1 /* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
2 * Patrick Schaaf <bof@bof.de>
3 * Copyright (C) 2003-2013 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License version 2 as
7 * published by the Free Software Foundation.
10 /* Kernel module for IP set management */
12 #include <linux/init.h>
13 #include <linux/module.h>
14 #include <linux/moduleparam.h>
16 #include <linux/skbuff.h>
17 #include <linux/spinlock.h>
18 #include <linux/rculist.h>
19 #include <net/netlink.h>
20 #include <net/net_namespace.h>
21 #include <net/netns/generic.h>
23 #include <linux/netfilter.h>
24 #include <linux/netfilter/x_tables.h>
25 #include <linux/netfilter/nfnetlink.h>
26 #include <linux/netfilter/ipset/ip_set.h>
28 static LIST_HEAD(ip_set_type_list); /* all registered set types */
29 static DEFINE_MUTEX(ip_set_type_mutex); /* protects ip_set_type_list */
30 static DEFINE_RWLOCK(ip_set_ref_lock); /* protects the set refs */
33 struct ip_set * __rcu *ip_set_list; /* all individual sets */
34 ip_set_id_t ip_set_max; /* max number of sets */
35 bool is_deleted; /* deleted by ip_set_net_exit */
36 bool is_destroyed; /* all sets are destroyed */
39 static unsigned int ip_set_net_id __read_mostly;
41 static inline struct ip_set_net *ip_set_pernet(struct net *net)
43 return net_generic(net, ip_set_net_id);
47 #define STRNCMP(a, b) (strncmp(a, b, IPSET_MAXNAMELEN) == 0)
49 static unsigned int max_sets;
51 module_param(max_sets, int, 0600);
52 MODULE_PARM_DESC(max_sets, "maximal number of sets");
53 MODULE_LICENSE("GPL");
54 MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
55 MODULE_DESCRIPTION("core IP set support");
56 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_IPSET);
58 /* When the nfnl mutex is held: */
59 #define ip_set_dereference(p) \
60 rcu_dereference_protected(p, lockdep_nfnl_is_held(NFNL_SUBSYS_IPSET))
61 #define ip_set(inst, id) \
62 ip_set_dereference((inst)->ip_set_list)[id]
64 /* The set types are implemented in modules and registered set types
65 * can be found in ip_set_type_list. Adding/deleting types is
66 * serialized by ip_set_type_mutex.
70 ip_set_type_lock(void)
72 mutex_lock(&ip_set_type_mutex);
76 ip_set_type_unlock(void)
78 mutex_unlock(&ip_set_type_mutex);
81 /* Register and deregister settype */
83 static struct ip_set_type *
84 find_set_type(const char *name, u8 family, u8 revision)
86 struct ip_set_type *type;
88 list_for_each_entry_rcu(type, &ip_set_type_list, list)
89 if (STRNCMP(type->name, name) &&
90 (type->family == family ||
91 type->family == NFPROTO_UNSPEC) &&
92 revision >= type->revision_min &&
93 revision <= type->revision_max)
98 /* Unlock, try to load a set type module and lock again */
100 load_settype(const char *name)
102 nfnl_unlock(NFNL_SUBSYS_IPSET);
103 pr_debug("try to load ip_set_%s\n", name);
104 if (request_module("ip_set_%s", name) < 0) {
105 pr_warn("Can't find ip_set type %s\n", name);
106 nfnl_lock(NFNL_SUBSYS_IPSET);
109 nfnl_lock(NFNL_SUBSYS_IPSET);
113 /* Find a set type and reference it */
114 #define find_set_type_get(name, family, revision, found) \
115 __find_set_type_get(name, family, revision, found, false)
118 __find_set_type_get(const char *name, u8 family, u8 revision,
119 struct ip_set_type **found, bool retry)
121 struct ip_set_type *type;
124 if (retry && !load_settype(name))
125 return -IPSET_ERR_FIND_TYPE;
128 *found = find_set_type(name, family, revision);
130 err = !try_module_get((*found)->me) ? -EFAULT : 0;
133 /* Make sure the type is already loaded
134 * but we don't support the revision
136 list_for_each_entry_rcu(type, &ip_set_type_list, list)
137 if (STRNCMP(type->name, name)) {
138 err = -IPSET_ERR_FIND_TYPE;
143 return retry ? -IPSET_ERR_FIND_TYPE :
144 __find_set_type_get(name, family, revision, found, true);
151 /* Find a given set type by name and family.
152 * If we succeeded, the supported minimal and maximum revisions are
155 #define find_set_type_minmax(name, family, min, max) \
156 __find_set_type_minmax(name, family, min, max, false)
159 __find_set_type_minmax(const char *name, u8 family, u8 *min, u8 *max,
162 struct ip_set_type *type;
165 if (retry && !load_settype(name))
166 return -IPSET_ERR_FIND_TYPE;
168 *min = 255; *max = 0;
170 list_for_each_entry_rcu(type, &ip_set_type_list, list)
171 if (STRNCMP(type->name, name) &&
172 (type->family == family ||
173 type->family == NFPROTO_UNSPEC)) {
175 if (type->revision_min < *min)
176 *min = type->revision_min;
177 if (type->revision_max > *max)
178 *max = type->revision_max;
184 return retry ? -IPSET_ERR_FIND_TYPE :
185 __find_set_type_minmax(name, family, min, max, true);
188 #define family_name(f) ((f) == NFPROTO_IPV4 ? "inet" : \
189 (f) == NFPROTO_IPV6 ? "inet6" : "any")
191 /* Register a set type structure. The type is identified by
192 * the unique triple of name, family and revision.
195 ip_set_type_register(struct ip_set_type *type)
199 if (type->protocol != IPSET_PROTOCOL) {
200 pr_warn("ip_set type %s, family %s, revision %u:%u uses wrong protocol version %u (want %u)\n",
201 type->name, family_name(type->family),
202 type->revision_min, type->revision_max,
203 type->protocol, IPSET_PROTOCOL);
208 if (find_set_type(type->name, type->family, type->revision_min)) {
210 pr_warn("ip_set type %s, family %s with revision min %u already registered!\n",
211 type->name, family_name(type->family),
213 ip_set_type_unlock();
216 list_add_rcu(&type->list, &ip_set_type_list);
217 pr_debug("type %s, family %s, revision %u:%u registered.\n",
218 type->name, family_name(type->family),
219 type->revision_min, type->revision_max);
220 ip_set_type_unlock();
224 EXPORT_SYMBOL_GPL(ip_set_type_register);
226 /* Unregister a set type. There's a small race with ip_set_create */
228 ip_set_type_unregister(struct ip_set_type *type)
231 if (!find_set_type(type->name, type->family, type->revision_min)) {
232 pr_warn("ip_set type %s, family %s with revision min %u not registered\n",
233 type->name, family_name(type->family),
235 ip_set_type_unlock();
238 list_del_rcu(&type->list);
239 pr_debug("type %s, family %s with revision min %u unregistered.\n",
240 type->name, family_name(type->family), type->revision_min);
241 ip_set_type_unlock();
245 EXPORT_SYMBOL_GPL(ip_set_type_unregister);
247 /* Utility functions */
249 ip_set_alloc(size_t size)
251 void *members = NULL;
253 if (size < KMALLOC_MAX_SIZE)
254 members = kzalloc(size, GFP_KERNEL | __GFP_NOWARN);
257 pr_debug("%p: allocated with kmalloc\n", members);
261 members = vzalloc(size);
264 pr_debug("%p: allocated with vmalloc\n", members);
268 EXPORT_SYMBOL_GPL(ip_set_alloc);
271 ip_set_free(void *members)
273 pr_debug("%p: free with %s\n", members,
274 is_vmalloc_addr(members) ? "vfree" : "kfree");
277 EXPORT_SYMBOL_GPL(ip_set_free);
280 flag_nested(const struct nlattr *nla)
282 return nla->nla_type & NLA_F_NESTED;
285 static const struct nla_policy ipaddr_policy[IPSET_ATTR_IPADDR_MAX + 1] = {
286 [IPSET_ATTR_IPADDR_IPV4] = { .type = NLA_U32 },
287 [IPSET_ATTR_IPADDR_IPV6] = { .type = NLA_BINARY,
288 .len = sizeof(struct in6_addr) },
292 ip_set_get_ipaddr4(struct nlattr *nla, __be32 *ipaddr)
294 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1];
296 if (unlikely(!flag_nested(nla)))
297 return -IPSET_ERR_PROTOCOL;
298 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla,
299 ipaddr_policy, NULL))
300 return -IPSET_ERR_PROTOCOL;
301 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV4)))
302 return -IPSET_ERR_PROTOCOL;
304 *ipaddr = nla_get_be32(tb[IPSET_ATTR_IPADDR_IPV4]);
307 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr4);
310 ip_set_get_ipaddr6(struct nlattr *nla, union nf_inet_addr *ipaddr)
312 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1];
314 if (unlikely(!flag_nested(nla)))
315 return -IPSET_ERR_PROTOCOL;
317 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla,
318 ipaddr_policy, NULL))
319 return -IPSET_ERR_PROTOCOL;
320 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV6)))
321 return -IPSET_ERR_PROTOCOL;
323 memcpy(ipaddr, nla_data(tb[IPSET_ATTR_IPADDR_IPV6]),
324 sizeof(struct in6_addr));
327 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr6);
329 typedef void (*destroyer)(struct ip_set *, void *);
330 /* ipset data extension types, in size order */
332 const struct ip_set_ext_type ip_set_extensions[] = {
333 [IPSET_EXT_ID_COUNTER] = {
334 .type = IPSET_EXT_COUNTER,
335 .flag = IPSET_FLAG_WITH_COUNTERS,
336 .len = sizeof(struct ip_set_counter),
337 .align = __alignof__(struct ip_set_counter),
339 [IPSET_EXT_ID_TIMEOUT] = {
340 .type = IPSET_EXT_TIMEOUT,
341 .len = sizeof(unsigned long),
342 .align = __alignof__(unsigned long),
344 [IPSET_EXT_ID_SKBINFO] = {
345 .type = IPSET_EXT_SKBINFO,
346 .flag = IPSET_FLAG_WITH_SKBINFO,
347 .len = sizeof(struct ip_set_skbinfo),
348 .align = __alignof__(struct ip_set_skbinfo),
350 [IPSET_EXT_ID_COMMENT] = {
351 .type = IPSET_EXT_COMMENT | IPSET_EXT_DESTROY,
352 .flag = IPSET_FLAG_WITH_COMMENT,
353 .len = sizeof(struct ip_set_comment),
354 .align = __alignof__(struct ip_set_comment),
355 .destroy = (destroyer) ip_set_comment_free,
358 EXPORT_SYMBOL_GPL(ip_set_extensions);
361 add_extension(enum ip_set_ext_id id, u32 flags, struct nlattr *tb[])
363 return ip_set_extensions[id].flag ?
364 (flags & ip_set_extensions[id].flag) :
365 !!tb[IPSET_ATTR_TIMEOUT];
369 ip_set_elem_len(struct ip_set *set, struct nlattr *tb[], size_t len,
372 enum ip_set_ext_id id;
375 if (tb[IPSET_ATTR_CADT_FLAGS])
376 cadt_flags = ip_set_get_h32(tb[IPSET_ATTR_CADT_FLAGS]);
377 if (cadt_flags & IPSET_FLAG_WITH_FORCEADD)
378 set->flags |= IPSET_CREATE_FLAG_FORCEADD;
381 for (id = 0; id < IPSET_EXT_ID_MAX; id++) {
382 if (!add_extension(id, cadt_flags, tb))
384 len = ALIGN(len, ip_set_extensions[id].align);
385 set->offset[id] = len;
386 set->extensions |= ip_set_extensions[id].type;
387 len += ip_set_extensions[id].len;
389 return ALIGN(len, align);
391 EXPORT_SYMBOL_GPL(ip_set_elem_len);
394 ip_set_get_extensions(struct ip_set *set, struct nlattr *tb[],
395 struct ip_set_ext *ext)
399 if (unlikely(!ip_set_optattr_netorder(tb, IPSET_ATTR_TIMEOUT) ||
400 !ip_set_optattr_netorder(tb, IPSET_ATTR_PACKETS) ||
401 !ip_set_optattr_netorder(tb, IPSET_ATTR_BYTES) ||
402 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBMARK) ||
403 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBPRIO) ||
404 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBQUEUE)))
405 return -IPSET_ERR_PROTOCOL;
407 if (tb[IPSET_ATTR_TIMEOUT]) {
408 if (!SET_WITH_TIMEOUT(set))
409 return -IPSET_ERR_TIMEOUT;
410 ext->timeout = ip_set_timeout_uget(tb[IPSET_ATTR_TIMEOUT]);
412 if (tb[IPSET_ATTR_BYTES] || tb[IPSET_ATTR_PACKETS]) {
413 if (!SET_WITH_COUNTER(set))
414 return -IPSET_ERR_COUNTER;
415 if (tb[IPSET_ATTR_BYTES])
416 ext->bytes = be64_to_cpu(nla_get_be64(
417 tb[IPSET_ATTR_BYTES]));
418 if (tb[IPSET_ATTR_PACKETS])
419 ext->packets = be64_to_cpu(nla_get_be64(
420 tb[IPSET_ATTR_PACKETS]));
422 if (tb[IPSET_ATTR_COMMENT]) {
423 if (!SET_WITH_COMMENT(set))
424 return -IPSET_ERR_COMMENT;
425 ext->comment = ip_set_comment_uget(tb[IPSET_ATTR_COMMENT]);
427 if (tb[IPSET_ATTR_SKBMARK]) {
428 if (!SET_WITH_SKBINFO(set))
429 return -IPSET_ERR_SKBINFO;
430 fullmark = be64_to_cpu(nla_get_be64(tb[IPSET_ATTR_SKBMARK]));
431 ext->skbinfo.skbmark = fullmark >> 32;
432 ext->skbinfo.skbmarkmask = fullmark & 0xffffffff;
434 if (tb[IPSET_ATTR_SKBPRIO]) {
435 if (!SET_WITH_SKBINFO(set))
436 return -IPSET_ERR_SKBINFO;
437 ext->skbinfo.skbprio =
438 be32_to_cpu(nla_get_be32(tb[IPSET_ATTR_SKBPRIO]));
440 if (tb[IPSET_ATTR_SKBQUEUE]) {
441 if (!SET_WITH_SKBINFO(set))
442 return -IPSET_ERR_SKBINFO;
443 ext->skbinfo.skbqueue =
444 be16_to_cpu(nla_get_be16(tb[IPSET_ATTR_SKBQUEUE]));
448 EXPORT_SYMBOL_GPL(ip_set_get_extensions);
451 ip_set_put_extensions(struct sk_buff *skb, const struct ip_set *set,
452 const void *e, bool active)
454 if (SET_WITH_TIMEOUT(set)) {
455 unsigned long *timeout = ext_timeout(e, set);
457 if (nla_put_net32(skb, IPSET_ATTR_TIMEOUT,
458 htonl(active ? ip_set_timeout_get(timeout)
462 if (SET_WITH_COUNTER(set) &&
463 ip_set_put_counter(skb, ext_counter(e, set)))
465 if (SET_WITH_COMMENT(set) &&
466 ip_set_put_comment(skb, ext_comment(e, set)))
468 if (SET_WITH_SKBINFO(set) &&
469 ip_set_put_skbinfo(skb, ext_skbinfo(e, set)))
473 EXPORT_SYMBOL_GPL(ip_set_put_extensions);
476 ip_set_match_extensions(struct ip_set *set, const struct ip_set_ext *ext,
477 struct ip_set_ext *mext, u32 flags, void *data)
479 if (SET_WITH_TIMEOUT(set) &&
480 ip_set_timeout_expired(ext_timeout(data, set)))
482 if (SET_WITH_COUNTER(set)) {
483 struct ip_set_counter *counter = ext_counter(data, set);
485 if (flags & IPSET_FLAG_MATCH_COUNTERS &&
486 !(ip_set_match_counter(ip_set_get_packets(counter),
487 mext->packets, mext->packets_op) &&
488 ip_set_match_counter(ip_set_get_bytes(counter),
489 mext->bytes, mext->bytes_op)))
491 ip_set_update_counter(counter, ext, flags);
493 if (SET_WITH_SKBINFO(set))
494 ip_set_get_skbinfo(ext_skbinfo(data, set),
498 EXPORT_SYMBOL_GPL(ip_set_match_extensions);
500 /* Creating/destroying/renaming/swapping affect the existence and
501 * the properties of a set. All of these can be executed from userspace
502 * only and serialized by the nfnl mutex indirectly from nfnetlink.
504 * Sets are identified by their index in ip_set_list and the index
505 * is used by the external references (set/SET netfilter modules).
507 * The set behind an index may change by swapping only, from userspace.
511 __ip_set_get(struct ip_set *set)
513 write_lock_bh(&ip_set_ref_lock);
515 write_unlock_bh(&ip_set_ref_lock);
519 __ip_set_put(struct ip_set *set)
521 write_lock_bh(&ip_set_ref_lock);
522 BUG_ON(set->ref == 0);
524 write_unlock_bh(&ip_set_ref_lock);
527 /* set->ref can be swapped out by ip_set_swap, netlink events (like dump) need
528 * a separate reference counter
531 __ip_set_put_netlink(struct ip_set *set)
533 write_lock_bh(&ip_set_ref_lock);
534 BUG_ON(set->ref_netlink == 0);
536 write_unlock_bh(&ip_set_ref_lock);
539 /* Add, del and test set entries from kernel.
541 * The set behind the index must exist and must be referenced
542 * so it can't be destroyed (or changed) under our foot.
545 static inline struct ip_set *
546 ip_set_rcu_get(struct net *net, ip_set_id_t index)
549 struct ip_set_net *inst = ip_set_pernet(net);
552 /* ip_set_list itself needs to be protected */
553 set = rcu_dereference(inst->ip_set_list)[index];
560 ip_set_test(ip_set_id_t index, const struct sk_buff *skb,
561 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
563 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
567 pr_debug("set %s, index %u\n", set->name, index);
569 if (opt->dim < set->type->dimension ||
570 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
574 ret = set->variant->kadt(set, skb, par, IPSET_TEST, opt);
575 rcu_read_unlock_bh();
577 if (ret == -EAGAIN) {
578 /* Type requests element to be completed */
579 pr_debug("element must be completed, ADD is triggered\n");
580 spin_lock_bh(&set->lock);
581 set->variant->kadt(set, skb, par, IPSET_ADD, opt);
582 spin_unlock_bh(&set->lock);
585 /* --return-nomatch: invert matched element */
586 if ((opt->cmdflags & IPSET_FLAG_RETURN_NOMATCH) &&
587 (set->type->features & IPSET_TYPE_NOMATCH) &&
588 (ret > 0 || ret == -ENOTEMPTY))
592 /* Convert error codes to nomatch */
593 return (ret < 0 ? 0 : ret);
595 EXPORT_SYMBOL_GPL(ip_set_test);
598 ip_set_add(ip_set_id_t index, const struct sk_buff *skb,
599 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
601 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
605 pr_debug("set %s, index %u\n", set->name, index);
607 if (opt->dim < set->type->dimension ||
608 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
609 return -IPSET_ERR_TYPE_MISMATCH;
611 spin_lock_bh(&set->lock);
612 ret = set->variant->kadt(set, skb, par, IPSET_ADD, opt);
613 spin_unlock_bh(&set->lock);
617 EXPORT_SYMBOL_GPL(ip_set_add);
620 ip_set_del(ip_set_id_t index, const struct sk_buff *skb,
621 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
623 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
627 pr_debug("set %s, index %u\n", set->name, index);
629 if (opt->dim < set->type->dimension ||
630 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
631 return -IPSET_ERR_TYPE_MISMATCH;
633 spin_lock_bh(&set->lock);
634 ret = set->variant->kadt(set, skb, par, IPSET_DEL, opt);
635 spin_unlock_bh(&set->lock);
639 EXPORT_SYMBOL_GPL(ip_set_del);
641 /* Find set by name, reference it once. The reference makes sure the
642 * thing pointed to, does not go away under our feet.
646 ip_set_get_byname(struct net *net, const char *name, struct ip_set **set)
648 ip_set_id_t i, index = IPSET_INVALID_ID;
650 struct ip_set_net *inst = ip_set_pernet(net);
653 for (i = 0; i < inst->ip_set_max; i++) {
654 s = rcu_dereference(inst->ip_set_list)[i];
655 if (s && STRNCMP(s->name, name)) {
666 EXPORT_SYMBOL_GPL(ip_set_get_byname);
668 /* If the given set pointer points to a valid set, decrement
669 * reference count by 1. The caller shall not assume the index
670 * to be valid, after calling this function.
675 __ip_set_put_byindex(struct ip_set_net *inst, ip_set_id_t index)
680 set = rcu_dereference(inst->ip_set_list)[index];
687 ip_set_put_byindex(struct net *net, ip_set_id_t index)
689 struct ip_set_net *inst = ip_set_pernet(net);
691 __ip_set_put_byindex(inst, index);
693 EXPORT_SYMBOL_GPL(ip_set_put_byindex);
695 /* Get the name of a set behind a set index.
696 * Set itself is protected by RCU, but its name isn't: to protect against
697 * renaming, grab ip_set_ref_lock as reader (see ip_set_rename()) and copy the
701 ip_set_name_byindex(struct net *net, ip_set_id_t index, char *name)
703 struct ip_set *set = ip_set_rcu_get(net, index);
707 read_lock_bh(&ip_set_ref_lock);
708 strncpy(name, set->name, IPSET_MAXNAMELEN);
709 read_unlock_bh(&ip_set_ref_lock);
711 EXPORT_SYMBOL_GPL(ip_set_name_byindex);
713 /* Routines to call by external subsystems, which do not
714 * call nfnl_lock for us.
717 /* Find set by index, reference it once. The reference makes sure the
718 * thing pointed to, does not go away under our feet.
720 * The nfnl mutex is used in the function.
723 ip_set_nfnl_get_byindex(struct net *net, ip_set_id_t index)
726 struct ip_set_net *inst = ip_set_pernet(net);
728 if (index >= inst->ip_set_max)
729 return IPSET_INVALID_ID;
731 nfnl_lock(NFNL_SUBSYS_IPSET);
732 set = ip_set(inst, index);
736 index = IPSET_INVALID_ID;
737 nfnl_unlock(NFNL_SUBSYS_IPSET);
741 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex);
743 /* If the given set pointer points to a valid set, decrement
744 * reference count by 1. The caller shall not assume the index
745 * to be valid, after calling this function.
747 * The nfnl mutex is used in the function.
750 ip_set_nfnl_put(struct net *net, ip_set_id_t index)
753 struct ip_set_net *inst = ip_set_pernet(net);
755 nfnl_lock(NFNL_SUBSYS_IPSET);
756 if (!inst->is_deleted) { /* already deleted from ip_set_net_exit() */
757 set = ip_set(inst, index);
761 nfnl_unlock(NFNL_SUBSYS_IPSET);
763 EXPORT_SYMBOL_GPL(ip_set_nfnl_put);
765 /* Communication protocol with userspace over netlink.
767 * The commands are serialized by the nfnl mutex.
771 protocol_failed(const struct nlattr * const tb[])
773 return !tb[IPSET_ATTR_PROTOCOL] ||
774 nla_get_u8(tb[IPSET_ATTR_PROTOCOL]) != IPSET_PROTOCOL;
778 flag_exist(const struct nlmsghdr *nlh)
780 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST;
783 static struct nlmsghdr *
784 start_msg(struct sk_buff *skb, u32 portid, u32 seq, unsigned int flags,
787 struct nlmsghdr *nlh;
788 struct nfgenmsg *nfmsg;
790 nlh = nlmsg_put(skb, portid, seq, nfnl_msg_type(NFNL_SUBSYS_IPSET, cmd),
791 sizeof(*nfmsg), flags);
795 nfmsg = nlmsg_data(nlh);
796 nfmsg->nfgen_family = NFPROTO_IPV4;
797 nfmsg->version = NFNETLINK_V0;
805 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = {
806 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
807 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
808 .len = IPSET_MAXNAMELEN - 1 },
809 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
810 .len = IPSET_MAXNAMELEN - 1},
811 [IPSET_ATTR_REVISION] = { .type = NLA_U8 },
812 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
813 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
816 static struct ip_set *
817 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id)
819 struct ip_set *set = NULL;
822 *id = IPSET_INVALID_ID;
823 for (i = 0; i < inst->ip_set_max; i++) {
824 set = ip_set(inst, i);
825 if (set && STRNCMP(set->name, name)) {
830 return (*id == IPSET_INVALID_ID ? NULL : set);
833 static inline struct ip_set *
834 find_set(struct ip_set_net *inst, const char *name)
838 return find_set_and_id(inst, name, &id);
842 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index,
848 *index = IPSET_INVALID_ID;
849 for (i = 0; i < inst->ip_set_max; i++) {
852 if (*index == IPSET_INVALID_ID)
854 } else if (STRNCMP(name, s->name)) {
860 if (*index == IPSET_INVALID_ID)
861 /* No free slot remained */
862 return -IPSET_ERR_MAX_SETS;
866 static int ip_set_none(struct net *net, struct sock *ctnl, struct sk_buff *skb,
867 const struct nlmsghdr *nlh,
868 const struct nlattr * const attr[],
869 struct netlink_ext_ack *extack)
874 static int ip_set_create(struct net *net, struct sock *ctnl,
875 struct sk_buff *skb, const struct nlmsghdr *nlh,
876 const struct nlattr * const attr[],
877 struct netlink_ext_ack *extack)
879 struct ip_set_net *inst = ip_set_pernet(net);
880 struct ip_set *set, *clash = NULL;
881 ip_set_id_t index = IPSET_INVALID_ID;
882 struct nlattr *tb[IPSET_ATTR_CREATE_MAX + 1] = {};
883 const char *name, *typename;
885 u32 flags = flag_exist(nlh);
888 if (unlikely(protocol_failed(attr) ||
889 !attr[IPSET_ATTR_SETNAME] ||
890 !attr[IPSET_ATTR_TYPENAME] ||
891 !attr[IPSET_ATTR_REVISION] ||
892 !attr[IPSET_ATTR_FAMILY] ||
893 (attr[IPSET_ATTR_DATA] &&
894 !flag_nested(attr[IPSET_ATTR_DATA]))))
895 return -IPSET_ERR_PROTOCOL;
897 name = nla_data(attr[IPSET_ATTR_SETNAME]);
898 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
899 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
900 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]);
901 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
902 name, typename, family_name(family), revision);
904 /* First, and without any locks, allocate and initialize
905 * a normal base set structure.
907 set = kzalloc(sizeof(*set), GFP_KERNEL);
910 spin_lock_init(&set->lock);
911 strlcpy(set->name, name, IPSET_MAXNAMELEN);
912 set->family = family;
913 set->revision = revision;
915 /* Next, check that we know the type, and take
916 * a reference on the type, to make sure it stays available
917 * while constructing our new set.
919 * After referencing the type, we try to create the type
920 * specific part of the set without holding any locks.
922 ret = find_set_type_get(typename, family, revision, &set->type);
926 /* Without holding any locks, create private part. */
927 if (attr[IPSET_ATTR_DATA] &&
928 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA],
929 set->type->create_policy, NULL)) {
930 ret = -IPSET_ERR_PROTOCOL;
934 ret = set->type->create(net, set, tb, flags);
938 /* BTW, ret==0 here. */
940 /* Here, we have a valid, constructed set and we are protected
941 * by the nfnl mutex. Find the first free index in ip_set_list
942 * and check clashing.
944 ret = find_free_id(inst, set->name, &index, &clash);
945 if (ret == -EEXIST) {
946 /* If this is the same set and requested, ignore error */
947 if ((flags & IPSET_FLAG_EXIST) &&
948 STRNCMP(set->type->name, clash->type->name) &&
949 set->type->family == clash->type->family &&
950 set->type->revision_min == clash->type->revision_min &&
951 set->type->revision_max == clash->type->revision_max &&
952 set->variant->same_set(set, clash))
955 } else if (ret == -IPSET_ERR_MAX_SETS) {
956 struct ip_set **list, **tmp;
957 ip_set_id_t i = inst->ip_set_max + IP_SET_INC;
959 if (i < inst->ip_set_max || i == IPSET_INVALID_ID)
963 list = kvcalloc(i, sizeof(struct ip_set *), GFP_KERNEL);
966 /* nfnl mutex is held, both lists are valid */
967 tmp = ip_set_dereference(inst->ip_set_list);
968 memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max);
969 rcu_assign_pointer(inst->ip_set_list, list);
970 /* Make sure all current packets have passed through */
973 index = inst->ip_set_max;
974 inst->ip_set_max = i;
981 /* Finally! Add our shiny new set to the list, and be done. */
982 pr_debug("create: '%s' created with index %u!\n", set->name, index);
983 ip_set(inst, index) = set;
988 set->variant->destroy(set);
990 module_put(set->type->me);
998 static const struct nla_policy
999 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = {
1000 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1001 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1002 .len = IPSET_MAXNAMELEN - 1 },
1006 ip_set_destroy_set(struct ip_set *set)
1008 pr_debug("set: %s\n", set->name);
1010 /* Must call it without holding any lock */
1011 set->variant->destroy(set);
1012 module_put(set->type->me);
1016 static int ip_set_destroy(struct net *net, struct sock *ctnl,
1017 struct sk_buff *skb, const struct nlmsghdr *nlh,
1018 const struct nlattr * const attr[],
1019 struct netlink_ext_ack *extack)
1021 struct ip_set_net *inst = ip_set_pernet(net);
1026 if (unlikely(protocol_failed(attr)))
1027 return -IPSET_ERR_PROTOCOL;
1029 /* Must wait for flush to be really finished in list:set */
1032 /* Commands are serialized and references are
1033 * protected by the ip_set_ref_lock.
1034 * External systems (i.e. xt_set) must call
1035 * ip_set_put|get_nfnl_* functions, that way we
1036 * can safely check references here.
1038 * list:set timer can only decrement the reference
1039 * counter, so if it's already zero, we can proceed
1040 * without holding the lock.
1042 read_lock_bh(&ip_set_ref_lock);
1043 if (!attr[IPSET_ATTR_SETNAME]) {
1044 for (i = 0; i < inst->ip_set_max; i++) {
1045 s = ip_set(inst, i);
1046 if (s && (s->ref || s->ref_netlink)) {
1047 ret = -IPSET_ERR_BUSY;
1051 inst->is_destroyed = true;
1052 read_unlock_bh(&ip_set_ref_lock);
1053 for (i = 0; i < inst->ip_set_max; i++) {
1054 s = ip_set(inst, i);
1056 ip_set(inst, i) = NULL;
1057 ip_set_destroy_set(s);
1060 /* Modified by ip_set_destroy() only, which is serialized */
1061 inst->is_destroyed = false;
1063 s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1068 } else if (s->ref || s->ref_netlink) {
1069 ret = -IPSET_ERR_BUSY;
1072 ip_set(inst, i) = NULL;
1073 read_unlock_bh(&ip_set_ref_lock);
1075 ip_set_destroy_set(s);
1079 read_unlock_bh(&ip_set_ref_lock);
1086 ip_set_flush_set(struct ip_set *set)
1088 pr_debug("set: %s\n", set->name);
1090 spin_lock_bh(&set->lock);
1091 set->variant->flush(set);
1092 spin_unlock_bh(&set->lock);
1095 static int ip_set_flush(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1096 const struct nlmsghdr *nlh,
1097 const struct nlattr * const attr[],
1098 struct netlink_ext_ack *extack)
1100 struct ip_set_net *inst = ip_set_pernet(net);
1104 if (unlikely(protocol_failed(attr)))
1105 return -IPSET_ERR_PROTOCOL;
1107 if (!attr[IPSET_ATTR_SETNAME]) {
1108 for (i = 0; i < inst->ip_set_max; i++) {
1109 s = ip_set(inst, i);
1111 ip_set_flush_set(s);
1114 s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1118 ip_set_flush_set(s);
1126 static const struct nla_policy
1127 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = {
1128 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1129 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1130 .len = IPSET_MAXNAMELEN - 1 },
1131 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING,
1132 .len = IPSET_MAXNAMELEN - 1 },
1135 static int ip_set_rename(struct net *net, struct sock *ctnl,
1136 struct sk_buff *skb, const struct nlmsghdr *nlh,
1137 const struct nlattr * const attr[],
1138 struct netlink_ext_ack *extack)
1140 struct ip_set_net *inst = ip_set_pernet(net);
1141 struct ip_set *set, *s;
1146 if (unlikely(protocol_failed(attr) ||
1147 !attr[IPSET_ATTR_SETNAME] ||
1148 !attr[IPSET_ATTR_SETNAME2]))
1149 return -IPSET_ERR_PROTOCOL;
1151 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1155 write_lock_bh(&ip_set_ref_lock);
1156 if (set->ref != 0) {
1157 ret = -IPSET_ERR_REFERENCED;
1161 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]);
1162 for (i = 0; i < inst->ip_set_max; i++) {
1163 s = ip_set(inst, i);
1164 if (s && STRNCMP(s->name, name2)) {
1165 ret = -IPSET_ERR_EXIST_SETNAME2;
1169 strncpy(set->name, name2, IPSET_MAXNAMELEN);
1172 write_unlock_bh(&ip_set_ref_lock);
1176 /* Swap two sets so that name/index points to the other.
1177 * References and set names are also swapped.
1179 * The commands are serialized by the nfnl mutex and references are
1180 * protected by the ip_set_ref_lock. The kernel interfaces
1181 * do not hold the mutex but the pointer settings are atomic
1182 * so the ip_set_list always contains valid pointers to the sets.
1185 static int ip_set_swap(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1186 const struct nlmsghdr *nlh,
1187 const struct nlattr * const attr[],
1188 struct netlink_ext_ack *extack)
1190 struct ip_set_net *inst = ip_set_pernet(net);
1191 struct ip_set *from, *to;
1192 ip_set_id_t from_id, to_id;
1193 char from_name[IPSET_MAXNAMELEN];
1195 if (unlikely(protocol_failed(attr) ||
1196 !attr[IPSET_ATTR_SETNAME] ||
1197 !attr[IPSET_ATTR_SETNAME2]))
1198 return -IPSET_ERR_PROTOCOL;
1200 from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1205 to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]),
1208 return -IPSET_ERR_EXIST_SETNAME2;
1210 /* Features must not change.
1211 * Not an artifical restriction anymore, as we must prevent
1212 * possible loops created by swapping in setlist type of sets.
1214 if (!(from->type->features == to->type->features &&
1215 from->family == to->family))
1216 return -IPSET_ERR_TYPE_MISMATCH;
1218 write_lock_bh(&ip_set_ref_lock);
1220 if (from->ref_netlink || to->ref_netlink) {
1221 write_unlock_bh(&ip_set_ref_lock);
1225 strncpy(from_name, from->name, IPSET_MAXNAMELEN);
1226 strncpy(from->name, to->name, IPSET_MAXNAMELEN);
1227 strncpy(to->name, from_name, IPSET_MAXNAMELEN);
1229 swap(from->ref, to->ref);
1230 ip_set(inst, from_id) = to;
1231 ip_set(inst, to_id) = from;
1232 write_unlock_bh(&ip_set_ref_lock);
1237 /* List/save set data */
1244 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1245 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1248 ip_set_dump_done(struct netlink_callback *cb)
1250 if (cb->args[IPSET_CB_ARG0]) {
1251 struct ip_set_net *inst =
1252 (struct ip_set_net *)cb->args[IPSET_CB_NET];
1253 ip_set_id_t index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1254 struct ip_set *set = ip_set(inst, index);
1256 if (set->variant->uref)
1257 set->variant->uref(set, cb, false);
1258 pr_debug("release set %s\n", set->name);
1259 __ip_set_put_netlink(set);
1265 dump_attrs(struct nlmsghdr *nlh)
1267 const struct nlattr *attr;
1270 pr_debug("dump nlmsg\n");
1271 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) {
1272 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len);
1277 dump_init(struct netlink_callback *cb, struct ip_set_net *inst)
1279 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb);
1280 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1281 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1282 struct nlattr *attr = (void *)nlh + min_len;
1286 /* Second pass, so parser can't fail */
1287 nla_parse(cda, IPSET_ATTR_CMD_MAX, attr, nlh->nlmsg_len - min_len,
1288 ip_set_setname_policy, NULL);
1290 if (cda[IPSET_ATTR_SETNAME]) {
1293 set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]),
1298 dump_type = DUMP_ONE;
1299 cb->args[IPSET_CB_INDEX] = index;
1301 dump_type = DUMP_ALL;
1304 if (cda[IPSET_ATTR_FLAGS]) {
1305 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]);
1307 dump_type |= (f << 16);
1309 cb->args[IPSET_CB_NET] = (unsigned long)inst;
1310 cb->args[IPSET_CB_DUMP] = dump_type;
1316 ip_set_dump_start(struct sk_buff *skb, struct netlink_callback *cb)
1318 ip_set_id_t index = IPSET_INVALID_ID, max;
1319 struct ip_set *set = NULL;
1320 struct nlmsghdr *nlh = NULL;
1321 unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0;
1322 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1323 u32 dump_type, dump_flags;
1327 if (!cb->args[IPSET_CB_DUMP]) {
1328 ret = dump_init(cb, inst);
1330 nlh = nlmsg_hdr(cb->skb);
1331 /* We have to create and send the error message
1334 if (nlh->nlmsg_flags & NLM_F_ACK)
1335 netlink_ack(cb->skb, nlh, ret, NULL);
1340 if (cb->args[IPSET_CB_INDEX] >= inst->ip_set_max)
1343 dump_type = DUMP_TYPE(cb->args[IPSET_CB_DUMP]);
1344 dump_flags = DUMP_FLAGS(cb->args[IPSET_CB_DUMP]);
1345 max = dump_type == DUMP_ONE ? cb->args[IPSET_CB_INDEX] + 1
1348 pr_debug("dump type, flag: %u %u index: %ld\n",
1349 dump_type, dump_flags, cb->args[IPSET_CB_INDEX]);
1350 for (; cb->args[IPSET_CB_INDEX] < max; cb->args[IPSET_CB_INDEX]++) {
1351 index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1352 write_lock_bh(&ip_set_ref_lock);
1353 set = ip_set(inst, index);
1354 is_destroyed = inst->is_destroyed;
1355 if (!set || is_destroyed) {
1356 write_unlock_bh(&ip_set_ref_lock);
1357 if (dump_type == DUMP_ONE) {
1362 /* All sets are just being destroyed */
1368 /* When dumping all sets, we must dump "sorted"
1369 * so that lists (unions of sets) are dumped last.
1371 if (dump_type != DUMP_ONE &&
1372 ((dump_type == DUMP_ALL) ==
1373 !!(set->type->features & IPSET_DUMP_LAST))) {
1374 write_unlock_bh(&ip_set_ref_lock);
1377 pr_debug("List set: %s\n", set->name);
1378 if (!cb->args[IPSET_CB_ARG0]) {
1379 /* Start listing: make sure set won't be destroyed */
1380 pr_debug("reference set\n");
1383 write_unlock_bh(&ip_set_ref_lock);
1384 nlh = start_msg(skb, NETLINK_CB(cb->skb).portid,
1385 cb->nlh->nlmsg_seq, flags,
1389 goto release_refcount;
1391 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1392 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name))
1393 goto nla_put_failure;
1394 if (dump_flags & IPSET_FLAG_LIST_SETNAME)
1396 switch (cb->args[IPSET_CB_ARG0]) {
1398 /* Core header data */
1399 if (nla_put_string(skb, IPSET_ATTR_TYPENAME,
1401 nla_put_u8(skb, IPSET_ATTR_FAMILY,
1403 nla_put_u8(skb, IPSET_ATTR_REVISION,
1405 goto nla_put_failure;
1406 ret = set->variant->head(set, skb);
1408 goto release_refcount;
1409 if (dump_flags & IPSET_FLAG_LIST_HEADER)
1411 if (set->variant->uref)
1412 set->variant->uref(set, cb, true);
1415 ret = set->variant->list(set, skb, cb);
1416 if (!cb->args[IPSET_CB_ARG0])
1417 /* Set is done, proceed with next one */
1419 goto release_refcount;
1422 /* If we dump all sets, continue with dumping last ones */
1423 if (dump_type == DUMP_ALL) {
1424 dump_type = DUMP_LAST;
1425 cb->args[IPSET_CB_DUMP] = dump_type | (dump_flags << 16);
1426 cb->args[IPSET_CB_INDEX] = 0;
1427 if (set && set->variant->uref)
1428 set->variant->uref(set, cb, false);
1436 if (dump_type == DUMP_ONE)
1437 cb->args[IPSET_CB_INDEX] = IPSET_INVALID_ID;
1439 cb->args[IPSET_CB_INDEX]++;
1441 /* If there was an error or set is done, release set */
1442 if (ret || !cb->args[IPSET_CB_ARG0]) {
1443 set = ip_set(inst, index);
1444 if (set->variant->uref)
1445 set->variant->uref(set, cb, false);
1446 pr_debug("release set %s\n", set->name);
1447 __ip_set_put_netlink(set);
1448 cb->args[IPSET_CB_ARG0] = 0;
1452 nlmsg_end(skb, nlh);
1453 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len);
1457 return ret < 0 ? ret : skb->len;
1460 static int ip_set_dump(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1461 const struct nlmsghdr *nlh,
1462 const struct nlattr * const attr[],
1463 struct netlink_ext_ack *extack)
1465 if (unlikely(protocol_failed(attr)))
1466 return -IPSET_ERR_PROTOCOL;
1469 struct netlink_dump_control c = {
1470 .dump = ip_set_dump_start,
1471 .done = ip_set_dump_done,
1473 return netlink_dump_start(ctnl, skb, nlh, &c);
1477 /* Add, del and test */
1479 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = {
1480 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1481 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1482 .len = IPSET_MAXNAMELEN - 1 },
1483 [IPSET_ATTR_LINENO] = { .type = NLA_U32 },
1484 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
1485 [IPSET_ATTR_ADT] = { .type = NLA_NESTED },
1489 call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set,
1490 struct nlattr *tb[], enum ipset_adt adt,
1491 u32 flags, bool use_lineno)
1495 bool eexist = flags & IPSET_FLAG_EXIST, retried = false;
1498 spin_lock_bh(&set->lock);
1499 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried);
1500 spin_unlock_bh(&set->lock);
1502 } while (ret == -EAGAIN &&
1503 set->variant->resize &&
1504 (ret = set->variant->resize(set, retried)) == 0);
1506 if (!ret || (ret == -IPSET_ERR_EXIST && eexist))
1508 if (lineno && use_lineno) {
1509 /* Error in restore/batch mode: send back lineno */
1510 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb);
1511 struct sk_buff *skb2;
1512 struct nlmsgerr *errmsg;
1513 size_t payload = min(SIZE_MAX,
1514 sizeof(*errmsg) + nlmsg_len(nlh));
1515 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1516 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1517 struct nlattr *cmdattr;
1520 skb2 = nlmsg_new(payload, GFP_KERNEL);
1523 rep = __nlmsg_put(skb2, NETLINK_CB(skb).portid,
1524 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0);
1525 errmsg = nlmsg_data(rep);
1526 errmsg->error = ret;
1527 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len);
1528 cmdattr = (void *)&errmsg->msg + min_len;
1530 nla_parse(cda, IPSET_ATTR_CMD_MAX, cmdattr,
1531 nlh->nlmsg_len - min_len, ip_set_adt_policy, NULL);
1533 errline = nla_data(cda[IPSET_ATTR_LINENO]);
1537 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid,
1539 /* Signal netlink not to send its ACK/errmsg. */
1546 static int ip_set_uadd(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1547 const struct nlmsghdr *nlh,
1548 const struct nlattr * const attr[],
1549 struct netlink_ext_ack *extack)
1551 struct ip_set_net *inst = ip_set_pernet(net);
1553 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1554 const struct nlattr *nla;
1555 u32 flags = flag_exist(nlh);
1559 if (unlikely(protocol_failed(attr) ||
1560 !attr[IPSET_ATTR_SETNAME] ||
1561 !((attr[IPSET_ATTR_DATA] != NULL) ^
1562 (attr[IPSET_ATTR_ADT] != NULL)) ||
1563 (attr[IPSET_ATTR_DATA] &&
1564 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1565 (attr[IPSET_ATTR_ADT] &&
1566 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1567 !attr[IPSET_ATTR_LINENO]))))
1568 return -IPSET_ERR_PROTOCOL;
1570 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1574 use_lineno = !!attr[IPSET_ATTR_LINENO];
1575 if (attr[IPSET_ATTR_DATA]) {
1576 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1577 attr[IPSET_ATTR_DATA],
1578 set->type->adt_policy, NULL))
1579 return -IPSET_ERR_PROTOCOL;
1580 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD, flags,
1585 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1586 memset(tb, 0, sizeof(tb));
1587 if (nla_type(nla) != IPSET_ATTR_DATA ||
1588 !flag_nested(nla) ||
1589 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1590 set->type->adt_policy, NULL))
1591 return -IPSET_ERR_PROTOCOL;
1592 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD,
1601 static int ip_set_udel(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1602 const struct nlmsghdr *nlh,
1603 const struct nlattr * const attr[],
1604 struct netlink_ext_ack *extack)
1606 struct ip_set_net *inst = ip_set_pernet(net);
1608 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1609 const struct nlattr *nla;
1610 u32 flags = flag_exist(nlh);
1614 if (unlikely(protocol_failed(attr) ||
1615 !attr[IPSET_ATTR_SETNAME] ||
1616 !((attr[IPSET_ATTR_DATA] != NULL) ^
1617 (attr[IPSET_ATTR_ADT] != NULL)) ||
1618 (attr[IPSET_ATTR_DATA] &&
1619 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1620 (attr[IPSET_ATTR_ADT] &&
1621 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1622 !attr[IPSET_ATTR_LINENO]))))
1623 return -IPSET_ERR_PROTOCOL;
1625 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1629 use_lineno = !!attr[IPSET_ATTR_LINENO];
1630 if (attr[IPSET_ATTR_DATA]) {
1631 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1632 attr[IPSET_ATTR_DATA],
1633 set->type->adt_policy, NULL))
1634 return -IPSET_ERR_PROTOCOL;
1635 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL, flags,
1640 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1641 memset(tb, 0, sizeof(*tb));
1642 if (nla_type(nla) != IPSET_ATTR_DATA ||
1643 !flag_nested(nla) ||
1644 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1645 set->type->adt_policy, NULL))
1646 return -IPSET_ERR_PROTOCOL;
1647 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL,
1656 static int ip_set_utest(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1657 const struct nlmsghdr *nlh,
1658 const struct nlattr * const attr[],
1659 struct netlink_ext_ack *extack)
1661 struct ip_set_net *inst = ip_set_pernet(net);
1663 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1666 if (unlikely(protocol_failed(attr) ||
1667 !attr[IPSET_ATTR_SETNAME] ||
1668 !attr[IPSET_ATTR_DATA] ||
1669 !flag_nested(attr[IPSET_ATTR_DATA])))
1670 return -IPSET_ERR_PROTOCOL;
1672 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1676 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA],
1677 set->type->adt_policy, NULL))
1678 return -IPSET_ERR_PROTOCOL;
1681 ret = set->variant->uadt(set, tb, IPSET_TEST, NULL, 0, 0);
1682 rcu_read_unlock_bh();
1683 /* Userspace can't trigger element to be re-added */
1687 return ret > 0 ? 0 : -IPSET_ERR_EXIST;
1690 /* Get headed data of a set */
1692 static int ip_set_header(struct net *net, struct sock *ctnl,
1693 struct sk_buff *skb, const struct nlmsghdr *nlh,
1694 const struct nlattr * const attr[],
1695 struct netlink_ext_ack *extack)
1697 struct ip_set_net *inst = ip_set_pernet(net);
1698 const struct ip_set *set;
1699 struct sk_buff *skb2;
1700 struct nlmsghdr *nlh2;
1703 if (unlikely(protocol_failed(attr) ||
1704 !attr[IPSET_ATTR_SETNAME]))
1705 return -IPSET_ERR_PROTOCOL;
1707 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1711 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1715 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1719 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1720 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) ||
1721 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) ||
1722 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
1723 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision))
1724 goto nla_put_failure;
1725 nlmsg_end(skb2, nlh2);
1727 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1734 nlmsg_cancel(skb2, nlh2);
1742 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = {
1743 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1744 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
1745 .len = IPSET_MAXNAMELEN - 1 },
1746 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
1749 static int ip_set_type(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1750 const struct nlmsghdr *nlh,
1751 const struct nlattr * const attr[],
1752 struct netlink_ext_ack *extack)
1754 struct sk_buff *skb2;
1755 struct nlmsghdr *nlh2;
1756 u8 family, min, max;
1757 const char *typename;
1760 if (unlikely(protocol_failed(attr) ||
1761 !attr[IPSET_ATTR_TYPENAME] ||
1762 !attr[IPSET_ATTR_FAMILY]))
1763 return -IPSET_ERR_PROTOCOL;
1765 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1766 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1767 ret = find_set_type_minmax(typename, family, &min, &max);
1771 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1775 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1779 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1780 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) ||
1781 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) ||
1782 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) ||
1783 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min))
1784 goto nla_put_failure;
1785 nlmsg_end(skb2, nlh2);
1787 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len);
1788 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1795 nlmsg_cancel(skb2, nlh2);
1801 /* Get protocol version */
1803 static const struct nla_policy
1804 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = {
1805 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1808 static int ip_set_protocol(struct net *net, struct sock *ctnl,
1809 struct sk_buff *skb, const struct nlmsghdr *nlh,
1810 const struct nlattr * const attr[],
1811 struct netlink_ext_ack *extack)
1813 struct sk_buff *skb2;
1814 struct nlmsghdr *nlh2;
1817 if (unlikely(!attr[IPSET_ATTR_PROTOCOL]))
1818 return -IPSET_ERR_PROTOCOL;
1820 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1824 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1825 IPSET_CMD_PROTOCOL);
1828 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL))
1829 goto nla_put_failure;
1830 nlmsg_end(skb2, nlh2);
1832 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1839 nlmsg_cancel(skb2, nlh2);
1845 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = {
1846 [IPSET_CMD_NONE] = {
1847 .call = ip_set_none,
1848 .attr_count = IPSET_ATTR_CMD_MAX,
1850 [IPSET_CMD_CREATE] = {
1851 .call = ip_set_create,
1852 .attr_count = IPSET_ATTR_CMD_MAX,
1853 .policy = ip_set_create_policy,
1855 [IPSET_CMD_DESTROY] = {
1856 .call = ip_set_destroy,
1857 .attr_count = IPSET_ATTR_CMD_MAX,
1858 .policy = ip_set_setname_policy,
1860 [IPSET_CMD_FLUSH] = {
1861 .call = ip_set_flush,
1862 .attr_count = IPSET_ATTR_CMD_MAX,
1863 .policy = ip_set_setname_policy,
1865 [IPSET_CMD_RENAME] = {
1866 .call = ip_set_rename,
1867 .attr_count = IPSET_ATTR_CMD_MAX,
1868 .policy = ip_set_setname2_policy,
1870 [IPSET_CMD_SWAP] = {
1871 .call = ip_set_swap,
1872 .attr_count = IPSET_ATTR_CMD_MAX,
1873 .policy = ip_set_setname2_policy,
1875 [IPSET_CMD_LIST] = {
1876 .call = ip_set_dump,
1877 .attr_count = IPSET_ATTR_CMD_MAX,
1878 .policy = ip_set_setname_policy,
1880 [IPSET_CMD_SAVE] = {
1881 .call = ip_set_dump,
1882 .attr_count = IPSET_ATTR_CMD_MAX,
1883 .policy = ip_set_setname_policy,
1886 .call = ip_set_uadd,
1887 .attr_count = IPSET_ATTR_CMD_MAX,
1888 .policy = ip_set_adt_policy,
1891 .call = ip_set_udel,
1892 .attr_count = IPSET_ATTR_CMD_MAX,
1893 .policy = ip_set_adt_policy,
1895 [IPSET_CMD_TEST] = {
1896 .call = ip_set_utest,
1897 .attr_count = IPSET_ATTR_CMD_MAX,
1898 .policy = ip_set_adt_policy,
1900 [IPSET_CMD_HEADER] = {
1901 .call = ip_set_header,
1902 .attr_count = IPSET_ATTR_CMD_MAX,
1903 .policy = ip_set_setname_policy,
1905 [IPSET_CMD_TYPE] = {
1906 .call = ip_set_type,
1907 .attr_count = IPSET_ATTR_CMD_MAX,
1908 .policy = ip_set_type_policy,
1910 [IPSET_CMD_PROTOCOL] = {
1911 .call = ip_set_protocol,
1912 .attr_count = IPSET_ATTR_CMD_MAX,
1913 .policy = ip_set_protocol_policy,
1917 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = {
1919 .subsys_id = NFNL_SUBSYS_IPSET,
1920 .cb_count = IPSET_MSG_MAX,
1921 .cb = ip_set_netlink_subsys_cb,
1924 /* Interface to iptables/ip6tables */
1927 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len)
1931 int copylen = *len, ret = 0;
1932 struct net *net = sock_net(sk);
1933 struct ip_set_net *inst = ip_set_pernet(net);
1935 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1937 if (optval != SO_IP_SET)
1939 if (*len < sizeof(unsigned int))
1942 data = vmalloc(*len);
1945 if (copy_from_user(data, user, *len) != 0) {
1951 if (*op < IP_SET_OP_VERSION) {
1952 /* Check the version at the beginning of operations */
1953 struct ip_set_req_version *req_version = data;
1955 if (*len < sizeof(struct ip_set_req_version)) {
1960 if (req_version->version != IPSET_PROTOCOL) {
1967 case IP_SET_OP_VERSION: {
1968 struct ip_set_req_version *req_version = data;
1970 if (*len != sizeof(struct ip_set_req_version)) {
1975 req_version->version = IPSET_PROTOCOL;
1976 ret = copy_to_user(user, req_version,
1977 sizeof(struct ip_set_req_version));
1980 case IP_SET_OP_GET_BYNAME: {
1981 struct ip_set_req_get_set *req_get = data;
1984 if (*len != sizeof(struct ip_set_req_get_set)) {
1988 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
1989 nfnl_lock(NFNL_SUBSYS_IPSET);
1990 find_set_and_id(inst, req_get->set.name, &id);
1991 req_get->set.index = id;
1992 nfnl_unlock(NFNL_SUBSYS_IPSET);
1995 case IP_SET_OP_GET_FNAME: {
1996 struct ip_set_req_get_set_family *req_get = data;
1999 if (*len != sizeof(struct ip_set_req_get_set_family)) {
2003 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
2004 nfnl_lock(NFNL_SUBSYS_IPSET);
2005 find_set_and_id(inst, req_get->set.name, &id);
2006 req_get->set.index = id;
2007 if (id != IPSET_INVALID_ID)
2008 req_get->family = ip_set(inst, id)->family;
2009 nfnl_unlock(NFNL_SUBSYS_IPSET);
2012 case IP_SET_OP_GET_BYINDEX: {
2013 struct ip_set_req_get_set *req_get = data;
2016 if (*len != sizeof(struct ip_set_req_get_set) ||
2017 req_get->set.index >= inst->ip_set_max) {
2021 nfnl_lock(NFNL_SUBSYS_IPSET);
2022 set = ip_set(inst, req_get->set.index);
2023 strncpy(req_get->set.name, set ? set->name : "",
2025 nfnl_unlock(NFNL_SUBSYS_IPSET);
2031 } /* end of switch(op) */
2034 ret = copy_to_user(user, data, copylen);
2043 static struct nf_sockopt_ops so_set __read_mostly = {
2045 .get_optmin = SO_IP_SET,
2046 .get_optmax = SO_IP_SET + 1,
2047 .get = ip_set_sockfn_get,
2048 .owner = THIS_MODULE,
2051 static int __net_init
2052 ip_set_net_init(struct net *net)
2054 struct ip_set_net *inst = ip_set_pernet(net);
2055 struct ip_set **list;
2057 inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX;
2058 if (inst->ip_set_max >= IPSET_INVALID_ID)
2059 inst->ip_set_max = IPSET_INVALID_ID - 1;
2061 list = kvcalloc(inst->ip_set_max, sizeof(struct ip_set *), GFP_KERNEL);
2064 inst->is_deleted = false;
2065 inst->is_destroyed = false;
2066 rcu_assign_pointer(inst->ip_set_list, list);
2070 static void __net_exit
2071 ip_set_net_exit(struct net *net)
2073 struct ip_set_net *inst = ip_set_pernet(net);
2075 struct ip_set *set = NULL;
2078 inst->is_deleted = true; /* flag for ip_set_nfnl_put */
2080 nfnl_lock(NFNL_SUBSYS_IPSET);
2081 for (i = 0; i < inst->ip_set_max; i++) {
2082 set = ip_set(inst, i);
2084 ip_set(inst, i) = NULL;
2085 ip_set_destroy_set(set);
2088 nfnl_unlock(NFNL_SUBSYS_IPSET);
2089 kvfree(rcu_dereference_protected(inst->ip_set_list, 1));
2092 static struct pernet_operations ip_set_net_ops = {
2093 .init = ip_set_net_init,
2094 .exit = ip_set_net_exit,
2095 .id = &ip_set_net_id,
2096 .size = sizeof(struct ip_set_net),
2102 int ret = register_pernet_subsys(&ip_set_net_ops);
2105 pr_err("ip_set: cannot register pernet_subsys.\n");
2109 ret = nfnetlink_subsys_register(&ip_set_netlink_subsys);
2111 pr_err("ip_set: cannot register with nfnetlink.\n");
2112 unregister_pernet_subsys(&ip_set_net_ops);
2116 ret = nf_register_sockopt(&so_set);
2118 pr_err("SO_SET registry failed: %d\n", ret);
2119 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2120 unregister_pernet_subsys(&ip_set_net_ops);
2130 nf_unregister_sockopt(&so_set);
2131 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2133 unregister_pernet_subsys(&ip_set_net_ops);
2134 pr_debug("these are the famous last words\n");
2137 module_init(ip_set_init);
2138 module_exit(ip_set_fini);
2140 MODULE_DESCRIPTION("ip_set: protocol " __stringify(IPSET_PROTOCOL));