ip_set_core.c revision 7cf7899d9ee31c88c86ea8459fc4db4bd11cc240
1/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu> 2 * Patrick Schaaf <bof@bof.de> 3 * Copyright (C) 2003-2011 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu> 4 * 5 * This program is free software; you can redistribute it and/or modify 6 * it under the terms of the GNU General Public License version 2 as 7 * published by the Free Software Foundation. 8 */ 9 10/* Kernel module for IP set management */ 11 12#include <linux/init.h> 13#include <linux/module.h> 14#include <linux/moduleparam.h> 15#include <linux/ip.h> 16#include <linux/skbuff.h> 17#include <linux/spinlock.h> 18#include <linux/netlink.h> 19#include <linux/rculist.h> 20#include <net/netlink.h> 21 22#include <linux/netfilter.h> 23#include <linux/netfilter/x_tables.h> 24#include <linux/netfilter/nfnetlink.h> 25#include <linux/netfilter/ipset/ip_set.h> 26 27static LIST_HEAD(ip_set_type_list); /* all registered set types */ 28static DEFINE_MUTEX(ip_set_type_mutex); /* protects ip_set_type_list */ 29static DEFINE_RWLOCK(ip_set_ref_lock); /* protects the set refs */ 30 31static struct ip_set **ip_set_list; /* all individual sets */ 32static ip_set_id_t ip_set_max = CONFIG_IP_SET_MAX; /* max number of sets */ 33 34#define STREQ(a, b) (strncmp(a, b, IPSET_MAXNAMELEN) == 0) 35 36static unsigned int max_sets; 37 38module_param(max_sets, int, 0600); 39MODULE_PARM_DESC(max_sets, "maximal number of sets"); 40MODULE_LICENSE("GPL"); 41MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>"); 42MODULE_DESCRIPTION("core IP set support"); 43MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_IPSET); 44 45/* 46 * The set types are implemented in modules and registered set types 47 * can be found in ip_set_type_list. Adding/deleting types is 48 * serialized by ip_set_type_mutex. 49 */ 50 51static inline void 52ip_set_type_lock(void) 53{ 54 mutex_lock(&ip_set_type_mutex); 55} 56 57static inline void 58ip_set_type_unlock(void) 59{ 60 mutex_unlock(&ip_set_type_mutex); 61} 62 63/* Register and deregister settype */ 64 65static struct ip_set_type * 66find_set_type(const char *name, u8 family, u8 revision) 67{ 68 struct ip_set_type *type; 69 70 list_for_each_entry_rcu(type, &ip_set_type_list, list) 71 if (STREQ(type->name, name) && 72 (type->family == family || type->family == NFPROTO_UNSPEC) && 73 revision >= type->revision_min && 74 revision <= type->revision_max) 75 return type; 76 return NULL; 77} 78 79/* Unlock, try to load a set type module and lock again */ 80static bool 81load_settype(const char *name) 82{ 83 nfnl_unlock(); 84 pr_debug("try to load ip_set_%s\n", name); 85 if (request_module("ip_set_%s", name) < 0) { 86 pr_warning("Can't find ip_set type %s\n", name); 87 nfnl_lock(); 88 return false; 89 } 90 nfnl_lock(); 91 return true; 92} 93 94/* Find a set type and reference it */ 95#define find_set_type_get(name, family, revision, found) \ 96 __find_set_type_get(name, family, revision, found, false) 97 98static int 99__find_set_type_get(const char *name, u8 family, u8 revision, 100 struct ip_set_type **found, bool retry) 101{ 102 struct ip_set_type *type; 103 int err; 104 105 if (retry && !load_settype(name)) 106 return -IPSET_ERR_FIND_TYPE; 107 108 rcu_read_lock(); 109 *found = find_set_type(name, family, revision); 110 if (*found) { 111 err = !try_module_get((*found)->me) ? -EFAULT : 0; 112 goto unlock; 113 } 114 /* Make sure the type is already loaded 115 * but we don't support the revision */ 116 list_for_each_entry_rcu(type, &ip_set_type_list, list) 117 if (STREQ(type->name, name)) { 118 err = -IPSET_ERR_FIND_TYPE; 119 goto unlock; 120 } 121 rcu_read_unlock(); 122 123 return retry ? -IPSET_ERR_FIND_TYPE : 124 __find_set_type_get(name, family, revision, found, true); 125 126unlock: 127 rcu_read_unlock(); 128 return err; 129} 130 131/* Find a given set type by name and family. 132 * If we succeeded, the supported minimal and maximum revisions are 133 * filled out. 134 */ 135#define find_set_type_minmax(name, family, min, max) \ 136 __find_set_type_minmax(name, family, min, max, false) 137 138static int 139__find_set_type_minmax(const char *name, u8 family, u8 *min, u8 *max, 140 bool retry) 141{ 142 struct ip_set_type *type; 143 bool found = false; 144 145 if (retry && !load_settype(name)) 146 return -IPSET_ERR_FIND_TYPE; 147 148 *min = 255; *max = 0; 149 rcu_read_lock(); 150 list_for_each_entry_rcu(type, &ip_set_type_list, list) 151 if (STREQ(type->name, name) && 152 (type->family == family || type->family == NFPROTO_UNSPEC)) { 153 found = true; 154 if (type->revision_min < *min) 155 *min = type->revision_min; 156 if (type->revision_max > *max) 157 *max = type->revision_max; 158 } 159 rcu_read_unlock(); 160 if (found) 161 return 0; 162 163 return retry ? -IPSET_ERR_FIND_TYPE : 164 __find_set_type_minmax(name, family, min, max, true); 165} 166 167#define family_name(f) ((f) == NFPROTO_IPV4 ? "inet" : \ 168 (f) == NFPROTO_IPV6 ? "inet6" : "any") 169 170/* Register a set type structure. The type is identified by 171 * the unique triple of name, family and revision. 172 */ 173int 174ip_set_type_register(struct ip_set_type *type) 175{ 176 int ret = 0; 177 178 if (type->protocol != IPSET_PROTOCOL) { 179 pr_warning("ip_set type %s, family %s, revision %u:%u uses " 180 "wrong protocol version %u (want %u)\n", 181 type->name, family_name(type->family), 182 type->revision_min, type->revision_max, 183 type->protocol, IPSET_PROTOCOL); 184 return -EINVAL; 185 } 186 187 ip_set_type_lock(); 188 if (find_set_type(type->name, type->family, type->revision_min)) { 189 /* Duplicate! */ 190 pr_warning("ip_set type %s, family %s with revision min %u " 191 "already registered!\n", type->name, 192 family_name(type->family), type->revision_min); 193 ret = -EINVAL; 194 goto unlock; 195 } 196 list_add_rcu(&type->list, &ip_set_type_list); 197 pr_debug("type %s, family %s, revision %u:%u registered.\n", 198 type->name, family_name(type->family), 199 type->revision_min, type->revision_max); 200unlock: 201 ip_set_type_unlock(); 202 return ret; 203} 204EXPORT_SYMBOL_GPL(ip_set_type_register); 205 206/* Unregister a set type. There's a small race with ip_set_create */ 207void 208ip_set_type_unregister(struct ip_set_type *type) 209{ 210 ip_set_type_lock(); 211 if (!find_set_type(type->name, type->family, type->revision_min)) { 212 pr_warning("ip_set type %s, family %s with revision min %u " 213 "not registered\n", type->name, 214 family_name(type->family), type->revision_min); 215 goto unlock; 216 } 217 list_del_rcu(&type->list); 218 pr_debug("type %s, family %s with revision min %u unregistered.\n", 219 type->name, family_name(type->family), type->revision_min); 220unlock: 221 ip_set_type_unlock(); 222 223 synchronize_rcu(); 224} 225EXPORT_SYMBOL_GPL(ip_set_type_unregister); 226 227/* Utility functions */ 228void * 229ip_set_alloc(size_t size) 230{ 231 void *members = NULL; 232 233 if (size < KMALLOC_MAX_SIZE) 234 members = kzalloc(size, GFP_KERNEL | __GFP_NOWARN); 235 236 if (members) { 237 pr_debug("%p: allocated with kmalloc\n", members); 238 return members; 239 } 240 241 members = vzalloc(size); 242 if (!members) 243 return NULL; 244 pr_debug("%p: allocated with vmalloc\n", members); 245 246 return members; 247} 248EXPORT_SYMBOL_GPL(ip_set_alloc); 249 250void 251ip_set_free(void *members) 252{ 253 pr_debug("%p: free with %s\n", members, 254 is_vmalloc_addr(members) ? "vfree" : "kfree"); 255 if (is_vmalloc_addr(members)) 256 vfree(members); 257 else 258 kfree(members); 259} 260EXPORT_SYMBOL_GPL(ip_set_free); 261 262static inline bool 263flag_nested(const struct nlattr *nla) 264{ 265 return nla->nla_type & NLA_F_NESTED; 266} 267 268static const struct nla_policy ipaddr_policy[IPSET_ATTR_IPADDR_MAX + 1] = { 269 [IPSET_ATTR_IPADDR_IPV4] = { .type = NLA_U32 }, 270 [IPSET_ATTR_IPADDR_IPV6] = { .type = NLA_BINARY, 271 .len = sizeof(struct in6_addr) }, 272}; 273 274int 275ip_set_get_ipaddr4(struct nlattr *nla, __be32 *ipaddr) 276{ 277 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX+1]; 278 279 if (unlikely(!flag_nested(nla))) 280 return -IPSET_ERR_PROTOCOL; 281 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla, ipaddr_policy)) 282 return -IPSET_ERR_PROTOCOL; 283 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV4))) 284 return -IPSET_ERR_PROTOCOL; 285 286 *ipaddr = nla_get_be32(tb[IPSET_ATTR_IPADDR_IPV4]); 287 return 0; 288} 289EXPORT_SYMBOL_GPL(ip_set_get_ipaddr4); 290 291int 292ip_set_get_ipaddr6(struct nlattr *nla, union nf_inet_addr *ipaddr) 293{ 294 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX+1]; 295 296 if (unlikely(!flag_nested(nla))) 297 return -IPSET_ERR_PROTOCOL; 298 299 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla, ipaddr_policy)) 300 return -IPSET_ERR_PROTOCOL; 301 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV6))) 302 return -IPSET_ERR_PROTOCOL; 303 304 memcpy(ipaddr, nla_data(tb[IPSET_ATTR_IPADDR_IPV6]), 305 sizeof(struct in6_addr)); 306 return 0; 307} 308EXPORT_SYMBOL_GPL(ip_set_get_ipaddr6); 309 310/* 311 * Creating/destroying/renaming/swapping affect the existence and 312 * the properties of a set. All of these can be executed from userspace 313 * only and serialized by the nfnl mutex indirectly from nfnetlink. 314 * 315 * Sets are identified by their index in ip_set_list and the index 316 * is used by the external references (set/SET netfilter modules). 317 * 318 * The set behind an index may change by swapping only, from userspace. 319 */ 320 321static inline void 322__ip_set_get(ip_set_id_t index) 323{ 324 write_lock_bh(&ip_set_ref_lock); 325 ip_set_list[index]->ref++; 326 write_unlock_bh(&ip_set_ref_lock); 327} 328 329static inline void 330__ip_set_put(ip_set_id_t index) 331{ 332 write_lock_bh(&ip_set_ref_lock); 333 BUG_ON(ip_set_list[index]->ref == 0); 334 ip_set_list[index]->ref--; 335 write_unlock_bh(&ip_set_ref_lock); 336} 337 338/* 339 * Add, del and test set entries from kernel. 340 * 341 * The set behind the index must exist and must be referenced 342 * so it can't be destroyed (or changed) under our foot. 343 */ 344 345int 346ip_set_test(ip_set_id_t index, const struct sk_buff *skb, 347 const struct xt_action_param *par, 348 const struct ip_set_adt_opt *opt) 349{ 350 struct ip_set *set = ip_set_list[index]; 351 int ret = 0; 352 353 BUG_ON(set == NULL); 354 pr_debug("set %s, index %u\n", set->name, index); 355 356 if (opt->dim < set->type->dimension || 357 !(opt->family == set->family || set->family == NFPROTO_UNSPEC)) 358 return 0; 359 360 read_lock_bh(&set->lock); 361 ret = set->variant->kadt(set, skb, par, IPSET_TEST, opt); 362 read_unlock_bh(&set->lock); 363 364 if (ret == -EAGAIN) { 365 /* Type requests element to be completed */ 366 pr_debug("element must be competed, ADD is triggered\n"); 367 write_lock_bh(&set->lock); 368 set->variant->kadt(set, skb, par, IPSET_ADD, opt); 369 write_unlock_bh(&set->lock); 370 ret = 1; 371 } 372 373 /* Convert error codes to nomatch */ 374 return (ret < 0 ? 0 : ret); 375} 376EXPORT_SYMBOL_GPL(ip_set_test); 377 378int 379ip_set_add(ip_set_id_t index, const struct sk_buff *skb, 380 const struct xt_action_param *par, 381 const struct ip_set_adt_opt *opt) 382{ 383 struct ip_set *set = ip_set_list[index]; 384 int ret; 385 386 BUG_ON(set == NULL); 387 pr_debug("set %s, index %u\n", set->name, index); 388 389 if (opt->dim < set->type->dimension || 390 !(opt->family == set->family || set->family == NFPROTO_UNSPEC)) 391 return 0; 392 393 write_lock_bh(&set->lock); 394 ret = set->variant->kadt(set, skb, par, IPSET_ADD, opt); 395 write_unlock_bh(&set->lock); 396 397 return ret; 398} 399EXPORT_SYMBOL_GPL(ip_set_add); 400 401int 402ip_set_del(ip_set_id_t index, const struct sk_buff *skb, 403 const struct xt_action_param *par, 404 const struct ip_set_adt_opt *opt) 405{ 406 struct ip_set *set = ip_set_list[index]; 407 int ret = 0; 408 409 BUG_ON(set == NULL); 410 pr_debug("set %s, index %u\n", set->name, index); 411 412 if (opt->dim < set->type->dimension || 413 !(opt->family == set->family || set->family == NFPROTO_UNSPEC)) 414 return 0; 415 416 write_lock_bh(&set->lock); 417 ret = set->variant->kadt(set, skb, par, IPSET_DEL, opt); 418 write_unlock_bh(&set->lock); 419 420 return ret; 421} 422EXPORT_SYMBOL_GPL(ip_set_del); 423 424/* 425 * Find set by name, reference it once. The reference makes sure the 426 * thing pointed to, does not go away under our feet. 427 * 428 */ 429ip_set_id_t 430ip_set_get_byname(const char *name, struct ip_set **set) 431{ 432 ip_set_id_t i, index = IPSET_INVALID_ID; 433 struct ip_set *s; 434 435 for (i = 0; i < ip_set_max; i++) { 436 s = ip_set_list[i]; 437 if (s != NULL && STREQ(s->name, name)) { 438 __ip_set_get(i); 439 index = i; 440 *set = s; 441 } 442 } 443 444 return index; 445} 446EXPORT_SYMBOL_GPL(ip_set_get_byname); 447 448/* 449 * If the given set pointer points to a valid set, decrement 450 * reference count by 1. The caller shall not assume the index 451 * to be valid, after calling this function. 452 * 453 */ 454void 455ip_set_put_byindex(ip_set_id_t index) 456{ 457 if (ip_set_list[index] != NULL) 458 __ip_set_put(index); 459} 460EXPORT_SYMBOL_GPL(ip_set_put_byindex); 461 462/* 463 * Get the name of a set behind a set index. 464 * We assume the set is referenced, so it does exist and 465 * can't be destroyed. The set cannot be renamed due to 466 * the referencing either. 467 * 468 */ 469const char * 470ip_set_name_byindex(ip_set_id_t index) 471{ 472 const struct ip_set *set = ip_set_list[index]; 473 474 BUG_ON(set == NULL); 475 BUG_ON(set->ref == 0); 476 477 /* Referenced, so it's safe */ 478 return set->name; 479} 480EXPORT_SYMBOL_GPL(ip_set_name_byindex); 481 482/* 483 * Routines to call by external subsystems, which do not 484 * call nfnl_lock for us. 485 */ 486 487/* 488 * Find set by name, reference it once. The reference makes sure the 489 * thing pointed to, does not go away under our feet. 490 * 491 * The nfnl mutex is used in the function. 492 */ 493ip_set_id_t 494ip_set_nfnl_get(const char *name) 495{ 496 struct ip_set *s; 497 ip_set_id_t index; 498 499 nfnl_lock(); 500 index = ip_set_get_byname(name, &s); 501 nfnl_unlock(); 502 503 return index; 504} 505EXPORT_SYMBOL_GPL(ip_set_nfnl_get); 506 507/* 508 * Find set by index, reference it once. The reference makes sure the 509 * thing pointed to, does not go away under our feet. 510 * 511 * The nfnl mutex is used in the function. 512 */ 513ip_set_id_t 514ip_set_nfnl_get_byindex(ip_set_id_t index) 515{ 516 if (index > ip_set_max) 517 return IPSET_INVALID_ID; 518 519 nfnl_lock(); 520 if (ip_set_list[index]) 521 __ip_set_get(index); 522 else 523 index = IPSET_INVALID_ID; 524 nfnl_unlock(); 525 526 return index; 527} 528EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex); 529 530/* 531 * If the given set pointer points to a valid set, decrement 532 * reference count by 1. The caller shall not assume the index 533 * to be valid, after calling this function. 534 * 535 * The nfnl mutex is used in the function. 536 */ 537void 538ip_set_nfnl_put(ip_set_id_t index) 539{ 540 nfnl_lock(); 541 ip_set_put_byindex(index); 542 nfnl_unlock(); 543} 544EXPORT_SYMBOL_GPL(ip_set_nfnl_put); 545 546/* 547 * Communication protocol with userspace over netlink. 548 * 549 * The commands are serialized by the nfnl mutex. 550 */ 551 552static inline bool 553protocol_failed(const struct nlattr * const tb[]) 554{ 555 return !tb[IPSET_ATTR_PROTOCOL] || 556 nla_get_u8(tb[IPSET_ATTR_PROTOCOL]) != IPSET_PROTOCOL; 557} 558 559static inline u32 560flag_exist(const struct nlmsghdr *nlh) 561{ 562 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST; 563} 564 565static struct nlmsghdr * 566start_msg(struct sk_buff *skb, u32 pid, u32 seq, unsigned int flags, 567 enum ipset_cmd cmd) 568{ 569 struct nlmsghdr *nlh; 570 struct nfgenmsg *nfmsg; 571 572 nlh = nlmsg_put(skb, pid, seq, cmd | (NFNL_SUBSYS_IPSET << 8), 573 sizeof(*nfmsg), flags); 574 if (nlh == NULL) 575 return NULL; 576 577 nfmsg = nlmsg_data(nlh); 578 nfmsg->nfgen_family = NFPROTO_IPV4; 579 nfmsg->version = NFNETLINK_V0; 580 nfmsg->res_id = 0; 581 582 return nlh; 583} 584 585/* Create a set */ 586 587static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = { 588 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 589 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 590 .len = IPSET_MAXNAMELEN - 1 }, 591 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING, 592 .len = IPSET_MAXNAMELEN - 1}, 593 [IPSET_ATTR_REVISION] = { .type = NLA_U8 }, 594 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 }, 595 [IPSET_ATTR_DATA] = { .type = NLA_NESTED }, 596}; 597 598static ip_set_id_t 599find_set_id(const char *name) 600{ 601 ip_set_id_t i, index = IPSET_INVALID_ID; 602 const struct ip_set *set; 603 604 for (i = 0; index == IPSET_INVALID_ID && i < ip_set_max; i++) { 605 set = ip_set_list[i]; 606 if (set != NULL && STREQ(set->name, name)) 607 index = i; 608 } 609 return index; 610} 611 612static inline struct ip_set * 613find_set(const char *name) 614{ 615 ip_set_id_t index = find_set_id(name); 616 617 return index == IPSET_INVALID_ID ? NULL : ip_set_list[index]; 618} 619 620static int 621find_free_id(const char *name, ip_set_id_t *index, struct ip_set **set) 622{ 623 ip_set_id_t i; 624 625 *index = IPSET_INVALID_ID; 626 for (i = 0; i < ip_set_max; i++) { 627 if (ip_set_list[i] == NULL) { 628 if (*index == IPSET_INVALID_ID) 629 *index = i; 630 } else if (STREQ(name, ip_set_list[i]->name)) { 631 /* Name clash */ 632 *set = ip_set_list[i]; 633 return -EEXIST; 634 } 635 } 636 if (*index == IPSET_INVALID_ID) 637 /* No free slot remained */ 638 return -IPSET_ERR_MAX_SETS; 639 return 0; 640} 641 642static int 643ip_set_create(struct sock *ctnl, struct sk_buff *skb, 644 const struct nlmsghdr *nlh, 645 const struct nlattr * const attr[]) 646{ 647 struct ip_set *set, *clash = NULL; 648 ip_set_id_t index = IPSET_INVALID_ID; 649 struct nlattr *tb[IPSET_ATTR_CREATE_MAX+1] = {}; 650 const char *name, *typename; 651 u8 family, revision; 652 u32 flags = flag_exist(nlh); 653 int ret = 0; 654 655 if (unlikely(protocol_failed(attr) || 656 attr[IPSET_ATTR_SETNAME] == NULL || 657 attr[IPSET_ATTR_TYPENAME] == NULL || 658 attr[IPSET_ATTR_REVISION] == NULL || 659 attr[IPSET_ATTR_FAMILY] == NULL || 660 (attr[IPSET_ATTR_DATA] != NULL && 661 !flag_nested(attr[IPSET_ATTR_DATA])))) 662 return -IPSET_ERR_PROTOCOL; 663 664 name = nla_data(attr[IPSET_ATTR_SETNAME]); 665 typename = nla_data(attr[IPSET_ATTR_TYPENAME]); 666 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]); 667 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]); 668 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n", 669 name, typename, family_name(family), revision); 670 671 /* 672 * First, and without any locks, allocate and initialize 673 * a normal base set structure. 674 */ 675 set = kzalloc(sizeof(struct ip_set), GFP_KERNEL); 676 if (!set) 677 return -ENOMEM; 678 rwlock_init(&set->lock); 679 strlcpy(set->name, name, IPSET_MAXNAMELEN); 680 set->family = family; 681 set->revision = revision; 682 683 /* 684 * Next, check that we know the type, and take 685 * a reference on the type, to make sure it stays available 686 * while constructing our new set. 687 * 688 * After referencing the type, we try to create the type 689 * specific part of the set without holding any locks. 690 */ 691 ret = find_set_type_get(typename, family, revision, &(set->type)); 692 if (ret) 693 goto out; 694 695 /* 696 * Without holding any locks, create private part. 697 */ 698 if (attr[IPSET_ATTR_DATA] && 699 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA], 700 set->type->create_policy)) { 701 ret = -IPSET_ERR_PROTOCOL; 702 goto put_out; 703 } 704 705 ret = set->type->create(set, tb, flags); 706 if (ret != 0) 707 goto put_out; 708 709 /* BTW, ret==0 here. */ 710 711 /* 712 * Here, we have a valid, constructed set and we are protected 713 * by the nfnl mutex. Find the first free index in ip_set_list 714 * and check clashing. 715 */ 716 if ((ret = find_free_id(set->name, &index, &clash)) != 0) { 717 /* If this is the same set and requested, ignore error */ 718 if (ret == -EEXIST && 719 (flags & IPSET_FLAG_EXIST) && 720 STREQ(set->type->name, clash->type->name) && 721 set->type->family == clash->type->family && 722 set->type->revision_min == clash->type->revision_min && 723 set->type->revision_max == clash->type->revision_max && 724 set->variant->same_set(set, clash)) 725 ret = 0; 726 goto cleanup; 727 } 728 729 /* 730 * Finally! Add our shiny new set to the list, and be done. 731 */ 732 pr_debug("create: '%s' created with index %u!\n", set->name, index); 733 ip_set_list[index] = set; 734 735 return ret; 736 737cleanup: 738 set->variant->destroy(set); 739put_out: 740 module_put(set->type->me); 741out: 742 kfree(set); 743 return ret; 744} 745 746/* Destroy sets */ 747 748static const struct nla_policy 749ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = { 750 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 751 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 752 .len = IPSET_MAXNAMELEN - 1 }, 753}; 754 755static void 756ip_set_destroy_set(ip_set_id_t index) 757{ 758 struct ip_set *set = ip_set_list[index]; 759 760 pr_debug("set: %s\n", set->name); 761 ip_set_list[index] = NULL; 762 763 /* Must call it without holding any lock */ 764 set->variant->destroy(set); 765 module_put(set->type->me); 766 kfree(set); 767} 768 769static int 770ip_set_destroy(struct sock *ctnl, struct sk_buff *skb, 771 const struct nlmsghdr *nlh, 772 const struct nlattr * const attr[]) 773{ 774 ip_set_id_t i; 775 int ret = 0; 776 777 if (unlikely(protocol_failed(attr))) 778 return -IPSET_ERR_PROTOCOL; 779 780 /* Commands are serialized and references are 781 * protected by the ip_set_ref_lock. 782 * External systems (i.e. xt_set) must call 783 * ip_set_put|get_nfnl_* functions, that way we 784 * can safely check references here. 785 * 786 * list:set timer can only decrement the reference 787 * counter, so if it's already zero, we can proceed 788 * without holding the lock. 789 */ 790 read_lock_bh(&ip_set_ref_lock); 791 if (!attr[IPSET_ATTR_SETNAME]) { 792 for (i = 0; i < ip_set_max; i++) { 793 if (ip_set_list[i] != NULL && ip_set_list[i]->ref) { 794 ret = -IPSET_ERR_BUSY; 795 goto out; 796 } 797 } 798 read_unlock_bh(&ip_set_ref_lock); 799 for (i = 0; i < ip_set_max; i++) { 800 if (ip_set_list[i] != NULL) 801 ip_set_destroy_set(i); 802 } 803 } else { 804 i = find_set_id(nla_data(attr[IPSET_ATTR_SETNAME])); 805 if (i == IPSET_INVALID_ID) { 806 ret = -ENOENT; 807 goto out; 808 } else if (ip_set_list[i]->ref) { 809 ret = -IPSET_ERR_BUSY; 810 goto out; 811 } 812 read_unlock_bh(&ip_set_ref_lock); 813 814 ip_set_destroy_set(i); 815 } 816 return 0; 817out: 818 read_unlock_bh(&ip_set_ref_lock); 819 return ret; 820} 821 822/* Flush sets */ 823 824static void 825ip_set_flush_set(struct ip_set *set) 826{ 827 pr_debug("set: %s\n", set->name); 828 829 write_lock_bh(&set->lock); 830 set->variant->flush(set); 831 write_unlock_bh(&set->lock); 832} 833 834static int 835ip_set_flush(struct sock *ctnl, struct sk_buff *skb, 836 const struct nlmsghdr *nlh, 837 const struct nlattr * const attr[]) 838{ 839 ip_set_id_t i; 840 841 if (unlikely(protocol_failed(attr))) 842 return -IPSET_ERR_PROTOCOL; 843 844 if (!attr[IPSET_ATTR_SETNAME]) { 845 for (i = 0; i < ip_set_max; i++) 846 if (ip_set_list[i] != NULL) 847 ip_set_flush_set(ip_set_list[i]); 848 } else { 849 i = find_set_id(nla_data(attr[IPSET_ATTR_SETNAME])); 850 if (i == IPSET_INVALID_ID) 851 return -ENOENT; 852 853 ip_set_flush_set(ip_set_list[i]); 854 } 855 856 return 0; 857} 858 859/* Rename a set */ 860 861static const struct nla_policy 862ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = { 863 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 864 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 865 .len = IPSET_MAXNAMELEN - 1 }, 866 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING, 867 .len = IPSET_MAXNAMELEN - 1 }, 868}; 869 870static int 871ip_set_rename(struct sock *ctnl, struct sk_buff *skb, 872 const struct nlmsghdr *nlh, 873 const struct nlattr * const attr[]) 874{ 875 struct ip_set *set; 876 const char *name2; 877 ip_set_id_t i; 878 int ret = 0; 879 880 if (unlikely(protocol_failed(attr) || 881 attr[IPSET_ATTR_SETNAME] == NULL || 882 attr[IPSET_ATTR_SETNAME2] == NULL)) 883 return -IPSET_ERR_PROTOCOL; 884 885 set = find_set(nla_data(attr[IPSET_ATTR_SETNAME])); 886 if (set == NULL) 887 return -ENOENT; 888 889 read_lock_bh(&ip_set_ref_lock); 890 if (set->ref != 0) { 891 ret = -IPSET_ERR_REFERENCED; 892 goto out; 893 } 894 895 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]); 896 for (i = 0; i < ip_set_max; i++) { 897 if (ip_set_list[i] != NULL && 898 STREQ(ip_set_list[i]->name, name2)) { 899 ret = -IPSET_ERR_EXIST_SETNAME2; 900 goto out; 901 } 902 } 903 strncpy(set->name, name2, IPSET_MAXNAMELEN); 904 905out: 906 read_unlock_bh(&ip_set_ref_lock); 907 return ret; 908} 909 910/* Swap two sets so that name/index points to the other. 911 * References and set names are also swapped. 912 * 913 * The commands are serialized by the nfnl mutex and references are 914 * protected by the ip_set_ref_lock. The kernel interfaces 915 * do not hold the mutex but the pointer settings are atomic 916 * so the ip_set_list always contains valid pointers to the sets. 917 */ 918 919static int 920ip_set_swap(struct sock *ctnl, struct sk_buff *skb, 921 const struct nlmsghdr *nlh, 922 const struct nlattr * const attr[]) 923{ 924 struct ip_set *from, *to; 925 ip_set_id_t from_id, to_id; 926 char from_name[IPSET_MAXNAMELEN]; 927 928 if (unlikely(protocol_failed(attr) || 929 attr[IPSET_ATTR_SETNAME] == NULL || 930 attr[IPSET_ATTR_SETNAME2] == NULL)) 931 return -IPSET_ERR_PROTOCOL; 932 933 from_id = find_set_id(nla_data(attr[IPSET_ATTR_SETNAME])); 934 if (from_id == IPSET_INVALID_ID) 935 return -ENOENT; 936 937 to_id = find_set_id(nla_data(attr[IPSET_ATTR_SETNAME2])); 938 if (to_id == IPSET_INVALID_ID) 939 return -IPSET_ERR_EXIST_SETNAME2; 940 941 from = ip_set_list[from_id]; 942 to = ip_set_list[to_id]; 943 944 /* Features must not change. 945 * Not an artificial restriction anymore, as we must prevent 946 * possible loops created by swapping in setlist type of sets. */ 947 if (!(from->type->features == to->type->features && 948 from->type->family == to->type->family)) 949 return -IPSET_ERR_TYPE_MISMATCH; 950 951 strncpy(from_name, from->name, IPSET_MAXNAMELEN); 952 strncpy(from->name, to->name, IPSET_MAXNAMELEN); 953 strncpy(to->name, from_name, IPSET_MAXNAMELEN); 954 955 write_lock_bh(&ip_set_ref_lock); 956 swap(from->ref, to->ref); 957 ip_set_list[from_id] = to; 958 ip_set_list[to_id] = from; 959 write_unlock_bh(&ip_set_ref_lock); 960 961 return 0; 962} 963 964/* List/save set data */ 965 966#define DUMP_INIT 0 967#define DUMP_ALL 1 968#define DUMP_ONE 2 969#define DUMP_LAST 3 970 971#define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF) 972#define DUMP_FLAGS(arg) (((u32)(arg)) >> 16) 973 974static int 975ip_set_dump_done(struct netlink_callback *cb) 976{ 977 if (cb->args[2]) { 978 pr_debug("release set %s\n", ip_set_list[cb->args[1]]->name); 979 ip_set_put_byindex((ip_set_id_t) cb->args[1]); 980 } 981 return 0; 982} 983 984static inline void 985dump_attrs(struct nlmsghdr *nlh) 986{ 987 const struct nlattr *attr; 988 int rem; 989 990 pr_debug("dump nlmsg\n"); 991 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) { 992 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len); 993 } 994} 995 996static int 997dump_init(struct netlink_callback *cb) 998{ 999 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb); 1000 int min_len = NLMSG_SPACE(sizeof(struct nfgenmsg)); 1001 struct nlattr *cda[IPSET_ATTR_CMD_MAX+1]; 1002 struct nlattr *attr = (void *)nlh + min_len; 1003 u32 dump_type; 1004 ip_set_id_t index; 1005 1006 /* Second pass, so parser can't fail */ 1007 nla_parse(cda, IPSET_ATTR_CMD_MAX, 1008 attr, nlh->nlmsg_len - min_len, ip_set_setname_policy); 1009 1010 /* cb->args[0] : dump single set/all sets 1011 * [1] : set index 1012 * [..]: type specific 1013 */ 1014 1015 if (cda[IPSET_ATTR_SETNAME]) { 1016 index = find_set_id(nla_data(cda[IPSET_ATTR_SETNAME])); 1017 if (index == IPSET_INVALID_ID) 1018 return -ENOENT; 1019 1020 dump_type = DUMP_ONE; 1021 cb->args[1] = index; 1022 } else 1023 dump_type = DUMP_ALL; 1024 1025 if (cda[IPSET_ATTR_FLAGS]) { 1026 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]); 1027 dump_type |= (f << 16); 1028 } 1029 cb->args[0] = dump_type; 1030 1031 return 0; 1032} 1033 1034static int 1035ip_set_dump_start(struct sk_buff *skb, struct netlink_callback *cb) 1036{ 1037 ip_set_id_t index = IPSET_INVALID_ID, max; 1038 struct ip_set *set = NULL; 1039 struct nlmsghdr *nlh = NULL; 1040 unsigned int flags = NETLINK_CB(cb->skb).pid ? NLM_F_MULTI : 0; 1041 u32 dump_type, dump_flags; 1042 int ret = 0; 1043 1044 if (!cb->args[0]) { 1045 ret = dump_init(cb); 1046 if (ret < 0) { 1047 nlh = nlmsg_hdr(cb->skb); 1048 /* We have to create and send the error message 1049 * manually :-( */ 1050 if (nlh->nlmsg_flags & NLM_F_ACK) 1051 netlink_ack(cb->skb, nlh, ret); 1052 return ret; 1053 } 1054 } 1055 1056 if (cb->args[1] >= ip_set_max) 1057 goto out; 1058 1059 dump_type = DUMP_TYPE(cb->args[0]); 1060 dump_flags = DUMP_FLAGS(cb->args[0]); 1061 max = dump_type == DUMP_ONE ? cb->args[1] + 1 : ip_set_max; 1062dump_last: 1063 pr_debug("args[0]: %u %u args[1]: %ld\n", 1064 dump_type, dump_flags, cb->args[1]); 1065 for (; cb->args[1] < max; cb->args[1]++) { 1066 index = (ip_set_id_t) cb->args[1]; 1067 set = ip_set_list[index]; 1068 if (set == NULL) { 1069 if (dump_type == DUMP_ONE) { 1070 ret = -ENOENT; 1071 goto out; 1072 } 1073 continue; 1074 } 1075 /* When dumping all sets, we must dump "sorted" 1076 * so that lists (unions of sets) are dumped last. 1077 */ 1078 if (dump_type != DUMP_ONE && 1079 ((dump_type == DUMP_ALL) == 1080 !!(set->type->features & IPSET_DUMP_LAST))) 1081 continue; 1082 pr_debug("List set: %s\n", set->name); 1083 if (!cb->args[2]) { 1084 /* Start listing: make sure set won't be destroyed */ 1085 pr_debug("reference set\n"); 1086 __ip_set_get(index); 1087 } 1088 nlh = start_msg(skb, NETLINK_CB(cb->skb).pid, 1089 cb->nlh->nlmsg_seq, flags, 1090 IPSET_CMD_LIST); 1091 if (!nlh) { 1092 ret = -EMSGSIZE; 1093 goto release_refcount; 1094 } 1095 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) || 1096 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name)) 1097 goto nla_put_failure; 1098 if (dump_flags & IPSET_FLAG_LIST_SETNAME) 1099 goto next_set; 1100 switch (cb->args[2]) { 1101 case 0: 1102 /* Core header data */ 1103 if (nla_put_string(skb, IPSET_ATTR_TYPENAME, 1104 set->type->name) || 1105 nla_put_u8(skb, IPSET_ATTR_FAMILY, 1106 set->family) || 1107 nla_put_u8(skb, IPSET_ATTR_REVISION, 1108 set->revision)) 1109 goto nla_put_failure; 1110 ret = set->variant->head(set, skb); 1111 if (ret < 0) 1112 goto release_refcount; 1113 if (dump_flags & IPSET_FLAG_LIST_HEADER) 1114 goto next_set; 1115 /* Fall through and add elements */ 1116 default: 1117 read_lock_bh(&set->lock); 1118 ret = set->variant->list(set, skb, cb); 1119 read_unlock_bh(&set->lock); 1120 if (!cb->args[2]) 1121 /* Set is done, proceed with next one */ 1122 goto next_set; 1123 goto release_refcount; 1124 } 1125 } 1126 /* If we dump all sets, continue with dumping last ones */ 1127 if (dump_type == DUMP_ALL) { 1128 dump_type = DUMP_LAST; 1129 cb->args[0] = dump_type | (dump_flags << 16); 1130 cb->args[1] = 0; 1131 goto dump_last; 1132 } 1133 goto out; 1134 1135nla_put_failure: 1136 ret = -EFAULT; 1137next_set: 1138 if (dump_type == DUMP_ONE) 1139 cb->args[1] = IPSET_INVALID_ID; 1140 else 1141 cb->args[1]++; 1142release_refcount: 1143 /* If there was an error or set is done, release set */ 1144 if (ret || !cb->args[2]) { 1145 pr_debug("release set %s\n", ip_set_list[index]->name); 1146 ip_set_put_byindex(index); 1147 cb->args[2] = 0; 1148 } 1149out: 1150 if (nlh) { 1151 nlmsg_end(skb, nlh); 1152 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len); 1153 dump_attrs(nlh); 1154 } 1155 1156 return ret < 0 ? ret : skb->len; 1157} 1158 1159static int 1160ip_set_dump(struct sock *ctnl, struct sk_buff *skb, 1161 const struct nlmsghdr *nlh, 1162 const struct nlattr * const attr[]) 1163{ 1164 if (unlikely(protocol_failed(attr))) 1165 return -IPSET_ERR_PROTOCOL; 1166 1167 { 1168 struct netlink_dump_control c = { 1169 .dump = ip_set_dump_start, 1170 .done = ip_set_dump_done, 1171 }; 1172 return netlink_dump_start(ctnl, skb, nlh, &c); 1173 } 1174} 1175 1176/* Add, del and test */ 1177 1178static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = { 1179 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1180 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 1181 .len = IPSET_MAXNAMELEN - 1 }, 1182 [IPSET_ATTR_LINENO] = { .type = NLA_U32 }, 1183 [IPSET_ATTR_DATA] = { .type = NLA_NESTED }, 1184 [IPSET_ATTR_ADT] = { .type = NLA_NESTED }, 1185}; 1186 1187static int 1188call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set, 1189 struct nlattr *tb[], enum ipset_adt adt, 1190 u32 flags, bool use_lineno) 1191{ 1192 int ret; 1193 u32 lineno = 0; 1194 bool eexist = flags & IPSET_FLAG_EXIST, retried = false; 1195 1196 do { 1197 write_lock_bh(&set->lock); 1198 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried); 1199 write_unlock_bh(&set->lock); 1200 retried = true; 1201 } while (ret == -EAGAIN && 1202 set->variant->resize && 1203 (ret = set->variant->resize(set, retried)) == 0); 1204 1205 if (!ret || (ret == -IPSET_ERR_EXIST && eexist)) 1206 return 0; 1207 if (lineno && use_lineno) { 1208 /* Error in restore/batch mode: send back lineno */ 1209 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb); 1210 struct sk_buff *skb2; 1211 struct nlmsgerr *errmsg; 1212 size_t payload = sizeof(*errmsg) + nlmsg_len(nlh); 1213 int min_len = NLMSG_SPACE(sizeof(struct nfgenmsg)); 1214 struct nlattr *cda[IPSET_ATTR_CMD_MAX+1]; 1215 struct nlattr *cmdattr; 1216 u32 *errline; 1217 1218 skb2 = nlmsg_new(payload, GFP_KERNEL); 1219 if (skb2 == NULL) 1220 return -ENOMEM; 1221 rep = __nlmsg_put(skb2, NETLINK_CB(skb).pid, 1222 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0); 1223 errmsg = nlmsg_data(rep); 1224 errmsg->error = ret; 1225 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len); 1226 cmdattr = (void *)&errmsg->msg + min_len; 1227 1228 nla_parse(cda, IPSET_ATTR_CMD_MAX, 1229 cmdattr, nlh->nlmsg_len - min_len, 1230 ip_set_adt_policy); 1231 1232 errline = nla_data(cda[IPSET_ATTR_LINENO]); 1233 1234 *errline = lineno; 1235 1236 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT); 1237 /* Signal netlink not to send its ACK/errmsg. */ 1238 return -EINTR; 1239 } 1240 1241 return ret; 1242} 1243 1244static int 1245ip_set_uadd(struct sock *ctnl, struct sk_buff *skb, 1246 const struct nlmsghdr *nlh, 1247 const struct nlattr * const attr[]) 1248{ 1249 struct ip_set *set; 1250 struct nlattr *tb[IPSET_ATTR_ADT_MAX+1] = {}; 1251 const struct nlattr *nla; 1252 u32 flags = flag_exist(nlh); 1253 bool use_lineno; 1254 int ret = 0; 1255 1256 if (unlikely(protocol_failed(attr) || 1257 attr[IPSET_ATTR_SETNAME] == NULL || 1258 !((attr[IPSET_ATTR_DATA] != NULL) ^ 1259 (attr[IPSET_ATTR_ADT] != NULL)) || 1260 (attr[IPSET_ATTR_DATA] != NULL && 1261 !flag_nested(attr[IPSET_ATTR_DATA])) || 1262 (attr[IPSET_ATTR_ADT] != NULL && 1263 (!flag_nested(attr[IPSET_ATTR_ADT]) || 1264 attr[IPSET_ATTR_LINENO] == NULL)))) 1265 return -IPSET_ERR_PROTOCOL; 1266 1267 set = find_set(nla_data(attr[IPSET_ATTR_SETNAME])); 1268 if (set == NULL) 1269 return -ENOENT; 1270 1271 use_lineno = !!attr[IPSET_ATTR_LINENO]; 1272 if (attr[IPSET_ATTR_DATA]) { 1273 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, 1274 attr[IPSET_ATTR_DATA], 1275 set->type->adt_policy)) 1276 return -IPSET_ERR_PROTOCOL; 1277 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD, flags, 1278 use_lineno); 1279 } else { 1280 int nla_rem; 1281 1282 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) { 1283 memset(tb, 0, sizeof(tb)); 1284 if (nla_type(nla) != IPSET_ATTR_DATA || 1285 !flag_nested(nla) || 1286 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla, 1287 set->type->adt_policy)) 1288 return -IPSET_ERR_PROTOCOL; 1289 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD, 1290 flags, use_lineno); 1291 if (ret < 0) 1292 return ret; 1293 } 1294 } 1295 return ret; 1296} 1297 1298static int 1299ip_set_udel(struct sock *ctnl, struct sk_buff *skb, 1300 const struct nlmsghdr *nlh, 1301 const struct nlattr * const attr[]) 1302{ 1303 struct ip_set *set; 1304 struct nlattr *tb[IPSET_ATTR_ADT_MAX+1] = {}; 1305 const struct nlattr *nla; 1306 u32 flags = flag_exist(nlh); 1307 bool use_lineno; 1308 int ret = 0; 1309 1310 if (unlikely(protocol_failed(attr) || 1311 attr[IPSET_ATTR_SETNAME] == NULL || 1312 !((attr[IPSET_ATTR_DATA] != NULL) ^ 1313 (attr[IPSET_ATTR_ADT] != NULL)) || 1314 (attr[IPSET_ATTR_DATA] != NULL && 1315 !flag_nested(attr[IPSET_ATTR_DATA])) || 1316 (attr[IPSET_ATTR_ADT] != NULL && 1317 (!flag_nested(attr[IPSET_ATTR_ADT]) || 1318 attr[IPSET_ATTR_LINENO] == NULL)))) 1319 return -IPSET_ERR_PROTOCOL; 1320 1321 set = find_set(nla_data(attr[IPSET_ATTR_SETNAME])); 1322 if (set == NULL) 1323 return -ENOENT; 1324 1325 use_lineno = !!attr[IPSET_ATTR_LINENO]; 1326 if (attr[IPSET_ATTR_DATA]) { 1327 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, 1328 attr[IPSET_ATTR_DATA], 1329 set->type->adt_policy)) 1330 return -IPSET_ERR_PROTOCOL; 1331 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL, flags, 1332 use_lineno); 1333 } else { 1334 int nla_rem; 1335 1336 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) { 1337 memset(tb, 0, sizeof(*tb)); 1338 if (nla_type(nla) != IPSET_ATTR_DATA || 1339 !flag_nested(nla) || 1340 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla, 1341 set->type->adt_policy)) 1342 return -IPSET_ERR_PROTOCOL; 1343 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL, 1344 flags, use_lineno); 1345 if (ret < 0) 1346 return ret; 1347 } 1348 } 1349 return ret; 1350} 1351 1352static int 1353ip_set_utest(struct sock *ctnl, struct sk_buff *skb, 1354 const struct nlmsghdr *nlh, 1355 const struct nlattr * const attr[]) 1356{ 1357 struct ip_set *set; 1358 struct nlattr *tb[IPSET_ATTR_ADT_MAX+1] = {}; 1359 int ret = 0; 1360 1361 if (unlikely(protocol_failed(attr) || 1362 attr[IPSET_ATTR_SETNAME] == NULL || 1363 attr[IPSET_ATTR_DATA] == NULL || 1364 !flag_nested(attr[IPSET_ATTR_DATA]))) 1365 return -IPSET_ERR_PROTOCOL; 1366 1367 set = find_set(nla_data(attr[IPSET_ATTR_SETNAME])); 1368 if (set == NULL) 1369 return -ENOENT; 1370 1371 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA], 1372 set->type->adt_policy)) 1373 return -IPSET_ERR_PROTOCOL; 1374 1375 read_lock_bh(&set->lock); 1376 ret = set->variant->uadt(set, tb, IPSET_TEST, NULL, 0, 0); 1377 read_unlock_bh(&set->lock); 1378 /* Userspace can't trigger element to be re-added */ 1379 if (ret == -EAGAIN) 1380 ret = 1; 1381 1382 return ret < 0 ? ret : ret > 0 ? 0 : -IPSET_ERR_EXIST; 1383} 1384 1385/* Get headed data of a set */ 1386 1387static int 1388ip_set_header(struct sock *ctnl, struct sk_buff *skb, 1389 const struct nlmsghdr *nlh, 1390 const struct nlattr * const attr[]) 1391{ 1392 const struct ip_set *set; 1393 struct sk_buff *skb2; 1394 struct nlmsghdr *nlh2; 1395 ip_set_id_t index; 1396 int ret = 0; 1397 1398 if (unlikely(protocol_failed(attr) || 1399 attr[IPSET_ATTR_SETNAME] == NULL)) 1400 return -IPSET_ERR_PROTOCOL; 1401 1402 index = find_set_id(nla_data(attr[IPSET_ATTR_SETNAME])); 1403 if (index == IPSET_INVALID_ID) 1404 return -ENOENT; 1405 set = ip_set_list[index]; 1406 1407 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 1408 if (skb2 == NULL) 1409 return -ENOMEM; 1410 1411 nlh2 = start_msg(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq, 0, 1412 IPSET_CMD_HEADER); 1413 if (!nlh2) 1414 goto nlmsg_failure; 1415 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) || 1416 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) || 1417 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) || 1418 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) || 1419 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision)) 1420 goto nla_put_failure; 1421 nlmsg_end(skb2, nlh2); 1422 1423 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT); 1424 if (ret < 0) 1425 return ret; 1426 1427 return 0; 1428 1429nla_put_failure: 1430 nlmsg_cancel(skb2, nlh2); 1431nlmsg_failure: 1432 kfree_skb(skb2); 1433 return -EMSGSIZE; 1434} 1435 1436/* Get type data */ 1437 1438static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = { 1439 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1440 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING, 1441 .len = IPSET_MAXNAMELEN - 1 }, 1442 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 }, 1443}; 1444 1445static int 1446ip_set_type(struct sock *ctnl, struct sk_buff *skb, 1447 const struct nlmsghdr *nlh, 1448 const struct nlattr * const attr[]) 1449{ 1450 struct sk_buff *skb2; 1451 struct nlmsghdr *nlh2; 1452 u8 family, min, max; 1453 const char *typename; 1454 int ret = 0; 1455 1456 if (unlikely(protocol_failed(attr) || 1457 attr[IPSET_ATTR_TYPENAME] == NULL || 1458 attr[IPSET_ATTR_FAMILY] == NULL)) 1459 return -IPSET_ERR_PROTOCOL; 1460 1461 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]); 1462 typename = nla_data(attr[IPSET_ATTR_TYPENAME]); 1463 ret = find_set_type_minmax(typename, family, &min, &max); 1464 if (ret) 1465 return ret; 1466 1467 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 1468 if (skb2 == NULL) 1469 return -ENOMEM; 1470 1471 nlh2 = start_msg(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq, 0, 1472 IPSET_CMD_TYPE); 1473 if (!nlh2) 1474 goto nlmsg_failure; 1475 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) || 1476 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) || 1477 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) || 1478 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) || 1479 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min)) 1480 goto nla_put_failure; 1481 nlmsg_end(skb2, nlh2); 1482 1483 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len); 1484 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT); 1485 if (ret < 0) 1486 return ret; 1487 1488 return 0; 1489 1490nla_put_failure: 1491 nlmsg_cancel(skb2, nlh2); 1492nlmsg_failure: 1493 kfree_skb(skb2); 1494 return -EMSGSIZE; 1495} 1496 1497/* Get protocol version */ 1498 1499static const struct nla_policy 1500ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = { 1501 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1502}; 1503 1504static int 1505ip_set_protocol(struct sock *ctnl, struct sk_buff *skb, 1506 const struct nlmsghdr *nlh, 1507 const struct nlattr * const attr[]) 1508{ 1509 struct sk_buff *skb2; 1510 struct nlmsghdr *nlh2; 1511 int ret = 0; 1512 1513 if (unlikely(attr[IPSET_ATTR_PROTOCOL] == NULL)) 1514 return -IPSET_ERR_PROTOCOL; 1515 1516 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 1517 if (skb2 == NULL) 1518 return -ENOMEM; 1519 1520 nlh2 = start_msg(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq, 0, 1521 IPSET_CMD_PROTOCOL); 1522 if (!nlh2) 1523 goto nlmsg_failure; 1524 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL)) 1525 goto nla_put_failure; 1526 nlmsg_end(skb2, nlh2); 1527 1528 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT); 1529 if (ret < 0) 1530 return ret; 1531 1532 return 0; 1533 1534nla_put_failure: 1535 nlmsg_cancel(skb2, nlh2); 1536nlmsg_failure: 1537 kfree_skb(skb2); 1538 return -EMSGSIZE; 1539} 1540 1541static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = { 1542 [IPSET_CMD_CREATE] = { 1543 .call = ip_set_create, 1544 .attr_count = IPSET_ATTR_CMD_MAX, 1545 .policy = ip_set_create_policy, 1546 }, 1547 [IPSET_CMD_DESTROY] = { 1548 .call = ip_set_destroy, 1549 .attr_count = IPSET_ATTR_CMD_MAX, 1550 .policy = ip_set_setname_policy, 1551 }, 1552 [IPSET_CMD_FLUSH] = { 1553 .call = ip_set_flush, 1554 .attr_count = IPSET_ATTR_CMD_MAX, 1555 .policy = ip_set_setname_policy, 1556 }, 1557 [IPSET_CMD_RENAME] = { 1558 .call = ip_set_rename, 1559 .attr_count = IPSET_ATTR_CMD_MAX, 1560 .policy = ip_set_setname2_policy, 1561 }, 1562 [IPSET_CMD_SWAP] = { 1563 .call = ip_set_swap, 1564 .attr_count = IPSET_ATTR_CMD_MAX, 1565 .policy = ip_set_setname2_policy, 1566 }, 1567 [IPSET_CMD_LIST] = { 1568 .call = ip_set_dump, 1569 .attr_count = IPSET_ATTR_CMD_MAX, 1570 .policy = ip_set_setname_policy, 1571 }, 1572 [IPSET_CMD_SAVE] = { 1573 .call = ip_set_dump, 1574 .attr_count = IPSET_ATTR_CMD_MAX, 1575 .policy = ip_set_setname_policy, 1576 }, 1577 [IPSET_CMD_ADD] = { 1578 .call = ip_set_uadd, 1579 .attr_count = IPSET_ATTR_CMD_MAX, 1580 .policy = ip_set_adt_policy, 1581 }, 1582 [IPSET_CMD_DEL] = { 1583 .call = ip_set_udel, 1584 .attr_count = IPSET_ATTR_CMD_MAX, 1585 .policy = ip_set_adt_policy, 1586 }, 1587 [IPSET_CMD_TEST] = { 1588 .call = ip_set_utest, 1589 .attr_count = IPSET_ATTR_CMD_MAX, 1590 .policy = ip_set_adt_policy, 1591 }, 1592 [IPSET_CMD_HEADER] = { 1593 .call = ip_set_header, 1594 .attr_count = IPSET_ATTR_CMD_MAX, 1595 .policy = ip_set_setname_policy, 1596 }, 1597 [IPSET_CMD_TYPE] = { 1598 .call = ip_set_type, 1599 .attr_count = IPSET_ATTR_CMD_MAX, 1600 .policy = ip_set_type_policy, 1601 }, 1602 [IPSET_CMD_PROTOCOL] = { 1603 .call = ip_set_protocol, 1604 .attr_count = IPSET_ATTR_CMD_MAX, 1605 .policy = ip_set_protocol_policy, 1606 }, 1607}; 1608 1609static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = { 1610 .name = "ip_set", 1611 .subsys_id = NFNL_SUBSYS_IPSET, 1612 .cb_count = IPSET_MSG_MAX, 1613 .cb = ip_set_netlink_subsys_cb, 1614}; 1615 1616/* Interface to iptables/ip6tables */ 1617 1618static int 1619ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len) 1620{ 1621 unsigned *op; 1622 void *data; 1623 int copylen = *len, ret = 0; 1624 1625 if (!capable(CAP_NET_ADMIN)) 1626 return -EPERM; 1627 if (optval != SO_IP_SET) 1628 return -EBADF; 1629 if (*len < sizeof(unsigned)) 1630 return -EINVAL; 1631 1632 data = vmalloc(*len); 1633 if (!data) 1634 return -ENOMEM; 1635 if (copy_from_user(data, user, *len) != 0) { 1636 ret = -EFAULT; 1637 goto done; 1638 } 1639 op = (unsigned *) data; 1640 1641 if (*op < IP_SET_OP_VERSION) { 1642 /* Check the version at the beginning of operations */ 1643 struct ip_set_req_version *req_version = data; 1644 if (req_version->version != IPSET_PROTOCOL) { 1645 ret = -EPROTO; 1646 goto done; 1647 } 1648 } 1649 1650 switch (*op) { 1651 case IP_SET_OP_VERSION: { 1652 struct ip_set_req_version *req_version = data; 1653 1654 if (*len != sizeof(struct ip_set_req_version)) { 1655 ret = -EINVAL; 1656 goto done; 1657 } 1658 1659 req_version->version = IPSET_PROTOCOL; 1660 ret = copy_to_user(user, req_version, 1661 sizeof(struct ip_set_req_version)); 1662 goto done; 1663 } 1664 case IP_SET_OP_GET_BYNAME: { 1665 struct ip_set_req_get_set *req_get = data; 1666 1667 if (*len != sizeof(struct ip_set_req_get_set)) { 1668 ret = -EINVAL; 1669 goto done; 1670 } 1671 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0'; 1672 nfnl_lock(); 1673 req_get->set.index = find_set_id(req_get->set.name); 1674 nfnl_unlock(); 1675 goto copy; 1676 } 1677 case IP_SET_OP_GET_BYINDEX: { 1678 struct ip_set_req_get_set *req_get = data; 1679 1680 if (*len != sizeof(struct ip_set_req_get_set) || 1681 req_get->set.index >= ip_set_max) { 1682 ret = -EINVAL; 1683 goto done; 1684 } 1685 nfnl_lock(); 1686 strncpy(req_get->set.name, 1687 ip_set_list[req_get->set.index] 1688 ? ip_set_list[req_get->set.index]->name : "", 1689 IPSET_MAXNAMELEN); 1690 nfnl_unlock(); 1691 goto copy; 1692 } 1693 default: 1694 ret = -EBADMSG; 1695 goto done; 1696 } /* end of switch(op) */ 1697 1698copy: 1699 ret = copy_to_user(user, data, copylen); 1700 1701done: 1702 vfree(data); 1703 if (ret > 0) 1704 ret = 0; 1705 return ret; 1706} 1707 1708static struct nf_sockopt_ops so_set __read_mostly = { 1709 .pf = PF_INET, 1710 .get_optmin = SO_IP_SET, 1711 .get_optmax = SO_IP_SET + 1, 1712 .get = &ip_set_sockfn_get, 1713 .owner = THIS_MODULE, 1714}; 1715 1716static int __init 1717ip_set_init(void) 1718{ 1719 int ret; 1720 1721 if (max_sets) 1722 ip_set_max = max_sets; 1723 if (ip_set_max >= IPSET_INVALID_ID) 1724 ip_set_max = IPSET_INVALID_ID - 1; 1725 1726 ip_set_list = kzalloc(sizeof(struct ip_set *) * ip_set_max, 1727 GFP_KERNEL); 1728 if (!ip_set_list) 1729 return -ENOMEM; 1730 1731 ret = nfnetlink_subsys_register(&ip_set_netlink_subsys); 1732 if (ret != 0) { 1733 pr_err("ip_set: cannot register with nfnetlink.\n"); 1734 kfree(ip_set_list); 1735 return ret; 1736 } 1737 ret = nf_register_sockopt(&so_set); 1738 if (ret != 0) { 1739 pr_err("SO_SET registry failed: %d\n", ret); 1740 nfnetlink_subsys_unregister(&ip_set_netlink_subsys); 1741 kfree(ip_set_list); 1742 return ret; 1743 } 1744 1745 pr_notice("ip_set: protocol %u\n", IPSET_PROTOCOL); 1746 return 0; 1747} 1748 1749static void __exit 1750ip_set_fini(void) 1751{ 1752 /* There can't be any existing set */ 1753 nf_unregister_sockopt(&so_set); 1754 nfnetlink_subsys_unregister(&ip_set_netlink_subsys); 1755 kfree(ip_set_list); 1756 pr_debug("these are the famous last words\n"); 1757} 1758 1759module_init(ip_set_init); 1760module_exit(ip_set_fini); 1761