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 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, ipaddr_policy))
299 return -IPSET_ERR_PROTOCOL;
300 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV4)))
301 return -IPSET_ERR_PROTOCOL;
303 *ipaddr = nla_get_be32(tb[IPSET_ATTR_IPADDR_IPV4]);
306 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr4);
309 ip_set_get_ipaddr6(struct nlattr *nla, union nf_inet_addr *ipaddr)
311 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1];
313 if (unlikely(!flag_nested(nla)))
314 return -IPSET_ERR_PROTOCOL;
316 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla, ipaddr_policy))
317 return -IPSET_ERR_PROTOCOL;
318 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV6)))
319 return -IPSET_ERR_PROTOCOL;
321 memcpy(ipaddr, nla_data(tb[IPSET_ATTR_IPADDR_IPV6]),
322 sizeof(struct in6_addr));
325 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr6);
327 typedef void (*destroyer)(void *);
328 /* ipset data extension types, in size order */
330 const struct ip_set_ext_type ip_set_extensions[] = {
331 [IPSET_EXT_ID_COUNTER] = {
332 .type = IPSET_EXT_COUNTER,
333 .flag = IPSET_FLAG_WITH_COUNTERS,
334 .len = sizeof(struct ip_set_counter),
335 .align = __alignof__(struct ip_set_counter),
337 [IPSET_EXT_ID_TIMEOUT] = {
338 .type = IPSET_EXT_TIMEOUT,
339 .len = sizeof(unsigned long),
340 .align = __alignof__(unsigned long),
342 [IPSET_EXT_ID_SKBINFO] = {
343 .type = IPSET_EXT_SKBINFO,
344 .flag = IPSET_FLAG_WITH_SKBINFO,
345 .len = sizeof(struct ip_set_skbinfo),
346 .align = __alignof__(struct ip_set_skbinfo),
348 [IPSET_EXT_ID_COMMENT] = {
349 .type = IPSET_EXT_COMMENT | IPSET_EXT_DESTROY,
350 .flag = IPSET_FLAG_WITH_COMMENT,
351 .len = sizeof(struct ip_set_comment),
352 .align = __alignof__(struct ip_set_comment),
353 .destroy = (destroyer) ip_set_comment_free,
356 EXPORT_SYMBOL_GPL(ip_set_extensions);
359 add_extension(enum ip_set_ext_id id, u32 flags, struct nlattr *tb[])
361 return ip_set_extensions[id].flag ?
362 (flags & ip_set_extensions[id].flag) :
363 !!tb[IPSET_ATTR_TIMEOUT];
367 ip_set_elem_len(struct ip_set *set, struct nlattr *tb[], size_t len,
370 enum ip_set_ext_id id;
373 if (tb[IPSET_ATTR_CADT_FLAGS])
374 cadt_flags = ip_set_get_h32(tb[IPSET_ATTR_CADT_FLAGS]);
375 if (cadt_flags & IPSET_FLAG_WITH_FORCEADD)
376 set->flags |= IPSET_CREATE_FLAG_FORCEADD;
379 for (id = 0; id < IPSET_EXT_ID_MAX; id++) {
380 if (!add_extension(id, cadt_flags, tb))
382 if (align < ip_set_extensions[id].align)
383 align = ip_set_extensions[id].align;
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->skbmark = fullmark >> 32;
432 ext->skbmarkmask = fullmark & 0xffffffff;
434 if (tb[IPSET_ATTR_SKBPRIO]) {
435 if (!SET_WITH_SKBINFO(set))
436 return -IPSET_ERR_SKBINFO;
437 ext->skbprio = be32_to_cpu(nla_get_be32(
438 tb[IPSET_ATTR_SKBPRIO]));
440 if (tb[IPSET_ATTR_SKBQUEUE]) {
441 if (!SET_WITH_SKBINFO(set))
442 return -IPSET_ERR_SKBINFO;
443 ext->skbqueue = be16_to_cpu(nla_get_be16(
444 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);
475 /* Creating/destroying/renaming/swapping affect the existence and
476 * the properties of a set. All of these can be executed from userspace
477 * only and serialized by the nfnl mutex indirectly from nfnetlink.
479 * Sets are identified by their index in ip_set_list and the index
480 * is used by the external references (set/SET netfilter modules).
482 * The set behind an index may change by swapping only, from userspace.
486 __ip_set_get(struct ip_set *set)
488 write_lock_bh(&ip_set_ref_lock);
490 write_unlock_bh(&ip_set_ref_lock);
494 __ip_set_put(struct ip_set *set)
496 write_lock_bh(&ip_set_ref_lock);
497 BUG_ON(set->ref == 0);
499 write_unlock_bh(&ip_set_ref_lock);
502 /* set->ref can be swapped out by ip_set_swap, netlink events (like dump) need
503 * a separate reference counter
506 __ip_set_get_netlink(struct ip_set *set)
508 write_lock_bh(&ip_set_ref_lock);
510 write_unlock_bh(&ip_set_ref_lock);
514 __ip_set_put_netlink(struct ip_set *set)
516 write_lock_bh(&ip_set_ref_lock);
517 BUG_ON(set->ref_netlink == 0);
519 write_unlock_bh(&ip_set_ref_lock);
522 /* Add, del and test set entries from kernel.
524 * The set behind the index must exist and must be referenced
525 * so it can't be destroyed (or changed) under our foot.
528 static inline struct ip_set *
529 ip_set_rcu_get(struct net *net, ip_set_id_t index)
532 struct ip_set_net *inst = ip_set_pernet(net);
535 /* ip_set_list itself needs to be protected */
536 set = rcu_dereference(inst->ip_set_list)[index];
543 ip_set_test(ip_set_id_t index, const struct sk_buff *skb,
544 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
546 struct ip_set *set = ip_set_rcu_get(par->net, index);
550 pr_debug("set %s, index %u\n", set->name, index);
552 if (opt->dim < set->type->dimension ||
553 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
557 ret = set->variant->kadt(set, skb, par, IPSET_TEST, opt);
558 rcu_read_unlock_bh();
560 if (ret == -EAGAIN) {
561 /* Type requests element to be completed */
562 pr_debug("element must be completed, ADD is triggered\n");
563 spin_lock_bh(&set->lock);
564 set->variant->kadt(set, skb, par, IPSET_ADD, opt);
565 spin_unlock_bh(&set->lock);
568 /* --return-nomatch: invert matched element */
569 if ((opt->cmdflags & IPSET_FLAG_RETURN_NOMATCH) &&
570 (set->type->features & IPSET_TYPE_NOMATCH) &&
571 (ret > 0 || ret == -ENOTEMPTY))
575 /* Convert error codes to nomatch */
576 return (ret < 0 ? 0 : ret);
578 EXPORT_SYMBOL_GPL(ip_set_test);
581 ip_set_add(ip_set_id_t index, const struct sk_buff *skb,
582 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
584 struct ip_set *set = ip_set_rcu_get(par->net, index);
588 pr_debug("set %s, index %u\n", set->name, index);
590 if (opt->dim < set->type->dimension ||
591 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
592 return -IPSET_ERR_TYPE_MISMATCH;
594 spin_lock_bh(&set->lock);
595 ret = set->variant->kadt(set, skb, par, IPSET_ADD, opt);
596 spin_unlock_bh(&set->lock);
600 EXPORT_SYMBOL_GPL(ip_set_add);
603 ip_set_del(ip_set_id_t index, const struct sk_buff *skb,
604 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
606 struct ip_set *set = ip_set_rcu_get(par->net, index);
610 pr_debug("set %s, index %u\n", set->name, index);
612 if (opt->dim < set->type->dimension ||
613 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
614 return -IPSET_ERR_TYPE_MISMATCH;
616 spin_lock_bh(&set->lock);
617 ret = set->variant->kadt(set, skb, par, IPSET_DEL, opt);
618 spin_unlock_bh(&set->lock);
622 EXPORT_SYMBOL_GPL(ip_set_del);
624 /* Find set by name, reference it once. The reference makes sure the
625 * thing pointed to, does not go away under our feet.
629 ip_set_get_byname(struct net *net, const char *name, struct ip_set **set)
631 ip_set_id_t i, index = IPSET_INVALID_ID;
633 struct ip_set_net *inst = ip_set_pernet(net);
636 for (i = 0; i < inst->ip_set_max; i++) {
637 s = rcu_dereference(inst->ip_set_list)[i];
638 if (s && STRNCMP(s->name, name)) {
649 EXPORT_SYMBOL_GPL(ip_set_get_byname);
651 /* If the given set pointer points to a valid set, decrement
652 * reference count by 1. The caller shall not assume the index
653 * to be valid, after calling this function.
658 __ip_set_put_byindex(struct ip_set_net *inst, ip_set_id_t index)
663 set = rcu_dereference(inst->ip_set_list)[index];
670 ip_set_put_byindex(struct net *net, ip_set_id_t index)
672 struct ip_set_net *inst = ip_set_pernet(net);
674 __ip_set_put_byindex(inst, index);
676 EXPORT_SYMBOL_GPL(ip_set_put_byindex);
678 /* Get the name of a set behind a set index.
679 * We assume the set is referenced, so it does exist and
680 * can't be destroyed. The set cannot be renamed due to
681 * the referencing either.
685 ip_set_name_byindex(struct net *net, ip_set_id_t index)
687 const struct ip_set *set = ip_set_rcu_get(net, index);
690 BUG_ON(set->ref == 0);
692 /* Referenced, so it's safe */
695 EXPORT_SYMBOL_GPL(ip_set_name_byindex);
697 /* Routines to call by external subsystems, which do not
698 * call nfnl_lock for us.
701 /* Find set by index, reference it once. The reference makes sure the
702 * thing pointed to, does not go away under our feet.
704 * The nfnl mutex is used in the function.
707 ip_set_nfnl_get_byindex(struct net *net, ip_set_id_t index)
710 struct ip_set_net *inst = ip_set_pernet(net);
712 if (index >= inst->ip_set_max)
713 return IPSET_INVALID_ID;
715 nfnl_lock(NFNL_SUBSYS_IPSET);
716 set = ip_set(inst, index);
720 index = IPSET_INVALID_ID;
721 nfnl_unlock(NFNL_SUBSYS_IPSET);
725 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex);
727 /* If the given set pointer points to a valid set, decrement
728 * reference count by 1. The caller shall not assume the index
729 * to be valid, after calling this function.
731 * The nfnl mutex is used in the function.
734 ip_set_nfnl_put(struct net *net, ip_set_id_t index)
737 struct ip_set_net *inst = ip_set_pernet(net);
739 nfnl_lock(NFNL_SUBSYS_IPSET);
740 if (!inst->is_deleted) { /* already deleted from ip_set_net_exit() */
741 set = ip_set(inst, index);
745 nfnl_unlock(NFNL_SUBSYS_IPSET);
747 EXPORT_SYMBOL_GPL(ip_set_nfnl_put);
749 /* Communication protocol with userspace over netlink.
751 * The commands are serialized by the nfnl mutex.
755 protocol_failed(const struct nlattr * const tb[])
757 return !tb[IPSET_ATTR_PROTOCOL] ||
758 nla_get_u8(tb[IPSET_ATTR_PROTOCOL]) != IPSET_PROTOCOL;
762 flag_exist(const struct nlmsghdr *nlh)
764 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST;
767 static struct nlmsghdr *
768 start_msg(struct sk_buff *skb, u32 portid, u32 seq, unsigned int flags,
771 struct nlmsghdr *nlh;
772 struct nfgenmsg *nfmsg;
774 nlh = nlmsg_put(skb, portid, seq, cmd | (NFNL_SUBSYS_IPSET << 8),
775 sizeof(*nfmsg), flags);
779 nfmsg = nlmsg_data(nlh);
780 nfmsg->nfgen_family = NFPROTO_IPV4;
781 nfmsg->version = NFNETLINK_V0;
789 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = {
790 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
791 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
792 .len = IPSET_MAXNAMELEN - 1 },
793 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
794 .len = IPSET_MAXNAMELEN - 1},
795 [IPSET_ATTR_REVISION] = { .type = NLA_U8 },
796 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
797 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
800 static struct ip_set *
801 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id)
803 struct ip_set *set = NULL;
806 *id = IPSET_INVALID_ID;
807 for (i = 0; i < inst->ip_set_max; i++) {
808 set = ip_set(inst, i);
809 if (set && STRNCMP(set->name, name)) {
814 return (*id == IPSET_INVALID_ID ? NULL : set);
817 static inline struct ip_set *
818 find_set(struct ip_set_net *inst, const char *name)
822 return find_set_and_id(inst, name, &id);
826 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index,
832 *index = IPSET_INVALID_ID;
833 for (i = 0; i < inst->ip_set_max; i++) {
836 if (*index == IPSET_INVALID_ID)
838 } else if (STRNCMP(name, s->name)) {
844 if (*index == IPSET_INVALID_ID)
845 /* No free slot remained */
846 return -IPSET_ERR_MAX_SETS;
850 static int ip_set_none(struct net *net, struct sock *ctnl, struct sk_buff *skb,
851 const struct nlmsghdr *nlh,
852 const struct nlattr * const attr[])
857 static int ip_set_create(struct net *net, struct sock *ctnl,
858 struct sk_buff *skb, const struct nlmsghdr *nlh,
859 const struct nlattr * const attr[])
861 struct ip_set_net *inst = ip_set_pernet(net);
862 struct ip_set *set, *clash = NULL;
863 ip_set_id_t index = IPSET_INVALID_ID;
864 struct nlattr *tb[IPSET_ATTR_CREATE_MAX + 1] = {};
865 const char *name, *typename;
867 u32 flags = flag_exist(nlh);
870 if (unlikely(protocol_failed(attr) ||
871 !attr[IPSET_ATTR_SETNAME] ||
872 !attr[IPSET_ATTR_TYPENAME] ||
873 !attr[IPSET_ATTR_REVISION] ||
874 !attr[IPSET_ATTR_FAMILY] ||
875 (attr[IPSET_ATTR_DATA] &&
876 !flag_nested(attr[IPSET_ATTR_DATA]))))
877 return -IPSET_ERR_PROTOCOL;
879 name = nla_data(attr[IPSET_ATTR_SETNAME]);
880 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
881 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
882 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]);
883 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
884 name, typename, family_name(family), revision);
886 /* First, and without any locks, allocate and initialize
887 * a normal base set structure.
889 set = kzalloc(sizeof(*set), GFP_KERNEL);
892 spin_lock_init(&set->lock);
893 strlcpy(set->name, name, IPSET_MAXNAMELEN);
894 set->family = family;
895 set->revision = revision;
897 /* Next, check that we know the type, and take
898 * a reference on the type, to make sure it stays available
899 * while constructing our new set.
901 * After referencing the type, we try to create the type
902 * specific part of the set without holding any locks.
904 ret = find_set_type_get(typename, family, revision, &set->type);
908 /* Without holding any locks, create private part. */
909 if (attr[IPSET_ATTR_DATA] &&
910 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA],
911 set->type->create_policy)) {
912 ret = -IPSET_ERR_PROTOCOL;
916 ret = set->type->create(net, set, tb, flags);
920 /* BTW, ret==0 here. */
922 /* Here, we have a valid, constructed set and we are protected
923 * by the nfnl mutex. Find the first free index in ip_set_list
924 * and check clashing.
926 ret = find_free_id(inst, set->name, &index, &clash);
927 if (ret == -EEXIST) {
928 /* If this is the same set and requested, ignore error */
929 if ((flags & IPSET_FLAG_EXIST) &&
930 STRNCMP(set->type->name, clash->type->name) &&
931 set->type->family == clash->type->family &&
932 set->type->revision_min == clash->type->revision_min &&
933 set->type->revision_max == clash->type->revision_max &&
934 set->variant->same_set(set, clash))
937 } else if (ret == -IPSET_ERR_MAX_SETS) {
938 struct ip_set **list, **tmp;
939 ip_set_id_t i = inst->ip_set_max + IP_SET_INC;
941 if (i < inst->ip_set_max || i == IPSET_INVALID_ID)
945 list = kcalloc(i, sizeof(struct ip_set *), GFP_KERNEL);
948 /* nfnl mutex is held, both lists are valid */
949 tmp = ip_set_dereference(inst->ip_set_list);
950 memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max);
951 rcu_assign_pointer(inst->ip_set_list, list);
952 /* Make sure all current packets have passed through */
955 index = inst->ip_set_max;
956 inst->ip_set_max = i;
963 /* Finally! Add our shiny new set to the list, and be done. */
964 pr_debug("create: '%s' created with index %u!\n", set->name, index);
965 ip_set(inst, index) = set;
970 set->variant->destroy(set);
972 module_put(set->type->me);
980 static const struct nla_policy
981 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = {
982 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
983 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
984 .len = IPSET_MAXNAMELEN - 1 },
988 ip_set_destroy_set(struct ip_set *set)
990 pr_debug("set: %s\n", set->name);
992 /* Must call it without holding any lock */
993 set->variant->destroy(set);
994 module_put(set->type->me);
998 static int ip_set_destroy(struct net *net, struct sock *ctnl,
999 struct sk_buff *skb, const struct nlmsghdr *nlh,
1000 const struct nlattr * const attr[])
1002 struct ip_set_net *inst = ip_set_pernet(net);
1007 if (unlikely(protocol_failed(attr)))
1008 return -IPSET_ERR_PROTOCOL;
1010 /* Must wait for flush to be really finished in list:set */
1013 /* Commands are serialized and references are
1014 * protected by the ip_set_ref_lock.
1015 * External systems (i.e. xt_set) must call
1016 * ip_set_put|get_nfnl_* functions, that way we
1017 * can safely check references here.
1019 * list:set timer can only decrement the reference
1020 * counter, so if it's already zero, we can proceed
1021 * without holding the lock.
1023 read_lock_bh(&ip_set_ref_lock);
1024 if (!attr[IPSET_ATTR_SETNAME]) {
1025 for (i = 0; i < inst->ip_set_max; i++) {
1026 s = ip_set(inst, i);
1027 if (s && (s->ref || s->ref_netlink)) {
1028 ret = -IPSET_ERR_BUSY;
1032 inst->is_destroyed = true;
1033 read_unlock_bh(&ip_set_ref_lock);
1034 for (i = 0; i < inst->ip_set_max; i++) {
1035 s = ip_set(inst, i);
1037 ip_set(inst, i) = NULL;
1038 ip_set_destroy_set(s);
1041 /* Modified by ip_set_destroy() only, which is serialized */
1042 inst->is_destroyed = false;
1044 s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1049 } else if (s->ref || s->ref_netlink) {
1050 ret = -IPSET_ERR_BUSY;
1053 ip_set(inst, i) = NULL;
1054 read_unlock_bh(&ip_set_ref_lock);
1056 ip_set_destroy_set(s);
1060 read_unlock_bh(&ip_set_ref_lock);
1067 ip_set_flush_set(struct ip_set *set)
1069 pr_debug("set: %s\n", set->name);
1071 spin_lock_bh(&set->lock);
1072 set->variant->flush(set);
1073 spin_unlock_bh(&set->lock);
1076 static int ip_set_flush(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1077 const struct nlmsghdr *nlh,
1078 const struct nlattr * const attr[])
1080 struct ip_set_net *inst = ip_set_pernet(net);
1084 if (unlikely(protocol_failed(attr)))
1085 return -IPSET_ERR_PROTOCOL;
1087 if (!attr[IPSET_ATTR_SETNAME]) {
1088 for (i = 0; i < inst->ip_set_max; i++) {
1089 s = ip_set(inst, i);
1091 ip_set_flush_set(s);
1094 s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1098 ip_set_flush_set(s);
1106 static const struct nla_policy
1107 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = {
1108 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1109 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1110 .len = IPSET_MAXNAMELEN - 1 },
1111 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING,
1112 .len = IPSET_MAXNAMELEN - 1 },
1115 static int ip_set_rename(struct net *net, struct sock *ctnl,
1116 struct sk_buff *skb, const struct nlmsghdr *nlh,
1117 const struct nlattr * const attr[])
1119 struct ip_set_net *inst = ip_set_pernet(net);
1120 struct ip_set *set, *s;
1125 if (unlikely(protocol_failed(attr) ||
1126 !attr[IPSET_ATTR_SETNAME] ||
1127 !attr[IPSET_ATTR_SETNAME2]))
1128 return -IPSET_ERR_PROTOCOL;
1130 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1134 read_lock_bh(&ip_set_ref_lock);
1135 if (set->ref != 0) {
1136 ret = -IPSET_ERR_REFERENCED;
1140 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]);
1141 for (i = 0; i < inst->ip_set_max; i++) {
1142 s = ip_set(inst, i);
1143 if (s && STRNCMP(s->name, name2)) {
1144 ret = -IPSET_ERR_EXIST_SETNAME2;
1148 strncpy(set->name, name2, IPSET_MAXNAMELEN);
1151 read_unlock_bh(&ip_set_ref_lock);
1155 /* Swap two sets so that name/index points to the other.
1156 * References and set names are also swapped.
1158 * The commands are serialized by the nfnl mutex and references are
1159 * protected by the ip_set_ref_lock. The kernel interfaces
1160 * do not hold the mutex but the pointer settings are atomic
1161 * so the ip_set_list always contains valid pointers to the sets.
1164 static int ip_set_swap(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1165 const struct nlmsghdr *nlh,
1166 const struct nlattr * const attr[])
1168 struct ip_set_net *inst = ip_set_pernet(net);
1169 struct ip_set *from, *to;
1170 ip_set_id_t from_id, to_id;
1171 char from_name[IPSET_MAXNAMELEN];
1173 if (unlikely(protocol_failed(attr) ||
1174 !attr[IPSET_ATTR_SETNAME] ||
1175 !attr[IPSET_ATTR_SETNAME2]))
1176 return -IPSET_ERR_PROTOCOL;
1178 from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1183 to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]),
1186 return -IPSET_ERR_EXIST_SETNAME2;
1188 /* Features must not change.
1189 * Not an artifical restriction anymore, as we must prevent
1190 * possible loops created by swapping in setlist type of sets.
1192 if (!(from->type->features == to->type->features &&
1193 from->family == to->family))
1194 return -IPSET_ERR_TYPE_MISMATCH;
1196 if (from->ref_netlink || to->ref_netlink)
1199 strncpy(from_name, from->name, IPSET_MAXNAMELEN);
1200 strncpy(from->name, to->name, IPSET_MAXNAMELEN);
1201 strncpy(to->name, from_name, IPSET_MAXNAMELEN);
1203 write_lock_bh(&ip_set_ref_lock);
1204 swap(from->ref, to->ref);
1205 ip_set(inst, from_id) = to;
1206 ip_set(inst, to_id) = from;
1207 write_unlock_bh(&ip_set_ref_lock);
1212 /* List/save set data */
1219 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1220 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1223 ip_set_dump_done(struct netlink_callback *cb)
1225 if (cb->args[IPSET_CB_ARG0]) {
1226 struct ip_set_net *inst =
1227 (struct ip_set_net *)cb->args[IPSET_CB_NET];
1228 ip_set_id_t index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1229 struct ip_set *set = ip_set(inst, index);
1231 if (set->variant->uref)
1232 set->variant->uref(set, cb, false);
1233 pr_debug("release set %s\n", set->name);
1234 __ip_set_put_netlink(set);
1240 dump_attrs(struct nlmsghdr *nlh)
1242 const struct nlattr *attr;
1245 pr_debug("dump nlmsg\n");
1246 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) {
1247 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len);
1252 dump_init(struct netlink_callback *cb, struct ip_set_net *inst)
1254 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb);
1255 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1256 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1257 struct nlattr *attr = (void *)nlh + min_len;
1261 /* Second pass, so parser can't fail */
1262 nla_parse(cda, IPSET_ATTR_CMD_MAX,
1263 attr, nlh->nlmsg_len - min_len, ip_set_setname_policy);
1265 if (cda[IPSET_ATTR_SETNAME]) {
1268 set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]),
1273 dump_type = DUMP_ONE;
1274 cb->args[IPSET_CB_INDEX] = index;
1276 dump_type = DUMP_ALL;
1279 if (cda[IPSET_ATTR_FLAGS]) {
1280 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]);
1282 dump_type |= (f << 16);
1284 cb->args[IPSET_CB_NET] = (unsigned long)inst;
1285 cb->args[IPSET_CB_DUMP] = dump_type;
1291 ip_set_dump_start(struct sk_buff *skb, struct netlink_callback *cb)
1293 ip_set_id_t index = IPSET_INVALID_ID, max;
1294 struct ip_set *set = NULL;
1295 struct nlmsghdr *nlh = NULL;
1296 unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0;
1297 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1298 u32 dump_type, dump_flags;
1302 if (!cb->args[IPSET_CB_DUMP]) {
1303 ret = dump_init(cb, inst);
1305 nlh = nlmsg_hdr(cb->skb);
1306 /* We have to create and send the error message
1309 if (nlh->nlmsg_flags & NLM_F_ACK)
1310 netlink_ack(cb->skb, nlh, ret);
1315 if (cb->args[IPSET_CB_INDEX] >= inst->ip_set_max)
1318 dump_type = DUMP_TYPE(cb->args[IPSET_CB_DUMP]);
1319 dump_flags = DUMP_FLAGS(cb->args[IPSET_CB_DUMP]);
1320 max = dump_type == DUMP_ONE ? cb->args[IPSET_CB_INDEX] + 1
1323 pr_debug("dump type, flag: %u %u index: %ld\n",
1324 dump_type, dump_flags, cb->args[IPSET_CB_INDEX]);
1325 for (; cb->args[IPSET_CB_INDEX] < max; cb->args[IPSET_CB_INDEX]++) {
1326 index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1327 write_lock_bh(&ip_set_ref_lock);
1328 set = ip_set(inst, index);
1329 is_destroyed = inst->is_destroyed;
1330 if (!set || is_destroyed) {
1331 write_unlock_bh(&ip_set_ref_lock);
1332 if (dump_type == DUMP_ONE) {
1337 /* All sets are just being destroyed */
1343 /* When dumping all sets, we must dump "sorted"
1344 * so that lists (unions of sets) are dumped last.
1346 if (dump_type != DUMP_ONE &&
1347 ((dump_type == DUMP_ALL) ==
1348 !!(set->type->features & IPSET_DUMP_LAST))) {
1349 write_unlock_bh(&ip_set_ref_lock);
1352 pr_debug("List set: %s\n", set->name);
1353 if (!cb->args[IPSET_CB_ARG0]) {
1354 /* Start listing: make sure set won't be destroyed */
1355 pr_debug("reference set\n");
1358 write_unlock_bh(&ip_set_ref_lock);
1359 nlh = start_msg(skb, NETLINK_CB(cb->skb).portid,
1360 cb->nlh->nlmsg_seq, flags,
1364 goto release_refcount;
1366 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1367 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name))
1368 goto nla_put_failure;
1369 if (dump_flags & IPSET_FLAG_LIST_SETNAME)
1371 switch (cb->args[IPSET_CB_ARG0]) {
1373 /* Core header data */
1374 if (nla_put_string(skb, IPSET_ATTR_TYPENAME,
1376 nla_put_u8(skb, IPSET_ATTR_FAMILY,
1378 nla_put_u8(skb, IPSET_ATTR_REVISION,
1380 goto nla_put_failure;
1381 ret = set->variant->head(set, skb);
1383 goto release_refcount;
1384 if (dump_flags & IPSET_FLAG_LIST_HEADER)
1386 if (set->variant->uref)
1387 set->variant->uref(set, cb, true);
1388 /* Fall through and add elements */
1391 ret = set->variant->list(set, skb, cb);
1392 rcu_read_unlock_bh();
1393 if (!cb->args[IPSET_CB_ARG0])
1394 /* Set is done, proceed with next one */
1396 goto release_refcount;
1399 /* If we dump all sets, continue with dumping last ones */
1400 if (dump_type == DUMP_ALL) {
1401 dump_type = DUMP_LAST;
1402 cb->args[IPSET_CB_DUMP] = dump_type | (dump_flags << 16);
1403 cb->args[IPSET_CB_INDEX] = 0;
1404 if (set && set->variant->uref)
1405 set->variant->uref(set, cb, false);
1413 if (dump_type == DUMP_ONE)
1414 cb->args[IPSET_CB_INDEX] = IPSET_INVALID_ID;
1416 cb->args[IPSET_CB_INDEX]++;
1418 /* If there was an error or set is done, release set */
1419 if (ret || !cb->args[IPSET_CB_ARG0]) {
1420 set = ip_set(inst, index);
1421 if (set->variant->uref)
1422 set->variant->uref(set, cb, false);
1423 pr_debug("release set %s\n", set->name);
1424 __ip_set_put_netlink(set);
1425 cb->args[IPSET_CB_ARG0] = 0;
1429 nlmsg_end(skb, nlh);
1430 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len);
1434 return ret < 0 ? ret : skb->len;
1437 static int ip_set_dump(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1438 const struct nlmsghdr *nlh,
1439 const struct nlattr * const attr[])
1441 if (unlikely(protocol_failed(attr)))
1442 return -IPSET_ERR_PROTOCOL;
1445 struct netlink_dump_control c = {
1446 .dump = ip_set_dump_start,
1447 .done = ip_set_dump_done,
1449 return netlink_dump_start(ctnl, skb, nlh, &c);
1453 /* Add, del and test */
1455 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = {
1456 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1457 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1458 .len = IPSET_MAXNAMELEN - 1 },
1459 [IPSET_ATTR_LINENO] = { .type = NLA_U32 },
1460 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
1461 [IPSET_ATTR_ADT] = { .type = NLA_NESTED },
1465 call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set,
1466 struct nlattr *tb[], enum ipset_adt adt,
1467 u32 flags, bool use_lineno)
1471 bool eexist = flags & IPSET_FLAG_EXIST, retried = false;
1474 spin_lock_bh(&set->lock);
1475 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried);
1476 spin_unlock_bh(&set->lock);
1478 } while (ret == -EAGAIN &&
1479 set->variant->resize &&
1480 (ret = set->variant->resize(set, retried)) == 0);
1482 if (!ret || (ret == -IPSET_ERR_EXIST && eexist))
1484 if (lineno && use_lineno) {
1485 /* Error in restore/batch mode: send back lineno */
1486 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb);
1487 struct sk_buff *skb2;
1488 struct nlmsgerr *errmsg;
1489 size_t payload = min(SIZE_MAX,
1490 sizeof(*errmsg) + nlmsg_len(nlh));
1491 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1492 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1493 struct nlattr *cmdattr;
1496 skb2 = nlmsg_new(payload, GFP_KERNEL);
1499 rep = __nlmsg_put(skb2, NETLINK_CB(skb).portid,
1500 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0);
1501 errmsg = nlmsg_data(rep);
1502 errmsg->error = ret;
1503 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len);
1504 cmdattr = (void *)&errmsg->msg + min_len;
1506 nla_parse(cda, IPSET_ATTR_CMD_MAX,
1507 cmdattr, nlh->nlmsg_len - min_len,
1510 errline = nla_data(cda[IPSET_ATTR_LINENO]);
1514 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid,
1516 /* Signal netlink not to send its ACK/errmsg. */
1523 static int ip_set_uadd(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1524 const struct nlmsghdr *nlh,
1525 const struct nlattr * const attr[])
1527 struct ip_set_net *inst = ip_set_pernet(net);
1529 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1530 const struct nlattr *nla;
1531 u32 flags = flag_exist(nlh);
1535 if (unlikely(protocol_failed(attr) ||
1536 !attr[IPSET_ATTR_SETNAME] ||
1537 !((attr[IPSET_ATTR_DATA] != NULL) ^
1538 (attr[IPSET_ATTR_ADT] != NULL)) ||
1539 (attr[IPSET_ATTR_DATA] &&
1540 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1541 (attr[IPSET_ATTR_ADT] &&
1542 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1543 !attr[IPSET_ATTR_LINENO]))))
1544 return -IPSET_ERR_PROTOCOL;
1546 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1550 use_lineno = !!attr[IPSET_ATTR_LINENO];
1551 if (attr[IPSET_ATTR_DATA]) {
1552 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1553 attr[IPSET_ATTR_DATA],
1554 set->type->adt_policy))
1555 return -IPSET_ERR_PROTOCOL;
1556 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD, flags,
1561 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1562 memset(tb, 0, sizeof(tb));
1563 if (nla_type(nla) != IPSET_ATTR_DATA ||
1564 !flag_nested(nla) ||
1565 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1566 set->type->adt_policy))
1567 return -IPSET_ERR_PROTOCOL;
1568 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD,
1577 static int ip_set_udel(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1578 const struct nlmsghdr *nlh,
1579 const struct nlattr * const attr[])
1581 struct ip_set_net *inst = ip_set_pernet(net);
1583 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1584 const struct nlattr *nla;
1585 u32 flags = flag_exist(nlh);
1589 if (unlikely(protocol_failed(attr) ||
1590 !attr[IPSET_ATTR_SETNAME] ||
1591 !((attr[IPSET_ATTR_DATA] != NULL) ^
1592 (attr[IPSET_ATTR_ADT] != NULL)) ||
1593 (attr[IPSET_ATTR_DATA] &&
1594 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1595 (attr[IPSET_ATTR_ADT] &&
1596 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1597 !attr[IPSET_ATTR_LINENO]))))
1598 return -IPSET_ERR_PROTOCOL;
1600 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1604 use_lineno = !!attr[IPSET_ATTR_LINENO];
1605 if (attr[IPSET_ATTR_DATA]) {
1606 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1607 attr[IPSET_ATTR_DATA],
1608 set->type->adt_policy))
1609 return -IPSET_ERR_PROTOCOL;
1610 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL, flags,
1615 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1616 memset(tb, 0, sizeof(*tb));
1617 if (nla_type(nla) != IPSET_ATTR_DATA ||
1618 !flag_nested(nla) ||
1619 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1620 set->type->adt_policy))
1621 return -IPSET_ERR_PROTOCOL;
1622 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL,
1631 static int ip_set_utest(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1632 const struct nlmsghdr *nlh,
1633 const struct nlattr * const attr[])
1635 struct ip_set_net *inst = ip_set_pernet(net);
1637 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1641 if (unlikely(protocol_failed(attr) ||
1642 !attr[IPSET_ATTR_SETNAME] ||
1643 !attr[IPSET_ATTR_DATA] ||
1644 !flag_nested(attr[IPSET_ATTR_DATA])))
1645 return -IPSET_ERR_PROTOCOL;
1647 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1651 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA],
1652 set->type->adt_policy))
1653 return -IPSET_ERR_PROTOCOL;
1656 ret = set->variant->uadt(set, tb, IPSET_TEST, &lineno, 0, 0);
1657 rcu_read_unlock_bh();
1658 /* Userspace can't trigger element to be re-added */
1662 return ret > 0 ? 0 : -IPSET_ERR_EXIST;
1665 /* Get headed data of a set */
1667 static int ip_set_header(struct net *net, struct sock *ctnl,
1668 struct sk_buff *skb, const struct nlmsghdr *nlh,
1669 const struct nlattr * const attr[])
1671 struct ip_set_net *inst = ip_set_pernet(net);
1672 const struct ip_set *set;
1673 struct sk_buff *skb2;
1674 struct nlmsghdr *nlh2;
1677 if (unlikely(protocol_failed(attr) ||
1678 !attr[IPSET_ATTR_SETNAME]))
1679 return -IPSET_ERR_PROTOCOL;
1681 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1685 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1689 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1693 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1694 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) ||
1695 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) ||
1696 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
1697 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision))
1698 goto nla_put_failure;
1699 nlmsg_end(skb2, nlh2);
1701 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1708 nlmsg_cancel(skb2, nlh2);
1716 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = {
1717 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1718 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
1719 .len = IPSET_MAXNAMELEN - 1 },
1720 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
1723 static int ip_set_type(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1724 const struct nlmsghdr *nlh,
1725 const struct nlattr * const attr[])
1727 struct sk_buff *skb2;
1728 struct nlmsghdr *nlh2;
1729 u8 family, min, max;
1730 const char *typename;
1733 if (unlikely(protocol_failed(attr) ||
1734 !attr[IPSET_ATTR_TYPENAME] ||
1735 !attr[IPSET_ATTR_FAMILY]))
1736 return -IPSET_ERR_PROTOCOL;
1738 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1739 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1740 ret = find_set_type_minmax(typename, family, &min, &max);
1744 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1748 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1752 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1753 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) ||
1754 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) ||
1755 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) ||
1756 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min))
1757 goto nla_put_failure;
1758 nlmsg_end(skb2, nlh2);
1760 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len);
1761 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1768 nlmsg_cancel(skb2, nlh2);
1774 /* Get protocol version */
1776 static const struct nla_policy
1777 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = {
1778 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1781 static int ip_set_protocol(struct net *net, struct sock *ctnl,
1782 struct sk_buff *skb, const struct nlmsghdr *nlh,
1783 const struct nlattr * const attr[])
1785 struct sk_buff *skb2;
1786 struct nlmsghdr *nlh2;
1789 if (unlikely(!attr[IPSET_ATTR_PROTOCOL]))
1790 return -IPSET_ERR_PROTOCOL;
1792 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1796 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1797 IPSET_CMD_PROTOCOL);
1800 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL))
1801 goto nla_put_failure;
1802 nlmsg_end(skb2, nlh2);
1804 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1811 nlmsg_cancel(skb2, nlh2);
1817 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = {
1818 [IPSET_CMD_NONE] = {
1819 .call = ip_set_none,
1820 .attr_count = IPSET_ATTR_CMD_MAX,
1822 [IPSET_CMD_CREATE] = {
1823 .call = ip_set_create,
1824 .attr_count = IPSET_ATTR_CMD_MAX,
1825 .policy = ip_set_create_policy,
1827 [IPSET_CMD_DESTROY] = {
1828 .call = ip_set_destroy,
1829 .attr_count = IPSET_ATTR_CMD_MAX,
1830 .policy = ip_set_setname_policy,
1832 [IPSET_CMD_FLUSH] = {
1833 .call = ip_set_flush,
1834 .attr_count = IPSET_ATTR_CMD_MAX,
1835 .policy = ip_set_setname_policy,
1837 [IPSET_CMD_RENAME] = {
1838 .call = ip_set_rename,
1839 .attr_count = IPSET_ATTR_CMD_MAX,
1840 .policy = ip_set_setname2_policy,
1842 [IPSET_CMD_SWAP] = {
1843 .call = ip_set_swap,
1844 .attr_count = IPSET_ATTR_CMD_MAX,
1845 .policy = ip_set_setname2_policy,
1847 [IPSET_CMD_LIST] = {
1848 .call = ip_set_dump,
1849 .attr_count = IPSET_ATTR_CMD_MAX,
1850 .policy = ip_set_setname_policy,
1852 [IPSET_CMD_SAVE] = {
1853 .call = ip_set_dump,
1854 .attr_count = IPSET_ATTR_CMD_MAX,
1855 .policy = ip_set_setname_policy,
1858 .call = ip_set_uadd,
1859 .attr_count = IPSET_ATTR_CMD_MAX,
1860 .policy = ip_set_adt_policy,
1863 .call = ip_set_udel,
1864 .attr_count = IPSET_ATTR_CMD_MAX,
1865 .policy = ip_set_adt_policy,
1867 [IPSET_CMD_TEST] = {
1868 .call = ip_set_utest,
1869 .attr_count = IPSET_ATTR_CMD_MAX,
1870 .policy = ip_set_adt_policy,
1872 [IPSET_CMD_HEADER] = {
1873 .call = ip_set_header,
1874 .attr_count = IPSET_ATTR_CMD_MAX,
1875 .policy = ip_set_setname_policy,
1877 [IPSET_CMD_TYPE] = {
1878 .call = ip_set_type,
1879 .attr_count = IPSET_ATTR_CMD_MAX,
1880 .policy = ip_set_type_policy,
1882 [IPSET_CMD_PROTOCOL] = {
1883 .call = ip_set_protocol,
1884 .attr_count = IPSET_ATTR_CMD_MAX,
1885 .policy = ip_set_protocol_policy,
1889 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = {
1891 .subsys_id = NFNL_SUBSYS_IPSET,
1892 .cb_count = IPSET_MSG_MAX,
1893 .cb = ip_set_netlink_subsys_cb,
1896 /* Interface to iptables/ip6tables */
1899 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len)
1903 int copylen = *len, ret = 0;
1904 struct net *net = sock_net(sk);
1905 struct ip_set_net *inst = ip_set_pernet(net);
1907 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1909 if (optval != SO_IP_SET)
1911 if (*len < sizeof(unsigned int))
1914 data = vmalloc(*len);
1917 if (copy_from_user(data, user, *len) != 0) {
1921 op = (unsigned int *)data;
1923 if (*op < IP_SET_OP_VERSION) {
1924 /* Check the version at the beginning of operations */
1925 struct ip_set_req_version *req_version = data;
1927 if (*len < sizeof(struct ip_set_req_version)) {
1932 if (req_version->version != IPSET_PROTOCOL) {
1939 case IP_SET_OP_VERSION: {
1940 struct ip_set_req_version *req_version = data;
1942 if (*len != sizeof(struct ip_set_req_version)) {
1947 req_version->version = IPSET_PROTOCOL;
1948 if (copy_to_user(user, req_version,
1949 sizeof(struct ip_set_req_version)))
1953 case IP_SET_OP_GET_BYNAME: {
1954 struct ip_set_req_get_set *req_get = data;
1957 if (*len != sizeof(struct ip_set_req_get_set)) {
1961 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
1962 nfnl_lock(NFNL_SUBSYS_IPSET);
1963 find_set_and_id(inst, req_get->set.name, &id);
1964 req_get->set.index = id;
1965 nfnl_unlock(NFNL_SUBSYS_IPSET);
1968 case IP_SET_OP_GET_FNAME: {
1969 struct ip_set_req_get_set_family *req_get = data;
1972 if (*len != sizeof(struct ip_set_req_get_set_family)) {
1976 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
1977 nfnl_lock(NFNL_SUBSYS_IPSET);
1978 find_set_and_id(inst, req_get->set.name, &id);
1979 req_get->set.index = id;
1980 if (id != IPSET_INVALID_ID)
1981 req_get->family = ip_set(inst, id)->family;
1982 nfnl_unlock(NFNL_SUBSYS_IPSET);
1985 case IP_SET_OP_GET_BYINDEX: {
1986 struct ip_set_req_get_set *req_get = data;
1989 if (*len != sizeof(struct ip_set_req_get_set) ||
1990 req_get->set.index >= inst->ip_set_max) {
1994 nfnl_lock(NFNL_SUBSYS_IPSET);
1995 set = ip_set(inst, req_get->set.index);
1996 strncpy(req_get->set.name, set ? set->name : "",
1998 nfnl_unlock(NFNL_SUBSYS_IPSET);
2004 } /* end of switch(op) */
2007 if (copy_to_user(user, data, copylen))
2017 static struct nf_sockopt_ops so_set __read_mostly = {
2019 .get_optmin = SO_IP_SET,
2020 .get_optmax = SO_IP_SET + 1,
2021 .get = &ip_set_sockfn_get,
2022 .owner = THIS_MODULE,
2025 static int __net_init
2026 ip_set_net_init(struct net *net)
2028 struct ip_set_net *inst = ip_set_pernet(net);
2029 struct ip_set **list;
2031 inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX;
2032 if (inst->ip_set_max >= IPSET_INVALID_ID)
2033 inst->ip_set_max = IPSET_INVALID_ID - 1;
2035 list = kcalloc(inst->ip_set_max, sizeof(struct ip_set *), GFP_KERNEL);
2038 inst->is_deleted = false;
2039 inst->is_destroyed = false;
2040 rcu_assign_pointer(inst->ip_set_list, list);
2044 static void __net_exit
2045 ip_set_net_exit(struct net *net)
2047 struct ip_set_net *inst = ip_set_pernet(net);
2049 struct ip_set *set = NULL;
2052 inst->is_deleted = true; /* flag for ip_set_nfnl_put */
2054 for (i = 0; i < inst->ip_set_max; i++) {
2055 set = ip_set(inst, i);
2057 ip_set(inst, i) = NULL;
2058 ip_set_destroy_set(set);
2061 kfree(rcu_dereference_protected(inst->ip_set_list, 1));
2064 static struct pernet_operations ip_set_net_ops = {
2065 .init = ip_set_net_init,
2066 .exit = ip_set_net_exit,
2067 .id = &ip_set_net_id,
2068 .size = sizeof(struct ip_set_net)
2074 int ret = nfnetlink_subsys_register(&ip_set_netlink_subsys);
2077 pr_err("ip_set: cannot register with nfnetlink.\n");
2080 ret = nf_register_sockopt(&so_set);
2082 pr_err("SO_SET registry failed: %d\n", ret);
2083 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2086 ret = register_pernet_subsys(&ip_set_net_ops);
2088 pr_err("ip_set: cannot register pernet_subsys.\n");
2089 nf_unregister_sockopt(&so_set);
2090 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2093 pr_info("ip_set: protocol %u\n", IPSET_PROTOCOL);
2100 unregister_pernet_subsys(&ip_set_net_ops);
2101 nf_unregister_sockopt(&so_set);
2102 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2103 pr_debug("these are the famous last words\n");
2106 module_init(ip_set_init);
2107 module_exit(ip_set_fini);