1 /* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
2 * Patrick Schaaf <bof@bof.de>
3 * Copyright (C) 2003-2013 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
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.
10 /* Kernel module for IP set management */
12 #include <linux/init.h>
13 #include <linux/module.h>
14 #include <linux/moduleparam.h>
16 #include <linux/skbuff.h>
17 #include <linux/spinlock.h>
18 #include <linux/rculist.h>
19 #include <net/netlink.h>
20 #include <net/net_namespace.h>
21 #include <net/netns/generic.h>
23 #include <linux/netfilter.h>
24 #include <linux/netfilter/x_tables.h>
25 #include <linux/netfilter/nfnetlink.h>
26 #include <linux/netfilter/ipset/ip_set.h>
28 static LIST_HEAD(ip_set_type_list); /* all registered set types */
29 static DEFINE_MUTEX(ip_set_type_mutex); /* protects ip_set_type_list */
30 static DEFINE_RWLOCK(ip_set_ref_lock); /* protects the set refs */
33 struct ip_set * __rcu *ip_set_list; /* all individual sets */
34 ip_set_id_t ip_set_max; /* max number of sets */
35 int is_deleted; /* deleted by ip_set_net_exit */
37 static int ip_set_net_id __read_mostly;
39 static inline struct ip_set_net *ip_set_pernet(struct net *net)
41 return net_generic(net, ip_set_net_id);
45 #define STREQ(a, b) (strncmp(a, b, IPSET_MAXNAMELEN) == 0)
47 static unsigned int max_sets;
49 module_param(max_sets, int, 0600);
50 MODULE_PARM_DESC(max_sets, "maximal number of sets");
51 MODULE_LICENSE("GPL");
52 MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
53 MODULE_DESCRIPTION("core IP set support");
54 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_IPSET);
56 /* When the nfnl mutex is held: */
57 #define ip_set_dereference(p) \
58 rcu_dereference_protected(p, 1)
59 #define ip_set(inst, id) \
60 ip_set_dereference((inst)->ip_set_list)[id]
63 * The set types are implemented in modules and registered set types
64 * can be found in ip_set_type_list. Adding/deleting types is
65 * serialized by ip_set_type_mutex.
69 ip_set_type_lock(void)
71 mutex_lock(&ip_set_type_mutex);
75 ip_set_type_unlock(void)
77 mutex_unlock(&ip_set_type_mutex);
80 /* Register and deregister settype */
82 static struct ip_set_type *
83 find_set_type(const char *name, u8 family, u8 revision)
85 struct ip_set_type *type;
87 list_for_each_entry_rcu(type, &ip_set_type_list, list)
88 if (STREQ(type->name, name) &&
89 (type->family == family ||
90 type->family == NFPROTO_UNSPEC) &&
91 revision >= type->revision_min &&
92 revision <= type->revision_max)
97 /* Unlock, try to load a set type module and lock again */
99 load_settype(const char *name)
101 nfnl_unlock(NFNL_SUBSYS_IPSET);
102 pr_debug("try to load ip_set_%s\n", name);
103 if (request_module("ip_set_%s", name) < 0) {
104 pr_warning("Can't find ip_set type %s\n", name);
105 nfnl_lock(NFNL_SUBSYS_IPSET);
108 nfnl_lock(NFNL_SUBSYS_IPSET);
112 /* Find a set type and reference it */
113 #define find_set_type_get(name, family, revision, found) \
114 __find_set_type_get(name, family, revision, found, false)
117 __find_set_type_get(const char *name, u8 family, u8 revision,
118 struct ip_set_type **found, bool retry)
120 struct ip_set_type *type;
123 if (retry && !load_settype(name))
124 return -IPSET_ERR_FIND_TYPE;
127 *found = find_set_type(name, family, revision);
129 err = !try_module_get((*found)->me) ? -EFAULT : 0;
132 /* Make sure the type is already loaded
133 * but we don't support the revision */
134 list_for_each_entry_rcu(type, &ip_set_type_list, list)
135 if (STREQ(type->name, name)) {
136 err = -IPSET_ERR_FIND_TYPE;
141 return retry ? -IPSET_ERR_FIND_TYPE :
142 __find_set_type_get(name, family, revision, found, true);
149 /* Find a given set type by name and family.
150 * If we succeeded, the supported minimal and maximum revisions are
153 #define find_set_type_minmax(name, family, min, max) \
154 __find_set_type_minmax(name, family, min, max, false)
157 __find_set_type_minmax(const char *name, u8 family, u8 *min, u8 *max,
160 struct ip_set_type *type;
163 if (retry && !load_settype(name))
164 return -IPSET_ERR_FIND_TYPE;
166 *min = 255; *max = 0;
168 list_for_each_entry_rcu(type, &ip_set_type_list, list)
169 if (STREQ(type->name, name) &&
170 (type->family == family ||
171 type->family == NFPROTO_UNSPEC)) {
173 if (type->revision_min < *min)
174 *min = type->revision_min;
175 if (type->revision_max > *max)
176 *max = type->revision_max;
182 return retry ? -IPSET_ERR_FIND_TYPE :
183 __find_set_type_minmax(name, family, min, max, true);
186 #define family_name(f) ((f) == NFPROTO_IPV4 ? "inet" : \
187 (f) == NFPROTO_IPV6 ? "inet6" : "any")
189 /* Register a set type structure. The type is identified by
190 * the unique triple of name, family and revision.
193 ip_set_type_register(struct ip_set_type *type)
197 if (type->protocol != IPSET_PROTOCOL) {
198 pr_warning("ip_set type %s, family %s, revision %u:%u uses "
199 "wrong protocol version %u (want %u)\n",
200 type->name, family_name(type->family),
201 type->revision_min, type->revision_max,
202 type->protocol, IPSET_PROTOCOL);
207 if (find_set_type(type->name, type->family, type->revision_min)) {
209 pr_warning("ip_set type %s, family %s with revision min %u "
210 "already registered!\n", type->name,
211 family_name(type->family), type->revision_min);
215 list_add_rcu(&type->list, &ip_set_type_list);
216 pr_debug("type %s, family %s, revision %u:%u registered.\n",
217 type->name, family_name(type->family),
218 type->revision_min, type->revision_max);
220 ip_set_type_unlock();
223 EXPORT_SYMBOL_GPL(ip_set_type_register);
225 /* Unregister a set type. There's a small race with ip_set_create */
227 ip_set_type_unregister(struct ip_set_type *type)
230 if (!find_set_type(type->name, type->family, type->revision_min)) {
231 pr_warning("ip_set type %s, family %s with revision min %u "
232 "not registered\n", type->name,
233 family_name(type->family), type->revision_min);
236 list_del_rcu(&type->list);
237 pr_debug("type %s, family %s with revision min %u unregistered.\n",
238 type->name, family_name(type->family), type->revision_min);
240 ip_set_type_unlock();
244 EXPORT_SYMBOL_GPL(ip_set_type_unregister);
246 /* Utility functions */
248 ip_set_alloc(size_t size)
250 void *members = NULL;
252 if (size < KMALLOC_MAX_SIZE)
253 members = kzalloc(size, GFP_KERNEL | __GFP_NOWARN);
256 pr_debug("%p: allocated with kmalloc\n", members);
260 members = vzalloc(size);
263 pr_debug("%p: allocated with vmalloc\n", members);
267 EXPORT_SYMBOL_GPL(ip_set_alloc);
270 ip_set_free(void *members)
272 pr_debug("%p: free with %s\n", members,
273 is_vmalloc_addr(members) ? "vfree" : "kfree");
274 if (is_vmalloc_addr(members))
279 EXPORT_SYMBOL_GPL(ip_set_free);
282 flag_nested(const struct nlattr *nla)
284 return nla->nla_type & NLA_F_NESTED;
287 static const struct nla_policy ipaddr_policy[IPSET_ATTR_IPADDR_MAX + 1] = {
288 [IPSET_ATTR_IPADDR_IPV4] = { .type = NLA_U32 },
289 [IPSET_ATTR_IPADDR_IPV6] = { .type = NLA_BINARY,
290 .len = sizeof(struct in6_addr) },
294 ip_set_get_ipaddr4(struct nlattr *nla, __be32 *ipaddr)
296 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX+1];
298 if (unlikely(!flag_nested(nla)))
299 return -IPSET_ERR_PROTOCOL;
300 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla, ipaddr_policy))
301 return -IPSET_ERR_PROTOCOL;
302 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV4)))
303 return -IPSET_ERR_PROTOCOL;
305 *ipaddr = nla_get_be32(tb[IPSET_ATTR_IPADDR_IPV4]);
308 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr4);
311 ip_set_get_ipaddr6(struct nlattr *nla, union nf_inet_addr *ipaddr)
313 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX+1];
315 if (unlikely(!flag_nested(nla)))
316 return -IPSET_ERR_PROTOCOL;
318 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla, ipaddr_policy))
319 return -IPSET_ERR_PROTOCOL;
320 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV6)))
321 return -IPSET_ERR_PROTOCOL;
323 memcpy(ipaddr, nla_data(tb[IPSET_ATTR_IPADDR_IPV6]),
324 sizeof(struct in6_addr));
327 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr6);
329 typedef void (*destroyer)(void *);
330 /* ipset data extension types, in size order */
332 const struct ip_set_ext_type ip_set_extensions[] = {
333 [IPSET_EXT_ID_COUNTER] = {
334 .type = IPSET_EXT_COUNTER,
335 .flag = IPSET_FLAG_WITH_COUNTERS,
336 .len = sizeof(struct ip_set_counter),
337 .align = __alignof__(struct ip_set_counter),
339 [IPSET_EXT_ID_TIMEOUT] = {
340 .type = IPSET_EXT_TIMEOUT,
341 .len = sizeof(unsigned long),
342 .align = __alignof__(unsigned long),
344 [IPSET_EXT_ID_COMMENT] = {
345 .type = IPSET_EXT_COMMENT | IPSET_EXT_DESTROY,
346 .flag = IPSET_FLAG_WITH_COMMENT,
347 .len = sizeof(struct ip_set_comment),
348 .align = __alignof__(struct ip_set_comment),
349 .destroy = (destroyer) ip_set_comment_free,
352 EXPORT_SYMBOL_GPL(ip_set_extensions);
355 add_extension(enum ip_set_ext_id id, u32 flags, struct nlattr *tb[])
357 return ip_set_extensions[id].flag ?
358 (flags & ip_set_extensions[id].flag) :
359 !!tb[IPSET_ATTR_TIMEOUT];
363 ip_set_elem_len(struct ip_set *set, struct nlattr *tb[], size_t len)
365 enum ip_set_ext_id id;
369 if (tb[IPSET_ATTR_CADT_FLAGS])
370 cadt_flags = ip_set_get_h32(tb[IPSET_ATTR_CADT_FLAGS]);
371 if (cadt_flags & IPSET_FLAG_WITH_FORCEADD)
372 set->flags |= IPSET_CREATE_FLAG_FORCEADD;
373 for (id = 0; id < IPSET_EXT_ID_MAX; id++) {
374 if (!add_extension(id, cadt_flags, tb))
376 offset += ALIGN(len + offset, ip_set_extensions[id].align);
377 set->offset[id] = offset;
378 set->extensions |= ip_set_extensions[id].type;
379 offset += ip_set_extensions[id].len;
383 EXPORT_SYMBOL_GPL(ip_set_elem_len);
386 ip_set_get_extensions(struct ip_set *set, struct nlattr *tb[],
387 struct ip_set_ext *ext)
389 if (tb[IPSET_ATTR_TIMEOUT]) {
390 if (!(set->extensions & IPSET_EXT_TIMEOUT))
391 return -IPSET_ERR_TIMEOUT;
392 ext->timeout = ip_set_timeout_uget(tb[IPSET_ATTR_TIMEOUT]);
394 if (tb[IPSET_ATTR_BYTES] || tb[IPSET_ATTR_PACKETS]) {
395 if (!(set->extensions & IPSET_EXT_COUNTER))
396 return -IPSET_ERR_COUNTER;
397 if (tb[IPSET_ATTR_BYTES])
398 ext->bytes = be64_to_cpu(nla_get_be64(
399 tb[IPSET_ATTR_BYTES]));
400 if (tb[IPSET_ATTR_PACKETS])
401 ext->packets = be64_to_cpu(nla_get_be64(
402 tb[IPSET_ATTR_PACKETS]));
404 if (tb[IPSET_ATTR_COMMENT]) {
405 if (!(set->extensions & IPSET_EXT_COMMENT))
406 return -IPSET_ERR_COMMENT;
407 ext->comment = ip_set_comment_uget(tb[IPSET_ATTR_COMMENT]);
412 EXPORT_SYMBOL_GPL(ip_set_get_extensions);
415 * Creating/destroying/renaming/swapping affect the existence and
416 * the properties of a set. All of these can be executed from userspace
417 * only and serialized by the nfnl mutex indirectly from nfnetlink.
419 * Sets are identified by their index in ip_set_list and the index
420 * is used by the external references (set/SET netfilter modules).
422 * The set behind an index may change by swapping only, from userspace.
426 __ip_set_get(struct ip_set *set)
428 write_lock_bh(&ip_set_ref_lock);
430 write_unlock_bh(&ip_set_ref_lock);
434 __ip_set_put(struct ip_set *set)
436 write_lock_bh(&ip_set_ref_lock);
437 BUG_ON(set->ref == 0);
439 write_unlock_bh(&ip_set_ref_lock);
443 * Add, del and test set entries from kernel.
445 * The set behind the index must exist and must be referenced
446 * so it can't be destroyed (or changed) under our foot.
449 static inline struct ip_set *
450 ip_set_rcu_get(struct net *net, ip_set_id_t index)
453 struct ip_set_net *inst = ip_set_pernet(net);
456 /* ip_set_list itself needs to be protected */
457 set = rcu_dereference(inst->ip_set_list)[index];
464 ip_set_test(ip_set_id_t index, const struct sk_buff *skb,
465 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
467 struct ip_set *set = ip_set_rcu_get(
468 dev_net(par->in ? par->in : par->out), index);
472 pr_debug("set %s, index %u\n", set->name, index);
474 if (opt->dim < set->type->dimension ||
475 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
478 read_lock_bh(&set->lock);
479 ret = set->variant->kadt(set, skb, par, IPSET_TEST, opt);
480 read_unlock_bh(&set->lock);
482 if (ret == -EAGAIN) {
483 /* Type requests element to be completed */
484 pr_debug("element must be competed, ADD is triggered\n");
485 write_lock_bh(&set->lock);
486 set->variant->kadt(set, skb, par, IPSET_ADD, opt);
487 write_unlock_bh(&set->lock);
490 /* --return-nomatch: invert matched element */
491 if ((opt->cmdflags & IPSET_FLAG_RETURN_NOMATCH) &&
492 (set->type->features & IPSET_TYPE_NOMATCH) &&
493 (ret > 0 || ret == -ENOTEMPTY))
497 /* Convert error codes to nomatch */
498 return (ret < 0 ? 0 : ret);
500 EXPORT_SYMBOL_GPL(ip_set_test);
503 ip_set_add(ip_set_id_t index, const struct sk_buff *skb,
504 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
506 struct ip_set *set = ip_set_rcu_get(
507 dev_net(par->in ? par->in : par->out), index);
511 pr_debug("set %s, index %u\n", set->name, index);
513 if (opt->dim < set->type->dimension ||
514 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
515 return -IPSET_ERR_TYPE_MISMATCH;
517 write_lock_bh(&set->lock);
518 ret = set->variant->kadt(set, skb, par, IPSET_ADD, opt);
519 write_unlock_bh(&set->lock);
523 EXPORT_SYMBOL_GPL(ip_set_add);
526 ip_set_del(ip_set_id_t index, const struct sk_buff *skb,
527 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
529 struct ip_set *set = ip_set_rcu_get(
530 dev_net(par->in ? par->in : par->out), index);
534 pr_debug("set %s, index %u\n", set->name, index);
536 if (opt->dim < set->type->dimension ||
537 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
538 return -IPSET_ERR_TYPE_MISMATCH;
540 write_lock_bh(&set->lock);
541 ret = set->variant->kadt(set, skb, par, IPSET_DEL, opt);
542 write_unlock_bh(&set->lock);
546 EXPORT_SYMBOL_GPL(ip_set_del);
549 * Find set by name, reference it once. The reference makes sure the
550 * thing pointed to, does not go away under our feet.
554 ip_set_get_byname(struct net *net, const char *name, struct ip_set **set)
556 ip_set_id_t i, index = IPSET_INVALID_ID;
558 struct ip_set_net *inst = ip_set_pernet(net);
561 for (i = 0; i < inst->ip_set_max; i++) {
562 s = rcu_dereference(inst->ip_set_list)[i];
563 if (s != NULL && STREQ(s->name, name)) {
574 EXPORT_SYMBOL_GPL(ip_set_get_byname);
577 * If the given set pointer points to a valid set, decrement
578 * reference count by 1. The caller shall not assume the index
579 * to be valid, after calling this function.
584 __ip_set_put_byindex(struct ip_set_net *inst, ip_set_id_t index)
589 set = rcu_dereference(inst->ip_set_list)[index];
596 ip_set_put_byindex(struct net *net, ip_set_id_t index)
598 struct ip_set_net *inst = ip_set_pernet(net);
600 __ip_set_put_byindex(inst, index);
602 EXPORT_SYMBOL_GPL(ip_set_put_byindex);
605 * Get the name of a set behind a set index.
606 * We assume the set is referenced, so it does exist and
607 * can't be destroyed. The set cannot be renamed due to
608 * the referencing either.
612 ip_set_name_byindex(struct net *net, ip_set_id_t index)
614 const struct ip_set *set = ip_set_rcu_get(net, index);
617 BUG_ON(set->ref == 0);
619 /* Referenced, so it's safe */
622 EXPORT_SYMBOL_GPL(ip_set_name_byindex);
625 * Routines to call by external subsystems, which do not
626 * call nfnl_lock for us.
630 * Find set by index, reference it once. The reference makes sure the
631 * thing pointed to, does not go away under our feet.
633 * The nfnl mutex is used in the function.
636 ip_set_nfnl_get_byindex(struct net *net, ip_set_id_t index)
639 struct ip_set_net *inst = ip_set_pernet(net);
641 if (index > inst->ip_set_max)
642 return IPSET_INVALID_ID;
644 nfnl_lock(NFNL_SUBSYS_IPSET);
645 set = ip_set(inst, index);
649 index = IPSET_INVALID_ID;
650 nfnl_unlock(NFNL_SUBSYS_IPSET);
654 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex);
657 * If the given set pointer points to a valid set, decrement
658 * reference count by 1. The caller shall not assume the index
659 * to be valid, after calling this function.
661 * The nfnl mutex is used in the function.
664 ip_set_nfnl_put(struct net *net, ip_set_id_t index)
667 struct ip_set_net *inst = ip_set_pernet(net);
669 nfnl_lock(NFNL_SUBSYS_IPSET);
670 if (!inst->is_deleted) { /* already deleted from ip_set_net_exit() */
671 set = ip_set(inst, index);
675 nfnl_unlock(NFNL_SUBSYS_IPSET);
677 EXPORT_SYMBOL_GPL(ip_set_nfnl_put);
680 * Communication protocol with userspace over netlink.
682 * The commands are serialized by the nfnl mutex.
686 protocol_failed(const struct nlattr * const tb[])
688 return !tb[IPSET_ATTR_PROTOCOL] ||
689 nla_get_u8(tb[IPSET_ATTR_PROTOCOL]) != IPSET_PROTOCOL;
693 flag_exist(const struct nlmsghdr *nlh)
695 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST;
698 static struct nlmsghdr *
699 start_msg(struct sk_buff *skb, u32 portid, u32 seq, unsigned int flags,
702 struct nlmsghdr *nlh;
703 struct nfgenmsg *nfmsg;
705 nlh = nlmsg_put(skb, portid, seq, cmd | (NFNL_SUBSYS_IPSET << 8),
706 sizeof(*nfmsg), flags);
710 nfmsg = nlmsg_data(nlh);
711 nfmsg->nfgen_family = NFPROTO_IPV4;
712 nfmsg->version = NFNETLINK_V0;
720 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = {
721 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
722 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
723 .len = IPSET_MAXNAMELEN - 1 },
724 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
725 .len = IPSET_MAXNAMELEN - 1},
726 [IPSET_ATTR_REVISION] = { .type = NLA_U8 },
727 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
728 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
731 static struct ip_set *
732 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id)
734 struct ip_set *set = NULL;
737 *id = IPSET_INVALID_ID;
738 for (i = 0; i < inst->ip_set_max; i++) {
739 set = ip_set(inst, i);
740 if (set != NULL && STREQ(set->name, name)) {
745 return (*id == IPSET_INVALID_ID ? NULL : set);
748 static inline struct ip_set *
749 find_set(struct ip_set_net *inst, const char *name)
753 return find_set_and_id(inst, name, &id);
757 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index,
763 *index = IPSET_INVALID_ID;
764 for (i = 0; i < inst->ip_set_max; i++) {
767 if (*index == IPSET_INVALID_ID)
769 } else if (STREQ(name, s->name)) {
775 if (*index == IPSET_INVALID_ID)
776 /* No free slot remained */
777 return -IPSET_ERR_MAX_SETS;
782 ip_set_none(struct sock *ctnl, struct sk_buff *skb,
783 const struct nlmsghdr *nlh,
784 const struct nlattr * const attr[])
790 ip_set_create(struct sock *ctnl, struct sk_buff *skb,
791 const struct nlmsghdr *nlh,
792 const struct nlattr * const attr[])
794 struct net *net = sock_net(ctnl);
795 struct ip_set_net *inst = ip_set_pernet(net);
796 struct ip_set *set, *clash = NULL;
797 ip_set_id_t index = IPSET_INVALID_ID;
798 struct nlattr *tb[IPSET_ATTR_CREATE_MAX+1] = {};
799 const char *name, *typename;
801 u32 flags = flag_exist(nlh);
804 if (unlikely(protocol_failed(attr) ||
805 attr[IPSET_ATTR_SETNAME] == NULL ||
806 attr[IPSET_ATTR_TYPENAME] == NULL ||
807 attr[IPSET_ATTR_REVISION] == NULL ||
808 attr[IPSET_ATTR_FAMILY] == NULL ||
809 (attr[IPSET_ATTR_DATA] != NULL &&
810 !flag_nested(attr[IPSET_ATTR_DATA]))))
811 return -IPSET_ERR_PROTOCOL;
813 name = nla_data(attr[IPSET_ATTR_SETNAME]);
814 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
815 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
816 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]);
817 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
818 name, typename, family_name(family), revision);
821 * First, and without any locks, allocate and initialize
822 * a normal base set structure.
824 set = kzalloc(sizeof(struct ip_set), GFP_KERNEL);
827 rwlock_init(&set->lock);
828 strlcpy(set->name, name, IPSET_MAXNAMELEN);
829 set->family = family;
830 set->revision = revision;
833 * Next, check that we know the type, and take
834 * a reference on the type, to make sure it stays available
835 * while constructing our new set.
837 * After referencing the type, we try to create the type
838 * specific part of the set without holding any locks.
840 ret = find_set_type_get(typename, family, revision, &(set->type));
845 * Without holding any locks, create private part.
847 if (attr[IPSET_ATTR_DATA] &&
848 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA],
849 set->type->create_policy)) {
850 ret = -IPSET_ERR_PROTOCOL;
854 ret = set->type->create(net, set, tb, flags);
858 /* BTW, ret==0 here. */
861 * Here, we have a valid, constructed set and we are protected
862 * by the nfnl mutex. Find the first free index in ip_set_list
863 * and check clashing.
865 ret = find_free_id(inst, set->name, &index, &clash);
866 if (ret == -EEXIST) {
867 /* If this is the same set and requested, ignore error */
868 if ((flags & IPSET_FLAG_EXIST) &&
869 STREQ(set->type->name, clash->type->name) &&
870 set->type->family == clash->type->family &&
871 set->type->revision_min == clash->type->revision_min &&
872 set->type->revision_max == clash->type->revision_max &&
873 set->variant->same_set(set, clash))
876 } else if (ret == -IPSET_ERR_MAX_SETS) {
877 struct ip_set **list, **tmp;
878 ip_set_id_t i = inst->ip_set_max + IP_SET_INC;
880 if (i < inst->ip_set_max || i == IPSET_INVALID_ID)
884 list = kzalloc(sizeof(struct ip_set *) * i, GFP_KERNEL);
887 /* nfnl mutex is held, both lists are valid */
888 tmp = ip_set_dereference(inst->ip_set_list);
889 memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max);
890 rcu_assign_pointer(inst->ip_set_list, list);
891 /* Make sure all current packets have passed through */
894 index = inst->ip_set_max;
895 inst->ip_set_max = i;
902 * Finally! Add our shiny new set to the list, and be done.
904 pr_debug("create: '%s' created with index %u!\n", set->name, index);
905 ip_set(inst, index) = set;
910 set->variant->destroy(set);
912 module_put(set->type->me);
920 static const struct nla_policy
921 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = {
922 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
923 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
924 .len = IPSET_MAXNAMELEN - 1 },
928 ip_set_destroy_set(struct ip_set_net *inst, ip_set_id_t index)
930 struct ip_set *set = ip_set(inst, index);
932 pr_debug("set: %s\n", set->name);
933 ip_set(inst, index) = NULL;
935 /* Must call it without holding any lock */
936 set->variant->destroy(set);
937 module_put(set->type->me);
942 ip_set_destroy(struct sock *ctnl, struct sk_buff *skb,
943 const struct nlmsghdr *nlh,
944 const struct nlattr * const attr[])
946 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
951 if (unlikely(protocol_failed(attr)))
952 return -IPSET_ERR_PROTOCOL;
954 /* Commands are serialized and references are
955 * protected by the ip_set_ref_lock.
956 * External systems (i.e. xt_set) must call
957 * ip_set_put|get_nfnl_* functions, that way we
958 * can safely check references here.
960 * list:set timer can only decrement the reference
961 * counter, so if it's already zero, we can proceed
962 * without holding the lock.
964 read_lock_bh(&ip_set_ref_lock);
965 if (!attr[IPSET_ATTR_SETNAME]) {
966 for (i = 0; i < inst->ip_set_max; i++) {
968 if (s != NULL && s->ref) {
969 ret = -IPSET_ERR_BUSY;
973 read_unlock_bh(&ip_set_ref_lock);
974 for (i = 0; i < inst->ip_set_max; i++) {
977 ip_set_destroy_set(inst, i);
980 s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
986 ret = -IPSET_ERR_BUSY;
989 read_unlock_bh(&ip_set_ref_lock);
991 ip_set_destroy_set(inst, i);
995 read_unlock_bh(&ip_set_ref_lock);
1002 ip_set_flush_set(struct ip_set *set)
1004 pr_debug("set: %s\n", set->name);
1006 write_lock_bh(&set->lock);
1007 set->variant->flush(set);
1008 write_unlock_bh(&set->lock);
1012 ip_set_flush(struct sock *ctnl, struct sk_buff *skb,
1013 const struct nlmsghdr *nlh,
1014 const struct nlattr * const attr[])
1016 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1020 if (unlikely(protocol_failed(attr)))
1021 return -IPSET_ERR_PROTOCOL;
1023 if (!attr[IPSET_ATTR_SETNAME]) {
1024 for (i = 0; i < inst->ip_set_max; i++) {
1025 s = ip_set(inst, i);
1027 ip_set_flush_set(s);
1030 s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1034 ip_set_flush_set(s);
1042 static const struct nla_policy
1043 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = {
1044 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1045 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1046 .len = IPSET_MAXNAMELEN - 1 },
1047 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING,
1048 .len = IPSET_MAXNAMELEN - 1 },
1052 ip_set_rename(struct sock *ctnl, struct sk_buff *skb,
1053 const struct nlmsghdr *nlh,
1054 const struct nlattr * const attr[])
1056 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1057 struct ip_set *set, *s;
1062 if (unlikely(protocol_failed(attr) ||
1063 attr[IPSET_ATTR_SETNAME] == NULL ||
1064 attr[IPSET_ATTR_SETNAME2] == NULL))
1065 return -IPSET_ERR_PROTOCOL;
1067 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1071 read_lock_bh(&ip_set_ref_lock);
1072 if (set->ref != 0) {
1073 ret = -IPSET_ERR_REFERENCED;
1077 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]);
1078 for (i = 0; i < inst->ip_set_max; i++) {
1079 s = ip_set(inst, i);
1080 if (s != NULL && STREQ(s->name, name2)) {
1081 ret = -IPSET_ERR_EXIST_SETNAME2;
1085 strncpy(set->name, name2, IPSET_MAXNAMELEN);
1088 read_unlock_bh(&ip_set_ref_lock);
1092 /* Swap two sets so that name/index points to the other.
1093 * References and set names are also swapped.
1095 * The commands are serialized by the nfnl mutex and references are
1096 * protected by the ip_set_ref_lock. The kernel interfaces
1097 * do not hold the mutex but the pointer settings are atomic
1098 * so the ip_set_list always contains valid pointers to the sets.
1102 ip_set_swap(struct sock *ctnl, struct sk_buff *skb,
1103 const struct nlmsghdr *nlh,
1104 const struct nlattr * const attr[])
1106 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1107 struct ip_set *from, *to;
1108 ip_set_id_t from_id, to_id;
1109 char from_name[IPSET_MAXNAMELEN];
1111 if (unlikely(protocol_failed(attr) ||
1112 attr[IPSET_ATTR_SETNAME] == NULL ||
1113 attr[IPSET_ATTR_SETNAME2] == NULL))
1114 return -IPSET_ERR_PROTOCOL;
1116 from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1121 to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]),
1124 return -IPSET_ERR_EXIST_SETNAME2;
1126 /* Features must not change.
1127 * Not an artificial restriction anymore, as we must prevent
1128 * possible loops created by swapping in setlist type of sets. */
1129 if (!(from->type->features == to->type->features &&
1130 from->family == to->family))
1131 return -IPSET_ERR_TYPE_MISMATCH;
1133 strncpy(from_name, from->name, IPSET_MAXNAMELEN);
1134 strncpy(from->name, to->name, IPSET_MAXNAMELEN);
1135 strncpy(to->name, from_name, IPSET_MAXNAMELEN);
1137 write_lock_bh(&ip_set_ref_lock);
1138 swap(from->ref, to->ref);
1139 ip_set(inst, from_id) = to;
1140 ip_set(inst, to_id) = from;
1141 write_unlock_bh(&ip_set_ref_lock);
1146 /* List/save set data */
1153 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1154 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1157 ip_set_dump_done(struct netlink_callback *cb)
1159 struct ip_set_net *inst = (struct ip_set_net *)cb->args[IPSET_CB_NET];
1160 if (cb->args[IPSET_CB_ARG0]) {
1161 pr_debug("release set %s\n",
1162 ip_set(inst, cb->args[IPSET_CB_INDEX])->name);
1163 __ip_set_put_byindex(inst,
1164 (ip_set_id_t) cb->args[IPSET_CB_INDEX]);
1170 dump_attrs(struct nlmsghdr *nlh)
1172 const struct nlattr *attr;
1175 pr_debug("dump nlmsg\n");
1176 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) {
1177 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len);
1182 dump_init(struct netlink_callback *cb, struct ip_set_net *inst)
1184 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb);
1185 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1186 struct nlattr *cda[IPSET_ATTR_CMD_MAX+1];
1187 struct nlattr *attr = (void *)nlh + min_len;
1191 /* Second pass, so parser can't fail */
1192 nla_parse(cda, IPSET_ATTR_CMD_MAX,
1193 attr, nlh->nlmsg_len - min_len, ip_set_setname_policy);
1195 /* cb->args[IPSET_CB_NET]: net namespace
1196 * [IPSET_CB_DUMP]: dump single set/all sets
1197 * [IPSET_CB_INDEX]: set index
1198 * [IPSET_CB_ARG0]: type specific
1201 if (cda[IPSET_ATTR_SETNAME]) {
1204 set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]),
1209 dump_type = DUMP_ONE;
1210 cb->args[IPSET_CB_INDEX] = index;
1212 dump_type = DUMP_ALL;
1214 if (cda[IPSET_ATTR_FLAGS]) {
1215 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]);
1216 dump_type |= (f << 16);
1218 cb->args[IPSET_CB_NET] = (unsigned long)inst;
1219 cb->args[IPSET_CB_DUMP] = dump_type;
1225 ip_set_dump_start(struct sk_buff *skb, struct netlink_callback *cb)
1227 ip_set_id_t index = IPSET_INVALID_ID, max;
1228 struct ip_set *set = NULL;
1229 struct nlmsghdr *nlh = NULL;
1230 unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0;
1231 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1232 u32 dump_type, dump_flags;
1235 if (!cb->args[IPSET_CB_DUMP]) {
1236 ret = dump_init(cb, inst);
1238 nlh = nlmsg_hdr(cb->skb);
1239 /* We have to create and send the error message
1241 if (nlh->nlmsg_flags & NLM_F_ACK)
1242 netlink_ack(cb->skb, nlh, ret);
1247 if (cb->args[IPSET_CB_INDEX] >= inst->ip_set_max)
1250 dump_type = DUMP_TYPE(cb->args[IPSET_CB_DUMP]);
1251 dump_flags = DUMP_FLAGS(cb->args[IPSET_CB_DUMP]);
1252 max = dump_type == DUMP_ONE ? cb->args[IPSET_CB_INDEX] + 1
1255 pr_debug("dump type, flag: %u %u index: %ld\n",
1256 dump_type, dump_flags, cb->args[IPSET_CB_INDEX]);
1257 for (; cb->args[IPSET_CB_INDEX] < max; cb->args[IPSET_CB_INDEX]++) {
1258 index = (ip_set_id_t) cb->args[IPSET_CB_INDEX];
1259 set = ip_set(inst, index);
1261 if (dump_type == DUMP_ONE) {
1267 /* When dumping all sets, we must dump "sorted"
1268 * so that lists (unions of sets) are dumped last.
1270 if (dump_type != DUMP_ONE &&
1271 ((dump_type == DUMP_ALL) ==
1272 !!(set->type->features & IPSET_DUMP_LAST)))
1274 pr_debug("List set: %s\n", set->name);
1275 if (!cb->args[IPSET_CB_ARG0]) {
1276 /* Start listing: make sure set won't be destroyed */
1277 pr_debug("reference set\n");
1280 nlh = start_msg(skb, NETLINK_CB(cb->skb).portid,
1281 cb->nlh->nlmsg_seq, flags,
1285 goto release_refcount;
1287 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1288 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name))
1289 goto nla_put_failure;
1290 if (dump_flags & IPSET_FLAG_LIST_SETNAME)
1292 switch (cb->args[IPSET_CB_ARG0]) {
1294 /* Core header data */
1295 if (nla_put_string(skb, IPSET_ATTR_TYPENAME,
1297 nla_put_u8(skb, IPSET_ATTR_FAMILY,
1299 nla_put_u8(skb, IPSET_ATTR_REVISION,
1301 goto nla_put_failure;
1302 ret = set->variant->head(set, skb);
1304 goto release_refcount;
1305 if (dump_flags & IPSET_FLAG_LIST_HEADER)
1307 /* Fall through and add elements */
1309 read_lock_bh(&set->lock);
1310 ret = set->variant->list(set, skb, cb);
1311 read_unlock_bh(&set->lock);
1312 if (!cb->args[IPSET_CB_ARG0])
1313 /* Set is done, proceed with next one */
1315 goto release_refcount;
1318 /* If we dump all sets, continue with dumping last ones */
1319 if (dump_type == DUMP_ALL) {
1320 dump_type = DUMP_LAST;
1321 cb->args[IPSET_CB_DUMP] = dump_type | (dump_flags << 16);
1322 cb->args[IPSET_CB_INDEX] = 0;
1330 if (dump_type == DUMP_ONE)
1331 cb->args[IPSET_CB_INDEX] = IPSET_INVALID_ID;
1333 cb->args[IPSET_CB_INDEX]++;
1335 /* If there was an error or set is done, release set */
1336 if (ret || !cb->args[IPSET_CB_ARG0]) {
1337 pr_debug("release set %s\n", ip_set(inst, index)->name);
1338 __ip_set_put_byindex(inst, index);
1339 cb->args[IPSET_CB_ARG0] = 0;
1343 nlmsg_end(skb, nlh);
1344 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len);
1348 return ret < 0 ? ret : skb->len;
1352 ip_set_dump(struct sock *ctnl, struct sk_buff *skb,
1353 const struct nlmsghdr *nlh,
1354 const struct nlattr * const attr[])
1356 if (unlikely(protocol_failed(attr)))
1357 return -IPSET_ERR_PROTOCOL;
1360 struct netlink_dump_control c = {
1361 .dump = ip_set_dump_start,
1362 .done = ip_set_dump_done,
1364 return netlink_dump_start(ctnl, skb, nlh, &c);
1368 /* Add, del and test */
1370 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = {
1371 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1372 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1373 .len = IPSET_MAXNAMELEN - 1 },
1374 [IPSET_ATTR_LINENO] = { .type = NLA_U32 },
1375 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
1376 [IPSET_ATTR_ADT] = { .type = NLA_NESTED },
1380 call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set,
1381 struct nlattr *tb[], enum ipset_adt adt,
1382 u32 flags, bool use_lineno)
1386 bool eexist = flags & IPSET_FLAG_EXIST, retried = false;
1389 write_lock_bh(&set->lock);
1390 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried);
1391 write_unlock_bh(&set->lock);
1393 } while (ret == -EAGAIN &&
1394 set->variant->resize &&
1395 (ret = set->variant->resize(set, retried)) == 0);
1397 if (!ret || (ret == -IPSET_ERR_EXIST && eexist))
1399 if (lineno && use_lineno) {
1400 /* Error in restore/batch mode: send back lineno */
1401 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb);
1402 struct sk_buff *skb2;
1403 struct nlmsgerr *errmsg;
1404 size_t payload = sizeof(*errmsg) + nlmsg_len(nlh);
1405 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1406 struct nlattr *cda[IPSET_ATTR_CMD_MAX+1];
1407 struct nlattr *cmdattr;
1410 skb2 = nlmsg_new(payload, GFP_KERNEL);
1413 rep = __nlmsg_put(skb2, NETLINK_CB(skb).portid,
1414 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0);
1415 errmsg = nlmsg_data(rep);
1416 errmsg->error = ret;
1417 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len);
1418 cmdattr = (void *)&errmsg->msg + min_len;
1420 nla_parse(cda, IPSET_ATTR_CMD_MAX,
1421 cmdattr, nlh->nlmsg_len - min_len,
1424 errline = nla_data(cda[IPSET_ATTR_LINENO]);
1428 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1429 /* Signal netlink not to send its ACK/errmsg. */
1437 ip_set_uadd(struct sock *ctnl, struct sk_buff *skb,
1438 const struct nlmsghdr *nlh,
1439 const struct nlattr * const attr[])
1441 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1443 struct nlattr *tb[IPSET_ATTR_ADT_MAX+1] = {};
1444 const struct nlattr *nla;
1445 u32 flags = flag_exist(nlh);
1449 if (unlikely(protocol_failed(attr) ||
1450 attr[IPSET_ATTR_SETNAME] == NULL ||
1451 !((attr[IPSET_ATTR_DATA] != NULL) ^
1452 (attr[IPSET_ATTR_ADT] != NULL)) ||
1453 (attr[IPSET_ATTR_DATA] != NULL &&
1454 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1455 (attr[IPSET_ATTR_ADT] != NULL &&
1456 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1457 attr[IPSET_ATTR_LINENO] == NULL))))
1458 return -IPSET_ERR_PROTOCOL;
1460 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1464 use_lineno = !!attr[IPSET_ATTR_LINENO];
1465 if (attr[IPSET_ATTR_DATA]) {
1466 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1467 attr[IPSET_ATTR_DATA],
1468 set->type->adt_policy))
1469 return -IPSET_ERR_PROTOCOL;
1470 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD, flags,
1475 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1476 memset(tb, 0, sizeof(tb));
1477 if (nla_type(nla) != IPSET_ATTR_DATA ||
1478 !flag_nested(nla) ||
1479 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1480 set->type->adt_policy))
1481 return -IPSET_ERR_PROTOCOL;
1482 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD,
1492 ip_set_udel(struct sock *ctnl, struct sk_buff *skb,
1493 const struct nlmsghdr *nlh,
1494 const struct nlattr * const attr[])
1496 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1498 struct nlattr *tb[IPSET_ATTR_ADT_MAX+1] = {};
1499 const struct nlattr *nla;
1500 u32 flags = flag_exist(nlh);
1504 if (unlikely(protocol_failed(attr) ||
1505 attr[IPSET_ATTR_SETNAME] == NULL ||
1506 !((attr[IPSET_ATTR_DATA] != NULL) ^
1507 (attr[IPSET_ATTR_ADT] != NULL)) ||
1508 (attr[IPSET_ATTR_DATA] != NULL &&
1509 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1510 (attr[IPSET_ATTR_ADT] != NULL &&
1511 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1512 attr[IPSET_ATTR_LINENO] == NULL))))
1513 return -IPSET_ERR_PROTOCOL;
1515 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1519 use_lineno = !!attr[IPSET_ATTR_LINENO];
1520 if (attr[IPSET_ATTR_DATA]) {
1521 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1522 attr[IPSET_ATTR_DATA],
1523 set->type->adt_policy))
1524 return -IPSET_ERR_PROTOCOL;
1525 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL, flags,
1530 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1531 memset(tb, 0, sizeof(*tb));
1532 if (nla_type(nla) != IPSET_ATTR_DATA ||
1533 !flag_nested(nla) ||
1534 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1535 set->type->adt_policy))
1536 return -IPSET_ERR_PROTOCOL;
1537 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL,
1547 ip_set_utest(struct sock *ctnl, struct sk_buff *skb,
1548 const struct nlmsghdr *nlh,
1549 const struct nlattr * const attr[])
1551 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1553 struct nlattr *tb[IPSET_ATTR_ADT_MAX+1] = {};
1556 if (unlikely(protocol_failed(attr) ||
1557 attr[IPSET_ATTR_SETNAME] == NULL ||
1558 attr[IPSET_ATTR_DATA] == NULL ||
1559 !flag_nested(attr[IPSET_ATTR_DATA])))
1560 return -IPSET_ERR_PROTOCOL;
1562 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1566 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA],
1567 set->type->adt_policy))
1568 return -IPSET_ERR_PROTOCOL;
1570 read_lock_bh(&set->lock);
1571 ret = set->variant->uadt(set, tb, IPSET_TEST, NULL, 0, 0);
1572 read_unlock_bh(&set->lock);
1573 /* Userspace can't trigger element to be re-added */
1577 return ret > 0 ? 0 : -IPSET_ERR_EXIST;
1580 /* Get headed data of a set */
1583 ip_set_header(struct sock *ctnl, struct sk_buff *skb,
1584 const struct nlmsghdr *nlh,
1585 const struct nlattr * const attr[])
1587 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1588 const struct ip_set *set;
1589 struct sk_buff *skb2;
1590 struct nlmsghdr *nlh2;
1593 if (unlikely(protocol_failed(attr) ||
1594 attr[IPSET_ATTR_SETNAME] == NULL))
1595 return -IPSET_ERR_PROTOCOL;
1597 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1601 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1605 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1609 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1610 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) ||
1611 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) ||
1612 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
1613 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision))
1614 goto nla_put_failure;
1615 nlmsg_end(skb2, nlh2);
1617 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1624 nlmsg_cancel(skb2, nlh2);
1632 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = {
1633 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1634 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
1635 .len = IPSET_MAXNAMELEN - 1 },
1636 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
1640 ip_set_type(struct sock *ctnl, struct sk_buff *skb,
1641 const struct nlmsghdr *nlh,
1642 const struct nlattr * const attr[])
1644 struct sk_buff *skb2;
1645 struct nlmsghdr *nlh2;
1646 u8 family, min, max;
1647 const char *typename;
1650 if (unlikely(protocol_failed(attr) ||
1651 attr[IPSET_ATTR_TYPENAME] == NULL ||
1652 attr[IPSET_ATTR_FAMILY] == NULL))
1653 return -IPSET_ERR_PROTOCOL;
1655 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1656 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1657 ret = find_set_type_minmax(typename, family, &min, &max);
1661 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1665 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1669 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1670 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) ||
1671 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) ||
1672 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) ||
1673 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min))
1674 goto nla_put_failure;
1675 nlmsg_end(skb2, nlh2);
1677 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len);
1678 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1685 nlmsg_cancel(skb2, nlh2);
1691 /* Get protocol version */
1693 static const struct nla_policy
1694 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = {
1695 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1699 ip_set_protocol(struct sock *ctnl, struct sk_buff *skb,
1700 const struct nlmsghdr *nlh,
1701 const struct nlattr * const attr[])
1703 struct sk_buff *skb2;
1704 struct nlmsghdr *nlh2;
1707 if (unlikely(attr[IPSET_ATTR_PROTOCOL] == NULL))
1708 return -IPSET_ERR_PROTOCOL;
1710 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1714 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1715 IPSET_CMD_PROTOCOL);
1718 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL))
1719 goto nla_put_failure;
1720 nlmsg_end(skb2, nlh2);
1722 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1729 nlmsg_cancel(skb2, nlh2);
1735 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = {
1736 [IPSET_CMD_NONE] = {
1737 .call = ip_set_none,
1738 .attr_count = IPSET_ATTR_CMD_MAX,
1740 [IPSET_CMD_CREATE] = {
1741 .call = ip_set_create,
1742 .attr_count = IPSET_ATTR_CMD_MAX,
1743 .policy = ip_set_create_policy,
1745 [IPSET_CMD_DESTROY] = {
1746 .call = ip_set_destroy,
1747 .attr_count = IPSET_ATTR_CMD_MAX,
1748 .policy = ip_set_setname_policy,
1750 [IPSET_CMD_FLUSH] = {
1751 .call = ip_set_flush,
1752 .attr_count = IPSET_ATTR_CMD_MAX,
1753 .policy = ip_set_setname_policy,
1755 [IPSET_CMD_RENAME] = {
1756 .call = ip_set_rename,
1757 .attr_count = IPSET_ATTR_CMD_MAX,
1758 .policy = ip_set_setname2_policy,
1760 [IPSET_CMD_SWAP] = {
1761 .call = ip_set_swap,
1762 .attr_count = IPSET_ATTR_CMD_MAX,
1763 .policy = ip_set_setname2_policy,
1765 [IPSET_CMD_LIST] = {
1766 .call = ip_set_dump,
1767 .attr_count = IPSET_ATTR_CMD_MAX,
1768 .policy = ip_set_setname_policy,
1770 [IPSET_CMD_SAVE] = {
1771 .call = ip_set_dump,
1772 .attr_count = IPSET_ATTR_CMD_MAX,
1773 .policy = ip_set_setname_policy,
1776 .call = ip_set_uadd,
1777 .attr_count = IPSET_ATTR_CMD_MAX,
1778 .policy = ip_set_adt_policy,
1781 .call = ip_set_udel,
1782 .attr_count = IPSET_ATTR_CMD_MAX,
1783 .policy = ip_set_adt_policy,
1785 [IPSET_CMD_TEST] = {
1786 .call = ip_set_utest,
1787 .attr_count = IPSET_ATTR_CMD_MAX,
1788 .policy = ip_set_adt_policy,
1790 [IPSET_CMD_HEADER] = {
1791 .call = ip_set_header,
1792 .attr_count = IPSET_ATTR_CMD_MAX,
1793 .policy = ip_set_setname_policy,
1795 [IPSET_CMD_TYPE] = {
1796 .call = ip_set_type,
1797 .attr_count = IPSET_ATTR_CMD_MAX,
1798 .policy = ip_set_type_policy,
1800 [IPSET_CMD_PROTOCOL] = {
1801 .call = ip_set_protocol,
1802 .attr_count = IPSET_ATTR_CMD_MAX,
1803 .policy = ip_set_protocol_policy,
1807 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = {
1809 .subsys_id = NFNL_SUBSYS_IPSET,
1810 .cb_count = IPSET_MSG_MAX,
1811 .cb = ip_set_netlink_subsys_cb,
1814 /* Interface to iptables/ip6tables */
1817 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len)
1821 int copylen = *len, ret = 0;
1822 struct net *net = sock_net(sk);
1823 struct ip_set_net *inst = ip_set_pernet(net);
1825 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1827 if (optval != SO_IP_SET)
1829 if (*len < sizeof(unsigned int))
1832 data = vmalloc(*len);
1835 if (copy_from_user(data, user, *len) != 0) {
1839 op = (unsigned int *) data;
1841 if (*op < IP_SET_OP_VERSION) {
1842 /* Check the version at the beginning of operations */
1843 struct ip_set_req_version *req_version = data;
1844 if (req_version->version != IPSET_PROTOCOL) {
1851 case IP_SET_OP_VERSION: {
1852 struct ip_set_req_version *req_version = data;
1854 if (*len != sizeof(struct ip_set_req_version)) {
1859 req_version->version = IPSET_PROTOCOL;
1860 ret = copy_to_user(user, req_version,
1861 sizeof(struct ip_set_req_version));
1864 case IP_SET_OP_GET_BYNAME: {
1865 struct ip_set_req_get_set *req_get = data;
1868 if (*len != sizeof(struct ip_set_req_get_set)) {
1872 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
1873 nfnl_lock(NFNL_SUBSYS_IPSET);
1874 find_set_and_id(inst, req_get->set.name, &id);
1875 req_get->set.index = id;
1876 nfnl_unlock(NFNL_SUBSYS_IPSET);
1879 case IP_SET_OP_GET_FNAME: {
1880 struct ip_set_req_get_set_family *req_get = data;
1883 if (*len != sizeof(struct ip_set_req_get_set_family)) {
1887 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
1888 nfnl_lock(NFNL_SUBSYS_IPSET);
1889 find_set_and_id(inst, req_get->set.name, &id);
1890 req_get->set.index = id;
1891 if (id != IPSET_INVALID_ID)
1892 req_get->family = ip_set(inst, id)->family;
1893 nfnl_unlock(NFNL_SUBSYS_IPSET);
1896 case IP_SET_OP_GET_BYINDEX: {
1897 struct ip_set_req_get_set *req_get = data;
1900 if (*len != sizeof(struct ip_set_req_get_set) ||
1901 req_get->set.index >= inst->ip_set_max) {
1905 nfnl_lock(NFNL_SUBSYS_IPSET);
1906 set = ip_set(inst, req_get->set.index);
1907 strncpy(req_get->set.name, set ? set->name : "",
1909 nfnl_unlock(NFNL_SUBSYS_IPSET);
1915 } /* end of switch(op) */
1918 ret = copy_to_user(user, data, copylen);
1927 static struct nf_sockopt_ops so_set __read_mostly = {
1929 .get_optmin = SO_IP_SET,
1930 .get_optmax = SO_IP_SET + 1,
1931 .get = &ip_set_sockfn_get,
1932 .owner = THIS_MODULE,
1935 static int __net_init
1936 ip_set_net_init(struct net *net)
1938 struct ip_set_net *inst = ip_set_pernet(net);
1939 struct ip_set **list;
1941 inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX;
1942 if (inst->ip_set_max >= IPSET_INVALID_ID)
1943 inst->ip_set_max = IPSET_INVALID_ID - 1;
1945 list = kzalloc(sizeof(struct ip_set *) * inst->ip_set_max, GFP_KERNEL);
1948 inst->is_deleted = 0;
1949 rcu_assign_pointer(inst->ip_set_list, list);
1953 static void __net_exit
1954 ip_set_net_exit(struct net *net)
1956 struct ip_set_net *inst = ip_set_pernet(net);
1958 struct ip_set *set = NULL;
1961 inst->is_deleted = 1; /* flag for ip_set_nfnl_put */
1963 for (i = 0; i < inst->ip_set_max; i++) {
1964 set = ip_set(inst, i);
1966 ip_set_destroy_set(inst, i);
1968 kfree(rcu_dereference_protected(inst->ip_set_list, 1));
1971 static struct pernet_operations ip_set_net_ops = {
1972 .init = ip_set_net_init,
1973 .exit = ip_set_net_exit,
1974 .id = &ip_set_net_id,
1975 .size = sizeof(struct ip_set_net)
1982 int ret = nfnetlink_subsys_register(&ip_set_netlink_subsys);
1984 pr_err("ip_set: cannot register with nfnetlink.\n");
1987 ret = nf_register_sockopt(&so_set);
1989 pr_err("SO_SET registry failed: %d\n", ret);
1990 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
1993 ret = register_pernet_subsys(&ip_set_net_ops);
1995 pr_err("ip_set: cannot register pernet_subsys.\n");
1996 nf_unregister_sockopt(&so_set);
1997 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2000 pr_info("ip_set: protocol %u\n", IPSET_PROTOCOL);
2007 unregister_pernet_subsys(&ip_set_net_ops);
2008 nf_unregister_sockopt(&so_set);
2009 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2010 pr_debug("these are the famous last words\n");
2013 module_init(ip_set_init);
2014 module_exit(ip_set_fini);