1 // SPDX-License-Identifier: GPL-2.0-only
2 /* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
3 * Patrick Schaaf <bof@bof.de>
4 * Copyright (C) 2003-2013 Jozsef Kadlecsik <kadlec@netfilter.org>
7 /* Kernel module for IP set management */
9 #include <linux/init.h>
10 #include <linux/module.h>
11 #include <linux/moduleparam.h>
13 #include <linux/skbuff.h>
14 #include <linux/spinlock.h>
15 #include <linux/rculist.h>
16 #include <net/netlink.h>
17 #include <net/net_namespace.h>
18 #include <net/netns/generic.h>
20 #include <linux/netfilter.h>
21 #include <linux/netfilter/x_tables.h>
22 #include <linux/netfilter/nfnetlink.h>
23 #include <linux/netfilter/ipset/ip_set.h>
25 static LIST_HEAD(ip_set_type_list); /* all registered set types */
26 static DEFINE_MUTEX(ip_set_type_mutex); /* protects ip_set_type_list */
27 static DEFINE_RWLOCK(ip_set_ref_lock); /* protects the set refs */
30 struct ip_set * __rcu *ip_set_list; /* all individual sets */
31 ip_set_id_t ip_set_max; /* max number of sets */
32 bool is_deleted; /* deleted by ip_set_net_exit */
33 bool is_destroyed; /* all sets are destroyed */
36 static unsigned int ip_set_net_id __read_mostly;
38 static struct ip_set_net *ip_set_pernet(struct net *net)
40 return net_generic(net, ip_set_net_id);
44 #define STRNCMP(a, b) (strncmp(a, b, IPSET_MAXNAMELEN) == 0)
46 static unsigned int max_sets;
48 module_param(max_sets, int, 0600);
49 MODULE_PARM_DESC(max_sets, "maximal number of sets");
50 MODULE_LICENSE("GPL");
51 MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@netfilter.org>");
52 MODULE_DESCRIPTION("core IP set support");
53 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_IPSET);
55 /* When the nfnl mutex or ip_set_ref_lock is held: */
56 #define ip_set_dereference(p) \
57 rcu_dereference_protected(p, \
58 lockdep_nfnl_is_held(NFNL_SUBSYS_IPSET) || \
59 lockdep_is_held(&ip_set_ref_lock))
60 #define ip_set(inst, id) \
61 ip_set_dereference((inst)->ip_set_list)[id]
62 #define ip_set_ref_netlink(inst,id) \
63 rcu_dereference_raw((inst)->ip_set_list)[id]
65 /* The set types are implemented in modules and registered set types
66 * can be found in ip_set_type_list. Adding/deleting types is
67 * serialized by ip_set_type_mutex.
71 ip_set_type_lock(void)
73 mutex_lock(&ip_set_type_mutex);
77 ip_set_type_unlock(void)
79 mutex_unlock(&ip_set_type_mutex);
82 /* Register and deregister settype */
84 static struct ip_set_type *
85 find_set_type(const char *name, u8 family, u8 revision)
87 struct ip_set_type *type;
89 list_for_each_entry_rcu(type, &ip_set_type_list, list,
90 lockdep_is_held(&ip_set_type_mutex))
91 if (STRNCMP(type->name, name) &&
92 (type->family == family ||
93 type->family == NFPROTO_UNSPEC) &&
94 revision >= type->revision_min &&
95 revision <= type->revision_max)
100 /* Unlock, try to load a set type module and lock again */
102 load_settype(const char *name)
104 nfnl_unlock(NFNL_SUBSYS_IPSET);
105 pr_debug("try to load ip_set_%s\n", name);
106 if (request_module("ip_set_%s", name) < 0) {
107 pr_warn("Can't find ip_set type %s\n", name);
108 nfnl_lock(NFNL_SUBSYS_IPSET);
111 nfnl_lock(NFNL_SUBSYS_IPSET);
115 /* Find a set type and reference it */
116 #define find_set_type_get(name, family, revision, found) \
117 __find_set_type_get(name, family, revision, found, false)
120 __find_set_type_get(const char *name, u8 family, u8 revision,
121 struct ip_set_type **found, bool retry)
123 struct ip_set_type *type;
126 if (retry && !load_settype(name))
127 return -IPSET_ERR_FIND_TYPE;
130 *found = find_set_type(name, family, revision);
132 err = !try_module_get((*found)->me) ? -EFAULT : 0;
135 /* Make sure the type is already loaded
136 * but we don't support the revision
138 list_for_each_entry_rcu(type, &ip_set_type_list, list)
139 if (STRNCMP(type->name, name)) {
140 err = -IPSET_ERR_FIND_TYPE;
145 return retry ? -IPSET_ERR_FIND_TYPE :
146 __find_set_type_get(name, family, revision, found, true);
153 /* Find a given set type by name and family.
154 * If we succeeded, the supported minimal and maximum revisions are
157 #define find_set_type_minmax(name, family, min, max) \
158 __find_set_type_minmax(name, family, min, max, false)
161 __find_set_type_minmax(const char *name, u8 family, u8 *min, u8 *max,
164 struct ip_set_type *type;
167 if (retry && !load_settype(name))
168 return -IPSET_ERR_FIND_TYPE;
170 *min = 255; *max = 0;
172 list_for_each_entry_rcu(type, &ip_set_type_list, list)
173 if (STRNCMP(type->name, name) &&
174 (type->family == family ||
175 type->family == NFPROTO_UNSPEC)) {
177 if (type->revision_min < *min)
178 *min = type->revision_min;
179 if (type->revision_max > *max)
180 *max = type->revision_max;
186 return retry ? -IPSET_ERR_FIND_TYPE :
187 __find_set_type_minmax(name, family, min, max, true);
190 #define family_name(f) ((f) == NFPROTO_IPV4 ? "inet" : \
191 (f) == NFPROTO_IPV6 ? "inet6" : "any")
193 /* Register a set type structure. The type is identified by
194 * the unique triple of name, family and revision.
197 ip_set_type_register(struct ip_set_type *type)
201 if (type->protocol != IPSET_PROTOCOL) {
202 pr_warn("ip_set type %s, family %s, revision %u:%u uses wrong protocol version %u (want %u)\n",
203 type->name, family_name(type->family),
204 type->revision_min, type->revision_max,
205 type->protocol, IPSET_PROTOCOL);
210 if (find_set_type(type->name, type->family, type->revision_min)) {
212 pr_warn("ip_set type %s, family %s with revision min %u already registered!\n",
213 type->name, family_name(type->family),
215 ip_set_type_unlock();
218 list_add_rcu(&type->list, &ip_set_type_list);
219 pr_debug("type %s, family %s, revision %u:%u registered.\n",
220 type->name, family_name(type->family),
221 type->revision_min, type->revision_max);
222 ip_set_type_unlock();
226 EXPORT_SYMBOL_GPL(ip_set_type_register);
228 /* Unregister a set type. There's a small race with ip_set_create */
230 ip_set_type_unregister(struct ip_set_type *type)
233 if (!find_set_type(type->name, type->family, type->revision_min)) {
234 pr_warn("ip_set type %s, family %s with revision min %u not registered\n",
235 type->name, family_name(type->family),
237 ip_set_type_unlock();
240 list_del_rcu(&type->list);
241 pr_debug("type %s, family %s with revision min %u unregistered.\n",
242 type->name, family_name(type->family), type->revision_min);
243 ip_set_type_unlock();
247 EXPORT_SYMBOL_GPL(ip_set_type_unregister);
249 /* Utility functions */
251 ip_set_alloc(size_t size)
253 return kvzalloc(size, GFP_KERNEL_ACCOUNT);
255 EXPORT_SYMBOL_GPL(ip_set_alloc);
258 ip_set_free(void *members)
260 pr_debug("%p: free with %s\n", members,
261 is_vmalloc_addr(members) ? "vfree" : "kfree");
264 EXPORT_SYMBOL_GPL(ip_set_free);
267 flag_nested(const struct nlattr *nla)
269 return nla->nla_type & NLA_F_NESTED;
272 static const struct nla_policy ipaddr_policy[IPSET_ATTR_IPADDR_MAX + 1] = {
273 [IPSET_ATTR_IPADDR_IPV4] = { .type = NLA_U32 },
274 [IPSET_ATTR_IPADDR_IPV6] = NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
278 ip_set_get_ipaddr4(struct nlattr *nla, __be32 *ipaddr)
280 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1];
282 if (unlikely(!flag_nested(nla)))
283 return -IPSET_ERR_PROTOCOL;
284 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla,
285 ipaddr_policy, NULL))
286 return -IPSET_ERR_PROTOCOL;
287 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV4)))
288 return -IPSET_ERR_PROTOCOL;
290 *ipaddr = nla_get_be32(tb[IPSET_ATTR_IPADDR_IPV4]);
293 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr4);
296 ip_set_get_ipaddr6(struct nlattr *nla, union nf_inet_addr *ipaddr)
298 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1];
300 if (unlikely(!flag_nested(nla)))
301 return -IPSET_ERR_PROTOCOL;
303 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla,
304 ipaddr_policy, NULL))
305 return -IPSET_ERR_PROTOCOL;
306 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV6)))
307 return -IPSET_ERR_PROTOCOL;
309 memcpy(ipaddr, nla_data(tb[IPSET_ATTR_IPADDR_IPV6]),
310 sizeof(struct in6_addr));
313 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr6);
316 ip_set_timeout_get(const unsigned long *timeout)
320 if (*timeout == IPSET_ELEM_PERMANENT)
323 t = jiffies_to_msecs(*timeout - jiffies) / MSEC_PER_SEC;
324 /* Zero value in userspace means no timeout */
325 return t == 0 ? 1 : t;
329 ip_set_comment_uget(struct nlattr *tb)
334 /* Called from uadd only, protected by the set spinlock.
335 * The kadt functions don't use the comment extensions in any way.
338 ip_set_init_comment(struct ip_set *set, struct ip_set_comment *comment,
339 const struct ip_set_ext *ext)
341 struct ip_set_comment_rcu *c = rcu_dereference_protected(comment->c, 1);
342 size_t len = ext->comment ? strlen(ext->comment) : 0;
345 set->ext_size -= sizeof(*c) + strlen(c->str) + 1;
347 rcu_assign_pointer(comment->c, NULL);
351 if (unlikely(len > IPSET_MAX_COMMENT_SIZE))
352 len = IPSET_MAX_COMMENT_SIZE;
353 c = kmalloc(sizeof(*c) + len + 1, GFP_ATOMIC);
356 strlcpy(c->str, ext->comment, len + 1);
357 set->ext_size += sizeof(*c) + strlen(c->str) + 1;
358 rcu_assign_pointer(comment->c, c);
360 EXPORT_SYMBOL_GPL(ip_set_init_comment);
362 /* Used only when dumping a set, protected by rcu_read_lock() */
364 ip_set_put_comment(struct sk_buff *skb, const struct ip_set_comment *comment)
366 struct ip_set_comment_rcu *c = rcu_dereference(comment->c);
370 return nla_put_string(skb, IPSET_ATTR_COMMENT, c->str);
373 /* Called from uadd/udel, flush or the garbage collectors protected
374 * by the set spinlock.
375 * Called when the set is destroyed and when there can't be any user
376 * of the set data anymore.
379 ip_set_comment_free(struct ip_set *set, void *ptr)
381 struct ip_set_comment *comment = ptr;
382 struct ip_set_comment_rcu *c;
384 c = rcu_dereference_protected(comment->c, 1);
387 set->ext_size -= sizeof(*c) + strlen(c->str) + 1;
389 rcu_assign_pointer(comment->c, NULL);
392 typedef void (*destroyer)(struct ip_set *, void *);
393 /* ipset data extension types, in size order */
395 const struct ip_set_ext_type ip_set_extensions[] = {
396 [IPSET_EXT_ID_COUNTER] = {
397 .type = IPSET_EXT_COUNTER,
398 .flag = IPSET_FLAG_WITH_COUNTERS,
399 .len = sizeof(struct ip_set_counter),
400 .align = __alignof__(struct ip_set_counter),
402 [IPSET_EXT_ID_TIMEOUT] = {
403 .type = IPSET_EXT_TIMEOUT,
404 .len = sizeof(unsigned long),
405 .align = __alignof__(unsigned long),
407 [IPSET_EXT_ID_SKBINFO] = {
408 .type = IPSET_EXT_SKBINFO,
409 .flag = IPSET_FLAG_WITH_SKBINFO,
410 .len = sizeof(struct ip_set_skbinfo),
411 .align = __alignof__(struct ip_set_skbinfo),
413 [IPSET_EXT_ID_COMMENT] = {
414 .type = IPSET_EXT_COMMENT | IPSET_EXT_DESTROY,
415 .flag = IPSET_FLAG_WITH_COMMENT,
416 .len = sizeof(struct ip_set_comment),
417 .align = __alignof__(struct ip_set_comment),
418 .destroy = ip_set_comment_free,
421 EXPORT_SYMBOL_GPL(ip_set_extensions);
424 add_extension(enum ip_set_ext_id id, u32 flags, struct nlattr *tb[])
426 return ip_set_extensions[id].flag ?
427 (flags & ip_set_extensions[id].flag) :
428 !!tb[IPSET_ATTR_TIMEOUT];
432 ip_set_elem_len(struct ip_set *set, struct nlattr *tb[], size_t len,
435 enum ip_set_ext_id id;
438 if (tb[IPSET_ATTR_CADT_FLAGS])
439 cadt_flags = ip_set_get_h32(tb[IPSET_ATTR_CADT_FLAGS]);
440 if (cadt_flags & IPSET_FLAG_WITH_FORCEADD)
441 set->flags |= IPSET_CREATE_FLAG_FORCEADD;
444 for (id = 0; id < IPSET_EXT_ID_MAX; id++) {
445 if (!add_extension(id, cadt_flags, tb))
447 if (align < ip_set_extensions[id].align)
448 align = ip_set_extensions[id].align;
449 len = ALIGN(len, ip_set_extensions[id].align);
450 set->offset[id] = len;
451 set->extensions |= ip_set_extensions[id].type;
452 len += ip_set_extensions[id].len;
454 return ALIGN(len, align);
456 EXPORT_SYMBOL_GPL(ip_set_elem_len);
459 ip_set_get_extensions(struct ip_set *set, struct nlattr *tb[],
460 struct ip_set_ext *ext)
464 if (unlikely(!ip_set_optattr_netorder(tb, IPSET_ATTR_TIMEOUT) ||
465 !ip_set_optattr_netorder(tb, IPSET_ATTR_PACKETS) ||
466 !ip_set_optattr_netorder(tb, IPSET_ATTR_BYTES) ||
467 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBMARK) ||
468 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBPRIO) ||
469 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBQUEUE)))
470 return -IPSET_ERR_PROTOCOL;
472 if (tb[IPSET_ATTR_TIMEOUT]) {
473 if (!SET_WITH_TIMEOUT(set))
474 return -IPSET_ERR_TIMEOUT;
475 ext->timeout = ip_set_timeout_uget(tb[IPSET_ATTR_TIMEOUT]);
477 if (tb[IPSET_ATTR_BYTES] || tb[IPSET_ATTR_PACKETS]) {
478 if (!SET_WITH_COUNTER(set))
479 return -IPSET_ERR_COUNTER;
480 if (tb[IPSET_ATTR_BYTES])
481 ext->bytes = be64_to_cpu(nla_get_be64(
482 tb[IPSET_ATTR_BYTES]));
483 if (tb[IPSET_ATTR_PACKETS])
484 ext->packets = be64_to_cpu(nla_get_be64(
485 tb[IPSET_ATTR_PACKETS]));
487 if (tb[IPSET_ATTR_COMMENT]) {
488 if (!SET_WITH_COMMENT(set))
489 return -IPSET_ERR_COMMENT;
490 ext->comment = ip_set_comment_uget(tb[IPSET_ATTR_COMMENT]);
492 if (tb[IPSET_ATTR_SKBMARK]) {
493 if (!SET_WITH_SKBINFO(set))
494 return -IPSET_ERR_SKBINFO;
495 fullmark = be64_to_cpu(nla_get_be64(tb[IPSET_ATTR_SKBMARK]));
496 ext->skbinfo.skbmark = fullmark >> 32;
497 ext->skbinfo.skbmarkmask = fullmark & 0xffffffff;
499 if (tb[IPSET_ATTR_SKBPRIO]) {
500 if (!SET_WITH_SKBINFO(set))
501 return -IPSET_ERR_SKBINFO;
502 ext->skbinfo.skbprio =
503 be32_to_cpu(nla_get_be32(tb[IPSET_ATTR_SKBPRIO]));
505 if (tb[IPSET_ATTR_SKBQUEUE]) {
506 if (!SET_WITH_SKBINFO(set))
507 return -IPSET_ERR_SKBINFO;
508 ext->skbinfo.skbqueue =
509 be16_to_cpu(nla_get_be16(tb[IPSET_ATTR_SKBQUEUE]));
513 EXPORT_SYMBOL_GPL(ip_set_get_extensions);
516 ip_set_get_bytes(const struct ip_set_counter *counter)
518 return (u64)atomic64_read(&(counter)->bytes);
522 ip_set_get_packets(const struct ip_set_counter *counter)
524 return (u64)atomic64_read(&(counter)->packets);
528 ip_set_put_counter(struct sk_buff *skb, const struct ip_set_counter *counter)
530 return nla_put_net64(skb, IPSET_ATTR_BYTES,
531 cpu_to_be64(ip_set_get_bytes(counter)),
533 nla_put_net64(skb, IPSET_ATTR_PACKETS,
534 cpu_to_be64(ip_set_get_packets(counter)),
539 ip_set_put_skbinfo(struct sk_buff *skb, const struct ip_set_skbinfo *skbinfo)
541 /* Send nonzero parameters only */
542 return ((skbinfo->skbmark || skbinfo->skbmarkmask) &&
543 nla_put_net64(skb, IPSET_ATTR_SKBMARK,
544 cpu_to_be64((u64)skbinfo->skbmark << 32 |
545 skbinfo->skbmarkmask),
548 nla_put_net32(skb, IPSET_ATTR_SKBPRIO,
549 cpu_to_be32(skbinfo->skbprio))) ||
550 (skbinfo->skbqueue &&
551 nla_put_net16(skb, IPSET_ATTR_SKBQUEUE,
552 cpu_to_be16(skbinfo->skbqueue)));
556 ip_set_put_extensions(struct sk_buff *skb, const struct ip_set *set,
557 const void *e, bool active)
559 if (SET_WITH_TIMEOUT(set)) {
560 unsigned long *timeout = ext_timeout(e, set);
562 if (nla_put_net32(skb, IPSET_ATTR_TIMEOUT,
563 htonl(active ? ip_set_timeout_get(timeout)
567 if (SET_WITH_COUNTER(set) &&
568 ip_set_put_counter(skb, ext_counter(e, set)))
570 if (SET_WITH_COMMENT(set) &&
571 ip_set_put_comment(skb, ext_comment(e, set)))
573 if (SET_WITH_SKBINFO(set) &&
574 ip_set_put_skbinfo(skb, ext_skbinfo(e, set)))
578 EXPORT_SYMBOL_GPL(ip_set_put_extensions);
581 ip_set_match_counter(u64 counter, u64 match, u8 op)
584 case IPSET_COUNTER_NONE:
586 case IPSET_COUNTER_EQ:
587 return counter == match;
588 case IPSET_COUNTER_NE:
589 return counter != match;
590 case IPSET_COUNTER_LT:
591 return counter < match;
592 case IPSET_COUNTER_GT:
593 return counter > match;
599 ip_set_add_bytes(u64 bytes, struct ip_set_counter *counter)
601 atomic64_add((long long)bytes, &(counter)->bytes);
605 ip_set_add_packets(u64 packets, struct ip_set_counter *counter)
607 atomic64_add((long long)packets, &(counter)->packets);
611 ip_set_update_counter(struct ip_set_counter *counter,
612 const struct ip_set_ext *ext, u32 flags)
614 if (ext->packets != ULLONG_MAX &&
615 !(flags & IPSET_FLAG_SKIP_COUNTER_UPDATE)) {
616 ip_set_add_bytes(ext->bytes, counter);
617 ip_set_add_packets(ext->packets, counter);
622 ip_set_get_skbinfo(struct ip_set_skbinfo *skbinfo,
623 const struct ip_set_ext *ext,
624 struct ip_set_ext *mext, u32 flags)
626 mext->skbinfo = *skbinfo;
630 ip_set_match_extensions(struct ip_set *set, const struct ip_set_ext *ext,
631 struct ip_set_ext *mext, u32 flags, void *data)
633 if (SET_WITH_TIMEOUT(set) &&
634 ip_set_timeout_expired(ext_timeout(data, set)))
636 if (SET_WITH_COUNTER(set)) {
637 struct ip_set_counter *counter = ext_counter(data, set);
639 ip_set_update_counter(counter, ext, flags);
641 if (flags & IPSET_FLAG_MATCH_COUNTERS &&
642 !(ip_set_match_counter(ip_set_get_packets(counter),
643 mext->packets, mext->packets_op) &&
644 ip_set_match_counter(ip_set_get_bytes(counter),
645 mext->bytes, mext->bytes_op)))
648 if (SET_WITH_SKBINFO(set))
649 ip_set_get_skbinfo(ext_skbinfo(data, set),
653 EXPORT_SYMBOL_GPL(ip_set_match_extensions);
655 /* Creating/destroying/renaming/swapping affect the existence and
656 * the properties of a set. All of these can be executed from userspace
657 * only and serialized by the nfnl mutex indirectly from nfnetlink.
659 * Sets are identified by their index in ip_set_list and the index
660 * is used by the external references (set/SET netfilter modules).
662 * The set behind an index may change by swapping only, from userspace.
666 __ip_set_get(struct ip_set *set)
668 write_lock_bh(&ip_set_ref_lock);
670 write_unlock_bh(&ip_set_ref_lock);
674 __ip_set_put(struct ip_set *set)
676 write_lock_bh(&ip_set_ref_lock);
677 BUG_ON(set->ref == 0);
679 write_unlock_bh(&ip_set_ref_lock);
682 /* set->ref can be swapped out by ip_set_swap, netlink events (like dump) need
683 * a separate reference counter
686 __ip_set_put_netlink(struct ip_set *set)
688 write_lock_bh(&ip_set_ref_lock);
689 BUG_ON(set->ref_netlink == 0);
691 write_unlock_bh(&ip_set_ref_lock);
694 /* Add, del and test set entries from kernel.
696 * The set behind the index must exist and must be referenced
697 * so it can't be destroyed (or changed) under our foot.
700 static struct ip_set *
701 ip_set_rcu_get(struct net *net, ip_set_id_t index)
704 struct ip_set_net *inst = ip_set_pernet(net);
707 /* ip_set_list itself needs to be protected */
708 set = rcu_dereference(inst->ip_set_list)[index];
715 ip_set_lock(struct ip_set *set)
717 if (!set->variant->region_lock)
718 spin_lock_bh(&set->lock);
722 ip_set_unlock(struct ip_set *set)
724 if (!set->variant->region_lock)
725 spin_unlock_bh(&set->lock);
729 ip_set_test(ip_set_id_t index, const struct sk_buff *skb,
730 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
732 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
736 pr_debug("set %s, index %u\n", set->name, index);
738 if (opt->dim < set->type->dimension ||
739 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
743 ret = set->variant->kadt(set, skb, par, IPSET_TEST, opt);
744 rcu_read_unlock_bh();
746 if (ret == -EAGAIN) {
747 /* Type requests element to be completed */
748 pr_debug("element must be completed, ADD is triggered\n");
750 set->variant->kadt(set, skb, par, IPSET_ADD, opt);
754 /* --return-nomatch: invert matched element */
755 if ((opt->cmdflags & IPSET_FLAG_RETURN_NOMATCH) &&
756 (set->type->features & IPSET_TYPE_NOMATCH) &&
757 (ret > 0 || ret == -ENOTEMPTY))
761 /* Convert error codes to nomatch */
762 return (ret < 0 ? 0 : ret);
764 EXPORT_SYMBOL_GPL(ip_set_test);
767 ip_set_add(ip_set_id_t index, const struct sk_buff *skb,
768 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
770 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
774 pr_debug("set %s, index %u\n", set->name, index);
776 if (opt->dim < set->type->dimension ||
777 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
778 return -IPSET_ERR_TYPE_MISMATCH;
781 ret = set->variant->kadt(set, skb, par, IPSET_ADD, opt);
786 EXPORT_SYMBOL_GPL(ip_set_add);
789 ip_set_del(ip_set_id_t index, const struct sk_buff *skb,
790 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
792 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
796 pr_debug("set %s, index %u\n", set->name, index);
798 if (opt->dim < set->type->dimension ||
799 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
800 return -IPSET_ERR_TYPE_MISMATCH;
803 ret = set->variant->kadt(set, skb, par, IPSET_DEL, opt);
808 EXPORT_SYMBOL_GPL(ip_set_del);
810 /* Find set by name, reference it once. The reference makes sure the
811 * thing pointed to, does not go away under our feet.
815 ip_set_get_byname(struct net *net, const char *name, struct ip_set **set)
817 ip_set_id_t i, index = IPSET_INVALID_ID;
819 struct ip_set_net *inst = ip_set_pernet(net);
822 for (i = 0; i < inst->ip_set_max; i++) {
823 s = rcu_dereference(inst->ip_set_list)[i];
824 if (s && STRNCMP(s->name, name)) {
835 EXPORT_SYMBOL_GPL(ip_set_get_byname);
837 /* If the given set pointer points to a valid set, decrement
838 * reference count by 1. The caller shall not assume the index
839 * to be valid, after calling this function.
844 __ip_set_put_byindex(struct ip_set_net *inst, ip_set_id_t index)
849 set = rcu_dereference(inst->ip_set_list)[index];
856 ip_set_put_byindex(struct net *net, ip_set_id_t index)
858 struct ip_set_net *inst = ip_set_pernet(net);
860 __ip_set_put_byindex(inst, index);
862 EXPORT_SYMBOL_GPL(ip_set_put_byindex);
864 /* Get the name of a set behind a set index.
865 * Set itself is protected by RCU, but its name isn't: to protect against
866 * renaming, grab ip_set_ref_lock as reader (see ip_set_rename()) and copy the
870 ip_set_name_byindex(struct net *net, ip_set_id_t index, char *name)
872 struct ip_set *set = ip_set_rcu_get(net, index);
876 read_lock_bh(&ip_set_ref_lock);
877 strncpy(name, set->name, IPSET_MAXNAMELEN);
878 read_unlock_bh(&ip_set_ref_lock);
880 EXPORT_SYMBOL_GPL(ip_set_name_byindex);
882 /* Routines to call by external subsystems, which do not
883 * call nfnl_lock for us.
886 /* Find set by index, reference it once. The reference makes sure the
887 * thing pointed to, does not go away under our feet.
889 * The nfnl mutex is used in the function.
892 ip_set_nfnl_get_byindex(struct net *net, ip_set_id_t index)
895 struct ip_set_net *inst = ip_set_pernet(net);
897 if (index >= inst->ip_set_max)
898 return IPSET_INVALID_ID;
900 nfnl_lock(NFNL_SUBSYS_IPSET);
901 set = ip_set(inst, index);
905 index = IPSET_INVALID_ID;
906 nfnl_unlock(NFNL_SUBSYS_IPSET);
910 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex);
912 /* If the given set pointer points to a valid set, decrement
913 * reference count by 1. The caller shall not assume the index
914 * to be valid, after calling this function.
916 * The nfnl mutex is used in the function.
919 ip_set_nfnl_put(struct net *net, ip_set_id_t index)
922 struct ip_set_net *inst = ip_set_pernet(net);
924 nfnl_lock(NFNL_SUBSYS_IPSET);
925 if (!inst->is_deleted) { /* already deleted from ip_set_net_exit() */
926 set = ip_set(inst, index);
930 nfnl_unlock(NFNL_SUBSYS_IPSET);
932 EXPORT_SYMBOL_GPL(ip_set_nfnl_put);
934 /* Communication protocol with userspace over netlink.
936 * The commands are serialized by the nfnl mutex.
939 static inline u8 protocol(const struct nlattr * const tb[])
941 return nla_get_u8(tb[IPSET_ATTR_PROTOCOL]);
945 protocol_failed(const struct nlattr * const tb[])
947 return !tb[IPSET_ATTR_PROTOCOL] || protocol(tb) != IPSET_PROTOCOL;
951 protocol_min_failed(const struct nlattr * const tb[])
953 return !tb[IPSET_ATTR_PROTOCOL] || protocol(tb) < IPSET_PROTOCOL_MIN;
957 flag_exist(const struct nlmsghdr *nlh)
959 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST;
962 static struct nlmsghdr *
963 start_msg(struct sk_buff *skb, u32 portid, u32 seq, unsigned int flags,
966 struct nlmsghdr *nlh;
967 struct nfgenmsg *nfmsg;
969 nlh = nlmsg_put(skb, portid, seq, nfnl_msg_type(NFNL_SUBSYS_IPSET, cmd),
970 sizeof(*nfmsg), flags);
974 nfmsg = nlmsg_data(nlh);
975 nfmsg->nfgen_family = NFPROTO_IPV4;
976 nfmsg->version = NFNETLINK_V0;
984 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = {
985 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
986 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
987 .len = IPSET_MAXNAMELEN - 1 },
988 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
989 .len = IPSET_MAXNAMELEN - 1},
990 [IPSET_ATTR_REVISION] = { .type = NLA_U8 },
991 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
992 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
995 static struct ip_set *
996 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id)
998 struct ip_set *set = NULL;
1001 *id = IPSET_INVALID_ID;
1002 for (i = 0; i < inst->ip_set_max; i++) {
1003 set = ip_set(inst, i);
1004 if (set && STRNCMP(set->name, name)) {
1009 return (*id == IPSET_INVALID_ID ? NULL : set);
1012 static inline struct ip_set *
1013 find_set(struct ip_set_net *inst, const char *name)
1017 return find_set_and_id(inst, name, &id);
1021 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index,
1022 struct ip_set **set)
1027 *index = IPSET_INVALID_ID;
1028 for (i = 0; i < inst->ip_set_max; i++) {
1029 s = ip_set(inst, i);
1031 if (*index == IPSET_INVALID_ID)
1033 } else if (STRNCMP(name, s->name)) {
1039 if (*index == IPSET_INVALID_ID)
1040 /* No free slot remained */
1041 return -IPSET_ERR_MAX_SETS;
1045 static int ip_set_none(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1046 const struct nlmsghdr *nlh,
1047 const struct nlattr * const attr[],
1048 struct netlink_ext_ack *extack)
1053 static int ip_set_create(struct net *net, struct sock *ctnl,
1054 struct sk_buff *skb, const struct nlmsghdr *nlh,
1055 const struct nlattr * const attr[],
1056 struct netlink_ext_ack *extack)
1058 struct ip_set_net *inst = ip_set_pernet(net);
1059 struct ip_set *set, *clash = NULL;
1060 ip_set_id_t index = IPSET_INVALID_ID;
1061 struct nlattr *tb[IPSET_ATTR_CREATE_MAX + 1] = {};
1062 const char *name, *typename;
1063 u8 family, revision;
1064 u32 flags = flag_exist(nlh);
1067 if (unlikely(protocol_min_failed(attr) ||
1068 !attr[IPSET_ATTR_SETNAME] ||
1069 !attr[IPSET_ATTR_TYPENAME] ||
1070 !attr[IPSET_ATTR_REVISION] ||
1071 !attr[IPSET_ATTR_FAMILY] ||
1072 (attr[IPSET_ATTR_DATA] &&
1073 !flag_nested(attr[IPSET_ATTR_DATA]))))
1074 return -IPSET_ERR_PROTOCOL;
1076 name = nla_data(attr[IPSET_ATTR_SETNAME]);
1077 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1078 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1079 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]);
1080 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
1081 name, typename, family_name(family), revision);
1083 /* First, and without any locks, allocate and initialize
1084 * a normal base set structure.
1086 set = kzalloc(sizeof(*set), GFP_KERNEL);
1089 spin_lock_init(&set->lock);
1090 strlcpy(set->name, name, IPSET_MAXNAMELEN);
1091 set->family = family;
1092 set->revision = revision;
1094 /* Next, check that we know the type, and take
1095 * a reference on the type, to make sure it stays available
1096 * while constructing our new set.
1098 * After referencing the type, we try to create the type
1099 * specific part of the set without holding any locks.
1101 ret = find_set_type_get(typename, family, revision, &set->type);
1105 /* Without holding any locks, create private part. */
1106 if (attr[IPSET_ATTR_DATA] &&
1107 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA],
1108 set->type->create_policy, NULL)) {
1109 ret = -IPSET_ERR_PROTOCOL;
1113 ret = set->type->create(net, set, tb, flags);
1117 /* BTW, ret==0 here. */
1119 /* Here, we have a valid, constructed set and we are protected
1120 * by the nfnl mutex. Find the first free index in ip_set_list
1121 * and check clashing.
1123 ret = find_free_id(inst, set->name, &index, &clash);
1124 if (ret == -EEXIST) {
1125 /* If this is the same set and requested, ignore error */
1126 if ((flags & IPSET_FLAG_EXIST) &&
1127 STRNCMP(set->type->name, clash->type->name) &&
1128 set->type->family == clash->type->family &&
1129 set->type->revision_min == clash->type->revision_min &&
1130 set->type->revision_max == clash->type->revision_max &&
1131 set->variant->same_set(set, clash))
1134 } else if (ret == -IPSET_ERR_MAX_SETS) {
1135 struct ip_set **list, **tmp;
1136 ip_set_id_t i = inst->ip_set_max + IP_SET_INC;
1138 if (i < inst->ip_set_max || i == IPSET_INVALID_ID)
1142 list = kvcalloc(i, sizeof(struct ip_set *), GFP_KERNEL);
1145 /* nfnl mutex is held, both lists are valid */
1146 tmp = ip_set_dereference(inst->ip_set_list);
1147 memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max);
1148 rcu_assign_pointer(inst->ip_set_list, list);
1149 /* Make sure all current packets have passed through */
1152 index = inst->ip_set_max;
1153 inst->ip_set_max = i;
1160 /* Finally! Add our shiny new set to the list, and be done. */
1161 pr_debug("create: '%s' created with index %u!\n", set->name, index);
1162 ip_set(inst, index) = set;
1167 set->variant->destroy(set);
1169 module_put(set->type->me);
1177 static const struct nla_policy
1178 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = {
1179 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1180 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1181 .len = IPSET_MAXNAMELEN - 1 },
1185 ip_set_destroy_set(struct ip_set *set)
1187 pr_debug("set: %s\n", set->name);
1189 /* Must call it without holding any lock */
1190 set->variant->destroy(set);
1191 module_put(set->type->me);
1195 static int ip_set_destroy(struct net *net, struct sock *ctnl,
1196 struct sk_buff *skb, const struct nlmsghdr *nlh,
1197 const struct nlattr * const attr[],
1198 struct netlink_ext_ack *extack)
1200 struct ip_set_net *inst = ip_set_pernet(net);
1205 if (unlikely(protocol_min_failed(attr)))
1206 return -IPSET_ERR_PROTOCOL;
1208 /* Must wait for flush to be really finished in list:set */
1211 /* Commands are serialized and references are
1212 * protected by the ip_set_ref_lock.
1213 * External systems (i.e. xt_set) must call
1214 * ip_set_put|get_nfnl_* functions, that way we
1215 * can safely check references here.
1217 * list:set timer can only decrement the reference
1218 * counter, so if it's already zero, we can proceed
1219 * without holding the lock.
1221 read_lock_bh(&ip_set_ref_lock);
1222 if (!attr[IPSET_ATTR_SETNAME]) {
1223 for (i = 0; i < inst->ip_set_max; i++) {
1224 s = ip_set(inst, i);
1225 if (s && (s->ref || s->ref_netlink)) {
1226 ret = -IPSET_ERR_BUSY;
1230 inst->is_destroyed = true;
1231 read_unlock_bh(&ip_set_ref_lock);
1232 for (i = 0; i < inst->ip_set_max; i++) {
1233 s = ip_set(inst, i);
1235 ip_set(inst, i) = NULL;
1236 ip_set_destroy_set(s);
1239 /* Modified by ip_set_destroy() only, which is serialized */
1240 inst->is_destroyed = false;
1242 s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1247 } else if (s->ref || s->ref_netlink) {
1248 ret = -IPSET_ERR_BUSY;
1251 ip_set(inst, i) = NULL;
1252 read_unlock_bh(&ip_set_ref_lock);
1254 ip_set_destroy_set(s);
1258 read_unlock_bh(&ip_set_ref_lock);
1265 ip_set_flush_set(struct ip_set *set)
1267 pr_debug("set: %s\n", set->name);
1270 set->variant->flush(set);
1274 static int ip_set_flush(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1275 const struct nlmsghdr *nlh,
1276 const struct nlattr * const attr[],
1277 struct netlink_ext_ack *extack)
1279 struct ip_set_net *inst = ip_set_pernet(net);
1283 if (unlikely(protocol_min_failed(attr)))
1284 return -IPSET_ERR_PROTOCOL;
1286 if (!attr[IPSET_ATTR_SETNAME]) {
1287 for (i = 0; i < inst->ip_set_max; i++) {
1288 s = ip_set(inst, i);
1290 ip_set_flush_set(s);
1293 s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1297 ip_set_flush_set(s);
1305 static const struct nla_policy
1306 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = {
1307 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1308 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1309 .len = IPSET_MAXNAMELEN - 1 },
1310 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING,
1311 .len = IPSET_MAXNAMELEN - 1 },
1314 static int ip_set_rename(struct net *net, struct sock *ctnl,
1315 struct sk_buff *skb, const struct nlmsghdr *nlh,
1316 const struct nlattr * const attr[],
1317 struct netlink_ext_ack *extack)
1319 struct ip_set_net *inst = ip_set_pernet(net);
1320 struct ip_set *set, *s;
1325 if (unlikely(protocol_min_failed(attr) ||
1326 !attr[IPSET_ATTR_SETNAME] ||
1327 !attr[IPSET_ATTR_SETNAME2]))
1328 return -IPSET_ERR_PROTOCOL;
1330 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1334 write_lock_bh(&ip_set_ref_lock);
1335 if (set->ref != 0 || set->ref_netlink != 0) {
1336 ret = -IPSET_ERR_REFERENCED;
1340 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]);
1341 for (i = 0; i < inst->ip_set_max; i++) {
1342 s = ip_set(inst, i);
1343 if (s && STRNCMP(s->name, name2)) {
1344 ret = -IPSET_ERR_EXIST_SETNAME2;
1348 strncpy(set->name, name2, IPSET_MAXNAMELEN);
1351 write_unlock_bh(&ip_set_ref_lock);
1355 /* Swap two sets so that name/index points to the other.
1356 * References and set names are also swapped.
1358 * The commands are serialized by the nfnl mutex and references are
1359 * protected by the ip_set_ref_lock. The kernel interfaces
1360 * do not hold the mutex but the pointer settings are atomic
1361 * so the ip_set_list always contains valid pointers to the sets.
1364 static int ip_set_swap(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1365 const struct nlmsghdr *nlh,
1366 const struct nlattr * const attr[],
1367 struct netlink_ext_ack *extack)
1369 struct ip_set_net *inst = ip_set_pernet(net);
1370 struct ip_set *from, *to;
1371 ip_set_id_t from_id, to_id;
1372 char from_name[IPSET_MAXNAMELEN];
1374 if (unlikely(protocol_min_failed(attr) ||
1375 !attr[IPSET_ATTR_SETNAME] ||
1376 !attr[IPSET_ATTR_SETNAME2]))
1377 return -IPSET_ERR_PROTOCOL;
1379 from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1384 to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]),
1387 return -IPSET_ERR_EXIST_SETNAME2;
1389 /* Features must not change.
1390 * Not an artifical restriction anymore, as we must prevent
1391 * possible loops created by swapping in setlist type of sets.
1393 if (!(from->type->features == to->type->features &&
1394 from->family == to->family))
1395 return -IPSET_ERR_TYPE_MISMATCH;
1397 write_lock_bh(&ip_set_ref_lock);
1399 if (from->ref_netlink || to->ref_netlink) {
1400 write_unlock_bh(&ip_set_ref_lock);
1404 strncpy(from_name, from->name, IPSET_MAXNAMELEN);
1405 strncpy(from->name, to->name, IPSET_MAXNAMELEN);
1406 strncpy(to->name, from_name, IPSET_MAXNAMELEN);
1408 swap(from->ref, to->ref);
1409 ip_set(inst, from_id) = to;
1410 ip_set(inst, to_id) = from;
1411 write_unlock_bh(&ip_set_ref_lock);
1416 /* List/save set data */
1423 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1424 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1427 ip_set_put_flags(struct sk_buff *skb, struct ip_set *set)
1431 if (SET_WITH_TIMEOUT(set))
1432 if (unlikely(nla_put_net32(skb, IPSET_ATTR_TIMEOUT,
1433 htonl(set->timeout))))
1435 if (SET_WITH_COUNTER(set))
1436 cadt_flags |= IPSET_FLAG_WITH_COUNTERS;
1437 if (SET_WITH_COMMENT(set))
1438 cadt_flags |= IPSET_FLAG_WITH_COMMENT;
1439 if (SET_WITH_SKBINFO(set))
1440 cadt_flags |= IPSET_FLAG_WITH_SKBINFO;
1441 if (SET_WITH_FORCEADD(set))
1442 cadt_flags |= IPSET_FLAG_WITH_FORCEADD;
1446 return nla_put_net32(skb, IPSET_ATTR_CADT_FLAGS, htonl(cadt_flags));
1448 EXPORT_SYMBOL_GPL(ip_set_put_flags);
1451 ip_set_dump_done(struct netlink_callback *cb)
1453 if (cb->args[IPSET_CB_ARG0]) {
1454 struct ip_set_net *inst =
1455 (struct ip_set_net *)cb->args[IPSET_CB_NET];
1456 ip_set_id_t index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1457 struct ip_set *set = ip_set_ref_netlink(inst, index);
1459 if (set->variant->uref)
1460 set->variant->uref(set, cb, false);
1461 pr_debug("release set %s\n", set->name);
1462 __ip_set_put_netlink(set);
1468 dump_attrs(struct nlmsghdr *nlh)
1470 const struct nlattr *attr;
1473 pr_debug("dump nlmsg\n");
1474 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) {
1475 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len);
1479 static const struct nla_policy
1480 ip_set_dump_policy[IPSET_ATTR_CMD_MAX + 1] = {
1481 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1482 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1483 .len = IPSET_MAXNAMELEN - 1 },
1484 [IPSET_ATTR_FLAGS] = { .type = NLA_U32 },
1488 ip_set_dump_start(struct netlink_callback *cb)
1490 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb);
1491 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1492 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1493 struct nlattr *attr = (void *)nlh + min_len;
1494 struct sk_buff *skb = cb->skb;
1495 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1499 ret = nla_parse(cda, IPSET_ATTR_CMD_MAX, attr,
1500 nlh->nlmsg_len - min_len,
1501 ip_set_dump_policy, NULL);
1505 cb->args[IPSET_CB_PROTO] = nla_get_u8(cda[IPSET_ATTR_PROTOCOL]);
1506 if (cda[IPSET_ATTR_SETNAME]) {
1510 set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]),
1516 dump_type = DUMP_ONE;
1517 cb->args[IPSET_CB_INDEX] = index;
1519 dump_type = DUMP_ALL;
1522 if (cda[IPSET_ATTR_FLAGS]) {
1523 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]);
1525 dump_type |= (f << 16);
1527 cb->args[IPSET_CB_NET] = (unsigned long)inst;
1528 cb->args[IPSET_CB_DUMP] = dump_type;
1533 /* We have to create and send the error message manually :-( */
1534 if (nlh->nlmsg_flags & NLM_F_ACK) {
1535 netlink_ack(cb->skb, nlh, ret, NULL);
1541 ip_set_dump_do(struct sk_buff *skb, struct netlink_callback *cb)
1543 ip_set_id_t index = IPSET_INVALID_ID, max;
1544 struct ip_set *set = NULL;
1545 struct nlmsghdr *nlh = NULL;
1546 unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0;
1547 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1548 u32 dump_type, dump_flags;
1552 if (!cb->args[IPSET_CB_DUMP])
1555 if (cb->args[IPSET_CB_INDEX] >= inst->ip_set_max)
1558 dump_type = DUMP_TYPE(cb->args[IPSET_CB_DUMP]);
1559 dump_flags = DUMP_FLAGS(cb->args[IPSET_CB_DUMP]);
1560 max = dump_type == DUMP_ONE ? cb->args[IPSET_CB_INDEX] + 1
1563 pr_debug("dump type, flag: %u %u index: %ld\n",
1564 dump_type, dump_flags, cb->args[IPSET_CB_INDEX]);
1565 for (; cb->args[IPSET_CB_INDEX] < max; cb->args[IPSET_CB_INDEX]++) {
1566 index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1567 write_lock_bh(&ip_set_ref_lock);
1568 set = ip_set(inst, index);
1569 is_destroyed = inst->is_destroyed;
1570 if (!set || is_destroyed) {
1571 write_unlock_bh(&ip_set_ref_lock);
1572 if (dump_type == DUMP_ONE) {
1577 /* All sets are just being destroyed */
1583 /* When dumping all sets, we must dump "sorted"
1584 * so that lists (unions of sets) are dumped last.
1586 if (dump_type != DUMP_ONE &&
1587 ((dump_type == DUMP_ALL) ==
1588 !!(set->type->features & IPSET_DUMP_LAST))) {
1589 write_unlock_bh(&ip_set_ref_lock);
1592 pr_debug("List set: %s\n", set->name);
1593 if (!cb->args[IPSET_CB_ARG0]) {
1594 /* Start listing: make sure set won't be destroyed */
1595 pr_debug("reference set\n");
1598 write_unlock_bh(&ip_set_ref_lock);
1599 nlh = start_msg(skb, NETLINK_CB(cb->skb).portid,
1600 cb->nlh->nlmsg_seq, flags,
1604 goto release_refcount;
1606 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL,
1607 cb->args[IPSET_CB_PROTO]) ||
1608 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name))
1609 goto nla_put_failure;
1610 if (dump_flags & IPSET_FLAG_LIST_SETNAME)
1612 switch (cb->args[IPSET_CB_ARG0]) {
1614 /* Core header data */
1615 if (nla_put_string(skb, IPSET_ATTR_TYPENAME,
1617 nla_put_u8(skb, IPSET_ATTR_FAMILY,
1619 nla_put_u8(skb, IPSET_ATTR_REVISION,
1621 goto nla_put_failure;
1622 if (cb->args[IPSET_CB_PROTO] > IPSET_PROTOCOL_MIN &&
1623 nla_put_net16(skb, IPSET_ATTR_INDEX, htons(index)))
1624 goto nla_put_failure;
1625 ret = set->variant->head(set, skb);
1627 goto release_refcount;
1628 if (dump_flags & IPSET_FLAG_LIST_HEADER)
1630 if (set->variant->uref)
1631 set->variant->uref(set, cb, true);
1634 ret = set->variant->list(set, skb, cb);
1635 if (!cb->args[IPSET_CB_ARG0])
1636 /* Set is done, proceed with next one */
1638 goto release_refcount;
1641 /* If we dump all sets, continue with dumping last ones */
1642 if (dump_type == DUMP_ALL) {
1643 dump_type = DUMP_LAST;
1644 cb->args[IPSET_CB_DUMP] = dump_type | (dump_flags << 16);
1645 cb->args[IPSET_CB_INDEX] = 0;
1646 if (set && set->variant->uref)
1647 set->variant->uref(set, cb, false);
1655 if (dump_type == DUMP_ONE)
1656 cb->args[IPSET_CB_INDEX] = IPSET_INVALID_ID;
1658 cb->args[IPSET_CB_INDEX]++;
1660 /* If there was an error or set is done, release set */
1661 if (ret || !cb->args[IPSET_CB_ARG0]) {
1662 set = ip_set_ref_netlink(inst, index);
1663 if (set->variant->uref)
1664 set->variant->uref(set, cb, false);
1665 pr_debug("release set %s\n", set->name);
1666 __ip_set_put_netlink(set);
1667 cb->args[IPSET_CB_ARG0] = 0;
1671 nlmsg_end(skb, nlh);
1672 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len);
1676 return ret < 0 ? ret : skb->len;
1679 static int ip_set_dump(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1680 const struct nlmsghdr *nlh,
1681 const struct nlattr * const attr[],
1682 struct netlink_ext_ack *extack)
1684 if (unlikely(protocol_min_failed(attr)))
1685 return -IPSET_ERR_PROTOCOL;
1688 struct netlink_dump_control c = {
1689 .start = ip_set_dump_start,
1690 .dump = ip_set_dump_do,
1691 .done = ip_set_dump_done,
1693 return netlink_dump_start(ctnl, skb, nlh, &c);
1697 /* Add, del and test */
1699 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = {
1700 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1701 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1702 .len = IPSET_MAXNAMELEN - 1 },
1703 [IPSET_ATTR_LINENO] = { .type = NLA_U32 },
1704 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
1705 [IPSET_ATTR_ADT] = { .type = NLA_NESTED },
1709 call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set,
1710 struct nlattr *tb[], enum ipset_adt adt,
1711 u32 flags, bool use_lineno)
1715 bool eexist = flags & IPSET_FLAG_EXIST, retried = false;
1719 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried);
1722 } while (ret == -EAGAIN &&
1723 set->variant->resize &&
1724 (ret = set->variant->resize(set, retried)) == 0);
1726 if (!ret || (ret == -IPSET_ERR_EXIST && eexist))
1728 if (lineno && use_lineno) {
1729 /* Error in restore/batch mode: send back lineno */
1730 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb);
1731 struct sk_buff *skb2;
1732 struct nlmsgerr *errmsg;
1733 size_t payload = min(SIZE_MAX,
1734 sizeof(*errmsg) + nlmsg_len(nlh));
1735 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1736 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1737 struct nlattr *cmdattr;
1740 skb2 = nlmsg_new(payload, GFP_KERNEL);
1743 rep = __nlmsg_put(skb2, NETLINK_CB(skb).portid,
1744 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0);
1745 errmsg = nlmsg_data(rep);
1746 errmsg->error = ret;
1747 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len);
1748 cmdattr = (void *)&errmsg->msg + min_len;
1750 ret = nla_parse(cda, IPSET_ATTR_CMD_MAX, cmdattr,
1751 nlh->nlmsg_len - min_len, ip_set_adt_policy,
1758 errline = nla_data(cda[IPSET_ATTR_LINENO]);
1762 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid,
1764 /* Signal netlink not to send its ACK/errmsg. */
1771 static int ip_set_ad(struct net *net, struct sock *ctnl,
1772 struct sk_buff *skb,
1774 const struct nlmsghdr *nlh,
1775 const struct nlattr * const attr[],
1776 struct netlink_ext_ack *extack)
1778 struct ip_set_net *inst = ip_set_pernet(net);
1780 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1781 const struct nlattr *nla;
1782 u32 flags = flag_exist(nlh);
1786 if (unlikely(protocol_min_failed(attr) ||
1787 !attr[IPSET_ATTR_SETNAME] ||
1788 !((attr[IPSET_ATTR_DATA] != NULL) ^
1789 (attr[IPSET_ATTR_ADT] != NULL)) ||
1790 (attr[IPSET_ATTR_DATA] &&
1791 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1792 (attr[IPSET_ATTR_ADT] &&
1793 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1794 !attr[IPSET_ATTR_LINENO]))))
1795 return -IPSET_ERR_PROTOCOL;
1797 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1801 use_lineno = !!attr[IPSET_ATTR_LINENO];
1802 if (attr[IPSET_ATTR_DATA]) {
1803 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1804 attr[IPSET_ATTR_DATA],
1805 set->type->adt_policy, NULL))
1806 return -IPSET_ERR_PROTOCOL;
1807 ret = call_ad(ctnl, skb, set, tb, adt, flags,
1812 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1813 if (nla_type(nla) != IPSET_ATTR_DATA ||
1814 !flag_nested(nla) ||
1815 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1816 set->type->adt_policy, NULL))
1817 return -IPSET_ERR_PROTOCOL;
1818 ret = call_ad(ctnl, skb, set, tb, adt,
1827 static int ip_set_uadd(struct net *net, struct sock *ctnl,
1828 struct sk_buff *skb, const struct nlmsghdr *nlh,
1829 const struct nlattr * const attr[],
1830 struct netlink_ext_ack *extack)
1832 return ip_set_ad(net, ctnl, skb,
1833 IPSET_ADD, nlh, attr, extack);
1836 static int ip_set_udel(struct net *net, struct sock *ctnl,
1837 struct sk_buff *skb, const struct nlmsghdr *nlh,
1838 const struct nlattr * const attr[],
1839 struct netlink_ext_ack *extack)
1841 return ip_set_ad(net, ctnl, skb,
1842 IPSET_DEL, nlh, attr, extack);
1845 static int ip_set_utest(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1846 const struct nlmsghdr *nlh,
1847 const struct nlattr * const attr[],
1848 struct netlink_ext_ack *extack)
1850 struct ip_set_net *inst = ip_set_pernet(net);
1852 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1856 if (unlikely(protocol_min_failed(attr) ||
1857 !attr[IPSET_ATTR_SETNAME] ||
1858 !attr[IPSET_ATTR_DATA] ||
1859 !flag_nested(attr[IPSET_ATTR_DATA])))
1860 return -IPSET_ERR_PROTOCOL;
1862 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1866 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA],
1867 set->type->adt_policy, NULL))
1868 return -IPSET_ERR_PROTOCOL;
1871 ret = set->variant->uadt(set, tb, IPSET_TEST, &lineno, 0, 0);
1872 rcu_read_unlock_bh();
1873 /* Userspace can't trigger element to be re-added */
1877 return ret > 0 ? 0 : -IPSET_ERR_EXIST;
1880 /* Get headed data of a set */
1882 static int ip_set_header(struct net *net, struct sock *ctnl,
1883 struct sk_buff *skb, const struct nlmsghdr *nlh,
1884 const struct nlattr * const attr[],
1885 struct netlink_ext_ack *extack)
1887 struct ip_set_net *inst = ip_set_pernet(net);
1888 const struct ip_set *set;
1889 struct sk_buff *skb2;
1890 struct nlmsghdr *nlh2;
1893 if (unlikely(protocol_min_failed(attr) ||
1894 !attr[IPSET_ATTR_SETNAME]))
1895 return -IPSET_ERR_PROTOCOL;
1897 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1901 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1905 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1909 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
1910 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) ||
1911 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) ||
1912 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
1913 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision))
1914 goto nla_put_failure;
1915 nlmsg_end(skb2, nlh2);
1917 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1924 nlmsg_cancel(skb2, nlh2);
1932 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = {
1933 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1934 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
1935 .len = IPSET_MAXNAMELEN - 1 },
1936 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
1939 static int ip_set_type(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1940 const struct nlmsghdr *nlh,
1941 const struct nlattr * const attr[],
1942 struct netlink_ext_ack *extack)
1944 struct sk_buff *skb2;
1945 struct nlmsghdr *nlh2;
1946 u8 family, min, max;
1947 const char *typename;
1950 if (unlikely(protocol_min_failed(attr) ||
1951 !attr[IPSET_ATTR_TYPENAME] ||
1952 !attr[IPSET_ATTR_FAMILY]))
1953 return -IPSET_ERR_PROTOCOL;
1955 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1956 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1957 ret = find_set_type_minmax(typename, family, &min, &max);
1961 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1965 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1969 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
1970 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) ||
1971 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) ||
1972 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) ||
1973 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min))
1974 goto nla_put_failure;
1975 nlmsg_end(skb2, nlh2);
1977 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len);
1978 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1985 nlmsg_cancel(skb2, nlh2);
1991 /* Get protocol version */
1993 static const struct nla_policy
1994 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = {
1995 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1998 static int ip_set_protocol(struct net *net, struct sock *ctnl,
1999 struct sk_buff *skb, const struct nlmsghdr *nlh,
2000 const struct nlattr * const attr[],
2001 struct netlink_ext_ack *extack)
2003 struct sk_buff *skb2;
2004 struct nlmsghdr *nlh2;
2007 if (unlikely(!attr[IPSET_ATTR_PROTOCOL]))
2008 return -IPSET_ERR_PROTOCOL;
2010 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
2014 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
2015 IPSET_CMD_PROTOCOL);
2018 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL))
2019 goto nla_put_failure;
2020 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL_MIN, IPSET_PROTOCOL_MIN))
2021 goto nla_put_failure;
2022 nlmsg_end(skb2, nlh2);
2024 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
2031 nlmsg_cancel(skb2, nlh2);
2037 /* Get set by name or index, from userspace */
2039 static int ip_set_byname(struct net *net, struct sock *ctnl,
2040 struct sk_buff *skb, const struct nlmsghdr *nlh,
2041 const struct nlattr * const attr[],
2042 struct netlink_ext_ack *extack)
2044 struct ip_set_net *inst = ip_set_pernet(net);
2045 struct sk_buff *skb2;
2046 struct nlmsghdr *nlh2;
2047 ip_set_id_t id = IPSET_INVALID_ID;
2048 const struct ip_set *set;
2051 if (unlikely(protocol_failed(attr) ||
2052 !attr[IPSET_ATTR_SETNAME]))
2053 return -IPSET_ERR_PROTOCOL;
2055 set = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]), &id);
2056 if (id == IPSET_INVALID_ID)
2059 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
2063 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
2064 IPSET_CMD_GET_BYNAME);
2067 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
2068 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
2069 nla_put_net16(skb2, IPSET_ATTR_INDEX, htons(id)))
2070 goto nla_put_failure;
2071 nlmsg_end(skb2, nlh2);
2073 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
2080 nlmsg_cancel(skb2, nlh2);
2086 static const struct nla_policy ip_set_index_policy[IPSET_ATTR_CMD_MAX + 1] = {
2087 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
2088 [IPSET_ATTR_INDEX] = { .type = NLA_U16 },
2091 static int ip_set_byindex(struct net *net, struct sock *ctnl,
2092 struct sk_buff *skb, const struct nlmsghdr *nlh,
2093 const struct nlattr * const attr[],
2094 struct netlink_ext_ack *extack)
2096 struct ip_set_net *inst = ip_set_pernet(net);
2097 struct sk_buff *skb2;
2098 struct nlmsghdr *nlh2;
2099 ip_set_id_t id = IPSET_INVALID_ID;
2100 const struct ip_set *set;
2103 if (unlikely(protocol_failed(attr) ||
2104 !attr[IPSET_ATTR_INDEX]))
2105 return -IPSET_ERR_PROTOCOL;
2107 id = ip_set_get_h16(attr[IPSET_ATTR_INDEX]);
2108 if (id >= inst->ip_set_max)
2110 set = ip_set(inst, id);
2114 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
2118 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
2119 IPSET_CMD_GET_BYINDEX);
2122 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
2123 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name))
2124 goto nla_put_failure;
2125 nlmsg_end(skb2, nlh2);
2127 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
2134 nlmsg_cancel(skb2, nlh2);
2140 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = {
2141 [IPSET_CMD_NONE] = {
2142 .call = ip_set_none,
2143 .attr_count = IPSET_ATTR_CMD_MAX,
2145 [IPSET_CMD_CREATE] = {
2146 .call = ip_set_create,
2147 .attr_count = IPSET_ATTR_CMD_MAX,
2148 .policy = ip_set_create_policy,
2150 [IPSET_CMD_DESTROY] = {
2151 .call = ip_set_destroy,
2152 .attr_count = IPSET_ATTR_CMD_MAX,
2153 .policy = ip_set_setname_policy,
2155 [IPSET_CMD_FLUSH] = {
2156 .call = ip_set_flush,
2157 .attr_count = IPSET_ATTR_CMD_MAX,
2158 .policy = ip_set_setname_policy,
2160 [IPSET_CMD_RENAME] = {
2161 .call = ip_set_rename,
2162 .attr_count = IPSET_ATTR_CMD_MAX,
2163 .policy = ip_set_setname2_policy,
2165 [IPSET_CMD_SWAP] = {
2166 .call = ip_set_swap,
2167 .attr_count = IPSET_ATTR_CMD_MAX,
2168 .policy = ip_set_setname2_policy,
2170 [IPSET_CMD_LIST] = {
2171 .call = ip_set_dump,
2172 .attr_count = IPSET_ATTR_CMD_MAX,
2173 .policy = ip_set_dump_policy,
2175 [IPSET_CMD_SAVE] = {
2176 .call = ip_set_dump,
2177 .attr_count = IPSET_ATTR_CMD_MAX,
2178 .policy = ip_set_setname_policy,
2181 .call = ip_set_uadd,
2182 .attr_count = IPSET_ATTR_CMD_MAX,
2183 .policy = ip_set_adt_policy,
2186 .call = ip_set_udel,
2187 .attr_count = IPSET_ATTR_CMD_MAX,
2188 .policy = ip_set_adt_policy,
2190 [IPSET_CMD_TEST] = {
2191 .call = ip_set_utest,
2192 .attr_count = IPSET_ATTR_CMD_MAX,
2193 .policy = ip_set_adt_policy,
2195 [IPSET_CMD_HEADER] = {
2196 .call = ip_set_header,
2197 .attr_count = IPSET_ATTR_CMD_MAX,
2198 .policy = ip_set_setname_policy,
2200 [IPSET_CMD_TYPE] = {
2201 .call = ip_set_type,
2202 .attr_count = IPSET_ATTR_CMD_MAX,
2203 .policy = ip_set_type_policy,
2205 [IPSET_CMD_PROTOCOL] = {
2206 .call = ip_set_protocol,
2207 .attr_count = IPSET_ATTR_CMD_MAX,
2208 .policy = ip_set_protocol_policy,
2210 [IPSET_CMD_GET_BYNAME] = {
2211 .call = ip_set_byname,
2212 .attr_count = IPSET_ATTR_CMD_MAX,
2213 .policy = ip_set_setname_policy,
2215 [IPSET_CMD_GET_BYINDEX] = {
2216 .call = ip_set_byindex,
2217 .attr_count = IPSET_ATTR_CMD_MAX,
2218 .policy = ip_set_index_policy,
2222 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = {
2224 .subsys_id = NFNL_SUBSYS_IPSET,
2225 .cb_count = IPSET_MSG_MAX,
2226 .cb = ip_set_netlink_subsys_cb,
2229 /* Interface to iptables/ip6tables */
2232 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len)
2236 int copylen = *len, ret = 0;
2237 struct net *net = sock_net(sk);
2238 struct ip_set_net *inst = ip_set_pernet(net);
2240 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2242 if (optval != SO_IP_SET)
2244 if (*len < sizeof(unsigned int))
2247 data = vmalloc(*len);
2250 if (copy_from_user(data, user, *len) != 0) {
2256 if (*op < IP_SET_OP_VERSION) {
2257 /* Check the version at the beginning of operations */
2258 struct ip_set_req_version *req_version = data;
2260 if (*len < sizeof(struct ip_set_req_version)) {
2265 if (req_version->version < IPSET_PROTOCOL_MIN) {
2272 case IP_SET_OP_VERSION: {
2273 struct ip_set_req_version *req_version = data;
2275 if (*len != sizeof(struct ip_set_req_version)) {
2280 req_version->version = IPSET_PROTOCOL;
2281 if (copy_to_user(user, req_version,
2282 sizeof(struct ip_set_req_version)))
2286 case IP_SET_OP_GET_BYNAME: {
2287 struct ip_set_req_get_set *req_get = data;
2290 if (*len != sizeof(struct ip_set_req_get_set)) {
2294 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
2295 nfnl_lock(NFNL_SUBSYS_IPSET);
2296 find_set_and_id(inst, req_get->set.name, &id);
2297 req_get->set.index = id;
2298 nfnl_unlock(NFNL_SUBSYS_IPSET);
2301 case IP_SET_OP_GET_FNAME: {
2302 struct ip_set_req_get_set_family *req_get = data;
2305 if (*len != sizeof(struct ip_set_req_get_set_family)) {
2309 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
2310 nfnl_lock(NFNL_SUBSYS_IPSET);
2311 find_set_and_id(inst, req_get->set.name, &id);
2312 req_get->set.index = id;
2313 if (id != IPSET_INVALID_ID)
2314 req_get->family = ip_set(inst, id)->family;
2315 nfnl_unlock(NFNL_SUBSYS_IPSET);
2318 case IP_SET_OP_GET_BYINDEX: {
2319 struct ip_set_req_get_set *req_get = data;
2322 if (*len != sizeof(struct ip_set_req_get_set) ||
2323 req_get->set.index >= inst->ip_set_max) {
2327 nfnl_lock(NFNL_SUBSYS_IPSET);
2328 set = ip_set(inst, req_get->set.index);
2329 ret = strscpy(req_get->set.name, set ? set->name : "",
2331 nfnl_unlock(NFNL_SUBSYS_IPSET);
2339 } /* end of switch(op) */
2342 if (copy_to_user(user, data, copylen))
2352 static struct nf_sockopt_ops so_set __read_mostly = {
2354 .get_optmin = SO_IP_SET,
2355 .get_optmax = SO_IP_SET + 1,
2356 .get = ip_set_sockfn_get,
2357 .owner = THIS_MODULE,
2360 static int __net_init
2361 ip_set_net_init(struct net *net)
2363 struct ip_set_net *inst = ip_set_pernet(net);
2364 struct ip_set **list;
2366 inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX;
2367 if (inst->ip_set_max >= IPSET_INVALID_ID)
2368 inst->ip_set_max = IPSET_INVALID_ID - 1;
2370 list = kvcalloc(inst->ip_set_max, sizeof(struct ip_set *), GFP_KERNEL);
2373 inst->is_deleted = false;
2374 inst->is_destroyed = false;
2375 rcu_assign_pointer(inst->ip_set_list, list);
2379 static void __net_exit
2380 ip_set_net_exit(struct net *net)
2382 struct ip_set_net *inst = ip_set_pernet(net);
2384 struct ip_set *set = NULL;
2387 inst->is_deleted = true; /* flag for ip_set_nfnl_put */
2389 nfnl_lock(NFNL_SUBSYS_IPSET);
2390 for (i = 0; i < inst->ip_set_max; i++) {
2391 set = ip_set(inst, i);
2393 ip_set(inst, i) = NULL;
2394 ip_set_destroy_set(set);
2397 nfnl_unlock(NFNL_SUBSYS_IPSET);
2398 kvfree(rcu_dereference_protected(inst->ip_set_list, 1));
2401 static struct pernet_operations ip_set_net_ops = {
2402 .init = ip_set_net_init,
2403 .exit = ip_set_net_exit,
2404 .id = &ip_set_net_id,
2405 .size = sizeof(struct ip_set_net),
2411 int ret = register_pernet_subsys(&ip_set_net_ops);
2414 pr_err("ip_set: cannot register pernet_subsys.\n");
2418 ret = nfnetlink_subsys_register(&ip_set_netlink_subsys);
2420 pr_err("ip_set: cannot register with nfnetlink.\n");
2421 unregister_pernet_subsys(&ip_set_net_ops);
2425 ret = nf_register_sockopt(&so_set);
2427 pr_err("SO_SET registry failed: %d\n", ret);
2428 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2429 unregister_pernet_subsys(&ip_set_net_ops);
2439 nf_unregister_sockopt(&so_set);
2440 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2442 unregister_pernet_subsys(&ip_set_net_ops);
2443 pr_debug("these are the famous last words\n");
2446 module_init(ip_set_init);
2447 module_exit(ip_set_fini);
2449 MODULE_DESCRIPTION("ip_set: protocol " __stringify(IPSET_PROTOCOL));