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 return nfnl_msg_put(skb, portid, seq,
967 nfnl_msg_type(NFNL_SUBSYS_IPSET, cmd), flags,
968 NFPROTO_IPV4, NFNETLINK_V0, 0);
973 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = {
974 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
975 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
976 .len = IPSET_MAXNAMELEN - 1 },
977 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
978 .len = IPSET_MAXNAMELEN - 1},
979 [IPSET_ATTR_REVISION] = { .type = NLA_U8 },
980 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
981 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
984 static struct ip_set *
985 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id)
987 struct ip_set *set = NULL;
990 *id = IPSET_INVALID_ID;
991 for (i = 0; i < inst->ip_set_max; i++) {
992 set = ip_set(inst, i);
993 if (set && STRNCMP(set->name, name)) {
998 return (*id == IPSET_INVALID_ID ? NULL : set);
1001 static inline struct ip_set *
1002 find_set(struct ip_set_net *inst, const char *name)
1006 return find_set_and_id(inst, name, &id);
1010 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index,
1011 struct ip_set **set)
1016 *index = IPSET_INVALID_ID;
1017 for (i = 0; i < inst->ip_set_max; i++) {
1018 s = ip_set(inst, i);
1020 if (*index == IPSET_INVALID_ID)
1022 } else if (STRNCMP(name, s->name)) {
1028 if (*index == IPSET_INVALID_ID)
1029 /* No free slot remained */
1030 return -IPSET_ERR_MAX_SETS;
1034 static int ip_set_none(struct sk_buff *skb, const struct nfnl_info *info,
1035 const struct nlattr * const attr[])
1040 static int ip_set_create(struct sk_buff *skb, const struct nfnl_info *info,
1041 const struct nlattr * const attr[])
1043 struct ip_set_net *inst = ip_set_pernet(info->net);
1044 struct ip_set *set, *clash = NULL;
1045 ip_set_id_t index = IPSET_INVALID_ID;
1046 struct nlattr *tb[IPSET_ATTR_CREATE_MAX + 1] = {};
1047 const char *name, *typename;
1048 u8 family, revision;
1049 u32 flags = flag_exist(info->nlh);
1052 if (unlikely(protocol_min_failed(attr) ||
1053 !attr[IPSET_ATTR_SETNAME] ||
1054 !attr[IPSET_ATTR_TYPENAME] ||
1055 !attr[IPSET_ATTR_REVISION] ||
1056 !attr[IPSET_ATTR_FAMILY] ||
1057 (attr[IPSET_ATTR_DATA] &&
1058 !flag_nested(attr[IPSET_ATTR_DATA]))))
1059 return -IPSET_ERR_PROTOCOL;
1061 name = nla_data(attr[IPSET_ATTR_SETNAME]);
1062 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1063 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1064 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]);
1065 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
1066 name, typename, family_name(family), revision);
1068 /* First, and without any locks, allocate and initialize
1069 * a normal base set structure.
1071 set = kzalloc(sizeof(*set), GFP_KERNEL);
1074 spin_lock_init(&set->lock);
1075 strlcpy(set->name, name, IPSET_MAXNAMELEN);
1076 set->family = family;
1077 set->revision = revision;
1079 /* Next, check that we know the type, and take
1080 * a reference on the type, to make sure it stays available
1081 * while constructing our new set.
1083 * After referencing the type, we try to create the type
1084 * specific part of the set without holding any locks.
1086 ret = find_set_type_get(typename, family, revision, &set->type);
1090 /* Without holding any locks, create private part. */
1091 if (attr[IPSET_ATTR_DATA] &&
1092 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA],
1093 set->type->create_policy, NULL)) {
1094 ret = -IPSET_ERR_PROTOCOL;
1097 /* Set create flags depending on the type revision */
1098 set->flags |= set->type->create_flags[revision];
1100 ret = set->type->create(info->net, set, tb, flags);
1104 /* BTW, ret==0 here. */
1106 /* Here, we have a valid, constructed set and we are protected
1107 * by the nfnl mutex. Find the first free index in ip_set_list
1108 * and check clashing.
1110 ret = find_free_id(inst, set->name, &index, &clash);
1111 if (ret == -EEXIST) {
1112 /* If this is the same set and requested, ignore error */
1113 if ((flags & IPSET_FLAG_EXIST) &&
1114 STRNCMP(set->type->name, clash->type->name) &&
1115 set->type->family == clash->type->family &&
1116 set->type->revision_min == clash->type->revision_min &&
1117 set->type->revision_max == clash->type->revision_max &&
1118 set->variant->same_set(set, clash))
1121 } else if (ret == -IPSET_ERR_MAX_SETS) {
1122 struct ip_set **list, **tmp;
1123 ip_set_id_t i = inst->ip_set_max + IP_SET_INC;
1125 if (i < inst->ip_set_max || i == IPSET_INVALID_ID)
1129 list = kvcalloc(i, sizeof(struct ip_set *), GFP_KERNEL);
1132 /* nfnl mutex is held, both lists are valid */
1133 tmp = ip_set_dereference(inst->ip_set_list);
1134 memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max);
1135 rcu_assign_pointer(inst->ip_set_list, list);
1136 /* Make sure all current packets have passed through */
1139 index = inst->ip_set_max;
1140 inst->ip_set_max = i;
1147 /* Finally! Add our shiny new set to the list, and be done. */
1148 pr_debug("create: '%s' created with index %u!\n", set->name, index);
1149 ip_set(inst, index) = set;
1154 set->variant->destroy(set);
1156 module_put(set->type->me);
1164 static const struct nla_policy
1165 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = {
1166 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1167 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1168 .len = IPSET_MAXNAMELEN - 1 },
1172 ip_set_destroy_set(struct ip_set *set)
1174 pr_debug("set: %s\n", set->name);
1176 /* Must call it without holding any lock */
1177 set->variant->destroy(set);
1178 module_put(set->type->me);
1182 static int ip_set_destroy(struct sk_buff *skb, const struct nfnl_info *info,
1183 const struct nlattr * const attr[])
1185 struct ip_set_net *inst = ip_set_pernet(info->net);
1190 if (unlikely(protocol_min_failed(attr)))
1191 return -IPSET_ERR_PROTOCOL;
1193 /* Must wait for flush to be really finished in list:set */
1196 /* Commands are serialized and references are
1197 * protected by the ip_set_ref_lock.
1198 * External systems (i.e. xt_set) must call
1199 * ip_set_put|get_nfnl_* functions, that way we
1200 * can safely check references here.
1202 * list:set timer can only decrement the reference
1203 * counter, so if it's already zero, we can proceed
1204 * without holding the lock.
1206 read_lock_bh(&ip_set_ref_lock);
1207 if (!attr[IPSET_ATTR_SETNAME]) {
1208 for (i = 0; i < inst->ip_set_max; i++) {
1209 s = ip_set(inst, i);
1210 if (s && (s->ref || s->ref_netlink)) {
1211 ret = -IPSET_ERR_BUSY;
1215 inst->is_destroyed = true;
1216 read_unlock_bh(&ip_set_ref_lock);
1217 for (i = 0; i < inst->ip_set_max; i++) {
1218 s = ip_set(inst, i);
1220 ip_set(inst, i) = NULL;
1221 ip_set_destroy_set(s);
1224 /* Modified by ip_set_destroy() only, which is serialized */
1225 inst->is_destroyed = false;
1227 u32 flags = flag_exist(info->nlh);
1228 s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1231 if (!(flags & IPSET_FLAG_EXIST))
1234 } else if (s->ref || s->ref_netlink) {
1235 ret = -IPSET_ERR_BUSY;
1238 ip_set(inst, i) = NULL;
1239 read_unlock_bh(&ip_set_ref_lock);
1241 ip_set_destroy_set(s);
1245 read_unlock_bh(&ip_set_ref_lock);
1252 ip_set_flush_set(struct ip_set *set)
1254 pr_debug("set: %s\n", set->name);
1257 set->variant->flush(set);
1261 static int ip_set_flush(struct sk_buff *skb, const struct nfnl_info *info,
1262 const struct nlattr * const attr[])
1264 struct ip_set_net *inst = ip_set_pernet(info->net);
1268 if (unlikely(protocol_min_failed(attr)))
1269 return -IPSET_ERR_PROTOCOL;
1271 if (!attr[IPSET_ATTR_SETNAME]) {
1272 for (i = 0; i < inst->ip_set_max; i++) {
1273 s = ip_set(inst, i);
1275 ip_set_flush_set(s);
1278 s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1282 ip_set_flush_set(s);
1290 static const struct nla_policy
1291 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = {
1292 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1293 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1294 .len = IPSET_MAXNAMELEN - 1 },
1295 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING,
1296 .len = IPSET_MAXNAMELEN - 1 },
1299 static int ip_set_rename(struct sk_buff *skb, const struct nfnl_info *info,
1300 const struct nlattr * const attr[])
1302 struct ip_set_net *inst = ip_set_pernet(info->net);
1303 struct ip_set *set, *s;
1308 if (unlikely(protocol_min_failed(attr) ||
1309 !attr[IPSET_ATTR_SETNAME] ||
1310 !attr[IPSET_ATTR_SETNAME2]))
1311 return -IPSET_ERR_PROTOCOL;
1313 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1317 write_lock_bh(&ip_set_ref_lock);
1318 if (set->ref != 0 || set->ref_netlink != 0) {
1319 ret = -IPSET_ERR_REFERENCED;
1323 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]);
1324 for (i = 0; i < inst->ip_set_max; i++) {
1325 s = ip_set(inst, i);
1326 if (s && STRNCMP(s->name, name2)) {
1327 ret = -IPSET_ERR_EXIST_SETNAME2;
1331 strncpy(set->name, name2, IPSET_MAXNAMELEN);
1334 write_unlock_bh(&ip_set_ref_lock);
1338 /* Swap two sets so that name/index points to the other.
1339 * References and set names are also swapped.
1341 * The commands are serialized by the nfnl mutex and references are
1342 * protected by the ip_set_ref_lock. The kernel interfaces
1343 * do not hold the mutex but the pointer settings are atomic
1344 * so the ip_set_list always contains valid pointers to the sets.
1347 static int ip_set_swap(struct sk_buff *skb, const struct nfnl_info *info,
1348 const struct nlattr * const attr[])
1350 struct ip_set_net *inst = ip_set_pernet(info->net);
1351 struct ip_set *from, *to;
1352 ip_set_id_t from_id, to_id;
1353 char from_name[IPSET_MAXNAMELEN];
1355 if (unlikely(protocol_min_failed(attr) ||
1356 !attr[IPSET_ATTR_SETNAME] ||
1357 !attr[IPSET_ATTR_SETNAME2]))
1358 return -IPSET_ERR_PROTOCOL;
1360 from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1365 to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]),
1368 return -IPSET_ERR_EXIST_SETNAME2;
1370 /* Features must not change.
1371 * Not an artifical restriction anymore, as we must prevent
1372 * possible loops created by swapping in setlist type of sets.
1374 if (!(from->type->features == to->type->features &&
1375 from->family == to->family))
1376 return -IPSET_ERR_TYPE_MISMATCH;
1378 write_lock_bh(&ip_set_ref_lock);
1380 if (from->ref_netlink || to->ref_netlink) {
1381 write_unlock_bh(&ip_set_ref_lock);
1385 strncpy(from_name, from->name, IPSET_MAXNAMELEN);
1386 strncpy(from->name, to->name, IPSET_MAXNAMELEN);
1387 strncpy(to->name, from_name, IPSET_MAXNAMELEN);
1389 swap(from->ref, to->ref);
1390 ip_set(inst, from_id) = to;
1391 ip_set(inst, to_id) = from;
1392 write_unlock_bh(&ip_set_ref_lock);
1397 /* List/save set data */
1404 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1405 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1408 ip_set_put_flags(struct sk_buff *skb, struct ip_set *set)
1412 if (SET_WITH_TIMEOUT(set))
1413 if (unlikely(nla_put_net32(skb, IPSET_ATTR_TIMEOUT,
1414 htonl(set->timeout))))
1416 if (SET_WITH_COUNTER(set))
1417 cadt_flags |= IPSET_FLAG_WITH_COUNTERS;
1418 if (SET_WITH_COMMENT(set))
1419 cadt_flags |= IPSET_FLAG_WITH_COMMENT;
1420 if (SET_WITH_SKBINFO(set))
1421 cadt_flags |= IPSET_FLAG_WITH_SKBINFO;
1422 if (SET_WITH_FORCEADD(set))
1423 cadt_flags |= IPSET_FLAG_WITH_FORCEADD;
1427 return nla_put_net32(skb, IPSET_ATTR_CADT_FLAGS, htonl(cadt_flags));
1429 EXPORT_SYMBOL_GPL(ip_set_put_flags);
1432 ip_set_dump_done(struct netlink_callback *cb)
1434 if (cb->args[IPSET_CB_ARG0]) {
1435 struct ip_set_net *inst =
1436 (struct ip_set_net *)cb->args[IPSET_CB_NET];
1437 ip_set_id_t index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1438 struct ip_set *set = ip_set_ref_netlink(inst, index);
1440 if (set->variant->uref)
1441 set->variant->uref(set, cb, false);
1442 pr_debug("release set %s\n", set->name);
1443 __ip_set_put_netlink(set);
1449 dump_attrs(struct nlmsghdr *nlh)
1451 const struct nlattr *attr;
1454 pr_debug("dump nlmsg\n");
1455 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) {
1456 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len);
1460 static const struct nla_policy
1461 ip_set_dump_policy[IPSET_ATTR_CMD_MAX + 1] = {
1462 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1463 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1464 .len = IPSET_MAXNAMELEN - 1 },
1465 [IPSET_ATTR_FLAGS] = { .type = NLA_U32 },
1469 ip_set_dump_start(struct netlink_callback *cb)
1471 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb);
1472 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1473 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1474 struct nlattr *attr = (void *)nlh + min_len;
1475 struct sk_buff *skb = cb->skb;
1476 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1480 ret = nla_parse(cda, IPSET_ATTR_CMD_MAX, attr,
1481 nlh->nlmsg_len - min_len,
1482 ip_set_dump_policy, NULL);
1486 cb->args[IPSET_CB_PROTO] = nla_get_u8(cda[IPSET_ATTR_PROTOCOL]);
1487 if (cda[IPSET_ATTR_SETNAME]) {
1491 set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]),
1497 dump_type = DUMP_ONE;
1498 cb->args[IPSET_CB_INDEX] = index;
1500 dump_type = DUMP_ALL;
1503 if (cda[IPSET_ATTR_FLAGS]) {
1504 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]);
1506 dump_type |= (f << 16);
1508 cb->args[IPSET_CB_NET] = (unsigned long)inst;
1509 cb->args[IPSET_CB_DUMP] = dump_type;
1514 /* We have to create and send the error message manually :-( */
1515 if (nlh->nlmsg_flags & NLM_F_ACK) {
1516 netlink_ack(cb->skb, nlh, ret, NULL);
1522 ip_set_dump_do(struct sk_buff *skb, struct netlink_callback *cb)
1524 ip_set_id_t index = IPSET_INVALID_ID, max;
1525 struct ip_set *set = NULL;
1526 struct nlmsghdr *nlh = NULL;
1527 unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0;
1528 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1529 u32 dump_type, dump_flags;
1533 if (!cb->args[IPSET_CB_DUMP])
1536 if (cb->args[IPSET_CB_INDEX] >= inst->ip_set_max)
1539 dump_type = DUMP_TYPE(cb->args[IPSET_CB_DUMP]);
1540 dump_flags = DUMP_FLAGS(cb->args[IPSET_CB_DUMP]);
1541 max = dump_type == DUMP_ONE ? cb->args[IPSET_CB_INDEX] + 1
1544 pr_debug("dump type, flag: %u %u index: %ld\n",
1545 dump_type, dump_flags, cb->args[IPSET_CB_INDEX]);
1546 for (; cb->args[IPSET_CB_INDEX] < max; cb->args[IPSET_CB_INDEX]++) {
1547 index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1548 write_lock_bh(&ip_set_ref_lock);
1549 set = ip_set(inst, index);
1550 is_destroyed = inst->is_destroyed;
1551 if (!set || is_destroyed) {
1552 write_unlock_bh(&ip_set_ref_lock);
1553 if (dump_type == DUMP_ONE) {
1558 /* All sets are just being destroyed */
1564 /* When dumping all sets, we must dump "sorted"
1565 * so that lists (unions of sets) are dumped last.
1567 if (dump_type != DUMP_ONE &&
1568 ((dump_type == DUMP_ALL) ==
1569 !!(set->type->features & IPSET_DUMP_LAST))) {
1570 write_unlock_bh(&ip_set_ref_lock);
1573 pr_debug("List set: %s\n", set->name);
1574 if (!cb->args[IPSET_CB_ARG0]) {
1575 /* Start listing: make sure set won't be destroyed */
1576 pr_debug("reference set\n");
1579 write_unlock_bh(&ip_set_ref_lock);
1580 nlh = start_msg(skb, NETLINK_CB(cb->skb).portid,
1581 cb->nlh->nlmsg_seq, flags,
1585 goto release_refcount;
1587 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL,
1588 cb->args[IPSET_CB_PROTO]) ||
1589 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name))
1590 goto nla_put_failure;
1591 if (dump_flags & IPSET_FLAG_LIST_SETNAME)
1593 switch (cb->args[IPSET_CB_ARG0]) {
1595 /* Core header data */
1596 if (nla_put_string(skb, IPSET_ATTR_TYPENAME,
1598 nla_put_u8(skb, IPSET_ATTR_FAMILY,
1600 nla_put_u8(skb, IPSET_ATTR_REVISION,
1602 goto nla_put_failure;
1603 if (cb->args[IPSET_CB_PROTO] > IPSET_PROTOCOL_MIN &&
1604 nla_put_net16(skb, IPSET_ATTR_INDEX, htons(index)))
1605 goto nla_put_failure;
1606 ret = set->variant->head(set, skb);
1608 goto release_refcount;
1609 if (dump_flags & IPSET_FLAG_LIST_HEADER)
1611 if (set->variant->uref)
1612 set->variant->uref(set, cb, true);
1615 ret = set->variant->list(set, skb, cb);
1616 if (!cb->args[IPSET_CB_ARG0])
1617 /* Set is done, proceed with next one */
1619 goto release_refcount;
1622 /* If we dump all sets, continue with dumping last ones */
1623 if (dump_type == DUMP_ALL) {
1624 dump_type = DUMP_LAST;
1625 cb->args[IPSET_CB_DUMP] = dump_type | (dump_flags << 16);
1626 cb->args[IPSET_CB_INDEX] = 0;
1627 if (set && set->variant->uref)
1628 set->variant->uref(set, cb, false);
1636 if (dump_type == DUMP_ONE)
1637 cb->args[IPSET_CB_INDEX] = IPSET_INVALID_ID;
1639 cb->args[IPSET_CB_INDEX]++;
1641 /* If there was an error or set is done, release set */
1642 if (ret || !cb->args[IPSET_CB_ARG0]) {
1643 set = ip_set_ref_netlink(inst, index);
1644 if (set->variant->uref)
1645 set->variant->uref(set, cb, false);
1646 pr_debug("release set %s\n", set->name);
1647 __ip_set_put_netlink(set);
1648 cb->args[IPSET_CB_ARG0] = 0;
1652 nlmsg_end(skb, nlh);
1653 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len);
1657 return ret < 0 ? ret : skb->len;
1660 static int ip_set_dump(struct sk_buff *skb, const struct nfnl_info *info,
1661 const struct nlattr * const attr[])
1663 if (unlikely(protocol_min_failed(attr)))
1664 return -IPSET_ERR_PROTOCOL;
1667 struct netlink_dump_control c = {
1668 .start = ip_set_dump_start,
1669 .dump = ip_set_dump_do,
1670 .done = ip_set_dump_done,
1672 return netlink_dump_start(info->sk, skb, info->nlh, &c);
1676 /* Add, del and test */
1678 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = {
1679 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1680 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1681 .len = IPSET_MAXNAMELEN - 1 },
1682 [IPSET_ATTR_LINENO] = { .type = NLA_U32 },
1683 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
1684 [IPSET_ATTR_ADT] = { .type = NLA_NESTED },
1688 call_ad(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1689 struct ip_set *set, struct nlattr *tb[], enum ipset_adt adt,
1690 u32 flags, bool use_lineno)
1694 bool eexist = flags & IPSET_FLAG_EXIST, retried = false;
1698 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried);
1701 } while (ret == -ERANGE ||
1703 set->variant->resize &&
1704 (ret = set->variant->resize(set, retried)) == 0));
1706 if (!ret || (ret == -IPSET_ERR_EXIST && eexist))
1708 if (lineno && use_lineno) {
1709 /* Error in restore/batch mode: send back lineno */
1710 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb);
1711 struct sk_buff *skb2;
1712 struct nlmsgerr *errmsg;
1713 size_t payload = min(SIZE_MAX,
1714 sizeof(*errmsg) + nlmsg_len(nlh));
1715 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1716 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1717 struct nlattr *cmdattr;
1720 skb2 = nlmsg_new(payload, GFP_KERNEL);
1723 rep = __nlmsg_put(skb2, NETLINK_CB(skb).portid,
1724 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0);
1725 errmsg = nlmsg_data(rep);
1726 errmsg->error = ret;
1727 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len);
1728 cmdattr = (void *)&errmsg->msg + min_len;
1730 ret = nla_parse(cda, IPSET_ATTR_CMD_MAX, cmdattr,
1731 nlh->nlmsg_len - min_len, ip_set_adt_policy,
1738 errline = nla_data(cda[IPSET_ATTR_LINENO]);
1742 nfnetlink_unicast(skb2, net, NETLINK_CB(skb).portid);
1743 /* Signal netlink not to send its ACK/errmsg. */
1750 static int ip_set_ad(struct net *net, struct sock *ctnl,
1751 struct sk_buff *skb,
1753 const struct nlmsghdr *nlh,
1754 const struct nlattr * const attr[],
1755 struct netlink_ext_ack *extack)
1757 struct ip_set_net *inst = ip_set_pernet(net);
1759 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1760 const struct nlattr *nla;
1761 u32 flags = flag_exist(nlh);
1765 if (unlikely(protocol_min_failed(attr) ||
1766 !attr[IPSET_ATTR_SETNAME] ||
1767 !((attr[IPSET_ATTR_DATA] != NULL) ^
1768 (attr[IPSET_ATTR_ADT] != NULL)) ||
1769 (attr[IPSET_ATTR_DATA] &&
1770 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1771 (attr[IPSET_ATTR_ADT] &&
1772 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1773 !attr[IPSET_ATTR_LINENO]))))
1774 return -IPSET_ERR_PROTOCOL;
1776 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1780 use_lineno = !!attr[IPSET_ATTR_LINENO];
1781 if (attr[IPSET_ATTR_DATA]) {
1782 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1783 attr[IPSET_ATTR_DATA],
1784 set->type->adt_policy, NULL))
1785 return -IPSET_ERR_PROTOCOL;
1786 ret = call_ad(net, ctnl, skb, set, tb, adt, flags,
1791 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1792 if (nla_type(nla) != IPSET_ATTR_DATA ||
1793 !flag_nested(nla) ||
1794 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1795 set->type->adt_policy, NULL))
1796 return -IPSET_ERR_PROTOCOL;
1797 ret = call_ad(net, ctnl, skb, set, tb, adt,
1806 static int ip_set_uadd(struct sk_buff *skb, const struct nfnl_info *info,
1807 const struct nlattr * const attr[])
1809 return ip_set_ad(info->net, info->sk, skb,
1810 IPSET_ADD, info->nlh, attr, info->extack);
1813 static int ip_set_udel(struct sk_buff *skb, const struct nfnl_info *info,
1814 const struct nlattr * const attr[])
1816 return ip_set_ad(info->net, info->sk, skb,
1817 IPSET_DEL, info->nlh, attr, info->extack);
1820 static int ip_set_utest(struct sk_buff *skb, const struct nfnl_info *info,
1821 const struct nlattr * const attr[])
1823 struct ip_set_net *inst = ip_set_pernet(info->net);
1825 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1829 if (unlikely(protocol_min_failed(attr) ||
1830 !attr[IPSET_ATTR_SETNAME] ||
1831 !attr[IPSET_ATTR_DATA] ||
1832 !flag_nested(attr[IPSET_ATTR_DATA])))
1833 return -IPSET_ERR_PROTOCOL;
1835 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1839 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA],
1840 set->type->adt_policy, NULL))
1841 return -IPSET_ERR_PROTOCOL;
1844 ret = set->variant->uadt(set, tb, IPSET_TEST, &lineno, 0, 0);
1845 rcu_read_unlock_bh();
1846 /* Userspace can't trigger element to be re-added */
1850 return ret > 0 ? 0 : -IPSET_ERR_EXIST;
1853 /* Get headed data of a set */
1855 static int ip_set_header(struct sk_buff *skb, const struct nfnl_info *info,
1856 const struct nlattr * const attr[])
1858 struct ip_set_net *inst = ip_set_pernet(info->net);
1859 const struct ip_set *set;
1860 struct sk_buff *skb2;
1861 struct nlmsghdr *nlh2;
1863 if (unlikely(protocol_min_failed(attr) ||
1864 !attr[IPSET_ATTR_SETNAME]))
1865 return -IPSET_ERR_PROTOCOL;
1867 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1871 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1875 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, info->nlh->nlmsg_seq, 0,
1879 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
1880 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) ||
1881 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) ||
1882 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
1883 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision))
1884 goto nla_put_failure;
1885 nlmsg_end(skb2, nlh2);
1887 return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid);
1890 nlmsg_cancel(skb2, nlh2);
1898 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = {
1899 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1900 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
1901 .len = IPSET_MAXNAMELEN - 1 },
1902 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
1905 static int ip_set_type(struct sk_buff *skb, const struct nfnl_info *info,
1906 const struct nlattr * const attr[])
1908 struct sk_buff *skb2;
1909 struct nlmsghdr *nlh2;
1910 u8 family, min, max;
1911 const char *typename;
1914 if (unlikely(protocol_min_failed(attr) ||
1915 !attr[IPSET_ATTR_TYPENAME] ||
1916 !attr[IPSET_ATTR_FAMILY]))
1917 return -IPSET_ERR_PROTOCOL;
1919 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1920 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1921 ret = find_set_type_minmax(typename, family, &min, &max);
1925 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1929 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, info->nlh->nlmsg_seq, 0,
1933 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
1934 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) ||
1935 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) ||
1936 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) ||
1937 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min))
1938 goto nla_put_failure;
1939 nlmsg_end(skb2, nlh2);
1941 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len);
1942 return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid);
1945 nlmsg_cancel(skb2, nlh2);
1951 /* Get protocol version */
1953 static const struct nla_policy
1954 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = {
1955 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1958 static int ip_set_protocol(struct sk_buff *skb, const struct nfnl_info *info,
1959 const struct nlattr * const attr[])
1961 struct sk_buff *skb2;
1962 struct nlmsghdr *nlh2;
1964 if (unlikely(!attr[IPSET_ATTR_PROTOCOL]))
1965 return -IPSET_ERR_PROTOCOL;
1967 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1971 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, info->nlh->nlmsg_seq, 0,
1972 IPSET_CMD_PROTOCOL);
1975 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL))
1976 goto nla_put_failure;
1977 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL_MIN, IPSET_PROTOCOL_MIN))
1978 goto nla_put_failure;
1979 nlmsg_end(skb2, nlh2);
1981 return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid);
1984 nlmsg_cancel(skb2, nlh2);
1990 /* Get set by name or index, from userspace */
1992 static int ip_set_byname(struct sk_buff *skb, const struct nfnl_info *info,
1993 const struct nlattr * const attr[])
1995 struct ip_set_net *inst = ip_set_pernet(info->net);
1996 struct sk_buff *skb2;
1997 struct nlmsghdr *nlh2;
1998 ip_set_id_t id = IPSET_INVALID_ID;
1999 const struct ip_set *set;
2001 if (unlikely(protocol_failed(attr) ||
2002 !attr[IPSET_ATTR_SETNAME]))
2003 return -IPSET_ERR_PROTOCOL;
2005 set = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]), &id);
2006 if (id == IPSET_INVALID_ID)
2009 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
2013 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, info->nlh->nlmsg_seq, 0,
2014 IPSET_CMD_GET_BYNAME);
2017 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
2018 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
2019 nla_put_net16(skb2, IPSET_ATTR_INDEX, htons(id)))
2020 goto nla_put_failure;
2021 nlmsg_end(skb2, nlh2);
2023 return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid);
2026 nlmsg_cancel(skb2, nlh2);
2032 static const struct nla_policy ip_set_index_policy[IPSET_ATTR_CMD_MAX + 1] = {
2033 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
2034 [IPSET_ATTR_INDEX] = { .type = NLA_U16 },
2037 static int ip_set_byindex(struct sk_buff *skb, const struct nfnl_info *info,
2038 const struct nlattr * const attr[])
2040 struct ip_set_net *inst = ip_set_pernet(info->net);
2041 struct sk_buff *skb2;
2042 struct nlmsghdr *nlh2;
2043 ip_set_id_t id = IPSET_INVALID_ID;
2044 const struct ip_set *set;
2046 if (unlikely(protocol_failed(attr) ||
2047 !attr[IPSET_ATTR_INDEX]))
2048 return -IPSET_ERR_PROTOCOL;
2050 id = ip_set_get_h16(attr[IPSET_ATTR_INDEX]);
2051 if (id >= inst->ip_set_max)
2053 set = ip_set(inst, id);
2057 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
2061 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, info->nlh->nlmsg_seq, 0,
2062 IPSET_CMD_GET_BYINDEX);
2065 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
2066 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name))
2067 goto nla_put_failure;
2068 nlmsg_end(skb2, nlh2);
2070 return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid);
2073 nlmsg_cancel(skb2, nlh2);
2079 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = {
2080 [IPSET_CMD_NONE] = {
2081 .call = ip_set_none,
2082 .type = NFNL_CB_MUTEX,
2083 .attr_count = IPSET_ATTR_CMD_MAX,
2085 [IPSET_CMD_CREATE] = {
2086 .call = ip_set_create,
2087 .type = NFNL_CB_MUTEX,
2088 .attr_count = IPSET_ATTR_CMD_MAX,
2089 .policy = ip_set_create_policy,
2091 [IPSET_CMD_DESTROY] = {
2092 .call = ip_set_destroy,
2093 .type = NFNL_CB_MUTEX,
2094 .attr_count = IPSET_ATTR_CMD_MAX,
2095 .policy = ip_set_setname_policy,
2097 [IPSET_CMD_FLUSH] = {
2098 .call = ip_set_flush,
2099 .type = NFNL_CB_MUTEX,
2100 .attr_count = IPSET_ATTR_CMD_MAX,
2101 .policy = ip_set_setname_policy,
2103 [IPSET_CMD_RENAME] = {
2104 .call = ip_set_rename,
2105 .type = NFNL_CB_MUTEX,
2106 .attr_count = IPSET_ATTR_CMD_MAX,
2107 .policy = ip_set_setname2_policy,
2109 [IPSET_CMD_SWAP] = {
2110 .call = ip_set_swap,
2111 .type = NFNL_CB_MUTEX,
2112 .attr_count = IPSET_ATTR_CMD_MAX,
2113 .policy = ip_set_setname2_policy,
2115 [IPSET_CMD_LIST] = {
2116 .call = ip_set_dump,
2117 .type = NFNL_CB_MUTEX,
2118 .attr_count = IPSET_ATTR_CMD_MAX,
2119 .policy = ip_set_dump_policy,
2121 [IPSET_CMD_SAVE] = {
2122 .call = ip_set_dump,
2123 .type = NFNL_CB_MUTEX,
2124 .attr_count = IPSET_ATTR_CMD_MAX,
2125 .policy = ip_set_setname_policy,
2128 .call = ip_set_uadd,
2129 .type = NFNL_CB_MUTEX,
2130 .attr_count = IPSET_ATTR_CMD_MAX,
2131 .policy = ip_set_adt_policy,
2134 .call = ip_set_udel,
2135 .type = NFNL_CB_MUTEX,
2136 .attr_count = IPSET_ATTR_CMD_MAX,
2137 .policy = ip_set_adt_policy,
2139 [IPSET_CMD_TEST] = {
2140 .call = ip_set_utest,
2141 .type = NFNL_CB_MUTEX,
2142 .attr_count = IPSET_ATTR_CMD_MAX,
2143 .policy = ip_set_adt_policy,
2145 [IPSET_CMD_HEADER] = {
2146 .call = ip_set_header,
2147 .type = NFNL_CB_MUTEX,
2148 .attr_count = IPSET_ATTR_CMD_MAX,
2149 .policy = ip_set_setname_policy,
2151 [IPSET_CMD_TYPE] = {
2152 .call = ip_set_type,
2153 .type = NFNL_CB_MUTEX,
2154 .attr_count = IPSET_ATTR_CMD_MAX,
2155 .policy = ip_set_type_policy,
2157 [IPSET_CMD_PROTOCOL] = {
2158 .call = ip_set_protocol,
2159 .type = NFNL_CB_MUTEX,
2160 .attr_count = IPSET_ATTR_CMD_MAX,
2161 .policy = ip_set_protocol_policy,
2163 [IPSET_CMD_GET_BYNAME] = {
2164 .call = ip_set_byname,
2165 .type = NFNL_CB_MUTEX,
2166 .attr_count = IPSET_ATTR_CMD_MAX,
2167 .policy = ip_set_setname_policy,
2169 [IPSET_CMD_GET_BYINDEX] = {
2170 .call = ip_set_byindex,
2171 .type = NFNL_CB_MUTEX,
2172 .attr_count = IPSET_ATTR_CMD_MAX,
2173 .policy = ip_set_index_policy,
2177 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = {
2179 .subsys_id = NFNL_SUBSYS_IPSET,
2180 .cb_count = IPSET_MSG_MAX,
2181 .cb = ip_set_netlink_subsys_cb,
2184 /* Interface to iptables/ip6tables */
2187 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len)
2191 int copylen = *len, ret = 0;
2192 struct net *net = sock_net(sk);
2193 struct ip_set_net *inst = ip_set_pernet(net);
2195 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2197 if (optval != SO_IP_SET)
2199 if (*len < sizeof(unsigned int))
2202 data = vmalloc(*len);
2205 if (copy_from_user(data, user, *len) != 0) {
2211 if (*op < IP_SET_OP_VERSION) {
2212 /* Check the version at the beginning of operations */
2213 struct ip_set_req_version *req_version = data;
2215 if (*len < sizeof(struct ip_set_req_version)) {
2220 if (req_version->version < IPSET_PROTOCOL_MIN) {
2227 case IP_SET_OP_VERSION: {
2228 struct ip_set_req_version *req_version = data;
2230 if (*len != sizeof(struct ip_set_req_version)) {
2235 req_version->version = IPSET_PROTOCOL;
2236 if (copy_to_user(user, req_version,
2237 sizeof(struct ip_set_req_version)))
2241 case IP_SET_OP_GET_BYNAME: {
2242 struct ip_set_req_get_set *req_get = data;
2245 if (*len != sizeof(struct ip_set_req_get_set)) {
2249 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
2250 nfnl_lock(NFNL_SUBSYS_IPSET);
2251 find_set_and_id(inst, req_get->set.name, &id);
2252 req_get->set.index = id;
2253 nfnl_unlock(NFNL_SUBSYS_IPSET);
2256 case IP_SET_OP_GET_FNAME: {
2257 struct ip_set_req_get_set_family *req_get = data;
2260 if (*len != sizeof(struct ip_set_req_get_set_family)) {
2264 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
2265 nfnl_lock(NFNL_SUBSYS_IPSET);
2266 find_set_and_id(inst, req_get->set.name, &id);
2267 req_get->set.index = id;
2268 if (id != IPSET_INVALID_ID)
2269 req_get->family = ip_set(inst, id)->family;
2270 nfnl_unlock(NFNL_SUBSYS_IPSET);
2273 case IP_SET_OP_GET_BYINDEX: {
2274 struct ip_set_req_get_set *req_get = data;
2277 if (*len != sizeof(struct ip_set_req_get_set) ||
2278 req_get->set.index >= inst->ip_set_max) {
2282 nfnl_lock(NFNL_SUBSYS_IPSET);
2283 set = ip_set(inst, req_get->set.index);
2284 ret = strscpy(req_get->set.name, set ? set->name : "",
2286 nfnl_unlock(NFNL_SUBSYS_IPSET);
2294 } /* end of switch(op) */
2297 if (copy_to_user(user, data, copylen))
2307 static struct nf_sockopt_ops so_set __read_mostly = {
2309 .get_optmin = SO_IP_SET,
2310 .get_optmax = SO_IP_SET + 1,
2311 .get = ip_set_sockfn_get,
2312 .owner = THIS_MODULE,
2315 static int __net_init
2316 ip_set_net_init(struct net *net)
2318 struct ip_set_net *inst = ip_set_pernet(net);
2319 struct ip_set **list;
2321 inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX;
2322 if (inst->ip_set_max >= IPSET_INVALID_ID)
2323 inst->ip_set_max = IPSET_INVALID_ID - 1;
2325 list = kvcalloc(inst->ip_set_max, sizeof(struct ip_set *), GFP_KERNEL);
2328 inst->is_deleted = false;
2329 inst->is_destroyed = false;
2330 rcu_assign_pointer(inst->ip_set_list, list);
2334 static void __net_exit
2335 ip_set_net_exit(struct net *net)
2337 struct ip_set_net *inst = ip_set_pernet(net);
2339 struct ip_set *set = NULL;
2342 inst->is_deleted = true; /* flag for ip_set_nfnl_put */
2344 nfnl_lock(NFNL_SUBSYS_IPSET);
2345 for (i = 0; i < inst->ip_set_max; i++) {
2346 set = ip_set(inst, i);
2348 ip_set(inst, i) = NULL;
2349 ip_set_destroy_set(set);
2352 nfnl_unlock(NFNL_SUBSYS_IPSET);
2353 kvfree(rcu_dereference_protected(inst->ip_set_list, 1));
2356 static struct pernet_operations ip_set_net_ops = {
2357 .init = ip_set_net_init,
2358 .exit = ip_set_net_exit,
2359 .id = &ip_set_net_id,
2360 .size = sizeof(struct ip_set_net),
2366 int ret = register_pernet_subsys(&ip_set_net_ops);
2369 pr_err("ip_set: cannot register pernet_subsys.\n");
2373 ret = nfnetlink_subsys_register(&ip_set_netlink_subsys);
2375 pr_err("ip_set: cannot register with nfnetlink.\n");
2376 unregister_pernet_subsys(&ip_set_net_ops);
2380 ret = nf_register_sockopt(&so_set);
2382 pr_err("SO_SET registry failed: %d\n", ret);
2383 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2384 unregister_pernet_subsys(&ip_set_net_ops);
2394 nf_unregister_sockopt(&so_set);
2395 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2397 unregister_pernet_subsys(&ip_set_net_ops);
2398 pr_debug("these are the famous last words\n");
2401 module_init(ip_set_init);
2402 module_exit(ip_set_fini);
2404 MODULE_DESCRIPTION("ip_set: protocol " __stringify(IPSET_PROTOCOL));