034 ipfw_obj_tentry tent; 2035}; 2036 2037static int 2038count_ext_entries(void *e, void *arg) 2039{ 2040 struct dump_args *da; 2041 2042 da = (struct dump_args *)arg; 2043 da->cnt++; 2044 2045 return (0); 2046} 2047 2048/* 2049 * Gets number of items from table either using 2050 * internal counter or calling algo callback for 2051 * externally-managed tables. 2052 * 2053 * Returns number of records. 2054 */ 2055static uint32_t 2056table_get_count(struct ip_fw_chain *ch, struct table_config *tc) 2057{ 2058 struct table_info *ti; 2059 struct table_algo *ta; 2060 struct dump_args da; 2061 2062 ti = KIDX_TO_TI(ch, tc->no.kidx); 2063 ta = tc->ta; 2064 2065 /* Use internal counter for self-managed tables */ 2066 if ((ta->flags & TA_FLAG_READONLY) == 0) 2067 return (tc->count); 2068 2069 /* Use callback to quickly get number of items */ 2070 if ((ta->flags & TA_FLAG_EXTCOUNTER) != 0) 2071 return (ta->get_count(tc->astate, ti)); 2072 2073 /* Count number of iterms ourselves */ 2074 memset(&da, 0, sizeof(da)); 2075 ta->foreach(tc->astate, ti, count_ext_entries, &da); 2076 2077 return (da.cnt); 2078} 2079 2080/* 2081 * Exports table @tc info into standard ipfw_xtable_info format. 2082 */ 2083static void 2084export_table_info(struct ip_fw_chain *ch, struct table_config *tc, 2085 ipfw_xtable_info *i) 2086{ 2087 struct table_info *ti; 2088 struct table_algo *ta; 2089 2090 i->type = tc->no.subtype; 2091 i->tflags = tc->tflags; 2092 i->vmask = tc->vmask; 2093 i->set = tc->no.set; 2094 i->kidx = tc->no.kidx; 2095 i->refcnt = tc->no.refcnt; 2096 i->count = table_get_count(ch, tc); 2097 i->limit = tc->limit; 2098 i->flags |= (tc->locked != 0) ? IPFW_TGFLAGS_LOCKED : 0; 2099 i->size = i->count * sizeof(ipfw_obj_tentry); 2100 i->size += sizeof(ipfw_obj_header) + sizeof(ipfw_xtable_info); 2101 strlcpy(i->tablename, tc->tablename, sizeof(i->tablename)); 2102 ti = KIDX_TO_TI(ch, tc->no.kidx); 2103 ta = tc->ta; 2104 if (ta->print_config != NULL) { 2105 /* Use algo function to print table config to string */ 2106 ta->print_config(tc->astate, ti, i->algoname, 2107 sizeof(i->algoname)); 2108 } else 2109 strlcpy(i->algoname, ta->name, sizeof(i->algoname)); 2110 /* Dump algo-specific data, if possible */ 2111 if (ta->dump_tinfo != NULL) { 2112 ta->dump_tinfo(tc->astate, ti, &i->ta_info); 2113 i->ta_info.flags |= IPFW_TATFLAGS_DATA; 2114 } 2115} 2116 2117struct dump_table_args { 2118 struct ip_fw_chain *ch; 2119 struct sockopt_data *sd; 2120}; 2121 2122static void 2123export_table_internal(struct namedobj_instance *ni, struct named_object *no, 2124 void *arg) 2125{ 2126 ipfw_xtable_info *i; 2127 struct dump_table_args *dta; 2128 2129 dta = (struct dump_table_args *)arg; 2130 2131 i = (ipfw_xtable_info *)ipfw_get_sopt_space(dta->sd, sizeof(*i)); 2132 KASSERT(i != NULL, ("previously checked buffer is not enough")); 2133 2134 export_table_info(dta->ch, (struct table_config *)no, i); 2135} 2136 2137/* 2138 * Export all tables as ipfw_xtable_info structures to 2139 * storage provided by @sd. 2140 * 2141 * If supplied buffer is too small, fills in required size 2142 * and returns ENOMEM. 2143 * Returns 0 on success. 2144 */ 2145static int 2146export_tables(struct ip_fw_chain *ch, ipfw_obj_lheader *olh, 2147 struct sockopt_data *sd) 2148{ 2149 uint32_t size; 2150 uint32_t count; 2151 struct dump_table_args dta; 2152 2153 count = ipfw_objhash_count(CHAIN_TO_NI(ch)); 2154 size = count * sizeof(ipfw_xtable_info) + sizeof(ipfw_obj_lheader); 2155 2156 /* Fill in header regadless of buffer size */ 2157 olh->count = count; 2158 olh->objsize = sizeof(ipfw_xtable_info); 2159 2160 if (size > olh->size) { 2161 olh->size = size; 2162 return (ENOMEM); 2163 } 2164 2165 olh->size = size; 2166 2167 dta.ch = ch; 2168 dta.sd = sd; 2169 2170 ipfw_objhash_foreach(CHAIN_TO_NI(ch), export_table_internal, &dta); 2171 2172 return (0); 2173} 2174 2175/* 2176 * Dumps all table data 2177 * Data layout (v1)(current): 2178 * Request: [ ipfw_obj_header ], size = ipfw_xtable_info.size 2179 * Reply: [ ipfw_obj_header ipfw_xtable_info ipfw_obj_tentry x N ] 2180 * 2181 * Returns 0 on success 2182 */ 2183static int 2184dump_table_v1(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 2185 struct sockopt_data *sd) 2186{ 2187 struct _ipfw_obj_header *oh; 2188 ipfw_xtable_info *i; 2189 struct tid_info ti; 2190 struct table_config *tc; 2191 struct table_algo *ta; 2192 struct dump_args da; 2193 uint32_t sz; 2194 2195 sz = sizeof(ipfw_obj_header) + sizeof(ipfw_xtable_info); 2196 oh = (struct _ipfw_obj_header *)ipfw_get_sopt_header(sd, sz); 2197 if (oh == NULL) 2198 return (EINVAL); 2199 2200 i = (ipfw_xtable_info *)(oh + 1); 2201 objheader_to_ti(oh, &ti); 2202 2203 IPFW_UH_RLOCK(ch); 2204 if ((tc = find_table(CHAIN_TO_NI(ch), &ti)) == NULL) { 2205 IPFW_UH_RUNLOCK(ch); 2206 return (ESRCH); 2207 } 2208 export_table_info(ch, tc, i); 2209 2210 if (sd->valsize < i->size) { 2211 2212 /* 2213 * Submitted buffer size is not enough. 2214 * WE've already filled in @i structure with 2215 * relevant table info including size, so we 2216 * can return. Buffer will be flushed automatically. 2217 */ 2218 IPFW_UH_RUNLOCK(ch); 2219 return (ENOMEM); 2220 } 2221 2222 /* 2223 * Do the actual dump in eXtended format 2224 */ 2225 memset(&da, 0, sizeof(da)); 2226 da.ch = ch; 2227 da.ti = KIDX_TO_TI(ch, tc->no.kidx); 2228 da.tc = tc; 2229 da.sd = sd; 2230 2231 ta = tc->ta; 2232 2233 ta->foreach(tc->astate, da.ti, dump_table_tentry, &da); 2234 IPFW_UH_RUNLOCK(ch); 2235 2236 return (da.error); 2237} 2238 2239/* 2240 * Dumps all table data 2241 * Data layout (version 0)(legacy): 2242 * Request: [ ipfw_xtable ], size = IP_FW_TABLE_XGETSIZE() 2243 * Reply: [ ipfw_xtable ipfw_table_xentry x N ] 2244 * 2245 * Returns 0 on success 2246 */ 2247static int 2248dump_table_v0(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 2249 struct sockopt_data *sd) 2250{ 2251 ipfw_xtable *xtbl; 2252 struct tid_info ti; 2253 struct table_config *tc; 2254 struct table_algo *ta; 2255 struct dump_args da; 2256 size_t sz, count; 2257 2258 xtbl = (ipfw_xtable *)ipfw_get_sopt_header(sd, sizeof(ipfw_xtable)); 2259 if (xtbl == NULL) 2260 return (EINVAL); 2261 2262 memset(&ti, 0, sizeof(ti)); 2263 ti.uidx = xtbl->tbl; 2264 2265 IPFW_UH_RLOCK(ch); 2266 if ((tc = find_table(CHAIN_TO_NI(ch), &ti)) == NULL) { 2267 IPFW_UH_RUNLOCK(ch); 2268 return (0); 2269 } 2270 count = table_get_count(ch, tc); 2271 sz = count * sizeof(ipfw_table_xentry) + sizeof(ipfw_xtable); 2272 2273 xtbl->cnt = count; 2274 xtbl->size = sz; 2275 xtbl->type = tc->no.subtype; 2276 xtbl->tbl = ti.uidx; 2277 2278 if (sd->valsize < sz) { 2279 2280 /* 2281 * Submitted buffer size is not enough. 2282 * WE've already filled in @i structure with 2283 * relevant table info including size, so we 2284 * can return. Buffer will be flushed automatically. 2285 */ 2286 IPFW_UH_RUNLOCK(ch); 2287 return (ENOMEM); 2288 } 2289 2290 /* Do the actual dump in eXtended format */ 2291 memset(&da, 0, sizeof(da)); 2292 da.ch = ch; 2293 da.ti = KIDX_TO_TI(ch, tc->no.kidx); 2294 da.tc = tc; 2295 da.sd = sd; 2296 2297 ta = tc->ta; 2298 2299 ta->foreach(tc->astate, da.ti, dump_table_xentry, &da); 2300 IPFW_UH_RUNLOCK(ch); 2301 2302 return (0); 2303} 2304 2305/* 2306 * Legacy function to retrieve number of items in table. 2307 */ 2308static int 2309get_table_size(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 2310 struct sockopt_data *sd) 2311{ 2312 uint32_t *tbl; 2313 struct tid_info ti; 2314 size_t sz; 2315 int error; 2316 2317 sz = sizeof(*op3) + sizeof(uint32_t); 2318 op3 = (ip_fw3_opheader *)ipfw_get_sopt_header(sd, sz); 2319 if (op3 == NULL) 2320 return (EINVAL); 2321 2322 tbl = (uint32_t *)(op3 + 1); 2323 memset(&ti, 0, sizeof(ti)); 2324 ti.uidx = *tbl; 2325 IPFW_UH_RLOCK(ch); 2326 error = ipfw_count_xtable(ch, &ti, tbl); 2327 IPFW_UH_RUNLOCK(ch); 2328 return (error); 2329} 2330 2331/* 2332 * Legacy IP_FW_TABLE_GETSIZE handler 2333 */ 2334int 2335ipfw_count_table(struct ip_fw_chain *ch, struct tid_info *ti, uint32_t *cnt) 2336{ 2337 struct table_config *tc; 2338 2339 if ((tc = find_table(CHAIN_TO_NI(ch), ti)) == NULL) 2340 return (ESRCH); 2341 *cnt = table_get_count(ch, tc); 2342 return (0); 2343} 2344 2345/* 2346 * Legacy IP_FW_TABLE_XGETSIZE handler 2347 */ 2348int 2349ipfw_count_xtable(struct ip_fw_chain *ch, struct tid_info *ti, uint32_t *cnt) 2350{ 2351 struct table_config *tc; 2352 uint32_t count; 2353 2354 if ((tc = find_table(CHAIN_TO_NI(ch), ti)) == NULL) { 2355 *cnt = 0; 2356 return (0); /* 'table all list' requires success */ 2357 } 2358 2359 count = table_get_count(ch, tc); 2360 *cnt = count * sizeof(ipfw_table_xentry); 2361 if (count > 0) 2362 *cnt += sizeof(ipfw_xtable); 2363 return (0); 2364} 2365 2366static int 2367dump_table_entry(void *e, void *arg) 2368{ 2369 struct dump_args *da; 2370 struct table_config *tc; 2371 struct table_algo *ta; 2372 ipfw_table_entry *ent; 2373 struct table_value *pval; 2374 int error; 2375 2376 da = (struct dump_args *)arg; 2377 2378 tc = da->tc; 2379 ta = tc->ta; 2380 2381 /* Out of memory, returning */ 2382 if (da->cnt == da->size) 2383 return (1); 2384 ent = da->ent++; 2385 ent->tbl = da->uidx; 2386 da->cnt++; 2387 2388 error = ta->dump_tentry(tc->astate, da->ti, e, &da->tent); 2389 if (error != 0) 2390 return (error); 2391 2392 ent->addr = da->tent.k.addr.s_addr; 2393 ent->masklen = da->tent.masklen; 2394 pval = get_table_value(da->ch, da->tc, da->tent.v.kidx); 2395 ent->value = ipfw_export_table_value_legacy(pval); 2396 2397 return (0); 2398} 2399 2400/* 2401 * Dumps table in pre-8.1 legacy format. 2402 */ 2403int 2404ipfw_dump_table_legacy(struct ip_fw_chain *ch, struct tid_info *ti, 2405 ipfw_table *tbl) 2406{ 2407 struct table_config *tc; 2408 struct table_algo *ta; 2409 struct dump_args da; 2410 2411 tbl->cnt = 0; 2412 2413 if ((tc = find_table(CHAIN_TO_NI(ch), ti)) == NULL) 2414 return (0); /* XXX: We should return ESRCH */ 2415 2416 ta = tc->ta; 2417 2418 /* This dump format supports IPv4 only */ 2419 if (tc->no.subtype != IPFW_TABLE_ADDR) 2420 return (0); 2421 2422 memset(&da, 0, sizeof(da)); 2423 da.ch = ch; 2424 da.ti = KIDX_TO_TI(ch, tc->no.kidx); 2425 da.tc = tc; 2426 da.ent = &tbl->ent[0]; 2427 da.size = tbl->size; 2428 2429 tbl->cnt = 0; 2430 ta->foreach(tc->astate, da.ti, dump_table_entry, &da); 2431 tbl->cnt = da.cnt; 2432 2433 return (0); 2434} 2435 2436/* 2437 * Dumps table entry in eXtended format (v1)(current). 2438 */ 2439static int 2440dump_table_tentry(void *e, void *arg) 2441{ 2442 struct dump_args *da; 2443 struct table_config *tc; 2444 struct table_algo *ta; 2445 struct table_value *pval; 2446 ipfw_obj_tentry *tent; 2447 int error; 2448 2449 da = (struct dump_args *)arg; 2450 2451 tc = da->tc; 2452 ta = tc->ta; 2453 2454 tent = (ipfw_obj_tentry *)ipfw_get_sopt_space(da->sd, sizeof(*tent)); 2455 /* Out of memory, returning */ 2456 if (tent == NULL) { 2457 da->error = ENOMEM; 2458 return (1); 2459 } 2460 tent->head.length = sizeof(ipfw_obj_tentry); 2461 tent->idx = da->uidx; 2462 2463 error = ta->dump_tentry(tc->astate, da->ti, e, tent); 2464 if (error != 0) 2465 return (error); 2466 2467 pval = get_table_value(da->ch, da->tc, tent->v.kidx); 2468 ipfw_export_table_value_v1(pval, &tent->v.value); 2469 2470 return (0); 2471} 2472 2473/* 2474 * Dumps table entry in eXtended format (v0). 2475 */ 2476static int 2477dump_table_xentry(void *e, void *arg) 2478{ 2479 struct dump_args *da; 2480 struct table_config *tc; 2481 struct table_algo *ta; 2482 ipfw_table_xentry *xent; 2483 ipfw_obj_tentry *tent; 2484 struct table_value *pval; 2485 int error; 2486 2487 da = (struct dump_args *)arg; 2488 2489 tc = da->tc; 2490 ta = tc->ta; 2491 2492 xent = (ipfw_table_xentry *)ipfw_get_sopt_space(da->sd, sizeof(*xent)); 2493 /* Out of memory, returning */ 2494 if (xent == NULL) 2495 return (1); 2496 xent->len = sizeof(ipfw_table_xentry); 2497 xent->tbl = da->uidx; 2498 2499 memset(&da->tent, 0, sizeof(da->tent)); 2500 tent = &da->tent; 2501 error = ta->dump_tentry(tc->astate, da->ti, e, tent); 2502 if (error != 0) 2503 return (error); 2504 2505 /* Convert current format to previous one */ 2506 xent->masklen = tent->masklen; 2507 pval = get_table_value(da->ch, da->tc, da->tent.v.kidx); 2508 xent->value = ipfw_export_table_value_legacy(pval); 2509 /* Apply some hacks */ 2510 if (tc->no.subtype == IPFW_TABLE_ADDR && tent->subtype == AF_INET) { 2511 xent->k.addr6.s6_addr32[3] = tent->k.addr.s_addr; 2512 xent->flags = IPFW_TCF_INET; 2513 } else 2514 memcpy(&xent->k, &tent->k, sizeof(xent->k)); 2515 2516 return (0); 2517} 2518 2519/* 2520 * Helper function to export table algo data 2521 * to tentry format before calling user function. 2522 * 2523 * Returns 0 on success. 2524 */ 2525static int 2526prepare_table_tentry(void *e, void *arg) 2527{ 2528 struct dump_args *da; 2529 struct table_config *tc; 2530 struct table_algo *ta; 2531 int error; 2532 2533 da = (struct dump_args *)arg; 2534 2535 tc = da->tc; 2536 ta = tc->ta; 2537 2538 error = ta->dump_tentry(tc->astate, da->ti, e, &da->tent); 2539 if (error != 0) 2540 return (error); 2541 2542 da->f(&da->tent, da->farg); 2543 2544 return (0); 2545} 2546 2547/* 2548 * Allow external consumers to read table entries in standard format. 2549 */ 2550int 2551ipfw_foreach_table_tentry(struct ip_fw_chain *ch, uint16_t kidx, 2552 ta_foreach_f *f, void *arg) 2553{ 2554 struct namedobj_instance *ni; 2555 struct table_config *tc; 2556 struct table_algo *ta; 2557 struct dump_args da; 2558 2559 ni = CHAIN_TO_NI(ch); 2560 2561 tc = (struct table_config *)ipfw_objhash_lookup_kidx(ni, kidx); 2562 if (tc == NULL) 2563 return (ESRCH); 2564 2565 ta = tc->ta; 2566 2567 memset(&da, 0, sizeof(da)); 2568 da.ch = ch; 2569 da.ti = KIDX_TO_TI(ch, tc->no.kidx); 2570 da.tc = tc; 2571 da.f = f; 2572 da.farg = arg; 2573 2574 ta->foreach(tc->astate, da.ti, prepare_table_tentry, &da); 2575 2576 return (0); 2577} 2578 2579/* 2580 * Table algorithms 2581 */ 2582 2583/* 2584 * Finds algorithm by index, table type or supplied name. 2585 * 2586 * Returns pointer to algo or NULL. 2587 */ 2588static struct table_algo * 2589find_table_algo(struct tables_config *tcfg, struct tid_info *ti, char *name) 2590{ 2591 int i, l; 2592 struct table_algo *ta; 2593 2594 if (ti->type > IPFW_TABLE_MAXTYPE) 2595 return (NULL); 2596 2597 /* Search by index */ 2598 if (ti->atype != 0) { 2599 if (ti->atype > tcfg->algo_count) 2600 return (NULL); 2601 return (tcfg->algo[ti->atype]); 2602 } 2603 2604 if (name == NULL) { 2605 /* Return default algorithm for given type if set */ 2606 return (tcfg->def_algo[ti->type]); 2607 } 2608 2609 /* Search by name */ 2610 /* TODO: better search */ 2611 for (i = 1; i <= tcfg->algo_count; i++) { 2612 ta = tcfg->algo[i]; 2613 2614 /* 2615 * One can supply additional algorithm 2616 * parameters so we compare only the first word 2617 * of supplied name: 2618 * 'addr:chash hsize=32' 2619 * '^^^^^^^^^' 2620 * 2621 */ 2622 l = strlen(ta->name); 2623 if (strncmp(name, ta->name, l) != 0) 2624 continue; 2625 if (name[l] != '\0' && name[l] != ' ') 2626 continue; 2627 /* Check if we're requesting proper table type */ 2628 if (ti->type != 0 && ti->type != ta->type) 2629 return (NULL); 2630 return (ta); 2631 } 2632 2633 return (NULL); 2634} 2635 2636/* 2637 * Register new table algo @ta. 2638 * Stores algo id inside @idx. 2639 * 2640 * Returns 0 on success. 2641 */ 2642int 2643ipfw_add_table_algo(struct ip_fw_chain *ch, struct table_algo *ta, size_t size, 2644 int *idx) 2645{ 2646 struct tables_config *tcfg; 2647 struct table_algo *ta_new; 2648 size_t sz; 2649 2650 if (size > sizeof(struct table_algo)) 2651 return (EINVAL); 2652 2653 /* Check for the required on-stack size for add/del */ 2654 sz = roundup2(ta->ta_buf_size, sizeof(void *)); 2655 if (sz > TA_BUF_SZ) 2656 return (EINVAL); 2657 2658 KASSERT(ta->type <= IPFW_TABLE_MAXTYPE,("Increase IPFW_TABLE_MAXTYPE")); 2659 2660 /* Copy algorithm data to stable storage. */ 2661 ta_new = malloc(sizeof(struct table_algo), M_IPFW, M_WAITOK | M_ZERO); 2662 memcpy(ta_new, ta, size); 2663 2664 tcfg = CHAIN_TO_TCFG(ch); 2665 2666 KASSERT(tcfg->algo_count < 255, ("Increase algo array size")); 2667 2668 tcfg->algo[++tcfg->algo_count] = ta_new; 2669 ta_new->idx = tcfg->algo_count; 2670 2671 /* Set algorithm as default one for given type */ 2672 if ((ta_new->flags & TA_FLAG_DEFAULT) != 0 && 2673 tcfg->def_algo[ta_new->type] == NULL) 2674 tcfg->def_algo[ta_new->type] = ta_new; 2675 2676 *idx = ta_new->idx; 2677 2678 return (0); 2679} 2680 2681/* 2682 * Unregisters table algo using @idx as id. 2683 * XXX: It is NOT safe to call this function in any place 2684 * other than ipfw instance destroy handler. 2685 */ 2686void 2687ipfw_del_table_algo(struct ip_fw_chain *ch, int idx) 2688{ 2689 struct tables_config *tcfg; 2690 struct table_algo *ta; 2691 2692 tcfg = CHAIN_TO_TCFG(ch); 2693 2694 KASSERT(idx <= tcfg->algo_count, ("algo idx %d out of range 1..%d", 2695 idx, tcfg->algo_count)); 2696 2697 ta = tcfg->algo[idx]; 2698 KASSERT(ta != NULL, ("algo idx %d is NULL", idx)); 2699 2700 if (tcfg->def_algo[ta->type] == ta) 2701 tcfg->def_algo[ta->type] = NULL; 2702 2703 free(ta, M_IPFW); 2704} 2705 2706/* 2707 * Lists all table algorithms currently available. 2708 * Data layout (v0)(current): 2709 * Request: [ ipfw_obj_lheader ], size = ipfw_obj_lheader.size 2710 * Reply: [ ipfw_obj_lheader ipfw_ta_info x N ] 2711 * 2712 * Returns 0 on success 2713 */ 2714static int 2715list_table_algo(struct ip_fw_chain *ch, ip_fw3_opheader *op3, 2716 struct sockopt_data *sd) 2717{ 2718 struct _ipfw_obj_lheader *olh; 2719 struct tables_config *tcfg; 2720 ipfw_ta_info *i; 2721 struct table_algo *ta; 2722 uint32_t count, n, size; 2723 2724 olh = (struct _ipfw_obj_lheader *)ipfw_get_sopt_header(sd,sizeof(*olh)); 2725 if (olh == NULL) 2726 return (EINVAL); 2727 if (sd->valsize < olh->size) 2728 return (EINVAL); 2729 2730 IPFW_UH_RLOCK(ch); 2731 tcfg = CHAIN_TO_TCFG(ch); 2732 count = tcfg->algo_count; 2733 size = count * sizeof(ipfw_ta_info) + sizeof(ipfw_obj_lheader); 2734 2735 /* Fill in header regadless of buffer size */ 2736 olh->count = count; 2737 olh->objsize = sizeof(ipfw_ta_info); 2738 2739 if (size > olh->size) { 2740 olh->size = size; 2741 IPFW_UH_RUNLOCK(ch); 2742 return (ENOMEM); 2743 } 2744 olh->size = size; 2745 2746 for (n = 1; n <= count; n++) { 2747 i = (ipfw_ta_info *)ipfw_get_sopt_space(sd, sizeof(*i)); 2748 KASSERT(i != NULL, ("previously checked buffer is not enough")); 2749 ta = tcfg->algo[n]; 2750 strlcpy(i->algoname, ta->name, sizeof(i->algoname)); 2751 i->type = ta->type; 2752 i->refcnt = ta->refcnt; 2753 } 2754 2755 IPFW_UH_RUNLOCK(ch); 2756 2757 return (0); 2758} 2759 2760static int 2761classify_srcdst(ipfw_insn *cmd, uint16_t *puidx, uint8_t *ptype) 2762{ 2763 /* Basic IPv4/IPv6 or u32 lookups */ 2764 *puidx = cmd->arg1; 2765 /* Assume ADDR by default */ 2766 *ptype = IPFW_TABLE_ADDR; 2767 int v; 2768 2769 if (F_LEN(cmd) > F_INSN_SIZE(ipfw_insn_u32)) { 2770 /* 2771 * generic lookup. The key must be 2772 * in 32bit big-endian format. 2773 */ 2774 v = ((ipfw_insn_u32 *)cmd)->d[1]; 2775 switch (v) { 2776 case 0: 2777 case 1: 2778 /* IPv4 src/dst */ 2779 break; 2780 case 2: 2781 case 3: 2782 /* src/dst port */ 2783 *ptype = IPFW_TABLE_NUMBER; 2784 break; 2785 case 4: 2786 /* uid/gid */ 2787 *ptype = IPFW_TABLE_NUMBER; 2788 break; 2789 case 5: 2790 /* jid */ 2791 *ptype = IPFW_TABLE_NUMBER; 2792 break; 2793 case 6: 2794 /* dscp */ 2795 *ptype = IPFW_TABLE_NUMBER; 2796 break; 2797 } 2798 } 2799 2800 return (0); 2801} 2802 2803static int 2804classify_via(ipfw_insn *cmd, uint16_t *puidx, uint8_t *ptype) 2805{ 2806 ipfw_insn_if *cmdif; 2807 2808 /* Interface table, possibly */ 2809 cmdif = (ipfw_insn_if *)cmd; 2810 if (cmdif->name[0] != '\1') 2811 return (1); 2812 2813 *ptype = IPFW_TABLE_INTERFACE; 2814 *puidx = cmdif->p.kidx; 2815 2816 return (0); 2817} 2818 2819static int 2820classify_flow(ipfw_insn *cmd, uint16_t *puidx, uint8_t *ptype) 2821{ 2822 2823 *puidx = cmd->arg1; 2824 *ptype = IPFW_TABLE_FLOW; 2825 2826 return (0); 2827} 2828 2829static void 2830update_arg1(ipfw_insn *cmd, uint16_t idx) 2831{ 2832 2833 cmd->arg1 = idx; 2834} 2835 2836static void 2837update_via(ipfw_insn *cmd, uint16_t idx) 2838{ 2839 ipfw_insn_if *cmdif; 2840 2841 cmdif = (ipfw_insn_if *)cmd; 2842 cmdif->p.kidx = idx; 2843} 2844 2845static int 2846table_findbyname(struct ip_fw_chain *ch, struct tid_info *ti, 2847 struct named_object **pno) 2848{ 2849 struct table_config *tc; 2850 int error; 2851 2852 IPFW_UH_WLOCK_ASSERT(ch); 2853 2854 error = find_table_err(CHAIN_TO_NI(ch), ti, &tc); 2855 if (error != 0) 2856 return (error); 2857 2858 *pno = &tc->no; 2859 return (0); 2860} 2861 2862/* XXX: sets-sets! */ 2863static struct named_object * 2864table_findbykidx(struct ip_fw_chain *ch, uint16_t idx) 2865{ 2866 struct namedobj_instance *ni; 2867 struct table_config *tc; 2868 2869 IPFW_UH_WLOCK_ASSERT(ch); 2870 ni = CHAIN_TO_NI(ch); 2871 tc = (struct table_config *)ipfw_objhash_lookup_kidx(ni, idx); 2872 KASSERT(tc != NULL, ("Table with index %d not found", idx)); 2873 2874 return (&tc->no); 2875} 2876 2877static struct opcode_obj_rewrite opcodes[] = { 2878 { 2879 O_IP_SRC_LOOKUP, IPFW_TLV_TBL_NAME, 2880 classify_srcdst, update_arg1, 2881 table_findbyname, table_findbykidx, create_table_compat 2882 }, 2883 { 2884 O_IP_DST_LOOKUP, IPFW_TLV_TBL_NAME, 2885 classify_srcdst, update_arg1, 2886 table_findbyname, table_findbykidx, create_table_compat 2887 }, 2888 { 2889 O_IP_FLOW_LOOKUP, IPFW_TLV_TBL_NAME, 2890 classify_flow, update_arg1, 2891 table_findbyname, table_findbykidx, create_table_compat 2892 }, 2893 { 2894 O_XMIT, IPFW_TLV_TBL_NAME, 2895 classify_via, update_via, 2896 table_findbyname, table_findbykidx, create_table_compat 2897 }, 2898 { 2899 O_RECV, IPFW_TLV_TBL_NAME, 2900 classify_via, update_via, 2901 table_findbyname, table_findbykidx, create_table_compat 2902 }, 2903 { 2904 O_VIA, IPFW_TLV_TBL_NAME, 2905 classify_via, update_via, 2906 table_findbyname, table_findbykidx, create_table_compat 2907 }, 2908}; 2909 2910 2911/* 2912 * Checks table name for validity. 2913 * Enforce basic length checks, the rest 2914 * should be done in userland. 2915 * 2916 * Returns 0 if name is considered valid. 2917 */ 2918static int 2919check_table_name(const char *name) 2920{ 2921 2922 /* 2923 * TODO: do some more complicated checks 2924 */ 2925 return (ipfw_check_object_name_generic(name)); 2926} 2927 2928/* 2929 * Finds table config based on either legacy index 2930 * or name in ntlv. 2931 * Note @ti structure contains unchecked data from userland. 2932 * 2933 * Returns 0 in success and fills in @tc with found config 2934 */ 2935static int 2936find_table_err(struct namedobj_instance *ni, struct tid_info *ti, 2937 struct table_config **tc) 2938{ 2939 char *name, bname[16]; 2940 struct named_object *no; 2941 ipfw_obj_ntlv *ntlv; 2942 uint32_t set; 2943 2944 if (ti->tlvs != NULL) { 2945 ntlv = ipfw_find_name_tlv_type(ti->tlvs, ti->tlen, ti->uidx, 2946 IPFW_TLV_TBL_NAME); 2947 if (ntlv == NULL) 2948 return (EINVAL); 2949 name = ntlv->name; 2950 2951 /* 2952 * Use set provided by @ti instead of @ntlv one. 2953 * This is needed due to different sets behavior 2954 * controlled by V_fw_tables_sets. 2955 */ 2956 set = ti->set; 2957 } else { 2958 snprintf(bname, sizeof(bname), "%d", ti->uidx); 2959 name = bname; 2960 set = 0; 2961 } 2962 2963 no = ipfw_objhash_lookup_name(ni, set, name); 2964 *tc = (struct table_config *)no; 2965 2966 return (0); 2967} 2968 2969/* 2970 * Finds table config based on either legacy index 2971 * or name in ntlv. 2972 * Note @ti structure contains unchecked data from userland. 2973 * 2974 * Returns pointer to table_config or NULL. 2975 */ 2976static struct table_config * 2977find_table(struct namedobj_instance *ni, struct tid_info *ti) 2978{ 2979 struct table_config *tc; 2980 2981 if (find_table_err(ni, ti, &tc) != 0) 2982 return (NULL); 2983 2984 return (tc); 2985} 2986 2987/* 2988 * Allocate new table config structure using 2989 * specified @algo and @aname. 2990 * 2991 * Returns pointer to config or NULL. 2992 */ 2993static struct table_config * 2994alloc_table_config(struct ip_fw_chain *ch, struct tid_info *ti, 2995 struct table_algo *ta, char *aname, uint8_t tflags) 2996{ 2997 char *name, bname[16]; 2998 struct table_config *tc; 2999 int error; 3000 ipfw_obj_ntlv *ntlv; 3001 uint32_t set; 3002 3003 if (ti->tlvs != NULL) { 3004 ntlv = ipfw_find_name_tlv_type(ti->tlvs, ti->tlen, ti->uidx, 3005 IPFW_TLV_TBL_NAME); 3006 if (ntlv == NULL) 3007 return (NULL); 3008 name = ntlv->name; 3009 set = ntlv->set; 3010 } else { 3011 /* Compat part: convert number to string representation */ 3012 snprintf(bname, sizeof(bname), "%d", ti->uidx); 3013 name = bname; 3014 set = 0; 3015 } 3016 3017 tc = malloc(sizeof(struct table_config), M_IPFW, M_WAITOK | M_ZERO); 3018 tc->no.name = tc->tablename; 3019 tc->no.subtype = ta->type; 3020 tc->no.set = set; 3021 tc->tflags = tflags; 3022 tc->ta = ta; 3023 strlcpy(tc->tablename, name, sizeof(tc->tablename)); 3024 /* Set "shared" value type by default */ 3025 tc->vshared = 1; 3026 3027 /* Preallocate data structures for new tables */ 3028 error = ta->init(ch, &tc->astate, &tc->ti_copy, aname, tflags); 3029 if (error != 0) { 3030 free(tc, M_IPFW); 3031 return (NULL); 3032 } 3033 3034 return (tc); 3035} 3036 3037/* 3038 * Destroys table state and config. 3039 */ 3040static void 3041free_table_config(struct namedobj_instance *ni, struct table_config *tc) 3042{ 3043 3044 KASSERT(tc->linked == 0, ("free() on linked config")); 3045 /* UH lock MUST NOT be held */ 3046 3047 /* 3048 * We're using ta without any locking/referencing. 3049 * TODO: fix this if we're going to use unloadable algos. 3050 */ 3051 tc->ta->destroy(tc->astate, &tc->ti_copy); 3052 free(tc, M_IPFW); 3053} 3054 3055/* 3056 * Links @tc to @chain table named instance. 3057 * Sets appropriate type/states in @chain table info. 3058 */ 3059static void 3060link_table(struct ip_fw_chain *ch, struct table_config *tc) 3061{ 3062 struct namedobj_instance *ni; 3063 struct table_info *ti; 3064 uint16_t kidx; 3065 3066 IPFW_UH_WLOCK_ASSERT(ch); 3067 IPFW_WLOCK_ASSERT(ch); 3068 3069 ni = CHAIN_TO_NI(ch); 3070 kidx = tc->no.kidx; 3071 3072 ipfw_objhash_add(ni, &tc->no); 3073 3074 ti = KIDX_TO_TI(ch, kidx); 3075 *ti = tc->ti_copy; 3076 3077 /* Notify algo on real @ti address */ 3078 if (tc->ta->change_ti != NULL) 3079 tc->ta->change_ti(tc->astate, ti); 3080 3081 tc->linked = 1; 3082 tc->ta->refcnt++; 3083} 3084 3085/* 3086 * Unlinks @tc from @chain table named instance. 3087 * Zeroes states in @chain and stores them in @tc. 3088 */ 3089static void 3090unlink_table(struct ip_fw_chain *ch, struct table_config *tc) 3091{ 3092 struct namedobj_instance *ni; 3093 struct table_info *ti; 3094 uint16_t kidx; 3095 3096 IPFW_UH_WLOCK_ASSERT(ch); 3097 IPFW_WLOCK_ASSERT(ch); 3098 3099 ni = CHAIN_TO_NI(ch); 3100 kidx = tc->no.kidx; 3101 3102 /* Clear state. @ti copy is already saved inside @tc */ 3103 ipfw_objhash_del(ni, &tc->no); 3104 ti = KIDX_TO_TI(ch, kidx); 3105 memset(ti, 0, sizeof(struct table_info)); 3106 tc->linked = 0; 3107 tc->ta->refcnt--; 3108 3109 /* Notify algo on real @ti address */ 3110 if (tc->ta->change_ti != NULL) 3111 tc->ta->change_ti(tc->astate, NULL); 3112} 3113 3114struct swap_table_args { 3115 int set; 3116 int new_set; 3117 int mv; 3118}; 3119 3120/* 3121 * Change set for each matching table. 3122 * 3123 * Ensure we dispatch each table once by setting/checking ochange 3124 * fields. 3125 */ 3126static void 3127swap_table_set(struct namedobj_instance *ni, struct named_object *no, 3128 void *arg) 3129{ 3130 struct table_config *tc; 3131 struct swap_table_args *sta; 3132 3133 tc = (struct table_config *)no; 3134 sta = (struct swap_table_args *)arg; 3135 3136 if (no->set != sta->set && (no->set != sta->new_set || sta->mv != 0)) 3137 return; 3138 3139 if (tc->ochanged != 0) 3140 return; 3141 3142 tc->ochanged = 1; 3143 ipfw_objhash_del(ni, no); 3144 if (no->set == sta->set) 3145 no->set = sta->new_set; 3146 else 3147 no->set = sta->set; 3148 ipfw_objhash_add(ni, no); 3149} 3150 3151/* 3152 * Cleans up ochange field for all tables. 3153 */ 3154static void 3155clean_table_set_data(struct namedobj_instance *ni, struct named_object *no, 3156 void *arg) 3157{ 3158 struct table_config *tc; 3159 struct swap_table_args *sta; 3160 3161 tc = (struct table_config *)no; 3162 sta = (struct swap_table_args *)arg; 3163 3164 tc->ochanged = 0; 3165} 3166 3167/* 3168 * Swaps tables within two sets. 3169 */ 3170void 3171ipfw_swap_tables_sets(struct ip_fw_chain *ch, uint32_t set, 3172 uint32_t new_set, int mv) 3173{ 3174 struct swap_table_args sta; 3175 3176 IPFW_UH_WLOCK_ASSERT(ch); 3177 3178 sta.set = set; 3179 sta.new_set = new_set; 3180 sta.mv = mv; 3181 3182 ipfw_objhash_foreach(CHAIN_TO_NI(ch), swap_table_set, &sta); 3183 ipfw_objhash_foreach(CHAIN_TO_NI(ch), clean_table_set_data, &sta); 3184} 3185 3186/* 3187 * Move all tables which are reference by rules in @rr to set @new_set. 3188 * Makes sure that all relevant tables are referenced ONLLY by given rules. 3189 * 3190 * Returns 0 on success, 3191 */ 3192int 3193ipfw_move_tables_sets(struct ip_fw_chain *ch, ipfw_range_tlv *rt, 3194 uint32_t new_set) 3195{ 3196 struct ip_fw *rule; 3197 struct table_config *tc; 3198 struct named_object *no; 3199 struct namedobj_instance *ni; 3200 int bad, i, l, cmdlen; 3201 uint16_t kidx; 3202 ipfw_insn *cmd; 3203 3204 IPFW_UH_WLOCK_ASSERT(ch); 3205 3206 ni = CHAIN_TO_NI(ch); 3207 3208 /* Stage 1: count number of references by given rules */ 3209 for (i = 0; i < ch->n_rules - 1; i++) { 3210 rule = ch->map[i]; 3211 if (ipfw_match_range(rule, rt) == 0) 3212 continue; 3213 3214 l = rule->cmd_len; 3215 cmd = rule->cmd; 3216 cmdlen = 0; 3217 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 3218 cmdlen = F_LEN(cmd); 3219 if (classify_opcode_kidx(cmd, &kidx) != 0) 3220 continue; 3221 no = ipfw_objhash_lookup_kidx(ni, kidx); 3222 KASSERT(no != NULL, 3223 ("objhash lookup failed on index %d", kidx)); 3224 tc = (struct table_config *)no; 3225 tc->ocount++; 3226 } 3227 3228 } 3229 3230 /* Stage 2: verify "ownership" */ 3231 bad = 0; 3232 for (i = 0; i < ch->n_rules - 1; i++) { 3233 rule = ch->map[i]; 3234 if (ipfw_match_range(rule, rt) == 0) 3235 continue; 3236 3237 l = rule->cmd_len; 3238 cmd = rule->cmd; 3239 cmdlen = 0; 3240 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 3241 cmdlen = F_LEN(cmd); 3242 if (classify_opcode_kidx(cmd, &kidx) != 0) 3243 continue; 3244 no = ipfw_objhash_lookup_kidx(ni, kidx); 3245 KASSERT(no != NULL, 3246 ("objhash lookup failed on index %d", kidx)); 3247 tc = (struct table_config *)no; 3248 if (tc->no.refcnt != tc->ocount) { 3249 3250 /* 3251 * Number of references differ: 3252 * Other rule(s) are holding reference to given 3253 * table, so it is not possible to change its set. 3254 * 3255 * Note that refcnt may account 3256 * references to some going-to-be-added rules. 3257 * Since we don't know their numbers (and event 3258 * if they will be added) it is perfectly OK 3259 * to return error here. 3260 */ 3261 bad = 1; 3262 break; 3263 } 3264 } 3265 3266 if (bad != 0) 3267 break; 3268 } 3269 3270 /* Stage 3: change set or cleanup */ 3271 for (i = 0; i < ch->n_rules - 1; i++) { 3272 rule = ch->map[i]; 3273 if (ipfw_match_range(rule, rt) == 0) 3274 continue; 3275 3276 l = rule->cmd_len; 3277 cmd = rule->cmd; 3278 cmdlen = 0; 3279 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) { 3280 cmdlen = F_LEN(cmd); 3281 if (classify_opcode_kidx(cmd, &kidx) != 0) 3282 continue; 3283 no = ipfw_objhash_lookup_kidx(ni, kidx); 3284 KASSERT(no != NULL, 3285 ("objhash lookup failed on index %d", kidx)); 3286 tc = (struct table_config *)no; 3287 3288 tc->ocount = 0; 3289 if (bad != 0) 3290 continue; 3291 3292 /* Actually change set. */ 3293 ipfw_objhash_del(ni, no); 3294 no->set = new_set; 3295 ipfw_objhash_add(ni, no); 3296 } 3297 } 3298 3299 return (bad); 3300} 3301 3302static struct ipfw_sopt_handler scodes[] = { 3303 { IP_FW_TABLE_XCREATE, 0, HDIR_SET, create_table }, 3304 { IP_FW_TABLE_XDESTROY, 0, HDIR_SET, flush_table_v0 }, 3305 { IP_FW_TABLE_XFLUSH, 0, HDIR_SET, flush_table_v0 }, 3306 { IP_FW_TABLE_XMODIFY, 0, HDIR_BOTH, modify_table }, 3307 { IP_FW_TABLE_XINFO, 0, HDIR_GET, describe_table }, 3308 { IP_FW_TABLES_XLIST, 0, HDIR_GET, list_tables }, 3309 { IP_FW_TABLE_XLIST, 0, HDIR_GET, dump_table_v0 }, 3310 { IP_FW_TABLE_XLIST, 1, HDIR_GET, dump_table_v1 }, 3311 { IP_FW_TABLE_XADD, 0, HDIR_BOTH, manage_table_ent_v0 }, 3312 { IP_FW_TABLE_XADD, 1, HDIR_BOTH, manage_table_ent_v1 }, 3313 { IP_FW_TABLE_XDEL, 0, HDIR_BOTH, manage_table_ent_v0 }, 3314 { IP_FW_TABLE_XDEL, 1, HDIR_BOTH, manage_table_ent_v1 }, 3315 { IP_FW_TABLE_XFIND, 0, HDIR_GET, find_table_entry }, 3316 { IP_FW_TABLE_XSWAP, 0, HDIR_SET, swap_table }, 3317 { IP_FW_TABLES_ALIST, 0, HDIR_GET, list_table_algo }, 3318 { IP_FW_TABLE_XGETSIZE, 0, HDIR_GET, get_table_size }, 3319}; 3320 3321static void 3322destroy_table_locked(struct namedobj_instance *ni, struct named_object *no, 3323 void *arg) 3324{ 3325 3326 unlink_table((struct ip_fw_chain *)arg, (struct table_config *)no); 3327 if (ipfw_objhash_free_idx(ni, no->kidx) != 0) 3328 printf("Error unlinking kidx %d from table %s\n", 3329 no->kidx, no->name); 3330 free_table_config(ni, (struct table_config *)no); 3331} 3332 3333/* 3334 * Shuts tables module down. 3335 */ 3336void 3337ipfw_destroy_tables(struct ip_fw_chain *ch, int last) 3338{ 3339 3340 IPFW_DEL_SOPT_HANDLER(last, scodes); 3341 IPFW_DEL_OBJ_REWRITER(last, opcodes); 3342 3343 /* Remove all tables from working set */ 3344 IPFW_UH_WLOCK(ch); 3345 IPFW_WLOCK(ch); 3346 ipfw_objhash_foreach(CHAIN_TO_NI(ch), destroy_table_locked, ch); 3347 IPFW_WUNLOCK(ch); 3348 IPFW_UH_WUNLOCK(ch); 3349 3350 /* Free pointers itself */ 3351 free(ch->tablestate, M_IPFW); 3352 3353 ipfw_table_value_destroy(ch, last); 3354 ipfw_table_algo_destroy(ch); 3355 3356 ipfw_objhash_destroy(CHAIN_TO_NI(ch)); 3357 free(CHAIN_TO_TCFG(ch), M_IPFW); 3358} 3359 3360/* 3361 * Starts tables module. 3362 */ 3363int 3364ipfw_init_tables(struct ip_fw_chain *ch, int first) 3365{ 3366 struct tables_config *tcfg; 3367 3368 /* Allocate pointers */ 3369 ch->tablestate = malloc(V_fw_tables_max * sizeof(struct table_info), 3370 M_IPFW, M_WAITOK | M_ZERO); 3371 3372 tcfg = malloc(sizeof(struct tables_config), M_IPFW, M_WAITOK | M_ZERO); 3373 tcfg->namehash = ipfw_objhash_create(V_fw_tables_max); 3374 ch->tblcfg = tcfg; 3375 3376 ipfw_table_value_init(ch, first); 3377 ipfw_table_algo_init(ch); 3378 3379 IPFW_ADD_OBJ_REWRITER(first, opcodes); 3380 IPFW_ADD_SOPT_HANDLER(first, scodes); 3381 return (0); 3382} 3383 3384 3385 3386