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 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN) 187 return ERR_PTR(-EINVAL); 188 189 mutex_lock(&xt[af].mutex); 190 list_for_each_entry(m, &xt[af].match, list) { 191 if (strcmp(m->name, name) == 0) { 192 if (m->revision == revision) { 193 if (try_module_get(m->me)) { 194 mutex_unlock(&xt[af].mutex); 195 return m; 196 } 197 } else 198 err = -EPROTOTYPE; /* Found something. */ 199 } 200 } 201 mutex_unlock(&xt[af].mutex); 202 203 if (af != NFPROTO_UNSPEC) 204 /* Try searching again in the family-independent list */ 205 return xt_find_match(NFPROTO_UNSPEC, name, revision); 206 207 return ERR_PTR(err); 208 } 209 EXPORT_SYMBOL(xt_find_match); 210 211 struct xt_match * 212 xt_request_find_match(uint8_t nfproto, const char *name, uint8_t revision) 213 { 214 struct xt_match *match; 215 216 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN) 217 return ERR_PTR(-EINVAL); 218 219 match = xt_find_match(nfproto, name, revision); 220 if (IS_ERR(match)) { 221 request_module("%st_%s", xt_prefix[nfproto], name); 222 match = xt_find_match(nfproto, name, revision); 223 } 224 225 return match; 226 } 227 EXPORT_SYMBOL_GPL(xt_request_find_match); 228 229 /* Find target, grabs ref. Returns ERR_PTR() on error. */ 230 static struct xt_target *xt_find_target(u8 af, const char *name, u8 revision) 231 { 232 struct xt_target *t; 233 int err = -ENOENT; 234 235 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN) 236 return ERR_PTR(-EINVAL); 237 238 mutex_lock(&xt[af].mutex); 239 list_for_each_entry(t, &xt[af].target, list) { 240 if (strcmp(t->name, name) == 0) { 241 if (t->revision == revision) { 242 if (try_module_get(t->me)) { 243 mutex_unlock(&xt[af].mutex); 244 return t; 245 } 246 } else 247 err = -EPROTOTYPE; /* Found something. */ 248 } 249 } 250 mutex_unlock(&xt[af].mutex); 251 252 if (af != NFPROTO_UNSPEC) 253 /* Try searching again in the family-independent list */ 254 return xt_find_target(NFPROTO_UNSPEC, name, revision); 255 256 return ERR_PTR(err); 257 } 258 259 struct xt_target *xt_request_find_target(u8 af, const char *name, u8 revision) 260 { 261 struct xt_target *target; 262 263 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN) 264 return ERR_PTR(-EINVAL); 265 266 target = xt_find_target(af, name, revision); 267 if (IS_ERR(target)) { 268 request_module("%st_%s", xt_prefix[af], name); 269 target = xt_find_target(af, name, revision); 270 } 271 272 return target; 273 } 274 EXPORT_SYMBOL_GPL(xt_request_find_target); 275 276 277 static int xt_obj_to_user(u16 __user *psize, u16 size, 278 void __user *pname, const char *name, 279 u8 __user *prev, u8 rev) 280 { 281 if (put_user(size, psize)) 282 return -EFAULT; 283 if (copy_to_user(pname, name, strlen(name) + 1)) 284 return -EFAULT; 285 if (put_user(rev, prev)) 286 return -EFAULT; 287 288 return 0; 289 } 290 291 #define XT_OBJ_TO_USER(U, K, TYPE, C_SIZE) \ 292 xt_obj_to_user(&U->u.TYPE##_size, C_SIZE ? : K->u.TYPE##_size, \ 293 U->u.user.name, K->u.kernel.TYPE->name, \ 294 &U->u.user.revision, K->u.kernel.TYPE->revision) 295 296 int xt_data_to_user(void __user *dst, const void *src, 297 int usersize, int size, int aligned_size) 298 { 299 usersize = usersize ? : size; 300 if (copy_to_user(dst, src, usersize)) 301 return -EFAULT; 302 if (usersize != aligned_size && 303 clear_user(dst + usersize, aligned_size - usersize)) 304 return -EFAULT; 305 306 return 0; 307 } 308 EXPORT_SYMBOL_GPL(xt_data_to_user); 309 310 #define XT_DATA_TO_USER(U, K, TYPE) \ 311 xt_data_to_user(U->data, K->data, \ 312 K->u.kernel.TYPE->usersize, \ 313 K->u.kernel.TYPE->TYPE##size, \ 314 XT_ALIGN(K->u.kernel.TYPE->TYPE##size)) 315 316 int xt_match_to_user(const struct xt_entry_match *m, 317 struct xt_entry_match __user *u) 318 { 319 return XT_OBJ_TO_USER(u, m, match, 0) || 320 XT_DATA_TO_USER(u, m, match); 321 } 322 EXPORT_SYMBOL_GPL(xt_match_to_user); 323 324 int xt_target_to_user(const struct xt_entry_target *t, 325 struct xt_entry_target __user *u) 326 { 327 return XT_OBJ_TO_USER(u, t, target, 0) || 328 XT_DATA_TO_USER(u, t, target); 329 } 330 EXPORT_SYMBOL_GPL(xt_target_to_user); 331 332 static int match_revfn(u8 af, const char *name, u8 revision, int *bestp) 333 { 334 const struct xt_match *m; 335 int have_rev = 0; 336 337 list_for_each_entry(m, &xt[af].match, list) { 338 if (strcmp(m->name, name) == 0) { 339 if (m->revision > *bestp) 340 *bestp = m->revision; 341 if (m->revision == revision) 342 have_rev = 1; 343 } 344 } 345 346 if (af != NFPROTO_UNSPEC && !have_rev) 347 return match_revfn(NFPROTO_UNSPEC, name, revision, bestp); 348 349 return have_rev; 350 } 351 352 static int target_revfn(u8 af, const char *name, u8 revision, int *bestp) 353 { 354 const struct xt_target *t; 355 int have_rev = 0; 356 357 list_for_each_entry(t, &xt[af].target, list) { 358 if (strcmp(t->name, name) == 0) { 359 if (t->revision > *bestp) 360 *bestp = t->revision; 361 if (t->revision == revision) 362 have_rev = 1; 363 } 364 } 365 366 if (af != NFPROTO_UNSPEC && !have_rev) 367 return target_revfn(NFPROTO_UNSPEC, name, revision, bestp); 368 369 return have_rev; 370 } 371 372 /* Returns true or false (if no such extension at all) */ 373 int xt_find_revision(u8 af, const char *name, u8 revision, int target, 374 int *err) 375 { 376 int have_rev, best = -1; 377 378 mutex_lock(&xt[af].mutex); 379 if (target == 1) 380 have_rev = target_revfn(af, name, revision, &best); 381 else 382 have_rev = match_revfn(af, name, revision, &best); 383 mutex_unlock(&xt[af].mutex); 384 385 /* Nothing at all? Return 0 to try loading module. */ 386 if (best == -1) { 387 *err = -ENOENT; 388 return 0; 389 } 390 391 *err = best; 392 if (!have_rev) 393 *err = -EPROTONOSUPPORT; 394 return 1; 395 } 396 EXPORT_SYMBOL_GPL(xt_find_revision); 397 398 static char * 399 textify_hooks(char *buf, size_t size, unsigned int mask, uint8_t nfproto) 400 { 401 static const char *const inetbr_names[] = { 402 "PREROUTING", "INPUT", "FORWARD", 403 "OUTPUT", "POSTROUTING", "BROUTING", 404 }; 405 static const char *const arp_names[] = { 406 "INPUT", "FORWARD", "OUTPUT", 407 }; 408 const char *const *names; 409 unsigned int i, max; 410 char *p = buf; 411 bool np = false; 412 int res; 413 414 names = (nfproto == NFPROTO_ARP) ? arp_names : inetbr_names; 415 max = (nfproto == NFPROTO_ARP) ? ARRAY_SIZE(arp_names) : 416 ARRAY_SIZE(inetbr_names); 417 *p = '\0'; 418 for (i = 0; i < max; ++i) { 419 if (!(mask & (1 << i))) 420 continue; 421 res = snprintf(p, size, "%s%s", np ? "/" : "", names[i]); 422 if (res > 0) { 423 size -= res; 424 p += res; 425 } 426 np = true; 427 } 428 429 return buf; 430 } 431 432 /** 433 * xt_check_proc_name - check that name is suitable for /proc file creation 434 * 435 * @name: file name candidate 436 * @size: length of buffer 437 * 438 * some x_tables modules wish to create a file in /proc. 439 * This function makes sure that the name is suitable for this 440 * purpose, it checks that name is NUL terminated and isn't a 'special' 441 * name, like "..". 442 * 443 * returns negative number on error or 0 if name is useable. 444 */ 445 int xt_check_proc_name(const char *name, unsigned int size) 446 { 447 if (name[0] == '\0') 448 return -EINVAL; 449 450 if (strnlen(name, size) == size) 451 return -ENAMETOOLONG; 452 453 if (strcmp(name, ".") == 0 || 454 strcmp(name, "..") == 0 || 455 strchr(name, '/')) 456 return -EINVAL; 457 458 return 0; 459 } 460 EXPORT_SYMBOL(xt_check_proc_name); 461 462 int xt_check_match(struct xt_mtchk_param *par, 463 unsigned int size, u16 proto, bool inv_proto) 464 { 465 int ret; 466 467 if (XT_ALIGN(par->match->matchsize) != size && 468 par->match->matchsize != -1) { 469 /* 470 * ebt_among is exempt from centralized matchsize checking 471 * because it uses a dynamic-size data set. 472 */ 473 pr_err_ratelimited("%s_tables: %s.%u match: invalid size %u (kernel) != (user) %u\n", 474 xt_prefix[par->family], par->match->name, 475 par->match->revision, 476 XT_ALIGN(par->match->matchsize), size); 477 return -EINVAL; 478 } 479 if (par->match->table != NULL && 480 strcmp(par->match->table, par->table) != 0) { 481 pr_info_ratelimited("%s_tables: %s match: only valid in %s table, not %s\n", 482 xt_prefix[par->family], par->match->name, 483 par->match->table, par->table); 484 return -EINVAL; 485 } 486 if (par->match->hooks && (par->hook_mask & ~par->match->hooks) != 0) { 487 char used[64], allow[64]; 488 489 pr_info_ratelimited("%s_tables: %s match: used from hooks %s, but only valid from %s\n", 490 xt_prefix[par->family], par->match->name, 491 textify_hooks(used, sizeof(used), 492 par->hook_mask, par->family), 493 textify_hooks(allow, sizeof(allow), 494 par->match->hooks, 495 par->family)); 496 return -EINVAL; 497 } 498 if (par->match->proto && (par->match->proto != proto || inv_proto)) { 499 pr_info_ratelimited("%s_tables: %s match: only valid for protocol %u\n", 500 xt_prefix[par->family], par->match->name, 501 par->match->proto); 502 return -EINVAL; 503 } 504 if (par->match->checkentry != NULL) { 505 ret = par->match->checkentry(par); 506 if (ret < 0) 507 return ret; 508 else if (ret > 0) 509 /* Flag up potential errors. */ 510 return -EIO; 511 } 512 return 0; 513 } 514 EXPORT_SYMBOL_GPL(xt_check_match); 515 516 /** xt_check_entry_match - check that matches end before start of target 517 * 518 * @match: beginning of xt_entry_match 519 * @target: beginning of this rules target (alleged end of matches) 520 * @alignment: alignment requirement of match structures 521 * 522 * Validates that all matches add up to the beginning of the target, 523 * and that each match covers at least the base structure size. 524 * 525 * Return: 0 on success, negative errno on failure. 526 */ 527 static int xt_check_entry_match(const char *match, const char *target, 528 const size_t alignment) 529 { 530 const struct xt_entry_match *pos; 531 int length = target - match; 532 533 if (length == 0) /* no matches */ 534 return 0; 535 536 pos = (struct xt_entry_match *)match; 537 do { 538 if ((unsigned long)pos % alignment) 539 return -EINVAL; 540 541 if (length < (int)sizeof(struct xt_entry_match)) 542 return -EINVAL; 543 544 if (pos->u.match_size < sizeof(struct xt_entry_match)) 545 return -EINVAL; 546 547 if (pos->u.match_size > length) 548 return -EINVAL; 549 550 length -= pos->u.match_size; 551 pos = ((void *)((char *)(pos) + (pos)->u.match_size)); 552 } while (length > 0); 553 554 return 0; 555 } 556 557 /** xt_check_table_hooks - check hook entry points are sane 558 * 559 * @info xt_table_info to check 560 * @valid_hooks - hook entry points that we can enter from 561 * 562 * Validates that the hook entry and underflows points are set up. 563 * 564 * Return: 0 on success, negative errno on failure. 565 */ 566 int xt_check_table_hooks(const struct xt_table_info *info, unsigned int valid_hooks) 567 { 568 const char *err = "unsorted underflow"; 569 unsigned int i, max_uflow, max_entry; 570 bool check_hooks = false; 571 572 BUILD_BUG_ON(ARRAY_SIZE(info->hook_entry) != ARRAY_SIZE(info->underflow)); 573 574 max_entry = 0; 575 max_uflow = 0; 576 577 for (i = 0; i < ARRAY_SIZE(info->hook_entry); i++) { 578 if (!(valid_hooks & (1 << i))) 579 continue; 580 581 if (info->hook_entry[i] == 0xFFFFFFFF) 582 return -EINVAL; 583 if (info->underflow[i] == 0xFFFFFFFF) 584 return -EINVAL; 585 586 if (check_hooks) { 587 if (max_uflow > info->underflow[i]) 588 goto error; 589 590 if (max_uflow == info->underflow[i]) { 591 err = "duplicate underflow"; 592 goto error; 593 } 594 if (max_entry > info->hook_entry[i]) { 595 err = "unsorted entry"; 596 goto error; 597 } 598 if (max_entry == info->hook_entry[i]) { 599 err = "duplicate entry"; 600 goto error; 601 } 602 } 603 max_entry = info->hook_entry[i]; 604 max_uflow = info->underflow[i]; 605 check_hooks = true; 606 } 607 608 return 0; 609 error: 610 pr_err_ratelimited("%s at hook %d\n", err, i); 611 return -EINVAL; 612 } 613 EXPORT_SYMBOL(xt_check_table_hooks); 614 615 static bool verdict_ok(int verdict) 616 { 617 if (verdict > 0) 618 return true; 619 620 if (verdict < 0) { 621 int v = -verdict - 1; 622 623 if (verdict == XT_RETURN) 624 return true; 625 626 switch (v) { 627 case NF_ACCEPT: return true; 628 case NF_DROP: return true; 629 case NF_QUEUE: return true; 630 default: 631 break; 632 } 633 634 return false; 635 } 636 637 return false; 638 } 639 640 static bool error_tg_ok(unsigned int usersize, unsigned int kernsize, 641 const char *msg, unsigned int msglen) 642 { 643 return usersize == kernsize && strnlen(msg, msglen) < msglen; 644 } 645 646 #ifdef CONFIG_COMPAT 647 int xt_compat_add_offset(u_int8_t af, unsigned int offset, int delta) 648 { 649 struct xt_af *xp = &xt[af]; 650 651 WARN_ON(!mutex_is_locked(&xt[af].compat_mutex)); 652 653 if (WARN_ON(!xp->compat_tab)) 654 return -ENOMEM; 655 656 if (xp->cur >= xp->number) 657 return -EINVAL; 658 659 if (xp->cur) 660 delta += xp->compat_tab[xp->cur - 1].delta; 661 xp->compat_tab[xp->cur].offset = offset; 662 xp->compat_tab[xp->cur].delta = delta; 663 xp->cur++; 664 return 0; 665 } 666 EXPORT_SYMBOL_GPL(xt_compat_add_offset); 667 668 void xt_compat_flush_offsets(u_int8_t af) 669 { 670 WARN_ON(!mutex_is_locked(&xt[af].compat_mutex)); 671 672 if (xt[af].compat_tab) { 673 vfree(xt[af].compat_tab); 674 xt[af].compat_tab = NULL; 675 xt[af].number = 0; 676 xt[af].cur = 0; 677 } 678 } 679 EXPORT_SYMBOL_GPL(xt_compat_flush_offsets); 680 681 int xt_compat_calc_jump(u_int8_t af, unsigned int offset) 682 { 683 struct compat_delta *tmp = xt[af].compat_tab; 684 int mid, left = 0, right = xt[af].cur - 1; 685 686 while (left <= right) { 687 mid = (left + right) >> 1; 688 if (offset > tmp[mid].offset) 689 left = mid + 1; 690 else if (offset < tmp[mid].offset) 691 right = mid - 1; 692 else 693 return mid ? tmp[mid - 1].delta : 0; 694 } 695 return left ? tmp[left - 1].delta : 0; 696 } 697 EXPORT_SYMBOL_GPL(xt_compat_calc_jump); 698 699 int xt_compat_init_offsets(u8 af, unsigned int number) 700 { 701 size_t mem; 702 703 WARN_ON(!mutex_is_locked(&xt[af].compat_mutex)); 704 705 if (!number || number > (INT_MAX / sizeof(struct compat_delta))) 706 return -EINVAL; 707 708 if (WARN_ON(xt[af].compat_tab)) 709 return -EINVAL; 710 711 mem = sizeof(struct compat_delta) * number; 712 if (mem > XT_MAX_TABLE_SIZE) 713 return -ENOMEM; 714 715 xt[af].compat_tab = vmalloc(mem); 716 if (!xt[af].compat_tab) 717 return -ENOMEM; 718 719 xt[af].number = number; 720 xt[af].cur = 0; 721 722 return 0; 723 } 724 EXPORT_SYMBOL(xt_compat_init_offsets); 725 726 int xt_compat_match_offset(const struct xt_match *match) 727 { 728 u_int16_t csize = match->compatsize ? : match->matchsize; 729 return XT_ALIGN(match->matchsize) - COMPAT_XT_ALIGN(csize); 730 } 731 EXPORT_SYMBOL_GPL(xt_compat_match_offset); 732 733 void xt_compat_match_from_user(struct xt_entry_match *m, void **dstptr, 734 unsigned int *size) 735 { 736 const struct xt_match *match = m->u.kernel.match; 737 struct compat_xt_entry_match *cm = (struct compat_xt_entry_match *)m; 738 int pad, off = xt_compat_match_offset(match); 739 u_int16_t msize = cm->u.user.match_size; 740 char name[sizeof(m->u.user.name)]; 741 742 m = *dstptr; 743 memcpy(m, cm, sizeof(*cm)); 744 if (match->compat_from_user) 745 match->compat_from_user(m->data, cm->data); 746 else 747 memcpy(m->data, cm->data, msize - sizeof(*cm)); 748 pad = XT_ALIGN(match->matchsize) - match->matchsize; 749 if (pad > 0) 750 memset(m->data + match->matchsize, 0, pad); 751 752 msize += off; 753 m->u.user.match_size = msize; 754 strlcpy(name, match->name, sizeof(name)); 755 module_put(match->me); 756 strncpy(m->u.user.name, name, sizeof(m->u.user.name)); 757 758 *size += off; 759 *dstptr += msize; 760 } 761 EXPORT_SYMBOL_GPL(xt_compat_match_from_user); 762 763 #define COMPAT_XT_DATA_TO_USER(U, K, TYPE, C_SIZE) \ 764 xt_data_to_user(U->data, K->data, \ 765 K->u.kernel.TYPE->usersize, \ 766 C_SIZE, \ 767 COMPAT_XT_ALIGN(C_SIZE)) 768 769 int xt_compat_match_to_user(const struct xt_entry_match *m, 770 void __user **dstptr, unsigned int *size) 771 { 772 const struct xt_match *match = m->u.kernel.match; 773 struct compat_xt_entry_match __user *cm = *dstptr; 774 int off = xt_compat_match_offset(match); 775 u_int16_t msize = m->u.user.match_size - off; 776 777 if (XT_OBJ_TO_USER(cm, m, match, msize)) 778 return -EFAULT; 779 780 if (match->compat_to_user) { 781 if (match->compat_to_user((void __user *)cm->data, m->data)) 782 return -EFAULT; 783 } else { 784 if (COMPAT_XT_DATA_TO_USER(cm, m, match, msize - sizeof(*cm))) 785 return -EFAULT; 786 } 787 788 *size -= off; 789 *dstptr += msize; 790 return 0; 791 } 792 EXPORT_SYMBOL_GPL(xt_compat_match_to_user); 793 794 /* non-compat version may have padding after verdict */ 795 struct compat_xt_standard_target { 796 struct compat_xt_entry_target t; 797 compat_uint_t verdict; 798 }; 799 800 struct compat_xt_error_target { 801 struct compat_xt_entry_target t; 802 char errorname[XT_FUNCTION_MAXNAMELEN]; 803 }; 804 805 int xt_compat_check_entry_offsets(const void *base, const char *elems, 806 unsigned int target_offset, 807 unsigned int next_offset) 808 { 809 long size_of_base_struct = elems - (const char *)base; 810 const struct compat_xt_entry_target *t; 811 const char *e = base; 812 813 if (target_offset < size_of_base_struct) 814 return -EINVAL; 815 816 if (target_offset + sizeof(*t) > next_offset) 817 return -EINVAL; 818 819 t = (void *)(e + target_offset); 820 if (t->u.target_size < sizeof(*t)) 821 return -EINVAL; 822 823 if (target_offset + t->u.target_size > next_offset) 824 return -EINVAL; 825 826 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0) { 827 const struct compat_xt_standard_target *st = (const void *)t; 828 829 if (COMPAT_XT_ALIGN(target_offset + sizeof(*st)) != next_offset) 830 return -EINVAL; 831 832 if (!verdict_ok(st->verdict)) 833 return -EINVAL; 834 } else if (strcmp(t->u.user.name, XT_ERROR_TARGET) == 0) { 835 const struct compat_xt_error_target *et = (const void *)t; 836 837 if (!error_tg_ok(t->u.target_size, sizeof(*et), 838 et->errorname, sizeof(et->errorname))) 839 return -EINVAL; 840 } 841 842 /* compat_xt_entry match has less strict alignment requirements, 843 * otherwise they are identical. In case of padding differences 844 * we need to add compat version of xt_check_entry_match. 845 */ 846 BUILD_BUG_ON(sizeof(struct compat_xt_entry_match) != sizeof(struct xt_entry_match)); 847 848 return xt_check_entry_match(elems, base + target_offset, 849 __alignof__(struct compat_xt_entry_match)); 850 } 851 EXPORT_SYMBOL(xt_compat_check_entry_offsets); 852 #endif /* CONFIG_COMPAT */ 853 854 /** 855 * xt_check_entry_offsets - validate arp/ip/ip6t_entry 856 * 857 * @base: pointer to arp/ip/ip6t_entry 858 * @elems: pointer to first xt_entry_match, i.e. ip(6)t_entry->elems 859 * @target_offset: the arp/ip/ip6_t->target_offset 860 * @next_offset: the arp/ip/ip6_t->next_offset 861 * 862 * validates that target_offset and next_offset are sane and that all 863 * match sizes (if any) align with the target offset. 864 * 865 * This function does not validate the targets or matches themselves, it 866 * only tests that all the offsets and sizes are correct, that all 867 * match structures are aligned, and that the last structure ends where 868 * the target structure begins. 869 * 870 * Also see xt_compat_check_entry_offsets for CONFIG_COMPAT version. 871 * 872 * The arp/ip/ip6t_entry structure @base must have passed following tests: 873 * - it must point to a valid memory location 874 * - base to base + next_offset must be accessible, i.e. not exceed allocated 875 * length. 876 * 877 * A well-formed entry looks like this: 878 * 879 * ip(6)t_entry match [mtdata] match [mtdata] target [tgdata] ip(6)t_entry 880 * e->elems[]-----' | | 881 * matchsize | | 882 * matchsize | | 883 * | | 884 * target_offset---------------------------------' | 885 * next_offset---------------------------------------------------' 886 * 887 * elems[]: flexible array member at end of ip(6)/arpt_entry struct. 888 * This is where matches (if any) and the target reside. 889 * target_offset: beginning of target. 890 * next_offset: start of the next rule; also: size of this rule. 891 * Since targets have a minimum size, target_offset + minlen <= next_offset. 892 * 893 * Every match stores its size, sum of sizes must not exceed target_offset. 894 * 895 * Return: 0 on success, negative errno on failure. 896 */ 897 int xt_check_entry_offsets(const void *base, 898 const char *elems, 899 unsigned int target_offset, 900 unsigned int next_offset) 901 { 902 long size_of_base_struct = elems - (const char *)base; 903 const struct xt_entry_target *t; 904 const char *e = base; 905 906 /* target start is within the ip/ip6/arpt_entry struct */ 907 if (target_offset < size_of_base_struct) 908 return -EINVAL; 909 910 if (target_offset + sizeof(*t) > next_offset) 911 return -EINVAL; 912 913 t = (void *)(e + target_offset); 914 if (t->u.target_size < sizeof(*t)) 915 return -EINVAL; 916 917 if (target_offset + t->u.target_size > next_offset) 918 return -EINVAL; 919 920 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0) { 921 const struct xt_standard_target *st = (const void *)t; 922 923 if (XT_ALIGN(target_offset + sizeof(*st)) != next_offset) 924 return -EINVAL; 925 926 if (!verdict_ok(st->verdict)) 927 return -EINVAL; 928 } else if (strcmp(t->u.user.name, XT_ERROR_TARGET) == 0) { 929 const struct xt_error_target *et = (const void *)t; 930 931 if (!error_tg_ok(t->u.target_size, sizeof(*et), 932 et->errorname, sizeof(et->errorname))) 933 return -EINVAL; 934 } 935 936 return xt_check_entry_match(elems, base + target_offset, 937 __alignof__(struct xt_entry_match)); 938 } 939 EXPORT_SYMBOL(xt_check_entry_offsets); 940 941 /** 942 * xt_alloc_entry_offsets - allocate array to store rule head offsets 943 * 944 * @size: number of entries 945 * 946 * Return: NULL or kmalloc'd or vmalloc'd array 947 */ 948 unsigned int *xt_alloc_entry_offsets(unsigned int size) 949 { 950 if (size > XT_MAX_TABLE_SIZE / sizeof(unsigned int)) 951 return NULL; 952 953 return kvmalloc_array(size, sizeof(unsigned int), GFP_KERNEL | __GFP_ZERO); 954 955 } 956 EXPORT_SYMBOL(xt_alloc_entry_offsets); 957 958 /** 959 * xt_find_jump_offset - check if target is a valid jump offset 960 * 961 * @offsets: array containing all valid rule start offsets of a rule blob 962 * @target: the jump target to search for 963 * @size: entries in @offset 964 */ 965 bool xt_find_jump_offset(const unsigned int *offsets, 966 unsigned int target, unsigned int size) 967 { 968 int m, low = 0, hi = size; 969 970 while (hi > low) { 971 m = (low + hi) / 2u; 972 973 if (offsets[m] > target) 974 hi = m; 975 else if (offsets[m] < target) 976 low = m + 1; 977 else 978 return true; 979 } 980 981 return false; 982 } 983 EXPORT_SYMBOL(xt_find_jump_offset); 984 985 int xt_check_target(struct xt_tgchk_param *par, 986 unsigned int size, u16 proto, bool inv_proto) 987 { 988 int ret; 989 990 if (XT_ALIGN(par->target->targetsize) != size) { 991 pr_err_ratelimited("%s_tables: %s.%u target: invalid size %u (kernel) != (user) %u\n", 992 xt_prefix[par->family], par->target->name, 993 par->target->revision, 994 XT_ALIGN(par->target->targetsize), size); 995 return -EINVAL; 996 } 997 if (par->target->table != NULL && 998 strcmp(par->target->table, par->table) != 0) { 999 pr_info_ratelimited("%s_tables: %s target: only valid in %s table, not %s\n", 1000 xt_prefix[par->family], par->target->name, 1001 par->target->table, par->table); 1002 return -EINVAL; 1003 } 1004 if (par->target->hooks && (par->hook_mask & ~par->target->hooks) != 0) { 1005 char used[64], allow[64]; 1006 1007 pr_info_ratelimited("%s_tables: %s target: used from hooks %s, but only usable from %s\n", 1008 xt_prefix[par->family], par->target->name, 1009 textify_hooks(used, sizeof(used), 1010 par->hook_mask, par->family), 1011 textify_hooks(allow, sizeof(allow), 1012 par->target->hooks, 1013 par->family)); 1014 return -EINVAL; 1015 } 1016 if (par->target->proto && (par->target->proto != proto || inv_proto)) { 1017 pr_info_ratelimited("%s_tables: %s target: only valid for protocol %u\n", 1018 xt_prefix[par->family], par->target->name, 1019 par->target->proto); 1020 return -EINVAL; 1021 } 1022 if (par->target->checkentry != NULL) { 1023 ret = par->target->checkentry(par); 1024 if (ret < 0) 1025 return ret; 1026 else if (ret > 0) 1027 /* Flag up potential errors. */ 1028 return -EIO; 1029 } 1030 return 0; 1031 } 1032 EXPORT_SYMBOL_GPL(xt_check_target); 1033 1034 /** 1035 * xt_copy_counters_from_user - copy counters and metadata from userspace 1036 * 1037 * @user: src pointer to userspace memory 1038 * @len: alleged size of userspace memory 1039 * @info: where to store the xt_counters_info metadata 1040 * @compat: true if we setsockopt call is done by 32bit task on 64bit kernel 1041 * 1042 * Copies counter meta data from @user and stores it in @info. 1043 * 1044 * vmallocs memory to hold the counters, then copies the counter data 1045 * from @user to the new memory and returns a pointer to it. 1046 * 1047 * If @compat is true, @info gets converted automatically to the 64bit 1048 * representation. 1049 * 1050 * The metadata associated with the counters is stored in @info. 1051 * 1052 * Return: returns pointer that caller has to test via IS_ERR(). 1053 * If IS_ERR is false, caller has to vfree the pointer. 1054 */ 1055 void *xt_copy_counters_from_user(const void __user *user, unsigned int len, 1056 struct xt_counters_info *info, bool compat) 1057 { 1058 void *mem; 1059 u64 size; 1060 1061 #ifdef CONFIG_COMPAT 1062 if (compat) { 1063 /* structures only differ in size due to alignment */ 1064 struct compat_xt_counters_info compat_tmp; 1065 1066 if (len <= sizeof(compat_tmp)) 1067 return ERR_PTR(-EINVAL); 1068 1069 len -= sizeof(compat_tmp); 1070 if (copy_from_user(&compat_tmp, user, sizeof(compat_tmp)) != 0) 1071 return ERR_PTR(-EFAULT); 1072 1073 memcpy(info->name, compat_tmp.name, sizeof(info->name) - 1); 1074 info->num_counters = compat_tmp.num_counters; 1075 user += sizeof(compat_tmp); 1076 } else 1077 #endif 1078 { 1079 if (len <= sizeof(*info)) 1080 return ERR_PTR(-EINVAL); 1081 1082 len -= sizeof(*info); 1083 if (copy_from_user(info, user, sizeof(*info)) != 0) 1084 return ERR_PTR(-EFAULT); 1085 1086 user += sizeof(*info); 1087 } 1088 info->name[sizeof(info->name) - 1] = '\0'; 1089 1090 size = sizeof(struct xt_counters); 1091 size *= info->num_counters; 1092 1093 if (size != (u64)len) 1094 return ERR_PTR(-EINVAL); 1095 1096 mem = vmalloc(len); 1097 if (!mem) 1098 return ERR_PTR(-ENOMEM); 1099 1100 if (copy_from_user(mem, user, len) == 0) 1101 return mem; 1102 1103 vfree(mem); 1104 return ERR_PTR(-EFAULT); 1105 } 1106 EXPORT_SYMBOL_GPL(xt_copy_counters_from_user); 1107 1108 #ifdef CONFIG_COMPAT 1109 int xt_compat_target_offset(const struct xt_target *target) 1110 { 1111 u_int16_t csize = target->compatsize ? : target->targetsize; 1112 return XT_ALIGN(target->targetsize) - COMPAT_XT_ALIGN(csize); 1113 } 1114 EXPORT_SYMBOL_GPL(xt_compat_target_offset); 1115 1116 void xt_compat_target_from_user(struct xt_entry_target *t, void **dstptr, 1117 unsigned int *size) 1118 { 1119 const struct xt_target *target = t->u.kernel.target; 1120 struct compat_xt_entry_target *ct = (struct compat_xt_entry_target *)t; 1121 int pad, off = xt_compat_target_offset(target); 1122 u_int16_t tsize = ct->u.user.target_size; 1123 char name[sizeof(t->u.user.name)]; 1124 1125 t = *dstptr; 1126 memcpy(t, ct, sizeof(*ct)); 1127 if (target->compat_from_user) 1128 target->compat_from_user(t->data, ct->data); 1129 else 1130 memcpy(t->data, ct->data, tsize - sizeof(*ct)); 1131 pad = XT_ALIGN(target->targetsize) - target->targetsize; 1132 if (pad > 0) 1133 memset(t->data + target->targetsize, 0, pad); 1134 1135 tsize += off; 1136 t->u.user.target_size = tsize; 1137 strlcpy(name, target->name, sizeof(name)); 1138 module_put(target->me); 1139 strncpy(t->u.user.name, name, sizeof(t->u.user.name)); 1140 1141 *size += off; 1142 *dstptr += tsize; 1143 } 1144 EXPORT_SYMBOL_GPL(xt_compat_target_from_user); 1145 1146 int xt_compat_target_to_user(const struct xt_entry_target *t, 1147 void __user **dstptr, unsigned int *size) 1148 { 1149 const struct xt_target *target = t->u.kernel.target; 1150 struct compat_xt_entry_target __user *ct = *dstptr; 1151 int off = xt_compat_target_offset(target); 1152 u_int16_t tsize = t->u.user.target_size - off; 1153 1154 if (XT_OBJ_TO_USER(ct, t, target, tsize)) 1155 return -EFAULT; 1156 1157 if (target->compat_to_user) { 1158 if (target->compat_to_user((void __user *)ct->data, t->data)) 1159 return -EFAULT; 1160 } else { 1161 if (COMPAT_XT_DATA_TO_USER(ct, t, target, tsize - sizeof(*ct))) 1162 return -EFAULT; 1163 } 1164 1165 *size -= off; 1166 *dstptr += tsize; 1167 return 0; 1168 } 1169 EXPORT_SYMBOL_GPL(xt_compat_target_to_user); 1170 #endif 1171 1172 struct xt_table_info *xt_alloc_table_info(unsigned int size) 1173 { 1174 struct xt_table_info *info = NULL; 1175 size_t sz = sizeof(*info) + size; 1176 1177 if (sz < sizeof(*info) || sz >= XT_MAX_TABLE_SIZE) 1178 return NULL; 1179 1180 info = kvmalloc(sz, GFP_KERNEL_ACCOUNT); 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(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 static void *xt_table_seq_start(struct seq_file *seq, loff_t *pos) 1493 { 1494 struct net *net = seq_file_net(seq); 1495 u_int8_t af = (unsigned long)PDE_DATA(file_inode(seq->file)); 1496 1497 mutex_lock(&xt[af].mutex); 1498 return seq_list_start(&net->xt.tables[af], *pos); 1499 } 1500 1501 static void *xt_table_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1502 { 1503 struct net *net = seq_file_net(seq); 1504 u_int8_t af = (unsigned long)PDE_DATA(file_inode(seq->file)); 1505 1506 return seq_list_next(v, &net->xt.tables[af], pos); 1507 } 1508 1509 static void xt_table_seq_stop(struct seq_file *seq, void *v) 1510 { 1511 u_int8_t af = (unsigned long)PDE_DATA(file_inode(seq->file)); 1512 1513 mutex_unlock(&xt[af].mutex); 1514 } 1515 1516 static int xt_table_seq_show(struct seq_file *seq, void *v) 1517 { 1518 struct xt_table *table = list_entry(v, struct xt_table, list); 1519 1520 if (*table->name) 1521 seq_printf(seq, "%s\n", table->name); 1522 return 0; 1523 } 1524 1525 static const struct seq_operations xt_table_seq_ops = { 1526 .start = xt_table_seq_start, 1527 .next = xt_table_seq_next, 1528 .stop = xt_table_seq_stop, 1529 .show = xt_table_seq_show, 1530 }; 1531 1532 /* 1533 * Traverse state for ip{,6}_{tables,matches} for helping crossing 1534 * the multi-AF mutexes. 1535 */ 1536 struct nf_mttg_trav { 1537 struct list_head *head, *curr; 1538 uint8_t class; 1539 }; 1540 1541 enum { 1542 MTTG_TRAV_INIT, 1543 MTTG_TRAV_NFP_UNSPEC, 1544 MTTG_TRAV_NFP_SPEC, 1545 MTTG_TRAV_DONE, 1546 }; 1547 1548 static void *xt_mttg_seq_next(struct seq_file *seq, void *v, loff_t *ppos, 1549 bool is_target) 1550 { 1551 static const uint8_t next_class[] = { 1552 [MTTG_TRAV_NFP_UNSPEC] = MTTG_TRAV_NFP_SPEC, 1553 [MTTG_TRAV_NFP_SPEC] = MTTG_TRAV_DONE, 1554 }; 1555 uint8_t nfproto = (unsigned long)PDE_DATA(file_inode(seq->file)); 1556 struct nf_mttg_trav *trav = seq->private; 1557 1558 switch (trav->class) { 1559 case MTTG_TRAV_INIT: 1560 trav->class = MTTG_TRAV_NFP_UNSPEC; 1561 mutex_lock(&xt[NFPROTO_UNSPEC].mutex); 1562 trav->head = trav->curr = is_target ? 1563 &xt[NFPROTO_UNSPEC].target : &xt[NFPROTO_UNSPEC].match; 1564 break; 1565 case MTTG_TRAV_NFP_UNSPEC: 1566 trav->curr = trav->curr->next; 1567 if (trav->curr != trav->head) 1568 break; 1569 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex); 1570 mutex_lock(&xt[nfproto].mutex); 1571 trav->head = trav->curr = is_target ? 1572 &xt[nfproto].target : &xt[nfproto].match; 1573 trav->class = next_class[trav->class]; 1574 break; 1575 case MTTG_TRAV_NFP_SPEC: 1576 trav->curr = trav->curr->next; 1577 if (trav->curr != trav->head) 1578 break; 1579 /* fall through */ 1580 default: 1581 return NULL; 1582 } 1583 1584 if (ppos != NULL) 1585 ++*ppos; 1586 return trav; 1587 } 1588 1589 static void *xt_mttg_seq_start(struct seq_file *seq, loff_t *pos, 1590 bool is_target) 1591 { 1592 struct nf_mttg_trav *trav = seq->private; 1593 unsigned int j; 1594 1595 trav->class = MTTG_TRAV_INIT; 1596 for (j = 0; j < *pos; ++j) 1597 if (xt_mttg_seq_next(seq, NULL, NULL, is_target) == NULL) 1598 return NULL; 1599 return trav; 1600 } 1601 1602 static void xt_mttg_seq_stop(struct seq_file *seq, void *v) 1603 { 1604 uint8_t nfproto = (unsigned long)PDE_DATA(file_inode(seq->file)); 1605 struct nf_mttg_trav *trav = seq->private; 1606 1607 switch (trav->class) { 1608 case MTTG_TRAV_NFP_UNSPEC: 1609 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex); 1610 break; 1611 case MTTG_TRAV_NFP_SPEC: 1612 mutex_unlock(&xt[nfproto].mutex); 1613 break; 1614 } 1615 } 1616 1617 static void *xt_match_seq_start(struct seq_file *seq, loff_t *pos) 1618 { 1619 return xt_mttg_seq_start(seq, pos, false); 1620 } 1621 1622 static void *xt_match_seq_next(struct seq_file *seq, void *v, loff_t *ppos) 1623 { 1624 return xt_mttg_seq_next(seq, v, ppos, false); 1625 } 1626 1627 static int xt_match_seq_show(struct seq_file *seq, void *v) 1628 { 1629 const struct nf_mttg_trav *trav = seq->private; 1630 const struct xt_match *match; 1631 1632 switch (trav->class) { 1633 case MTTG_TRAV_NFP_UNSPEC: 1634 case MTTG_TRAV_NFP_SPEC: 1635 if (trav->curr == trav->head) 1636 return 0; 1637 match = list_entry(trav->curr, struct xt_match, list); 1638 if (*match->name) 1639 seq_printf(seq, "%s\n", match->name); 1640 } 1641 return 0; 1642 } 1643 1644 static const struct seq_operations xt_match_seq_ops = { 1645 .start = xt_match_seq_start, 1646 .next = xt_match_seq_next, 1647 .stop = xt_mttg_seq_stop, 1648 .show = xt_match_seq_show, 1649 }; 1650 1651 static void *xt_target_seq_start(struct seq_file *seq, loff_t *pos) 1652 { 1653 return xt_mttg_seq_start(seq, pos, true); 1654 } 1655 1656 static void *xt_target_seq_next(struct seq_file *seq, void *v, loff_t *ppos) 1657 { 1658 return xt_mttg_seq_next(seq, v, ppos, true); 1659 } 1660 1661 static int xt_target_seq_show(struct seq_file *seq, void *v) 1662 { 1663 const struct nf_mttg_trav *trav = seq->private; 1664 const struct xt_target *target; 1665 1666 switch (trav->class) { 1667 case MTTG_TRAV_NFP_UNSPEC: 1668 case MTTG_TRAV_NFP_SPEC: 1669 if (trav->curr == trav->head) 1670 return 0; 1671 target = list_entry(trav->curr, struct xt_target, list); 1672 if (*target->name) 1673 seq_printf(seq, "%s\n", target->name); 1674 } 1675 return 0; 1676 } 1677 1678 static const struct seq_operations xt_target_seq_ops = { 1679 .start = xt_target_seq_start, 1680 .next = xt_target_seq_next, 1681 .stop = xt_mttg_seq_stop, 1682 .show = xt_target_seq_show, 1683 }; 1684 1685 #define FORMAT_TABLES "_tables_names" 1686 #define FORMAT_MATCHES "_tables_matches" 1687 #define FORMAT_TARGETS "_tables_targets" 1688 1689 #endif /* CONFIG_PROC_FS */ 1690 1691 /** 1692 * xt_hook_ops_alloc - set up hooks for a new table 1693 * @table: table with metadata needed to set up hooks 1694 * @fn: Hook function 1695 * 1696 * This function will create the nf_hook_ops that the x_table needs 1697 * to hand to xt_hook_link_net(). 1698 */ 1699 struct nf_hook_ops * 1700 xt_hook_ops_alloc(const struct xt_table *table, nf_hookfn *fn) 1701 { 1702 unsigned int hook_mask = table->valid_hooks; 1703 uint8_t i, num_hooks = hweight32(hook_mask); 1704 uint8_t hooknum; 1705 struct nf_hook_ops *ops; 1706 1707 if (!num_hooks) 1708 return ERR_PTR(-EINVAL); 1709 1710 ops = kcalloc(num_hooks, sizeof(*ops), GFP_KERNEL); 1711 if (ops == NULL) 1712 return ERR_PTR(-ENOMEM); 1713 1714 for (i = 0, hooknum = 0; i < num_hooks && hook_mask != 0; 1715 hook_mask >>= 1, ++hooknum) { 1716 if (!(hook_mask & 1)) 1717 continue; 1718 ops[i].hook = fn; 1719 ops[i].pf = table->af; 1720 ops[i].hooknum = hooknum; 1721 ops[i].priority = table->priority; 1722 ++i; 1723 } 1724 1725 return ops; 1726 } 1727 EXPORT_SYMBOL_GPL(xt_hook_ops_alloc); 1728 1729 int xt_proto_init(struct net *net, u_int8_t af) 1730 { 1731 #ifdef CONFIG_PROC_FS 1732 char buf[XT_FUNCTION_MAXNAMELEN]; 1733 struct proc_dir_entry *proc; 1734 kuid_t root_uid; 1735 kgid_t root_gid; 1736 #endif 1737 1738 if (af >= ARRAY_SIZE(xt_prefix)) 1739 return -EINVAL; 1740 1741 1742 #ifdef CONFIG_PROC_FS 1743 root_uid = make_kuid(net->user_ns, 0); 1744 root_gid = make_kgid(net->user_ns, 0); 1745 1746 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1747 strlcat(buf, FORMAT_TABLES, sizeof(buf)); 1748 proc = proc_create_net_data(buf, 0440, net->proc_net, &xt_table_seq_ops, 1749 sizeof(struct seq_net_private), 1750 (void *)(unsigned long)af); 1751 if (!proc) 1752 goto out; 1753 if (uid_valid(root_uid) && gid_valid(root_gid)) 1754 proc_set_user(proc, root_uid, root_gid); 1755 1756 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1757 strlcat(buf, FORMAT_MATCHES, sizeof(buf)); 1758 proc = proc_create_seq_private(buf, 0440, net->proc_net, 1759 &xt_match_seq_ops, sizeof(struct nf_mttg_trav), 1760 (void *)(unsigned long)af); 1761 if (!proc) 1762 goto out_remove_tables; 1763 if (uid_valid(root_uid) && gid_valid(root_gid)) 1764 proc_set_user(proc, root_uid, root_gid); 1765 1766 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1767 strlcat(buf, FORMAT_TARGETS, sizeof(buf)); 1768 proc = proc_create_seq_private(buf, 0440, net->proc_net, 1769 &xt_target_seq_ops, sizeof(struct nf_mttg_trav), 1770 (void *)(unsigned long)af); 1771 if (!proc) 1772 goto out_remove_matches; 1773 if (uid_valid(root_uid) && gid_valid(root_gid)) 1774 proc_set_user(proc, root_uid, root_gid); 1775 #endif 1776 1777 return 0; 1778 1779 #ifdef CONFIG_PROC_FS 1780 out_remove_matches: 1781 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1782 strlcat(buf, FORMAT_MATCHES, sizeof(buf)); 1783 remove_proc_entry(buf, net->proc_net); 1784 1785 out_remove_tables: 1786 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1787 strlcat(buf, FORMAT_TABLES, sizeof(buf)); 1788 remove_proc_entry(buf, net->proc_net); 1789 out: 1790 return -1; 1791 #endif 1792 } 1793 EXPORT_SYMBOL_GPL(xt_proto_init); 1794 1795 void xt_proto_fini(struct net *net, u_int8_t af) 1796 { 1797 #ifdef CONFIG_PROC_FS 1798 char buf[XT_FUNCTION_MAXNAMELEN]; 1799 1800 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1801 strlcat(buf, FORMAT_TABLES, sizeof(buf)); 1802 remove_proc_entry(buf, net->proc_net); 1803 1804 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1805 strlcat(buf, FORMAT_TARGETS, sizeof(buf)); 1806 remove_proc_entry(buf, net->proc_net); 1807 1808 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1809 strlcat(buf, FORMAT_MATCHES, sizeof(buf)); 1810 remove_proc_entry(buf, net->proc_net); 1811 #endif /*CONFIG_PROC_FS*/ 1812 } 1813 EXPORT_SYMBOL_GPL(xt_proto_fini); 1814 1815 /** 1816 * xt_percpu_counter_alloc - allocate x_tables rule counter 1817 * 1818 * @state: pointer to xt_percpu allocation state 1819 * @counter: pointer to counter struct inside the ip(6)/arpt_entry struct 1820 * 1821 * On SMP, the packet counter [ ip(6)t_entry->counters.pcnt ] will then 1822 * contain the address of the real (percpu) counter. 1823 * 1824 * Rule evaluation needs to use xt_get_this_cpu_counter() helper 1825 * to fetch the real percpu counter. 1826 * 1827 * To speed up allocation and improve data locality, a 4kb block is 1828 * allocated. Freeing any counter may free an entire block, so all 1829 * counters allocated using the same state must be freed at the same 1830 * time. 1831 * 1832 * xt_percpu_counter_alloc_state contains the base address of the 1833 * allocated page and the current sub-offset. 1834 * 1835 * returns false on error. 1836 */ 1837 bool xt_percpu_counter_alloc(struct xt_percpu_counter_alloc_state *state, 1838 struct xt_counters *counter) 1839 { 1840 BUILD_BUG_ON(XT_PCPU_BLOCK_SIZE < (sizeof(*counter) * 2)); 1841 1842 if (nr_cpu_ids <= 1) 1843 return true; 1844 1845 if (!state->mem) { 1846 state->mem = __alloc_percpu(XT_PCPU_BLOCK_SIZE, 1847 XT_PCPU_BLOCK_SIZE); 1848 if (!state->mem) 1849 return false; 1850 } 1851 counter->pcnt = (__force unsigned long)(state->mem + state->off); 1852 state->off += sizeof(*counter); 1853 if (state->off > (XT_PCPU_BLOCK_SIZE - sizeof(*counter))) { 1854 state->mem = NULL; 1855 state->off = 0; 1856 } 1857 return true; 1858 } 1859 EXPORT_SYMBOL_GPL(xt_percpu_counter_alloc); 1860 1861 void xt_percpu_counter_free(struct xt_counters *counters) 1862 { 1863 unsigned long pcnt = counters->pcnt; 1864 1865 if (nr_cpu_ids > 1 && (pcnt & (XT_PCPU_BLOCK_SIZE - 1)) == 0) 1866 free_percpu((void __percpu *)pcnt); 1867 } 1868 EXPORT_SYMBOL_GPL(xt_percpu_counter_free); 1869 1870 static int __net_init xt_net_init(struct net *net) 1871 { 1872 int i; 1873 1874 for (i = 0; i < NFPROTO_NUMPROTO; i++) 1875 INIT_LIST_HEAD(&net->xt.tables[i]); 1876 return 0; 1877 } 1878 1879 static void __net_exit xt_net_exit(struct net *net) 1880 { 1881 int i; 1882 1883 for (i = 0; i < NFPROTO_NUMPROTO; i++) 1884 WARN_ON_ONCE(!list_empty(&net->xt.tables[i])); 1885 } 1886 1887 static struct pernet_operations xt_net_ops = { 1888 .init = xt_net_init, 1889 .exit = xt_net_exit, 1890 }; 1891 1892 static int __init xt_init(void) 1893 { 1894 unsigned int i; 1895 int rv; 1896 1897 for_each_possible_cpu(i) { 1898 seqcount_init(&per_cpu(xt_recseq, i)); 1899 } 1900 1901 xt = kcalloc(NFPROTO_NUMPROTO, sizeof(struct xt_af), GFP_KERNEL); 1902 if (!xt) 1903 return -ENOMEM; 1904 1905 for (i = 0; i < NFPROTO_NUMPROTO; i++) { 1906 mutex_init(&xt[i].mutex); 1907 #ifdef CONFIG_COMPAT 1908 mutex_init(&xt[i].compat_mutex); 1909 xt[i].compat_tab = NULL; 1910 #endif 1911 INIT_LIST_HEAD(&xt[i].target); 1912 INIT_LIST_HEAD(&xt[i].match); 1913 } 1914 rv = register_pernet_subsys(&xt_net_ops); 1915 if (rv < 0) 1916 kfree(xt); 1917 return rv; 1918 } 1919 1920 static void __exit xt_fini(void) 1921 { 1922 unregister_pernet_subsys(&xt_net_ops); 1923 kfree(xt); 1924 } 1925 1926 module_init(xt_init); 1927 module_exit(xt_fini); 1928 1929