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 unsigned 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 or ip_set_ref_lock is held: */
59 #define ip_set_dereference(p) \
60 rcu_dereference_protected(p, \
61 lockdep_nfnl_is_held(NFNL_SUBSYS_IPSET) || \
62 lockdep_is_held(&ip_set_ref_lock))
63 #define ip_set(inst, id) \
64 ip_set_dereference((inst)->ip_set_list)[id]
65 #define ip_set_ref_netlink(inst,id) \
66 rcu_dereference_raw((inst)->ip_set_list)[id]
68 /* The set types are implemented in modules and registered set types
69 * can be found in ip_set_type_list. Adding/deleting types is
70 * serialized by ip_set_type_mutex.
74 ip_set_type_lock(void)
76 mutex_lock(&ip_set_type_mutex);
80 ip_set_type_unlock(void)
82 mutex_unlock(&ip_set_type_mutex);
85 /* Register and deregister settype */
87 static struct ip_set_type *
88 find_set_type(const char *name, u8 family, u8 revision)
90 struct ip_set_type *type;
92 list_for_each_entry_rcu(type, &ip_set_type_list, list)
93 if (STRNCMP(type->name, name) &&
94 (type->family == family ||
95 type->family == NFPROTO_UNSPEC) &&
96 revision >= type->revision_min &&
97 revision <= type->revision_max)
102 /* Unlock, try to load a set type module and lock again */
104 load_settype(const char *name)
106 nfnl_unlock(NFNL_SUBSYS_IPSET);
107 pr_debug("try to load ip_set_%s\n", name);
108 if (request_module("ip_set_%s", name) < 0) {
109 pr_warn("Can't find ip_set type %s\n", name);
110 nfnl_lock(NFNL_SUBSYS_IPSET);
113 nfnl_lock(NFNL_SUBSYS_IPSET);
117 /* Find a set type and reference it */
118 #define find_set_type_get(name, family, revision, found) \
119 __find_set_type_get(name, family, revision, found, false)
122 __find_set_type_get(const char *name, u8 family, u8 revision,
123 struct ip_set_type **found, bool retry)
125 struct ip_set_type *type;
128 if (retry && !load_settype(name))
129 return -IPSET_ERR_FIND_TYPE;
132 *found = find_set_type(name, family, revision);
134 err = !try_module_get((*found)->me) ? -EFAULT : 0;
137 /* Make sure the type is already loaded
138 * but we don't support the revision
140 list_for_each_entry_rcu(type, &ip_set_type_list, list)
141 if (STRNCMP(type->name, name)) {
142 err = -IPSET_ERR_FIND_TYPE;
147 return retry ? -IPSET_ERR_FIND_TYPE :
148 __find_set_type_get(name, family, revision, found, true);
155 /* Find a given set type by name and family.
156 * If we succeeded, the supported minimal and maximum revisions are
159 #define find_set_type_minmax(name, family, min, max) \
160 __find_set_type_minmax(name, family, min, max, false)
163 __find_set_type_minmax(const char *name, u8 family, u8 *min, u8 *max,
166 struct ip_set_type *type;
169 if (retry && !load_settype(name))
170 return -IPSET_ERR_FIND_TYPE;
172 *min = 255; *max = 0;
174 list_for_each_entry_rcu(type, &ip_set_type_list, list)
175 if (STRNCMP(type->name, name) &&
176 (type->family == family ||
177 type->family == NFPROTO_UNSPEC)) {
179 if (type->revision_min < *min)
180 *min = type->revision_min;
181 if (type->revision_max > *max)
182 *max = type->revision_max;
188 return retry ? -IPSET_ERR_FIND_TYPE :
189 __find_set_type_minmax(name, family, min, max, true);
192 #define family_name(f) ((f) == NFPROTO_IPV4 ? "inet" : \
193 (f) == NFPROTO_IPV6 ? "inet6" : "any")
195 /* Register a set type structure. The type is identified by
196 * the unique triple of name, family and revision.
199 ip_set_type_register(struct ip_set_type *type)
203 if (type->protocol != IPSET_PROTOCOL) {
204 pr_warn("ip_set type %s, family %s, revision %u:%u uses wrong protocol version %u (want %u)\n",
205 type->name, family_name(type->family),
206 type->revision_min, type->revision_max,
207 type->protocol, IPSET_PROTOCOL);
212 if (find_set_type(type->name, type->family, type->revision_min)) {
214 pr_warn("ip_set type %s, family %s with revision min %u already registered!\n",
215 type->name, family_name(type->family),
217 ip_set_type_unlock();
220 list_add_rcu(&type->list, &ip_set_type_list);
221 pr_debug("type %s, family %s, revision %u:%u registered.\n",
222 type->name, family_name(type->family),
223 type->revision_min, type->revision_max);
224 ip_set_type_unlock();
228 EXPORT_SYMBOL_GPL(ip_set_type_register);
230 /* Unregister a set type. There's a small race with ip_set_create */
232 ip_set_type_unregister(struct ip_set_type *type)
235 if (!find_set_type(type->name, type->family, type->revision_min)) {
236 pr_warn("ip_set type %s, family %s with revision min %u not registered\n",
237 type->name, family_name(type->family),
239 ip_set_type_unlock();
242 list_del_rcu(&type->list);
243 pr_debug("type %s, family %s with revision min %u unregistered.\n",
244 type->name, family_name(type->family), type->revision_min);
245 ip_set_type_unlock();
249 EXPORT_SYMBOL_GPL(ip_set_type_unregister);
251 /* Utility functions */
253 ip_set_alloc(size_t size)
255 void *members = NULL;
257 if (size < KMALLOC_MAX_SIZE)
258 members = kzalloc(size, GFP_KERNEL | __GFP_NOWARN);
261 pr_debug("%p: allocated with kmalloc\n", members);
265 members = vzalloc(size);
268 pr_debug("%p: allocated with vmalloc\n", members);
272 EXPORT_SYMBOL_GPL(ip_set_alloc);
275 ip_set_free(void *members)
277 pr_debug("%p: free with %s\n", members,
278 is_vmalloc_addr(members) ? "vfree" : "kfree");
281 EXPORT_SYMBOL_GPL(ip_set_free);
284 flag_nested(const struct nlattr *nla)
286 return nla->nla_type & NLA_F_NESTED;
289 static const struct nla_policy ipaddr_policy[IPSET_ATTR_IPADDR_MAX + 1] = {
290 [IPSET_ATTR_IPADDR_IPV4] = { .type = NLA_U32 },
291 [IPSET_ATTR_IPADDR_IPV6] = { .type = NLA_BINARY,
292 .len = sizeof(struct in6_addr) },
296 ip_set_get_ipaddr4(struct nlattr *nla, __be32 *ipaddr)
298 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1];
300 if (unlikely(!flag_nested(nla)))
301 return -IPSET_ERR_PROTOCOL;
302 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla,
303 ipaddr_policy, NULL))
304 return -IPSET_ERR_PROTOCOL;
305 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV4)))
306 return -IPSET_ERR_PROTOCOL;
308 *ipaddr = nla_get_be32(tb[IPSET_ATTR_IPADDR_IPV4]);
311 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr4);
314 ip_set_get_ipaddr6(struct nlattr *nla, union nf_inet_addr *ipaddr)
316 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1];
318 if (unlikely(!flag_nested(nla)))
319 return -IPSET_ERR_PROTOCOL;
321 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla,
322 ipaddr_policy, NULL))
323 return -IPSET_ERR_PROTOCOL;
324 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV6)))
325 return -IPSET_ERR_PROTOCOL;
327 memcpy(ipaddr, nla_data(tb[IPSET_ATTR_IPADDR_IPV6]),
328 sizeof(struct in6_addr));
331 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr6);
333 typedef void (*destroyer)(struct ip_set *, void *);
334 /* ipset data extension types, in size order */
336 const struct ip_set_ext_type ip_set_extensions[] = {
337 [IPSET_EXT_ID_COUNTER] = {
338 .type = IPSET_EXT_COUNTER,
339 .flag = IPSET_FLAG_WITH_COUNTERS,
340 .len = sizeof(struct ip_set_counter),
341 .align = __alignof__(struct ip_set_counter),
343 [IPSET_EXT_ID_TIMEOUT] = {
344 .type = IPSET_EXT_TIMEOUT,
345 .len = sizeof(unsigned long),
346 .align = __alignof__(unsigned long),
348 [IPSET_EXT_ID_SKBINFO] = {
349 .type = IPSET_EXT_SKBINFO,
350 .flag = IPSET_FLAG_WITH_SKBINFO,
351 .len = sizeof(struct ip_set_skbinfo),
352 .align = __alignof__(struct ip_set_skbinfo),
354 [IPSET_EXT_ID_COMMENT] = {
355 .type = IPSET_EXT_COMMENT | IPSET_EXT_DESTROY,
356 .flag = IPSET_FLAG_WITH_COMMENT,
357 .len = sizeof(struct ip_set_comment),
358 .align = __alignof__(struct ip_set_comment),
359 .destroy = (destroyer) ip_set_comment_free,
362 EXPORT_SYMBOL_GPL(ip_set_extensions);
365 add_extension(enum ip_set_ext_id id, u32 flags, struct nlattr *tb[])
367 return ip_set_extensions[id].flag ?
368 (flags & ip_set_extensions[id].flag) :
369 !!tb[IPSET_ATTR_TIMEOUT];
373 ip_set_elem_len(struct ip_set *set, struct nlattr *tb[], size_t len,
376 enum ip_set_ext_id id;
379 if (tb[IPSET_ATTR_CADT_FLAGS])
380 cadt_flags = ip_set_get_h32(tb[IPSET_ATTR_CADT_FLAGS]);
381 if (cadt_flags & IPSET_FLAG_WITH_FORCEADD)
382 set->flags |= IPSET_CREATE_FLAG_FORCEADD;
385 for (id = 0; id < IPSET_EXT_ID_MAX; id++) {
386 if (!add_extension(id, cadt_flags, tb))
388 if (align < ip_set_extensions[id].align)
389 align = ip_set_extensions[id].align;
390 len = ALIGN(len, ip_set_extensions[id].align);
391 set->offset[id] = len;
392 set->extensions |= ip_set_extensions[id].type;
393 len += ip_set_extensions[id].len;
395 return ALIGN(len, align);
397 EXPORT_SYMBOL_GPL(ip_set_elem_len);
400 ip_set_get_extensions(struct ip_set *set, struct nlattr *tb[],
401 struct ip_set_ext *ext)
405 if (unlikely(!ip_set_optattr_netorder(tb, IPSET_ATTR_TIMEOUT) ||
406 !ip_set_optattr_netorder(tb, IPSET_ATTR_PACKETS) ||
407 !ip_set_optattr_netorder(tb, IPSET_ATTR_BYTES) ||
408 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBMARK) ||
409 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBPRIO) ||
410 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBQUEUE)))
411 return -IPSET_ERR_PROTOCOL;
413 if (tb[IPSET_ATTR_TIMEOUT]) {
414 if (!SET_WITH_TIMEOUT(set))
415 return -IPSET_ERR_TIMEOUT;
416 ext->timeout = ip_set_timeout_uget(tb[IPSET_ATTR_TIMEOUT]);
418 if (tb[IPSET_ATTR_BYTES] || tb[IPSET_ATTR_PACKETS]) {
419 if (!SET_WITH_COUNTER(set))
420 return -IPSET_ERR_COUNTER;
421 if (tb[IPSET_ATTR_BYTES])
422 ext->bytes = be64_to_cpu(nla_get_be64(
423 tb[IPSET_ATTR_BYTES]));
424 if (tb[IPSET_ATTR_PACKETS])
425 ext->packets = be64_to_cpu(nla_get_be64(
426 tb[IPSET_ATTR_PACKETS]));
428 if (tb[IPSET_ATTR_COMMENT]) {
429 if (!SET_WITH_COMMENT(set))
430 return -IPSET_ERR_COMMENT;
431 ext->comment = ip_set_comment_uget(tb[IPSET_ATTR_COMMENT]);
433 if (tb[IPSET_ATTR_SKBMARK]) {
434 if (!SET_WITH_SKBINFO(set))
435 return -IPSET_ERR_SKBINFO;
436 fullmark = be64_to_cpu(nla_get_be64(tb[IPSET_ATTR_SKBMARK]));
437 ext->skbinfo.skbmark = fullmark >> 32;
438 ext->skbinfo.skbmarkmask = fullmark & 0xffffffff;
440 if (tb[IPSET_ATTR_SKBPRIO]) {
441 if (!SET_WITH_SKBINFO(set))
442 return -IPSET_ERR_SKBINFO;
443 ext->skbinfo.skbprio =
444 be32_to_cpu(nla_get_be32(tb[IPSET_ATTR_SKBPRIO]));
446 if (tb[IPSET_ATTR_SKBQUEUE]) {
447 if (!SET_WITH_SKBINFO(set))
448 return -IPSET_ERR_SKBINFO;
449 ext->skbinfo.skbqueue =
450 be16_to_cpu(nla_get_be16(tb[IPSET_ATTR_SKBQUEUE]));
454 EXPORT_SYMBOL_GPL(ip_set_get_extensions);
457 ip_set_put_extensions(struct sk_buff *skb, const struct ip_set *set,
458 const void *e, bool active)
460 if (SET_WITH_TIMEOUT(set)) {
461 unsigned long *timeout = ext_timeout(e, set);
463 if (nla_put_net32(skb, IPSET_ATTR_TIMEOUT,
464 htonl(active ? ip_set_timeout_get(timeout)
468 if (SET_WITH_COUNTER(set) &&
469 ip_set_put_counter(skb, ext_counter(e, set)))
471 if (SET_WITH_COMMENT(set) &&
472 ip_set_put_comment(skb, ext_comment(e, set)))
474 if (SET_WITH_SKBINFO(set) &&
475 ip_set_put_skbinfo(skb, ext_skbinfo(e, set)))
479 EXPORT_SYMBOL_GPL(ip_set_put_extensions);
482 ip_set_match_extensions(struct ip_set *set, const struct ip_set_ext *ext,
483 struct ip_set_ext *mext, u32 flags, void *data)
485 if (SET_WITH_TIMEOUT(set) &&
486 ip_set_timeout_expired(ext_timeout(data, set)))
488 if (SET_WITH_COUNTER(set)) {
489 struct ip_set_counter *counter = ext_counter(data, set);
491 ip_set_update_counter(counter, ext, flags);
493 if (flags & IPSET_FLAG_MATCH_COUNTERS &&
494 !(ip_set_match_counter(ip_set_get_packets(counter),
495 mext->packets, mext->packets_op) &&
496 ip_set_match_counter(ip_set_get_bytes(counter),
497 mext->bytes, mext->bytes_op)))
500 if (SET_WITH_SKBINFO(set))
501 ip_set_get_skbinfo(ext_skbinfo(data, set),
505 EXPORT_SYMBOL_GPL(ip_set_match_extensions);
507 /* Creating/destroying/renaming/swapping affect the existence and
508 * the properties of a set. All of these can be executed from userspace
509 * only and serialized by the nfnl mutex indirectly from nfnetlink.
511 * Sets are identified by their index in ip_set_list and the index
512 * is used by the external references (set/SET netfilter modules).
514 * The set behind an index may change by swapping only, from userspace.
518 __ip_set_get(struct ip_set *set)
520 write_lock_bh(&ip_set_ref_lock);
522 write_unlock_bh(&ip_set_ref_lock);
526 __ip_set_put(struct ip_set *set)
528 write_lock_bh(&ip_set_ref_lock);
529 BUG_ON(set->ref == 0);
531 write_unlock_bh(&ip_set_ref_lock);
534 /* set->ref can be swapped out by ip_set_swap, netlink events (like dump) need
535 * a separate reference counter
538 __ip_set_put_netlink(struct ip_set *set)
540 write_lock_bh(&ip_set_ref_lock);
541 BUG_ON(set->ref_netlink == 0);
543 write_unlock_bh(&ip_set_ref_lock);
546 /* Add, del and test set entries from kernel.
548 * The set behind the index must exist and must be referenced
549 * so it can't be destroyed (or changed) under our foot.
552 static inline struct ip_set *
553 ip_set_rcu_get(struct net *net, ip_set_id_t index)
556 struct ip_set_net *inst = ip_set_pernet(net);
559 /* ip_set_list itself needs to be protected */
560 set = rcu_dereference(inst->ip_set_list)[index];
567 ip_set_test(ip_set_id_t index, const struct sk_buff *skb,
568 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
570 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
574 pr_debug("set %s, index %u\n", set->name, index);
576 if (opt->dim < set->type->dimension ||
577 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
581 ret = set->variant->kadt(set, skb, par, IPSET_TEST, opt);
582 rcu_read_unlock_bh();
584 if (ret == -EAGAIN) {
585 /* Type requests element to be completed */
586 pr_debug("element must be completed, ADD is triggered\n");
587 spin_lock_bh(&set->lock);
588 set->variant->kadt(set, skb, par, IPSET_ADD, opt);
589 spin_unlock_bh(&set->lock);
592 /* --return-nomatch: invert matched element */
593 if ((opt->cmdflags & IPSET_FLAG_RETURN_NOMATCH) &&
594 (set->type->features & IPSET_TYPE_NOMATCH) &&
595 (ret > 0 || ret == -ENOTEMPTY))
599 /* Convert error codes to nomatch */
600 return (ret < 0 ? 0 : ret);
602 EXPORT_SYMBOL_GPL(ip_set_test);
605 ip_set_add(ip_set_id_t index, const struct sk_buff *skb,
606 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
608 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
612 pr_debug("set %s, index %u\n", set->name, index);
614 if (opt->dim < set->type->dimension ||
615 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
616 return -IPSET_ERR_TYPE_MISMATCH;
618 spin_lock_bh(&set->lock);
619 ret = set->variant->kadt(set, skb, par, IPSET_ADD, opt);
620 spin_unlock_bh(&set->lock);
624 EXPORT_SYMBOL_GPL(ip_set_add);
627 ip_set_del(ip_set_id_t index, const struct sk_buff *skb,
628 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
630 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
634 pr_debug("set %s, index %u\n", set->name, index);
636 if (opt->dim < set->type->dimension ||
637 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
638 return -IPSET_ERR_TYPE_MISMATCH;
640 spin_lock_bh(&set->lock);
641 ret = set->variant->kadt(set, skb, par, IPSET_DEL, opt);
642 spin_unlock_bh(&set->lock);
646 EXPORT_SYMBOL_GPL(ip_set_del);
648 /* Find set by name, reference it once. The reference makes sure the
649 * thing pointed to, does not go away under our feet.
653 ip_set_get_byname(struct net *net, const char *name, struct ip_set **set)
655 ip_set_id_t i, index = IPSET_INVALID_ID;
657 struct ip_set_net *inst = ip_set_pernet(net);
660 for (i = 0; i < inst->ip_set_max; i++) {
661 s = rcu_dereference(inst->ip_set_list)[i];
662 if (s && STRNCMP(s->name, name)) {
673 EXPORT_SYMBOL_GPL(ip_set_get_byname);
675 /* If the given set pointer points to a valid set, decrement
676 * reference count by 1. The caller shall not assume the index
677 * to be valid, after calling this function.
682 __ip_set_put_byindex(struct ip_set_net *inst, ip_set_id_t index)
687 set = rcu_dereference(inst->ip_set_list)[index];
694 ip_set_put_byindex(struct net *net, ip_set_id_t index)
696 struct ip_set_net *inst = ip_set_pernet(net);
698 __ip_set_put_byindex(inst, index);
700 EXPORT_SYMBOL_GPL(ip_set_put_byindex);
702 /* Get the name of a set behind a set index.
703 * Set itself is protected by RCU, but its name isn't: to protect against
704 * renaming, grab ip_set_ref_lock as reader (see ip_set_rename()) and copy the
708 ip_set_name_byindex(struct net *net, ip_set_id_t index, char *name)
710 struct ip_set *set = ip_set_rcu_get(net, index);
714 read_lock_bh(&ip_set_ref_lock);
715 strncpy(name, set->name, IPSET_MAXNAMELEN);
716 read_unlock_bh(&ip_set_ref_lock);
718 EXPORT_SYMBOL_GPL(ip_set_name_byindex);
720 /* Routines to call by external subsystems, which do not
721 * call nfnl_lock for us.
724 /* Find set by index, reference it once. The reference makes sure the
725 * thing pointed to, does not go away under our feet.
727 * The nfnl mutex is used in the function.
730 ip_set_nfnl_get_byindex(struct net *net, ip_set_id_t index)
733 struct ip_set_net *inst = ip_set_pernet(net);
735 if (index >= inst->ip_set_max)
736 return IPSET_INVALID_ID;
738 nfnl_lock(NFNL_SUBSYS_IPSET);
739 set = ip_set(inst, index);
743 index = IPSET_INVALID_ID;
744 nfnl_unlock(NFNL_SUBSYS_IPSET);
748 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex);
750 /* If the given set pointer points to a valid set, decrement
751 * reference count by 1. The caller shall not assume the index
752 * to be valid, after calling this function.
754 * The nfnl mutex is used in the function.
757 ip_set_nfnl_put(struct net *net, ip_set_id_t index)
760 struct ip_set_net *inst = ip_set_pernet(net);
762 nfnl_lock(NFNL_SUBSYS_IPSET);
763 if (!inst->is_deleted) { /* already deleted from ip_set_net_exit() */
764 set = ip_set(inst, index);
768 nfnl_unlock(NFNL_SUBSYS_IPSET);
770 EXPORT_SYMBOL_GPL(ip_set_nfnl_put);
772 /* Communication protocol with userspace over netlink.
774 * The commands are serialized by the nfnl mutex.
778 protocol_failed(const struct nlattr * const tb[])
780 return !tb[IPSET_ATTR_PROTOCOL] ||
781 nla_get_u8(tb[IPSET_ATTR_PROTOCOL]) != IPSET_PROTOCOL;
785 flag_exist(const struct nlmsghdr *nlh)
787 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST;
790 static struct nlmsghdr *
791 start_msg(struct sk_buff *skb, u32 portid, u32 seq, unsigned int flags,
794 struct nlmsghdr *nlh;
795 struct nfgenmsg *nfmsg;
797 nlh = nlmsg_put(skb, portid, seq, nfnl_msg_type(NFNL_SUBSYS_IPSET, cmd),
798 sizeof(*nfmsg), flags);
802 nfmsg = nlmsg_data(nlh);
803 nfmsg->nfgen_family = NFPROTO_IPV4;
804 nfmsg->version = NFNETLINK_V0;
812 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = {
813 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
814 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
815 .len = IPSET_MAXNAMELEN - 1 },
816 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
817 .len = IPSET_MAXNAMELEN - 1},
818 [IPSET_ATTR_REVISION] = { .type = NLA_U8 },
819 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
820 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
823 static struct ip_set *
824 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id)
826 struct ip_set *set = NULL;
829 *id = IPSET_INVALID_ID;
830 for (i = 0; i < inst->ip_set_max; i++) {
831 set = ip_set(inst, i);
832 if (set && STRNCMP(set->name, name)) {
837 return (*id == IPSET_INVALID_ID ? NULL : set);
840 static inline struct ip_set *
841 find_set(struct ip_set_net *inst, const char *name)
845 return find_set_and_id(inst, name, &id);
849 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index,
855 *index = IPSET_INVALID_ID;
856 for (i = 0; i < inst->ip_set_max; i++) {
859 if (*index == IPSET_INVALID_ID)
861 } else if (STRNCMP(name, s->name)) {
867 if (*index == IPSET_INVALID_ID)
868 /* No free slot remained */
869 return -IPSET_ERR_MAX_SETS;
873 static int ip_set_none(struct net *net, struct sock *ctnl, struct sk_buff *skb,
874 const struct nlmsghdr *nlh,
875 const struct nlattr * const attr[],
876 struct netlink_ext_ack *extack)
881 static int ip_set_create(struct net *net, struct sock *ctnl,
882 struct sk_buff *skb, const struct nlmsghdr *nlh,
883 const struct nlattr * const attr[],
884 struct netlink_ext_ack *extack)
886 struct ip_set_net *inst = ip_set_pernet(net);
887 struct ip_set *set, *clash = NULL;
888 ip_set_id_t index = IPSET_INVALID_ID;
889 struct nlattr *tb[IPSET_ATTR_CREATE_MAX + 1] = {};
890 const char *name, *typename;
892 u32 flags = flag_exist(nlh);
895 if (unlikely(protocol_failed(attr) ||
896 !attr[IPSET_ATTR_SETNAME] ||
897 !attr[IPSET_ATTR_TYPENAME] ||
898 !attr[IPSET_ATTR_REVISION] ||
899 !attr[IPSET_ATTR_FAMILY] ||
900 (attr[IPSET_ATTR_DATA] &&
901 !flag_nested(attr[IPSET_ATTR_DATA]))))
902 return -IPSET_ERR_PROTOCOL;
904 name = nla_data(attr[IPSET_ATTR_SETNAME]);
905 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
906 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
907 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]);
908 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
909 name, typename, family_name(family), revision);
911 /* First, and without any locks, allocate and initialize
912 * a normal base set structure.
914 set = kzalloc(sizeof(*set), GFP_KERNEL);
917 spin_lock_init(&set->lock);
918 strlcpy(set->name, name, IPSET_MAXNAMELEN);
919 set->family = family;
920 set->revision = revision;
922 /* Next, check that we know the type, and take
923 * a reference on the type, to make sure it stays available
924 * while constructing our new set.
926 * After referencing the type, we try to create the type
927 * specific part of the set without holding any locks.
929 ret = find_set_type_get(typename, family, revision, &set->type);
933 /* Without holding any locks, create private part. */
934 if (attr[IPSET_ATTR_DATA] &&
935 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA],
936 set->type->create_policy, NULL)) {
937 ret = -IPSET_ERR_PROTOCOL;
941 ret = set->type->create(net, set, tb, flags);
945 /* BTW, ret==0 here. */
947 /* Here, we have a valid, constructed set and we are protected
948 * by the nfnl mutex. Find the first free index in ip_set_list
949 * and check clashing.
951 ret = find_free_id(inst, set->name, &index, &clash);
952 if (ret == -EEXIST) {
953 /* If this is the same set and requested, ignore error */
954 if ((flags & IPSET_FLAG_EXIST) &&
955 STRNCMP(set->type->name, clash->type->name) &&
956 set->type->family == clash->type->family &&
957 set->type->revision_min == clash->type->revision_min &&
958 set->type->revision_max == clash->type->revision_max &&
959 set->variant->same_set(set, clash))
962 } else if (ret == -IPSET_ERR_MAX_SETS) {
963 struct ip_set **list, **tmp;
964 ip_set_id_t i = inst->ip_set_max + IP_SET_INC;
966 if (i < inst->ip_set_max || i == IPSET_INVALID_ID)
970 list = kvcalloc(i, sizeof(struct ip_set *), GFP_KERNEL);
973 /* nfnl mutex is held, both lists are valid */
974 tmp = ip_set_dereference(inst->ip_set_list);
975 memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max);
976 rcu_assign_pointer(inst->ip_set_list, list);
977 /* Make sure all current packets have passed through */
980 index = inst->ip_set_max;
981 inst->ip_set_max = i;
988 /* Finally! Add our shiny new set to the list, and be done. */
989 pr_debug("create: '%s' created with index %u!\n", set->name, index);
990 ip_set(inst, index) = set;
995 set->variant->destroy(set);
997 module_put(set->type->me);
1005 static const struct nla_policy
1006 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = {
1007 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1008 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1009 .len = IPSET_MAXNAMELEN - 1 },
1013 ip_set_destroy_set(struct ip_set *set)
1015 pr_debug("set: %s\n", set->name);
1017 /* Must call it without holding any lock */
1018 set->variant->destroy(set);
1019 module_put(set->type->me);
1023 static int ip_set_destroy(struct net *net, struct sock *ctnl,
1024 struct sk_buff *skb, const struct nlmsghdr *nlh,
1025 const struct nlattr * const attr[],
1026 struct netlink_ext_ack *extack)
1028 struct ip_set_net *inst = ip_set_pernet(net);
1033 if (unlikely(protocol_failed(attr)))
1034 return -IPSET_ERR_PROTOCOL;
1036 /* Must wait for flush to be really finished in list:set */
1039 /* Commands are serialized and references are
1040 * protected by the ip_set_ref_lock.
1041 * External systems (i.e. xt_set) must call
1042 * ip_set_put|get_nfnl_* functions, that way we
1043 * can safely check references here.
1045 * list:set timer can only decrement the reference
1046 * counter, so if it's already zero, we can proceed
1047 * without holding the lock.
1049 read_lock_bh(&ip_set_ref_lock);
1050 if (!attr[IPSET_ATTR_SETNAME]) {
1051 for (i = 0; i < inst->ip_set_max; i++) {
1052 s = ip_set(inst, i);
1053 if (s && (s->ref || s->ref_netlink)) {
1054 ret = -IPSET_ERR_BUSY;
1058 inst->is_destroyed = true;
1059 read_unlock_bh(&ip_set_ref_lock);
1060 for (i = 0; i < inst->ip_set_max; i++) {
1061 s = ip_set(inst, i);
1063 ip_set(inst, i) = NULL;
1064 ip_set_destroy_set(s);
1067 /* Modified by ip_set_destroy() only, which is serialized */
1068 inst->is_destroyed = false;
1070 s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1075 } else if (s->ref || s->ref_netlink) {
1076 ret = -IPSET_ERR_BUSY;
1079 ip_set(inst, i) = NULL;
1080 read_unlock_bh(&ip_set_ref_lock);
1082 ip_set_destroy_set(s);
1086 read_unlock_bh(&ip_set_ref_lock);
1093 ip_set_flush_set(struct ip_set *set)
1095 pr_debug("set: %s\n", set->name);
1097 spin_lock_bh(&set->lock);
1098 set->variant->flush(set);
1099 spin_unlock_bh(&set->lock);
1102 static int ip_set_flush(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1103 const struct nlmsghdr *nlh,
1104 const struct nlattr * const attr[],
1105 struct netlink_ext_ack *extack)
1107 struct ip_set_net *inst = ip_set_pernet(net);
1111 if (unlikely(protocol_failed(attr)))
1112 return -IPSET_ERR_PROTOCOL;
1114 if (!attr[IPSET_ATTR_SETNAME]) {
1115 for (i = 0; i < inst->ip_set_max; i++) {
1116 s = ip_set(inst, i);
1118 ip_set_flush_set(s);
1121 s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1125 ip_set_flush_set(s);
1133 static const struct nla_policy
1134 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = {
1135 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1136 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1137 .len = IPSET_MAXNAMELEN - 1 },
1138 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING,
1139 .len = IPSET_MAXNAMELEN - 1 },
1142 static int ip_set_rename(struct net *net, struct sock *ctnl,
1143 struct sk_buff *skb, const struct nlmsghdr *nlh,
1144 const struct nlattr * const attr[],
1145 struct netlink_ext_ack *extack)
1147 struct ip_set_net *inst = ip_set_pernet(net);
1148 struct ip_set *set, *s;
1153 if (unlikely(protocol_failed(attr) ||
1154 !attr[IPSET_ATTR_SETNAME] ||
1155 !attr[IPSET_ATTR_SETNAME2]))
1156 return -IPSET_ERR_PROTOCOL;
1158 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1162 write_lock_bh(&ip_set_ref_lock);
1163 if (set->ref != 0 || set->ref_netlink != 0) {
1164 ret = -IPSET_ERR_REFERENCED;
1168 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]);
1169 for (i = 0; i < inst->ip_set_max; i++) {
1170 s = ip_set(inst, i);
1171 if (s && STRNCMP(s->name, name2)) {
1172 ret = -IPSET_ERR_EXIST_SETNAME2;
1176 strncpy(set->name, name2, IPSET_MAXNAMELEN);
1179 write_unlock_bh(&ip_set_ref_lock);
1183 /* Swap two sets so that name/index points to the other.
1184 * References and set names are also swapped.
1186 * The commands are serialized by the nfnl mutex and references are
1187 * protected by the ip_set_ref_lock. The kernel interfaces
1188 * do not hold the mutex but the pointer settings are atomic
1189 * so the ip_set_list always contains valid pointers to the sets.
1192 static int ip_set_swap(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1193 const struct nlmsghdr *nlh,
1194 const struct nlattr * const attr[],
1195 struct netlink_ext_ack *extack)
1197 struct ip_set_net *inst = ip_set_pernet(net);
1198 struct ip_set *from, *to;
1199 ip_set_id_t from_id, to_id;
1200 char from_name[IPSET_MAXNAMELEN];
1202 if (unlikely(protocol_failed(attr) ||
1203 !attr[IPSET_ATTR_SETNAME] ||
1204 !attr[IPSET_ATTR_SETNAME2]))
1205 return -IPSET_ERR_PROTOCOL;
1207 from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1212 to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]),
1215 return -IPSET_ERR_EXIST_SETNAME2;
1217 /* Features must not change.
1218 * Not an artifical restriction anymore, as we must prevent
1219 * possible loops created by swapping in setlist type of sets.
1221 if (!(from->type->features == to->type->features &&
1222 from->family == to->family))
1223 return -IPSET_ERR_TYPE_MISMATCH;
1225 write_lock_bh(&ip_set_ref_lock);
1227 if (from->ref_netlink || to->ref_netlink) {
1228 write_unlock_bh(&ip_set_ref_lock);
1232 strncpy(from_name, from->name, IPSET_MAXNAMELEN);
1233 strncpy(from->name, to->name, IPSET_MAXNAMELEN);
1234 strncpy(to->name, from_name, IPSET_MAXNAMELEN);
1236 swap(from->ref, to->ref);
1237 ip_set(inst, from_id) = to;
1238 ip_set(inst, to_id) = from;
1239 write_unlock_bh(&ip_set_ref_lock);
1244 /* List/save set data */
1251 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1252 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1255 ip_set_dump_done(struct netlink_callback *cb)
1257 if (cb->args[IPSET_CB_ARG0]) {
1258 struct ip_set_net *inst =
1259 (struct ip_set_net *)cb->args[IPSET_CB_NET];
1260 ip_set_id_t index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1261 struct ip_set *set = ip_set_ref_netlink(inst, index);
1263 if (set->variant->uref)
1264 set->variant->uref(set, cb, false);
1265 pr_debug("release set %s\n", set->name);
1266 __ip_set_put_netlink(set);
1272 dump_attrs(struct nlmsghdr *nlh)
1274 const struct nlattr *attr;
1277 pr_debug("dump nlmsg\n");
1278 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) {
1279 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len);
1284 dump_init(struct netlink_callback *cb, struct ip_set_net *inst)
1286 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb);
1287 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1288 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1289 struct nlattr *attr = (void *)nlh + min_len;
1293 /* Second pass, so parser can't fail */
1294 nla_parse(cda, IPSET_ATTR_CMD_MAX, attr, nlh->nlmsg_len - min_len,
1295 ip_set_setname_policy, NULL);
1297 if (cda[IPSET_ATTR_SETNAME]) {
1300 set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]),
1305 dump_type = DUMP_ONE;
1306 cb->args[IPSET_CB_INDEX] = index;
1308 dump_type = DUMP_ALL;
1311 if (cda[IPSET_ATTR_FLAGS]) {
1312 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]);
1314 dump_type |= (f << 16);
1316 cb->args[IPSET_CB_NET] = (unsigned long)inst;
1317 cb->args[IPSET_CB_DUMP] = dump_type;
1323 ip_set_dump_start(struct sk_buff *skb, struct netlink_callback *cb)
1325 ip_set_id_t index = IPSET_INVALID_ID, max;
1326 struct ip_set *set = NULL;
1327 struct nlmsghdr *nlh = NULL;
1328 unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0;
1329 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1330 u32 dump_type, dump_flags;
1334 if (!cb->args[IPSET_CB_DUMP]) {
1335 ret = dump_init(cb, inst);
1337 nlh = nlmsg_hdr(cb->skb);
1338 /* We have to create and send the error message
1341 if (nlh->nlmsg_flags & NLM_F_ACK)
1342 netlink_ack(cb->skb, nlh, ret, NULL);
1347 if (cb->args[IPSET_CB_INDEX] >= inst->ip_set_max)
1350 dump_type = DUMP_TYPE(cb->args[IPSET_CB_DUMP]);
1351 dump_flags = DUMP_FLAGS(cb->args[IPSET_CB_DUMP]);
1352 max = dump_type == DUMP_ONE ? cb->args[IPSET_CB_INDEX] + 1
1355 pr_debug("dump type, flag: %u %u index: %ld\n",
1356 dump_type, dump_flags, cb->args[IPSET_CB_INDEX]);
1357 for (; cb->args[IPSET_CB_INDEX] < max; cb->args[IPSET_CB_INDEX]++) {
1358 index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1359 write_lock_bh(&ip_set_ref_lock);
1360 set = ip_set(inst, index);
1361 is_destroyed = inst->is_destroyed;
1362 if (!set || is_destroyed) {
1363 write_unlock_bh(&ip_set_ref_lock);
1364 if (dump_type == DUMP_ONE) {
1369 /* All sets are just being destroyed */
1375 /* When dumping all sets, we must dump "sorted"
1376 * so that lists (unions of sets) are dumped last.
1378 if (dump_type != DUMP_ONE &&
1379 ((dump_type == DUMP_ALL) ==
1380 !!(set->type->features & IPSET_DUMP_LAST))) {
1381 write_unlock_bh(&ip_set_ref_lock);
1384 pr_debug("List set: %s\n", set->name);
1385 if (!cb->args[IPSET_CB_ARG0]) {
1386 /* Start listing: make sure set won't be destroyed */
1387 pr_debug("reference set\n");
1390 write_unlock_bh(&ip_set_ref_lock);
1391 nlh = start_msg(skb, NETLINK_CB(cb->skb).portid,
1392 cb->nlh->nlmsg_seq, flags,
1396 goto release_refcount;
1398 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1399 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name))
1400 goto nla_put_failure;
1401 if (dump_flags & IPSET_FLAG_LIST_SETNAME)
1403 switch (cb->args[IPSET_CB_ARG0]) {
1405 /* Core header data */
1406 if (nla_put_string(skb, IPSET_ATTR_TYPENAME,
1408 nla_put_u8(skb, IPSET_ATTR_FAMILY,
1410 nla_put_u8(skb, IPSET_ATTR_REVISION,
1412 goto nla_put_failure;
1413 ret = set->variant->head(set, skb);
1415 goto release_refcount;
1416 if (dump_flags & IPSET_FLAG_LIST_HEADER)
1418 if (set->variant->uref)
1419 set->variant->uref(set, cb, true);
1422 ret = set->variant->list(set, skb, cb);
1423 if (!cb->args[IPSET_CB_ARG0])
1424 /* Set is done, proceed with next one */
1426 goto release_refcount;
1429 /* If we dump all sets, continue with dumping last ones */
1430 if (dump_type == DUMP_ALL) {
1431 dump_type = DUMP_LAST;
1432 cb->args[IPSET_CB_DUMP] = dump_type | (dump_flags << 16);
1433 cb->args[IPSET_CB_INDEX] = 0;
1434 if (set && set->variant->uref)
1435 set->variant->uref(set, cb, false);
1443 if (dump_type == DUMP_ONE)
1444 cb->args[IPSET_CB_INDEX] = IPSET_INVALID_ID;
1446 cb->args[IPSET_CB_INDEX]++;
1448 /* If there was an error or set is done, release set */
1449 if (ret || !cb->args[IPSET_CB_ARG0]) {
1450 set = ip_set_ref_netlink(inst, index);
1451 if (set->variant->uref)
1452 set->variant->uref(set, cb, false);
1453 pr_debug("release set %s\n", set->name);
1454 __ip_set_put_netlink(set);
1455 cb->args[IPSET_CB_ARG0] = 0;
1459 nlmsg_end(skb, nlh);
1460 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len);
1464 return ret < 0 ? ret : skb->len;
1467 static int ip_set_dump(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1468 const struct nlmsghdr *nlh,
1469 const struct nlattr * const attr[],
1470 struct netlink_ext_ack *extack)
1472 if (unlikely(protocol_failed(attr)))
1473 return -IPSET_ERR_PROTOCOL;
1476 struct netlink_dump_control c = {
1477 .dump = ip_set_dump_start,
1478 .done = ip_set_dump_done,
1480 return netlink_dump_start(ctnl, skb, nlh, &c);
1484 /* Add, del and test */
1486 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = {
1487 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1488 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1489 .len = IPSET_MAXNAMELEN - 1 },
1490 [IPSET_ATTR_LINENO] = { .type = NLA_U32 },
1491 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
1492 [IPSET_ATTR_ADT] = { .type = NLA_NESTED },
1496 call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set,
1497 struct nlattr *tb[], enum ipset_adt adt,
1498 u32 flags, bool use_lineno)
1502 bool eexist = flags & IPSET_FLAG_EXIST, retried = false;
1505 spin_lock_bh(&set->lock);
1506 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried);
1507 spin_unlock_bh(&set->lock);
1509 } while (ret == -EAGAIN &&
1510 set->variant->resize &&
1511 (ret = set->variant->resize(set, retried)) == 0);
1513 if (!ret || (ret == -IPSET_ERR_EXIST && eexist))
1515 if (lineno && use_lineno) {
1516 /* Error in restore/batch mode: send back lineno */
1517 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb);
1518 struct sk_buff *skb2;
1519 struct nlmsgerr *errmsg;
1520 size_t payload = min(SIZE_MAX,
1521 sizeof(*errmsg) + nlmsg_len(nlh));
1522 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1523 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1524 struct nlattr *cmdattr;
1527 skb2 = nlmsg_new(payload, GFP_KERNEL);
1530 rep = __nlmsg_put(skb2, NETLINK_CB(skb).portid,
1531 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0);
1532 errmsg = nlmsg_data(rep);
1533 errmsg->error = ret;
1534 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len);
1535 cmdattr = (void *)&errmsg->msg + min_len;
1537 nla_parse(cda, IPSET_ATTR_CMD_MAX, cmdattr,
1538 nlh->nlmsg_len - min_len, ip_set_adt_policy, NULL);
1540 errline = nla_data(cda[IPSET_ATTR_LINENO]);
1544 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid,
1546 /* Signal netlink not to send its ACK/errmsg. */
1553 static int ip_set_uadd(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1554 const struct nlmsghdr *nlh,
1555 const struct nlattr * const attr[],
1556 struct netlink_ext_ack *extack)
1558 struct ip_set_net *inst = ip_set_pernet(net);
1560 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1561 const struct nlattr *nla;
1562 u32 flags = flag_exist(nlh);
1566 if (unlikely(protocol_failed(attr) ||
1567 !attr[IPSET_ATTR_SETNAME] ||
1568 !((attr[IPSET_ATTR_DATA] != NULL) ^
1569 (attr[IPSET_ATTR_ADT] != NULL)) ||
1570 (attr[IPSET_ATTR_DATA] &&
1571 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1572 (attr[IPSET_ATTR_ADT] &&
1573 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1574 !attr[IPSET_ATTR_LINENO]))))
1575 return -IPSET_ERR_PROTOCOL;
1577 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1581 use_lineno = !!attr[IPSET_ATTR_LINENO];
1582 if (attr[IPSET_ATTR_DATA]) {
1583 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1584 attr[IPSET_ATTR_DATA],
1585 set->type->adt_policy, NULL))
1586 return -IPSET_ERR_PROTOCOL;
1587 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD, flags,
1592 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1593 memset(tb, 0, sizeof(tb));
1594 if (nla_type(nla) != IPSET_ATTR_DATA ||
1595 !flag_nested(nla) ||
1596 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1597 set->type->adt_policy, NULL))
1598 return -IPSET_ERR_PROTOCOL;
1599 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD,
1608 static int ip_set_udel(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1609 const struct nlmsghdr *nlh,
1610 const struct nlattr * const attr[],
1611 struct netlink_ext_ack *extack)
1613 struct ip_set_net *inst = ip_set_pernet(net);
1615 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1616 const struct nlattr *nla;
1617 u32 flags = flag_exist(nlh);
1621 if (unlikely(protocol_failed(attr) ||
1622 !attr[IPSET_ATTR_SETNAME] ||
1623 !((attr[IPSET_ATTR_DATA] != NULL) ^
1624 (attr[IPSET_ATTR_ADT] != NULL)) ||
1625 (attr[IPSET_ATTR_DATA] &&
1626 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1627 (attr[IPSET_ATTR_ADT] &&
1628 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1629 !attr[IPSET_ATTR_LINENO]))))
1630 return -IPSET_ERR_PROTOCOL;
1632 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1636 use_lineno = !!attr[IPSET_ATTR_LINENO];
1637 if (attr[IPSET_ATTR_DATA]) {
1638 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1639 attr[IPSET_ATTR_DATA],
1640 set->type->adt_policy, NULL))
1641 return -IPSET_ERR_PROTOCOL;
1642 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL, flags,
1647 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1648 memset(tb, 0, sizeof(*tb));
1649 if (nla_type(nla) != IPSET_ATTR_DATA ||
1650 !flag_nested(nla) ||
1651 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1652 set->type->adt_policy, NULL))
1653 return -IPSET_ERR_PROTOCOL;
1654 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL,
1663 static int ip_set_utest(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1664 const struct nlmsghdr *nlh,
1665 const struct nlattr * const attr[],
1666 struct netlink_ext_ack *extack)
1668 struct ip_set_net *inst = ip_set_pernet(net);
1670 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1674 if (unlikely(protocol_failed(attr) ||
1675 !attr[IPSET_ATTR_SETNAME] ||
1676 !attr[IPSET_ATTR_DATA] ||
1677 !flag_nested(attr[IPSET_ATTR_DATA])))
1678 return -IPSET_ERR_PROTOCOL;
1680 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1684 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA],
1685 set->type->adt_policy, NULL))
1686 return -IPSET_ERR_PROTOCOL;
1689 ret = set->variant->uadt(set, tb, IPSET_TEST, &lineno, 0, 0);
1690 rcu_read_unlock_bh();
1691 /* Userspace can't trigger element to be re-added */
1695 return ret > 0 ? 0 : -IPSET_ERR_EXIST;
1698 /* Get headed data of a set */
1700 static int ip_set_header(struct net *net, struct sock *ctnl,
1701 struct sk_buff *skb, const struct nlmsghdr *nlh,
1702 const struct nlattr * const attr[],
1703 struct netlink_ext_ack *extack)
1705 struct ip_set_net *inst = ip_set_pernet(net);
1706 const struct ip_set *set;
1707 struct sk_buff *skb2;
1708 struct nlmsghdr *nlh2;
1711 if (unlikely(protocol_failed(attr) ||
1712 !attr[IPSET_ATTR_SETNAME]))
1713 return -IPSET_ERR_PROTOCOL;
1715 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1719 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1723 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1727 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1728 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) ||
1729 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) ||
1730 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
1731 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision))
1732 goto nla_put_failure;
1733 nlmsg_end(skb2, nlh2);
1735 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1742 nlmsg_cancel(skb2, nlh2);
1750 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = {
1751 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1752 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
1753 .len = IPSET_MAXNAMELEN - 1 },
1754 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
1757 static int ip_set_type(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1758 const struct nlmsghdr *nlh,
1759 const struct nlattr * const attr[],
1760 struct netlink_ext_ack *extack)
1762 struct sk_buff *skb2;
1763 struct nlmsghdr *nlh2;
1764 u8 family, min, max;
1765 const char *typename;
1768 if (unlikely(protocol_failed(attr) ||
1769 !attr[IPSET_ATTR_TYPENAME] ||
1770 !attr[IPSET_ATTR_FAMILY]))
1771 return -IPSET_ERR_PROTOCOL;
1773 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1774 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1775 ret = find_set_type_minmax(typename, family, &min, &max);
1779 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1783 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1787 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1788 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) ||
1789 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) ||
1790 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) ||
1791 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min))
1792 goto nla_put_failure;
1793 nlmsg_end(skb2, nlh2);
1795 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len);
1796 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1803 nlmsg_cancel(skb2, nlh2);
1809 /* Get protocol version */
1811 static const struct nla_policy
1812 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = {
1813 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1816 static int ip_set_protocol(struct net *net, struct sock *ctnl,
1817 struct sk_buff *skb, const struct nlmsghdr *nlh,
1818 const struct nlattr * const attr[],
1819 struct netlink_ext_ack *extack)
1821 struct sk_buff *skb2;
1822 struct nlmsghdr *nlh2;
1825 if (unlikely(!attr[IPSET_ATTR_PROTOCOL]))
1826 return -IPSET_ERR_PROTOCOL;
1828 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1832 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1833 IPSET_CMD_PROTOCOL);
1836 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL))
1837 goto nla_put_failure;
1838 nlmsg_end(skb2, nlh2);
1840 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1847 nlmsg_cancel(skb2, nlh2);
1853 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = {
1854 [IPSET_CMD_NONE] = {
1855 .call = ip_set_none,
1856 .attr_count = IPSET_ATTR_CMD_MAX,
1858 [IPSET_CMD_CREATE] = {
1859 .call = ip_set_create,
1860 .attr_count = IPSET_ATTR_CMD_MAX,
1861 .policy = ip_set_create_policy,
1863 [IPSET_CMD_DESTROY] = {
1864 .call = ip_set_destroy,
1865 .attr_count = IPSET_ATTR_CMD_MAX,
1866 .policy = ip_set_setname_policy,
1868 [IPSET_CMD_FLUSH] = {
1869 .call = ip_set_flush,
1870 .attr_count = IPSET_ATTR_CMD_MAX,
1871 .policy = ip_set_setname_policy,
1873 [IPSET_CMD_RENAME] = {
1874 .call = ip_set_rename,
1875 .attr_count = IPSET_ATTR_CMD_MAX,
1876 .policy = ip_set_setname2_policy,
1878 [IPSET_CMD_SWAP] = {
1879 .call = ip_set_swap,
1880 .attr_count = IPSET_ATTR_CMD_MAX,
1881 .policy = ip_set_setname2_policy,
1883 [IPSET_CMD_LIST] = {
1884 .call = ip_set_dump,
1885 .attr_count = IPSET_ATTR_CMD_MAX,
1886 .policy = ip_set_setname_policy,
1888 [IPSET_CMD_SAVE] = {
1889 .call = ip_set_dump,
1890 .attr_count = IPSET_ATTR_CMD_MAX,
1891 .policy = ip_set_setname_policy,
1894 .call = ip_set_uadd,
1895 .attr_count = IPSET_ATTR_CMD_MAX,
1896 .policy = ip_set_adt_policy,
1899 .call = ip_set_udel,
1900 .attr_count = IPSET_ATTR_CMD_MAX,
1901 .policy = ip_set_adt_policy,
1903 [IPSET_CMD_TEST] = {
1904 .call = ip_set_utest,
1905 .attr_count = IPSET_ATTR_CMD_MAX,
1906 .policy = ip_set_adt_policy,
1908 [IPSET_CMD_HEADER] = {
1909 .call = ip_set_header,
1910 .attr_count = IPSET_ATTR_CMD_MAX,
1911 .policy = ip_set_setname_policy,
1913 [IPSET_CMD_TYPE] = {
1914 .call = ip_set_type,
1915 .attr_count = IPSET_ATTR_CMD_MAX,
1916 .policy = ip_set_type_policy,
1918 [IPSET_CMD_PROTOCOL] = {
1919 .call = ip_set_protocol,
1920 .attr_count = IPSET_ATTR_CMD_MAX,
1921 .policy = ip_set_protocol_policy,
1925 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = {
1927 .subsys_id = NFNL_SUBSYS_IPSET,
1928 .cb_count = IPSET_MSG_MAX,
1929 .cb = ip_set_netlink_subsys_cb,
1932 /* Interface to iptables/ip6tables */
1935 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len)
1939 int copylen = *len, ret = 0;
1940 struct net *net = sock_net(sk);
1941 struct ip_set_net *inst = ip_set_pernet(net);
1943 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1945 if (optval != SO_IP_SET)
1947 if (*len < sizeof(unsigned int))
1950 data = vmalloc(*len);
1953 if (copy_from_user(data, user, *len) != 0) {
1959 if (*op < IP_SET_OP_VERSION) {
1960 /* Check the version at the beginning of operations */
1961 struct ip_set_req_version *req_version = data;
1963 if (*len < sizeof(struct ip_set_req_version)) {
1968 if (req_version->version != IPSET_PROTOCOL) {
1975 case IP_SET_OP_VERSION: {
1976 struct ip_set_req_version *req_version = data;
1978 if (*len != sizeof(struct ip_set_req_version)) {
1983 req_version->version = IPSET_PROTOCOL;
1984 if (copy_to_user(user, req_version,
1985 sizeof(struct ip_set_req_version)))
1989 case IP_SET_OP_GET_BYNAME: {
1990 struct ip_set_req_get_set *req_get = data;
1993 if (*len != sizeof(struct ip_set_req_get_set)) {
1997 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
1998 nfnl_lock(NFNL_SUBSYS_IPSET);
1999 find_set_and_id(inst, req_get->set.name, &id);
2000 req_get->set.index = id;
2001 nfnl_unlock(NFNL_SUBSYS_IPSET);
2004 case IP_SET_OP_GET_FNAME: {
2005 struct ip_set_req_get_set_family *req_get = data;
2008 if (*len != sizeof(struct ip_set_req_get_set_family)) {
2012 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
2013 nfnl_lock(NFNL_SUBSYS_IPSET);
2014 find_set_and_id(inst, req_get->set.name, &id);
2015 req_get->set.index = id;
2016 if (id != IPSET_INVALID_ID)
2017 req_get->family = ip_set(inst, id)->family;
2018 nfnl_unlock(NFNL_SUBSYS_IPSET);
2021 case IP_SET_OP_GET_BYINDEX: {
2022 struct ip_set_req_get_set *req_get = data;
2025 if (*len != sizeof(struct ip_set_req_get_set) ||
2026 req_get->set.index >= inst->ip_set_max) {
2030 nfnl_lock(NFNL_SUBSYS_IPSET);
2031 set = ip_set(inst, req_get->set.index);
2032 strncpy(req_get->set.name, set ? set->name : "",
2034 nfnl_unlock(NFNL_SUBSYS_IPSET);
2040 } /* end of switch(op) */
2043 if (copy_to_user(user, data, copylen))
2053 static struct nf_sockopt_ops so_set __read_mostly = {
2055 .get_optmin = SO_IP_SET,
2056 .get_optmax = SO_IP_SET + 1,
2057 .get = ip_set_sockfn_get,
2058 .owner = THIS_MODULE,
2061 static int __net_init
2062 ip_set_net_init(struct net *net)
2064 struct ip_set_net *inst = ip_set_pernet(net);
2065 struct ip_set **list;
2067 inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX;
2068 if (inst->ip_set_max >= IPSET_INVALID_ID)
2069 inst->ip_set_max = IPSET_INVALID_ID - 1;
2071 list = kvcalloc(inst->ip_set_max, sizeof(struct ip_set *), GFP_KERNEL);
2074 inst->is_deleted = false;
2075 inst->is_destroyed = false;
2076 rcu_assign_pointer(inst->ip_set_list, list);
2080 static void __net_exit
2081 ip_set_net_exit(struct net *net)
2083 struct ip_set_net *inst = ip_set_pernet(net);
2085 struct ip_set *set = NULL;
2088 inst->is_deleted = true; /* flag for ip_set_nfnl_put */
2090 nfnl_lock(NFNL_SUBSYS_IPSET);
2091 for (i = 0; i < inst->ip_set_max; i++) {
2092 set = ip_set(inst, i);
2094 ip_set(inst, i) = NULL;
2095 ip_set_destroy_set(set);
2098 nfnl_unlock(NFNL_SUBSYS_IPSET);
2099 kvfree(rcu_dereference_protected(inst->ip_set_list, 1));
2102 static struct pernet_operations ip_set_net_ops = {
2103 .init = ip_set_net_init,
2104 .exit = ip_set_net_exit,
2105 .id = &ip_set_net_id,
2106 .size = sizeof(struct ip_set_net),
2112 int ret = register_pernet_subsys(&ip_set_net_ops);
2115 pr_err("ip_set: cannot register pernet_subsys.\n");
2119 ret = nfnetlink_subsys_register(&ip_set_netlink_subsys);
2121 pr_err("ip_set: cannot register with nfnetlink.\n");
2122 unregister_pernet_subsys(&ip_set_net_ops);
2126 ret = nf_register_sockopt(&so_set);
2128 pr_err("SO_SET registry failed: %d\n", ret);
2129 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2130 unregister_pernet_subsys(&ip_set_net_ops);
2140 nf_unregister_sockopt(&so_set);
2141 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2143 unregister_pernet_subsys(&ip_set_net_ops);
2144 pr_debug("these are the famous last words\n");
2147 module_init(ip_set_init);
2148 module_exit(ip_set_fini);
2150 MODULE_DESCRIPTION("ip_set: protocol " __stringify(IPSET_PROTOCOL));