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, 1)
61 #define ip_set(inst, id) \
62 ip_set_dereference((inst)->ip_set_list)[id]
64 /* The set types are implemented in modules and registered set types
65 * can be found in ip_set_type_list. Adding/deleting types is
66 * serialized by ip_set_type_mutex.
70 ip_set_type_lock(void)
72 mutex_lock(&ip_set_type_mutex);
76 ip_set_type_unlock(void)
78 mutex_unlock(&ip_set_type_mutex);
81 /* Register and deregister settype */
83 static struct ip_set_type *
84 find_set_type(const char *name, u8 family, u8 revision)
86 struct ip_set_type *type;
88 list_for_each_entry_rcu(type, &ip_set_type_list, list)
89 if (STRNCMP(type->name, name) &&
90 (type->family == family ||
91 type->family == NFPROTO_UNSPEC) &&
92 revision >= type->revision_min &&
93 revision <= type->revision_max)
98 /* Unlock, try to load a set type module and lock again */
100 load_settype(const char *name)
102 nfnl_unlock(NFNL_SUBSYS_IPSET);
103 pr_debug("try to load ip_set_%s\n", name);
104 if (request_module("ip_set_%s", name) < 0) {
105 pr_warn("Can't find ip_set type %s\n", name);
106 nfnl_lock(NFNL_SUBSYS_IPSET);
109 nfnl_lock(NFNL_SUBSYS_IPSET);
113 /* Find a set type and reference it */
114 #define find_set_type_get(name, family, revision, found) \
115 __find_set_type_get(name, family, revision, found, false)
118 __find_set_type_get(const char *name, u8 family, u8 revision,
119 struct ip_set_type **found, bool retry)
121 struct ip_set_type *type;
124 if (retry && !load_settype(name))
125 return -IPSET_ERR_FIND_TYPE;
128 *found = find_set_type(name, family, revision);
130 err = !try_module_get((*found)->me) ? -EFAULT : 0;
133 /* Make sure the type is already loaded
134 * but we don't support the revision
136 list_for_each_entry_rcu(type, &ip_set_type_list, list)
137 if (STRNCMP(type->name, name)) {
138 err = -IPSET_ERR_FIND_TYPE;
143 return retry ? -IPSET_ERR_FIND_TYPE :
144 __find_set_type_get(name, family, revision, found, true);
151 /* Find a given set type by name and family.
152 * If we succeeded, the supported minimal and maximum revisions are
155 #define find_set_type_minmax(name, family, min, max) \
156 __find_set_type_minmax(name, family, min, max, false)
159 __find_set_type_minmax(const char *name, u8 family, u8 *min, u8 *max,
162 struct ip_set_type *type;
165 if (retry && !load_settype(name))
166 return -IPSET_ERR_FIND_TYPE;
168 *min = 255; *max = 0;
170 list_for_each_entry_rcu(type, &ip_set_type_list, list)
171 if (STRNCMP(type->name, name) &&
172 (type->family == family ||
173 type->family == NFPROTO_UNSPEC)) {
175 if (type->revision_min < *min)
176 *min = type->revision_min;
177 if (type->revision_max > *max)
178 *max = type->revision_max;
184 return retry ? -IPSET_ERR_FIND_TYPE :
185 __find_set_type_minmax(name, family, min, max, true);
188 #define family_name(f) ((f) == NFPROTO_IPV4 ? "inet" : \
189 (f) == NFPROTO_IPV6 ? "inet6" : "any")
191 /* Register a set type structure. The type is identified by
192 * the unique triple of name, family and revision.
195 ip_set_type_register(struct ip_set_type *type)
199 if (type->protocol != IPSET_PROTOCOL) {
200 pr_warn("ip_set type %s, family %s, revision %u:%u uses wrong protocol version %u (want %u)\n",
201 type->name, family_name(type->family),
202 type->revision_min, type->revision_max,
203 type->protocol, IPSET_PROTOCOL);
208 if (find_set_type(type->name, type->family, type->revision_min)) {
210 pr_warn("ip_set type %s, family %s with revision min %u already registered!\n",
211 type->name, family_name(type->family),
213 ip_set_type_unlock();
216 list_add_rcu(&type->list, &ip_set_type_list);
217 pr_debug("type %s, family %s, revision %u:%u registered.\n",
218 type->name, family_name(type->family),
219 type->revision_min, type->revision_max);
220 ip_set_type_unlock();
224 EXPORT_SYMBOL_GPL(ip_set_type_register);
226 /* Unregister a set type. There's a small race with ip_set_create */
228 ip_set_type_unregister(struct ip_set_type *type)
231 if (!find_set_type(type->name, type->family, type->revision_min)) {
232 pr_warn("ip_set type %s, family %s with revision min %u not registered\n",
233 type->name, family_name(type->family),
235 ip_set_type_unlock();
238 list_del_rcu(&type->list);
239 pr_debug("type %s, family %s with revision min %u unregistered.\n",
240 type->name, family_name(type->family), type->revision_min);
241 ip_set_type_unlock();
245 EXPORT_SYMBOL_GPL(ip_set_type_unregister);
247 /* Utility functions */
249 ip_set_alloc(size_t size)
251 void *members = NULL;
253 if (size < KMALLOC_MAX_SIZE)
254 members = kzalloc(size, GFP_KERNEL | __GFP_NOWARN);
257 pr_debug("%p: allocated with kmalloc\n", members);
261 members = vzalloc(size);
264 pr_debug("%p: allocated with vmalloc\n", members);
268 EXPORT_SYMBOL_GPL(ip_set_alloc);
271 ip_set_free(void *members)
273 pr_debug("%p: free with %s\n", members,
274 is_vmalloc_addr(members) ? "vfree" : "kfree");
277 EXPORT_SYMBOL_GPL(ip_set_free);
280 flag_nested(const struct nlattr *nla)
282 return nla->nla_type & NLA_F_NESTED;
285 static const struct nla_policy ipaddr_policy[IPSET_ATTR_IPADDR_MAX + 1] = {
286 [IPSET_ATTR_IPADDR_IPV4] = { .type = NLA_U32 },
287 [IPSET_ATTR_IPADDR_IPV6] = { .type = NLA_BINARY,
288 .len = sizeof(struct in6_addr) },
292 ip_set_get_ipaddr4(struct nlattr *nla, __be32 *ipaddr)
294 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1];
296 if (unlikely(!flag_nested(nla)))
297 return -IPSET_ERR_PROTOCOL;
298 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla,
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 if (align < ip_set_extensions[id].align)
385 align = ip_set_extensions[id].align;
386 len = ALIGN(len, ip_set_extensions[id].align);
387 set->offset[id] = len;
388 set->extensions |= ip_set_extensions[id].type;
389 len += ip_set_extensions[id].len;
391 return ALIGN(len, align);
393 EXPORT_SYMBOL_GPL(ip_set_elem_len);
396 ip_set_get_extensions(struct ip_set *set, struct nlattr *tb[],
397 struct ip_set_ext *ext)
401 if (unlikely(!ip_set_optattr_netorder(tb, IPSET_ATTR_TIMEOUT) ||
402 !ip_set_optattr_netorder(tb, IPSET_ATTR_PACKETS) ||
403 !ip_set_optattr_netorder(tb, IPSET_ATTR_BYTES) ||
404 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBMARK) ||
405 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBPRIO) ||
406 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBQUEUE)))
407 return -IPSET_ERR_PROTOCOL;
409 if (tb[IPSET_ATTR_TIMEOUT]) {
410 if (!SET_WITH_TIMEOUT(set))
411 return -IPSET_ERR_TIMEOUT;
412 ext->timeout = ip_set_timeout_uget(tb[IPSET_ATTR_TIMEOUT]);
414 if (tb[IPSET_ATTR_BYTES] || tb[IPSET_ATTR_PACKETS]) {
415 if (!SET_WITH_COUNTER(set))
416 return -IPSET_ERR_COUNTER;
417 if (tb[IPSET_ATTR_BYTES])
418 ext->bytes = be64_to_cpu(nla_get_be64(
419 tb[IPSET_ATTR_BYTES]));
420 if (tb[IPSET_ATTR_PACKETS])
421 ext->packets = be64_to_cpu(nla_get_be64(
422 tb[IPSET_ATTR_PACKETS]));
424 if (tb[IPSET_ATTR_COMMENT]) {
425 if (!SET_WITH_COMMENT(set))
426 return -IPSET_ERR_COMMENT;
427 ext->comment = ip_set_comment_uget(tb[IPSET_ATTR_COMMENT]);
429 if (tb[IPSET_ATTR_SKBMARK]) {
430 if (!SET_WITH_SKBINFO(set))
431 return -IPSET_ERR_SKBINFO;
432 fullmark = be64_to_cpu(nla_get_be64(tb[IPSET_ATTR_SKBMARK]));
433 ext->skbinfo.skbmark = fullmark >> 32;
434 ext->skbinfo.skbmarkmask = fullmark & 0xffffffff;
436 if (tb[IPSET_ATTR_SKBPRIO]) {
437 if (!SET_WITH_SKBINFO(set))
438 return -IPSET_ERR_SKBINFO;
439 ext->skbinfo.skbprio =
440 be32_to_cpu(nla_get_be32(tb[IPSET_ATTR_SKBPRIO]));
442 if (tb[IPSET_ATTR_SKBQUEUE]) {
443 if (!SET_WITH_SKBINFO(set))
444 return -IPSET_ERR_SKBINFO;
445 ext->skbinfo.skbqueue =
446 be16_to_cpu(nla_get_be16(tb[IPSET_ATTR_SKBQUEUE]));
450 EXPORT_SYMBOL_GPL(ip_set_get_extensions);
453 ip_set_put_extensions(struct sk_buff *skb, const struct ip_set *set,
454 const void *e, bool active)
456 if (SET_WITH_TIMEOUT(set)) {
457 unsigned long *timeout = ext_timeout(e, set);
459 if (nla_put_net32(skb, IPSET_ATTR_TIMEOUT,
460 htonl(active ? ip_set_timeout_get(timeout)
464 if (SET_WITH_COUNTER(set) &&
465 ip_set_put_counter(skb, ext_counter(e, set)))
467 if (SET_WITH_COMMENT(set) &&
468 ip_set_put_comment(skb, ext_comment(e, set)))
470 if (SET_WITH_SKBINFO(set) &&
471 ip_set_put_skbinfo(skb, ext_skbinfo(e, set)))
475 EXPORT_SYMBOL_GPL(ip_set_put_extensions);
477 /* Creating/destroying/renaming/swapping affect the existence and
478 * the properties of a set. All of these can be executed from userspace
479 * only and serialized by the nfnl mutex indirectly from nfnetlink.
481 * Sets are identified by their index in ip_set_list and the index
482 * is used by the external references (set/SET netfilter modules).
484 * The set behind an index may change by swapping only, from userspace.
488 __ip_set_get(struct ip_set *set)
490 write_lock_bh(&ip_set_ref_lock);
492 write_unlock_bh(&ip_set_ref_lock);
496 __ip_set_put(struct ip_set *set)
498 write_lock_bh(&ip_set_ref_lock);
499 BUG_ON(set->ref == 0);
501 write_unlock_bh(&ip_set_ref_lock);
504 /* set->ref can be swapped out by ip_set_swap, netlink events (like dump) need
505 * a separate reference counter
508 __ip_set_put_netlink(struct ip_set *set)
510 write_lock_bh(&ip_set_ref_lock);
511 BUG_ON(set->ref_netlink == 0);
513 write_unlock_bh(&ip_set_ref_lock);
516 /* Add, del and test set entries from kernel.
518 * The set behind the index must exist and must be referenced
519 * so it can't be destroyed (or changed) under our foot.
522 static inline struct ip_set *
523 ip_set_rcu_get(struct net *net, ip_set_id_t index)
526 struct ip_set_net *inst = ip_set_pernet(net);
529 /* ip_set_list itself needs to be protected */
530 set = rcu_dereference(inst->ip_set_list)[index];
537 ip_set_test(ip_set_id_t index, const struct sk_buff *skb,
538 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
540 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
544 pr_debug("set %s, index %u\n", set->name, index);
546 if (opt->dim < set->type->dimension ||
547 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
551 ret = set->variant->kadt(set, skb, par, IPSET_TEST, opt);
552 rcu_read_unlock_bh();
554 if (ret == -EAGAIN) {
555 /* Type requests element to be completed */
556 pr_debug("element must be completed, ADD is triggered\n");
557 spin_lock_bh(&set->lock);
558 set->variant->kadt(set, skb, par, IPSET_ADD, opt);
559 spin_unlock_bh(&set->lock);
562 /* --return-nomatch: invert matched element */
563 if ((opt->cmdflags & IPSET_FLAG_RETURN_NOMATCH) &&
564 (set->type->features & IPSET_TYPE_NOMATCH) &&
565 (ret > 0 || ret == -ENOTEMPTY))
569 /* Convert error codes to nomatch */
570 return (ret < 0 ? 0 : ret);
572 EXPORT_SYMBOL_GPL(ip_set_test);
575 ip_set_add(ip_set_id_t index, const struct sk_buff *skb,
576 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
578 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
582 pr_debug("set %s, index %u\n", set->name, index);
584 if (opt->dim < set->type->dimension ||
585 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
586 return -IPSET_ERR_TYPE_MISMATCH;
588 spin_lock_bh(&set->lock);
589 ret = set->variant->kadt(set, skb, par, IPSET_ADD, opt);
590 spin_unlock_bh(&set->lock);
594 EXPORT_SYMBOL_GPL(ip_set_add);
597 ip_set_del(ip_set_id_t index, const struct sk_buff *skb,
598 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
600 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
604 pr_debug("set %s, index %u\n", set->name, index);
606 if (opt->dim < set->type->dimension ||
607 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
608 return -IPSET_ERR_TYPE_MISMATCH;
610 spin_lock_bh(&set->lock);
611 ret = set->variant->kadt(set, skb, par, IPSET_DEL, opt);
612 spin_unlock_bh(&set->lock);
616 EXPORT_SYMBOL_GPL(ip_set_del);
618 /* Find set by name, reference it once. The reference makes sure the
619 * thing pointed to, does not go away under our feet.
623 ip_set_get_byname(struct net *net, const char *name, struct ip_set **set)
625 ip_set_id_t i, index = IPSET_INVALID_ID;
627 struct ip_set_net *inst = ip_set_pernet(net);
630 for (i = 0; i < inst->ip_set_max; i++) {
631 s = rcu_dereference(inst->ip_set_list)[i];
632 if (s && STRNCMP(s->name, name)) {
643 EXPORT_SYMBOL_GPL(ip_set_get_byname);
645 /* If the given set pointer points to a valid set, decrement
646 * reference count by 1. The caller shall not assume the index
647 * to be valid, after calling this function.
652 __ip_set_put_byindex(struct ip_set_net *inst, ip_set_id_t index)
657 set = rcu_dereference(inst->ip_set_list)[index];
664 ip_set_put_byindex(struct net *net, ip_set_id_t index)
666 struct ip_set_net *inst = ip_set_pernet(net);
668 __ip_set_put_byindex(inst, index);
670 EXPORT_SYMBOL_GPL(ip_set_put_byindex);
672 /* Get the name of a set behind a set index.
673 * Set itself is protected by RCU, but its name isn't: to protect against
674 * renaming, grab ip_set_ref_lock as reader (see ip_set_rename()) and copy the
678 ip_set_name_byindex(struct net *net, ip_set_id_t index, char *name)
680 struct ip_set *set = ip_set_rcu_get(net, index);
684 read_lock_bh(&ip_set_ref_lock);
685 strncpy(name, set->name, IPSET_MAXNAMELEN);
686 read_unlock_bh(&ip_set_ref_lock);
688 EXPORT_SYMBOL_GPL(ip_set_name_byindex);
690 /* Routines to call by external subsystems, which do not
691 * call nfnl_lock for us.
694 /* Find set by index, reference it once. The reference makes sure the
695 * thing pointed to, does not go away under our feet.
697 * The nfnl mutex is used in the function.
700 ip_set_nfnl_get_byindex(struct net *net, ip_set_id_t index)
703 struct ip_set_net *inst = ip_set_pernet(net);
705 if (index >= inst->ip_set_max)
706 return IPSET_INVALID_ID;
708 nfnl_lock(NFNL_SUBSYS_IPSET);
709 set = ip_set(inst, index);
713 index = IPSET_INVALID_ID;
714 nfnl_unlock(NFNL_SUBSYS_IPSET);
718 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex);
720 /* If the given set pointer points to a valid set, decrement
721 * reference count by 1. The caller shall not assume the index
722 * to be valid, after calling this function.
724 * The nfnl mutex is used in the function.
727 ip_set_nfnl_put(struct net *net, ip_set_id_t index)
730 struct ip_set_net *inst = ip_set_pernet(net);
732 nfnl_lock(NFNL_SUBSYS_IPSET);
733 if (!inst->is_deleted) { /* already deleted from ip_set_net_exit() */
734 set = ip_set(inst, index);
738 nfnl_unlock(NFNL_SUBSYS_IPSET);
740 EXPORT_SYMBOL_GPL(ip_set_nfnl_put);
742 /* Communication protocol with userspace over netlink.
744 * The commands are serialized by the nfnl mutex.
748 protocol_failed(const struct nlattr * const tb[])
750 return !tb[IPSET_ATTR_PROTOCOL] ||
751 nla_get_u8(tb[IPSET_ATTR_PROTOCOL]) != IPSET_PROTOCOL;
755 flag_exist(const struct nlmsghdr *nlh)
757 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST;
760 static struct nlmsghdr *
761 start_msg(struct sk_buff *skb, u32 portid, u32 seq, unsigned int flags,
764 struct nlmsghdr *nlh;
765 struct nfgenmsg *nfmsg;
767 nlh = nlmsg_put(skb, portid, seq, nfnl_msg_type(NFNL_SUBSYS_IPSET, cmd),
768 sizeof(*nfmsg), flags);
772 nfmsg = nlmsg_data(nlh);
773 nfmsg->nfgen_family = NFPROTO_IPV4;
774 nfmsg->version = NFNETLINK_V0;
782 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = {
783 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
784 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
785 .len = IPSET_MAXNAMELEN - 1 },
786 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
787 .len = IPSET_MAXNAMELEN - 1},
788 [IPSET_ATTR_REVISION] = { .type = NLA_U8 },
789 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
790 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
793 static struct ip_set *
794 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id)
796 struct ip_set *set = NULL;
799 *id = IPSET_INVALID_ID;
800 for (i = 0; i < inst->ip_set_max; i++) {
801 set = ip_set(inst, i);
802 if (set && STRNCMP(set->name, name)) {
807 return (*id == IPSET_INVALID_ID ? NULL : set);
810 static inline struct ip_set *
811 find_set(struct ip_set_net *inst, const char *name)
815 return find_set_and_id(inst, name, &id);
819 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index,
825 *index = IPSET_INVALID_ID;
826 for (i = 0; i < inst->ip_set_max; i++) {
829 if (*index == IPSET_INVALID_ID)
831 } else if (STRNCMP(name, s->name)) {
837 if (*index == IPSET_INVALID_ID)
838 /* No free slot remained */
839 return -IPSET_ERR_MAX_SETS;
843 static int ip_set_none(struct net *net, struct sock *ctnl, struct sk_buff *skb,
844 const struct nlmsghdr *nlh,
845 const struct nlattr * const attr[],
846 struct netlink_ext_ack *extack)
851 static int ip_set_create(struct net *net, struct sock *ctnl,
852 struct sk_buff *skb, const struct nlmsghdr *nlh,
853 const struct nlattr * const attr[],
854 struct netlink_ext_ack *extack)
856 struct ip_set_net *inst = ip_set_pernet(net);
857 struct ip_set *set, *clash = NULL;
858 ip_set_id_t index = IPSET_INVALID_ID;
859 struct nlattr *tb[IPSET_ATTR_CREATE_MAX + 1] = {};
860 const char *name, *typename;
862 u32 flags = flag_exist(nlh);
865 if (unlikely(protocol_failed(attr) ||
866 !attr[IPSET_ATTR_SETNAME] ||
867 !attr[IPSET_ATTR_TYPENAME] ||
868 !attr[IPSET_ATTR_REVISION] ||
869 !attr[IPSET_ATTR_FAMILY] ||
870 (attr[IPSET_ATTR_DATA] &&
871 !flag_nested(attr[IPSET_ATTR_DATA]))))
872 return -IPSET_ERR_PROTOCOL;
874 name = nla_data(attr[IPSET_ATTR_SETNAME]);
875 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
876 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
877 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]);
878 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
879 name, typename, family_name(family), revision);
881 /* First, and without any locks, allocate and initialize
882 * a normal base set structure.
884 set = kzalloc(sizeof(*set), GFP_KERNEL);
887 spin_lock_init(&set->lock);
888 strlcpy(set->name, name, IPSET_MAXNAMELEN);
889 set->family = family;
890 set->revision = revision;
892 /* Next, check that we know the type, and take
893 * a reference on the type, to make sure it stays available
894 * while constructing our new set.
896 * After referencing the type, we try to create the type
897 * specific part of the set without holding any locks.
899 ret = find_set_type_get(typename, family, revision, &set->type);
903 /* Without holding any locks, create private part. */
904 if (attr[IPSET_ATTR_DATA] &&
905 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA],
906 set->type->create_policy, NULL)) {
907 ret = -IPSET_ERR_PROTOCOL;
911 ret = set->type->create(net, set, tb, flags);
915 /* BTW, ret==0 here. */
917 /* Here, we have a valid, constructed set and we are protected
918 * by the nfnl mutex. Find the first free index in ip_set_list
919 * and check clashing.
921 ret = find_free_id(inst, set->name, &index, &clash);
922 if (ret == -EEXIST) {
923 /* If this is the same set and requested, ignore error */
924 if ((flags & IPSET_FLAG_EXIST) &&
925 STRNCMP(set->type->name, clash->type->name) &&
926 set->type->family == clash->type->family &&
927 set->type->revision_min == clash->type->revision_min &&
928 set->type->revision_max == clash->type->revision_max &&
929 set->variant->same_set(set, clash))
932 } else if (ret == -IPSET_ERR_MAX_SETS) {
933 struct ip_set **list, **tmp;
934 ip_set_id_t i = inst->ip_set_max + IP_SET_INC;
936 if (i < inst->ip_set_max || i == IPSET_INVALID_ID)
940 list = kcalloc(i, sizeof(struct ip_set *), GFP_KERNEL);
943 /* nfnl mutex is held, both lists are valid */
944 tmp = ip_set_dereference(inst->ip_set_list);
945 memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max);
946 rcu_assign_pointer(inst->ip_set_list, list);
947 /* Make sure all current packets have passed through */
950 index = inst->ip_set_max;
951 inst->ip_set_max = i;
958 /* Finally! Add our shiny new set to the list, and be done. */
959 pr_debug("create: '%s' created with index %u!\n", set->name, index);
960 ip_set(inst, index) = set;
965 set->variant->destroy(set);
967 module_put(set->type->me);
975 static const struct nla_policy
976 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = {
977 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
978 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
979 .len = IPSET_MAXNAMELEN - 1 },
983 ip_set_destroy_set(struct ip_set *set)
985 pr_debug("set: %s\n", set->name);
987 /* Must call it without holding any lock */
988 set->variant->destroy(set);
989 module_put(set->type->me);
993 static int ip_set_destroy(struct net *net, struct sock *ctnl,
994 struct sk_buff *skb, const struct nlmsghdr *nlh,
995 const struct nlattr * const attr[],
996 struct netlink_ext_ack *extack)
998 struct ip_set_net *inst = ip_set_pernet(net);
1003 if (unlikely(protocol_failed(attr)))
1004 return -IPSET_ERR_PROTOCOL;
1006 /* Must wait for flush to be really finished in list:set */
1009 /* Commands are serialized and references are
1010 * protected by the ip_set_ref_lock.
1011 * External systems (i.e. xt_set) must call
1012 * ip_set_put|get_nfnl_* functions, that way we
1013 * can safely check references here.
1015 * list:set timer can only decrement the reference
1016 * counter, so if it's already zero, we can proceed
1017 * without holding the lock.
1019 read_lock_bh(&ip_set_ref_lock);
1020 if (!attr[IPSET_ATTR_SETNAME]) {
1021 for (i = 0; i < inst->ip_set_max; i++) {
1022 s = ip_set(inst, i);
1023 if (s && (s->ref || s->ref_netlink)) {
1024 ret = -IPSET_ERR_BUSY;
1028 inst->is_destroyed = true;
1029 read_unlock_bh(&ip_set_ref_lock);
1030 for (i = 0; i < inst->ip_set_max; i++) {
1031 s = ip_set(inst, i);
1033 ip_set(inst, i) = NULL;
1034 ip_set_destroy_set(s);
1037 /* Modified by ip_set_destroy() only, which is serialized */
1038 inst->is_destroyed = false;
1040 s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1045 } else if (s->ref || s->ref_netlink) {
1046 ret = -IPSET_ERR_BUSY;
1049 ip_set(inst, i) = NULL;
1050 read_unlock_bh(&ip_set_ref_lock);
1052 ip_set_destroy_set(s);
1056 read_unlock_bh(&ip_set_ref_lock);
1063 ip_set_flush_set(struct ip_set *set)
1065 pr_debug("set: %s\n", set->name);
1067 spin_lock_bh(&set->lock);
1068 set->variant->flush(set);
1069 spin_unlock_bh(&set->lock);
1072 static int ip_set_flush(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1073 const struct nlmsghdr *nlh,
1074 const struct nlattr * const attr[],
1075 struct netlink_ext_ack *extack)
1077 struct ip_set_net *inst = ip_set_pernet(net);
1081 if (unlikely(protocol_failed(attr)))
1082 return -IPSET_ERR_PROTOCOL;
1084 if (!attr[IPSET_ATTR_SETNAME]) {
1085 for (i = 0; i < inst->ip_set_max; i++) {
1086 s = ip_set(inst, i);
1088 ip_set_flush_set(s);
1091 s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1095 ip_set_flush_set(s);
1103 static const struct nla_policy
1104 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = {
1105 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1106 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1107 .len = IPSET_MAXNAMELEN - 1 },
1108 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING,
1109 .len = IPSET_MAXNAMELEN - 1 },
1112 static int ip_set_rename(struct net *net, struct sock *ctnl,
1113 struct sk_buff *skb, const struct nlmsghdr *nlh,
1114 const struct nlattr * const attr[],
1115 struct netlink_ext_ack *extack)
1117 struct ip_set_net *inst = ip_set_pernet(net);
1118 struct ip_set *set, *s;
1123 if (unlikely(protocol_failed(attr) ||
1124 !attr[IPSET_ATTR_SETNAME] ||
1125 !attr[IPSET_ATTR_SETNAME2]))
1126 return -IPSET_ERR_PROTOCOL;
1128 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1132 write_lock_bh(&ip_set_ref_lock);
1133 if (set->ref != 0 || set->ref_netlink != 0) {
1134 ret = -IPSET_ERR_REFERENCED;
1138 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]);
1139 for (i = 0; i < inst->ip_set_max; i++) {
1140 s = ip_set(inst, i);
1141 if (s && STRNCMP(s->name, name2)) {
1142 ret = -IPSET_ERR_EXIST_SETNAME2;
1146 strncpy(set->name, name2, IPSET_MAXNAMELEN);
1149 write_unlock_bh(&ip_set_ref_lock);
1153 /* Swap two sets so that name/index points to the other.
1154 * References and set names are also swapped.
1156 * The commands are serialized by the nfnl mutex and references are
1157 * protected by the ip_set_ref_lock. The kernel interfaces
1158 * do not hold the mutex but the pointer settings are atomic
1159 * so the ip_set_list always contains valid pointers to the sets.
1162 static int ip_set_swap(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1163 const struct nlmsghdr *nlh,
1164 const struct nlattr * const attr[],
1165 struct netlink_ext_ack *extack)
1167 struct ip_set_net *inst = ip_set_pernet(net);
1168 struct ip_set *from, *to;
1169 ip_set_id_t from_id, to_id;
1170 char from_name[IPSET_MAXNAMELEN];
1172 if (unlikely(protocol_failed(attr) ||
1173 !attr[IPSET_ATTR_SETNAME] ||
1174 !attr[IPSET_ATTR_SETNAME2]))
1175 return -IPSET_ERR_PROTOCOL;
1177 from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1182 to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]),
1185 return -IPSET_ERR_EXIST_SETNAME2;
1187 /* Features must not change.
1188 * Not an artifical restriction anymore, as we must prevent
1189 * possible loops created by swapping in setlist type of sets.
1191 if (!(from->type->features == to->type->features &&
1192 from->family == to->family))
1193 return -IPSET_ERR_TYPE_MISMATCH;
1195 write_lock_bh(&ip_set_ref_lock);
1197 if (from->ref_netlink || to->ref_netlink) {
1198 write_unlock_bh(&ip_set_ref_lock);
1202 strncpy(from_name, from->name, IPSET_MAXNAMELEN);
1203 strncpy(from->name, to->name, IPSET_MAXNAMELEN);
1204 strncpy(to->name, from_name, IPSET_MAXNAMELEN);
1206 swap(from->ref, to->ref);
1207 ip_set(inst, from_id) = to;
1208 ip_set(inst, to_id) = from;
1209 write_unlock_bh(&ip_set_ref_lock);
1214 /* List/save set data */
1221 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1222 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1225 ip_set_dump_done(struct netlink_callback *cb)
1227 if (cb->args[IPSET_CB_ARG0]) {
1228 struct ip_set_net *inst =
1229 (struct ip_set_net *)cb->args[IPSET_CB_NET];
1230 ip_set_id_t index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1231 struct ip_set *set = ip_set(inst, index);
1233 if (set->variant->uref)
1234 set->variant->uref(set, cb, false);
1235 pr_debug("release set %s\n", set->name);
1236 __ip_set_put_netlink(set);
1242 dump_attrs(struct nlmsghdr *nlh)
1244 const struct nlattr *attr;
1247 pr_debug("dump nlmsg\n");
1248 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) {
1249 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len);
1254 dump_init(struct netlink_callback *cb, struct ip_set_net *inst)
1256 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb);
1257 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1258 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1259 struct nlattr *attr = (void *)nlh + min_len;
1263 /* Second pass, so parser can't fail */
1264 nla_parse(cda, IPSET_ATTR_CMD_MAX, attr, nlh->nlmsg_len - min_len,
1265 ip_set_setname_policy, NULL);
1267 if (cda[IPSET_ATTR_SETNAME]) {
1270 set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]),
1275 dump_type = DUMP_ONE;
1276 cb->args[IPSET_CB_INDEX] = index;
1278 dump_type = DUMP_ALL;
1281 if (cda[IPSET_ATTR_FLAGS]) {
1282 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]);
1284 dump_type |= (f << 16);
1286 cb->args[IPSET_CB_NET] = (unsigned long)inst;
1287 cb->args[IPSET_CB_DUMP] = dump_type;
1293 ip_set_dump_start(struct sk_buff *skb, struct netlink_callback *cb)
1295 ip_set_id_t index = IPSET_INVALID_ID, max;
1296 struct ip_set *set = NULL;
1297 struct nlmsghdr *nlh = NULL;
1298 unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0;
1299 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1300 u32 dump_type, dump_flags;
1304 if (!cb->args[IPSET_CB_DUMP]) {
1305 ret = dump_init(cb, inst);
1307 nlh = nlmsg_hdr(cb->skb);
1308 /* We have to create and send the error message
1311 if (nlh->nlmsg_flags & NLM_F_ACK)
1312 netlink_ack(cb->skb, nlh, ret, NULL);
1317 if (cb->args[IPSET_CB_INDEX] >= inst->ip_set_max)
1320 dump_type = DUMP_TYPE(cb->args[IPSET_CB_DUMP]);
1321 dump_flags = DUMP_FLAGS(cb->args[IPSET_CB_DUMP]);
1322 max = dump_type == DUMP_ONE ? cb->args[IPSET_CB_INDEX] + 1
1325 pr_debug("dump type, flag: %u %u index: %ld\n",
1326 dump_type, dump_flags, cb->args[IPSET_CB_INDEX]);
1327 for (; cb->args[IPSET_CB_INDEX] < max; cb->args[IPSET_CB_INDEX]++) {
1328 index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1329 write_lock_bh(&ip_set_ref_lock);
1330 set = ip_set(inst, index);
1331 is_destroyed = inst->is_destroyed;
1332 if (!set || is_destroyed) {
1333 write_unlock_bh(&ip_set_ref_lock);
1334 if (dump_type == DUMP_ONE) {
1339 /* All sets are just being destroyed */
1345 /* When dumping all sets, we must dump "sorted"
1346 * so that lists (unions of sets) are dumped last.
1348 if (dump_type != DUMP_ONE &&
1349 ((dump_type == DUMP_ALL) ==
1350 !!(set->type->features & IPSET_DUMP_LAST))) {
1351 write_unlock_bh(&ip_set_ref_lock);
1354 pr_debug("List set: %s\n", set->name);
1355 if (!cb->args[IPSET_CB_ARG0]) {
1356 /* Start listing: make sure set won't be destroyed */
1357 pr_debug("reference set\n");
1360 write_unlock_bh(&ip_set_ref_lock);
1361 nlh = start_msg(skb, NETLINK_CB(cb->skb).portid,
1362 cb->nlh->nlmsg_seq, flags,
1366 goto release_refcount;
1368 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1369 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name))
1370 goto nla_put_failure;
1371 if (dump_flags & IPSET_FLAG_LIST_SETNAME)
1373 switch (cb->args[IPSET_CB_ARG0]) {
1375 /* Core header data */
1376 if (nla_put_string(skb, IPSET_ATTR_TYPENAME,
1378 nla_put_u8(skb, IPSET_ATTR_FAMILY,
1380 nla_put_u8(skb, IPSET_ATTR_REVISION,
1382 goto nla_put_failure;
1383 ret = set->variant->head(set, skb);
1385 goto release_refcount;
1386 if (dump_flags & IPSET_FLAG_LIST_HEADER)
1388 if (set->variant->uref)
1389 set->variant->uref(set, cb, true);
1390 /* Fall through and add elements */
1393 ret = set->variant->list(set, skb, cb);
1394 rcu_read_unlock_bh();
1395 if (!cb->args[IPSET_CB_ARG0])
1396 /* Set is done, proceed with next one */
1398 goto release_refcount;
1401 /* If we dump all sets, continue with dumping last ones */
1402 if (dump_type == DUMP_ALL) {
1403 dump_type = DUMP_LAST;
1404 cb->args[IPSET_CB_DUMP] = dump_type | (dump_flags << 16);
1405 cb->args[IPSET_CB_INDEX] = 0;
1406 if (set && set->variant->uref)
1407 set->variant->uref(set, cb, false);
1415 if (dump_type == DUMP_ONE)
1416 cb->args[IPSET_CB_INDEX] = IPSET_INVALID_ID;
1418 cb->args[IPSET_CB_INDEX]++;
1420 /* If there was an error or set is done, release set */
1421 if (ret || !cb->args[IPSET_CB_ARG0]) {
1422 set = ip_set(inst, index);
1423 if (set->variant->uref)
1424 set->variant->uref(set, cb, false);
1425 pr_debug("release set %s\n", set->name);
1426 __ip_set_put_netlink(set);
1427 cb->args[IPSET_CB_ARG0] = 0;
1431 nlmsg_end(skb, nlh);
1432 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len);
1436 return ret < 0 ? ret : skb->len;
1439 static int ip_set_dump(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1440 const struct nlmsghdr *nlh,
1441 const struct nlattr * const attr[],
1442 struct netlink_ext_ack *extack)
1444 if (unlikely(protocol_failed(attr)))
1445 return -IPSET_ERR_PROTOCOL;
1448 struct netlink_dump_control c = {
1449 .dump = ip_set_dump_start,
1450 .done = ip_set_dump_done,
1452 return netlink_dump_start(ctnl, skb, nlh, &c);
1456 /* Add, del and test */
1458 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = {
1459 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1460 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1461 .len = IPSET_MAXNAMELEN - 1 },
1462 [IPSET_ATTR_LINENO] = { .type = NLA_U32 },
1463 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
1464 [IPSET_ATTR_ADT] = { .type = NLA_NESTED },
1468 call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set,
1469 struct nlattr *tb[], enum ipset_adt adt,
1470 u32 flags, bool use_lineno)
1474 bool eexist = flags & IPSET_FLAG_EXIST, retried = false;
1477 spin_lock_bh(&set->lock);
1478 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried);
1479 spin_unlock_bh(&set->lock);
1481 } while (ret == -EAGAIN &&
1482 set->variant->resize &&
1483 (ret = set->variant->resize(set, retried)) == 0);
1485 if (!ret || (ret == -IPSET_ERR_EXIST && eexist))
1487 if (lineno && use_lineno) {
1488 /* Error in restore/batch mode: send back lineno */
1489 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb);
1490 struct sk_buff *skb2;
1491 struct nlmsgerr *errmsg;
1492 size_t payload = min(SIZE_MAX,
1493 sizeof(*errmsg) + nlmsg_len(nlh));
1494 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1495 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1496 struct nlattr *cmdattr;
1499 skb2 = nlmsg_new(payload, GFP_KERNEL);
1502 rep = __nlmsg_put(skb2, NETLINK_CB(skb).portid,
1503 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0);
1504 errmsg = nlmsg_data(rep);
1505 errmsg->error = ret;
1506 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len);
1507 cmdattr = (void *)&errmsg->msg + min_len;
1509 nla_parse(cda, IPSET_ATTR_CMD_MAX, cmdattr,
1510 nlh->nlmsg_len - min_len, ip_set_adt_policy, NULL);
1512 errline = nla_data(cda[IPSET_ATTR_LINENO]);
1516 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid,
1518 /* Signal netlink not to send its ACK/errmsg. */
1525 static int ip_set_uadd(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1526 const struct nlmsghdr *nlh,
1527 const struct nlattr * const attr[],
1528 struct netlink_ext_ack *extack)
1530 struct ip_set_net *inst = ip_set_pernet(net);
1532 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1533 const struct nlattr *nla;
1534 u32 flags = flag_exist(nlh);
1538 if (unlikely(protocol_failed(attr) ||
1539 !attr[IPSET_ATTR_SETNAME] ||
1540 !((attr[IPSET_ATTR_DATA] != NULL) ^
1541 (attr[IPSET_ATTR_ADT] != NULL)) ||
1542 (attr[IPSET_ATTR_DATA] &&
1543 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1544 (attr[IPSET_ATTR_ADT] &&
1545 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1546 !attr[IPSET_ATTR_LINENO]))))
1547 return -IPSET_ERR_PROTOCOL;
1549 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1553 use_lineno = !!attr[IPSET_ATTR_LINENO];
1554 if (attr[IPSET_ATTR_DATA]) {
1555 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1556 attr[IPSET_ATTR_DATA],
1557 set->type->adt_policy, NULL))
1558 return -IPSET_ERR_PROTOCOL;
1559 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD, flags,
1564 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1565 memset(tb, 0, sizeof(tb));
1566 if (nla_type(nla) != IPSET_ATTR_DATA ||
1567 !flag_nested(nla) ||
1568 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1569 set->type->adt_policy, NULL))
1570 return -IPSET_ERR_PROTOCOL;
1571 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD,
1580 static int ip_set_udel(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1581 const struct nlmsghdr *nlh,
1582 const struct nlattr * const attr[],
1583 struct netlink_ext_ack *extack)
1585 struct ip_set_net *inst = ip_set_pernet(net);
1587 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1588 const struct nlattr *nla;
1589 u32 flags = flag_exist(nlh);
1593 if (unlikely(protocol_failed(attr) ||
1594 !attr[IPSET_ATTR_SETNAME] ||
1595 !((attr[IPSET_ATTR_DATA] != NULL) ^
1596 (attr[IPSET_ATTR_ADT] != NULL)) ||
1597 (attr[IPSET_ATTR_DATA] &&
1598 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1599 (attr[IPSET_ATTR_ADT] &&
1600 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1601 !attr[IPSET_ATTR_LINENO]))))
1602 return -IPSET_ERR_PROTOCOL;
1604 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1608 use_lineno = !!attr[IPSET_ATTR_LINENO];
1609 if (attr[IPSET_ATTR_DATA]) {
1610 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1611 attr[IPSET_ATTR_DATA],
1612 set->type->adt_policy, NULL))
1613 return -IPSET_ERR_PROTOCOL;
1614 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL, flags,
1619 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1620 memset(tb, 0, sizeof(*tb));
1621 if (nla_type(nla) != IPSET_ATTR_DATA ||
1622 !flag_nested(nla) ||
1623 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1624 set->type->adt_policy, NULL))
1625 return -IPSET_ERR_PROTOCOL;
1626 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL,
1635 static int ip_set_utest(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1636 const struct nlmsghdr *nlh,
1637 const struct nlattr * const attr[],
1638 struct netlink_ext_ack *extack)
1640 struct ip_set_net *inst = ip_set_pernet(net);
1642 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1646 if (unlikely(protocol_failed(attr) ||
1647 !attr[IPSET_ATTR_SETNAME] ||
1648 !attr[IPSET_ATTR_DATA] ||
1649 !flag_nested(attr[IPSET_ATTR_DATA])))
1650 return -IPSET_ERR_PROTOCOL;
1652 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1656 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA],
1657 set->type->adt_policy, NULL))
1658 return -IPSET_ERR_PROTOCOL;
1661 ret = set->variant->uadt(set, tb, IPSET_TEST, &lineno, 0, 0);
1662 rcu_read_unlock_bh();
1663 /* Userspace can't trigger element to be re-added */
1667 return ret > 0 ? 0 : -IPSET_ERR_EXIST;
1670 /* Get headed data of a set */
1672 static int ip_set_header(struct net *net, struct sock *ctnl,
1673 struct sk_buff *skb, const struct nlmsghdr *nlh,
1674 const struct nlattr * const attr[],
1675 struct netlink_ext_ack *extack)
1677 struct ip_set_net *inst = ip_set_pernet(net);
1678 const struct ip_set *set;
1679 struct sk_buff *skb2;
1680 struct nlmsghdr *nlh2;
1683 if (unlikely(protocol_failed(attr) ||
1684 !attr[IPSET_ATTR_SETNAME]))
1685 return -IPSET_ERR_PROTOCOL;
1687 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1691 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1695 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1699 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1700 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) ||
1701 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) ||
1702 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
1703 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision))
1704 goto nla_put_failure;
1705 nlmsg_end(skb2, nlh2);
1707 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1714 nlmsg_cancel(skb2, nlh2);
1722 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = {
1723 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1724 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
1725 .len = IPSET_MAXNAMELEN - 1 },
1726 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
1729 static int ip_set_type(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1730 const struct nlmsghdr *nlh,
1731 const struct nlattr * const attr[],
1732 struct netlink_ext_ack *extack)
1734 struct sk_buff *skb2;
1735 struct nlmsghdr *nlh2;
1736 u8 family, min, max;
1737 const char *typename;
1740 if (unlikely(protocol_failed(attr) ||
1741 !attr[IPSET_ATTR_TYPENAME] ||
1742 !attr[IPSET_ATTR_FAMILY]))
1743 return -IPSET_ERR_PROTOCOL;
1745 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1746 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1747 ret = find_set_type_minmax(typename, family, &min, &max);
1751 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1755 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1759 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1760 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) ||
1761 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) ||
1762 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) ||
1763 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min))
1764 goto nla_put_failure;
1765 nlmsg_end(skb2, nlh2);
1767 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len);
1768 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1775 nlmsg_cancel(skb2, nlh2);
1781 /* Get protocol version */
1783 static const struct nla_policy
1784 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = {
1785 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1788 static int ip_set_protocol(struct net *net, struct sock *ctnl,
1789 struct sk_buff *skb, const struct nlmsghdr *nlh,
1790 const struct nlattr * const attr[],
1791 struct netlink_ext_ack *extack)
1793 struct sk_buff *skb2;
1794 struct nlmsghdr *nlh2;
1797 if (unlikely(!attr[IPSET_ATTR_PROTOCOL]))
1798 return -IPSET_ERR_PROTOCOL;
1800 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1804 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1805 IPSET_CMD_PROTOCOL);
1808 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL))
1809 goto nla_put_failure;
1810 nlmsg_end(skb2, nlh2);
1812 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1819 nlmsg_cancel(skb2, nlh2);
1825 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = {
1826 [IPSET_CMD_NONE] = {
1827 .call = ip_set_none,
1828 .attr_count = IPSET_ATTR_CMD_MAX,
1830 [IPSET_CMD_CREATE] = {
1831 .call = ip_set_create,
1832 .attr_count = IPSET_ATTR_CMD_MAX,
1833 .policy = ip_set_create_policy,
1835 [IPSET_CMD_DESTROY] = {
1836 .call = ip_set_destroy,
1837 .attr_count = IPSET_ATTR_CMD_MAX,
1838 .policy = ip_set_setname_policy,
1840 [IPSET_CMD_FLUSH] = {
1841 .call = ip_set_flush,
1842 .attr_count = IPSET_ATTR_CMD_MAX,
1843 .policy = ip_set_setname_policy,
1845 [IPSET_CMD_RENAME] = {
1846 .call = ip_set_rename,
1847 .attr_count = IPSET_ATTR_CMD_MAX,
1848 .policy = ip_set_setname2_policy,
1850 [IPSET_CMD_SWAP] = {
1851 .call = ip_set_swap,
1852 .attr_count = IPSET_ATTR_CMD_MAX,
1853 .policy = ip_set_setname2_policy,
1855 [IPSET_CMD_LIST] = {
1856 .call = ip_set_dump,
1857 .attr_count = IPSET_ATTR_CMD_MAX,
1858 .policy = ip_set_setname_policy,
1860 [IPSET_CMD_SAVE] = {
1861 .call = ip_set_dump,
1862 .attr_count = IPSET_ATTR_CMD_MAX,
1863 .policy = ip_set_setname_policy,
1866 .call = ip_set_uadd,
1867 .attr_count = IPSET_ATTR_CMD_MAX,
1868 .policy = ip_set_adt_policy,
1871 .call = ip_set_udel,
1872 .attr_count = IPSET_ATTR_CMD_MAX,
1873 .policy = ip_set_adt_policy,
1875 [IPSET_CMD_TEST] = {
1876 .call = ip_set_utest,
1877 .attr_count = IPSET_ATTR_CMD_MAX,
1878 .policy = ip_set_adt_policy,
1880 [IPSET_CMD_HEADER] = {
1881 .call = ip_set_header,
1882 .attr_count = IPSET_ATTR_CMD_MAX,
1883 .policy = ip_set_setname_policy,
1885 [IPSET_CMD_TYPE] = {
1886 .call = ip_set_type,
1887 .attr_count = IPSET_ATTR_CMD_MAX,
1888 .policy = ip_set_type_policy,
1890 [IPSET_CMD_PROTOCOL] = {
1891 .call = ip_set_protocol,
1892 .attr_count = IPSET_ATTR_CMD_MAX,
1893 .policy = ip_set_protocol_policy,
1897 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = {
1899 .subsys_id = NFNL_SUBSYS_IPSET,
1900 .cb_count = IPSET_MSG_MAX,
1901 .cb = ip_set_netlink_subsys_cb,
1904 /* Interface to iptables/ip6tables */
1907 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len)
1911 int copylen = *len, ret = 0;
1912 struct net *net = sock_net(sk);
1913 struct ip_set_net *inst = ip_set_pernet(net);
1915 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1917 if (optval != SO_IP_SET)
1919 if (*len < sizeof(unsigned int))
1922 data = vmalloc(*len);
1925 if (copy_from_user(data, user, *len) != 0) {
1931 if (*op < IP_SET_OP_VERSION) {
1932 /* Check the version at the beginning of operations */
1933 struct ip_set_req_version *req_version = data;
1935 if (*len < sizeof(struct ip_set_req_version)) {
1940 if (req_version->version != IPSET_PROTOCOL) {
1947 case IP_SET_OP_VERSION: {
1948 struct ip_set_req_version *req_version = data;
1950 if (*len != sizeof(struct ip_set_req_version)) {
1955 req_version->version = IPSET_PROTOCOL;
1956 if (copy_to_user(user, req_version,
1957 sizeof(struct ip_set_req_version)))
1961 case IP_SET_OP_GET_BYNAME: {
1962 struct ip_set_req_get_set *req_get = data;
1965 if (*len != sizeof(struct ip_set_req_get_set)) {
1969 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
1970 nfnl_lock(NFNL_SUBSYS_IPSET);
1971 find_set_and_id(inst, req_get->set.name, &id);
1972 req_get->set.index = id;
1973 nfnl_unlock(NFNL_SUBSYS_IPSET);
1976 case IP_SET_OP_GET_FNAME: {
1977 struct ip_set_req_get_set_family *req_get = data;
1980 if (*len != sizeof(struct ip_set_req_get_set_family)) {
1984 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
1985 nfnl_lock(NFNL_SUBSYS_IPSET);
1986 find_set_and_id(inst, req_get->set.name, &id);
1987 req_get->set.index = id;
1988 if (id != IPSET_INVALID_ID)
1989 req_get->family = ip_set(inst, id)->family;
1990 nfnl_unlock(NFNL_SUBSYS_IPSET);
1993 case IP_SET_OP_GET_BYINDEX: {
1994 struct ip_set_req_get_set *req_get = data;
1997 if (*len != sizeof(struct ip_set_req_get_set) ||
1998 req_get->set.index >= inst->ip_set_max) {
2002 nfnl_lock(NFNL_SUBSYS_IPSET);
2003 set = ip_set(inst, req_get->set.index);
2004 strncpy(req_get->set.name, set ? set->name : "",
2006 nfnl_unlock(NFNL_SUBSYS_IPSET);
2012 } /* end of switch(op) */
2015 if (copy_to_user(user, data, copylen))
2025 static struct nf_sockopt_ops so_set __read_mostly = {
2027 .get_optmin = SO_IP_SET,
2028 .get_optmax = SO_IP_SET + 1,
2029 .get = ip_set_sockfn_get,
2030 .owner = THIS_MODULE,
2033 static int __net_init
2034 ip_set_net_init(struct net *net)
2036 struct ip_set_net *inst = ip_set_pernet(net);
2037 struct ip_set **list;
2039 inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX;
2040 if (inst->ip_set_max >= IPSET_INVALID_ID)
2041 inst->ip_set_max = IPSET_INVALID_ID - 1;
2043 list = kcalloc(inst->ip_set_max, sizeof(struct ip_set *), GFP_KERNEL);
2046 inst->is_deleted = false;
2047 inst->is_destroyed = false;
2048 rcu_assign_pointer(inst->ip_set_list, list);
2052 static void __net_exit
2053 ip_set_net_exit(struct net *net)
2055 struct ip_set_net *inst = ip_set_pernet(net);
2057 struct ip_set *set = NULL;
2060 inst->is_deleted = true; /* flag for ip_set_nfnl_put */
2062 nfnl_lock(NFNL_SUBSYS_IPSET);
2063 for (i = 0; i < inst->ip_set_max; i++) {
2064 set = ip_set(inst, i);
2066 ip_set(inst, i) = NULL;
2067 ip_set_destroy_set(set);
2070 nfnl_unlock(NFNL_SUBSYS_IPSET);
2071 kfree(rcu_dereference_protected(inst->ip_set_list, 1));
2074 static struct pernet_operations ip_set_net_ops = {
2075 .init = ip_set_net_init,
2076 .exit = ip_set_net_exit,
2077 .id = &ip_set_net_id,
2078 .size = sizeof(struct ip_set_net)
2084 int ret = register_pernet_subsys(&ip_set_net_ops);
2087 pr_err("ip_set: cannot register pernet_subsys.\n");
2091 ret = nfnetlink_subsys_register(&ip_set_netlink_subsys);
2093 pr_err("ip_set: cannot register with nfnetlink.\n");
2094 unregister_pernet_subsys(&ip_set_net_ops);
2098 ret = nf_register_sockopt(&so_set);
2100 pr_err("SO_SET registry failed: %d\n", ret);
2101 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2102 unregister_pernet_subsys(&ip_set_net_ops);
2106 pr_info("ip_set: protocol %u\n", IPSET_PROTOCOL);
2113 nf_unregister_sockopt(&so_set);
2114 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2116 unregister_pernet_subsys(&ip_set_net_ops);
2117 pr_debug("these are the famous last words\n");
2120 module_init(ip_set_init);
2121 module_exit(ip_set_fini);