1 /* 2 * x_tables core - Backend for {ip,ip6,arp}_tables 3 * 4 * Copyright (C) 2006-2006 Harald Welte <laforge@netfilter.org> 5 * Copyright (C) 2006-2012 Patrick McHardy <kaber@trash.net> 6 * 7 * Based on existing ip_tables code which is 8 * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling 9 * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org> 10 * 11 * This program is free software; you can redistribute it and/or modify 12 * it under the terms of the GNU General Public License version 2 as 13 * published by the Free Software Foundation. 14 * 15 */ 16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 17 #include <linux/kernel.h> 18 #include <linux/module.h> 19 #include <linux/socket.h> 20 #include <linux/net.h> 21 #include <linux/proc_fs.h> 22 #include <linux/seq_file.h> 23 #include <linux/string.h> 24 #include <linux/vmalloc.h> 25 #include <linux/mutex.h> 26 #include <linux/mm.h> 27 #include <linux/slab.h> 28 #include <linux/audit.h> 29 #include <linux/user_namespace.h> 30 #include <net/net_namespace.h> 31 32 #include <linux/netfilter/x_tables.h> 33 #include <linux/netfilter_arp.h> 34 #include <linux/netfilter_ipv4/ip_tables.h> 35 #include <linux/netfilter_ipv6/ip6_tables.h> 36 #include <linux/netfilter_arp/arp_tables.h> 37 38 MODULE_LICENSE("GPL"); 39 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>"); 40 MODULE_DESCRIPTION("{ip,ip6,arp,eb}_tables backend module"); 41 42 #define SMP_ALIGN(x) (((x) + SMP_CACHE_BYTES-1) & ~(SMP_CACHE_BYTES-1)) 43 44 struct compat_delta { 45 unsigned int offset; /* offset in kernel */ 46 int delta; /* delta in 32bit user land */ 47 }; 48 49 struct xt_af { 50 struct mutex mutex; 51 struct list_head match; 52 struct list_head target; 53 #ifdef CONFIG_COMPAT 54 struct mutex compat_mutex; 55 struct compat_delta *compat_tab; 56 unsigned int number; /* number of slots in compat_tab[] */ 57 unsigned int cur; /* number of used slots in compat_tab[] */ 58 #endif 59 }; 60 61 static struct xt_af *xt; 62 63 static const char *const xt_prefix[NFPROTO_NUMPROTO] = { 64 [NFPROTO_UNSPEC] = "x", 65 [NFPROTO_IPV4] = "ip", 66 [NFPROTO_ARP] = "arp", 67 [NFPROTO_BRIDGE] = "eb", 68 [NFPROTO_IPV6] = "ip6", 69 }; 70 71 /* Registration hooks for targets. */ 72 int xt_register_target(struct xt_target *target) 73 { 74 u_int8_t af = target->family; 75 76 mutex_lock(&xt[af].mutex); 77 list_add(&target->list, &xt[af].target); 78 mutex_unlock(&xt[af].mutex); 79 return 0; 80 } 81 EXPORT_SYMBOL(xt_register_target); 82 83 void 84 xt_unregister_target(struct xt_target *target) 85 { 86 u_int8_t af = target->family; 87 88 mutex_lock(&xt[af].mutex); 89 list_del(&target->list); 90 mutex_unlock(&xt[af].mutex); 91 } 92 EXPORT_SYMBOL(xt_unregister_target); 93 94 int 95 xt_register_targets(struct xt_target *target, unsigned int n) 96 { 97 unsigned int i; 98 int err = 0; 99 100 for (i = 0; i < n; i++) { 101 err = xt_register_target(&target[i]); 102 if (err) 103 goto err; 104 } 105 return err; 106 107 err: 108 if (i > 0) 109 xt_unregister_targets(target, i); 110 return err; 111 } 112 EXPORT_SYMBOL(xt_register_targets); 113 114 void 115 xt_unregister_targets(struct xt_target *target, unsigned int n) 116 { 117 while (n-- > 0) 118 xt_unregister_target(&target[n]); 119 } 120 EXPORT_SYMBOL(xt_unregister_targets); 121 122 int xt_register_match(struct xt_match *match) 123 { 124 u_int8_t af = match->family; 125 126 mutex_lock(&xt[af].mutex); 127 list_add(&match->list, &xt[af].match); 128 mutex_unlock(&xt[af].mutex); 129 return 0; 130 } 131 EXPORT_SYMBOL(xt_register_match); 132 133 void 134 xt_unregister_match(struct xt_match *match) 135 { 136 u_int8_t af = match->family; 137 138 mutex_lock(&xt[af].mutex); 139 list_del(&match->list); 140 mutex_unlock(&xt[af].mutex); 141 } 142 EXPORT_SYMBOL(xt_unregister_match); 143 144 int 145 xt_register_matches(struct xt_match *match, unsigned int n) 146 { 147 unsigned int i; 148 int err = 0; 149 150 for (i = 0; i < n; i++) { 151 err = xt_register_match(&match[i]); 152 if (err) 153 goto err; 154 } 155 return err; 156 157 err: 158 if (i > 0) 159 xt_unregister_matches(match, i); 160 return err; 161 } 162 EXPORT_SYMBOL(xt_register_matches); 163 164 void 165 xt_unregister_matches(struct xt_match *match, unsigned int n) 166 { 167 while (n-- > 0) 168 xt_unregister_match(&match[n]); 169 } 170 EXPORT_SYMBOL(xt_unregister_matches); 171 172 173 /* 174 * These are weird, but module loading must not be done with mutex 175 * held (since they will register), and we have to have a single 176 * function to use. 177 */ 178 179 /* Find match, grabs ref. Returns ERR_PTR() on error. */ 180 struct xt_match *xt_find_match(u8 af, const char *name, u8 revision) 181 { 182 struct xt_match *m; 183 int err = -ENOENT; 184 185 mutex_lock(&xt[af].mutex); 186 list_for_each_entry(m, &xt[af].match, list) { 187 if (strcmp(m->name, name) == 0) { 188 if (m->revision == revision) { 189 if (try_module_get(m->me)) { 190 mutex_unlock(&xt[af].mutex); 191 return m; 192 } 193 } else 194 err = -EPROTOTYPE; /* Found something. */ 195 } 196 } 197 mutex_unlock(&xt[af].mutex); 198 199 if (af != NFPROTO_UNSPEC) 200 /* Try searching again in the family-independent list */ 201 return xt_find_match(NFPROTO_UNSPEC, name, revision); 202 203 return ERR_PTR(err); 204 } 205 EXPORT_SYMBOL(xt_find_match); 206 207 struct xt_match * 208 xt_request_find_match(uint8_t nfproto, const char *name, uint8_t revision) 209 { 210 struct xt_match *match; 211 212 match = xt_find_match(nfproto, name, revision); 213 if (IS_ERR(match)) { 214 request_module("%st_%s", xt_prefix[nfproto], name); 215 match = xt_find_match(nfproto, name, revision); 216 } 217 218 return match; 219 } 220 EXPORT_SYMBOL_GPL(xt_request_find_match); 221 222 /* Find target, grabs ref. Returns ERR_PTR() on error. */ 223 struct xt_target *xt_find_target(u8 af, const char *name, u8 revision) 224 { 225 struct xt_target *t; 226 int err = -ENOENT; 227 228 mutex_lock(&xt[af].mutex); 229 list_for_each_entry(t, &xt[af].target, list) { 230 if (strcmp(t->name, name) == 0) { 231 if (t->revision == revision) { 232 if (try_module_get(t->me)) { 233 mutex_unlock(&xt[af].mutex); 234 return t; 235 } 236 } else 237 err = -EPROTOTYPE; /* Found something. */ 238 } 239 } 240 mutex_unlock(&xt[af].mutex); 241 242 if (af != NFPROTO_UNSPEC) 243 /* Try searching again in the family-independent list */ 244 return xt_find_target(NFPROTO_UNSPEC, name, revision); 245 246 return ERR_PTR(err); 247 } 248 EXPORT_SYMBOL(xt_find_target); 249 250 struct xt_target *xt_request_find_target(u8 af, const char *name, u8 revision) 251 { 252 struct xt_target *target; 253 254 target = xt_find_target(af, name, revision); 255 if (IS_ERR(target)) { 256 request_module("%st_%s", xt_prefix[af], name); 257 target = xt_find_target(af, name, revision); 258 } 259 260 return target; 261 } 262 EXPORT_SYMBOL_GPL(xt_request_find_target); 263 264 static int match_revfn(u8 af, const char *name, u8 revision, int *bestp) 265 { 266 const struct xt_match *m; 267 int have_rev = 0; 268 269 list_for_each_entry(m, &xt[af].match, list) { 270 if (strcmp(m->name, name) == 0) { 271 if (m->revision > *bestp) 272 *bestp = m->revision; 273 if (m->revision == revision) 274 have_rev = 1; 275 } 276 } 277 278 if (af != NFPROTO_UNSPEC && !have_rev) 279 return match_revfn(NFPROTO_UNSPEC, name, revision, bestp); 280 281 return have_rev; 282 } 283 284 static int target_revfn(u8 af, const char *name, u8 revision, int *bestp) 285 { 286 const struct xt_target *t; 287 int have_rev = 0; 288 289 list_for_each_entry(t, &xt[af].target, list) { 290 if (strcmp(t->name, name) == 0) { 291 if (t->revision > *bestp) 292 *bestp = t->revision; 293 if (t->revision == revision) 294 have_rev = 1; 295 } 296 } 297 298 if (af != NFPROTO_UNSPEC && !have_rev) 299 return target_revfn(NFPROTO_UNSPEC, name, revision, bestp); 300 301 return have_rev; 302 } 303 304 /* Returns true or false (if no such extension at all) */ 305 int xt_find_revision(u8 af, const char *name, u8 revision, int target, 306 int *err) 307 { 308 int have_rev, best = -1; 309 310 mutex_lock(&xt[af].mutex); 311 if (target == 1) 312 have_rev = target_revfn(af, name, revision, &best); 313 else 314 have_rev = match_revfn(af, name, revision, &best); 315 mutex_unlock(&xt[af].mutex); 316 317 /* Nothing at all? Return 0 to try loading module. */ 318 if (best == -1) { 319 *err = -ENOENT; 320 return 0; 321 } 322 323 *err = best; 324 if (!have_rev) 325 *err = -EPROTONOSUPPORT; 326 return 1; 327 } 328 EXPORT_SYMBOL_GPL(xt_find_revision); 329 330 static char * 331 textify_hooks(char *buf, size_t size, unsigned int mask, uint8_t nfproto) 332 { 333 static const char *const inetbr_names[] = { 334 "PREROUTING", "INPUT", "FORWARD", 335 "OUTPUT", "POSTROUTING", "BROUTING", 336 }; 337 static const char *const arp_names[] = { 338 "INPUT", "FORWARD", "OUTPUT", 339 }; 340 const char *const *names; 341 unsigned int i, max; 342 char *p = buf; 343 bool np = false; 344 int res; 345 346 names = (nfproto == NFPROTO_ARP) ? arp_names : inetbr_names; 347 max = (nfproto == NFPROTO_ARP) ? ARRAY_SIZE(arp_names) : 348 ARRAY_SIZE(inetbr_names); 349 *p = '\0'; 350 for (i = 0; i < max; ++i) { 351 if (!(mask & (1 << i))) 352 continue; 353 res = snprintf(p, size, "%s%s", np ? "/" : "", names[i]); 354 if (res > 0) { 355 size -= res; 356 p += res; 357 } 358 np = true; 359 } 360 361 return buf; 362 } 363 364 int xt_check_match(struct xt_mtchk_param *par, 365 unsigned int size, u_int8_t proto, bool inv_proto) 366 { 367 int ret; 368 369 if (XT_ALIGN(par->match->matchsize) != size && 370 par->match->matchsize != -1) { 371 /* 372 * ebt_among is exempt from centralized matchsize checking 373 * because it uses a dynamic-size data set. 374 */ 375 pr_err("%s_tables: %s.%u match: invalid size " 376 "%u (kernel) != (user) %u\n", 377 xt_prefix[par->family], par->match->name, 378 par->match->revision, 379 XT_ALIGN(par->match->matchsize), size); 380 return -EINVAL; 381 } 382 if (par->match->table != NULL && 383 strcmp(par->match->table, par->table) != 0) { 384 pr_err("%s_tables: %s match: only valid in %s table, not %s\n", 385 xt_prefix[par->family], par->match->name, 386 par->match->table, par->table); 387 return -EINVAL; 388 } 389 if (par->match->hooks && (par->hook_mask & ~par->match->hooks) != 0) { 390 char used[64], allow[64]; 391 392 pr_err("%s_tables: %s match: used from hooks %s, but only " 393 "valid from %s\n", 394 xt_prefix[par->family], par->match->name, 395 textify_hooks(used, sizeof(used), par->hook_mask, 396 par->family), 397 textify_hooks(allow, sizeof(allow), par->match->hooks, 398 par->family)); 399 return -EINVAL; 400 } 401 if (par->match->proto && (par->match->proto != proto || inv_proto)) { 402 pr_err("%s_tables: %s match: only valid for protocol %u\n", 403 xt_prefix[par->family], par->match->name, 404 par->match->proto); 405 return -EINVAL; 406 } 407 if (par->match->checkentry != NULL) { 408 ret = par->match->checkentry(par); 409 if (ret < 0) 410 return ret; 411 else if (ret > 0) 412 /* Flag up potential errors. */ 413 return -EIO; 414 } 415 return 0; 416 } 417 EXPORT_SYMBOL_GPL(xt_check_match); 418 419 #ifdef CONFIG_COMPAT 420 int xt_compat_add_offset(u_int8_t af, unsigned int offset, int delta) 421 { 422 struct xt_af *xp = &xt[af]; 423 424 if (!xp->compat_tab) { 425 if (!xp->number) 426 return -EINVAL; 427 xp->compat_tab = vmalloc(sizeof(struct compat_delta) * xp->number); 428 if (!xp->compat_tab) 429 return -ENOMEM; 430 xp->cur = 0; 431 } 432 433 if (xp->cur >= xp->number) 434 return -EINVAL; 435 436 if (xp->cur) 437 delta += xp->compat_tab[xp->cur - 1].delta; 438 xp->compat_tab[xp->cur].offset = offset; 439 xp->compat_tab[xp->cur].delta = delta; 440 xp->cur++; 441 return 0; 442 } 443 EXPORT_SYMBOL_GPL(xt_compat_add_offset); 444 445 void xt_compat_flush_offsets(u_int8_t af) 446 { 447 if (xt[af].compat_tab) { 448 vfree(xt[af].compat_tab); 449 xt[af].compat_tab = NULL; 450 xt[af].number = 0; 451 xt[af].cur = 0; 452 } 453 } 454 EXPORT_SYMBOL_GPL(xt_compat_flush_offsets); 455 456 int xt_compat_calc_jump(u_int8_t af, unsigned int offset) 457 { 458 struct compat_delta *tmp = xt[af].compat_tab; 459 int mid, left = 0, right = xt[af].cur - 1; 460 461 while (left <= right) { 462 mid = (left + right) >> 1; 463 if (offset > tmp[mid].offset) 464 left = mid + 1; 465 else if (offset < tmp[mid].offset) 466 right = mid - 1; 467 else 468 return mid ? tmp[mid - 1].delta : 0; 469 } 470 return left ? tmp[left - 1].delta : 0; 471 } 472 EXPORT_SYMBOL_GPL(xt_compat_calc_jump); 473 474 void xt_compat_init_offsets(u_int8_t af, unsigned int number) 475 { 476 xt[af].number = number; 477 xt[af].cur = 0; 478 } 479 EXPORT_SYMBOL(xt_compat_init_offsets); 480 481 int xt_compat_match_offset(const struct xt_match *match) 482 { 483 u_int16_t csize = match->compatsize ? : match->matchsize; 484 return XT_ALIGN(match->matchsize) - COMPAT_XT_ALIGN(csize); 485 } 486 EXPORT_SYMBOL_GPL(xt_compat_match_offset); 487 488 int xt_compat_match_from_user(struct xt_entry_match *m, void **dstptr, 489 unsigned int *size) 490 { 491 const struct xt_match *match = m->u.kernel.match; 492 struct compat_xt_entry_match *cm = (struct compat_xt_entry_match *)m; 493 int pad, off = xt_compat_match_offset(match); 494 u_int16_t msize = cm->u.user.match_size; 495 496 m = *dstptr; 497 memcpy(m, cm, sizeof(*cm)); 498 if (match->compat_from_user) 499 match->compat_from_user(m->data, cm->data); 500 else 501 memcpy(m->data, cm->data, msize - sizeof(*cm)); 502 pad = XT_ALIGN(match->matchsize) - match->matchsize; 503 if (pad > 0) 504 memset(m->data + match->matchsize, 0, pad); 505 506 msize += off; 507 m->u.user.match_size = msize; 508 509 *size += off; 510 *dstptr += msize; 511 return 0; 512 } 513 EXPORT_SYMBOL_GPL(xt_compat_match_from_user); 514 515 int xt_compat_match_to_user(const struct xt_entry_match *m, 516 void __user **dstptr, unsigned int *size) 517 { 518 const struct xt_match *match = m->u.kernel.match; 519 struct compat_xt_entry_match __user *cm = *dstptr; 520 int off = xt_compat_match_offset(match); 521 u_int16_t msize = m->u.user.match_size - off; 522 523 if (copy_to_user(cm, m, sizeof(*cm)) || 524 put_user(msize, &cm->u.user.match_size) || 525 copy_to_user(cm->u.user.name, m->u.kernel.match->name, 526 strlen(m->u.kernel.match->name) + 1)) 527 return -EFAULT; 528 529 if (match->compat_to_user) { 530 if (match->compat_to_user((void __user *)cm->data, m->data)) 531 return -EFAULT; 532 } else { 533 if (copy_to_user(cm->data, m->data, msize - sizeof(*cm))) 534 return -EFAULT; 535 } 536 537 *size -= off; 538 *dstptr += msize; 539 return 0; 540 } 541 EXPORT_SYMBOL_GPL(xt_compat_match_to_user); 542 #endif /* CONFIG_COMPAT */ 543 544 int xt_check_target(struct xt_tgchk_param *par, 545 unsigned int size, u_int8_t proto, bool inv_proto) 546 { 547 int ret; 548 549 if (XT_ALIGN(par->target->targetsize) != size) { 550 pr_err("%s_tables: %s.%u target: invalid size " 551 "%u (kernel) != (user) %u\n", 552 xt_prefix[par->family], par->target->name, 553 par->target->revision, 554 XT_ALIGN(par->target->targetsize), size); 555 return -EINVAL; 556 } 557 if (par->target->table != NULL && 558 strcmp(par->target->table, par->table) != 0) { 559 pr_err("%s_tables: %s target: only valid in %s table, not %s\n", 560 xt_prefix[par->family], par->target->name, 561 par->target->table, par->table); 562 return -EINVAL; 563 } 564 if (par->target->hooks && (par->hook_mask & ~par->target->hooks) != 0) { 565 char used[64], allow[64]; 566 567 pr_err("%s_tables: %s target: used from hooks %s, but only " 568 "usable from %s\n", 569 xt_prefix[par->family], par->target->name, 570 textify_hooks(used, sizeof(used), par->hook_mask, 571 par->family), 572 textify_hooks(allow, sizeof(allow), par->target->hooks, 573 par->family)); 574 return -EINVAL; 575 } 576 if (par->target->proto && (par->target->proto != proto || inv_proto)) { 577 pr_err("%s_tables: %s target: only valid for protocol %u\n", 578 xt_prefix[par->family], par->target->name, 579 par->target->proto); 580 return -EINVAL; 581 } 582 if (par->target->checkentry != NULL) { 583 ret = par->target->checkentry(par); 584 if (ret < 0) 585 return ret; 586 else if (ret > 0) 587 /* Flag up potential errors. */ 588 return -EIO; 589 } 590 return 0; 591 } 592 EXPORT_SYMBOL_GPL(xt_check_target); 593 594 #ifdef CONFIG_COMPAT 595 int xt_compat_target_offset(const struct xt_target *target) 596 { 597 u_int16_t csize = target->compatsize ? : target->targetsize; 598 return XT_ALIGN(target->targetsize) - COMPAT_XT_ALIGN(csize); 599 } 600 EXPORT_SYMBOL_GPL(xt_compat_target_offset); 601 602 void xt_compat_target_from_user(struct xt_entry_target *t, void **dstptr, 603 unsigned int *size) 604 { 605 const struct xt_target *target = t->u.kernel.target; 606 struct compat_xt_entry_target *ct = (struct compat_xt_entry_target *)t; 607 int pad, off = xt_compat_target_offset(target); 608 u_int16_t tsize = ct->u.user.target_size; 609 610 t = *dstptr; 611 memcpy(t, ct, sizeof(*ct)); 612 if (target->compat_from_user) 613 target->compat_from_user(t->data, ct->data); 614 else 615 memcpy(t->data, ct->data, tsize - sizeof(*ct)); 616 pad = XT_ALIGN(target->targetsize) - target->targetsize; 617 if (pad > 0) 618 memset(t->data + target->targetsize, 0, pad); 619 620 tsize += off; 621 t->u.user.target_size = tsize; 622 623 *size += off; 624 *dstptr += tsize; 625 } 626 EXPORT_SYMBOL_GPL(xt_compat_target_from_user); 627 628 int xt_compat_target_to_user(const struct xt_entry_target *t, 629 void __user **dstptr, unsigned int *size) 630 { 631 const struct xt_target *target = t->u.kernel.target; 632 struct compat_xt_entry_target __user *ct = *dstptr; 633 int off = xt_compat_target_offset(target); 634 u_int16_t tsize = t->u.user.target_size - off; 635 636 if (copy_to_user(ct, t, sizeof(*ct)) || 637 put_user(tsize, &ct->u.user.target_size) || 638 copy_to_user(ct->u.user.name, t->u.kernel.target->name, 639 strlen(t->u.kernel.target->name) + 1)) 640 return -EFAULT; 641 642 if (target->compat_to_user) { 643 if (target->compat_to_user((void __user *)ct->data, t->data)) 644 return -EFAULT; 645 } else { 646 if (copy_to_user(ct->data, t->data, tsize - sizeof(*ct))) 647 return -EFAULT; 648 } 649 650 *size -= off; 651 *dstptr += tsize; 652 return 0; 653 } 654 EXPORT_SYMBOL_GPL(xt_compat_target_to_user); 655 #endif 656 657 struct xt_table_info *xt_alloc_table_info(unsigned int size) 658 { 659 struct xt_table_info *info = NULL; 660 size_t sz = sizeof(*info) + size; 661 662 if (sz < sizeof(*info)) 663 return NULL; 664 665 /* Pedantry: prevent them from hitting BUG() in vmalloc.c --RR */ 666 if ((SMP_ALIGN(size) >> PAGE_SHIFT) + 2 > totalram_pages) 667 return NULL; 668 669 if (sz <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) 670 info = kmalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_NORETRY); 671 if (!info) { 672 info = vmalloc(sz); 673 if (!info) 674 return NULL; 675 } 676 memset(info, 0, sizeof(*info)); 677 info->size = size; 678 return info; 679 } 680 EXPORT_SYMBOL(xt_alloc_table_info); 681 682 void xt_free_table_info(struct xt_table_info *info) 683 { 684 int cpu; 685 686 if (info->jumpstack != NULL) { 687 for_each_possible_cpu(cpu) 688 kvfree(info->jumpstack[cpu]); 689 kvfree(info->jumpstack); 690 } 691 692 kvfree(info); 693 } 694 EXPORT_SYMBOL(xt_free_table_info); 695 696 /* Find table by name, grabs mutex & ref. Returns ERR_PTR() on error. */ 697 struct xt_table *xt_find_table_lock(struct net *net, u_int8_t af, 698 const char *name) 699 { 700 struct xt_table *t, *found = NULL; 701 702 mutex_lock(&xt[af].mutex); 703 list_for_each_entry(t, &net->xt.tables[af], list) 704 if (strcmp(t->name, name) == 0 && try_module_get(t->me)) 705 return t; 706 707 if (net == &init_net) 708 goto out; 709 710 /* Table doesn't exist in this netns, re-try init */ 711 list_for_each_entry(t, &init_net.xt.tables[af], list) { 712 if (strcmp(t->name, name)) 713 continue; 714 if (!try_module_get(t->me)) 715 return NULL; 716 717 mutex_unlock(&xt[af].mutex); 718 if (t->table_init(net) != 0) { 719 module_put(t->me); 720 return NULL; 721 } 722 723 found = t; 724 725 mutex_lock(&xt[af].mutex); 726 break; 727 } 728 729 if (!found) 730 goto out; 731 732 /* and once again: */ 733 list_for_each_entry(t, &net->xt.tables[af], list) 734 if (strcmp(t->name, name) == 0) 735 return t; 736 737 module_put(found->me); 738 out: 739 mutex_unlock(&xt[af].mutex); 740 return NULL; 741 } 742 EXPORT_SYMBOL_GPL(xt_find_table_lock); 743 744 void xt_table_unlock(struct xt_table *table) 745 { 746 mutex_unlock(&xt[table->af].mutex); 747 } 748 EXPORT_SYMBOL_GPL(xt_table_unlock); 749 750 #ifdef CONFIG_COMPAT 751 void xt_compat_lock(u_int8_t af) 752 { 753 mutex_lock(&xt[af].compat_mutex); 754 } 755 EXPORT_SYMBOL_GPL(xt_compat_lock); 756 757 void xt_compat_unlock(u_int8_t af) 758 { 759 mutex_unlock(&xt[af].compat_mutex); 760 } 761 EXPORT_SYMBOL_GPL(xt_compat_unlock); 762 #endif 763 764 DEFINE_PER_CPU(seqcount_t, xt_recseq); 765 EXPORT_PER_CPU_SYMBOL_GPL(xt_recseq); 766 767 struct static_key xt_tee_enabled __read_mostly; 768 EXPORT_SYMBOL_GPL(xt_tee_enabled); 769 770 static int xt_jumpstack_alloc(struct xt_table_info *i) 771 { 772 unsigned int size; 773 int cpu; 774 775 size = sizeof(void **) * nr_cpu_ids; 776 if (size > PAGE_SIZE) 777 i->jumpstack = vzalloc(size); 778 else 779 i->jumpstack = kzalloc(size, GFP_KERNEL); 780 if (i->jumpstack == NULL) 781 return -ENOMEM; 782 783 /* ruleset without jumps -- no stack needed */ 784 if (i->stacksize == 0) 785 return 0; 786 787 /* Jumpstack needs to be able to record two full callchains, one 788 * from the first rule set traversal, plus one table reentrancy 789 * via -j TEE without clobbering the callchain that brought us to 790 * TEE target. 791 * 792 * This is done by allocating two jumpstacks per cpu, on reentry 793 * the upper half of the stack is used. 794 * 795 * see the jumpstack setup in ipt_do_table() for more details. 796 */ 797 size = sizeof(void *) * i->stacksize * 2u; 798 for_each_possible_cpu(cpu) { 799 if (size > PAGE_SIZE) 800 i->jumpstack[cpu] = vmalloc_node(size, 801 cpu_to_node(cpu)); 802 else 803 i->jumpstack[cpu] = kmalloc_node(size, 804 GFP_KERNEL, cpu_to_node(cpu)); 805 if (i->jumpstack[cpu] == NULL) 806 /* 807 * Freeing will be done later on by the callers. The 808 * chain is: xt_replace_table -> __do_replace -> 809 * do_replace -> xt_free_table_info. 810 */ 811 return -ENOMEM; 812 } 813 814 return 0; 815 } 816 817 struct xt_table_info * 818 xt_replace_table(struct xt_table *table, 819 unsigned int num_counters, 820 struct xt_table_info *newinfo, 821 int *error) 822 { 823 struct xt_table_info *private; 824 int ret; 825 826 ret = xt_jumpstack_alloc(newinfo); 827 if (ret < 0) { 828 *error = ret; 829 return NULL; 830 } 831 832 /* Do the substitution. */ 833 local_bh_disable(); 834 private = table->private; 835 836 /* Check inside lock: is the old number correct? */ 837 if (num_counters != private->number) { 838 pr_debug("num_counters != table->private->number (%u/%u)\n", 839 num_counters, private->number); 840 local_bh_enable(); 841 *error = -EAGAIN; 842 return NULL; 843 } 844 845 newinfo->initial_entries = private->initial_entries; 846 /* 847 * Ensure contents of newinfo are visible before assigning to 848 * private. 849 */ 850 smp_wmb(); 851 table->private = newinfo; 852 853 /* 854 * Even though table entries have now been swapped, other CPU's 855 * may still be using the old entries. This is okay, because 856 * resynchronization happens because of the locking done 857 * during the get_counters() routine. 858 */ 859 local_bh_enable(); 860 861 #ifdef CONFIG_AUDIT 862 if (audit_enabled) { 863 struct audit_buffer *ab; 864 865 ab = audit_log_start(current->audit_context, GFP_KERNEL, 866 AUDIT_NETFILTER_CFG); 867 if (ab) { 868 audit_log_format(ab, "table=%s family=%u entries=%u", 869 table->name, table->af, 870 private->number); 871 audit_log_end(ab); 872 } 873 } 874 #endif 875 876 return private; 877 } 878 EXPORT_SYMBOL_GPL(xt_replace_table); 879 880 struct xt_table *xt_register_table(struct net *net, 881 const struct xt_table *input_table, 882 struct xt_table_info *bootstrap, 883 struct xt_table_info *newinfo) 884 { 885 int ret; 886 struct xt_table_info *private; 887 struct xt_table *t, *table; 888 889 /* Don't add one object to multiple lists. */ 890 table = kmemdup(input_table, sizeof(struct xt_table), GFP_KERNEL); 891 if (!table) { 892 ret = -ENOMEM; 893 goto out; 894 } 895 896 mutex_lock(&xt[table->af].mutex); 897 /* Don't autoload: we'd eat our tail... */ 898 list_for_each_entry(t, &net->xt.tables[table->af], list) { 899 if (strcmp(t->name, table->name) == 0) { 900 ret = -EEXIST; 901 goto unlock; 902 } 903 } 904 905 /* Simplifies replace_table code. */ 906 table->private = bootstrap; 907 908 if (!xt_replace_table(table, 0, newinfo, &ret)) 909 goto unlock; 910 911 private = table->private; 912 pr_debug("table->private->number = %u\n", private->number); 913 914 /* save number of initial entries */ 915 private->initial_entries = private->number; 916 917 list_add(&table->list, &net->xt.tables[table->af]); 918 mutex_unlock(&xt[table->af].mutex); 919 return table; 920 921 unlock: 922 mutex_unlock(&xt[table->af].mutex); 923 kfree(table); 924 out: 925 return ERR_PTR(ret); 926 } 927 EXPORT_SYMBOL_GPL(xt_register_table); 928 929 void *xt_unregister_table(struct xt_table *table) 930 { 931 struct xt_table_info *private; 932 933 mutex_lock(&xt[table->af].mutex); 934 private = table->private; 935 list_del(&table->list); 936 mutex_unlock(&xt[table->af].mutex); 937 kfree(table); 938 939 return private; 940 } 941 EXPORT_SYMBOL_GPL(xt_unregister_table); 942 943 #ifdef CONFIG_PROC_FS 944 struct xt_names_priv { 945 struct seq_net_private p; 946 u_int8_t af; 947 }; 948 static void *xt_table_seq_start(struct seq_file *seq, loff_t *pos) 949 { 950 struct xt_names_priv *priv = seq->private; 951 struct net *net = seq_file_net(seq); 952 u_int8_t af = priv->af; 953 954 mutex_lock(&xt[af].mutex); 955 return seq_list_start(&net->xt.tables[af], *pos); 956 } 957 958 static void *xt_table_seq_next(struct seq_file *seq, void *v, loff_t *pos) 959 { 960 struct xt_names_priv *priv = seq->private; 961 struct net *net = seq_file_net(seq); 962 u_int8_t af = priv->af; 963 964 return seq_list_next(v, &net->xt.tables[af], pos); 965 } 966 967 static void xt_table_seq_stop(struct seq_file *seq, void *v) 968 { 969 struct xt_names_priv *priv = seq->private; 970 u_int8_t af = priv->af; 971 972 mutex_unlock(&xt[af].mutex); 973 } 974 975 static int xt_table_seq_show(struct seq_file *seq, void *v) 976 { 977 struct xt_table *table = list_entry(v, struct xt_table, list); 978 979 if (*table->name) 980 seq_printf(seq, "%s\n", table->name); 981 return 0; 982 } 983 984 static const struct seq_operations xt_table_seq_ops = { 985 .start = xt_table_seq_start, 986 .next = xt_table_seq_next, 987 .stop = xt_table_seq_stop, 988 .show = xt_table_seq_show, 989 }; 990 991 static int xt_table_open(struct inode *inode, struct file *file) 992 { 993 int ret; 994 struct xt_names_priv *priv; 995 996 ret = seq_open_net(inode, file, &xt_table_seq_ops, 997 sizeof(struct xt_names_priv)); 998 if (!ret) { 999 priv = ((struct seq_file *)file->private_data)->private; 1000 priv->af = (unsigned long)PDE_DATA(inode); 1001 } 1002 return ret; 1003 } 1004 1005 static const struct file_operations xt_table_ops = { 1006 .owner = THIS_MODULE, 1007 .open = xt_table_open, 1008 .read = seq_read, 1009 .llseek = seq_lseek, 1010 .release = seq_release_net, 1011 }; 1012 1013 /* 1014 * Traverse state for ip{,6}_{tables,matches} for helping crossing 1015 * the multi-AF mutexes. 1016 */ 1017 struct nf_mttg_trav { 1018 struct list_head *head, *curr; 1019 uint8_t class, nfproto; 1020 }; 1021 1022 enum { 1023 MTTG_TRAV_INIT, 1024 MTTG_TRAV_NFP_UNSPEC, 1025 MTTG_TRAV_NFP_SPEC, 1026 MTTG_TRAV_DONE, 1027 }; 1028 1029 static void *xt_mttg_seq_next(struct seq_file *seq, void *v, loff_t *ppos, 1030 bool is_target) 1031 { 1032 static const uint8_t next_class[] = { 1033 [MTTG_TRAV_NFP_UNSPEC] = MTTG_TRAV_NFP_SPEC, 1034 [MTTG_TRAV_NFP_SPEC] = MTTG_TRAV_DONE, 1035 }; 1036 struct nf_mttg_trav *trav = seq->private; 1037 1038 switch (trav->class) { 1039 case MTTG_TRAV_INIT: 1040 trav->class = MTTG_TRAV_NFP_UNSPEC; 1041 mutex_lock(&xt[NFPROTO_UNSPEC].mutex); 1042 trav->head = trav->curr = is_target ? 1043 &xt[NFPROTO_UNSPEC].target : &xt[NFPROTO_UNSPEC].match; 1044 break; 1045 case MTTG_TRAV_NFP_UNSPEC: 1046 trav->curr = trav->curr->next; 1047 if (trav->curr != trav->head) 1048 break; 1049 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex); 1050 mutex_lock(&xt[trav->nfproto].mutex); 1051 trav->head = trav->curr = is_target ? 1052 &xt[trav->nfproto].target : &xt[trav->nfproto].match; 1053 trav->class = next_class[trav->class]; 1054 break; 1055 case MTTG_TRAV_NFP_SPEC: 1056 trav->curr = trav->curr->next; 1057 if (trav->curr != trav->head) 1058 break; 1059 /* fallthru, _stop will unlock */ 1060 default: 1061 return NULL; 1062 } 1063 1064 if (ppos != NULL) 1065 ++*ppos; 1066 return trav; 1067 } 1068 1069 static void *xt_mttg_seq_start(struct seq_file *seq, loff_t *pos, 1070 bool is_target) 1071 { 1072 struct nf_mttg_trav *trav = seq->private; 1073 unsigned int j; 1074 1075 trav->class = MTTG_TRAV_INIT; 1076 for (j = 0; j < *pos; ++j) 1077 if (xt_mttg_seq_next(seq, NULL, NULL, is_target) == NULL) 1078 return NULL; 1079 return trav; 1080 } 1081 1082 static void xt_mttg_seq_stop(struct seq_file *seq, void *v) 1083 { 1084 struct nf_mttg_trav *trav = seq->private; 1085 1086 switch (trav->class) { 1087 case MTTG_TRAV_NFP_UNSPEC: 1088 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex); 1089 break; 1090 case MTTG_TRAV_NFP_SPEC: 1091 mutex_unlock(&xt[trav->nfproto].mutex); 1092 break; 1093 } 1094 } 1095 1096 static void *xt_match_seq_start(struct seq_file *seq, loff_t *pos) 1097 { 1098 return xt_mttg_seq_start(seq, pos, false); 1099 } 1100 1101 static void *xt_match_seq_next(struct seq_file *seq, void *v, loff_t *ppos) 1102 { 1103 return xt_mttg_seq_next(seq, v, ppos, false); 1104 } 1105 1106 static int xt_match_seq_show(struct seq_file *seq, void *v) 1107 { 1108 const struct nf_mttg_trav *trav = seq->private; 1109 const struct xt_match *match; 1110 1111 switch (trav->class) { 1112 case MTTG_TRAV_NFP_UNSPEC: 1113 case MTTG_TRAV_NFP_SPEC: 1114 if (trav->curr == trav->head) 1115 return 0; 1116 match = list_entry(trav->curr, struct xt_match, list); 1117 if (*match->name) 1118 seq_printf(seq, "%s\n", match->name); 1119 } 1120 return 0; 1121 } 1122 1123 static const struct seq_operations xt_match_seq_ops = { 1124 .start = xt_match_seq_start, 1125 .next = xt_match_seq_next, 1126 .stop = xt_mttg_seq_stop, 1127 .show = xt_match_seq_show, 1128 }; 1129 1130 static int xt_match_open(struct inode *inode, struct file *file) 1131 { 1132 struct nf_mttg_trav *trav; 1133 trav = __seq_open_private(file, &xt_match_seq_ops, sizeof(*trav)); 1134 if (!trav) 1135 return -ENOMEM; 1136 1137 trav->nfproto = (unsigned long)PDE_DATA(inode); 1138 return 0; 1139 } 1140 1141 static const struct file_operations xt_match_ops = { 1142 .owner = THIS_MODULE, 1143 .open = xt_match_open, 1144 .read = seq_read, 1145 .llseek = seq_lseek, 1146 .release = seq_release_private, 1147 }; 1148 1149 static void *xt_target_seq_start(struct seq_file *seq, loff_t *pos) 1150 { 1151 return xt_mttg_seq_start(seq, pos, true); 1152 } 1153 1154 static void *xt_target_seq_next(struct seq_file *seq, void *v, loff_t *ppos) 1155 { 1156 return xt_mttg_seq_next(seq, v, ppos, true); 1157 } 1158 1159 static int xt_target_seq_show(struct seq_file *seq, void *v) 1160 { 1161 const struct nf_mttg_trav *trav = seq->private; 1162 const struct xt_target *target; 1163 1164 switch (trav->class) { 1165 case MTTG_TRAV_NFP_UNSPEC: 1166 case MTTG_TRAV_NFP_SPEC: 1167 if (trav->curr == trav->head) 1168 return 0; 1169 target = list_entry(trav->curr, struct xt_target, list); 1170 if (*target->name) 1171 seq_printf(seq, "%s\n", target->name); 1172 } 1173 return 0; 1174 } 1175 1176 static const struct seq_operations xt_target_seq_ops = { 1177 .start = xt_target_seq_start, 1178 .next = xt_target_seq_next, 1179 .stop = xt_mttg_seq_stop, 1180 .show = xt_target_seq_show, 1181 }; 1182 1183 static int xt_target_open(struct inode *inode, struct file *file) 1184 { 1185 struct nf_mttg_trav *trav; 1186 trav = __seq_open_private(file, &xt_target_seq_ops, sizeof(*trav)); 1187 if (!trav) 1188 return -ENOMEM; 1189 1190 trav->nfproto = (unsigned long)PDE_DATA(inode); 1191 return 0; 1192 } 1193 1194 static const struct file_operations xt_target_ops = { 1195 .owner = THIS_MODULE, 1196 .open = xt_target_open, 1197 .read = seq_read, 1198 .llseek = seq_lseek, 1199 .release = seq_release_private, 1200 }; 1201 1202 #define FORMAT_TABLES "_tables_names" 1203 #define FORMAT_MATCHES "_tables_matches" 1204 #define FORMAT_TARGETS "_tables_targets" 1205 1206 #endif /* CONFIG_PROC_FS */ 1207 1208 /** 1209 * xt_hook_ops_alloc - set up hooks for a new table 1210 * @table: table with metadata needed to set up hooks 1211 * @fn: Hook function 1212 * 1213 * This function will create the nf_hook_ops that the x_table needs 1214 * to hand to xt_hook_link_net(). 1215 */ 1216 struct nf_hook_ops * 1217 xt_hook_ops_alloc(const struct xt_table *table, nf_hookfn *fn) 1218 { 1219 unsigned int hook_mask = table->valid_hooks; 1220 uint8_t i, num_hooks = hweight32(hook_mask); 1221 uint8_t hooknum; 1222 struct nf_hook_ops *ops; 1223 1224 ops = kmalloc(sizeof(*ops) * num_hooks, GFP_KERNEL); 1225 if (ops == NULL) 1226 return ERR_PTR(-ENOMEM); 1227 1228 for (i = 0, hooknum = 0; i < num_hooks && hook_mask != 0; 1229 hook_mask >>= 1, ++hooknum) { 1230 if (!(hook_mask & 1)) 1231 continue; 1232 ops[i].hook = fn; 1233 ops[i].pf = table->af; 1234 ops[i].hooknum = hooknum; 1235 ops[i].priority = table->priority; 1236 ++i; 1237 } 1238 1239 return ops; 1240 } 1241 EXPORT_SYMBOL_GPL(xt_hook_ops_alloc); 1242 1243 int xt_proto_init(struct net *net, u_int8_t af) 1244 { 1245 #ifdef CONFIG_PROC_FS 1246 char buf[XT_FUNCTION_MAXNAMELEN]; 1247 struct proc_dir_entry *proc; 1248 kuid_t root_uid; 1249 kgid_t root_gid; 1250 #endif 1251 1252 if (af >= ARRAY_SIZE(xt_prefix)) 1253 return -EINVAL; 1254 1255 1256 #ifdef CONFIG_PROC_FS 1257 root_uid = make_kuid(net->user_ns, 0); 1258 root_gid = make_kgid(net->user_ns, 0); 1259 1260 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1261 strlcat(buf, FORMAT_TABLES, sizeof(buf)); 1262 proc = proc_create_data(buf, 0440, net->proc_net, &xt_table_ops, 1263 (void *)(unsigned long)af); 1264 if (!proc) 1265 goto out; 1266 if (uid_valid(root_uid) && gid_valid(root_gid)) 1267 proc_set_user(proc, root_uid, root_gid); 1268 1269 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1270 strlcat(buf, FORMAT_MATCHES, sizeof(buf)); 1271 proc = proc_create_data(buf, 0440, net->proc_net, &xt_match_ops, 1272 (void *)(unsigned long)af); 1273 if (!proc) 1274 goto out_remove_tables; 1275 if (uid_valid(root_uid) && gid_valid(root_gid)) 1276 proc_set_user(proc, root_uid, root_gid); 1277 1278 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1279 strlcat(buf, FORMAT_TARGETS, sizeof(buf)); 1280 proc = proc_create_data(buf, 0440, net->proc_net, &xt_target_ops, 1281 (void *)(unsigned long)af); 1282 if (!proc) 1283 goto out_remove_matches; 1284 if (uid_valid(root_uid) && gid_valid(root_gid)) 1285 proc_set_user(proc, root_uid, root_gid); 1286 #endif 1287 1288 return 0; 1289 1290 #ifdef CONFIG_PROC_FS 1291 out_remove_matches: 1292 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1293 strlcat(buf, FORMAT_MATCHES, sizeof(buf)); 1294 remove_proc_entry(buf, net->proc_net); 1295 1296 out_remove_tables: 1297 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1298 strlcat(buf, FORMAT_TABLES, sizeof(buf)); 1299 remove_proc_entry(buf, net->proc_net); 1300 out: 1301 return -1; 1302 #endif 1303 } 1304 EXPORT_SYMBOL_GPL(xt_proto_init); 1305 1306 void xt_proto_fini(struct net *net, u_int8_t af) 1307 { 1308 #ifdef CONFIG_PROC_FS 1309 char buf[XT_FUNCTION_MAXNAMELEN]; 1310 1311 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1312 strlcat(buf, FORMAT_TABLES, sizeof(buf)); 1313 remove_proc_entry(buf, net->proc_net); 1314 1315 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1316 strlcat(buf, FORMAT_TARGETS, sizeof(buf)); 1317 remove_proc_entry(buf, net->proc_net); 1318 1319 strlcpy(buf, xt_prefix[af], sizeof(buf)); 1320 strlcat(buf, FORMAT_MATCHES, sizeof(buf)); 1321 remove_proc_entry(buf, net->proc_net); 1322 #endif /*CONFIG_PROC_FS*/ 1323 } 1324 EXPORT_SYMBOL_GPL(xt_proto_fini); 1325 1326 static int __net_init xt_net_init(struct net *net) 1327 { 1328 int i; 1329 1330 for (i = 0; i < NFPROTO_NUMPROTO; i++) 1331 INIT_LIST_HEAD(&net->xt.tables[i]); 1332 return 0; 1333 } 1334 1335 static struct pernet_operations xt_net_ops = { 1336 .init = xt_net_init, 1337 }; 1338 1339 static int __init xt_init(void) 1340 { 1341 unsigned int i; 1342 int rv; 1343 1344 for_each_possible_cpu(i) { 1345 seqcount_init(&per_cpu(xt_recseq, i)); 1346 } 1347 1348 xt = kmalloc(sizeof(struct xt_af) * NFPROTO_NUMPROTO, GFP_KERNEL); 1349 if (!xt) 1350 return -ENOMEM; 1351 1352 for (i = 0; i < NFPROTO_NUMPROTO; i++) { 1353 mutex_init(&xt[i].mutex); 1354 #ifdef CONFIG_COMPAT 1355 mutex_init(&xt[i].compat_mutex); 1356 xt[i].compat_tab = NULL; 1357 #endif 1358 INIT_LIST_HEAD(&xt[i].target); 1359 INIT_LIST_HEAD(&xt[i].match); 1360 } 1361 rv = register_pernet_subsys(&xt_net_ops); 1362 if (rv < 0) 1363 kfree(xt); 1364 return rv; 1365 } 1366 1367 static void __exit xt_fini(void) 1368 { 1369 unregister_pernet_subsys(&xt_net_ops); 1370 kfree(xt); 1371 } 1372 1373 module_init(xt_init); 1374 module_exit(xt_fini); 1375 1376