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