FILE
x_tables.c
File preview with syntax highlighting.
linux-6.19.14/net/netfilter/x_tables.c
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 35MODULE_LICENSE("GPL"); 36MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>"); 37MODULE_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 42struct 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 54static struct list_head xt_templates[NFPROTO_NUMPROTO]; 55 56struct xt_pernet { 57 struct list_head tables[NFPROTO_NUMPROTO]; 58}; 59 60struct compat_delta { 61 unsigned int offset; /* offset in kernel */ 62 int delta; /* delta in 32bit user land */ 63}; 64 65struct 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 77static unsigned int xt_pernet_id __read_mostly; 78static struct xt_af *xt __read_mostly; 79 80static 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. */ 89int 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} 98EXPORT_SYMBOL(xt_register_target); 99 100void 101xt_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} 109EXPORT_SYMBOL(xt_unregister_target); 110 111int 112xt_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 124err: 125 if (i > 0) 126 xt_unregister_targets(target, i); 127 return err; 128} 129EXPORT_SYMBOL(xt_register_targets); 130 131void 132xt_unregister_targets(struct xt_target *target, unsigned int n) 133{ 134 while (n-- > 0) 135 xt_unregister_target(&target[n]); 136} 137EXPORT_SYMBOL(xt_unregister_targets); 138 139int 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} 148EXPORT_SYMBOL(xt_register_match); 149 150void 151xt_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} 159EXPORT_SYMBOL(xt_unregister_match); 160 161int 162xt_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 174err: 175 if (i > 0) 176 xt_unregister_matches(match, i); 177 return err; 178} 179EXPORT_SYMBOL(xt_register_matches); 180 181void 182xt_unregister_matches(struct xt_match *match, unsigned int n) 183{ 184 while (n-- > 0) 185 xt_unregister_match(&match[n]); 186} 187EXPORT_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. */ 197struct 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} 225EXPORT_SYMBOL(xt_find_match); 226 227struct xt_match * 228xt_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} 243EXPORT_SYMBOL_GPL(xt_request_find_match); 244 245/* Find target, grabs ref. Returns ERR_PTR() on error. */ 246static 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 275struct 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} 290EXPORT_SYMBOL_GPL(xt_request_find_target); 291 292 293static 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 312int 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} 324EXPORT_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 332int 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} 338EXPORT_SYMBOL_GPL(xt_match_to_user); 339 340int 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} 346EXPORT_SYMBOL_GPL(xt_target_to_user); 347 348static 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 370static 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) */ 393int 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} 414EXPORT_SYMBOL_GPL(xt_find_revision); 415 416static char * 417textify_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 */ 463int 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} 478EXPORT_SYMBOL(xt_check_proc_name); 479 480int 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 } 504 505 /* NFPROTO_UNSPEC implies NF_INET_* hooks which do not overlap with 506 * NF_ARP_IN,OUT,FORWARD, allow explicit extensions with NFPROTO_ARP 507 * support. 508 */ 509 if (par->family == NFPROTO_ARP && 510 par->match->family != NFPROTO_ARP) { 511 pr_info_ratelimited("%s_tables: %s match: not valid for this family\n", 512 xt_prefix[par->family], par->match->name); 513 return -EINVAL; 514 } 515 if (par->match->hooks && (par->hook_mask & ~par->match->hooks) != 0) { 516 char used[64], allow[64]; 517 518 pr_info_ratelimited("%s_tables: %s match: used from hooks %s, but only valid from %s\n", 519 xt_prefix[par->family], par->match->name, 520 textify_hooks(used, sizeof(used), 521 par->hook_mask, par->family), 522 textify_hooks(allow, sizeof(allow), 523 par->match->hooks, 524 par->family)); 525 return -EINVAL; 526 } 527 if (par->match->proto && (par->match->proto != proto || inv_proto)) { 528 pr_info_ratelimited("%s_tables: %s match: only valid for protocol %u\n", 529 xt_prefix[par->family], par->match->name, 530 par->match->proto); 531 return -EINVAL; 532 } 533 if (par->match->checkentry != NULL) { 534 ret = par->match->checkentry(par); 535 if (ret < 0) 536 return ret; 537 else if (ret > 0) 538 /* Flag up potential errors. */ 539 return -EIO; 540 } 541 return 0; 542} 543EXPORT_SYMBOL_GPL(xt_check_match); 544 545/** xt_check_entry_match - check that matches end before start of target 546 * 547 * @match: beginning of xt_entry_match 548 * @target: beginning of this rules target (alleged end of matches) 549 * @alignment: alignment requirement of match structures 550 * 551 * Validates that all matches add up to the beginning of the target, 552 * and that each match covers at least the base structure size. 553 * 554 * Return: 0 on success, negative errno on failure. 555 */ 556static int xt_check_entry_match(const char *match, const char *target, 557 const size_t alignment) 558{ 559 const struct xt_entry_match *pos; 560 int length = target - match; 561 562 if (length == 0) /* no matches */ 563 return 0; 564 565 pos = (struct xt_entry_match *)match; 566 do { 567 if ((unsigned long)pos % alignment) 568 return -EINVAL; 569 570 if (length < (int)sizeof(struct xt_entry_match)) 571 return -EINVAL; 572 573 if (pos->u.match_size < sizeof(struct xt_entry_match)) 574 return -EINVAL; 575 576 if (pos->u.match_size > length) 577 return -EINVAL; 578 579 length -= pos->u.match_size; 580 pos = ((void *)((char *)(pos) + (pos)->u.match_size)); 581 } while (length > 0); 582 583 return 0; 584} 585 586/** xt_check_table_hooks - check hook entry points are sane 587 * 588 * @info xt_table_info to check 589 * @valid_hooks - hook entry points that we can enter from 590 * 591 * Validates that the hook entry and underflows points are set up. 592 * 593 * Return: 0 on success, negative errno on failure. 594 */ 595int xt_check_table_hooks(const struct xt_table_info *info, unsigned int valid_hooks) 596{ 597 const char *err = "unsorted underflow"; 598 unsigned int i, max_uflow, max_entry; 599 bool check_hooks = false; 600 601 BUILD_BUG_ON(ARRAY_SIZE(info->hook_entry) != ARRAY_SIZE(info->underflow)); 602 603 max_entry = 0; 604 max_uflow = 0; 605 606 for (i = 0; i < ARRAY_SIZE(info->hook_entry); i++) { 607 if (!(valid_hooks & (1 << i))) 608 continue; 609 610 if (info->hook_entry[i] == 0xFFFFFFFF) 611 return -EINVAL; 612 if (info->underflow[i] == 0xFFFFFFFF) 613 return -EINVAL; 614 615 if (check_hooks) { 616 if (max_uflow > info->underflow[i]) 617 goto error; 618 619 if (max_uflow == info->underflow[i]) { 620 err = "duplicate underflow"; 621 goto error; 622 } 623 if (max_entry > info->hook_entry[i]) { 624 err = "unsorted entry"; 625 goto error; 626 } 627 if (max_entry == info->hook_entry[i]) { 628 err = "duplicate entry"; 629 goto error; 630 } 631 } 632 max_entry = info->hook_entry[i]; 633 max_uflow = info->underflow[i]; 634 check_hooks = true; 635 } 636 637 return 0; 638error: 639 pr_err_ratelimited("%s at hook %d\n", err, i); 640 return -EINVAL; 641} 642EXPORT_SYMBOL(xt_check_table_hooks); 643 644static bool verdict_ok(int verdict) 645{ 646 if (verdict > 0) 647 return true; 648 649 if (verdict < 0) { 650 int v = -verdict - 1; 651 652 if (verdict == XT_RETURN) 653 return true; 654 655 switch (v) { 656 case NF_ACCEPT: return true; 657 case NF_DROP: return true; 658 case NF_QUEUE: return true; 659 default: 660 break; 661 } 662 663 return false; 664 } 665 666 return false; 667} 668 669static bool error_tg_ok(unsigned int usersize, unsigned int kernsize, 670 const char *msg, unsigned int msglen) 671{ 672 return usersize == kernsize && strnlen(msg, msglen) < msglen; 673} 674 675#ifdef CONFIG_NETFILTER_XTABLES_COMPAT 676int xt_compat_add_offset(u_int8_t af, unsigned int offset, int delta) 677{ 678 struct xt_af *xp = &xt[af]; 679 680 WARN_ON(!mutex_is_locked(&xt[af].compat_mutex)); 681 682 if (WARN_ON(!xp->compat_tab)) 683 return -ENOMEM; 684 685 if (xp->cur >= xp->number) 686 return -EINVAL; 687 688 if (xp->cur) 689 delta += xp->compat_tab[xp->cur - 1].delta; 690 xp->compat_tab[xp->cur].offset = offset; 691 xp->compat_tab[xp->cur].delta = delta; 692 xp->cur++; 693 return 0; 694} 695EXPORT_SYMBOL_GPL(xt_compat_add_offset); 696 697void xt_compat_flush_offsets(u_int8_t af) 698{ 699 WARN_ON(!mutex_is_locked(&xt[af].compat_mutex)); 700 701 if (xt[af].compat_tab) { 702 vfree(xt[af].compat_tab); 703 xt[af].compat_tab = NULL; 704 xt[af].number = 0; 705 xt[af].cur = 0; 706 } 707} 708EXPORT_SYMBOL_GPL(xt_compat_flush_offsets); 709 710int xt_compat_calc_jump(u_int8_t af, unsigned int offset) 711{ 712 struct compat_delta *tmp = xt[af].compat_tab; 713 int mid, left = 0, right = xt[af].cur - 1; 714 715 while (left <= right) { 716 mid = (left + right) >> 1; 717 if (offset > tmp[mid].offset) 718 left = mid + 1; 719 else if (offset < tmp[mid].offset) 720 right = mid - 1; 721 else 722 return mid ? tmp[mid - 1].delta : 0; 723 } 724 return left ? tmp[left - 1].delta : 0; 725} 726EXPORT_SYMBOL_GPL(xt_compat_calc_jump); 727 728int xt_compat_init_offsets(u8 af, unsigned int number) 729{ 730 size_t mem; 731 732 WARN_ON(!mutex_is_locked(&xt[af].compat_mutex)); 733 734 if (!number || number > (INT_MAX / sizeof(struct compat_delta))) 735 return -EINVAL; 736 737 if (WARN_ON(xt[af].compat_tab)) 738 return -EINVAL; 739 740 mem = sizeof(struct compat_delta) * number; 741 if (mem > XT_MAX_TABLE_SIZE) 742 return -ENOMEM; 743 744 xt[af].compat_tab = vmalloc(mem); 745 if (!xt[af].compat_tab) 746 return -ENOMEM; 747 748 xt[af].number = number; 749 xt[af].cur = 0; 750 751 return 0; 752} 753EXPORT_SYMBOL(xt_compat_init_offsets); 754 755int xt_compat_match_offset(const struct xt_match *match) 756{ 757 u_int16_t csize = match->compatsize ? : match->matchsize; 758 return XT_ALIGN(match->matchsize) - COMPAT_XT_ALIGN(csize); 759} 760EXPORT_SYMBOL_GPL(xt_compat_match_offset); 761 762void xt_compat_match_from_user(struct xt_entry_match *m, void **dstptr, 763 unsigned int *size) 764{ 765 const struct xt_match *match = m->u.kernel.match; 766 struct compat_xt_entry_match *cm = (struct compat_xt_entry_match *)m; 767 int off = xt_compat_match_offset(match); 768 u_int16_t msize = cm->u.user.match_size; 769 char name[sizeof(m->u.user.name)]; 770 771 m = *dstptr; 772 memcpy(m, cm, sizeof(*cm)); 773 if (match->compat_from_user) 774 match->compat_from_user(m->data, cm->data); 775 else 776 memcpy(m->data, cm->data, msize - sizeof(*cm)); 777 778 msize += off; 779 m->u.user.match_size = msize; 780 strscpy(name, match->name, sizeof(name)); 781 module_put(match->me); 782 strscpy_pad(m->u.user.name, name, sizeof(m->u.user.name)); 783 784 *size += off; 785 *dstptr += msize; 786} 787EXPORT_SYMBOL_GPL(xt_compat_match_from_user); 788 789#define COMPAT_XT_DATA_TO_USER(U, K, TYPE, C_SIZE) \ 790 xt_data_to_user(U->data, K->data, \ 791 K->u.kernel.TYPE->usersize, \ 792 C_SIZE, \ 793 COMPAT_XT_ALIGN(C_SIZE)) 794 795int xt_compat_match_to_user(const struct xt_entry_match *m, 796 void __user **dstptr, unsigned int *size) 797{ 798 const struct xt_match *match = m->u.kernel.match; 799 struct compat_xt_entry_match __user *cm = *dstptr; 800 int off = xt_compat_match_offset(match); 801 u_int16_t msize = m->u.user.match_size - off; 802 803 if (XT_OBJ_TO_USER(cm, m, match, msize)) 804 return -EFAULT; 805 806 if (match->compat_to_user) { 807 if (match->compat_to_user((void __user *)cm->data, m->data)) 808 return -EFAULT; 809 } else { 810 if (COMPAT_XT_DATA_TO_USER(cm, m, match, msize - sizeof(*cm))) 811 return -EFAULT; 812 } 813 814 *size -= off; 815 *dstptr += msize; 816 return 0; 817} 818EXPORT_SYMBOL_GPL(xt_compat_match_to_user); 819 820/* non-compat version may have padding after verdict */ 821struct compat_xt_standard_target { 822 struct compat_xt_entry_target t; 823 compat_uint_t verdict; 824}; 825 826struct compat_xt_error_target { 827 struct compat_xt_entry_target t; 828 char errorname[XT_FUNCTION_MAXNAMELEN]; 829}; 830 831int xt_compat_check_entry_offsets(const void *base, const char *elems, 832 unsigned int target_offset, 833 unsigned int next_offset) 834{ 835 long size_of_base_struct = elems - (const char *)base; 836 const struct compat_xt_entry_target *t; 837 const char *e = base; 838 839 if (target_offset < size_of_base_struct) 840 return -EINVAL; 841 842 if (target_offset + sizeof(*t) > next_offset) 843 return -EINVAL; 844 845 t = (void *)(e + target_offset); 846 if (t->u.target_size < sizeof(*t)) 847 return -EINVAL; 848 849 if (target_offset + t->u.target_size > next_offset) 850 return -EINVAL; 851 852 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0) { 853 const struct compat_xt_standard_target *st = (const void *)t; 854 855 if (COMPAT_XT_ALIGN(target_offset + sizeof(*st)) != next_offset) 856 return -EINVAL; 857 858 if (!verdict_ok(st->verdict)) 859 return -EINVAL; 860 } else if (strcmp(t->u.user.name, XT_ERROR_TARGET) == 0) { 861 const struct compat_xt_error_target *et = (const void *)t; 862 863 if (!error_tg_ok(t->u.target_size, sizeof(*et), 864 et->errorname, sizeof(et->errorname))) 865 return -EINVAL; 866 } 867 868 /* compat_xt_entry match has less strict alignment requirements, 869 * otherwise they are identical. In case of padding differences 870 * we need to add compat version of xt_check_entry_match. 871 */ 872 BUILD_BUG_ON(sizeof(struct compat_xt_entry_match) != sizeof(struct xt_entry_match)); 873 874 return xt_check_entry_match(elems, base + target_offset, 875 __alignof__(struct compat_xt_entry_match)); 876} 877EXPORT_SYMBOL(xt_compat_check_entry_offsets); 878#endif /* CONFIG_NETFILTER_XTABLES_COMPAT */ 879 880/** 881 * xt_check_entry_offsets - validate arp/ip/ip6t_entry 882 * 883 * @base: pointer to arp/ip/ip6t_entry 884 * @elems: pointer to first xt_entry_match, i.e. ip(6)t_entry->elems 885 * @target_offset: the arp/ip/ip6_t->target_offset 886 * @next_offset: the arp/ip/ip6_t->next_offset 887 * 888 * validates that target_offset and next_offset are sane and that all 889 * match sizes (if any) align with the target offset. 890 * 891 * This function does not validate the targets or matches themselves, it 892 * only tests that all the offsets and sizes are correct, that all 893 * match structures are aligned, and that the last structure ends where 894 * the target structure begins. 895 * 896 * Also see xt_compat_check_entry_offsets for CONFIG_NETFILTER_XTABLES_COMPAT version. 897 * 898 * The arp/ip/ip6t_entry structure @base must have passed following tests: 899 * - it must point to a valid memory location 900 * - base to base + next_offset must be accessible, i.e. not exceed allocated 901 * length. 902 * 903 * A well-formed entry looks like this: 904 * 905 * ip(6)t_entry match [mtdata] match [mtdata] target [tgdata] ip(6)t_entry 906 * e->elems[]-----' | | 907 * matchsize | | 908 * matchsize | | 909 * | | 910 * target_offset---------------------------------' | 911 * next_offset---------------------------------------------------' 912 * 913 * elems[]: flexible array member at end of ip(6)/arpt_entry struct. 914 * This is where matches (if any) and the target reside. 915 * target_offset: beginning of target. 916 * next_offset: start of the next rule; also: size of this rule. 917 * Since targets have a minimum size, target_offset + minlen <= next_offset. 918 * 919 * Every match stores its size, sum of sizes must not exceed target_offset. 920 * 921 * Return: 0 on success, negative errno on failure. 922 */ 923int xt_check_entry_offsets(const void *base, 924 const char *elems, 925 unsigned int target_offset, 926 unsigned int next_offset) 927{ 928 long size_of_base_struct = elems - (const char *)base; 929 const struct xt_entry_target *t; 930 const char *e = base; 931 932 /* target start is within the ip/ip6/arpt_entry struct */ 933 if (target_offset < size_of_base_struct) 934 return -EINVAL; 935 936 if (target_offset + sizeof(*t) > next_offset) 937 return -EINVAL; 938 939 t = (void *)(e + target_offset); 940 if (t->u.target_size < sizeof(*t)) 941 return -EINVAL; 942 943 if (target_offset + t->u.target_size > next_offset) 944 return -EINVAL; 945 946 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0) { 947 const struct xt_standard_target *st = (const void *)t; 948 949 if (XT_ALIGN(target_offset + sizeof(*st)) != next_offset) 950 return -EINVAL; 951 952 if (!verdict_ok(st->verdict)) 953 return -EINVAL; 954 } else if (strcmp(t->u.user.name, XT_ERROR_TARGET) == 0) { 955 const struct xt_error_target *et = (const void *)t; 956 957 if (!error_tg_ok(t->u.target_size, sizeof(*et), 958 et->errorname, sizeof(et->errorname))) 959 return -EINVAL; 960 } 961 962 return xt_check_entry_match(elems, base + target_offset, 963 __alignof__(struct xt_entry_match)); 964} 965EXPORT_SYMBOL(xt_check_entry_offsets); 966 967/** 968 * xt_alloc_entry_offsets - allocate array to store rule head offsets 969 * 970 * @size: number of entries 971 * 972 * Return: NULL or zeroed kmalloc'd or vmalloc'd array 973 */ 974unsigned int *xt_alloc_entry_offsets(unsigned int size) 975{ 976 if (size > XT_MAX_TABLE_SIZE / sizeof(unsigned int)) 977 return NULL; 978 979 return kvcalloc(size, sizeof(unsigned int), GFP_KERNEL); 980 981} 982EXPORT_SYMBOL(xt_alloc_entry_offsets); 983 984/** 985 * xt_find_jump_offset - check if target is a valid jump offset 986 * 987 * @offsets: array containing all valid rule start offsets of a rule blob 988 * @target: the jump target to search for 989 * @size: entries in @offset 990 */ 991bool xt_find_jump_offset(const unsigned int *offsets, 992 unsigned int target, unsigned int size) 993{ 994 int m, low = 0, hi = size; 995 996 while (hi > low) { 997 m = (low + hi) / 2u; 998 999 if (offsets[m] > target) 1000 hi = m; 1001 else if (offsets[m] < target) 1002 low = m + 1; 1003 else 1004 return true; 1005 } 1006 1007 return false; 1008} 1009EXPORT_SYMBOL(xt_find_jump_offset); 1010 1011int xt_check_target(struct xt_tgchk_param *par, 1012 unsigned int size, u16 proto, bool inv_proto) 1013{ 1014 int ret; 1015 1016 if (XT_ALIGN(par->target->targetsize) != size) { 1017 pr_err_ratelimited("%s_tables: %s.%u target: invalid size %u (kernel) != (user) %u\n", 1018 xt_prefix[par->family], par->target->name, 1019 par->target->revision, 1020 XT_ALIGN(par->target->targetsize), size); 1021 return -EINVAL; 1022 } 1023 if (par->target->table != NULL && 1024 strcmp(par->target->table, par->table) != 0) { 1025 pr_info_ratelimited("%s_tables: %s target: only valid in %s table, not %s\n", 1026 xt_prefix[par->family], par->target->name, 1027 par->target->table, par->table); 1028 return -EINVAL; 1029 } 1030 1031 /* NFPROTO_UNSPEC implies NF_INET_* hooks which do not overlap with 1032 * NF_ARP_IN,OUT,FORWARD, allow explicit extensions with NFPROTO_ARP 1033 * support. 1034 */ 1035 if (par->family == NFPROTO_ARP && 1036 par->target->family != NFPROTO_ARP) { 1037 pr_info_ratelimited("%s_tables: %s target: not valid for this family\n", 1038 xt_prefix[par->family], par->target->name); 1039 return -EINVAL; 1040 } 1041 1042 if (par->target->hooks && (par->hook_mask & ~par->target->hooks) != 0) { 1043 char used[64], allow[64]; 1044 1045 pr_info_ratelimited("%s_tables: %s target: used from hooks %s, but only usable from %s\n", 1046 xt_prefix[par->family], par->target->name, 1047 textify_hooks(used, sizeof(used), 1048 par->hook_mask, par->family), 1049 textify_hooks(allow, sizeof(allow), 1050 par->target->hooks, 1051 par->family)); 1052 return -EINVAL; 1053 } 1054 if (par->target->proto && (par->target->proto != proto || inv_proto)) { 1055 pr_info_ratelimited("%s_tables: %s target: only valid for protocol %u\n", 1056 xt_prefix[par->family], par->target->name, 1057 par->target->proto); 1058 return -EINVAL; 1059 } 1060 if (par->target->checkentry != NULL) { 1061 ret = par->target->checkentry(par); 1062 if (ret < 0) 1063 return ret; 1064 else if (ret > 0) 1065 /* Flag up potential errors. */ 1066 return -EIO; 1067 } 1068 return 0; 1069} 1070EXPORT_SYMBOL_GPL(xt_check_target); 1071 1072/** 1073 * xt_copy_counters - copy counters and metadata from a sockptr_t 1074 * 1075 * @arg: src sockptr 1076 * @len: alleged size of userspace memory 1077 * @info: where to store the xt_counters_info metadata 1078 * 1079 * Copies counter meta data from @user and stores it in @info. 1080 * 1081 * vmallocs memory to hold the counters, then copies the counter data 1082 * from @user to the new memory and returns a pointer to it. 1083 * 1084 * If called from a compat syscall, @info gets converted automatically to the 1085 * 64bit representation. 1086 * 1087 * The metadata associated with the counters is stored in @info. 1088 * 1089 * Return: returns pointer that caller has to test via IS_ERR(). 1090 * If IS_ERR is false, caller has to vfree the pointer. 1091 */ 1092void *xt_copy_counters(sockptr_t arg, unsigned int len, 1093 struct xt_counters_info *info) 1094{ 1095 size_t offset; 1096 void *mem; 1097 u64 size; 1098 1099#ifdef CONFIG_NETFILTER_XTABLES_COMPAT 1100 if (in_compat_syscall()) { 1101 /* structures only differ in size due to alignment */ 1102 struct compat_xt_counters_info compat_tmp; 1103 1104 if (len <= sizeof(compat_tmp)) 1105 return ERR_PTR(-EINVAL); 1106 1107 len -= sizeof(compat_tmp); 1108 if (copy_from_sockptr(&compat_tmp, arg, sizeof(compat_tmp)) != 0) 1109 return ERR_PTR(-EFAULT); 1110 1111 memcpy(info->name, compat_tmp.name, sizeof(info->name) - 1); 1112 info->num_counters = compat_tmp.num_counters; 1113 offset = sizeof(compat_tmp); 1114 } else 1115#endif 1116 { 1117 if (len <= sizeof(*info)) 1118 return ERR_PTR(-EINVAL); 1119 1120 len -= sizeof(*info); 1121 if (copy_from_sockptr(info, arg, sizeof(*info)) != 0) 1122 return ERR_PTR(-EFAULT); 1123 1124 offset = sizeof(*info); 1125 } 1126 info->name[sizeof(info->name) - 1] = '\0'; 1127 1128 size = sizeof(struct xt_counters); 1129 size *= info->num_counters; 1130 1131 if (size != (u64)len) 1132 return ERR_PTR(-EINVAL); 1133 1134 mem = vmalloc(len); 1135 if (!mem) 1136 return ERR_PTR(-ENOMEM); 1137 1138 if (copy_from_sockptr_offset(mem, arg, offset, len) == 0) 1139 return mem; 1140 1141 vfree(mem); 1142 return ERR_PTR(-EFAULT); 1143} 1144EXPORT_SYMBOL_GPL(xt_copy_counters); 1145 1146#ifdef CONFIG_NETFILTER_XTABLES_COMPAT 1147int xt_compat_target_offset(const struct xt_target *target) 1148{ 1149 u_int16_t csize = target->compatsize ? : target->targetsize; 1150 return XT_ALIGN(target->targetsize) - COMPAT_XT_ALIGN(csize); 1151} 1152EXPORT_SYMBOL_GPL(xt_compat_target_offset); 1153 1154void xt_compat_target_from_user(struct xt_entry_target *t, void **dstptr, 1155 unsigned int *size) 1156{ 1157 const struct xt_target *target = t->u.kernel.target; 1158 struct compat_xt_entry_target *ct = (struct compat_xt_entry_target *)t; 1159 int off = xt_compat_target_offset(target); 1160 u_int16_t tsize = ct->u.user.target_size; 1161 char name[sizeof(t->u.user.name)]; 1162 1163 t = *dstptr; 1164 memcpy(t, ct, sizeof(*ct)); 1165 if (target->compat_from_user) 1166 target->compat_from_user(t->data, ct->data); 1167 else 1168 unsafe_memcpy(t->data, ct->data, tsize - sizeof(*ct), 1169 /* UAPI 0-sized destination */); 1170 1171 tsize += off; 1172 t->u.user.target_size = tsize; 1173 strscpy(name, target->name, sizeof(name)); 1174 module_put(target->me); 1175 strscpy_pad(t->u.user.name, name, sizeof(t->u.user.name)); 1176 1177 *size += off; 1178 *dstptr += tsize; 1179} 1180EXPORT_SYMBOL_GPL(xt_compat_target_from_user); 1181 1182int xt_compat_target_to_user(const struct xt_entry_target *t, 1183 void __user **dstptr, unsigned int *size) 1184{ 1185 const struct xt_target *target = t->u.kernel.target; 1186 struct compat_xt_entry_target __user *ct = *dstptr; 1187 int off = xt_compat_target_offset(target); 1188 u_int16_t tsize = t->u.user.target_size - off; 1189 1190 if (XT_OBJ_TO_USER(ct, t, target, tsize)) 1191 return -EFAULT; 1192 1193 if (target->compat_to_user) { 1194 if (target->compat_to_user((void __user *)ct->data, t->data)) 1195 return -EFAULT; 1196 } else { 1197 if (COMPAT_XT_DATA_TO_USER(ct, t, target, tsize - sizeof(*ct))) 1198 return -EFAULT; 1199 } 1200 1201 *size -= off; 1202 *dstptr += tsize; 1203 return 0; 1204} 1205EXPORT_SYMBOL_GPL(xt_compat_target_to_user); 1206#endif 1207 1208struct xt_table_info *xt_alloc_table_info(unsigned int size) 1209{ 1210 struct xt_table_info *info = NULL; 1211 size_t sz = sizeof(*info) + size; 1212 1213 if (sz < sizeof(*info) || sz >= XT_MAX_TABLE_SIZE) 1214 return NULL; 1215 1216 info = kvmalloc(sz, GFP_KERNEL_ACCOUNT); 1217 if (!info) 1218 return NULL; 1219 1220 memset(info, 0, sizeof(*info)); 1221 info->size = size; 1222 return info; 1223} 1224EXPORT_SYMBOL(xt_alloc_table_info); 1225 1226void xt_free_table_info(struct xt_table_info *info) 1227{ 1228 int cpu; 1229 1230 if (info->jumpstack != NULL) { 1231 for_each_possible_cpu(cpu) 1232 kvfree(info->jumpstack[cpu]); 1233 kvfree(info->jumpstack); 1234 } 1235 1236 kvfree(info); 1237} 1238EXPORT_SYMBOL(xt_free_table_info); 1239 1240struct xt_table *xt_find_table(struct net *net, u8 af, const char *name) 1241{ 1242 struct xt_pernet *xt_net = net_generic(net, xt_pernet_id); 1243 struct xt_table *t; 1244 1245 mutex_lock(&xt[af].mutex); 1246 list_for_each_entry(t, &xt_net->tables[af], list) { 1247 if (strcmp(t->name, name) == 0) { 1248 mutex_unlock(&xt[af].mutex); 1249 return t; 1250 } 1251 } 1252 mutex_unlock(&xt[af].mutex); 1253 return NULL; 1254} 1255EXPORT_SYMBOL(xt_find_table); 1256 1257/* Find table by name, grabs mutex & ref. Returns ERR_PTR on error. */ 1258struct xt_table *xt_find_table_lock(struct net *net, u_int8_t af, 1259 const char *name) 1260{ 1261 struct xt_pernet *xt_net = net_generic(net, xt_pernet_id); 1262 struct module *owner = NULL; 1263 struct xt_template *tmpl; 1264 struct xt_table *t; 1265 1266 mutex_lock(&xt[af].mutex); 1267 list_for_each_entry(t, &xt_net->tables[af], list) 1268 if (strcmp(t->name, name) == 0 && try_module_get(t->me)) 1269 return t; 1270 1271 /* Table doesn't exist in this netns, check larval list */ 1272 list_for_each_entry(tmpl, &xt_templates[af], list) { 1273 int err; 1274 1275 if (strcmp(tmpl->name, name)) 1276 continue; 1277 if (!try_module_get(tmpl->me)) 1278 goto out; 1279 1280 owner = tmpl->me; 1281 1282 mutex_unlock(&xt[af].mutex); 1283 err = tmpl->table_init(net); 1284 if (err < 0) { 1285 module_put(owner); 1286 return ERR_PTR(err); 1287 } 1288 1289 mutex_lock(&xt[af].mutex); 1290 break; 1291 } 1292 1293 /* and once again: */ 1294 list_for_each_entry(t, &xt_net->tables[af], list) 1295 if (strcmp(t->name, name) == 0 && owner == t->me) 1296 return t; 1297 1298 module_put(owner); 1299 out: 1300 mutex_unlock(&xt[af].mutex); 1301 return ERR_PTR(-ENOENT); 1302} 1303EXPORT_SYMBOL_GPL(xt_find_table_lock); 1304 1305struct xt_table *xt_request_find_table_lock(struct net *net, u_int8_t af, 1306 const char *name) 1307{ 1308 struct xt_table *t = xt_find_table_lock(net, af, name); 1309 1310#ifdef CONFIG_MODULES 1311 if (IS_ERR(t)) { 1312 int err = request_module("%stable_%s", xt_prefix[af], name); 1313 if (err < 0) 1314 return ERR_PTR(err); 1315 t = xt_find_table_lock(net, af, name); 1316 } 1317#endif 1318 1319 return t; 1320} 1321EXPORT_SYMBOL_GPL(xt_request_find_table_lock); 1322 1323void xt_table_unlock(struct xt_table *table) 1324{ 1325 mutex_unlock(&xt[table->af].mutex); 1326} 1327EXPORT_SYMBOL_GPL(xt_table_unlock); 1328 1329#ifdef CONFIG_NETFILTER_XTABLES_COMPAT 1330void xt_compat_lock(u_int8_t af) 1331{ 1332 mutex_lock(&xt[af].compat_mutex); 1333} 1334EXPORT_SYMBOL_GPL(xt_compat_lock); 1335 1336void xt_compat_unlock(u_int8_t af) 1337{ 1338 mutex_unlock(&xt[af].compat_mutex); 1339} 1340EXPORT_SYMBOL_GPL(xt_compat_unlock); 1341#endif 1342 1343struct static_key xt_tee_enabled __read_mostly; 1344EXPORT_SYMBOL_GPL(xt_tee_enabled); 1345 1346#ifdef CONFIG_NETFILTER_XTABLES_LEGACY 1347DEFINE_PER_CPU(seqcount_t, xt_recseq); 1348EXPORT_PER_CPU_SYMBOL_GPL(xt_recseq); 1349 1350static int xt_jumpstack_alloc(struct xt_table_info *i) 1351{ 1352 unsigned int size; 1353 int cpu; 1354 1355 size = sizeof(void **) * nr_cpu_ids; 1356 if (size > PAGE_SIZE) 1357 i->jumpstack = kvzalloc(size, GFP_KERNEL); 1358 else 1359 i->jumpstack = kzalloc(size, GFP_KERNEL); 1360 if (i->jumpstack == NULL) 1361 return -ENOMEM; 1362 1363 /* ruleset without jumps -- no stack needed */ 1364 if (i->stacksize == 0) 1365 return 0; 1366 1367 /* Jumpstack needs to be able to record two full callchains, one 1368 * from the first rule set traversal, plus one table reentrancy 1369 * via -j TEE without clobbering the callchain that brought us to 1370 * TEE target. 1371 * 1372 * This is done by allocating two jumpstacks per cpu, on reentry 1373 * the upper half of the stack is used. 1374 * 1375 * see the jumpstack setup in ipt_do_table() for more details. 1376 */ 1377 size = sizeof(void *) * i->stacksize * 2u; 1378 for_each_possible_cpu(cpu) { 1379 i->jumpstack[cpu] = kvmalloc_node(size, GFP_KERNEL, 1380 cpu_to_node(cpu)); 1381 if (i->jumpstack[cpu] == NULL) 1382 /* 1383 * Freeing will be done later on by the callers. The 1384 * chain is: xt_replace_table -> __do_replace -> 1385 * do_replace -> xt_free_table_info. 1386 */ 1387 return -ENOMEM; 1388 } 1389 1390 return 0; 1391} 1392 1393struct xt_counters *xt_counters_alloc(unsigned int counters) 1394{ 1395 struct xt_counters *mem; 1396 1397 if (counters == 0 || counters > INT_MAX / sizeof(*mem)) 1398 return NULL; 1399 1400 counters *= sizeof(*mem); 1401 if (counters > XT_MAX_TABLE_SIZE) 1402 return NULL; 1403 1404 return vzalloc(counters); 1405} 1406EXPORT_SYMBOL(xt_counters_alloc); 1407 1408struct xt_table_info * 1409xt_replace_table(struct xt_table *table, 1410 unsigned int num_counters, 1411 struct xt_table_info *newinfo, 1412 int *error) 1413{ 1414 struct xt_table_info *private; 1415 unsigned int cpu; 1416 int ret; 1417 1418 ret = xt_jumpstack_alloc(newinfo); 1419 if (ret < 0) { 1420 *error = ret; 1421 return NULL; 1422 } 1423 1424 /* Do the substitution. */ 1425 local_bh_disable(); 1426 private = table->private; 1427 1428 /* Check inside lock: is the old number correct? */ 1429 if (num_counters != private->number) { 1430 pr_debug("num_counters != table->private->number (%u/%u)\n", 1431 num_counters, private->number); 1432 local_bh_enable(); 1433 *error = -EAGAIN; 1434 return NULL; 1435 } 1436 1437 newinfo->initial_entries = private->initial_entries; 1438 /* 1439 * Ensure contents of newinfo are visible before assigning to 1440 * private. 1441 */ 1442 smp_wmb(); 1443 table->private = newinfo; 1444 1445 /* make sure all cpus see new ->private value */ 1446 smp_mb(); 1447 1448 /* 1449 * Even though table entries have now been swapped, other CPU's 1450 * may still be using the old entries... 1451 */ 1452 local_bh_enable(); 1453 1454 /* ... so wait for even xt_recseq on all cpus */ 1455 for_each_possible_cpu(cpu) { 1456 seqcount_t *s = &per_cpu(xt_recseq, cpu); 1457 u32 seq = raw_read_seqcount(s); 1458 1459 if (seq & 1) { 1460 do { 1461 cond_resched(); 1462 cpu_relax(); 1463 } while (seq == raw_read_seqcount(s)); 1464 } 1465 } 1466 1467 audit_log_nfcfg(table->name, table->af, private->number, 1468 !private->number ? AUDIT_XT_OP_REGISTER : 1469 AUDIT_XT_OP_REPLACE, 1470 GFP_KERNEL); 1471 return private; 1472} 1473EXPORT_SYMBOL_GPL(xt_replace_table); 1474 1475struct xt_table *xt_register_table(struct net *net, 1476 const struct xt_table *input_table, 1477 struct xt_table_info *bootstrap, 1478 struct xt_table_info *newinfo) 1479{ 1480 struct xt_pernet *xt_net = net_generic(net, xt_pernet_id); 1481 struct xt_table_info *private; 1482 struct xt_table *t, *table; 1483 int ret; 1484 1485 /* Don't add one object to multiple lists. */ 1486 table = kmemdup(input_table, sizeof(struct xt_table), GFP_KERNEL); 1487 if (!table) { 1488 ret = -ENOMEM; 1489 goto out; 1490 } 1491 1492 mutex_lock(&xt[table->af].mutex); 1493 /* Don't autoload: we'd eat our tail... */ 1494 list_for_each_entry(t, &xt_net->tables[table->af], list) { 1495 if (strcmp(t->name, table->name) == 0) { 1496 ret = -EEXIST; 1497 goto unlock; 1498 } 1499 } 1500 1501 /* Simplifies replace_table code. */ 1502 table->private = bootstrap; 1503 1504 if (!xt_replace_table(table, 0, newinfo, &ret)) 1505 goto unlock; 1506 1507 private = table->private; 1508 pr_debug("table->private->number = %u\n", private->number); 1509 1510 /* save number of initial entries */ 1511 private->initial_entries = private->number; 1512 1513 list_add(&table->list, &xt_net->tables[table->af]); 1514 mutex_unlock(&xt[table->af].mutex); 1515 return table; 1516 1517unlock: 1518 mutex_unlock(&xt[table->af].mutex); 1519 kfree(table); 1520out: 1521 return ERR_PTR(ret); 1522} 1523EXPORT_SYMBOL_GPL(xt_register_table); 1524 1525void *xt_unregister_table(struct xt_table *table) 1526{ 1527 struct xt_table_info *private; 1528 1529 mutex_lock(&xt[table->af].mutex); 1530 private = table->private; 1531 list_del(&table->list); 1532 mutex_unlock(&xt[table->af].mutex); 1533 audit_log_nfcfg(table->name, table->af, private->number, 1534 AUDIT_XT_OP_UNREGISTER, GFP_KERNEL); 1535 kfree(table->ops); 1536 kfree(table); 1537 1538 return private; 1539} 1540EXPORT_SYMBOL_GPL(xt_unregister_table); 1541#endif 1542 1543#ifdef CONFIG_PROC_FS 1544static void *xt_table_seq_start(struct seq_file *seq, loff_t *pos) 1545{ 1546 u8 af = (unsigned long)pde_data(file_inode(seq->file)); 1547 struct net *net = seq_file_net(seq); 1548 struct xt_pernet *xt_net; 1549 1550 xt_net = net_generic(net, xt_pernet_id); 1551 1552 mutex_lock(&xt[af].mutex); 1553 return seq_list_start(&xt_net->tables[af], *pos); 1554} 1555 1556static void *xt_table_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1557{ 1558 u8 af = (unsigned long)pde_data(file_inode(seq->file)); 1559 struct net *net = seq_file_net(seq); 1560 struct xt_pernet *xt_net; 1561 1562 xt_net = net_generic(net, xt_pernet_id); 1563 1564 return seq_list_next(v, &xt_net->tables[af], pos); 1565} 1566 1567static void xt_table_seq_stop(struct seq_file *seq, void *v) 1568{ 1569 u_int8_t af = (unsigned long)pde_data(file_inode(seq->file)); 1570 1571 mutex_unlock(&xt[af].mutex); 1572} 1573 1574static int xt_table_seq_show(struct seq_file *seq, void *v) 1575{ 1576 struct xt_table *table = list_entry(v, struct xt_table, list); 1577 1578 if (*table->name) 1579 seq_printf(seq, "%s\n", table->name); 1580 return 0; 1581} 1582 1583static const struct seq_operations xt_table_seq_ops = { 1584 .start = xt_table_seq_start, 1585 .next = xt_table_seq_next, 1586 .stop = xt_table_seq_stop, 1587 .show = xt_table_seq_show, 1588}; 1589 1590/* 1591 * Traverse state for ip{,6}_{tables,matches} for helping crossing 1592 * the multi-AF mutexes. 1593 */ 1594struct nf_mttg_trav { 1595 struct list_head *head, *curr; 1596 uint8_t class; 1597}; 1598 1599enum { 1600 MTTG_TRAV_INIT, 1601 MTTG_TRAV_NFP_UNSPEC, 1602 MTTG_TRAV_NFP_SPEC, 1603 MTTG_TRAV_DONE, 1604}; 1605 1606static void *xt_mttg_seq_next(struct seq_file *seq, void *v, loff_t *ppos, 1607 bool is_target) 1608{ 1609 static const uint8_t next_class[] = { 1610 [MTTG_TRAV_NFP_UNSPEC] = MTTG_TRAV_NFP_SPEC, 1611 [MTTG_TRAV_NFP_SPEC] = MTTG_TRAV_DONE, 1612 }; 1613 uint8_t nfproto = (unsigned long)pde_data(file_inode(seq->file)); 1614 struct nf_mttg_trav *trav = seq->private; 1615 1616 if (ppos != NULL) 1617 ++(*ppos); 1618 1619 switch (trav->class) { 1620 case MTTG_TRAV_INIT: 1621 trav->class = MTTG_TRAV_NFP_UNSPEC; 1622 mutex_lock(&xt[NFPROTO_UNSPEC].mutex); 1623 trav->head = trav->curr = is_target ? 1624 &xt[NFPROTO_UNSPEC].target : &xt[NFPROTO_UNSPEC].match; 1625 break; 1626 case MTTG_TRAV_NFP_UNSPEC: 1627 trav->curr = trav->curr->next; 1628 if (trav->curr != trav->head) 1629 break; 1630 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex); 1631 mutex_lock(&xt[nfproto].mutex); 1632 trav->head = trav->curr = is_target ? 1633 &xt[nfproto].target : &xt[nfproto].match; 1634 trav->class = next_class[trav->class]; 1635 break; 1636 case MTTG_TRAV_NFP_SPEC: 1637 trav->curr = trav->curr->next; 1638 if (trav->curr != trav->head) 1639 break; 1640 fallthrough; 1641 default: 1642 return NULL; 1643 } 1644 return trav; 1645} 1646 1647static void *xt_mttg_seq_start(struct seq_file *seq, loff_t *pos, 1648 bool is_target) 1649{ 1650 struct nf_mttg_trav *trav = seq->private; 1651 unsigned int j; 1652 1653 trav->class = MTTG_TRAV_INIT; 1654 for (j = 0; j < *pos; ++j) 1655 if (xt_mttg_seq_next(seq, NULL, NULL, is_target) == NULL) 1656 return NULL; 1657 return trav; 1658} 1659 1660static void xt_mttg_seq_stop(struct seq_file *seq, void *v) 1661{ 1662 uint8_t nfproto = (unsigned long)pde_data(file_inode(seq->file)); 1663 struct nf_mttg_trav *trav = seq->private; 1664 1665 switch (trav->class) { 1666 case MTTG_TRAV_NFP_UNSPEC: 1667 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex); 1668 break; 1669 case MTTG_TRAV_NFP_SPEC: 1670 mutex_unlock(&xt[nfproto].mutex); 1671 break; 1672 } 1673} 1674 1675static void *xt_match_seq_start(struct seq_file *seq, loff_t *pos) 1676{ 1677 return xt_mttg_seq_start(seq, pos, false); 1678} 1679 1680static void *xt_match_seq_next(struct seq_file *seq, void *v, loff_t *ppos) 1681{ 1682 return xt_mttg_seq_next(seq, v, ppos, false); 1683} 1684 1685static int xt_match_seq_show(struct seq_file *seq, void *v) 1686{ 1687 const struct nf_mttg_trav *trav = seq->private; 1688 const struct xt_match *match; 1689 1690 switch (trav->class) { 1691 case MTTG_TRAV_NFP_UNSPEC: 1692 case MTTG_TRAV_NFP_SPEC: 1693 if (trav->curr == trav->head) 1694 return 0; 1695 match = list_entry(trav->curr, struct xt_match, list); 1696 if (*match->name) 1697 seq_printf(seq, "%s\n", match->name); 1698 } 1699 return 0; 1700} 1701 1702static const struct seq_operations xt_match_seq_ops = { 1703 .start = xt_match_seq_start, 1704 .next = xt_match_seq_next, 1705 .stop = xt_mttg_seq_stop, 1706 .show = xt_match_seq_show, 1707}; 1708 1709static void *xt_target_seq_start(struct seq_file *seq, loff_t *pos) 1710{ 1711 return xt_mttg_seq_start(seq, pos, true); 1712} 1713 1714static void *xt_target_seq_next(struct seq_file *seq, void *v, loff_t *ppos) 1715{ 1716 return xt_mttg_seq_next(seq, v, ppos, true); 1717} 1718 1719static int xt_target_seq_show(struct seq_file *seq, void *v) 1720{ 1721 const struct nf_mttg_trav *trav = seq->private; 1722 const struct xt_target *target; 1723 1724 switch (trav->class) { 1725 case MTTG_TRAV_NFP_UNSPEC: 1726 case MTTG_TRAV_NFP_SPEC: 1727 if (trav->curr == trav->head) 1728 return 0; 1729 target = list_entry(trav->curr, struct xt_target, list); 1730 if (*target->name) 1731 seq_printf(seq, "%s\n", target->name); 1732 } 1733 return 0; 1734} 1735 1736static const struct seq_operations xt_target_seq_ops = { 1737 .start = xt_target_seq_start, 1738 .next = xt_target_seq_next, 1739 .stop = xt_mttg_seq_stop, 1740 .show = xt_target_seq_show, 1741}; 1742 1743#define FORMAT_TABLES "_tables_names" 1744#define FORMAT_MATCHES "_tables_matches" 1745#define FORMAT_TARGETS "_tables_targets" 1746 1747#endif /* CONFIG_PROC_FS */ 1748 1749/** 1750 * xt_hook_ops_alloc - set up hooks for a new table 1751 * @table: table with metadata needed to set up hooks 1752 * @fn: Hook function 1753 * 1754 * This function will create the nf_hook_ops that the x_table needs 1755 * to hand to xt_hook_link_net(). 1756 */ 1757struct nf_hook_ops * 1758xt_hook_ops_alloc(const struct xt_table *table, nf_hookfn *fn) 1759{ 1760 unsigned int hook_mask = table->valid_hooks; 1761 uint8_t i, num_hooks = hweight32(hook_mask); 1762 uint8_t hooknum; 1763 struct nf_hook_ops *ops; 1764 1765 if (!num_hooks) 1766 return ERR_PTR(-EINVAL); 1767 1768 ops = kcalloc(num_hooks, sizeof(*ops), GFP_KERNEL); 1769 if (ops == NULL) 1770 return ERR_PTR(-ENOMEM); 1771 1772 for (i = 0, hooknum = 0; i < num_hooks && hook_mask != 0; 1773 hook_mask >>= 1, ++hooknum) { 1774 if (!(hook_mask & 1)) 1775 continue; 1776 ops[i].hook = fn; 1777 ops[i].pf = table->af; 1778 ops[i].hooknum = hooknum; 1779 ops[i].priority = table->priority; 1780 ++i; 1781 } 1782 1783 return ops; 1784} 1785EXPORT_SYMBOL_GPL(xt_hook_ops_alloc); 1786 1787int xt_register_template(const struct xt_table *table, 1788 int (*table_init)(struct net *net)) 1789{ 1790 int ret = -EBUSY, af = table->af; 1791 struct xt_template *t; 1792 1793 mutex_lock(&xt[af].mutex); 1794 1795 list_for_each_entry(t, &xt_templates[af], list) { 1796 if (WARN_ON_ONCE(strcmp(table->name, t->name) == 0)) 1797 goto out_unlock; 1798 } 1799 1800 ret = -ENOMEM; 1801 t = kzalloc(sizeof(*t), GFP_KERNEL); 1802 if (!t) 1803 goto out_unlock; 1804 1805 BUILD_BUG_ON(sizeof(t->name) != sizeof(table->name)); 1806 1807 strscpy(t->name, table->name, sizeof(t->name)); 1808 t->table_init = table_init; 1809 t->me = table->me; 1810 list_add(&t->list, &xt_templates[af]); 1811 ret = 0; 1812out_unlock: 1813 mutex_unlock(&xt[af].mutex); 1814 return ret; 1815} 1816EXPORT_SYMBOL_GPL(xt_register_template); 1817 1818void xt_unregister_template(const struct xt_table *table) 1819{ 1820 struct xt_template *t; 1821 int af = table->af; 1822 1823 mutex_lock(&xt[af].mutex); 1824 list_for_each_entry(t, &xt_templates[af], list) { 1825 if (strcmp(table->name, t->name)) 1826 continue; 1827 1828 list_del(&t->list); 1829 mutex_unlock(&xt[af].mutex); 1830 kfree(t); 1831 return; 1832 } 1833 1834 mutex_unlock(&xt[af].mutex); 1835 WARN_ON_ONCE(1); 1836} 1837EXPORT_SYMBOL_GPL(xt_unregister_template); 1838 1839int xt_proto_init(struct net *net, u_int8_t af) 1840{ 1841#ifdef CONFIG_PROC_FS 1842 char buf[XT_FUNCTION_MAXNAMELEN]; 1843 struct proc_dir_entry *proc; 1844 kuid_t root_uid; 1845 kgid_t root_gid; 1846#endif 1847 1848 if (af >= ARRAY_SIZE(xt_prefix)) 1849 return -EINVAL; 1850 1851 1852#ifdef CONFIG_PROC_FS 1853 root_uid = make_kuid(net->user_ns, 0); 1854 root_gid = make_kgid(net->user_ns, 0); 1855 1856 strscpy(buf, xt_prefix[af], sizeof(buf)); 1857 strlcat(buf, FORMAT_TABLES, sizeof(buf)); 1858 proc = proc_create_net_data(buf, 0440, net->proc_net, &xt_table_seq_ops, 1859 sizeof(struct seq_net_private), 1860 (void *)(unsigned long)af); 1861 if (!proc) 1862 goto out; 1863 if (uid_valid(root_uid) && gid_valid(root_gid)) 1864 proc_set_user(proc, root_uid, root_gid); 1865 1866 strscpy(buf, xt_prefix[af], sizeof(buf)); 1867 strlcat(buf, FORMAT_MATCHES, sizeof(buf)); 1868 proc = proc_create_seq_private(buf, 0440, net->proc_net, 1869 &xt_match_seq_ops, sizeof(struct nf_mttg_trav), 1870 (void *)(unsigned long)af); 1871 if (!proc) 1872 goto out_remove_tables; 1873 if (uid_valid(root_uid) && gid_valid(root_gid)) 1874 proc_set_user(proc, root_uid, root_gid); 1875 1876 strscpy(buf, xt_prefix[af], sizeof(buf)); 1877 strlcat(buf, FORMAT_TARGETS, sizeof(buf)); 1878 proc = proc_create_seq_private(buf, 0440, net->proc_net, 1879 &xt_target_seq_ops, sizeof(struct nf_mttg_trav), 1880 (void *)(unsigned long)af); 1881 if (!proc) 1882 goto out_remove_matches; 1883 if (uid_valid(root_uid) && gid_valid(root_gid)) 1884 proc_set_user(proc, root_uid, root_gid); 1885#endif 1886 1887 return 0; 1888 1889#ifdef CONFIG_PROC_FS 1890out_remove_matches: 1891 strscpy(buf, xt_prefix[af], sizeof(buf)); 1892 strlcat(buf, FORMAT_MATCHES, sizeof(buf)); 1893 remove_proc_entry(buf, net->proc_net); 1894 1895out_remove_tables: 1896 strscpy(buf, xt_prefix[af], sizeof(buf)); 1897 strlcat(buf, FORMAT_TABLES, sizeof(buf)); 1898 remove_proc_entry(buf, net->proc_net); 1899out: 1900 return -1; 1901#endif 1902} 1903EXPORT_SYMBOL_GPL(xt_proto_init); 1904 1905void xt_proto_fini(struct net *net, u_int8_t af) 1906{ 1907#ifdef CONFIG_PROC_FS 1908 char buf[XT_FUNCTION_MAXNAMELEN]; 1909 1910 strscpy(buf, xt_prefix[af], sizeof(buf)); 1911 strlcat(buf, FORMAT_TABLES, sizeof(buf)); 1912 remove_proc_entry(buf, net->proc_net); 1913 1914 strscpy(buf, xt_prefix[af], sizeof(buf)); 1915 strlcat(buf, FORMAT_TARGETS, sizeof(buf)); 1916 remove_proc_entry(buf, net->proc_net); 1917 1918 strscpy(buf, xt_prefix[af], sizeof(buf)); 1919 strlcat(buf, FORMAT_MATCHES, sizeof(buf)); 1920 remove_proc_entry(buf, net->proc_net); 1921#endif /*CONFIG_PROC_FS*/ 1922} 1923EXPORT_SYMBOL_GPL(xt_proto_fini); 1924 1925#ifdef CONFIG_NETFILTER_XTABLES_LEGACY 1926/** 1927 * xt_percpu_counter_alloc - allocate x_tables rule counter 1928 * 1929 * @state: pointer to xt_percpu allocation state 1930 * @counter: pointer to counter struct inside the ip(6)/arpt_entry struct 1931 * 1932 * On SMP, the packet counter [ ip(6)t_entry->counters.pcnt ] will then 1933 * contain the address of the real (percpu) counter. 1934 * 1935 * Rule evaluation needs to use xt_get_this_cpu_counter() helper 1936 * to fetch the real percpu counter. 1937 * 1938 * To speed up allocation and improve data locality, a 4kb block is 1939 * allocated. Freeing any counter may free an entire block, so all 1940 * counters allocated using the same state must be freed at the same 1941 * time. 1942 * 1943 * xt_percpu_counter_alloc_state contains the base address of the 1944 * allocated page and the current sub-offset. 1945 * 1946 * returns false on error. 1947 */ 1948bool xt_percpu_counter_alloc(struct xt_percpu_counter_alloc_state *state, 1949 struct xt_counters *counter) 1950{ 1951 BUILD_BUG_ON(XT_PCPU_BLOCK_SIZE < (sizeof(*counter) * 2)); 1952 1953 if (nr_cpu_ids <= 1) 1954 return true; 1955 1956 if (!state->mem) { 1957 state->mem = __alloc_percpu(XT_PCPU_BLOCK_SIZE, 1958 XT_PCPU_BLOCK_SIZE); 1959 if (!state->mem) 1960 return false; 1961 } 1962 counter->pcnt = (__force unsigned long)(state->mem + state->off); 1963 state->off += sizeof(*counter); 1964 if (state->off > (XT_PCPU_BLOCK_SIZE - sizeof(*counter))) { 1965 state->mem = NULL; 1966 state->off = 0; 1967 } 1968 return true; 1969} 1970EXPORT_SYMBOL_GPL(xt_percpu_counter_alloc); 1971 1972void xt_percpu_counter_free(struct xt_counters *counters) 1973{ 1974 unsigned long pcnt = counters->pcnt; 1975 1976 if (nr_cpu_ids > 1 && (pcnt & (XT_PCPU_BLOCK_SIZE - 1)) == 0) 1977 free_percpu((void __percpu *)pcnt); 1978} 1979EXPORT_SYMBOL_GPL(xt_percpu_counter_free); 1980#endif 1981 1982static int __net_init xt_net_init(struct net *net) 1983{ 1984 struct xt_pernet *xt_net = net_generic(net, xt_pernet_id); 1985 int i; 1986 1987 for (i = 0; i < NFPROTO_NUMPROTO; i++) 1988 INIT_LIST_HEAD(&xt_net->tables[i]); 1989 return 0; 1990} 1991 1992static void __net_exit xt_net_exit(struct net *net) 1993{ 1994 struct xt_pernet *xt_net = net_generic(net, xt_pernet_id); 1995 int i; 1996 1997 for (i = 0; i < NFPROTO_NUMPROTO; i++) 1998 WARN_ON_ONCE(!list_empty(&xt_net->tables[i])); 1999} 2000 2001static struct pernet_operations xt_net_ops = { 2002 .init = xt_net_init, 2003 .exit = xt_net_exit, 2004 .id = &xt_pernet_id, 2005 .size = sizeof(struct xt_pernet), 2006}; 2007 2008static int __init xt_init(void) 2009{ 2010 unsigned int i; 2011 int rv; 2012 2013 if (IS_ENABLED(CONFIG_NETFILTER_XTABLES_LEGACY)) { 2014 for_each_possible_cpu(i) { 2015 seqcount_init(&per_cpu(xt_recseq, i)); 2016 } 2017 } 2018 2019 xt = kcalloc(NFPROTO_NUMPROTO, sizeof(struct xt_af), GFP_KERNEL); 2020 if (!xt) 2021 return -ENOMEM; 2022 2023 for (i = 0; i < NFPROTO_NUMPROTO; i++) { 2024 mutex_init(&xt[i].mutex); 2025#ifdef CONFIG_NETFILTER_XTABLES_COMPAT 2026 mutex_init(&xt[i].compat_mutex); 2027 xt[i].compat_tab = NULL; 2028#endif 2029 INIT_LIST_HEAD(&xt[i].target); 2030 INIT_LIST_HEAD(&xt[i].match); 2031 INIT_LIST_HEAD(&xt_templates[i]); 2032 } 2033 rv = register_pernet_subsys(&xt_net_ops); 2034 if (rv < 0) 2035 kfree(xt); 2036 return rv; 2037} 2038 2039static void __exit xt_fini(void) 2040{ 2041 unregister_pernet_subsys(&xt_net_ops); 2042 kfree(xt); 2043} 2044 2045module_init(xt_init); 2046module_exit(xt_fini);
~/
Read-only file preview.