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 bool is_deleted; /* deleted by ip_set_net_exit */
36 bool is_destroyed; /* all sets are destroyed */
39 static int ip_set_net_id __read_mostly;
41 static inline struct ip_set_net *ip_set_pernet(struct net *net)
43 return net_generic(net, ip_set_net_id);
47 #define STRNCMP(a, b) (strncmp(a, b, IPSET_MAXNAMELEN) == 0)
49 static unsigned int max_sets;
51 module_param(max_sets, int, 0600);
52 MODULE_PARM_DESC(max_sets, "maximal number of sets");
53 MODULE_LICENSE("GPL");
54 MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
55 MODULE_DESCRIPTION("core IP set support");
56 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_IPSET);
58 /* When the nfnl mutex is held: */
59 #define ip_set_dereference(p) \
60 rcu_dereference_protected(p, 1)
61 #define ip_set(inst, id) \
62 ip_set_dereference((inst)->ip_set_list)[id]
64 /* The set types are implemented in modules and registered set types
65 * can be found in ip_set_type_list. Adding/deleting types is
66 * serialized by ip_set_type_mutex.
70 ip_set_type_lock(void)
72 mutex_lock(&ip_set_type_mutex);
76 ip_set_type_unlock(void)
78 mutex_unlock(&ip_set_type_mutex);
81 /* Register and deregister settype */
83 static struct ip_set_type *
84 find_set_type(const char *name, u8 family, u8 revision)
86 struct ip_set_type *type;
88 list_for_each_entry_rcu(type, &ip_set_type_list, list)
89 if (STRNCMP(type->name, name) &&
90 (type->family == family ||
91 type->family == NFPROTO_UNSPEC) &&
92 revision >= type->revision_min &&
93 revision <= type->revision_max)
98 /* Unlock, try to load a set type module and lock again */
100 load_settype(const char *name)
102 nfnl_unlock(NFNL_SUBSYS_IPSET);
103 pr_debug("try to load ip_set_%s\n", name);
104 if (request_module("ip_set_%s", name) < 0) {
105 pr_warn("Can't find ip_set type %s\n", name);
106 nfnl_lock(NFNL_SUBSYS_IPSET);
109 nfnl_lock(NFNL_SUBSYS_IPSET);
113 /* Find a set type and reference it */
114 #define find_set_type_get(name, family, revision, found) \
115 __find_set_type_get(name, family, revision, found, false)
118 __find_set_type_get(const char *name, u8 family, u8 revision,
119 struct ip_set_type **found, bool retry)
121 struct ip_set_type *type;
124 if (retry && !load_settype(name))
125 return -IPSET_ERR_FIND_TYPE;
128 *found = find_set_type(name, family, revision);
130 err = !try_module_get((*found)->me) ? -EFAULT : 0;
133 /* Make sure the type is already loaded
134 * but we don't support the revision
136 list_for_each_entry_rcu(type, &ip_set_type_list, list)
137 if (STRNCMP(type->name, name)) {
138 err = -IPSET_ERR_FIND_TYPE;
143 return retry ? -IPSET_ERR_FIND_TYPE :
144 __find_set_type_get(name, family, revision, found, true);
151 /* Find a given set type by name and family.
152 * If we succeeded, the supported minimal and maximum revisions are
155 #define find_set_type_minmax(name, family, min, max) \
156 __find_set_type_minmax(name, family, min, max, false)
159 __find_set_type_minmax(const char *name, u8 family, u8 *min, u8 *max,
162 struct ip_set_type *type;
165 if (retry && !load_settype(name))
166 return -IPSET_ERR_FIND_TYPE;
168 *min = 255; *max = 0;
170 list_for_each_entry_rcu(type, &ip_set_type_list, list)
171 if (STRNCMP(type->name, name) &&
172 (type->family == family ||
173 type->family == NFPROTO_UNSPEC)) {
175 if (type->revision_min < *min)
176 *min = type->revision_min;
177 if (type->revision_max > *max)
178 *max = type->revision_max;
184 return retry ? -IPSET_ERR_FIND_TYPE :
185 __find_set_type_minmax(name, family, min, max, true);
188 #define family_name(f) ((f) == NFPROTO_IPV4 ? "inet" : \
189 (f) == NFPROTO_IPV6 ? "inet6" : "any")
191 /* Register a set type structure. The type is identified by
192 * the unique triple of name, family and revision.
195 ip_set_type_register(struct ip_set_type *type)
199 if (type->protocol != IPSET_PROTOCOL) {
200 pr_warn("ip_set type %s, family %s, revision %u:%u uses wrong protocol version %u (want %u)\n",
201 type->name, family_name(type->family),
202 type->revision_min, type->revision_max,
203 type->protocol, IPSET_PROTOCOL);
208 if (find_set_type(type->name, type->family, type->revision_min)) {
210 pr_warn("ip_set type %s, family %s with revision min %u already registered!\n",
211 type->name, family_name(type->family),
213 ip_set_type_unlock();
216 list_add_rcu(&type->list, &ip_set_type_list);
217 pr_debug("type %s, family %s, revision %u:%u registered.\n",
218 type->name, family_name(type->family),
219 type->revision_min, type->revision_max);
220 ip_set_type_unlock();
224 EXPORT_SYMBOL_GPL(ip_set_type_register);
226 /* Unregister a set type. There's a small race with ip_set_create */
228 ip_set_type_unregister(struct ip_set_type *type)
231 if (!find_set_type(type->name, type->family, type->revision_min)) {
232 pr_warn("ip_set type %s, family %s with revision min %u not registered\n",
233 type->name, family_name(type->family),
235 ip_set_type_unlock();
238 list_del_rcu(&type->list);
239 pr_debug("type %s, family %s with revision min %u unregistered.\n",
240 type->name, family_name(type->family), type->revision_min);
241 ip_set_type_unlock();
245 EXPORT_SYMBOL_GPL(ip_set_type_unregister);
247 /* Utility functions */
249 ip_set_alloc(size_t size)
251 void *members = NULL;
253 if (size < KMALLOC_MAX_SIZE)
254 members = kzalloc(size, GFP_KERNEL | __GFP_NOWARN);
257 pr_debug("%p: allocated with kmalloc\n", members);
261 members = vzalloc(size);
264 pr_debug("%p: allocated with vmalloc\n", members);
268 EXPORT_SYMBOL_GPL(ip_set_alloc);
271 ip_set_free(void *members)
273 pr_debug("%p: free with %s\n", members,
274 is_vmalloc_addr(members) ? "vfree" : "kfree");
277 EXPORT_SYMBOL_GPL(ip_set_free);
280 flag_nested(const struct nlattr *nla)
282 return nla->nla_type & NLA_F_NESTED;
285 static const struct nla_policy ipaddr_policy[IPSET_ATTR_IPADDR_MAX + 1] = {
286 [IPSET_ATTR_IPADDR_IPV4] = { .type = NLA_U32 },
287 [IPSET_ATTR_IPADDR_IPV6] = { .type = NLA_BINARY,
288 .len = sizeof(struct in6_addr) },
292 ip_set_get_ipaddr4(struct nlattr *nla, __be32 *ipaddr)
294 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1];
296 if (unlikely(!flag_nested(nla)))
297 return -IPSET_ERR_PROTOCOL;
298 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla, ipaddr_policy))
299 return -IPSET_ERR_PROTOCOL;
300 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV4)))
301 return -IPSET_ERR_PROTOCOL;
303 *ipaddr = nla_get_be32(tb[IPSET_ATTR_IPADDR_IPV4]);
306 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr4);
309 ip_set_get_ipaddr6(struct nlattr *nla, union nf_inet_addr *ipaddr)
311 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1];
313 if (unlikely(!flag_nested(nla)))
314 return -IPSET_ERR_PROTOCOL;
316 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla, ipaddr_policy))
317 return -IPSET_ERR_PROTOCOL;
318 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV6)))
319 return -IPSET_ERR_PROTOCOL;
321 memcpy(ipaddr, nla_data(tb[IPSET_ATTR_IPADDR_IPV6]),
322 sizeof(struct in6_addr));
325 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr6);
327 typedef void (*destroyer)(void *);
328 /* ipset data extension types, in size order */
330 const struct ip_set_ext_type ip_set_extensions[] = {
331 [IPSET_EXT_ID_COUNTER] = {
332 .type = IPSET_EXT_COUNTER,
333 .flag = IPSET_FLAG_WITH_COUNTERS,
334 .len = sizeof(struct ip_set_counter),
335 .align = __alignof__(struct ip_set_counter),
337 [IPSET_EXT_ID_TIMEOUT] = {
338 .type = IPSET_EXT_TIMEOUT,
339 .len = sizeof(unsigned long),
340 .align = __alignof__(unsigned long),
342 [IPSET_EXT_ID_SKBINFO] = {
343 .type = IPSET_EXT_SKBINFO,
344 .flag = IPSET_FLAG_WITH_SKBINFO,
345 .len = sizeof(struct ip_set_skbinfo),
346 .align = __alignof__(struct ip_set_skbinfo),
348 [IPSET_EXT_ID_COMMENT] = {
349 .type = IPSET_EXT_COMMENT | IPSET_EXT_DESTROY,
350 .flag = IPSET_FLAG_WITH_COMMENT,
351 .len = sizeof(struct ip_set_comment),
352 .align = __alignof__(struct ip_set_comment),
353 .destroy = (destroyer) ip_set_comment_free,
356 EXPORT_SYMBOL_GPL(ip_set_extensions);
359 add_extension(enum ip_set_ext_id id, u32 flags, struct nlattr *tb[])
361 return ip_set_extensions[id].flag ?
362 (flags & ip_set_extensions[id].flag) :
363 !!tb[IPSET_ATTR_TIMEOUT];
367 ip_set_elem_len(struct ip_set *set, struct nlattr *tb[], size_t len,
370 enum ip_set_ext_id id;
373 if (tb[IPSET_ATTR_CADT_FLAGS])
374 cadt_flags = ip_set_get_h32(tb[IPSET_ATTR_CADT_FLAGS]);
375 if (cadt_flags & IPSET_FLAG_WITH_FORCEADD)
376 set->flags |= IPSET_CREATE_FLAG_FORCEADD;
379 for (id = 0; id < IPSET_EXT_ID_MAX; id++) {
380 if (!add_extension(id, cadt_flags, tb))
382 len = ALIGN(len, ip_set_extensions[id].align);
383 set->offset[id] = len;
384 set->extensions |= ip_set_extensions[id].type;
385 len += ip_set_extensions[id].len;
387 return ALIGN(len, align);
389 EXPORT_SYMBOL_GPL(ip_set_elem_len);
392 ip_set_get_extensions(struct ip_set *set, struct nlattr *tb[],
393 struct ip_set_ext *ext)
397 if (unlikely(!ip_set_optattr_netorder(tb, IPSET_ATTR_TIMEOUT) ||
398 !ip_set_optattr_netorder(tb, IPSET_ATTR_PACKETS) ||
399 !ip_set_optattr_netorder(tb, IPSET_ATTR_BYTES) ||
400 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBMARK) ||
401 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBPRIO) ||
402 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBQUEUE)))
403 return -IPSET_ERR_PROTOCOL;
405 if (tb[IPSET_ATTR_TIMEOUT]) {
406 if (!SET_WITH_TIMEOUT(set))
407 return -IPSET_ERR_TIMEOUT;
408 ext->timeout = ip_set_timeout_uget(tb[IPSET_ATTR_TIMEOUT]);
410 if (tb[IPSET_ATTR_BYTES] || tb[IPSET_ATTR_PACKETS]) {
411 if (!SET_WITH_COUNTER(set))
412 return -IPSET_ERR_COUNTER;
413 if (tb[IPSET_ATTR_BYTES])
414 ext->bytes = be64_to_cpu(nla_get_be64(
415 tb[IPSET_ATTR_BYTES]));
416 if (tb[IPSET_ATTR_PACKETS])
417 ext->packets = be64_to_cpu(nla_get_be64(
418 tb[IPSET_ATTR_PACKETS]));
420 if (tb[IPSET_ATTR_COMMENT]) {
421 if (!SET_WITH_COMMENT(set))
422 return -IPSET_ERR_COMMENT;
423 ext->comment = ip_set_comment_uget(tb[IPSET_ATTR_COMMENT]);
425 if (tb[IPSET_ATTR_SKBMARK]) {
426 if (!SET_WITH_SKBINFO(set))
427 return -IPSET_ERR_SKBINFO;
428 fullmark = be64_to_cpu(nla_get_be64(tb[IPSET_ATTR_SKBMARK]));
429 ext->skbmark = fullmark >> 32;
430 ext->skbmarkmask = fullmark & 0xffffffff;
432 if (tb[IPSET_ATTR_SKBPRIO]) {
433 if (!SET_WITH_SKBINFO(set))
434 return -IPSET_ERR_SKBINFO;
435 ext->skbprio = be32_to_cpu(nla_get_be32(
436 tb[IPSET_ATTR_SKBPRIO]));
438 if (tb[IPSET_ATTR_SKBQUEUE]) {
439 if (!SET_WITH_SKBINFO(set))
440 return -IPSET_ERR_SKBINFO;
441 ext->skbqueue = be16_to_cpu(nla_get_be16(
442 tb[IPSET_ATTR_SKBQUEUE]));
446 EXPORT_SYMBOL_GPL(ip_set_get_extensions);
449 ip_set_put_extensions(struct sk_buff *skb, const struct ip_set *set,
450 const void *e, bool active)
452 if (SET_WITH_TIMEOUT(set)) {
453 unsigned long *timeout = ext_timeout(e, set);
455 if (nla_put_net32(skb, IPSET_ATTR_TIMEOUT,
456 htonl(active ? ip_set_timeout_get(timeout)
460 if (SET_WITH_COUNTER(set) &&
461 ip_set_put_counter(skb, ext_counter(e, set)))
463 if (SET_WITH_COMMENT(set) &&
464 ip_set_put_comment(skb, ext_comment(e, set)))
466 if (SET_WITH_SKBINFO(set) &&
467 ip_set_put_skbinfo(skb, ext_skbinfo(e, set)))
471 EXPORT_SYMBOL_GPL(ip_set_put_extensions);
473 /* Creating/destroying/renaming/swapping affect the existence and
474 * the properties of a set. All of these can be executed from userspace
475 * only and serialized by the nfnl mutex indirectly from nfnetlink.
477 * Sets are identified by their index in ip_set_list and the index
478 * is used by the external references (set/SET netfilter modules).
480 * The set behind an index may change by swapping only, from userspace.
484 __ip_set_get(struct ip_set *set)
486 write_lock_bh(&ip_set_ref_lock);
488 write_unlock_bh(&ip_set_ref_lock);
492 __ip_set_put(struct ip_set *set)
494 write_lock_bh(&ip_set_ref_lock);
495 BUG_ON(set->ref == 0);
497 write_unlock_bh(&ip_set_ref_lock);
500 /* Add, del and test set entries from kernel.
502 * The set behind the index must exist and must be referenced
503 * so it can't be destroyed (or changed) under our foot.
506 static inline struct ip_set *
507 ip_set_rcu_get(struct net *net, ip_set_id_t index)
510 struct ip_set_net *inst = ip_set_pernet(net);
513 /* ip_set_list itself needs to be protected */
514 set = rcu_dereference(inst->ip_set_list)[index];
521 ip_set_test(ip_set_id_t index, const struct sk_buff *skb,
522 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
524 struct ip_set *set = ip_set_rcu_get(par->net, index);
528 pr_debug("set %s, index %u\n", set->name, index);
530 if (opt->dim < set->type->dimension ||
531 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
535 ret = set->variant->kadt(set, skb, par, IPSET_TEST, opt);
536 rcu_read_unlock_bh();
538 if (ret == -EAGAIN) {
539 /* Type requests element to be completed */
540 pr_debug("element must be completed, ADD is triggered\n");
541 spin_lock_bh(&set->lock);
542 set->variant->kadt(set, skb, par, IPSET_ADD, opt);
543 spin_unlock_bh(&set->lock);
546 /* --return-nomatch: invert matched element */
547 if ((opt->cmdflags & IPSET_FLAG_RETURN_NOMATCH) &&
548 (set->type->features & IPSET_TYPE_NOMATCH) &&
549 (ret > 0 || ret == -ENOTEMPTY))
553 /* Convert error codes to nomatch */
554 return (ret < 0 ? 0 : ret);
556 EXPORT_SYMBOL_GPL(ip_set_test);
559 ip_set_add(ip_set_id_t index, const struct sk_buff *skb,
560 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
562 struct ip_set *set = ip_set_rcu_get(par->net, index);
566 pr_debug("set %s, index %u\n", set->name, index);
568 if (opt->dim < set->type->dimension ||
569 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
570 return -IPSET_ERR_TYPE_MISMATCH;
572 spin_lock_bh(&set->lock);
573 ret = set->variant->kadt(set, skb, par, IPSET_ADD, opt);
574 spin_unlock_bh(&set->lock);
578 EXPORT_SYMBOL_GPL(ip_set_add);
581 ip_set_del(ip_set_id_t index, const struct sk_buff *skb,
582 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
584 struct ip_set *set = ip_set_rcu_get(par->net, index);
588 pr_debug("set %s, index %u\n", set->name, index);
590 if (opt->dim < set->type->dimension ||
591 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
592 return -IPSET_ERR_TYPE_MISMATCH;
594 spin_lock_bh(&set->lock);
595 ret = set->variant->kadt(set, skb, par, IPSET_DEL, opt);
596 spin_unlock_bh(&set->lock);
600 EXPORT_SYMBOL_GPL(ip_set_del);
602 /* Find set by name, reference it once. The reference makes sure the
603 * thing pointed to, does not go away under our feet.
607 ip_set_get_byname(struct net *net, const char *name, struct ip_set **set)
609 ip_set_id_t i, index = IPSET_INVALID_ID;
611 struct ip_set_net *inst = ip_set_pernet(net);
614 for (i = 0; i < inst->ip_set_max; i++) {
615 s = rcu_dereference(inst->ip_set_list)[i];
616 if (s && STRNCMP(s->name, name)) {
627 EXPORT_SYMBOL_GPL(ip_set_get_byname);
629 /* If the given set pointer points to a valid set, decrement
630 * reference count by 1. The caller shall not assume the index
631 * to be valid, after calling this function.
636 __ip_set_put_byindex(struct ip_set_net *inst, ip_set_id_t index)
641 set = rcu_dereference(inst->ip_set_list)[index];
648 ip_set_put_byindex(struct net *net, ip_set_id_t index)
650 struct ip_set_net *inst = ip_set_pernet(net);
652 __ip_set_put_byindex(inst, index);
654 EXPORT_SYMBOL_GPL(ip_set_put_byindex);
656 /* Get the name of a set behind a set index.
657 * We assume the set is referenced, so it does exist and
658 * can't be destroyed. The set cannot be renamed due to
659 * the referencing either.
663 ip_set_name_byindex(struct net *net, ip_set_id_t index)
665 const struct ip_set *set = ip_set_rcu_get(net, index);
668 BUG_ON(set->ref == 0);
670 /* Referenced, so it's safe */
673 EXPORT_SYMBOL_GPL(ip_set_name_byindex);
675 /* Routines to call by external subsystems, which do not
676 * call nfnl_lock for us.
679 /* Find set by index, reference it once. The reference makes sure the
680 * thing pointed to, does not go away under our feet.
682 * The nfnl mutex is used in the function.
685 ip_set_nfnl_get_byindex(struct net *net, ip_set_id_t index)
688 struct ip_set_net *inst = ip_set_pernet(net);
690 if (index >= inst->ip_set_max)
691 return IPSET_INVALID_ID;
693 nfnl_lock(NFNL_SUBSYS_IPSET);
694 set = ip_set(inst, index);
698 index = IPSET_INVALID_ID;
699 nfnl_unlock(NFNL_SUBSYS_IPSET);
703 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex);
705 /* If the given set pointer points to a valid set, decrement
706 * reference count by 1. The caller shall not assume the index
707 * to be valid, after calling this function.
709 * The nfnl mutex is used in the function.
712 ip_set_nfnl_put(struct net *net, ip_set_id_t index)
715 struct ip_set_net *inst = ip_set_pernet(net);
717 nfnl_lock(NFNL_SUBSYS_IPSET);
718 if (!inst->is_deleted) { /* already deleted from ip_set_net_exit() */
719 set = ip_set(inst, index);
723 nfnl_unlock(NFNL_SUBSYS_IPSET);
725 EXPORT_SYMBOL_GPL(ip_set_nfnl_put);
727 /* Communication protocol with userspace over netlink.
729 * The commands are serialized by the nfnl mutex.
733 protocol_failed(const struct nlattr * const tb[])
735 return !tb[IPSET_ATTR_PROTOCOL] ||
736 nla_get_u8(tb[IPSET_ATTR_PROTOCOL]) != IPSET_PROTOCOL;
740 flag_exist(const struct nlmsghdr *nlh)
742 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST;
745 static struct nlmsghdr *
746 start_msg(struct sk_buff *skb, u32 portid, u32 seq, unsigned int flags,
749 struct nlmsghdr *nlh;
750 struct nfgenmsg *nfmsg;
752 nlh = nlmsg_put(skb, portid, seq, cmd | (NFNL_SUBSYS_IPSET << 8),
753 sizeof(*nfmsg), flags);
757 nfmsg = nlmsg_data(nlh);
758 nfmsg->nfgen_family = NFPROTO_IPV4;
759 nfmsg->version = NFNETLINK_V0;
767 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = {
768 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
769 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
770 .len = IPSET_MAXNAMELEN - 1 },
771 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
772 .len = IPSET_MAXNAMELEN - 1},
773 [IPSET_ATTR_REVISION] = { .type = NLA_U8 },
774 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
775 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
778 static struct ip_set *
779 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id)
781 struct ip_set *set = NULL;
784 *id = IPSET_INVALID_ID;
785 for (i = 0; i < inst->ip_set_max; i++) {
786 set = ip_set(inst, i);
787 if (set && STRNCMP(set->name, name)) {
792 return (*id == IPSET_INVALID_ID ? NULL : set);
795 static inline struct ip_set *
796 find_set(struct ip_set_net *inst, const char *name)
800 return find_set_and_id(inst, name, &id);
804 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index,
810 *index = IPSET_INVALID_ID;
811 for (i = 0; i < inst->ip_set_max; i++) {
814 if (*index == IPSET_INVALID_ID)
816 } else if (STRNCMP(name, s->name)) {
822 if (*index == IPSET_INVALID_ID)
823 /* No free slot remained */
824 return -IPSET_ERR_MAX_SETS;
829 ip_set_none(struct sock *ctnl, struct sk_buff *skb,
830 const struct nlmsghdr *nlh,
831 const struct nlattr * const attr[])
837 ip_set_create(struct sock *ctnl, struct sk_buff *skb,
838 const struct nlmsghdr *nlh,
839 const struct nlattr * const attr[])
841 struct net *net = sock_net(ctnl);
842 struct ip_set_net *inst = ip_set_pernet(net);
843 struct ip_set *set, *clash = NULL;
844 ip_set_id_t index = IPSET_INVALID_ID;
845 struct nlattr *tb[IPSET_ATTR_CREATE_MAX + 1] = {};
846 const char *name, *typename;
848 u32 flags = flag_exist(nlh);
851 if (unlikely(protocol_failed(attr) ||
852 !attr[IPSET_ATTR_SETNAME] ||
853 !attr[IPSET_ATTR_TYPENAME] ||
854 !attr[IPSET_ATTR_REVISION] ||
855 !attr[IPSET_ATTR_FAMILY] ||
856 (attr[IPSET_ATTR_DATA] &&
857 !flag_nested(attr[IPSET_ATTR_DATA]))))
858 return -IPSET_ERR_PROTOCOL;
860 name = nla_data(attr[IPSET_ATTR_SETNAME]);
861 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
862 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
863 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]);
864 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
865 name, typename, family_name(family), revision);
867 /* First, and without any locks, allocate and initialize
868 * a normal base set structure.
870 set = kzalloc(sizeof(*set), GFP_KERNEL);
873 spin_lock_init(&set->lock);
874 strlcpy(set->name, name, IPSET_MAXNAMELEN);
875 set->family = family;
876 set->revision = revision;
878 /* Next, check that we know the type, and take
879 * a reference on the type, to make sure it stays available
880 * while constructing our new set.
882 * After referencing the type, we try to create the type
883 * specific part of the set without holding any locks.
885 ret = find_set_type_get(typename, family, revision, &set->type);
889 /* Without holding any locks, create private part. */
890 if (attr[IPSET_ATTR_DATA] &&
891 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA],
892 set->type->create_policy)) {
893 ret = -IPSET_ERR_PROTOCOL;
897 ret = set->type->create(net, set, tb, flags);
901 /* BTW, ret==0 here. */
903 /* Here, we have a valid, constructed set and we are protected
904 * by the nfnl mutex. Find the first free index in ip_set_list
905 * and check clashing.
907 ret = find_free_id(inst, set->name, &index, &clash);
908 if (ret == -EEXIST) {
909 /* If this is the same set and requested, ignore error */
910 if ((flags & IPSET_FLAG_EXIST) &&
911 STRNCMP(set->type->name, clash->type->name) &&
912 set->type->family == clash->type->family &&
913 set->type->revision_min == clash->type->revision_min &&
914 set->type->revision_max == clash->type->revision_max &&
915 set->variant->same_set(set, clash))
918 } else if (ret == -IPSET_ERR_MAX_SETS) {
919 struct ip_set **list, **tmp;
920 ip_set_id_t i = inst->ip_set_max + IP_SET_INC;
922 if (i < inst->ip_set_max || i == IPSET_INVALID_ID)
926 list = kcalloc(i, sizeof(struct ip_set *), GFP_KERNEL);
929 /* nfnl mutex is held, both lists are valid */
930 tmp = ip_set_dereference(inst->ip_set_list);
931 memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max);
932 rcu_assign_pointer(inst->ip_set_list, list);
933 /* Make sure all current packets have passed through */
936 index = inst->ip_set_max;
937 inst->ip_set_max = i;
944 /* Finally! Add our shiny new set to the list, and be done. */
945 pr_debug("create: '%s' created with index %u!\n", set->name, index);
946 ip_set(inst, index) = set;
951 set->variant->destroy(set);
953 module_put(set->type->me);
961 static const struct nla_policy
962 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = {
963 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
964 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
965 .len = IPSET_MAXNAMELEN - 1 },
969 ip_set_destroy_set(struct ip_set *set)
971 pr_debug("set: %s\n", set->name);
973 /* Must call it without holding any lock */
974 set->variant->destroy(set);
975 module_put(set->type->me);
980 ip_set_destroy(struct sock *ctnl, struct sk_buff *skb,
981 const struct nlmsghdr *nlh,
982 const struct nlattr * const attr[])
984 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
989 if (unlikely(protocol_failed(attr)))
990 return -IPSET_ERR_PROTOCOL;
992 /* Commands are serialized and references are
993 * protected by the ip_set_ref_lock.
994 * External systems (i.e. xt_set) must call
995 * ip_set_put|get_nfnl_* functions, that way we
996 * can safely check references here.
998 * list:set timer can only decrement the reference
999 * counter, so if it's already zero, we can proceed
1000 * without holding the lock.
1002 read_lock_bh(&ip_set_ref_lock);
1003 if (!attr[IPSET_ATTR_SETNAME]) {
1004 for (i = 0; i < inst->ip_set_max; i++) {
1005 s = ip_set(inst, i);
1007 ret = -IPSET_ERR_BUSY;
1011 inst->is_destroyed = true;
1012 read_unlock_bh(&ip_set_ref_lock);
1013 for (i = 0; i < inst->ip_set_max; i++) {
1014 s = ip_set(inst, i);
1016 ip_set(inst, i) = NULL;
1017 ip_set_destroy_set(s);
1020 /* Modified by ip_set_destroy() only, which is serialized */
1021 inst->is_destroyed = false;
1023 s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1028 } else if (s->ref) {
1029 ret = -IPSET_ERR_BUSY;
1032 ip_set(inst, i) = NULL;
1033 read_unlock_bh(&ip_set_ref_lock);
1035 ip_set_destroy_set(s);
1039 read_unlock_bh(&ip_set_ref_lock);
1046 ip_set_flush_set(struct ip_set *set)
1048 pr_debug("set: %s\n", set->name);
1050 spin_lock_bh(&set->lock);
1051 set->variant->flush(set);
1052 spin_unlock_bh(&set->lock);
1056 ip_set_flush(struct sock *ctnl, struct sk_buff *skb,
1057 const struct nlmsghdr *nlh,
1058 const struct nlattr * const attr[])
1060 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1064 if (unlikely(protocol_failed(attr)))
1065 return -IPSET_ERR_PROTOCOL;
1067 if (!attr[IPSET_ATTR_SETNAME]) {
1068 for (i = 0; i < inst->ip_set_max; i++) {
1069 s = ip_set(inst, i);
1071 ip_set_flush_set(s);
1074 s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1078 ip_set_flush_set(s);
1086 static const struct nla_policy
1087 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = {
1088 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1089 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1090 .len = IPSET_MAXNAMELEN - 1 },
1091 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING,
1092 .len = IPSET_MAXNAMELEN - 1 },
1096 ip_set_rename(struct sock *ctnl, struct sk_buff *skb,
1097 const struct nlmsghdr *nlh,
1098 const struct nlattr * const attr[])
1100 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1101 struct ip_set *set, *s;
1106 if (unlikely(protocol_failed(attr) ||
1107 !attr[IPSET_ATTR_SETNAME] ||
1108 !attr[IPSET_ATTR_SETNAME2]))
1109 return -IPSET_ERR_PROTOCOL;
1111 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1115 read_lock_bh(&ip_set_ref_lock);
1116 if (set->ref != 0) {
1117 ret = -IPSET_ERR_REFERENCED;
1121 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]);
1122 for (i = 0; i < inst->ip_set_max; i++) {
1123 s = ip_set(inst, i);
1124 if (s && STRNCMP(s->name, name2)) {
1125 ret = -IPSET_ERR_EXIST_SETNAME2;
1129 strncpy(set->name, name2, IPSET_MAXNAMELEN);
1132 read_unlock_bh(&ip_set_ref_lock);
1136 /* Swap two sets so that name/index points to the other.
1137 * References and set names are also swapped.
1139 * The commands are serialized by the nfnl mutex and references are
1140 * protected by the ip_set_ref_lock. The kernel interfaces
1141 * do not hold the mutex but the pointer settings are atomic
1142 * so the ip_set_list always contains valid pointers to the sets.
1146 ip_set_swap(struct sock *ctnl, struct sk_buff *skb,
1147 const struct nlmsghdr *nlh,
1148 const struct nlattr * const attr[])
1150 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1151 struct ip_set *from, *to;
1152 ip_set_id_t from_id, to_id;
1153 char from_name[IPSET_MAXNAMELEN];
1155 if (unlikely(protocol_failed(attr) ||
1156 !attr[IPSET_ATTR_SETNAME] ||
1157 !attr[IPSET_ATTR_SETNAME2]))
1158 return -IPSET_ERR_PROTOCOL;
1160 from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1165 to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]),
1168 return -IPSET_ERR_EXIST_SETNAME2;
1170 /* Features must not change.
1171 * Not an artifical restriction anymore, as we must prevent
1172 * possible loops created by swapping in setlist type of sets.
1174 if (!(from->type->features == to->type->features &&
1175 from->family == to->family))
1176 return -IPSET_ERR_TYPE_MISMATCH;
1178 strncpy(from_name, from->name, IPSET_MAXNAMELEN);
1179 strncpy(from->name, to->name, IPSET_MAXNAMELEN);
1180 strncpy(to->name, from_name, IPSET_MAXNAMELEN);
1182 write_lock_bh(&ip_set_ref_lock);
1183 swap(from->ref, to->ref);
1184 ip_set(inst, from_id) = to;
1185 ip_set(inst, to_id) = from;
1186 write_unlock_bh(&ip_set_ref_lock);
1191 /* List/save set data */
1198 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1199 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1202 ip_set_dump_done(struct netlink_callback *cb)
1204 if (cb->args[IPSET_CB_ARG0]) {
1205 struct ip_set_net *inst =
1206 (struct ip_set_net *)cb->args[IPSET_CB_NET];
1207 ip_set_id_t index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1208 struct ip_set *set = ip_set(inst, index);
1210 if (set->variant->uref)
1211 set->variant->uref(set, cb, false);
1212 pr_debug("release set %s\n", set->name);
1213 __ip_set_put_byindex(inst, index);
1219 dump_attrs(struct nlmsghdr *nlh)
1221 const struct nlattr *attr;
1224 pr_debug("dump nlmsg\n");
1225 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) {
1226 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len);
1231 dump_init(struct netlink_callback *cb, struct ip_set_net *inst)
1233 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb);
1234 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1235 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1236 struct nlattr *attr = (void *)nlh + min_len;
1240 /* Second pass, so parser can't fail */
1241 nla_parse(cda, IPSET_ATTR_CMD_MAX,
1242 attr, nlh->nlmsg_len - min_len, ip_set_setname_policy);
1244 if (cda[IPSET_ATTR_SETNAME]) {
1247 set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]),
1252 dump_type = DUMP_ONE;
1253 cb->args[IPSET_CB_INDEX] = index;
1255 dump_type = DUMP_ALL;
1258 if (cda[IPSET_ATTR_FLAGS]) {
1259 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]);
1261 dump_type |= (f << 16);
1263 cb->args[IPSET_CB_NET] = (unsigned long)inst;
1264 cb->args[IPSET_CB_DUMP] = dump_type;
1270 ip_set_dump_start(struct sk_buff *skb, struct netlink_callback *cb)
1272 ip_set_id_t index = IPSET_INVALID_ID, max;
1273 struct ip_set *set = NULL;
1274 struct nlmsghdr *nlh = NULL;
1275 unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0;
1276 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1277 u32 dump_type, dump_flags;
1281 if (!cb->args[IPSET_CB_DUMP]) {
1282 ret = dump_init(cb, inst);
1284 nlh = nlmsg_hdr(cb->skb);
1285 /* We have to create and send the error message
1288 if (nlh->nlmsg_flags & NLM_F_ACK)
1289 netlink_ack(cb->skb, nlh, ret);
1294 if (cb->args[IPSET_CB_INDEX] >= inst->ip_set_max)
1297 dump_type = DUMP_TYPE(cb->args[IPSET_CB_DUMP]);
1298 dump_flags = DUMP_FLAGS(cb->args[IPSET_CB_DUMP]);
1299 max = dump_type == DUMP_ONE ? cb->args[IPSET_CB_INDEX] + 1
1302 pr_debug("dump type, flag: %u %u index: %ld\n",
1303 dump_type, dump_flags, cb->args[IPSET_CB_INDEX]);
1304 for (; cb->args[IPSET_CB_INDEX] < max; cb->args[IPSET_CB_INDEX]++) {
1305 index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1306 write_lock_bh(&ip_set_ref_lock);
1307 set = ip_set(inst, index);
1308 is_destroyed = inst->is_destroyed;
1309 if (!set || is_destroyed) {
1310 write_unlock_bh(&ip_set_ref_lock);
1311 if (dump_type == DUMP_ONE) {
1316 /* All sets are just being destroyed */
1322 /* When dumping all sets, we must dump "sorted"
1323 * so that lists (unions of sets) are dumped last.
1325 if (dump_type != DUMP_ONE &&
1326 ((dump_type == DUMP_ALL) ==
1327 !!(set->type->features & IPSET_DUMP_LAST))) {
1328 write_unlock_bh(&ip_set_ref_lock);
1331 pr_debug("List set: %s\n", set->name);
1332 if (!cb->args[IPSET_CB_ARG0]) {
1333 /* Start listing: make sure set won't be destroyed */
1334 pr_debug("reference set\n");
1337 write_unlock_bh(&ip_set_ref_lock);
1338 nlh = start_msg(skb, NETLINK_CB(cb->skb).portid,
1339 cb->nlh->nlmsg_seq, flags,
1343 goto release_refcount;
1345 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1346 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name))
1347 goto nla_put_failure;
1348 if (dump_flags & IPSET_FLAG_LIST_SETNAME)
1350 switch (cb->args[IPSET_CB_ARG0]) {
1352 /* Core header data */
1353 if (nla_put_string(skb, IPSET_ATTR_TYPENAME,
1355 nla_put_u8(skb, IPSET_ATTR_FAMILY,
1357 nla_put_u8(skb, IPSET_ATTR_REVISION,
1359 goto nla_put_failure;
1360 ret = set->variant->head(set, skb);
1362 goto release_refcount;
1363 if (dump_flags & IPSET_FLAG_LIST_HEADER)
1365 if (set->variant->uref)
1366 set->variant->uref(set, cb, true);
1367 /* Fall through and add elements */
1370 ret = set->variant->list(set, skb, cb);
1371 rcu_read_unlock_bh();
1372 if (!cb->args[IPSET_CB_ARG0])
1373 /* Set is done, proceed with next one */
1375 goto release_refcount;
1378 /* If we dump all sets, continue with dumping last ones */
1379 if (dump_type == DUMP_ALL) {
1380 dump_type = DUMP_LAST;
1381 cb->args[IPSET_CB_DUMP] = dump_type | (dump_flags << 16);
1382 cb->args[IPSET_CB_INDEX] = 0;
1383 if (set && set->variant->uref)
1384 set->variant->uref(set, cb, false);
1392 if (dump_type == DUMP_ONE)
1393 cb->args[IPSET_CB_INDEX] = IPSET_INVALID_ID;
1395 cb->args[IPSET_CB_INDEX]++;
1397 /* If there was an error or set is done, release set */
1398 if (ret || !cb->args[IPSET_CB_ARG0]) {
1399 set = ip_set(inst, index);
1400 if (set->variant->uref)
1401 set->variant->uref(set, cb, false);
1402 pr_debug("release set %s\n", set->name);
1403 __ip_set_put_byindex(inst, index);
1404 cb->args[IPSET_CB_ARG0] = 0;
1408 nlmsg_end(skb, nlh);
1409 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len);
1413 return ret < 0 ? ret : skb->len;
1417 ip_set_dump(struct sock *ctnl, struct sk_buff *skb,
1418 const struct nlmsghdr *nlh,
1419 const struct nlattr * const attr[])
1421 if (unlikely(protocol_failed(attr)))
1422 return -IPSET_ERR_PROTOCOL;
1425 struct netlink_dump_control c = {
1426 .dump = ip_set_dump_start,
1427 .done = ip_set_dump_done,
1429 return netlink_dump_start(ctnl, skb, nlh, &c);
1433 /* Add, del and test */
1435 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = {
1436 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1437 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1438 .len = IPSET_MAXNAMELEN - 1 },
1439 [IPSET_ATTR_LINENO] = { .type = NLA_U32 },
1440 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
1441 [IPSET_ATTR_ADT] = { .type = NLA_NESTED },
1445 call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set,
1446 struct nlattr *tb[], enum ipset_adt adt,
1447 u32 flags, bool use_lineno)
1451 bool eexist = flags & IPSET_FLAG_EXIST, retried = false;
1454 spin_lock_bh(&set->lock);
1455 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried);
1456 spin_unlock_bh(&set->lock);
1458 } while (ret == -EAGAIN &&
1459 set->variant->resize &&
1460 (ret = set->variant->resize(set, retried)) == 0);
1462 if (!ret || (ret == -IPSET_ERR_EXIST && eexist))
1464 if (lineno && use_lineno) {
1465 /* Error in restore/batch mode: send back lineno */
1466 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb);
1467 struct sk_buff *skb2;
1468 struct nlmsgerr *errmsg;
1469 size_t payload = min(SIZE_MAX,
1470 sizeof(*errmsg) + nlmsg_len(nlh));
1471 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1472 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1473 struct nlattr *cmdattr;
1476 skb2 = nlmsg_new(payload, GFP_KERNEL);
1479 rep = __nlmsg_put(skb2, NETLINK_CB(skb).portid,
1480 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0);
1481 errmsg = nlmsg_data(rep);
1482 errmsg->error = ret;
1483 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len);
1484 cmdattr = (void *)&errmsg->msg + min_len;
1486 nla_parse(cda, IPSET_ATTR_CMD_MAX,
1487 cmdattr, nlh->nlmsg_len - min_len,
1490 errline = nla_data(cda[IPSET_ATTR_LINENO]);
1494 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid,
1496 /* Signal netlink not to send its ACK/errmsg. */
1504 ip_set_uadd(struct sock *ctnl, struct sk_buff *skb,
1505 const struct nlmsghdr *nlh,
1506 const struct nlattr * const attr[])
1508 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1510 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1511 const struct nlattr *nla;
1512 u32 flags = flag_exist(nlh);
1516 if (unlikely(protocol_failed(attr) ||
1517 !attr[IPSET_ATTR_SETNAME] ||
1518 !((attr[IPSET_ATTR_DATA] != NULL) ^
1519 (attr[IPSET_ATTR_ADT] != NULL)) ||
1520 (attr[IPSET_ATTR_DATA] &&
1521 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1522 (attr[IPSET_ATTR_ADT] &&
1523 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1524 !attr[IPSET_ATTR_LINENO]))))
1525 return -IPSET_ERR_PROTOCOL;
1527 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1531 use_lineno = !!attr[IPSET_ATTR_LINENO];
1532 if (attr[IPSET_ATTR_DATA]) {
1533 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1534 attr[IPSET_ATTR_DATA],
1535 set->type->adt_policy))
1536 return -IPSET_ERR_PROTOCOL;
1537 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD, flags,
1542 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1543 memset(tb, 0, sizeof(tb));
1544 if (nla_type(nla) != IPSET_ATTR_DATA ||
1545 !flag_nested(nla) ||
1546 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1547 set->type->adt_policy))
1548 return -IPSET_ERR_PROTOCOL;
1549 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD,
1559 ip_set_udel(struct sock *ctnl, struct sk_buff *skb,
1560 const struct nlmsghdr *nlh,
1561 const struct nlattr * const attr[])
1563 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1565 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1566 const struct nlattr *nla;
1567 u32 flags = flag_exist(nlh);
1571 if (unlikely(protocol_failed(attr) ||
1572 !attr[IPSET_ATTR_SETNAME] ||
1573 !((attr[IPSET_ATTR_DATA] != NULL) ^
1574 (attr[IPSET_ATTR_ADT] != NULL)) ||
1575 (attr[IPSET_ATTR_DATA] &&
1576 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1577 (attr[IPSET_ATTR_ADT] &&
1578 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1579 !attr[IPSET_ATTR_LINENO]))))
1580 return -IPSET_ERR_PROTOCOL;
1582 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1586 use_lineno = !!attr[IPSET_ATTR_LINENO];
1587 if (attr[IPSET_ATTR_DATA]) {
1588 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1589 attr[IPSET_ATTR_DATA],
1590 set->type->adt_policy))
1591 return -IPSET_ERR_PROTOCOL;
1592 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL, flags,
1597 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1598 memset(tb, 0, sizeof(*tb));
1599 if (nla_type(nla) != IPSET_ATTR_DATA ||
1600 !flag_nested(nla) ||
1601 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1602 set->type->adt_policy))
1603 return -IPSET_ERR_PROTOCOL;
1604 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL,
1614 ip_set_utest(struct sock *ctnl, struct sk_buff *skb,
1615 const struct nlmsghdr *nlh,
1616 const struct nlattr * const attr[])
1618 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1620 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1623 if (unlikely(protocol_failed(attr) ||
1624 !attr[IPSET_ATTR_SETNAME] ||
1625 !attr[IPSET_ATTR_DATA] ||
1626 !flag_nested(attr[IPSET_ATTR_DATA])))
1627 return -IPSET_ERR_PROTOCOL;
1629 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1633 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA],
1634 set->type->adt_policy))
1635 return -IPSET_ERR_PROTOCOL;
1638 ret = set->variant->uadt(set, tb, IPSET_TEST, NULL, 0, 0);
1639 rcu_read_unlock_bh();
1640 /* Userspace can't trigger element to be re-added */
1644 return ret > 0 ? 0 : -IPSET_ERR_EXIST;
1647 /* Get headed data of a set */
1650 ip_set_header(struct sock *ctnl, struct sk_buff *skb,
1651 const struct nlmsghdr *nlh,
1652 const struct nlattr * const attr[])
1654 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1655 const struct ip_set *set;
1656 struct sk_buff *skb2;
1657 struct nlmsghdr *nlh2;
1660 if (unlikely(protocol_failed(attr) ||
1661 !attr[IPSET_ATTR_SETNAME]))
1662 return -IPSET_ERR_PROTOCOL;
1664 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1668 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1672 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1676 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1677 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) ||
1678 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) ||
1679 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
1680 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision))
1681 goto nla_put_failure;
1682 nlmsg_end(skb2, nlh2);
1684 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1691 nlmsg_cancel(skb2, nlh2);
1699 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = {
1700 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1701 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
1702 .len = IPSET_MAXNAMELEN - 1 },
1703 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
1707 ip_set_type(struct sock *ctnl, struct sk_buff *skb,
1708 const struct nlmsghdr *nlh,
1709 const struct nlattr * const attr[])
1711 struct sk_buff *skb2;
1712 struct nlmsghdr *nlh2;
1713 u8 family, min, max;
1714 const char *typename;
1717 if (unlikely(protocol_failed(attr) ||
1718 !attr[IPSET_ATTR_TYPENAME] ||
1719 !attr[IPSET_ATTR_FAMILY]))
1720 return -IPSET_ERR_PROTOCOL;
1722 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1723 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1724 ret = find_set_type_minmax(typename, family, &min, &max);
1728 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1732 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1736 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1737 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) ||
1738 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) ||
1739 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) ||
1740 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min))
1741 goto nla_put_failure;
1742 nlmsg_end(skb2, nlh2);
1744 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len);
1745 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1752 nlmsg_cancel(skb2, nlh2);
1758 /* Get protocol version */
1760 static const struct nla_policy
1761 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = {
1762 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1766 ip_set_protocol(struct sock *ctnl, struct sk_buff *skb,
1767 const struct nlmsghdr *nlh,
1768 const struct nlattr * const attr[])
1770 struct sk_buff *skb2;
1771 struct nlmsghdr *nlh2;
1774 if (unlikely(!attr[IPSET_ATTR_PROTOCOL]))
1775 return -IPSET_ERR_PROTOCOL;
1777 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1781 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1782 IPSET_CMD_PROTOCOL);
1785 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL))
1786 goto nla_put_failure;
1787 nlmsg_end(skb2, nlh2);
1789 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1796 nlmsg_cancel(skb2, nlh2);
1802 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = {
1803 [IPSET_CMD_NONE] = {
1804 .call = ip_set_none,
1805 .attr_count = IPSET_ATTR_CMD_MAX,
1807 [IPSET_CMD_CREATE] = {
1808 .call = ip_set_create,
1809 .attr_count = IPSET_ATTR_CMD_MAX,
1810 .policy = ip_set_create_policy,
1812 [IPSET_CMD_DESTROY] = {
1813 .call = ip_set_destroy,
1814 .attr_count = IPSET_ATTR_CMD_MAX,
1815 .policy = ip_set_setname_policy,
1817 [IPSET_CMD_FLUSH] = {
1818 .call = ip_set_flush,
1819 .attr_count = IPSET_ATTR_CMD_MAX,
1820 .policy = ip_set_setname_policy,
1822 [IPSET_CMD_RENAME] = {
1823 .call = ip_set_rename,
1824 .attr_count = IPSET_ATTR_CMD_MAX,
1825 .policy = ip_set_setname2_policy,
1827 [IPSET_CMD_SWAP] = {
1828 .call = ip_set_swap,
1829 .attr_count = IPSET_ATTR_CMD_MAX,
1830 .policy = ip_set_setname2_policy,
1832 [IPSET_CMD_LIST] = {
1833 .call = ip_set_dump,
1834 .attr_count = IPSET_ATTR_CMD_MAX,
1835 .policy = ip_set_setname_policy,
1837 [IPSET_CMD_SAVE] = {
1838 .call = ip_set_dump,
1839 .attr_count = IPSET_ATTR_CMD_MAX,
1840 .policy = ip_set_setname_policy,
1843 .call = ip_set_uadd,
1844 .attr_count = IPSET_ATTR_CMD_MAX,
1845 .policy = ip_set_adt_policy,
1848 .call = ip_set_udel,
1849 .attr_count = IPSET_ATTR_CMD_MAX,
1850 .policy = ip_set_adt_policy,
1852 [IPSET_CMD_TEST] = {
1853 .call = ip_set_utest,
1854 .attr_count = IPSET_ATTR_CMD_MAX,
1855 .policy = ip_set_adt_policy,
1857 [IPSET_CMD_HEADER] = {
1858 .call = ip_set_header,
1859 .attr_count = IPSET_ATTR_CMD_MAX,
1860 .policy = ip_set_setname_policy,
1862 [IPSET_CMD_TYPE] = {
1863 .call = ip_set_type,
1864 .attr_count = IPSET_ATTR_CMD_MAX,
1865 .policy = ip_set_type_policy,
1867 [IPSET_CMD_PROTOCOL] = {
1868 .call = ip_set_protocol,
1869 .attr_count = IPSET_ATTR_CMD_MAX,
1870 .policy = ip_set_protocol_policy,
1874 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = {
1876 .subsys_id = NFNL_SUBSYS_IPSET,
1877 .cb_count = IPSET_MSG_MAX,
1878 .cb = ip_set_netlink_subsys_cb,
1881 /* Interface to iptables/ip6tables */
1884 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len)
1888 int copylen = *len, ret = 0;
1889 struct net *net = sock_net(sk);
1890 struct ip_set_net *inst = ip_set_pernet(net);
1892 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1894 if (optval != SO_IP_SET)
1896 if (*len < sizeof(unsigned int))
1899 data = vmalloc(*len);
1902 if (copy_from_user(data, user, *len) != 0) {
1906 op = (unsigned int *)data;
1908 if (*op < IP_SET_OP_VERSION) {
1909 /* Check the version at the beginning of operations */
1910 struct ip_set_req_version *req_version = data;
1912 if (*len < sizeof(struct ip_set_req_version)) {
1917 if (req_version->version != IPSET_PROTOCOL) {
1924 case IP_SET_OP_VERSION: {
1925 struct ip_set_req_version *req_version = data;
1927 if (*len != sizeof(struct ip_set_req_version)) {
1932 req_version->version = IPSET_PROTOCOL;
1933 ret = copy_to_user(user, req_version,
1934 sizeof(struct ip_set_req_version));
1937 case IP_SET_OP_GET_BYNAME: {
1938 struct ip_set_req_get_set *req_get = data;
1941 if (*len != sizeof(struct ip_set_req_get_set)) {
1945 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
1946 nfnl_lock(NFNL_SUBSYS_IPSET);
1947 find_set_and_id(inst, req_get->set.name, &id);
1948 req_get->set.index = id;
1949 nfnl_unlock(NFNL_SUBSYS_IPSET);
1952 case IP_SET_OP_GET_FNAME: {
1953 struct ip_set_req_get_set_family *req_get = data;
1956 if (*len != sizeof(struct ip_set_req_get_set_family)) {
1960 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
1961 nfnl_lock(NFNL_SUBSYS_IPSET);
1962 find_set_and_id(inst, req_get->set.name, &id);
1963 req_get->set.index = id;
1964 if (id != IPSET_INVALID_ID)
1965 req_get->family = ip_set(inst, id)->family;
1966 nfnl_unlock(NFNL_SUBSYS_IPSET);
1969 case IP_SET_OP_GET_BYINDEX: {
1970 struct ip_set_req_get_set *req_get = data;
1973 if (*len != sizeof(struct ip_set_req_get_set) ||
1974 req_get->set.index >= inst->ip_set_max) {
1978 nfnl_lock(NFNL_SUBSYS_IPSET);
1979 set = ip_set(inst, req_get->set.index);
1980 strncpy(req_get->set.name, set ? set->name : "",
1982 nfnl_unlock(NFNL_SUBSYS_IPSET);
1988 } /* end of switch(op) */
1991 ret = copy_to_user(user, data, copylen);
2000 static struct nf_sockopt_ops so_set __read_mostly = {
2002 .get_optmin = SO_IP_SET,
2003 .get_optmax = SO_IP_SET + 1,
2004 .get = &ip_set_sockfn_get,
2005 .owner = THIS_MODULE,
2008 static int __net_init
2009 ip_set_net_init(struct net *net)
2011 struct ip_set_net *inst = ip_set_pernet(net);
2012 struct ip_set **list;
2014 inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX;
2015 if (inst->ip_set_max >= IPSET_INVALID_ID)
2016 inst->ip_set_max = IPSET_INVALID_ID - 1;
2018 list = kcalloc(inst->ip_set_max, sizeof(struct ip_set *), GFP_KERNEL);
2021 inst->is_deleted = false;
2022 inst->is_destroyed = false;
2023 rcu_assign_pointer(inst->ip_set_list, list);
2027 static void __net_exit
2028 ip_set_net_exit(struct net *net)
2030 struct ip_set_net *inst = ip_set_pernet(net);
2032 struct ip_set *set = NULL;
2035 inst->is_deleted = true; /* flag for ip_set_nfnl_put */
2037 for (i = 0; i < inst->ip_set_max; i++) {
2038 set = ip_set(inst, i);
2040 ip_set(inst, i) = NULL;
2041 ip_set_destroy_set(set);
2044 kfree(rcu_dereference_protected(inst->ip_set_list, 1));
2047 static struct pernet_operations ip_set_net_ops = {
2048 .init = ip_set_net_init,
2049 .exit = ip_set_net_exit,
2050 .id = &ip_set_net_id,
2051 .size = sizeof(struct ip_set_net)
2057 int ret = nfnetlink_subsys_register(&ip_set_netlink_subsys);
2060 pr_err("ip_set: cannot register with nfnetlink.\n");
2063 ret = nf_register_sockopt(&so_set);
2065 pr_err("SO_SET registry failed: %d\n", ret);
2066 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2069 ret = register_pernet_subsys(&ip_set_net_ops);
2071 pr_err("ip_set: cannot register pernet_subsys.\n");
2072 nf_unregister_sockopt(&so_set);
2073 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2076 pr_info("ip_set: protocol %u\n", IPSET_PROTOCOL);
2083 unregister_pernet_subsys(&ip_set_net_ops);
2084 nf_unregister_sockopt(&so_set);
2085 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2086 pr_debug("these are the famous last words\n");
2089 module_init(ip_set_init);
2090 module_exit(ip_set_fini);