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 /* Add, del and test set entries from kernel.
502 * The set behind the index must exist and must be referenced
503 * so it can't be destroyed (or changed) under our foot.
506 static inline struct ip_set
*
507 ip_set_rcu_get(struct net
*net
, ip_set_id_t index
)
510 struct ip_set_net
*inst
= ip_set_pernet(net
);
513 /* ip_set_list itself needs to be protected */
514 set
= rcu_dereference(inst
->ip_set_list
)[index
];
521 ip_set_test(ip_set_id_t index
, const struct sk_buff
*skb
,
522 const struct xt_action_param
*par
, struct ip_set_adt_opt
*opt
)
524 struct ip_set
*set
= ip_set_rcu_get(par
->net
, index
);
528 pr_debug("set %s, index %u\n", set
->name
, index
);
530 if (opt
->dim
< set
->type
->dimension
||
531 !(opt
->family
== set
->family
|| set
->family
== NFPROTO_UNSPEC
))
535 ret
= set
->variant
->kadt(set
, skb
, par
, IPSET_TEST
, opt
);
536 rcu_read_unlock_bh();
538 if (ret
== -EAGAIN
) {
539 /* Type requests element to be completed */
540 pr_debug("element must be completed, ADD is triggered\n");
541 spin_lock_bh(&set
->lock
);
542 set
->variant
->kadt(set
, skb
, par
, IPSET_ADD
, opt
);
543 spin_unlock_bh(&set
->lock
);
546 /* --return-nomatch: invert matched element */
547 if ((opt
->cmdflags
& IPSET_FLAG_RETURN_NOMATCH
) &&
548 (set
->type
->features
& IPSET_TYPE_NOMATCH
) &&
549 (ret
> 0 || ret
== -ENOTEMPTY
))
553 /* Convert error codes to nomatch */
554 return (ret
< 0 ? 0 : ret
);
556 EXPORT_SYMBOL_GPL(ip_set_test
);
559 ip_set_add(ip_set_id_t index
, const struct sk_buff
*skb
,
560 const struct xt_action_param
*par
, struct ip_set_adt_opt
*opt
)
562 struct ip_set
*set
= ip_set_rcu_get(par
->net
, index
);
566 pr_debug("set %s, index %u\n", set
->name
, index
);
568 if (opt
->dim
< set
->type
->dimension
||
569 !(opt
->family
== set
->family
|| set
->family
== NFPROTO_UNSPEC
))
570 return -IPSET_ERR_TYPE_MISMATCH
;
572 spin_lock_bh(&set
->lock
);
573 ret
= set
->variant
->kadt(set
, skb
, par
, IPSET_ADD
, opt
);
574 spin_unlock_bh(&set
->lock
);
578 EXPORT_SYMBOL_GPL(ip_set_add
);
581 ip_set_del(ip_set_id_t index
, const struct sk_buff
*skb
,
582 const struct xt_action_param
*par
, struct ip_set_adt_opt
*opt
)
584 struct ip_set
*set
= ip_set_rcu_get(par
->net
, index
);
588 pr_debug("set %s, index %u\n", set
->name
, index
);
590 if (opt
->dim
< set
->type
->dimension
||
591 !(opt
->family
== set
->family
|| set
->family
== NFPROTO_UNSPEC
))
592 return -IPSET_ERR_TYPE_MISMATCH
;
594 spin_lock_bh(&set
->lock
);
595 ret
= set
->variant
->kadt(set
, skb
, par
, IPSET_DEL
, opt
);
596 spin_unlock_bh(&set
->lock
);
600 EXPORT_SYMBOL_GPL(ip_set_del
);
602 /* Find set by name, reference it once. The reference makes sure the
603 * thing pointed to, does not go away under our feet.
607 ip_set_get_byname(struct net
*net
, const char *name
, struct ip_set
**set
)
609 ip_set_id_t i
, index
= IPSET_INVALID_ID
;
611 struct ip_set_net
*inst
= ip_set_pernet(net
);
614 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
615 s
= rcu_dereference(inst
->ip_set_list
)[i
];
616 if (s
&& STRNCMP(s
->name
, name
)) {
627 EXPORT_SYMBOL_GPL(ip_set_get_byname
);
629 /* If the given set pointer points to a valid set, decrement
630 * reference count by 1. The caller shall not assume the index
631 * to be valid, after calling this function.
636 __ip_set_put_byindex(struct ip_set_net
*inst
, ip_set_id_t index
)
641 set
= rcu_dereference(inst
->ip_set_list
)[index
];
648 ip_set_put_byindex(struct net
*net
, ip_set_id_t index
)
650 struct ip_set_net
*inst
= ip_set_pernet(net
);
652 __ip_set_put_byindex(inst
, index
);
654 EXPORT_SYMBOL_GPL(ip_set_put_byindex
);
656 /* Get the name of a set behind a set index.
657 * We assume the set is referenced, so it does exist and
658 * can't be destroyed. The set cannot be renamed due to
659 * the referencing either.
663 ip_set_name_byindex(struct net
*net
, ip_set_id_t index
)
665 const struct ip_set
*set
= ip_set_rcu_get(net
, index
);
668 BUG_ON(set
->ref
== 0);
670 /* Referenced, so it's safe */
673 EXPORT_SYMBOL_GPL(ip_set_name_byindex
);
675 /* Routines to call by external subsystems, which do not
676 * call nfnl_lock for us.
679 /* Find set by index, reference it once. The reference makes sure the
680 * thing pointed to, does not go away under our feet.
682 * The nfnl mutex is used in the function.
685 ip_set_nfnl_get_byindex(struct net
*net
, ip_set_id_t index
)
688 struct ip_set_net
*inst
= ip_set_pernet(net
);
690 if (index
>= inst
->ip_set_max
)
691 return IPSET_INVALID_ID
;
693 nfnl_lock(NFNL_SUBSYS_IPSET
);
694 set
= ip_set(inst
, index
);
698 index
= IPSET_INVALID_ID
;
699 nfnl_unlock(NFNL_SUBSYS_IPSET
);
703 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex
);
705 /* If the given set pointer points to a valid set, decrement
706 * reference count by 1. The caller shall not assume the index
707 * to be valid, after calling this function.
709 * The nfnl mutex is used in the function.
712 ip_set_nfnl_put(struct net
*net
, ip_set_id_t index
)
715 struct ip_set_net
*inst
= ip_set_pernet(net
);
717 nfnl_lock(NFNL_SUBSYS_IPSET
);
718 if (!inst
->is_deleted
) { /* already deleted from ip_set_net_exit() */
719 set
= ip_set(inst
, index
);
723 nfnl_unlock(NFNL_SUBSYS_IPSET
);
725 EXPORT_SYMBOL_GPL(ip_set_nfnl_put
);
727 /* Communication protocol with userspace over netlink.
729 * The commands are serialized by the nfnl mutex.
733 protocol_failed(const struct nlattr
* const tb
[])
735 return !tb
[IPSET_ATTR_PROTOCOL
] ||
736 nla_get_u8(tb
[IPSET_ATTR_PROTOCOL
]) != IPSET_PROTOCOL
;
740 flag_exist(const struct nlmsghdr
*nlh
)
742 return nlh
->nlmsg_flags
& NLM_F_EXCL
? 0 : IPSET_FLAG_EXIST
;
745 static struct nlmsghdr
*
746 start_msg(struct sk_buff
*skb
, u32 portid
, u32 seq
, unsigned int flags
,
749 struct nlmsghdr
*nlh
;
750 struct nfgenmsg
*nfmsg
;
752 nlh
= nlmsg_put(skb
, portid
, seq
, cmd
| (NFNL_SUBSYS_IPSET
<< 8),
753 sizeof(*nfmsg
), flags
);
757 nfmsg
= nlmsg_data(nlh
);
758 nfmsg
->nfgen_family
= NFPROTO_IPV4
;
759 nfmsg
->version
= NFNETLINK_V0
;
767 static const struct nla_policy ip_set_create_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
768 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
769 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
770 .len
= IPSET_MAXNAMELEN
- 1 },
771 [IPSET_ATTR_TYPENAME
] = { .type
= NLA_NUL_STRING
,
772 .len
= IPSET_MAXNAMELEN
- 1},
773 [IPSET_ATTR_REVISION
] = { .type
= NLA_U8
},
774 [IPSET_ATTR_FAMILY
] = { .type
= NLA_U8
},
775 [IPSET_ATTR_DATA
] = { .type
= NLA_NESTED
},
778 static struct ip_set
*
779 find_set_and_id(struct ip_set_net
*inst
, const char *name
, ip_set_id_t
*id
)
781 struct ip_set
*set
= NULL
;
784 *id
= IPSET_INVALID_ID
;
785 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
786 set
= ip_set(inst
, i
);
787 if (set
&& STRNCMP(set
->name
, name
)) {
792 return (*id
== IPSET_INVALID_ID
? NULL
: set
);
795 static inline struct ip_set
*
796 find_set(struct ip_set_net
*inst
, const char *name
)
800 return find_set_and_id(inst
, name
, &id
);
804 find_free_id(struct ip_set_net
*inst
, const char *name
, ip_set_id_t
*index
,
810 *index
= IPSET_INVALID_ID
;
811 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
814 if (*index
== IPSET_INVALID_ID
)
816 } else if (STRNCMP(name
, s
->name
)) {
822 if (*index
== IPSET_INVALID_ID
)
823 /* No free slot remained */
824 return -IPSET_ERR_MAX_SETS
;
828 static int ip_set_none(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
829 const struct nlmsghdr
*nlh
,
830 const struct nlattr
* const attr
[])
835 static int ip_set_create(struct net
*net
, struct sock
*ctnl
,
836 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
837 const struct nlattr
* const attr
[])
839 struct ip_set_net
*inst
= ip_set_pernet(net
);
840 struct ip_set
*set
, *clash
= NULL
;
841 ip_set_id_t index
= IPSET_INVALID_ID
;
842 struct nlattr
*tb
[IPSET_ATTR_CREATE_MAX
+ 1] = {};
843 const char *name
, *typename
;
845 u32 flags
= flag_exist(nlh
);
848 if (unlikely(protocol_failed(attr
) ||
849 !attr
[IPSET_ATTR_SETNAME
] ||
850 !attr
[IPSET_ATTR_TYPENAME
] ||
851 !attr
[IPSET_ATTR_REVISION
] ||
852 !attr
[IPSET_ATTR_FAMILY
] ||
853 (attr
[IPSET_ATTR_DATA
] &&
854 !flag_nested(attr
[IPSET_ATTR_DATA
]))))
855 return -IPSET_ERR_PROTOCOL
;
857 name
= nla_data(attr
[IPSET_ATTR_SETNAME
]);
858 typename
= nla_data(attr
[IPSET_ATTR_TYPENAME
]);
859 family
= nla_get_u8(attr
[IPSET_ATTR_FAMILY
]);
860 revision
= nla_get_u8(attr
[IPSET_ATTR_REVISION
]);
861 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
862 name
, typename
, family_name(family
), revision
);
864 /* First, and without any locks, allocate and initialize
865 * a normal base set structure.
867 set
= kzalloc(sizeof(*set
), GFP_KERNEL
);
870 spin_lock_init(&set
->lock
);
871 strlcpy(set
->name
, name
, IPSET_MAXNAMELEN
);
872 set
->family
= family
;
873 set
->revision
= revision
;
875 /* Next, check that we know the type, and take
876 * a reference on the type, to make sure it stays available
877 * while constructing our new set.
879 * After referencing the type, we try to create the type
880 * specific part of the set without holding any locks.
882 ret
= find_set_type_get(typename
, family
, revision
, &set
->type
);
886 /* Without holding any locks, create private part. */
887 if (attr
[IPSET_ATTR_DATA
] &&
888 nla_parse_nested(tb
, IPSET_ATTR_CREATE_MAX
, attr
[IPSET_ATTR_DATA
],
889 set
->type
->create_policy
)) {
890 ret
= -IPSET_ERR_PROTOCOL
;
894 ret
= set
->type
->create(net
, set
, tb
, flags
);
898 /* BTW, ret==0 here. */
900 /* Here, we have a valid, constructed set and we are protected
901 * by the nfnl mutex. Find the first free index in ip_set_list
902 * and check clashing.
904 ret
= find_free_id(inst
, set
->name
, &index
, &clash
);
905 if (ret
== -EEXIST
) {
906 /* If this is the same set and requested, ignore error */
907 if ((flags
& IPSET_FLAG_EXIST
) &&
908 STRNCMP(set
->type
->name
, clash
->type
->name
) &&
909 set
->type
->family
== clash
->type
->family
&&
910 set
->type
->revision_min
== clash
->type
->revision_min
&&
911 set
->type
->revision_max
== clash
->type
->revision_max
&&
912 set
->variant
->same_set(set
, clash
))
915 } else if (ret
== -IPSET_ERR_MAX_SETS
) {
916 struct ip_set
**list
, **tmp
;
917 ip_set_id_t i
= inst
->ip_set_max
+ IP_SET_INC
;
919 if (i
< inst
->ip_set_max
|| i
== IPSET_INVALID_ID
)
923 list
= kcalloc(i
, sizeof(struct ip_set
*), GFP_KERNEL
);
926 /* nfnl mutex is held, both lists are valid */
927 tmp
= ip_set_dereference(inst
->ip_set_list
);
928 memcpy(list
, tmp
, sizeof(struct ip_set
*) * inst
->ip_set_max
);
929 rcu_assign_pointer(inst
->ip_set_list
, list
);
930 /* Make sure all current packets have passed through */
933 index
= inst
->ip_set_max
;
934 inst
->ip_set_max
= i
;
941 /* Finally! Add our shiny new set to the list, and be done. */
942 pr_debug("create: '%s' created with index %u!\n", set
->name
, index
);
943 ip_set(inst
, index
) = set
;
948 set
->variant
->destroy(set
);
950 module_put(set
->type
->me
);
958 static const struct nla_policy
959 ip_set_setname_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
960 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
961 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
962 .len
= IPSET_MAXNAMELEN
- 1 },
966 ip_set_destroy_set(struct ip_set
*set
)
968 pr_debug("set: %s\n", set
->name
);
970 /* Must call it without holding any lock */
971 set
->variant
->destroy(set
);
972 module_put(set
->type
->me
);
976 static int ip_set_destroy(struct net
*net
, struct sock
*ctnl
,
977 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
978 const struct nlattr
* const attr
[])
980 struct ip_set_net
*inst
= ip_set_pernet(net
);
985 if (unlikely(protocol_failed(attr
)))
986 return -IPSET_ERR_PROTOCOL
;
988 /* Must wait for flush to be really finished in list:set */
991 /* Commands are serialized and references are
992 * protected by the ip_set_ref_lock.
993 * External systems (i.e. xt_set) must call
994 * ip_set_put|get_nfnl_* functions, that way we
995 * can safely check references here.
997 * list:set timer can only decrement the reference
998 * counter, so if it's already zero, we can proceed
999 * without holding the lock.
1001 read_lock_bh(&ip_set_ref_lock
);
1002 if (!attr
[IPSET_ATTR_SETNAME
]) {
1003 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1004 s
= ip_set(inst
, i
);
1006 ret
= -IPSET_ERR_BUSY
;
1010 inst
->is_destroyed
= true;
1011 read_unlock_bh(&ip_set_ref_lock
);
1012 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1013 s
= ip_set(inst
, i
);
1015 ip_set(inst
, i
) = NULL
;
1016 ip_set_destroy_set(s
);
1019 /* Modified by ip_set_destroy() only, which is serialized */
1020 inst
->is_destroyed
= false;
1022 s
= find_set_and_id(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]),
1027 } else if (s
->ref
) {
1028 ret
= -IPSET_ERR_BUSY
;
1031 ip_set(inst
, i
) = NULL
;
1032 read_unlock_bh(&ip_set_ref_lock
);
1034 ip_set_destroy_set(s
);
1038 read_unlock_bh(&ip_set_ref_lock
);
1045 ip_set_flush_set(struct ip_set
*set
)
1047 pr_debug("set: %s\n", set
->name
);
1049 spin_lock_bh(&set
->lock
);
1050 set
->variant
->flush(set
);
1051 spin_unlock_bh(&set
->lock
);
1054 static int ip_set_flush(struct net
*net
, 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(net
);
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 },
1093 static int ip_set_rename(struct net
*net
, struct sock
*ctnl
,
1094 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1095 const struct nlattr
* const attr
[])
1097 struct ip_set_net
*inst
= ip_set_pernet(net
);
1098 struct ip_set
*set
, *s
;
1103 if (unlikely(protocol_failed(attr
) ||
1104 !attr
[IPSET_ATTR_SETNAME
] ||
1105 !attr
[IPSET_ATTR_SETNAME2
]))
1106 return -IPSET_ERR_PROTOCOL
;
1108 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1112 read_lock_bh(&ip_set_ref_lock
);
1113 if (set
->ref
!= 0) {
1114 ret
= -IPSET_ERR_REFERENCED
;
1118 name2
= nla_data(attr
[IPSET_ATTR_SETNAME2
]);
1119 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1120 s
= ip_set(inst
, i
);
1121 if (s
&& STRNCMP(s
->name
, name2
)) {
1122 ret
= -IPSET_ERR_EXIST_SETNAME2
;
1126 strncpy(set
->name
, name2
, IPSET_MAXNAMELEN
);
1129 read_unlock_bh(&ip_set_ref_lock
);
1133 /* Swap two sets so that name/index points to the other.
1134 * References and set names are also swapped.
1136 * The commands are serialized by the nfnl mutex and references are
1137 * protected by the ip_set_ref_lock. The kernel interfaces
1138 * do not hold the mutex but the pointer settings are atomic
1139 * so the ip_set_list always contains valid pointers to the sets.
1142 static int ip_set_swap(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1143 const struct nlmsghdr
*nlh
,
1144 const struct nlattr
* const attr
[])
1146 struct ip_set_net
*inst
= ip_set_pernet(net
);
1147 struct ip_set
*from
, *to
;
1148 ip_set_id_t from_id
, to_id
;
1149 char from_name
[IPSET_MAXNAMELEN
];
1151 if (unlikely(protocol_failed(attr
) ||
1152 !attr
[IPSET_ATTR_SETNAME
] ||
1153 !attr
[IPSET_ATTR_SETNAME2
]))
1154 return -IPSET_ERR_PROTOCOL
;
1156 from
= find_set_and_id(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]),
1161 to
= find_set_and_id(inst
, nla_data(attr
[IPSET_ATTR_SETNAME2
]),
1164 return -IPSET_ERR_EXIST_SETNAME2
;
1166 /* Features must not change.
1167 * Not an artifical restriction anymore, as we must prevent
1168 * possible loops created by swapping in setlist type of sets.
1170 if (!(from
->type
->features
== to
->type
->features
&&
1171 from
->family
== to
->family
))
1172 return -IPSET_ERR_TYPE_MISMATCH
;
1174 strncpy(from_name
, from
->name
, IPSET_MAXNAMELEN
);
1175 strncpy(from
->name
, to
->name
, IPSET_MAXNAMELEN
);
1176 strncpy(to
->name
, from_name
, IPSET_MAXNAMELEN
);
1178 write_lock_bh(&ip_set_ref_lock
);
1179 swap(from
->ref
, to
->ref
);
1180 ip_set(inst
, from_id
) = to
;
1181 ip_set(inst
, to_id
) = from
;
1182 write_unlock_bh(&ip_set_ref_lock
);
1187 /* List/save set data */
1194 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1195 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1198 ip_set_dump_done(struct netlink_callback
*cb
)
1200 if (cb
->args
[IPSET_CB_ARG0
]) {
1201 struct ip_set_net
*inst
=
1202 (struct ip_set_net
*)cb
->args
[IPSET_CB_NET
];
1203 ip_set_id_t index
= (ip_set_id_t
)cb
->args
[IPSET_CB_INDEX
];
1204 struct ip_set
*set
= ip_set(inst
, index
);
1206 if (set
->variant
->uref
)
1207 set
->variant
->uref(set
, cb
, false);
1208 pr_debug("release set %s\n", set
->name
);
1209 __ip_set_put_byindex(inst
, index
);
1215 dump_attrs(struct nlmsghdr
*nlh
)
1217 const struct nlattr
*attr
;
1220 pr_debug("dump nlmsg\n");
1221 nlmsg_for_each_attr(attr
, nlh
, sizeof(struct nfgenmsg
), rem
) {
1222 pr_debug("type: %u, len %u\n", nla_type(attr
), attr
->nla_len
);
1227 dump_init(struct netlink_callback
*cb
, struct ip_set_net
*inst
)
1229 struct nlmsghdr
*nlh
= nlmsg_hdr(cb
->skb
);
1230 int min_len
= nlmsg_total_size(sizeof(struct nfgenmsg
));
1231 struct nlattr
*cda
[IPSET_ATTR_CMD_MAX
+ 1];
1232 struct nlattr
*attr
= (void *)nlh
+ min_len
;
1236 /* Second pass, so parser can't fail */
1237 nla_parse(cda
, IPSET_ATTR_CMD_MAX
,
1238 attr
, nlh
->nlmsg_len
- min_len
, ip_set_setname_policy
);
1240 if (cda
[IPSET_ATTR_SETNAME
]) {
1243 set
= find_set_and_id(inst
, nla_data(cda
[IPSET_ATTR_SETNAME
]),
1248 dump_type
= DUMP_ONE
;
1249 cb
->args
[IPSET_CB_INDEX
] = index
;
1251 dump_type
= DUMP_ALL
;
1254 if (cda
[IPSET_ATTR_FLAGS
]) {
1255 u32 f
= ip_set_get_h32(cda
[IPSET_ATTR_FLAGS
]);
1257 dump_type
|= (f
<< 16);
1259 cb
->args
[IPSET_CB_NET
] = (unsigned long)inst
;
1260 cb
->args
[IPSET_CB_DUMP
] = dump_type
;
1266 ip_set_dump_start(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1268 ip_set_id_t index
= IPSET_INVALID_ID
, max
;
1269 struct ip_set
*set
= NULL
;
1270 struct nlmsghdr
*nlh
= NULL
;
1271 unsigned int flags
= NETLINK_CB(cb
->skb
).portid
? NLM_F_MULTI
: 0;
1272 struct ip_set_net
*inst
= ip_set_pernet(sock_net(skb
->sk
));
1273 u32 dump_type
, dump_flags
;
1277 if (!cb
->args
[IPSET_CB_DUMP
]) {
1278 ret
= dump_init(cb
, inst
);
1280 nlh
= nlmsg_hdr(cb
->skb
);
1281 /* We have to create and send the error message
1284 if (nlh
->nlmsg_flags
& NLM_F_ACK
)
1285 netlink_ack(cb
->skb
, nlh
, ret
);
1290 if (cb
->args
[IPSET_CB_INDEX
] >= inst
->ip_set_max
)
1293 dump_type
= DUMP_TYPE(cb
->args
[IPSET_CB_DUMP
]);
1294 dump_flags
= DUMP_FLAGS(cb
->args
[IPSET_CB_DUMP
]);
1295 max
= dump_type
== DUMP_ONE
? cb
->args
[IPSET_CB_INDEX
] + 1
1298 pr_debug("dump type, flag: %u %u index: %ld\n",
1299 dump_type
, dump_flags
, cb
->args
[IPSET_CB_INDEX
]);
1300 for (; cb
->args
[IPSET_CB_INDEX
] < max
; cb
->args
[IPSET_CB_INDEX
]++) {
1301 index
= (ip_set_id_t
)cb
->args
[IPSET_CB_INDEX
];
1302 write_lock_bh(&ip_set_ref_lock
);
1303 set
= ip_set(inst
, index
);
1304 is_destroyed
= inst
->is_destroyed
;
1305 if (!set
|| is_destroyed
) {
1306 write_unlock_bh(&ip_set_ref_lock
);
1307 if (dump_type
== DUMP_ONE
) {
1312 /* All sets are just being destroyed */
1318 /* When dumping all sets, we must dump "sorted"
1319 * so that lists (unions of sets) are dumped last.
1321 if (dump_type
!= DUMP_ONE
&&
1322 ((dump_type
== DUMP_ALL
) ==
1323 !!(set
->type
->features
& IPSET_DUMP_LAST
))) {
1324 write_unlock_bh(&ip_set_ref_lock
);
1327 pr_debug("List set: %s\n", set
->name
);
1328 if (!cb
->args
[IPSET_CB_ARG0
]) {
1329 /* Start listing: make sure set won't be destroyed */
1330 pr_debug("reference set\n");
1333 write_unlock_bh(&ip_set_ref_lock
);
1334 nlh
= start_msg(skb
, NETLINK_CB(cb
->skb
).portid
,
1335 cb
->nlh
->nlmsg_seq
, flags
,
1339 goto release_refcount
;
1341 if (nla_put_u8(skb
, IPSET_ATTR_PROTOCOL
, IPSET_PROTOCOL
) ||
1342 nla_put_string(skb
, IPSET_ATTR_SETNAME
, set
->name
))
1343 goto nla_put_failure
;
1344 if (dump_flags
& IPSET_FLAG_LIST_SETNAME
)
1346 switch (cb
->args
[IPSET_CB_ARG0
]) {
1348 /* Core header data */
1349 if (nla_put_string(skb
, IPSET_ATTR_TYPENAME
,
1351 nla_put_u8(skb
, IPSET_ATTR_FAMILY
,
1353 nla_put_u8(skb
, IPSET_ATTR_REVISION
,
1355 goto nla_put_failure
;
1356 ret
= set
->variant
->head(set
, skb
);
1358 goto release_refcount
;
1359 if (dump_flags
& IPSET_FLAG_LIST_HEADER
)
1361 if (set
->variant
->uref
)
1362 set
->variant
->uref(set
, cb
, true);
1363 /* Fall through and add elements */
1366 ret
= set
->variant
->list(set
, skb
, cb
);
1367 rcu_read_unlock_bh();
1368 if (!cb
->args
[IPSET_CB_ARG0
])
1369 /* Set is done, proceed with next one */
1371 goto release_refcount
;
1374 /* If we dump all sets, continue with dumping last ones */
1375 if (dump_type
== DUMP_ALL
) {
1376 dump_type
= DUMP_LAST
;
1377 cb
->args
[IPSET_CB_DUMP
] = dump_type
| (dump_flags
<< 16);
1378 cb
->args
[IPSET_CB_INDEX
] = 0;
1379 if (set
&& set
->variant
->uref
)
1380 set
->variant
->uref(set
, cb
, false);
1388 if (dump_type
== DUMP_ONE
)
1389 cb
->args
[IPSET_CB_INDEX
] = IPSET_INVALID_ID
;
1391 cb
->args
[IPSET_CB_INDEX
]++;
1393 /* If there was an error or set is done, release set */
1394 if (ret
|| !cb
->args
[IPSET_CB_ARG0
]) {
1395 set
= ip_set(inst
, index
);
1396 if (set
->variant
->uref
)
1397 set
->variant
->uref(set
, cb
, false);
1398 pr_debug("release set %s\n", set
->name
);
1399 __ip_set_put_byindex(inst
, index
);
1400 cb
->args
[IPSET_CB_ARG0
] = 0;
1404 nlmsg_end(skb
, nlh
);
1405 pr_debug("nlmsg_len: %u\n", nlh
->nlmsg_len
);
1409 return ret
< 0 ? ret
: skb
->len
;
1412 static int ip_set_dump(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1413 const struct nlmsghdr
*nlh
,
1414 const struct nlattr
* const attr
[])
1416 if (unlikely(protocol_failed(attr
)))
1417 return -IPSET_ERR_PROTOCOL
;
1420 struct netlink_dump_control c
= {
1421 .dump
= ip_set_dump_start
,
1422 .done
= ip_set_dump_done
,
1424 return netlink_dump_start(ctnl
, skb
, nlh
, &c
);
1428 /* Add, del and test */
1430 static const struct nla_policy ip_set_adt_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1431 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1432 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
1433 .len
= IPSET_MAXNAMELEN
- 1 },
1434 [IPSET_ATTR_LINENO
] = { .type
= NLA_U32
},
1435 [IPSET_ATTR_DATA
] = { .type
= NLA_NESTED
},
1436 [IPSET_ATTR_ADT
] = { .type
= NLA_NESTED
},
1440 call_ad(struct sock
*ctnl
, struct sk_buff
*skb
, struct ip_set
*set
,
1441 struct nlattr
*tb
[], enum ipset_adt adt
,
1442 u32 flags
, bool use_lineno
)
1446 bool eexist
= flags
& IPSET_FLAG_EXIST
, retried
= false;
1449 spin_lock_bh(&set
->lock
);
1450 ret
= set
->variant
->uadt(set
, tb
, adt
, &lineno
, flags
, retried
);
1451 spin_unlock_bh(&set
->lock
);
1453 } while (ret
== -EAGAIN
&&
1454 set
->variant
->resize
&&
1455 (ret
= set
->variant
->resize(set
, retried
)) == 0);
1457 if (!ret
|| (ret
== -IPSET_ERR_EXIST
&& eexist
))
1459 if (lineno
&& use_lineno
) {
1460 /* Error in restore/batch mode: send back lineno */
1461 struct nlmsghdr
*rep
, *nlh
= nlmsg_hdr(skb
);
1462 struct sk_buff
*skb2
;
1463 struct nlmsgerr
*errmsg
;
1464 size_t payload
= min(SIZE_MAX
,
1465 sizeof(*errmsg
) + nlmsg_len(nlh
));
1466 int min_len
= nlmsg_total_size(sizeof(struct nfgenmsg
));
1467 struct nlattr
*cda
[IPSET_ATTR_CMD_MAX
+ 1];
1468 struct nlattr
*cmdattr
;
1471 skb2
= nlmsg_new(payload
, GFP_KERNEL
);
1474 rep
= __nlmsg_put(skb2
, NETLINK_CB(skb
).portid
,
1475 nlh
->nlmsg_seq
, NLMSG_ERROR
, payload
, 0);
1476 errmsg
= nlmsg_data(rep
);
1477 errmsg
->error
= ret
;
1478 memcpy(&errmsg
->msg
, nlh
, nlh
->nlmsg_len
);
1479 cmdattr
= (void *)&errmsg
->msg
+ min_len
;
1481 nla_parse(cda
, IPSET_ATTR_CMD_MAX
,
1482 cmdattr
, nlh
->nlmsg_len
- min_len
,
1485 errline
= nla_data(cda
[IPSET_ATTR_LINENO
]);
1489 netlink_unicast(ctnl
, skb2
, NETLINK_CB(skb
).portid
,
1491 /* Signal netlink not to send its ACK/errmsg. */
1498 static int ip_set_uadd(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1499 const struct nlmsghdr
*nlh
,
1500 const struct nlattr
* const attr
[])
1502 struct ip_set_net
*inst
= ip_set_pernet(net
);
1504 struct nlattr
*tb
[IPSET_ATTR_ADT_MAX
+ 1] = {};
1505 const struct nlattr
*nla
;
1506 u32 flags
= flag_exist(nlh
);
1510 if (unlikely(protocol_failed(attr
) ||
1511 !attr
[IPSET_ATTR_SETNAME
] ||
1512 !((attr
[IPSET_ATTR_DATA
] != NULL
) ^
1513 (attr
[IPSET_ATTR_ADT
] != NULL
)) ||
1514 (attr
[IPSET_ATTR_DATA
] &&
1515 !flag_nested(attr
[IPSET_ATTR_DATA
])) ||
1516 (attr
[IPSET_ATTR_ADT
] &&
1517 (!flag_nested(attr
[IPSET_ATTR_ADT
]) ||
1518 !attr
[IPSET_ATTR_LINENO
]))))
1519 return -IPSET_ERR_PROTOCOL
;
1521 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1525 use_lineno
= !!attr
[IPSET_ATTR_LINENO
];
1526 if (attr
[IPSET_ATTR_DATA
]) {
1527 if (nla_parse_nested(tb
, IPSET_ATTR_ADT_MAX
,
1528 attr
[IPSET_ATTR_DATA
],
1529 set
->type
->adt_policy
))
1530 return -IPSET_ERR_PROTOCOL
;
1531 ret
= call_ad(ctnl
, skb
, set
, tb
, IPSET_ADD
, flags
,
1536 nla_for_each_nested(nla
, attr
[IPSET_ATTR_ADT
], nla_rem
) {
1537 memset(tb
, 0, sizeof(tb
));
1538 if (nla_type(nla
) != IPSET_ATTR_DATA
||
1539 !flag_nested(nla
) ||
1540 nla_parse_nested(tb
, IPSET_ATTR_ADT_MAX
, nla
,
1541 set
->type
->adt_policy
))
1542 return -IPSET_ERR_PROTOCOL
;
1543 ret
= call_ad(ctnl
, skb
, set
, tb
, IPSET_ADD
,
1552 static int ip_set_udel(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1553 const struct nlmsghdr
*nlh
,
1554 const struct nlattr
* const attr
[])
1556 struct ip_set_net
*inst
= ip_set_pernet(net
);
1558 struct nlattr
*tb
[IPSET_ATTR_ADT_MAX
+ 1] = {};
1559 const struct nlattr
*nla
;
1560 u32 flags
= flag_exist(nlh
);
1564 if (unlikely(protocol_failed(attr
) ||
1565 !attr
[IPSET_ATTR_SETNAME
] ||
1566 !((attr
[IPSET_ATTR_DATA
] != NULL
) ^
1567 (attr
[IPSET_ATTR_ADT
] != NULL
)) ||
1568 (attr
[IPSET_ATTR_DATA
] &&
1569 !flag_nested(attr
[IPSET_ATTR_DATA
])) ||
1570 (attr
[IPSET_ATTR_ADT
] &&
1571 (!flag_nested(attr
[IPSET_ATTR_ADT
]) ||
1572 !attr
[IPSET_ATTR_LINENO
]))))
1573 return -IPSET_ERR_PROTOCOL
;
1575 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1579 use_lineno
= !!attr
[IPSET_ATTR_LINENO
];
1580 if (attr
[IPSET_ATTR_DATA
]) {
1581 if (nla_parse_nested(tb
, IPSET_ATTR_ADT_MAX
,
1582 attr
[IPSET_ATTR_DATA
],
1583 set
->type
->adt_policy
))
1584 return -IPSET_ERR_PROTOCOL
;
1585 ret
= call_ad(ctnl
, skb
, set
, tb
, IPSET_DEL
, flags
,
1590 nla_for_each_nested(nla
, attr
[IPSET_ATTR_ADT
], nla_rem
) {
1591 memset(tb
, 0, sizeof(*tb
));
1592 if (nla_type(nla
) != IPSET_ATTR_DATA
||
1593 !flag_nested(nla
) ||
1594 nla_parse_nested(tb
, IPSET_ATTR_ADT_MAX
, nla
,
1595 set
->type
->adt_policy
))
1596 return -IPSET_ERR_PROTOCOL
;
1597 ret
= call_ad(ctnl
, skb
, set
, tb
, IPSET_DEL
,
1606 static int ip_set_utest(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1607 const struct nlmsghdr
*nlh
,
1608 const struct nlattr
* const attr
[])
1610 struct ip_set_net
*inst
= ip_set_pernet(net
);
1612 struct nlattr
*tb
[IPSET_ATTR_ADT_MAX
+ 1] = {};
1615 if (unlikely(protocol_failed(attr
) ||
1616 !attr
[IPSET_ATTR_SETNAME
] ||
1617 !attr
[IPSET_ATTR_DATA
] ||
1618 !flag_nested(attr
[IPSET_ATTR_DATA
])))
1619 return -IPSET_ERR_PROTOCOL
;
1621 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1625 if (nla_parse_nested(tb
, IPSET_ATTR_ADT_MAX
, attr
[IPSET_ATTR_DATA
],
1626 set
->type
->adt_policy
))
1627 return -IPSET_ERR_PROTOCOL
;
1630 ret
= set
->variant
->uadt(set
, tb
, IPSET_TEST
, NULL
, 0, 0);
1631 rcu_read_unlock_bh();
1632 /* Userspace can't trigger element to be re-added */
1636 return ret
> 0 ? 0 : -IPSET_ERR_EXIST
;
1639 /* Get headed data of a set */
1641 static int ip_set_header(struct net
*net
, struct sock
*ctnl
,
1642 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1643 const struct nlattr
* const attr
[])
1645 struct ip_set_net
*inst
= ip_set_pernet(net
);
1646 const struct ip_set
*set
;
1647 struct sk_buff
*skb2
;
1648 struct nlmsghdr
*nlh2
;
1651 if (unlikely(protocol_failed(attr
) ||
1652 !attr
[IPSET_ATTR_SETNAME
]))
1653 return -IPSET_ERR_PROTOCOL
;
1655 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1659 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
1663 nlh2
= start_msg(skb2
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
1667 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, IPSET_PROTOCOL
) ||
1668 nla_put_string(skb2
, IPSET_ATTR_SETNAME
, set
->name
) ||
1669 nla_put_string(skb2
, IPSET_ATTR_TYPENAME
, set
->type
->name
) ||
1670 nla_put_u8(skb2
, IPSET_ATTR_FAMILY
, set
->family
) ||
1671 nla_put_u8(skb2
, IPSET_ATTR_REVISION
, set
->revision
))
1672 goto nla_put_failure
;
1673 nlmsg_end(skb2
, nlh2
);
1675 ret
= netlink_unicast(ctnl
, skb2
, NETLINK_CB(skb
).portid
, MSG_DONTWAIT
);
1682 nlmsg_cancel(skb2
, nlh2
);
1690 static const struct nla_policy ip_set_type_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1691 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1692 [IPSET_ATTR_TYPENAME
] = { .type
= NLA_NUL_STRING
,
1693 .len
= IPSET_MAXNAMELEN
- 1 },
1694 [IPSET_ATTR_FAMILY
] = { .type
= NLA_U8
},
1697 static int ip_set_type(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1698 const struct nlmsghdr
*nlh
,
1699 const struct nlattr
* const attr
[])
1701 struct sk_buff
*skb2
;
1702 struct nlmsghdr
*nlh2
;
1703 u8 family
, min
, max
;
1704 const char *typename
;
1707 if (unlikely(protocol_failed(attr
) ||
1708 !attr
[IPSET_ATTR_TYPENAME
] ||
1709 !attr
[IPSET_ATTR_FAMILY
]))
1710 return -IPSET_ERR_PROTOCOL
;
1712 family
= nla_get_u8(attr
[IPSET_ATTR_FAMILY
]);
1713 typename
= nla_data(attr
[IPSET_ATTR_TYPENAME
]);
1714 ret
= find_set_type_minmax(typename
, family
, &min
, &max
);
1718 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
1722 nlh2
= start_msg(skb2
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
1726 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, IPSET_PROTOCOL
) ||
1727 nla_put_string(skb2
, IPSET_ATTR_TYPENAME
, typename
) ||
1728 nla_put_u8(skb2
, IPSET_ATTR_FAMILY
, family
) ||
1729 nla_put_u8(skb2
, IPSET_ATTR_REVISION
, max
) ||
1730 nla_put_u8(skb2
, IPSET_ATTR_REVISION_MIN
, min
))
1731 goto nla_put_failure
;
1732 nlmsg_end(skb2
, nlh2
);
1734 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2
->nlmsg_len
);
1735 ret
= netlink_unicast(ctnl
, skb2
, NETLINK_CB(skb
).portid
, MSG_DONTWAIT
);
1742 nlmsg_cancel(skb2
, nlh2
);
1748 /* Get protocol version */
1750 static const struct nla_policy
1751 ip_set_protocol_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1752 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1755 static int ip_set_protocol(struct net
*net
, struct sock
*ctnl
,
1756 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1757 const struct nlattr
* const attr
[])
1759 struct sk_buff
*skb2
;
1760 struct nlmsghdr
*nlh2
;
1763 if (unlikely(!attr
[IPSET_ATTR_PROTOCOL
]))
1764 return -IPSET_ERR_PROTOCOL
;
1766 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
1770 nlh2
= start_msg(skb2
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
1771 IPSET_CMD_PROTOCOL
);
1774 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, IPSET_PROTOCOL
))
1775 goto nla_put_failure
;
1776 nlmsg_end(skb2
, nlh2
);
1778 ret
= netlink_unicast(ctnl
, skb2
, NETLINK_CB(skb
).portid
, MSG_DONTWAIT
);
1785 nlmsg_cancel(skb2
, nlh2
);
1791 static const struct nfnl_callback ip_set_netlink_subsys_cb
[IPSET_MSG_MAX
] = {
1792 [IPSET_CMD_NONE
] = {
1793 .call
= ip_set_none
,
1794 .attr_count
= IPSET_ATTR_CMD_MAX
,
1796 [IPSET_CMD_CREATE
] = {
1797 .call
= ip_set_create
,
1798 .attr_count
= IPSET_ATTR_CMD_MAX
,
1799 .policy
= ip_set_create_policy
,
1801 [IPSET_CMD_DESTROY
] = {
1802 .call
= ip_set_destroy
,
1803 .attr_count
= IPSET_ATTR_CMD_MAX
,
1804 .policy
= ip_set_setname_policy
,
1806 [IPSET_CMD_FLUSH
] = {
1807 .call
= ip_set_flush
,
1808 .attr_count
= IPSET_ATTR_CMD_MAX
,
1809 .policy
= ip_set_setname_policy
,
1811 [IPSET_CMD_RENAME
] = {
1812 .call
= ip_set_rename
,
1813 .attr_count
= IPSET_ATTR_CMD_MAX
,
1814 .policy
= ip_set_setname2_policy
,
1816 [IPSET_CMD_SWAP
] = {
1817 .call
= ip_set_swap
,
1818 .attr_count
= IPSET_ATTR_CMD_MAX
,
1819 .policy
= ip_set_setname2_policy
,
1821 [IPSET_CMD_LIST
] = {
1822 .call
= ip_set_dump
,
1823 .attr_count
= IPSET_ATTR_CMD_MAX
,
1824 .policy
= ip_set_setname_policy
,
1826 [IPSET_CMD_SAVE
] = {
1827 .call
= ip_set_dump
,
1828 .attr_count
= IPSET_ATTR_CMD_MAX
,
1829 .policy
= ip_set_setname_policy
,
1832 .call
= ip_set_uadd
,
1833 .attr_count
= IPSET_ATTR_CMD_MAX
,
1834 .policy
= ip_set_adt_policy
,
1837 .call
= ip_set_udel
,
1838 .attr_count
= IPSET_ATTR_CMD_MAX
,
1839 .policy
= ip_set_adt_policy
,
1841 [IPSET_CMD_TEST
] = {
1842 .call
= ip_set_utest
,
1843 .attr_count
= IPSET_ATTR_CMD_MAX
,
1844 .policy
= ip_set_adt_policy
,
1846 [IPSET_CMD_HEADER
] = {
1847 .call
= ip_set_header
,
1848 .attr_count
= IPSET_ATTR_CMD_MAX
,
1849 .policy
= ip_set_setname_policy
,
1851 [IPSET_CMD_TYPE
] = {
1852 .call
= ip_set_type
,
1853 .attr_count
= IPSET_ATTR_CMD_MAX
,
1854 .policy
= ip_set_type_policy
,
1856 [IPSET_CMD_PROTOCOL
] = {
1857 .call
= ip_set_protocol
,
1858 .attr_count
= IPSET_ATTR_CMD_MAX
,
1859 .policy
= ip_set_protocol_policy
,
1863 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly
= {
1865 .subsys_id
= NFNL_SUBSYS_IPSET
,
1866 .cb_count
= IPSET_MSG_MAX
,
1867 .cb
= ip_set_netlink_subsys_cb
,
1870 /* Interface to iptables/ip6tables */
1873 ip_set_sockfn_get(struct sock
*sk
, int optval
, void __user
*user
, int *len
)
1877 int copylen
= *len
, ret
= 0;
1878 struct net
*net
= sock_net(sk
);
1879 struct ip_set_net
*inst
= ip_set_pernet(net
);
1881 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1883 if (optval
!= SO_IP_SET
)
1885 if (*len
< sizeof(unsigned int))
1888 data
= vmalloc(*len
);
1891 if (copy_from_user(data
, user
, *len
) != 0) {
1895 op
= (unsigned int *)data
;
1897 if (*op
< IP_SET_OP_VERSION
) {
1898 /* Check the version at the beginning of operations */
1899 struct ip_set_req_version
*req_version
= data
;
1901 if (*len
< sizeof(struct ip_set_req_version
)) {
1906 if (req_version
->version
!= IPSET_PROTOCOL
) {
1913 case IP_SET_OP_VERSION
: {
1914 struct ip_set_req_version
*req_version
= data
;
1916 if (*len
!= sizeof(struct ip_set_req_version
)) {
1921 req_version
->version
= IPSET_PROTOCOL
;
1922 ret
= copy_to_user(user
, req_version
,
1923 sizeof(struct ip_set_req_version
));
1926 case IP_SET_OP_GET_BYNAME
: {
1927 struct ip_set_req_get_set
*req_get
= data
;
1930 if (*len
!= sizeof(struct ip_set_req_get_set
)) {
1934 req_get
->set
.name
[IPSET_MAXNAMELEN
- 1] = '\0';
1935 nfnl_lock(NFNL_SUBSYS_IPSET
);
1936 find_set_and_id(inst
, req_get
->set
.name
, &id
);
1937 req_get
->set
.index
= id
;
1938 nfnl_unlock(NFNL_SUBSYS_IPSET
);
1941 case IP_SET_OP_GET_FNAME
: {
1942 struct ip_set_req_get_set_family
*req_get
= data
;
1945 if (*len
!= sizeof(struct ip_set_req_get_set_family
)) {
1949 req_get
->set
.name
[IPSET_MAXNAMELEN
- 1] = '\0';
1950 nfnl_lock(NFNL_SUBSYS_IPSET
);
1951 find_set_and_id(inst
, req_get
->set
.name
, &id
);
1952 req_get
->set
.index
= id
;
1953 if (id
!= IPSET_INVALID_ID
)
1954 req_get
->family
= ip_set(inst
, id
)->family
;
1955 nfnl_unlock(NFNL_SUBSYS_IPSET
);
1958 case IP_SET_OP_GET_BYINDEX
: {
1959 struct ip_set_req_get_set
*req_get
= data
;
1962 if (*len
!= sizeof(struct ip_set_req_get_set
) ||
1963 req_get
->set
.index
>= inst
->ip_set_max
) {
1967 nfnl_lock(NFNL_SUBSYS_IPSET
);
1968 set
= ip_set(inst
, req_get
->set
.index
);
1969 strncpy(req_get
->set
.name
, set
? set
->name
: "",
1971 nfnl_unlock(NFNL_SUBSYS_IPSET
);
1977 } /* end of switch(op) */
1980 ret
= copy_to_user(user
, data
, copylen
);
1989 static struct nf_sockopt_ops so_set __read_mostly
= {
1991 .get_optmin
= SO_IP_SET
,
1992 .get_optmax
= SO_IP_SET
+ 1,
1993 .get
= &ip_set_sockfn_get
,
1994 .owner
= THIS_MODULE
,
1997 static int __net_init
1998 ip_set_net_init(struct net
*net
)
2000 struct ip_set_net
*inst
= ip_set_pernet(net
);
2001 struct ip_set
**list
;
2003 inst
->ip_set_max
= max_sets
? max_sets
: CONFIG_IP_SET_MAX
;
2004 if (inst
->ip_set_max
>= IPSET_INVALID_ID
)
2005 inst
->ip_set_max
= IPSET_INVALID_ID
- 1;
2007 list
= kcalloc(inst
->ip_set_max
, sizeof(struct ip_set
*), GFP_KERNEL
);
2010 inst
->is_deleted
= false;
2011 inst
->is_destroyed
= false;
2012 rcu_assign_pointer(inst
->ip_set_list
, list
);
2016 static void __net_exit
2017 ip_set_net_exit(struct net
*net
)
2019 struct ip_set_net
*inst
= ip_set_pernet(net
);
2021 struct ip_set
*set
= NULL
;
2024 inst
->is_deleted
= true; /* flag for ip_set_nfnl_put */
2026 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
2027 set
= ip_set(inst
, i
);
2029 ip_set(inst
, i
) = NULL
;
2030 ip_set_destroy_set(set
);
2033 kfree(rcu_dereference_protected(inst
->ip_set_list
, 1));
2036 static struct pernet_operations ip_set_net_ops
= {
2037 .init
= ip_set_net_init
,
2038 .exit
= ip_set_net_exit
,
2039 .id
= &ip_set_net_id
,
2040 .size
= sizeof(struct ip_set_net
)
2046 int ret
= nfnetlink_subsys_register(&ip_set_netlink_subsys
);
2049 pr_err("ip_set: cannot register with nfnetlink.\n");
2052 ret
= nf_register_sockopt(&so_set
);
2054 pr_err("SO_SET registry failed: %d\n", ret
);
2055 nfnetlink_subsys_unregister(&ip_set_netlink_subsys
);
2058 ret
= register_pernet_subsys(&ip_set_net_ops
);
2060 pr_err("ip_set: cannot register pernet_subsys.\n");
2061 nf_unregister_sockopt(&so_set
);
2062 nfnetlink_subsys_unregister(&ip_set_netlink_subsys
);
2065 pr_info("ip_set: protocol %u\n", IPSET_PROTOCOL
);
2072 unregister_pernet_subsys(&ip_set_net_ops
);
2073 nf_unregister_sockopt(&so_set
);
2074 nfnetlink_subsys_unregister(&ip_set_netlink_subsys
);
2075 pr_debug("these are the famous last words\n");
2078 module_init(ip_set_init
);
2079 module_exit(ip_set_fini
);