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 len
= ALIGN(len
, ip_set_extensions
[id
].align
);
383 set
->offset
[id
] = len
;
384 set
->extensions
|= ip_set_extensions
[id
].type
;
385 len
+= ip_set_extensions
[id
].len
;
387 return ALIGN(len
, align
);
389 EXPORT_SYMBOL_GPL(ip_set_elem_len
);
392 ip_set_get_extensions(struct ip_set
*set
, struct nlattr
*tb
[],
393 struct ip_set_ext
*ext
)
397 if (unlikely(!ip_set_optattr_netorder(tb
, IPSET_ATTR_TIMEOUT
) ||
398 !ip_set_optattr_netorder(tb
, IPSET_ATTR_PACKETS
) ||
399 !ip_set_optattr_netorder(tb
, IPSET_ATTR_BYTES
) ||
400 !ip_set_optattr_netorder(tb
, IPSET_ATTR_SKBMARK
) ||
401 !ip_set_optattr_netorder(tb
, IPSET_ATTR_SKBPRIO
) ||
402 !ip_set_optattr_netorder(tb
, IPSET_ATTR_SKBQUEUE
)))
403 return -IPSET_ERR_PROTOCOL
;
405 if (tb
[IPSET_ATTR_TIMEOUT
]) {
406 if (!SET_WITH_TIMEOUT(set
))
407 return -IPSET_ERR_TIMEOUT
;
408 ext
->timeout
= ip_set_timeout_uget(tb
[IPSET_ATTR_TIMEOUT
]);
410 if (tb
[IPSET_ATTR_BYTES
] || tb
[IPSET_ATTR_PACKETS
]) {
411 if (!SET_WITH_COUNTER(set
))
412 return -IPSET_ERR_COUNTER
;
413 if (tb
[IPSET_ATTR_BYTES
])
414 ext
->bytes
= be64_to_cpu(nla_get_be64(
415 tb
[IPSET_ATTR_BYTES
]));
416 if (tb
[IPSET_ATTR_PACKETS
])
417 ext
->packets
= be64_to_cpu(nla_get_be64(
418 tb
[IPSET_ATTR_PACKETS
]));
420 if (tb
[IPSET_ATTR_COMMENT
]) {
421 if (!SET_WITH_COMMENT(set
))
422 return -IPSET_ERR_COMMENT
;
423 ext
->comment
= ip_set_comment_uget(tb
[IPSET_ATTR_COMMENT
]);
425 if (tb
[IPSET_ATTR_SKBMARK
]) {
426 if (!SET_WITH_SKBINFO(set
))
427 return -IPSET_ERR_SKBINFO
;
428 fullmark
= be64_to_cpu(nla_get_be64(tb
[IPSET_ATTR_SKBMARK
]));
429 ext
->skbmark
= fullmark
>> 32;
430 ext
->skbmarkmask
= fullmark
& 0xffffffff;
432 if (tb
[IPSET_ATTR_SKBPRIO
]) {
433 if (!SET_WITH_SKBINFO(set
))
434 return -IPSET_ERR_SKBINFO
;
435 ext
->skbprio
= be32_to_cpu(nla_get_be32(
436 tb
[IPSET_ATTR_SKBPRIO
]));
438 if (tb
[IPSET_ATTR_SKBQUEUE
]) {
439 if (!SET_WITH_SKBINFO(set
))
440 return -IPSET_ERR_SKBINFO
;
441 ext
->skbqueue
= be16_to_cpu(nla_get_be16(
442 tb
[IPSET_ATTR_SKBQUEUE
]));
446 EXPORT_SYMBOL_GPL(ip_set_get_extensions
);
449 ip_set_put_extensions(struct sk_buff
*skb
, const struct ip_set
*set
,
450 const void *e
, bool active
)
452 if (SET_WITH_TIMEOUT(set
)) {
453 unsigned long *timeout
= ext_timeout(e
, set
);
455 if (nla_put_net32(skb
, IPSET_ATTR_TIMEOUT
,
456 htonl(active
? ip_set_timeout_get(timeout
)
460 if (SET_WITH_COUNTER(set
) &&
461 ip_set_put_counter(skb
, ext_counter(e
, set
)))
463 if (SET_WITH_COMMENT(set
) &&
464 ip_set_put_comment(skb
, ext_comment(e
, set
)))
466 if (SET_WITH_SKBINFO(set
) &&
467 ip_set_put_skbinfo(skb
, ext_skbinfo(e
, set
)))
471 EXPORT_SYMBOL_GPL(ip_set_put_extensions
);
473 /* Creating/destroying/renaming/swapping affect the existence and
474 * the properties of a set. All of these can be executed from userspace
475 * only and serialized by the nfnl mutex indirectly from nfnetlink.
477 * Sets are identified by their index in ip_set_list and the index
478 * is used by the external references (set/SET netfilter modules).
480 * The set behind an index may change by swapping only, from userspace.
484 __ip_set_get(struct ip_set
*set
)
486 write_lock_bh(&ip_set_ref_lock
);
488 write_unlock_bh(&ip_set_ref_lock
);
492 __ip_set_put(struct ip_set
*set
)
494 write_lock_bh(&ip_set_ref_lock
);
495 BUG_ON(set
->ref
== 0);
497 write_unlock_bh(&ip_set_ref_lock
);
500 /* set->ref can be swapped out by ip_set_swap, netlink events (like dump) need
501 * a separate reference counter
504 __ip_set_get_netlink(struct ip_set
*set
)
506 write_lock_bh(&ip_set_ref_lock
);
508 write_unlock_bh(&ip_set_ref_lock
);
512 __ip_set_put_netlink(struct ip_set
*set
)
514 write_lock_bh(&ip_set_ref_lock
);
515 BUG_ON(set
->ref_netlink
== 0);
517 write_unlock_bh(&ip_set_ref_lock
);
520 /* Add, del and test set entries from kernel.
522 * The set behind the index must exist and must be referenced
523 * so it can't be destroyed (or changed) under our foot.
526 static inline struct ip_set
*
527 ip_set_rcu_get(struct net
*net
, ip_set_id_t index
)
530 struct ip_set_net
*inst
= ip_set_pernet(net
);
533 /* ip_set_list itself needs to be protected */
534 set
= rcu_dereference(inst
->ip_set_list
)[index
];
541 ip_set_test(ip_set_id_t index
, const struct sk_buff
*skb
,
542 const struct xt_action_param
*par
, struct ip_set_adt_opt
*opt
)
544 struct ip_set
*set
= ip_set_rcu_get(xt_net(par
), index
);
548 pr_debug("set %s, index %u\n", set
->name
, index
);
550 if (opt
->dim
< set
->type
->dimension
||
551 !(opt
->family
== set
->family
|| set
->family
== NFPROTO_UNSPEC
))
555 ret
= set
->variant
->kadt(set
, skb
, par
, IPSET_TEST
, opt
);
556 rcu_read_unlock_bh();
558 if (ret
== -EAGAIN
) {
559 /* Type requests element to be completed */
560 pr_debug("element must be completed, ADD is triggered\n");
561 spin_lock_bh(&set
->lock
);
562 set
->variant
->kadt(set
, skb
, par
, IPSET_ADD
, opt
);
563 spin_unlock_bh(&set
->lock
);
566 /* --return-nomatch: invert matched element */
567 if ((opt
->cmdflags
& IPSET_FLAG_RETURN_NOMATCH
) &&
568 (set
->type
->features
& IPSET_TYPE_NOMATCH
) &&
569 (ret
> 0 || ret
== -ENOTEMPTY
))
573 /* Convert error codes to nomatch */
574 return (ret
< 0 ? 0 : ret
);
576 EXPORT_SYMBOL_GPL(ip_set_test
);
579 ip_set_add(ip_set_id_t index
, const struct sk_buff
*skb
,
580 const struct xt_action_param
*par
, struct ip_set_adt_opt
*opt
)
582 struct ip_set
*set
= ip_set_rcu_get(xt_net(par
), index
);
586 pr_debug("set %s, index %u\n", set
->name
, index
);
588 if (opt
->dim
< set
->type
->dimension
||
589 !(opt
->family
== set
->family
|| set
->family
== NFPROTO_UNSPEC
))
590 return -IPSET_ERR_TYPE_MISMATCH
;
592 spin_lock_bh(&set
->lock
);
593 ret
= set
->variant
->kadt(set
, skb
, par
, IPSET_ADD
, opt
);
594 spin_unlock_bh(&set
->lock
);
598 EXPORT_SYMBOL_GPL(ip_set_add
);
601 ip_set_del(ip_set_id_t index
, const struct sk_buff
*skb
,
602 const struct xt_action_param
*par
, struct ip_set_adt_opt
*opt
)
604 struct ip_set
*set
= ip_set_rcu_get(xt_net(par
), index
);
608 pr_debug("set %s, index %u\n", set
->name
, index
);
610 if (opt
->dim
< set
->type
->dimension
||
611 !(opt
->family
== set
->family
|| set
->family
== NFPROTO_UNSPEC
))
612 return -IPSET_ERR_TYPE_MISMATCH
;
614 spin_lock_bh(&set
->lock
);
615 ret
= set
->variant
->kadt(set
, skb
, par
, IPSET_DEL
, opt
);
616 spin_unlock_bh(&set
->lock
);
620 EXPORT_SYMBOL_GPL(ip_set_del
);
622 /* Find set by name, reference it once. The reference makes sure the
623 * thing pointed to, does not go away under our feet.
627 ip_set_get_byname(struct net
*net
, const char *name
, struct ip_set
**set
)
629 ip_set_id_t i
, index
= IPSET_INVALID_ID
;
631 struct ip_set_net
*inst
= ip_set_pernet(net
);
634 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
635 s
= rcu_dereference(inst
->ip_set_list
)[i
];
636 if (s
&& STRNCMP(s
->name
, name
)) {
647 EXPORT_SYMBOL_GPL(ip_set_get_byname
);
649 /* If the given set pointer points to a valid set, decrement
650 * reference count by 1. The caller shall not assume the index
651 * to be valid, after calling this function.
656 __ip_set_put_byindex(struct ip_set_net
*inst
, ip_set_id_t index
)
661 set
= rcu_dereference(inst
->ip_set_list
)[index
];
668 ip_set_put_byindex(struct net
*net
, ip_set_id_t index
)
670 struct ip_set_net
*inst
= ip_set_pernet(net
);
672 __ip_set_put_byindex(inst
, index
);
674 EXPORT_SYMBOL_GPL(ip_set_put_byindex
);
676 /* Get the name of a set behind a set index.
677 * We assume the set is referenced, so it does exist and
678 * can't be destroyed. The set cannot be renamed due to
679 * the referencing either.
683 ip_set_name_byindex(struct net
*net
, ip_set_id_t index
)
685 const struct ip_set
*set
= ip_set_rcu_get(net
, index
);
688 BUG_ON(set
->ref
== 0);
690 /* Referenced, so it's safe */
693 EXPORT_SYMBOL_GPL(ip_set_name_byindex
);
695 /* Routines to call by external subsystems, which do not
696 * call nfnl_lock for us.
699 /* Find set by index, reference it once. The reference makes sure the
700 * thing pointed to, does not go away under our feet.
702 * The nfnl mutex is used in the function.
705 ip_set_nfnl_get_byindex(struct net
*net
, ip_set_id_t index
)
708 struct ip_set_net
*inst
= ip_set_pernet(net
);
710 if (index
>= inst
->ip_set_max
)
711 return IPSET_INVALID_ID
;
713 nfnl_lock(NFNL_SUBSYS_IPSET
);
714 set
= ip_set(inst
, index
);
718 index
= IPSET_INVALID_ID
;
719 nfnl_unlock(NFNL_SUBSYS_IPSET
);
723 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex
);
725 /* If the given set pointer points to a valid set, decrement
726 * reference count by 1. The caller shall not assume the index
727 * to be valid, after calling this function.
729 * The nfnl mutex is used in the function.
732 ip_set_nfnl_put(struct net
*net
, ip_set_id_t index
)
735 struct ip_set_net
*inst
= ip_set_pernet(net
);
737 nfnl_lock(NFNL_SUBSYS_IPSET
);
738 if (!inst
->is_deleted
) { /* already deleted from ip_set_net_exit() */
739 set
= ip_set(inst
, index
);
743 nfnl_unlock(NFNL_SUBSYS_IPSET
);
745 EXPORT_SYMBOL_GPL(ip_set_nfnl_put
);
747 /* Communication protocol with userspace over netlink.
749 * The commands are serialized by the nfnl mutex.
753 protocol_failed(const struct nlattr
* const tb
[])
755 return !tb
[IPSET_ATTR_PROTOCOL
] ||
756 nla_get_u8(tb
[IPSET_ATTR_PROTOCOL
]) != IPSET_PROTOCOL
;
760 flag_exist(const struct nlmsghdr
*nlh
)
762 return nlh
->nlmsg_flags
& NLM_F_EXCL
? 0 : IPSET_FLAG_EXIST
;
765 static struct nlmsghdr
*
766 start_msg(struct sk_buff
*skb
, u32 portid
, u32 seq
, unsigned int flags
,
769 struct nlmsghdr
*nlh
;
770 struct nfgenmsg
*nfmsg
;
772 nlh
= nlmsg_put(skb
, portid
, seq
, cmd
| (NFNL_SUBSYS_IPSET
<< 8),
773 sizeof(*nfmsg
), flags
);
777 nfmsg
= nlmsg_data(nlh
);
778 nfmsg
->nfgen_family
= NFPROTO_IPV4
;
779 nfmsg
->version
= NFNETLINK_V0
;
787 static const struct nla_policy ip_set_create_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
788 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
789 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
790 .len
= IPSET_MAXNAMELEN
- 1 },
791 [IPSET_ATTR_TYPENAME
] = { .type
= NLA_NUL_STRING
,
792 .len
= IPSET_MAXNAMELEN
- 1},
793 [IPSET_ATTR_REVISION
] = { .type
= NLA_U8
},
794 [IPSET_ATTR_FAMILY
] = { .type
= NLA_U8
},
795 [IPSET_ATTR_DATA
] = { .type
= NLA_NESTED
},
798 static struct ip_set
*
799 find_set_and_id(struct ip_set_net
*inst
, const char *name
, ip_set_id_t
*id
)
801 struct ip_set
*set
= NULL
;
804 *id
= IPSET_INVALID_ID
;
805 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
806 set
= ip_set(inst
, i
);
807 if (set
&& STRNCMP(set
->name
, name
)) {
812 return (*id
== IPSET_INVALID_ID
? NULL
: set
);
815 static inline struct ip_set
*
816 find_set(struct ip_set_net
*inst
, const char *name
)
820 return find_set_and_id(inst
, name
, &id
);
824 find_free_id(struct ip_set_net
*inst
, const char *name
, ip_set_id_t
*index
,
830 *index
= IPSET_INVALID_ID
;
831 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
834 if (*index
== IPSET_INVALID_ID
)
836 } else if (STRNCMP(name
, s
->name
)) {
842 if (*index
== IPSET_INVALID_ID
)
843 /* No free slot remained */
844 return -IPSET_ERR_MAX_SETS
;
848 static int ip_set_none(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
849 const struct nlmsghdr
*nlh
,
850 const struct nlattr
* const attr
[])
855 static int ip_set_create(struct net
*net
, struct sock
*ctnl
,
856 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
857 const struct nlattr
* const attr
[])
859 struct ip_set_net
*inst
= ip_set_pernet(net
);
860 struct ip_set
*set
, *clash
= NULL
;
861 ip_set_id_t index
= IPSET_INVALID_ID
;
862 struct nlattr
*tb
[IPSET_ATTR_CREATE_MAX
+ 1] = {};
863 const char *name
, *typename
;
865 u32 flags
= flag_exist(nlh
);
868 if (unlikely(protocol_failed(attr
) ||
869 !attr
[IPSET_ATTR_SETNAME
] ||
870 !attr
[IPSET_ATTR_TYPENAME
] ||
871 !attr
[IPSET_ATTR_REVISION
] ||
872 !attr
[IPSET_ATTR_FAMILY
] ||
873 (attr
[IPSET_ATTR_DATA
] &&
874 !flag_nested(attr
[IPSET_ATTR_DATA
]))))
875 return -IPSET_ERR_PROTOCOL
;
877 name
= nla_data(attr
[IPSET_ATTR_SETNAME
]);
878 typename
= nla_data(attr
[IPSET_ATTR_TYPENAME
]);
879 family
= nla_get_u8(attr
[IPSET_ATTR_FAMILY
]);
880 revision
= nla_get_u8(attr
[IPSET_ATTR_REVISION
]);
881 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
882 name
, typename
, family_name(family
), revision
);
884 /* First, and without any locks, allocate and initialize
885 * a normal base set structure.
887 set
= kzalloc(sizeof(*set
), GFP_KERNEL
);
890 spin_lock_init(&set
->lock
);
891 strlcpy(set
->name
, name
, IPSET_MAXNAMELEN
);
892 set
->family
= family
;
893 set
->revision
= revision
;
895 /* Next, check that we know the type, and take
896 * a reference on the type, to make sure it stays available
897 * while constructing our new set.
899 * After referencing the type, we try to create the type
900 * specific part of the set without holding any locks.
902 ret
= find_set_type_get(typename
, family
, revision
, &set
->type
);
906 /* Without holding any locks, create private part. */
907 if (attr
[IPSET_ATTR_DATA
] &&
908 nla_parse_nested(tb
, IPSET_ATTR_CREATE_MAX
, attr
[IPSET_ATTR_DATA
],
909 set
->type
->create_policy
)) {
910 ret
= -IPSET_ERR_PROTOCOL
;
914 ret
= set
->type
->create(net
, set
, tb
, flags
);
918 /* BTW, ret==0 here. */
920 /* Here, we have a valid, constructed set and we are protected
921 * by the nfnl mutex. Find the first free index in ip_set_list
922 * and check clashing.
924 ret
= find_free_id(inst
, set
->name
, &index
, &clash
);
925 if (ret
== -EEXIST
) {
926 /* If this is the same set and requested, ignore error */
927 if ((flags
& IPSET_FLAG_EXIST
) &&
928 STRNCMP(set
->type
->name
, clash
->type
->name
) &&
929 set
->type
->family
== clash
->type
->family
&&
930 set
->type
->revision_min
== clash
->type
->revision_min
&&
931 set
->type
->revision_max
== clash
->type
->revision_max
&&
932 set
->variant
->same_set(set
, clash
))
935 } else if (ret
== -IPSET_ERR_MAX_SETS
) {
936 struct ip_set
**list
, **tmp
;
937 ip_set_id_t i
= inst
->ip_set_max
+ IP_SET_INC
;
939 if (i
< inst
->ip_set_max
|| i
== IPSET_INVALID_ID
)
943 list
= kcalloc(i
, sizeof(struct ip_set
*), GFP_KERNEL
);
946 /* nfnl mutex is held, both lists are valid */
947 tmp
= ip_set_dereference(inst
->ip_set_list
);
948 memcpy(list
, tmp
, sizeof(struct ip_set
*) * inst
->ip_set_max
);
949 rcu_assign_pointer(inst
->ip_set_list
, list
);
950 /* Make sure all current packets have passed through */
953 index
= inst
->ip_set_max
;
954 inst
->ip_set_max
= i
;
961 /* Finally! Add our shiny new set to the list, and be done. */
962 pr_debug("create: '%s' created with index %u!\n", set
->name
, index
);
963 ip_set(inst
, index
) = set
;
968 set
->variant
->destroy(set
);
970 module_put(set
->type
->me
);
978 static const struct nla_policy
979 ip_set_setname_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
980 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
981 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
982 .len
= IPSET_MAXNAMELEN
- 1 },
986 ip_set_destroy_set(struct ip_set
*set
)
988 pr_debug("set: %s\n", set
->name
);
990 /* Must call it without holding any lock */
991 set
->variant
->destroy(set
);
992 module_put(set
->type
->me
);
996 static int ip_set_destroy(struct net
*net
, struct sock
*ctnl
,
997 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
998 const struct nlattr
* const attr
[])
1000 struct ip_set_net
*inst
= ip_set_pernet(net
);
1005 if (unlikely(protocol_failed(attr
)))
1006 return -IPSET_ERR_PROTOCOL
;
1008 /* Must wait for flush to be really finished in list:set */
1011 /* Commands are serialized and references are
1012 * protected by the ip_set_ref_lock.
1013 * External systems (i.e. xt_set) must call
1014 * ip_set_put|get_nfnl_* functions, that way we
1015 * can safely check references here.
1017 * list:set timer can only decrement the reference
1018 * counter, so if it's already zero, we can proceed
1019 * without holding the lock.
1021 read_lock_bh(&ip_set_ref_lock
);
1022 if (!attr
[IPSET_ATTR_SETNAME
]) {
1023 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1024 s
= ip_set(inst
, i
);
1025 if (s
&& (s
->ref
|| s
->ref_netlink
)) {
1026 ret
= -IPSET_ERR_BUSY
;
1030 inst
->is_destroyed
= true;
1031 read_unlock_bh(&ip_set_ref_lock
);
1032 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1033 s
= ip_set(inst
, i
);
1035 ip_set(inst
, i
) = NULL
;
1036 ip_set_destroy_set(s
);
1039 /* Modified by ip_set_destroy() only, which is serialized */
1040 inst
->is_destroyed
= false;
1042 s
= find_set_and_id(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]),
1047 } else if (s
->ref
|| s
->ref_netlink
) {
1048 ret
= -IPSET_ERR_BUSY
;
1051 ip_set(inst
, i
) = NULL
;
1052 read_unlock_bh(&ip_set_ref_lock
);
1054 ip_set_destroy_set(s
);
1058 read_unlock_bh(&ip_set_ref_lock
);
1065 ip_set_flush_set(struct ip_set
*set
)
1067 pr_debug("set: %s\n", set
->name
);
1069 spin_lock_bh(&set
->lock
);
1070 set
->variant
->flush(set
);
1071 spin_unlock_bh(&set
->lock
);
1074 static int ip_set_flush(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1075 const struct nlmsghdr
*nlh
,
1076 const struct nlattr
* const attr
[])
1078 struct ip_set_net
*inst
= ip_set_pernet(net
);
1082 if (unlikely(protocol_failed(attr
)))
1083 return -IPSET_ERR_PROTOCOL
;
1085 if (!attr
[IPSET_ATTR_SETNAME
]) {
1086 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1087 s
= ip_set(inst
, i
);
1089 ip_set_flush_set(s
);
1092 s
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1096 ip_set_flush_set(s
);
1104 static const struct nla_policy
1105 ip_set_setname2_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1106 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1107 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
1108 .len
= IPSET_MAXNAMELEN
- 1 },
1109 [IPSET_ATTR_SETNAME2
] = { .type
= NLA_NUL_STRING
,
1110 .len
= IPSET_MAXNAMELEN
- 1 },
1113 static int ip_set_rename(struct net
*net
, struct sock
*ctnl
,
1114 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1115 const struct nlattr
* const attr
[])
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 read_lock_bh(&ip_set_ref_lock
);
1133 if (set
->ref
!= 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 read_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
[])
1166 struct ip_set_net
*inst
= ip_set_pernet(net
);
1167 struct ip_set
*from
, *to
;
1168 ip_set_id_t from_id
, to_id
;
1169 char from_name
[IPSET_MAXNAMELEN
];
1171 if (unlikely(protocol_failed(attr
) ||
1172 !attr
[IPSET_ATTR_SETNAME
] ||
1173 !attr
[IPSET_ATTR_SETNAME2
]))
1174 return -IPSET_ERR_PROTOCOL
;
1176 from
= find_set_and_id(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]),
1181 to
= find_set_and_id(inst
, nla_data(attr
[IPSET_ATTR_SETNAME2
]),
1184 return -IPSET_ERR_EXIST_SETNAME2
;
1186 /* Features must not change.
1187 * Not an artifical restriction anymore, as we must prevent
1188 * possible loops created by swapping in setlist type of sets.
1190 if (!(from
->type
->features
== to
->type
->features
&&
1191 from
->family
== to
->family
))
1192 return -IPSET_ERR_TYPE_MISMATCH
;
1194 if (from
->ref_netlink
|| to
->ref_netlink
)
1197 strncpy(from_name
, from
->name
, IPSET_MAXNAMELEN
);
1198 strncpy(from
->name
, to
->name
, IPSET_MAXNAMELEN
);
1199 strncpy(to
->name
, from_name
, IPSET_MAXNAMELEN
);
1201 write_lock_bh(&ip_set_ref_lock
);
1202 swap(from
->ref
, to
->ref
);
1203 ip_set(inst
, from_id
) = to
;
1204 ip_set(inst
, to_id
) = from
;
1205 write_unlock_bh(&ip_set_ref_lock
);
1210 /* List/save set data */
1217 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1218 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1221 ip_set_dump_done(struct netlink_callback
*cb
)
1223 if (cb
->args
[IPSET_CB_ARG0
]) {
1224 struct ip_set_net
*inst
=
1225 (struct ip_set_net
*)cb
->args
[IPSET_CB_NET
];
1226 ip_set_id_t index
= (ip_set_id_t
)cb
->args
[IPSET_CB_INDEX
];
1227 struct ip_set
*set
= ip_set(inst
, index
);
1229 if (set
->variant
->uref
)
1230 set
->variant
->uref(set
, cb
, false);
1231 pr_debug("release set %s\n", set
->name
);
1232 __ip_set_put_netlink(set
);
1238 dump_attrs(struct nlmsghdr
*nlh
)
1240 const struct nlattr
*attr
;
1243 pr_debug("dump nlmsg\n");
1244 nlmsg_for_each_attr(attr
, nlh
, sizeof(struct nfgenmsg
), rem
) {
1245 pr_debug("type: %u, len %u\n", nla_type(attr
), attr
->nla_len
);
1250 dump_init(struct netlink_callback
*cb
, struct ip_set_net
*inst
)
1252 struct nlmsghdr
*nlh
= nlmsg_hdr(cb
->skb
);
1253 int min_len
= nlmsg_total_size(sizeof(struct nfgenmsg
));
1254 struct nlattr
*cda
[IPSET_ATTR_CMD_MAX
+ 1];
1255 struct nlattr
*attr
= (void *)nlh
+ min_len
;
1259 /* Second pass, so parser can't fail */
1260 nla_parse(cda
, IPSET_ATTR_CMD_MAX
,
1261 attr
, nlh
->nlmsg_len
- min_len
, ip_set_setname_policy
);
1263 if (cda
[IPSET_ATTR_SETNAME
]) {
1266 set
= find_set_and_id(inst
, nla_data(cda
[IPSET_ATTR_SETNAME
]),
1271 dump_type
= DUMP_ONE
;
1272 cb
->args
[IPSET_CB_INDEX
] = index
;
1274 dump_type
= DUMP_ALL
;
1277 if (cda
[IPSET_ATTR_FLAGS
]) {
1278 u32 f
= ip_set_get_h32(cda
[IPSET_ATTR_FLAGS
]);
1280 dump_type
|= (f
<< 16);
1282 cb
->args
[IPSET_CB_NET
] = (unsigned long)inst
;
1283 cb
->args
[IPSET_CB_DUMP
] = dump_type
;
1289 ip_set_dump_start(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1291 ip_set_id_t index
= IPSET_INVALID_ID
, max
;
1292 struct ip_set
*set
= NULL
;
1293 struct nlmsghdr
*nlh
= NULL
;
1294 unsigned int flags
= NETLINK_CB(cb
->skb
).portid
? NLM_F_MULTI
: 0;
1295 struct ip_set_net
*inst
= ip_set_pernet(sock_net(skb
->sk
));
1296 u32 dump_type
, dump_flags
;
1300 if (!cb
->args
[IPSET_CB_DUMP
]) {
1301 ret
= dump_init(cb
, inst
);
1303 nlh
= nlmsg_hdr(cb
->skb
);
1304 /* We have to create and send the error message
1307 if (nlh
->nlmsg_flags
& NLM_F_ACK
)
1308 netlink_ack(cb
->skb
, nlh
, ret
);
1313 if (cb
->args
[IPSET_CB_INDEX
] >= inst
->ip_set_max
)
1316 dump_type
= DUMP_TYPE(cb
->args
[IPSET_CB_DUMP
]);
1317 dump_flags
= DUMP_FLAGS(cb
->args
[IPSET_CB_DUMP
]);
1318 max
= dump_type
== DUMP_ONE
? cb
->args
[IPSET_CB_INDEX
] + 1
1321 pr_debug("dump type, flag: %u %u index: %ld\n",
1322 dump_type
, dump_flags
, cb
->args
[IPSET_CB_INDEX
]);
1323 for (; cb
->args
[IPSET_CB_INDEX
] < max
; cb
->args
[IPSET_CB_INDEX
]++) {
1324 index
= (ip_set_id_t
)cb
->args
[IPSET_CB_INDEX
];
1325 write_lock_bh(&ip_set_ref_lock
);
1326 set
= ip_set(inst
, index
);
1327 is_destroyed
= inst
->is_destroyed
;
1328 if (!set
|| is_destroyed
) {
1329 write_unlock_bh(&ip_set_ref_lock
);
1330 if (dump_type
== DUMP_ONE
) {
1335 /* All sets are just being destroyed */
1341 /* When dumping all sets, we must dump "sorted"
1342 * so that lists (unions of sets) are dumped last.
1344 if (dump_type
!= DUMP_ONE
&&
1345 ((dump_type
== DUMP_ALL
) ==
1346 !!(set
->type
->features
& IPSET_DUMP_LAST
))) {
1347 write_unlock_bh(&ip_set_ref_lock
);
1350 pr_debug("List set: %s\n", set
->name
);
1351 if (!cb
->args
[IPSET_CB_ARG0
]) {
1352 /* Start listing: make sure set won't be destroyed */
1353 pr_debug("reference set\n");
1356 write_unlock_bh(&ip_set_ref_lock
);
1357 nlh
= start_msg(skb
, NETLINK_CB(cb
->skb
).portid
,
1358 cb
->nlh
->nlmsg_seq
, flags
,
1362 goto release_refcount
;
1364 if (nla_put_u8(skb
, IPSET_ATTR_PROTOCOL
, IPSET_PROTOCOL
) ||
1365 nla_put_string(skb
, IPSET_ATTR_SETNAME
, set
->name
))
1366 goto nla_put_failure
;
1367 if (dump_flags
& IPSET_FLAG_LIST_SETNAME
)
1369 switch (cb
->args
[IPSET_CB_ARG0
]) {
1371 /* Core header data */
1372 if (nla_put_string(skb
, IPSET_ATTR_TYPENAME
,
1374 nla_put_u8(skb
, IPSET_ATTR_FAMILY
,
1376 nla_put_u8(skb
, IPSET_ATTR_REVISION
,
1378 goto nla_put_failure
;
1379 ret
= set
->variant
->head(set
, skb
);
1381 goto release_refcount
;
1382 if (dump_flags
& IPSET_FLAG_LIST_HEADER
)
1384 if (set
->variant
->uref
)
1385 set
->variant
->uref(set
, cb
, true);
1386 /* Fall through and add elements */
1389 ret
= set
->variant
->list(set
, skb
, cb
);
1390 rcu_read_unlock_bh();
1391 if (!cb
->args
[IPSET_CB_ARG0
])
1392 /* Set is done, proceed with next one */
1394 goto release_refcount
;
1397 /* If we dump all sets, continue with dumping last ones */
1398 if (dump_type
== DUMP_ALL
) {
1399 dump_type
= DUMP_LAST
;
1400 cb
->args
[IPSET_CB_DUMP
] = dump_type
| (dump_flags
<< 16);
1401 cb
->args
[IPSET_CB_INDEX
] = 0;
1402 if (set
&& set
->variant
->uref
)
1403 set
->variant
->uref(set
, cb
, false);
1411 if (dump_type
== DUMP_ONE
)
1412 cb
->args
[IPSET_CB_INDEX
] = IPSET_INVALID_ID
;
1414 cb
->args
[IPSET_CB_INDEX
]++;
1416 /* If there was an error or set is done, release set */
1417 if (ret
|| !cb
->args
[IPSET_CB_ARG0
]) {
1418 set
= ip_set(inst
, index
);
1419 if (set
->variant
->uref
)
1420 set
->variant
->uref(set
, cb
, false);
1421 pr_debug("release set %s\n", set
->name
);
1422 __ip_set_put_netlink(set
);
1423 cb
->args
[IPSET_CB_ARG0
] = 0;
1427 nlmsg_end(skb
, nlh
);
1428 pr_debug("nlmsg_len: %u\n", nlh
->nlmsg_len
);
1432 return ret
< 0 ? ret
: skb
->len
;
1435 static int ip_set_dump(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1436 const struct nlmsghdr
*nlh
,
1437 const struct nlattr
* const attr
[])
1439 if (unlikely(protocol_failed(attr
)))
1440 return -IPSET_ERR_PROTOCOL
;
1443 struct netlink_dump_control c
= {
1444 .dump
= ip_set_dump_start
,
1445 .done
= ip_set_dump_done
,
1447 return netlink_dump_start(ctnl
, skb
, nlh
, &c
);
1451 /* Add, del and test */
1453 static const struct nla_policy ip_set_adt_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1454 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1455 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
1456 .len
= IPSET_MAXNAMELEN
- 1 },
1457 [IPSET_ATTR_LINENO
] = { .type
= NLA_U32
},
1458 [IPSET_ATTR_DATA
] = { .type
= NLA_NESTED
},
1459 [IPSET_ATTR_ADT
] = { .type
= NLA_NESTED
},
1463 call_ad(struct sock
*ctnl
, struct sk_buff
*skb
, struct ip_set
*set
,
1464 struct nlattr
*tb
[], enum ipset_adt adt
,
1465 u32 flags
, bool use_lineno
)
1469 bool eexist
= flags
& IPSET_FLAG_EXIST
, retried
= false;
1472 spin_lock_bh(&set
->lock
);
1473 ret
= set
->variant
->uadt(set
, tb
, adt
, &lineno
, flags
, retried
);
1474 spin_unlock_bh(&set
->lock
);
1476 } while (ret
== -EAGAIN
&&
1477 set
->variant
->resize
&&
1478 (ret
= set
->variant
->resize(set
, retried
)) == 0);
1480 if (!ret
|| (ret
== -IPSET_ERR_EXIST
&& eexist
))
1482 if (lineno
&& use_lineno
) {
1483 /* Error in restore/batch mode: send back lineno */
1484 struct nlmsghdr
*rep
, *nlh
= nlmsg_hdr(skb
);
1485 struct sk_buff
*skb2
;
1486 struct nlmsgerr
*errmsg
;
1487 size_t payload
= min(SIZE_MAX
,
1488 sizeof(*errmsg
) + nlmsg_len(nlh
));
1489 int min_len
= nlmsg_total_size(sizeof(struct nfgenmsg
));
1490 struct nlattr
*cda
[IPSET_ATTR_CMD_MAX
+ 1];
1491 struct nlattr
*cmdattr
;
1494 skb2
= nlmsg_new(payload
, GFP_KERNEL
);
1497 rep
= __nlmsg_put(skb2
, NETLINK_CB(skb
).portid
,
1498 nlh
->nlmsg_seq
, NLMSG_ERROR
, payload
, 0);
1499 errmsg
= nlmsg_data(rep
);
1500 errmsg
->error
= ret
;
1501 memcpy(&errmsg
->msg
, nlh
, nlh
->nlmsg_len
);
1502 cmdattr
= (void *)&errmsg
->msg
+ min_len
;
1504 nla_parse(cda
, IPSET_ATTR_CMD_MAX
,
1505 cmdattr
, nlh
->nlmsg_len
- min_len
,
1508 errline
= nla_data(cda
[IPSET_ATTR_LINENO
]);
1512 netlink_unicast(ctnl
, skb2
, NETLINK_CB(skb
).portid
,
1514 /* Signal netlink not to send its ACK/errmsg. */
1521 static int ip_set_uadd(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1522 const struct nlmsghdr
*nlh
,
1523 const struct nlattr
* const attr
[])
1525 struct ip_set_net
*inst
= ip_set_pernet(net
);
1527 struct nlattr
*tb
[IPSET_ATTR_ADT_MAX
+ 1] = {};
1528 const struct nlattr
*nla
;
1529 u32 flags
= flag_exist(nlh
);
1533 if (unlikely(protocol_failed(attr
) ||
1534 !attr
[IPSET_ATTR_SETNAME
] ||
1535 !((attr
[IPSET_ATTR_DATA
] != NULL
) ^
1536 (attr
[IPSET_ATTR_ADT
] != NULL
)) ||
1537 (attr
[IPSET_ATTR_DATA
] &&
1538 !flag_nested(attr
[IPSET_ATTR_DATA
])) ||
1539 (attr
[IPSET_ATTR_ADT
] &&
1540 (!flag_nested(attr
[IPSET_ATTR_ADT
]) ||
1541 !attr
[IPSET_ATTR_LINENO
]))))
1542 return -IPSET_ERR_PROTOCOL
;
1544 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1548 use_lineno
= !!attr
[IPSET_ATTR_LINENO
];
1549 if (attr
[IPSET_ATTR_DATA
]) {
1550 if (nla_parse_nested(tb
, IPSET_ATTR_ADT_MAX
,
1551 attr
[IPSET_ATTR_DATA
],
1552 set
->type
->adt_policy
))
1553 return -IPSET_ERR_PROTOCOL
;
1554 ret
= call_ad(ctnl
, skb
, set
, tb
, IPSET_ADD
, flags
,
1559 nla_for_each_nested(nla
, attr
[IPSET_ATTR_ADT
], nla_rem
) {
1560 memset(tb
, 0, sizeof(tb
));
1561 if (nla_type(nla
) != IPSET_ATTR_DATA
||
1562 !flag_nested(nla
) ||
1563 nla_parse_nested(tb
, IPSET_ATTR_ADT_MAX
, nla
,
1564 set
->type
->adt_policy
))
1565 return -IPSET_ERR_PROTOCOL
;
1566 ret
= call_ad(ctnl
, skb
, set
, tb
, IPSET_ADD
,
1575 static int ip_set_udel(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1576 const struct nlmsghdr
*nlh
,
1577 const struct nlattr
* const attr
[])
1579 struct ip_set_net
*inst
= ip_set_pernet(net
);
1581 struct nlattr
*tb
[IPSET_ATTR_ADT_MAX
+ 1] = {};
1582 const struct nlattr
*nla
;
1583 u32 flags
= flag_exist(nlh
);
1587 if (unlikely(protocol_failed(attr
) ||
1588 !attr
[IPSET_ATTR_SETNAME
] ||
1589 !((attr
[IPSET_ATTR_DATA
] != NULL
) ^
1590 (attr
[IPSET_ATTR_ADT
] != NULL
)) ||
1591 (attr
[IPSET_ATTR_DATA
] &&
1592 !flag_nested(attr
[IPSET_ATTR_DATA
])) ||
1593 (attr
[IPSET_ATTR_ADT
] &&
1594 (!flag_nested(attr
[IPSET_ATTR_ADT
]) ||
1595 !attr
[IPSET_ATTR_LINENO
]))))
1596 return -IPSET_ERR_PROTOCOL
;
1598 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1602 use_lineno
= !!attr
[IPSET_ATTR_LINENO
];
1603 if (attr
[IPSET_ATTR_DATA
]) {
1604 if (nla_parse_nested(tb
, IPSET_ATTR_ADT_MAX
,
1605 attr
[IPSET_ATTR_DATA
],
1606 set
->type
->adt_policy
))
1607 return -IPSET_ERR_PROTOCOL
;
1608 ret
= call_ad(ctnl
, skb
, set
, tb
, IPSET_DEL
, flags
,
1613 nla_for_each_nested(nla
, attr
[IPSET_ATTR_ADT
], nla_rem
) {
1614 memset(tb
, 0, sizeof(*tb
));
1615 if (nla_type(nla
) != IPSET_ATTR_DATA
||
1616 !flag_nested(nla
) ||
1617 nla_parse_nested(tb
, IPSET_ATTR_ADT_MAX
, nla
,
1618 set
->type
->adt_policy
))
1619 return -IPSET_ERR_PROTOCOL
;
1620 ret
= call_ad(ctnl
, skb
, set
, tb
, IPSET_DEL
,
1629 static int ip_set_utest(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1630 const struct nlmsghdr
*nlh
,
1631 const struct nlattr
* const attr
[])
1633 struct ip_set_net
*inst
= ip_set_pernet(net
);
1635 struct nlattr
*tb
[IPSET_ATTR_ADT_MAX
+ 1] = {};
1638 if (unlikely(protocol_failed(attr
) ||
1639 !attr
[IPSET_ATTR_SETNAME
] ||
1640 !attr
[IPSET_ATTR_DATA
] ||
1641 !flag_nested(attr
[IPSET_ATTR_DATA
])))
1642 return -IPSET_ERR_PROTOCOL
;
1644 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1648 if (nla_parse_nested(tb
, IPSET_ATTR_ADT_MAX
, attr
[IPSET_ATTR_DATA
],
1649 set
->type
->adt_policy
))
1650 return -IPSET_ERR_PROTOCOL
;
1653 ret
= set
->variant
->uadt(set
, tb
, IPSET_TEST
, NULL
, 0, 0);
1654 rcu_read_unlock_bh();
1655 /* Userspace can't trigger element to be re-added */
1659 return ret
> 0 ? 0 : -IPSET_ERR_EXIST
;
1662 /* Get headed data of a set */
1664 static int ip_set_header(struct net
*net
, struct sock
*ctnl
,
1665 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1666 const struct nlattr
* const attr
[])
1668 struct ip_set_net
*inst
= ip_set_pernet(net
);
1669 const struct ip_set
*set
;
1670 struct sk_buff
*skb2
;
1671 struct nlmsghdr
*nlh2
;
1674 if (unlikely(protocol_failed(attr
) ||
1675 !attr
[IPSET_ATTR_SETNAME
]))
1676 return -IPSET_ERR_PROTOCOL
;
1678 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1682 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
1686 nlh2
= start_msg(skb2
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
1690 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, IPSET_PROTOCOL
) ||
1691 nla_put_string(skb2
, IPSET_ATTR_SETNAME
, set
->name
) ||
1692 nla_put_string(skb2
, IPSET_ATTR_TYPENAME
, set
->type
->name
) ||
1693 nla_put_u8(skb2
, IPSET_ATTR_FAMILY
, set
->family
) ||
1694 nla_put_u8(skb2
, IPSET_ATTR_REVISION
, set
->revision
))
1695 goto nla_put_failure
;
1696 nlmsg_end(skb2
, nlh2
);
1698 ret
= netlink_unicast(ctnl
, skb2
, NETLINK_CB(skb
).portid
, MSG_DONTWAIT
);
1705 nlmsg_cancel(skb2
, nlh2
);
1713 static const struct nla_policy ip_set_type_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1714 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1715 [IPSET_ATTR_TYPENAME
] = { .type
= NLA_NUL_STRING
,
1716 .len
= IPSET_MAXNAMELEN
- 1 },
1717 [IPSET_ATTR_FAMILY
] = { .type
= NLA_U8
},
1720 static int ip_set_type(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1721 const struct nlmsghdr
*nlh
,
1722 const struct nlattr
* const attr
[])
1724 struct sk_buff
*skb2
;
1725 struct nlmsghdr
*nlh2
;
1726 u8 family
, min
, max
;
1727 const char *typename
;
1730 if (unlikely(protocol_failed(attr
) ||
1731 !attr
[IPSET_ATTR_TYPENAME
] ||
1732 !attr
[IPSET_ATTR_FAMILY
]))
1733 return -IPSET_ERR_PROTOCOL
;
1735 family
= nla_get_u8(attr
[IPSET_ATTR_FAMILY
]);
1736 typename
= nla_data(attr
[IPSET_ATTR_TYPENAME
]);
1737 ret
= find_set_type_minmax(typename
, family
, &min
, &max
);
1741 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
1745 nlh2
= start_msg(skb2
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
1749 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, IPSET_PROTOCOL
) ||
1750 nla_put_string(skb2
, IPSET_ATTR_TYPENAME
, typename
) ||
1751 nla_put_u8(skb2
, IPSET_ATTR_FAMILY
, family
) ||
1752 nla_put_u8(skb2
, IPSET_ATTR_REVISION
, max
) ||
1753 nla_put_u8(skb2
, IPSET_ATTR_REVISION_MIN
, min
))
1754 goto nla_put_failure
;
1755 nlmsg_end(skb2
, nlh2
);
1757 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2
->nlmsg_len
);
1758 ret
= netlink_unicast(ctnl
, skb2
, NETLINK_CB(skb
).portid
, MSG_DONTWAIT
);
1765 nlmsg_cancel(skb2
, nlh2
);
1771 /* Get protocol version */
1773 static const struct nla_policy
1774 ip_set_protocol_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1775 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1778 static int ip_set_protocol(struct net
*net
, struct sock
*ctnl
,
1779 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1780 const struct nlattr
* const attr
[])
1782 struct sk_buff
*skb2
;
1783 struct nlmsghdr
*nlh2
;
1786 if (unlikely(!attr
[IPSET_ATTR_PROTOCOL
]))
1787 return -IPSET_ERR_PROTOCOL
;
1789 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
1793 nlh2
= start_msg(skb2
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
1794 IPSET_CMD_PROTOCOL
);
1797 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, IPSET_PROTOCOL
))
1798 goto nla_put_failure
;
1799 nlmsg_end(skb2
, nlh2
);
1801 ret
= netlink_unicast(ctnl
, skb2
, NETLINK_CB(skb
).portid
, MSG_DONTWAIT
);
1808 nlmsg_cancel(skb2
, nlh2
);
1814 static const struct nfnl_callback ip_set_netlink_subsys_cb
[IPSET_MSG_MAX
] = {
1815 [IPSET_CMD_NONE
] = {
1816 .call
= ip_set_none
,
1817 .attr_count
= IPSET_ATTR_CMD_MAX
,
1819 [IPSET_CMD_CREATE
] = {
1820 .call
= ip_set_create
,
1821 .attr_count
= IPSET_ATTR_CMD_MAX
,
1822 .policy
= ip_set_create_policy
,
1824 [IPSET_CMD_DESTROY
] = {
1825 .call
= ip_set_destroy
,
1826 .attr_count
= IPSET_ATTR_CMD_MAX
,
1827 .policy
= ip_set_setname_policy
,
1829 [IPSET_CMD_FLUSH
] = {
1830 .call
= ip_set_flush
,
1831 .attr_count
= IPSET_ATTR_CMD_MAX
,
1832 .policy
= ip_set_setname_policy
,
1834 [IPSET_CMD_RENAME
] = {
1835 .call
= ip_set_rename
,
1836 .attr_count
= IPSET_ATTR_CMD_MAX
,
1837 .policy
= ip_set_setname2_policy
,
1839 [IPSET_CMD_SWAP
] = {
1840 .call
= ip_set_swap
,
1841 .attr_count
= IPSET_ATTR_CMD_MAX
,
1842 .policy
= ip_set_setname2_policy
,
1844 [IPSET_CMD_LIST
] = {
1845 .call
= ip_set_dump
,
1846 .attr_count
= IPSET_ATTR_CMD_MAX
,
1847 .policy
= ip_set_setname_policy
,
1849 [IPSET_CMD_SAVE
] = {
1850 .call
= ip_set_dump
,
1851 .attr_count
= IPSET_ATTR_CMD_MAX
,
1852 .policy
= ip_set_setname_policy
,
1855 .call
= ip_set_uadd
,
1856 .attr_count
= IPSET_ATTR_CMD_MAX
,
1857 .policy
= ip_set_adt_policy
,
1860 .call
= ip_set_udel
,
1861 .attr_count
= IPSET_ATTR_CMD_MAX
,
1862 .policy
= ip_set_adt_policy
,
1864 [IPSET_CMD_TEST
] = {
1865 .call
= ip_set_utest
,
1866 .attr_count
= IPSET_ATTR_CMD_MAX
,
1867 .policy
= ip_set_adt_policy
,
1869 [IPSET_CMD_HEADER
] = {
1870 .call
= ip_set_header
,
1871 .attr_count
= IPSET_ATTR_CMD_MAX
,
1872 .policy
= ip_set_setname_policy
,
1874 [IPSET_CMD_TYPE
] = {
1875 .call
= ip_set_type
,
1876 .attr_count
= IPSET_ATTR_CMD_MAX
,
1877 .policy
= ip_set_type_policy
,
1879 [IPSET_CMD_PROTOCOL
] = {
1880 .call
= ip_set_protocol
,
1881 .attr_count
= IPSET_ATTR_CMD_MAX
,
1882 .policy
= ip_set_protocol_policy
,
1886 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly
= {
1888 .subsys_id
= NFNL_SUBSYS_IPSET
,
1889 .cb_count
= IPSET_MSG_MAX
,
1890 .cb
= ip_set_netlink_subsys_cb
,
1893 /* Interface to iptables/ip6tables */
1896 ip_set_sockfn_get(struct sock
*sk
, int optval
, void __user
*user
, int *len
)
1900 int copylen
= *len
, ret
= 0;
1901 struct net
*net
= sock_net(sk
);
1902 struct ip_set_net
*inst
= ip_set_pernet(net
);
1904 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1906 if (optval
!= SO_IP_SET
)
1908 if (*len
< sizeof(unsigned int))
1911 data
= vmalloc(*len
);
1914 if (copy_from_user(data
, user
, *len
) != 0) {
1918 op
= (unsigned int *)data
;
1920 if (*op
< IP_SET_OP_VERSION
) {
1921 /* Check the version at the beginning of operations */
1922 struct ip_set_req_version
*req_version
= data
;
1924 if (*len
< sizeof(struct ip_set_req_version
)) {
1929 if (req_version
->version
!= IPSET_PROTOCOL
) {
1936 case IP_SET_OP_VERSION
: {
1937 struct ip_set_req_version
*req_version
= data
;
1939 if (*len
!= sizeof(struct ip_set_req_version
)) {
1944 req_version
->version
= IPSET_PROTOCOL
;
1945 ret
= copy_to_user(user
, req_version
,
1946 sizeof(struct ip_set_req_version
));
1949 case IP_SET_OP_GET_BYNAME
: {
1950 struct ip_set_req_get_set
*req_get
= data
;
1953 if (*len
!= sizeof(struct ip_set_req_get_set
)) {
1957 req_get
->set
.name
[IPSET_MAXNAMELEN
- 1] = '\0';
1958 nfnl_lock(NFNL_SUBSYS_IPSET
);
1959 find_set_and_id(inst
, req_get
->set
.name
, &id
);
1960 req_get
->set
.index
= id
;
1961 nfnl_unlock(NFNL_SUBSYS_IPSET
);
1964 case IP_SET_OP_GET_FNAME
: {
1965 struct ip_set_req_get_set_family
*req_get
= data
;
1968 if (*len
!= sizeof(struct ip_set_req_get_set_family
)) {
1972 req_get
->set
.name
[IPSET_MAXNAMELEN
- 1] = '\0';
1973 nfnl_lock(NFNL_SUBSYS_IPSET
);
1974 find_set_and_id(inst
, req_get
->set
.name
, &id
);
1975 req_get
->set
.index
= id
;
1976 if (id
!= IPSET_INVALID_ID
)
1977 req_get
->family
= ip_set(inst
, id
)->family
;
1978 nfnl_unlock(NFNL_SUBSYS_IPSET
);
1981 case IP_SET_OP_GET_BYINDEX
: {
1982 struct ip_set_req_get_set
*req_get
= data
;
1985 if (*len
!= sizeof(struct ip_set_req_get_set
) ||
1986 req_get
->set
.index
>= inst
->ip_set_max
) {
1990 nfnl_lock(NFNL_SUBSYS_IPSET
);
1991 set
= ip_set(inst
, req_get
->set
.index
);
1992 strncpy(req_get
->set
.name
, set
? set
->name
: "",
1994 nfnl_unlock(NFNL_SUBSYS_IPSET
);
2000 } /* end of switch(op) */
2003 ret
= copy_to_user(user
, data
, copylen
);
2012 static struct nf_sockopt_ops so_set __read_mostly
= {
2014 .get_optmin
= SO_IP_SET
,
2015 .get_optmax
= SO_IP_SET
+ 1,
2016 .get
= &ip_set_sockfn_get
,
2017 .owner
= THIS_MODULE
,
2020 static int __net_init
2021 ip_set_net_init(struct net
*net
)
2023 struct ip_set_net
*inst
= ip_set_pernet(net
);
2024 struct ip_set
**list
;
2026 inst
->ip_set_max
= max_sets
? max_sets
: CONFIG_IP_SET_MAX
;
2027 if (inst
->ip_set_max
>= IPSET_INVALID_ID
)
2028 inst
->ip_set_max
= IPSET_INVALID_ID
- 1;
2030 list
= kcalloc(inst
->ip_set_max
, sizeof(struct ip_set
*), GFP_KERNEL
);
2033 inst
->is_deleted
= false;
2034 inst
->is_destroyed
= false;
2035 rcu_assign_pointer(inst
->ip_set_list
, list
);
2039 static void __net_exit
2040 ip_set_net_exit(struct net
*net
)
2042 struct ip_set_net
*inst
= ip_set_pernet(net
);
2044 struct ip_set
*set
= NULL
;
2047 inst
->is_deleted
= true; /* flag for ip_set_nfnl_put */
2049 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
2050 set
= ip_set(inst
, i
);
2052 ip_set(inst
, i
) = NULL
;
2053 ip_set_destroy_set(set
);
2056 kfree(rcu_dereference_protected(inst
->ip_set_list
, 1));
2059 static struct pernet_operations ip_set_net_ops
= {
2060 .init
= ip_set_net_init
,
2061 .exit
= ip_set_net_exit
,
2062 .id
= &ip_set_net_id
,
2063 .size
= sizeof(struct ip_set_net
)
2069 int ret
= nfnetlink_subsys_register(&ip_set_netlink_subsys
);
2072 pr_err("ip_set: cannot register with nfnetlink.\n");
2075 ret
= nf_register_sockopt(&so_set
);
2077 pr_err("SO_SET registry failed: %d\n", ret
);
2078 nfnetlink_subsys_unregister(&ip_set_netlink_subsys
);
2081 ret
= register_pernet_subsys(&ip_set_net_ops
);
2083 pr_err("ip_set: cannot register pernet_subsys.\n");
2084 nf_unregister_sockopt(&so_set
);
2085 nfnetlink_subsys_unregister(&ip_set_netlink_subsys
);
2088 pr_info("ip_set: protocol %u\n", IPSET_PROTOCOL
);
2095 unregister_pernet_subsys(&ip_set_net_ops
);
2096 nf_unregister_sockopt(&so_set
);
2097 nfnetlink_subsys_unregister(&ip_set_netlink_subsys
);
2098 pr_debug("these are the famous last words\n");
2101 module_init(ip_set_init
);
2102 module_exit(ip_set_fini
);