1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * x_tables core - Backend for {ip,ip6,arp}_tables 4 * 5 * Copyright (C) 2006-2006 Harald Welte <laforge@netfilter.org> 6 * Copyright (C) 2006-2012 Patrick McHardy <kaber@trash.net> 7 * 8 * Based on existing ip_tables code which is 9 * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling 10 * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org> 11 */ 12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 13 #include <linux/kernel.h> 14 #include <linux/module.h> 15 #include <linux/socket.h> 16 #include <linux/net.h> 17 #include <linux/proc_fs.h> 18 #include <linux/seq_file.h> 19 #include <linux/string.h> 20 #include <linux/vmalloc.h> 21 #include <linux/mutex.h> 22 #include <linux/mm.h> 23 #include <linux/slab.h> 24 #include <linux/audit.h> 25 #include <linux/user_namespace.h> 26 #include <net/net_namespace.h> 27 #include <net/netns/generic.h> 28 29 #include <linux/netfilter/x_tables.h> 30 #include <linux/netfilter_arp.h> 31 #include <linux/netfilter_ipv4/ip_tables.h> 32 #include <linux/netfilter_ipv6/ip6_tables.h> 33 #include <linux/netfilter_arp/arp_tables.h> 34 35 MODULE_LICENSE("GPL"); 36 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>"); 37 MODULE_DESCRIPTION("{ip,ip6,arp,eb}_tables backend module"); 38 39 #define XT_PCPU_BLOCK_SIZE 4096 40 #define XT_MAX_TABLE_SIZE (512 * 1024 * 1024) 41 42 struct xt_template { 43 struct list_head list; 44 45 /* called when table is needed in the given netns */ 46 int (*table_init)(struct net *net); 47 48 struct module *me; 49 50 /* A unique name... */ 51 char name[XT_TABLE_MAXNAMELEN]; 52 }; 53 54 static struct list_head xt_templates[NFPROTO_NUMPROTO]; 55 56 struct xt_pernet { 57 struct list_head tables[NFPROTO_NUMPROTO]; 58 }; 59 60 struct compat_delta { 61 unsigned int offset; /* offset in kernel */ 62 int delta; /* delta in 32bit user land */ 63 }; 64 65 struct xt_af { 66 struct mutex mutex; 67 struct list_head match; 68 struct list_head target; 69 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT 70 struct mutex compat_mutex; 71 struct compat_delta *compat_tab; 72 unsigned int number; /* number of slots in compat_tab[] */ 73 unsigned int cur; /* number of used slots in compat_tab[] */ 74 #endif 75 }; 76 77 static unsigned int xt_pernet_id __read_mostly; 78 static struct xt_af *xt __read_mostly; 79 80 static const char *const xt_prefix[NFPROTO_NUMPROTO] = { 81 [NFPROTO_UNSPEC] = "x", 82 [NFPROTO_IPV4] = "ip", 83 [NFPROTO_ARP] = "arp", 84 [NFPROTO_BRIDGE] = "eb", 85 [NFPROTO_IPV6] = "ip6", 86 }; 87 88 /* Registration hooks for targets. */ 89 int xt_register_target(struct xt_target *target) 90 { 91 u_int8_t af = target->family; 92 93 mutex_lock(&xt[af].mutex); 94 list_add(&target->list, &xt[af].target); 95 mutex_unlock(&xt[af].mutex); 96 return 0; 97 } 98 EXPORT_SYMBOL(xt_register_target); 99 100 void 101 xt_unregister_target(struct xt_target *target) 102 { 103 u_int8_t af = target->family; 104 105 mutex_lock(&xt[af].mutex); 106 list_del(&target->list); 107 mutex_unlock(&xt[af].mutex); 108 } 109 EXPORT_SYMBOL(xt_unregister_target); 110 111 int 112 xt_register_targets(struct xt_target *target, unsigned int n) 113 { 114 unsigned int i; 115 int err = 0; 116 117 for (i = 0; i < n; i++) { 118 err = xt_register_target(&target[i]); 119 if (err) 120 goto err; 121 } 122 return err; 123 124 err: 125 if (i > 0) 126 xt_unregister_targets(target, i); 127 return err; 128 } 129 EXPORT_SYMBOL(xt_register_targets); 130 131 void 132 xt_unregister_targets(struct xt_target *target, unsigned int n) 133 { 134 while (n-- > 0) 135 xt_unregister_target(&target[n]); 136 } 137 EXPORT_SYMBOL(xt_unregister_targets); 138 139 int xt_register_match(struct xt_match *match) 140 { 141 u_int8_t af = match->family; 142 143 mutex_lock(&xt[af].mutex); 144 list_add(&match->list, &xt[af].match); 145 mutex_unlock(&xt[af].mutex); 146 return 0; 147 } 148 EXPORT_SYMBOL(xt_register_match); 149 150 void 151 xt_unregister_match(struct xt_match *match) 152 { 153 u_int8_t af = match->family; 154 155 mutex_lock(&xt[af].mutex); 156 list_del(&match->list); 157 mutex_unlock(&xt[af].mutex); 158 } 159 EXPORT_SYMBOL(xt_unregister_match); 160 161 int 162 xt_register_matches(struct xt_match *match, unsigned int n) 163 { 164 unsigned int i; 165 int err = 0; 166 167 for (i = 0; i < n; i++) { 168 err = xt_register_match(&match[i]); 169 if (err) 170 goto err; 171 } 172 return err; 173 174 err: 175 if (i > 0) 176 xt_unregister_matches(match, i); 177 return err; 178 } 179 EXPORT_SYMBOL(xt_register_matches); 180 181 void 182 xt_unregister_matches(struct xt_match *match, unsigned int n) 183 { 184 while (n-- > 0) 185 xt_unregister_match(&match[n]); 186 } 187 EXPORT_SYMBOL(xt_unregister_matches); 188 189 190 /* 191 * These are weird, but module loading must not be done with mutex 192 * held (since they will register), and we have to have a single 193 * function to use. 194 */ 195 196 /* Find match, grabs ref. Returns ERR_PTR() on error. */ 197 struct xt_match *xt_find_match(u8 af, const char *name, u8 revision) 198 { 199 struct xt_match *m; 200 int err = -ENOENT; 201 202 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN) 203 return ERR_PTR(-EINVAL); 204 205 mutex_lock(&xt[af].mutex); 206 list_for_each_entry(m, &xt[af].match, list) { 207 if (strcmp(m->name, name) == 0) { 208 if (m->revision == revision) { 209 if (try_module_get(m->me)) { 210 mutex_unlock(&xt[af].mutex); 211 return m; 212 } 213 } else 214 err = -EPROTOTYPE; /* Found something. */ 215 } 216 } 217 mutex_unlock(&xt[af].mutex); 218 219 if (af != NFPROTO_UNSPEC) 220 /* Try searching again in the family-independent list */ 221 return xt_find_match(NFPROTO_UNSPEC, name, revision); 222 223 return ERR_PTR(err); 224 } 225 EXPORT_SYMBOL(xt_find_match); 226 227 struct xt_match * 228 xt_request_find_match(uint8_t nfproto, const char *name, uint8_t revision) 229 { 230 struct xt_match *match; 231 232 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN) 233 return ERR_PTR(-EINVAL); 234 235 match = xt_find_match(nfproto, name, revision); 236 if (IS_ERR(match)) { 237 request_module("%st_%s", xt_prefix[nfproto], name); 238 match = xt_find_match(nfproto, name, revision); 239 } 240 241 return match; 242 } 243 EXPORT_SYMBOL_GPL(xt_request_find_match); 244 245 /* Find target, grabs ref. Returns ERR_PTR() on error. */ 246 static struct xt_target *xt_find_target(u8 af, const char *name, u8 revision) 247 { 248 struct xt_target *t; 249 int err = -ENOENT; 250 251 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN) 252 return ERR_PTR(-EINVAL); 253 254 mutex_lock(&xt[af].mutex); 255 list_for_each_entry(t, &xt[af].target, list) { 256 if (strcmp(t->name, name) == 0) { 257 if (t->revision == revision) { 258 if (try_module_get(t->me)) { 259 mutex_unlock(&xt[af].mutex); 260 return t; 261 } 262 } else 263 err = -EPROTOTYPE; /* Found something. */ 264 } 265 } 266 mutex_unlock(&xt[af].mutex); 267 268 if (af != NFPROTO_UNSPEC) 269 /* Try searching again in the family-independent list */ 270 return xt_find_target(NFPROTO_UNSPEC, name, revision); 271 272 return ERR_PTR(err); 273 } 274 275 struct xt_target *xt_request_find_target(u8 af, const char *name, u8 revision) 276 { 277 struct xt_target *target; 278 279 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN) 280 return ERR_PTR(-EINVAL); 281 282 target = xt_find_target(af, name, revision); 283 if (IS_ERR(target)) { 284 request_module("%st_%s", xt_prefix[af], name); 285 target = xt_find_target(af, name, revision); 286 } 287 288 return target; 289 } 290 EXPORT_SYMBOL_GPL(xt_request_find_target); 291 292 293 static int xt_obj_to_user(u16 __user *psize, u16 size, 294 void __user *pname, const char *name, 295 u8 __user *prev, u8 rev) 296 { 297 if (put_user(size, psize)) 298 return -EFAULT; 299 if (copy_to_user(pname, name, strlen(name) + 1)) 300 return -EFAULT; 301 if (put_user(rev, prev)) 302 return -EFAULT; 303 304 return 0; 305 } 306 307 #define XT_OBJ_TO_USER(U, K, TYPE, C_SIZE) \ 308 xt_obj_to_user(&U->u.TYPE##_size, C_SIZE ? : K->u.TYPE##_size, \ 309 U->u.user.name, K->u.kernel.TYPE->name, \ 310 &U->u.user.revision, K->u.kernel.TYPE->revision) 311 312 int xt_data_to_user(void __user *dst, const void *src, 313 int usersize, int size, int aligned_size) 314 { 315 usersize = usersize ? : size; 316 if (copy_to_user(dst, src, usersize)) 317 return -EFAULT; 318 if (usersize != aligned_size && 319 clear_user(dst + usersize, aligned_size - usersize)) 320 return -EFAULT; 321 322 return 0; 323 } 324 EXPORT_SYMBOL_GPL(xt_data_to_user); 325 326 #define XT_DATA_TO_USER(U, K, TYPE) \ 327 xt_data_to_user(U->data, K->data, \ 328 K->u.kernel.TYPE->usersize, \ 329 K->u.kernel.TYPE->TYPE##size, \ 330 XT_ALIGN(K->u.kernel.TYPE->TYPE##size)) 331 332 int xt_match_to_user(const struct xt_entry_match *m, 333 struct xt_entry_match __user *u) 334 { 335 return XT_OBJ_TO_USER(u, m, match, 0) || 336 XT_DATA_TO_USER(u, m, match); 337 } 338 EXPORT_SYMBOL_GPL(xt_match_to_user); 339 340 int xt_target_to_user(const struct xt_entry_target *t, 341 struct xt_entry_target __user *u) 342 { 343 return XT_OBJ_TO_USER(u, t, target, 0) || 344 XT_DATA_TO_USER(u, t, target); 345 } 346 EXPORT_SYMBOL_GPL(xt_target_to_user); 347 348 static int match_revfn(u8 af, const char *name, u8 revision, int *bestp) 349 { 350 const struct xt_match *m; 351 int have_rev = 0; 352 353 mutex_lock(&xt[af].mutex); 354 list_for_each_entry(m, &xt[af].match, list) { 355 if (strcmp(m->name, name) == 0) { 356 if (m->revision > *bestp) 357 *bestp = m->revision; 358 if (m->revision == revision) 359 have_rev = 1; 360 } 361 } 362 mutex_unlock(&xt[af].mutex); 363 364 if (af != NFPROTO_UNSPEC && !have_rev) 365 return match_revfn(NFPROTO_UNSPEC, name, revision, bestp); 366 367 return have_rev; 368 } 369 370 static int target_revfn(u8 af, const char *name, u8 revision, int *bestp) 371 { 372 const struct xt_target *t; 373 int have_rev = 0; 374 375 mutex_lock(&xt[af].mutex); 376 list_for_each_entry(t, &xt[af].target, list) { 377 if (strcmp(t->name, name) == 0) { 378 if (t->revision > *bestp) 379 *bestp = t->revision; 380 if (t->revision == revision) 381 have_rev = 1; 382 } 383 } 384 mutex_unlock(&xt[af].mutex); 385 386 if (af != NFPROTO_UNSPEC && !have_rev) 387 return target_revfn(NFPROTO_UNSPEC, name, revision, bestp); 388 389 return have_rev; 390 } 391 392 /* Returns true or false (if no such extension at all) */ 393 int xt_find_revision(u8 af, const char *name, u8 revision, int target, 394 int *err) 395 { 396 int have_rev, best = -1; 397 398 if (target == 1) 399 have_rev = target_revfn(af, name, revision, &best); 400 else 401 have_rev = match_revfn(af, name, revision, &best); 402 403 /* Nothing at all? Return 0 to try loading module. */ 404 if (best == -1) { 405 *err = -ENOENT; 406 return 0; 407 } 408 409 *err = best; 410 if (!have_rev) 411 *err = -EPROTONOSUPPORT; 412 return 1; 413 } 414 EXPORT_SYMBOL_GPL(xt_find_revision); 415 416 static char * 417 textify_hooks(char *buf, size_t size, unsigned int mask, uint8_t nfproto) 418 { 419 static const char *const inetbr_names[] = { 420 "PREROUTING", "INPUT", "FORWARD", 421 "OUTPUT", "POSTROUTING", "BROUTING", 422 }; 423 static const char *const arp_names[] = { 424 "INPUT", "FORWARD", "OUTPUT", 425 }; 426 const char *const *names; 427 unsigned int i, max; 428 char *p = buf; 429 bool np = false; 430 int res; 431 432 names = (nfproto == NFPROTO_ARP) ? arp_names : inetbr_names; 433 max = (nfproto == NFPROTO_ARP) ? ARRAY_SIZE(arp_names) : 434 ARRAY_SIZE(inetbr_names); 435 *p = '\0'; 436 for (i = 0; i < max; ++i) { 437 if (!(mask & (1 << i))) 438 continue; 439 res = snprintf(p, size, "%s%s", np ? "/" : "", names[i]); 440 if (res > 0) { 441 size -= res; 442 p += res; 443 } 444 np = true; 445 } 446 447 return buf; 448 } 449 450 /** 451 * xt_check_proc_name - check that name is suitable for /proc file creation 452 * 453 * @name: file name candidate 454 * @size: length of buffer 455 * 456 * some x_tables modules wish to create a file in /proc. 457 * This function makes sure that the name is suitable for this 458 * purpose, it checks that name is NUL terminated and isn't a 'special' 459 * name, like "..". 460 * 461 * returns negative number on error or 0 if name is useable. 462 */ 463 int xt_check_proc_name(const char *name, unsigned int size) 464 { 465 if (name[0] == '\0') 466 return -EINVAL; 467 468 if (strnlen(name, size) == size) 469 return -ENAMETOOLONG; 470 471 if (strcmp(name, ".") == 0 || 472 strcmp(name, "..") == 0 || 473 strchr(name, '/')) 474 return -EINVAL; 475 476 return 0; 477 } 478 EXPORT_SYMBOL(xt_check_proc_name); 479 480 int xt_check_match(struct xt_mtchk_param *par, 481 unsigned int size, u16 proto, bool inv_proto) 482 { 483 int ret; 484 485 if (XT_ALIGN(par->match->matchsize) != size && 486 par->match->matchsize != -1) { 487 /* 488 * ebt_among is exempt from centralized matchsize checking 489 * because it uses a dynamic-size data set. 490 */ 491 pr_err_ratelimited("%s_tables: %s.%u match: invalid size %u (kernel) != (user) %u\n", 492 xt_prefix[par->family], par->match->name, 493 par->match->revision, 494 XT_ALIGN(par->match->matchsize), size); 495 return -EINVAL; 496 } 497 if (par->match->table != NULL && 498 strcmp(par->match->table, par->table) != 0) { 499 pr_info_ratelimited("%s_tables: %s match: only valid in %s table, not %s\n", 500 xt_prefix[par->family], par->match->name, 501 par->match->table, par->table); 502 return -EINVAL; 503 } 504 if (par->match->hooks && (par->hook_mask & ~par->match->hooks) != 0) { 505 char used[64], allow[64]; 506 507 pr_info_ratelimited("%s_tables: %s match: used from hooks %s, but only valid from %s\n", 508 xt_prefix[par->family], par->match->name, 509 textify_hooks(used, sizeof(used), 510 par->hook_mask, par->family), 511 textify_hooks(allow, sizeof(allow), 512 par->match->hooks, 513 par->family)); 514 return -EINVAL; 515 } 516 if (par->match->proto && (par->match->proto != proto || inv_proto)) { 517 pr_info_ratelimited("%s_tables: %s match: only valid for protocol %u\n", 518 xt_prefix[par->family], par->match->name, 519 par->match->proto); 520 return -EINVAL; 521 } 522 if (par->match->checkentry != NULL) { 523 ret = par->match->checkentry(par); 524 if (ret < 0) 525 return ret; 526 else if (ret > 0) 527 /* Flag up potential errors. */ 528 return -EIO; 529 } 530 return 0; 531 } 532 EXPORT_SYMBOL_GPL(xt_check_match); 533 534 /** xt_check_entry_match - check that matches end before start of target 535 * 536 * @match: beginning of xt_entry_match 537 * @target: beginning of this rules target (alleged end of matches) 538 * @alignment: alignment requirement of match structures 539 * 540 * Validates that all matches add up to the beginning of the target, 541 * and that each match covers at least the base structure size. 542 * 543 * Return: 0 on success, negative errno on failure. 544 */ 545 static int xt_check_entry_match(const char *match, const char *target, 546 const size_t alignment) 547 { 548 const struct xt_entry_match *pos; 549 int length = target - match; 550 551 if (length == 0) /* no matches */ 552 return 0; 553 554 pos = (struct xt_entry_match *)match; 555 do { 556 if ((unsigned long)pos % alignment) 557 return -EINVAL; 558 559 if (length < (int)sizeof(struct xt_entry_match)) 560 return -EINVAL; 561 562 if (pos->u.match_size < sizeof(struct xt_entry_match)) 563 return -EINVAL; 564 565 if (pos->u.match_size > length) 566 return -EINVAL; 567 568 length -= pos->u.match_size; 569 pos = ((void *)((char *)(pos) + (pos)->u.match_size)); 570 } while (length > 0); 571 572 return 0; 573 } 574 575 /** xt_check_table_hooks - check hook entry points are sane 576 * 577 * @info xt_table_info to check 578 * @valid_hooks - hook entry points that we can enter from 579 * 580 * Validates that the hook entry and underflows points are set up. 581 * 582 * Return: 0 on success, negative errno on failure. 583 */ 584 int xt_check_table_hooks(const struct xt_table_info *info, unsigned int valid_hooks) 585 { 586 const char *err = "unsorted underflow"; 587 unsigned int i, max_uflow, max_entry; 588 bool check_hooks = false; 589 590 BUILD_BUG_ON(ARRAY_SIZE(info->hook_entry) != ARRAY_SIZE(info->underflow)); 591 592 max_entry = 0; 593 max_uflow = 0; 594 595 for (i = 0; i < ARRAY_SIZE(info->hook_entry); i++) { 596 if (!(valid_hooks & (1 << i))) 597 continue; 598 599 if (info->hook_entry[i] == 0xFFFFFFFF) 600 return -EINVAL; 601 if (info->underflow[i] == 0xFFFFFFFF) 602 return -EINVAL; 603 604 if (check_hooks) { 605 if (max_uflow > info->underflow[i]) 606 goto error; 607 608 if (max_uflow == info->underflow[i]) { 609 err = "duplicate underflow"; 610 goto error; 611 } 612 if (max_entry > info->hook_entry[i]) { 613 err = "unsorted entry"; 614 goto error; 615 } 616 if (max_entry == info->hook_entry[i]) { 617 err = "duplicate entry"; 618 goto error; 619 } 620 } 621 max_entry = info->hook_entry[i]; 622 max_uflow = info->underflow[i]; 623 check_hooks = true; 624 } 625 626 return 0; 627 error: 628 pr_err_ratelimited("%s at hook %d\n", err, i); 629 return -EINVAL; 630 } 631 EXPORT_SYMBOL(xt_check_table_hooks); 632 633 static bool verdict_ok(int verdict) 634 { 635 if (verdict > 0) 636 return true; 637 638 if (verdict < 0) { 639 int v = -verdict - 1; 640 641 if (verdict == XT_RETURN) 642 return true; 643 644 switch (v) { 645 case NF_ACCEPT: return true; 646 case NF_DROP: return true; 647 case NF_QUEUE: return true; 648 default: 649 break; 650 } 651 652 return false; 653 } 654 655 return false; 656 } 657 658 static bool error_tg_ok(unsigned int usersize, unsigned int kernsize, 659 const char *msg, unsigned int msglen) 660 { 661 return usersize == kernsize && strnlen(msg, msglen) < msglen; 662 } 663 664 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT 665 int xt_compat_add_offset(u_int8_t af, unsigned int offset, int delta) 666 { 667 struct xt_af *xp = &xt[af]; 668 669 WARN_ON(!mutex_is_locked(&xt[af].compat_mutex)); 670 671 if (WARN_ON(!xp->compat_tab)) 672 return -ENOMEM; 673 674 if (xp->cur >= xp->number) 675 return -EINVAL; 676 677 if (xp->cur) 678 delta += xp->compat_tab[xp->cur - 1].delta; 679 xp->compat_tab[xp->cur].offset = offset; 680 xp->compat_tab[xp->cur].delta = delta; 681 xp->cur++; 682 return 0; 683 } 684 EXPORT_SYMBOL_GPL(xt_compat_add_offset); 685 686 void xt_compat_flush_offsets(u_int8_t af) 687 { 688 WARN_ON(!mutex_is_locked(&xt[af].compat_mutex)); 689 690 if (xt[af].compat_tab) { 691 vfree(xt[af].compat_tab); 692 xt[af].compat_tab = NULL; 693 xt[af].number = 0; 694 xt[af].cur = 0; 695 } 696 } 697 EXPORT_SYMBOL_GPL(xt_compat_flush_offsets); 698 699 int xt_compat_calc_jump(u_int8_t af, unsigned int offset) 700 { 701 struct compat_delta *tmp = xt[af].compat_tab; 702 int mid, left = 0, right = xt[af].cur - 1; 703 704 while (left <= right) { 705 mid = (left + right) >> 1; 706 if (offset > tmp[mid].offset) 707 left = mid + 1; 708 else if (offset < tmp[mid].offset) 709 right = mid - 1; 710 else 711 return mid ? tmp[mid - 1].delta : 0; 712 } 713 return left ? tmp[left - 1].delta : 0; 714 } 715 EXPORT_SYMBOL_GPL(xt_compat_calc_jump); 716 717 int xt_compat_init_offsets(u8 af, unsigned int number) 718 { 719 size_t mem; 720 721 WARN_ON(!mutex_is_locked(&xt[af].compat_mutex)); 722 723 if (!number || number > (INT_MAX / sizeof(struct compat_delta))) 724 return -EINVAL; 725 726 if (WARN_ON(xt[af].compat_tab)) 727 return -EINVAL; 728 729 mem = sizeof(struct compat_delta) * number; 730 if (mem > XT_MAX_TABLE_SIZE) 731 return -ENOMEM; 732 733 xt[af].compat_tab = vmalloc(mem); 734 if (!xt[af].compat_tab) 735 return -ENOMEM; 736 737 xt[af].number = number; 738 xt[af].cur = 0; 739 740 return 0; 741 } 742 EXPORT_SYMBOL(xt_compat_init_offsets); 743 744 int xt_compat_match_offset(const struct xt_match *match) 745 { 746 u_int16_t csize = match->compatsize ? : match->matchsize; 747 return XT_ALIGN(match->matchsize) - COMPAT_XT_ALIGN(csize); 748 } 749 EXPORT_SYMBOL_GPL(xt_compat_match_offset); 750 751 void xt_compat_match_from_user(struct xt_entry_match *m, void **dstptr, 752 unsigned int *size) 753 { 754 const struct xt_match *match = m->u.kernel.match; 755 struct compat_xt_entry_match *cm = (struct compat_xt_entry_match *)m; 756 int off = xt_compat_match_offset(match); 757 u_int16_t msize = cm->u.user.match_size; 758 char name[sizeof(m->u.user.name)]; 759 760 m = *dstptr; 761 memcpy(m, cm, sizeof(*cm)); 762 if (match->compat_from_user) 763 match->compat_from_user(m->data, cm->data); 764 else 765 memcpy(m->data, cm->data, msize - sizeof(*cm)); 766 767 msize += off; 768 m->u.user.match_size = msize; 769 strscpy(name, match->name, sizeof(name)); 770 module_put(match->me); 771 strscpy_pad(m->u.user.name, name, sizeof(m->u.user.name)); 772 773 *size += off; 774 *dstptr += msize; 775 } 776 EXPORT_SYMBOL_GPL(xt_compat_match_from_user); 777 778 #define COMPAT_XT_DATA_TO_USER(U, K, TYPE, C_SIZE) \ 779 xt_data_to_user(U->data, K->data, \ 780 K->u.kernel.TYPE->usersize, \ 781 C_SIZE, \ 782 COMPAT_XT_ALIGN(C_SIZE)) 783 784 int xt_compat_match_to_user(const struct xt_entry_match *m, 785 void __user **dstptr, unsigned int *size) 786 { 787 const struct xt_match *match = m->u.kernel.match; 788 struct compat_xt_entry_match __user *cm = *dstptr; 789 int off = xt_compat_match_offset(match); 790 u_int16_t msize = m->u.user.match_size - off; 791 792 if (XT_OBJ_TO_USER(cm, m, match, msize)) 793 return -EFAULT; 794 795 if (match->compat_to_user) { 796 if (match->compat_to_user((void __user *)cm->data, m->data)) 797 return -EFAULT; 798 } else { 799 if (COMPAT_XT_DATA_TO_USER(cm, m, match, msize - sizeof(*cm))) 800 return -EFAULT; 801 } 802 803 *size -= off; 804 *dstptr += msize; 805 return 0; 806 } 807 EXPORT_SYMBOL_GPL(xt_compat_match_to_user); 808 809 /* non-compat version may have padding after verdict */ 810 struct compat_xt_standard_target { 811 struct compat_xt_entry_target t; 812 compat_uint_t verdict; 813 }; 814 815 struct compat_xt_error_target { 816 struct compat_xt_entry_target t; 817 char errorname[XT_FUNCTION_MAXNAMELEN]; 818 }; 819 820 int xt_compat_check_entry_offsets(const void *base, const char *elems, 821 unsigned int target_offset, 822 unsigned int next_offset) 823 { 824 long size_of_base_struct = elems - (const char *)base; 825 const struct compat_xt_entry_target *t; 826 const char *e = base; 827 828 if (target_offset < size_of_base_struct) 829 return -EINVAL; 830 831 if (target_offset + sizeof(*t) > next_offset) 832 return -EINVAL; 833 834 t = (void *)(e + target_offset); 835 if (t->u.target_size < sizeof(*t)) 836 return -EINVAL; 837 838 if (target_offset + t->u.target_size > next_offset) 839 return -EINVAL; 840 841 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0) { 842 const struct compat_xt_standard_target *st = (const void *)t; 843 844 if (COMPAT_XT_ALIGN(target_offset + sizeof(*st)) != next_offset) 845 return -EINVAL; 846 847 if (!verdict_ok(st->verdict)) 848 return -EINVAL; 849 } else if (strcmp(t->u.user.name, XT_ERROR_TARGET) == 0) { 850 const struct compat_xt_error_target *et = (const void *)t; 851 852 if (!error_tg_ok(t->u.target_size, sizeof(*et), 853 et->errorname, sizeof(et->errorname))) 854 return -EINVAL; 855 } 856 857 /* compat_xt_entry match has less strict alignment requirements, 858 * otherwise they are identical. In case of padding differences 859 * we need to add compat version of xt_check_entry_match. 860 */ 861 BUILD_BUG_ON(sizeof(struct compat_xt_entry_match) != sizeof(struct xt_entry_match)); 862 863 return xt_check_entry_match(elems, base + target_offset, 864 __alignof__(struct compat_xt_entry_match)); 865 } 866 EXPORT_SYMBOL(xt_compat_check_entry_offsets); 867 #endif /* CONFIG_NETFILTER_XTABLES_COMPAT */ 868 869 /** 870 * xt_check_entry_offsets - validate arp/ip/ip6t_entry 871 * 872 * @base: pointer to arp/ip/ip6t_entry 873 * @elems: pointer to first xt_entry_match, i.e. ip(6)t_entry->elems 874 * @target_offset: the arp/ip/ip6_t->target_offset 875 * @next_offset: the arp/ip/ip6_t->next_offset 876 * 877 * validates that target_offset and next_offset are sane and that all 878 * match sizes (if any) align with the target offset. 879 * 880 * This function does not validate the targets or matches themselves, it 881 * only tests that all the offsets and sizes are correct, that all 882 * match structures are aligned, and that the last structure ends where 883 * the target structure begins. 884 * 885 * Also see xt_compat_check_entry_offsets for CONFIG_NETFILTER_XTABLES_COMPAT version. 886 * 887 * The arp/ip/ip6t_entry structure @base must have passed following tests: 888 * - it must point to a valid memory location 889 * - base to base + next_offset must be accessible, i.e. not exceed allocated 890 * length. 891 * 892 * A well-formed entry looks like this: 893 * 894 * ip(6)t_entry match [mtdata] match [mtdata] target [tgdata] ip(6)t_entry 895 * e->elems[]-----' | | 896 * matchsize | | 897 * matchsize | | 898 * | | 899 * target_offset---------------------------------' | 900 * next_offset---------------------------------------------------' 901 * 902 * elems[]: flexible array member at end of ip(6)/arpt_entry struct. 903 * This is where matches (if any) and the target reside. 904 * target_offset: beginning of target. 905 * next_offset: start of the next rule; also: size of this rule. 906 * Since targets have a minimum size, target_offset + minlen <= next_offset. 907 * 908 * Every match stores its size, sum of sizes must not exceed target_offset. 909 * 910 * Return: 0 on success, negative errno on failure. 911 */ 912 int xt_check_entry_offsets(const void *base, 913 const char *elems, 914 unsigned int target_offset, 915 unsigned int next_offset) 916 { 917 long size_of_base_struct = elems - (const char *)base; 918 const struct xt_entry_target *t; 919 const char *e = base; 920 921 /* target start is within the ip/ip6/arpt_entry struct */ 922 if (target_offset < size_of_base_struct) 923 return -EINVAL; 924 925 if (target_offset + sizeof(*t) > next_offset) 926 return -EINVAL; 927 928 t = (void *)(e + target_offset); 929 if (t->u.target_size < sizeof(*t)) 930 return -EINVAL; 931 932 if (target_offset + t->u.target_size > next_offset) 933 return -EINVAL; 934 935 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0) { 936 const struct xt_standard_target *st = (const void *)t; 937 938 if (XT_ALIGN(target_offset + sizeof(*st)) != next_offset) 939 return -EINVAL; 940 941 if (!verdict_ok(st->verdict)) 942 return -EINVAL; 943 } else if (strcmp(t->u.user.name, XT_ERROR_TARGET) == 0) { 944 const struct xt_error_target *et = (const void *)t; 945 946 if (!error_tg_ok(t->u.target_size, sizeof(*et), 947 et->errorname, sizeof(et->errorname))) 948 return -EINVAL; 949 } 950 951 return xt_check_entry_match(elems, base + target_offset, 952 __alignof__(struct xt_entry_match)); 953 } 954 EXPORT_SYMBOL(xt_check_entry_offsets); 955 956 /** 957 * xt_alloc_entry_offsets - allocate array to store rule head offsets 958 * 959 * @size: number of entries 960 * 961 * Return: NULL or zeroed kmalloc'd or vmalloc'd array 962 */ 963 unsigned int *xt_alloc_entry_offsets(unsigned int size) 964 { 965 if (size > XT_MAX_TABLE_SIZE / sizeof(unsigned int)) 966 return NULL; 967 968 return kvcalloc(size, sizeof(unsigned int), GFP_KERNEL); 969 970 } 971 EXPORT_SYMBOL(xt_alloc_entry_offsets); 972 973 /** 974 * xt_find_jump_offset - check if target is a valid jump offset 975 * 976 * @offsets: array containing all valid rule start offsets of a rule blob 977 * @target: the jump target to search for 978 * @size: entries in @offset 979 */ 980 bool xt_find_jump_offset(const unsigned int *offsets, 981 unsigned int target, unsigned int size) 982 { 983 int m, low = 0, hi = size; 984 985 while (hi > low) { 986 m = (low + hi) / 2u; 987 988 if (offsets[m] > target) 989 hi = m; 990 else if (offsets[m] < target) 991 low = m + 1; 992 else 993 return true; 994 } 995 996 return false; 997 } 998 EXPORT_SYMBOL(xt_find_jump_offset); 999 1000 int xt_check_target(struct xt_tgchk_param *par, 1001 unsigned int size, u16 proto, bool inv_proto) 1002 { 1003 int ret; 1004 1005 if (XT_ALIGN(par->target->targetsize) != size) { 1006 pr_err_ratelimited("%s_tables: %s.%u target: invalid size %u (kernel) != (user) %u\n", 1007 xt_prefix[par->family], par->target->name, 1008 par->target->revision, 1009 XT_ALIGN(par->target->targetsize), size); 1010 return -EINVAL; 1011 } 1012 if (par->target->table != NULL && 1013 strcmp(par->target->table, par->table) != 0) { 1014 pr_info_ratelimited("%s_tables: %s target: only valid in %s table, not %s\n", 1015 xt_prefix[par->family], par->target->name, 1016 par->target->table, par->table); 1017 return -EINVAL; 1018 } 1019 if (par->target->hooks && (par->hook_mask & ~par->target->hooks) != 0) { 1020 char used[64], allow[64]; 1021 1022 pr_info_ratelimited("%s_tables: %s target: used from hooks %s, but only usable from %s\n", 1023 xt_prefix[par->family], par->target->name, 1024 textify_hooks(used, sizeof(used), 1025 par->hook_mask, par->family), 1026 textify_hooks(allow, sizeof(allow), 1027 par->target->hooks, 1028 par->family)); 1029 return -EINVAL; 1030 } 1031 if (par->target->proto && (par->target->proto != proto || inv_proto)) { 1032 pr_info_ratelimited("%s_tables: %s target: only valid for protocol %u\n", 1033 xt_prefix[par->family], par->target->name, 1034 par->target->proto); 1035 return -EINVAL; 1036 } 1037 if (par->target->checkentry != NULL) { 1038 ret = par->target->checkentry(par); 1039 if (ret < 0) 1040 return ret; 1041 else if (ret > 0) 1042 /* Flag up potential errors. */ 1043 return -EIO; 1044 } 1045 return 0; 1046 } 1047 EXPORT_SYMBOL_GPL(xt_check_target); 1048 1049 /** 1050 * xt_copy_counters - copy counters and metadata from a sockptr_t 1051 * 1052 * @arg: src sockptr 1053 * @len: alleged size of userspace memory 1054 * @info: where to store the xt_counters_info metadata 1055 * 1056 * Copies counter meta data from @user and stores it in @info. 1057 * 1058 * vmallocs memory to hold the counters, then copies the counter data 1059 * from @user to the new memory and returns a pointer to it. 1060 * 1061 * If called from a compat syscall, @info gets converted automatically to the 1062 * 64bit representation. 1063 * 1064 * The metadata associated with the counters is stored in @info. 1065 * 1066 * Return: returns pointer that caller has to test via IS_ERR(). 1067 * If IS_ERR is false, caller has to vfree the pointer. 1068 */ 1069 void *xt_copy_counters(sockptr_t arg, unsigned int len, 1070 struct xt_counters_info *info) 1071 { 1072 size_t offset; 1073 void *mem; 1074 u64 size; 1075 1076 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT 1077 if (in_compat_syscall()) { 1078 /* structures only differ in size due to alignment */ 1079 struct compat_xt_counters_info compat_tmp; 1080 1081 if (len <= sizeof(compat_tmp)) 1082 return ERR_PTR(-EINVAL); 1083 1084 len -= sizeof(compat_tmp); 1085 if (copy_from_sockptr(&compat_tmp, arg, sizeof(compat_tmp)) != 0) 1086 return ERR_PTR(-EFAULT); 1087 1088 memcpy(info->name, compat_tmp.name, sizeof(info->name) - 1); 1089 info->num_counters = compat_tmp.num_counters; 1090 offset = sizeof(compat_tmp); 1091 } else 1092 #endif 1093 { 1094 if (len <= sizeof(*info)) 1095 return ERR_PTR(-EINVAL); 1096 1097 len -= sizeof(*info); 1098 if (copy_from_sockptr(info, arg, sizeof(*info)) != 0) 1099 return ERR_PTR(-EFAULT); 1100 1101 offset = sizeof(*info); 1102 } 1103 info->name[sizeof(info->name) - 1] = '\0'; 1104 1105 size = sizeof(struct xt_counters); 1106 size *= info->num_counters; 1107 1108 if (size != (u64)len) 1109 return ERR_PTR(-EINVAL); 1110 1111 mem = vmalloc(len); 1112 if (!mem) 1113 return ERR_PTR(-ENOMEM); 1114 1115 if (copy_from_sockptr_offset(mem, arg, offset, len) == 0) 1116 return mem; 1117 1118 vfree(mem); 1119 return ERR_PTR(-EFAULT); 1120 } 1121 EXPORT_SYMBOL_GPL(xt_copy_counters); 1122 1123 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT 1124 int xt_compat_target_offset(const struct xt_target *target) 1125 { 1126 u_int16_t csize = target->compatsize ? : target->targetsize; 1127 return XT_ALIGN(target->targetsize) - COMPAT_XT_ALIGN(csize); 1128 } 1129 EXPORT_SYMBOL_GPL(xt_compat_target_offset); 1130 1131 void xt_compat_target_from_user(struct xt_entry_target *t, void **dstptr, 1132 unsigned int *size) 1133 { 1134 const struct xt_target *target = t->u.kernel.target; 1135 struct compat_xt_entry_target *ct = (struct compat_xt_entry_target *)t; 1136 int off = xt_compat_target_offset(target); 1137 u_int16_t tsize = ct->u.user.target_size; 1138 char name[sizeof(t->u.user.name)]; 1139 1140 t = *dstptr; 1141 memcpy(t, ct, sizeof(*ct)); 1142 if (target->compat_from_user) 1143 target->compat_from_user(t->data, ct->data); 1144 else 1145 memcpy(t->data, ct->data, tsize - sizeof(*ct)); 1146 1147 tsize += off; 1148 t->u.user.target_size = tsize; 1149 strscpy(name, target->name, sizeof(name)); 1150 module_put(target->me); 1151 strscpy_pad(t->u.user.name, name, sizeof(t->u.user.name)); 1152 1153 *size += off; 1154 *dstptr += tsize; 1155 } 1156 EXPORT_SYMBOL_GPL(xt_compat_target_from_user); 1157 1158 int xt_compat_target_to_user(const struct xt_entry_target *t, 1159 void __user **dstptr, unsigned int *size) 1160 { 1161 const struct xt_target *target = t->u.kernel.target; 1162 struct compat_xt_entry_target __user *ct = *dstptr; 1163 int off = xt_compat_target_offset(target); 1164 u_int16_t tsize = t->u.user.target_size - off; 1165 1166 if (XT_OBJ_TO_USER(ct, t, target, tsize)) 1167 return -EFAULT; 1168 1169 if (target->compat_to_user) { 1170 if (target->compat_to_user((void __user *)ct->data, t->data)) 1171 return -EFAULT; 1172 } else { 1173 if (COMPAT_XT_DATA_TO_USER(ct, t, target, tsize - sizeof(*ct))) 1174 return -EFAULT; 1175 } 1176 1177 *size -= off; 1178 *dstptr += tsize; 1179 return 0; 1180 } 1181 EXPORT_SYMBOL_GPL(xt_compat_target_to_user); 1182 #endif 1183 1184 struct xt_table_info *xt_alloc_table_info(unsigned int size) 1185 { 1186 struct xt_table_info *info = NULL; 1187 size_t sz = sizeof(*info) + size; 1188 1189 if (sz < sizeof(*info) || sz >= XT_MAX_TABLE_SIZE) 1190 return NULL; 1191 1192 info = kvmalloc(sz, GFP_KERNEL_ACCOUNT); 1193 if (!info) 1194 return NULL; 1195 1196 memset(info, 0, sizeof(*info)); 1197 info->size = size; 1198 return info; 1199 } 1200 EXPORT_SYMBOL(xt_alloc_table_info); 1201 1202 void xt_free_table_info(struct xt_table_info *info) 1203 { 1204 int cpu; 1205 1206 if (info->jumpstack != NULL) { 1207 for_each_possible_cpu(cpu) 1208 kvfree(info->jumpstack[cpu]); 1209 kvfree(info->jumpstack); 1210 } 1211 1212 kvfree(info); 1213 } 1214 EXPORT_SYMBOL(xt_free_table_info); 1215 1216 struct xt_table *xt_find_table(struct net *net, u8 af, const char *name) 1217 { 1218 struct xt_pernet *xt_net = net_generic(net, xt_pernet_id); 1219 struct xt_table *t; 1220 1221 mutex_lock(&xt[af].mutex); 1222 list_for_each_entry(t, &xt_net->tables[af], list) { 1223 if (strcmp(t->name, name) == 0) { 1224 mutex_unlock(&xt[af].mutex); 1225 return t; 1226 } 1227 } 1228 mutex_unlock(&xt[af].mutex); 1229 return NULL; 1230 } 1231 EXPORT_SYMBOL(xt_find_table); 1232 1233 /* Find table by name, grabs mutex & ref. Returns ERR_PTR on error. */ 1234 struct xt_table *xt_find_table_lock(struct net *net, u_int8_t af, 1235 const char *name) 1236 { 1237 struct xt_pernet *xt_net = net_generic(net, xt_pernet_id); 1238 struct module *owner = NULL; 1239 struct xt_template *tmpl; 1240 struct xt_table *t; 1241 1242 mutex_lock(&xt[af].mutex); 1243 list_for_each_entry(t, &xt_net->tables[af], list) 1244 if (strcmp(t->name, name) == 0 && try_module_get(t->me)) 1245 return t; 1246 1247 /* Table doesn't exist in this netns, check larval list */ 1248 list_for_each_entry(tmpl, &xt_templates[af], list) { 1249 int err; 1250 1251 if (strcmp(tmpl->name, name)) 1252 continue; 1253 if (!try_module_get(tmpl->me)) 1254 goto out; 1255 1256 owner = tmpl->me; 1257 1258 mutex_unlock(&xt[af].mutex); 1259 err = tmpl->table_init(net); 1260 if (err < 0) { 1261 module_put(owner); 1262 return ERR_PTR(err); 1263 } 1264 1265 mutex_lock(&xt[af].mutex); 1266 break; 1267 } 1268 1269 /* and once again: */ 1270 list_for_each_entry(t, &xt_net->tables[af], list) 1271 if (strcmp(t->name, name) == 0) 1272 return t; 1273 1274 module_put(owner); 1275 out: 1276 mutex_unlock(&xt[af].mutex); 1277 return ERR_PTR(-ENOENT); 1278 } 1279 EXPORT_SYMBOL_GPL(xt_find_table_lock); 1280 1281 struct xt_table *xt_request_find_table_lock(struct net *net, u_int8_t af, 1282 const char *name) 1283 { 1284 struct xt_table *t = xt_find_table_lock(net, af, name); 1285 1286 #ifdef CONFIG_MODULES 1287 if (IS_ERR(t)) { 1288 int err = request_module("%stable_%s", xt_prefix[af], name); 1289 if (err < 0) 1290 return ERR_PTR(err); 1291 t = xt_find_table_lock(net, af, name); 1292 } 1293 #endif 1294 1295 return t; 1296 } 1297 EXPORT_SYMBOL_GPL(xt_request_find_table_lock); 1298 1299 void xt_table_unlock(struct xt_table *table) 1300 { 1301 mutex_unlock(&xt[table->af].mutex); 1302 } 1303 EXPORT_SYMBOL_GPL(xt_table_unlock); 1304 1305 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT 1306 void xt_compat_lock(u_int8_t af) 1307 { 1308 mutex_lock(&xt[af].compat_mutex); 1309 } 1310 EXPORT_SYMBOL_GPL(xt_compat_lock); 1311 1312 void xt_compat_unlock(u_int8_t af) 1313 { 1314 mutex_unlock(&xt[af].compat_mutex); 1315 } 1316 EXPORT_SYMBOL_GPL(xt_compat_unlock); 1317 #endif 1318 1319 DEFINE_PER_CPU(seqcount_t, xt_recseq); 1320 EXPORT_PER_CPU_SYMBOL_GPL(xt_recseq); 1321 1322 struct static_key xt_tee_enabled __read_mostly; 1323 EXPORT_SYMBOL_GPL(xt_tee_enabled); 1324 1325 static int xt_jumpstack_alloc(struct xt_table_info *i) 1326 { 1327 unsigned int size; 1328 int cpu; 1329 1330 size = sizeof(void **) * nr_cpu_ids; 1331 if (size > PAGE_SIZE) 1332 i->jumpstack = kvzalloc(size, GFP_KERNEL); 1333 else 1334 i->jumpstack = kzalloc(size, GFP_KERNEL); 1335 if (i->jumpstack == NULL) 1336 return -ENOMEM; 1337 1338 /* ruleset without jumps -- no stack needed */ 1339 if (i->stacksize == 0) 1340 return 0; 1341 1342 /* Jumpstack needs to be able to record two full callchains, one 1343 * from the first rule set traversal, plus one table reentrancy 1344 * via -j TEE without clobbering the callchain that brought us to 1345 * TEE target. 1346 * 1347 * This is done by allocating two jumpstacks per cpu, on reentry 1348 * the upper half of the stack is used. 1349 * 1350 * see the jumpstack setup in ipt_do_table() for more details. 1351 */ 1352 size = sizeof(void *) * i->stacksize * 2u; 1353 for_each_possible_cpu(cpu) { 1354 i->jumpstack[cpu] = kvmalloc_node(size, GFP_KERNEL, 1355 cpu_to_node(cpu)); 1356 if (i->jumpstack[cpu] == NULL) 1357 /* 1358 * Freeing will be done later on by the callers. The 1359 * chain is: xt_replace_table -> __do_replace -> 1360 * do_replace -> xt_free_table_info. 1361 */ 1362 return -ENOMEM; 1363 } 1364 1365 return 0; 1366 } 1367 1368 struct xt_counters *xt_counters_alloc(unsigned int counters) 1369 { 1370 struct xt_counters *mem; 1371 1372 if (counters == 0 || counters > INT_MAX / sizeof(*mem)) 1373 return NULL; 1374 1375 counters *= sizeof(*mem); 1376 if (counters > XT_MAX_TABLE_SIZE) 1377 return NULL; 1378 1379 return vzalloc(counters); 1380 } 1381 EXPORT_SYMBOL(xt_counters_alloc); 1382 1383 struct xt_table_info * 1384 xt_replace_table(struct xt_table *table, 1385 unsigned int num_counters, 1386 struct xt_table_info *newinfo, 1387 int *error) 1388 { 1389 struct xt_table_info *private; 1390 unsigned int cpu; 1391 int ret; 1392 1393 ret = xt_jumpstack_alloc(newinfo); 1394 if (ret < 0) { 1395 *error = ret; 1396 return NULL; 1397 } 1398 1399 /* Do the substitution. */ 1400 local_bh_disable(); 1401 private = table->private; 1402 1403 /* Check inside lock: is the old number correct? */ 1404 if (num_counters != private->number) { 1405 pr_debug("num_counters != table->private->number (%u/%u)\n", 1406 num_counters, private->number); 1407 local_bh_enable(); 1408 *error = -EAGAIN; 1409 return NULL; 1410 } 1411 1412 newinfo->initial_entries = private->initial_entries; 1413 /* 1414 * Ensure contents of newinfo are visible before assigning to 1415 * private. 1416 */ 1417 smp_wmb(); 1418 table->private = newinfo; 1419 1420 /* make sure all cpus see new ->private value */ 1421 smp_mb(); 1422 1423 /* 1424 * Even though table entries have now been swapped, other CPU's 1425 * may still be using the old entries... 1426 */ 1427 local_bh_enable(); 1428 1429 /* ... so wait for even xt_recseq on all cpus */ 1430 for_each_possible_cpu(cpu) { 1431 seqcount_t *s = &per_cpu(xt_recseq, cpu); 1432 u32 seq = raw_read_seqcount(s); 1433 1434 if (seq & 1) { 1435 do { 1436 cond_resched(); 1437 cpu_relax(); 1438 } while (seq == raw_read_seqcount(s)); 1439 } 1440 } 1441 1442 audit_log_nfcfg(table->name, table->af, private->number, 1443 !private->number ? AUDIT_XT_OP_REGISTER : 1444 AUDIT_XT_OP_REPLACE, 1445 GFP_KERNEL); 1446 return private; 1447 } 1448 EXPORT_SYMBOL_GPL(xt_replace_table); 1449 1450 struct xt_table *xt_register_table(struct net *net, 1451 const struct xt_table *input_table, 1452 struct xt_table_info *bootstrap, 1453 struct xt_table_info *newinfo) 1454 { 1455 struct xt_pernet *xt_net = net_generic(net, xt_pernet_id); 1456 struct xt_table_info *private; 1457 struct xt_table *t, *table; 1458 int ret; 1459 1460 /* Don't add one object to multiple lists. */ 1461 table = kmemdup(input_table, sizeof(struct xt_table), GFP_KERNEL); 1462 if (!table) { 1463 ret = -ENOMEM; 1464 goto out; 1465 } 1466 1467 mutex_lock(&xt[table->af].mutex); 1468 /* Don't autoload: we'd eat our tail... */ 1469 list_for_each_entry(t, &xt_net->tables[table->af], list) { 1470 if (strcmp(t->name, table->name) == 0) { 1471 ret = -EEXIST; 1472 goto unlock; 1473 } 1474 } 1475 1476 /* Simplifies replace_table code. */ 1477 table->private = bootstrap; 1478 1479 if (!xt_replace_table(table, 0, newinfo, &ret)) 1480 goto unlock; 1481 1482 private = table->private; 1483 pr_debug("table->private->number = %u\n", private->number); 1484 1485 /* save number of initial entries */ 1486 private->initial_entries = private->number; 1487 1488 list_add(&table->list, &xt_net->tables[table->af]); 1489 mutex_unlock(&xt[table->af].mutex); 1490 return table; 1491 1492 unlock: 1493 mutex_unlock(&xt[table->af].mutex); 1494 kfree(table); 1495 out: 1496 return ERR_PTR(ret); 1497 } 1498 EXPORT_SYMBOL_GPL(xt_register_table); 1499 1500 void *xt_unregister_table(struct xt_table *table) 1501 { 1502 struct xt_table_info *private; 1503 1504 mutex_lock(&xt[table->af].mutex); 1505 private = table->private; 1506 list_del(&table->list); 1507 mutex_unlock(&xt[table->af].mutex); 1508 audit_log_nfcfg(table->name, table->af, private->number, 1509 AUDIT_XT_OP_UNREGISTER, GFP_KERNEL); 1510 kfree(table->ops); 1511 kfree(table); 1512 1513 return private; 1514 } 1515 EXPORT_SYMBOL_GPL(xt_unregister_table); 1516 1517 #ifdef CONFIG_PROC_FS 1518 static void *xt_table_seq_start(struct seq_file *seq, loff_t *pos) 1519 { 1520 u8 af = (unsigned long)pde_data(file_inode(seq->file)); 1521 struct net *net = seq_file_net(seq); 1522 struct xt_pernet *xt_net; 1523 1524 xt_net = net_generic(net, xt_pernet_id); 1525 1526 mutex_lock(&xt[af].mutex); 1527 return seq_list_start(&xt_net->tables[af], *pos); 1528 } 1529 1530 static void *xt_table_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1531 { 1532 u8 af = (unsigned long)pde_data(file_inode(seq->file)); 1533 struct net *net = seq_file_net(seq); 1534 struct xt_pernet *xt_net; 1535 1536 xt_net = net_generic(net, xt_pernet_id); 1537 1538 return seq_list_next(v, &xt_net->tables[af], pos); 1539 } 1540 1541 static void xt_table_seq_stop(struct seq_file *seq, void *v) 1542 { 1543 u_int8_t af = (unsigned long)pde_data(file_inode(seq->file)); 1544 1545 mutex_unlock(&xt[af].mutex); 1546 } 1547 1548 static int xt_table_seq_show(struct seq_file *seq, void *v) 1549 { 1550 struct xt_table *table = list_entry(v, struct xt_table, list); 1551 1552 if (*table->name) 1553 seq_printf(seq, "%s\n", table->name); 1554 return 0; 1555 } 1556 1557 static const struct seq_operations xt_table_seq_ops = { 1558 .start = xt_table_seq_start, 1559 .next = xt_table_seq_next, 1560 .stop = xt_table_seq_stop, 1561 .show = xt_table_seq_show, 1562 }; 1563 1564 /* 1565 * Traverse state for ip{,6}_{tables,matches} for helping crossing 1566 * the multi-AF mutexes. 1567 */ 1568 struct nf_mttg_trav { 1569 struct list_head *head, *curr; 1570 uint8_t class; 1571 }; 1572 1573 enum { 1574 MTTG_TRAV_INIT, 1575 MTTG_TRAV_NFP_UNSPEC, 1576 MTTG_TRAV_NFP_SPEC, 1577 MTTG_TRAV_DONE, 1578 }; 1579 1580 static void *xt_mttg_seq_next(struct seq_file *seq, void *v, loff_t *ppos, 1581 bool is_target) 1582 { 1583 static const uint8_t next_class[] = { 1584 [MTTG_TRAV_NFP_UNSPEC] = MTTG_TRAV_NFP_SPEC, 1585 [MTTG_TRAV_NFP_SPEC] = MTTG_TRAV_DONE, 1586 }; 1587 uint8_t nfproto = (unsigned long)pde_data(file_inode(seq->file)); 1588 struct nf_mttg_trav *trav = seq->private; 1589 1590 if (ppos != NULL) 1591 ++(*ppos); 1592 1593 switch (trav->class) { 1594 case MTTG_TRAV_INIT: 1595 trav->class = MTTG_TRAV_NFP_UNSPEC; 1596 mutex_lock(&xt[NFPROTO_UNSPEC].mutex); 1597 trav->head = trav->curr = is_target ? 1598 &xt[NFPROTO_UNSPEC].target : &xt[NFPROTO_UNSPEC].match; 1599 break; 1600 case MTTG_TRAV_NFP_UNSPEC: 1601 trav->curr = trav->curr->next; 1602 if (trav->curr != trav->head) 1603 break; 1604 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex); 1605 mutex_lock(&xt[nfproto].mutex); 1606 trav->head = trav->curr = is_target ? 1607 &xt[nfproto].target : &xt[nfproto].match; 1608 trav->class = next_class[trav->class]; 1609 break; 1610 case MTTG_TRAV_NFP_SPEC: 1611 trav->curr = trav->curr->next; 1612 if (trav->curr != trav->head) 1613 break; 1614 fallthrough; 1615 default: 1616 return NULL; 1617 } 1618 return trav; 1619 } 1620 1621 static void *xt_mttg_seq_start(struct seq_file *seq, loff_t *pos, 1622 bool is_target) 1623 { 1624 struct nf_mttg_trav *trav = seq->private; 1625 unsigned int j; 1626 1627 trav->class = MTTG_TRAV_INIT; 1628 for (j = 0; j < *pos; ++j) 1629 if (xt_mttg_seq_next(seq, NULL, NULL, is_target) == NULL) 1630 return NULL; 1631 return trav; 1632 } 1633 1634 static void xt_mttg_seq_stop(struct seq_file *seq, void *v) 1635 { 1636 uint8_t nfproto = (unsigned long)pde_data(file_inode(seq->file)); 1637 struct nf_mttg_trav *trav = seq->private; 1638 1639 switch (trav->class) { 1640 case MTTG_TRAV_NFP_UNSPEC: 1641 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex); 1642 break; 1643 case MTTG_TRAV_NFP_SPEC: 1644 mutex_unlock(&xt[nfproto].mutex); 1645 break; 1646 } 1647 } 1648 1649 static void *xt_match_seq_start(struct seq_file *seq, loff_t *pos) 1650 { 1651 return xt_mttg_seq_start(seq, pos, false); 1652 } 1653 1654 static void *xt_match_seq_next(struct seq_file *seq, void *v, loff_t *ppos) 1655 { 1656 return xt_mttg_seq_next(seq, v, ppos, false); 1657 } 1658 1659 static int xt_match_seq_show(struct seq_file *seq, void *v) 1660 { 1661 const struct nf_mttg_trav *trav = seq->private; 1662 const struct xt_match *match; 1663 1664 switch (trav->class) { 1665 case MTTG_TRAV_NFP_UNSPEC: 1666 case MTTG_TRAV_NFP_SPEC: 1667 if (trav->curr == trav->head) 1668 return 0; 1669 match = list_entry(trav->curr, struct xt_match, list); 1670 if (*match->name) 1671 seq_printf(seq, "%s\n", match->name); 1672 } 1673 return 0; 1674 } 1675 1676 static const struct seq_operations xt_match_seq_ops = { 1677 .start = xt_match_seq_start, 1678 .next = xt_match_seq_next, 1679 .stop = xt_mttg_seq_stop, 1680 .show = xt_match_seq_show, 1681 }; 1682 1683 static void *xt_target_seq_start(struct seq_file *seq, loff_t *pos) 1684 { 1685 return xt_mttg_seq_start(seq, pos, true); 1686 } 1687 1688 static void *xt_target_seq_next(struct seq_file *seq, void *v, loff_t *ppos) 1689 { 1690 return xt_mttg_seq_next(seq, v, ppos, true); 1691 } 1692 1693 static int xt_target_seq_show(struct seq_file *seq, void *v) 1694 { 1695 const struct nf_mttg_trav *trav = seq->private; 1696 const struct xt_target *target; 1697 1698 switch (trav->class) { 1699 case MTTG_TRAV_NFP_UNSPEC: 1700 case MTTG_TRAV_NFP_SPEC: 1701 if (trav->curr == trav->head) 1702 return 0; 1703 target = list_entry(trav->curr, struct xt_target, list); 1704 if (*target->name) 1705 seq_printf(seq, "%s\n", target->name); 1706 } 1707 return 0; 1708 } 1709 1710 static const struct seq_operations xt_target_seq_ops = { 1711 .start = xt_target_seq_start, 1712 .next = xt_target_seq_next, 1713 .stop = xt_mttg_seq_stop, 1714 .show = xt_target_seq_show, 1715 }; 1716 1717 #define FORMAT_TABLES "_tables_names" 1718 #define FORMAT_MATCHES "_tables_matches" 1719 #define FORMAT_TARGETS "_tables_targets" 1720 1721 #endif /* CONFIG_PROC_FS */ 1722 1723 /** 1724 * xt_hook_ops_alloc - set up hooks for a new table 1725 * @table: table with metadata needed to set up hooks 1726 * @fn: Hook function 1727 * 1728 * This function will create the nf_hook_ops that the x_table needs 1729 * to hand to xt_hook_link_net(). 1730 */ 1731 struct nf_hook_ops * 1732 xt_hook_ops_alloc(const struct xt_table *table, nf_hookfn *fn) 1733 { 1734 unsigned int hook_mask = table->valid_hooks; 1735 uint8_t i, num_hooks = hweight32(hook_mask); 1736 uint8_t hooknum; 1737 struct nf_hook_ops *ops; 1738 1739 if (!num_hooks) 1740 return ERR_PTR(-EINVAL); 1741 1742 ops = kcalloc(num_hooks, sizeof(*ops), GFP_KERNEL); 1743 if (ops == NULL) 1744 return ERR_PTR(-ENOMEM); 1745 1746 for (i = 0, hooknum = 0; i < num_hooks && hook_mask != 0; 1747 hook_mask >>= 1, ++hooknum) { 1748 if (!(hook_mask & 1)) 1749 continue; 1750 ops[i].hook = fn; 1751 ops[i].pf = table->af; 1752 ops[i].hooknum = hooknum; 1753 ops[i].priority = table->priority; 1754 ++i; 1755 } 1756 1757 return ops; 1758 } 1759 EXPORT_SYMBOL_GPL(xt_hook_ops_alloc); 1760 1761 int xt_register_template(const struct xt_table *table, 1762 int (*table_init)(struct net *net)) 1763 { 1764 int ret = -EEXIST, af = table->af; 1765 struct xt_template *t; 1766 1767 mutex_lock(&xt[af].mutex); 1768 1769 list_for_each_entry(t, &xt_templates[af], list) { 1770 if (WARN_ON_ONCE(strcmp(table->name, t->name) == 0)) 1771 goto out_unlock; 1772 } 1773 1774 ret = -ENOMEM; 1775 t = kzalloc(sizeof(*t), GFP_KERNEL); 1776 if (!t) 1777 goto out_unlock; 1778 1779 BUILD_BUG_ON(sizeof(t->name) != sizeof(table->name)); 1780 1781 strscpy(t->name, table->name, sizeof(t->name)); 1782 t->table_init = table_init; 1783 t->me = table->me; 1784 list_add(&t->list, &xt_templates[af]); 1785 ret = 0; 1786 out_unlock: 1787 mutex_unlock(&xt[af].mutex); 1788 return ret; 1789 } 1790 EXPORT_SYMBOL_GPL(xt_register_template); 1791 1792 void xt_unregister_template(const struct xt_table *table) 1793 { 1794 struct xt_template *t; 1795 int af = table->af; 1796 1797 mutex_lock(&xt[af].mutex); 1798 list_for_each_entry(t, &xt_templates[af], list) { 1799 if (strcmp(table->name, t->name)) 1800 continue; 1801 1802 list_del(&t->list); 1803 mutex_unlock(&xt[af].mutex); 1804 kfree(t); 1805 return; 1806 } 1807 1808 mutex_unlock(&xt[af].mutex); 1809 WARN_ON_ONCE(1); 1810 } 1811 EXPORT_SYMBOL_GPL(xt_unregister_template); 1812 1813 int xt_proto_init(struct net *net, u_int8_t af) 1814 { 1815 #ifdef CONFIG_PROC_FS 1816 char buf[XT_FUNCTION_MAXNAMELEN]; 1817 struct proc_dir_entry *proc; 1818 kuid_t root_uid; 1819 kgid_t root_gid; 1820 #endif 1821 1822 if (af >= ARRAY_SIZE(xt_prefix)) 1823 return -EINVAL; 1824 1825 1826 #ifdef CONFIG_PROC_FS 1827 root_uid = make_kuid(net->user_ns, 0); 1828 root_gid = make_kgid(net->user_ns, 0); 1829 1830 strscpy(buf, xt_prefix[af], sizeof(buf)); 1831 strlcat(buf, FORMAT_TABLES, sizeof(buf)); 1832 proc = proc_create_net_data(buf, 0440, net->proc_net, &xt_table_seq_ops, 1833 sizeof(struct seq_net_private), 1834 (void *)(unsigned long)af); 1835 if (!proc) 1836 goto out; 1837 if (uid_valid(root_uid) && gid_valid(root_gid)) 1838 proc_set_user(proc, root_uid, root_gid); 1839 1840 strscpy(buf, xt_prefix[af], sizeof(buf)); 1841 strlcat(buf, FORMAT_MATCHES, sizeof(buf)); 1842 proc = proc_create_seq_private(buf, 0440, net->proc_net, 1843 &xt_match_seq_ops, sizeof(struct nf_mttg_trav), 1844 (void *)(unsigned long)af); 1845 if (!proc) 1846 goto out_remove_tables; 1847 if (uid_valid(root_uid) && gid_valid(root_gid)) 1848 proc_set_user(proc, root_uid, root_gid); 1849 1850 strscpy(buf, xt_prefix[af], sizeof(buf)); 1851 strlcat(buf, FORMAT_TARGETS, sizeof(buf)); 1852 proc = proc_create_seq_private(buf, 0440, net->proc_net, 1853 &xt_target_seq_ops, sizeof(struct nf_mttg_trav), 1854 (void *)(unsigned long)af); 1855 if (!proc) 1856 goto out_remove_matches; 1857 if (uid_valid(root_uid) && gid_valid(root_gid)) 1858 proc_set_user(proc, root_uid, root_gid); 1859 #endif 1860 1861 return 0; 1862 1863 #ifdef CONFIG_PROC_FS 1864 out_remove_matches: 1865 strscpy(buf, xt_prefix[af], sizeof(buf)); 1866 strlcat(buf, FORMAT_MATCHES, sizeof(buf)); 1867 remove_proc_entry(buf, net->proc_net); 1868 1869 out_remove_tables: 1870 strscpy(buf, xt_prefix[af], sizeof(buf)); 1871 strlcat(buf, FORMAT_TABLES, sizeof(buf)); 1872 remove_proc_entry(buf, net->proc_net); 1873 out: 1874 return -1; 1875 #endif 1876 } 1877 EXPORT_SYMBOL_GPL(xt_proto_init); 1878 1879 void xt_proto_fini(struct net *net, u_int8_t af) 1880 { 1881 #ifdef CONFIG_PROC_FS 1882 char buf[XT_FUNCTION_MAXNAMELEN]; 1883 1884 strscpy(buf, xt_prefix[af], sizeof(buf)); 1885 strlcat(buf, FORMAT_TABLES, sizeof(buf)); 1886 remove_proc_entry(buf, net->proc_net); 1887 1888 strscpy(buf, xt_prefix[af], sizeof(buf)); 1889 strlcat(buf, FORMAT_TARGETS, sizeof(buf)); 1890 remove_proc_entry(buf, net->proc_net); 1891 1892 strscpy(buf, xt_prefix[af], sizeof(buf)); 1893 strlcat(buf, FORMAT_MATCHES, sizeof(buf)); 1894 remove_proc_entry(buf, net->proc_net); 1895 #endif /*CONFIG_PROC_FS*/ 1896 } 1897 EXPORT_SYMBOL_GPL(xt_proto_fini); 1898 1899 /** 1900 * xt_percpu_counter_alloc - allocate x_tables rule counter 1901 * 1902 * @state: pointer to xt_percpu allocation state 1903 * @counter: pointer to counter struct inside the ip(6)/arpt_entry struct 1904 * 1905 * On SMP, the packet counter [ ip(6)t_entry->counters.pcnt ] will then 1906 * contain the address of the real (percpu) counter. 1907 * 1908 * Rule evaluation needs to use xt_get_this_cpu_counter() helper 1909 * to fetch the real percpu counter. 1910 * 1911 * To speed up allocation and improve data locality, a 4kb block is 1912 * allocated. Freeing any counter may free an entire block, so all 1913 * counters allocated using the same state must be freed at the same 1914 * time. 1915 * 1916 * xt_percpu_counter_alloc_state contains the base address of the 1917 * allocated page and the current sub-offset. 1918 * 1919 * returns false on error. 1920 */ 1921 bool xt_percpu_counter_alloc(struct xt_percpu_counter_alloc_state *state, 1922 struct xt_counters *counter) 1923 { 1924 BUILD_BUG_ON(XT_PCPU_BLOCK_SIZE < (sizeof(*counter) * 2)); 1925 1926 if (nr_cpu_ids <= 1) 1927 return true; 1928 1929 if (!state->mem) { 1930 state->mem = __alloc_percpu(XT_PCPU_BLOCK_SIZE, 1931 XT_PCPU_BLOCK_SIZE); 1932 if (!state->mem) 1933 return false; 1934 } 1935 counter->pcnt = (__force unsigned long)(state->mem + state->off); 1936 state->off += sizeof(*counter); 1937 if (state->off > (XT_PCPU_BLOCK_SIZE - sizeof(*counter))) { 1938 state->mem = NULL; 1939 state->off = 0; 1940 } 1941 return true; 1942 } 1943 EXPORT_SYMBOL_GPL(xt_percpu_counter_alloc); 1944 1945 void xt_percpu_counter_free(struct xt_counters *counters) 1946 { 1947 unsigned long pcnt = counters->pcnt; 1948 1949 if (nr_cpu_ids > 1 && (pcnt & (XT_PCPU_BLOCK_SIZE - 1)) == 0) 1950 free_percpu((void __percpu *)pcnt); 1951 } 1952 EXPORT_SYMBOL_GPL(xt_percpu_counter_free); 1953 1954 static int __net_init xt_net_init(struct net *net) 1955 { 1956 struct xt_pernet *xt_net = net_generic(net, xt_pernet_id); 1957 int i; 1958 1959 for (i = 0; i < NFPROTO_NUMPROTO; i++) 1960 INIT_LIST_HEAD(&xt_net->tables[i]); 1961 return 0; 1962 } 1963 1964 static void __net_exit xt_net_exit(struct net *net) 1965 { 1966 struct xt_pernet *xt_net = net_generic(net, xt_pernet_id); 1967 int i; 1968 1969 for (i = 0; i < NFPROTO_NUMPROTO; i++) 1970 WARN_ON_ONCE(!list_empty(&xt_net->tables[i])); 1971 } 1972 1973 static struct pernet_operations xt_net_ops = { 1974 .init = xt_net_init, 1975 .exit = xt_net_exit, 1976 .id = &xt_pernet_id, 1977 .size = sizeof(struct xt_pernet), 1978 }; 1979 1980 static int __init xt_init(void) 1981 { 1982 unsigned int i; 1983 int rv; 1984 1985 for_each_possible_cpu(i) { 1986 seqcount_init(&per_cpu(xt_recseq, i)); 1987 } 1988 1989 xt = kcalloc(NFPROTO_NUMPROTO, sizeof(struct xt_af), GFP_KERNEL); 1990 if (!xt) 1991 return -ENOMEM; 1992 1993 for (i = 0; i < NFPROTO_NUMPROTO; i++) { 1994 mutex_init(&xt[i].mutex); 1995 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT 1996 mutex_init(&xt[i].compat_mutex); 1997 xt[i].compat_tab = NULL; 1998 #endif 1999 INIT_LIST_HEAD(&xt[i].target); 2000 INIT_LIST_HEAD(&xt[i].match); 2001 INIT_LIST_HEAD(&xt_templates[i]); 2002 } 2003 rv = register_pernet_subsys(&xt_net_ops); 2004 if (rv < 0) 2005 kfree(xt); 2006 return rv; 2007 } 2008 2009 static void __exit xt_fini(void) 2010 { 2011 unregister_pernet_subsys(&xt_net_ops); 2012 kfree(xt); 2013 } 2014 2015 module_init(xt_init); 2016 module_exit(xt_fini); 2017