1 /* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu> 2 * Patrick Schaaf <bof@bof.de> 3 * Copyright (C) 2003-2013 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu> 4 * 5 * This program is free software; you can redistribute it and/or modify 6 * it under the terms of the GNU General Public License version 2 as 7 * published by the Free Software Foundation. 8 */ 9 10 /* Kernel module for IP set management */ 11 12 #include <linux/init.h> 13 #include <linux/module.h> 14 #include <linux/moduleparam.h> 15 #include <linux/ip.h> 16 #include <linux/skbuff.h> 17 #include <linux/spinlock.h> 18 #include <linux/rculist.h> 19 #include <net/netlink.h> 20 #include <net/net_namespace.h> 21 #include <net/netns/generic.h> 22 23 #include <linux/netfilter.h> 24 #include <linux/netfilter/x_tables.h> 25 #include <linux/netfilter/nfnetlink.h> 26 #include <linux/netfilter/ipset/ip_set.h> 27 28 static LIST_HEAD(ip_set_type_list); /* all registered set types */ 29 static DEFINE_MUTEX(ip_set_type_mutex); /* protects ip_set_type_list */ 30 static DEFINE_RWLOCK(ip_set_ref_lock); /* protects the set refs */ 31 32 struct ip_set_net { 33 struct ip_set * __rcu *ip_set_list; /* all individual sets */ 34 ip_set_id_t ip_set_max; /* max number of sets */ 35 bool is_deleted; /* deleted by ip_set_net_exit */ 36 bool is_destroyed; /* all sets are destroyed */ 37 }; 38 39 static unsigned int ip_set_net_id __read_mostly; 40 41 static inline struct ip_set_net *ip_set_pernet(struct net *net) 42 { 43 return net_generic(net, ip_set_net_id); 44 } 45 46 #define IP_SET_INC 64 47 #define STRNCMP(a, b) (strncmp(a, b, IPSET_MAXNAMELEN) == 0) 48 49 static unsigned int max_sets; 50 51 module_param(max_sets, int, 0600); 52 MODULE_PARM_DESC(max_sets, "maximal number of sets"); 53 MODULE_LICENSE("GPL"); 54 MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>"); 55 MODULE_DESCRIPTION("core IP set support"); 56 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_IPSET); 57 58 /* When the nfnl mutex is held: */ 59 #define ip_set_dereference(p) \ 60 rcu_dereference_protected(p, 1) 61 #define ip_set(inst, id) \ 62 ip_set_dereference((inst)->ip_set_list)[id] 63 64 /* The set types are implemented in modules and registered set types 65 * can be found in ip_set_type_list. Adding/deleting types is 66 * serialized by ip_set_type_mutex. 67 */ 68 69 static inline void 70 ip_set_type_lock(void) 71 { 72 mutex_lock(&ip_set_type_mutex); 73 } 74 75 static inline void 76 ip_set_type_unlock(void) 77 { 78 mutex_unlock(&ip_set_type_mutex); 79 } 80 81 /* Register and deregister settype */ 82 83 static struct ip_set_type * 84 find_set_type(const char *name, u8 family, u8 revision) 85 { 86 struct ip_set_type *type; 87 88 list_for_each_entry_rcu(type, &ip_set_type_list, list) 89 if (STRNCMP(type->name, name) && 90 (type->family == family || 91 type->family == NFPROTO_UNSPEC) && 92 revision >= type->revision_min && 93 revision <= type->revision_max) 94 return type; 95 return NULL; 96 } 97 98 /* Unlock, try to load a set type module and lock again */ 99 static bool 100 load_settype(const char *name) 101 { 102 nfnl_unlock(NFNL_SUBSYS_IPSET); 103 pr_debug("try to load ip_set_%s\n", name); 104 if (request_module("ip_set_%s", name) < 0) { 105 pr_warn("Can't find ip_set type %s\n", name); 106 nfnl_lock(NFNL_SUBSYS_IPSET); 107 return false; 108 } 109 nfnl_lock(NFNL_SUBSYS_IPSET); 110 return true; 111 } 112 113 /* Find a set type and reference it */ 114 #define find_set_type_get(name, family, revision, found) \ 115 __find_set_type_get(name, family, revision, found, false) 116 117 static int 118 __find_set_type_get(const char *name, u8 family, u8 revision, 119 struct ip_set_type **found, bool retry) 120 { 121 struct ip_set_type *type; 122 int err; 123 124 if (retry && !load_settype(name)) 125 return -IPSET_ERR_FIND_TYPE; 126 127 rcu_read_lock(); 128 *found = find_set_type(name, family, revision); 129 if (*found) { 130 err = !try_module_get((*found)->me) ? -EFAULT : 0; 131 goto unlock; 132 } 133 /* Make sure the type is already loaded 134 * but we don't support the revision 135 */ 136 list_for_each_entry_rcu(type, &ip_set_type_list, list) 137 if (STRNCMP(type->name, name)) { 138 err = -IPSET_ERR_FIND_TYPE; 139 goto unlock; 140 } 141 rcu_read_unlock(); 142 143 return retry ? -IPSET_ERR_FIND_TYPE : 144 __find_set_type_get(name, family, revision, found, true); 145 146 unlock: 147 rcu_read_unlock(); 148 return err; 149 } 150 151 /* Find a given set type by name and family. 152 * If we succeeded, the supported minimal and maximum revisions are 153 * filled out. 154 */ 155 #define find_set_type_minmax(name, family, min, max) \ 156 __find_set_type_minmax(name, family, min, max, false) 157 158 static int 159 __find_set_type_minmax(const char *name, u8 family, u8 *min, u8 *max, 160 bool retry) 161 { 162 struct ip_set_type *type; 163 bool found = false; 164 165 if (retry && !load_settype(name)) 166 return -IPSET_ERR_FIND_TYPE; 167 168 *min = 255; *max = 0; 169 rcu_read_lock(); 170 list_for_each_entry_rcu(type, &ip_set_type_list, list) 171 if (STRNCMP(type->name, name) && 172 (type->family == family || 173 type->family == NFPROTO_UNSPEC)) { 174 found = true; 175 if (type->revision_min < *min) 176 *min = type->revision_min; 177 if (type->revision_max > *max) 178 *max = type->revision_max; 179 } 180 rcu_read_unlock(); 181 if (found) 182 return 0; 183 184 return retry ? -IPSET_ERR_FIND_TYPE : 185 __find_set_type_minmax(name, family, min, max, true); 186 } 187 188 #define family_name(f) ((f) == NFPROTO_IPV4 ? "inet" : \ 189 (f) == NFPROTO_IPV6 ? "inet6" : "any") 190 191 /* Register a set type structure. The type is identified by 192 * the unique triple of name, family and revision. 193 */ 194 int 195 ip_set_type_register(struct ip_set_type *type) 196 { 197 int ret = 0; 198 199 if (type->protocol != IPSET_PROTOCOL) { 200 pr_warn("ip_set type %s, family %s, revision %u:%u uses wrong protocol version %u (want %u)\n", 201 type->name, family_name(type->family), 202 type->revision_min, type->revision_max, 203 type->protocol, IPSET_PROTOCOL); 204 return -EINVAL; 205 } 206 207 ip_set_type_lock(); 208 if (find_set_type(type->name, type->family, type->revision_min)) { 209 /* Duplicate! */ 210 pr_warn("ip_set type %s, family %s with revision min %u already registered!\n", 211 type->name, family_name(type->family), 212 type->revision_min); 213 ip_set_type_unlock(); 214 return -EINVAL; 215 } 216 list_add_rcu(&type->list, &ip_set_type_list); 217 pr_debug("type %s, family %s, revision %u:%u registered.\n", 218 type->name, family_name(type->family), 219 type->revision_min, type->revision_max); 220 ip_set_type_unlock(); 221 222 return ret; 223 } 224 EXPORT_SYMBOL_GPL(ip_set_type_register); 225 226 /* Unregister a set type. There's a small race with ip_set_create */ 227 void 228 ip_set_type_unregister(struct ip_set_type *type) 229 { 230 ip_set_type_lock(); 231 if (!find_set_type(type->name, type->family, type->revision_min)) { 232 pr_warn("ip_set type %s, family %s with revision min %u not registered\n", 233 type->name, family_name(type->family), 234 type->revision_min); 235 ip_set_type_unlock(); 236 return; 237 } 238 list_del_rcu(&type->list); 239 pr_debug("type %s, family %s with revision min %u unregistered.\n", 240 type->name, family_name(type->family), type->revision_min); 241 ip_set_type_unlock(); 242 243 synchronize_rcu(); 244 } 245 EXPORT_SYMBOL_GPL(ip_set_type_unregister); 246 247 /* Utility functions */ 248 void * 249 ip_set_alloc(size_t size) 250 { 251 void *members = NULL; 252 253 if (size < KMALLOC_MAX_SIZE) 254 members = kzalloc(size, GFP_KERNEL | __GFP_NOWARN); 255 256 if (members) { 257 pr_debug("%p: allocated with kmalloc\n", members); 258 return members; 259 } 260 261 members = vzalloc(size); 262 if (!members) 263 return NULL; 264 pr_debug("%p: allocated with vmalloc\n", members); 265 266 return members; 267 } 268 EXPORT_SYMBOL_GPL(ip_set_alloc); 269 270 void 271 ip_set_free(void *members) 272 { 273 pr_debug("%p: free with %s\n", members, 274 is_vmalloc_addr(members) ? "vfree" : "kfree"); 275 kvfree(members); 276 } 277 EXPORT_SYMBOL_GPL(ip_set_free); 278 279 static inline bool 280 flag_nested(const struct nlattr *nla) 281 { 282 return nla->nla_type & NLA_F_NESTED; 283 } 284 285 static const struct nla_policy ipaddr_policy[IPSET_ATTR_IPADDR_MAX + 1] = { 286 [IPSET_ATTR_IPADDR_IPV4] = { .type = NLA_U32 }, 287 [IPSET_ATTR_IPADDR_IPV6] = { .type = NLA_BINARY, 288 .len = sizeof(struct in6_addr) }, 289 }; 290 291 int 292 ip_set_get_ipaddr4(struct nlattr *nla, __be32 *ipaddr) 293 { 294 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1]; 295 296 if (unlikely(!flag_nested(nla))) 297 return -IPSET_ERR_PROTOCOL; 298 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla, 299 ipaddr_policy, NULL)) 300 return -IPSET_ERR_PROTOCOL; 301 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV4))) 302 return -IPSET_ERR_PROTOCOL; 303 304 *ipaddr = nla_get_be32(tb[IPSET_ATTR_IPADDR_IPV4]); 305 return 0; 306 } 307 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr4); 308 309 int 310 ip_set_get_ipaddr6(struct nlattr *nla, union nf_inet_addr *ipaddr) 311 { 312 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1]; 313 314 if (unlikely(!flag_nested(nla))) 315 return -IPSET_ERR_PROTOCOL; 316 317 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla, 318 ipaddr_policy, NULL)) 319 return -IPSET_ERR_PROTOCOL; 320 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV6))) 321 return -IPSET_ERR_PROTOCOL; 322 323 memcpy(ipaddr, nla_data(tb[IPSET_ATTR_IPADDR_IPV6]), 324 sizeof(struct in6_addr)); 325 return 0; 326 } 327 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr6); 328 329 typedef void (*destroyer)(struct ip_set *, void *); 330 /* ipset data extension types, in size order */ 331 332 const struct ip_set_ext_type ip_set_extensions[] = { 333 [IPSET_EXT_ID_COUNTER] = { 334 .type = IPSET_EXT_COUNTER, 335 .flag = IPSET_FLAG_WITH_COUNTERS, 336 .len = sizeof(struct ip_set_counter), 337 .align = __alignof__(struct ip_set_counter), 338 }, 339 [IPSET_EXT_ID_TIMEOUT] = { 340 .type = IPSET_EXT_TIMEOUT, 341 .len = sizeof(unsigned long), 342 .align = __alignof__(unsigned long), 343 }, 344 [IPSET_EXT_ID_SKBINFO] = { 345 .type = IPSET_EXT_SKBINFO, 346 .flag = IPSET_FLAG_WITH_SKBINFO, 347 .len = sizeof(struct ip_set_skbinfo), 348 .align = __alignof__(struct ip_set_skbinfo), 349 }, 350 [IPSET_EXT_ID_COMMENT] = { 351 .type = IPSET_EXT_COMMENT | IPSET_EXT_DESTROY, 352 .flag = IPSET_FLAG_WITH_COMMENT, 353 .len = sizeof(struct ip_set_comment), 354 .align = __alignof__(struct ip_set_comment), 355 .destroy = (destroyer) ip_set_comment_free, 356 }, 357 }; 358 EXPORT_SYMBOL_GPL(ip_set_extensions); 359 360 static inline bool 361 add_extension(enum ip_set_ext_id id, u32 flags, struct nlattr *tb[]) 362 { 363 return ip_set_extensions[id].flag ? 364 (flags & ip_set_extensions[id].flag) : 365 !!tb[IPSET_ATTR_TIMEOUT]; 366 } 367 368 size_t 369 ip_set_elem_len(struct ip_set *set, struct nlattr *tb[], size_t len, 370 size_t align) 371 { 372 enum ip_set_ext_id id; 373 u32 cadt_flags = 0; 374 375 if (tb[IPSET_ATTR_CADT_FLAGS]) 376 cadt_flags = ip_set_get_h32(tb[IPSET_ATTR_CADT_FLAGS]); 377 if (cadt_flags & IPSET_FLAG_WITH_FORCEADD) 378 set->flags |= IPSET_CREATE_FLAG_FORCEADD; 379 if (!align) 380 align = 1; 381 for (id = 0; id < IPSET_EXT_ID_MAX; id++) { 382 if (!add_extension(id, cadt_flags, tb)) 383 continue; 384 len = ALIGN(len, ip_set_extensions[id].align); 385 set->offset[id] = len; 386 set->extensions |= ip_set_extensions[id].type; 387 len += ip_set_extensions[id].len; 388 } 389 return ALIGN(len, align); 390 } 391 EXPORT_SYMBOL_GPL(ip_set_elem_len); 392 393 int 394 ip_set_get_extensions(struct ip_set *set, struct nlattr *tb[], 395 struct ip_set_ext *ext) 396 { 397 u64 fullmark; 398 399 if (unlikely(!ip_set_optattr_netorder(tb, IPSET_ATTR_TIMEOUT) || 400 !ip_set_optattr_netorder(tb, IPSET_ATTR_PACKETS) || 401 !ip_set_optattr_netorder(tb, IPSET_ATTR_BYTES) || 402 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBMARK) || 403 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBPRIO) || 404 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBQUEUE))) 405 return -IPSET_ERR_PROTOCOL; 406 407 if (tb[IPSET_ATTR_TIMEOUT]) { 408 if (!SET_WITH_TIMEOUT(set)) 409 return -IPSET_ERR_TIMEOUT; 410 ext->timeout = ip_set_timeout_uget(tb[IPSET_ATTR_TIMEOUT]); 411 } 412 if (tb[IPSET_ATTR_BYTES] || tb[IPSET_ATTR_PACKETS]) { 413 if (!SET_WITH_COUNTER(set)) 414 return -IPSET_ERR_COUNTER; 415 if (tb[IPSET_ATTR_BYTES]) 416 ext->bytes = be64_to_cpu(nla_get_be64( 417 tb[IPSET_ATTR_BYTES])); 418 if (tb[IPSET_ATTR_PACKETS]) 419 ext->packets = be64_to_cpu(nla_get_be64( 420 tb[IPSET_ATTR_PACKETS])); 421 } 422 if (tb[IPSET_ATTR_COMMENT]) { 423 if (!SET_WITH_COMMENT(set)) 424 return -IPSET_ERR_COMMENT; 425 ext->comment = ip_set_comment_uget(tb[IPSET_ATTR_COMMENT]); 426 } 427 if (tb[IPSET_ATTR_SKBMARK]) { 428 if (!SET_WITH_SKBINFO(set)) 429 return -IPSET_ERR_SKBINFO; 430 fullmark = be64_to_cpu(nla_get_be64(tb[IPSET_ATTR_SKBMARK])); 431 ext->skbinfo.skbmark = fullmark >> 32; 432 ext->skbinfo.skbmarkmask = fullmark & 0xffffffff; 433 } 434 if (tb[IPSET_ATTR_SKBPRIO]) { 435 if (!SET_WITH_SKBINFO(set)) 436 return -IPSET_ERR_SKBINFO; 437 ext->skbinfo.skbprio = 438 be32_to_cpu(nla_get_be32(tb[IPSET_ATTR_SKBPRIO])); 439 } 440 if (tb[IPSET_ATTR_SKBQUEUE]) { 441 if (!SET_WITH_SKBINFO(set)) 442 return -IPSET_ERR_SKBINFO; 443 ext->skbinfo.skbqueue = 444 be16_to_cpu(nla_get_be16(tb[IPSET_ATTR_SKBQUEUE])); 445 } 446 return 0; 447 } 448 EXPORT_SYMBOL_GPL(ip_set_get_extensions); 449 450 int 451 ip_set_put_extensions(struct sk_buff *skb, const struct ip_set *set, 452 const void *e, bool active) 453 { 454 if (SET_WITH_TIMEOUT(set)) { 455 unsigned long *timeout = ext_timeout(e, set); 456 457 if (nla_put_net32(skb, IPSET_ATTR_TIMEOUT, 458 htonl(active ? ip_set_timeout_get(timeout) 459 : *timeout))) 460 return -EMSGSIZE; 461 } 462 if (SET_WITH_COUNTER(set) && 463 ip_set_put_counter(skb, ext_counter(e, set))) 464 return -EMSGSIZE; 465 if (SET_WITH_COMMENT(set) && 466 ip_set_put_comment(skb, ext_comment(e, set))) 467 return -EMSGSIZE; 468 if (SET_WITH_SKBINFO(set) && 469 ip_set_put_skbinfo(skb, ext_skbinfo(e, set))) 470 return -EMSGSIZE; 471 return 0; 472 } 473 EXPORT_SYMBOL_GPL(ip_set_put_extensions); 474 475 /* Creating/destroying/renaming/swapping affect the existence and 476 * the properties of a set. All of these can be executed from userspace 477 * only and serialized by the nfnl mutex indirectly from nfnetlink. 478 * 479 * Sets are identified by their index in ip_set_list and the index 480 * is used by the external references (set/SET netfilter modules). 481 * 482 * The set behind an index may change by swapping only, from userspace. 483 */ 484 485 static inline void 486 __ip_set_get(struct ip_set *set) 487 { 488 write_lock_bh(&ip_set_ref_lock); 489 set->ref++; 490 write_unlock_bh(&ip_set_ref_lock); 491 } 492 493 static inline void 494 __ip_set_put(struct ip_set *set) 495 { 496 write_lock_bh(&ip_set_ref_lock); 497 BUG_ON(set->ref == 0); 498 set->ref--; 499 write_unlock_bh(&ip_set_ref_lock); 500 } 501 502 /* set->ref can be swapped out by ip_set_swap, netlink events (like dump) need 503 * a separate reference counter 504 */ 505 static inline void 506 __ip_set_put_netlink(struct ip_set *set) 507 { 508 write_lock_bh(&ip_set_ref_lock); 509 BUG_ON(set->ref_netlink == 0); 510 set->ref_netlink--; 511 write_unlock_bh(&ip_set_ref_lock); 512 } 513 514 /* Add, del and test set entries from kernel. 515 * 516 * The set behind the index must exist and must be referenced 517 * so it can't be destroyed (or changed) under our foot. 518 */ 519 520 static inline struct ip_set * 521 ip_set_rcu_get(struct net *net, ip_set_id_t index) 522 { 523 struct ip_set *set; 524 struct ip_set_net *inst = ip_set_pernet(net); 525 526 rcu_read_lock(); 527 /* ip_set_list itself needs to be protected */ 528 set = rcu_dereference(inst->ip_set_list)[index]; 529 rcu_read_unlock(); 530 531 return set; 532 } 533 534 int 535 ip_set_test(ip_set_id_t index, const struct sk_buff *skb, 536 const struct xt_action_param *par, struct ip_set_adt_opt *opt) 537 { 538 struct ip_set *set = ip_set_rcu_get(xt_net(par), index); 539 int ret = 0; 540 541 BUG_ON(!set); 542 pr_debug("set %s, index %u\n", set->name, index); 543 544 if (opt->dim < set->type->dimension || 545 !(opt->family == set->family || set->family == NFPROTO_UNSPEC)) 546 return 0; 547 548 rcu_read_lock_bh(); 549 ret = set->variant->kadt(set, skb, par, IPSET_TEST, opt); 550 rcu_read_unlock_bh(); 551 552 if (ret == -EAGAIN) { 553 /* Type requests element to be completed */ 554 pr_debug("element must be completed, ADD is triggered\n"); 555 spin_lock_bh(&set->lock); 556 set->variant->kadt(set, skb, par, IPSET_ADD, opt); 557 spin_unlock_bh(&set->lock); 558 ret = 1; 559 } else { 560 /* --return-nomatch: invert matched element */ 561 if ((opt->cmdflags & IPSET_FLAG_RETURN_NOMATCH) && 562 (set->type->features & IPSET_TYPE_NOMATCH) && 563 (ret > 0 || ret == -ENOTEMPTY)) 564 ret = -ret; 565 } 566 567 /* Convert error codes to nomatch */ 568 return (ret < 0 ? 0 : ret); 569 } 570 EXPORT_SYMBOL_GPL(ip_set_test); 571 572 int 573 ip_set_add(ip_set_id_t index, const struct sk_buff *skb, 574 const struct xt_action_param *par, struct ip_set_adt_opt *opt) 575 { 576 struct ip_set *set = ip_set_rcu_get(xt_net(par), index); 577 int ret; 578 579 BUG_ON(!set); 580 pr_debug("set %s, index %u\n", set->name, index); 581 582 if (opt->dim < set->type->dimension || 583 !(opt->family == set->family || set->family == NFPROTO_UNSPEC)) 584 return -IPSET_ERR_TYPE_MISMATCH; 585 586 spin_lock_bh(&set->lock); 587 ret = set->variant->kadt(set, skb, par, IPSET_ADD, opt); 588 spin_unlock_bh(&set->lock); 589 590 return ret; 591 } 592 EXPORT_SYMBOL_GPL(ip_set_add); 593 594 int 595 ip_set_del(ip_set_id_t index, const struct sk_buff *skb, 596 const struct xt_action_param *par, struct ip_set_adt_opt *opt) 597 { 598 struct ip_set *set = ip_set_rcu_get(xt_net(par), index); 599 int ret = 0; 600 601 BUG_ON(!set); 602 pr_debug("set %s, index %u\n", set->name, index); 603 604 if (opt->dim < set->type->dimension || 605 !(opt->family == set->family || set->family == NFPROTO_UNSPEC)) 606 return -IPSET_ERR_TYPE_MISMATCH; 607 608 spin_lock_bh(&set->lock); 609 ret = set->variant->kadt(set, skb, par, IPSET_DEL, opt); 610 spin_unlock_bh(&set->lock); 611 612 return ret; 613 } 614 EXPORT_SYMBOL_GPL(ip_set_del); 615 616 /* Find set by name, reference it once. The reference makes sure the 617 * thing pointed to, does not go away under our feet. 618 * 619 */ 620 ip_set_id_t 621 ip_set_get_byname(struct net *net, const char *name, struct ip_set **set) 622 { 623 ip_set_id_t i, index = IPSET_INVALID_ID; 624 struct ip_set *s; 625 struct ip_set_net *inst = ip_set_pernet(net); 626 627 rcu_read_lock(); 628 for (i = 0; i < inst->ip_set_max; i++) { 629 s = rcu_dereference(inst->ip_set_list)[i]; 630 if (s && STRNCMP(s->name, name)) { 631 __ip_set_get(s); 632 index = i; 633 *set = s; 634 break; 635 } 636 } 637 rcu_read_unlock(); 638 639 return index; 640 } 641 EXPORT_SYMBOL_GPL(ip_set_get_byname); 642 643 /* If the given set pointer points to a valid set, decrement 644 * reference count by 1. The caller shall not assume the index 645 * to be valid, after calling this function. 646 * 647 */ 648 649 static inline void 650 __ip_set_put_byindex(struct ip_set_net *inst, ip_set_id_t index) 651 { 652 struct ip_set *set; 653 654 rcu_read_lock(); 655 set = rcu_dereference(inst->ip_set_list)[index]; 656 if (set) 657 __ip_set_put(set); 658 rcu_read_unlock(); 659 } 660 661 void 662 ip_set_put_byindex(struct net *net, ip_set_id_t index) 663 { 664 struct ip_set_net *inst = ip_set_pernet(net); 665 666 __ip_set_put_byindex(inst, index); 667 } 668 EXPORT_SYMBOL_GPL(ip_set_put_byindex); 669 670 /* Get the name of a set behind a set index. 671 * We assume the set is referenced, so it does exist and 672 * can't be destroyed. The set cannot be renamed due to 673 * the referencing either. 674 * 675 */ 676 const char * 677 ip_set_name_byindex(struct net *net, ip_set_id_t index) 678 { 679 const struct ip_set *set = ip_set_rcu_get(net, index); 680 681 BUG_ON(!set); 682 BUG_ON(set->ref == 0); 683 684 /* Referenced, so it's safe */ 685 return set->name; 686 } 687 EXPORT_SYMBOL_GPL(ip_set_name_byindex); 688 689 /* Routines to call by external subsystems, which do not 690 * call nfnl_lock for us. 691 */ 692 693 /* Find set by index, reference it once. The reference makes sure the 694 * thing pointed to, does not go away under our feet. 695 * 696 * The nfnl mutex is used in the function. 697 */ 698 ip_set_id_t 699 ip_set_nfnl_get_byindex(struct net *net, ip_set_id_t index) 700 { 701 struct ip_set *set; 702 struct ip_set_net *inst = ip_set_pernet(net); 703 704 if (index >= inst->ip_set_max) 705 return IPSET_INVALID_ID; 706 707 nfnl_lock(NFNL_SUBSYS_IPSET); 708 set = ip_set(inst, index); 709 if (set) 710 __ip_set_get(set); 711 else 712 index = IPSET_INVALID_ID; 713 nfnl_unlock(NFNL_SUBSYS_IPSET); 714 715 return index; 716 } 717 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex); 718 719 /* If the given set pointer points to a valid set, decrement 720 * reference count by 1. The caller shall not assume the index 721 * to be valid, after calling this function. 722 * 723 * The nfnl mutex is used in the function. 724 */ 725 void 726 ip_set_nfnl_put(struct net *net, ip_set_id_t index) 727 { 728 struct ip_set *set; 729 struct ip_set_net *inst = ip_set_pernet(net); 730 731 nfnl_lock(NFNL_SUBSYS_IPSET); 732 if (!inst->is_deleted) { /* already deleted from ip_set_net_exit() */ 733 set = ip_set(inst, index); 734 if (set) 735 __ip_set_put(set); 736 } 737 nfnl_unlock(NFNL_SUBSYS_IPSET); 738 } 739 EXPORT_SYMBOL_GPL(ip_set_nfnl_put); 740 741 /* Communication protocol with userspace over netlink. 742 * 743 * The commands are serialized by the nfnl mutex. 744 */ 745 746 static inline bool 747 protocol_failed(const struct nlattr * const tb[]) 748 { 749 return !tb[IPSET_ATTR_PROTOCOL] || 750 nla_get_u8(tb[IPSET_ATTR_PROTOCOL]) != IPSET_PROTOCOL; 751 } 752 753 static inline u32 754 flag_exist(const struct nlmsghdr *nlh) 755 { 756 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST; 757 } 758 759 static struct nlmsghdr * 760 start_msg(struct sk_buff *skb, u32 portid, u32 seq, unsigned int flags, 761 enum ipset_cmd cmd) 762 { 763 struct nlmsghdr *nlh; 764 struct nfgenmsg *nfmsg; 765 766 nlh = nlmsg_put(skb, portid, seq, nfnl_msg_type(NFNL_SUBSYS_IPSET, cmd), 767 sizeof(*nfmsg), flags); 768 if (!nlh) 769 return NULL; 770 771 nfmsg = nlmsg_data(nlh); 772 nfmsg->nfgen_family = NFPROTO_IPV4; 773 nfmsg->version = NFNETLINK_V0; 774 nfmsg->res_id = 0; 775 776 return nlh; 777 } 778 779 /* Create a set */ 780 781 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = { 782 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 783 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 784 .len = IPSET_MAXNAMELEN - 1 }, 785 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING, 786 .len = IPSET_MAXNAMELEN - 1}, 787 [IPSET_ATTR_REVISION] = { .type = NLA_U8 }, 788 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 }, 789 [IPSET_ATTR_DATA] = { .type = NLA_NESTED }, 790 }; 791 792 static struct ip_set * 793 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id) 794 { 795 struct ip_set *set = NULL; 796 ip_set_id_t i; 797 798 *id = IPSET_INVALID_ID; 799 for (i = 0; i < inst->ip_set_max; i++) { 800 set = ip_set(inst, i); 801 if (set && STRNCMP(set->name, name)) { 802 *id = i; 803 break; 804 } 805 } 806 return (*id == IPSET_INVALID_ID ? NULL : set); 807 } 808 809 static inline struct ip_set * 810 find_set(struct ip_set_net *inst, const char *name) 811 { 812 ip_set_id_t id; 813 814 return find_set_and_id(inst, name, &id); 815 } 816 817 static int 818 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index, 819 struct ip_set **set) 820 { 821 struct ip_set *s; 822 ip_set_id_t i; 823 824 *index = IPSET_INVALID_ID; 825 for (i = 0; i < inst->ip_set_max; i++) { 826 s = ip_set(inst, i); 827 if (!s) { 828 if (*index == IPSET_INVALID_ID) 829 *index = i; 830 } else if (STRNCMP(name, s->name)) { 831 /* Name clash */ 832 *set = s; 833 return -EEXIST; 834 } 835 } 836 if (*index == IPSET_INVALID_ID) 837 /* No free slot remained */ 838 return -IPSET_ERR_MAX_SETS; 839 return 0; 840 } 841 842 static int ip_set_none(struct net *net, struct sock *ctnl, struct sk_buff *skb, 843 const struct nlmsghdr *nlh, 844 const struct nlattr * const attr[]) 845 { 846 return -EOPNOTSUPP; 847 } 848 849 static int ip_set_create(struct net *net, struct sock *ctnl, 850 struct sk_buff *skb, const struct nlmsghdr *nlh, 851 const struct nlattr * const attr[]) 852 { 853 struct ip_set_net *inst = ip_set_pernet(net); 854 struct ip_set *set, *clash = NULL; 855 ip_set_id_t index = IPSET_INVALID_ID; 856 struct nlattr *tb[IPSET_ATTR_CREATE_MAX + 1] = {}; 857 const char *name, *typename; 858 u8 family, revision; 859 u32 flags = flag_exist(nlh); 860 int ret = 0; 861 862 if (unlikely(protocol_failed(attr) || 863 !attr[IPSET_ATTR_SETNAME] || 864 !attr[IPSET_ATTR_TYPENAME] || 865 !attr[IPSET_ATTR_REVISION] || 866 !attr[IPSET_ATTR_FAMILY] || 867 (attr[IPSET_ATTR_DATA] && 868 !flag_nested(attr[IPSET_ATTR_DATA])))) 869 return -IPSET_ERR_PROTOCOL; 870 871 name = nla_data(attr[IPSET_ATTR_SETNAME]); 872 typename = nla_data(attr[IPSET_ATTR_TYPENAME]); 873 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]); 874 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]); 875 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n", 876 name, typename, family_name(family), revision); 877 878 /* First, and without any locks, allocate and initialize 879 * a normal base set structure. 880 */ 881 set = kzalloc(sizeof(*set), GFP_KERNEL); 882 if (!set) 883 return -ENOMEM; 884 spin_lock_init(&set->lock); 885 strlcpy(set->name, name, IPSET_MAXNAMELEN); 886 set->family = family; 887 set->revision = revision; 888 889 /* Next, check that we know the type, and take 890 * a reference on the type, to make sure it stays available 891 * while constructing our new set. 892 * 893 * After referencing the type, we try to create the type 894 * specific part of the set without holding any locks. 895 */ 896 ret = find_set_type_get(typename, family, revision, &set->type); 897 if (ret) 898 goto out; 899 900 /* Without holding any locks, create private part. */ 901 if (attr[IPSET_ATTR_DATA] && 902 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA], 903 set->type->create_policy, NULL)) { 904 ret = -IPSET_ERR_PROTOCOL; 905 goto put_out; 906 } 907 908 ret = set->type->create(net, set, tb, flags); 909 if (ret != 0) 910 goto put_out; 911 912 /* BTW, ret==0 here. */ 913 914 /* Here, we have a valid, constructed set and we are protected 915 * by the nfnl mutex. Find the first free index in ip_set_list 916 * and check clashing. 917 */ 918 ret = find_free_id(inst, set->name, &index, &clash); 919 if (ret == -EEXIST) { 920 /* If this is the same set and requested, ignore error */ 921 if ((flags & IPSET_FLAG_EXIST) && 922 STRNCMP(set->type->name, clash->type->name) && 923 set->type->family == clash->type->family && 924 set->type->revision_min == clash->type->revision_min && 925 set->type->revision_max == clash->type->revision_max && 926 set->variant->same_set(set, clash)) 927 ret = 0; 928 goto cleanup; 929 } else if (ret == -IPSET_ERR_MAX_SETS) { 930 struct ip_set **list, **tmp; 931 ip_set_id_t i = inst->ip_set_max + IP_SET_INC; 932 933 if (i < inst->ip_set_max || i == IPSET_INVALID_ID) 934 /* Wraparound */ 935 goto cleanup; 936 937 list = kcalloc(i, sizeof(struct ip_set *), GFP_KERNEL); 938 if (!list) 939 goto cleanup; 940 /* nfnl mutex is held, both lists are valid */ 941 tmp = ip_set_dereference(inst->ip_set_list); 942 memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max); 943 rcu_assign_pointer(inst->ip_set_list, list); 944 /* Make sure all current packets have passed through */ 945 synchronize_net(); 946 /* Use new list */ 947 index = inst->ip_set_max; 948 inst->ip_set_max = i; 949 kfree(tmp); 950 ret = 0; 951 } else if (ret) { 952 goto cleanup; 953 } 954 955 /* Finally! Add our shiny new set to the list, and be done. */ 956 pr_debug("create: '%s' created with index %u!\n", set->name, index); 957 ip_set(inst, index) = set; 958 959 return ret; 960 961 cleanup: 962 set->variant->destroy(set); 963 put_out: 964 module_put(set->type->me); 965 out: 966 kfree(set); 967 return ret; 968 } 969 970 /* Destroy sets */ 971 972 static const struct nla_policy 973 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = { 974 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 975 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 976 .len = IPSET_MAXNAMELEN - 1 }, 977 }; 978 979 static void 980 ip_set_destroy_set(struct ip_set *set) 981 { 982 pr_debug("set: %s\n", set->name); 983 984 /* Must call it without holding any lock */ 985 set->variant->destroy(set); 986 module_put(set->type->me); 987 kfree(set); 988 } 989 990 static int ip_set_destroy(struct net *net, struct sock *ctnl, 991 struct sk_buff *skb, const struct nlmsghdr *nlh, 992 const struct nlattr * const attr[]) 993 { 994 struct ip_set_net *inst = ip_set_pernet(net); 995 struct ip_set *s; 996 ip_set_id_t i; 997 int ret = 0; 998 999 if (unlikely(protocol_failed(attr))) 1000 return -IPSET_ERR_PROTOCOL; 1001 1002 /* Must wait for flush to be really finished in list:set */ 1003 rcu_barrier(); 1004 1005 /* Commands are serialized and references are 1006 * protected by the ip_set_ref_lock. 1007 * External systems (i.e. xt_set) must call 1008 * ip_set_put|get_nfnl_* functions, that way we 1009 * can safely check references here. 1010 * 1011 * list:set timer can only decrement the reference 1012 * counter, so if it's already zero, we can proceed 1013 * without holding the lock. 1014 */ 1015 read_lock_bh(&ip_set_ref_lock); 1016 if (!attr[IPSET_ATTR_SETNAME]) { 1017 for (i = 0; i < inst->ip_set_max; i++) { 1018 s = ip_set(inst, i); 1019 if (s && (s->ref || s->ref_netlink)) { 1020 ret = -IPSET_ERR_BUSY; 1021 goto out; 1022 } 1023 } 1024 inst->is_destroyed = true; 1025 read_unlock_bh(&ip_set_ref_lock); 1026 for (i = 0; i < inst->ip_set_max; i++) { 1027 s = ip_set(inst, i); 1028 if (s) { 1029 ip_set(inst, i) = NULL; 1030 ip_set_destroy_set(s); 1031 } 1032 } 1033 /* Modified by ip_set_destroy() only, which is serialized */ 1034 inst->is_destroyed = false; 1035 } else { 1036 s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]), 1037 &i); 1038 if (!s) { 1039 ret = -ENOENT; 1040 goto out; 1041 } else if (s->ref || s->ref_netlink) { 1042 ret = -IPSET_ERR_BUSY; 1043 goto out; 1044 } 1045 ip_set(inst, i) = NULL; 1046 read_unlock_bh(&ip_set_ref_lock); 1047 1048 ip_set_destroy_set(s); 1049 } 1050 return 0; 1051 out: 1052 read_unlock_bh(&ip_set_ref_lock); 1053 return ret; 1054 } 1055 1056 /* Flush sets */ 1057 1058 static void 1059 ip_set_flush_set(struct ip_set *set) 1060 { 1061 pr_debug("set: %s\n", set->name); 1062 1063 spin_lock_bh(&set->lock); 1064 set->variant->flush(set); 1065 spin_unlock_bh(&set->lock); 1066 } 1067 1068 static int ip_set_flush(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1069 const struct nlmsghdr *nlh, 1070 const struct nlattr * const attr[]) 1071 { 1072 struct ip_set_net *inst = ip_set_pernet(net); 1073 struct ip_set *s; 1074 ip_set_id_t i; 1075 1076 if (unlikely(protocol_failed(attr))) 1077 return -IPSET_ERR_PROTOCOL; 1078 1079 if (!attr[IPSET_ATTR_SETNAME]) { 1080 for (i = 0; i < inst->ip_set_max; i++) { 1081 s = ip_set(inst, i); 1082 if (s) 1083 ip_set_flush_set(s); 1084 } 1085 } else { 1086 s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME])); 1087 if (!s) 1088 return -ENOENT; 1089 1090 ip_set_flush_set(s); 1091 } 1092 1093 return 0; 1094 } 1095 1096 /* Rename a set */ 1097 1098 static const struct nla_policy 1099 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = { 1100 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1101 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 1102 .len = IPSET_MAXNAMELEN - 1 }, 1103 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING, 1104 .len = IPSET_MAXNAMELEN - 1 }, 1105 }; 1106 1107 static int ip_set_rename(struct net *net, struct sock *ctnl, 1108 struct sk_buff *skb, const struct nlmsghdr *nlh, 1109 const struct nlattr * const attr[]) 1110 { 1111 struct ip_set_net *inst = ip_set_pernet(net); 1112 struct ip_set *set, *s; 1113 const char *name2; 1114 ip_set_id_t i; 1115 int ret = 0; 1116 1117 if (unlikely(protocol_failed(attr) || 1118 !attr[IPSET_ATTR_SETNAME] || 1119 !attr[IPSET_ATTR_SETNAME2])) 1120 return -IPSET_ERR_PROTOCOL; 1121 1122 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME])); 1123 if (!set) 1124 return -ENOENT; 1125 1126 read_lock_bh(&ip_set_ref_lock); 1127 if (set->ref != 0) { 1128 ret = -IPSET_ERR_REFERENCED; 1129 goto out; 1130 } 1131 1132 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]); 1133 for (i = 0; i < inst->ip_set_max; i++) { 1134 s = ip_set(inst, i); 1135 if (s && STRNCMP(s->name, name2)) { 1136 ret = -IPSET_ERR_EXIST_SETNAME2; 1137 goto out; 1138 } 1139 } 1140 strncpy(set->name, name2, IPSET_MAXNAMELEN); 1141 1142 out: 1143 read_unlock_bh(&ip_set_ref_lock); 1144 return ret; 1145 } 1146 1147 /* Swap two sets so that name/index points to the other. 1148 * References and set names are also swapped. 1149 * 1150 * The commands are serialized by the nfnl mutex and references are 1151 * protected by the ip_set_ref_lock. The kernel interfaces 1152 * do not hold the mutex but the pointer settings are atomic 1153 * so the ip_set_list always contains valid pointers to the sets. 1154 */ 1155 1156 static int ip_set_swap(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1157 const struct nlmsghdr *nlh, 1158 const struct nlattr * const attr[]) 1159 { 1160 struct ip_set_net *inst = ip_set_pernet(net); 1161 struct ip_set *from, *to; 1162 ip_set_id_t from_id, to_id; 1163 char from_name[IPSET_MAXNAMELEN]; 1164 1165 if (unlikely(protocol_failed(attr) || 1166 !attr[IPSET_ATTR_SETNAME] || 1167 !attr[IPSET_ATTR_SETNAME2])) 1168 return -IPSET_ERR_PROTOCOL; 1169 1170 from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]), 1171 &from_id); 1172 if (!from) 1173 return -ENOENT; 1174 1175 to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]), 1176 &to_id); 1177 if (!to) 1178 return -IPSET_ERR_EXIST_SETNAME2; 1179 1180 /* Features must not change. 1181 * Not an artifical restriction anymore, as we must prevent 1182 * possible loops created by swapping in setlist type of sets. 1183 */ 1184 if (!(from->type->features == to->type->features && 1185 from->family == to->family)) 1186 return -IPSET_ERR_TYPE_MISMATCH; 1187 1188 if (from->ref_netlink || to->ref_netlink) 1189 return -EBUSY; 1190 1191 strncpy(from_name, from->name, IPSET_MAXNAMELEN); 1192 strncpy(from->name, to->name, IPSET_MAXNAMELEN); 1193 strncpy(to->name, from_name, IPSET_MAXNAMELEN); 1194 1195 write_lock_bh(&ip_set_ref_lock); 1196 swap(from->ref, to->ref); 1197 ip_set(inst, from_id) = to; 1198 ip_set(inst, to_id) = from; 1199 write_unlock_bh(&ip_set_ref_lock); 1200 1201 return 0; 1202 } 1203 1204 /* List/save set data */ 1205 1206 #define DUMP_INIT 0 1207 #define DUMP_ALL 1 1208 #define DUMP_ONE 2 1209 #define DUMP_LAST 3 1210 1211 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF) 1212 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16) 1213 1214 static int 1215 ip_set_dump_done(struct netlink_callback *cb) 1216 { 1217 if (cb->args[IPSET_CB_ARG0]) { 1218 struct ip_set_net *inst = 1219 (struct ip_set_net *)cb->args[IPSET_CB_NET]; 1220 ip_set_id_t index = (ip_set_id_t)cb->args[IPSET_CB_INDEX]; 1221 struct ip_set *set = ip_set(inst, index); 1222 1223 if (set->variant->uref) 1224 set->variant->uref(set, cb, false); 1225 pr_debug("release set %s\n", set->name); 1226 __ip_set_put_netlink(set); 1227 } 1228 return 0; 1229 } 1230 1231 static inline void 1232 dump_attrs(struct nlmsghdr *nlh) 1233 { 1234 const struct nlattr *attr; 1235 int rem; 1236 1237 pr_debug("dump nlmsg\n"); 1238 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) { 1239 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len); 1240 } 1241 } 1242 1243 static int 1244 dump_init(struct netlink_callback *cb, struct ip_set_net *inst) 1245 { 1246 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb); 1247 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg)); 1248 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1]; 1249 struct nlattr *attr = (void *)nlh + min_len; 1250 u32 dump_type; 1251 ip_set_id_t index; 1252 1253 /* Second pass, so parser can't fail */ 1254 nla_parse(cda, IPSET_ATTR_CMD_MAX, attr, nlh->nlmsg_len - min_len, 1255 ip_set_setname_policy, NULL); 1256 1257 if (cda[IPSET_ATTR_SETNAME]) { 1258 struct ip_set *set; 1259 1260 set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]), 1261 &index); 1262 if (!set) 1263 return -ENOENT; 1264 1265 dump_type = DUMP_ONE; 1266 cb->args[IPSET_CB_INDEX] = index; 1267 } else { 1268 dump_type = DUMP_ALL; 1269 } 1270 1271 if (cda[IPSET_ATTR_FLAGS]) { 1272 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]); 1273 1274 dump_type |= (f << 16); 1275 } 1276 cb->args[IPSET_CB_NET] = (unsigned long)inst; 1277 cb->args[IPSET_CB_DUMP] = dump_type; 1278 1279 return 0; 1280 } 1281 1282 static int 1283 ip_set_dump_start(struct sk_buff *skb, struct netlink_callback *cb) 1284 { 1285 ip_set_id_t index = IPSET_INVALID_ID, max; 1286 struct ip_set *set = NULL; 1287 struct nlmsghdr *nlh = NULL; 1288 unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0; 1289 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk)); 1290 u32 dump_type, dump_flags; 1291 bool is_destroyed; 1292 int ret = 0; 1293 1294 if (!cb->args[IPSET_CB_DUMP]) { 1295 ret = dump_init(cb, inst); 1296 if (ret < 0) { 1297 nlh = nlmsg_hdr(cb->skb); 1298 /* We have to create and send the error message 1299 * manually :-( 1300 */ 1301 if (nlh->nlmsg_flags & NLM_F_ACK) 1302 netlink_ack(cb->skb, nlh, ret, NULL); 1303 return ret; 1304 } 1305 } 1306 1307 if (cb->args[IPSET_CB_INDEX] >= inst->ip_set_max) 1308 goto out; 1309 1310 dump_type = DUMP_TYPE(cb->args[IPSET_CB_DUMP]); 1311 dump_flags = DUMP_FLAGS(cb->args[IPSET_CB_DUMP]); 1312 max = dump_type == DUMP_ONE ? cb->args[IPSET_CB_INDEX] + 1 1313 : inst->ip_set_max; 1314 dump_last: 1315 pr_debug("dump type, flag: %u %u index: %ld\n", 1316 dump_type, dump_flags, cb->args[IPSET_CB_INDEX]); 1317 for (; cb->args[IPSET_CB_INDEX] < max; cb->args[IPSET_CB_INDEX]++) { 1318 index = (ip_set_id_t)cb->args[IPSET_CB_INDEX]; 1319 write_lock_bh(&ip_set_ref_lock); 1320 set = ip_set(inst, index); 1321 is_destroyed = inst->is_destroyed; 1322 if (!set || is_destroyed) { 1323 write_unlock_bh(&ip_set_ref_lock); 1324 if (dump_type == DUMP_ONE) { 1325 ret = -ENOENT; 1326 goto out; 1327 } 1328 if (is_destroyed) { 1329 /* All sets are just being destroyed */ 1330 ret = 0; 1331 goto out; 1332 } 1333 continue; 1334 } 1335 /* When dumping all sets, we must dump "sorted" 1336 * so that lists (unions of sets) are dumped last. 1337 */ 1338 if (dump_type != DUMP_ONE && 1339 ((dump_type == DUMP_ALL) == 1340 !!(set->type->features & IPSET_DUMP_LAST))) { 1341 write_unlock_bh(&ip_set_ref_lock); 1342 continue; 1343 } 1344 pr_debug("List set: %s\n", set->name); 1345 if (!cb->args[IPSET_CB_ARG0]) { 1346 /* Start listing: make sure set won't be destroyed */ 1347 pr_debug("reference set\n"); 1348 set->ref_netlink++; 1349 } 1350 write_unlock_bh(&ip_set_ref_lock); 1351 nlh = start_msg(skb, NETLINK_CB(cb->skb).portid, 1352 cb->nlh->nlmsg_seq, flags, 1353 IPSET_CMD_LIST); 1354 if (!nlh) { 1355 ret = -EMSGSIZE; 1356 goto release_refcount; 1357 } 1358 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) || 1359 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name)) 1360 goto nla_put_failure; 1361 if (dump_flags & IPSET_FLAG_LIST_SETNAME) 1362 goto next_set; 1363 switch (cb->args[IPSET_CB_ARG0]) { 1364 case 0: 1365 /* Core header data */ 1366 if (nla_put_string(skb, IPSET_ATTR_TYPENAME, 1367 set->type->name) || 1368 nla_put_u8(skb, IPSET_ATTR_FAMILY, 1369 set->family) || 1370 nla_put_u8(skb, IPSET_ATTR_REVISION, 1371 set->revision)) 1372 goto nla_put_failure; 1373 ret = set->variant->head(set, skb); 1374 if (ret < 0) 1375 goto release_refcount; 1376 if (dump_flags & IPSET_FLAG_LIST_HEADER) 1377 goto next_set; 1378 if (set->variant->uref) 1379 set->variant->uref(set, cb, true); 1380 /* Fall through and add elements */ 1381 default: 1382 rcu_read_lock_bh(); 1383 ret = set->variant->list(set, skb, cb); 1384 rcu_read_unlock_bh(); 1385 if (!cb->args[IPSET_CB_ARG0]) 1386 /* Set is done, proceed with next one */ 1387 goto next_set; 1388 goto release_refcount; 1389 } 1390 } 1391 /* If we dump all sets, continue with dumping last ones */ 1392 if (dump_type == DUMP_ALL) { 1393 dump_type = DUMP_LAST; 1394 cb->args[IPSET_CB_DUMP] = dump_type | (dump_flags << 16); 1395 cb->args[IPSET_CB_INDEX] = 0; 1396 if (set && set->variant->uref) 1397 set->variant->uref(set, cb, false); 1398 goto dump_last; 1399 } 1400 goto out; 1401 1402 nla_put_failure: 1403 ret = -EFAULT; 1404 next_set: 1405 if (dump_type == DUMP_ONE) 1406 cb->args[IPSET_CB_INDEX] = IPSET_INVALID_ID; 1407 else 1408 cb->args[IPSET_CB_INDEX]++; 1409 release_refcount: 1410 /* If there was an error or set is done, release set */ 1411 if (ret || !cb->args[IPSET_CB_ARG0]) { 1412 set = ip_set(inst, index); 1413 if (set->variant->uref) 1414 set->variant->uref(set, cb, false); 1415 pr_debug("release set %s\n", set->name); 1416 __ip_set_put_netlink(set); 1417 cb->args[IPSET_CB_ARG0] = 0; 1418 } 1419 out: 1420 if (nlh) { 1421 nlmsg_end(skb, nlh); 1422 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len); 1423 dump_attrs(nlh); 1424 } 1425 1426 return ret < 0 ? ret : skb->len; 1427 } 1428 1429 static int ip_set_dump(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1430 const struct nlmsghdr *nlh, 1431 const struct nlattr * const attr[]) 1432 { 1433 if (unlikely(protocol_failed(attr))) 1434 return -IPSET_ERR_PROTOCOL; 1435 1436 { 1437 struct netlink_dump_control c = { 1438 .dump = ip_set_dump_start, 1439 .done = ip_set_dump_done, 1440 }; 1441 return netlink_dump_start(ctnl, skb, nlh, &c); 1442 } 1443 } 1444 1445 /* Add, del and test */ 1446 1447 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = { 1448 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1449 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 1450 .len = IPSET_MAXNAMELEN - 1 }, 1451 [IPSET_ATTR_LINENO] = { .type = NLA_U32 }, 1452 [IPSET_ATTR_DATA] = { .type = NLA_NESTED }, 1453 [IPSET_ATTR_ADT] = { .type = NLA_NESTED }, 1454 }; 1455 1456 static int 1457 call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set, 1458 struct nlattr *tb[], enum ipset_adt adt, 1459 u32 flags, bool use_lineno) 1460 { 1461 int ret; 1462 u32 lineno = 0; 1463 bool eexist = flags & IPSET_FLAG_EXIST, retried = false; 1464 1465 do { 1466 spin_lock_bh(&set->lock); 1467 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried); 1468 spin_unlock_bh(&set->lock); 1469 retried = true; 1470 } while (ret == -EAGAIN && 1471 set->variant->resize && 1472 (ret = set->variant->resize(set, retried)) == 0); 1473 1474 if (!ret || (ret == -IPSET_ERR_EXIST && eexist)) 1475 return 0; 1476 if (lineno && use_lineno) { 1477 /* Error in restore/batch mode: send back lineno */ 1478 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb); 1479 struct sk_buff *skb2; 1480 struct nlmsgerr *errmsg; 1481 size_t payload = min(SIZE_MAX, 1482 sizeof(*errmsg) + nlmsg_len(nlh)); 1483 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg)); 1484 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1]; 1485 struct nlattr *cmdattr; 1486 u32 *errline; 1487 1488 skb2 = nlmsg_new(payload, GFP_KERNEL); 1489 if (!skb2) 1490 return -ENOMEM; 1491 rep = __nlmsg_put(skb2, NETLINK_CB(skb).portid, 1492 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0); 1493 errmsg = nlmsg_data(rep); 1494 errmsg->error = ret; 1495 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len); 1496 cmdattr = (void *)&errmsg->msg + min_len; 1497 1498 nla_parse(cda, IPSET_ATTR_CMD_MAX, cmdattr, 1499 nlh->nlmsg_len - min_len, ip_set_adt_policy, NULL); 1500 1501 errline = nla_data(cda[IPSET_ATTR_LINENO]); 1502 1503 *errline = lineno; 1504 1505 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, 1506 MSG_DONTWAIT); 1507 /* Signal netlink not to send its ACK/errmsg. */ 1508 return -EINTR; 1509 } 1510 1511 return ret; 1512 } 1513 1514 static int ip_set_uadd(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1515 const struct nlmsghdr *nlh, 1516 const struct nlattr * const attr[]) 1517 { 1518 struct ip_set_net *inst = ip_set_pernet(net); 1519 struct ip_set *set; 1520 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {}; 1521 const struct nlattr *nla; 1522 u32 flags = flag_exist(nlh); 1523 bool use_lineno; 1524 int ret = 0; 1525 1526 if (unlikely(protocol_failed(attr) || 1527 !attr[IPSET_ATTR_SETNAME] || 1528 !((attr[IPSET_ATTR_DATA] != NULL) ^ 1529 (attr[IPSET_ATTR_ADT] != NULL)) || 1530 (attr[IPSET_ATTR_DATA] && 1531 !flag_nested(attr[IPSET_ATTR_DATA])) || 1532 (attr[IPSET_ATTR_ADT] && 1533 (!flag_nested(attr[IPSET_ATTR_ADT]) || 1534 !attr[IPSET_ATTR_LINENO])))) 1535 return -IPSET_ERR_PROTOCOL; 1536 1537 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME])); 1538 if (!set) 1539 return -ENOENT; 1540 1541 use_lineno = !!attr[IPSET_ATTR_LINENO]; 1542 if (attr[IPSET_ATTR_DATA]) { 1543 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, 1544 attr[IPSET_ATTR_DATA], 1545 set->type->adt_policy, NULL)) 1546 return -IPSET_ERR_PROTOCOL; 1547 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD, flags, 1548 use_lineno); 1549 } else { 1550 int nla_rem; 1551 1552 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) { 1553 memset(tb, 0, sizeof(tb)); 1554 if (nla_type(nla) != IPSET_ATTR_DATA || 1555 !flag_nested(nla) || 1556 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla, 1557 set->type->adt_policy, NULL)) 1558 return -IPSET_ERR_PROTOCOL; 1559 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD, 1560 flags, use_lineno); 1561 if (ret < 0) 1562 return ret; 1563 } 1564 } 1565 return ret; 1566 } 1567 1568 static int ip_set_udel(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1569 const struct nlmsghdr *nlh, 1570 const struct nlattr * const attr[]) 1571 { 1572 struct ip_set_net *inst = ip_set_pernet(net); 1573 struct ip_set *set; 1574 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {}; 1575 const struct nlattr *nla; 1576 u32 flags = flag_exist(nlh); 1577 bool use_lineno; 1578 int ret = 0; 1579 1580 if (unlikely(protocol_failed(attr) || 1581 !attr[IPSET_ATTR_SETNAME] || 1582 !((attr[IPSET_ATTR_DATA] != NULL) ^ 1583 (attr[IPSET_ATTR_ADT] != NULL)) || 1584 (attr[IPSET_ATTR_DATA] && 1585 !flag_nested(attr[IPSET_ATTR_DATA])) || 1586 (attr[IPSET_ATTR_ADT] && 1587 (!flag_nested(attr[IPSET_ATTR_ADT]) || 1588 !attr[IPSET_ATTR_LINENO])))) 1589 return -IPSET_ERR_PROTOCOL; 1590 1591 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME])); 1592 if (!set) 1593 return -ENOENT; 1594 1595 use_lineno = !!attr[IPSET_ATTR_LINENO]; 1596 if (attr[IPSET_ATTR_DATA]) { 1597 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, 1598 attr[IPSET_ATTR_DATA], 1599 set->type->adt_policy, NULL)) 1600 return -IPSET_ERR_PROTOCOL; 1601 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL, flags, 1602 use_lineno); 1603 } else { 1604 int nla_rem; 1605 1606 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) { 1607 memset(tb, 0, sizeof(*tb)); 1608 if (nla_type(nla) != IPSET_ATTR_DATA || 1609 !flag_nested(nla) || 1610 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla, 1611 set->type->adt_policy, NULL)) 1612 return -IPSET_ERR_PROTOCOL; 1613 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL, 1614 flags, use_lineno); 1615 if (ret < 0) 1616 return ret; 1617 } 1618 } 1619 return ret; 1620 } 1621 1622 static int ip_set_utest(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1623 const struct nlmsghdr *nlh, 1624 const struct nlattr * const attr[]) 1625 { 1626 struct ip_set_net *inst = ip_set_pernet(net); 1627 struct ip_set *set; 1628 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {}; 1629 int ret = 0; 1630 1631 if (unlikely(protocol_failed(attr) || 1632 !attr[IPSET_ATTR_SETNAME] || 1633 !attr[IPSET_ATTR_DATA] || 1634 !flag_nested(attr[IPSET_ATTR_DATA]))) 1635 return -IPSET_ERR_PROTOCOL; 1636 1637 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME])); 1638 if (!set) 1639 return -ENOENT; 1640 1641 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA], 1642 set->type->adt_policy, NULL)) 1643 return -IPSET_ERR_PROTOCOL; 1644 1645 rcu_read_lock_bh(); 1646 ret = set->variant->uadt(set, tb, IPSET_TEST, NULL, 0, 0); 1647 rcu_read_unlock_bh(); 1648 /* Userspace can't trigger element to be re-added */ 1649 if (ret == -EAGAIN) 1650 ret = 1; 1651 1652 return ret > 0 ? 0 : -IPSET_ERR_EXIST; 1653 } 1654 1655 /* Get headed data of a set */ 1656 1657 static int ip_set_header(struct net *net, struct sock *ctnl, 1658 struct sk_buff *skb, const struct nlmsghdr *nlh, 1659 const struct nlattr * const attr[]) 1660 { 1661 struct ip_set_net *inst = ip_set_pernet(net); 1662 const struct ip_set *set; 1663 struct sk_buff *skb2; 1664 struct nlmsghdr *nlh2; 1665 int ret = 0; 1666 1667 if (unlikely(protocol_failed(attr) || 1668 !attr[IPSET_ATTR_SETNAME])) 1669 return -IPSET_ERR_PROTOCOL; 1670 1671 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME])); 1672 if (!set) 1673 return -ENOENT; 1674 1675 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 1676 if (!skb2) 1677 return -ENOMEM; 1678 1679 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 1680 IPSET_CMD_HEADER); 1681 if (!nlh2) 1682 goto nlmsg_failure; 1683 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) || 1684 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) || 1685 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) || 1686 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) || 1687 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision)) 1688 goto nla_put_failure; 1689 nlmsg_end(skb2, nlh2); 1690 1691 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT); 1692 if (ret < 0) 1693 return ret; 1694 1695 return 0; 1696 1697 nla_put_failure: 1698 nlmsg_cancel(skb2, nlh2); 1699 nlmsg_failure: 1700 kfree_skb(skb2); 1701 return -EMSGSIZE; 1702 } 1703 1704 /* Get type data */ 1705 1706 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = { 1707 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1708 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING, 1709 .len = IPSET_MAXNAMELEN - 1 }, 1710 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 }, 1711 }; 1712 1713 static int ip_set_type(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1714 const struct nlmsghdr *nlh, 1715 const struct nlattr * const attr[]) 1716 { 1717 struct sk_buff *skb2; 1718 struct nlmsghdr *nlh2; 1719 u8 family, min, max; 1720 const char *typename; 1721 int ret = 0; 1722 1723 if (unlikely(protocol_failed(attr) || 1724 !attr[IPSET_ATTR_TYPENAME] || 1725 !attr[IPSET_ATTR_FAMILY])) 1726 return -IPSET_ERR_PROTOCOL; 1727 1728 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]); 1729 typename = nla_data(attr[IPSET_ATTR_TYPENAME]); 1730 ret = find_set_type_minmax(typename, family, &min, &max); 1731 if (ret) 1732 return ret; 1733 1734 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 1735 if (!skb2) 1736 return -ENOMEM; 1737 1738 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 1739 IPSET_CMD_TYPE); 1740 if (!nlh2) 1741 goto nlmsg_failure; 1742 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) || 1743 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) || 1744 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) || 1745 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) || 1746 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min)) 1747 goto nla_put_failure; 1748 nlmsg_end(skb2, nlh2); 1749 1750 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len); 1751 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT); 1752 if (ret < 0) 1753 return ret; 1754 1755 return 0; 1756 1757 nla_put_failure: 1758 nlmsg_cancel(skb2, nlh2); 1759 nlmsg_failure: 1760 kfree_skb(skb2); 1761 return -EMSGSIZE; 1762 } 1763 1764 /* Get protocol version */ 1765 1766 static const struct nla_policy 1767 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = { 1768 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1769 }; 1770 1771 static int ip_set_protocol(struct net *net, struct sock *ctnl, 1772 struct sk_buff *skb, const struct nlmsghdr *nlh, 1773 const struct nlattr * const attr[]) 1774 { 1775 struct sk_buff *skb2; 1776 struct nlmsghdr *nlh2; 1777 int ret = 0; 1778 1779 if (unlikely(!attr[IPSET_ATTR_PROTOCOL])) 1780 return -IPSET_ERR_PROTOCOL; 1781 1782 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 1783 if (!skb2) 1784 return -ENOMEM; 1785 1786 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 1787 IPSET_CMD_PROTOCOL); 1788 if (!nlh2) 1789 goto nlmsg_failure; 1790 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL)) 1791 goto nla_put_failure; 1792 nlmsg_end(skb2, nlh2); 1793 1794 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT); 1795 if (ret < 0) 1796 return ret; 1797 1798 return 0; 1799 1800 nla_put_failure: 1801 nlmsg_cancel(skb2, nlh2); 1802 nlmsg_failure: 1803 kfree_skb(skb2); 1804 return -EMSGSIZE; 1805 } 1806 1807 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = { 1808 [IPSET_CMD_NONE] = { 1809 .call = ip_set_none, 1810 .attr_count = IPSET_ATTR_CMD_MAX, 1811 }, 1812 [IPSET_CMD_CREATE] = { 1813 .call = ip_set_create, 1814 .attr_count = IPSET_ATTR_CMD_MAX, 1815 .policy = ip_set_create_policy, 1816 }, 1817 [IPSET_CMD_DESTROY] = { 1818 .call = ip_set_destroy, 1819 .attr_count = IPSET_ATTR_CMD_MAX, 1820 .policy = ip_set_setname_policy, 1821 }, 1822 [IPSET_CMD_FLUSH] = { 1823 .call = ip_set_flush, 1824 .attr_count = IPSET_ATTR_CMD_MAX, 1825 .policy = ip_set_setname_policy, 1826 }, 1827 [IPSET_CMD_RENAME] = { 1828 .call = ip_set_rename, 1829 .attr_count = IPSET_ATTR_CMD_MAX, 1830 .policy = ip_set_setname2_policy, 1831 }, 1832 [IPSET_CMD_SWAP] = { 1833 .call = ip_set_swap, 1834 .attr_count = IPSET_ATTR_CMD_MAX, 1835 .policy = ip_set_setname2_policy, 1836 }, 1837 [IPSET_CMD_LIST] = { 1838 .call = ip_set_dump, 1839 .attr_count = IPSET_ATTR_CMD_MAX, 1840 .policy = ip_set_setname_policy, 1841 }, 1842 [IPSET_CMD_SAVE] = { 1843 .call = ip_set_dump, 1844 .attr_count = IPSET_ATTR_CMD_MAX, 1845 .policy = ip_set_setname_policy, 1846 }, 1847 [IPSET_CMD_ADD] = { 1848 .call = ip_set_uadd, 1849 .attr_count = IPSET_ATTR_CMD_MAX, 1850 .policy = ip_set_adt_policy, 1851 }, 1852 [IPSET_CMD_DEL] = { 1853 .call = ip_set_udel, 1854 .attr_count = IPSET_ATTR_CMD_MAX, 1855 .policy = ip_set_adt_policy, 1856 }, 1857 [IPSET_CMD_TEST] = { 1858 .call = ip_set_utest, 1859 .attr_count = IPSET_ATTR_CMD_MAX, 1860 .policy = ip_set_adt_policy, 1861 }, 1862 [IPSET_CMD_HEADER] = { 1863 .call = ip_set_header, 1864 .attr_count = IPSET_ATTR_CMD_MAX, 1865 .policy = ip_set_setname_policy, 1866 }, 1867 [IPSET_CMD_TYPE] = { 1868 .call = ip_set_type, 1869 .attr_count = IPSET_ATTR_CMD_MAX, 1870 .policy = ip_set_type_policy, 1871 }, 1872 [IPSET_CMD_PROTOCOL] = { 1873 .call = ip_set_protocol, 1874 .attr_count = IPSET_ATTR_CMD_MAX, 1875 .policy = ip_set_protocol_policy, 1876 }, 1877 }; 1878 1879 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = { 1880 .name = "ip_set", 1881 .subsys_id = NFNL_SUBSYS_IPSET, 1882 .cb_count = IPSET_MSG_MAX, 1883 .cb = ip_set_netlink_subsys_cb, 1884 }; 1885 1886 /* Interface to iptables/ip6tables */ 1887 1888 static int 1889 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len) 1890 { 1891 unsigned int *op; 1892 void *data; 1893 int copylen = *len, ret = 0; 1894 struct net *net = sock_net(sk); 1895 struct ip_set_net *inst = ip_set_pernet(net); 1896 1897 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1898 return -EPERM; 1899 if (optval != SO_IP_SET) 1900 return -EBADF; 1901 if (*len < sizeof(unsigned int)) 1902 return -EINVAL; 1903 1904 data = vmalloc(*len); 1905 if (!data) 1906 return -ENOMEM; 1907 if (copy_from_user(data, user, *len) != 0) { 1908 ret = -EFAULT; 1909 goto done; 1910 } 1911 op = data; 1912 1913 if (*op < IP_SET_OP_VERSION) { 1914 /* Check the version at the beginning of operations */ 1915 struct ip_set_req_version *req_version = data; 1916 1917 if (*len < sizeof(struct ip_set_req_version)) { 1918 ret = -EINVAL; 1919 goto done; 1920 } 1921 1922 if (req_version->version != IPSET_PROTOCOL) { 1923 ret = -EPROTO; 1924 goto done; 1925 } 1926 } 1927 1928 switch (*op) { 1929 case IP_SET_OP_VERSION: { 1930 struct ip_set_req_version *req_version = data; 1931 1932 if (*len != sizeof(struct ip_set_req_version)) { 1933 ret = -EINVAL; 1934 goto done; 1935 } 1936 1937 req_version->version = IPSET_PROTOCOL; 1938 ret = copy_to_user(user, req_version, 1939 sizeof(struct ip_set_req_version)); 1940 goto done; 1941 } 1942 case IP_SET_OP_GET_BYNAME: { 1943 struct ip_set_req_get_set *req_get = data; 1944 ip_set_id_t id; 1945 1946 if (*len != sizeof(struct ip_set_req_get_set)) { 1947 ret = -EINVAL; 1948 goto done; 1949 } 1950 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0'; 1951 nfnl_lock(NFNL_SUBSYS_IPSET); 1952 find_set_and_id(inst, req_get->set.name, &id); 1953 req_get->set.index = id; 1954 nfnl_unlock(NFNL_SUBSYS_IPSET); 1955 goto copy; 1956 } 1957 case IP_SET_OP_GET_FNAME: { 1958 struct ip_set_req_get_set_family *req_get = data; 1959 ip_set_id_t id; 1960 1961 if (*len != sizeof(struct ip_set_req_get_set_family)) { 1962 ret = -EINVAL; 1963 goto done; 1964 } 1965 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0'; 1966 nfnl_lock(NFNL_SUBSYS_IPSET); 1967 find_set_and_id(inst, req_get->set.name, &id); 1968 req_get->set.index = id; 1969 if (id != IPSET_INVALID_ID) 1970 req_get->family = ip_set(inst, id)->family; 1971 nfnl_unlock(NFNL_SUBSYS_IPSET); 1972 goto copy; 1973 } 1974 case IP_SET_OP_GET_BYINDEX: { 1975 struct ip_set_req_get_set *req_get = data; 1976 struct ip_set *set; 1977 1978 if (*len != sizeof(struct ip_set_req_get_set) || 1979 req_get->set.index >= inst->ip_set_max) { 1980 ret = -EINVAL; 1981 goto done; 1982 } 1983 nfnl_lock(NFNL_SUBSYS_IPSET); 1984 set = ip_set(inst, req_get->set.index); 1985 strncpy(req_get->set.name, set ? set->name : "", 1986 IPSET_MAXNAMELEN); 1987 nfnl_unlock(NFNL_SUBSYS_IPSET); 1988 goto copy; 1989 } 1990 default: 1991 ret = -EBADMSG; 1992 goto done; 1993 } /* end of switch(op) */ 1994 1995 copy: 1996 ret = copy_to_user(user, data, copylen); 1997 1998 done: 1999 vfree(data); 2000 if (ret > 0) 2001 ret = 0; 2002 return ret; 2003 } 2004 2005 static struct nf_sockopt_ops so_set __read_mostly = { 2006 .pf = PF_INET, 2007 .get_optmin = SO_IP_SET, 2008 .get_optmax = SO_IP_SET + 1, 2009 .get = ip_set_sockfn_get, 2010 .owner = THIS_MODULE, 2011 }; 2012 2013 static int __net_init 2014 ip_set_net_init(struct net *net) 2015 { 2016 struct ip_set_net *inst = ip_set_pernet(net); 2017 struct ip_set **list; 2018 2019 inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX; 2020 if (inst->ip_set_max >= IPSET_INVALID_ID) 2021 inst->ip_set_max = IPSET_INVALID_ID - 1; 2022 2023 list = kcalloc(inst->ip_set_max, sizeof(struct ip_set *), GFP_KERNEL); 2024 if (!list) 2025 return -ENOMEM; 2026 inst->is_deleted = false; 2027 inst->is_destroyed = false; 2028 rcu_assign_pointer(inst->ip_set_list, list); 2029 return 0; 2030 } 2031 2032 static void __net_exit 2033 ip_set_net_exit(struct net *net) 2034 { 2035 struct ip_set_net *inst = ip_set_pernet(net); 2036 2037 struct ip_set *set = NULL; 2038 ip_set_id_t i; 2039 2040 inst->is_deleted = true; /* flag for ip_set_nfnl_put */ 2041 2042 for (i = 0; i < inst->ip_set_max; i++) { 2043 set = ip_set(inst, i); 2044 if (set) { 2045 ip_set(inst, i) = NULL; 2046 ip_set_destroy_set(set); 2047 } 2048 } 2049 kfree(rcu_dereference_protected(inst->ip_set_list, 1)); 2050 } 2051 2052 static struct pernet_operations ip_set_net_ops = { 2053 .init = ip_set_net_init, 2054 .exit = ip_set_net_exit, 2055 .id = &ip_set_net_id, 2056 .size = sizeof(struct ip_set_net) 2057 }; 2058 2059 static int __init 2060 ip_set_init(void) 2061 { 2062 int ret = nfnetlink_subsys_register(&ip_set_netlink_subsys); 2063 2064 if (ret != 0) { 2065 pr_err("ip_set: cannot register with nfnetlink.\n"); 2066 return ret; 2067 } 2068 ret = nf_register_sockopt(&so_set); 2069 if (ret != 0) { 2070 pr_err("SO_SET registry failed: %d\n", ret); 2071 nfnetlink_subsys_unregister(&ip_set_netlink_subsys); 2072 return ret; 2073 } 2074 ret = register_pernet_subsys(&ip_set_net_ops); 2075 if (ret) { 2076 pr_err("ip_set: cannot register pernet_subsys.\n"); 2077 nf_unregister_sockopt(&so_set); 2078 nfnetlink_subsys_unregister(&ip_set_netlink_subsys); 2079 return ret; 2080 } 2081 pr_info("ip_set: protocol %u\n", IPSET_PROTOCOL); 2082 return 0; 2083 } 2084 2085 static void __exit 2086 ip_set_fini(void) 2087 { 2088 unregister_pernet_subsys(&ip_set_net_ops); 2089 nf_unregister_sockopt(&so_set); 2090 nfnetlink_subsys_unregister(&ip_set_netlink_subsys); 2091 pr_debug("these are the famous last words\n"); 2092 } 2093 2094 module_init(ip_set_init); 2095 module_exit(ip_set_fini); 2096