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 or ip_set_ref_lock is held: */ 59 #define ip_set_dereference(p) \ 60 rcu_dereference_protected(p, \ 61 lockdep_nfnl_is_held(NFNL_SUBSYS_IPSET) || \ 62 lockdep_is_held(&ip_set_ref_lock)) 63 #define ip_set(inst, id) \ 64 ip_set_dereference((inst)->ip_set_list)[id] 65 #define ip_set_ref_netlink(inst,id) \ 66 rcu_dereference_raw((inst)->ip_set_list)[id] 67 68 /* The set types are implemented in modules and registered set types 69 * can be found in ip_set_type_list. Adding/deleting types is 70 * serialized by ip_set_type_mutex. 71 */ 72 73 static inline void 74 ip_set_type_lock(void) 75 { 76 mutex_lock(&ip_set_type_mutex); 77 } 78 79 static inline void 80 ip_set_type_unlock(void) 81 { 82 mutex_unlock(&ip_set_type_mutex); 83 } 84 85 /* Register and deregister settype */ 86 87 static struct ip_set_type * 88 find_set_type(const char *name, u8 family, u8 revision) 89 { 90 struct ip_set_type *type; 91 92 list_for_each_entry_rcu(type, &ip_set_type_list, list) 93 if (STRNCMP(type->name, name) && 94 (type->family == family || 95 type->family == NFPROTO_UNSPEC) && 96 revision >= type->revision_min && 97 revision <= type->revision_max) 98 return type; 99 return NULL; 100 } 101 102 /* Unlock, try to load a set type module and lock again */ 103 static bool 104 load_settype(const char *name) 105 { 106 nfnl_unlock(NFNL_SUBSYS_IPSET); 107 pr_debug("try to load ip_set_%s\n", name); 108 if (request_module("ip_set_%s", name) < 0) { 109 pr_warn("Can't find ip_set type %s\n", name); 110 nfnl_lock(NFNL_SUBSYS_IPSET); 111 return false; 112 } 113 nfnl_lock(NFNL_SUBSYS_IPSET); 114 return true; 115 } 116 117 /* Find a set type and reference it */ 118 #define find_set_type_get(name, family, revision, found) \ 119 __find_set_type_get(name, family, revision, found, false) 120 121 static int 122 __find_set_type_get(const char *name, u8 family, u8 revision, 123 struct ip_set_type **found, bool retry) 124 { 125 struct ip_set_type *type; 126 int err; 127 128 if (retry && !load_settype(name)) 129 return -IPSET_ERR_FIND_TYPE; 130 131 rcu_read_lock(); 132 *found = find_set_type(name, family, revision); 133 if (*found) { 134 err = !try_module_get((*found)->me) ? -EFAULT : 0; 135 goto unlock; 136 } 137 /* Make sure the type is already loaded 138 * but we don't support the revision 139 */ 140 list_for_each_entry_rcu(type, &ip_set_type_list, list) 141 if (STRNCMP(type->name, name)) { 142 err = -IPSET_ERR_FIND_TYPE; 143 goto unlock; 144 } 145 rcu_read_unlock(); 146 147 return retry ? -IPSET_ERR_FIND_TYPE : 148 __find_set_type_get(name, family, revision, found, true); 149 150 unlock: 151 rcu_read_unlock(); 152 return err; 153 } 154 155 /* Find a given set type by name and family. 156 * If we succeeded, the supported minimal and maximum revisions are 157 * filled out. 158 */ 159 #define find_set_type_minmax(name, family, min, max) \ 160 __find_set_type_minmax(name, family, min, max, false) 161 162 static int 163 __find_set_type_minmax(const char *name, u8 family, u8 *min, u8 *max, 164 bool retry) 165 { 166 struct ip_set_type *type; 167 bool found = false; 168 169 if (retry && !load_settype(name)) 170 return -IPSET_ERR_FIND_TYPE; 171 172 *min = 255; *max = 0; 173 rcu_read_lock(); 174 list_for_each_entry_rcu(type, &ip_set_type_list, list) 175 if (STRNCMP(type->name, name) && 176 (type->family == family || 177 type->family == NFPROTO_UNSPEC)) { 178 found = true; 179 if (type->revision_min < *min) 180 *min = type->revision_min; 181 if (type->revision_max > *max) 182 *max = type->revision_max; 183 } 184 rcu_read_unlock(); 185 if (found) 186 return 0; 187 188 return retry ? -IPSET_ERR_FIND_TYPE : 189 __find_set_type_minmax(name, family, min, max, true); 190 } 191 192 #define family_name(f) ((f) == NFPROTO_IPV4 ? "inet" : \ 193 (f) == NFPROTO_IPV6 ? "inet6" : "any") 194 195 /* Register a set type structure. The type is identified by 196 * the unique triple of name, family and revision. 197 */ 198 int 199 ip_set_type_register(struct ip_set_type *type) 200 { 201 int ret = 0; 202 203 if (type->protocol != IPSET_PROTOCOL) { 204 pr_warn("ip_set type %s, family %s, revision %u:%u uses wrong protocol version %u (want %u)\n", 205 type->name, family_name(type->family), 206 type->revision_min, type->revision_max, 207 type->protocol, IPSET_PROTOCOL); 208 return -EINVAL; 209 } 210 211 ip_set_type_lock(); 212 if (find_set_type(type->name, type->family, type->revision_min)) { 213 /* Duplicate! */ 214 pr_warn("ip_set type %s, family %s with revision min %u already registered!\n", 215 type->name, family_name(type->family), 216 type->revision_min); 217 ip_set_type_unlock(); 218 return -EINVAL; 219 } 220 list_add_rcu(&type->list, &ip_set_type_list); 221 pr_debug("type %s, family %s, revision %u:%u registered.\n", 222 type->name, family_name(type->family), 223 type->revision_min, type->revision_max); 224 ip_set_type_unlock(); 225 226 return ret; 227 } 228 EXPORT_SYMBOL_GPL(ip_set_type_register); 229 230 /* Unregister a set type. There's a small race with ip_set_create */ 231 void 232 ip_set_type_unregister(struct ip_set_type *type) 233 { 234 ip_set_type_lock(); 235 if (!find_set_type(type->name, type->family, type->revision_min)) { 236 pr_warn("ip_set type %s, family %s with revision min %u not registered\n", 237 type->name, family_name(type->family), 238 type->revision_min); 239 ip_set_type_unlock(); 240 return; 241 } 242 list_del_rcu(&type->list); 243 pr_debug("type %s, family %s with revision min %u unregistered.\n", 244 type->name, family_name(type->family), type->revision_min); 245 ip_set_type_unlock(); 246 247 synchronize_rcu(); 248 } 249 EXPORT_SYMBOL_GPL(ip_set_type_unregister); 250 251 /* Utility functions */ 252 void * 253 ip_set_alloc(size_t size) 254 { 255 void *members = NULL; 256 257 if (size < KMALLOC_MAX_SIZE) 258 members = kzalloc(size, GFP_KERNEL | __GFP_NOWARN); 259 260 if (members) { 261 pr_debug("%p: allocated with kmalloc\n", members); 262 return members; 263 } 264 265 members = vzalloc(size); 266 if (!members) 267 return NULL; 268 pr_debug("%p: allocated with vmalloc\n", members); 269 270 return members; 271 } 272 EXPORT_SYMBOL_GPL(ip_set_alloc); 273 274 void 275 ip_set_free(void *members) 276 { 277 pr_debug("%p: free with %s\n", members, 278 is_vmalloc_addr(members) ? "vfree" : "kfree"); 279 kvfree(members); 280 } 281 EXPORT_SYMBOL_GPL(ip_set_free); 282 283 static inline bool 284 flag_nested(const struct nlattr *nla) 285 { 286 return nla->nla_type & NLA_F_NESTED; 287 } 288 289 static const struct nla_policy ipaddr_policy[IPSET_ATTR_IPADDR_MAX + 1] = { 290 [IPSET_ATTR_IPADDR_IPV4] = { .type = NLA_U32 }, 291 [IPSET_ATTR_IPADDR_IPV6] = { .type = NLA_BINARY, 292 .len = sizeof(struct in6_addr) }, 293 }; 294 295 int 296 ip_set_get_ipaddr4(struct nlattr *nla, __be32 *ipaddr) 297 { 298 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1]; 299 300 if (unlikely(!flag_nested(nla))) 301 return -IPSET_ERR_PROTOCOL; 302 if (nla_parse_nested_deprecated(tb, IPSET_ATTR_IPADDR_MAX, nla, ipaddr_policy, NULL)) 303 return -IPSET_ERR_PROTOCOL; 304 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV4))) 305 return -IPSET_ERR_PROTOCOL; 306 307 *ipaddr = nla_get_be32(tb[IPSET_ATTR_IPADDR_IPV4]); 308 return 0; 309 } 310 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr4); 311 312 int 313 ip_set_get_ipaddr6(struct nlattr *nla, union nf_inet_addr *ipaddr) 314 { 315 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1]; 316 317 if (unlikely(!flag_nested(nla))) 318 return -IPSET_ERR_PROTOCOL; 319 320 if (nla_parse_nested_deprecated(tb, IPSET_ATTR_IPADDR_MAX, nla, ipaddr_policy, NULL)) 321 return -IPSET_ERR_PROTOCOL; 322 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV6))) 323 return -IPSET_ERR_PROTOCOL; 324 325 memcpy(ipaddr, nla_data(tb[IPSET_ATTR_IPADDR_IPV6]), 326 sizeof(struct in6_addr)); 327 return 0; 328 } 329 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr6); 330 331 typedef void (*destroyer)(struct ip_set *, void *); 332 /* ipset data extension types, in size order */ 333 334 const struct ip_set_ext_type ip_set_extensions[] = { 335 [IPSET_EXT_ID_COUNTER] = { 336 .type = IPSET_EXT_COUNTER, 337 .flag = IPSET_FLAG_WITH_COUNTERS, 338 .len = sizeof(struct ip_set_counter), 339 .align = __alignof__(struct ip_set_counter), 340 }, 341 [IPSET_EXT_ID_TIMEOUT] = { 342 .type = IPSET_EXT_TIMEOUT, 343 .len = sizeof(unsigned long), 344 .align = __alignof__(unsigned long), 345 }, 346 [IPSET_EXT_ID_SKBINFO] = { 347 .type = IPSET_EXT_SKBINFO, 348 .flag = IPSET_FLAG_WITH_SKBINFO, 349 .len = sizeof(struct ip_set_skbinfo), 350 .align = __alignof__(struct ip_set_skbinfo), 351 }, 352 [IPSET_EXT_ID_COMMENT] = { 353 .type = IPSET_EXT_COMMENT | IPSET_EXT_DESTROY, 354 .flag = IPSET_FLAG_WITH_COMMENT, 355 .len = sizeof(struct ip_set_comment), 356 .align = __alignof__(struct ip_set_comment), 357 .destroy = (destroyer) ip_set_comment_free, 358 }, 359 }; 360 EXPORT_SYMBOL_GPL(ip_set_extensions); 361 362 static inline bool 363 add_extension(enum ip_set_ext_id id, u32 flags, struct nlattr *tb[]) 364 { 365 return ip_set_extensions[id].flag ? 366 (flags & ip_set_extensions[id].flag) : 367 !!tb[IPSET_ATTR_TIMEOUT]; 368 } 369 370 size_t 371 ip_set_elem_len(struct ip_set *set, struct nlattr *tb[], size_t len, 372 size_t align) 373 { 374 enum ip_set_ext_id id; 375 u32 cadt_flags = 0; 376 377 if (tb[IPSET_ATTR_CADT_FLAGS]) 378 cadt_flags = ip_set_get_h32(tb[IPSET_ATTR_CADT_FLAGS]); 379 if (cadt_flags & IPSET_FLAG_WITH_FORCEADD) 380 set->flags |= IPSET_CREATE_FLAG_FORCEADD; 381 if (!align) 382 align = 1; 383 for (id = 0; id < IPSET_EXT_ID_MAX; id++) { 384 if (!add_extension(id, cadt_flags, tb)) 385 continue; 386 len = ALIGN(len, ip_set_extensions[id].align); 387 set->offset[id] = len; 388 set->extensions |= ip_set_extensions[id].type; 389 len += ip_set_extensions[id].len; 390 } 391 return ALIGN(len, align); 392 } 393 EXPORT_SYMBOL_GPL(ip_set_elem_len); 394 395 int 396 ip_set_get_extensions(struct ip_set *set, struct nlattr *tb[], 397 struct ip_set_ext *ext) 398 { 399 u64 fullmark; 400 401 if (unlikely(!ip_set_optattr_netorder(tb, IPSET_ATTR_TIMEOUT) || 402 !ip_set_optattr_netorder(tb, IPSET_ATTR_PACKETS) || 403 !ip_set_optattr_netorder(tb, IPSET_ATTR_BYTES) || 404 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBMARK) || 405 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBPRIO) || 406 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBQUEUE))) 407 return -IPSET_ERR_PROTOCOL; 408 409 if (tb[IPSET_ATTR_TIMEOUT]) { 410 if (!SET_WITH_TIMEOUT(set)) 411 return -IPSET_ERR_TIMEOUT; 412 ext->timeout = ip_set_timeout_uget(tb[IPSET_ATTR_TIMEOUT]); 413 } 414 if (tb[IPSET_ATTR_BYTES] || tb[IPSET_ATTR_PACKETS]) { 415 if (!SET_WITH_COUNTER(set)) 416 return -IPSET_ERR_COUNTER; 417 if (tb[IPSET_ATTR_BYTES]) 418 ext->bytes = be64_to_cpu(nla_get_be64( 419 tb[IPSET_ATTR_BYTES])); 420 if (tb[IPSET_ATTR_PACKETS]) 421 ext->packets = be64_to_cpu(nla_get_be64( 422 tb[IPSET_ATTR_PACKETS])); 423 } 424 if (tb[IPSET_ATTR_COMMENT]) { 425 if (!SET_WITH_COMMENT(set)) 426 return -IPSET_ERR_COMMENT; 427 ext->comment = ip_set_comment_uget(tb[IPSET_ATTR_COMMENT]); 428 } 429 if (tb[IPSET_ATTR_SKBMARK]) { 430 if (!SET_WITH_SKBINFO(set)) 431 return -IPSET_ERR_SKBINFO; 432 fullmark = be64_to_cpu(nla_get_be64(tb[IPSET_ATTR_SKBMARK])); 433 ext->skbinfo.skbmark = fullmark >> 32; 434 ext->skbinfo.skbmarkmask = fullmark & 0xffffffff; 435 } 436 if (tb[IPSET_ATTR_SKBPRIO]) { 437 if (!SET_WITH_SKBINFO(set)) 438 return -IPSET_ERR_SKBINFO; 439 ext->skbinfo.skbprio = 440 be32_to_cpu(nla_get_be32(tb[IPSET_ATTR_SKBPRIO])); 441 } 442 if (tb[IPSET_ATTR_SKBQUEUE]) { 443 if (!SET_WITH_SKBINFO(set)) 444 return -IPSET_ERR_SKBINFO; 445 ext->skbinfo.skbqueue = 446 be16_to_cpu(nla_get_be16(tb[IPSET_ATTR_SKBQUEUE])); 447 } 448 return 0; 449 } 450 EXPORT_SYMBOL_GPL(ip_set_get_extensions); 451 452 int 453 ip_set_put_extensions(struct sk_buff *skb, const struct ip_set *set, 454 const void *e, bool active) 455 { 456 if (SET_WITH_TIMEOUT(set)) { 457 unsigned long *timeout = ext_timeout(e, set); 458 459 if (nla_put_net32(skb, IPSET_ATTR_TIMEOUT, 460 htonl(active ? ip_set_timeout_get(timeout) 461 : *timeout))) 462 return -EMSGSIZE; 463 } 464 if (SET_WITH_COUNTER(set) && 465 ip_set_put_counter(skb, ext_counter(e, set))) 466 return -EMSGSIZE; 467 if (SET_WITH_COMMENT(set) && 468 ip_set_put_comment(skb, ext_comment(e, set))) 469 return -EMSGSIZE; 470 if (SET_WITH_SKBINFO(set) && 471 ip_set_put_skbinfo(skb, ext_skbinfo(e, set))) 472 return -EMSGSIZE; 473 return 0; 474 } 475 EXPORT_SYMBOL_GPL(ip_set_put_extensions); 476 477 bool 478 ip_set_match_extensions(struct ip_set *set, const struct ip_set_ext *ext, 479 struct ip_set_ext *mext, u32 flags, void *data) 480 { 481 if (SET_WITH_TIMEOUT(set) && 482 ip_set_timeout_expired(ext_timeout(data, set))) 483 return false; 484 if (SET_WITH_COUNTER(set)) { 485 struct ip_set_counter *counter = ext_counter(data, set); 486 487 if (flags & IPSET_FLAG_MATCH_COUNTERS && 488 !(ip_set_match_counter(ip_set_get_packets(counter), 489 mext->packets, mext->packets_op) && 490 ip_set_match_counter(ip_set_get_bytes(counter), 491 mext->bytes, mext->bytes_op))) 492 return false; 493 ip_set_update_counter(counter, ext, flags); 494 } 495 if (SET_WITH_SKBINFO(set)) 496 ip_set_get_skbinfo(ext_skbinfo(data, set), 497 ext, mext, flags); 498 return true; 499 } 500 EXPORT_SYMBOL_GPL(ip_set_match_extensions); 501 502 /* Creating/destroying/renaming/swapping affect the existence and 503 * the properties of a set. All of these can be executed from userspace 504 * only and serialized by the nfnl mutex indirectly from nfnetlink. 505 * 506 * Sets are identified by their index in ip_set_list and the index 507 * is used by the external references (set/SET netfilter modules). 508 * 509 * The set behind an index may change by swapping only, from userspace. 510 */ 511 512 static inline void 513 __ip_set_get(struct ip_set *set) 514 { 515 write_lock_bh(&ip_set_ref_lock); 516 set->ref++; 517 write_unlock_bh(&ip_set_ref_lock); 518 } 519 520 static inline void 521 __ip_set_put(struct ip_set *set) 522 { 523 write_lock_bh(&ip_set_ref_lock); 524 BUG_ON(set->ref == 0); 525 set->ref--; 526 write_unlock_bh(&ip_set_ref_lock); 527 } 528 529 /* set->ref can be swapped out by ip_set_swap, netlink events (like dump) need 530 * a separate reference counter 531 */ 532 static inline void 533 __ip_set_put_netlink(struct ip_set *set) 534 { 535 write_lock_bh(&ip_set_ref_lock); 536 BUG_ON(set->ref_netlink == 0); 537 set->ref_netlink--; 538 write_unlock_bh(&ip_set_ref_lock); 539 } 540 541 /* Add, del and test set entries from kernel. 542 * 543 * The set behind the index must exist and must be referenced 544 * so it can't be destroyed (or changed) under our foot. 545 */ 546 547 static inline struct ip_set * 548 ip_set_rcu_get(struct net *net, ip_set_id_t index) 549 { 550 struct ip_set *set; 551 struct ip_set_net *inst = ip_set_pernet(net); 552 553 rcu_read_lock(); 554 /* ip_set_list itself needs to be protected */ 555 set = rcu_dereference(inst->ip_set_list)[index]; 556 rcu_read_unlock(); 557 558 return set; 559 } 560 561 int 562 ip_set_test(ip_set_id_t index, const struct sk_buff *skb, 563 const struct xt_action_param *par, struct ip_set_adt_opt *opt) 564 { 565 struct ip_set *set = ip_set_rcu_get(xt_net(par), index); 566 int ret = 0; 567 568 BUG_ON(!set); 569 pr_debug("set %s, index %u\n", set->name, index); 570 571 if (opt->dim < set->type->dimension || 572 !(opt->family == set->family || set->family == NFPROTO_UNSPEC)) 573 return 0; 574 575 rcu_read_lock_bh(); 576 ret = set->variant->kadt(set, skb, par, IPSET_TEST, opt); 577 rcu_read_unlock_bh(); 578 579 if (ret == -EAGAIN) { 580 /* Type requests element to be completed */ 581 pr_debug("element must be completed, ADD is triggered\n"); 582 spin_lock_bh(&set->lock); 583 set->variant->kadt(set, skb, par, IPSET_ADD, opt); 584 spin_unlock_bh(&set->lock); 585 ret = 1; 586 } else { 587 /* --return-nomatch: invert matched element */ 588 if ((opt->cmdflags & IPSET_FLAG_RETURN_NOMATCH) && 589 (set->type->features & IPSET_TYPE_NOMATCH) && 590 (ret > 0 || ret == -ENOTEMPTY)) 591 ret = -ret; 592 } 593 594 /* Convert error codes to nomatch */ 595 return (ret < 0 ? 0 : ret); 596 } 597 EXPORT_SYMBOL_GPL(ip_set_test); 598 599 int 600 ip_set_add(ip_set_id_t index, const struct sk_buff *skb, 601 const struct xt_action_param *par, struct ip_set_adt_opt *opt) 602 { 603 struct ip_set *set = ip_set_rcu_get(xt_net(par), index); 604 int ret; 605 606 BUG_ON(!set); 607 pr_debug("set %s, index %u\n", set->name, index); 608 609 if (opt->dim < set->type->dimension || 610 !(opt->family == set->family || set->family == NFPROTO_UNSPEC)) 611 return -IPSET_ERR_TYPE_MISMATCH; 612 613 spin_lock_bh(&set->lock); 614 ret = set->variant->kadt(set, skb, par, IPSET_ADD, opt); 615 spin_unlock_bh(&set->lock); 616 617 return ret; 618 } 619 EXPORT_SYMBOL_GPL(ip_set_add); 620 621 int 622 ip_set_del(ip_set_id_t index, const struct sk_buff *skb, 623 const struct xt_action_param *par, struct ip_set_adt_opt *opt) 624 { 625 struct ip_set *set = ip_set_rcu_get(xt_net(par), index); 626 int ret = 0; 627 628 BUG_ON(!set); 629 pr_debug("set %s, index %u\n", set->name, index); 630 631 if (opt->dim < set->type->dimension || 632 !(opt->family == set->family || set->family == NFPROTO_UNSPEC)) 633 return -IPSET_ERR_TYPE_MISMATCH; 634 635 spin_lock_bh(&set->lock); 636 ret = set->variant->kadt(set, skb, par, IPSET_DEL, opt); 637 spin_unlock_bh(&set->lock); 638 639 return ret; 640 } 641 EXPORT_SYMBOL_GPL(ip_set_del); 642 643 /* Find set by name, reference it once. The reference makes sure the 644 * thing pointed to, does not go away under our feet. 645 * 646 */ 647 ip_set_id_t 648 ip_set_get_byname(struct net *net, const char *name, struct ip_set **set) 649 { 650 ip_set_id_t i, index = IPSET_INVALID_ID; 651 struct ip_set *s; 652 struct ip_set_net *inst = ip_set_pernet(net); 653 654 rcu_read_lock(); 655 for (i = 0; i < inst->ip_set_max; i++) { 656 s = rcu_dereference(inst->ip_set_list)[i]; 657 if (s && STRNCMP(s->name, name)) { 658 __ip_set_get(s); 659 index = i; 660 *set = s; 661 break; 662 } 663 } 664 rcu_read_unlock(); 665 666 return index; 667 } 668 EXPORT_SYMBOL_GPL(ip_set_get_byname); 669 670 /* If the given set pointer points to a valid set, decrement 671 * reference count by 1. The caller shall not assume the index 672 * to be valid, after calling this function. 673 * 674 */ 675 676 static inline void 677 __ip_set_put_byindex(struct ip_set_net *inst, ip_set_id_t index) 678 { 679 struct ip_set *set; 680 681 rcu_read_lock(); 682 set = rcu_dereference(inst->ip_set_list)[index]; 683 if (set) 684 __ip_set_put(set); 685 rcu_read_unlock(); 686 } 687 688 void 689 ip_set_put_byindex(struct net *net, ip_set_id_t index) 690 { 691 struct ip_set_net *inst = ip_set_pernet(net); 692 693 __ip_set_put_byindex(inst, index); 694 } 695 EXPORT_SYMBOL_GPL(ip_set_put_byindex); 696 697 /* Get the name of a set behind a set index. 698 * Set itself is protected by RCU, but its name isn't: to protect against 699 * renaming, grab ip_set_ref_lock as reader (see ip_set_rename()) and copy the 700 * name. 701 */ 702 void 703 ip_set_name_byindex(struct net *net, ip_set_id_t index, char *name) 704 { 705 struct ip_set *set = ip_set_rcu_get(net, index); 706 707 BUG_ON(!set); 708 709 read_lock_bh(&ip_set_ref_lock); 710 strncpy(name, set->name, IPSET_MAXNAMELEN); 711 read_unlock_bh(&ip_set_ref_lock); 712 } 713 EXPORT_SYMBOL_GPL(ip_set_name_byindex); 714 715 /* Routines to call by external subsystems, which do not 716 * call nfnl_lock for us. 717 */ 718 719 /* Find set by index, reference it once. The reference makes sure the 720 * thing pointed to, does not go away under our feet. 721 * 722 * The nfnl mutex is used in the function. 723 */ 724 ip_set_id_t 725 ip_set_nfnl_get_byindex(struct net *net, ip_set_id_t index) 726 { 727 struct ip_set *set; 728 struct ip_set_net *inst = ip_set_pernet(net); 729 730 if (index >= inst->ip_set_max) 731 return IPSET_INVALID_ID; 732 733 nfnl_lock(NFNL_SUBSYS_IPSET); 734 set = ip_set(inst, index); 735 if (set) 736 __ip_set_get(set); 737 else 738 index = IPSET_INVALID_ID; 739 nfnl_unlock(NFNL_SUBSYS_IPSET); 740 741 return index; 742 } 743 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex); 744 745 /* If the given set pointer points to a valid set, decrement 746 * reference count by 1. The caller shall not assume the index 747 * to be valid, after calling this function. 748 * 749 * The nfnl mutex is used in the function. 750 */ 751 void 752 ip_set_nfnl_put(struct net *net, ip_set_id_t index) 753 { 754 struct ip_set *set; 755 struct ip_set_net *inst = ip_set_pernet(net); 756 757 nfnl_lock(NFNL_SUBSYS_IPSET); 758 if (!inst->is_deleted) { /* already deleted from ip_set_net_exit() */ 759 set = ip_set(inst, index); 760 if (set) 761 __ip_set_put(set); 762 } 763 nfnl_unlock(NFNL_SUBSYS_IPSET); 764 } 765 EXPORT_SYMBOL_GPL(ip_set_nfnl_put); 766 767 /* Communication protocol with userspace over netlink. 768 * 769 * The commands are serialized by the nfnl mutex. 770 */ 771 772 static inline u8 protocol(const struct nlattr * const tb[]) 773 { 774 return nla_get_u8(tb[IPSET_ATTR_PROTOCOL]); 775 } 776 777 static inline bool 778 protocol_failed(const struct nlattr * const tb[]) 779 { 780 return !tb[IPSET_ATTR_PROTOCOL] || protocol(tb) != IPSET_PROTOCOL; 781 } 782 783 static inline bool 784 protocol_min_failed(const struct nlattr * const tb[]) 785 { 786 return !tb[IPSET_ATTR_PROTOCOL] || protocol(tb) < IPSET_PROTOCOL_MIN; 787 } 788 789 static inline u32 790 flag_exist(const struct nlmsghdr *nlh) 791 { 792 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST; 793 } 794 795 static struct nlmsghdr * 796 start_msg(struct sk_buff *skb, u32 portid, u32 seq, unsigned int flags, 797 enum ipset_cmd cmd) 798 { 799 struct nlmsghdr *nlh; 800 struct nfgenmsg *nfmsg; 801 802 nlh = nlmsg_put(skb, portid, seq, nfnl_msg_type(NFNL_SUBSYS_IPSET, cmd), 803 sizeof(*nfmsg), flags); 804 if (!nlh) 805 return NULL; 806 807 nfmsg = nlmsg_data(nlh); 808 nfmsg->nfgen_family = NFPROTO_IPV4; 809 nfmsg->version = NFNETLINK_V0; 810 nfmsg->res_id = 0; 811 812 return nlh; 813 } 814 815 /* Create a set */ 816 817 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = { 818 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 819 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 820 .len = IPSET_MAXNAMELEN - 1 }, 821 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING, 822 .len = IPSET_MAXNAMELEN - 1}, 823 [IPSET_ATTR_REVISION] = { .type = NLA_U8 }, 824 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 }, 825 [IPSET_ATTR_DATA] = { .type = NLA_NESTED }, 826 }; 827 828 static struct ip_set * 829 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id) 830 { 831 struct ip_set *set = NULL; 832 ip_set_id_t i; 833 834 *id = IPSET_INVALID_ID; 835 for (i = 0; i < inst->ip_set_max; i++) { 836 set = ip_set(inst, i); 837 if (set && STRNCMP(set->name, name)) { 838 *id = i; 839 break; 840 } 841 } 842 return (*id == IPSET_INVALID_ID ? NULL : set); 843 } 844 845 static inline struct ip_set * 846 find_set(struct ip_set_net *inst, const char *name) 847 { 848 ip_set_id_t id; 849 850 return find_set_and_id(inst, name, &id); 851 } 852 853 static int 854 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index, 855 struct ip_set **set) 856 { 857 struct ip_set *s; 858 ip_set_id_t i; 859 860 *index = IPSET_INVALID_ID; 861 for (i = 0; i < inst->ip_set_max; i++) { 862 s = ip_set(inst, i); 863 if (!s) { 864 if (*index == IPSET_INVALID_ID) 865 *index = i; 866 } else if (STRNCMP(name, s->name)) { 867 /* Name clash */ 868 *set = s; 869 return -EEXIST; 870 } 871 } 872 if (*index == IPSET_INVALID_ID) 873 /* No free slot remained */ 874 return -IPSET_ERR_MAX_SETS; 875 return 0; 876 } 877 878 static int ip_set_none(struct net *net, struct sock *ctnl, struct sk_buff *skb, 879 const struct nlmsghdr *nlh, 880 const struct nlattr * const attr[], 881 struct netlink_ext_ack *extack) 882 { 883 return -EOPNOTSUPP; 884 } 885 886 static int ip_set_create(struct net *net, struct sock *ctnl, 887 struct sk_buff *skb, const struct nlmsghdr *nlh, 888 const struct nlattr * const attr[], 889 struct netlink_ext_ack *extack) 890 { 891 struct ip_set_net *inst = ip_set_pernet(net); 892 struct ip_set *set, *clash = NULL; 893 ip_set_id_t index = IPSET_INVALID_ID; 894 struct nlattr *tb[IPSET_ATTR_CREATE_MAX + 1] = {}; 895 const char *name, *typename; 896 u8 family, revision; 897 u32 flags = flag_exist(nlh); 898 int ret = 0; 899 900 if (unlikely(protocol_min_failed(attr) || 901 !attr[IPSET_ATTR_SETNAME] || 902 !attr[IPSET_ATTR_TYPENAME] || 903 !attr[IPSET_ATTR_REVISION] || 904 !attr[IPSET_ATTR_FAMILY] || 905 (attr[IPSET_ATTR_DATA] && 906 !flag_nested(attr[IPSET_ATTR_DATA])))) 907 return -IPSET_ERR_PROTOCOL; 908 909 name = nla_data(attr[IPSET_ATTR_SETNAME]); 910 typename = nla_data(attr[IPSET_ATTR_TYPENAME]); 911 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]); 912 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]); 913 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n", 914 name, typename, family_name(family), revision); 915 916 /* First, and without any locks, allocate and initialize 917 * a normal base set structure. 918 */ 919 set = kzalloc(sizeof(*set), GFP_KERNEL); 920 if (!set) 921 return -ENOMEM; 922 spin_lock_init(&set->lock); 923 strlcpy(set->name, name, IPSET_MAXNAMELEN); 924 set->family = family; 925 set->revision = revision; 926 927 /* Next, check that we know the type, and take 928 * a reference on the type, to make sure it stays available 929 * while constructing our new set. 930 * 931 * After referencing the type, we try to create the type 932 * specific part of the set without holding any locks. 933 */ 934 ret = find_set_type_get(typename, family, revision, &set->type); 935 if (ret) 936 goto out; 937 938 /* Without holding any locks, create private part. */ 939 if (attr[IPSET_ATTR_DATA] && 940 nla_parse_nested_deprecated(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA], set->type->create_policy, NULL)) { 941 ret = -IPSET_ERR_PROTOCOL; 942 goto put_out; 943 } 944 945 ret = set->type->create(net, set, tb, flags); 946 if (ret != 0) 947 goto put_out; 948 949 /* BTW, ret==0 here. */ 950 951 /* Here, we have a valid, constructed set and we are protected 952 * by the nfnl mutex. Find the first free index in ip_set_list 953 * and check clashing. 954 */ 955 ret = find_free_id(inst, set->name, &index, &clash); 956 if (ret == -EEXIST) { 957 /* If this is the same set and requested, ignore error */ 958 if ((flags & IPSET_FLAG_EXIST) && 959 STRNCMP(set->type->name, clash->type->name) && 960 set->type->family == clash->type->family && 961 set->type->revision_min == clash->type->revision_min && 962 set->type->revision_max == clash->type->revision_max && 963 set->variant->same_set(set, clash)) 964 ret = 0; 965 goto cleanup; 966 } else if (ret == -IPSET_ERR_MAX_SETS) { 967 struct ip_set **list, **tmp; 968 ip_set_id_t i = inst->ip_set_max + IP_SET_INC; 969 970 if (i < inst->ip_set_max || i == IPSET_INVALID_ID) 971 /* Wraparound */ 972 goto cleanup; 973 974 list = kvcalloc(i, sizeof(struct ip_set *), GFP_KERNEL); 975 if (!list) 976 goto cleanup; 977 /* nfnl mutex is held, both lists are valid */ 978 tmp = ip_set_dereference(inst->ip_set_list); 979 memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max); 980 rcu_assign_pointer(inst->ip_set_list, list); 981 /* Make sure all current packets have passed through */ 982 synchronize_net(); 983 /* Use new list */ 984 index = inst->ip_set_max; 985 inst->ip_set_max = i; 986 kvfree(tmp); 987 ret = 0; 988 } else if (ret) { 989 goto cleanup; 990 } 991 992 /* Finally! Add our shiny new set to the list, and be done. */ 993 pr_debug("create: '%s' created with index %u!\n", set->name, index); 994 ip_set(inst, index) = set; 995 996 return ret; 997 998 cleanup: 999 set->variant->destroy(set); 1000 put_out: 1001 module_put(set->type->me); 1002 out: 1003 kfree(set); 1004 return ret; 1005 } 1006 1007 /* Destroy sets */ 1008 1009 static const struct nla_policy 1010 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = { 1011 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1012 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 1013 .len = IPSET_MAXNAMELEN - 1 }, 1014 }; 1015 1016 static void 1017 ip_set_destroy_set(struct ip_set *set) 1018 { 1019 pr_debug("set: %s\n", set->name); 1020 1021 /* Must call it without holding any lock */ 1022 set->variant->destroy(set); 1023 module_put(set->type->me); 1024 kfree(set); 1025 } 1026 1027 static int ip_set_destroy(struct net *net, struct sock *ctnl, 1028 struct sk_buff *skb, const struct nlmsghdr *nlh, 1029 const struct nlattr * const attr[], 1030 struct netlink_ext_ack *extack) 1031 { 1032 struct ip_set_net *inst = ip_set_pernet(net); 1033 struct ip_set *s; 1034 ip_set_id_t i; 1035 int ret = 0; 1036 1037 if (unlikely(protocol_min_failed(attr))) 1038 return -IPSET_ERR_PROTOCOL; 1039 1040 /* Must wait for flush to be really finished in list:set */ 1041 rcu_barrier(); 1042 1043 /* Commands are serialized and references are 1044 * protected by the ip_set_ref_lock. 1045 * External systems (i.e. xt_set) must call 1046 * ip_set_put|get_nfnl_* functions, that way we 1047 * can safely check references here. 1048 * 1049 * list:set timer can only decrement the reference 1050 * counter, so if it's already zero, we can proceed 1051 * without holding the lock. 1052 */ 1053 read_lock_bh(&ip_set_ref_lock); 1054 if (!attr[IPSET_ATTR_SETNAME]) { 1055 for (i = 0; i < inst->ip_set_max; i++) { 1056 s = ip_set(inst, i); 1057 if (s && (s->ref || s->ref_netlink)) { 1058 ret = -IPSET_ERR_BUSY; 1059 goto out; 1060 } 1061 } 1062 inst->is_destroyed = true; 1063 read_unlock_bh(&ip_set_ref_lock); 1064 for (i = 0; i < inst->ip_set_max; i++) { 1065 s = ip_set(inst, i); 1066 if (s) { 1067 ip_set(inst, i) = NULL; 1068 ip_set_destroy_set(s); 1069 } 1070 } 1071 /* Modified by ip_set_destroy() only, which is serialized */ 1072 inst->is_destroyed = false; 1073 } else { 1074 s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]), 1075 &i); 1076 if (!s) { 1077 ret = -ENOENT; 1078 goto out; 1079 } else if (s->ref || s->ref_netlink) { 1080 ret = -IPSET_ERR_BUSY; 1081 goto out; 1082 } 1083 ip_set(inst, i) = NULL; 1084 read_unlock_bh(&ip_set_ref_lock); 1085 1086 ip_set_destroy_set(s); 1087 } 1088 return 0; 1089 out: 1090 read_unlock_bh(&ip_set_ref_lock); 1091 return ret; 1092 } 1093 1094 /* Flush sets */ 1095 1096 static void 1097 ip_set_flush_set(struct ip_set *set) 1098 { 1099 pr_debug("set: %s\n", set->name); 1100 1101 spin_lock_bh(&set->lock); 1102 set->variant->flush(set); 1103 spin_unlock_bh(&set->lock); 1104 } 1105 1106 static int ip_set_flush(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1107 const struct nlmsghdr *nlh, 1108 const struct nlattr * const attr[], 1109 struct netlink_ext_ack *extack) 1110 { 1111 struct ip_set_net *inst = ip_set_pernet(net); 1112 struct ip_set *s; 1113 ip_set_id_t i; 1114 1115 if (unlikely(protocol_min_failed(attr))) 1116 return -IPSET_ERR_PROTOCOL; 1117 1118 if (!attr[IPSET_ATTR_SETNAME]) { 1119 for (i = 0; i < inst->ip_set_max; i++) { 1120 s = ip_set(inst, i); 1121 if (s) 1122 ip_set_flush_set(s); 1123 } 1124 } else { 1125 s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME])); 1126 if (!s) 1127 return -ENOENT; 1128 1129 ip_set_flush_set(s); 1130 } 1131 1132 return 0; 1133 } 1134 1135 /* Rename a set */ 1136 1137 static const struct nla_policy 1138 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = { 1139 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1140 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 1141 .len = IPSET_MAXNAMELEN - 1 }, 1142 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING, 1143 .len = IPSET_MAXNAMELEN - 1 }, 1144 }; 1145 1146 static int ip_set_rename(struct net *net, struct sock *ctnl, 1147 struct sk_buff *skb, const struct nlmsghdr *nlh, 1148 const struct nlattr * const attr[], 1149 struct netlink_ext_ack *extack) 1150 { 1151 struct ip_set_net *inst = ip_set_pernet(net); 1152 struct ip_set *set, *s; 1153 const char *name2; 1154 ip_set_id_t i; 1155 int ret = 0; 1156 1157 if (unlikely(protocol_min_failed(attr) || 1158 !attr[IPSET_ATTR_SETNAME] || 1159 !attr[IPSET_ATTR_SETNAME2])) 1160 return -IPSET_ERR_PROTOCOL; 1161 1162 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME])); 1163 if (!set) 1164 return -ENOENT; 1165 1166 write_lock_bh(&ip_set_ref_lock); 1167 if (set->ref != 0) { 1168 ret = -IPSET_ERR_REFERENCED; 1169 goto out; 1170 } 1171 1172 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]); 1173 for (i = 0; i < inst->ip_set_max; i++) { 1174 s = ip_set(inst, i); 1175 if (s && STRNCMP(s->name, name2)) { 1176 ret = -IPSET_ERR_EXIST_SETNAME2; 1177 goto out; 1178 } 1179 } 1180 strncpy(set->name, name2, IPSET_MAXNAMELEN); 1181 1182 out: 1183 write_unlock_bh(&ip_set_ref_lock); 1184 return ret; 1185 } 1186 1187 /* Swap two sets so that name/index points to the other. 1188 * References and set names are also swapped. 1189 * 1190 * The commands are serialized by the nfnl mutex and references are 1191 * protected by the ip_set_ref_lock. The kernel interfaces 1192 * do not hold the mutex but the pointer settings are atomic 1193 * so the ip_set_list always contains valid pointers to the sets. 1194 */ 1195 1196 static int ip_set_swap(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1197 const struct nlmsghdr *nlh, 1198 const struct nlattr * const attr[], 1199 struct netlink_ext_ack *extack) 1200 { 1201 struct ip_set_net *inst = ip_set_pernet(net); 1202 struct ip_set *from, *to; 1203 ip_set_id_t from_id, to_id; 1204 char from_name[IPSET_MAXNAMELEN]; 1205 1206 if (unlikely(protocol_min_failed(attr) || 1207 !attr[IPSET_ATTR_SETNAME] || 1208 !attr[IPSET_ATTR_SETNAME2])) 1209 return -IPSET_ERR_PROTOCOL; 1210 1211 from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]), 1212 &from_id); 1213 if (!from) 1214 return -ENOENT; 1215 1216 to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]), 1217 &to_id); 1218 if (!to) 1219 return -IPSET_ERR_EXIST_SETNAME2; 1220 1221 /* Features must not change. 1222 * Not an artifical restriction anymore, as we must prevent 1223 * possible loops created by swapping in setlist type of sets. 1224 */ 1225 if (!(from->type->features == to->type->features && 1226 from->family == to->family)) 1227 return -IPSET_ERR_TYPE_MISMATCH; 1228 1229 write_lock_bh(&ip_set_ref_lock); 1230 1231 if (from->ref_netlink || to->ref_netlink) { 1232 write_unlock_bh(&ip_set_ref_lock); 1233 return -EBUSY; 1234 } 1235 1236 strncpy(from_name, from->name, IPSET_MAXNAMELEN); 1237 strncpy(from->name, to->name, IPSET_MAXNAMELEN); 1238 strncpy(to->name, from_name, IPSET_MAXNAMELEN); 1239 1240 swap(from->ref, to->ref); 1241 ip_set(inst, from_id) = to; 1242 ip_set(inst, to_id) = from; 1243 write_unlock_bh(&ip_set_ref_lock); 1244 1245 return 0; 1246 } 1247 1248 /* List/save set data */ 1249 1250 #define DUMP_INIT 0 1251 #define DUMP_ALL 1 1252 #define DUMP_ONE 2 1253 #define DUMP_LAST 3 1254 1255 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF) 1256 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16) 1257 1258 static int 1259 ip_set_dump_done(struct netlink_callback *cb) 1260 { 1261 if (cb->args[IPSET_CB_ARG0]) { 1262 struct ip_set_net *inst = 1263 (struct ip_set_net *)cb->args[IPSET_CB_NET]; 1264 ip_set_id_t index = (ip_set_id_t)cb->args[IPSET_CB_INDEX]; 1265 struct ip_set *set = ip_set_ref_netlink(inst, index); 1266 1267 if (set->variant->uref) 1268 set->variant->uref(set, cb, false); 1269 pr_debug("release set %s\n", set->name); 1270 __ip_set_put_netlink(set); 1271 } 1272 return 0; 1273 } 1274 1275 static inline void 1276 dump_attrs(struct nlmsghdr *nlh) 1277 { 1278 const struct nlattr *attr; 1279 int rem; 1280 1281 pr_debug("dump nlmsg\n"); 1282 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) { 1283 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len); 1284 } 1285 } 1286 1287 static int 1288 dump_init(struct netlink_callback *cb, struct ip_set_net *inst) 1289 { 1290 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb); 1291 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg)); 1292 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1]; 1293 struct nlattr *attr = (void *)nlh + min_len; 1294 u32 dump_type; 1295 ip_set_id_t index; 1296 1297 /* Second pass, so parser can't fail */ 1298 nla_parse_deprecated(cda, IPSET_ATTR_CMD_MAX, attr, 1299 nlh->nlmsg_len - min_len, ip_set_setname_policy, 1300 NULL); 1301 1302 cb->args[IPSET_CB_PROTO] = nla_get_u8(cda[IPSET_ATTR_PROTOCOL]); 1303 if (cda[IPSET_ATTR_SETNAME]) { 1304 struct ip_set *set; 1305 1306 set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]), 1307 &index); 1308 if (!set) 1309 return -ENOENT; 1310 1311 dump_type = DUMP_ONE; 1312 cb->args[IPSET_CB_INDEX] = index; 1313 } else { 1314 dump_type = DUMP_ALL; 1315 } 1316 1317 if (cda[IPSET_ATTR_FLAGS]) { 1318 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]); 1319 1320 dump_type |= (f << 16); 1321 } 1322 cb->args[IPSET_CB_NET] = (unsigned long)inst; 1323 cb->args[IPSET_CB_DUMP] = dump_type; 1324 1325 return 0; 1326 } 1327 1328 static int 1329 ip_set_dump_start(struct sk_buff *skb, struct netlink_callback *cb) 1330 { 1331 ip_set_id_t index = IPSET_INVALID_ID, max; 1332 struct ip_set *set = NULL; 1333 struct nlmsghdr *nlh = NULL; 1334 unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0; 1335 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk)); 1336 u32 dump_type, dump_flags; 1337 bool is_destroyed; 1338 int ret = 0; 1339 1340 if (!cb->args[IPSET_CB_DUMP]) { 1341 ret = dump_init(cb, inst); 1342 if (ret < 0) { 1343 nlh = nlmsg_hdr(cb->skb); 1344 /* We have to create and send the error message 1345 * manually :-( 1346 */ 1347 if (nlh->nlmsg_flags & NLM_F_ACK) 1348 netlink_ack(cb->skb, nlh, ret, NULL); 1349 return ret; 1350 } 1351 } 1352 1353 if (cb->args[IPSET_CB_INDEX] >= inst->ip_set_max) 1354 goto out; 1355 1356 dump_type = DUMP_TYPE(cb->args[IPSET_CB_DUMP]); 1357 dump_flags = DUMP_FLAGS(cb->args[IPSET_CB_DUMP]); 1358 max = dump_type == DUMP_ONE ? cb->args[IPSET_CB_INDEX] + 1 1359 : inst->ip_set_max; 1360 dump_last: 1361 pr_debug("dump type, flag: %u %u index: %ld\n", 1362 dump_type, dump_flags, cb->args[IPSET_CB_INDEX]); 1363 for (; cb->args[IPSET_CB_INDEX] < max; cb->args[IPSET_CB_INDEX]++) { 1364 index = (ip_set_id_t)cb->args[IPSET_CB_INDEX]; 1365 write_lock_bh(&ip_set_ref_lock); 1366 set = ip_set(inst, index); 1367 is_destroyed = inst->is_destroyed; 1368 if (!set || is_destroyed) { 1369 write_unlock_bh(&ip_set_ref_lock); 1370 if (dump_type == DUMP_ONE) { 1371 ret = -ENOENT; 1372 goto out; 1373 } 1374 if (is_destroyed) { 1375 /* All sets are just being destroyed */ 1376 ret = 0; 1377 goto out; 1378 } 1379 continue; 1380 } 1381 /* When dumping all sets, we must dump "sorted" 1382 * so that lists (unions of sets) are dumped last. 1383 */ 1384 if (dump_type != DUMP_ONE && 1385 ((dump_type == DUMP_ALL) == 1386 !!(set->type->features & IPSET_DUMP_LAST))) { 1387 write_unlock_bh(&ip_set_ref_lock); 1388 continue; 1389 } 1390 pr_debug("List set: %s\n", set->name); 1391 if (!cb->args[IPSET_CB_ARG0]) { 1392 /* Start listing: make sure set won't be destroyed */ 1393 pr_debug("reference set\n"); 1394 set->ref_netlink++; 1395 } 1396 write_unlock_bh(&ip_set_ref_lock); 1397 nlh = start_msg(skb, NETLINK_CB(cb->skb).portid, 1398 cb->nlh->nlmsg_seq, flags, 1399 IPSET_CMD_LIST); 1400 if (!nlh) { 1401 ret = -EMSGSIZE; 1402 goto release_refcount; 1403 } 1404 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL, 1405 cb->args[IPSET_CB_PROTO]) || 1406 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name)) 1407 goto nla_put_failure; 1408 if (dump_flags & IPSET_FLAG_LIST_SETNAME) 1409 goto next_set; 1410 switch (cb->args[IPSET_CB_ARG0]) { 1411 case 0: 1412 /* Core header data */ 1413 if (nla_put_string(skb, IPSET_ATTR_TYPENAME, 1414 set->type->name) || 1415 nla_put_u8(skb, IPSET_ATTR_FAMILY, 1416 set->family) || 1417 nla_put_u8(skb, IPSET_ATTR_REVISION, 1418 set->revision)) 1419 goto nla_put_failure; 1420 if (cb->args[IPSET_CB_PROTO] > IPSET_PROTOCOL_MIN && 1421 nla_put_net16(skb, IPSET_ATTR_INDEX, htons(index))) 1422 goto nla_put_failure; 1423 ret = set->variant->head(set, skb); 1424 if (ret < 0) 1425 goto release_refcount; 1426 if (dump_flags & IPSET_FLAG_LIST_HEADER) 1427 goto next_set; 1428 if (set->variant->uref) 1429 set->variant->uref(set, cb, true); 1430 /* fall through */ 1431 default: 1432 ret = set->variant->list(set, skb, cb); 1433 if (!cb->args[IPSET_CB_ARG0]) 1434 /* Set is done, proceed with next one */ 1435 goto next_set; 1436 goto release_refcount; 1437 } 1438 } 1439 /* If we dump all sets, continue with dumping last ones */ 1440 if (dump_type == DUMP_ALL) { 1441 dump_type = DUMP_LAST; 1442 cb->args[IPSET_CB_DUMP] = dump_type | (dump_flags << 16); 1443 cb->args[IPSET_CB_INDEX] = 0; 1444 if (set && set->variant->uref) 1445 set->variant->uref(set, cb, false); 1446 goto dump_last; 1447 } 1448 goto out; 1449 1450 nla_put_failure: 1451 ret = -EFAULT; 1452 next_set: 1453 if (dump_type == DUMP_ONE) 1454 cb->args[IPSET_CB_INDEX] = IPSET_INVALID_ID; 1455 else 1456 cb->args[IPSET_CB_INDEX]++; 1457 release_refcount: 1458 /* If there was an error or set is done, release set */ 1459 if (ret || !cb->args[IPSET_CB_ARG0]) { 1460 set = ip_set_ref_netlink(inst, index); 1461 if (set->variant->uref) 1462 set->variant->uref(set, cb, false); 1463 pr_debug("release set %s\n", set->name); 1464 __ip_set_put_netlink(set); 1465 cb->args[IPSET_CB_ARG0] = 0; 1466 } 1467 out: 1468 if (nlh) { 1469 nlmsg_end(skb, nlh); 1470 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len); 1471 dump_attrs(nlh); 1472 } 1473 1474 return ret < 0 ? ret : skb->len; 1475 } 1476 1477 static int ip_set_dump(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1478 const struct nlmsghdr *nlh, 1479 const struct nlattr * const attr[], 1480 struct netlink_ext_ack *extack) 1481 { 1482 if (unlikely(protocol_min_failed(attr))) 1483 return -IPSET_ERR_PROTOCOL; 1484 1485 { 1486 struct netlink_dump_control c = { 1487 .dump = ip_set_dump_start, 1488 .done = ip_set_dump_done, 1489 }; 1490 return netlink_dump_start(ctnl, skb, nlh, &c); 1491 } 1492 } 1493 1494 /* Add, del and test */ 1495 1496 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = { 1497 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1498 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 1499 .len = IPSET_MAXNAMELEN - 1 }, 1500 [IPSET_ATTR_LINENO] = { .type = NLA_U32 }, 1501 [IPSET_ATTR_DATA] = { .type = NLA_NESTED }, 1502 [IPSET_ATTR_ADT] = { .type = NLA_NESTED }, 1503 }; 1504 1505 static int 1506 call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set, 1507 struct nlattr *tb[], enum ipset_adt adt, 1508 u32 flags, bool use_lineno) 1509 { 1510 int ret; 1511 u32 lineno = 0; 1512 bool eexist = flags & IPSET_FLAG_EXIST, retried = false; 1513 1514 do { 1515 spin_lock_bh(&set->lock); 1516 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried); 1517 spin_unlock_bh(&set->lock); 1518 retried = true; 1519 } while (ret == -EAGAIN && 1520 set->variant->resize && 1521 (ret = set->variant->resize(set, retried)) == 0); 1522 1523 if (!ret || (ret == -IPSET_ERR_EXIST && eexist)) 1524 return 0; 1525 if (lineno && use_lineno) { 1526 /* Error in restore/batch mode: send back lineno */ 1527 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb); 1528 struct sk_buff *skb2; 1529 struct nlmsgerr *errmsg; 1530 size_t payload = min(SIZE_MAX, 1531 sizeof(*errmsg) + nlmsg_len(nlh)); 1532 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg)); 1533 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1]; 1534 struct nlattr *cmdattr; 1535 u32 *errline; 1536 1537 skb2 = nlmsg_new(payload, GFP_KERNEL); 1538 if (!skb2) 1539 return -ENOMEM; 1540 rep = __nlmsg_put(skb2, NETLINK_CB(skb).portid, 1541 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0); 1542 errmsg = nlmsg_data(rep); 1543 errmsg->error = ret; 1544 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len); 1545 cmdattr = (void *)&errmsg->msg + min_len; 1546 1547 nla_parse_deprecated(cda, IPSET_ATTR_CMD_MAX, cmdattr, 1548 nlh->nlmsg_len - min_len, 1549 ip_set_adt_policy, NULL); 1550 1551 errline = nla_data(cda[IPSET_ATTR_LINENO]); 1552 1553 *errline = lineno; 1554 1555 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, 1556 MSG_DONTWAIT); 1557 /* Signal netlink not to send its ACK/errmsg. */ 1558 return -EINTR; 1559 } 1560 1561 return ret; 1562 } 1563 1564 static int ip_set_uadd(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1565 const struct nlmsghdr *nlh, 1566 const struct nlattr * const attr[], 1567 struct netlink_ext_ack *extack) 1568 { 1569 struct ip_set_net *inst = ip_set_pernet(net); 1570 struct ip_set *set; 1571 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {}; 1572 const struct nlattr *nla; 1573 u32 flags = flag_exist(nlh); 1574 bool use_lineno; 1575 int ret = 0; 1576 1577 if (unlikely(protocol_min_failed(attr) || 1578 !attr[IPSET_ATTR_SETNAME] || 1579 !((attr[IPSET_ATTR_DATA] != NULL) ^ 1580 (attr[IPSET_ATTR_ADT] != NULL)) || 1581 (attr[IPSET_ATTR_DATA] && 1582 !flag_nested(attr[IPSET_ATTR_DATA])) || 1583 (attr[IPSET_ATTR_ADT] && 1584 (!flag_nested(attr[IPSET_ATTR_ADT]) || 1585 !attr[IPSET_ATTR_LINENO])))) 1586 return -IPSET_ERR_PROTOCOL; 1587 1588 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME])); 1589 if (!set) 1590 return -ENOENT; 1591 1592 use_lineno = !!attr[IPSET_ATTR_LINENO]; 1593 if (attr[IPSET_ATTR_DATA]) { 1594 if (nla_parse_nested_deprecated(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA], set->type->adt_policy, NULL)) 1595 return -IPSET_ERR_PROTOCOL; 1596 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD, flags, 1597 use_lineno); 1598 } else { 1599 int nla_rem; 1600 1601 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) { 1602 memset(tb, 0, sizeof(tb)); 1603 if (nla_type(nla) != IPSET_ATTR_DATA || 1604 !flag_nested(nla) || 1605 nla_parse_nested_deprecated(tb, IPSET_ATTR_ADT_MAX, nla, set->type->adt_policy, NULL)) 1606 return -IPSET_ERR_PROTOCOL; 1607 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD, 1608 flags, use_lineno); 1609 if (ret < 0) 1610 return ret; 1611 } 1612 } 1613 return ret; 1614 } 1615 1616 static int ip_set_udel(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1617 const struct nlmsghdr *nlh, 1618 const struct nlattr * const attr[], 1619 struct netlink_ext_ack *extack) 1620 { 1621 struct ip_set_net *inst = ip_set_pernet(net); 1622 struct ip_set *set; 1623 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {}; 1624 const struct nlattr *nla; 1625 u32 flags = flag_exist(nlh); 1626 bool use_lineno; 1627 int ret = 0; 1628 1629 if (unlikely(protocol_min_failed(attr) || 1630 !attr[IPSET_ATTR_SETNAME] || 1631 !((attr[IPSET_ATTR_DATA] != NULL) ^ 1632 (attr[IPSET_ATTR_ADT] != NULL)) || 1633 (attr[IPSET_ATTR_DATA] && 1634 !flag_nested(attr[IPSET_ATTR_DATA])) || 1635 (attr[IPSET_ATTR_ADT] && 1636 (!flag_nested(attr[IPSET_ATTR_ADT]) || 1637 !attr[IPSET_ATTR_LINENO])))) 1638 return -IPSET_ERR_PROTOCOL; 1639 1640 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME])); 1641 if (!set) 1642 return -ENOENT; 1643 1644 use_lineno = !!attr[IPSET_ATTR_LINENO]; 1645 if (attr[IPSET_ATTR_DATA]) { 1646 if (nla_parse_nested_deprecated(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA], set->type->adt_policy, NULL)) 1647 return -IPSET_ERR_PROTOCOL; 1648 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL, flags, 1649 use_lineno); 1650 } else { 1651 int nla_rem; 1652 1653 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) { 1654 memset(tb, 0, sizeof(*tb)); 1655 if (nla_type(nla) != IPSET_ATTR_DATA || 1656 !flag_nested(nla) || 1657 nla_parse_nested_deprecated(tb, IPSET_ATTR_ADT_MAX, nla, set->type->adt_policy, NULL)) 1658 return -IPSET_ERR_PROTOCOL; 1659 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL, 1660 flags, use_lineno); 1661 if (ret < 0) 1662 return ret; 1663 } 1664 } 1665 return ret; 1666 } 1667 1668 static int ip_set_utest(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1669 const struct nlmsghdr *nlh, 1670 const struct nlattr * const attr[], 1671 struct netlink_ext_ack *extack) 1672 { 1673 struct ip_set_net *inst = ip_set_pernet(net); 1674 struct ip_set *set; 1675 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {}; 1676 int ret = 0; 1677 1678 if (unlikely(protocol_min_failed(attr) || 1679 !attr[IPSET_ATTR_SETNAME] || 1680 !attr[IPSET_ATTR_DATA] || 1681 !flag_nested(attr[IPSET_ATTR_DATA]))) 1682 return -IPSET_ERR_PROTOCOL; 1683 1684 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME])); 1685 if (!set) 1686 return -ENOENT; 1687 1688 if (nla_parse_nested_deprecated(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA], set->type->adt_policy, NULL)) 1689 return -IPSET_ERR_PROTOCOL; 1690 1691 rcu_read_lock_bh(); 1692 ret = set->variant->uadt(set, tb, IPSET_TEST, NULL, 0, 0); 1693 rcu_read_unlock_bh(); 1694 /* Userspace can't trigger element to be re-added */ 1695 if (ret == -EAGAIN) 1696 ret = 1; 1697 1698 return ret > 0 ? 0 : -IPSET_ERR_EXIST; 1699 } 1700 1701 /* Get headed data of a set */ 1702 1703 static int ip_set_header(struct net *net, struct sock *ctnl, 1704 struct sk_buff *skb, const struct nlmsghdr *nlh, 1705 const struct nlattr * const attr[], 1706 struct netlink_ext_ack *extack) 1707 { 1708 struct ip_set_net *inst = ip_set_pernet(net); 1709 const struct ip_set *set; 1710 struct sk_buff *skb2; 1711 struct nlmsghdr *nlh2; 1712 int ret = 0; 1713 1714 if (unlikely(protocol_min_failed(attr) || 1715 !attr[IPSET_ATTR_SETNAME])) 1716 return -IPSET_ERR_PROTOCOL; 1717 1718 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME])); 1719 if (!set) 1720 return -ENOENT; 1721 1722 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 1723 if (!skb2) 1724 return -ENOMEM; 1725 1726 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 1727 IPSET_CMD_HEADER); 1728 if (!nlh2) 1729 goto nlmsg_failure; 1730 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) || 1731 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) || 1732 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) || 1733 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) || 1734 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision)) 1735 goto nla_put_failure; 1736 nlmsg_end(skb2, nlh2); 1737 1738 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT); 1739 if (ret < 0) 1740 return ret; 1741 1742 return 0; 1743 1744 nla_put_failure: 1745 nlmsg_cancel(skb2, nlh2); 1746 nlmsg_failure: 1747 kfree_skb(skb2); 1748 return -EMSGSIZE; 1749 } 1750 1751 /* Get type data */ 1752 1753 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = { 1754 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1755 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING, 1756 .len = IPSET_MAXNAMELEN - 1 }, 1757 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 }, 1758 }; 1759 1760 static int ip_set_type(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1761 const struct nlmsghdr *nlh, 1762 const struct nlattr * const attr[], 1763 struct netlink_ext_ack *extack) 1764 { 1765 struct sk_buff *skb2; 1766 struct nlmsghdr *nlh2; 1767 u8 family, min, max; 1768 const char *typename; 1769 int ret = 0; 1770 1771 if (unlikely(protocol_min_failed(attr) || 1772 !attr[IPSET_ATTR_TYPENAME] || 1773 !attr[IPSET_ATTR_FAMILY])) 1774 return -IPSET_ERR_PROTOCOL; 1775 1776 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]); 1777 typename = nla_data(attr[IPSET_ATTR_TYPENAME]); 1778 ret = find_set_type_minmax(typename, family, &min, &max); 1779 if (ret) 1780 return ret; 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_TYPE); 1788 if (!nlh2) 1789 goto nlmsg_failure; 1790 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) || 1791 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) || 1792 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) || 1793 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) || 1794 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min)) 1795 goto nla_put_failure; 1796 nlmsg_end(skb2, nlh2); 1797 1798 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len); 1799 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT); 1800 if (ret < 0) 1801 return ret; 1802 1803 return 0; 1804 1805 nla_put_failure: 1806 nlmsg_cancel(skb2, nlh2); 1807 nlmsg_failure: 1808 kfree_skb(skb2); 1809 return -EMSGSIZE; 1810 } 1811 1812 /* Get protocol version */ 1813 1814 static const struct nla_policy 1815 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = { 1816 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1817 }; 1818 1819 static int ip_set_protocol(struct net *net, struct sock *ctnl, 1820 struct sk_buff *skb, const struct nlmsghdr *nlh, 1821 const struct nlattr * const attr[], 1822 struct netlink_ext_ack *extack) 1823 { 1824 struct sk_buff *skb2; 1825 struct nlmsghdr *nlh2; 1826 int ret = 0; 1827 1828 if (unlikely(!attr[IPSET_ATTR_PROTOCOL])) 1829 return -IPSET_ERR_PROTOCOL; 1830 1831 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 1832 if (!skb2) 1833 return -ENOMEM; 1834 1835 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 1836 IPSET_CMD_PROTOCOL); 1837 if (!nlh2) 1838 goto nlmsg_failure; 1839 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL)) 1840 goto nla_put_failure; 1841 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL_MIN, IPSET_PROTOCOL_MIN)) 1842 goto nla_put_failure; 1843 nlmsg_end(skb2, nlh2); 1844 1845 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT); 1846 if (ret < 0) 1847 return ret; 1848 1849 return 0; 1850 1851 nla_put_failure: 1852 nlmsg_cancel(skb2, nlh2); 1853 nlmsg_failure: 1854 kfree_skb(skb2); 1855 return -EMSGSIZE; 1856 } 1857 1858 /* Get set by name or index, from userspace */ 1859 1860 static int ip_set_byname(struct net *net, struct sock *ctnl, 1861 struct sk_buff *skb, const struct nlmsghdr *nlh, 1862 const struct nlattr * const attr[], 1863 struct netlink_ext_ack *extack) 1864 { 1865 struct ip_set_net *inst = ip_set_pernet(net); 1866 struct sk_buff *skb2; 1867 struct nlmsghdr *nlh2; 1868 ip_set_id_t id = IPSET_INVALID_ID; 1869 const struct ip_set *set; 1870 int ret = 0; 1871 1872 if (unlikely(protocol_failed(attr) || 1873 !attr[IPSET_ATTR_SETNAME])) 1874 return -IPSET_ERR_PROTOCOL; 1875 1876 set = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]), &id); 1877 if (id == IPSET_INVALID_ID) 1878 return -ENOENT; 1879 1880 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 1881 if (!skb2) 1882 return -ENOMEM; 1883 1884 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 1885 IPSET_CMD_GET_BYNAME); 1886 if (!nlh2) 1887 goto nlmsg_failure; 1888 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) || 1889 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) || 1890 nla_put_net16(skb2, IPSET_ATTR_INDEX, htons(id))) 1891 goto nla_put_failure; 1892 nlmsg_end(skb2, nlh2); 1893 1894 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT); 1895 if (ret < 0) 1896 return ret; 1897 1898 return 0; 1899 1900 nla_put_failure: 1901 nlmsg_cancel(skb2, nlh2); 1902 nlmsg_failure: 1903 kfree_skb(skb2); 1904 return -EMSGSIZE; 1905 } 1906 1907 static const struct nla_policy ip_set_index_policy[IPSET_ATTR_CMD_MAX + 1] = { 1908 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1909 [IPSET_ATTR_INDEX] = { .type = NLA_U16 }, 1910 }; 1911 1912 static int ip_set_byindex(struct net *net, struct sock *ctnl, 1913 struct sk_buff *skb, const struct nlmsghdr *nlh, 1914 const struct nlattr * const attr[], 1915 struct netlink_ext_ack *extack) 1916 { 1917 struct ip_set_net *inst = ip_set_pernet(net); 1918 struct sk_buff *skb2; 1919 struct nlmsghdr *nlh2; 1920 ip_set_id_t id = IPSET_INVALID_ID; 1921 const struct ip_set *set; 1922 int ret = 0; 1923 1924 if (unlikely(protocol_failed(attr) || 1925 !attr[IPSET_ATTR_INDEX])) 1926 return -IPSET_ERR_PROTOCOL; 1927 1928 id = ip_set_get_h16(attr[IPSET_ATTR_INDEX]); 1929 if (id >= inst->ip_set_max) 1930 return -ENOENT; 1931 set = ip_set(inst, id); 1932 if (set == NULL) 1933 return -ENOENT; 1934 1935 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 1936 if (!skb2) 1937 return -ENOMEM; 1938 1939 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 1940 IPSET_CMD_GET_BYINDEX); 1941 if (!nlh2) 1942 goto nlmsg_failure; 1943 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) || 1944 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name)) 1945 goto nla_put_failure; 1946 nlmsg_end(skb2, nlh2); 1947 1948 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT); 1949 if (ret < 0) 1950 return ret; 1951 1952 return 0; 1953 1954 nla_put_failure: 1955 nlmsg_cancel(skb2, nlh2); 1956 nlmsg_failure: 1957 kfree_skb(skb2); 1958 return -EMSGSIZE; 1959 } 1960 1961 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = { 1962 [IPSET_CMD_NONE] = { 1963 .call = ip_set_none, 1964 .attr_count = IPSET_ATTR_CMD_MAX, 1965 }, 1966 [IPSET_CMD_CREATE] = { 1967 .call = ip_set_create, 1968 .attr_count = IPSET_ATTR_CMD_MAX, 1969 .policy = ip_set_create_policy, 1970 }, 1971 [IPSET_CMD_DESTROY] = { 1972 .call = ip_set_destroy, 1973 .attr_count = IPSET_ATTR_CMD_MAX, 1974 .policy = ip_set_setname_policy, 1975 }, 1976 [IPSET_CMD_FLUSH] = { 1977 .call = ip_set_flush, 1978 .attr_count = IPSET_ATTR_CMD_MAX, 1979 .policy = ip_set_setname_policy, 1980 }, 1981 [IPSET_CMD_RENAME] = { 1982 .call = ip_set_rename, 1983 .attr_count = IPSET_ATTR_CMD_MAX, 1984 .policy = ip_set_setname2_policy, 1985 }, 1986 [IPSET_CMD_SWAP] = { 1987 .call = ip_set_swap, 1988 .attr_count = IPSET_ATTR_CMD_MAX, 1989 .policy = ip_set_setname2_policy, 1990 }, 1991 [IPSET_CMD_LIST] = { 1992 .call = ip_set_dump, 1993 .attr_count = IPSET_ATTR_CMD_MAX, 1994 .policy = ip_set_setname_policy, 1995 }, 1996 [IPSET_CMD_SAVE] = { 1997 .call = ip_set_dump, 1998 .attr_count = IPSET_ATTR_CMD_MAX, 1999 .policy = ip_set_setname_policy, 2000 }, 2001 [IPSET_CMD_ADD] = { 2002 .call = ip_set_uadd, 2003 .attr_count = IPSET_ATTR_CMD_MAX, 2004 .policy = ip_set_adt_policy, 2005 }, 2006 [IPSET_CMD_DEL] = { 2007 .call = ip_set_udel, 2008 .attr_count = IPSET_ATTR_CMD_MAX, 2009 .policy = ip_set_adt_policy, 2010 }, 2011 [IPSET_CMD_TEST] = { 2012 .call = ip_set_utest, 2013 .attr_count = IPSET_ATTR_CMD_MAX, 2014 .policy = ip_set_adt_policy, 2015 }, 2016 [IPSET_CMD_HEADER] = { 2017 .call = ip_set_header, 2018 .attr_count = IPSET_ATTR_CMD_MAX, 2019 .policy = ip_set_setname_policy, 2020 }, 2021 [IPSET_CMD_TYPE] = { 2022 .call = ip_set_type, 2023 .attr_count = IPSET_ATTR_CMD_MAX, 2024 .policy = ip_set_type_policy, 2025 }, 2026 [IPSET_CMD_PROTOCOL] = { 2027 .call = ip_set_protocol, 2028 .attr_count = IPSET_ATTR_CMD_MAX, 2029 .policy = ip_set_protocol_policy, 2030 }, 2031 [IPSET_CMD_GET_BYNAME] = { 2032 .call = ip_set_byname, 2033 .attr_count = IPSET_ATTR_CMD_MAX, 2034 .policy = ip_set_setname_policy, 2035 }, 2036 [IPSET_CMD_GET_BYINDEX] = { 2037 .call = ip_set_byindex, 2038 .attr_count = IPSET_ATTR_CMD_MAX, 2039 .policy = ip_set_index_policy, 2040 }, 2041 }; 2042 2043 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = { 2044 .name = "ip_set", 2045 .subsys_id = NFNL_SUBSYS_IPSET, 2046 .cb_count = IPSET_MSG_MAX, 2047 .cb = ip_set_netlink_subsys_cb, 2048 }; 2049 2050 /* Interface to iptables/ip6tables */ 2051 2052 static int 2053 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len) 2054 { 2055 unsigned int *op; 2056 void *data; 2057 int copylen = *len, ret = 0; 2058 struct net *net = sock_net(sk); 2059 struct ip_set_net *inst = ip_set_pernet(net); 2060 2061 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 2062 return -EPERM; 2063 if (optval != SO_IP_SET) 2064 return -EBADF; 2065 if (*len < sizeof(unsigned int)) 2066 return -EINVAL; 2067 2068 data = vmalloc(*len); 2069 if (!data) 2070 return -ENOMEM; 2071 if (copy_from_user(data, user, *len) != 0) { 2072 ret = -EFAULT; 2073 goto done; 2074 } 2075 op = data; 2076 2077 if (*op < IP_SET_OP_VERSION) { 2078 /* Check the version at the beginning of operations */ 2079 struct ip_set_req_version *req_version = data; 2080 2081 if (*len < sizeof(struct ip_set_req_version)) { 2082 ret = -EINVAL; 2083 goto done; 2084 } 2085 2086 if (req_version->version < IPSET_PROTOCOL_MIN) { 2087 ret = -EPROTO; 2088 goto done; 2089 } 2090 } 2091 2092 switch (*op) { 2093 case IP_SET_OP_VERSION: { 2094 struct ip_set_req_version *req_version = data; 2095 2096 if (*len != sizeof(struct ip_set_req_version)) { 2097 ret = -EINVAL; 2098 goto done; 2099 } 2100 2101 req_version->version = IPSET_PROTOCOL; 2102 ret = copy_to_user(user, req_version, 2103 sizeof(struct ip_set_req_version)); 2104 goto done; 2105 } 2106 case IP_SET_OP_GET_BYNAME: { 2107 struct ip_set_req_get_set *req_get = data; 2108 ip_set_id_t id; 2109 2110 if (*len != sizeof(struct ip_set_req_get_set)) { 2111 ret = -EINVAL; 2112 goto done; 2113 } 2114 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0'; 2115 nfnl_lock(NFNL_SUBSYS_IPSET); 2116 find_set_and_id(inst, req_get->set.name, &id); 2117 req_get->set.index = id; 2118 nfnl_unlock(NFNL_SUBSYS_IPSET); 2119 goto copy; 2120 } 2121 case IP_SET_OP_GET_FNAME: { 2122 struct ip_set_req_get_set_family *req_get = data; 2123 ip_set_id_t id; 2124 2125 if (*len != sizeof(struct ip_set_req_get_set_family)) { 2126 ret = -EINVAL; 2127 goto done; 2128 } 2129 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0'; 2130 nfnl_lock(NFNL_SUBSYS_IPSET); 2131 find_set_and_id(inst, req_get->set.name, &id); 2132 req_get->set.index = id; 2133 if (id != IPSET_INVALID_ID) 2134 req_get->family = ip_set(inst, id)->family; 2135 nfnl_unlock(NFNL_SUBSYS_IPSET); 2136 goto copy; 2137 } 2138 case IP_SET_OP_GET_BYINDEX: { 2139 struct ip_set_req_get_set *req_get = data; 2140 struct ip_set *set; 2141 2142 if (*len != sizeof(struct ip_set_req_get_set) || 2143 req_get->set.index >= inst->ip_set_max) { 2144 ret = -EINVAL; 2145 goto done; 2146 } 2147 nfnl_lock(NFNL_SUBSYS_IPSET); 2148 set = ip_set(inst, req_get->set.index); 2149 ret = strscpy(req_get->set.name, set ? set->name : "", 2150 IPSET_MAXNAMELEN); 2151 nfnl_unlock(NFNL_SUBSYS_IPSET); 2152 if (ret < 0) 2153 goto done; 2154 goto copy; 2155 } 2156 default: 2157 ret = -EBADMSG; 2158 goto done; 2159 } /* end of switch(op) */ 2160 2161 copy: 2162 ret = copy_to_user(user, data, copylen); 2163 2164 done: 2165 vfree(data); 2166 if (ret > 0) 2167 ret = 0; 2168 return ret; 2169 } 2170 2171 static struct nf_sockopt_ops so_set __read_mostly = { 2172 .pf = PF_INET, 2173 .get_optmin = SO_IP_SET, 2174 .get_optmax = SO_IP_SET + 1, 2175 .get = ip_set_sockfn_get, 2176 .owner = THIS_MODULE, 2177 }; 2178 2179 static int __net_init 2180 ip_set_net_init(struct net *net) 2181 { 2182 struct ip_set_net *inst = ip_set_pernet(net); 2183 struct ip_set **list; 2184 2185 inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX; 2186 if (inst->ip_set_max >= IPSET_INVALID_ID) 2187 inst->ip_set_max = IPSET_INVALID_ID - 1; 2188 2189 list = kvcalloc(inst->ip_set_max, sizeof(struct ip_set *), GFP_KERNEL); 2190 if (!list) 2191 return -ENOMEM; 2192 inst->is_deleted = false; 2193 inst->is_destroyed = false; 2194 rcu_assign_pointer(inst->ip_set_list, list); 2195 return 0; 2196 } 2197 2198 static void __net_exit 2199 ip_set_net_exit(struct net *net) 2200 { 2201 struct ip_set_net *inst = ip_set_pernet(net); 2202 2203 struct ip_set *set = NULL; 2204 ip_set_id_t i; 2205 2206 inst->is_deleted = true; /* flag for ip_set_nfnl_put */ 2207 2208 nfnl_lock(NFNL_SUBSYS_IPSET); 2209 for (i = 0; i < inst->ip_set_max; i++) { 2210 set = ip_set(inst, i); 2211 if (set) { 2212 ip_set(inst, i) = NULL; 2213 ip_set_destroy_set(set); 2214 } 2215 } 2216 nfnl_unlock(NFNL_SUBSYS_IPSET); 2217 kvfree(rcu_dereference_protected(inst->ip_set_list, 1)); 2218 } 2219 2220 static struct pernet_operations ip_set_net_ops = { 2221 .init = ip_set_net_init, 2222 .exit = ip_set_net_exit, 2223 .id = &ip_set_net_id, 2224 .size = sizeof(struct ip_set_net), 2225 }; 2226 2227 static int __init 2228 ip_set_init(void) 2229 { 2230 int ret = register_pernet_subsys(&ip_set_net_ops); 2231 2232 if (ret) { 2233 pr_err("ip_set: cannot register pernet_subsys.\n"); 2234 return ret; 2235 } 2236 2237 ret = nfnetlink_subsys_register(&ip_set_netlink_subsys); 2238 if (ret != 0) { 2239 pr_err("ip_set: cannot register with nfnetlink.\n"); 2240 unregister_pernet_subsys(&ip_set_net_ops); 2241 return ret; 2242 } 2243 2244 ret = nf_register_sockopt(&so_set); 2245 if (ret != 0) { 2246 pr_err("SO_SET registry failed: %d\n", ret); 2247 nfnetlink_subsys_unregister(&ip_set_netlink_subsys); 2248 unregister_pernet_subsys(&ip_set_net_ops); 2249 return ret; 2250 } 2251 2252 return 0; 2253 } 2254 2255 static void __exit 2256 ip_set_fini(void) 2257 { 2258 nf_unregister_sockopt(&so_set); 2259 nfnetlink_subsys_unregister(&ip_set_netlink_subsys); 2260 2261 unregister_pernet_subsys(&ip_set_net_ops); 2262 pr_debug("these are the famous last words\n"); 2263 } 2264 2265 module_init(ip_set_init); 2266 module_exit(ip_set_fini); 2267 2268 MODULE_DESCRIPTION("ip_set: protocol " __stringify(IPSET_PROTOCOL)); 2269