ip_set_core.c revision f1e00b39797944bf25addaf543839feeb25fbdc5
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 <linux/version.h> 21#include <net/netlink.h> 22 23#include <linux/netfilter.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 == AF_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 int 81try_to_load_type(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 -IPSET_ERR_FIND_TYPE; 89 } 90 nfnl_lock(); 91 return -EAGAIN; 92} 93 94/* Find a set type and reference it */ 95static int 96find_set_type_get(const char *name, u8 family, u8 revision, 97 struct ip_set_type **found) 98{ 99 struct ip_set_type *type; 100 int err; 101 102 rcu_read_lock(); 103 *found = find_set_type(name, family, revision); 104 if (*found) { 105 err = !try_module_get((*found)->me) ? -EFAULT : 0; 106 goto unlock; 107 } 108 /* Make sure the type is loaded but we don't support the revision */ 109 list_for_each_entry_rcu(type, &ip_set_type_list, list) 110 if (STREQ(type->name, name)) { 111 err = -IPSET_ERR_FIND_TYPE; 112 goto unlock; 113 } 114 rcu_read_unlock(); 115 116 return try_to_load_type(name); 117 118unlock: 119 rcu_read_unlock(); 120 return err; 121} 122 123/* Find a given set type by name and family. 124 * If we succeeded, the supported minimal and maximum revisions are 125 * filled out. 126 */ 127static int 128find_set_type_minmax(const char *name, u8 family, u8 *min, u8 *max) 129{ 130 struct ip_set_type *type; 131 bool found = false; 132 133 *min = 255; *max = 0; 134 rcu_read_lock(); 135 list_for_each_entry_rcu(type, &ip_set_type_list, list) 136 if (STREQ(type->name, name) && 137 (type->family == family || type->family == AF_UNSPEC)) { 138 found = true; 139 if (type->revision_min < *min) 140 *min = type->revision_min; 141 if (type->revision_max > *max) 142 *max = type->revision_max; 143 } 144 rcu_read_unlock(); 145 if (found) 146 return 0; 147 148 return try_to_load_type(name); 149} 150 151#define family_name(f) ((f) == AF_INET ? "inet" : \ 152 (f) == AF_INET6 ? "inet6" : "any") 153 154/* Register a set type structure. The type is identified by 155 * the unique triple of name, family and revision. 156 */ 157int 158ip_set_type_register(struct ip_set_type *type) 159{ 160 int ret = 0; 161 162 if (type->protocol != IPSET_PROTOCOL) { 163 pr_warning("ip_set type %s, family %s, revision %u:%u uses " 164 "wrong protocol version %u (want %u)\n", 165 type->name, family_name(type->family), 166 type->revision_min, type->revision_max, 167 type->protocol, IPSET_PROTOCOL); 168 return -EINVAL; 169 } 170 171 ip_set_type_lock(); 172 if (find_set_type(type->name, type->family, type->revision_min)) { 173 /* Duplicate! */ 174 pr_warning("ip_set type %s, family %s with revision min %u " 175 "already registered!\n", type->name, 176 family_name(type->family), type->revision_min); 177 ret = -EINVAL; 178 goto unlock; 179 } 180 list_add_rcu(&type->list, &ip_set_type_list); 181 pr_debug("type %s, family %s, revision %u:%u registered.\n", 182 type->name, family_name(type->family), 183 type->revision_min, type->revision_max); 184unlock: 185 ip_set_type_unlock(); 186 return ret; 187} 188EXPORT_SYMBOL_GPL(ip_set_type_register); 189 190/* Unregister a set type. There's a small race with ip_set_create */ 191void 192ip_set_type_unregister(struct ip_set_type *type) 193{ 194 ip_set_type_lock(); 195 if (!find_set_type(type->name, type->family, type->revision_min)) { 196 pr_warning("ip_set type %s, family %s with revision min %u " 197 "not registered\n", type->name, 198 family_name(type->family), type->revision_min); 199 goto unlock; 200 } 201 list_del_rcu(&type->list); 202 pr_debug("type %s, family %s with revision min %u unregistered.\n", 203 type->name, family_name(type->family), type->revision_min); 204unlock: 205 ip_set_type_unlock(); 206 207 synchronize_rcu(); 208} 209EXPORT_SYMBOL_GPL(ip_set_type_unregister); 210 211/* Utility functions */ 212void * 213ip_set_alloc(size_t size) 214{ 215 void *members = NULL; 216 217 if (size < KMALLOC_MAX_SIZE) 218 members = kzalloc(size, GFP_KERNEL | __GFP_NOWARN); 219 220 if (members) { 221 pr_debug("%p: allocated with kmalloc\n", members); 222 return members; 223 } 224 225 members = vzalloc(size); 226 if (!members) 227 return NULL; 228 pr_debug("%p: allocated with vmalloc\n", members); 229 230 return members; 231} 232EXPORT_SYMBOL_GPL(ip_set_alloc); 233 234void 235ip_set_free(void *members) 236{ 237 pr_debug("%p: free with %s\n", members, 238 is_vmalloc_addr(members) ? "vfree" : "kfree"); 239 if (is_vmalloc_addr(members)) 240 vfree(members); 241 else 242 kfree(members); 243} 244EXPORT_SYMBOL_GPL(ip_set_free); 245 246static inline bool 247flag_nested(const struct nlattr *nla) 248{ 249 return nla->nla_type & NLA_F_NESTED; 250} 251 252static const struct nla_policy ipaddr_policy[IPSET_ATTR_IPADDR_MAX + 1] = { 253 [IPSET_ATTR_IPADDR_IPV4] = { .type = NLA_U32 }, 254 [IPSET_ATTR_IPADDR_IPV6] = { .type = NLA_BINARY, 255 .len = sizeof(struct in6_addr) }, 256}; 257 258int 259ip_set_get_ipaddr4(struct nlattr *nla, __be32 *ipaddr) 260{ 261 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX+1]; 262 263 if (unlikely(!flag_nested(nla))) 264 return -IPSET_ERR_PROTOCOL; 265 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla, ipaddr_policy)) 266 return -IPSET_ERR_PROTOCOL; 267 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV4))) 268 return -IPSET_ERR_PROTOCOL; 269 270 *ipaddr = nla_get_be32(tb[IPSET_ATTR_IPADDR_IPV4]); 271 return 0; 272} 273EXPORT_SYMBOL_GPL(ip_set_get_ipaddr4); 274 275int 276ip_set_get_ipaddr6(struct nlattr *nla, union nf_inet_addr *ipaddr) 277{ 278 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX+1]; 279 280 if (unlikely(!flag_nested(nla))) 281 return -IPSET_ERR_PROTOCOL; 282 283 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla, ipaddr_policy)) 284 return -IPSET_ERR_PROTOCOL; 285 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV6))) 286 return -IPSET_ERR_PROTOCOL; 287 288 memcpy(ipaddr, nla_data(tb[IPSET_ATTR_IPADDR_IPV6]), 289 sizeof(struct in6_addr)); 290 return 0; 291} 292EXPORT_SYMBOL_GPL(ip_set_get_ipaddr6); 293 294/* 295 * Creating/destroying/renaming/swapping affect the existence and 296 * the properties of a set. All of these can be executed from userspace 297 * only and serialized by the nfnl mutex indirectly from nfnetlink. 298 * 299 * Sets are identified by their index in ip_set_list and the index 300 * is used by the external references (set/SET netfilter modules). 301 * 302 * The set behind an index may change by swapping only, from userspace. 303 */ 304 305static inline void 306__ip_set_get(ip_set_id_t index) 307{ 308 write_lock_bh(&ip_set_ref_lock); 309 ip_set_list[index]->ref++; 310 write_unlock_bh(&ip_set_ref_lock); 311} 312 313static inline void 314__ip_set_put(ip_set_id_t index) 315{ 316 write_lock_bh(&ip_set_ref_lock); 317 BUG_ON(ip_set_list[index]->ref == 0); 318 ip_set_list[index]->ref--; 319 write_unlock_bh(&ip_set_ref_lock); 320} 321 322/* 323 * Add, del and test set entries from kernel. 324 * 325 * The set behind the index must exist and must be referenced 326 * so it can't be destroyed (or changed) under our foot. 327 */ 328 329int 330ip_set_test(ip_set_id_t index, const struct sk_buff *skb, 331 const struct ip_set_adt_opt *opt) 332{ 333 struct ip_set *set = ip_set_list[index]; 334 int ret = 0; 335 336 BUG_ON(set == NULL); 337 pr_debug("set %s, index %u\n", set->name, index); 338 339 if (opt->dim < set->type->dimension || 340 !(opt->family == set->family || set->family == AF_UNSPEC)) 341 return 0; 342 343 read_lock_bh(&set->lock); 344 ret = set->variant->kadt(set, skb, IPSET_TEST, opt); 345 read_unlock_bh(&set->lock); 346 347 if (ret == -EAGAIN) { 348 /* Type requests element to be completed */ 349 pr_debug("element must be competed, ADD is triggered\n"); 350 write_lock_bh(&set->lock); 351 set->variant->kadt(set, skb, IPSET_ADD, opt); 352 write_unlock_bh(&set->lock); 353 ret = 1; 354 } 355 356 /* Convert error codes to nomatch */ 357 return (ret < 0 ? 0 : ret); 358} 359EXPORT_SYMBOL_GPL(ip_set_test); 360 361int 362ip_set_add(ip_set_id_t index, const struct sk_buff *skb, 363 const struct ip_set_adt_opt *opt) 364{ 365 struct ip_set *set = ip_set_list[index]; 366 int ret; 367 368 BUG_ON(set == NULL); 369 pr_debug("set %s, index %u\n", set->name, index); 370 371 if (opt->dim < set->type->dimension || 372 !(opt->family == set->family || set->family == AF_UNSPEC)) 373 return 0; 374 375 write_lock_bh(&set->lock); 376 ret = set->variant->kadt(set, skb, IPSET_ADD, opt); 377 write_unlock_bh(&set->lock); 378 379 return ret; 380} 381EXPORT_SYMBOL_GPL(ip_set_add); 382 383int 384ip_set_del(ip_set_id_t index, const struct sk_buff *skb, 385 const struct ip_set_adt_opt *opt) 386{ 387 struct ip_set *set = ip_set_list[index]; 388 int ret = 0; 389 390 BUG_ON(set == NULL); 391 pr_debug("set %s, index %u\n", set->name, index); 392 393 if (opt->dim < set->type->dimension || 394 !(opt->family == set->family || set->family == AF_UNSPEC)) 395 return 0; 396 397 write_lock_bh(&set->lock); 398 ret = set->variant->kadt(set, skb, IPSET_DEL, opt); 399 write_unlock_bh(&set->lock); 400 401 return ret; 402} 403EXPORT_SYMBOL_GPL(ip_set_del); 404 405/* 406 * Find set by name, reference it once. The reference makes sure the 407 * thing pointed to, does not go away under our feet. 408 * 409 */ 410ip_set_id_t 411ip_set_get_byname(const char *name, struct ip_set **set) 412{ 413 ip_set_id_t i, index = IPSET_INVALID_ID; 414 struct ip_set *s; 415 416 for (i = 0; i < ip_set_max; i++) { 417 s = ip_set_list[i]; 418 if (s != NULL && STREQ(s->name, name)) { 419 __ip_set_get(i); 420 index = i; 421 *set = s; 422 } 423 } 424 425 return index; 426} 427EXPORT_SYMBOL_GPL(ip_set_get_byname); 428 429/* 430 * If the given set pointer points to a valid set, decrement 431 * reference count by 1. The caller shall not assume the index 432 * to be valid, after calling this function. 433 * 434 */ 435void 436ip_set_put_byindex(ip_set_id_t index) 437{ 438 if (ip_set_list[index] != NULL) 439 __ip_set_put(index); 440} 441EXPORT_SYMBOL_GPL(ip_set_put_byindex); 442 443/* 444 * Get the name of a set behind a set index. 445 * We assume the set is referenced, so it does exist and 446 * can't be destroyed. The set cannot be renamed due to 447 * the referencing either. 448 * 449 */ 450const char * 451ip_set_name_byindex(ip_set_id_t index) 452{ 453 const struct ip_set *set = ip_set_list[index]; 454 455 BUG_ON(set == NULL); 456 BUG_ON(set->ref == 0); 457 458 /* Referenced, so it's safe */ 459 return set->name; 460} 461EXPORT_SYMBOL_GPL(ip_set_name_byindex); 462 463/* 464 * Routines to call by external subsystems, which do not 465 * call nfnl_lock for us. 466 */ 467 468/* 469 * Find set by name, reference it once. The reference makes sure the 470 * thing pointed to, does not go away under our feet. 471 * 472 * The nfnl mutex is used in the function. 473 */ 474ip_set_id_t 475ip_set_nfnl_get(const char *name) 476{ 477 struct ip_set *s; 478 ip_set_id_t index; 479 480 nfnl_lock(); 481 index = ip_set_get_byname(name, &s); 482 nfnl_unlock(); 483 484 return index; 485} 486EXPORT_SYMBOL_GPL(ip_set_nfnl_get); 487 488/* 489 * Find set by index, reference it once. The reference makes sure the 490 * thing pointed to, does not go away under our feet. 491 * 492 * The nfnl mutex is used in the function. 493 */ 494ip_set_id_t 495ip_set_nfnl_get_byindex(ip_set_id_t index) 496{ 497 if (index > ip_set_max) 498 return IPSET_INVALID_ID; 499 500 nfnl_lock(); 501 if (ip_set_list[index]) 502 __ip_set_get(index); 503 else 504 index = IPSET_INVALID_ID; 505 nfnl_unlock(); 506 507 return index; 508} 509EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex); 510 511/* 512 * If the given set pointer points to a valid set, decrement 513 * reference count by 1. The caller shall not assume the index 514 * to be valid, after calling this function. 515 * 516 * The nfnl mutex is used in the function. 517 */ 518void 519ip_set_nfnl_put(ip_set_id_t index) 520{ 521 nfnl_lock(); 522 ip_set_put_byindex(index); 523 nfnl_unlock(); 524} 525EXPORT_SYMBOL_GPL(ip_set_nfnl_put); 526 527/* 528 * Communication protocol with userspace over netlink. 529 * 530 * The commands are serialized by the nfnl mutex. 531 */ 532 533static inline bool 534protocol_failed(const struct nlattr * const tb[]) 535{ 536 return !tb[IPSET_ATTR_PROTOCOL] || 537 nla_get_u8(tb[IPSET_ATTR_PROTOCOL]) != IPSET_PROTOCOL; 538} 539 540static inline u32 541flag_exist(const struct nlmsghdr *nlh) 542{ 543 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST; 544} 545 546static struct nlmsghdr * 547start_msg(struct sk_buff *skb, u32 pid, u32 seq, unsigned int flags, 548 enum ipset_cmd cmd) 549{ 550 struct nlmsghdr *nlh; 551 struct nfgenmsg *nfmsg; 552 553 nlh = nlmsg_put(skb, pid, seq, cmd | (NFNL_SUBSYS_IPSET << 8), 554 sizeof(*nfmsg), flags); 555 if (nlh == NULL) 556 return NULL; 557 558 nfmsg = nlmsg_data(nlh); 559 nfmsg->nfgen_family = AF_INET; 560 nfmsg->version = NFNETLINK_V0; 561 nfmsg->res_id = 0; 562 563 return nlh; 564} 565 566/* Create a set */ 567 568static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = { 569 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 570 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 571 .len = IPSET_MAXNAMELEN - 1 }, 572 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING, 573 .len = IPSET_MAXNAMELEN - 1}, 574 [IPSET_ATTR_REVISION] = { .type = NLA_U8 }, 575 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 }, 576 [IPSET_ATTR_DATA] = { .type = NLA_NESTED }, 577}; 578 579static ip_set_id_t 580find_set_id(const char *name) 581{ 582 ip_set_id_t i, index = IPSET_INVALID_ID; 583 const struct ip_set *set; 584 585 for (i = 0; index == IPSET_INVALID_ID && i < ip_set_max; i++) { 586 set = ip_set_list[i]; 587 if (set != NULL && STREQ(set->name, name)) 588 index = i; 589 } 590 return index; 591} 592 593static inline struct ip_set * 594find_set(const char *name) 595{ 596 ip_set_id_t index = find_set_id(name); 597 598 return index == IPSET_INVALID_ID ? NULL : ip_set_list[index]; 599} 600 601static int 602find_free_id(const char *name, ip_set_id_t *index, struct ip_set **set) 603{ 604 ip_set_id_t i; 605 606 *index = IPSET_INVALID_ID; 607 for (i = 0; i < ip_set_max; i++) { 608 if (ip_set_list[i] == NULL) { 609 if (*index == IPSET_INVALID_ID) 610 *index = i; 611 } else if (STREQ(name, ip_set_list[i]->name)) { 612 /* Name clash */ 613 *set = ip_set_list[i]; 614 return -EEXIST; 615 } 616 } 617 if (*index == IPSET_INVALID_ID) 618 /* No free slot remained */ 619 return -IPSET_ERR_MAX_SETS; 620 return 0; 621} 622 623static int 624ip_set_create(struct sock *ctnl, struct sk_buff *skb, 625 const struct nlmsghdr *nlh, 626 const struct nlattr * const attr[]) 627{ 628 struct ip_set *set, *clash = NULL; 629 ip_set_id_t index = IPSET_INVALID_ID; 630 struct nlattr *tb[IPSET_ATTR_CREATE_MAX+1] = {}; 631 const char *name, *typename; 632 u8 family, revision; 633 u32 flags = flag_exist(nlh); 634 int ret = 0; 635 636 if (unlikely(protocol_failed(attr) || 637 attr[IPSET_ATTR_SETNAME] == NULL || 638 attr[IPSET_ATTR_TYPENAME] == NULL || 639 attr[IPSET_ATTR_REVISION] == NULL || 640 attr[IPSET_ATTR_FAMILY] == NULL || 641 (attr[IPSET_ATTR_DATA] != NULL && 642 !flag_nested(attr[IPSET_ATTR_DATA])))) 643 return -IPSET_ERR_PROTOCOL; 644 645 name = nla_data(attr[IPSET_ATTR_SETNAME]); 646 typename = nla_data(attr[IPSET_ATTR_TYPENAME]); 647 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]); 648 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]); 649 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n", 650 name, typename, family_name(family), revision); 651 652 /* 653 * First, and without any locks, allocate and initialize 654 * a normal base set structure. 655 */ 656 set = kzalloc(sizeof(struct ip_set), GFP_KERNEL); 657 if (!set) 658 return -ENOMEM; 659 rwlock_init(&set->lock); 660 strlcpy(set->name, name, IPSET_MAXNAMELEN); 661 set->family = family; 662 set->revision = revision; 663 664 /* 665 * Next, check that we know the type, and take 666 * a reference on the type, to make sure it stays available 667 * while constructing our new set. 668 * 669 * After referencing the type, we try to create the type 670 * specific part of the set without holding any locks. 671 */ 672 ret = find_set_type_get(typename, family, revision, &(set->type)); 673 if (ret) 674 goto out; 675 676 /* 677 * Without holding any locks, create private part. 678 */ 679 if (attr[IPSET_ATTR_DATA] && 680 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA], 681 set->type->create_policy)) { 682 ret = -IPSET_ERR_PROTOCOL; 683 goto put_out; 684 } 685 686 ret = set->type->create(set, tb, flags); 687 if (ret != 0) 688 goto put_out; 689 690 /* BTW, ret==0 here. */ 691 692 /* 693 * Here, we have a valid, constructed set and we are protected 694 * by the nfnl mutex. Find the first free index in ip_set_list 695 * and check clashing. 696 */ 697 if ((ret = find_free_id(set->name, &index, &clash)) != 0) { 698 /* If this is the same set and requested, ignore error */ 699 if (ret == -EEXIST && 700 (flags & IPSET_FLAG_EXIST) && 701 STREQ(set->type->name, clash->type->name) && 702 set->type->family == clash->type->family && 703 set->type->revision_min == clash->type->revision_min && 704 set->type->revision_max == clash->type->revision_max && 705 set->variant->same_set(set, clash)) 706 ret = 0; 707 goto cleanup; 708 } 709 710 /* 711 * Finally! Add our shiny new set to the list, and be done. 712 */ 713 pr_debug("create: '%s' created with index %u!\n", set->name, index); 714 ip_set_list[index] = set; 715 716 return ret; 717 718cleanup: 719 set->variant->destroy(set); 720put_out: 721 module_put(set->type->me); 722out: 723 kfree(set); 724 return ret; 725} 726 727/* Destroy sets */ 728 729static const struct nla_policy 730ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = { 731 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 732 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 733 .len = IPSET_MAXNAMELEN - 1 }, 734}; 735 736static void 737ip_set_destroy_set(ip_set_id_t index) 738{ 739 struct ip_set *set = ip_set_list[index]; 740 741 pr_debug("set: %s\n", set->name); 742 ip_set_list[index] = NULL; 743 744 /* Must call it without holding any lock */ 745 set->variant->destroy(set); 746 module_put(set->type->me); 747 kfree(set); 748} 749 750static int 751ip_set_destroy(struct sock *ctnl, struct sk_buff *skb, 752 const struct nlmsghdr *nlh, 753 const struct nlattr * const attr[]) 754{ 755 ip_set_id_t i; 756 int ret = 0; 757 758 if (unlikely(protocol_failed(attr))) 759 return -IPSET_ERR_PROTOCOL; 760 761 /* Commands are serialized and references are 762 * protected by the ip_set_ref_lock. 763 * External systems (i.e. xt_set) must call 764 * ip_set_put|get_nfnl_* functions, that way we 765 * can safely check references here. 766 * 767 * list:set timer can only decrement the reference 768 * counter, so if it's already zero, we can proceed 769 * without holding the lock. 770 */ 771 read_lock_bh(&ip_set_ref_lock); 772 if (!attr[IPSET_ATTR_SETNAME]) { 773 for (i = 0; i < ip_set_max; i++) { 774 if (ip_set_list[i] != NULL && ip_set_list[i]->ref) { 775 ret = IPSET_ERR_BUSY; 776 goto out; 777 } 778 } 779 read_unlock_bh(&ip_set_ref_lock); 780 for (i = 0; i < ip_set_max; i++) { 781 if (ip_set_list[i] != NULL) 782 ip_set_destroy_set(i); 783 } 784 } else { 785 i = find_set_id(nla_data(attr[IPSET_ATTR_SETNAME])); 786 if (i == IPSET_INVALID_ID) { 787 ret = -ENOENT; 788 goto out; 789 } else if (ip_set_list[i]->ref) { 790 ret = -IPSET_ERR_BUSY; 791 goto out; 792 } 793 read_unlock_bh(&ip_set_ref_lock); 794 795 ip_set_destroy_set(i); 796 } 797 return 0; 798out: 799 read_unlock_bh(&ip_set_ref_lock); 800 return ret; 801} 802 803/* Flush sets */ 804 805static void 806ip_set_flush_set(struct ip_set *set) 807{ 808 pr_debug("set: %s\n", set->name); 809 810 write_lock_bh(&set->lock); 811 set->variant->flush(set); 812 write_unlock_bh(&set->lock); 813} 814 815static int 816ip_set_flush(struct sock *ctnl, struct sk_buff *skb, 817 const struct nlmsghdr *nlh, 818 const struct nlattr * const attr[]) 819{ 820 ip_set_id_t i; 821 822 if (unlikely(protocol_failed(attr))) 823 return -IPSET_ERR_PROTOCOL; 824 825 if (!attr[IPSET_ATTR_SETNAME]) { 826 for (i = 0; i < ip_set_max; i++) 827 if (ip_set_list[i] != NULL) 828 ip_set_flush_set(ip_set_list[i]); 829 } else { 830 i = find_set_id(nla_data(attr[IPSET_ATTR_SETNAME])); 831 if (i == IPSET_INVALID_ID) 832 return -ENOENT; 833 834 ip_set_flush_set(ip_set_list[i]); 835 } 836 837 return 0; 838} 839 840/* Rename a set */ 841 842static const struct nla_policy 843ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = { 844 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 845 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 846 .len = IPSET_MAXNAMELEN - 1 }, 847 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING, 848 .len = IPSET_MAXNAMELEN - 1 }, 849}; 850 851static int 852ip_set_rename(struct sock *ctnl, struct sk_buff *skb, 853 const struct nlmsghdr *nlh, 854 const struct nlattr * const attr[]) 855{ 856 struct ip_set *set; 857 const char *name2; 858 ip_set_id_t i; 859 int ret = 0; 860 861 if (unlikely(protocol_failed(attr) || 862 attr[IPSET_ATTR_SETNAME] == NULL || 863 attr[IPSET_ATTR_SETNAME2] == NULL)) 864 return -IPSET_ERR_PROTOCOL; 865 866 set = find_set(nla_data(attr[IPSET_ATTR_SETNAME])); 867 if (set == NULL) 868 return -ENOENT; 869 870 read_lock_bh(&ip_set_ref_lock); 871 if (set->ref != 0) { 872 ret = -IPSET_ERR_REFERENCED; 873 goto out; 874 } 875 876 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]); 877 for (i = 0; i < ip_set_max; i++) { 878 if (ip_set_list[i] != NULL && 879 STREQ(ip_set_list[i]->name, name2)) { 880 ret = -IPSET_ERR_EXIST_SETNAME2; 881 goto out; 882 } 883 } 884 strncpy(set->name, name2, IPSET_MAXNAMELEN); 885 886out: 887 read_unlock_bh(&ip_set_ref_lock); 888 return ret; 889} 890 891/* Swap two sets so that name/index points to the other. 892 * References and set names are also swapped. 893 * 894 * The commands are serialized by the nfnl mutex and references are 895 * protected by the ip_set_ref_lock. The kernel interfaces 896 * do not hold the mutex but the pointer settings are atomic 897 * so the ip_set_list always contains valid pointers to the sets. 898 */ 899 900static int 901ip_set_swap(struct sock *ctnl, struct sk_buff *skb, 902 const struct nlmsghdr *nlh, 903 const struct nlattr * const attr[]) 904{ 905 struct ip_set *from, *to; 906 ip_set_id_t from_id, to_id; 907 char from_name[IPSET_MAXNAMELEN]; 908 909 if (unlikely(protocol_failed(attr) || 910 attr[IPSET_ATTR_SETNAME] == NULL || 911 attr[IPSET_ATTR_SETNAME2] == NULL)) 912 return -IPSET_ERR_PROTOCOL; 913 914 from_id = find_set_id(nla_data(attr[IPSET_ATTR_SETNAME])); 915 if (from_id == IPSET_INVALID_ID) 916 return -ENOENT; 917 918 to_id = find_set_id(nla_data(attr[IPSET_ATTR_SETNAME2])); 919 if (to_id == IPSET_INVALID_ID) 920 return -IPSET_ERR_EXIST_SETNAME2; 921 922 from = ip_set_list[from_id]; 923 to = ip_set_list[to_id]; 924 925 /* Features must not change. 926 * Not an artificial restriction anymore, as we must prevent 927 * possible loops created by swapping in setlist type of sets. */ 928 if (!(from->type->features == to->type->features && 929 from->type->family == to->type->family)) 930 return -IPSET_ERR_TYPE_MISMATCH; 931 932 strncpy(from_name, from->name, IPSET_MAXNAMELEN); 933 strncpy(from->name, to->name, IPSET_MAXNAMELEN); 934 strncpy(to->name, from_name, IPSET_MAXNAMELEN); 935 936 write_lock_bh(&ip_set_ref_lock); 937 swap(from->ref, to->ref); 938 ip_set_list[from_id] = to; 939 ip_set_list[to_id] = from; 940 write_unlock_bh(&ip_set_ref_lock); 941 942 return 0; 943} 944 945/* List/save set data */ 946 947#define DUMP_INIT 0 948#define DUMP_ALL 1 949#define DUMP_ONE 2 950#define DUMP_LAST 3 951 952#define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF) 953#define DUMP_FLAGS(arg) (((u32)(arg)) >> 16) 954 955static int 956ip_set_dump_done(struct netlink_callback *cb) 957{ 958 if (cb->args[2]) { 959 pr_debug("release set %s\n", ip_set_list[cb->args[1]]->name); 960 ip_set_put_byindex((ip_set_id_t) cb->args[1]); 961 } 962 return 0; 963} 964 965static inline void 966dump_attrs(struct nlmsghdr *nlh) 967{ 968 const struct nlattr *attr; 969 int rem; 970 971 pr_debug("dump nlmsg\n"); 972 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) { 973 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len); 974 } 975} 976 977static int 978dump_init(struct netlink_callback *cb) 979{ 980 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb); 981 int min_len = NLMSG_SPACE(sizeof(struct nfgenmsg)); 982 struct nlattr *cda[IPSET_ATTR_CMD_MAX+1]; 983 struct nlattr *attr = (void *)nlh + min_len; 984 u32 dump_type; 985 ip_set_id_t index; 986 987 /* Second pass, so parser can't fail */ 988 nla_parse(cda, IPSET_ATTR_CMD_MAX, 989 attr, nlh->nlmsg_len - min_len, ip_set_setname_policy); 990 991 /* cb->args[0] : dump single set/all sets 992 * [1] : set index 993 * [..]: type specific 994 */ 995 996 if (cda[IPSET_ATTR_SETNAME]) { 997 index = find_set_id(nla_data(cda[IPSET_ATTR_SETNAME])); 998 if (index == IPSET_INVALID_ID) 999 return -ENOENT; 1000 1001 dump_type = DUMP_ONE; 1002 cb->args[1] = index; 1003 } else 1004 dump_type = DUMP_ALL; 1005 1006 if (cda[IPSET_ATTR_FLAGS]) { 1007 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]); 1008 dump_type |= (f << 16); 1009 } 1010 cb->args[0] = dump_type; 1011 1012 return 0; 1013} 1014 1015static int 1016ip_set_dump_start(struct sk_buff *skb, struct netlink_callback *cb) 1017{ 1018 ip_set_id_t index = IPSET_INVALID_ID, max; 1019 struct ip_set *set = NULL; 1020 struct nlmsghdr *nlh = NULL; 1021 unsigned int flags = NETLINK_CB(cb->skb).pid ? NLM_F_MULTI : 0; 1022 u32 dump_type, dump_flags; 1023 int ret = 0; 1024 1025 if (!cb->args[0]) { 1026 ret = dump_init(cb); 1027 if (ret < 0) { 1028 nlh = nlmsg_hdr(cb->skb); 1029 /* We have to create and send the error message 1030 * manually :-( */ 1031 if (nlh->nlmsg_flags & NLM_F_ACK) 1032 netlink_ack(cb->skb, nlh, ret); 1033 return ret; 1034 } 1035 } 1036 1037 if (cb->args[1] >= ip_set_max) 1038 goto out; 1039 1040 dump_type = DUMP_TYPE(cb->args[0]); 1041 dump_flags = DUMP_FLAGS(cb->args[0]); 1042 max = dump_type == DUMP_ONE ? cb->args[1] + 1 : ip_set_max; 1043dump_last: 1044 pr_debug("args[0]: %u %u args[1]: %ld\n", 1045 dump_type, dump_flags, cb->args[1]); 1046 for (; cb->args[1] < max; cb->args[1]++) { 1047 index = (ip_set_id_t) cb->args[1]; 1048 set = ip_set_list[index]; 1049 if (set == NULL) { 1050 if (dump_type == DUMP_ONE) { 1051 ret = -ENOENT; 1052 goto out; 1053 } 1054 continue; 1055 } 1056 /* When dumping all sets, we must dump "sorted" 1057 * so that lists (unions of sets) are dumped last. 1058 */ 1059 if (dump_type != DUMP_ONE && 1060 ((dump_type == DUMP_ALL) == 1061 !!(set->type->features & IPSET_DUMP_LAST))) 1062 continue; 1063 pr_debug("List set: %s\n", set->name); 1064 if (!cb->args[2]) { 1065 /* Start listing: make sure set won't be destroyed */ 1066 pr_debug("reference set\n"); 1067 __ip_set_get(index); 1068 } 1069 nlh = start_msg(skb, NETLINK_CB(cb->skb).pid, 1070 cb->nlh->nlmsg_seq, flags, 1071 IPSET_CMD_LIST); 1072 if (!nlh) { 1073 ret = -EMSGSIZE; 1074 goto release_refcount; 1075 } 1076 NLA_PUT_U8(skb, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL); 1077 NLA_PUT_STRING(skb, IPSET_ATTR_SETNAME, set->name); 1078 if (dump_flags & IPSET_FLAG_LIST_SETNAME) 1079 goto next_set; 1080 switch (cb->args[2]) { 1081 case 0: 1082 /* Core header data */ 1083 NLA_PUT_STRING(skb, IPSET_ATTR_TYPENAME, 1084 set->type->name); 1085 NLA_PUT_U8(skb, IPSET_ATTR_FAMILY, 1086 set->family); 1087 NLA_PUT_U8(skb, IPSET_ATTR_REVISION, 1088 set->revision); 1089 ret = set->variant->head(set, skb); 1090 if (ret < 0) 1091 goto release_refcount; 1092 if (dump_flags & IPSET_FLAG_LIST_HEADER) 1093 goto next_set; 1094 /* Fall through and add elements */ 1095 default: 1096 read_lock_bh(&set->lock); 1097 ret = set->variant->list(set, skb, cb); 1098 read_unlock_bh(&set->lock); 1099 if (!cb->args[2]) 1100 /* Set is done, proceed with next one */ 1101 goto next_set; 1102 goto release_refcount; 1103 } 1104 } 1105 /* If we dump all sets, continue with dumping last ones */ 1106 if (dump_type == DUMP_ALL) { 1107 dump_type = DUMP_LAST; 1108 cb->args[0] = dump_type | (dump_flags << 16); 1109 cb->args[1] = 0; 1110 goto dump_last; 1111 } 1112 goto out; 1113 1114nla_put_failure: 1115 ret = -EFAULT; 1116next_set: 1117 if (dump_type == DUMP_ONE) 1118 cb->args[1] = IPSET_INVALID_ID; 1119 else 1120 cb->args[1]++; 1121release_refcount: 1122 /* If there was an error or set is done, release set */ 1123 if (ret || !cb->args[2]) { 1124 pr_debug("release set %s\n", ip_set_list[index]->name); 1125 ip_set_put_byindex(index); 1126 } 1127out: 1128 if (nlh) { 1129 nlmsg_end(skb, nlh); 1130 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len); 1131 dump_attrs(nlh); 1132 } 1133 1134 return ret < 0 ? ret : skb->len; 1135} 1136 1137static int 1138ip_set_dump(struct sock *ctnl, struct sk_buff *skb, 1139 const struct nlmsghdr *nlh, 1140 const struct nlattr * const attr[]) 1141{ 1142 if (unlikely(protocol_failed(attr))) 1143 return -IPSET_ERR_PROTOCOL; 1144 1145 return netlink_dump_start(ctnl, skb, nlh, 1146 ip_set_dump_start, 1147 ip_set_dump_done, 0); 1148} 1149 1150/* Add, del and test */ 1151 1152static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = { 1153 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1154 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING, 1155 .len = IPSET_MAXNAMELEN - 1 }, 1156 [IPSET_ATTR_LINENO] = { .type = NLA_U32 }, 1157 [IPSET_ATTR_DATA] = { .type = NLA_NESTED }, 1158 [IPSET_ATTR_ADT] = { .type = NLA_NESTED }, 1159}; 1160 1161static int 1162call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set, 1163 struct nlattr *tb[], enum ipset_adt adt, 1164 u32 flags, bool use_lineno) 1165{ 1166 int ret; 1167 u32 lineno = 0; 1168 bool eexist = flags & IPSET_FLAG_EXIST, retried = false; 1169 1170 do { 1171 write_lock_bh(&set->lock); 1172 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried); 1173 write_unlock_bh(&set->lock); 1174 retried = true; 1175 } while (ret == -EAGAIN && 1176 set->variant->resize && 1177 (ret = set->variant->resize(set, retried)) == 0); 1178 1179 if (!ret || (ret == -IPSET_ERR_EXIST && eexist)) 1180 return 0; 1181 if (lineno && use_lineno) { 1182 /* Error in restore/batch mode: send back lineno */ 1183 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb); 1184 struct sk_buff *skb2; 1185 struct nlmsgerr *errmsg; 1186 size_t payload = sizeof(*errmsg) + nlmsg_len(nlh); 1187 int min_len = NLMSG_SPACE(sizeof(struct nfgenmsg)); 1188 struct nlattr *cda[IPSET_ATTR_CMD_MAX+1]; 1189 struct nlattr *cmdattr; 1190 u32 *errline; 1191 1192 skb2 = nlmsg_new(payload, GFP_KERNEL); 1193 if (skb2 == NULL) 1194 return -ENOMEM; 1195 rep = __nlmsg_put(skb2, NETLINK_CB(skb).pid, 1196 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0); 1197 errmsg = nlmsg_data(rep); 1198 errmsg->error = ret; 1199 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len); 1200 cmdattr = (void *)&errmsg->msg + min_len; 1201 1202 nla_parse(cda, IPSET_ATTR_CMD_MAX, 1203 cmdattr, nlh->nlmsg_len - min_len, 1204 ip_set_adt_policy); 1205 1206 errline = nla_data(cda[IPSET_ATTR_LINENO]); 1207 1208 *errline = lineno; 1209 1210 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT); 1211 /* Signal netlink not to send its ACK/errmsg. */ 1212 return -EINTR; 1213 } 1214 1215 return ret; 1216} 1217 1218static int 1219ip_set_uadd(struct sock *ctnl, struct sk_buff *skb, 1220 const struct nlmsghdr *nlh, 1221 const struct nlattr * const attr[]) 1222{ 1223 struct ip_set *set; 1224 struct nlattr *tb[IPSET_ATTR_ADT_MAX+1] = {}; 1225 const struct nlattr *nla; 1226 u32 flags = flag_exist(nlh); 1227 bool use_lineno; 1228 int ret = 0; 1229 1230 if (unlikely(protocol_failed(attr) || 1231 attr[IPSET_ATTR_SETNAME] == NULL || 1232 !((attr[IPSET_ATTR_DATA] != NULL) ^ 1233 (attr[IPSET_ATTR_ADT] != NULL)) || 1234 (attr[IPSET_ATTR_DATA] != NULL && 1235 !flag_nested(attr[IPSET_ATTR_DATA])) || 1236 (attr[IPSET_ATTR_ADT] != NULL && 1237 (!flag_nested(attr[IPSET_ATTR_ADT]) || 1238 attr[IPSET_ATTR_LINENO] == NULL)))) 1239 return -IPSET_ERR_PROTOCOL; 1240 1241 set = find_set(nla_data(attr[IPSET_ATTR_SETNAME])); 1242 if (set == NULL) 1243 return -ENOENT; 1244 1245 use_lineno = !!attr[IPSET_ATTR_LINENO]; 1246 if (attr[IPSET_ATTR_DATA]) { 1247 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, 1248 attr[IPSET_ATTR_DATA], 1249 set->type->adt_policy)) 1250 return -IPSET_ERR_PROTOCOL; 1251 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD, flags, 1252 use_lineno); 1253 } else { 1254 int nla_rem; 1255 1256 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) { 1257 memset(tb, 0, sizeof(tb)); 1258 if (nla_type(nla) != IPSET_ATTR_DATA || 1259 !flag_nested(nla) || 1260 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla, 1261 set->type->adt_policy)) 1262 return -IPSET_ERR_PROTOCOL; 1263 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD, 1264 flags, use_lineno); 1265 if (ret < 0) 1266 return ret; 1267 } 1268 } 1269 return ret; 1270} 1271 1272static int 1273ip_set_udel(struct sock *ctnl, struct sk_buff *skb, 1274 const struct nlmsghdr *nlh, 1275 const struct nlattr * const attr[]) 1276{ 1277 struct ip_set *set; 1278 struct nlattr *tb[IPSET_ATTR_ADT_MAX+1] = {}; 1279 const struct nlattr *nla; 1280 u32 flags = flag_exist(nlh); 1281 bool use_lineno; 1282 int ret = 0; 1283 1284 if (unlikely(protocol_failed(attr) || 1285 attr[IPSET_ATTR_SETNAME] == NULL || 1286 !((attr[IPSET_ATTR_DATA] != NULL) ^ 1287 (attr[IPSET_ATTR_ADT] != NULL)) || 1288 (attr[IPSET_ATTR_DATA] != NULL && 1289 !flag_nested(attr[IPSET_ATTR_DATA])) || 1290 (attr[IPSET_ATTR_ADT] != NULL && 1291 (!flag_nested(attr[IPSET_ATTR_ADT]) || 1292 attr[IPSET_ATTR_LINENO] == NULL)))) 1293 return -IPSET_ERR_PROTOCOL; 1294 1295 set = find_set(nla_data(attr[IPSET_ATTR_SETNAME])); 1296 if (set == NULL) 1297 return -ENOENT; 1298 1299 use_lineno = !!attr[IPSET_ATTR_LINENO]; 1300 if (attr[IPSET_ATTR_DATA]) { 1301 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, 1302 attr[IPSET_ATTR_DATA], 1303 set->type->adt_policy)) 1304 return -IPSET_ERR_PROTOCOL; 1305 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL, flags, 1306 use_lineno); 1307 } else { 1308 int nla_rem; 1309 1310 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) { 1311 memset(tb, 0, sizeof(*tb)); 1312 if (nla_type(nla) != IPSET_ATTR_DATA || 1313 !flag_nested(nla) || 1314 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla, 1315 set->type->adt_policy)) 1316 return -IPSET_ERR_PROTOCOL; 1317 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL, 1318 flags, use_lineno); 1319 if (ret < 0) 1320 return ret; 1321 } 1322 } 1323 return ret; 1324} 1325 1326static int 1327ip_set_utest(struct sock *ctnl, struct sk_buff *skb, 1328 const struct nlmsghdr *nlh, 1329 const struct nlattr * const attr[]) 1330{ 1331 struct ip_set *set; 1332 struct nlattr *tb[IPSET_ATTR_ADT_MAX+1] = {}; 1333 int ret = 0; 1334 1335 if (unlikely(protocol_failed(attr) || 1336 attr[IPSET_ATTR_SETNAME] == NULL || 1337 attr[IPSET_ATTR_DATA] == NULL || 1338 !flag_nested(attr[IPSET_ATTR_DATA]))) 1339 return -IPSET_ERR_PROTOCOL; 1340 1341 set = find_set(nla_data(attr[IPSET_ATTR_SETNAME])); 1342 if (set == NULL) 1343 return -ENOENT; 1344 1345 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA], 1346 set->type->adt_policy)) 1347 return -IPSET_ERR_PROTOCOL; 1348 1349 read_lock_bh(&set->lock); 1350 ret = set->variant->uadt(set, tb, IPSET_TEST, NULL, 0, 0); 1351 read_unlock_bh(&set->lock); 1352 /* Userspace can't trigger element to be re-added */ 1353 if (ret == -EAGAIN) 1354 ret = 1; 1355 1356 return ret < 0 ? ret : ret > 0 ? 0 : -IPSET_ERR_EXIST; 1357} 1358 1359/* Get headed data of a set */ 1360 1361static int 1362ip_set_header(struct sock *ctnl, struct sk_buff *skb, 1363 const struct nlmsghdr *nlh, 1364 const struct nlattr * const attr[]) 1365{ 1366 const struct ip_set *set; 1367 struct sk_buff *skb2; 1368 struct nlmsghdr *nlh2; 1369 ip_set_id_t index; 1370 int ret = 0; 1371 1372 if (unlikely(protocol_failed(attr) || 1373 attr[IPSET_ATTR_SETNAME] == NULL)) 1374 return -IPSET_ERR_PROTOCOL; 1375 1376 index = find_set_id(nla_data(attr[IPSET_ATTR_SETNAME])); 1377 if (index == IPSET_INVALID_ID) 1378 return -ENOENT; 1379 set = ip_set_list[index]; 1380 1381 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 1382 if (skb2 == NULL) 1383 return -ENOMEM; 1384 1385 nlh2 = start_msg(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq, 0, 1386 IPSET_CMD_HEADER); 1387 if (!nlh2) 1388 goto nlmsg_failure; 1389 NLA_PUT_U8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL); 1390 NLA_PUT_STRING(skb2, IPSET_ATTR_SETNAME, set->name); 1391 NLA_PUT_STRING(skb2, IPSET_ATTR_TYPENAME, set->type->name); 1392 NLA_PUT_U8(skb2, IPSET_ATTR_FAMILY, set->family); 1393 NLA_PUT_U8(skb2, IPSET_ATTR_REVISION, set->revision); 1394 nlmsg_end(skb2, nlh2); 1395 1396 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT); 1397 if (ret < 0) 1398 return ret; 1399 1400 return 0; 1401 1402nla_put_failure: 1403 nlmsg_cancel(skb2, nlh2); 1404nlmsg_failure: 1405 kfree_skb(skb2); 1406 return -EMSGSIZE; 1407} 1408 1409/* Get type data */ 1410 1411static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = { 1412 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1413 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING, 1414 .len = IPSET_MAXNAMELEN - 1 }, 1415 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 }, 1416}; 1417 1418static int 1419ip_set_type(struct sock *ctnl, struct sk_buff *skb, 1420 const struct nlmsghdr *nlh, 1421 const struct nlattr * const attr[]) 1422{ 1423 struct sk_buff *skb2; 1424 struct nlmsghdr *nlh2; 1425 u8 family, min, max; 1426 const char *typename; 1427 int ret = 0; 1428 1429 if (unlikely(protocol_failed(attr) || 1430 attr[IPSET_ATTR_TYPENAME] == NULL || 1431 attr[IPSET_ATTR_FAMILY] == NULL)) 1432 return -IPSET_ERR_PROTOCOL; 1433 1434 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]); 1435 typename = nla_data(attr[IPSET_ATTR_TYPENAME]); 1436 ret = find_set_type_minmax(typename, family, &min, &max); 1437 if (ret) 1438 return ret; 1439 1440 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 1441 if (skb2 == NULL) 1442 return -ENOMEM; 1443 1444 nlh2 = start_msg(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq, 0, 1445 IPSET_CMD_TYPE); 1446 if (!nlh2) 1447 goto nlmsg_failure; 1448 NLA_PUT_U8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL); 1449 NLA_PUT_STRING(skb2, IPSET_ATTR_TYPENAME, typename); 1450 NLA_PUT_U8(skb2, IPSET_ATTR_FAMILY, family); 1451 NLA_PUT_U8(skb2, IPSET_ATTR_REVISION, max); 1452 NLA_PUT_U8(skb2, IPSET_ATTR_REVISION_MIN, min); 1453 nlmsg_end(skb2, nlh2); 1454 1455 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len); 1456 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT); 1457 if (ret < 0) 1458 return ret; 1459 1460 return 0; 1461 1462nla_put_failure: 1463 nlmsg_cancel(skb2, nlh2); 1464nlmsg_failure: 1465 kfree_skb(skb2); 1466 return -EMSGSIZE; 1467} 1468 1469/* Get protocol version */ 1470 1471static const struct nla_policy 1472ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = { 1473 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 }, 1474}; 1475 1476static int 1477ip_set_protocol(struct sock *ctnl, struct sk_buff *skb, 1478 const struct nlmsghdr *nlh, 1479 const struct nlattr * const attr[]) 1480{ 1481 struct sk_buff *skb2; 1482 struct nlmsghdr *nlh2; 1483 int ret = 0; 1484 1485 if (unlikely(attr[IPSET_ATTR_PROTOCOL] == NULL)) 1486 return -IPSET_ERR_PROTOCOL; 1487 1488 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 1489 if (skb2 == NULL) 1490 return -ENOMEM; 1491 1492 nlh2 = start_msg(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq, 0, 1493 IPSET_CMD_PROTOCOL); 1494 if (!nlh2) 1495 goto nlmsg_failure; 1496 NLA_PUT_U8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL); 1497 nlmsg_end(skb2, nlh2); 1498 1499 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT); 1500 if (ret < 0) 1501 return ret; 1502 1503 return 0; 1504 1505nla_put_failure: 1506 nlmsg_cancel(skb2, nlh2); 1507nlmsg_failure: 1508 kfree_skb(skb2); 1509 return -EMSGSIZE; 1510} 1511 1512static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = { 1513 [IPSET_CMD_CREATE] = { 1514 .call = ip_set_create, 1515 .attr_count = IPSET_ATTR_CMD_MAX, 1516 .policy = ip_set_create_policy, 1517 }, 1518 [IPSET_CMD_DESTROY] = { 1519 .call = ip_set_destroy, 1520 .attr_count = IPSET_ATTR_CMD_MAX, 1521 .policy = ip_set_setname_policy, 1522 }, 1523 [IPSET_CMD_FLUSH] = { 1524 .call = ip_set_flush, 1525 .attr_count = IPSET_ATTR_CMD_MAX, 1526 .policy = ip_set_setname_policy, 1527 }, 1528 [IPSET_CMD_RENAME] = { 1529 .call = ip_set_rename, 1530 .attr_count = IPSET_ATTR_CMD_MAX, 1531 .policy = ip_set_setname2_policy, 1532 }, 1533 [IPSET_CMD_SWAP] = { 1534 .call = ip_set_swap, 1535 .attr_count = IPSET_ATTR_CMD_MAX, 1536 .policy = ip_set_setname2_policy, 1537 }, 1538 [IPSET_CMD_LIST] = { 1539 .call = ip_set_dump, 1540 .attr_count = IPSET_ATTR_CMD_MAX, 1541 .policy = ip_set_setname_policy, 1542 }, 1543 [IPSET_CMD_SAVE] = { 1544 .call = ip_set_dump, 1545 .attr_count = IPSET_ATTR_CMD_MAX, 1546 .policy = ip_set_setname_policy, 1547 }, 1548 [IPSET_CMD_ADD] = { 1549 .call = ip_set_uadd, 1550 .attr_count = IPSET_ATTR_CMD_MAX, 1551 .policy = ip_set_adt_policy, 1552 }, 1553 [IPSET_CMD_DEL] = { 1554 .call = ip_set_udel, 1555 .attr_count = IPSET_ATTR_CMD_MAX, 1556 .policy = ip_set_adt_policy, 1557 }, 1558 [IPSET_CMD_TEST] = { 1559 .call = ip_set_utest, 1560 .attr_count = IPSET_ATTR_CMD_MAX, 1561 .policy = ip_set_adt_policy, 1562 }, 1563 [IPSET_CMD_HEADER] = { 1564 .call = ip_set_header, 1565 .attr_count = IPSET_ATTR_CMD_MAX, 1566 .policy = ip_set_setname_policy, 1567 }, 1568 [IPSET_CMD_TYPE] = { 1569 .call = ip_set_type, 1570 .attr_count = IPSET_ATTR_CMD_MAX, 1571 .policy = ip_set_type_policy, 1572 }, 1573 [IPSET_CMD_PROTOCOL] = { 1574 .call = ip_set_protocol, 1575 .attr_count = IPSET_ATTR_CMD_MAX, 1576 .policy = ip_set_protocol_policy, 1577 }, 1578}; 1579 1580static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = { 1581 .name = "ip_set", 1582 .subsys_id = NFNL_SUBSYS_IPSET, 1583 .cb_count = IPSET_MSG_MAX, 1584 .cb = ip_set_netlink_subsys_cb, 1585}; 1586 1587/* Interface to iptables/ip6tables */ 1588 1589static int 1590ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len) 1591{ 1592 unsigned *op; 1593 void *data; 1594 int copylen = *len, ret = 0; 1595 1596 if (!capable(CAP_NET_ADMIN)) 1597 return -EPERM; 1598 if (optval != SO_IP_SET) 1599 return -EBADF; 1600 if (*len < sizeof(unsigned)) 1601 return -EINVAL; 1602 1603 data = vmalloc(*len); 1604 if (!data) 1605 return -ENOMEM; 1606 if (copy_from_user(data, user, *len) != 0) { 1607 ret = -EFAULT; 1608 goto done; 1609 } 1610 op = (unsigned *) data; 1611 1612 if (*op < IP_SET_OP_VERSION) { 1613 /* Check the version at the beginning of operations */ 1614 struct ip_set_req_version *req_version = data; 1615 if (req_version->version != IPSET_PROTOCOL) { 1616 ret = -EPROTO; 1617 goto done; 1618 } 1619 } 1620 1621 switch (*op) { 1622 case IP_SET_OP_VERSION: { 1623 struct ip_set_req_version *req_version = data; 1624 1625 if (*len != sizeof(struct ip_set_req_version)) { 1626 ret = -EINVAL; 1627 goto done; 1628 } 1629 1630 req_version->version = IPSET_PROTOCOL; 1631 ret = copy_to_user(user, req_version, 1632 sizeof(struct ip_set_req_version)); 1633 goto done; 1634 } 1635 case IP_SET_OP_GET_BYNAME: { 1636 struct ip_set_req_get_set *req_get = data; 1637 1638 if (*len != sizeof(struct ip_set_req_get_set)) { 1639 ret = -EINVAL; 1640 goto done; 1641 } 1642 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0'; 1643 nfnl_lock(); 1644 req_get->set.index = find_set_id(req_get->set.name); 1645 nfnl_unlock(); 1646 goto copy; 1647 } 1648 case IP_SET_OP_GET_BYINDEX: { 1649 struct ip_set_req_get_set *req_get = data; 1650 1651 if (*len != sizeof(struct ip_set_req_get_set) || 1652 req_get->set.index >= ip_set_max) { 1653 ret = -EINVAL; 1654 goto done; 1655 } 1656 nfnl_lock(); 1657 strncpy(req_get->set.name, 1658 ip_set_list[req_get->set.index] 1659 ? ip_set_list[req_get->set.index]->name : "", 1660 IPSET_MAXNAMELEN); 1661 nfnl_unlock(); 1662 goto copy; 1663 } 1664 default: 1665 ret = -EBADMSG; 1666 goto done; 1667 } /* end of switch(op) */ 1668 1669copy: 1670 ret = copy_to_user(user, data, copylen); 1671 1672done: 1673 vfree(data); 1674 if (ret > 0) 1675 ret = 0; 1676 return ret; 1677} 1678 1679static struct nf_sockopt_ops so_set __read_mostly = { 1680 .pf = PF_INET, 1681 .get_optmin = SO_IP_SET, 1682 .get_optmax = SO_IP_SET + 1, 1683 .get = &ip_set_sockfn_get, 1684 .owner = THIS_MODULE, 1685}; 1686 1687static int __init 1688ip_set_init(void) 1689{ 1690 int ret; 1691 1692 if (max_sets) 1693 ip_set_max = max_sets; 1694 if (ip_set_max >= IPSET_INVALID_ID) 1695 ip_set_max = IPSET_INVALID_ID - 1; 1696 1697 ip_set_list = kzalloc(sizeof(struct ip_set *) * ip_set_max, 1698 GFP_KERNEL); 1699 if (!ip_set_list) { 1700 pr_err("ip_set: Unable to create ip_set_list\n"); 1701 return -ENOMEM; 1702 } 1703 1704 ret = nfnetlink_subsys_register(&ip_set_netlink_subsys); 1705 if (ret != 0) { 1706 pr_err("ip_set: cannot register with nfnetlink.\n"); 1707 kfree(ip_set_list); 1708 return ret; 1709 } 1710 ret = nf_register_sockopt(&so_set); 1711 if (ret != 0) { 1712 pr_err("SO_SET registry failed: %d\n", ret); 1713 nfnetlink_subsys_unregister(&ip_set_netlink_subsys); 1714 kfree(ip_set_list); 1715 return ret; 1716 } 1717 1718 pr_notice("ip_set: protocol %u\n", IPSET_PROTOCOL); 1719 return 0; 1720} 1721 1722static void __exit 1723ip_set_fini(void) 1724{ 1725 /* There can't be any existing set */ 1726 nf_unregister_sockopt(&so_set); 1727 nfnetlink_subsys_unregister(&ip_set_netlink_subsys); 1728 kfree(ip_set_list); 1729 pr_debug("these are the famous last words\n"); 1730} 1731 1732module_init(ip_set_init); 1733module_exit(ip_set_fini); 1734