1 /* 2 * x_tables core - Backend for {ip,ip6,arp}_tables 3 * 4 * Copyright (C) 2006-2006 Harald Welte <laforge@netfilter.org> 5 * Copyright (C) 2006-2012 Patrick McHardy <kaber@trash.net> 6 * 7 * Based on existing ip_tables code which is 8 * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling 9 * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org> 10 * 11 * This program is free software; you can redistribute it and/or modify 12 * it under the terms of the GNU General Public License version 2 as 13 * published by the Free Software Foundation. 14 * 15 */ 16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 17 #include <linux/kernel.h> 18 #include <linux/module.h> 19 #include <linux/socket.h> 20 #include <linux/net.h> 21 #include <linux/proc_fs.h> 22 #include <linux/seq_file.h> 23 #include <linux/string.h> 24 #include <linux/vmalloc.h> 25 #include <linux/mutex.h> 26 #include <linux/mm.h> 27 #include <linux/slab.h> 28 #include <linux/audit.h> 29 #include <linux/user_namespace.h> 30 #include <net/net_namespace.h> 31 32 #include <linux/netfilter/x_tables.h> 33 #include <linux/netfilter_arp.h> 34 #include <linux/netfilter_ipv4/ip_tables.h> 35 #include <linux/netfilter_ipv6/ip6_tables.h> 36 #include <linux/netfilter_arp/arp_tables.h> 37 38 MODULE_LICENSE("GPL"); 39 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>"); 40 MODULE_DESCRIPTION("{ip,ip6,arp,eb}_tables backend module"); 41 42 #define XT_PCPU_BLOCK_SIZE 4096 43 #define XT_MAX_TABLE_SIZE (512 * 1024 * 1024) 44 45 struct compat_delta { 46 unsigned int offset; /* offset in kernel */ 47 int delta; /* delta in 32bit user land */ 48 }; 49 50 struct xt_af { 51 struct mutex mutex; 52 struct list_head match; 53 struct list_head target; 54 #ifdef CONFIG_COMPAT 55 struct mutex compat_mutex; 56 struct compat_delta *compat_tab; 57 unsigned int number; /* number of slots in compat_tab[] */ 58 unsigned int cur; /* number of used slots in compat_tab[] */ 59 #endif 60 }; 61 62 static struct xt_af *xt; 63 64 static const char *const xt_prefix[NFPROTO_NUMPROTO] = { 65 [NFPROTO_UNSPEC] = "x", 66 [NFPROTO_IPV4] = "ip", 67 [NFPROTO_ARP] = "arp", 68 [NFPROTO_BRIDGE] = "eb", 69 [NFPROTO_IPV6] = "ip6", 70 }; 71 72 /* Registration hooks for targets. */ 73 int xt_register_target(struct xt_target *target) 74 { 75 u_int8_t af = target->family; 76 77 mutex_lock(&xt[af].mutex); 78 list_add(&target->list, &xt[af].target); 79 mutex_unlock(&xt[af].mutex); 80 return 0; 81 } 82 EXPORT_SYMBOL(xt_register_target); 83 84 void 85 xt_unregister_target(struct xt_target *target) 86 { 87 u_int8_t af = target->family; 88 89 mutex_lock(&xt[af].mutex); 90 list_del(&target->list); 91 mutex_unlock(&xt[af].mutex); 92 } 93 EXPORT_SYMBOL(xt_unregister_target); 94 95 int 96 xt_register_targets(struct xt_target *target, unsigned int n) 97 { 98 unsigned int i; 99 int err = 0; 100 101 for (i = 0; i < n; i++) { 102 err = xt_register_target(&target[i]); 103 if (err) 104 goto err; 105 } 106 return err; 107 108 err: 109 if (i > 0) 110 xt_unregister_targets(target, i); 111 return err; 112 } 113 EXPORT_SYMBOL(xt_register_targets); 114 115 void 116 xt_unregister_targets(struct xt_target *target, unsigned int n) 117 { 118 while (n-- > 0) 119 xt_unregister_target(&target[n]); 120 } 121 EXPORT_SYMBOL(xt_unregister_targets); 122 123 int xt_register_match(struct xt_match *match) 124 { 125 u_int8_t af = match->family; 126 127 mutex_lock(&xt[af].mutex); 128 list_add(&match->list, &xt[af].match); 129 mutex_unlock(&xt[af].mutex); 130 return 0; 131 } 132 EXPORT_SYMBOL(xt_register_match); 133 134 void 135 xt_unregister_match(struct xt_match *match) 136 { 137 u_int8_t af = match->family; 138 139 mutex_lock(&xt[af].mutex); 140 list_del(&match->list); 141 mutex_unlock(&xt[af].mutex); 142 } 143 EXPORT_SYMBOL(xt_unregister_match); 144 145 int 146 xt_register_matches(struct xt_match *match, unsigned int n) 147 { 148 unsigned int i; 149 int err = 0; 150 151 for (i = 0; i < n; i++) { 152 err = xt_register_match(&match[i]); 153 if (err) 154 goto err; 155 } 156 return err; 157 158 err: 159 if (i > 0) 160 xt_unregister_matches(match, i); 161 return err; 162 } 163 EXPORT_SYMBOL(xt_register_matches); 164 165 void 166 xt_unregister_matches(struct xt_match *match, unsigned int n) 167 { 168 while (n-- > 0) 169 xt_unregister_match(&match[n]); 170 } 171 EXPORT_SYMBOL(xt_unregister_matches); 172 173 174 /* 175 * These are weird, but module loading must not be done with mutex 176 * held (since they will register), and we have to have a single 177 * function to use. 178 */ 179 180 /* Find match, grabs ref. Returns ERR_PTR() on error. */ 181 struct xt_match *xt_find_match(u8 af, const char *name, u8 revision) 182 { 183 struct xt_match *m; 184 int err = -ENOENT; 185 186 mutex_lock(&xt[af].mutex); 187 list_for_each_entry(m, &xt[af].match, list) { 188 if (strcmp(m->name, name) == 0) { 189 if (m->revision == revision) { 190 if (try_module_get(m->me)) { 191 mutex_unlock(&xt[af].mutex); 192 return m; 193 } 194 } else 195 err = -EPROTOTYPE; /* Found something. */ 196 } 197 } 198 mutex_unlock(&xt[af].mutex); 199 200 if (af != NFPROTO_UNSPEC) 201 /* Try searching again in the family-independent list */ 202 return xt_find_match(NFPROTO_UNSPEC, name, revision); 203 204 return ERR_PTR(err); 205 } 206 EXPORT_SYMBOL(xt_find_match); 207 208 struct xt_match * 209 xt_request_find_match(uint8_t nfproto, const char *name, uint8_t revision) 210 { 211 struct xt_match *match; 212 213 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN) 214 return ERR_PTR(-EINVAL); 215 216 match = xt_find_match(nfproto, name, revision); 217 if (IS_ERR(match)) { 218 request_module("%st_%s", xt_prefix[nfproto], name); 219 match = xt_find_match(nfproto, name, revision); 220 } 221 222 return match; 223 } 224 EXPORT_SYMBOL_GPL(xt_request_find_match); 225 226 /* Find target, grabs ref. Returns ERR_PTR() on error. */ 227 struct xt_target *xt_find_target(u8 af, const char *name, u8 revision) 228 { 229 struct xt_target *t; 230 int err = -ENOENT; 231 232 mutex_lock(&xt[af].mutex); 233 list_for_each_entry(t, &xt[af].target, list) { 234 if (strcmp(t->name, name) == 0) { 235 if (t->revision == revision) { 236 if (try_module_get(t->me)) { 237 mutex_unlock(&xt[af].mutex); 238 return t; 239 } 240 } else 241 err = -EPROTOTYPE; /* Found something. */ 242 } 243 } 244 mutex_unlock(&xt[af].mutex); 245 246 if (af != NFPROTO_UNSPEC) 247 /* Try searching again in the family-independent list */ 248 return xt_find_target(NFPROTO_UNSPEC, name, revision); 249 250 return ERR_PTR(err); 251 } 252 EXPORT_SYMBOL(xt_find_target); 253 254 struct xt_target *xt_request_find_target(u8 af, const char *name, u8 revision) 255 { 256 struct xt_target *target; 257 258 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN) 259 return ERR_PTR(-EINVAL); 260 261 target = xt_find_target(af, name, revision); 262 if (IS_ERR(target)) { 263 request_module("%st_%s", xt_prefix[af], name); 264 target = xt_find_target(af, name, revision); 265 } 266 267 return target; 268 } 269 EXPORT_SYMBOL_GPL(xt_request_find_target); 270 271 272 static int xt_obj_to_user(u16 __user *psize, u16 size, 273 void __user *pname, const char *name, 274 u8 __user *prev, u8 rev) 275 { 276 if (put_user(size, psize)) 277 return -EFAULT; 278 if (copy_to_user(pname, name, strlen(name) + 1)) 279 return -EFAULT; 280 if (put_user(rev, prev)) 281 return -EFAULT; 282 283 return 0; 284 } 285 286 #define XT_OBJ_TO_USER(U, K, TYPE, C_SIZE) \ 287 xt_obj_to_user(&U->u.TYPE##_size, C_SIZE ? : K->u.TYPE##_size, \ 288 U->u.user.name, K->u.kernel.TYPE->name, \ 289 &U->u.user.revision, K->u.kernel.TYPE->revision) 290 291 int xt_data_to_user(void __user *dst, const void *src, 292 int usersize, int size, int aligned_size) 293 { 294 usersize = usersize ? : size; 295 if (copy_to_user(dst, src, usersize)) 296 return -EFAULT; 297 if (usersize != aligned_size && 298 clear_user(dst + usersize, aligned_size - usersize)) 299 return -EFAULT; 300 301 return 0; 302 } 303 EXPORT_SYMBOL_GPL(xt_data_to_user); 304 305 #define XT_DATA_TO_USER(U, K, TYPE) \ 306 xt_data_to_user(U->data, K->data, \ 307 K->u.kernel.TYPE->usersize, \ 308 K->u.kernel.TYPE->TYPE##size, \ 309 XT_ALIGN(K->u.kernel.TYPE->TYPE##size)) 310 311 int xt_match_to_user(const struct xt_entry_match *m, 312 struct xt_entry_match __user *u) 313 { 314 return XT_OBJ_TO_USER(u, m, match, 0) || 315 XT_DATA_TO_USER(u, m, match); 316 } 317 EXPORT_SYMBOL_GPL(xt_match_to_user); 318 319 int xt_target_to_user(const struct xt_entry_target *t, 320 struct xt_entry_target __user *u) 321 { 322 return XT_OBJ_TO_USER(u, t, target, 0) || 323 XT_DATA_TO_USER(u, t, target); 324 } 325 EXPORT_SYMBOL_GPL(xt_target_to_user); 326 327 static int match_revfn(u8 af, const char *name, u8 revision, int *bestp) 328 { 329 const struct xt_match *m; 330 int have_rev = 0; 331 332 list_for_each_entry(m, &xt[af].match, list) { 333 if (strcmp(m->name, name) == 0) { 334 if (m->revision > *bestp) 335 *bestp = m->revision; 336 if (m->revision == revision) 337 have_rev = 1; 338 } 339 } 340 341 if (af != NFPROTO_UNSPEC && !have_rev) 342 return match_revfn(NFPROTO_UNSPEC, name, revision, bestp); 343 344 return have_rev; 345 } 346 347 static int target_revfn(u8 af, const char *name, u8 revision, int *bestp) 348 { 349 const struct xt_target *t; 350 int have_rev = 0; 351 352 list_for_each_entry(t, &xt[af].target, list) { 353 if (strcmp(t->name, name) == 0) { 354 if (t->revision > *bestp) 355 *bestp = t->revision; 356 if (t->revision == revision) 357 have_rev = 1; 358 } 359 } 360 361 if (af != NFPROTO_UNSPEC && !have_rev) 362 return target_revfn(NFPROTO_UNSPEC, name, revision, bestp); 363 364 return have_rev; 365 } 366 367 /* Returns true or false (if no such extension at all) */ 368 int xt_find_revision(u8 af, const char *name, u8 revision, int target, 369 int *err) 370 { 371 int have_rev, best = -1; 372 373 mutex_lock(&xt[af].mutex); 374 if (target == 1) 375 have_rev = target_revfn(af, name, revision, &best); 376 else 377 have_rev = match_revfn(af, name, revision, &best); 378 mutex_unlock(&xt[af].mutex); 379 380 /* Nothing at all? Return 0 to try loading module. */ 381 if (best == -1) { 382 *err = -ENOENT; 383 return 0; 384 } 385 386 *err = best; 387 if (!have_rev) 388 *err = -EPROTONOSUPPORT; 389 return 1; 390 } 391 EXPORT_SYMBOL_GPL(xt_find_revision); 392 393 static char * 394 textify_hooks(char *buf, size_t size, unsigned int mask, uint8_t nfproto) 395 { 396 static const char *const inetbr_names[] = { 397 "PREROUTING", "INPUT", "FORWARD", 398 "OUTPUT", "POSTROUTING", "BROUTING", 399 }; 400 static const char *const arp_names[] = { 401 "INPUT", "FORWARD", "OUTPUT", 402 }; 403 const char *const *names; 404 unsigned int i, max; 405 char *p = buf; 406 bool np = false; 407 int res; 408 409 names = (nfproto == NFPROTO_ARP) ? arp_names : inetbr_names; 410 max = (nfproto == NFPROTO_ARP) ? ARRAY_SIZE(arp_names) : 411 ARRAY_SIZE(inetbr_names); 412 *p = '\0'; 413 for (i = 0; i < max; ++i) { 414 if (!(mask & (1 << i))) 415 continue; 416 res = snprintf(p, size, "%s%s", np ? "/" : "", names[i]); 417 if (res > 0) { 418 size -= res; 419 p += res; 420 } 421 np = true; 422 } 423 424 return buf; 425 } 426 427 /** 428 * xt_check_proc_name - check that name is suitable for /proc file creation 429 * 430 * @name: file name candidate 431 * @size: length of buffer 432 * 433 * some x_tables modules wish to create a file in /proc. 434 * This function makes sure that the name is suitable for this 435 * purpose, it checks that name is NUL terminated and isn't a 'special' 436 * name, like "..". 437 * 438 * returns negative number on error or 0 if name is useable. 439 */ 440 int xt_check_proc_name(const char *name, unsigned int size) 441 { 442 if (name[0] == '\0') 443 return -EINVAL; 444 445 if (strnlen(name, size) == size) 446 return -ENAMETOOLONG; 447 448 if (strcmp(name, ".") == 0 || 449 strcmp(name, "..") == 0 || 450 strchr(name, '/')) 451 return -EINVAL; 452 453 return 0; 454 } 455 EXPORT_SYMBOL(xt_check_proc_name); 456 457 int xt_check_match(struct xt_mtchk_param *par, 458 unsigned int size, u_int8_t proto, bool inv_proto) 459 { 460 int ret; 461 462 if (XT_ALIGN(par->match->matchsize) != size && 463 par->match->matchsize != -1) { 464 /* 465 * ebt_among is exempt from centralized matchsize checking 466 * because it uses a dynamic-size data set. 467 */ 468 pr_err_ratelimited("%s_tables: %s.%u match: invalid size %u (kernel) != (user) %u\n", 469 xt_prefix[par->family], par->match->name, 470 par->match->revision, 471 XT_ALIGN(par->match->matchsize), size); 472 return -EINVAL; 473 } 474 if (par->match->table != NULL && 475 strcmp(par->match->table, par->table) != 0) { 476 pr_info_ratelimited("%s_tables: %s match: only valid in %s table, not %s\n", 477 xt_prefix[par->family], par->match->name, 478 par->match->table, par->table); 479 return -EINVAL; 480 } 481 if (par->match->hooks && (par->hook_mask & ~par->match->hooks) != 0) { 482 char used[64], allow[64]; 483 484 pr_info_ratelimited("%s_tables: %s match: used from hooks %s, but only valid from %s\n", 485 xt_prefix[par->family], par->match->name, 486 textify_hooks(used, sizeof(used), 487 par->hook_mask, par->family), 488 textify_hooks(allow, sizeof(allow), 489 par->match->hooks, 490 par->family)); 491 return -EINVAL; 492 } 493 if (par->match->proto && (par->match->proto != proto || inv_proto)) { 494 pr_info_ratelimited("%s_tables: %s match: only valid for protocol %u\n", 495 xt_prefix[par->family], par->match->name, 496 par->match->proto); 497 return -EINVAL; 498 } 499 if (par->match->checkentry != NULL) { 500 ret = par->match->checkentry(par); 501 if (ret < 0) 502 return ret; 503 else if (ret > 0) 504 /* Flag up potential errors. */ 505 return -EIO; 506 } 507 return 0; 508 } 509 EXPORT_SYMBOL_GPL(xt_check_match); 510 511 /** xt_check_entry_match - check that matches end before start of target 512 * 513 * @match: beginning of xt_entry_match 514 * @target: beginning of this rules target (alleged end of matches) 515 * @alignment: alignment requirement of match structures 516 * 517 * Validates that all matches add up to the beginning of the target, 518 * and that each match covers at least the base structure size. 519 * 520 * Return: 0 on success, negative errno on failure. 521 */ 522 static int xt_check_entry_match(const char *match, const char *target, 523 const size_t alignment) 524 { 525 const struct xt_entry_match *pos; 526 int length = target - match; 527 528 if (length == 0) /* no matches */ 529 return 0; 530 531 pos = (struct xt_entry_match *)match; 532 do { 533 if ((unsigned long)pos % alignment) 534 return -EINVAL; 535 536 if (length < (int)sizeof(struct xt_entry_match)) 537 return -EINVAL; 538 539 if (pos->u.match_size < sizeof(struct xt_entry_match)) 540 return -EINVAL; 541 542 if (pos->u.match_size > length) 543 return -EINVAL; 544 545 length -= pos->u.match_size; 546 pos = ((void *)((char *)(pos) + (pos)->u.match_size)); 547 } while (length > 0); 548 549 return 0; 550 } 551 552 /** xt_check_table_hooks - check hook entry points are sane 553 * 554 * @info xt_table_info to check 555 * @valid_hooks - hook entry points that we can enter from 556 * 557 * Validates that the hook entry and underflows points are set up. 558 * 559 * Return: 0 on success, negative errno on failure. 560 */ 561 int xt_check_table_hooks(const struct xt_table_info *info, unsigned int valid_hooks) 562 { 563 const char *err = "unsorted underflow"; 564 unsigned int i, max_uflow, max_entry; 565 bool check_hooks = false; 566 567 BUILD_BUG_ON(ARRAY_SIZE(info->hook_entry) != ARRAY_SIZE(info->underflow)); 568 569 max_entry = 0; 570 max_uflow = 0; 571 572 for (i = 0; i < ARRAY_SIZE(info->hook_entry); i++) { 573 if (!(valid_hooks & (1 << i))) 574 continue; 575 576 if (info->hook_entry[i] == 0xFFFFFFFF) 577 return -EINVAL; 578 if (info->underflow[i] == 0xFFFFFFFF) 579 return -EINVAL; 580 581 if (check_hooks) { 582 if (max_uflow > info->underflow[i]) 583 goto error; 584 585 if (max_uflow == info->underflow[i]) { 586 err = "duplicate underflow"; 587 goto error; 588 } 589 if (max_entry > info->hook_entry[i]) { 590 err = "unsorted entry"; 591 goto error; 592 } 593 if (max_entry == info->hook_entry[i]) { 594 err = "duplicate entry"; 595 goto error; 596 } 597 } 598 max_entry = info->hook_entry[i]; 599 max_uflow = info->underflow[i]; 600 check_hooks = true; 601 } 602 603 return 0; 604 error: 605 pr_err_ratelimited("%s at hook %d\n", err, i); 606 return -EINVAL; 607 } 608 EXPORT_SYMBOL(xt_check_table_hooks); 609 610 static bool verdict_ok(int verdict) 611 { 612 if (verdict > 0) 613 return true; 614 615 if (verdict < 0) { 616 int v = -verdict - 1; 617 618 if (verdict == XT_RETURN) 619 return true; 620 621 switch (v) { 622 case NF_ACCEPT: return true; 623 case NF_DROP: return true; 624 case NF_QUEUE: return true; 625 default: 626 break; 627 } 628 629 return false; 630 } 631 632 return false; 633 } 634 635 static bool error_tg_ok(unsigned int usersize, unsigned int kernsize, 636 const char *msg, unsigned int msglen) 637 { 638 return usersize == kernsize && strnlen(msg, msglen) < msglen; 639 } 640 641 #ifdef CONFIG_COMPAT 642 int xt_compat_add_offset(u_int8_t af, unsigned int offset, int delta) 643 { 644 struct xt_af *xp = &xt[af]; 645 646 WARN_ON(!mutex_is_locked(&xt[af].compat_mutex)); 647 648 if (WARN_ON(!xp->compat_tab)) 649 return -ENOMEM; 650 651 if (xp->cur >= xp->number) 652 return -EINVAL; 653 654 if (xp->cur) 655 delta += xp->compat_tab[xp->cur - 1].delta; 656 xp->compat_tab[xp->cur].offset = offset; 657 xp->compat_tab[xp->cur].delta = delta; 658 xp->cur++; 659 return 0; 660 } 661 EXPORT_SYMBOL_GPL(xt_compat_add_offset); 662 663 void xt_compat_flush_offsets(u_int8_t af) 664 { 665 WARN_ON(!mutex_is_locked(&xt[af].compat_mutex)); 666 667 if (xt[af].compat_tab) { 668 vfree(xt[af].compat_tab); 669 xt[af].compat_tab = NULL; 670 xt[af].number = 0; 671 xt[af].cur = 0; 672 } 673 } 674 EXPORT_SYMBOL_GPL(xt_compat_flush_offsets); 675 676 int xt_compat_calc_jump(u_int8_t af, unsigned int offset) 677 { 678 struct compat_delta *tmp = xt[af].compat_tab; 679 int mid, left = 0, right = xt[af].cur - 1; 680 681 while (left <= right) { 682 mid = (left + right) >> 1; 683 if (offset > tmp[mid].offset) 684 left = mid + 1; 685 else if (offset < tmp[mid].offset) 686 right = mid - 1; 687 else 688 return mid ? tmp[mid - 1].delta : 0; 689 } 690 return left ? tmp[left - 1].delta : 0; 691 } 692 EXPORT_SYMBOL_GPL(xt_compat_calc_jump); 693 694 int xt_compat_init_offsets(u8 af, unsigned int number) 695 { 696 size_t mem; 697 698 WARN_ON(!mutex_is_locked(&xt[af].compat_mutex)); 699 700 if (!number || number > (INT_MAX / sizeof(struct compat_delta))) 701 return -EINVAL; 702 703 if (WARN_ON(xt[af].compat_tab)) 704 return -EINVAL; 705 706 mem = sizeof(struct compat_delta) * number; 707 if (mem > XT_MAX_TABLE_SIZE) 708 return -ENOMEM; 709 710 xt[af].compat_tab = vmalloc(mem); 711 if (!xt[af].compat_tab) 712 return -ENOMEM; 713 714 xt[af].number = number; 715 xt[af].cur = 0; 716 717 return 0; 718 } 719 EXPORT_SYMBOL(xt_compat_init_offsets); 720 721 int xt_compat_match_offset(const struct xt_match *match) 722 { 723 u_int16_t csize = match->compatsize ? : match->matchsize; 724 return XT_ALIGN(match->matchsize) - COMPAT_XT_ALIGN(csize); 725 } 726 EXPORT_SYMBOL_GPL(xt_compat_match_offset); 727 728 void xt_compat_match_from_user(struct xt_entry_match *m, void **dstptr, 729 unsigned int *size) 730 { 731 const struct xt_match *match = m->u.kernel.match; 732 struct compat_xt_entry_match *cm = (struct compat_xt_entry_match *)m; 733 int pad, off = xt_compat_match_offset(match); 734 u_int16_t msize = cm->u.user.match_size; 735 char name[sizeof(m->u.user.name)]; 736 737 m = *dstptr; 738 memcpy(m, cm, sizeof(*cm)); 739 if (match->compat_from_user) 740 match->compat_from_user(m->data, cm->data); 741 else 742 memcpy(m->data, cm->data, msize - sizeof(*cm)); 743 pad = XT_ALIGN(match->matchsize) - match->matchsize; 744 if (pad > 0) 745 memset(m->data + match->matchsize, 0, pad); 746 747 msize += off; 748 m->u.user.match_size = msize; 749 strlcpy(name, match->name, sizeof(name)); 750 module_put(match->me); 751 strncpy(m->u.user.name, name, sizeof(m->u.user.name)); 752 753 *size += off; 754 *dstptr += msize; 755 } 756 EXPORT_SYMBOL_GPL(xt_compat_match_from_user); 757 758 #define COMPAT_XT_DATA_TO_USER(U, K, TYPE, C_SIZE) \ 759 xt_data_to_user(U->data, K->data, \ 760 K->u.kernel.TYPE->usersize, \ 761 C_SIZE, \ 762 COMPAT_XT_ALIGN(C_SIZE)) 763 764 int xt_compat_match_to_user(const struct xt_entry_match *m, 765 void __user **dstptr, unsigned int *size) 766 { 767 const struct xt_match *match = m->u.kernel.match; 768 struct compat_xt_entry_match __user *cm = *dstptr; 769 int off = xt_compat_match_offset(match); 770 u_int16_t msize = m->u.user.match_size - off; 771 772 if (XT_OBJ_TO_USER(cm, m, match, msize)) 773 return -EFAULT; 774 775 if (match->compat_to_user) { 776 if (match->compat_to_user((void __user *)cm->data, m->data)) 777 return -EFAULT; 778 } else { 779 if (COMPAT_XT_DATA_TO_USER(cm, m, match, msize - sizeof(*cm))) 780 return -EFAULT; 781 } 782 783 *size -= off; 784 *dstptr += msize; 785 return 0; 786 } 787 EXPORT_SYMBOL_GPL(xt_compat_match_to_user); 788 789 /* non-compat version may have padding after verdict */ 790 struct compat_xt_standard_target { 791 struct compat_xt_entry_target t; 792 compat_uint_t verdict; 793 }; 794 795 struct compat_xt_error_target { 796 struct compat_xt_entry_target t; 797 char errorname[XT_FUNCTION_MAXNAMELEN]; 798 }; 799 800 int xt_compat_check_entry_offsets(const void *base, const char *elems, 801 unsigned int target_offset, 802 unsigned int next_offset) 803 { 804 long size_of_base_struct = elems - (const char *)base; 805 const struct compat_xt_entry_target *t; 806 const char *e = base; 807 808 if (target_offset < size_of_base_struct) 809 return -EINVAL; 810 811 if (target_offset + sizeof(*t) > next_offset) 812 return -EINVAL; 813 814 t = (void *)(e + target_offset); 815 if (t->u.target_size < sizeof(*t)) 816 return -EINVAL; 817 818 if (target_offset + t->u.target_size > next_offset) 819 return -EINVAL; 820 821 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0) { 822 const struct compat_xt_standard_target *st = (const void *)t; 823 824 if (COMPAT_XT_ALIGN(target_offset + sizeof(*st)) != next_offset) 825 return -EINVAL; 826 827 if (!verdict_ok(st->verdict)) 828 return -EINVAL; 829 } else if (strcmp(t->u.user.name, XT_ERROR_TARGET) == 0) { 830 const struct compat_xt_error_target *et = (const void *)t; 831 832 if (!error_tg_ok(t->u.target_size, sizeof(*et), 833 et->errorname, sizeof(et->errorname))) 834 return -EINVAL; 835 } 836 837 /* compat_xt_entry match has less strict alignment requirements, 838 * otherwise they are identical. In case of padding differences 839 * we need to add compat version of xt_check_entry_match. 840 */ 841 BUILD_BUG_ON(sizeof(struct compat_xt_entry_match) != sizeof(struct xt_entry_match)); 842 843 return xt_check_entry_match(elems, base + target_offset, 844 __alignof__(struct compat_xt_entry_match)); 845 } 846 EXPORT_SYMBOL(xt_compat_check_entry_offsets); 847 #endif /* CONFIG_COMPAT */ 848 849 /** 850 * xt_check_entry_offsets - validate arp/ip/ip6t_entry 851 * 852 * @base: pointer to arp/ip/ip6t_entry 853 * @elems: pointer to first xt_entry_match, i.e. ip(6)t_entry->elems 854 * @target_offset: the arp/ip/ip6_t->target_offset 855 * @next_offset: the arp/ip/ip6_t->next_offset 856 * 857 * validates that target_offset and next_offset are sane and that all 858 * match sizes (if any) align with the target offset. 859 * 860 * This function does not validate the targets or matches themselves, it 861 * only tests that all the offsets and sizes are correct, that all 862 * match structures are aligned, and that the last structure ends where 863 * the target structure begins. 864 * 865 * Also see xt_compat_check_entry_offsets for CONFIG_COMPAT version. 866 * 867 * The arp/ip/ip6t_entry structure @base must have passed following tests: 868 * - it must point to a valid memory location 869 * - base to base + next_offset must be accessible, i.e. not exceed allocated 870 * length. 871 * 872 * A well-formed entry looks like this: 873 * 874 * ip(6)t_entry match [mtdata] match [mtdata] target [tgdata] ip(6)t_entry 875 * e->elems[]-----' | | 876 * matchsize | | 877 * matchsize | | 878 * | | 879 * target_offset---------------------------------' | 880 * next_offset---------------------------------------------------' 881 * 882 * elems[]: flexible array member at end of ip(6)/arpt_entry struct. 883 * This is where matches (if any) and the target reside. 884 * target_offset: beginning of target. 885 * next_offset: start of the next rule; also: size of this rule. 886 * Since targets have a minimum size, target_offset + minlen <= next_offset. 887 * 888 * Every match stores its size, sum of sizes must not exceed target_offset. 889 * 890 * Return: 0 on success, negative errno on failure. 891 */ 892 int xt_check_entry_offsets(const void *base, 893 const char *elems, 894 unsigned int target_offset, 895 unsigned int next_offset) 896 { 897 long size_of_base_struct = elems - (const char *)base; 898 const struct xt_entry_target *t; 899 const char *e = base; 900 901 /* target start is within the ip/ip6/arpt_entry struct */ 902 if (target_offset < size_of_base_struct) 903 return -EINVAL; 904 905 if (target_offset + sizeof(*t) > next_offset) 906 return -EINVAL; 907 908 t = (void *)(e + target_offset); 909 if (t->u.target_size < sizeof(*t)) 910 return -EINVAL; 911 912 if (target_offset + t->u.target_size > next_offset) 913 return -EINVAL; 914 915 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0) { 916 const struct xt_standard_target *st = (const void *)t; 917 918 if (XT_ALIGN(target_offset + sizeof(*st)) != next_offset) 919 return -EINVAL; 920 921 if (!verdict_ok(st->verdict)) 922 return -EINVAL; 923 } else if (strcmp(t->u.user.name, XT_ERROR_TARGET) == 0) { 924 const struct xt_error_target *et = (const void *)t; 925 926 if (!error_tg_ok(t->u.target_size, sizeof(*et), 927 et->errorname, sizeof(et->errorname))) 928 return -EINVAL; 929 } 930 931 return xt_check_entry_match(elems, base + target_offset, 932 __alignof__(struct xt_entry_match)); 933 } 934 EXPORT_SYMBOL(xt_check_entry_offsets); 935 936 /** 937 * xt_alloc_entry_offsets - allocate array to store rule head offsets 938 * 939 * @size: number of entries 940 * 941 * Return: NULL or kmalloc'd or vmalloc'd array 942 */ 943 unsigned int *xt_alloc_entry_offsets(unsigned int size) 944 { 945 if (size > XT_MAX_TABLE_SIZE / sizeof(unsigned int)) 946 return NULL; 947 948 return kvmalloc_array(size, sizeof(unsigned int), GFP_KERNEL | __GFP_ZERO); 949 950 } 951 EXPORT_SYMBOL(xt_alloc_entry_offsets); 952 953 /** 954 * xt_find_jump_offset - check if target is a valid jump offset 955 * 956 * @offsets: array containing all valid rule start offsets of a rule blob 957 * @target: the jump target to search for 958 * @size: entries in @offset 959 */ 960 bool xt_find_jump_offset(const unsigned int *offsets, 961 unsigned int target, unsigned int size) 962 { 963 int m, low = 0, hi = size; 964 965 while (hi > low) { 966 m = (low + hi) / 2u; 967 968 if (offsets[m] > target) 969 hi = m; 970 else if (offsets[m] < target) 971 low = m + 1; 972 else 973 return true; 974 } 975 976 return false; 977 } 978 EXPORT_SYMBOL(xt_find_jump_offset); 979 980 int xt_check_target(struct xt_tgchk_param *par, 981 unsigned int size, u_int8_t proto, bool inv_proto) 982 { 983 int ret; 984 985 if (XT_ALIGN(par->target->targetsize) != size) { 986 pr_err_ratelimited("%s_tables: %s.%u target: invalid size %u (kernel) != (user) %u\n", 987 xt_prefix[par->family], par->target->name, 988 par->target->revision, 989 XT_ALIGN(par->target->targetsize), size); 990 return -EINVAL; 991 } 992 if (par->target->table != NULL && 993 strcmp(par->target->table, par->table) != 0) { 994 pr_info_ratelimited("%s_tables: %s target: only valid in %s table, not %s\n", 995 xt_prefix[par->family], par->target->name, 996 par->target->table, par->table); 997 return -EINVAL; 998 } 999 if (par->target->hooks && (par->hook_mask & ~par->target->hooks) != 0) { 1000 char used[64], allow[64]; 1001 1002 pr_info_ratelimited("%s_tables: %s target: used from hooks %s, but only usable from %s\n", 1003 xt_prefix[par->family], par->target->name, 1004 textify_hooks(used, sizeof(used), 1005 par->hook_mask, par->family), 1006 textify_hooks(allow, sizeof(allow), 1007 par->target->hooks, 1008 par->family)); 1009 return -EINVAL; 1010 } 1011 if (par->target->proto && (par->target->proto != proto || inv_proto)) { 1012 pr_info_ratelimited("%s_tables: %s target: only valid for protocol %u\n", 1013 xt_prefix[par->family], par->target->name, 1014 par->target->proto); 1015 return -EINVAL; 1016 } 1017 if (par->target->checkentry != NULL) { 1018 ret = par->target->checkentry(par); 1019 if (ret < 0) 1020 return ret; 1021 else if (ret > 0) 1022 /* Flag up potential errors. */ 1023 return -EIO; 1024 } 1025 return 0; 1026 } 1027 EXPORT_SYMBOL_GPL(xt_check_target); 1028 1029 /** 1030 * xt_copy_counters_from_user - copy counters and metadata from userspace 1031 * 1032 * @user: src pointer to userspace memory 1033 * @len: alleged size of userspace memory 1034 * @info: where to store the xt_counters_info metadata 1035 * @compat: true if we setsockopt call is done by 32bit task on 64bit kernel 1036 * 1037 * Copies counter meta data from @user and stores it in @info. 1038 * 1039 * vmallocs memory to hold the counters, then copies the counter data 1040 * from @user to the new memory and returns a pointer to it. 1041 * 1042 * If @compat is true, @info gets converted automatically to the 64bit 1043 * representation. 1044 * 1045 * The metadata associated with the counters is stored in @info. 1046 * 1047 * Return: returns pointer that caller has to test via IS_ERR(). 1048 * If IS_ERR is false, caller has to vfree the pointer. 1049 */ 1050 void *xt_copy_counters_from_user(const void __user *user, unsigned int len, 1051 struct xt_counters_info *info, bool compat) 1052 { 1053 void *mem; 1054 u64 size; 1055 1056 #ifdef CONFIG_COMPAT 1057 if (compat) { 1058 /* structures only differ in size due to alignment */ 1059 struct compat_xt_counters_info compat_tmp; 1060 1061 if (len <= sizeof(compat_tmp)) 1062 return ERR_PTR(-EINVAL); 1063 1064 len -= sizeof(compat_tmp); 1065 if (copy_from_user(&compat_tmp, user, sizeof(compat_tmp)) != 0) 1066 return ERR_PTR(-EFAULT); 1067 1068 memcpy(info->name, compat_tmp.name, sizeof(info->name) - 1); 1069 info->num_counters = compat_tmp.num_counters; 1070 user += sizeof(compat_tmp); 1071 } else 1072 #endif 1073 { 1074 if (len <= sizeof(*info)) 1075 return ERR_PTR(-EINVAL); 1076 1077 len -= sizeof(*info); 1078 if (copy_from_user(info, user, sizeof(*info)) != 0) 1079 return ERR_PTR(-EFAULT); 1080 1081 user += sizeof(*info); 1082 } 1083 info->name[sizeof(info->name) - 1] = '\0'; 1084 1085 size = sizeof(struct xt_counters); 1086 size *= info->num_counters; 1087 1088 if (size != (u64)len) 1089 return ERR_PTR(-EINVAL); 1090 1091 mem = vmalloc(len); 1092 if (!mem) 1093 return ERR_PTR(-ENOMEM); 1094 1095 if (copy_from_user(mem, user, len) == 0) 1096 return mem; 1097 1098 vfree(mem); 1099 return ERR_PTR(-EFAULT); 1100 } 1101 EXPORT_SYMBOL_GPL(xt_copy_counters_from_user); 1102 1103 #ifdef CONFIG_COMPAT 1104 int xt_compat_target_offset(const struct xt_target *target) 1105 { 1106 u_int16_t csize = target->compatsize ? : target->targetsize; 1107 return XT_ALIGN(target->targetsize) - COMPAT_XT_ALIGN(csize); 1108 } 1109 EXPORT_SYMBOL_GPL(xt_compat_target_offset); 1110 1111 void xt_compat_target_from_user(struct xt_entry_target *t, void **dstptr, 1112 unsigned int *size) 1113 { 1114 const struct xt_target *target = t->u.kernel.target; 1115 struct compat_xt_entry_target *ct = (struct compat_xt_entry_target *)t; 1116 int pad, off = xt_compat_target_offset(target); 1117 u_int16_t tsize = ct->u.user.target_size; 1118 char name[sizeof(t->u.user.name)]; 1119 1120 t = *dstptr; 1121 memcpy(t, ct, sizeof(*ct)); 1122 if (target->compat_from_user) 1123 target->compat_from_user(t->data, ct->data); 1124 else 1125 memcpy(t->data, ct->data, tsize - sizeof(*ct)); 1126 pad = XT_ALIGN(target->targetsize) - target->targetsize; 1127 if (pad > 0) 1128 memset(t->data + target->targetsize, 0, pad); 1129 1130 tsize += off; 1131 t->u.user.target_size = tsize; 1132 strlcpy(name, target->name, sizeof(name)); 1133 module_put(target->me); 1134 strncpy(t->u.user.name, name, sizeof(t->u.user.name)); 1135 1136 *size += off; 1137 *dstptr += tsize; 1138 } 1139 EXPORT_SYMBOL_GPL(xt_compat_target_from_user); 1140 1141 int xt_compat_target_to_user(const struct xt_entry_target *t, 1142 void __user **dstptr, unsigned int *size) 1143 { 1144 const struct xt_target *target = t->u.kernel.target; 1145 struct compat_xt_entry_target __user *ct = *dstptr; 1146 int off = xt_compat_target_offset(target); 1147 u_int16_t tsize = t->u.user.target_size - off; 1148 1149 if (XT_OBJ_TO_USER(ct, t, target, tsize)) 1150 return -EFAULT; 1151 1152 if (target->compat_to_user) { 1153 if (target->compat_to_user((void __user *)ct->data, t->data)) 1154 return -EFAULT; 1155 } else { 1156 if (COMPAT_XT_DATA_TO_USER(ct, t, target, tsize - sizeof(*ct))) 1157 return -EFAULT; 1158 } 1159 1160 *size -= off; 1161 *dstptr += tsize; 1162 return 0; 1163 } 1164 EXPORT_SYMBOL_GPL(xt_compat_target_to_user); 1165 #endif 1166 1167 struct xt_table_info *xt_alloc_table_info(unsigned int size) 1168 { 1169 struct xt_table_info *info = NULL; 1170 size_t sz = sizeof(*info) + size; 1171 1172 if (sz < sizeof(*info) || sz >= XT_MAX_TABLE_SIZE) 1173 return NULL; 1174 1175 /* __GFP_NORETRY is not fully supported by kvmalloc but it should 1176 * work reasonably well if sz is too large and bail out rather 1177 * than shoot all processes down before realizing there is nothing 1178 * more to reclaim. 1179 */ 1180 info = kvmalloc(sz, GFP_KERNEL | __GFP_NORETRY); 1181 if (!info) 1182 return NULL; 1183 1184 memset(info, 0, sizeof(*info)); 1185 info->size = size; 1186 return info; 1187 } 1188 EXPORT_SYMBOL(xt_alloc_table_info); 1189 1190 void xt_free_table_info(struct xt_table_info *info) 1191 { 1192 int cpu; 1193 1194 if (info->jumpstack != NULL) { 1195 for_each_possible_cpu(cpu) 1196 kvfree(info->jumpstack[cpu]); 1197 kvfree(info->jumpstack); 1198 } 1199 1200 kvfree(info); 1201 } 1202 EXPORT_SYMBOL(xt_free_table_info); 1203 1204 /* Find table by name, grabs mutex & ref. Returns ERR_PTR on error. */ 1205 struct xt_table *xt_find_table_lock(struct net *net, u_int8_t af, 1206 const char *name) 1207 { 1208 struct xt_table *t, *found = NULL; 1209 1210 mutex_lock(&xt[af].mutex); 1211 list_for_each_entry(t, &net->xt.tables[af], list) 1212 if (strcmp(t->name, name) == 0 && try_module_get(t->me)) 1213 return t; 1214 1215 if (net == &init_net) 1216 goto out; 1217 1218 /* Table doesn't exist in this netns, re-try init */ 1219 list_for_each_entry(t, &init_net.xt.tables[af], list) { 1220 int err; 1221 1222 if (strcmp(t->name, name)) 1223 continue; 1224 if (!try_module_get(t->me)) 1225 goto out; 1226 mutex_unlock(&xt[af].mutex); 1227 err = t->table_init(net); 1228 if (err < 0) { 1229 module_put(t->me); 1230 return ERR_PTR(err); 1231 } 1232 1233 found = t; 1234 1235 mutex_lock(&xt[af].mutex); 1236 break; 1237 } 1238 1239 if (!found) 1240 goto out; 1241 1242 /* and once again: */ 1243 list_for_each_entry(t, &net->xt.tables[af], list) 1244 if (strcmp(t->name, name) == 0) 1245 return t; 1246 1247 module_put(found->me); 1248 out: 1249 mutex_unlock(&xt[af].mutex); 1250 return ERR_PTR(-ENOENT); 1251 } 1252 EXPORT_SYMBOL_GPL(xt_find_table_lock); 1253 1254 struct xt_table *xt_request_find_table_lock(struct net *net, u_int8_t af, 1255 const char *name) 1256 { 1257 struct xt_table *t = xt_find_table_lock(net, af, name); 1258 1259 #ifdef CONFIG_MODULES 1260 if (IS_ERR(t)) { 1261 int err = request_module("%stable_%s", xt_prefix[af], name); 1262 if (err < 0) 1263 return ERR_PTR(err); 1264 t = xt_find_table_lock(net, af, name); 1265 } 1266 #endif 1267 1268 return t; 1269 } 1270 EXPORT_SYMBOL_GPL(xt_request_find_table_lock); 1271 1272 void xt_table_unlock(struct xt_table *table) 1273 { 1274 mutex_unlock(&xt[table->af].mutex); 1275 } 1276 EXPORT_SYMBOL_GPL(xt_table_unlock); 1277 1278 #ifdef CONFIG_COMPAT 1279 void xt_compat_lock(u_int8_t af) 1280 { 1281 mutex_lock(&xt[af].compat_mutex); 1282 } 1283 EXPORT_SYMBOL_GPL(xt_compat_lock); 1284 1285 void xt_compat_unlock(u_int8_t af) 1286 { 1287 mutex_unlock(&xt[af].compat_mutex); 1288 } 1289 EXPORT_SYMBOL_GPL(xt_compat_unlock); 1290 #endif 1291 1292 DEFINE_PER_CPU(seqcount_t, xt_recseq); 1293 EXPORT_PER_CPU_SYMBOL_GPL(xt_recseq); 1294 1295 struct static_key xt_tee_enabled __read_mostly; 1296 EXPORT_SYMBOL_GPL(xt_tee_enabled); 1297 1298 static int xt_jumpstack_alloc(struct xt_table_info *i) 1299 { 1300 unsigned int size; 1301 int cpu; 1302 1303 size = sizeof(void **) * nr_cpu_ids; 1304 if (size > PAGE_SIZE) 1305 i->jumpstack = kvzalloc(size, GFP_KERNEL); 1306 else 1307 i->jumpstack = kzalloc(size, GFP_KERNEL); 1308 if (i->jumpstack == NULL) 1309 return -ENOMEM; 1310 1311 /* ruleset without jumps -- no stack needed */ 1312 if (i->stacksize == 0) 1313 return 0; 1314 1315 /* Jumpstack needs to be able to record two full callchains, one 1316 * from the first rule set traversal, plus one table reentrancy 1317 * via -j TEE without clobbering the callchain that brought us to 1318 * TEE target. 1319 * 1320 * This is done by allocating two jumpstacks per cpu, on reentry 1321 * the upper half of the stack is used. 1322 * 1323 * see the jumpstack setup in ipt_do_table() for more details. 1324 */ 1325 size = sizeof(void *) * i->stacksize * 2u; 1326 for_each_possible_cpu(cpu) { 1327 i->jumpstack[cpu] = kvmalloc_node(size, GFP_KERNEL, 1328 cpu_to_node(cpu)); 1329 if (i->jumpstack[cpu] == NULL) 1330 /* 1331 * Freeing will be done later on by the callers. The 1332 * chain is: xt_replace_table -> __do_replace -> 1333 * do_replace -> xt_free_table_info. 1334 */ 1335 return -ENOMEM; 1336 } 1337 1338 return 0; 1339 } 1340 1341 struct xt_counters *xt_counters_alloc(unsigned int counters) 1342 { 1343 struct xt_counters *mem; 1344 1345 if (counters == 0 || counters > INT_MAX / sizeof(*mem)) 1346 return NULL; 1347 1348 counters *= sizeof(*mem); 1349 if (counters > XT_MAX_TABLE_SIZE) 1350 return NULL; 1351 1352 return vzalloc(counters); 1353 } 1354 EXPORT_SYMBOL(xt_counters_alloc); 1355 1356 struct xt_table_info * 1357 xt_replace_table(struct xt_table *table, 1358 unsigned int num_counters, 1359 struct xt_table_info *newinfo, 1360 int *error) 1361 { 1362 struct xt_table_info *private; 1363 unsigned int cpu; 1364 int ret; 1365 1366 ret = xt_jumpstack_alloc(newinfo); 1367 if (ret < 0) { 1368 *error = ret; 1369 return NULL; 1370 } 1371 1372 /* Do the substitution. */ 1373 local_bh_disable(); 1374 private = table->private; 1375 1376 /* Check inside lock: is the old number correct? */ 1377 if (num_counters != private->number) { 1378 pr_debug("num_counters != table->private->number (%u/%u)\n", 1379 num_counters, private->number); 1380 local_bh_enable(); 1381 *error = -EAGAIN; 1382 return NULL; 1383 } 1384 1385 newinfo->initial_entries = private->initial_entries; 1386 /* 1387 * Ensure contents of newinfo are visible before assigning to 1388 * private. 1389 */ 1390 smp_wmb(); 1391 table->private = newinfo; 1392 1393 /* make sure all cpus see new ->private value */ 1394 smp_wmb(); 1395 1396 /* 1397 * Even though table entries have now been swapped, other CPU's 1398 * may still be using the old entries... 1399 */ 1400 local_bh_enable(); 1401 1402 /* ... so wait for even xt_recseq on all cpus */ 1403 for_each_possible_cpu(cpu) { 1404 seqcount_t *s = &per_cpu(xt_recseq, cpu); 1405 u32 seq = raw_read_seqcount(s); 1406 1407 if (seq & 1) { 1408 do { 1409 cond_resched(); 1410 cpu_relax(); 1411 } while (seq == raw_read_seqcount(s)); 1412 } 1413 } 1414 1415 #ifdef CONFIG_AUDIT 1416 if (audit_enabled) { 1417 audit_log(current->audit_context, GFP_KERNEL, 1418 AUDIT_NETFILTER_CFG, 1419 "table=%s family=%u entries=%u", 1420 table->name, table->af, private->number); 1421 } 1422 #endif 1423 1424 return private; 1425 } 1426 EXPORT_SYMBOL_GPL(xt_replace_table); 1427 1428 struct xt_table *xt_register_table(struct net *net, 1429 const struct xt_table *input_table, 1430 struct xt_table_info *bootstrap, 1431 struct xt_table_info *newinfo) 1432 { 1433 int ret; 1434 struct xt_table_info *private; 1435 struct xt_table *t, *table; 1436 1437 /* Don't add one object to multiple lists. */ 1438 table = kmemdup(input_table, sizeof(struct xt_table), GFP_KERNEL); 1439 if (!table) { 1440 ret = -ENOMEM; 1441 goto out; 1442 } 1443 1444 mutex_lock(&xt[table->af].mutex); 1445 /* Don't autoload: we'd eat our tail... */ 1446 list_for_each_entry(t, &net->xt.tables[table->af], list) { 1447 if (strcmp(t->name, table->name) == 0) { 1448 ret = -EEXIST; 1449 goto unlock; 1450 } 1451 } 1452 1453 /* Simplifies replace_table code. */ 1454 table->private = bootstrap; 1455 1456 if (!xt_replace_table(table, 0, newinfo, &ret)) 1457 goto unlock; 1458 1459 private = table->private; 1460 pr_debug("table->private->number = %u\n", private->number); 1461 1462 /* save number of initial entries */ 1463 private->initial_entries = private->number; 1464 1465 list_add(&table->list, &net->xt.tables[table->af]); 1466 mutex_unlock(&xt[table->af].mutex); 1467 return table; 1468 1469 unlock: 1470 mutex_unlock(&xt[table->af].mutex); 1471 kfree(table); 1472 out: 1473 return ERR_PTR(ret); 1474 } 1475 EXPORT_SYMBOL_GPL(xt_register_table); 1476 1477 void *xt_unregister_table(struct xt_table *table) 1478 { 1479 struct xt_table_info *private; 1480 1481 mutex_lock(&xt[table->af].mutex); 1482 private = table->private; 1483 list_del(&table->list); 1484 mutex_unlock(&xt[table->af].mutex); 1485 kfree(table); 1486 1487 return private; 1488 } 1489 EXPORT_SYMBOL_GPL(xt_unregister_table); 1490 1491 #ifdef CONFIG_PROC_FS 1492 struct xt_names_priv { 1493 struct seq_net_private p; 1494 u_int8_t af; 1495 }; 1496 static void *xt_table_seq_start(struct seq_file *seq, loff_t *pos) 1497 { 1498 struct xt_names_priv *priv = seq->private; 1499 struct net *net = seq_file_net(seq); 1500 u_int8_t af = priv->af; 1501 1502 mutex_lock(&xt[af].mutex); 1503 return seq_list_start(&net->xt.tables[af], *pos); 1504 } 1505 1506 static void *xt_table_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1507 { 1508 struct xt_names_priv *priv = seq->private; 1509 struct net *net = seq_file_net(seq); 1510 u_int8_t af = priv->af; 1511 1512 return seq_list_next(v, &net->xt.tables[af], pos); 1513 } 1514 1515 static void xt_table_seq_stop(struct seq_file *seq, void *v) 1516 { 1517 struct xt_names_priv *priv = seq->private; 1518 u_int8_t af = priv->af; 1519 1520 mutex_unlock(&xt[af].mutex); 1521 } 1522 1523 static int xt_table_seq_show(struct seq_file *seq, void *v) 1524 { 1525 struct xt_table *table = list_entry(v, struct xt_table, list); 1526 1527 if (*table->name) 1528 seq_printf(seq, "%s\n", table->name); 1529 return 0; 1530 } 1531 1532 static const struct seq_operations xt_table_seq_ops = { 1533 .start = xt_table_seq_start, 1534 .next = xt_table_seq_next, 1535 .stop = xt_table_seq_stop, 1536 .show = xt_table_seq_show, 1537 }; 1538 1539 static int xt_table_open(struct inode *inode, struct file *file) 1540 { 1541 int ret; 1542 struct xt_names_priv *priv; 1543 1544 ret = seq_open_net(inode, file, &xt_table_seq_ops, 1545 sizeof(struct xt_names_priv)); 1546 if (!ret) { 1547 priv = ((struct seq_file *)file->private_data)->private; 1548 priv->af = (unsigned long)PDE_DATA(inode); 1549 } 1550 return ret; 1551 } 1552 1553 static const struct file_operations xt_table_ops = { 1554 .open = xt_table_open, 1555 .read = seq_read, 1556 .llseek = seq_lseek, 1557 .release = seq_release_net, 1558 }; 1559 1560 /* 1561 * Traverse state for ip{,6}_{tables,matches} for helping crossing 1562 * the multi-AF mutexes. 1563 */ 1564 struct nf_mttg_trav { 1565 struct list_head *head, *curr; 1566 uint8_t class, nfproto; 1567 }; 1568 1569 enum { 1570 MTTG_TRAV_INIT, 1571 MTTG_TRAV_NFP_UNSPEC, 1572 MTTG_TRAV_NFP_SPEC, 1573 MTTG_TRAV_DONE, 1574 }; 1575 1576 static void *xt_mttg_seq_next(struct seq_file *seq, void *v, loff_t *ppos, 1577 bool is_target) 1578 { 1579 static const uint8_t next_class[] = { 1580 [MTTG_TRAV_NFP_UNSPEC] = MTTG_TRAV_NFP_SPEC, 1581 [MTTG_TRAV_NFP_SPEC] = MTTG_TRAV_DONE, 1582 }; 1583 struct nf_mttg_trav *trav = seq->private; 1584 1585 switch (trav->class) { 1586 case MTTG_TRAV_INIT: 1587 trav->class = MTTG_TRAV_NFP_UNSPEC; 1588 mutex_lock(&xt[NFPROTO_UNSPEC].mutex); 1589 trav->head = trav->curr = is_target ? 1590 &xt[NFPROTO_UNSPEC].target : &xt[NFPROTO_UNSPEC].match; 1591 break; 1592 case MTTG_TRAV_NFP_UNSPEC: 1593 trav->curr = trav->curr->next; 1594 if (trav->curr != trav->head) 1595 break; 1596 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex); 1597 mutex_lock(&xt[trav->nfproto].mutex); 1598 trav->head = trav->curr = is_target ? 1599 &xt[trav->nfproto].target : &xt[trav->nfproto].match; 1600 trav->class = next_class[trav->class]; 1601 break; 1602 case MTTG_TRAV_NFP_SPEC: 1603 trav->curr = trav->curr->next; 1604 if (trav->curr != trav->head) 1605 break; 1606 /* fall through */ 1607 default: 1608 return NULL; 1609 } 1610 1611 if (ppos != NULL) 1612 ++*ppos; 1613 return trav; 1614 } 1615 1616 static void *xt_mttg_seq_start(struct seq_file *seq, loff_t *pos, 1617 bool is_target) 1618 { 1619 struct nf_mttg_trav *trav = seq->private; 1620 unsigned int j; 1621 1622 trav->class = MTTG_TRAV_INIT; 1623 for (j = 0; j < *pos; ++j) 1624 if (xt_mttg_seq_next(seq, NULL, NULL, is_target) == NULL) 1625 return NULL; 1626 return trav; 1627 } 1628 1629 static void xt_mttg_seq_stop(struct seq_file *seq, void *v) 1630 { 1631 struct nf_mttg_trav *trav = seq->private; 1632 1633 switch (trav->class) { 1634 case MTTG_TRAV_NFP_UNSPEC: 1635 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex); 1636 break; 1637 case MTTG_TRAV_NFP_SPEC: 1638 mutex_unlock(&xt[trav->nfproto].mutex); 1639 break; 1640 } 1641 } 1642 1643 static void *xt_match_seq_start(struct seq_file *seq, loff_t *pos) 1644 { 1645 return xt_mttg_seq_start(seq, pos, false); 1646 } 1647 1648 static void *xt_match_seq_next(struct seq_file *seq, void *v, loff_t *ppos) 1649 { 1650 return xt_mttg_seq_next(seq, v, ppos, false); 1651 } 1652 1653 static int xt_match_seq_show(struct seq_file *seq, void *v) 1654 { 1655 const struct nf_mttg_trav *trav = seq->private; 1656 const struct xt_match *match; 1657 1658 switch (trav->class) { 1659 case MTTG_TRAV_NFP_UNSPEC: 1660 case MTTG_TRAV_NFP_SPEC: 1661 if (trav->curr == trav->head) 1662 return 0; 1663 match = list_entry(trav->curr, struct xt_match, list); 1664 if (*match->name) 1665 seq_printf(seq, "%s\n", match->name); 1666 } 1667 return 0; 1668 } 1669 1670 static const struct seq_operations xt_match_seq_ops = { 1671 .start = xt_match_seq_start, 1672 .next = xt_match_seq_next, 1673 .stop = xt_mttg_seq_stop, 1674 .show = xt_match_seq_show, 1675 }; 1676 1677 static int xt_match_open(struct inode *inode, struct file *file) 1678 { 1679 struct nf_mttg_trav *trav; 1680 trav = __seq_open_private(file, &xt_match_seq_ops, sizeof(*trav)); 1681 if (!trav) 1682 return -ENOMEM; 1683 1684 trav->nfproto = (unsigned long)PDE_DATA(inode); 1685 return 0; 1686 } 1687 1688 static const struct file_operations xt_match_ops = { 1689 .open = xt_match_open, 1690 .read = seq_read, 1691 .llseek = seq_lseek, 1692 .release = seq_release_private, 1693 }; 1694 1695 static void *xt_target_seq_start(struct seq_file *seq, loff_t *pos) 1696 { 1697 return xt_mttg_seq_start(seq, pos, true); 1698 } 1699 1700 static void *xt_target_seq_next(struct seq_file *seq, void *v, loff_t *ppos) 1701 { 1702 return xt_mttg_seq_next(seq, v, ppos, true); 1703 } 1704 1705 static int xt_target_seq_show(struct seq_file *seq, void *v) 1706 { 1707 const struct nf_mttg_trav *trav = seq->private; 1708 const struct xt_target *target; 1709 1710 switch (trav->class) { 1711 case MTTG_TRAV_NFP_UNSPEC: 1712 case MTTG_TRAV_NFP_SPEC: 1713 if (trav->curr == trav->head) 1714 return 0; 1715 target = list_entry(trav->curr, struct xt_target, list); 1716 if (*target->name) 1717 seq_printf(seq, "%s\n", target->name); 1718 } 1719 return 0; 1720 } 1721 1722 static const struct seq_operations xt_target_seq_ops = { 1723 .start = xt_target_seq_start, 1724 .next = xt_target_seq_next, 1725 .stop = xt_mttg_seq_stop, 1726 .show = xt_target_seq_show, 1727 }; 1728 1729 static int xt_target_open(struct inode *inode, struct file *file) 1730 { 1731 struct nf_mttg_trav *trav; 1732 trav = __seq_open_private(file, &xt_target_seq_ops, sizeof(*trav)); 1733 if (!trav) 1734 return -ENOMEM; 1735 1736 trav->nfproto = (unsigned long)PDE_DATA(inode); 1737 return 0; 1738 } 1739 1740 static const struct file_operations xt_target_ops = { 1741 .open = xt_target_open, 1742 .read = seq_read, 1743 .llseek = seq_lseek, 1744 .release = seq_release_private, 1745 }; 1746 1747 #define FORMAT_TABLES "_tables_names" 1748 #define FORMAT_MATCHES "_tables_matches" 1749 #define FORMAT_TARGETS "_tables_targets" 1750 1751 #endif /* CONFIG_PROC_FS */ 1752 1753 /** 1754 * xt_hook_ops_alloc - set up hooks for a new table 1755 * @table: table with metadata needed to set up hooks 1756 * @fn: Hook function 1757 * 1758 * This function will create the nf_hook_ops that the x_table needs 1759 * to hand to xt_hook_link_net(). 1760 */ 1761 struct nf_hook_ops * 1762 xt_hook_ops_alloc(const struct xt_table *table, nf_hookfn *fn) 1763 { 1764 unsigned int hook_mask = table->valid_hooks; 1765 uint8_t i, num_hooks = hweight32(hook_mask); 1766 uint8_t hooknum; 1767 struct nf_hook_ops *ops; 1768 1769 if (!num_hooks) 1770 return ERR_PTR(-EINVAL); 1771 1772 ops = kcalloc(num_hooks, sizeof(*ops), GFP_KERNEL); 1773 if (ops == NULL) 1774 return ERR_PTR(-ENOMEM); 1775 1776 for (i = 0, hooknum = 0; i < num_hooks && hook_mask != 0; 1777 hook_mask >>= 1, ++hooknum) { 1778 if (!(hook_mask & 1)) 1779 continue; 1780 ops[i].hook = fn; 1781 ops[i].pf = table->af; 1782 ops[i].hooknum = hooknum; 1783 ops[i].priority = table->priority; 1784 ++i; 1785 } 1786 1787 return ops; 1788 } 1789 EXPORT_SYMBOL_GPL(xt_hook_ops_alloc); 1790 1791 int xt_proto_init(struct net *net, u_int8_t af) 1792 { 1793 #ifdef CONFIG_PROC_FS 1794 char buf[XT_FUNCTION_MAXNAMELEN]; 1795 struct proc_dir_entry *proc; 1796 kuid_t root_uid; 1797 kgid_t root_gid; 1798 #endif 1799 1800 if (af >= ARRAY_SIZE(xt_prefix)) 1801 return -EINVAL; 1802 1803 1804 #ifdef CONFIG_PROC_FS 1805 root_uid = make_kuid(net->user_ns, 0); 1806 root_gid = make_kgid(net->user_ns, 0); 1807 1808 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1809 strlcat(buf, FORMAT_TABLES, sizeof(buf)); 1810 proc = proc_create_data(buf, 0440, net->proc_net, &xt_table_ops, 1811 (void *)(unsigned long)af); 1812 if (!proc) 1813 goto out; 1814 if (uid_valid(root_uid) && gid_valid(root_gid)) 1815 proc_set_user(proc, root_uid, root_gid); 1816 1817 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1818 strlcat(buf, FORMAT_MATCHES, sizeof(buf)); 1819 proc = proc_create_data(buf, 0440, net->proc_net, &xt_match_ops, 1820 (void *)(unsigned long)af); 1821 if (!proc) 1822 goto out_remove_tables; 1823 if (uid_valid(root_uid) && gid_valid(root_gid)) 1824 proc_set_user(proc, root_uid, root_gid); 1825 1826 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1827 strlcat(buf, FORMAT_TARGETS, sizeof(buf)); 1828 proc = proc_create_data(buf, 0440, net->proc_net, &xt_target_ops, 1829 (void *)(unsigned long)af); 1830 if (!proc) 1831 goto out_remove_matches; 1832 if (uid_valid(root_uid) && gid_valid(root_gid)) 1833 proc_set_user(proc, root_uid, root_gid); 1834 #endif 1835 1836 return 0; 1837 1838 #ifdef CONFIG_PROC_FS 1839 out_remove_matches: 1840 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1841 strlcat(buf, FORMAT_MATCHES, sizeof(buf)); 1842 remove_proc_entry(buf, net->proc_net); 1843 1844 out_remove_tables: 1845 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1846 strlcat(buf, FORMAT_TABLES, sizeof(buf)); 1847 remove_proc_entry(buf, net->proc_net); 1848 out: 1849 return -1; 1850 #endif 1851 } 1852 EXPORT_SYMBOL_GPL(xt_proto_init); 1853 1854 void xt_proto_fini(struct net *net, u_int8_t af) 1855 { 1856 #ifdef CONFIG_PROC_FS 1857 char buf[XT_FUNCTION_MAXNAMELEN]; 1858 1859 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1860 strlcat(buf, FORMAT_TABLES, sizeof(buf)); 1861 remove_proc_entry(buf, net->proc_net); 1862 1863 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1864 strlcat(buf, FORMAT_TARGETS, sizeof(buf)); 1865 remove_proc_entry(buf, net->proc_net); 1866 1867 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1868 strlcat(buf, FORMAT_MATCHES, sizeof(buf)); 1869 remove_proc_entry(buf, net->proc_net); 1870 #endif /*CONFIG_PROC_FS*/ 1871 } 1872 EXPORT_SYMBOL_GPL(xt_proto_fini); 1873 1874 /** 1875 * xt_percpu_counter_alloc - allocate x_tables rule counter 1876 * 1877 * @state: pointer to xt_percpu allocation state 1878 * @counter: pointer to counter struct inside the ip(6)/arpt_entry struct 1879 * 1880 * On SMP, the packet counter [ ip(6)t_entry->counters.pcnt ] will then 1881 * contain the address of the real (percpu) counter. 1882 * 1883 * Rule evaluation needs to use xt_get_this_cpu_counter() helper 1884 * to fetch the real percpu counter. 1885 * 1886 * To speed up allocation and improve data locality, a 4kb block is 1887 * allocated. Freeing any counter may free an entire block, so all 1888 * counters allocated using the same state must be freed at the same 1889 * time. 1890 * 1891 * xt_percpu_counter_alloc_state contains the base address of the 1892 * allocated page and the current sub-offset. 1893 * 1894 * returns false on error. 1895 */ 1896 bool xt_percpu_counter_alloc(struct xt_percpu_counter_alloc_state *state, 1897 struct xt_counters *counter) 1898 { 1899 BUILD_BUG_ON(XT_PCPU_BLOCK_SIZE < (sizeof(*counter) * 2)); 1900 1901 if (nr_cpu_ids <= 1) 1902 return true; 1903 1904 if (!state->mem) { 1905 state->mem = __alloc_percpu(XT_PCPU_BLOCK_SIZE, 1906 XT_PCPU_BLOCK_SIZE); 1907 if (!state->mem) 1908 return false; 1909 } 1910 counter->pcnt = (__force unsigned long)(state->mem + state->off); 1911 state->off += sizeof(*counter); 1912 if (state->off > (XT_PCPU_BLOCK_SIZE - sizeof(*counter))) { 1913 state->mem = NULL; 1914 state->off = 0; 1915 } 1916 return true; 1917 } 1918 EXPORT_SYMBOL_GPL(xt_percpu_counter_alloc); 1919 1920 void xt_percpu_counter_free(struct xt_counters *counters) 1921 { 1922 unsigned long pcnt = counters->pcnt; 1923 1924 if (nr_cpu_ids > 1 && (pcnt & (XT_PCPU_BLOCK_SIZE - 1)) == 0) 1925 free_percpu((void __percpu *)pcnt); 1926 } 1927 EXPORT_SYMBOL_GPL(xt_percpu_counter_free); 1928 1929 static int __net_init xt_net_init(struct net *net) 1930 { 1931 int i; 1932 1933 for (i = 0; i < NFPROTO_NUMPROTO; i++) 1934 INIT_LIST_HEAD(&net->xt.tables[i]); 1935 return 0; 1936 } 1937 1938 static void __net_exit xt_net_exit(struct net *net) 1939 { 1940 int i; 1941 1942 for (i = 0; i < NFPROTO_NUMPROTO; i++) 1943 WARN_ON_ONCE(!list_empty(&net->xt.tables[i])); 1944 } 1945 1946 static struct pernet_operations xt_net_ops = { 1947 .init = xt_net_init, 1948 .exit = xt_net_exit, 1949 }; 1950 1951 static int __init xt_init(void) 1952 { 1953 unsigned int i; 1954 int rv; 1955 1956 for_each_possible_cpu(i) { 1957 seqcount_init(&per_cpu(xt_recseq, i)); 1958 } 1959 1960 xt = kmalloc(sizeof(struct xt_af) * NFPROTO_NUMPROTO, GFP_KERNEL); 1961 if (!xt) 1962 return -ENOMEM; 1963 1964 for (i = 0; i < NFPROTO_NUMPROTO; i++) { 1965 mutex_init(&xt[i].mutex); 1966 #ifdef CONFIG_COMPAT 1967 mutex_init(&xt[i].compat_mutex); 1968 xt[i].compat_tab = NULL; 1969 #endif 1970 INIT_LIST_HEAD(&xt[i].target); 1971 INIT_LIST_HEAD(&xt[i].match); 1972 } 1973 rv = register_pernet_subsys(&xt_net_ops); 1974 if (rv < 0) 1975 kfree(xt); 1976 return rv; 1977 } 1978 1979 static void __exit xt_fini(void) 1980 { 1981 unregister_pernet_subsys(&xt_net_ops); 1982 kfree(xt); 1983 } 1984 1985 module_init(xt_init); 1986 module_exit(xt_fini); 1987 1988