1 // SPDX-License-Identifier: GPL-2.0-only 2 /* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu> 3 * Patrick Schaaf <bof@bof.de> 4 * Copyright (C) 2003-2013 Jozsef Kadlecsik <kadlec@netfilter.org> 5 */ 6 7 /* Kernel module for IP set management */ 8 9 #include <linux/init.h> 10 #include <linux/module.h> 11 #include <linux/moduleparam.h> 12 #include <linux/ip.h> 13 #include <linux/skbuff.h> 14 #include <linux/spinlock.h> 15 #include <linux/rculist.h> 16 #include <net/netlink.h> 17 #include <net/net_namespace.h> 18 #include <net/netns/generic.h> 19 20 #include <linux/netfilter.h> 21 #include <linux/netfilter/x_tables.h> 22 #include <linux/netfilter/nfnetlink.h> 23 #include <linux/netfilter/ipset/ip_set.h> 24 25 static LIST_HEAD(ip_set_type_list); /* all registered set types */ 26 static DEFINE_MUTEX(ip_set_type_mutex); /* protects ip_set_type_list */ 27 static DEFINE_RWLOCK(ip_set_ref_lock); /* protects the set refs */ 28 29 struct ip_set_net { 30 struct ip_set * __rcu *ip_set_list; /* all individual sets */ 31 ip_set_id_t ip_set_max; /* max number of sets */ 32 bool is_deleted; /* deleted by ip_set_net_exit */ 33 bool is_destroyed; /* all sets are destroyed */ 34 }; 35 36 static unsigned int ip_set_net_id __read_mostly; 37 38 static struct ip_set_net *ip_set_pernet(struct net *net) 39 { 40 return net_generic(net, ip_set_net_id); 41 } 42 43 #define IP_SET_INC 64 44 #define STRNCMP(a, b) (strncmp(a, b, IPSET_MAXNAMELEN) == 0) 45 46 static unsigned int max_sets; 47 48 module_param(max_sets, int, 0600); 49 MODULE_PARM_DESC(max_sets, "maximal number of sets"); 50 MODULE_LICENSE("GPL"); 51 MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@netfilter.org>"); 52 MODULE_DESCRIPTION("core IP set support"); 53 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_IPSET); 54 55 /* When the nfnl mutex or ip_set_ref_lock is held: */ 56 #define ip_set_dereference(p) \ 57 rcu_dereference_protected(p, \ 58 lockdep_nfnl_is_held(NFNL_SUBSYS_IPSET) || \ 59 lockdep_is_held(&ip_set_ref_lock)) 60 #define ip_set(inst, id) \ 61 ip_set_dereference((inst)->ip_set_list)[id] 62 #define ip_set_ref_netlink(inst,id) \ 63 rcu_dereference_raw((inst)->ip_set_list)[id] 64 65 /* The set types are implemented in modules and registered set types 66 * can be found in ip_set_type_list. Adding/deleting types is 67 * serialized by ip_set_type_mutex. 68 */ 69 70 static void 71 ip_set_type_lock(void) 72 { 73 mutex_lock(&ip_set_type_mutex); 74 } 75 76 static void 77 ip_set_type_unlock(void) 78 { 79 mutex_unlock(&ip_set_type_mutex); 80 } 81 82 /* Register and deregister settype */ 83 84 static struct ip_set_type * 85 find_set_type(const char *name, u8 family, u8 revision) 86 { 87 struct ip_set_type *type; 88 89 list_for_each_entry_rcu(type, &ip_set_type_list, list) 90 if (STRNCMP(type->name, name) && 91 (type->family == family || 92 type->family == NFPROTO_UNSPEC) && 93 revision >= type->revision_min && 94 revision <= type->revision_max) 95 return type; 96 return NULL; 97 } 98 99 /* Unlock, try to load a set type module and lock again */ 100 static bool 101 load_settype(const char *name) 102 { 103 nfnl_unlock(NFNL_SUBSYS_IPSET); 104 pr_debug("try to load ip_set_%s\n", name); 105 if (request_module("ip_set_%s", name) < 0) { 106 pr_warn("Can't find ip_set type %s\n", name); 107 nfnl_lock(NFNL_SUBSYS_IPSET); 108 return false; 109 } 110 nfnl_lock(NFNL_SUBSYS_IPSET); 111 return true; 112 } 113 114 /* Find a set type and reference it */ 115 #define find_set_type_get(name, family, revision, found) \ 116 __find_set_type_get(name, family, revision, found, false) 117 118 static int 119 __find_set_type_get(const char *name, u8 family, u8 revision, 120 struct ip_set_type **found, bool retry) 121 { 122 struct ip_set_type *type; 123 int err; 124 125 if (retry && !load_settype(name)) 126 return -IPSET_ERR_FIND_TYPE; 127 128 rcu_read_lock(); 129 *found = find_set_type(name, family, revision); 130 if (*found) { 131 err = !try_module_get((*found)->me) ? -EFAULT : 0; 132 goto unlock; 133 } 134 /* Make sure the type is already loaded 135 * but we don't support the revision 136 */ 137 list_for_each_entry_rcu(type, &ip_set_type_list, list) 138 if (STRNCMP(type->name, name)) { 139 err = -IPSET_ERR_FIND_TYPE; 140 goto unlock; 141 } 142 rcu_read_unlock(); 143 144 return retry ? -IPSET_ERR_FIND_TYPE : 145 __find_set_type_get(name, family, revision, found, true); 146 147 unlock: 148 rcu_read_unlock(); 149 return err; 150 } 151 152 /* Find a given set type by name and family. 153 * If we succeeded, the supported minimal and maximum revisions are 154 * filled out. 155 */ 156 #define find_set_type_minmax(name, family, min, max) \ 157 __find_set_type_minmax(name, family, min, max, false) 158 159 static int 160 __find_set_type_minmax(const char *name, u8 family, u8 *min, u8 *max, 161 bool retry) 162 { 163 struct ip_set_type *type; 164 bool found = false; 165 166 if (retry && !load_settype(name)) 167 return -IPSET_ERR_FIND_TYPE; 168 169 *min = 255; *max = 0; 170 rcu_read_lock(); 171 list_for_each_entry_rcu(type, &ip_set_type_list, list) 172 if (STRNCMP(type->name, name) && 173 (type->family == family || 174 type->family == NFPROTO_UNSPEC)) { 175 found = true; 176 if (type->revision_min < *min) 177 *min = type->revision_min; 178 if (type->revision_max > *max) 179 *max = type->revision_max; 180 } 181 rcu_read_unlock(); 182 if (found) 183 return 0; 184 185 return retry ? -IPSET_ERR_FIND_TYPE : 186 __find_set_type_minmax(name, family, min, max, true); 187 } 188 189 #define family_name(f) ((f) == NFPROTO_IPV4 ? "inet" : \ 190 (f) == NFPROTO_IPV6 ? "inet6" : "any") 191 192 /* Register a set type structure. The type is identified by 193 * the unique triple of name, family and revision. 194 */ 195 int 196 ip_set_type_register(struct ip_set_type *type) 197 { 198 int ret = 0; 199 200 if (type->protocol != IPSET_PROTOCOL) { 201 pr_warn("ip_set type %s, family %s, revision %u:%u uses wrong protocol version %u (want %u)\n", 202 type->name, family_name(type->family), 203 type->revision_min, type->revision_max, 204 type->protocol, IPSET_PROTOCOL); 205 return -EINVAL; 206 } 207 208 ip_set_type_lock(); 209 if (find_set_type(type->name, type->family, type->revision_min)) { 210 /* Duplicate! */ 211 pr_warn("ip_set type %s, family %s with revision min %u already registered!\n", 212 type->name, family_name(type->family), 213 type->revision_min); 214 ip_set_type_unlock(); 215 return -EINVAL; 216 } 217 list_add_rcu(&type->list, &ip_set_type_list); 218 pr_debug("type %s, family %s, revision %u:%u registered.\n", 219 type->name, family_name(type->family), 220 type->revision_min, type->revision_max); 221 ip_set_type_unlock(); 222 223 return ret; 224 } 225 EXPORT_SYMBOL_GPL(ip_set_type_register); 226 227 /* Unregister a set type. There's a small race with ip_set_create */ 228 void 229 ip_set_type_unregister(struct ip_set_type *type) 230 { 231 ip_set_type_lock(); 232 if (!find_set_type(type->name, type->family, type->revision_min)) { 233 pr_warn("ip_set type %s, family %s with revision min %u not registered\n", 234 type->name, family_name(type->family), 235 type->revision_min); 236 ip_set_type_unlock(); 237 return; 238 } 239 list_del_rcu(&type->list); 240 pr_debug("type %s, family %s with revision min %u unregistered.\n", 241 type->name, family_name(type->family), type->revision_min); 242 ip_set_type_unlock(); 243 244 synchronize_rcu(); 245 } 246 EXPORT_SYMBOL_GPL(ip_set_type_unregister); 247 248 /* Utility functions */ 249 void * 250 ip_set_alloc(size_t size) 251 { 252 void *members = NULL; 253 254 if (size < KMALLOC_MAX_SIZE) 255 members = kzalloc(size, GFP_KERNEL | __GFP_NOWARN); 256 257 if (members) { 258 pr_debug("%p: allocated with kmalloc\n", members); 259 return members; 260 } 261 262 members = vzalloc(size); 263 if (!members) 264 return NULL; 265 pr_debug("%p: allocated with vmalloc\n", members); 266 267 return members; 268 } 269 EXPORT_SYMBOL_GPL(ip_set_alloc); 270 271 void 272 ip_set_free(void *members) 273 { 274 pr_debug("%p: free with %s\n", members, 275 is_vmalloc_addr(members) ? "vfree" : "kfree"); 276 kvfree(members); 277 } 278 EXPORT_SYMBOL_GPL(ip_set_free); 279 280 static bool 281 flag_nested(const struct nlattr *nla) 282 { 283 return nla->nla_type & NLA_F_NESTED; 284 } 285 286 static const struct nla_policy ipaddr_policy[IPSET_ATTR_IPADDR_MAX + 1] = { 287 [IPSET_ATTR_IPADDR_IPV4] = { .type = NLA_U32 }, 288 [IPSET_ATTR_IPADDR_IPV6] = { .type = NLA_BINARY, 289 .len = sizeof(struct in6_addr) }, 290 }; 291 292 int 293 ip_set_get_ipaddr4(struct nlattr *nla, __be32 *ipaddr) 294 { 295 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1]; 296 297 if (unlikely(!flag_nested(nla))) 298 return -IPSET_ERR_PROTOCOL; 299 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla, 300 ipaddr_policy, NULL)) 301 return -IPSET_ERR_PROTOCOL; 302 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV4))) 303 return -IPSET_ERR_PROTOCOL; 304 305 *ipaddr = nla_get_be32(tb[IPSET_ATTR_IPADDR_IPV4]); 306 return 0; 307 } 308 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr4); 309 310 int 311 ip_set_get_ipaddr6(struct nlattr *nla, union nf_inet_addr *ipaddr) 312 { 313 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1]; 314 315 if (unlikely(!flag_nested(nla))) 316 return -IPSET_ERR_PROTOCOL; 317 318 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla, 319 ipaddr_policy, NULL)) 320 return -IPSET_ERR_PROTOCOL; 321 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV6))) 322 return -IPSET_ERR_PROTOCOL; 323 324 memcpy(ipaddr, nla_data(tb[IPSET_ATTR_IPADDR_IPV6]), 325 sizeof(struct in6_addr)); 326 return 0; 327 } 328 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr6); 329 330 static u32 331 ip_set_timeout_get(const unsigned long *timeout) 332 { 333 u32 t; 334 335 if (*timeout == IPSET_ELEM_PERMANENT) 336 return 0; 337 338 t = jiffies_to_msecs(*timeout - jiffies) / MSEC_PER_SEC; 339 /* Zero value in userspace means no timeout */ 340 return t == 0 ? 1 : t; 341 } 342 343 static char * 344 ip_set_comment_uget(struct nlattr *tb) 345 { 346 return nla_data(tb); 347 } 348 349 /* Called from uadd only, protected by the set spinlock. 350 * The kadt functions don't use the comment extensions in any way. 351 */ 352 void 353 ip_set_init_comment(struct ip_set *set, struct ip_set_comment *comment, 354 const struct ip_set_ext *ext) 355 { 356 struct ip_set_comment_rcu *c = rcu_dereference_protected(comment->c, 1); 357 size_t len = ext->comment ? strlen(ext->comment) : 0; 358 359 if (unlikely(c)) { 360 set->ext_size -= sizeof(*c) + strlen(c->str) + 1; 361 kfree_rcu(c, rcu); 362 rcu_assign_pointer(comment->c, NULL); 363 } 364 if (!len) 365 return; 366 if (unlikely(len > IPSET_MAX_COMMENT_SIZE)) 367 len = IPSET_MAX_COMMENT_SIZE; 368 c = kmalloc(sizeof(*c) + len + 1, GFP_ATOMIC); 369 if (unlikely(!c)) 370 return; 371 strlcpy(c->str, ext->comment, len + 1); 372 set->ext_size += sizeof(*c) + strlen(c->str) + 1; 373 rcu_assign_pointer(comment->c, c); 374 } 375 EXPORT_SYMBOL_GPL(ip_set_init_comment); 376 377 /* Used only when dumping a set, protected by rcu_read_lock() */ 378 static int 379 ip_set_put_comment(struct sk_buff *skb, const struct ip_set_comment *comment) 380 { 381 struct ip_set_comment_rcu *c = rcu_dereference(comment->c); 382 383 if (!c) 384 return 0; 385 return nla_put_string(skb, IPSET_ATTR_COMMENT, c->str); 386 } 387 388 /* Called from uadd/udel, flush or the garbage collectors protected 389 * by the set spinlock. 390 * Called when the set is destroyed and when there can't be any user 391 * of the set data anymore. 392 */ 393 static void 394 ip_set_comment_free(struct ip_set *set, void *ptr) 395 { 396 struct ip_set_comment *comment = ptr; 397 struct ip_set_comment_rcu *c; 398 399 c = rcu_dereference_protected(comment->c, 1); 400 if (unlikely(!c)) 401 return; 402 set->ext_size -= sizeof(*c) + strlen(c->str) + 1; 403 kfree_rcu(c, rcu); 404 rcu_assign_pointer(comment->c, NULL); 405 } 406 407 typedef void (*destroyer)(struct ip_set *, void *); 408 /* ipset data extension types, in size order */ 409 410 const struct ip_set_ext_type ip_set_extensions[] = { 411 [IPSET_EXT_ID_COUNTER] = { 412 .type = IPSET_EXT_COUNTER, 413 .flag = IPSET_FLAG_WITH_COUNTERS, 414 .len = sizeof(struct ip_set_counter), 415 .align = __alignof__(struct ip_set_counter), 416 }, 417 [IPSET_EXT_ID_TIMEOUT] = { 418 .type = IPSET_EXT_TIMEOUT, 419 .len = sizeof(unsigned long), 420 .align = __alignof__(unsigned long), 421 }, 422 [IPSET_EXT_ID_SKBINFO] = { 423 .type = IPSET_EXT_SKBINFO, 424 .flag = IPSET_FLAG_WITH_SKBINFO, 425 .len = sizeof(struct ip_set_skbinfo), 426 .align = __alignof__(struct ip_set_skbinfo), 427 }, 428 [IPSET_EXT_ID_COMMENT] = { 429 .type = IPSET_EXT_COMMENT | IPSET_EXT_DESTROY, 430 .flag = IPSET_FLAG_WITH_COMMENT, 431 .len = sizeof(struct ip_set_comment), 432 .align = __alignof__(struct ip_set_comment), 433 .destroy = ip_set_comment_free, 434 }, 435 }; 436 EXPORT_SYMBOL_GPL(ip_set_extensions); 437 438 static bool 439 add_extension(enum ip_set_ext_id id, u32 flags, struct nlattr *tb[]) 440 { 441 return ip_set_extensions[id].flag ? 442 (flags & ip_set_extensions[id].flag) : 443 !!tb[IPSET_ATTR_TIMEOUT]; 444 } 445 446 size_t 447 ip_set_elem_len(struct ip_set *set, struct nlattr *tb[], size_t len, 448 size_t align) 449 { 450 enum ip_set_ext_id id; 451 u32 cadt_flags = 0; 452 453 if (tb[IPSET_ATTR_CADT_FLAGS]) 454 cadt_flags = ip_set_get_h32(tb[IPSET_ATTR_CADT_FLAGS]); 455 if (cadt_flags & IPSET_FLAG_WITH_FORCEADD) 456 set->flags |= IPSET_CREATE_FLAG_FORCEADD; 457 if (!align) 458 align = 1; 459 for (id = 0; id < IPSET_EXT_ID_MAX; id++) { 460 if (!add_extension(id, cadt_flags, tb)) 461 continue; 462 len = ALIGN(len, ip_set_extensions[id].align); 463 set->offset[id] = len; 464 set->extensions |= ip_set_extensions[id].type; 465 len += ip_set_extensions[id].len; 466 } 467 return ALIGN(len, align); 468 } 469 EXPORT_SYMBOL_GPL(ip_set_elem_len); 470 471 int 472 ip_set_get_extensions(struct ip_set *set, struct nlattr *tb[], 473 struct ip_set_ext *ext) 474 { 475 u64 fullmark; 476 477 if (unlikely(!ip_set_optattr_netorder(tb, IPSET_ATTR_TIMEOUT) || 478 !ip_set_optattr_netorder(tb, IPSET_ATTR_PACKETS) || 479 !ip_set_optattr_netorder(tb, IPSET_ATTR_BYTES) || 480 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBMARK) || 481 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBPRIO) || 482 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBQUEUE))) 483 return -IPSET_ERR_PROTOCOL; 484 485 if (tb[IPSET_ATTR_TIMEOUT]) { 486 if (!SET_WITH_TIMEOUT(set)) 487 return -IPSET_ERR_TIMEOUT; 488 ext->timeout = ip_set_timeout_uget(tb[IPSET_ATTR_TIMEOUT]); 489 } 490 if (tb[IPSET_ATTR_BYTES] || tb[IPSET_ATTR_PACKETS]) { 491 if (!SET_WITH_COUNTER(set)) 492 return -IPSET_ERR_COUNTER; 493 if (tb[IPSET_ATTR_BYTES]) 494 ext->bytes = be64_to_cpu(nla_get_be64( 495 tb[IPSET_ATTR_BYTES])); 496 if (tb[IPSET_ATTR_PACKETS]) 497 ext->packets = be64_to_cpu(nla_get_be64( 498 tb[IPSET_ATTR_PACKETS])); 499 } 500 if (tb[IPSET_ATTR_COMMENT]) { 501 if (!SET_WITH_COMMENT(set)) 502 return -IPSET_ERR_COMMENT; 503 ext->comment = ip_set_comment_uget(tb[IPSET_ATTR_COMMENT]); 504 } 505 if (tb[IPSET_ATTR_SKBMARK]) { 506 if (!SET_WITH_SKBINFO(set)) 507 return -IPSET_ERR_SKBINFO; 508 fullmark = be64_to_cpu(nla_get_be64(tb[IPSET_ATTR_SKBMARK])); 509 ext->skbinfo.skbmark = fullmark >> 32; 510 ext->skbinfo.skbmarkmask = fullmark & 0xffffffff; 511 } 512 if (tb[IPSET_ATTR_SKBPRIO]) { 513 if (!SET_WITH_SKBINFO(set)) 514 return -IPSET_ERR_SKBINFO; 515 ext->skbinfo.skbprio = 516 be32_to_cpu(nla_get_be32(tb[IPSET_ATTR_SKBPRIO])); 517 } 518 if (tb[IPSET_ATTR_SKBQUEUE]) { 519 if (!SET_WITH_SKBINFO(set)) 520 return -IPSET_ERR_SKBINFO; 521 ext->skbinfo.skbqueue = 522 be16_to_cpu(nla_get_be16(tb[IPSET_ATTR_SKBQUEUE])); 523 } 524 return 0; 525 } 526 EXPORT_SYMBOL_GPL(ip_set_get_extensions); 527 528 static u64 529 ip_set_get_bytes(const struct ip_set_counter *counter) 530 { 531 return (u64)atomic64_read(&(counter)->bytes); 532 } 533 534 static u64 535 ip_set_get_packets(const struct ip_set_counter *counter) 536 { 537 return (u64)atomic64_read(&(counter)->packets); 538 } 539 540 static bool 541 ip_set_put_counter(struct sk_buff *skb, const struct ip_set_counter *counter) 542 { 543 return nla_put_net64(skb, IPSET_ATTR_BYTES, 544 cpu_to_be64(ip_set_get_bytes(counter)), 545 IPSET_ATTR_PAD) || 546 nla_put_net64(skb, IPSET_ATTR_PACKETS, 547 cpu_to_be64(ip_set_get_packets(counter)), 548 IPSET_ATTR_PAD); 549 } 550 551 static bool 552 ip_set_put_skbinfo(struct sk_buff *skb, const struct ip_set_skbinfo *skbinfo) 553 { 554 /* Send nonzero parameters only */ 555 return ((skbinfo->skbmark || skbinfo->skbmarkmask) && 556 nla_put_net64(skb, IPSET_ATTR_SKBMARK, 557 cpu_to_be64((u64)skbinfo->skbmark << 32 | 558 skbinfo->skbmarkmask), 559 IPSET_ATTR_PAD)) || 560 (skbinfo->skbprio && 561 nla_put_net32(skb, IPSET_ATTR_SKBPRIO, 562 cpu_to_be32(skbinfo->skbprio))) || 563 (skbinfo->skbqueue && 564 nla_put_net16(skb, IPSET_ATTR_SKBQUEUE, 565 cpu_to_be16(skbinfo->skbqueue))); 566 } 567 568 int 569 ip_set_put_extensions(struct sk_buff *skb, const struct ip_set *set, 570 const void *e, bool active) 571 { 572 if (SET_WITH_TIMEOUT(set)) { 573 unsigned long *timeout = ext_timeout(e, set); 574 575 if (nla_put_net32(skb, IPSET_ATTR_TIMEOUT, 576 htonl(active ? ip_set_timeout_get(timeout) 577 : *timeout))) 578 return -EMSGSIZE; 579 } 580 if (SET_WITH_COUNTER(set) && 581 ip_set_put_counter(skb, ext_counter(e, set))) 582 return -EMSGSIZE; 583 if (SET_WITH_COMMENT(set) && 584 ip_set_put_comment(skb, ext_comment(e, set))) 585 return -EMSGSIZE; 586 if (SET_WITH_SKBINFO(set) && 587 ip_set_put_skbinfo(skb, ext_skbinfo(e, set))) 588 return -EMSGSIZE; 589 return 0; 590 } 591 EXPORT_SYMBOL_GPL(ip_set_put_extensions); 592 593 static bool 594 ip_set_match_counter(u64 counter, u64 match, u8 op) 595 { 596 switch (op) { 597 case IPSET_COUNTER_NONE: 598 return true; 599 case IPSET_COUNTER_EQ: 600 return counter == match; 601 case IPSET_COUNTER_NE: 602 return counter != match; 603 case IPSET_COUNTER_LT: 604 return counter < match; 605 case IPSET_COUNTER_GT: 606 return counter > match; 607 } 608 return false; 609 } 610 611 static void 612 ip_set_add_bytes(u64 bytes, struct ip_set_counter *counter) 613 { 614 atomic64_add((long long)bytes, &(counter)->bytes); 615 } 616 617 static void 618 ip_set_add_packets(u64 packets, struct ip_set_counter *counter) 619 { 620 atomic64_add((long long)packets, &(counter)->packets); 621 } 622 623 static void 624 ip_set_update_counter(struct ip_set_counter *counter, 625 const struct ip_set_ext *ext, u32 flags) 626 { 627 if (ext->packets != ULLONG_MAX && 628 !(flags & IPSET_FLAG_SKIP_COUNTER_UPDATE)) { 629 ip_set_add_bytes(ext->bytes, counter); 630 ip_set_add_packets(ext->packets, counter); 631 } 632 } 633 634 static void 635 ip_set_get_skbinfo(struct ip_set_skbinfo *skbinfo, 636 const struct ip_set_ext *ext, 637 struct ip_set_ext *mext, u32 flags) 638 { 639 mext->skbinfo = *skbinfo; 640 } 641 642 bool 643 ip_set_match_extensions(struct ip_set *set, const struct ip_set_ext *ext, 644 struct ip_set_ext *mext, u32 flags, void *data) 645 { 646 if (SET_WITH_TIMEOUT(set) && 647 ip_set_timeout_expired(ext_timeout(data, set))) 648 return false; 649 if (SET_WITH_COUNTER(set)) { 650 struct ip_set_counter *counter = ext_counter(data, set); 651 652 if (flags & IPSET_FLAG_MATCH_COUNTERS && 653 !(ip_set_match_counter(ip_set_get_packets(counter), 654 mext->packets, mext->packets_op) && 655 ip_set_match_counter(ip_set_get_bytes(counter), 656 mext->bytes, mext->bytes_op))) 657 return false; 658 ip_set_update_counter(counter, ext, flags); 659 } 660 if (SET_WITH_SKBINFO(set)) 661 ip_set_get_skbinfo(ext_skbinfo(data, set), 662 ext, mext, flags); 663 return true; 664 } 665 EXPORT_SYMBOL_GPL(ip_set_match_extensions); 666 667 /* Creating/destroying/renaming/swapping affect the existence and 668 * the properties of a set. All of these can be executed from userspace 669 * only and serialized by the nfnl mutex indirectly from nfnetlink. 670 * 671 * Sets are identified by their index in ip_set_list and the index 672 * is used by the external references (set/SET netfilter modules). 673 * 674 * The set behind an index may change by swapping only, from userspace. 675 */ 676 677 static void 678 __ip_set_get(struct ip_set *set) 679 { 680 write_lock_bh(&ip_set_ref_lock); 681 set->ref++; 682 write_unlock_bh(&ip_set_ref_lock); 683 } 684 685 static void 686 __ip_set_put(struct ip_set *set) 687 { 688 write_lock_bh(&ip_set_ref_lock); 689 BUG_ON(set->ref == 0); 690 set->ref--; 691 write_unlock_bh(&ip_set_ref_lock); 692 } 693 694 /* set->ref can be swapped out by ip_set_swap, netlink events (like dump) need 695 * a separate reference counter 696 */ 697 static void 698 __ip_set_put_netlink(struct ip_set *set) 699 { 700 write_lock_bh(&ip_set_ref_lock); 701 BUG_ON(set->ref_netlink == 0); 702 set->ref_netlink--; 703 write_unlock_bh(&ip_set_ref_lock); 704 } 705 706 /* Add, del and test set entries from kernel. 707 * 708 * The set behind the index must exist and must be referenced 709 * so it can't be destroyed (or changed) under our foot. 710 */ 711 712 static struct ip_set * 713 ip_set_rcu_get(struct net *net, ip_set_id_t index) 714 { 715 struct ip_set *set; 716 struct ip_set_net *inst = ip_set_pernet(net); 717 718 rcu_read_lock(); 719 /* ip_set_list itself needs to be protected */ 720 set = rcu_dereference(inst->ip_set_list)[index]; 721 rcu_read_unlock(); 722 723 return set; 724 } 725 726 static inline void 727 ip_set_lock(struct ip_set *set) 728 { 729 if (!set->variant->region_lock) 730 spin_lock_bh(&set->lock); 731 } 732 733 static inline void 734 ip_set_unlock(struct ip_set *set) 735 { 736 if (!set->variant->region_lock) 737 spin_unlock_bh(&set->lock); 738 } 739 740 int 741 ip_set_test(ip_set_id_t index, const struct sk_buff *skb, 742 const struct xt_action_param *par, struct ip_set_adt_opt *opt) 743 { 744 struct ip_set *set = ip_set_rcu_get(xt_net(par), index); 745 int ret = 0; 746 747 BUG_ON(!set); 748 pr_debug("set %s, index %u\n", set->name, index); 749 750 if (opt->dim < set->type->dimension || 751 !(opt->family == set->family || set->family == NFPROTO_UNSPEC)) 752 return 0; 753 754 rcu_read_lock_bh(); 755 ret = set->variant->kadt(set, skb, par, IPSET_TEST, opt); 756 rcu_read_unlock_bh(); 757 758 if (ret == -EAGAIN) { 759 /* Type requests element to be completed */ 760 pr_debug("element must be completed, ADD is triggered\n"); 761 ip_set_lock(set); 762 set->variant->kadt(set, skb, par, IPSET_ADD, opt); 763 ip_set_unlock(set); 764 ret = 1; 765 } else { 766 /* --return-nomatch: invert matched element */ 767 if ((opt->cmdflags & IPSET_FLAG_RETURN_NOMATCH) && 768 (set->type->features & IPSET_TYPE_NOMATCH) && 769 (ret > 0 || ret == -ENOTEMPTY)) 770 ret = -ret; 771 } 772 773 /* Convert error codes to nomatch */ 774 return (ret < 0 ? 0 : ret); 775 } 776 EXPORT_SYMBOL_GPL(ip_set_test); 777 778 int 779 ip_set_add(ip_set_id_t index, const struct sk_buff *skb, 780 const struct xt_action_param *par, struct ip_set_adt_opt *opt) 781 { 782 struct ip_set *set = ip_set_rcu_get(xt_net(par), index); 783 int ret; 784 785 BUG_ON(!set); 786 pr_debug("set %s, index %u\n", set->name, index); 787 788 if (opt->dim < set->type->dimension || 789 !(opt->family == set->family || set->family == NFPROTO_UNSPEC)) 790 return -IPSET_ERR_TYPE_MISMATCH; 791 792 ip_set_lock(set); 793 ret = set->variant->kadt(set, skb, par, IPSET_ADD, opt); 794 ip_set_unlock(set); 795 796 return ret; 797 } 798 EXPORT_SYMBOL_GPL(ip_set_add); 799 800 int 801 ip_set_del(ip_set_id_t index, const struct sk_buff *skb, 802 const struct xt_action_param *par, struct ip_set_adt_opt *opt) 803 { 804 struct ip_set *set = ip_set_rcu_get(xt_net(par), index); 805 int ret = 0; 806 807 BUG_ON(!set); 808 pr_debug("set %s, index %u\n", set->name, index); 809 810 if (opt->dim < set->type->dimension || 811 !(opt->family == set->family || set->family == NFPROTO_UNSPEC)) 812 return -IPSET_ERR_TYPE_MISMATCH; 813 814 ip_set_lock(set); 815 ret = set->variant->kadt(set, skb, par, IPSET_DEL, opt); 816 ip_set_unlock(set); 817 818 return ret; 819 } 820 EXPORT_SYMBOL_GPL(ip_set_del); 821 822 /* Find set by name, reference it once. The reference makes sure the 823 * thing pointed to, does not go away under our feet. 824 * 825 */ 826 ip_set_id_t 827 ip_set_get_byname(struct net *net, const char *name, struct ip_set **set) 828 { 829 ip_set_id_t i, index = IPSET_INVALID_ID; 830 struct ip_set *s; 831 struct ip_set_net *inst = ip_set_pernet(net); 832 833 rcu_read_lock(); 834 for (i = 0; i < inst->ip_set_max; i++) { 835 s = rcu_dereference(inst->ip_set_list)[i]; 836 if (s && STRNCMP(s->name, name)) { 837 __ip_set_get(s); 838 index = i; 839 *set = s; 840 break; 841 } 842 } 843 rcu_read_unlock(); 844 845 return index; 846 } 847 EXPORT_SYMBOL_GPL(ip_set_get_byname); 848 849 /* If the given set pointer points to a valid set, decrement 850 * reference count by 1. The caller shall not assume the index 851 * to be valid, after calling this function. 852 * 853 */ 854 855 static void 856 __ip_set_put_byindex(struct ip_set_net *inst, ip_set_id_t index) 857 { 858 struct ip_set *set; 859 860 rcu_read_lock(); 861 set = rcu_dereference(inst->ip_set_list)[index]; 862 if (set) 863 __ip_set_put(set); 864 rcu_read_unlock(); 865 } 866 867 void 868 ip_set_put_byindex(struct net *net, ip_set_id_t index) 869 { 870 struct ip_set_net *inst = ip_set_pernet(net); 871 872 __ip_set_put_byindex(inst, index); 873 } 874 EXPORT_SYMBOL_GPL(ip_set_put_byindex); 875 876 /* Get the name of a set behind a set index. 877 * Set itself is protected by RCU, but its name isn't: to protect against 878 * renaming, grab ip_set_ref_lock as reader (see ip_set_rename()) and copy the 879 * name. 880 */ 881 void 882 ip_set_name_byindex(struct net *net, ip_set_id_t index, char *name) 883 { 884 struct ip_set *set = ip_set_rcu_get(net, index); 885 886 BUG_ON(!set); 887 888 read_lock_bh(&ip_set_ref_lock); 889 strncpy(name, set->name, IPSET_MAXNAMELEN); 890 read_unlock_bh(&ip_set_ref_lock); 891 } 892 EXPORT_SYMBOL_GPL(ip_set_name_byindex); 893 894 /* Routines to call by external subsystems, which do not 895 * call nfnl_lock for us. 896 */ 897 898 /* Find set by index, reference it once. The reference makes sure the 899 * thing pointed to, does not go away under our feet. 900 * 901 * The nfnl mutex is used in the function. 902 */ 903 ip_set_id_t 904 ip_set_nfnl_get_byindex(struct net *net, ip_set_id_t index) 905 { 906 struct ip_set *set; 907 struct ip_set_net *inst = ip_set_pernet(net); 908 909 if (index >= inst->ip_set_max) 910 return IPSET_INVALID_ID; 911 912 nfnl_lock(NFNL_SUBSYS_IPSET); 913 set = ip_set(inst, index); 914 if (set) 915 __ip_set_get(set); 916 else 917 index = IPSET_INVALID_ID; 918 nfnl_unlock(NFNL_SUBSYS_IPSET); 919 920 return index; 921 } 922 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex); 923 924 /* If the given set pointer points to a valid set, decrement 925 * reference count by 1. The caller shall not assume the index 926 * to be valid, after calling this function. 927 * 928 * The nfnl mutex is used in the function. 929 */ 930 void 931 ip_set_nfnl_put(struct net *net, ip_set_id_t index) 932 { 933 struct ip_set *set; 934 struct ip_set_net *inst = ip_set_pernet(net); 935 936 nfnl_lock(NFNL_SUBSYS_IPSET); 937 if (!inst->is_deleted) { /* already deleted from ip_set_net_exit() */ 938 set = ip_set(inst, index); 939 if (set) 940 __ip_set_put(set); 941 } 942 nfnl_unlock(NFNL_SUBSYS_IPSET); 943 } 944 EXPORT_SYMBOL_GPL(ip_set_nfnl_put); 945 946 /* Communication protocol with userspace over netlink. 947 * 948 * The commands are serialized by the nfnl mutex. 949 */ 950 951 static inline u8 protocol(const struct nlattr * const tb[]) 952 { 953 return nla_get_u8(tb[IPSET_ATTR_PROTOCOL]); 954 } 955 956 static inline bool 957 protocol_failed(const struct nlattr * const tb[]) 958 { 959 return !tb[IPSET_ATTR_PROTOCOL] || protocol(tb) != IPSET_PROTOCOL; 960 } 961 962 static inline bool 963 protocol_min_failed(const struct nlattr * const tb[]) 964 { 965 return !tb[IPSET_ATTR_PROTOCOL] || protocol(tb) < IPSET_PROTOCOL_MIN; 966 } 967 968 static inline u32 969 flag_exist(const struct nlmsghdr *nlh) 970 { 971 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST; 972 } 973 974 static struct nlmsghdr * 975 start_msg(struct sk_buff *skb, u32 portid, u32 seq, unsigned int flags, 976 enum ipset_cmd cmd) 977 { 978 struct nlmsghdr *nlh; 979 struct nfgenmsg *nfmsg; 980 981 nlh = nlmsg_put(skb, portid, seq, nfnl_msg_type(NFNL_SUBSYS_IPSET, cmd), 982 sizeof(*nfmsg), flags); 983 if (!nlh) 984 return NULL; 985 986 nfmsg = nlmsg_data(nlh); 987 nfmsg->nfgen_family = NFPROTO_IPV4; 988 nfmsg->version = NFNETLINK_V0; 989 nfmsg->res_id = 0; 990 991 return nlh; 992 } 993 994 /* Create a set */ 995 996 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = { 997 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 998 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 999 .len = IPSET_MAXNAMELEN - 1 }, 1000 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING, 1001 .len = IPSET_MAXNAMELEN - 1}, 1002 [IPSET_ATTR_REVISION] = { .type = NLA_U8 }, 1003 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 }, 1004 [IPSET_ATTR_DATA] = { .type = NLA_NESTED }, 1005 }; 1006 1007 static struct ip_set * 1008 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id) 1009 { 1010 struct ip_set *set = NULL; 1011 ip_set_id_t i; 1012 1013 *id = IPSET_INVALID_ID; 1014 for (i = 0; i < inst->ip_set_max; i++) { 1015 set = ip_set(inst, i); 1016 if (set && STRNCMP(set->name, name)) { 1017 *id = i; 1018 break; 1019 } 1020 } 1021 return (*id == IPSET_INVALID_ID ? NULL : set); 1022 } 1023 1024 static inline struct ip_set * 1025 find_set(struct ip_set_net *inst, const char *name) 1026 { 1027 ip_set_id_t id; 1028 1029 return find_set_and_id(inst, name, &id); 1030 } 1031 1032 static int 1033 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index, 1034 struct ip_set **set) 1035 { 1036 struct ip_set *s; 1037 ip_set_id_t i; 1038 1039 *index = IPSET_INVALID_ID; 1040 for (i = 0; i < inst->ip_set_max; i++) { 1041 s = ip_set(inst, i); 1042 if (!s) { 1043 if (*index == IPSET_INVALID_ID) 1044 *index = i; 1045 } else if (STRNCMP(name, s->name)) { 1046 /* Name clash */ 1047 *set = s; 1048 return -EEXIST; 1049 } 1050 } 1051 if (*index == IPSET_INVALID_ID) 1052 /* No free slot remained */ 1053 return -IPSET_ERR_MAX_SETS; 1054 return 0; 1055 } 1056 1057 static int ip_set_none(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1058 const struct nlmsghdr *nlh, 1059 const struct nlattr * const attr[], 1060 struct netlink_ext_ack *extack) 1061 { 1062 return -EOPNOTSUPP; 1063 } 1064 1065 static int ip_set_create(struct net *net, struct sock *ctnl, 1066 struct sk_buff *skb, const struct nlmsghdr *nlh, 1067 const struct nlattr * const attr[], 1068 struct netlink_ext_ack *extack) 1069 { 1070 struct ip_set_net *inst = ip_set_pernet(net); 1071 struct ip_set *set, *clash = NULL; 1072 ip_set_id_t index = IPSET_INVALID_ID; 1073 struct nlattr *tb[IPSET_ATTR_CREATE_MAX + 1] = {}; 1074 const char *name, *typename; 1075 u8 family, revision; 1076 u32 flags = flag_exist(nlh); 1077 int ret = 0; 1078 1079 if (unlikely(protocol_min_failed(attr) || 1080 !attr[IPSET_ATTR_SETNAME] || 1081 !attr[IPSET_ATTR_TYPENAME] || 1082 !attr[IPSET_ATTR_REVISION] || 1083 !attr[IPSET_ATTR_FAMILY] || 1084 (attr[IPSET_ATTR_DATA] && 1085 !flag_nested(attr[IPSET_ATTR_DATA])))) 1086 return -IPSET_ERR_PROTOCOL; 1087 1088 name = nla_data(attr[IPSET_ATTR_SETNAME]); 1089 typename = nla_data(attr[IPSET_ATTR_TYPENAME]); 1090 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]); 1091 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]); 1092 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n", 1093 name, typename, family_name(family), revision); 1094 1095 /* First, and without any locks, allocate and initialize 1096 * a normal base set structure. 1097 */ 1098 set = kzalloc(sizeof(*set), GFP_KERNEL); 1099 if (!set) 1100 return -ENOMEM; 1101 spin_lock_init(&set->lock); 1102 strlcpy(set->name, name, IPSET_MAXNAMELEN); 1103 set->family = family; 1104 set->revision = revision; 1105 1106 /* Next, check that we know the type, and take 1107 * a reference on the type, to make sure it stays available 1108 * while constructing our new set. 1109 * 1110 * After referencing the type, we try to create the type 1111 * specific part of the set without holding any locks. 1112 */ 1113 ret = find_set_type_get(typename, family, revision, &set->type); 1114 if (ret) 1115 goto out; 1116 1117 /* Without holding any locks, create private part. */ 1118 if (attr[IPSET_ATTR_DATA] && 1119 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA], 1120 set->type->create_policy, NULL)) { 1121 ret = -IPSET_ERR_PROTOCOL; 1122 goto put_out; 1123 } 1124 1125 ret = set->type->create(net, set, tb, flags); 1126 if (ret != 0) 1127 goto put_out; 1128 1129 /* BTW, ret==0 here. */ 1130 1131 /* Here, we have a valid, constructed set and we are protected 1132 * by the nfnl mutex. Find the first free index in ip_set_list 1133 * and check clashing. 1134 */ 1135 ret = find_free_id(inst, set->name, &index, &clash); 1136 if (ret == -EEXIST) { 1137 /* If this is the same set and requested, ignore error */ 1138 if ((flags & IPSET_FLAG_EXIST) && 1139 STRNCMP(set->type->name, clash->type->name) && 1140 set->type->family == clash->type->family && 1141 set->type->revision_min == clash->type->revision_min && 1142 set->type->revision_max == clash->type->revision_max && 1143 set->variant->same_set(set, clash)) 1144 ret = 0; 1145 goto cleanup; 1146 } else if (ret == -IPSET_ERR_MAX_SETS) { 1147 struct ip_set **list, **tmp; 1148 ip_set_id_t i = inst->ip_set_max + IP_SET_INC; 1149 1150 if (i < inst->ip_set_max || i == IPSET_INVALID_ID) 1151 /* Wraparound */ 1152 goto cleanup; 1153 1154 list = kvcalloc(i, sizeof(struct ip_set *), GFP_KERNEL); 1155 if (!list) 1156 goto cleanup; 1157 /* nfnl mutex is held, both lists are valid */ 1158 tmp = ip_set_dereference(inst->ip_set_list); 1159 memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max); 1160 rcu_assign_pointer(inst->ip_set_list, list); 1161 /* Make sure all current packets have passed through */ 1162 synchronize_net(); 1163 /* Use new list */ 1164 index = inst->ip_set_max; 1165 inst->ip_set_max = i; 1166 kvfree(tmp); 1167 ret = 0; 1168 } else if (ret) { 1169 goto cleanup; 1170 } 1171 1172 /* Finally! Add our shiny new set to the list, and be done. */ 1173 pr_debug("create: '%s' created with index %u!\n", set->name, index); 1174 ip_set(inst, index) = set; 1175 1176 return ret; 1177 1178 cleanup: 1179 set->variant->destroy(set); 1180 put_out: 1181 module_put(set->type->me); 1182 out: 1183 kfree(set); 1184 return ret; 1185 } 1186 1187 /* Destroy sets */ 1188 1189 static const struct nla_policy 1190 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = { 1191 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1192 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 1193 .len = IPSET_MAXNAMELEN - 1 }, 1194 }; 1195 1196 static void 1197 ip_set_destroy_set(struct ip_set *set) 1198 { 1199 pr_debug("set: %s\n", set->name); 1200 1201 /* Must call it without holding any lock */ 1202 set->variant->destroy(set); 1203 module_put(set->type->me); 1204 kfree(set); 1205 } 1206 1207 static int ip_set_destroy(struct net *net, struct sock *ctnl, 1208 struct sk_buff *skb, const struct nlmsghdr *nlh, 1209 const struct nlattr * const attr[], 1210 struct netlink_ext_ack *extack) 1211 { 1212 struct ip_set_net *inst = ip_set_pernet(net); 1213 struct ip_set *s; 1214 ip_set_id_t i; 1215 int ret = 0; 1216 1217 if (unlikely(protocol_min_failed(attr))) 1218 return -IPSET_ERR_PROTOCOL; 1219 1220 /* Must wait for flush to be really finished in list:set */ 1221 rcu_barrier(); 1222 1223 /* Commands are serialized and references are 1224 * protected by the ip_set_ref_lock. 1225 * External systems (i.e. xt_set) must call 1226 * ip_set_put|get_nfnl_* functions, that way we 1227 * can safely check references here. 1228 * 1229 * list:set timer can only decrement the reference 1230 * counter, so if it's already zero, we can proceed 1231 * without holding the lock. 1232 */ 1233 read_lock_bh(&ip_set_ref_lock); 1234 if (!attr[IPSET_ATTR_SETNAME]) { 1235 for (i = 0; i < inst->ip_set_max; i++) { 1236 s = ip_set(inst, i); 1237 if (s && (s->ref || s->ref_netlink)) { 1238 ret = -IPSET_ERR_BUSY; 1239 goto out; 1240 } 1241 } 1242 inst->is_destroyed = true; 1243 read_unlock_bh(&ip_set_ref_lock); 1244 for (i = 0; i < inst->ip_set_max; i++) { 1245 s = ip_set(inst, i); 1246 if (s) { 1247 ip_set(inst, i) = NULL; 1248 ip_set_destroy_set(s); 1249 } 1250 } 1251 /* Modified by ip_set_destroy() only, which is serialized */ 1252 inst->is_destroyed = false; 1253 } else { 1254 s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]), 1255 &i); 1256 if (!s) { 1257 ret = -ENOENT; 1258 goto out; 1259 } else if (s->ref || s->ref_netlink) { 1260 ret = -IPSET_ERR_BUSY; 1261 goto out; 1262 } 1263 ip_set(inst, i) = NULL; 1264 read_unlock_bh(&ip_set_ref_lock); 1265 1266 ip_set_destroy_set(s); 1267 } 1268 return 0; 1269 out: 1270 read_unlock_bh(&ip_set_ref_lock); 1271 return ret; 1272 } 1273 1274 /* Flush sets */ 1275 1276 static void 1277 ip_set_flush_set(struct ip_set *set) 1278 { 1279 pr_debug("set: %s\n", set->name); 1280 1281 ip_set_lock(set); 1282 set->variant->flush(set); 1283 ip_set_unlock(set); 1284 } 1285 1286 static int ip_set_flush(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1287 const struct nlmsghdr *nlh, 1288 const struct nlattr * const attr[], 1289 struct netlink_ext_ack *extack) 1290 { 1291 struct ip_set_net *inst = ip_set_pernet(net); 1292 struct ip_set *s; 1293 ip_set_id_t i; 1294 1295 if (unlikely(protocol_min_failed(attr))) 1296 return -IPSET_ERR_PROTOCOL; 1297 1298 if (!attr[IPSET_ATTR_SETNAME]) { 1299 for (i = 0; i < inst->ip_set_max; i++) { 1300 s = ip_set(inst, i); 1301 if (s) 1302 ip_set_flush_set(s); 1303 } 1304 } else { 1305 s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME])); 1306 if (!s) 1307 return -ENOENT; 1308 1309 ip_set_flush_set(s); 1310 } 1311 1312 return 0; 1313 } 1314 1315 /* Rename a set */ 1316 1317 static const struct nla_policy 1318 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = { 1319 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1320 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 1321 .len = IPSET_MAXNAMELEN - 1 }, 1322 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING, 1323 .len = IPSET_MAXNAMELEN - 1 }, 1324 }; 1325 1326 static int ip_set_rename(struct net *net, struct sock *ctnl, 1327 struct sk_buff *skb, const struct nlmsghdr *nlh, 1328 const struct nlattr * const attr[], 1329 struct netlink_ext_ack *extack) 1330 { 1331 struct ip_set_net *inst = ip_set_pernet(net); 1332 struct ip_set *set, *s; 1333 const char *name2; 1334 ip_set_id_t i; 1335 int ret = 0; 1336 1337 if (unlikely(protocol_min_failed(attr) || 1338 !attr[IPSET_ATTR_SETNAME] || 1339 !attr[IPSET_ATTR_SETNAME2])) 1340 return -IPSET_ERR_PROTOCOL; 1341 1342 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME])); 1343 if (!set) 1344 return -ENOENT; 1345 1346 write_lock_bh(&ip_set_ref_lock); 1347 if (set->ref != 0 || set->ref_netlink != 0) { 1348 ret = -IPSET_ERR_REFERENCED; 1349 goto out; 1350 } 1351 1352 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]); 1353 for (i = 0; i < inst->ip_set_max; i++) { 1354 s = ip_set(inst, i); 1355 if (s && STRNCMP(s->name, name2)) { 1356 ret = -IPSET_ERR_EXIST_SETNAME2; 1357 goto out; 1358 } 1359 } 1360 strncpy(set->name, name2, IPSET_MAXNAMELEN); 1361 1362 out: 1363 write_unlock_bh(&ip_set_ref_lock); 1364 return ret; 1365 } 1366 1367 /* Swap two sets so that name/index points to the other. 1368 * References and set names are also swapped. 1369 * 1370 * The commands are serialized by the nfnl mutex and references are 1371 * protected by the ip_set_ref_lock. The kernel interfaces 1372 * do not hold the mutex but the pointer settings are atomic 1373 * so the ip_set_list always contains valid pointers to the sets. 1374 */ 1375 1376 static int ip_set_swap(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1377 const struct nlmsghdr *nlh, 1378 const struct nlattr * const attr[], 1379 struct netlink_ext_ack *extack) 1380 { 1381 struct ip_set_net *inst = ip_set_pernet(net); 1382 struct ip_set *from, *to; 1383 ip_set_id_t from_id, to_id; 1384 char from_name[IPSET_MAXNAMELEN]; 1385 1386 if (unlikely(protocol_min_failed(attr) || 1387 !attr[IPSET_ATTR_SETNAME] || 1388 !attr[IPSET_ATTR_SETNAME2])) 1389 return -IPSET_ERR_PROTOCOL; 1390 1391 from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]), 1392 &from_id); 1393 if (!from) 1394 return -ENOENT; 1395 1396 to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]), 1397 &to_id); 1398 if (!to) 1399 return -IPSET_ERR_EXIST_SETNAME2; 1400 1401 /* Features must not change. 1402 * Not an artifical restriction anymore, as we must prevent 1403 * possible loops created by swapping in setlist type of sets. 1404 */ 1405 if (!(from->type->features == to->type->features && 1406 from->family == to->family)) 1407 return -IPSET_ERR_TYPE_MISMATCH; 1408 1409 write_lock_bh(&ip_set_ref_lock); 1410 1411 if (from->ref_netlink || to->ref_netlink) { 1412 write_unlock_bh(&ip_set_ref_lock); 1413 return -EBUSY; 1414 } 1415 1416 strncpy(from_name, from->name, IPSET_MAXNAMELEN); 1417 strncpy(from->name, to->name, IPSET_MAXNAMELEN); 1418 strncpy(to->name, from_name, IPSET_MAXNAMELEN); 1419 1420 swap(from->ref, to->ref); 1421 ip_set(inst, from_id) = to; 1422 ip_set(inst, to_id) = from; 1423 write_unlock_bh(&ip_set_ref_lock); 1424 1425 return 0; 1426 } 1427 1428 /* List/save set data */ 1429 1430 #define DUMP_INIT 0 1431 #define DUMP_ALL 1 1432 #define DUMP_ONE 2 1433 #define DUMP_LAST 3 1434 1435 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF) 1436 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16) 1437 1438 int 1439 ip_set_put_flags(struct sk_buff *skb, struct ip_set *set) 1440 { 1441 u32 cadt_flags = 0; 1442 1443 if (SET_WITH_TIMEOUT(set)) 1444 if (unlikely(nla_put_net32(skb, IPSET_ATTR_TIMEOUT, 1445 htonl(set->timeout)))) 1446 return -EMSGSIZE; 1447 if (SET_WITH_COUNTER(set)) 1448 cadt_flags |= IPSET_FLAG_WITH_COUNTERS; 1449 if (SET_WITH_COMMENT(set)) 1450 cadt_flags |= IPSET_FLAG_WITH_COMMENT; 1451 if (SET_WITH_SKBINFO(set)) 1452 cadt_flags |= IPSET_FLAG_WITH_SKBINFO; 1453 if (SET_WITH_FORCEADD(set)) 1454 cadt_flags |= IPSET_FLAG_WITH_FORCEADD; 1455 1456 if (!cadt_flags) 1457 return 0; 1458 return nla_put_net32(skb, IPSET_ATTR_CADT_FLAGS, htonl(cadt_flags)); 1459 } 1460 EXPORT_SYMBOL_GPL(ip_set_put_flags); 1461 1462 static int 1463 ip_set_dump_done(struct netlink_callback *cb) 1464 { 1465 if (cb->args[IPSET_CB_ARG0]) { 1466 struct ip_set_net *inst = 1467 (struct ip_set_net *)cb->args[IPSET_CB_NET]; 1468 ip_set_id_t index = (ip_set_id_t)cb->args[IPSET_CB_INDEX]; 1469 struct ip_set *set = ip_set_ref_netlink(inst, index); 1470 1471 if (set->variant->uref) 1472 set->variant->uref(set, cb, false); 1473 pr_debug("release set %s\n", set->name); 1474 __ip_set_put_netlink(set); 1475 } 1476 return 0; 1477 } 1478 1479 static inline void 1480 dump_attrs(struct nlmsghdr *nlh) 1481 { 1482 const struct nlattr *attr; 1483 int rem; 1484 1485 pr_debug("dump nlmsg\n"); 1486 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) { 1487 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len); 1488 } 1489 } 1490 1491 static const struct nla_policy 1492 ip_set_dump_policy[IPSET_ATTR_CMD_MAX + 1] = { 1493 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1494 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 1495 .len = IPSET_MAXNAMELEN - 1 }, 1496 [IPSET_ATTR_FLAGS] = { .type = NLA_U32 }, 1497 }; 1498 1499 static int 1500 ip_set_dump_start(struct netlink_callback *cb) 1501 { 1502 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb); 1503 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg)); 1504 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1]; 1505 struct nlattr *attr = (void *)nlh + min_len; 1506 struct sk_buff *skb = cb->skb; 1507 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk)); 1508 u32 dump_type; 1509 int ret; 1510 1511 ret = nla_parse(cda, IPSET_ATTR_CMD_MAX, attr, 1512 nlh->nlmsg_len - min_len, 1513 ip_set_dump_policy, NULL); 1514 if (ret) 1515 goto error; 1516 1517 cb->args[IPSET_CB_PROTO] = nla_get_u8(cda[IPSET_ATTR_PROTOCOL]); 1518 if (cda[IPSET_ATTR_SETNAME]) { 1519 ip_set_id_t index; 1520 struct ip_set *set; 1521 1522 set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]), 1523 &index); 1524 if (!set) { 1525 ret = -ENOENT; 1526 goto error; 1527 } 1528 dump_type = DUMP_ONE; 1529 cb->args[IPSET_CB_INDEX] = index; 1530 } else { 1531 dump_type = DUMP_ALL; 1532 } 1533 1534 if (cda[IPSET_ATTR_FLAGS]) { 1535 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]); 1536 1537 dump_type |= (f << 16); 1538 } 1539 cb->args[IPSET_CB_NET] = (unsigned long)inst; 1540 cb->args[IPSET_CB_DUMP] = dump_type; 1541 1542 return 0; 1543 1544 error: 1545 /* We have to create and send the error message manually :-( */ 1546 if (nlh->nlmsg_flags & NLM_F_ACK) { 1547 netlink_ack(cb->skb, nlh, ret, NULL); 1548 } 1549 return ret; 1550 } 1551 1552 static int 1553 ip_set_dump_do(struct sk_buff *skb, struct netlink_callback *cb) 1554 { 1555 ip_set_id_t index = IPSET_INVALID_ID, max; 1556 struct ip_set *set = NULL; 1557 struct nlmsghdr *nlh = NULL; 1558 unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0; 1559 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk)); 1560 u32 dump_type, dump_flags; 1561 bool is_destroyed; 1562 int ret = 0; 1563 1564 if (!cb->args[IPSET_CB_DUMP]) 1565 return -EINVAL; 1566 1567 if (cb->args[IPSET_CB_INDEX] >= inst->ip_set_max) 1568 goto out; 1569 1570 dump_type = DUMP_TYPE(cb->args[IPSET_CB_DUMP]); 1571 dump_flags = DUMP_FLAGS(cb->args[IPSET_CB_DUMP]); 1572 max = dump_type == DUMP_ONE ? cb->args[IPSET_CB_INDEX] + 1 1573 : inst->ip_set_max; 1574 dump_last: 1575 pr_debug("dump type, flag: %u %u index: %ld\n", 1576 dump_type, dump_flags, cb->args[IPSET_CB_INDEX]); 1577 for (; cb->args[IPSET_CB_INDEX] < max; cb->args[IPSET_CB_INDEX]++) { 1578 index = (ip_set_id_t)cb->args[IPSET_CB_INDEX]; 1579 write_lock_bh(&ip_set_ref_lock); 1580 set = ip_set(inst, index); 1581 is_destroyed = inst->is_destroyed; 1582 if (!set || is_destroyed) { 1583 write_unlock_bh(&ip_set_ref_lock); 1584 if (dump_type == DUMP_ONE) { 1585 ret = -ENOENT; 1586 goto out; 1587 } 1588 if (is_destroyed) { 1589 /* All sets are just being destroyed */ 1590 ret = 0; 1591 goto out; 1592 } 1593 continue; 1594 } 1595 /* When dumping all sets, we must dump "sorted" 1596 * so that lists (unions of sets) are dumped last. 1597 */ 1598 if (dump_type != DUMP_ONE && 1599 ((dump_type == DUMP_ALL) == 1600 !!(set->type->features & IPSET_DUMP_LAST))) { 1601 write_unlock_bh(&ip_set_ref_lock); 1602 continue; 1603 } 1604 pr_debug("List set: %s\n", set->name); 1605 if (!cb->args[IPSET_CB_ARG0]) { 1606 /* Start listing: make sure set won't be destroyed */ 1607 pr_debug("reference set\n"); 1608 set->ref_netlink++; 1609 } 1610 write_unlock_bh(&ip_set_ref_lock); 1611 nlh = start_msg(skb, NETLINK_CB(cb->skb).portid, 1612 cb->nlh->nlmsg_seq, flags, 1613 IPSET_CMD_LIST); 1614 if (!nlh) { 1615 ret = -EMSGSIZE; 1616 goto release_refcount; 1617 } 1618 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL, 1619 cb->args[IPSET_CB_PROTO]) || 1620 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name)) 1621 goto nla_put_failure; 1622 if (dump_flags & IPSET_FLAG_LIST_SETNAME) 1623 goto next_set; 1624 switch (cb->args[IPSET_CB_ARG0]) { 1625 case 0: 1626 /* Core header data */ 1627 if (nla_put_string(skb, IPSET_ATTR_TYPENAME, 1628 set->type->name) || 1629 nla_put_u8(skb, IPSET_ATTR_FAMILY, 1630 set->family) || 1631 nla_put_u8(skb, IPSET_ATTR_REVISION, 1632 set->revision)) 1633 goto nla_put_failure; 1634 if (cb->args[IPSET_CB_PROTO] > IPSET_PROTOCOL_MIN && 1635 nla_put_net16(skb, IPSET_ATTR_INDEX, htons(index))) 1636 goto nla_put_failure; 1637 ret = set->variant->head(set, skb); 1638 if (ret < 0) 1639 goto release_refcount; 1640 if (dump_flags & IPSET_FLAG_LIST_HEADER) 1641 goto next_set; 1642 if (set->variant->uref) 1643 set->variant->uref(set, cb, true); 1644 /* fall through */ 1645 default: 1646 ret = set->variant->list(set, skb, cb); 1647 if (!cb->args[IPSET_CB_ARG0]) 1648 /* Set is done, proceed with next one */ 1649 goto next_set; 1650 goto release_refcount; 1651 } 1652 } 1653 /* If we dump all sets, continue with dumping last ones */ 1654 if (dump_type == DUMP_ALL) { 1655 dump_type = DUMP_LAST; 1656 cb->args[IPSET_CB_DUMP] = dump_type | (dump_flags << 16); 1657 cb->args[IPSET_CB_INDEX] = 0; 1658 if (set && set->variant->uref) 1659 set->variant->uref(set, cb, false); 1660 goto dump_last; 1661 } 1662 goto out; 1663 1664 nla_put_failure: 1665 ret = -EFAULT; 1666 next_set: 1667 if (dump_type == DUMP_ONE) 1668 cb->args[IPSET_CB_INDEX] = IPSET_INVALID_ID; 1669 else 1670 cb->args[IPSET_CB_INDEX]++; 1671 release_refcount: 1672 /* If there was an error or set is done, release set */ 1673 if (ret || !cb->args[IPSET_CB_ARG0]) { 1674 set = ip_set_ref_netlink(inst, index); 1675 if (set->variant->uref) 1676 set->variant->uref(set, cb, false); 1677 pr_debug("release set %s\n", set->name); 1678 __ip_set_put_netlink(set); 1679 cb->args[IPSET_CB_ARG0] = 0; 1680 } 1681 out: 1682 if (nlh) { 1683 nlmsg_end(skb, nlh); 1684 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len); 1685 dump_attrs(nlh); 1686 } 1687 1688 return ret < 0 ? ret : skb->len; 1689 } 1690 1691 static int ip_set_dump(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1692 const struct nlmsghdr *nlh, 1693 const struct nlattr * const attr[], 1694 struct netlink_ext_ack *extack) 1695 { 1696 if (unlikely(protocol_min_failed(attr))) 1697 return -IPSET_ERR_PROTOCOL; 1698 1699 { 1700 struct netlink_dump_control c = { 1701 .start = ip_set_dump_start, 1702 .dump = ip_set_dump_do, 1703 .done = ip_set_dump_done, 1704 }; 1705 return netlink_dump_start(ctnl, skb, nlh, &c); 1706 } 1707 } 1708 1709 /* Add, del and test */ 1710 1711 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = { 1712 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1713 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 1714 .len = IPSET_MAXNAMELEN - 1 }, 1715 [IPSET_ATTR_LINENO] = { .type = NLA_U32 }, 1716 [IPSET_ATTR_DATA] = { .type = NLA_NESTED }, 1717 [IPSET_ATTR_ADT] = { .type = NLA_NESTED }, 1718 }; 1719 1720 static int 1721 call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set, 1722 struct nlattr *tb[], enum ipset_adt adt, 1723 u32 flags, bool use_lineno) 1724 { 1725 int ret; 1726 u32 lineno = 0; 1727 bool eexist = flags & IPSET_FLAG_EXIST, retried = false; 1728 1729 do { 1730 ip_set_lock(set); 1731 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried); 1732 ip_set_unlock(set); 1733 retried = true; 1734 } while (ret == -EAGAIN && 1735 set->variant->resize && 1736 (ret = set->variant->resize(set, retried)) == 0); 1737 1738 if (!ret || (ret == -IPSET_ERR_EXIST && eexist)) 1739 return 0; 1740 if (lineno && use_lineno) { 1741 /* Error in restore/batch mode: send back lineno */ 1742 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb); 1743 struct sk_buff *skb2; 1744 struct nlmsgerr *errmsg; 1745 size_t payload = min(SIZE_MAX, 1746 sizeof(*errmsg) + nlmsg_len(nlh)); 1747 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg)); 1748 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1]; 1749 struct nlattr *cmdattr; 1750 u32 *errline; 1751 1752 skb2 = nlmsg_new(payload, GFP_KERNEL); 1753 if (!skb2) 1754 return -ENOMEM; 1755 rep = __nlmsg_put(skb2, NETLINK_CB(skb).portid, 1756 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0); 1757 errmsg = nlmsg_data(rep); 1758 errmsg->error = ret; 1759 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len); 1760 cmdattr = (void *)&errmsg->msg + min_len; 1761 1762 ret = nla_parse(cda, IPSET_ATTR_CMD_MAX, cmdattr, 1763 nlh->nlmsg_len - min_len, ip_set_adt_policy, 1764 NULL); 1765 1766 if (ret) { 1767 nlmsg_free(skb2); 1768 return ret; 1769 } 1770 errline = nla_data(cda[IPSET_ATTR_LINENO]); 1771 1772 *errline = lineno; 1773 1774 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, 1775 MSG_DONTWAIT); 1776 /* Signal netlink not to send its ACK/errmsg. */ 1777 return -EINTR; 1778 } 1779 1780 return ret; 1781 } 1782 1783 static int ip_set_ad(struct net *net, struct sock *ctnl, 1784 struct sk_buff *skb, 1785 enum ipset_adt adt, 1786 const struct nlmsghdr *nlh, 1787 const struct nlattr * const attr[], 1788 struct netlink_ext_ack *extack) 1789 { 1790 struct ip_set_net *inst = ip_set_pernet(net); 1791 struct ip_set *set; 1792 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {}; 1793 const struct nlattr *nla; 1794 u32 flags = flag_exist(nlh); 1795 bool use_lineno; 1796 int ret = 0; 1797 1798 if (unlikely(protocol_min_failed(attr) || 1799 !attr[IPSET_ATTR_SETNAME] || 1800 !((attr[IPSET_ATTR_DATA] != NULL) ^ 1801 (attr[IPSET_ATTR_ADT] != NULL)) || 1802 (attr[IPSET_ATTR_DATA] && 1803 !flag_nested(attr[IPSET_ATTR_DATA])) || 1804 (attr[IPSET_ATTR_ADT] && 1805 (!flag_nested(attr[IPSET_ATTR_ADT]) || 1806 !attr[IPSET_ATTR_LINENO])))) 1807 return -IPSET_ERR_PROTOCOL; 1808 1809 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME])); 1810 if (!set) 1811 return -ENOENT; 1812 1813 use_lineno = !!attr[IPSET_ATTR_LINENO]; 1814 if (attr[IPSET_ATTR_DATA]) { 1815 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, 1816 attr[IPSET_ATTR_DATA], 1817 set->type->adt_policy, NULL)) 1818 return -IPSET_ERR_PROTOCOL; 1819 ret = call_ad(ctnl, skb, set, tb, adt, flags, 1820 use_lineno); 1821 } else { 1822 int nla_rem; 1823 1824 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) { 1825 if (nla_type(nla) != IPSET_ATTR_DATA || 1826 !flag_nested(nla) || 1827 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla, 1828 set->type->adt_policy, NULL)) 1829 return -IPSET_ERR_PROTOCOL; 1830 ret = call_ad(ctnl, skb, set, tb, adt, 1831 flags, use_lineno); 1832 if (ret < 0) 1833 return ret; 1834 } 1835 } 1836 return ret; 1837 } 1838 1839 static int ip_set_uadd(struct net *net, struct sock *ctnl, 1840 struct sk_buff *skb, const struct nlmsghdr *nlh, 1841 const struct nlattr * const attr[], 1842 struct netlink_ext_ack *extack) 1843 { 1844 return ip_set_ad(net, ctnl, skb, 1845 IPSET_ADD, nlh, attr, extack); 1846 } 1847 1848 static int ip_set_udel(struct net *net, struct sock *ctnl, 1849 struct sk_buff *skb, const struct nlmsghdr *nlh, 1850 const struct nlattr * const attr[], 1851 struct netlink_ext_ack *extack) 1852 { 1853 return ip_set_ad(net, ctnl, skb, 1854 IPSET_DEL, nlh, attr, extack); 1855 } 1856 1857 static int ip_set_utest(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1858 const struct nlmsghdr *nlh, 1859 const struct nlattr * const attr[], 1860 struct netlink_ext_ack *extack) 1861 { 1862 struct ip_set_net *inst = ip_set_pernet(net); 1863 struct ip_set *set; 1864 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {}; 1865 int ret = 0; 1866 u32 lineno; 1867 1868 if (unlikely(protocol_min_failed(attr) || 1869 !attr[IPSET_ATTR_SETNAME] || 1870 !attr[IPSET_ATTR_DATA] || 1871 !flag_nested(attr[IPSET_ATTR_DATA]))) 1872 return -IPSET_ERR_PROTOCOL; 1873 1874 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME])); 1875 if (!set) 1876 return -ENOENT; 1877 1878 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA], 1879 set->type->adt_policy, NULL)) 1880 return -IPSET_ERR_PROTOCOL; 1881 1882 rcu_read_lock_bh(); 1883 ret = set->variant->uadt(set, tb, IPSET_TEST, &lineno, 0, 0); 1884 rcu_read_unlock_bh(); 1885 /* Userspace can't trigger element to be re-added */ 1886 if (ret == -EAGAIN) 1887 ret = 1; 1888 1889 return ret > 0 ? 0 : -IPSET_ERR_EXIST; 1890 } 1891 1892 /* Get headed data of a set */ 1893 1894 static int ip_set_header(struct net *net, struct sock *ctnl, 1895 struct sk_buff *skb, const struct nlmsghdr *nlh, 1896 const struct nlattr * const attr[], 1897 struct netlink_ext_ack *extack) 1898 { 1899 struct ip_set_net *inst = ip_set_pernet(net); 1900 const struct ip_set *set; 1901 struct sk_buff *skb2; 1902 struct nlmsghdr *nlh2; 1903 int ret = 0; 1904 1905 if (unlikely(protocol_min_failed(attr) || 1906 !attr[IPSET_ATTR_SETNAME])) 1907 return -IPSET_ERR_PROTOCOL; 1908 1909 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME])); 1910 if (!set) 1911 return -ENOENT; 1912 1913 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 1914 if (!skb2) 1915 return -ENOMEM; 1916 1917 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 1918 IPSET_CMD_HEADER); 1919 if (!nlh2) 1920 goto nlmsg_failure; 1921 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) || 1922 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) || 1923 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) || 1924 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) || 1925 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision)) 1926 goto nla_put_failure; 1927 nlmsg_end(skb2, nlh2); 1928 1929 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT); 1930 if (ret < 0) 1931 return ret; 1932 1933 return 0; 1934 1935 nla_put_failure: 1936 nlmsg_cancel(skb2, nlh2); 1937 nlmsg_failure: 1938 kfree_skb(skb2); 1939 return -EMSGSIZE; 1940 } 1941 1942 /* Get type data */ 1943 1944 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = { 1945 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1946 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING, 1947 .len = IPSET_MAXNAMELEN - 1 }, 1948 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 }, 1949 }; 1950 1951 static int ip_set_type(struct net *net, struct sock *ctnl, struct sk_buff *skb, 1952 const struct nlmsghdr *nlh, 1953 const struct nlattr * const attr[], 1954 struct netlink_ext_ack *extack) 1955 { 1956 struct sk_buff *skb2; 1957 struct nlmsghdr *nlh2; 1958 u8 family, min, max; 1959 const char *typename; 1960 int ret = 0; 1961 1962 if (unlikely(protocol_min_failed(attr) || 1963 !attr[IPSET_ATTR_TYPENAME] || 1964 !attr[IPSET_ATTR_FAMILY])) 1965 return -IPSET_ERR_PROTOCOL; 1966 1967 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]); 1968 typename = nla_data(attr[IPSET_ATTR_TYPENAME]); 1969 ret = find_set_type_minmax(typename, family, &min, &max); 1970 if (ret) 1971 return ret; 1972 1973 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 1974 if (!skb2) 1975 return -ENOMEM; 1976 1977 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 1978 IPSET_CMD_TYPE); 1979 if (!nlh2) 1980 goto nlmsg_failure; 1981 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) || 1982 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) || 1983 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) || 1984 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) || 1985 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min)) 1986 goto nla_put_failure; 1987 nlmsg_end(skb2, nlh2); 1988 1989 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len); 1990 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT); 1991 if (ret < 0) 1992 return ret; 1993 1994 return 0; 1995 1996 nla_put_failure: 1997 nlmsg_cancel(skb2, nlh2); 1998 nlmsg_failure: 1999 kfree_skb(skb2); 2000 return -EMSGSIZE; 2001 } 2002 2003 /* Get protocol version */ 2004 2005 static const struct nla_policy 2006 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = { 2007 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 2008 }; 2009 2010 static int ip_set_protocol(struct net *net, struct sock *ctnl, 2011 struct sk_buff *skb, const struct nlmsghdr *nlh, 2012 const struct nlattr * const attr[], 2013 struct netlink_ext_ack *extack) 2014 { 2015 struct sk_buff *skb2; 2016 struct nlmsghdr *nlh2; 2017 int ret = 0; 2018 2019 if (unlikely(!attr[IPSET_ATTR_PROTOCOL])) 2020 return -IPSET_ERR_PROTOCOL; 2021 2022 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 2023 if (!skb2) 2024 return -ENOMEM; 2025 2026 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 2027 IPSET_CMD_PROTOCOL); 2028 if (!nlh2) 2029 goto nlmsg_failure; 2030 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL)) 2031 goto nla_put_failure; 2032 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL_MIN, IPSET_PROTOCOL_MIN)) 2033 goto nla_put_failure; 2034 nlmsg_end(skb2, nlh2); 2035 2036 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT); 2037 if (ret < 0) 2038 return ret; 2039 2040 return 0; 2041 2042 nla_put_failure: 2043 nlmsg_cancel(skb2, nlh2); 2044 nlmsg_failure: 2045 kfree_skb(skb2); 2046 return -EMSGSIZE; 2047 } 2048 2049 /* Get set by name or index, from userspace */ 2050 2051 static int ip_set_byname(struct net *net, struct sock *ctnl, 2052 struct sk_buff *skb, const struct nlmsghdr *nlh, 2053 const struct nlattr * const attr[], 2054 struct netlink_ext_ack *extack) 2055 { 2056 struct ip_set_net *inst = ip_set_pernet(net); 2057 struct sk_buff *skb2; 2058 struct nlmsghdr *nlh2; 2059 ip_set_id_t id = IPSET_INVALID_ID; 2060 const struct ip_set *set; 2061 int ret = 0; 2062 2063 if (unlikely(protocol_failed(attr) || 2064 !attr[IPSET_ATTR_SETNAME])) 2065 return -IPSET_ERR_PROTOCOL; 2066 2067 set = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]), &id); 2068 if (id == IPSET_INVALID_ID) 2069 return -ENOENT; 2070 2071 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 2072 if (!skb2) 2073 return -ENOMEM; 2074 2075 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 2076 IPSET_CMD_GET_BYNAME); 2077 if (!nlh2) 2078 goto nlmsg_failure; 2079 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) || 2080 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) || 2081 nla_put_net16(skb2, IPSET_ATTR_INDEX, htons(id))) 2082 goto nla_put_failure; 2083 nlmsg_end(skb2, nlh2); 2084 2085 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT); 2086 if (ret < 0) 2087 return ret; 2088 2089 return 0; 2090 2091 nla_put_failure: 2092 nlmsg_cancel(skb2, nlh2); 2093 nlmsg_failure: 2094 kfree_skb(skb2); 2095 return -EMSGSIZE; 2096 } 2097 2098 static const struct nla_policy ip_set_index_policy[IPSET_ATTR_CMD_MAX + 1] = { 2099 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 2100 [IPSET_ATTR_INDEX] = { .type = NLA_U16 }, 2101 }; 2102 2103 static int ip_set_byindex(struct net *net, struct sock *ctnl, 2104 struct sk_buff *skb, const struct nlmsghdr *nlh, 2105 const struct nlattr * const attr[], 2106 struct netlink_ext_ack *extack) 2107 { 2108 struct ip_set_net *inst = ip_set_pernet(net); 2109 struct sk_buff *skb2; 2110 struct nlmsghdr *nlh2; 2111 ip_set_id_t id = IPSET_INVALID_ID; 2112 const struct ip_set *set; 2113 int ret = 0; 2114 2115 if (unlikely(protocol_failed(attr) || 2116 !attr[IPSET_ATTR_INDEX])) 2117 return -IPSET_ERR_PROTOCOL; 2118 2119 id = ip_set_get_h16(attr[IPSET_ATTR_INDEX]); 2120 if (id >= inst->ip_set_max) 2121 return -ENOENT; 2122 set = ip_set(inst, id); 2123 if (set == NULL) 2124 return -ENOENT; 2125 2126 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 2127 if (!skb2) 2128 return -ENOMEM; 2129 2130 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 2131 IPSET_CMD_GET_BYINDEX); 2132 if (!nlh2) 2133 goto nlmsg_failure; 2134 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) || 2135 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name)) 2136 goto nla_put_failure; 2137 nlmsg_end(skb2, nlh2); 2138 2139 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT); 2140 if (ret < 0) 2141 return ret; 2142 2143 return 0; 2144 2145 nla_put_failure: 2146 nlmsg_cancel(skb2, nlh2); 2147 nlmsg_failure: 2148 kfree_skb(skb2); 2149 return -EMSGSIZE; 2150 } 2151 2152 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = { 2153 [IPSET_CMD_NONE] = { 2154 .call = ip_set_none, 2155 .attr_count = IPSET_ATTR_CMD_MAX, 2156 }, 2157 [IPSET_CMD_CREATE] = { 2158 .call = ip_set_create, 2159 .attr_count = IPSET_ATTR_CMD_MAX, 2160 .policy = ip_set_create_policy, 2161 }, 2162 [IPSET_CMD_DESTROY] = { 2163 .call = ip_set_destroy, 2164 .attr_count = IPSET_ATTR_CMD_MAX, 2165 .policy = ip_set_setname_policy, 2166 }, 2167 [IPSET_CMD_FLUSH] = { 2168 .call = ip_set_flush, 2169 .attr_count = IPSET_ATTR_CMD_MAX, 2170 .policy = ip_set_setname_policy, 2171 }, 2172 [IPSET_CMD_RENAME] = { 2173 .call = ip_set_rename, 2174 .attr_count = IPSET_ATTR_CMD_MAX, 2175 .policy = ip_set_setname2_policy, 2176 }, 2177 [IPSET_CMD_SWAP] = { 2178 .call = ip_set_swap, 2179 .attr_count = IPSET_ATTR_CMD_MAX, 2180 .policy = ip_set_setname2_policy, 2181 }, 2182 [IPSET_CMD_LIST] = { 2183 .call = ip_set_dump, 2184 .attr_count = IPSET_ATTR_CMD_MAX, 2185 .policy = ip_set_dump_policy, 2186 }, 2187 [IPSET_CMD_SAVE] = { 2188 .call = ip_set_dump, 2189 .attr_count = IPSET_ATTR_CMD_MAX, 2190 .policy = ip_set_setname_policy, 2191 }, 2192 [IPSET_CMD_ADD] = { 2193 .call = ip_set_uadd, 2194 .attr_count = IPSET_ATTR_CMD_MAX, 2195 .policy = ip_set_adt_policy, 2196 }, 2197 [IPSET_CMD_DEL] = { 2198 .call = ip_set_udel, 2199 .attr_count = IPSET_ATTR_CMD_MAX, 2200 .policy = ip_set_adt_policy, 2201 }, 2202 [IPSET_CMD_TEST] = { 2203 .call = ip_set_utest, 2204 .attr_count = IPSET_ATTR_CMD_MAX, 2205 .policy = ip_set_adt_policy, 2206 }, 2207 [IPSET_CMD_HEADER] = { 2208 .call = ip_set_header, 2209 .attr_count = IPSET_ATTR_CMD_MAX, 2210 .policy = ip_set_setname_policy, 2211 }, 2212 [IPSET_CMD_TYPE] = { 2213 .call = ip_set_type, 2214 .attr_count = IPSET_ATTR_CMD_MAX, 2215 .policy = ip_set_type_policy, 2216 }, 2217 [IPSET_CMD_PROTOCOL] = { 2218 .call = ip_set_protocol, 2219 .attr_count = IPSET_ATTR_CMD_MAX, 2220 .policy = ip_set_protocol_policy, 2221 }, 2222 [IPSET_CMD_GET_BYNAME] = { 2223 .call = ip_set_byname, 2224 .attr_count = IPSET_ATTR_CMD_MAX, 2225 .policy = ip_set_setname_policy, 2226 }, 2227 [IPSET_CMD_GET_BYINDEX] = { 2228 .call = ip_set_byindex, 2229 .attr_count = IPSET_ATTR_CMD_MAX, 2230 .policy = ip_set_index_policy, 2231 }, 2232 }; 2233 2234 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = { 2235 .name = "ip_set", 2236 .subsys_id = NFNL_SUBSYS_IPSET, 2237 .cb_count = IPSET_MSG_MAX, 2238 .cb = ip_set_netlink_subsys_cb, 2239 }; 2240 2241 /* Interface to iptables/ip6tables */ 2242 2243 static int 2244 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len) 2245 { 2246 unsigned int *op; 2247 void *data; 2248 int copylen = *len, ret = 0; 2249 struct net *net = sock_net(sk); 2250 struct ip_set_net *inst = ip_set_pernet(net); 2251 2252 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 2253 return -EPERM; 2254 if (optval != SO_IP_SET) 2255 return -EBADF; 2256 if (*len < sizeof(unsigned int)) 2257 return -EINVAL; 2258 2259 data = vmalloc(*len); 2260 if (!data) 2261 return -ENOMEM; 2262 if (copy_from_user(data, user, *len) != 0) { 2263 ret = -EFAULT; 2264 goto done; 2265 } 2266 op = data; 2267 2268 if (*op < IP_SET_OP_VERSION) { 2269 /* Check the version at the beginning of operations */ 2270 struct ip_set_req_version *req_version = data; 2271 2272 if (*len < sizeof(struct ip_set_req_version)) { 2273 ret = -EINVAL; 2274 goto done; 2275 } 2276 2277 if (req_version->version < IPSET_PROTOCOL_MIN) { 2278 ret = -EPROTO; 2279 goto done; 2280 } 2281 } 2282 2283 switch (*op) { 2284 case IP_SET_OP_VERSION: { 2285 struct ip_set_req_version *req_version = data; 2286 2287 if (*len != sizeof(struct ip_set_req_version)) { 2288 ret = -EINVAL; 2289 goto done; 2290 } 2291 2292 req_version->version = IPSET_PROTOCOL; 2293 if (copy_to_user(user, req_version, 2294 sizeof(struct ip_set_req_version))) 2295 ret = -EFAULT; 2296 goto done; 2297 } 2298 case IP_SET_OP_GET_BYNAME: { 2299 struct ip_set_req_get_set *req_get = data; 2300 ip_set_id_t id; 2301 2302 if (*len != sizeof(struct ip_set_req_get_set)) { 2303 ret = -EINVAL; 2304 goto done; 2305 } 2306 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0'; 2307 nfnl_lock(NFNL_SUBSYS_IPSET); 2308 find_set_and_id(inst, req_get->set.name, &id); 2309 req_get->set.index = id; 2310 nfnl_unlock(NFNL_SUBSYS_IPSET); 2311 goto copy; 2312 } 2313 case IP_SET_OP_GET_FNAME: { 2314 struct ip_set_req_get_set_family *req_get = data; 2315 ip_set_id_t id; 2316 2317 if (*len != sizeof(struct ip_set_req_get_set_family)) { 2318 ret = -EINVAL; 2319 goto done; 2320 } 2321 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0'; 2322 nfnl_lock(NFNL_SUBSYS_IPSET); 2323 find_set_and_id(inst, req_get->set.name, &id); 2324 req_get->set.index = id; 2325 if (id != IPSET_INVALID_ID) 2326 req_get->family = ip_set(inst, id)->family; 2327 nfnl_unlock(NFNL_SUBSYS_IPSET); 2328 goto copy; 2329 } 2330 case IP_SET_OP_GET_BYINDEX: { 2331 struct ip_set_req_get_set *req_get = data; 2332 struct ip_set *set; 2333 2334 if (*len != sizeof(struct ip_set_req_get_set) || 2335 req_get->set.index >= inst->ip_set_max) { 2336 ret = -EINVAL; 2337 goto done; 2338 } 2339 nfnl_lock(NFNL_SUBSYS_IPSET); 2340 set = ip_set(inst, req_get->set.index); 2341 ret = strscpy(req_get->set.name, set ? set->name : "", 2342 IPSET_MAXNAMELEN); 2343 nfnl_unlock(NFNL_SUBSYS_IPSET); 2344 if (ret < 0) 2345 goto done; 2346 goto copy; 2347 } 2348 default: 2349 ret = -EBADMSG; 2350 goto done; 2351 } /* end of switch(op) */ 2352 2353 copy: 2354 if (copy_to_user(user, data, copylen)) 2355 ret = -EFAULT; 2356 2357 done: 2358 vfree(data); 2359 if (ret > 0) 2360 ret = 0; 2361 return ret; 2362 } 2363 2364 static struct nf_sockopt_ops so_set __read_mostly = { 2365 .pf = PF_INET, 2366 .get_optmin = SO_IP_SET, 2367 .get_optmax = SO_IP_SET + 1, 2368 .get = ip_set_sockfn_get, 2369 .owner = THIS_MODULE, 2370 }; 2371 2372 static int __net_init 2373 ip_set_net_init(struct net *net) 2374 { 2375 struct ip_set_net *inst = ip_set_pernet(net); 2376 struct ip_set **list; 2377 2378 inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX; 2379 if (inst->ip_set_max >= IPSET_INVALID_ID) 2380 inst->ip_set_max = IPSET_INVALID_ID - 1; 2381 2382 list = kvcalloc(inst->ip_set_max, sizeof(struct ip_set *), GFP_KERNEL); 2383 if (!list) 2384 return -ENOMEM; 2385 inst->is_deleted = false; 2386 inst->is_destroyed = false; 2387 rcu_assign_pointer(inst->ip_set_list, list); 2388 return 0; 2389 } 2390 2391 static void __net_exit 2392 ip_set_net_exit(struct net *net) 2393 { 2394 struct ip_set_net *inst = ip_set_pernet(net); 2395 2396 struct ip_set *set = NULL; 2397 ip_set_id_t i; 2398 2399 inst->is_deleted = true; /* flag for ip_set_nfnl_put */ 2400 2401 nfnl_lock(NFNL_SUBSYS_IPSET); 2402 for (i = 0; i < inst->ip_set_max; i++) { 2403 set = ip_set(inst, i); 2404 if (set) { 2405 ip_set(inst, i) = NULL; 2406 ip_set_destroy_set(set); 2407 } 2408 } 2409 nfnl_unlock(NFNL_SUBSYS_IPSET); 2410 kvfree(rcu_dereference_protected(inst->ip_set_list, 1)); 2411 } 2412 2413 static struct pernet_operations ip_set_net_ops = { 2414 .init = ip_set_net_init, 2415 .exit = ip_set_net_exit, 2416 .id = &ip_set_net_id, 2417 .size = sizeof(struct ip_set_net), 2418 }; 2419 2420 static int __init 2421 ip_set_init(void) 2422 { 2423 int ret = register_pernet_subsys(&ip_set_net_ops); 2424 2425 if (ret) { 2426 pr_err("ip_set: cannot register pernet_subsys.\n"); 2427 return ret; 2428 } 2429 2430 ret = nfnetlink_subsys_register(&ip_set_netlink_subsys); 2431 if (ret != 0) { 2432 pr_err("ip_set: cannot register with nfnetlink.\n"); 2433 unregister_pernet_subsys(&ip_set_net_ops); 2434 return ret; 2435 } 2436 2437 ret = nf_register_sockopt(&so_set); 2438 if (ret != 0) { 2439 pr_err("SO_SET registry failed: %d\n", ret); 2440 nfnetlink_subsys_unregister(&ip_set_netlink_subsys); 2441 unregister_pernet_subsys(&ip_set_net_ops); 2442 return ret; 2443 } 2444 2445 return 0; 2446 } 2447 2448 static void __exit 2449 ip_set_fini(void) 2450 { 2451 nf_unregister_sockopt(&so_set); 2452 nfnetlink_subsys_unregister(&ip_set_netlink_subsys); 2453 2454 unregister_pernet_subsys(&ip_set_net_ops); 2455 pr_debug("these are the famous last words\n"); 2456 } 2457 2458 module_init(ip_set_init); 2459 module_exit(ip_set_fini); 2460 2461 MODULE_DESCRIPTION("ip_set: protocol " __stringify(IPSET_PROTOCOL)); 2462