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
)
369 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
;
377 for (id
= 0; id
< IPSET_EXT_ID_MAX
; id
++) {
378 if (!add_extension(id
, cadt_flags
, tb
))
380 offset
= ALIGN(offset
, ip_set_extensions
[id
].align
);
381 set
->offset
[id
] = offset
;
382 set
->extensions
|= ip_set_extensions
[id
].type
;
383 offset
+= ip_set_extensions
[id
].len
;
387 EXPORT_SYMBOL_GPL(ip_set_elem_len
);
390 ip_set_get_extensions(struct ip_set
*set
, struct nlattr
*tb
[],
391 struct ip_set_ext
*ext
)
395 if (unlikely(!ip_set_optattr_netorder(tb
, IPSET_ATTR_TIMEOUT
) ||
396 !ip_set_optattr_netorder(tb
, IPSET_ATTR_PACKETS
) ||
397 !ip_set_optattr_netorder(tb
, IPSET_ATTR_BYTES
) ||
398 !ip_set_optattr_netorder(tb
, IPSET_ATTR_SKBMARK
) ||
399 !ip_set_optattr_netorder(tb
, IPSET_ATTR_SKBPRIO
) ||
400 !ip_set_optattr_netorder(tb
, IPSET_ATTR_SKBQUEUE
)))
401 return -IPSET_ERR_PROTOCOL
;
403 if (tb
[IPSET_ATTR_TIMEOUT
]) {
404 if (!SET_WITH_TIMEOUT(set
))
405 return -IPSET_ERR_TIMEOUT
;
406 ext
->timeout
= ip_set_timeout_uget(tb
[IPSET_ATTR_TIMEOUT
]);
408 if (tb
[IPSET_ATTR_BYTES
] || tb
[IPSET_ATTR_PACKETS
]) {
409 if (!SET_WITH_COUNTER(set
))
410 return -IPSET_ERR_COUNTER
;
411 if (tb
[IPSET_ATTR_BYTES
])
412 ext
->bytes
= be64_to_cpu(nla_get_be64(
413 tb
[IPSET_ATTR_BYTES
]));
414 if (tb
[IPSET_ATTR_PACKETS
])
415 ext
->packets
= be64_to_cpu(nla_get_be64(
416 tb
[IPSET_ATTR_PACKETS
]));
418 if (tb
[IPSET_ATTR_COMMENT
]) {
419 if (!SET_WITH_COMMENT(set
))
420 return -IPSET_ERR_COMMENT
;
421 ext
->comment
= ip_set_comment_uget(tb
[IPSET_ATTR_COMMENT
]);
423 if (tb
[IPSET_ATTR_SKBMARK
]) {
424 if (!SET_WITH_SKBINFO(set
))
425 return -IPSET_ERR_SKBINFO
;
426 fullmark
= be64_to_cpu(nla_get_be64(tb
[IPSET_ATTR_SKBMARK
]));
427 ext
->skbmark
= fullmark
>> 32;
428 ext
->skbmarkmask
= fullmark
& 0xffffffff;
430 if (tb
[IPSET_ATTR_SKBPRIO
]) {
431 if (!SET_WITH_SKBINFO(set
))
432 return -IPSET_ERR_SKBINFO
;
433 ext
->skbprio
= be32_to_cpu(nla_get_be32(
434 tb
[IPSET_ATTR_SKBPRIO
]));
436 if (tb
[IPSET_ATTR_SKBQUEUE
]) {
437 if (!SET_WITH_SKBINFO(set
))
438 return -IPSET_ERR_SKBINFO
;
439 ext
->skbqueue
= be16_to_cpu(nla_get_be16(
440 tb
[IPSET_ATTR_SKBQUEUE
]));
444 EXPORT_SYMBOL_GPL(ip_set_get_extensions
);
447 ip_set_put_extensions(struct sk_buff
*skb
, const struct ip_set
*set
,
448 const void *e
, bool active
)
450 if (SET_WITH_TIMEOUT(set
)) {
451 unsigned long *timeout
= ext_timeout(e
, set
);
453 if (nla_put_net32(skb
, IPSET_ATTR_TIMEOUT
,
454 htonl(active
? ip_set_timeout_get(timeout
)
458 if (SET_WITH_COUNTER(set
) &&
459 ip_set_put_counter(skb
, ext_counter(e
, set
)))
461 if (SET_WITH_COMMENT(set
) &&
462 ip_set_put_comment(skb
, ext_comment(e
, set
)))
464 if (SET_WITH_SKBINFO(set
) &&
465 ip_set_put_skbinfo(skb
, ext_skbinfo(e
, set
)))
469 EXPORT_SYMBOL_GPL(ip_set_put_extensions
);
471 /* Creating/destroying/renaming/swapping affect the existence and
472 * the properties of a set. All of these can be executed from userspace
473 * only and serialized by the nfnl mutex indirectly from nfnetlink.
475 * Sets are identified by their index in ip_set_list and the index
476 * is used by the external references (set/SET netfilter modules).
478 * The set behind an index may change by swapping only, from userspace.
482 __ip_set_get(struct ip_set
*set
)
484 write_lock_bh(&ip_set_ref_lock
);
486 write_unlock_bh(&ip_set_ref_lock
);
490 __ip_set_put(struct ip_set
*set
)
492 write_lock_bh(&ip_set_ref_lock
);
493 BUG_ON(set
->ref
== 0);
495 write_unlock_bh(&ip_set_ref_lock
);
498 /* Add, del and test set entries from kernel.
500 * The set behind the index must exist and must be referenced
501 * so it can't be destroyed (or changed) under our foot.
504 static inline struct ip_set
*
505 ip_set_rcu_get(struct net
*net
, ip_set_id_t index
)
508 struct ip_set_net
*inst
= ip_set_pernet(net
);
511 /* ip_set_list itself needs to be protected */
512 set
= rcu_dereference(inst
->ip_set_list
)[index
];
519 ip_set_test(ip_set_id_t index
, const struct sk_buff
*skb
,
520 const struct xt_action_param
*par
, struct ip_set_adt_opt
*opt
)
522 struct ip_set
*set
= ip_set_rcu_get(par
->net
, index
);
526 pr_debug("set %s, index %u\n", set
->name
, index
);
528 if (opt
->dim
< set
->type
->dimension
||
529 !(opt
->family
== set
->family
|| set
->family
== NFPROTO_UNSPEC
))
533 ret
= set
->variant
->kadt(set
, skb
, par
, IPSET_TEST
, opt
);
534 rcu_read_unlock_bh();
536 if (ret
== -EAGAIN
) {
537 /* Type requests element to be completed */
538 pr_debug("element must be completed, ADD is triggered\n");
539 spin_lock_bh(&set
->lock
);
540 set
->variant
->kadt(set
, skb
, par
, IPSET_ADD
, opt
);
541 spin_unlock_bh(&set
->lock
);
544 /* --return-nomatch: invert matched element */
545 if ((opt
->cmdflags
& IPSET_FLAG_RETURN_NOMATCH
) &&
546 (set
->type
->features
& IPSET_TYPE_NOMATCH
) &&
547 (ret
> 0 || ret
== -ENOTEMPTY
))
551 /* Convert error codes to nomatch */
552 return (ret
< 0 ? 0 : ret
);
554 EXPORT_SYMBOL_GPL(ip_set_test
);
557 ip_set_add(ip_set_id_t index
, const struct sk_buff
*skb
,
558 const struct xt_action_param
*par
, struct ip_set_adt_opt
*opt
)
560 struct ip_set
*set
= ip_set_rcu_get(par
->net
, index
);
564 pr_debug("set %s, index %u\n", set
->name
, index
);
566 if (opt
->dim
< set
->type
->dimension
||
567 !(opt
->family
== set
->family
|| set
->family
== NFPROTO_UNSPEC
))
568 return -IPSET_ERR_TYPE_MISMATCH
;
570 spin_lock_bh(&set
->lock
);
571 ret
= set
->variant
->kadt(set
, skb
, par
, IPSET_ADD
, opt
);
572 spin_unlock_bh(&set
->lock
);
576 EXPORT_SYMBOL_GPL(ip_set_add
);
579 ip_set_del(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(par
->net
, 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_DEL
, opt
);
594 spin_unlock_bh(&set
->lock
);
598 EXPORT_SYMBOL_GPL(ip_set_del
);
600 /* Find set by name, reference it once. The reference makes sure the
601 * thing pointed to, does not go away under our feet.
605 ip_set_get_byname(struct net
*net
, const char *name
, struct ip_set
**set
)
607 ip_set_id_t i
, index
= IPSET_INVALID_ID
;
609 struct ip_set_net
*inst
= ip_set_pernet(net
);
612 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
613 s
= rcu_dereference(inst
->ip_set_list
)[i
];
614 if (s
&& STRNCMP(s
->name
, name
)) {
625 EXPORT_SYMBOL_GPL(ip_set_get_byname
);
627 /* If the given set pointer points to a valid set, decrement
628 * reference count by 1. The caller shall not assume the index
629 * to be valid, after calling this function.
634 __ip_set_put_byindex(struct ip_set_net
*inst
, ip_set_id_t index
)
639 set
= rcu_dereference(inst
->ip_set_list
)[index
];
646 ip_set_put_byindex(struct net
*net
, ip_set_id_t index
)
648 struct ip_set_net
*inst
= ip_set_pernet(net
);
650 __ip_set_put_byindex(inst
, index
);
652 EXPORT_SYMBOL_GPL(ip_set_put_byindex
);
654 /* Get the name of a set behind a set index.
655 * We assume the set is referenced, so it does exist and
656 * can't be destroyed. The set cannot be renamed due to
657 * the referencing either.
661 ip_set_name_byindex(struct net
*net
, ip_set_id_t index
)
663 const struct ip_set
*set
= ip_set_rcu_get(net
, index
);
666 BUG_ON(set
->ref
== 0);
668 /* Referenced, so it's safe */
671 EXPORT_SYMBOL_GPL(ip_set_name_byindex
);
673 /* Routines to call by external subsystems, which do not
674 * call nfnl_lock for us.
677 /* Find set by index, reference it once. The reference makes sure the
678 * thing pointed to, does not go away under our feet.
680 * The nfnl mutex is used in the function.
683 ip_set_nfnl_get_byindex(struct net
*net
, ip_set_id_t index
)
686 struct ip_set_net
*inst
= ip_set_pernet(net
);
688 if (index
>= inst
->ip_set_max
)
689 return IPSET_INVALID_ID
;
691 nfnl_lock(NFNL_SUBSYS_IPSET
);
692 set
= ip_set(inst
, index
);
696 index
= IPSET_INVALID_ID
;
697 nfnl_unlock(NFNL_SUBSYS_IPSET
);
701 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex
);
703 /* If the given set pointer points to a valid set, decrement
704 * reference count by 1. The caller shall not assume the index
705 * to be valid, after calling this function.
707 * The nfnl mutex is used in the function.
710 ip_set_nfnl_put(struct net
*net
, ip_set_id_t index
)
713 struct ip_set_net
*inst
= ip_set_pernet(net
);
715 nfnl_lock(NFNL_SUBSYS_IPSET
);
716 if (!inst
->is_deleted
) { /* already deleted from ip_set_net_exit() */
717 set
= ip_set(inst
, index
);
721 nfnl_unlock(NFNL_SUBSYS_IPSET
);
723 EXPORT_SYMBOL_GPL(ip_set_nfnl_put
);
725 /* Communication protocol with userspace over netlink.
727 * The commands are serialized by the nfnl mutex.
731 protocol_failed(const struct nlattr
* const tb
[])
733 return !tb
[IPSET_ATTR_PROTOCOL
] ||
734 nla_get_u8(tb
[IPSET_ATTR_PROTOCOL
]) != IPSET_PROTOCOL
;
738 flag_exist(const struct nlmsghdr
*nlh
)
740 return nlh
->nlmsg_flags
& NLM_F_EXCL
? 0 : IPSET_FLAG_EXIST
;
743 static struct nlmsghdr
*
744 start_msg(struct sk_buff
*skb
, u32 portid
, u32 seq
, unsigned int flags
,
747 struct nlmsghdr
*nlh
;
748 struct nfgenmsg
*nfmsg
;
750 nlh
= nlmsg_put(skb
, portid
, seq
, cmd
| (NFNL_SUBSYS_IPSET
<< 8),
751 sizeof(*nfmsg
), flags
);
755 nfmsg
= nlmsg_data(nlh
);
756 nfmsg
->nfgen_family
= NFPROTO_IPV4
;
757 nfmsg
->version
= NFNETLINK_V0
;
765 static const struct nla_policy ip_set_create_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
766 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
767 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
768 .len
= IPSET_MAXNAMELEN
- 1 },
769 [IPSET_ATTR_TYPENAME
] = { .type
= NLA_NUL_STRING
,
770 .len
= IPSET_MAXNAMELEN
- 1},
771 [IPSET_ATTR_REVISION
] = { .type
= NLA_U8
},
772 [IPSET_ATTR_FAMILY
] = { .type
= NLA_U8
},
773 [IPSET_ATTR_DATA
] = { .type
= NLA_NESTED
},
776 static struct ip_set
*
777 find_set_and_id(struct ip_set_net
*inst
, const char *name
, ip_set_id_t
*id
)
779 struct ip_set
*set
= NULL
;
782 *id
= IPSET_INVALID_ID
;
783 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
784 set
= ip_set(inst
, i
);
785 if (set
&& STRNCMP(set
->name
, name
)) {
790 return (*id
== IPSET_INVALID_ID
? NULL
: set
);
793 static inline struct ip_set
*
794 find_set(struct ip_set_net
*inst
, const char *name
)
798 return find_set_and_id(inst
, name
, &id
);
802 find_free_id(struct ip_set_net
*inst
, const char *name
, ip_set_id_t
*index
,
808 *index
= IPSET_INVALID_ID
;
809 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
812 if (*index
== IPSET_INVALID_ID
)
814 } else if (STRNCMP(name
, s
->name
)) {
820 if (*index
== IPSET_INVALID_ID
)
821 /* No free slot remained */
822 return -IPSET_ERR_MAX_SETS
;
827 ip_set_none(struct sock
*ctnl
, struct sk_buff
*skb
,
828 const struct nlmsghdr
*nlh
,
829 const struct nlattr
* const attr
[])
835 ip_set_create(struct sock
*ctnl
, struct sk_buff
*skb
,
836 const struct nlmsghdr
*nlh
,
837 const struct nlattr
* const attr
[])
839 struct net
*net
= sock_net(ctnl
);
840 struct ip_set_net
*inst
= ip_set_pernet(net
);
841 struct ip_set
*set
, *clash
= NULL
;
842 ip_set_id_t index
= IPSET_INVALID_ID
;
843 struct nlattr
*tb
[IPSET_ATTR_CREATE_MAX
+ 1] = {};
844 const char *name
, *typename
;
846 u32 flags
= flag_exist(nlh
);
849 if (unlikely(protocol_failed(attr
) ||
850 !attr
[IPSET_ATTR_SETNAME
] ||
851 !attr
[IPSET_ATTR_TYPENAME
] ||
852 !attr
[IPSET_ATTR_REVISION
] ||
853 !attr
[IPSET_ATTR_FAMILY
] ||
854 (attr
[IPSET_ATTR_DATA
] &&
855 !flag_nested(attr
[IPSET_ATTR_DATA
]))))
856 return -IPSET_ERR_PROTOCOL
;
858 name
= nla_data(attr
[IPSET_ATTR_SETNAME
]);
859 typename
= nla_data(attr
[IPSET_ATTR_TYPENAME
]);
860 family
= nla_get_u8(attr
[IPSET_ATTR_FAMILY
]);
861 revision
= nla_get_u8(attr
[IPSET_ATTR_REVISION
]);
862 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
863 name
, typename
, family_name(family
), revision
);
865 /* First, and without any locks, allocate and initialize
866 * a normal base set structure.
868 set
= kzalloc(sizeof(*set
), GFP_KERNEL
);
871 spin_lock_init(&set
->lock
);
872 strlcpy(set
->name
, name
, IPSET_MAXNAMELEN
);
873 set
->family
= family
;
874 set
->revision
= revision
;
876 /* Next, check that we know the type, and take
877 * a reference on the type, to make sure it stays available
878 * while constructing our new set.
880 * After referencing the type, we try to create the type
881 * specific part of the set without holding any locks.
883 ret
= find_set_type_get(typename
, family
, revision
, &set
->type
);
887 /* Without holding any locks, create private part. */
888 if (attr
[IPSET_ATTR_DATA
] &&
889 nla_parse_nested(tb
, IPSET_ATTR_CREATE_MAX
, attr
[IPSET_ATTR_DATA
],
890 set
->type
->create_policy
)) {
891 ret
= -IPSET_ERR_PROTOCOL
;
895 ret
= set
->type
->create(net
, set
, tb
, flags
);
899 /* BTW, ret==0 here. */
901 /* Here, we have a valid, constructed set and we are protected
902 * by the nfnl mutex. Find the first free index in ip_set_list
903 * and check clashing.
905 ret
= find_free_id(inst
, set
->name
, &index
, &clash
);
906 if (ret
== -EEXIST
) {
907 /* If this is the same set and requested, ignore error */
908 if ((flags
& IPSET_FLAG_EXIST
) &&
909 STRNCMP(set
->type
->name
, clash
->type
->name
) &&
910 set
->type
->family
== clash
->type
->family
&&
911 set
->type
->revision_min
== clash
->type
->revision_min
&&
912 set
->type
->revision_max
== clash
->type
->revision_max
&&
913 set
->variant
->same_set(set
, clash
))
916 } else if (ret
== -IPSET_ERR_MAX_SETS
) {
917 struct ip_set
**list
, **tmp
;
918 ip_set_id_t i
= inst
->ip_set_max
+ IP_SET_INC
;
920 if (i
< inst
->ip_set_max
|| i
== IPSET_INVALID_ID
)
924 list
= kcalloc(i
, sizeof(struct ip_set
*), GFP_KERNEL
);
927 /* nfnl mutex is held, both lists are valid */
928 tmp
= ip_set_dereference(inst
->ip_set_list
);
929 memcpy(list
, tmp
, sizeof(struct ip_set
*) * inst
->ip_set_max
);
930 rcu_assign_pointer(inst
->ip_set_list
, list
);
931 /* Make sure all current packets have passed through */
934 index
= inst
->ip_set_max
;
935 inst
->ip_set_max
= i
;
942 /* Finally! Add our shiny new set to the list, and be done. */
943 pr_debug("create: '%s' created with index %u!\n", set
->name
, index
);
944 ip_set(inst
, index
) = set
;
949 set
->variant
->destroy(set
);
951 module_put(set
->type
->me
);
959 static const struct nla_policy
960 ip_set_setname_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
961 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
962 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
963 .len
= IPSET_MAXNAMELEN
- 1 },
967 ip_set_destroy_set(struct ip_set
*set
)
969 pr_debug("set: %s\n", set
->name
);
971 /* Must call it without holding any lock */
972 set
->variant
->destroy(set
);
973 module_put(set
->type
->me
);
978 ip_set_destroy(struct sock
*ctnl
, struct sk_buff
*skb
,
979 const struct nlmsghdr
*nlh
,
980 const struct nlattr
* const attr
[])
982 struct ip_set_net
*inst
= ip_set_pernet(sock_net(ctnl
));
987 if (unlikely(protocol_failed(attr
)))
988 return -IPSET_ERR_PROTOCOL
;
990 /* Commands are serialized and references are
991 * protected by the ip_set_ref_lock.
992 * External systems (i.e. xt_set) must call
993 * ip_set_put|get_nfnl_* functions, that way we
994 * can safely check references here.
996 * list:set timer can only decrement the reference
997 * counter, so if it's already zero, we can proceed
998 * without holding the lock.
1000 read_lock_bh(&ip_set_ref_lock
);
1001 if (!attr
[IPSET_ATTR_SETNAME
]) {
1002 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1003 s
= ip_set(inst
, i
);
1005 ret
= -IPSET_ERR_BUSY
;
1009 inst
->is_destroyed
= true;
1010 read_unlock_bh(&ip_set_ref_lock
);
1011 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1012 s
= ip_set(inst
, i
);
1014 ip_set(inst
, i
) = NULL
;
1015 ip_set_destroy_set(s
);
1018 /* Modified by ip_set_destroy() only, which is serialized */
1019 inst
->is_destroyed
= false;
1021 s
= find_set_and_id(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]),
1026 } else if (s
->ref
) {
1027 ret
= -IPSET_ERR_BUSY
;
1030 ip_set(inst
, i
) = NULL
;
1031 read_unlock_bh(&ip_set_ref_lock
);
1033 ip_set_destroy_set(s
);
1037 read_unlock_bh(&ip_set_ref_lock
);
1044 ip_set_flush_set(struct ip_set
*set
)
1046 pr_debug("set: %s\n", set
->name
);
1048 spin_lock_bh(&set
->lock
);
1049 set
->variant
->flush(set
);
1050 spin_unlock_bh(&set
->lock
);
1054 ip_set_flush(struct sock
*ctnl
, struct sk_buff
*skb
,
1055 const struct nlmsghdr
*nlh
,
1056 const struct nlattr
* const attr
[])
1058 struct ip_set_net
*inst
= ip_set_pernet(sock_net(ctnl
));
1062 if (unlikely(protocol_failed(attr
)))
1063 return -IPSET_ERR_PROTOCOL
;
1065 if (!attr
[IPSET_ATTR_SETNAME
]) {
1066 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1067 s
= ip_set(inst
, i
);
1069 ip_set_flush_set(s
);
1072 s
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1076 ip_set_flush_set(s
);
1084 static const struct nla_policy
1085 ip_set_setname2_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1086 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1087 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
1088 .len
= IPSET_MAXNAMELEN
- 1 },
1089 [IPSET_ATTR_SETNAME2
] = { .type
= NLA_NUL_STRING
,
1090 .len
= IPSET_MAXNAMELEN
- 1 },
1094 ip_set_rename(struct sock
*ctnl
, struct sk_buff
*skb
,
1095 const struct nlmsghdr
*nlh
,
1096 const struct nlattr
* const attr
[])
1098 struct ip_set_net
*inst
= ip_set_pernet(sock_net(ctnl
));
1099 struct ip_set
*set
, *s
;
1104 if (unlikely(protocol_failed(attr
) ||
1105 !attr
[IPSET_ATTR_SETNAME
] ||
1106 !attr
[IPSET_ATTR_SETNAME2
]))
1107 return -IPSET_ERR_PROTOCOL
;
1109 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1113 read_lock_bh(&ip_set_ref_lock
);
1114 if (set
->ref
!= 0) {
1115 ret
= -IPSET_ERR_REFERENCED
;
1119 name2
= nla_data(attr
[IPSET_ATTR_SETNAME2
]);
1120 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1121 s
= ip_set(inst
, i
);
1122 if (s
&& STRNCMP(s
->name
, name2
)) {
1123 ret
= -IPSET_ERR_EXIST_SETNAME2
;
1127 strncpy(set
->name
, name2
, IPSET_MAXNAMELEN
);
1130 read_unlock_bh(&ip_set_ref_lock
);
1134 /* Swap two sets so that name/index points to the other.
1135 * References and set names are also swapped.
1137 * The commands are serialized by the nfnl mutex and references are
1138 * protected by the ip_set_ref_lock. The kernel interfaces
1139 * do not hold the mutex but the pointer settings are atomic
1140 * so the ip_set_list always contains valid pointers to the sets.
1144 ip_set_swap(struct sock
*ctnl
, struct sk_buff
*skb
,
1145 const struct nlmsghdr
*nlh
,
1146 const struct nlattr
* const attr
[])
1148 struct ip_set_net
*inst
= ip_set_pernet(sock_net(ctnl
));
1149 struct ip_set
*from
, *to
;
1150 ip_set_id_t from_id
, to_id
;
1151 char from_name
[IPSET_MAXNAMELEN
];
1153 if (unlikely(protocol_failed(attr
) ||
1154 !attr
[IPSET_ATTR_SETNAME
] ||
1155 !attr
[IPSET_ATTR_SETNAME2
]))
1156 return -IPSET_ERR_PROTOCOL
;
1158 from
= find_set_and_id(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]),
1163 to
= find_set_and_id(inst
, nla_data(attr
[IPSET_ATTR_SETNAME2
]),
1166 return -IPSET_ERR_EXIST_SETNAME2
;
1168 /* Features must not change.
1169 * Not an artifical restriction anymore, as we must prevent
1170 * possible loops created by swapping in setlist type of sets.
1172 if (!(from
->type
->features
== to
->type
->features
&&
1173 from
->family
== to
->family
))
1174 return -IPSET_ERR_TYPE_MISMATCH
;
1176 strncpy(from_name
, from
->name
, IPSET_MAXNAMELEN
);
1177 strncpy(from
->name
, to
->name
, IPSET_MAXNAMELEN
);
1178 strncpy(to
->name
, from_name
, IPSET_MAXNAMELEN
);
1180 write_lock_bh(&ip_set_ref_lock
);
1181 swap(from
->ref
, to
->ref
);
1182 ip_set(inst
, from_id
) = to
;
1183 ip_set(inst
, to_id
) = from
;
1184 write_unlock_bh(&ip_set_ref_lock
);
1189 /* List/save set data */
1196 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1197 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1200 ip_set_dump_done(struct netlink_callback
*cb
)
1202 if (cb
->args
[IPSET_CB_ARG0
]) {
1203 struct ip_set_net
*inst
=
1204 (struct ip_set_net
*)cb
->args
[IPSET_CB_NET
];
1205 ip_set_id_t index
= (ip_set_id_t
)cb
->args
[IPSET_CB_INDEX
];
1206 struct ip_set
*set
= ip_set(inst
, index
);
1208 if (set
->variant
->uref
)
1209 set
->variant
->uref(set
, cb
, false);
1210 pr_debug("release set %s\n", set
->name
);
1211 __ip_set_put_byindex(inst
, index
);
1217 dump_attrs(struct nlmsghdr
*nlh
)
1219 const struct nlattr
*attr
;
1222 pr_debug("dump nlmsg\n");
1223 nlmsg_for_each_attr(attr
, nlh
, sizeof(struct nfgenmsg
), rem
) {
1224 pr_debug("type: %u, len %u\n", nla_type(attr
), attr
->nla_len
);
1229 dump_init(struct netlink_callback
*cb
, struct ip_set_net
*inst
)
1231 struct nlmsghdr
*nlh
= nlmsg_hdr(cb
->skb
);
1232 int min_len
= nlmsg_total_size(sizeof(struct nfgenmsg
));
1233 struct nlattr
*cda
[IPSET_ATTR_CMD_MAX
+ 1];
1234 struct nlattr
*attr
= (void *)nlh
+ min_len
;
1238 /* Second pass, so parser can't fail */
1239 nla_parse(cda
, IPSET_ATTR_CMD_MAX
,
1240 attr
, nlh
->nlmsg_len
- min_len
, ip_set_setname_policy
);
1242 if (cda
[IPSET_ATTR_SETNAME
]) {
1245 set
= find_set_and_id(inst
, nla_data(cda
[IPSET_ATTR_SETNAME
]),
1250 dump_type
= DUMP_ONE
;
1251 cb
->args
[IPSET_CB_INDEX
] = index
;
1253 dump_type
= DUMP_ALL
;
1256 if (cda
[IPSET_ATTR_FLAGS
]) {
1257 u32 f
= ip_set_get_h32(cda
[IPSET_ATTR_FLAGS
]);
1259 dump_type
|= (f
<< 16);
1261 cb
->args
[IPSET_CB_NET
] = (unsigned long)inst
;
1262 cb
->args
[IPSET_CB_DUMP
] = dump_type
;
1268 ip_set_dump_start(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1270 ip_set_id_t index
= IPSET_INVALID_ID
, max
;
1271 struct ip_set
*set
= NULL
;
1272 struct nlmsghdr
*nlh
= NULL
;
1273 unsigned int flags
= NETLINK_CB(cb
->skb
).portid
? NLM_F_MULTI
: 0;
1274 struct ip_set_net
*inst
= ip_set_pernet(sock_net(skb
->sk
));
1275 u32 dump_type
, dump_flags
;
1279 if (!cb
->args
[IPSET_CB_DUMP
]) {
1280 ret
= dump_init(cb
, inst
);
1282 nlh
= nlmsg_hdr(cb
->skb
);
1283 /* We have to create and send the error message
1286 if (nlh
->nlmsg_flags
& NLM_F_ACK
)
1287 netlink_ack(cb
->skb
, nlh
, ret
);
1292 if (cb
->args
[IPSET_CB_INDEX
] >= inst
->ip_set_max
)
1295 dump_type
= DUMP_TYPE(cb
->args
[IPSET_CB_DUMP
]);
1296 dump_flags
= DUMP_FLAGS(cb
->args
[IPSET_CB_DUMP
]);
1297 max
= dump_type
== DUMP_ONE
? cb
->args
[IPSET_CB_INDEX
] + 1
1300 pr_debug("dump type, flag: %u %u index: %ld\n",
1301 dump_type
, dump_flags
, cb
->args
[IPSET_CB_INDEX
]);
1302 for (; cb
->args
[IPSET_CB_INDEX
] < max
; cb
->args
[IPSET_CB_INDEX
]++) {
1303 index
= (ip_set_id_t
)cb
->args
[IPSET_CB_INDEX
];
1304 write_lock_bh(&ip_set_ref_lock
);
1305 set
= ip_set(inst
, index
);
1306 is_destroyed
= inst
->is_destroyed
;
1307 if (!set
|| is_destroyed
) {
1308 write_unlock_bh(&ip_set_ref_lock
);
1309 if (dump_type
== DUMP_ONE
) {
1314 /* All sets are just being destroyed */
1320 /* When dumping all sets, we must dump "sorted"
1321 * so that lists (unions of sets) are dumped last.
1323 if (dump_type
!= DUMP_ONE
&&
1324 ((dump_type
== DUMP_ALL
) ==
1325 !!(set
->type
->features
& IPSET_DUMP_LAST
))) {
1326 write_unlock_bh(&ip_set_ref_lock
);
1329 pr_debug("List set: %s\n", set
->name
);
1330 if (!cb
->args
[IPSET_CB_ARG0
]) {
1331 /* Start listing: make sure set won't be destroyed */
1332 pr_debug("reference set\n");
1335 write_unlock_bh(&ip_set_ref_lock
);
1336 nlh
= start_msg(skb
, NETLINK_CB(cb
->skb
).portid
,
1337 cb
->nlh
->nlmsg_seq
, flags
,
1341 goto release_refcount
;
1343 if (nla_put_u8(skb
, IPSET_ATTR_PROTOCOL
, IPSET_PROTOCOL
) ||
1344 nla_put_string(skb
, IPSET_ATTR_SETNAME
, set
->name
))
1345 goto nla_put_failure
;
1346 if (dump_flags
& IPSET_FLAG_LIST_SETNAME
)
1348 switch (cb
->args
[IPSET_CB_ARG0
]) {
1350 /* Core header data */
1351 if (nla_put_string(skb
, IPSET_ATTR_TYPENAME
,
1353 nla_put_u8(skb
, IPSET_ATTR_FAMILY
,
1355 nla_put_u8(skb
, IPSET_ATTR_REVISION
,
1357 goto nla_put_failure
;
1358 ret
= set
->variant
->head(set
, skb
);
1360 goto release_refcount
;
1361 if (dump_flags
& IPSET_FLAG_LIST_HEADER
)
1363 if (set
->variant
->uref
)
1364 set
->variant
->uref(set
, cb
, true);
1365 /* Fall through and add elements */
1368 ret
= set
->variant
->list(set
, skb
, cb
);
1369 rcu_read_unlock_bh();
1370 if (!cb
->args
[IPSET_CB_ARG0
])
1371 /* Set is done, proceed with next one */
1373 goto release_refcount
;
1376 /* If we dump all sets, continue with dumping last ones */
1377 if (dump_type
== DUMP_ALL
) {
1378 dump_type
= DUMP_LAST
;
1379 cb
->args
[IPSET_CB_DUMP
] = dump_type
| (dump_flags
<< 16);
1380 cb
->args
[IPSET_CB_INDEX
] = 0;
1381 if (set
&& set
->variant
->uref
)
1382 set
->variant
->uref(set
, cb
, false);
1390 if (dump_type
== DUMP_ONE
)
1391 cb
->args
[IPSET_CB_INDEX
] = IPSET_INVALID_ID
;
1393 cb
->args
[IPSET_CB_INDEX
]++;
1395 /* If there was an error or set is done, release set */
1396 if (ret
|| !cb
->args
[IPSET_CB_ARG0
]) {
1397 set
= ip_set(inst
, index
);
1398 if (set
->variant
->uref
)
1399 set
->variant
->uref(set
, cb
, false);
1400 pr_debug("release set %s\n", set
->name
);
1401 __ip_set_put_byindex(inst
, index
);
1402 cb
->args
[IPSET_CB_ARG0
] = 0;
1406 nlmsg_end(skb
, nlh
);
1407 pr_debug("nlmsg_len: %u\n", nlh
->nlmsg_len
);
1411 return ret
< 0 ? ret
: skb
->len
;
1415 ip_set_dump(struct sock
*ctnl
, struct sk_buff
*skb
,
1416 const struct nlmsghdr
*nlh
,
1417 const struct nlattr
* const attr
[])
1419 if (unlikely(protocol_failed(attr
)))
1420 return -IPSET_ERR_PROTOCOL
;
1423 struct netlink_dump_control c
= {
1424 .dump
= ip_set_dump_start
,
1425 .done
= ip_set_dump_done
,
1427 return netlink_dump_start(ctnl
, skb
, nlh
, &c
);
1431 /* Add, del and test */
1433 static const struct nla_policy ip_set_adt_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1434 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1435 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
1436 .len
= IPSET_MAXNAMELEN
- 1 },
1437 [IPSET_ATTR_LINENO
] = { .type
= NLA_U32
},
1438 [IPSET_ATTR_DATA
] = { .type
= NLA_NESTED
},
1439 [IPSET_ATTR_ADT
] = { .type
= NLA_NESTED
},
1443 call_ad(struct sock
*ctnl
, struct sk_buff
*skb
, struct ip_set
*set
,
1444 struct nlattr
*tb
[], enum ipset_adt adt
,
1445 u32 flags
, bool use_lineno
)
1449 bool eexist
= flags
& IPSET_FLAG_EXIST
, retried
= false;
1452 spin_lock_bh(&set
->lock
);
1453 ret
= set
->variant
->uadt(set
, tb
, adt
, &lineno
, flags
, retried
);
1454 spin_unlock_bh(&set
->lock
);
1456 } while (ret
== -EAGAIN
&&
1457 set
->variant
->resize
&&
1458 (ret
= set
->variant
->resize(set
, retried
)) == 0);
1460 if (!ret
|| (ret
== -IPSET_ERR_EXIST
&& eexist
))
1462 if (lineno
&& use_lineno
) {
1463 /* Error in restore/batch mode: send back lineno */
1464 struct nlmsghdr
*rep
, *nlh
= nlmsg_hdr(skb
);
1465 struct sk_buff
*skb2
;
1466 struct nlmsgerr
*errmsg
;
1467 size_t payload
= min(SIZE_MAX
,
1468 sizeof(*errmsg
) + nlmsg_len(nlh
));
1469 int min_len
= nlmsg_total_size(sizeof(struct nfgenmsg
));
1470 struct nlattr
*cda
[IPSET_ATTR_CMD_MAX
+ 1];
1471 struct nlattr
*cmdattr
;
1474 skb2
= nlmsg_new(payload
, GFP_KERNEL
);
1477 rep
= __nlmsg_put(skb2
, NETLINK_CB(skb
).portid
,
1478 nlh
->nlmsg_seq
, NLMSG_ERROR
, payload
, 0);
1479 errmsg
= nlmsg_data(rep
);
1480 errmsg
->error
= ret
;
1481 memcpy(&errmsg
->msg
, nlh
, nlh
->nlmsg_len
);
1482 cmdattr
= (void *)&errmsg
->msg
+ min_len
;
1484 nla_parse(cda
, IPSET_ATTR_CMD_MAX
,
1485 cmdattr
, nlh
->nlmsg_len
- min_len
,
1488 errline
= nla_data(cda
[IPSET_ATTR_LINENO
]);
1492 netlink_unicast(ctnl
, skb2
, NETLINK_CB(skb
).portid
,
1494 /* Signal netlink not to send its ACK/errmsg. */
1502 ip_set_uadd(struct sock
*ctnl
, struct sk_buff
*skb
,
1503 const struct nlmsghdr
*nlh
,
1504 const struct nlattr
* const attr
[])
1506 struct ip_set_net
*inst
= ip_set_pernet(sock_net(ctnl
));
1508 struct nlattr
*tb
[IPSET_ATTR_ADT_MAX
+ 1] = {};
1509 const struct nlattr
*nla
;
1510 u32 flags
= flag_exist(nlh
);
1514 if (unlikely(protocol_failed(attr
) ||
1515 !attr
[IPSET_ATTR_SETNAME
] ||
1516 !((attr
[IPSET_ATTR_DATA
] != NULL
) ^
1517 (attr
[IPSET_ATTR_ADT
] != NULL
)) ||
1518 (attr
[IPSET_ATTR_DATA
] &&
1519 !flag_nested(attr
[IPSET_ATTR_DATA
])) ||
1520 (attr
[IPSET_ATTR_ADT
] &&
1521 (!flag_nested(attr
[IPSET_ATTR_ADT
]) ||
1522 !attr
[IPSET_ATTR_LINENO
]))))
1523 return -IPSET_ERR_PROTOCOL
;
1525 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1529 use_lineno
= !!attr
[IPSET_ATTR_LINENO
];
1530 if (attr
[IPSET_ATTR_DATA
]) {
1531 if (nla_parse_nested(tb
, IPSET_ATTR_ADT_MAX
,
1532 attr
[IPSET_ATTR_DATA
],
1533 set
->type
->adt_policy
))
1534 return -IPSET_ERR_PROTOCOL
;
1535 ret
= call_ad(ctnl
, skb
, set
, tb
, IPSET_ADD
, flags
,
1540 nla_for_each_nested(nla
, attr
[IPSET_ATTR_ADT
], nla_rem
) {
1541 memset(tb
, 0, sizeof(tb
));
1542 if (nla_type(nla
) != IPSET_ATTR_DATA
||
1543 !flag_nested(nla
) ||
1544 nla_parse_nested(tb
, IPSET_ATTR_ADT_MAX
, nla
,
1545 set
->type
->adt_policy
))
1546 return -IPSET_ERR_PROTOCOL
;
1547 ret
= call_ad(ctnl
, skb
, set
, tb
, IPSET_ADD
,
1557 ip_set_udel(struct sock
*ctnl
, struct sk_buff
*skb
,
1558 const struct nlmsghdr
*nlh
,
1559 const struct nlattr
* const attr
[])
1561 struct ip_set_net
*inst
= ip_set_pernet(sock_net(ctnl
));
1563 struct nlattr
*tb
[IPSET_ATTR_ADT_MAX
+ 1] = {};
1564 const struct nlattr
*nla
;
1565 u32 flags
= flag_exist(nlh
);
1569 if (unlikely(protocol_failed(attr
) ||
1570 !attr
[IPSET_ATTR_SETNAME
] ||
1571 !((attr
[IPSET_ATTR_DATA
] != NULL
) ^
1572 (attr
[IPSET_ATTR_ADT
] != NULL
)) ||
1573 (attr
[IPSET_ATTR_DATA
] &&
1574 !flag_nested(attr
[IPSET_ATTR_DATA
])) ||
1575 (attr
[IPSET_ATTR_ADT
] &&
1576 (!flag_nested(attr
[IPSET_ATTR_ADT
]) ||
1577 !attr
[IPSET_ATTR_LINENO
]))))
1578 return -IPSET_ERR_PROTOCOL
;
1580 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1584 use_lineno
= !!attr
[IPSET_ATTR_LINENO
];
1585 if (attr
[IPSET_ATTR_DATA
]) {
1586 if (nla_parse_nested(tb
, IPSET_ATTR_ADT_MAX
,
1587 attr
[IPSET_ATTR_DATA
],
1588 set
->type
->adt_policy
))
1589 return -IPSET_ERR_PROTOCOL
;
1590 ret
= call_ad(ctnl
, skb
, set
, tb
, IPSET_DEL
, flags
,
1595 nla_for_each_nested(nla
, attr
[IPSET_ATTR_ADT
], nla_rem
) {
1596 memset(tb
, 0, sizeof(*tb
));
1597 if (nla_type(nla
) != IPSET_ATTR_DATA
||
1598 !flag_nested(nla
) ||
1599 nla_parse_nested(tb
, IPSET_ATTR_ADT_MAX
, nla
,
1600 set
->type
->adt_policy
))
1601 return -IPSET_ERR_PROTOCOL
;
1602 ret
= call_ad(ctnl
, skb
, set
, tb
, IPSET_DEL
,
1612 ip_set_utest(struct sock
*ctnl
, struct sk_buff
*skb
,
1613 const struct nlmsghdr
*nlh
,
1614 const struct nlattr
* const attr
[])
1616 struct ip_set_net
*inst
= ip_set_pernet(sock_net(ctnl
));
1618 struct nlattr
*tb
[IPSET_ATTR_ADT_MAX
+ 1] = {};
1621 if (unlikely(protocol_failed(attr
) ||
1622 !attr
[IPSET_ATTR_SETNAME
] ||
1623 !attr
[IPSET_ATTR_DATA
] ||
1624 !flag_nested(attr
[IPSET_ATTR_DATA
])))
1625 return -IPSET_ERR_PROTOCOL
;
1627 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1631 if (nla_parse_nested(tb
, IPSET_ATTR_ADT_MAX
, attr
[IPSET_ATTR_DATA
],
1632 set
->type
->adt_policy
))
1633 return -IPSET_ERR_PROTOCOL
;
1636 ret
= set
->variant
->uadt(set
, tb
, IPSET_TEST
, NULL
, 0, 0);
1637 rcu_read_unlock_bh();
1638 /* Userspace can't trigger element to be re-added */
1642 return ret
> 0 ? 0 : -IPSET_ERR_EXIST
;
1645 /* Get headed data of a set */
1648 ip_set_header(struct sock
*ctnl
, struct sk_buff
*skb
,
1649 const struct nlmsghdr
*nlh
,
1650 const struct nlattr
* const attr
[])
1652 struct ip_set_net
*inst
= ip_set_pernet(sock_net(ctnl
));
1653 const struct ip_set
*set
;
1654 struct sk_buff
*skb2
;
1655 struct nlmsghdr
*nlh2
;
1658 if (unlikely(protocol_failed(attr
) ||
1659 !attr
[IPSET_ATTR_SETNAME
]))
1660 return -IPSET_ERR_PROTOCOL
;
1662 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1666 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
1670 nlh2
= start_msg(skb2
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
1674 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, IPSET_PROTOCOL
) ||
1675 nla_put_string(skb2
, IPSET_ATTR_SETNAME
, set
->name
) ||
1676 nla_put_string(skb2
, IPSET_ATTR_TYPENAME
, set
->type
->name
) ||
1677 nla_put_u8(skb2
, IPSET_ATTR_FAMILY
, set
->family
) ||
1678 nla_put_u8(skb2
, IPSET_ATTR_REVISION
, set
->revision
))
1679 goto nla_put_failure
;
1680 nlmsg_end(skb2
, nlh2
);
1682 ret
= netlink_unicast(ctnl
, skb2
, NETLINK_CB(skb
).portid
, MSG_DONTWAIT
);
1689 nlmsg_cancel(skb2
, nlh2
);
1697 static const struct nla_policy ip_set_type_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1698 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1699 [IPSET_ATTR_TYPENAME
] = { .type
= NLA_NUL_STRING
,
1700 .len
= IPSET_MAXNAMELEN
- 1 },
1701 [IPSET_ATTR_FAMILY
] = { .type
= NLA_U8
},
1705 ip_set_type(struct sock
*ctnl
, struct sk_buff
*skb
,
1706 const struct nlmsghdr
*nlh
,
1707 const struct nlattr
* const attr
[])
1709 struct sk_buff
*skb2
;
1710 struct nlmsghdr
*nlh2
;
1711 u8 family
, min
, max
;
1712 const char *typename
;
1715 if (unlikely(protocol_failed(attr
) ||
1716 !attr
[IPSET_ATTR_TYPENAME
] ||
1717 !attr
[IPSET_ATTR_FAMILY
]))
1718 return -IPSET_ERR_PROTOCOL
;
1720 family
= nla_get_u8(attr
[IPSET_ATTR_FAMILY
]);
1721 typename
= nla_data(attr
[IPSET_ATTR_TYPENAME
]);
1722 ret
= find_set_type_minmax(typename
, family
, &min
, &max
);
1726 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
1730 nlh2
= start_msg(skb2
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
1734 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, IPSET_PROTOCOL
) ||
1735 nla_put_string(skb2
, IPSET_ATTR_TYPENAME
, typename
) ||
1736 nla_put_u8(skb2
, IPSET_ATTR_FAMILY
, family
) ||
1737 nla_put_u8(skb2
, IPSET_ATTR_REVISION
, max
) ||
1738 nla_put_u8(skb2
, IPSET_ATTR_REVISION_MIN
, min
))
1739 goto nla_put_failure
;
1740 nlmsg_end(skb2
, nlh2
);
1742 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2
->nlmsg_len
);
1743 ret
= netlink_unicast(ctnl
, skb2
, NETLINK_CB(skb
).portid
, MSG_DONTWAIT
);
1750 nlmsg_cancel(skb2
, nlh2
);
1756 /* Get protocol version */
1758 static const struct nla_policy
1759 ip_set_protocol_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1760 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1764 ip_set_protocol(struct sock
*ctnl
, struct sk_buff
*skb
,
1765 const struct nlmsghdr
*nlh
,
1766 const struct nlattr
* const attr
[])
1768 struct sk_buff
*skb2
;
1769 struct nlmsghdr
*nlh2
;
1772 if (unlikely(!attr
[IPSET_ATTR_PROTOCOL
]))
1773 return -IPSET_ERR_PROTOCOL
;
1775 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
1779 nlh2
= start_msg(skb2
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
1780 IPSET_CMD_PROTOCOL
);
1783 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, IPSET_PROTOCOL
))
1784 goto nla_put_failure
;
1785 nlmsg_end(skb2
, nlh2
);
1787 ret
= netlink_unicast(ctnl
, skb2
, NETLINK_CB(skb
).portid
, MSG_DONTWAIT
);
1794 nlmsg_cancel(skb2
, nlh2
);
1800 static const struct nfnl_callback ip_set_netlink_subsys_cb
[IPSET_MSG_MAX
] = {
1801 [IPSET_CMD_NONE
] = {
1802 .call
= ip_set_none
,
1803 .attr_count
= IPSET_ATTR_CMD_MAX
,
1805 [IPSET_CMD_CREATE
] = {
1806 .call
= ip_set_create
,
1807 .attr_count
= IPSET_ATTR_CMD_MAX
,
1808 .policy
= ip_set_create_policy
,
1810 [IPSET_CMD_DESTROY
] = {
1811 .call
= ip_set_destroy
,
1812 .attr_count
= IPSET_ATTR_CMD_MAX
,
1813 .policy
= ip_set_setname_policy
,
1815 [IPSET_CMD_FLUSH
] = {
1816 .call
= ip_set_flush
,
1817 .attr_count
= IPSET_ATTR_CMD_MAX
,
1818 .policy
= ip_set_setname_policy
,
1820 [IPSET_CMD_RENAME
] = {
1821 .call
= ip_set_rename
,
1822 .attr_count
= IPSET_ATTR_CMD_MAX
,
1823 .policy
= ip_set_setname2_policy
,
1825 [IPSET_CMD_SWAP
] = {
1826 .call
= ip_set_swap
,
1827 .attr_count
= IPSET_ATTR_CMD_MAX
,
1828 .policy
= ip_set_setname2_policy
,
1830 [IPSET_CMD_LIST
] = {
1831 .call
= ip_set_dump
,
1832 .attr_count
= IPSET_ATTR_CMD_MAX
,
1833 .policy
= ip_set_setname_policy
,
1835 [IPSET_CMD_SAVE
] = {
1836 .call
= ip_set_dump
,
1837 .attr_count
= IPSET_ATTR_CMD_MAX
,
1838 .policy
= ip_set_setname_policy
,
1841 .call
= ip_set_uadd
,
1842 .attr_count
= IPSET_ATTR_CMD_MAX
,
1843 .policy
= ip_set_adt_policy
,
1846 .call
= ip_set_udel
,
1847 .attr_count
= IPSET_ATTR_CMD_MAX
,
1848 .policy
= ip_set_adt_policy
,
1850 [IPSET_CMD_TEST
] = {
1851 .call
= ip_set_utest
,
1852 .attr_count
= IPSET_ATTR_CMD_MAX
,
1853 .policy
= ip_set_adt_policy
,
1855 [IPSET_CMD_HEADER
] = {
1856 .call
= ip_set_header
,
1857 .attr_count
= IPSET_ATTR_CMD_MAX
,
1858 .policy
= ip_set_setname_policy
,
1860 [IPSET_CMD_TYPE
] = {
1861 .call
= ip_set_type
,
1862 .attr_count
= IPSET_ATTR_CMD_MAX
,
1863 .policy
= ip_set_type_policy
,
1865 [IPSET_CMD_PROTOCOL
] = {
1866 .call
= ip_set_protocol
,
1867 .attr_count
= IPSET_ATTR_CMD_MAX
,
1868 .policy
= ip_set_protocol_policy
,
1872 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly
= {
1874 .subsys_id
= NFNL_SUBSYS_IPSET
,
1875 .cb_count
= IPSET_MSG_MAX
,
1876 .cb
= ip_set_netlink_subsys_cb
,
1879 /* Interface to iptables/ip6tables */
1882 ip_set_sockfn_get(struct sock
*sk
, int optval
, void __user
*user
, int *len
)
1886 int copylen
= *len
, ret
= 0;
1887 struct net
*net
= sock_net(sk
);
1888 struct ip_set_net
*inst
= ip_set_pernet(net
);
1890 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1892 if (optval
!= SO_IP_SET
)
1894 if (*len
< sizeof(unsigned int))
1897 data
= vmalloc(*len
);
1900 if (copy_from_user(data
, user
, *len
) != 0) {
1904 op
= (unsigned int *)data
;
1906 if (*op
< IP_SET_OP_VERSION
) {
1907 /* Check the version at the beginning of operations */
1908 struct ip_set_req_version
*req_version
= data
;
1910 if (*len
< sizeof(struct ip_set_req_version
)) {
1915 if (req_version
->version
!= IPSET_PROTOCOL
) {
1922 case IP_SET_OP_VERSION
: {
1923 struct ip_set_req_version
*req_version
= data
;
1925 if (*len
!= sizeof(struct ip_set_req_version
)) {
1930 req_version
->version
= IPSET_PROTOCOL
;
1931 ret
= copy_to_user(user
, req_version
,
1932 sizeof(struct ip_set_req_version
));
1935 case IP_SET_OP_GET_BYNAME
: {
1936 struct ip_set_req_get_set
*req_get
= data
;
1939 if (*len
!= sizeof(struct ip_set_req_get_set
)) {
1943 req_get
->set
.name
[IPSET_MAXNAMELEN
- 1] = '\0';
1944 nfnl_lock(NFNL_SUBSYS_IPSET
);
1945 find_set_and_id(inst
, req_get
->set
.name
, &id
);
1946 req_get
->set
.index
= id
;
1947 nfnl_unlock(NFNL_SUBSYS_IPSET
);
1950 case IP_SET_OP_GET_FNAME
: {
1951 struct ip_set_req_get_set_family
*req_get
= data
;
1954 if (*len
!= sizeof(struct ip_set_req_get_set_family
)) {
1958 req_get
->set
.name
[IPSET_MAXNAMELEN
- 1] = '\0';
1959 nfnl_lock(NFNL_SUBSYS_IPSET
);
1960 find_set_and_id(inst
, req_get
->set
.name
, &id
);
1961 req_get
->set
.index
= id
;
1962 if (id
!= IPSET_INVALID_ID
)
1963 req_get
->family
= ip_set(inst
, id
)->family
;
1964 nfnl_unlock(NFNL_SUBSYS_IPSET
);
1967 case IP_SET_OP_GET_BYINDEX
: {
1968 struct ip_set_req_get_set
*req_get
= data
;
1971 if (*len
!= sizeof(struct ip_set_req_get_set
) ||
1972 req_get
->set
.index
>= inst
->ip_set_max
) {
1976 nfnl_lock(NFNL_SUBSYS_IPSET
);
1977 set
= ip_set(inst
, req_get
->set
.index
);
1978 strncpy(req_get
->set
.name
, set
? set
->name
: "",
1980 nfnl_unlock(NFNL_SUBSYS_IPSET
);
1986 } /* end of switch(op) */
1989 ret
= copy_to_user(user
, data
, copylen
);
1998 static struct nf_sockopt_ops so_set __read_mostly
= {
2000 .get_optmin
= SO_IP_SET
,
2001 .get_optmax
= SO_IP_SET
+ 1,
2002 .get
= &ip_set_sockfn_get
,
2003 .owner
= THIS_MODULE
,
2006 static int __net_init
2007 ip_set_net_init(struct net
*net
)
2009 struct ip_set_net
*inst
= ip_set_pernet(net
);
2010 struct ip_set
**list
;
2012 inst
->ip_set_max
= max_sets
? max_sets
: CONFIG_IP_SET_MAX
;
2013 if (inst
->ip_set_max
>= IPSET_INVALID_ID
)
2014 inst
->ip_set_max
= IPSET_INVALID_ID
- 1;
2016 list
= kcalloc(inst
->ip_set_max
, sizeof(struct ip_set
*), GFP_KERNEL
);
2019 inst
->is_deleted
= false;
2020 inst
->is_destroyed
= false;
2021 rcu_assign_pointer(inst
->ip_set_list
, list
);
2025 static void __net_exit
2026 ip_set_net_exit(struct net
*net
)
2028 struct ip_set_net
*inst
= ip_set_pernet(net
);
2030 struct ip_set
*set
= NULL
;
2033 inst
->is_deleted
= true; /* flag for ip_set_nfnl_put */
2035 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
2036 set
= ip_set(inst
, i
);
2038 ip_set(inst
, i
) = NULL
;
2039 ip_set_destroy_set(set
);
2042 kfree(rcu_dereference_protected(inst
->ip_set_list
, 1));
2045 static struct pernet_operations ip_set_net_ops
= {
2046 .init
= ip_set_net_init
,
2047 .exit
= ip_set_net_exit
,
2048 .id
= &ip_set_net_id
,
2049 .size
= sizeof(struct ip_set_net
)
2055 int ret
= nfnetlink_subsys_register(&ip_set_netlink_subsys
);
2058 pr_err("ip_set: cannot register with nfnetlink.\n");
2061 ret
= nf_register_sockopt(&so_set
);
2063 pr_err("SO_SET registry failed: %d\n", ret
);
2064 nfnetlink_subsys_unregister(&ip_set_netlink_subsys
);
2067 ret
= register_pernet_subsys(&ip_set_net_ops
);
2069 pr_err("ip_set: cannot register pernet_subsys.\n");
2070 nf_unregister_sockopt(&so_set
);
2071 nfnetlink_subsys_unregister(&ip_set_netlink_subsys
);
2074 pr_info("ip_set: protocol %u\n", IPSET_PROTOCOL
);
2081 unregister_pernet_subsys(&ip_set_net_ops
);
2082 nf_unregister_sockopt(&so_set
);
2083 nfnetlink_subsys_unregister(&ip_set_netlink_subsys
);
2084 pr_debug("these are the famous last words\n");
2087 module_init(ip_set_init
);
2088 module_exit(ip_set_fini
);