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 index, 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_byindex(struct net *net, ip_set_id_t index)
637 struct ip_set_net *inst = ip_set_pernet(net);
639 if (index > inst->ip_set_max)
640 return IPSET_INVALID_ID;
642 nfnl_lock(NFNL_SUBSYS_IPSET);
643 set = nfnl_set(inst, index);
647 index = IPSET_INVALID_ID;
648 nfnl_unlock(NFNL_SUBSYS_IPSET);
652 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex);
655 * If the given set pointer points to a valid set, decrement
656 * reference count by 1. The caller shall not assume the index
657 * to be valid, after calling this function.
659 * The nfnl mutex is used in the function.
662 ip_set_nfnl_put(struct net *net, ip_set_id_t index)
665 struct ip_set_net *inst = ip_set_pernet(net);
667 nfnl_lock(NFNL_SUBSYS_IPSET);
668 if (!inst->is_deleted) { /* already deleted from ip_set_net_exit() */
669 set = nfnl_set(inst, index);
673 nfnl_unlock(NFNL_SUBSYS_IPSET);
675 EXPORT_SYMBOL_GPL(ip_set_nfnl_put);
678 * Communication protocol with userspace over netlink.
680 * The commands are serialized by the nfnl mutex.
684 protocol_failed(const struct nlattr * const tb[])
686 return !tb[IPSET_ATTR_PROTOCOL] ||
687 nla_get_u8(tb[IPSET_ATTR_PROTOCOL]) != IPSET_PROTOCOL;
691 flag_exist(const struct nlmsghdr *nlh)
693 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST;
696 static struct nlmsghdr *
697 start_msg(struct sk_buff *skb, u32 portid, u32 seq, unsigned int flags,
700 struct nlmsghdr *nlh;
701 struct nfgenmsg *nfmsg;
703 nlh = nlmsg_put(skb, portid, seq, cmd | (NFNL_SUBSYS_IPSET << 8),
704 sizeof(*nfmsg), flags);
708 nfmsg = nlmsg_data(nlh);
709 nfmsg->nfgen_family = NFPROTO_IPV4;
710 nfmsg->version = NFNETLINK_V0;
718 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = {
719 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
720 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
721 .len = IPSET_MAXNAMELEN - 1 },
722 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
723 .len = IPSET_MAXNAMELEN - 1},
724 [IPSET_ATTR_REVISION] = { .type = NLA_U8 },
725 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
726 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
729 static struct ip_set *
730 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id)
732 struct ip_set *set = NULL;
735 *id = IPSET_INVALID_ID;
736 for (i = 0; i < inst->ip_set_max; i++) {
737 set = nfnl_set(inst, i);
738 if (set != NULL && STREQ(set->name, name)) {
743 return (*id == IPSET_INVALID_ID ? NULL : set);
746 static inline struct ip_set *
747 find_set(struct ip_set_net *inst, const char *name)
751 return find_set_and_id(inst, name, &id);
755 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index,
761 *index = IPSET_INVALID_ID;
762 for (i = 0; i < inst->ip_set_max; i++) {
763 s = nfnl_set(inst, i);
765 if (*index == IPSET_INVALID_ID)
767 } else if (STREQ(name, s->name)) {
773 if (*index == IPSET_INVALID_ID)
774 /* No free slot remained */
775 return -IPSET_ERR_MAX_SETS;
780 ip_set_none(struct sock *ctnl, struct sk_buff *skb,
781 const struct nlmsghdr *nlh,
782 const struct nlattr * const attr[])
788 ip_set_create(struct sock *ctnl, struct sk_buff *skb,
789 const struct nlmsghdr *nlh,
790 const struct nlattr * const attr[])
792 struct net *net = sock_net(ctnl);
793 struct ip_set_net *inst = ip_set_pernet(net);
794 struct ip_set *set, *clash = NULL;
795 ip_set_id_t index = IPSET_INVALID_ID;
796 struct nlattr *tb[IPSET_ATTR_CREATE_MAX+1] = {};
797 const char *name, *typename;
799 u32 flags = flag_exist(nlh);
802 if (unlikely(protocol_failed(attr) ||
803 attr[IPSET_ATTR_SETNAME] == NULL ||
804 attr[IPSET_ATTR_TYPENAME] == NULL ||
805 attr[IPSET_ATTR_REVISION] == NULL ||
806 attr[IPSET_ATTR_FAMILY] == NULL ||
807 (attr[IPSET_ATTR_DATA] != NULL &&
808 !flag_nested(attr[IPSET_ATTR_DATA]))))
809 return -IPSET_ERR_PROTOCOL;
811 name = nla_data(attr[IPSET_ATTR_SETNAME]);
812 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
813 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
814 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]);
815 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
816 name, typename, family_name(family), revision);
819 * First, and without any locks, allocate and initialize
820 * a normal base set structure.
822 set = kzalloc(sizeof(struct ip_set), GFP_KERNEL);
825 rwlock_init(&set->lock);
826 strlcpy(set->name, name, IPSET_MAXNAMELEN);
827 set->family = family;
828 set->revision = revision;
831 * Next, check that we know the type, and take
832 * a reference on the type, to make sure it stays available
833 * while constructing our new set.
835 * After referencing the type, we try to create the type
836 * specific part of the set without holding any locks.
838 ret = find_set_type_get(typename, family, revision, &(set->type));
843 * Without holding any locks, create private part.
845 if (attr[IPSET_ATTR_DATA] &&
846 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA],
847 set->type->create_policy)) {
848 ret = -IPSET_ERR_PROTOCOL;
852 ret = set->type->create(net, set, tb, flags);
856 /* BTW, ret==0 here. */
859 * Here, we have a valid, constructed set and we are protected
860 * by the nfnl mutex. Find the first free index in ip_set_list
861 * and check clashing.
863 ret = find_free_id(inst, set->name, &index, &clash);
864 if (ret == -EEXIST) {
865 /* If this is the same set and requested, ignore error */
866 if ((flags & IPSET_FLAG_EXIST) &&
867 STREQ(set->type->name, clash->type->name) &&
868 set->type->family == clash->type->family &&
869 set->type->revision_min == clash->type->revision_min &&
870 set->type->revision_max == clash->type->revision_max &&
871 set->variant->same_set(set, clash))
874 } else if (ret == -IPSET_ERR_MAX_SETS) {
875 struct ip_set **list, **tmp;
876 ip_set_id_t i = inst->ip_set_max + IP_SET_INC;
878 if (i < inst->ip_set_max || i == IPSET_INVALID_ID)
882 list = kzalloc(sizeof(struct ip_set *) * i, GFP_KERNEL);
885 /* nfnl mutex is held, both lists are valid */
886 tmp = nfnl_dereference(inst->ip_set_list);
887 memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max);
888 rcu_assign_pointer(inst->ip_set_list, list);
889 /* Make sure all current packets have passed through */
892 index = inst->ip_set_max;
893 inst->ip_set_max = i;
900 * Finally! Add our shiny new set to the list, and be done.
902 pr_debug("create: '%s' created with index %u!\n", set->name, index);
903 nfnl_set(inst, index) = set;
908 set->variant->destroy(set);
910 module_put(set->type->me);
918 static const struct nla_policy
919 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = {
920 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
921 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
922 .len = IPSET_MAXNAMELEN - 1 },
926 ip_set_destroy_set(struct ip_set_net *inst, ip_set_id_t index)
928 struct ip_set *set = nfnl_set(inst, index);
930 pr_debug("set: %s\n", set->name);
931 nfnl_set(inst, index) = NULL;
933 /* Must call it without holding any lock */
934 set->variant->destroy(set);
935 module_put(set->type->me);
940 ip_set_destroy(struct sock *ctnl, struct sk_buff *skb,
941 const struct nlmsghdr *nlh,
942 const struct nlattr * const attr[])
944 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
949 if (unlikely(protocol_failed(attr)))
950 return -IPSET_ERR_PROTOCOL;
952 /* Commands are serialized and references are
953 * protected by the ip_set_ref_lock.
954 * External systems (i.e. xt_set) must call
955 * ip_set_put|get_nfnl_* functions, that way we
956 * can safely check references here.
958 * list:set timer can only decrement the reference
959 * counter, so if it's already zero, we can proceed
960 * without holding the lock.
962 read_lock_bh(&ip_set_ref_lock);
963 if (!attr[IPSET_ATTR_SETNAME]) {
964 for (i = 0; i < inst->ip_set_max; i++) {
965 s = nfnl_set(inst, i);
966 if (s != NULL && s->ref) {
967 ret = -IPSET_ERR_BUSY;
971 read_unlock_bh(&ip_set_ref_lock);
972 for (i = 0; i < inst->ip_set_max; i++) {
973 s = nfnl_set(inst, i);
975 ip_set_destroy_set(inst, i);
978 s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
984 ret = -IPSET_ERR_BUSY;
987 read_unlock_bh(&ip_set_ref_lock);
989 ip_set_destroy_set(inst, i);
993 read_unlock_bh(&ip_set_ref_lock);
1000 ip_set_flush_set(struct ip_set *set)
1002 pr_debug("set: %s\n", set->name);
1004 write_lock_bh(&set->lock);
1005 set->variant->flush(set);
1006 write_unlock_bh(&set->lock);
1010 ip_set_flush(struct sock *ctnl, struct sk_buff *skb,
1011 const struct nlmsghdr *nlh,
1012 const struct nlattr * const attr[])
1014 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1018 if (unlikely(protocol_failed(attr)))
1019 return -IPSET_ERR_PROTOCOL;
1021 if (!attr[IPSET_ATTR_SETNAME]) {
1022 for (i = 0; i < inst->ip_set_max; i++) {
1023 s = nfnl_set(inst, i);
1025 ip_set_flush_set(s);
1028 s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1032 ip_set_flush_set(s);
1040 static const struct nla_policy
1041 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = {
1042 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1043 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1044 .len = IPSET_MAXNAMELEN - 1 },
1045 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING,
1046 .len = IPSET_MAXNAMELEN - 1 },
1050 ip_set_rename(struct sock *ctnl, struct sk_buff *skb,
1051 const struct nlmsghdr *nlh,
1052 const struct nlattr * const attr[])
1054 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1055 struct ip_set *set, *s;
1060 if (unlikely(protocol_failed(attr) ||
1061 attr[IPSET_ATTR_SETNAME] == NULL ||
1062 attr[IPSET_ATTR_SETNAME2] == NULL))
1063 return -IPSET_ERR_PROTOCOL;
1065 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1069 read_lock_bh(&ip_set_ref_lock);
1070 if (set->ref != 0) {
1071 ret = -IPSET_ERR_REFERENCED;
1075 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]);
1076 for (i = 0; i < inst->ip_set_max; i++) {
1077 s = nfnl_set(inst, i);
1078 if (s != NULL && STREQ(s->name, name2)) {
1079 ret = -IPSET_ERR_EXIST_SETNAME2;
1083 strncpy(set->name, name2, IPSET_MAXNAMELEN);
1086 read_unlock_bh(&ip_set_ref_lock);
1090 /* Swap two sets so that name/index points to the other.
1091 * References and set names are also swapped.
1093 * The commands are serialized by the nfnl mutex and references are
1094 * protected by the ip_set_ref_lock. The kernel interfaces
1095 * do not hold the mutex but the pointer settings are atomic
1096 * so the ip_set_list always contains valid pointers to the sets.
1100 ip_set_swap(struct sock *ctnl, struct sk_buff *skb,
1101 const struct nlmsghdr *nlh,
1102 const struct nlattr * const attr[])
1104 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1105 struct ip_set *from, *to;
1106 ip_set_id_t from_id, to_id;
1107 char from_name[IPSET_MAXNAMELEN];
1109 if (unlikely(protocol_failed(attr) ||
1110 attr[IPSET_ATTR_SETNAME] == NULL ||
1111 attr[IPSET_ATTR_SETNAME2] == NULL))
1112 return -IPSET_ERR_PROTOCOL;
1114 from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1119 to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]),
1122 return -IPSET_ERR_EXIST_SETNAME2;
1124 /* Features must not change.
1125 * Not an artificial restriction anymore, as we must prevent
1126 * possible loops created by swapping in setlist type of sets. */
1127 if (!(from->type->features == to->type->features &&
1128 from->family == to->family))
1129 return -IPSET_ERR_TYPE_MISMATCH;
1131 strncpy(from_name, from->name, IPSET_MAXNAMELEN);
1132 strncpy(from->name, to->name, IPSET_MAXNAMELEN);
1133 strncpy(to->name, from_name, IPSET_MAXNAMELEN);
1135 write_lock_bh(&ip_set_ref_lock);
1136 swap(from->ref, to->ref);
1137 nfnl_set(inst, from_id) = to;
1138 nfnl_set(inst, to_id) = from;
1139 write_unlock_bh(&ip_set_ref_lock);
1144 /* List/save set data */
1151 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1152 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1155 ip_set_dump_done(struct netlink_callback *cb)
1157 struct ip_set_net *inst = (struct ip_set_net *)cb->args[IPSET_CB_NET];
1158 if (cb->args[IPSET_CB_ARG0]) {
1159 pr_debug("release set %s\n",
1160 nfnl_set(inst, cb->args[IPSET_CB_INDEX])->name);
1161 __ip_set_put_byindex(inst,
1162 (ip_set_id_t) cb->args[IPSET_CB_INDEX]);
1168 dump_attrs(struct nlmsghdr *nlh)
1170 const struct nlattr *attr;
1173 pr_debug("dump nlmsg\n");
1174 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) {
1175 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len);
1180 dump_init(struct netlink_callback *cb, struct ip_set_net *inst)
1182 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb);
1183 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1184 struct nlattr *cda[IPSET_ATTR_CMD_MAX+1];
1185 struct nlattr *attr = (void *)nlh + min_len;
1189 /* Second pass, so parser can't fail */
1190 nla_parse(cda, IPSET_ATTR_CMD_MAX,
1191 attr, nlh->nlmsg_len - min_len, ip_set_setname_policy);
1193 /* cb->args[IPSET_CB_NET]: net namespace
1194 * [IPSET_CB_DUMP]: dump single set/all sets
1195 * [IPSET_CB_INDEX]: set index
1196 * [IPSET_CB_ARG0]: type specific
1199 if (cda[IPSET_ATTR_SETNAME]) {
1202 set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]),
1207 dump_type = DUMP_ONE;
1208 cb->args[IPSET_CB_INDEX] = index;
1210 dump_type = DUMP_ALL;
1212 if (cda[IPSET_ATTR_FLAGS]) {
1213 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]);
1214 dump_type |= (f << 16);
1216 cb->args[IPSET_CB_NET] = (unsigned long)inst;
1217 cb->args[IPSET_CB_DUMP] = dump_type;
1223 ip_set_dump_start(struct sk_buff *skb, struct netlink_callback *cb)
1225 ip_set_id_t index = IPSET_INVALID_ID, max;
1226 struct ip_set *set = NULL;
1227 struct nlmsghdr *nlh = NULL;
1228 unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0;
1229 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1230 u32 dump_type, dump_flags;
1233 if (!cb->args[IPSET_CB_DUMP]) {
1234 ret = dump_init(cb, inst);
1236 nlh = nlmsg_hdr(cb->skb);
1237 /* We have to create and send the error message
1239 if (nlh->nlmsg_flags & NLM_F_ACK)
1240 netlink_ack(cb->skb, nlh, ret);
1245 if (cb->args[IPSET_CB_INDEX] >= inst->ip_set_max)
1248 dump_type = DUMP_TYPE(cb->args[IPSET_CB_DUMP]);
1249 dump_flags = DUMP_FLAGS(cb->args[IPSET_CB_DUMP]);
1250 max = dump_type == DUMP_ONE ? cb->args[IPSET_CB_INDEX] + 1
1253 pr_debug("dump type, flag: %u %u index: %ld\n",
1254 dump_type, dump_flags, cb->args[IPSET_CB_INDEX]);
1255 for (; cb->args[IPSET_CB_INDEX] < max; cb->args[IPSET_CB_INDEX]++) {
1256 index = (ip_set_id_t) cb->args[IPSET_CB_INDEX];
1257 set = nfnl_set(inst, index);
1259 if (dump_type == DUMP_ONE) {
1265 /* When dumping all sets, we must dump "sorted"
1266 * so that lists (unions of sets) are dumped last.
1268 if (dump_type != DUMP_ONE &&
1269 ((dump_type == DUMP_ALL) ==
1270 !!(set->type->features & IPSET_DUMP_LAST)))
1272 pr_debug("List set: %s\n", set->name);
1273 if (!cb->args[IPSET_CB_ARG0]) {
1274 /* Start listing: make sure set won't be destroyed */
1275 pr_debug("reference set\n");
1278 nlh = start_msg(skb, NETLINK_CB(cb->skb).portid,
1279 cb->nlh->nlmsg_seq, flags,
1283 goto release_refcount;
1285 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1286 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name))
1287 goto nla_put_failure;
1288 if (dump_flags & IPSET_FLAG_LIST_SETNAME)
1290 switch (cb->args[IPSET_CB_ARG0]) {
1292 /* Core header data */
1293 if (nla_put_string(skb, IPSET_ATTR_TYPENAME,
1295 nla_put_u8(skb, IPSET_ATTR_FAMILY,
1297 nla_put_u8(skb, IPSET_ATTR_REVISION,
1299 goto nla_put_failure;
1300 ret = set->variant->head(set, skb);
1302 goto release_refcount;
1303 if (dump_flags & IPSET_FLAG_LIST_HEADER)
1305 /* Fall through and add elements */
1307 read_lock_bh(&set->lock);
1308 ret = set->variant->list(set, skb, cb);
1309 read_unlock_bh(&set->lock);
1310 if (!cb->args[IPSET_CB_ARG0])
1311 /* Set is done, proceed with next one */
1313 goto release_refcount;
1316 /* If we dump all sets, continue with dumping last ones */
1317 if (dump_type == DUMP_ALL) {
1318 dump_type = DUMP_LAST;
1319 cb->args[IPSET_CB_DUMP] = dump_type | (dump_flags << 16);
1320 cb->args[IPSET_CB_INDEX] = 0;
1328 if (dump_type == DUMP_ONE)
1329 cb->args[IPSET_CB_INDEX] = IPSET_INVALID_ID;
1331 cb->args[IPSET_CB_INDEX]++;
1333 /* If there was an error or set is done, release set */
1334 if (ret || !cb->args[IPSET_CB_ARG0]) {
1335 pr_debug("release set %s\n", nfnl_set(inst, index)->name);
1336 __ip_set_put_byindex(inst, index);
1337 cb->args[IPSET_CB_ARG0] = 0;
1341 nlmsg_end(skb, nlh);
1342 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len);
1346 return ret < 0 ? ret : skb->len;
1350 ip_set_dump(struct sock *ctnl, struct sk_buff *skb,
1351 const struct nlmsghdr *nlh,
1352 const struct nlattr * const attr[])
1354 if (unlikely(protocol_failed(attr)))
1355 return -IPSET_ERR_PROTOCOL;
1358 struct netlink_dump_control c = {
1359 .dump = ip_set_dump_start,
1360 .done = ip_set_dump_done,
1362 return netlink_dump_start(ctnl, skb, nlh, &c);
1366 /* Add, del and test */
1368 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = {
1369 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1370 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1371 .len = IPSET_MAXNAMELEN - 1 },
1372 [IPSET_ATTR_LINENO] = { .type = NLA_U32 },
1373 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
1374 [IPSET_ATTR_ADT] = { .type = NLA_NESTED },
1378 call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set,
1379 struct nlattr *tb[], enum ipset_adt adt,
1380 u32 flags, bool use_lineno)
1384 bool eexist = flags & IPSET_FLAG_EXIST, retried = false;
1387 write_lock_bh(&set->lock);
1388 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried);
1389 write_unlock_bh(&set->lock);
1391 } while (ret == -EAGAIN &&
1392 set->variant->resize &&
1393 (ret = set->variant->resize(set, retried)) == 0);
1395 if (!ret || (ret == -IPSET_ERR_EXIST && eexist))
1397 if (lineno && use_lineno) {
1398 /* Error in restore/batch mode: send back lineno */
1399 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb);
1400 struct sk_buff *skb2;
1401 struct nlmsgerr *errmsg;
1402 size_t payload = sizeof(*errmsg) + nlmsg_len(nlh);
1403 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1404 struct nlattr *cda[IPSET_ATTR_CMD_MAX+1];
1405 struct nlattr *cmdattr;
1408 skb2 = nlmsg_new(payload, GFP_KERNEL);
1411 rep = __nlmsg_put(skb2, NETLINK_CB(skb).portid,
1412 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0);
1413 errmsg = nlmsg_data(rep);
1414 errmsg->error = ret;
1415 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len);
1416 cmdattr = (void *)&errmsg->msg + min_len;
1418 nla_parse(cda, IPSET_ATTR_CMD_MAX,
1419 cmdattr, nlh->nlmsg_len - min_len,
1422 errline = nla_data(cda[IPSET_ATTR_LINENO]);
1426 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1427 /* Signal netlink not to send its ACK/errmsg. */
1435 ip_set_uadd(struct sock *ctnl, struct sk_buff *skb,
1436 const struct nlmsghdr *nlh,
1437 const struct nlattr * const attr[])
1439 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1441 struct nlattr *tb[IPSET_ATTR_ADT_MAX+1] = {};
1442 const struct nlattr *nla;
1443 u32 flags = flag_exist(nlh);
1447 if (unlikely(protocol_failed(attr) ||
1448 attr[IPSET_ATTR_SETNAME] == NULL ||
1449 !((attr[IPSET_ATTR_DATA] != NULL) ^
1450 (attr[IPSET_ATTR_ADT] != NULL)) ||
1451 (attr[IPSET_ATTR_DATA] != NULL &&
1452 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1453 (attr[IPSET_ATTR_ADT] != NULL &&
1454 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1455 attr[IPSET_ATTR_LINENO] == NULL))))
1456 return -IPSET_ERR_PROTOCOL;
1458 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1462 use_lineno = !!attr[IPSET_ATTR_LINENO];
1463 if (attr[IPSET_ATTR_DATA]) {
1464 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1465 attr[IPSET_ATTR_DATA],
1466 set->type->adt_policy))
1467 return -IPSET_ERR_PROTOCOL;
1468 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD, flags,
1473 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1474 memset(tb, 0, sizeof(tb));
1475 if (nla_type(nla) != IPSET_ATTR_DATA ||
1476 !flag_nested(nla) ||
1477 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1478 set->type->adt_policy))
1479 return -IPSET_ERR_PROTOCOL;
1480 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD,
1490 ip_set_udel(struct sock *ctnl, struct sk_buff *skb,
1491 const struct nlmsghdr *nlh,
1492 const struct nlattr * const attr[])
1494 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1496 struct nlattr *tb[IPSET_ATTR_ADT_MAX+1] = {};
1497 const struct nlattr *nla;
1498 u32 flags = flag_exist(nlh);
1502 if (unlikely(protocol_failed(attr) ||
1503 attr[IPSET_ATTR_SETNAME] == NULL ||
1504 !((attr[IPSET_ATTR_DATA] != NULL) ^
1505 (attr[IPSET_ATTR_ADT] != NULL)) ||
1506 (attr[IPSET_ATTR_DATA] != NULL &&
1507 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1508 (attr[IPSET_ATTR_ADT] != NULL &&
1509 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1510 attr[IPSET_ATTR_LINENO] == NULL))))
1511 return -IPSET_ERR_PROTOCOL;
1513 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1517 use_lineno = !!attr[IPSET_ATTR_LINENO];
1518 if (attr[IPSET_ATTR_DATA]) {
1519 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1520 attr[IPSET_ATTR_DATA],
1521 set->type->adt_policy))
1522 return -IPSET_ERR_PROTOCOL;
1523 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL, flags,
1528 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1529 memset(tb, 0, sizeof(*tb));
1530 if (nla_type(nla) != IPSET_ATTR_DATA ||
1531 !flag_nested(nla) ||
1532 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1533 set->type->adt_policy))
1534 return -IPSET_ERR_PROTOCOL;
1535 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL,
1545 ip_set_utest(struct sock *ctnl, struct sk_buff *skb,
1546 const struct nlmsghdr *nlh,
1547 const struct nlattr * const attr[])
1549 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1551 struct nlattr *tb[IPSET_ATTR_ADT_MAX+1] = {};
1554 if (unlikely(protocol_failed(attr) ||
1555 attr[IPSET_ATTR_SETNAME] == NULL ||
1556 attr[IPSET_ATTR_DATA] == NULL ||
1557 !flag_nested(attr[IPSET_ATTR_DATA])))
1558 return -IPSET_ERR_PROTOCOL;
1560 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1564 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA],
1565 set->type->adt_policy))
1566 return -IPSET_ERR_PROTOCOL;
1568 read_lock_bh(&set->lock);
1569 ret = set->variant->uadt(set, tb, IPSET_TEST, NULL, 0, 0);
1570 read_unlock_bh(&set->lock);
1571 /* Userspace can't trigger element to be re-added */
1575 return ret > 0 ? 0 : -IPSET_ERR_EXIST;
1578 /* Get headed data of a set */
1581 ip_set_header(struct sock *ctnl, struct sk_buff *skb,
1582 const struct nlmsghdr *nlh,
1583 const struct nlattr * const attr[])
1585 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1586 const struct ip_set *set;
1587 struct sk_buff *skb2;
1588 struct nlmsghdr *nlh2;
1591 if (unlikely(protocol_failed(attr) ||
1592 attr[IPSET_ATTR_SETNAME] == NULL))
1593 return -IPSET_ERR_PROTOCOL;
1595 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1599 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1603 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1607 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1608 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) ||
1609 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) ||
1610 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
1611 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision))
1612 goto nla_put_failure;
1613 nlmsg_end(skb2, nlh2);
1615 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1622 nlmsg_cancel(skb2, nlh2);
1630 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = {
1631 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1632 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
1633 .len = IPSET_MAXNAMELEN - 1 },
1634 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
1638 ip_set_type(struct sock *ctnl, struct sk_buff *skb,
1639 const struct nlmsghdr *nlh,
1640 const struct nlattr * const attr[])
1642 struct sk_buff *skb2;
1643 struct nlmsghdr *nlh2;
1644 u8 family, min, max;
1645 const char *typename;
1648 if (unlikely(protocol_failed(attr) ||
1649 attr[IPSET_ATTR_TYPENAME] == NULL ||
1650 attr[IPSET_ATTR_FAMILY] == NULL))
1651 return -IPSET_ERR_PROTOCOL;
1653 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1654 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1655 ret = find_set_type_minmax(typename, family, &min, &max);
1659 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1663 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1667 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1668 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) ||
1669 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) ||
1670 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) ||
1671 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min))
1672 goto nla_put_failure;
1673 nlmsg_end(skb2, nlh2);
1675 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len);
1676 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1683 nlmsg_cancel(skb2, nlh2);
1689 /* Get protocol version */
1691 static const struct nla_policy
1692 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = {
1693 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1697 ip_set_protocol(struct sock *ctnl, struct sk_buff *skb,
1698 const struct nlmsghdr *nlh,
1699 const struct nlattr * const attr[])
1701 struct sk_buff *skb2;
1702 struct nlmsghdr *nlh2;
1705 if (unlikely(attr[IPSET_ATTR_PROTOCOL] == NULL))
1706 return -IPSET_ERR_PROTOCOL;
1708 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1712 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1713 IPSET_CMD_PROTOCOL);
1716 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL))
1717 goto nla_put_failure;
1718 nlmsg_end(skb2, nlh2);
1720 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1727 nlmsg_cancel(skb2, nlh2);
1733 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = {
1734 [IPSET_CMD_NONE] = {
1735 .call = ip_set_none,
1736 .attr_count = IPSET_ATTR_CMD_MAX,
1738 [IPSET_CMD_CREATE] = {
1739 .call = ip_set_create,
1740 .attr_count = IPSET_ATTR_CMD_MAX,
1741 .policy = ip_set_create_policy,
1743 [IPSET_CMD_DESTROY] = {
1744 .call = ip_set_destroy,
1745 .attr_count = IPSET_ATTR_CMD_MAX,
1746 .policy = ip_set_setname_policy,
1748 [IPSET_CMD_FLUSH] = {
1749 .call = ip_set_flush,
1750 .attr_count = IPSET_ATTR_CMD_MAX,
1751 .policy = ip_set_setname_policy,
1753 [IPSET_CMD_RENAME] = {
1754 .call = ip_set_rename,
1755 .attr_count = IPSET_ATTR_CMD_MAX,
1756 .policy = ip_set_setname2_policy,
1758 [IPSET_CMD_SWAP] = {
1759 .call = ip_set_swap,
1760 .attr_count = IPSET_ATTR_CMD_MAX,
1761 .policy = ip_set_setname2_policy,
1763 [IPSET_CMD_LIST] = {
1764 .call = ip_set_dump,
1765 .attr_count = IPSET_ATTR_CMD_MAX,
1766 .policy = ip_set_setname_policy,
1768 [IPSET_CMD_SAVE] = {
1769 .call = ip_set_dump,
1770 .attr_count = IPSET_ATTR_CMD_MAX,
1771 .policy = ip_set_setname_policy,
1774 .call = ip_set_uadd,
1775 .attr_count = IPSET_ATTR_CMD_MAX,
1776 .policy = ip_set_adt_policy,
1779 .call = ip_set_udel,
1780 .attr_count = IPSET_ATTR_CMD_MAX,
1781 .policy = ip_set_adt_policy,
1783 [IPSET_CMD_TEST] = {
1784 .call = ip_set_utest,
1785 .attr_count = IPSET_ATTR_CMD_MAX,
1786 .policy = ip_set_adt_policy,
1788 [IPSET_CMD_HEADER] = {
1789 .call = ip_set_header,
1790 .attr_count = IPSET_ATTR_CMD_MAX,
1791 .policy = ip_set_setname_policy,
1793 [IPSET_CMD_TYPE] = {
1794 .call = ip_set_type,
1795 .attr_count = IPSET_ATTR_CMD_MAX,
1796 .policy = ip_set_type_policy,
1798 [IPSET_CMD_PROTOCOL] = {
1799 .call = ip_set_protocol,
1800 .attr_count = IPSET_ATTR_CMD_MAX,
1801 .policy = ip_set_protocol_policy,
1805 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = {
1807 .subsys_id = NFNL_SUBSYS_IPSET,
1808 .cb_count = IPSET_MSG_MAX,
1809 .cb = ip_set_netlink_subsys_cb,
1812 /* Interface to iptables/ip6tables */
1815 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len)
1819 int copylen = *len, ret = 0;
1820 struct net *net = sock_net(sk);
1821 struct ip_set_net *inst = ip_set_pernet(net);
1823 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1825 if (optval != SO_IP_SET)
1827 if (*len < sizeof(unsigned int))
1830 data = vmalloc(*len);
1833 if (copy_from_user(data, user, *len) != 0) {
1837 op = (unsigned int *) data;
1839 if (*op < IP_SET_OP_VERSION) {
1840 /* Check the version at the beginning of operations */
1841 struct ip_set_req_version *req_version = data;
1842 if (req_version->version != IPSET_PROTOCOL) {
1849 case IP_SET_OP_VERSION: {
1850 struct ip_set_req_version *req_version = data;
1852 if (*len != sizeof(struct ip_set_req_version)) {
1857 req_version->version = IPSET_PROTOCOL;
1858 ret = copy_to_user(user, req_version,
1859 sizeof(struct ip_set_req_version));
1862 case IP_SET_OP_GET_BYNAME: {
1863 struct ip_set_req_get_set *req_get = data;
1866 if (*len != sizeof(struct ip_set_req_get_set)) {
1870 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
1871 nfnl_lock(NFNL_SUBSYS_IPSET);
1872 find_set_and_id(inst, req_get->set.name, &id);
1873 req_get->set.index = id;
1874 nfnl_unlock(NFNL_SUBSYS_IPSET);
1877 case IP_SET_OP_GET_FNAME: {
1878 struct ip_set_req_get_set_family *req_get = data;
1881 if (*len != sizeof(struct ip_set_req_get_set_family)) {
1885 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
1886 nfnl_lock(NFNL_SUBSYS_IPSET);
1887 find_set_and_id(inst, req_get->set.name, &id);
1888 req_get->set.index = id;
1889 if (id != IPSET_INVALID_ID)
1890 req_get->family = nfnl_set(inst, id)->family;
1891 nfnl_unlock(NFNL_SUBSYS_IPSET);
1894 case IP_SET_OP_GET_BYINDEX: {
1895 struct ip_set_req_get_set *req_get = data;
1898 if (*len != sizeof(struct ip_set_req_get_set) ||
1899 req_get->set.index >= inst->ip_set_max) {
1903 nfnl_lock(NFNL_SUBSYS_IPSET);
1904 set = nfnl_set(inst, req_get->set.index);
1905 strncpy(req_get->set.name, set ? set->name : "",
1907 nfnl_unlock(NFNL_SUBSYS_IPSET);
1913 } /* end of switch(op) */
1916 ret = copy_to_user(user, data, copylen);
1925 static struct nf_sockopt_ops so_set __read_mostly = {
1927 .get_optmin = SO_IP_SET,
1928 .get_optmax = SO_IP_SET + 1,
1929 .get = &ip_set_sockfn_get,
1930 .owner = THIS_MODULE,
1933 static int __net_init
1934 ip_set_net_init(struct net *net)
1936 struct ip_set_net *inst = ip_set_pernet(net);
1937 struct ip_set **list;
1939 inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX;
1940 if (inst->ip_set_max >= IPSET_INVALID_ID)
1941 inst->ip_set_max = IPSET_INVALID_ID - 1;
1943 list = kzalloc(sizeof(struct ip_set *) * inst->ip_set_max, GFP_KERNEL);
1946 inst->is_deleted = 0;
1947 rcu_assign_pointer(inst->ip_set_list, list);
1948 pr_notice("ip_set: protocol %u\n", IPSET_PROTOCOL);
1952 static void __net_exit
1953 ip_set_net_exit(struct net *net)
1955 struct ip_set_net *inst = ip_set_pernet(net);
1957 struct ip_set *set = NULL;
1960 inst->is_deleted = 1; /* flag for ip_set_nfnl_put */
1962 for (i = 0; i < inst->ip_set_max; i++) {
1963 set = nfnl_set(inst, i);
1965 ip_set_destroy_set(inst, i);
1967 kfree(rcu_dereference_protected(inst->ip_set_list, 1));
1970 static struct pernet_operations ip_set_net_ops = {
1971 .init = ip_set_net_init,
1972 .exit = ip_set_net_exit,
1973 .id = &ip_set_net_id,
1974 .size = sizeof(struct ip_set_net)
1981 int ret = nfnetlink_subsys_register(&ip_set_netlink_subsys);
1983 pr_err("ip_set: cannot register with nfnetlink.\n");
1986 ret = nf_register_sockopt(&so_set);
1988 pr_err("SO_SET registry failed: %d\n", ret);
1989 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
1992 ret = register_pernet_subsys(&ip_set_net_ops);
1994 pr_err("ip_set: cannot register pernet_subsys.\n");
1995 nf_unregister_sockopt(&so_set);
1996 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2005 unregister_pernet_subsys(&ip_set_net_ops);
2006 nf_unregister_sockopt(&so_set);
2007 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2008 pr_debug("these are the famous last words\n");
2011 module_init(ip_set_init);
2012 module_exit(ip_set_fini);