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 nfnl_dereference(p) \
58 rcu_dereference_protected(p, 1)
59 #define nfnl_set(inst, id) \
60 nfnl_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 for (id = 0; id < IPSET_EXT_ID_MAX; id++) {
372 if (!add_extension(id, cadt_flags, tb))
374 offset += ALIGN(len + offset, ip_set_extensions[id].align);
375 set->offset[id] = offset;
376 set->extensions |= ip_set_extensions[id].type;
377 offset += ip_set_extensions[id].len;
381 EXPORT_SYMBOL_GPL(ip_set_elem_len);
384 ip_set_get_extensions(struct ip_set *set, struct nlattr *tb[],
385 struct ip_set_ext *ext)
387 if (tb[IPSET_ATTR_TIMEOUT]) {
388 if (!(set->extensions & IPSET_EXT_TIMEOUT))
389 return -IPSET_ERR_TIMEOUT;
390 ext->timeout = ip_set_timeout_uget(tb[IPSET_ATTR_TIMEOUT]);
392 if (tb[IPSET_ATTR_BYTES] || tb[IPSET_ATTR_PACKETS]) {
393 if (!(set->extensions & IPSET_EXT_COUNTER))
394 return -IPSET_ERR_COUNTER;
395 if (tb[IPSET_ATTR_BYTES])
396 ext->bytes = be64_to_cpu(nla_get_be64(
397 tb[IPSET_ATTR_BYTES]));
398 if (tb[IPSET_ATTR_PACKETS])
399 ext->packets = be64_to_cpu(nla_get_be64(
400 tb[IPSET_ATTR_PACKETS]));
402 if (tb[IPSET_ATTR_COMMENT]) {
403 if (!(set->extensions & IPSET_EXT_COMMENT))
404 return -IPSET_ERR_COMMENT;
405 ext->comment = ip_set_comment_uget(tb[IPSET_ATTR_COMMENT]);
410 EXPORT_SYMBOL_GPL(ip_set_get_extensions);
413 * Creating/destroying/renaming/swapping affect the existence and
414 * the properties of a set. All of these can be executed from userspace
415 * only and serialized by the nfnl mutex indirectly from nfnetlink.
417 * Sets are identified by their index in ip_set_list and the index
418 * is used by the external references (set/SET netfilter modules).
420 * The set behind an index may change by swapping only, from userspace.
424 __ip_set_get(struct ip_set *set)
426 write_lock_bh(&ip_set_ref_lock);
428 write_unlock_bh(&ip_set_ref_lock);
432 __ip_set_put(struct ip_set *set)
434 write_lock_bh(&ip_set_ref_lock);
435 BUG_ON(set->ref == 0);
437 write_unlock_bh(&ip_set_ref_lock);
441 * Add, del and test set entries from kernel.
443 * The set behind the index must exist and must be referenced
444 * so it can't be destroyed (or changed) under our foot.
447 static inline struct ip_set *
448 ip_set_rcu_get(struct net *net, ip_set_id_t index)
451 struct ip_set_net *inst = ip_set_pernet(net);
454 /* ip_set_list itself needs to be protected */
455 set = rcu_dereference(inst->ip_set_list)[index];
462 ip_set_test(ip_set_id_t index, const struct sk_buff *skb,
463 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
465 struct ip_set *set = ip_set_rcu_get(
466 dev_net(par->in ? par->in : par->out), index);
470 pr_debug("set %s, index %u\n", set->name, index);
472 if (opt->dim < set->type->dimension ||
473 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
476 read_lock_bh(&set->lock);
477 ret = set->variant->kadt(set, skb, par, IPSET_TEST, opt);
478 read_unlock_bh(&set->lock);
480 if (ret == -EAGAIN) {
481 /* Type requests element to be completed */
482 pr_debug("element must be competed, ADD is triggered\n");
483 write_lock_bh(&set->lock);
484 set->variant->kadt(set, skb, par, IPSET_ADD, opt);
485 write_unlock_bh(&set->lock);
488 /* --return-nomatch: invert matched element */
489 if ((opt->cmdflags & IPSET_FLAG_RETURN_NOMATCH) &&
490 (set->type->features & IPSET_TYPE_NOMATCH) &&
491 (ret > 0 || ret == -ENOTEMPTY))
495 /* Convert error codes to nomatch */
496 return (ret < 0 ? 0 : ret);
498 EXPORT_SYMBOL_GPL(ip_set_test);
501 ip_set_add(ip_set_id_t index, const struct sk_buff *skb,
502 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
504 struct ip_set *set = ip_set_rcu_get(
505 dev_net(par->in ? par->in : par->out), index);
509 pr_debug("set %s, index %u\n", set->name, index);
511 if (opt->dim < set->type->dimension ||
512 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
515 write_lock_bh(&set->lock);
516 ret = set->variant->kadt(set, skb, par, IPSET_ADD, opt);
517 write_unlock_bh(&set->lock);
521 EXPORT_SYMBOL_GPL(ip_set_add);
524 ip_set_del(ip_set_id_t index, const struct sk_buff *skb,
525 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
527 struct ip_set *set = ip_set_rcu_get(
528 dev_net(par->in ? par->in : par->out), index);
532 pr_debug("set %s, index %u\n", set->name, index);
534 if (opt->dim < set->type->dimension ||
535 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
538 write_lock_bh(&set->lock);
539 ret = set->variant->kadt(set, skb, par, IPSET_DEL, opt);
540 write_unlock_bh(&set->lock);
544 EXPORT_SYMBOL_GPL(ip_set_del);
547 * Find set by name, reference it once. The reference makes sure the
548 * thing pointed to, does not go away under our feet.
552 ip_set_get_byname(struct net *net, const char *name, struct ip_set **set)
554 ip_set_id_t i, index = IPSET_INVALID_ID;
556 struct ip_set_net *inst = ip_set_pernet(net);
559 for (i = 0; i < inst->ip_set_max; i++) {
560 s = rcu_dereference(inst->ip_set_list)[i];
561 if (s != NULL && STREQ(s->name, name)) {
572 EXPORT_SYMBOL_GPL(ip_set_get_byname);
575 * If the given set pointer points to a valid set, decrement
576 * reference count by 1. The caller shall not assume the index
577 * to be valid, after calling this function.
582 __ip_set_put_byindex(struct ip_set_net *inst, ip_set_id_t index)
587 set = rcu_dereference(inst->ip_set_list)[index];
594 ip_set_put_byindex(struct net *net, ip_set_id_t index)
596 struct ip_set_net *inst = ip_set_pernet(net);
598 __ip_set_put_byindex(inst, index);
600 EXPORT_SYMBOL_GPL(ip_set_put_byindex);
603 * Get the name of a set behind a set index.
604 * We assume the set is referenced, so it does exist and
605 * can't be destroyed. The set cannot be renamed due to
606 * the referencing either.
610 ip_set_name_byindex(struct net *net, ip_set_id_t index)
612 const struct ip_set *set = ip_set_rcu_get(net, index);
615 BUG_ON(set->ref == 0);
617 /* Referenced, so it's safe */
620 EXPORT_SYMBOL_GPL(ip_set_name_byindex);
623 * Routines to call by external subsystems, which do not
624 * call nfnl_lock for us.
628 * Find set by name, reference it once. The reference makes sure the
629 * thing pointed to, does not go away under our feet.
631 * The nfnl mutex is used in the function.
634 ip_set_nfnl_get(struct net *net, const char *name)
636 ip_set_id_t i, index = IPSET_INVALID_ID;
638 struct ip_set_net *inst = ip_set_pernet(net);
640 nfnl_lock(NFNL_SUBSYS_IPSET);
641 for (i = 0; i < inst->ip_set_max; i++) {
642 s = nfnl_set(inst, i);
643 if (s != NULL && STREQ(s->name, name)) {
649 nfnl_unlock(NFNL_SUBSYS_IPSET);
653 EXPORT_SYMBOL_GPL(ip_set_nfnl_get);
656 * Find set by index, reference it once. The reference makes sure the
657 * thing pointed to, does not go away under our feet.
659 * The nfnl mutex is used in the function.
662 ip_set_nfnl_get_byindex(struct net *net, ip_set_id_t index)
665 struct ip_set_net *inst = ip_set_pernet(net);
667 if (index > inst->ip_set_max)
668 return IPSET_INVALID_ID;
670 nfnl_lock(NFNL_SUBSYS_IPSET);
671 set = nfnl_set(inst, index);
675 index = IPSET_INVALID_ID;
676 nfnl_unlock(NFNL_SUBSYS_IPSET);
680 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex);
683 * If the given set pointer points to a valid set, decrement
684 * reference count by 1. The caller shall not assume the index
685 * to be valid, after calling this function.
687 * The nfnl mutex is used in the function.
690 ip_set_nfnl_put(struct net *net, ip_set_id_t index)
693 struct ip_set_net *inst = ip_set_pernet(net);
695 nfnl_lock(NFNL_SUBSYS_IPSET);
696 if (!inst->is_deleted) { /* already deleted from ip_set_net_exit() */
697 set = nfnl_set(inst, index);
701 nfnl_unlock(NFNL_SUBSYS_IPSET);
703 EXPORT_SYMBOL_GPL(ip_set_nfnl_put);
706 * Communication protocol with userspace over netlink.
708 * The commands are serialized by the nfnl mutex.
712 protocol_failed(const struct nlattr * const tb[])
714 return !tb[IPSET_ATTR_PROTOCOL] ||
715 nla_get_u8(tb[IPSET_ATTR_PROTOCOL]) != IPSET_PROTOCOL;
719 flag_exist(const struct nlmsghdr *nlh)
721 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST;
724 static struct nlmsghdr *
725 start_msg(struct sk_buff *skb, u32 portid, u32 seq, unsigned int flags,
728 struct nlmsghdr *nlh;
729 struct nfgenmsg *nfmsg;
731 nlh = nlmsg_put(skb, portid, seq, cmd | (NFNL_SUBSYS_IPSET << 8),
732 sizeof(*nfmsg), flags);
736 nfmsg = nlmsg_data(nlh);
737 nfmsg->nfgen_family = NFPROTO_IPV4;
738 nfmsg->version = NFNETLINK_V0;
746 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = {
747 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
748 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
749 .len = IPSET_MAXNAMELEN - 1 },
750 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
751 .len = IPSET_MAXNAMELEN - 1},
752 [IPSET_ATTR_REVISION] = { .type = NLA_U8 },
753 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
754 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
757 static struct ip_set *
758 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id)
760 struct ip_set *set = NULL;
763 *id = IPSET_INVALID_ID;
764 for (i = 0; i < inst->ip_set_max; i++) {
765 set = nfnl_set(inst, i);
766 if (set != NULL && STREQ(set->name, name)) {
771 return (*id == IPSET_INVALID_ID ? NULL : set);
774 static inline struct ip_set *
775 find_set(struct ip_set_net *inst, const char *name)
779 return find_set_and_id(inst, name, &id);
783 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index,
789 *index = IPSET_INVALID_ID;
790 for (i = 0; i < inst->ip_set_max; i++) {
791 s = nfnl_set(inst, i);
793 if (*index == IPSET_INVALID_ID)
795 } else if (STREQ(name, s->name)) {
801 if (*index == IPSET_INVALID_ID)
802 /* No free slot remained */
803 return -IPSET_ERR_MAX_SETS;
808 ip_set_none(struct sock *ctnl, struct sk_buff *skb,
809 const struct nlmsghdr *nlh,
810 const struct nlattr * const attr[])
816 ip_set_create(struct sock *ctnl, struct sk_buff *skb,
817 const struct nlmsghdr *nlh,
818 const struct nlattr * const attr[])
820 struct net *net = sock_net(ctnl);
821 struct ip_set_net *inst = ip_set_pernet(net);
822 struct ip_set *set, *clash = NULL;
823 ip_set_id_t index = IPSET_INVALID_ID;
824 struct nlattr *tb[IPSET_ATTR_CREATE_MAX+1] = {};
825 const char *name, *typename;
827 u32 flags = flag_exist(nlh);
830 if (unlikely(protocol_failed(attr) ||
831 attr[IPSET_ATTR_SETNAME] == NULL ||
832 attr[IPSET_ATTR_TYPENAME] == NULL ||
833 attr[IPSET_ATTR_REVISION] == NULL ||
834 attr[IPSET_ATTR_FAMILY] == NULL ||
835 (attr[IPSET_ATTR_DATA] != NULL &&
836 !flag_nested(attr[IPSET_ATTR_DATA]))))
837 return -IPSET_ERR_PROTOCOL;
839 name = nla_data(attr[IPSET_ATTR_SETNAME]);
840 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
841 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
842 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]);
843 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
844 name, typename, family_name(family), revision);
847 * First, and without any locks, allocate and initialize
848 * a normal base set structure.
850 set = kzalloc(sizeof(struct ip_set), GFP_KERNEL);
853 rwlock_init(&set->lock);
854 strlcpy(set->name, name, IPSET_MAXNAMELEN);
855 set->family = family;
856 set->revision = revision;
859 * Next, check that we know the type, and take
860 * a reference on the type, to make sure it stays available
861 * while constructing our new set.
863 * After referencing the type, we try to create the type
864 * specific part of the set without holding any locks.
866 ret = find_set_type_get(typename, family, revision, &(set->type));
871 * Without holding any locks, create private part.
873 if (attr[IPSET_ATTR_DATA] &&
874 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA],
875 set->type->create_policy)) {
876 ret = -IPSET_ERR_PROTOCOL;
880 ret = set->type->create(net, set, tb, flags);
884 /* BTW, ret==0 here. */
887 * Here, we have a valid, constructed set and we are protected
888 * by the nfnl mutex. Find the first free index in ip_set_list
889 * and check clashing.
891 ret = find_free_id(inst, set->name, &index, &clash);
892 if (ret == -EEXIST) {
893 /* If this is the same set and requested, ignore error */
894 if ((flags & IPSET_FLAG_EXIST) &&
895 STREQ(set->type->name, clash->type->name) &&
896 set->type->family == clash->type->family &&
897 set->type->revision_min == clash->type->revision_min &&
898 set->type->revision_max == clash->type->revision_max &&
899 set->variant->same_set(set, clash))
902 } else if (ret == -IPSET_ERR_MAX_SETS) {
903 struct ip_set **list, **tmp;
904 ip_set_id_t i = inst->ip_set_max + IP_SET_INC;
906 if (i < inst->ip_set_max || i == IPSET_INVALID_ID)
910 list = kzalloc(sizeof(struct ip_set *) * i, GFP_KERNEL);
913 /* nfnl mutex is held, both lists are valid */
914 tmp = nfnl_dereference(inst->ip_set_list);
915 memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max);
916 rcu_assign_pointer(inst->ip_set_list, list);
917 /* Make sure all current packets have passed through */
920 index = inst->ip_set_max;
921 inst->ip_set_max = i;
928 * Finally! Add our shiny new set to the list, and be done.
930 pr_debug("create: '%s' created with index %u!\n", set->name, index);
931 nfnl_set(inst, index) = set;
936 set->variant->destroy(set);
938 module_put(set->type->me);
946 static const struct nla_policy
947 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = {
948 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
949 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
950 .len = IPSET_MAXNAMELEN - 1 },
954 ip_set_destroy_set(struct ip_set_net *inst, ip_set_id_t index)
956 struct ip_set *set = nfnl_set(inst, index);
958 pr_debug("set: %s\n", set->name);
959 nfnl_set(inst, index) = NULL;
961 /* Must call it without holding any lock */
962 set->variant->destroy(set);
963 module_put(set->type->me);
968 ip_set_destroy(struct sock *ctnl, struct sk_buff *skb,
969 const struct nlmsghdr *nlh,
970 const struct nlattr * const attr[])
972 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
977 if (unlikely(protocol_failed(attr)))
978 return -IPSET_ERR_PROTOCOL;
980 /* Commands are serialized and references are
981 * protected by the ip_set_ref_lock.
982 * External systems (i.e. xt_set) must call
983 * ip_set_put|get_nfnl_* functions, that way we
984 * can safely check references here.
986 * list:set timer can only decrement the reference
987 * counter, so if it's already zero, we can proceed
988 * without holding the lock.
990 read_lock_bh(&ip_set_ref_lock);
991 if (!attr[IPSET_ATTR_SETNAME]) {
992 for (i = 0; i < inst->ip_set_max; i++) {
993 s = nfnl_set(inst, i);
994 if (s != NULL && s->ref) {
995 ret = -IPSET_ERR_BUSY;
999 read_unlock_bh(&ip_set_ref_lock);
1000 for (i = 0; i < inst->ip_set_max; i++) {
1001 s = nfnl_set(inst, i);
1003 ip_set_destroy_set(inst, i);
1006 s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1011 } else if (s->ref) {
1012 ret = -IPSET_ERR_BUSY;
1015 read_unlock_bh(&ip_set_ref_lock);
1017 ip_set_destroy_set(inst, i);
1021 read_unlock_bh(&ip_set_ref_lock);
1028 ip_set_flush_set(struct ip_set *set)
1030 pr_debug("set: %s\n", set->name);
1032 write_lock_bh(&set->lock);
1033 set->variant->flush(set);
1034 write_unlock_bh(&set->lock);
1038 ip_set_flush(struct sock *ctnl, struct sk_buff *skb,
1039 const struct nlmsghdr *nlh,
1040 const struct nlattr * const attr[])
1042 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1046 if (unlikely(protocol_failed(attr)))
1047 return -IPSET_ERR_PROTOCOL;
1049 if (!attr[IPSET_ATTR_SETNAME]) {
1050 for (i = 0; i < inst->ip_set_max; i++) {
1051 s = nfnl_set(inst, i);
1053 ip_set_flush_set(s);
1056 s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1060 ip_set_flush_set(s);
1068 static const struct nla_policy
1069 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = {
1070 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1071 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1072 .len = IPSET_MAXNAMELEN - 1 },
1073 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING,
1074 .len = IPSET_MAXNAMELEN - 1 },
1078 ip_set_rename(struct sock *ctnl, struct sk_buff *skb,
1079 const struct nlmsghdr *nlh,
1080 const struct nlattr * const attr[])
1082 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1083 struct ip_set *set, *s;
1088 if (unlikely(protocol_failed(attr) ||
1089 attr[IPSET_ATTR_SETNAME] == NULL ||
1090 attr[IPSET_ATTR_SETNAME2] == NULL))
1091 return -IPSET_ERR_PROTOCOL;
1093 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1097 read_lock_bh(&ip_set_ref_lock);
1098 if (set->ref != 0) {
1099 ret = -IPSET_ERR_REFERENCED;
1103 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]);
1104 for (i = 0; i < inst->ip_set_max; i++) {
1105 s = nfnl_set(inst, i);
1106 if (s != NULL && STREQ(s->name, name2)) {
1107 ret = -IPSET_ERR_EXIST_SETNAME2;
1111 strncpy(set->name, name2, IPSET_MAXNAMELEN);
1114 read_unlock_bh(&ip_set_ref_lock);
1118 /* Swap two sets so that name/index points to the other.
1119 * References and set names are also swapped.
1121 * The commands are serialized by the nfnl mutex and references are
1122 * protected by the ip_set_ref_lock. The kernel interfaces
1123 * do not hold the mutex but the pointer settings are atomic
1124 * so the ip_set_list always contains valid pointers to the sets.
1128 ip_set_swap(struct sock *ctnl, struct sk_buff *skb,
1129 const struct nlmsghdr *nlh,
1130 const struct nlattr * const attr[])
1132 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1133 struct ip_set *from, *to;
1134 ip_set_id_t from_id, to_id;
1135 char from_name[IPSET_MAXNAMELEN];
1137 if (unlikely(protocol_failed(attr) ||
1138 attr[IPSET_ATTR_SETNAME] == NULL ||
1139 attr[IPSET_ATTR_SETNAME2] == NULL))
1140 return -IPSET_ERR_PROTOCOL;
1142 from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1147 to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]),
1150 return -IPSET_ERR_EXIST_SETNAME2;
1152 /* Features must not change.
1153 * Not an artificial restriction anymore, as we must prevent
1154 * possible loops created by swapping in setlist type of sets. */
1155 if (!(from->type->features == to->type->features &&
1156 from->family == to->family))
1157 return -IPSET_ERR_TYPE_MISMATCH;
1159 strncpy(from_name, from->name, IPSET_MAXNAMELEN);
1160 strncpy(from->name, to->name, IPSET_MAXNAMELEN);
1161 strncpy(to->name, from_name, IPSET_MAXNAMELEN);
1163 write_lock_bh(&ip_set_ref_lock);
1164 swap(from->ref, to->ref);
1165 nfnl_set(inst, from_id) = to;
1166 nfnl_set(inst, to_id) = from;
1167 write_unlock_bh(&ip_set_ref_lock);
1172 /* List/save set data */
1179 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1180 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1183 ip_set_dump_done(struct netlink_callback *cb)
1185 struct ip_set_net *inst = (struct ip_set_net *)cb->data;
1187 pr_debug("release set %s\n", nfnl_set(inst, cb->args[1])->name);
1188 __ip_set_put_byindex(inst, (ip_set_id_t) cb->args[1]);
1194 dump_attrs(struct nlmsghdr *nlh)
1196 const struct nlattr *attr;
1199 pr_debug("dump nlmsg\n");
1200 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) {
1201 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len);
1206 dump_init(struct netlink_callback *cb)
1208 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb);
1209 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1210 struct nlattr *cda[IPSET_ATTR_CMD_MAX+1];
1211 struct nlattr *attr = (void *)nlh + min_len;
1214 struct ip_set_net *inst = (struct ip_set_net *)cb->data;
1216 /* Second pass, so parser can't fail */
1217 nla_parse(cda, IPSET_ATTR_CMD_MAX,
1218 attr, nlh->nlmsg_len - min_len, ip_set_setname_policy);
1220 /* cb->args[0] : dump single set/all sets
1222 * [..]: type specific
1225 if (cda[IPSET_ATTR_SETNAME]) {
1228 set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]),
1233 dump_type = DUMP_ONE;
1234 cb->args[1] = index;
1236 dump_type = DUMP_ALL;
1238 if (cda[IPSET_ATTR_FLAGS]) {
1239 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]);
1240 dump_type |= (f << 16);
1242 cb->args[0] = dump_type;
1248 ip_set_dump_start(struct sk_buff *skb, struct netlink_callback *cb)
1250 ip_set_id_t index = IPSET_INVALID_ID, max;
1251 struct ip_set *set = NULL;
1252 struct nlmsghdr *nlh = NULL;
1253 unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0;
1254 u32 dump_type, dump_flags;
1256 struct ip_set_net *inst = (struct ip_set_net *)cb->data;
1259 ret = dump_init(cb);
1261 nlh = nlmsg_hdr(cb->skb);
1262 /* We have to create and send the error message
1264 if (nlh->nlmsg_flags & NLM_F_ACK)
1265 netlink_ack(cb->skb, nlh, ret);
1270 if (cb->args[1] >= inst->ip_set_max)
1273 dump_type = DUMP_TYPE(cb->args[0]);
1274 dump_flags = DUMP_FLAGS(cb->args[0]);
1275 max = dump_type == DUMP_ONE ? cb->args[1] + 1 : inst->ip_set_max;
1277 pr_debug("args[0]: %u %u args[1]: %ld\n",
1278 dump_type, dump_flags, cb->args[1]);
1279 for (; cb->args[1] < max; cb->args[1]++) {
1280 index = (ip_set_id_t) cb->args[1];
1281 set = nfnl_set(inst, index);
1283 if (dump_type == DUMP_ONE) {
1289 /* When dumping all sets, we must dump "sorted"
1290 * so that lists (unions of sets) are dumped last.
1292 if (dump_type != DUMP_ONE &&
1293 ((dump_type == DUMP_ALL) ==
1294 !!(set->type->features & IPSET_DUMP_LAST)))
1296 pr_debug("List set: %s\n", set->name);
1298 /* Start listing: make sure set won't be destroyed */
1299 pr_debug("reference set\n");
1302 nlh = start_msg(skb, NETLINK_CB(cb->skb).portid,
1303 cb->nlh->nlmsg_seq, flags,
1307 goto release_refcount;
1309 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1310 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name))
1311 goto nla_put_failure;
1312 if (dump_flags & IPSET_FLAG_LIST_SETNAME)
1314 switch (cb->args[2]) {
1316 /* Core header data */
1317 if (nla_put_string(skb, IPSET_ATTR_TYPENAME,
1319 nla_put_u8(skb, IPSET_ATTR_FAMILY,
1321 nla_put_u8(skb, IPSET_ATTR_REVISION,
1323 goto nla_put_failure;
1324 ret = set->variant->head(set, skb);
1326 goto release_refcount;
1327 if (dump_flags & IPSET_FLAG_LIST_HEADER)
1329 /* Fall through and add elements */
1331 read_lock_bh(&set->lock);
1332 ret = set->variant->list(set, skb, cb);
1333 read_unlock_bh(&set->lock);
1335 /* Set is done, proceed with next one */
1337 goto release_refcount;
1340 /* If we dump all sets, continue with dumping last ones */
1341 if (dump_type == DUMP_ALL) {
1342 dump_type = DUMP_LAST;
1343 cb->args[0] = dump_type | (dump_flags << 16);
1352 if (dump_type == DUMP_ONE)
1353 cb->args[1] = IPSET_INVALID_ID;
1357 /* If there was an error or set is done, release set */
1358 if (ret || !cb->args[2]) {
1359 pr_debug("release set %s\n", nfnl_set(inst, index)->name);
1360 __ip_set_put_byindex(inst, index);
1365 nlmsg_end(skb, nlh);
1366 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len);
1370 return ret < 0 ? ret : skb->len;
1374 ip_set_dump(struct sock *ctnl, struct sk_buff *skb,
1375 const struct nlmsghdr *nlh,
1376 const struct nlattr * const attr[])
1378 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1380 if (unlikely(protocol_failed(attr)))
1381 return -IPSET_ERR_PROTOCOL;
1384 struct netlink_dump_control c = {
1385 .dump = ip_set_dump_start,
1386 .done = ip_set_dump_done,
1387 .data = (void *)inst
1389 return netlink_dump_start(ctnl, skb, nlh, &c);
1393 /* Add, del and test */
1395 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = {
1396 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1397 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1398 .len = IPSET_MAXNAMELEN - 1 },
1399 [IPSET_ATTR_LINENO] = { .type = NLA_U32 },
1400 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
1401 [IPSET_ATTR_ADT] = { .type = NLA_NESTED },
1405 call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set,
1406 struct nlattr *tb[], enum ipset_adt adt,
1407 u32 flags, bool use_lineno)
1411 bool eexist = flags & IPSET_FLAG_EXIST, retried = false;
1414 write_lock_bh(&set->lock);
1415 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried);
1416 write_unlock_bh(&set->lock);
1418 } while (ret == -EAGAIN &&
1419 set->variant->resize &&
1420 (ret = set->variant->resize(set, retried)) == 0);
1422 if (!ret || (ret == -IPSET_ERR_EXIST && eexist))
1424 if (lineno && use_lineno) {
1425 /* Error in restore/batch mode: send back lineno */
1426 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb);
1427 struct sk_buff *skb2;
1428 struct nlmsgerr *errmsg;
1429 size_t payload = sizeof(*errmsg) + nlmsg_len(nlh);
1430 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1431 struct nlattr *cda[IPSET_ATTR_CMD_MAX+1];
1432 struct nlattr *cmdattr;
1435 skb2 = nlmsg_new(payload, GFP_KERNEL);
1438 rep = __nlmsg_put(skb2, NETLINK_CB(skb).portid,
1439 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0);
1440 errmsg = nlmsg_data(rep);
1441 errmsg->error = ret;
1442 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len);
1443 cmdattr = (void *)&errmsg->msg + min_len;
1445 nla_parse(cda, IPSET_ATTR_CMD_MAX,
1446 cmdattr, nlh->nlmsg_len - min_len,
1449 errline = nla_data(cda[IPSET_ATTR_LINENO]);
1453 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1454 /* Signal netlink not to send its ACK/errmsg. */
1462 ip_set_uadd(struct sock *ctnl, struct sk_buff *skb,
1463 const struct nlmsghdr *nlh,
1464 const struct nlattr * const attr[])
1466 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1468 struct nlattr *tb[IPSET_ATTR_ADT_MAX+1] = {};
1469 const struct nlattr *nla;
1470 u32 flags = flag_exist(nlh);
1474 if (unlikely(protocol_failed(attr) ||
1475 attr[IPSET_ATTR_SETNAME] == NULL ||
1476 !((attr[IPSET_ATTR_DATA] != NULL) ^
1477 (attr[IPSET_ATTR_ADT] != NULL)) ||
1478 (attr[IPSET_ATTR_DATA] != NULL &&
1479 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1480 (attr[IPSET_ATTR_ADT] != NULL &&
1481 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1482 attr[IPSET_ATTR_LINENO] == NULL))))
1483 return -IPSET_ERR_PROTOCOL;
1485 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1489 use_lineno = !!attr[IPSET_ATTR_LINENO];
1490 if (attr[IPSET_ATTR_DATA]) {
1491 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1492 attr[IPSET_ATTR_DATA],
1493 set->type->adt_policy))
1494 return -IPSET_ERR_PROTOCOL;
1495 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD, flags,
1500 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1501 memset(tb, 0, sizeof(tb));
1502 if (nla_type(nla) != IPSET_ATTR_DATA ||
1503 !flag_nested(nla) ||
1504 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1505 set->type->adt_policy))
1506 return -IPSET_ERR_PROTOCOL;
1507 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD,
1517 ip_set_udel(struct sock *ctnl, struct sk_buff *skb,
1518 const struct nlmsghdr *nlh,
1519 const struct nlattr * const attr[])
1521 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1523 struct nlattr *tb[IPSET_ATTR_ADT_MAX+1] = {};
1524 const struct nlattr *nla;
1525 u32 flags = flag_exist(nlh);
1529 if (unlikely(protocol_failed(attr) ||
1530 attr[IPSET_ATTR_SETNAME] == NULL ||
1531 !((attr[IPSET_ATTR_DATA] != NULL) ^
1532 (attr[IPSET_ATTR_ADT] != NULL)) ||
1533 (attr[IPSET_ATTR_DATA] != NULL &&
1534 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1535 (attr[IPSET_ATTR_ADT] != NULL &&
1536 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1537 attr[IPSET_ATTR_LINENO] == NULL))))
1538 return -IPSET_ERR_PROTOCOL;
1540 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1544 use_lineno = !!attr[IPSET_ATTR_LINENO];
1545 if (attr[IPSET_ATTR_DATA]) {
1546 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1547 attr[IPSET_ATTR_DATA],
1548 set->type->adt_policy))
1549 return -IPSET_ERR_PROTOCOL;
1550 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL, flags,
1555 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1556 memset(tb, 0, sizeof(*tb));
1557 if (nla_type(nla) != IPSET_ATTR_DATA ||
1558 !flag_nested(nla) ||
1559 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1560 set->type->adt_policy))
1561 return -IPSET_ERR_PROTOCOL;
1562 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL,
1572 ip_set_utest(struct sock *ctnl, struct sk_buff *skb,
1573 const struct nlmsghdr *nlh,
1574 const struct nlattr * const attr[])
1576 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1578 struct nlattr *tb[IPSET_ATTR_ADT_MAX+1] = {};
1581 if (unlikely(protocol_failed(attr) ||
1582 attr[IPSET_ATTR_SETNAME] == NULL ||
1583 attr[IPSET_ATTR_DATA] == NULL ||
1584 !flag_nested(attr[IPSET_ATTR_DATA])))
1585 return -IPSET_ERR_PROTOCOL;
1587 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1591 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA],
1592 set->type->adt_policy))
1593 return -IPSET_ERR_PROTOCOL;
1595 read_lock_bh(&set->lock);
1596 ret = set->variant->uadt(set, tb, IPSET_TEST, NULL, 0, 0);
1597 read_unlock_bh(&set->lock);
1598 /* Userspace can't trigger element to be re-added */
1602 return ret > 0 ? 0 : -IPSET_ERR_EXIST;
1605 /* Get headed data of a set */
1608 ip_set_header(struct sock *ctnl, struct sk_buff *skb,
1609 const struct nlmsghdr *nlh,
1610 const struct nlattr * const attr[])
1612 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1613 const struct ip_set *set;
1614 struct sk_buff *skb2;
1615 struct nlmsghdr *nlh2;
1618 if (unlikely(protocol_failed(attr) ||
1619 attr[IPSET_ATTR_SETNAME] == NULL))
1620 return -IPSET_ERR_PROTOCOL;
1622 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1626 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1630 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1634 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1635 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) ||
1636 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) ||
1637 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
1638 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision))
1639 goto nla_put_failure;
1640 nlmsg_end(skb2, nlh2);
1642 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1649 nlmsg_cancel(skb2, nlh2);
1657 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = {
1658 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1659 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
1660 .len = IPSET_MAXNAMELEN - 1 },
1661 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
1665 ip_set_type(struct sock *ctnl, struct sk_buff *skb,
1666 const struct nlmsghdr *nlh,
1667 const struct nlattr * const attr[])
1669 struct sk_buff *skb2;
1670 struct nlmsghdr *nlh2;
1671 u8 family, min, max;
1672 const char *typename;
1675 if (unlikely(protocol_failed(attr) ||
1676 attr[IPSET_ATTR_TYPENAME] == NULL ||
1677 attr[IPSET_ATTR_FAMILY] == NULL))
1678 return -IPSET_ERR_PROTOCOL;
1680 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1681 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1682 ret = find_set_type_minmax(typename, family, &min, &max);
1686 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1690 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1694 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1695 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) ||
1696 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) ||
1697 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) ||
1698 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min))
1699 goto nla_put_failure;
1700 nlmsg_end(skb2, nlh2);
1702 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len);
1703 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1710 nlmsg_cancel(skb2, nlh2);
1716 /* Get protocol version */
1718 static const struct nla_policy
1719 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = {
1720 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1724 ip_set_protocol(struct sock *ctnl, struct sk_buff *skb,
1725 const struct nlmsghdr *nlh,
1726 const struct nlattr * const attr[])
1728 struct sk_buff *skb2;
1729 struct nlmsghdr *nlh2;
1732 if (unlikely(attr[IPSET_ATTR_PROTOCOL] == NULL))
1733 return -IPSET_ERR_PROTOCOL;
1735 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1739 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1740 IPSET_CMD_PROTOCOL);
1743 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL))
1744 goto nla_put_failure;
1745 nlmsg_end(skb2, nlh2);
1747 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1754 nlmsg_cancel(skb2, nlh2);
1760 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = {
1761 [IPSET_CMD_NONE] = {
1762 .call = ip_set_none,
1763 .attr_count = IPSET_ATTR_CMD_MAX,
1765 [IPSET_CMD_CREATE] = {
1766 .call = ip_set_create,
1767 .attr_count = IPSET_ATTR_CMD_MAX,
1768 .policy = ip_set_create_policy,
1770 [IPSET_CMD_DESTROY] = {
1771 .call = ip_set_destroy,
1772 .attr_count = IPSET_ATTR_CMD_MAX,
1773 .policy = ip_set_setname_policy,
1775 [IPSET_CMD_FLUSH] = {
1776 .call = ip_set_flush,
1777 .attr_count = IPSET_ATTR_CMD_MAX,
1778 .policy = ip_set_setname_policy,
1780 [IPSET_CMD_RENAME] = {
1781 .call = ip_set_rename,
1782 .attr_count = IPSET_ATTR_CMD_MAX,
1783 .policy = ip_set_setname2_policy,
1785 [IPSET_CMD_SWAP] = {
1786 .call = ip_set_swap,
1787 .attr_count = IPSET_ATTR_CMD_MAX,
1788 .policy = ip_set_setname2_policy,
1790 [IPSET_CMD_LIST] = {
1791 .call = ip_set_dump,
1792 .attr_count = IPSET_ATTR_CMD_MAX,
1793 .policy = ip_set_setname_policy,
1795 [IPSET_CMD_SAVE] = {
1796 .call = ip_set_dump,
1797 .attr_count = IPSET_ATTR_CMD_MAX,
1798 .policy = ip_set_setname_policy,
1801 .call = ip_set_uadd,
1802 .attr_count = IPSET_ATTR_CMD_MAX,
1803 .policy = ip_set_adt_policy,
1806 .call = ip_set_udel,
1807 .attr_count = IPSET_ATTR_CMD_MAX,
1808 .policy = ip_set_adt_policy,
1810 [IPSET_CMD_TEST] = {
1811 .call = ip_set_utest,
1812 .attr_count = IPSET_ATTR_CMD_MAX,
1813 .policy = ip_set_adt_policy,
1815 [IPSET_CMD_HEADER] = {
1816 .call = ip_set_header,
1817 .attr_count = IPSET_ATTR_CMD_MAX,
1818 .policy = ip_set_setname_policy,
1820 [IPSET_CMD_TYPE] = {
1821 .call = ip_set_type,
1822 .attr_count = IPSET_ATTR_CMD_MAX,
1823 .policy = ip_set_type_policy,
1825 [IPSET_CMD_PROTOCOL] = {
1826 .call = ip_set_protocol,
1827 .attr_count = IPSET_ATTR_CMD_MAX,
1828 .policy = ip_set_protocol_policy,
1832 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = {
1834 .subsys_id = NFNL_SUBSYS_IPSET,
1835 .cb_count = IPSET_MSG_MAX,
1836 .cb = ip_set_netlink_subsys_cb,
1839 /* Interface to iptables/ip6tables */
1842 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len)
1846 int copylen = *len, ret = 0;
1847 struct net *net = sock_net(sk);
1848 struct ip_set_net *inst = ip_set_pernet(net);
1850 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1852 if (optval != SO_IP_SET)
1854 if (*len < sizeof(unsigned int))
1857 data = vmalloc(*len);
1860 if (copy_from_user(data, user, *len) != 0) {
1864 op = (unsigned int *) data;
1866 if (*op < IP_SET_OP_VERSION) {
1867 /* Check the version at the beginning of operations */
1868 struct ip_set_req_version *req_version = data;
1869 if (req_version->version != IPSET_PROTOCOL) {
1876 case IP_SET_OP_VERSION: {
1877 struct ip_set_req_version *req_version = data;
1879 if (*len != sizeof(struct ip_set_req_version)) {
1884 req_version->version = IPSET_PROTOCOL;
1885 ret = copy_to_user(user, req_version,
1886 sizeof(struct ip_set_req_version));
1889 case IP_SET_OP_GET_BYNAME: {
1890 struct ip_set_req_get_set *req_get = data;
1893 if (*len != sizeof(struct ip_set_req_get_set)) {
1897 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
1898 nfnl_lock(NFNL_SUBSYS_IPSET);
1899 find_set_and_id(inst, req_get->set.name, &id);
1900 req_get->set.index = id;
1901 nfnl_unlock(NFNL_SUBSYS_IPSET);
1904 case IP_SET_OP_GET_FNAME: {
1905 struct ip_set_req_get_set_family *req_get = data;
1908 if (*len != sizeof(struct ip_set_req_get_set_family)) {
1912 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
1913 nfnl_lock(NFNL_SUBSYS_IPSET);
1914 find_set_and_id(inst, req_get->set.name, &id);
1915 req_get->set.index = id;
1916 if (id != IPSET_INVALID_ID)
1917 req_get->family = nfnl_set(inst, id)->family;
1918 nfnl_unlock(NFNL_SUBSYS_IPSET);
1921 case IP_SET_OP_GET_BYINDEX: {
1922 struct ip_set_req_get_set *req_get = data;
1925 if (*len != sizeof(struct ip_set_req_get_set) ||
1926 req_get->set.index >= inst->ip_set_max) {
1930 nfnl_lock(NFNL_SUBSYS_IPSET);
1931 set = nfnl_set(inst, req_get->set.index);
1932 strncpy(req_get->set.name, set ? set->name : "",
1934 nfnl_unlock(NFNL_SUBSYS_IPSET);
1940 } /* end of switch(op) */
1943 ret = copy_to_user(user, data, copylen);
1952 static struct nf_sockopt_ops so_set __read_mostly = {
1954 .get_optmin = SO_IP_SET,
1955 .get_optmax = SO_IP_SET + 1,
1956 .get = &ip_set_sockfn_get,
1957 .owner = THIS_MODULE,
1960 static int __net_init
1961 ip_set_net_init(struct net *net)
1963 struct ip_set_net *inst = ip_set_pernet(net);
1965 struct ip_set **list;
1967 inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX;
1968 if (inst->ip_set_max >= IPSET_INVALID_ID)
1969 inst->ip_set_max = IPSET_INVALID_ID - 1;
1971 list = kzalloc(sizeof(struct ip_set *) * inst->ip_set_max, GFP_KERNEL);
1974 inst->is_deleted = 0;
1975 rcu_assign_pointer(inst->ip_set_list, list);
1976 pr_notice("ip_set: protocol %u\n", IPSET_PROTOCOL);
1980 static void __net_exit
1981 ip_set_net_exit(struct net *net)
1983 struct ip_set_net *inst = ip_set_pernet(net);
1985 struct ip_set *set = NULL;
1988 inst->is_deleted = 1; /* flag for ip_set_nfnl_put */
1990 for (i = 0; i < inst->ip_set_max; i++) {
1991 set = nfnl_set(inst, i);
1993 ip_set_destroy_set(inst, i);
1995 kfree(rcu_dereference_protected(inst->ip_set_list, 1));
1998 static struct pernet_operations ip_set_net_ops = {
1999 .init = ip_set_net_init,
2000 .exit = ip_set_net_exit,
2001 .id = &ip_set_net_id,
2002 .size = sizeof(struct ip_set_net)
2009 int ret = nfnetlink_subsys_register(&ip_set_netlink_subsys);
2011 pr_err("ip_set: cannot register with nfnetlink.\n");
2014 ret = nf_register_sockopt(&so_set);
2016 pr_err("SO_SET registry failed: %d\n", ret);
2017 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2020 ret = register_pernet_subsys(&ip_set_net_ops);
2022 pr_err("ip_set: cannot register pernet_subsys.\n");
2023 nf_unregister_sockopt(&so_set);
2024 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2033 unregister_pernet_subsys(&ip_set_net_ops);
2034 nf_unregister_sockopt(&so_set);
2035 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2036 pr_debug("these are the famous last words\n");
2039 module_init(ip_set_init);
2040 module_exit(ip_set_fini);