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 if (align < ip_set_extensions[id].align)
383 align = ip_set_extensions[id].align;
384 len = ALIGN(len, ip_set_extensions[id].align);
385 set->offset[id] = len;
386 set->extensions |= ip_set_extensions[id].type;
387 len += ip_set_extensions[id].len;
389 return ALIGN(len, align);
391 EXPORT_SYMBOL_GPL(ip_set_elem_len);
394 ip_set_get_extensions(struct ip_set *set, struct nlattr *tb[],
395 struct ip_set_ext *ext)
399 if (unlikely(!ip_set_optattr_netorder(tb, IPSET_ATTR_TIMEOUT) ||
400 !ip_set_optattr_netorder(tb, IPSET_ATTR_PACKETS) ||
401 !ip_set_optattr_netorder(tb, IPSET_ATTR_BYTES) ||
402 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBMARK) ||
403 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBPRIO) ||
404 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBQUEUE)))
405 return -IPSET_ERR_PROTOCOL;
407 if (tb[IPSET_ATTR_TIMEOUT]) {
408 if (!SET_WITH_TIMEOUT(set))
409 return -IPSET_ERR_TIMEOUT;
410 ext->timeout = ip_set_timeout_uget(tb[IPSET_ATTR_TIMEOUT]);
412 if (tb[IPSET_ATTR_BYTES] || tb[IPSET_ATTR_PACKETS]) {
413 if (!SET_WITH_COUNTER(set))
414 return -IPSET_ERR_COUNTER;
415 if (tb[IPSET_ATTR_BYTES])
416 ext->bytes = be64_to_cpu(nla_get_be64(
417 tb[IPSET_ATTR_BYTES]));
418 if (tb[IPSET_ATTR_PACKETS])
419 ext->packets = be64_to_cpu(nla_get_be64(
420 tb[IPSET_ATTR_PACKETS]));
422 if (tb[IPSET_ATTR_COMMENT]) {
423 if (!SET_WITH_COMMENT(set))
424 return -IPSET_ERR_COMMENT;
425 ext->comment = ip_set_comment_uget(tb[IPSET_ATTR_COMMENT]);
427 if (tb[IPSET_ATTR_SKBMARK]) {
428 if (!SET_WITH_SKBINFO(set))
429 return -IPSET_ERR_SKBINFO;
430 fullmark = be64_to_cpu(nla_get_be64(tb[IPSET_ATTR_SKBMARK]));
431 ext->skbmark = fullmark >> 32;
432 ext->skbmarkmask = fullmark & 0xffffffff;
434 if (tb[IPSET_ATTR_SKBPRIO]) {
435 if (!SET_WITH_SKBINFO(set))
436 return -IPSET_ERR_SKBINFO;
437 ext->skbprio = be32_to_cpu(nla_get_be32(
438 tb[IPSET_ATTR_SKBPRIO]));
440 if (tb[IPSET_ATTR_SKBQUEUE]) {
441 if (!SET_WITH_SKBINFO(set))
442 return -IPSET_ERR_SKBINFO;
443 ext->skbqueue = be16_to_cpu(nla_get_be16(
444 tb[IPSET_ATTR_SKBQUEUE]));
448 EXPORT_SYMBOL_GPL(ip_set_get_extensions);
451 ip_set_put_extensions(struct sk_buff *skb, const struct ip_set *set,
452 const void *e, bool active)
454 if (SET_WITH_TIMEOUT(set)) {
455 unsigned long *timeout = ext_timeout(e, set);
457 if (nla_put_net32(skb, IPSET_ATTR_TIMEOUT,
458 htonl(active ? ip_set_timeout_get(timeout)
462 if (SET_WITH_COUNTER(set) &&
463 ip_set_put_counter(skb, ext_counter(e, set)))
465 if (SET_WITH_COMMENT(set) &&
466 ip_set_put_comment(skb, ext_comment(e, set)))
468 if (SET_WITH_SKBINFO(set) &&
469 ip_set_put_skbinfo(skb, ext_skbinfo(e, set)))
473 EXPORT_SYMBOL_GPL(ip_set_put_extensions);
475 /* Creating/destroying/renaming/swapping affect the existence and
476 * the properties of a set. All of these can be executed from userspace
477 * only and serialized by the nfnl mutex indirectly from nfnetlink.
479 * Sets are identified by their index in ip_set_list and the index
480 * is used by the external references (set/SET netfilter modules).
482 * The set behind an index may change by swapping only, from userspace.
486 __ip_set_get(struct ip_set *set)
488 write_lock_bh(&ip_set_ref_lock);
490 write_unlock_bh(&ip_set_ref_lock);
494 __ip_set_put(struct ip_set *set)
496 write_lock_bh(&ip_set_ref_lock);
497 BUG_ON(set->ref == 0);
499 write_unlock_bh(&ip_set_ref_lock);
502 /* Add, del and test set entries from kernel.
504 * The set behind the index must exist and must be referenced
505 * so it can't be destroyed (or changed) under our foot.
508 static inline struct ip_set *
509 ip_set_rcu_get(struct net *net, ip_set_id_t index)
512 struct ip_set_net *inst = ip_set_pernet(net);
515 /* ip_set_list itself needs to be protected */
516 set = rcu_dereference(inst->ip_set_list)[index];
523 ip_set_test(ip_set_id_t index, const struct sk_buff *skb,
524 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
526 struct ip_set *set = ip_set_rcu_get(par->net, index);
530 pr_debug("set %s, index %u\n", set->name, index);
532 if (opt->dim < set->type->dimension ||
533 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
537 ret = set->variant->kadt(set, skb, par, IPSET_TEST, opt);
538 rcu_read_unlock_bh();
540 if (ret == -EAGAIN) {
541 /* Type requests element to be completed */
542 pr_debug("element must be completed, ADD is triggered\n");
543 spin_lock_bh(&set->lock);
544 set->variant->kadt(set, skb, par, IPSET_ADD, opt);
545 spin_unlock_bh(&set->lock);
548 /* --return-nomatch: invert matched element */
549 if ((opt->cmdflags & IPSET_FLAG_RETURN_NOMATCH) &&
550 (set->type->features & IPSET_TYPE_NOMATCH) &&
551 (ret > 0 || ret == -ENOTEMPTY))
555 /* Convert error codes to nomatch */
556 return (ret < 0 ? 0 : ret);
558 EXPORT_SYMBOL_GPL(ip_set_test);
561 ip_set_add(ip_set_id_t index, const struct sk_buff *skb,
562 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
564 struct ip_set *set = ip_set_rcu_get(par->net, index);
568 pr_debug("set %s, index %u\n", set->name, index);
570 if (opt->dim < set->type->dimension ||
571 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
572 return -IPSET_ERR_TYPE_MISMATCH;
574 spin_lock_bh(&set->lock);
575 ret = set->variant->kadt(set, skb, par, IPSET_ADD, opt);
576 spin_unlock_bh(&set->lock);
580 EXPORT_SYMBOL_GPL(ip_set_add);
583 ip_set_del(ip_set_id_t index, const struct sk_buff *skb,
584 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
586 struct ip_set *set = ip_set_rcu_get(par->net, index);
590 pr_debug("set %s, index %u\n", set->name, index);
592 if (opt->dim < set->type->dimension ||
593 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
594 return -IPSET_ERR_TYPE_MISMATCH;
596 spin_lock_bh(&set->lock);
597 ret = set->variant->kadt(set, skb, par, IPSET_DEL, opt);
598 spin_unlock_bh(&set->lock);
602 EXPORT_SYMBOL_GPL(ip_set_del);
604 /* Find set by name, reference it once. The reference makes sure the
605 * thing pointed to, does not go away under our feet.
609 ip_set_get_byname(struct net *net, const char *name, struct ip_set **set)
611 ip_set_id_t i, index = IPSET_INVALID_ID;
613 struct ip_set_net *inst = ip_set_pernet(net);
616 for (i = 0; i < inst->ip_set_max; i++) {
617 s = rcu_dereference(inst->ip_set_list)[i];
618 if (s && STRNCMP(s->name, name)) {
629 EXPORT_SYMBOL_GPL(ip_set_get_byname);
631 /* If the given set pointer points to a valid set, decrement
632 * reference count by 1. The caller shall not assume the index
633 * to be valid, after calling this function.
638 __ip_set_put_byindex(struct ip_set_net *inst, ip_set_id_t index)
643 set = rcu_dereference(inst->ip_set_list)[index];
650 ip_set_put_byindex(struct net *net, ip_set_id_t index)
652 struct ip_set_net *inst = ip_set_pernet(net);
654 __ip_set_put_byindex(inst, index);
656 EXPORT_SYMBOL_GPL(ip_set_put_byindex);
658 /* Get the name of a set behind a set index.
659 * We assume the set is referenced, so it does exist and
660 * can't be destroyed. The set cannot be renamed due to
661 * the referencing either.
665 ip_set_name_byindex(struct net *net, ip_set_id_t index)
667 const struct ip_set *set = ip_set_rcu_get(net, index);
670 BUG_ON(set->ref == 0);
672 /* Referenced, so it's safe */
675 EXPORT_SYMBOL_GPL(ip_set_name_byindex);
677 /* Routines to call by external subsystems, which do not
678 * call nfnl_lock for us.
681 /* Find set by index, reference it once. The reference makes sure the
682 * thing pointed to, does not go away under our feet.
684 * The nfnl mutex is used in the function.
687 ip_set_nfnl_get_byindex(struct net *net, ip_set_id_t index)
690 struct ip_set_net *inst = ip_set_pernet(net);
692 if (index >= inst->ip_set_max)
693 return IPSET_INVALID_ID;
695 nfnl_lock(NFNL_SUBSYS_IPSET);
696 set = ip_set(inst, index);
700 index = IPSET_INVALID_ID;
701 nfnl_unlock(NFNL_SUBSYS_IPSET);
705 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex);
707 /* If the given set pointer points to a valid set, decrement
708 * reference count by 1. The caller shall not assume the index
709 * to be valid, after calling this function.
711 * The nfnl mutex is used in the function.
714 ip_set_nfnl_put(struct net *net, ip_set_id_t index)
717 struct ip_set_net *inst = ip_set_pernet(net);
719 nfnl_lock(NFNL_SUBSYS_IPSET);
720 if (!inst->is_deleted) { /* already deleted from ip_set_net_exit() */
721 set = ip_set(inst, index);
725 nfnl_unlock(NFNL_SUBSYS_IPSET);
727 EXPORT_SYMBOL_GPL(ip_set_nfnl_put);
729 /* Communication protocol with userspace over netlink.
731 * The commands are serialized by the nfnl mutex.
735 protocol_failed(const struct nlattr * const tb[])
737 return !tb[IPSET_ATTR_PROTOCOL] ||
738 nla_get_u8(tb[IPSET_ATTR_PROTOCOL]) != IPSET_PROTOCOL;
742 flag_exist(const struct nlmsghdr *nlh)
744 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST;
747 static struct nlmsghdr *
748 start_msg(struct sk_buff *skb, u32 portid, u32 seq, unsigned int flags,
751 struct nlmsghdr *nlh;
752 struct nfgenmsg *nfmsg;
754 nlh = nlmsg_put(skb, portid, seq, cmd | (NFNL_SUBSYS_IPSET << 8),
755 sizeof(*nfmsg), flags);
759 nfmsg = nlmsg_data(nlh);
760 nfmsg->nfgen_family = NFPROTO_IPV4;
761 nfmsg->version = NFNETLINK_V0;
769 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = {
770 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
771 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
772 .len = IPSET_MAXNAMELEN - 1 },
773 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
774 .len = IPSET_MAXNAMELEN - 1},
775 [IPSET_ATTR_REVISION] = { .type = NLA_U8 },
776 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
777 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
780 static struct ip_set *
781 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id)
783 struct ip_set *set = NULL;
786 *id = IPSET_INVALID_ID;
787 for (i = 0; i < inst->ip_set_max; i++) {
788 set = ip_set(inst, i);
789 if (set && STRNCMP(set->name, name)) {
794 return (*id == IPSET_INVALID_ID ? NULL : set);
797 static inline struct ip_set *
798 find_set(struct ip_set_net *inst, const char *name)
802 return find_set_and_id(inst, name, &id);
806 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index,
812 *index = IPSET_INVALID_ID;
813 for (i = 0; i < inst->ip_set_max; i++) {
816 if (*index == IPSET_INVALID_ID)
818 } else if (STRNCMP(name, s->name)) {
824 if (*index == IPSET_INVALID_ID)
825 /* No free slot remained */
826 return -IPSET_ERR_MAX_SETS;
831 ip_set_none(struct sock *ctnl, struct sk_buff *skb,
832 const struct nlmsghdr *nlh,
833 const struct nlattr * const attr[])
839 ip_set_create(struct sock *ctnl, struct sk_buff *skb,
840 const struct nlmsghdr *nlh,
841 const struct nlattr * const attr[])
843 struct net *net = sock_net(ctnl);
844 struct ip_set_net *inst = ip_set_pernet(net);
845 struct ip_set *set, *clash = NULL;
846 ip_set_id_t index = IPSET_INVALID_ID;
847 struct nlattr *tb[IPSET_ATTR_CREATE_MAX + 1] = {};
848 const char *name, *typename;
850 u32 flags = flag_exist(nlh);
853 if (unlikely(protocol_failed(attr) ||
854 !attr[IPSET_ATTR_SETNAME] ||
855 !attr[IPSET_ATTR_TYPENAME] ||
856 !attr[IPSET_ATTR_REVISION] ||
857 !attr[IPSET_ATTR_FAMILY] ||
858 (attr[IPSET_ATTR_DATA] &&
859 !flag_nested(attr[IPSET_ATTR_DATA]))))
860 return -IPSET_ERR_PROTOCOL;
862 name = nla_data(attr[IPSET_ATTR_SETNAME]);
863 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
864 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
865 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]);
866 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
867 name, typename, family_name(family), revision);
869 /* First, and without any locks, allocate and initialize
870 * a normal base set structure.
872 set = kzalloc(sizeof(*set), GFP_KERNEL);
875 spin_lock_init(&set->lock);
876 strlcpy(set->name, name, IPSET_MAXNAMELEN);
877 set->family = family;
878 set->revision = revision;
880 /* Next, check that we know the type, and take
881 * a reference on the type, to make sure it stays available
882 * while constructing our new set.
884 * After referencing the type, we try to create the type
885 * specific part of the set without holding any locks.
887 ret = find_set_type_get(typename, family, revision, &set->type);
891 /* Without holding any locks, create private part. */
892 if (attr[IPSET_ATTR_DATA] &&
893 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA],
894 set->type->create_policy)) {
895 ret = -IPSET_ERR_PROTOCOL;
899 ret = set->type->create(net, set, tb, flags);
903 /* BTW, ret==0 here. */
905 /* Here, we have a valid, constructed set and we are protected
906 * by the nfnl mutex. Find the first free index in ip_set_list
907 * and check clashing.
909 ret = find_free_id(inst, set->name, &index, &clash);
910 if (ret == -EEXIST) {
911 /* If this is the same set and requested, ignore error */
912 if ((flags & IPSET_FLAG_EXIST) &&
913 STRNCMP(set->type->name, clash->type->name) &&
914 set->type->family == clash->type->family &&
915 set->type->revision_min == clash->type->revision_min &&
916 set->type->revision_max == clash->type->revision_max &&
917 set->variant->same_set(set, clash))
920 } else if (ret == -IPSET_ERR_MAX_SETS) {
921 struct ip_set **list, **tmp;
922 ip_set_id_t i = inst->ip_set_max + IP_SET_INC;
924 if (i < inst->ip_set_max || i == IPSET_INVALID_ID)
928 list = kcalloc(i, sizeof(struct ip_set *), GFP_KERNEL);
931 /* nfnl mutex is held, both lists are valid */
932 tmp = ip_set_dereference(inst->ip_set_list);
933 memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max);
934 rcu_assign_pointer(inst->ip_set_list, list);
935 /* Make sure all current packets have passed through */
938 index = inst->ip_set_max;
939 inst->ip_set_max = i;
946 /* Finally! Add our shiny new set to the list, and be done. */
947 pr_debug("create: '%s' created with index %u!\n", set->name, index);
948 ip_set(inst, index) = set;
953 set->variant->destroy(set);
955 module_put(set->type->me);
963 static const struct nla_policy
964 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = {
965 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
966 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
967 .len = IPSET_MAXNAMELEN - 1 },
971 ip_set_destroy_set(struct ip_set *set)
973 pr_debug("set: %s\n", set->name);
975 /* Must call it without holding any lock */
976 set->variant->destroy(set);
977 module_put(set->type->me);
982 ip_set_destroy(struct sock *ctnl, struct sk_buff *skb,
983 const struct nlmsghdr *nlh,
984 const struct nlattr * const attr[])
986 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
991 if (unlikely(protocol_failed(attr)))
992 return -IPSET_ERR_PROTOCOL;
994 /* Commands are serialized and references are
995 * protected by the ip_set_ref_lock.
996 * External systems (i.e. xt_set) must call
997 * ip_set_put|get_nfnl_* functions, that way we
998 * can safely check references here.
1000 * list:set timer can only decrement the reference
1001 * counter, so if it's already zero, we can proceed
1002 * without holding the lock.
1004 read_lock_bh(&ip_set_ref_lock);
1005 if (!attr[IPSET_ATTR_SETNAME]) {
1006 for (i = 0; i < inst->ip_set_max; i++) {
1007 s = ip_set(inst, i);
1009 ret = -IPSET_ERR_BUSY;
1013 inst->is_destroyed = true;
1014 read_unlock_bh(&ip_set_ref_lock);
1015 for (i = 0; i < inst->ip_set_max; i++) {
1016 s = ip_set(inst, i);
1018 ip_set(inst, i) = NULL;
1019 ip_set_destroy_set(s);
1022 /* Modified by ip_set_destroy() only, which is serialized */
1023 inst->is_destroyed = false;
1025 s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1030 } else if (s->ref) {
1031 ret = -IPSET_ERR_BUSY;
1034 ip_set(inst, i) = NULL;
1035 read_unlock_bh(&ip_set_ref_lock);
1037 ip_set_destroy_set(s);
1041 read_unlock_bh(&ip_set_ref_lock);
1048 ip_set_flush_set(struct ip_set *set)
1050 pr_debug("set: %s\n", set->name);
1052 spin_lock_bh(&set->lock);
1053 set->variant->flush(set);
1054 spin_unlock_bh(&set->lock);
1058 ip_set_flush(struct sock *ctnl, struct sk_buff *skb,
1059 const struct nlmsghdr *nlh,
1060 const struct nlattr * const attr[])
1062 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1066 if (unlikely(protocol_failed(attr)))
1067 return -IPSET_ERR_PROTOCOL;
1069 if (!attr[IPSET_ATTR_SETNAME]) {
1070 for (i = 0; i < inst->ip_set_max; i++) {
1071 s = ip_set(inst, i);
1073 ip_set_flush_set(s);
1076 s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1080 ip_set_flush_set(s);
1088 static const struct nla_policy
1089 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = {
1090 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1091 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1092 .len = IPSET_MAXNAMELEN - 1 },
1093 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING,
1094 .len = IPSET_MAXNAMELEN - 1 },
1098 ip_set_rename(struct sock *ctnl, struct sk_buff *skb,
1099 const struct nlmsghdr *nlh,
1100 const struct nlattr * const attr[])
1102 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1103 struct ip_set *set, *s;
1108 if (unlikely(protocol_failed(attr) ||
1109 !attr[IPSET_ATTR_SETNAME] ||
1110 !attr[IPSET_ATTR_SETNAME2]))
1111 return -IPSET_ERR_PROTOCOL;
1113 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1117 read_lock_bh(&ip_set_ref_lock);
1118 if (set->ref != 0) {
1119 ret = -IPSET_ERR_REFERENCED;
1123 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]);
1124 for (i = 0; i < inst->ip_set_max; i++) {
1125 s = ip_set(inst, i);
1126 if (s && STRNCMP(s->name, name2)) {
1127 ret = -IPSET_ERR_EXIST_SETNAME2;
1131 strncpy(set->name, name2, IPSET_MAXNAMELEN);
1134 read_unlock_bh(&ip_set_ref_lock);
1138 /* Swap two sets so that name/index points to the other.
1139 * References and set names are also swapped.
1141 * The commands are serialized by the nfnl mutex and references are
1142 * protected by the ip_set_ref_lock. The kernel interfaces
1143 * do not hold the mutex but the pointer settings are atomic
1144 * so the ip_set_list always contains valid pointers to the sets.
1148 ip_set_swap(struct sock *ctnl, struct sk_buff *skb,
1149 const struct nlmsghdr *nlh,
1150 const struct nlattr * const attr[])
1152 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1153 struct ip_set *from, *to;
1154 ip_set_id_t from_id, to_id;
1155 char from_name[IPSET_MAXNAMELEN];
1157 if (unlikely(protocol_failed(attr) ||
1158 !attr[IPSET_ATTR_SETNAME] ||
1159 !attr[IPSET_ATTR_SETNAME2]))
1160 return -IPSET_ERR_PROTOCOL;
1162 from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1167 to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]),
1170 return -IPSET_ERR_EXIST_SETNAME2;
1172 /* Features must not change.
1173 * Not an artifical restriction anymore, as we must prevent
1174 * possible loops created by swapping in setlist type of sets.
1176 if (!(from->type->features == to->type->features &&
1177 from->family == to->family))
1178 return -IPSET_ERR_TYPE_MISMATCH;
1180 strncpy(from_name, from->name, IPSET_MAXNAMELEN);
1181 strncpy(from->name, to->name, IPSET_MAXNAMELEN);
1182 strncpy(to->name, from_name, IPSET_MAXNAMELEN);
1184 write_lock_bh(&ip_set_ref_lock);
1185 swap(from->ref, to->ref);
1186 ip_set(inst, from_id) = to;
1187 ip_set(inst, to_id) = from;
1188 write_unlock_bh(&ip_set_ref_lock);
1193 /* List/save set data */
1200 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1201 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1204 ip_set_dump_done(struct netlink_callback *cb)
1206 if (cb->args[IPSET_CB_ARG0]) {
1207 struct ip_set_net *inst =
1208 (struct ip_set_net *)cb->args[IPSET_CB_NET];
1209 ip_set_id_t index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1210 struct ip_set *set = ip_set(inst, index);
1212 if (set->variant->uref)
1213 set->variant->uref(set, cb, false);
1214 pr_debug("release set %s\n", set->name);
1215 __ip_set_put_byindex(inst, index);
1221 dump_attrs(struct nlmsghdr *nlh)
1223 const struct nlattr *attr;
1226 pr_debug("dump nlmsg\n");
1227 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) {
1228 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len);
1233 dump_init(struct netlink_callback *cb, struct ip_set_net *inst)
1235 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb);
1236 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1237 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1238 struct nlattr *attr = (void *)nlh + min_len;
1242 /* Second pass, so parser can't fail */
1243 nla_parse(cda, IPSET_ATTR_CMD_MAX,
1244 attr, nlh->nlmsg_len - min_len, ip_set_setname_policy);
1246 if (cda[IPSET_ATTR_SETNAME]) {
1249 set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]),
1254 dump_type = DUMP_ONE;
1255 cb->args[IPSET_CB_INDEX] = index;
1257 dump_type = DUMP_ALL;
1260 if (cda[IPSET_ATTR_FLAGS]) {
1261 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]);
1263 dump_type |= (f << 16);
1265 cb->args[IPSET_CB_NET] = (unsigned long)inst;
1266 cb->args[IPSET_CB_DUMP] = dump_type;
1272 ip_set_dump_start(struct sk_buff *skb, struct netlink_callback *cb)
1274 ip_set_id_t index = IPSET_INVALID_ID, max;
1275 struct ip_set *set = NULL;
1276 struct nlmsghdr *nlh = NULL;
1277 unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0;
1278 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1279 u32 dump_type, dump_flags;
1283 if (!cb->args[IPSET_CB_DUMP]) {
1284 ret = dump_init(cb, inst);
1286 nlh = nlmsg_hdr(cb->skb);
1287 /* We have to create and send the error message
1290 if (nlh->nlmsg_flags & NLM_F_ACK)
1291 netlink_ack(cb->skb, nlh, ret);
1296 if (cb->args[IPSET_CB_INDEX] >= inst->ip_set_max)
1299 dump_type = DUMP_TYPE(cb->args[IPSET_CB_DUMP]);
1300 dump_flags = DUMP_FLAGS(cb->args[IPSET_CB_DUMP]);
1301 max = dump_type == DUMP_ONE ? cb->args[IPSET_CB_INDEX] + 1
1304 pr_debug("dump type, flag: %u %u index: %ld\n",
1305 dump_type, dump_flags, cb->args[IPSET_CB_INDEX]);
1306 for (; cb->args[IPSET_CB_INDEX] < max; cb->args[IPSET_CB_INDEX]++) {
1307 index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1308 write_lock_bh(&ip_set_ref_lock);
1309 set = ip_set(inst, index);
1310 is_destroyed = inst->is_destroyed;
1311 if (!set || is_destroyed) {
1312 write_unlock_bh(&ip_set_ref_lock);
1313 if (dump_type == DUMP_ONE) {
1318 /* All sets are just being destroyed */
1324 /* When dumping all sets, we must dump "sorted"
1325 * so that lists (unions of sets) are dumped last.
1327 if (dump_type != DUMP_ONE &&
1328 ((dump_type == DUMP_ALL) ==
1329 !!(set->type->features & IPSET_DUMP_LAST))) {
1330 write_unlock_bh(&ip_set_ref_lock);
1333 pr_debug("List set: %s\n", set->name);
1334 if (!cb->args[IPSET_CB_ARG0]) {
1335 /* Start listing: make sure set won't be destroyed */
1336 pr_debug("reference set\n");
1339 write_unlock_bh(&ip_set_ref_lock);
1340 nlh = start_msg(skb, NETLINK_CB(cb->skb).portid,
1341 cb->nlh->nlmsg_seq, flags,
1345 goto release_refcount;
1347 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1348 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name))
1349 goto nla_put_failure;
1350 if (dump_flags & IPSET_FLAG_LIST_SETNAME)
1352 switch (cb->args[IPSET_CB_ARG0]) {
1354 /* Core header data */
1355 if (nla_put_string(skb, IPSET_ATTR_TYPENAME,
1357 nla_put_u8(skb, IPSET_ATTR_FAMILY,
1359 nla_put_u8(skb, IPSET_ATTR_REVISION,
1361 goto nla_put_failure;
1362 ret = set->variant->head(set, skb);
1364 goto release_refcount;
1365 if (dump_flags & IPSET_FLAG_LIST_HEADER)
1367 if (set->variant->uref)
1368 set->variant->uref(set, cb, true);
1369 /* Fall through and add elements */
1372 ret = set->variant->list(set, skb, cb);
1373 rcu_read_unlock_bh();
1374 if (!cb->args[IPSET_CB_ARG0])
1375 /* Set is done, proceed with next one */
1377 goto release_refcount;
1380 /* If we dump all sets, continue with dumping last ones */
1381 if (dump_type == DUMP_ALL) {
1382 dump_type = DUMP_LAST;
1383 cb->args[IPSET_CB_DUMP] = dump_type | (dump_flags << 16);
1384 cb->args[IPSET_CB_INDEX] = 0;
1385 if (set && set->variant->uref)
1386 set->variant->uref(set, cb, false);
1394 if (dump_type == DUMP_ONE)
1395 cb->args[IPSET_CB_INDEX] = IPSET_INVALID_ID;
1397 cb->args[IPSET_CB_INDEX]++;
1399 /* If there was an error or set is done, release set */
1400 if (ret || !cb->args[IPSET_CB_ARG0]) {
1401 set = ip_set(inst, index);
1402 if (set->variant->uref)
1403 set->variant->uref(set, cb, false);
1404 pr_debug("release set %s\n", set->name);
1405 __ip_set_put_byindex(inst, index);
1406 cb->args[IPSET_CB_ARG0] = 0;
1410 nlmsg_end(skb, nlh);
1411 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len);
1415 return ret < 0 ? ret : skb->len;
1419 ip_set_dump(struct sock *ctnl, struct sk_buff *skb,
1420 const struct nlmsghdr *nlh,
1421 const struct nlattr * const attr[])
1423 if (unlikely(protocol_failed(attr)))
1424 return -IPSET_ERR_PROTOCOL;
1427 struct netlink_dump_control c = {
1428 .dump = ip_set_dump_start,
1429 .done = ip_set_dump_done,
1431 return netlink_dump_start(ctnl, skb, nlh, &c);
1435 /* Add, del and test */
1437 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = {
1438 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1439 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1440 .len = IPSET_MAXNAMELEN - 1 },
1441 [IPSET_ATTR_LINENO] = { .type = NLA_U32 },
1442 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
1443 [IPSET_ATTR_ADT] = { .type = NLA_NESTED },
1447 call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set,
1448 struct nlattr *tb[], enum ipset_adt adt,
1449 u32 flags, bool use_lineno)
1453 bool eexist = flags & IPSET_FLAG_EXIST, retried = false;
1456 spin_lock_bh(&set->lock);
1457 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried);
1458 spin_unlock_bh(&set->lock);
1460 } while (ret == -EAGAIN &&
1461 set->variant->resize &&
1462 (ret = set->variant->resize(set, retried)) == 0);
1464 if (!ret || (ret == -IPSET_ERR_EXIST && eexist))
1466 if (lineno && use_lineno) {
1467 /* Error in restore/batch mode: send back lineno */
1468 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb);
1469 struct sk_buff *skb2;
1470 struct nlmsgerr *errmsg;
1471 size_t payload = min(SIZE_MAX,
1472 sizeof(*errmsg) + nlmsg_len(nlh));
1473 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1474 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1475 struct nlattr *cmdattr;
1478 skb2 = nlmsg_new(payload, GFP_KERNEL);
1481 rep = __nlmsg_put(skb2, NETLINK_CB(skb).portid,
1482 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0);
1483 errmsg = nlmsg_data(rep);
1484 errmsg->error = ret;
1485 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len);
1486 cmdattr = (void *)&errmsg->msg + min_len;
1488 nla_parse(cda, IPSET_ATTR_CMD_MAX,
1489 cmdattr, nlh->nlmsg_len - min_len,
1492 errline = nla_data(cda[IPSET_ATTR_LINENO]);
1496 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid,
1498 /* Signal netlink not to send its ACK/errmsg. */
1506 ip_set_uadd(struct sock *ctnl, struct sk_buff *skb,
1507 const struct nlmsghdr *nlh,
1508 const struct nlattr * const attr[])
1510 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1512 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1513 const struct nlattr *nla;
1514 u32 flags = flag_exist(nlh);
1518 if (unlikely(protocol_failed(attr) ||
1519 !attr[IPSET_ATTR_SETNAME] ||
1520 !((attr[IPSET_ATTR_DATA] != NULL) ^
1521 (attr[IPSET_ATTR_ADT] != NULL)) ||
1522 (attr[IPSET_ATTR_DATA] &&
1523 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1524 (attr[IPSET_ATTR_ADT] &&
1525 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1526 !attr[IPSET_ATTR_LINENO]))))
1527 return -IPSET_ERR_PROTOCOL;
1529 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1533 use_lineno = !!attr[IPSET_ATTR_LINENO];
1534 if (attr[IPSET_ATTR_DATA]) {
1535 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1536 attr[IPSET_ATTR_DATA],
1537 set->type->adt_policy))
1538 return -IPSET_ERR_PROTOCOL;
1539 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD, flags,
1544 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1545 memset(tb, 0, sizeof(tb));
1546 if (nla_type(nla) != IPSET_ATTR_DATA ||
1547 !flag_nested(nla) ||
1548 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1549 set->type->adt_policy))
1550 return -IPSET_ERR_PROTOCOL;
1551 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD,
1561 ip_set_udel(struct sock *ctnl, struct sk_buff *skb,
1562 const struct nlmsghdr *nlh,
1563 const struct nlattr * const attr[])
1565 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1567 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1568 const struct nlattr *nla;
1569 u32 flags = flag_exist(nlh);
1573 if (unlikely(protocol_failed(attr) ||
1574 !attr[IPSET_ATTR_SETNAME] ||
1575 !((attr[IPSET_ATTR_DATA] != NULL) ^
1576 (attr[IPSET_ATTR_ADT] != NULL)) ||
1577 (attr[IPSET_ATTR_DATA] &&
1578 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1579 (attr[IPSET_ATTR_ADT] &&
1580 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1581 !attr[IPSET_ATTR_LINENO]))))
1582 return -IPSET_ERR_PROTOCOL;
1584 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1588 use_lineno = !!attr[IPSET_ATTR_LINENO];
1589 if (attr[IPSET_ATTR_DATA]) {
1590 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1591 attr[IPSET_ATTR_DATA],
1592 set->type->adt_policy))
1593 return -IPSET_ERR_PROTOCOL;
1594 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL, flags,
1599 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1600 memset(tb, 0, sizeof(*tb));
1601 if (nla_type(nla) != IPSET_ATTR_DATA ||
1602 !flag_nested(nla) ||
1603 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1604 set->type->adt_policy))
1605 return -IPSET_ERR_PROTOCOL;
1606 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL,
1616 ip_set_utest(struct sock *ctnl, struct sk_buff *skb,
1617 const struct nlmsghdr *nlh,
1618 const struct nlattr * const attr[])
1620 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1622 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1626 if (unlikely(protocol_failed(attr) ||
1627 !attr[IPSET_ATTR_SETNAME] ||
1628 !attr[IPSET_ATTR_DATA] ||
1629 !flag_nested(attr[IPSET_ATTR_DATA])))
1630 return -IPSET_ERR_PROTOCOL;
1632 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1636 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA],
1637 set->type->adt_policy))
1638 return -IPSET_ERR_PROTOCOL;
1641 ret = set->variant->uadt(set, tb, IPSET_TEST, &lineno, 0, 0);
1642 rcu_read_unlock_bh();
1643 /* Userspace can't trigger element to be re-added */
1647 return ret > 0 ? 0 : -IPSET_ERR_EXIST;
1650 /* Get headed data of a set */
1653 ip_set_header(struct sock *ctnl, struct sk_buff *skb,
1654 const struct nlmsghdr *nlh,
1655 const struct nlattr * const attr[])
1657 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1658 const struct ip_set *set;
1659 struct sk_buff *skb2;
1660 struct nlmsghdr *nlh2;
1663 if (unlikely(protocol_failed(attr) ||
1664 !attr[IPSET_ATTR_SETNAME]))
1665 return -IPSET_ERR_PROTOCOL;
1667 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1671 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1675 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1679 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1680 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) ||
1681 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) ||
1682 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
1683 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision))
1684 goto nla_put_failure;
1685 nlmsg_end(skb2, nlh2);
1687 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1694 nlmsg_cancel(skb2, nlh2);
1702 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = {
1703 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1704 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
1705 .len = IPSET_MAXNAMELEN - 1 },
1706 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
1710 ip_set_type(struct sock *ctnl, struct sk_buff *skb,
1711 const struct nlmsghdr *nlh,
1712 const struct nlattr * const attr[])
1714 struct sk_buff *skb2;
1715 struct nlmsghdr *nlh2;
1716 u8 family, min, max;
1717 const char *typename;
1720 if (unlikely(protocol_failed(attr) ||
1721 !attr[IPSET_ATTR_TYPENAME] ||
1722 !attr[IPSET_ATTR_FAMILY]))
1723 return -IPSET_ERR_PROTOCOL;
1725 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1726 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1727 ret = find_set_type_minmax(typename, family, &min, &max);
1731 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1735 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1739 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1740 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) ||
1741 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) ||
1742 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) ||
1743 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min))
1744 goto nla_put_failure;
1745 nlmsg_end(skb2, nlh2);
1747 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len);
1748 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1755 nlmsg_cancel(skb2, nlh2);
1761 /* Get protocol version */
1763 static const struct nla_policy
1764 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = {
1765 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1769 ip_set_protocol(struct sock *ctnl, struct sk_buff *skb,
1770 const struct nlmsghdr *nlh,
1771 const struct nlattr * const attr[])
1773 struct sk_buff *skb2;
1774 struct nlmsghdr *nlh2;
1777 if (unlikely(!attr[IPSET_ATTR_PROTOCOL]))
1778 return -IPSET_ERR_PROTOCOL;
1780 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1784 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1785 IPSET_CMD_PROTOCOL);
1788 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL))
1789 goto nla_put_failure;
1790 nlmsg_end(skb2, nlh2);
1792 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1799 nlmsg_cancel(skb2, nlh2);
1805 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = {
1806 [IPSET_CMD_NONE] = {
1807 .call = ip_set_none,
1808 .attr_count = IPSET_ATTR_CMD_MAX,
1810 [IPSET_CMD_CREATE] = {
1811 .call = ip_set_create,
1812 .attr_count = IPSET_ATTR_CMD_MAX,
1813 .policy = ip_set_create_policy,
1815 [IPSET_CMD_DESTROY] = {
1816 .call = ip_set_destroy,
1817 .attr_count = IPSET_ATTR_CMD_MAX,
1818 .policy = ip_set_setname_policy,
1820 [IPSET_CMD_FLUSH] = {
1821 .call = ip_set_flush,
1822 .attr_count = IPSET_ATTR_CMD_MAX,
1823 .policy = ip_set_setname_policy,
1825 [IPSET_CMD_RENAME] = {
1826 .call = ip_set_rename,
1827 .attr_count = IPSET_ATTR_CMD_MAX,
1828 .policy = ip_set_setname2_policy,
1830 [IPSET_CMD_SWAP] = {
1831 .call = ip_set_swap,
1832 .attr_count = IPSET_ATTR_CMD_MAX,
1833 .policy = ip_set_setname2_policy,
1835 [IPSET_CMD_LIST] = {
1836 .call = ip_set_dump,
1837 .attr_count = IPSET_ATTR_CMD_MAX,
1838 .policy = ip_set_setname_policy,
1840 [IPSET_CMD_SAVE] = {
1841 .call = ip_set_dump,
1842 .attr_count = IPSET_ATTR_CMD_MAX,
1843 .policy = ip_set_setname_policy,
1846 .call = ip_set_uadd,
1847 .attr_count = IPSET_ATTR_CMD_MAX,
1848 .policy = ip_set_adt_policy,
1851 .call = ip_set_udel,
1852 .attr_count = IPSET_ATTR_CMD_MAX,
1853 .policy = ip_set_adt_policy,
1855 [IPSET_CMD_TEST] = {
1856 .call = ip_set_utest,
1857 .attr_count = IPSET_ATTR_CMD_MAX,
1858 .policy = ip_set_adt_policy,
1860 [IPSET_CMD_HEADER] = {
1861 .call = ip_set_header,
1862 .attr_count = IPSET_ATTR_CMD_MAX,
1863 .policy = ip_set_setname_policy,
1865 [IPSET_CMD_TYPE] = {
1866 .call = ip_set_type,
1867 .attr_count = IPSET_ATTR_CMD_MAX,
1868 .policy = ip_set_type_policy,
1870 [IPSET_CMD_PROTOCOL] = {
1871 .call = ip_set_protocol,
1872 .attr_count = IPSET_ATTR_CMD_MAX,
1873 .policy = ip_set_protocol_policy,
1877 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = {
1879 .subsys_id = NFNL_SUBSYS_IPSET,
1880 .cb_count = IPSET_MSG_MAX,
1881 .cb = ip_set_netlink_subsys_cb,
1884 /* Interface to iptables/ip6tables */
1887 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len)
1891 int copylen = *len, ret = 0;
1892 struct net *net = sock_net(sk);
1893 struct ip_set_net *inst = ip_set_pernet(net);
1895 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1897 if (optval != SO_IP_SET)
1899 if (*len < sizeof(unsigned int))
1902 data = vmalloc(*len);
1905 if (copy_from_user(data, user, *len) != 0) {
1909 op = (unsigned int *)data;
1911 if (*op < IP_SET_OP_VERSION) {
1912 /* Check the version at the beginning of operations */
1913 struct ip_set_req_version *req_version = data;
1915 if (*len < sizeof(struct ip_set_req_version)) {
1920 if (req_version->version != IPSET_PROTOCOL) {
1927 case IP_SET_OP_VERSION: {
1928 struct ip_set_req_version *req_version = data;
1930 if (*len != sizeof(struct ip_set_req_version)) {
1935 req_version->version = IPSET_PROTOCOL;
1936 if (copy_to_user(user, req_version,
1937 sizeof(struct ip_set_req_version)))
1941 case IP_SET_OP_GET_BYNAME: {
1942 struct ip_set_req_get_set *req_get = data;
1945 if (*len != sizeof(struct ip_set_req_get_set)) {
1949 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
1950 nfnl_lock(NFNL_SUBSYS_IPSET);
1951 find_set_and_id(inst, req_get->set.name, &id);
1952 req_get->set.index = id;
1953 nfnl_unlock(NFNL_SUBSYS_IPSET);
1956 case IP_SET_OP_GET_FNAME: {
1957 struct ip_set_req_get_set_family *req_get = data;
1960 if (*len != sizeof(struct ip_set_req_get_set_family)) {
1964 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
1965 nfnl_lock(NFNL_SUBSYS_IPSET);
1966 find_set_and_id(inst, req_get->set.name, &id);
1967 req_get->set.index = id;
1968 if (id != IPSET_INVALID_ID)
1969 req_get->family = ip_set(inst, id)->family;
1970 nfnl_unlock(NFNL_SUBSYS_IPSET);
1973 case IP_SET_OP_GET_BYINDEX: {
1974 struct ip_set_req_get_set *req_get = data;
1977 if (*len != sizeof(struct ip_set_req_get_set) ||
1978 req_get->set.index >= inst->ip_set_max) {
1982 nfnl_lock(NFNL_SUBSYS_IPSET);
1983 set = ip_set(inst, req_get->set.index);
1984 strncpy(req_get->set.name, set ? set->name : "",
1986 nfnl_unlock(NFNL_SUBSYS_IPSET);
1992 } /* end of switch(op) */
1995 if (copy_to_user(user, data, copylen))
2005 static struct nf_sockopt_ops so_set __read_mostly = {
2007 .get_optmin = SO_IP_SET,
2008 .get_optmax = SO_IP_SET + 1,
2009 .get = &ip_set_sockfn_get,
2010 .owner = THIS_MODULE,
2013 static int __net_init
2014 ip_set_net_init(struct net *net)
2016 struct ip_set_net *inst = ip_set_pernet(net);
2017 struct ip_set **list;
2019 inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX;
2020 if (inst->ip_set_max >= IPSET_INVALID_ID)
2021 inst->ip_set_max = IPSET_INVALID_ID - 1;
2023 list = kcalloc(inst->ip_set_max, sizeof(struct ip_set *), GFP_KERNEL);
2026 inst->is_deleted = false;
2027 inst->is_destroyed = false;
2028 rcu_assign_pointer(inst->ip_set_list, list);
2032 static void __net_exit
2033 ip_set_net_exit(struct net *net)
2035 struct ip_set_net *inst = ip_set_pernet(net);
2037 struct ip_set *set = NULL;
2040 inst->is_deleted = true; /* flag for ip_set_nfnl_put */
2042 for (i = 0; i < inst->ip_set_max; i++) {
2043 set = ip_set(inst, i);
2045 ip_set(inst, i) = NULL;
2046 ip_set_destroy_set(set);
2049 kfree(rcu_dereference_protected(inst->ip_set_list, 1));
2052 static struct pernet_operations ip_set_net_ops = {
2053 .init = ip_set_net_init,
2054 .exit = ip_set_net_exit,
2055 .id = &ip_set_net_id,
2056 .size = sizeof(struct ip_set_net)
2062 int ret = nfnetlink_subsys_register(&ip_set_netlink_subsys);
2065 pr_err("ip_set: cannot register with nfnetlink.\n");
2068 ret = nf_register_sockopt(&so_set);
2070 pr_err("SO_SET registry failed: %d\n", ret);
2071 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2074 ret = register_pernet_subsys(&ip_set_net_ops);
2076 pr_err("ip_set: cannot register pernet_subsys.\n");
2077 nf_unregister_sockopt(&so_set);
2078 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2081 pr_info("ip_set: protocol %u\n", IPSET_PROTOCOL);
2088 unregister_pernet_subsys(&ip_set_net_ops);
2089 nf_unregister_sockopt(&so_set);
2090 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2091 pr_debug("these are the famous last words\n");
2094 module_init(ip_set_init);
2095 module_exit(ip_set_fini);