1 // SPDX-License-Identifier: GPL-2.0-only
2 /* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
3 * Patrick Schaaf <bof@bof.de>
4 * Copyright (C) 2003-2013 Jozsef Kadlecsik <kadlec@netfilter.org>
7 /* Kernel module for IP set management */
9 #include <linux/init.h>
10 #include <linux/module.h>
11 #include <linux/moduleparam.h>
13 #include <linux/skbuff.h>
14 #include <linux/spinlock.h>
15 #include <linux/rculist.h>
16 #include <net/netlink.h>
17 #include <net/net_namespace.h>
18 #include <net/netns/generic.h>
20 #include <linux/netfilter.h>
21 #include <linux/netfilter/x_tables.h>
22 #include <linux/netfilter/nfnetlink.h>
23 #include <linux/netfilter/ipset/ip_set.h>
25 static LIST_HEAD(ip_set_type_list
); /* all registered set types */
26 static DEFINE_MUTEX(ip_set_type_mutex
); /* protects ip_set_type_list */
27 static DEFINE_RWLOCK(ip_set_ref_lock
); /* protects the set refs */
30 struct ip_set
* __rcu
*ip_set_list
; /* all individual sets */
31 ip_set_id_t ip_set_max
; /* max number of sets */
32 bool is_deleted
; /* deleted by ip_set_net_exit */
33 bool is_destroyed
; /* all sets are destroyed */
36 static unsigned int ip_set_net_id __read_mostly
;
38 static struct ip_set_net
*ip_set_pernet(struct net
*net
)
40 return net_generic(net
, ip_set_net_id
);
44 #define STRNCMP(a, b) (strncmp(a, b, IPSET_MAXNAMELEN) == 0)
46 static unsigned int max_sets
;
48 module_param(max_sets
, int, 0600);
49 MODULE_PARM_DESC(max_sets
, "maximal number of sets");
50 MODULE_LICENSE("GPL");
51 MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@netfilter.org>");
52 MODULE_DESCRIPTION("core IP set support");
53 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_IPSET
);
55 /* When the nfnl mutex or ip_set_ref_lock is held: */
56 #define ip_set_dereference(p) \
57 rcu_dereference_protected(p, \
58 lockdep_nfnl_is_held(NFNL_SUBSYS_IPSET) || \
59 lockdep_is_held(&ip_set_ref_lock))
60 #define ip_set(inst, id) \
61 ip_set_dereference((inst)->ip_set_list)[id]
62 #define ip_set_ref_netlink(inst,id) \
63 rcu_dereference_raw((inst)->ip_set_list)[id]
65 /* The set types are implemented in modules and registered set types
66 * can be found in ip_set_type_list. Adding/deleting types is
67 * serialized by ip_set_type_mutex.
71 ip_set_type_lock(void)
73 mutex_lock(&ip_set_type_mutex
);
77 ip_set_type_unlock(void)
79 mutex_unlock(&ip_set_type_mutex
);
82 /* Register and deregister settype */
84 static struct ip_set_type
*
85 find_set_type(const char *name
, u8 family
, u8 revision
)
87 struct ip_set_type
*type
;
89 list_for_each_entry_rcu(type
, &ip_set_type_list
, list
)
90 if (STRNCMP(type
->name
, name
) &&
91 (type
->family
== family
||
92 type
->family
== NFPROTO_UNSPEC
) &&
93 revision
>= type
->revision_min
&&
94 revision
<= type
->revision_max
)
99 /* Unlock, try to load a set type module and lock again */
101 load_settype(const char *name
)
103 nfnl_unlock(NFNL_SUBSYS_IPSET
);
104 pr_debug("try to load ip_set_%s\n", name
);
105 if (request_module("ip_set_%s", name
) < 0) {
106 pr_warn("Can't find ip_set type %s\n", name
);
107 nfnl_lock(NFNL_SUBSYS_IPSET
);
110 nfnl_lock(NFNL_SUBSYS_IPSET
);
114 /* Find a set type and reference it */
115 #define find_set_type_get(name, family, revision, found) \
116 __find_set_type_get(name, family, revision, found, false)
119 __find_set_type_get(const char *name
, u8 family
, u8 revision
,
120 struct ip_set_type
**found
, bool retry
)
122 struct ip_set_type
*type
;
125 if (retry
&& !load_settype(name
))
126 return -IPSET_ERR_FIND_TYPE
;
129 *found
= find_set_type(name
, family
, revision
);
131 err
= !try_module_get((*found
)->me
) ? -EFAULT
: 0;
134 /* Make sure the type is already loaded
135 * but we don't support the revision
137 list_for_each_entry_rcu(type
, &ip_set_type_list
, list
)
138 if (STRNCMP(type
->name
, name
)) {
139 err
= -IPSET_ERR_FIND_TYPE
;
144 return retry
? -IPSET_ERR_FIND_TYPE
:
145 __find_set_type_get(name
, family
, revision
, found
, true);
152 /* Find a given set type by name and family.
153 * If we succeeded, the supported minimal and maximum revisions are
156 #define find_set_type_minmax(name, family, min, max) \
157 __find_set_type_minmax(name, family, min, max, false)
160 __find_set_type_minmax(const char *name
, u8 family
, u8
*min
, u8
*max
,
163 struct ip_set_type
*type
;
166 if (retry
&& !load_settype(name
))
167 return -IPSET_ERR_FIND_TYPE
;
169 *min
= 255; *max
= 0;
171 list_for_each_entry_rcu(type
, &ip_set_type_list
, list
)
172 if (STRNCMP(type
->name
, name
) &&
173 (type
->family
== family
||
174 type
->family
== NFPROTO_UNSPEC
)) {
176 if (type
->revision_min
< *min
)
177 *min
= type
->revision_min
;
178 if (type
->revision_max
> *max
)
179 *max
= type
->revision_max
;
185 return retry
? -IPSET_ERR_FIND_TYPE
:
186 __find_set_type_minmax(name
, family
, min
, max
, true);
189 #define family_name(f) ((f) == NFPROTO_IPV4 ? "inet" : \
190 (f) == NFPROTO_IPV6 ? "inet6" : "any")
192 /* Register a set type structure. The type is identified by
193 * the unique triple of name, family and revision.
196 ip_set_type_register(struct ip_set_type
*type
)
200 if (type
->protocol
!= IPSET_PROTOCOL
) {
201 pr_warn("ip_set type %s, family %s, revision %u:%u uses wrong protocol version %u (want %u)\n",
202 type
->name
, family_name(type
->family
),
203 type
->revision_min
, type
->revision_max
,
204 type
->protocol
, IPSET_PROTOCOL
);
209 if (find_set_type(type
->name
, type
->family
, type
->revision_min
)) {
211 pr_warn("ip_set type %s, family %s with revision min %u already registered!\n",
212 type
->name
, family_name(type
->family
),
214 ip_set_type_unlock();
217 list_add_rcu(&type
->list
, &ip_set_type_list
);
218 pr_debug("type %s, family %s, revision %u:%u registered.\n",
219 type
->name
, family_name(type
->family
),
220 type
->revision_min
, type
->revision_max
);
221 ip_set_type_unlock();
225 EXPORT_SYMBOL_GPL(ip_set_type_register
);
227 /* Unregister a set type. There's a small race with ip_set_create */
229 ip_set_type_unregister(struct ip_set_type
*type
)
232 if (!find_set_type(type
->name
, type
->family
, type
->revision_min
)) {
233 pr_warn("ip_set type %s, family %s with revision min %u not registered\n",
234 type
->name
, family_name(type
->family
),
236 ip_set_type_unlock();
239 list_del_rcu(&type
->list
);
240 pr_debug("type %s, family %s with revision min %u unregistered.\n",
241 type
->name
, family_name(type
->family
), type
->revision_min
);
242 ip_set_type_unlock();
246 EXPORT_SYMBOL_GPL(ip_set_type_unregister
);
248 /* Utility functions */
250 ip_set_alloc(size_t size
)
252 void *members
= NULL
;
254 if (size
< KMALLOC_MAX_SIZE
)
255 members
= kzalloc(size
, GFP_KERNEL
| __GFP_NOWARN
);
258 pr_debug("%p: allocated with kmalloc\n", members
);
262 members
= vzalloc(size
);
265 pr_debug("%p: allocated with vmalloc\n", members
);
269 EXPORT_SYMBOL_GPL(ip_set_alloc
);
272 ip_set_free(void *members
)
274 pr_debug("%p: free with %s\n", members
,
275 is_vmalloc_addr(members
) ? "vfree" : "kfree");
278 EXPORT_SYMBOL_GPL(ip_set_free
);
281 flag_nested(const struct nlattr
*nla
)
283 return nla
->nla_type
& NLA_F_NESTED
;
286 static const struct nla_policy ipaddr_policy
[IPSET_ATTR_IPADDR_MAX
+ 1] = {
287 [IPSET_ATTR_IPADDR_IPV4
] = { .type
= NLA_U32
},
288 [IPSET_ATTR_IPADDR_IPV6
] = { .type
= NLA_BINARY
,
289 .len
= sizeof(struct in6_addr
) },
293 ip_set_get_ipaddr4(struct nlattr
*nla
, __be32
*ipaddr
)
295 struct nlattr
*tb
[IPSET_ATTR_IPADDR_MAX
+ 1];
297 if (unlikely(!flag_nested(nla
)))
298 return -IPSET_ERR_PROTOCOL
;
299 if (nla_parse_nested(tb
, IPSET_ATTR_IPADDR_MAX
, nla
,
300 ipaddr_policy
, NULL
))
301 return -IPSET_ERR_PROTOCOL
;
302 if (unlikely(!ip_set_attr_netorder(tb
, IPSET_ATTR_IPADDR_IPV4
)))
303 return -IPSET_ERR_PROTOCOL
;
305 *ipaddr
= nla_get_be32(tb
[IPSET_ATTR_IPADDR_IPV4
]);
308 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr4
);
311 ip_set_get_ipaddr6(struct nlattr
*nla
, union nf_inet_addr
*ipaddr
)
313 struct nlattr
*tb
[IPSET_ATTR_IPADDR_MAX
+ 1];
315 if (unlikely(!flag_nested(nla
)))
316 return -IPSET_ERR_PROTOCOL
;
318 if (nla_parse_nested(tb
, IPSET_ATTR_IPADDR_MAX
, nla
,
319 ipaddr_policy
, NULL
))
320 return -IPSET_ERR_PROTOCOL
;
321 if (unlikely(!ip_set_attr_netorder(tb
, IPSET_ATTR_IPADDR_IPV6
)))
322 return -IPSET_ERR_PROTOCOL
;
324 memcpy(ipaddr
, nla_data(tb
[IPSET_ATTR_IPADDR_IPV6
]),
325 sizeof(struct in6_addr
));
328 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr6
);
331 ip_set_timeout_get(const unsigned long *timeout
)
335 if (*timeout
== IPSET_ELEM_PERMANENT
)
338 t
= jiffies_to_msecs(*timeout
- jiffies
) / MSEC_PER_SEC
;
339 /* Zero value in userspace means no timeout */
340 return t
== 0 ? 1 : t
;
344 ip_set_comment_uget(struct nlattr
*tb
)
349 /* Called from uadd only, protected by the set spinlock.
350 * The kadt functions don't use the comment extensions in any way.
353 ip_set_init_comment(struct ip_set
*set
, struct ip_set_comment
*comment
,
354 const struct ip_set_ext
*ext
)
356 struct ip_set_comment_rcu
*c
= rcu_dereference_protected(comment
->c
, 1);
357 size_t len
= ext
->comment
? strlen(ext
->comment
) : 0;
360 set
->ext_size
-= sizeof(*c
) + strlen(c
->str
) + 1;
362 rcu_assign_pointer(comment
->c
, NULL
);
366 if (unlikely(len
> IPSET_MAX_COMMENT_SIZE
))
367 len
= IPSET_MAX_COMMENT_SIZE
;
368 c
= kmalloc(sizeof(*c
) + len
+ 1, GFP_ATOMIC
);
371 strlcpy(c
->str
, ext
->comment
, len
+ 1);
372 set
->ext_size
+= sizeof(*c
) + strlen(c
->str
) + 1;
373 rcu_assign_pointer(comment
->c
, c
);
375 EXPORT_SYMBOL_GPL(ip_set_init_comment
);
377 /* Used only when dumping a set, protected by rcu_read_lock() */
379 ip_set_put_comment(struct sk_buff
*skb
, const struct ip_set_comment
*comment
)
381 struct ip_set_comment_rcu
*c
= rcu_dereference(comment
->c
);
385 return nla_put_string(skb
, IPSET_ATTR_COMMENT
, c
->str
);
388 /* Called from uadd/udel, flush or the garbage collectors protected
389 * by the set spinlock.
390 * Called when the set is destroyed and when there can't be any user
391 * of the set data anymore.
394 ip_set_comment_free(struct ip_set
*set
, void *ptr
)
396 struct ip_set_comment
*comment
= ptr
;
397 struct ip_set_comment_rcu
*c
;
399 c
= rcu_dereference_protected(comment
->c
, 1);
402 set
->ext_size
-= sizeof(*c
) + strlen(c
->str
) + 1;
404 rcu_assign_pointer(comment
->c
, NULL
);
407 typedef void (*destroyer
)(struct ip_set
*, void *);
408 /* ipset data extension types, in size order */
410 const struct ip_set_ext_type ip_set_extensions
[] = {
411 [IPSET_EXT_ID_COUNTER
] = {
412 .type
= IPSET_EXT_COUNTER
,
413 .flag
= IPSET_FLAG_WITH_COUNTERS
,
414 .len
= sizeof(struct ip_set_counter
),
415 .align
= __alignof__(struct ip_set_counter
),
417 [IPSET_EXT_ID_TIMEOUT
] = {
418 .type
= IPSET_EXT_TIMEOUT
,
419 .len
= sizeof(unsigned long),
420 .align
= __alignof__(unsigned long),
422 [IPSET_EXT_ID_SKBINFO
] = {
423 .type
= IPSET_EXT_SKBINFO
,
424 .flag
= IPSET_FLAG_WITH_SKBINFO
,
425 .len
= sizeof(struct ip_set_skbinfo
),
426 .align
= __alignof__(struct ip_set_skbinfo
),
428 [IPSET_EXT_ID_COMMENT
] = {
429 .type
= IPSET_EXT_COMMENT
| IPSET_EXT_DESTROY
,
430 .flag
= IPSET_FLAG_WITH_COMMENT
,
431 .len
= sizeof(struct ip_set_comment
),
432 .align
= __alignof__(struct ip_set_comment
),
433 .destroy
= ip_set_comment_free
,
436 EXPORT_SYMBOL_GPL(ip_set_extensions
);
439 add_extension(enum ip_set_ext_id id
, u32 flags
, struct nlattr
*tb
[])
441 return ip_set_extensions
[id
].flag
?
442 (flags
& ip_set_extensions
[id
].flag
) :
443 !!tb
[IPSET_ATTR_TIMEOUT
];
447 ip_set_elem_len(struct ip_set
*set
, struct nlattr
*tb
[], size_t len
,
450 enum ip_set_ext_id id
;
453 if (tb
[IPSET_ATTR_CADT_FLAGS
])
454 cadt_flags
= ip_set_get_h32(tb
[IPSET_ATTR_CADT_FLAGS
]);
455 if (cadt_flags
& IPSET_FLAG_WITH_FORCEADD
)
456 set
->flags
|= IPSET_CREATE_FLAG_FORCEADD
;
459 for (id
= 0; id
< IPSET_EXT_ID_MAX
; id
++) {
460 if (!add_extension(id
, cadt_flags
, tb
))
462 len
= ALIGN(len
, ip_set_extensions
[id
].align
);
463 set
->offset
[id
] = len
;
464 set
->extensions
|= ip_set_extensions
[id
].type
;
465 len
+= ip_set_extensions
[id
].len
;
467 return ALIGN(len
, align
);
469 EXPORT_SYMBOL_GPL(ip_set_elem_len
);
472 ip_set_get_extensions(struct ip_set
*set
, struct nlattr
*tb
[],
473 struct ip_set_ext
*ext
)
477 if (unlikely(!ip_set_optattr_netorder(tb
, IPSET_ATTR_TIMEOUT
) ||
478 !ip_set_optattr_netorder(tb
, IPSET_ATTR_PACKETS
) ||
479 !ip_set_optattr_netorder(tb
, IPSET_ATTR_BYTES
) ||
480 !ip_set_optattr_netorder(tb
, IPSET_ATTR_SKBMARK
) ||
481 !ip_set_optattr_netorder(tb
, IPSET_ATTR_SKBPRIO
) ||
482 !ip_set_optattr_netorder(tb
, IPSET_ATTR_SKBQUEUE
)))
483 return -IPSET_ERR_PROTOCOL
;
485 if (tb
[IPSET_ATTR_TIMEOUT
]) {
486 if (!SET_WITH_TIMEOUT(set
))
487 return -IPSET_ERR_TIMEOUT
;
488 ext
->timeout
= ip_set_timeout_uget(tb
[IPSET_ATTR_TIMEOUT
]);
490 if (tb
[IPSET_ATTR_BYTES
] || tb
[IPSET_ATTR_PACKETS
]) {
491 if (!SET_WITH_COUNTER(set
))
492 return -IPSET_ERR_COUNTER
;
493 if (tb
[IPSET_ATTR_BYTES
])
494 ext
->bytes
= be64_to_cpu(nla_get_be64(
495 tb
[IPSET_ATTR_BYTES
]));
496 if (tb
[IPSET_ATTR_PACKETS
])
497 ext
->packets
= be64_to_cpu(nla_get_be64(
498 tb
[IPSET_ATTR_PACKETS
]));
500 if (tb
[IPSET_ATTR_COMMENT
]) {
501 if (!SET_WITH_COMMENT(set
))
502 return -IPSET_ERR_COMMENT
;
503 ext
->comment
= ip_set_comment_uget(tb
[IPSET_ATTR_COMMENT
]);
505 if (tb
[IPSET_ATTR_SKBMARK
]) {
506 if (!SET_WITH_SKBINFO(set
))
507 return -IPSET_ERR_SKBINFO
;
508 fullmark
= be64_to_cpu(nla_get_be64(tb
[IPSET_ATTR_SKBMARK
]));
509 ext
->skbinfo
.skbmark
= fullmark
>> 32;
510 ext
->skbinfo
.skbmarkmask
= fullmark
& 0xffffffff;
512 if (tb
[IPSET_ATTR_SKBPRIO
]) {
513 if (!SET_WITH_SKBINFO(set
))
514 return -IPSET_ERR_SKBINFO
;
515 ext
->skbinfo
.skbprio
=
516 be32_to_cpu(nla_get_be32(tb
[IPSET_ATTR_SKBPRIO
]));
518 if (tb
[IPSET_ATTR_SKBQUEUE
]) {
519 if (!SET_WITH_SKBINFO(set
))
520 return -IPSET_ERR_SKBINFO
;
521 ext
->skbinfo
.skbqueue
=
522 be16_to_cpu(nla_get_be16(tb
[IPSET_ATTR_SKBQUEUE
]));
526 EXPORT_SYMBOL_GPL(ip_set_get_extensions
);
529 ip_set_get_bytes(const struct ip_set_counter
*counter
)
531 return (u64
)atomic64_read(&(counter
)->bytes
);
535 ip_set_get_packets(const struct ip_set_counter
*counter
)
537 return (u64
)atomic64_read(&(counter
)->packets
);
541 ip_set_put_counter(struct sk_buff
*skb
, const struct ip_set_counter
*counter
)
543 return nla_put_net64(skb
, IPSET_ATTR_BYTES
,
544 cpu_to_be64(ip_set_get_bytes(counter
)),
546 nla_put_net64(skb
, IPSET_ATTR_PACKETS
,
547 cpu_to_be64(ip_set_get_packets(counter
)),
552 ip_set_put_skbinfo(struct sk_buff
*skb
, const struct ip_set_skbinfo
*skbinfo
)
554 /* Send nonzero parameters only */
555 return ((skbinfo
->skbmark
|| skbinfo
->skbmarkmask
) &&
556 nla_put_net64(skb
, IPSET_ATTR_SKBMARK
,
557 cpu_to_be64((u64
)skbinfo
->skbmark
<< 32 |
558 skbinfo
->skbmarkmask
),
561 nla_put_net32(skb
, IPSET_ATTR_SKBPRIO
,
562 cpu_to_be32(skbinfo
->skbprio
))) ||
563 (skbinfo
->skbqueue
&&
564 nla_put_net16(skb
, IPSET_ATTR_SKBQUEUE
,
565 cpu_to_be16(skbinfo
->skbqueue
)));
569 ip_set_put_extensions(struct sk_buff
*skb
, const struct ip_set
*set
,
570 const void *e
, bool active
)
572 if (SET_WITH_TIMEOUT(set
)) {
573 unsigned long *timeout
= ext_timeout(e
, set
);
575 if (nla_put_net32(skb
, IPSET_ATTR_TIMEOUT
,
576 htonl(active
? ip_set_timeout_get(timeout
)
580 if (SET_WITH_COUNTER(set
) &&
581 ip_set_put_counter(skb
, ext_counter(e
, set
)))
583 if (SET_WITH_COMMENT(set
) &&
584 ip_set_put_comment(skb
, ext_comment(e
, set
)))
586 if (SET_WITH_SKBINFO(set
) &&
587 ip_set_put_skbinfo(skb
, ext_skbinfo(e
, set
)))
591 EXPORT_SYMBOL_GPL(ip_set_put_extensions
);
594 ip_set_match_counter(u64 counter
, u64 match
, u8 op
)
597 case IPSET_COUNTER_NONE
:
599 case IPSET_COUNTER_EQ
:
600 return counter
== match
;
601 case IPSET_COUNTER_NE
:
602 return counter
!= match
;
603 case IPSET_COUNTER_LT
:
604 return counter
< match
;
605 case IPSET_COUNTER_GT
:
606 return counter
> match
;
612 ip_set_add_bytes(u64 bytes
, struct ip_set_counter
*counter
)
614 atomic64_add((long long)bytes
, &(counter
)->bytes
);
618 ip_set_add_packets(u64 packets
, struct ip_set_counter
*counter
)
620 atomic64_add((long long)packets
, &(counter
)->packets
);
624 ip_set_update_counter(struct ip_set_counter
*counter
,
625 const struct ip_set_ext
*ext
, u32 flags
)
627 if (ext
->packets
!= ULLONG_MAX
&&
628 !(flags
& IPSET_FLAG_SKIP_COUNTER_UPDATE
)) {
629 ip_set_add_bytes(ext
->bytes
, counter
);
630 ip_set_add_packets(ext
->packets
, counter
);
635 ip_set_get_skbinfo(struct ip_set_skbinfo
*skbinfo
,
636 const struct ip_set_ext
*ext
,
637 struct ip_set_ext
*mext
, u32 flags
)
639 mext
->skbinfo
= *skbinfo
;
643 ip_set_match_extensions(struct ip_set
*set
, const struct ip_set_ext
*ext
,
644 struct ip_set_ext
*mext
, u32 flags
, void *data
)
646 if (SET_WITH_TIMEOUT(set
) &&
647 ip_set_timeout_expired(ext_timeout(data
, set
)))
649 if (SET_WITH_COUNTER(set
)) {
650 struct ip_set_counter
*counter
= ext_counter(data
, set
);
652 if (flags
& IPSET_FLAG_MATCH_COUNTERS
&&
653 !(ip_set_match_counter(ip_set_get_packets(counter
),
654 mext
->packets
, mext
->packets_op
) &&
655 ip_set_match_counter(ip_set_get_bytes(counter
),
656 mext
->bytes
, mext
->bytes_op
)))
658 ip_set_update_counter(counter
, ext
, flags
);
660 if (SET_WITH_SKBINFO(set
))
661 ip_set_get_skbinfo(ext_skbinfo(data
, set
),
665 EXPORT_SYMBOL_GPL(ip_set_match_extensions
);
667 /* Creating/destroying/renaming/swapping affect the existence and
668 * the properties of a set. All of these can be executed from userspace
669 * only and serialized by the nfnl mutex indirectly from nfnetlink.
671 * Sets are identified by their index in ip_set_list and the index
672 * is used by the external references (set/SET netfilter modules).
674 * The set behind an index may change by swapping only, from userspace.
678 __ip_set_get(struct ip_set
*set
)
680 write_lock_bh(&ip_set_ref_lock
);
682 write_unlock_bh(&ip_set_ref_lock
);
686 __ip_set_put(struct ip_set
*set
)
688 write_lock_bh(&ip_set_ref_lock
);
689 BUG_ON(set
->ref
== 0);
691 write_unlock_bh(&ip_set_ref_lock
);
694 /* set->ref can be swapped out by ip_set_swap, netlink events (like dump) need
695 * a separate reference counter
698 __ip_set_put_netlink(struct ip_set
*set
)
700 write_lock_bh(&ip_set_ref_lock
);
701 BUG_ON(set
->ref_netlink
== 0);
703 write_unlock_bh(&ip_set_ref_lock
);
706 /* Add, del and test set entries from kernel.
708 * The set behind the index must exist and must be referenced
709 * so it can't be destroyed (or changed) under our foot.
712 static struct ip_set
*
713 ip_set_rcu_get(struct net
*net
, ip_set_id_t index
)
716 struct ip_set_net
*inst
= ip_set_pernet(net
);
719 /* ip_set_list itself needs to be protected */
720 set
= rcu_dereference(inst
->ip_set_list
)[index
];
727 ip_set_lock(struct ip_set
*set
)
729 if (!set
->variant
->region_lock
)
730 spin_lock_bh(&set
->lock
);
734 ip_set_unlock(struct ip_set
*set
)
736 if (!set
->variant
->region_lock
)
737 spin_unlock_bh(&set
->lock
);
741 ip_set_test(ip_set_id_t index
, const struct sk_buff
*skb
,
742 const struct xt_action_param
*par
, struct ip_set_adt_opt
*opt
)
744 struct ip_set
*set
= ip_set_rcu_get(xt_net(par
), index
);
748 pr_debug("set %s, index %u\n", set
->name
, index
);
750 if (opt
->dim
< set
->type
->dimension
||
751 !(opt
->family
== set
->family
|| set
->family
== NFPROTO_UNSPEC
))
755 ret
= set
->variant
->kadt(set
, skb
, par
, IPSET_TEST
, opt
);
756 rcu_read_unlock_bh();
758 if (ret
== -EAGAIN
) {
759 /* Type requests element to be completed */
760 pr_debug("element must be completed, ADD is triggered\n");
762 set
->variant
->kadt(set
, skb
, par
, IPSET_ADD
, opt
);
766 /* --return-nomatch: invert matched element */
767 if ((opt
->cmdflags
& IPSET_FLAG_RETURN_NOMATCH
) &&
768 (set
->type
->features
& IPSET_TYPE_NOMATCH
) &&
769 (ret
> 0 || ret
== -ENOTEMPTY
))
773 /* Convert error codes to nomatch */
774 return (ret
< 0 ? 0 : ret
);
776 EXPORT_SYMBOL_GPL(ip_set_test
);
779 ip_set_add(ip_set_id_t index
, const struct sk_buff
*skb
,
780 const struct xt_action_param
*par
, struct ip_set_adt_opt
*opt
)
782 struct ip_set
*set
= ip_set_rcu_get(xt_net(par
), index
);
786 pr_debug("set %s, index %u\n", set
->name
, index
);
788 if (opt
->dim
< set
->type
->dimension
||
789 !(opt
->family
== set
->family
|| set
->family
== NFPROTO_UNSPEC
))
790 return -IPSET_ERR_TYPE_MISMATCH
;
793 ret
= set
->variant
->kadt(set
, skb
, par
, IPSET_ADD
, opt
);
798 EXPORT_SYMBOL_GPL(ip_set_add
);
801 ip_set_del(ip_set_id_t index
, const struct sk_buff
*skb
,
802 const struct xt_action_param
*par
, struct ip_set_adt_opt
*opt
)
804 struct ip_set
*set
= ip_set_rcu_get(xt_net(par
), index
);
808 pr_debug("set %s, index %u\n", set
->name
, index
);
810 if (opt
->dim
< set
->type
->dimension
||
811 !(opt
->family
== set
->family
|| set
->family
== NFPROTO_UNSPEC
))
812 return -IPSET_ERR_TYPE_MISMATCH
;
815 ret
= set
->variant
->kadt(set
, skb
, par
, IPSET_DEL
, opt
);
820 EXPORT_SYMBOL_GPL(ip_set_del
);
822 /* Find set by name, reference it once. The reference makes sure the
823 * thing pointed to, does not go away under our feet.
827 ip_set_get_byname(struct net
*net
, const char *name
, struct ip_set
**set
)
829 ip_set_id_t i
, index
= IPSET_INVALID_ID
;
831 struct ip_set_net
*inst
= ip_set_pernet(net
);
834 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
835 s
= rcu_dereference(inst
->ip_set_list
)[i
];
836 if (s
&& STRNCMP(s
->name
, name
)) {
847 EXPORT_SYMBOL_GPL(ip_set_get_byname
);
849 /* If the given set pointer points to a valid set, decrement
850 * reference count by 1. The caller shall not assume the index
851 * to be valid, after calling this function.
856 __ip_set_put_byindex(struct ip_set_net
*inst
, ip_set_id_t index
)
861 set
= rcu_dereference(inst
->ip_set_list
)[index
];
868 ip_set_put_byindex(struct net
*net
, ip_set_id_t index
)
870 struct ip_set_net
*inst
= ip_set_pernet(net
);
872 __ip_set_put_byindex(inst
, index
);
874 EXPORT_SYMBOL_GPL(ip_set_put_byindex
);
876 /* Get the name of a set behind a set index.
877 * Set itself is protected by RCU, but its name isn't: to protect against
878 * renaming, grab ip_set_ref_lock as reader (see ip_set_rename()) and copy the
882 ip_set_name_byindex(struct net
*net
, ip_set_id_t index
, char *name
)
884 struct ip_set
*set
= ip_set_rcu_get(net
, index
);
888 read_lock_bh(&ip_set_ref_lock
);
889 strncpy(name
, set
->name
, IPSET_MAXNAMELEN
);
890 read_unlock_bh(&ip_set_ref_lock
);
892 EXPORT_SYMBOL_GPL(ip_set_name_byindex
);
894 /* Routines to call by external subsystems, which do not
895 * call nfnl_lock for us.
898 /* Find set by index, reference it once. The reference makes sure the
899 * thing pointed to, does not go away under our feet.
901 * The nfnl mutex is used in the function.
904 ip_set_nfnl_get_byindex(struct net
*net
, ip_set_id_t index
)
907 struct ip_set_net
*inst
= ip_set_pernet(net
);
909 if (index
>= inst
->ip_set_max
)
910 return IPSET_INVALID_ID
;
912 nfnl_lock(NFNL_SUBSYS_IPSET
);
913 set
= ip_set(inst
, index
);
917 index
= IPSET_INVALID_ID
;
918 nfnl_unlock(NFNL_SUBSYS_IPSET
);
922 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex
);
924 /* If the given set pointer points to a valid set, decrement
925 * reference count by 1. The caller shall not assume the index
926 * to be valid, after calling this function.
928 * The nfnl mutex is used in the function.
931 ip_set_nfnl_put(struct net
*net
, ip_set_id_t index
)
934 struct ip_set_net
*inst
= ip_set_pernet(net
);
936 nfnl_lock(NFNL_SUBSYS_IPSET
);
937 if (!inst
->is_deleted
) { /* already deleted from ip_set_net_exit() */
938 set
= ip_set(inst
, index
);
942 nfnl_unlock(NFNL_SUBSYS_IPSET
);
944 EXPORT_SYMBOL_GPL(ip_set_nfnl_put
);
946 /* Communication protocol with userspace over netlink.
948 * The commands are serialized by the nfnl mutex.
951 static inline u8
protocol(const struct nlattr
* const tb
[])
953 return nla_get_u8(tb
[IPSET_ATTR_PROTOCOL
]);
957 protocol_failed(const struct nlattr
* const tb
[])
959 return !tb
[IPSET_ATTR_PROTOCOL
] || protocol(tb
) != IPSET_PROTOCOL
;
963 protocol_min_failed(const struct nlattr
* const tb
[])
965 return !tb
[IPSET_ATTR_PROTOCOL
] || protocol(tb
) < IPSET_PROTOCOL_MIN
;
969 flag_exist(const struct nlmsghdr
*nlh
)
971 return nlh
->nlmsg_flags
& NLM_F_EXCL
? 0 : IPSET_FLAG_EXIST
;
974 static struct nlmsghdr
*
975 start_msg(struct sk_buff
*skb
, u32 portid
, u32 seq
, unsigned int flags
,
978 struct nlmsghdr
*nlh
;
979 struct nfgenmsg
*nfmsg
;
981 nlh
= nlmsg_put(skb
, portid
, seq
, nfnl_msg_type(NFNL_SUBSYS_IPSET
, cmd
),
982 sizeof(*nfmsg
), flags
);
986 nfmsg
= nlmsg_data(nlh
);
987 nfmsg
->nfgen_family
= NFPROTO_IPV4
;
988 nfmsg
->version
= NFNETLINK_V0
;
996 static const struct nla_policy ip_set_create_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
997 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
998 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
999 .len
= IPSET_MAXNAMELEN
- 1 },
1000 [IPSET_ATTR_TYPENAME
] = { .type
= NLA_NUL_STRING
,
1001 .len
= IPSET_MAXNAMELEN
- 1},
1002 [IPSET_ATTR_REVISION
] = { .type
= NLA_U8
},
1003 [IPSET_ATTR_FAMILY
] = { .type
= NLA_U8
},
1004 [IPSET_ATTR_DATA
] = { .type
= NLA_NESTED
},
1007 static struct ip_set
*
1008 find_set_and_id(struct ip_set_net
*inst
, const char *name
, ip_set_id_t
*id
)
1010 struct ip_set
*set
= NULL
;
1013 *id
= IPSET_INVALID_ID
;
1014 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1015 set
= ip_set(inst
, i
);
1016 if (set
&& STRNCMP(set
->name
, name
)) {
1021 return (*id
== IPSET_INVALID_ID
? NULL
: set
);
1024 static inline struct ip_set
*
1025 find_set(struct ip_set_net
*inst
, const char *name
)
1029 return find_set_and_id(inst
, name
, &id
);
1033 find_free_id(struct ip_set_net
*inst
, const char *name
, ip_set_id_t
*index
,
1034 struct ip_set
**set
)
1039 *index
= IPSET_INVALID_ID
;
1040 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1041 s
= ip_set(inst
, i
);
1043 if (*index
== IPSET_INVALID_ID
)
1045 } else if (STRNCMP(name
, s
->name
)) {
1051 if (*index
== IPSET_INVALID_ID
)
1052 /* No free slot remained */
1053 return -IPSET_ERR_MAX_SETS
;
1057 static int ip_set_none(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1058 const struct nlmsghdr
*nlh
,
1059 const struct nlattr
* const attr
[],
1060 struct netlink_ext_ack
*extack
)
1065 static int ip_set_create(struct net
*net
, struct sock
*ctnl
,
1066 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1067 const struct nlattr
* const attr
[],
1068 struct netlink_ext_ack
*extack
)
1070 struct ip_set_net
*inst
= ip_set_pernet(net
);
1071 struct ip_set
*set
, *clash
= NULL
;
1072 ip_set_id_t index
= IPSET_INVALID_ID
;
1073 struct nlattr
*tb
[IPSET_ATTR_CREATE_MAX
+ 1] = {};
1074 const char *name
, *typename
;
1075 u8 family
, revision
;
1076 u32 flags
= flag_exist(nlh
);
1079 if (unlikely(protocol_min_failed(attr
) ||
1080 !attr
[IPSET_ATTR_SETNAME
] ||
1081 !attr
[IPSET_ATTR_TYPENAME
] ||
1082 !attr
[IPSET_ATTR_REVISION
] ||
1083 !attr
[IPSET_ATTR_FAMILY
] ||
1084 (attr
[IPSET_ATTR_DATA
] &&
1085 !flag_nested(attr
[IPSET_ATTR_DATA
]))))
1086 return -IPSET_ERR_PROTOCOL
;
1088 name
= nla_data(attr
[IPSET_ATTR_SETNAME
]);
1089 typename
= nla_data(attr
[IPSET_ATTR_TYPENAME
]);
1090 family
= nla_get_u8(attr
[IPSET_ATTR_FAMILY
]);
1091 revision
= nla_get_u8(attr
[IPSET_ATTR_REVISION
]);
1092 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
1093 name
, typename
, family_name(family
), revision
);
1095 /* First, and without any locks, allocate and initialize
1096 * a normal base set structure.
1098 set
= kzalloc(sizeof(*set
), GFP_KERNEL
);
1101 spin_lock_init(&set
->lock
);
1102 strlcpy(set
->name
, name
, IPSET_MAXNAMELEN
);
1103 set
->family
= family
;
1104 set
->revision
= revision
;
1106 /* Next, check that we know the type, and take
1107 * a reference on the type, to make sure it stays available
1108 * while constructing our new set.
1110 * After referencing the type, we try to create the type
1111 * specific part of the set without holding any locks.
1113 ret
= find_set_type_get(typename
, family
, revision
, &set
->type
);
1117 /* Without holding any locks, create private part. */
1118 if (attr
[IPSET_ATTR_DATA
] &&
1119 nla_parse_nested(tb
, IPSET_ATTR_CREATE_MAX
, attr
[IPSET_ATTR_DATA
],
1120 set
->type
->create_policy
, NULL
)) {
1121 ret
= -IPSET_ERR_PROTOCOL
;
1125 ret
= set
->type
->create(net
, set
, tb
, flags
);
1129 /* BTW, ret==0 here. */
1131 /* Here, we have a valid, constructed set and we are protected
1132 * by the nfnl mutex. Find the first free index in ip_set_list
1133 * and check clashing.
1135 ret
= find_free_id(inst
, set
->name
, &index
, &clash
);
1136 if (ret
== -EEXIST
) {
1137 /* If this is the same set and requested, ignore error */
1138 if ((flags
& IPSET_FLAG_EXIST
) &&
1139 STRNCMP(set
->type
->name
, clash
->type
->name
) &&
1140 set
->type
->family
== clash
->type
->family
&&
1141 set
->type
->revision_min
== clash
->type
->revision_min
&&
1142 set
->type
->revision_max
== clash
->type
->revision_max
&&
1143 set
->variant
->same_set(set
, clash
))
1146 } else if (ret
== -IPSET_ERR_MAX_SETS
) {
1147 struct ip_set
**list
, **tmp
;
1148 ip_set_id_t i
= inst
->ip_set_max
+ IP_SET_INC
;
1150 if (i
< inst
->ip_set_max
|| i
== IPSET_INVALID_ID
)
1154 list
= kvcalloc(i
, sizeof(struct ip_set
*), GFP_KERNEL
);
1157 /* nfnl mutex is held, both lists are valid */
1158 tmp
= ip_set_dereference(inst
->ip_set_list
);
1159 memcpy(list
, tmp
, sizeof(struct ip_set
*) * inst
->ip_set_max
);
1160 rcu_assign_pointer(inst
->ip_set_list
, list
);
1161 /* Make sure all current packets have passed through */
1164 index
= inst
->ip_set_max
;
1165 inst
->ip_set_max
= i
;
1172 /* Finally! Add our shiny new set to the list, and be done. */
1173 pr_debug("create: '%s' created with index %u!\n", set
->name
, index
);
1174 ip_set(inst
, index
) = set
;
1179 set
->variant
->destroy(set
);
1181 module_put(set
->type
->me
);
1189 static const struct nla_policy
1190 ip_set_setname_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1191 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1192 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
1193 .len
= IPSET_MAXNAMELEN
- 1 },
1197 ip_set_destroy_set(struct ip_set
*set
)
1199 pr_debug("set: %s\n", set
->name
);
1201 /* Must call it without holding any lock */
1202 set
->variant
->destroy(set
);
1203 module_put(set
->type
->me
);
1207 static int ip_set_destroy(struct net
*net
, struct sock
*ctnl
,
1208 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1209 const struct nlattr
* const attr
[],
1210 struct netlink_ext_ack
*extack
)
1212 struct ip_set_net
*inst
= ip_set_pernet(net
);
1217 if (unlikely(protocol_min_failed(attr
)))
1218 return -IPSET_ERR_PROTOCOL
;
1220 /* Must wait for flush to be really finished in list:set */
1223 /* Commands are serialized and references are
1224 * protected by the ip_set_ref_lock.
1225 * External systems (i.e. xt_set) must call
1226 * ip_set_put|get_nfnl_* functions, that way we
1227 * can safely check references here.
1229 * list:set timer can only decrement the reference
1230 * counter, so if it's already zero, we can proceed
1231 * without holding the lock.
1233 read_lock_bh(&ip_set_ref_lock
);
1234 if (!attr
[IPSET_ATTR_SETNAME
]) {
1235 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1236 s
= ip_set(inst
, i
);
1237 if (s
&& (s
->ref
|| s
->ref_netlink
)) {
1238 ret
= -IPSET_ERR_BUSY
;
1242 inst
->is_destroyed
= true;
1243 read_unlock_bh(&ip_set_ref_lock
);
1244 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1245 s
= ip_set(inst
, i
);
1247 ip_set(inst
, i
) = NULL
;
1248 ip_set_destroy_set(s
);
1251 /* Modified by ip_set_destroy() only, which is serialized */
1252 inst
->is_destroyed
= false;
1254 s
= find_set_and_id(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]),
1259 } else if (s
->ref
|| s
->ref_netlink
) {
1260 ret
= -IPSET_ERR_BUSY
;
1263 ip_set(inst
, i
) = NULL
;
1264 read_unlock_bh(&ip_set_ref_lock
);
1266 ip_set_destroy_set(s
);
1270 read_unlock_bh(&ip_set_ref_lock
);
1277 ip_set_flush_set(struct ip_set
*set
)
1279 pr_debug("set: %s\n", set
->name
);
1282 set
->variant
->flush(set
);
1286 static int ip_set_flush(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1287 const struct nlmsghdr
*nlh
,
1288 const struct nlattr
* const attr
[],
1289 struct netlink_ext_ack
*extack
)
1291 struct ip_set_net
*inst
= ip_set_pernet(net
);
1295 if (unlikely(protocol_min_failed(attr
)))
1296 return -IPSET_ERR_PROTOCOL
;
1298 if (!attr
[IPSET_ATTR_SETNAME
]) {
1299 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1300 s
= ip_set(inst
, i
);
1302 ip_set_flush_set(s
);
1305 s
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1309 ip_set_flush_set(s
);
1317 static const struct nla_policy
1318 ip_set_setname2_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1319 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1320 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
1321 .len
= IPSET_MAXNAMELEN
- 1 },
1322 [IPSET_ATTR_SETNAME2
] = { .type
= NLA_NUL_STRING
,
1323 .len
= IPSET_MAXNAMELEN
- 1 },
1326 static int ip_set_rename(struct net
*net
, struct sock
*ctnl
,
1327 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1328 const struct nlattr
* const attr
[],
1329 struct netlink_ext_ack
*extack
)
1331 struct ip_set_net
*inst
= ip_set_pernet(net
);
1332 struct ip_set
*set
, *s
;
1337 if (unlikely(protocol_min_failed(attr
) ||
1338 !attr
[IPSET_ATTR_SETNAME
] ||
1339 !attr
[IPSET_ATTR_SETNAME2
]))
1340 return -IPSET_ERR_PROTOCOL
;
1342 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1346 write_lock_bh(&ip_set_ref_lock
);
1347 if (set
->ref
!= 0 || set
->ref_netlink
!= 0) {
1348 ret
= -IPSET_ERR_REFERENCED
;
1352 name2
= nla_data(attr
[IPSET_ATTR_SETNAME2
]);
1353 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1354 s
= ip_set(inst
, i
);
1355 if (s
&& STRNCMP(s
->name
, name2
)) {
1356 ret
= -IPSET_ERR_EXIST_SETNAME2
;
1360 strncpy(set
->name
, name2
, IPSET_MAXNAMELEN
);
1363 write_unlock_bh(&ip_set_ref_lock
);
1367 /* Swap two sets so that name/index points to the other.
1368 * References and set names are also swapped.
1370 * The commands are serialized by the nfnl mutex and references are
1371 * protected by the ip_set_ref_lock. The kernel interfaces
1372 * do not hold the mutex but the pointer settings are atomic
1373 * so the ip_set_list always contains valid pointers to the sets.
1376 static int ip_set_swap(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1377 const struct nlmsghdr
*nlh
,
1378 const struct nlattr
* const attr
[],
1379 struct netlink_ext_ack
*extack
)
1381 struct ip_set_net
*inst
= ip_set_pernet(net
);
1382 struct ip_set
*from
, *to
;
1383 ip_set_id_t from_id
, to_id
;
1384 char from_name
[IPSET_MAXNAMELEN
];
1386 if (unlikely(protocol_min_failed(attr
) ||
1387 !attr
[IPSET_ATTR_SETNAME
] ||
1388 !attr
[IPSET_ATTR_SETNAME2
]))
1389 return -IPSET_ERR_PROTOCOL
;
1391 from
= find_set_and_id(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]),
1396 to
= find_set_and_id(inst
, nla_data(attr
[IPSET_ATTR_SETNAME2
]),
1399 return -IPSET_ERR_EXIST_SETNAME2
;
1401 /* Features must not change.
1402 * Not an artifical restriction anymore, as we must prevent
1403 * possible loops created by swapping in setlist type of sets.
1405 if (!(from
->type
->features
== to
->type
->features
&&
1406 from
->family
== to
->family
))
1407 return -IPSET_ERR_TYPE_MISMATCH
;
1409 write_lock_bh(&ip_set_ref_lock
);
1411 if (from
->ref_netlink
|| to
->ref_netlink
) {
1412 write_unlock_bh(&ip_set_ref_lock
);
1416 strncpy(from_name
, from
->name
, IPSET_MAXNAMELEN
);
1417 strncpy(from
->name
, to
->name
, IPSET_MAXNAMELEN
);
1418 strncpy(to
->name
, from_name
, IPSET_MAXNAMELEN
);
1420 swap(from
->ref
, to
->ref
);
1421 ip_set(inst
, from_id
) = to
;
1422 ip_set(inst
, to_id
) = from
;
1423 write_unlock_bh(&ip_set_ref_lock
);
1428 /* List/save set data */
1435 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1436 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1439 ip_set_put_flags(struct sk_buff
*skb
, struct ip_set
*set
)
1443 if (SET_WITH_TIMEOUT(set
))
1444 if (unlikely(nla_put_net32(skb
, IPSET_ATTR_TIMEOUT
,
1445 htonl(set
->timeout
))))
1447 if (SET_WITH_COUNTER(set
))
1448 cadt_flags
|= IPSET_FLAG_WITH_COUNTERS
;
1449 if (SET_WITH_COMMENT(set
))
1450 cadt_flags
|= IPSET_FLAG_WITH_COMMENT
;
1451 if (SET_WITH_SKBINFO(set
))
1452 cadt_flags
|= IPSET_FLAG_WITH_SKBINFO
;
1453 if (SET_WITH_FORCEADD(set
))
1454 cadt_flags
|= IPSET_FLAG_WITH_FORCEADD
;
1458 return nla_put_net32(skb
, IPSET_ATTR_CADT_FLAGS
, htonl(cadt_flags
));
1460 EXPORT_SYMBOL_GPL(ip_set_put_flags
);
1463 ip_set_dump_done(struct netlink_callback
*cb
)
1465 if (cb
->args
[IPSET_CB_ARG0
]) {
1466 struct ip_set_net
*inst
=
1467 (struct ip_set_net
*)cb
->args
[IPSET_CB_NET
];
1468 ip_set_id_t index
= (ip_set_id_t
)cb
->args
[IPSET_CB_INDEX
];
1469 struct ip_set
*set
= ip_set_ref_netlink(inst
, index
);
1471 if (set
->variant
->uref
)
1472 set
->variant
->uref(set
, cb
, false);
1473 pr_debug("release set %s\n", set
->name
);
1474 __ip_set_put_netlink(set
);
1480 dump_attrs(struct nlmsghdr
*nlh
)
1482 const struct nlattr
*attr
;
1485 pr_debug("dump nlmsg\n");
1486 nlmsg_for_each_attr(attr
, nlh
, sizeof(struct nfgenmsg
), rem
) {
1487 pr_debug("type: %u, len %u\n", nla_type(attr
), attr
->nla_len
);
1491 static const struct nla_policy
1492 ip_set_dump_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1493 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1494 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
1495 .len
= IPSET_MAXNAMELEN
- 1 },
1496 [IPSET_ATTR_FLAGS
] = { .type
= NLA_U32
},
1500 ip_set_dump_start(struct netlink_callback
*cb
)
1502 struct nlmsghdr
*nlh
= nlmsg_hdr(cb
->skb
);
1503 int min_len
= nlmsg_total_size(sizeof(struct nfgenmsg
));
1504 struct nlattr
*cda
[IPSET_ATTR_CMD_MAX
+ 1];
1505 struct nlattr
*attr
= (void *)nlh
+ min_len
;
1506 struct sk_buff
*skb
= cb
->skb
;
1507 struct ip_set_net
*inst
= ip_set_pernet(sock_net(skb
->sk
));
1511 ret
= nla_parse(cda
, IPSET_ATTR_CMD_MAX
, attr
,
1512 nlh
->nlmsg_len
- min_len
,
1513 ip_set_dump_policy
, NULL
);
1517 cb
->args
[IPSET_CB_PROTO
] = nla_get_u8(cda
[IPSET_ATTR_PROTOCOL
]);
1518 if (cda
[IPSET_ATTR_SETNAME
]) {
1522 set
= find_set_and_id(inst
, nla_data(cda
[IPSET_ATTR_SETNAME
]),
1528 dump_type
= DUMP_ONE
;
1529 cb
->args
[IPSET_CB_INDEX
] = index
;
1531 dump_type
= DUMP_ALL
;
1534 if (cda
[IPSET_ATTR_FLAGS
]) {
1535 u32 f
= ip_set_get_h32(cda
[IPSET_ATTR_FLAGS
]);
1537 dump_type
|= (f
<< 16);
1539 cb
->args
[IPSET_CB_NET
] = (unsigned long)inst
;
1540 cb
->args
[IPSET_CB_DUMP
] = dump_type
;
1545 /* We have to create and send the error message manually :-( */
1546 if (nlh
->nlmsg_flags
& NLM_F_ACK
) {
1547 netlink_ack(cb
->skb
, nlh
, ret
, NULL
);
1553 ip_set_dump_do(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1555 ip_set_id_t index
= IPSET_INVALID_ID
, max
;
1556 struct ip_set
*set
= NULL
;
1557 struct nlmsghdr
*nlh
= NULL
;
1558 unsigned int flags
= NETLINK_CB(cb
->skb
).portid
? NLM_F_MULTI
: 0;
1559 struct ip_set_net
*inst
= ip_set_pernet(sock_net(skb
->sk
));
1560 u32 dump_type
, dump_flags
;
1564 if (!cb
->args
[IPSET_CB_DUMP
])
1567 if (cb
->args
[IPSET_CB_INDEX
] >= inst
->ip_set_max
)
1570 dump_type
= DUMP_TYPE(cb
->args
[IPSET_CB_DUMP
]);
1571 dump_flags
= DUMP_FLAGS(cb
->args
[IPSET_CB_DUMP
]);
1572 max
= dump_type
== DUMP_ONE
? cb
->args
[IPSET_CB_INDEX
] + 1
1575 pr_debug("dump type, flag: %u %u index: %ld\n",
1576 dump_type
, dump_flags
, cb
->args
[IPSET_CB_INDEX
]);
1577 for (; cb
->args
[IPSET_CB_INDEX
] < max
; cb
->args
[IPSET_CB_INDEX
]++) {
1578 index
= (ip_set_id_t
)cb
->args
[IPSET_CB_INDEX
];
1579 write_lock_bh(&ip_set_ref_lock
);
1580 set
= ip_set(inst
, index
);
1581 is_destroyed
= inst
->is_destroyed
;
1582 if (!set
|| is_destroyed
) {
1583 write_unlock_bh(&ip_set_ref_lock
);
1584 if (dump_type
== DUMP_ONE
) {
1589 /* All sets are just being destroyed */
1595 /* When dumping all sets, we must dump "sorted"
1596 * so that lists (unions of sets) are dumped last.
1598 if (dump_type
!= DUMP_ONE
&&
1599 ((dump_type
== DUMP_ALL
) ==
1600 !!(set
->type
->features
& IPSET_DUMP_LAST
))) {
1601 write_unlock_bh(&ip_set_ref_lock
);
1604 pr_debug("List set: %s\n", set
->name
);
1605 if (!cb
->args
[IPSET_CB_ARG0
]) {
1606 /* Start listing: make sure set won't be destroyed */
1607 pr_debug("reference set\n");
1610 write_unlock_bh(&ip_set_ref_lock
);
1611 nlh
= start_msg(skb
, NETLINK_CB(cb
->skb
).portid
,
1612 cb
->nlh
->nlmsg_seq
, flags
,
1616 goto release_refcount
;
1618 if (nla_put_u8(skb
, IPSET_ATTR_PROTOCOL
,
1619 cb
->args
[IPSET_CB_PROTO
]) ||
1620 nla_put_string(skb
, IPSET_ATTR_SETNAME
, set
->name
))
1621 goto nla_put_failure
;
1622 if (dump_flags
& IPSET_FLAG_LIST_SETNAME
)
1624 switch (cb
->args
[IPSET_CB_ARG0
]) {
1626 /* Core header data */
1627 if (nla_put_string(skb
, IPSET_ATTR_TYPENAME
,
1629 nla_put_u8(skb
, IPSET_ATTR_FAMILY
,
1631 nla_put_u8(skb
, IPSET_ATTR_REVISION
,
1633 goto nla_put_failure
;
1634 if (cb
->args
[IPSET_CB_PROTO
] > IPSET_PROTOCOL_MIN
&&
1635 nla_put_net16(skb
, IPSET_ATTR_INDEX
, htons(index
)))
1636 goto nla_put_failure
;
1637 ret
= set
->variant
->head(set
, skb
);
1639 goto release_refcount
;
1640 if (dump_flags
& IPSET_FLAG_LIST_HEADER
)
1642 if (set
->variant
->uref
)
1643 set
->variant
->uref(set
, cb
, true);
1646 ret
= set
->variant
->list(set
, skb
, cb
);
1647 if (!cb
->args
[IPSET_CB_ARG0
])
1648 /* Set is done, proceed with next one */
1650 goto release_refcount
;
1653 /* If we dump all sets, continue with dumping last ones */
1654 if (dump_type
== DUMP_ALL
) {
1655 dump_type
= DUMP_LAST
;
1656 cb
->args
[IPSET_CB_DUMP
] = dump_type
| (dump_flags
<< 16);
1657 cb
->args
[IPSET_CB_INDEX
] = 0;
1658 if (set
&& set
->variant
->uref
)
1659 set
->variant
->uref(set
, cb
, false);
1667 if (dump_type
== DUMP_ONE
)
1668 cb
->args
[IPSET_CB_INDEX
] = IPSET_INVALID_ID
;
1670 cb
->args
[IPSET_CB_INDEX
]++;
1672 /* If there was an error or set is done, release set */
1673 if (ret
|| !cb
->args
[IPSET_CB_ARG0
]) {
1674 set
= ip_set_ref_netlink(inst
, index
);
1675 if (set
->variant
->uref
)
1676 set
->variant
->uref(set
, cb
, false);
1677 pr_debug("release set %s\n", set
->name
);
1678 __ip_set_put_netlink(set
);
1679 cb
->args
[IPSET_CB_ARG0
] = 0;
1683 nlmsg_end(skb
, nlh
);
1684 pr_debug("nlmsg_len: %u\n", nlh
->nlmsg_len
);
1688 return ret
< 0 ? ret
: skb
->len
;
1691 static int ip_set_dump(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1692 const struct nlmsghdr
*nlh
,
1693 const struct nlattr
* const attr
[],
1694 struct netlink_ext_ack
*extack
)
1696 if (unlikely(protocol_min_failed(attr
)))
1697 return -IPSET_ERR_PROTOCOL
;
1700 struct netlink_dump_control c
= {
1701 .start
= ip_set_dump_start
,
1702 .dump
= ip_set_dump_do
,
1703 .done
= ip_set_dump_done
,
1705 return netlink_dump_start(ctnl
, skb
, nlh
, &c
);
1709 /* Add, del and test */
1711 static const struct nla_policy ip_set_adt_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1712 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1713 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
1714 .len
= IPSET_MAXNAMELEN
- 1 },
1715 [IPSET_ATTR_LINENO
] = { .type
= NLA_U32
},
1716 [IPSET_ATTR_DATA
] = { .type
= NLA_NESTED
},
1717 [IPSET_ATTR_ADT
] = { .type
= NLA_NESTED
},
1721 call_ad(struct sock
*ctnl
, struct sk_buff
*skb
, struct ip_set
*set
,
1722 struct nlattr
*tb
[], enum ipset_adt adt
,
1723 u32 flags
, bool use_lineno
)
1727 bool eexist
= flags
& IPSET_FLAG_EXIST
, retried
= false;
1731 ret
= set
->variant
->uadt(set
, tb
, adt
, &lineno
, flags
, retried
);
1734 } while (ret
== -EAGAIN
&&
1735 set
->variant
->resize
&&
1736 (ret
= set
->variant
->resize(set
, retried
)) == 0);
1738 if (!ret
|| (ret
== -IPSET_ERR_EXIST
&& eexist
))
1740 if (lineno
&& use_lineno
) {
1741 /* Error in restore/batch mode: send back lineno */
1742 struct nlmsghdr
*rep
, *nlh
= nlmsg_hdr(skb
);
1743 struct sk_buff
*skb2
;
1744 struct nlmsgerr
*errmsg
;
1745 size_t payload
= min(SIZE_MAX
,
1746 sizeof(*errmsg
) + nlmsg_len(nlh
));
1747 int min_len
= nlmsg_total_size(sizeof(struct nfgenmsg
));
1748 struct nlattr
*cda
[IPSET_ATTR_CMD_MAX
+ 1];
1749 struct nlattr
*cmdattr
;
1752 skb2
= nlmsg_new(payload
, GFP_KERNEL
);
1755 rep
= __nlmsg_put(skb2
, NETLINK_CB(skb
).portid
,
1756 nlh
->nlmsg_seq
, NLMSG_ERROR
, payload
, 0);
1757 errmsg
= nlmsg_data(rep
);
1758 errmsg
->error
= ret
;
1759 memcpy(&errmsg
->msg
, nlh
, nlh
->nlmsg_len
);
1760 cmdattr
= (void *)&errmsg
->msg
+ min_len
;
1762 ret
= nla_parse(cda
, IPSET_ATTR_CMD_MAX
, cmdattr
,
1763 nlh
->nlmsg_len
- min_len
, ip_set_adt_policy
,
1770 errline
= nla_data(cda
[IPSET_ATTR_LINENO
]);
1774 netlink_unicast(ctnl
, skb2
, NETLINK_CB(skb
).portid
,
1776 /* Signal netlink not to send its ACK/errmsg. */
1783 static int ip_set_ad(struct net
*net
, struct sock
*ctnl
,
1784 struct sk_buff
*skb
,
1786 const struct nlmsghdr
*nlh
,
1787 const struct nlattr
* const attr
[],
1788 struct netlink_ext_ack
*extack
)
1790 struct ip_set_net
*inst
= ip_set_pernet(net
);
1792 struct nlattr
*tb
[IPSET_ATTR_ADT_MAX
+ 1] = {};
1793 const struct nlattr
*nla
;
1794 u32 flags
= flag_exist(nlh
);
1798 if (unlikely(protocol_min_failed(attr
) ||
1799 !attr
[IPSET_ATTR_SETNAME
] ||
1800 !((attr
[IPSET_ATTR_DATA
] != NULL
) ^
1801 (attr
[IPSET_ATTR_ADT
] != NULL
)) ||
1802 (attr
[IPSET_ATTR_DATA
] &&
1803 !flag_nested(attr
[IPSET_ATTR_DATA
])) ||
1804 (attr
[IPSET_ATTR_ADT
] &&
1805 (!flag_nested(attr
[IPSET_ATTR_ADT
]) ||
1806 !attr
[IPSET_ATTR_LINENO
]))))
1807 return -IPSET_ERR_PROTOCOL
;
1809 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1813 use_lineno
= !!attr
[IPSET_ATTR_LINENO
];
1814 if (attr
[IPSET_ATTR_DATA
]) {
1815 if (nla_parse_nested(tb
, IPSET_ATTR_ADT_MAX
,
1816 attr
[IPSET_ATTR_DATA
],
1817 set
->type
->adt_policy
, NULL
))
1818 return -IPSET_ERR_PROTOCOL
;
1819 ret
= call_ad(ctnl
, skb
, set
, tb
, adt
, flags
,
1824 nla_for_each_nested(nla
, attr
[IPSET_ATTR_ADT
], nla_rem
) {
1825 if (nla_type(nla
) != IPSET_ATTR_DATA
||
1826 !flag_nested(nla
) ||
1827 nla_parse_nested(tb
, IPSET_ATTR_ADT_MAX
, nla
,
1828 set
->type
->adt_policy
, NULL
))
1829 return -IPSET_ERR_PROTOCOL
;
1830 ret
= call_ad(ctnl
, skb
, set
, tb
, adt
,
1839 static int ip_set_uadd(struct net
*net
, struct sock
*ctnl
,
1840 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1841 const struct nlattr
* const attr
[],
1842 struct netlink_ext_ack
*extack
)
1844 return ip_set_ad(net
, ctnl
, skb
,
1845 IPSET_ADD
, nlh
, attr
, extack
);
1848 static int ip_set_udel(struct net
*net
, struct sock
*ctnl
,
1849 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1850 const struct nlattr
* const attr
[],
1851 struct netlink_ext_ack
*extack
)
1853 return ip_set_ad(net
, ctnl
, skb
,
1854 IPSET_DEL
, nlh
, attr
, extack
);
1857 static int ip_set_utest(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1858 const struct nlmsghdr
*nlh
,
1859 const struct nlattr
* const attr
[],
1860 struct netlink_ext_ack
*extack
)
1862 struct ip_set_net
*inst
= ip_set_pernet(net
);
1864 struct nlattr
*tb
[IPSET_ATTR_ADT_MAX
+ 1] = {};
1868 if (unlikely(protocol_min_failed(attr
) ||
1869 !attr
[IPSET_ATTR_SETNAME
] ||
1870 !attr
[IPSET_ATTR_DATA
] ||
1871 !flag_nested(attr
[IPSET_ATTR_DATA
])))
1872 return -IPSET_ERR_PROTOCOL
;
1874 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1878 if (nla_parse_nested(tb
, IPSET_ATTR_ADT_MAX
, attr
[IPSET_ATTR_DATA
],
1879 set
->type
->adt_policy
, NULL
))
1880 return -IPSET_ERR_PROTOCOL
;
1883 ret
= set
->variant
->uadt(set
, tb
, IPSET_TEST
, &lineno
, 0, 0);
1884 rcu_read_unlock_bh();
1885 /* Userspace can't trigger element to be re-added */
1889 return ret
> 0 ? 0 : -IPSET_ERR_EXIST
;
1892 /* Get headed data of a set */
1894 static int ip_set_header(struct net
*net
, struct sock
*ctnl
,
1895 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1896 const struct nlattr
* const attr
[],
1897 struct netlink_ext_ack
*extack
)
1899 struct ip_set_net
*inst
= ip_set_pernet(net
);
1900 const struct ip_set
*set
;
1901 struct sk_buff
*skb2
;
1902 struct nlmsghdr
*nlh2
;
1905 if (unlikely(protocol_min_failed(attr
) ||
1906 !attr
[IPSET_ATTR_SETNAME
]))
1907 return -IPSET_ERR_PROTOCOL
;
1909 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1913 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
1917 nlh2
= start_msg(skb2
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
1921 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, protocol(attr
)) ||
1922 nla_put_string(skb2
, IPSET_ATTR_SETNAME
, set
->name
) ||
1923 nla_put_string(skb2
, IPSET_ATTR_TYPENAME
, set
->type
->name
) ||
1924 nla_put_u8(skb2
, IPSET_ATTR_FAMILY
, set
->family
) ||
1925 nla_put_u8(skb2
, IPSET_ATTR_REVISION
, set
->revision
))
1926 goto nla_put_failure
;
1927 nlmsg_end(skb2
, nlh2
);
1929 ret
= netlink_unicast(ctnl
, skb2
, NETLINK_CB(skb
).portid
, MSG_DONTWAIT
);
1936 nlmsg_cancel(skb2
, nlh2
);
1944 static const struct nla_policy ip_set_type_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1945 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1946 [IPSET_ATTR_TYPENAME
] = { .type
= NLA_NUL_STRING
,
1947 .len
= IPSET_MAXNAMELEN
- 1 },
1948 [IPSET_ATTR_FAMILY
] = { .type
= NLA_U8
},
1951 static int ip_set_type(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1952 const struct nlmsghdr
*nlh
,
1953 const struct nlattr
* const attr
[],
1954 struct netlink_ext_ack
*extack
)
1956 struct sk_buff
*skb2
;
1957 struct nlmsghdr
*nlh2
;
1958 u8 family
, min
, max
;
1959 const char *typename
;
1962 if (unlikely(protocol_min_failed(attr
) ||
1963 !attr
[IPSET_ATTR_TYPENAME
] ||
1964 !attr
[IPSET_ATTR_FAMILY
]))
1965 return -IPSET_ERR_PROTOCOL
;
1967 family
= nla_get_u8(attr
[IPSET_ATTR_FAMILY
]);
1968 typename
= nla_data(attr
[IPSET_ATTR_TYPENAME
]);
1969 ret
= find_set_type_minmax(typename
, family
, &min
, &max
);
1973 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
1977 nlh2
= start_msg(skb2
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
1981 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, protocol(attr
)) ||
1982 nla_put_string(skb2
, IPSET_ATTR_TYPENAME
, typename
) ||
1983 nla_put_u8(skb2
, IPSET_ATTR_FAMILY
, family
) ||
1984 nla_put_u8(skb2
, IPSET_ATTR_REVISION
, max
) ||
1985 nla_put_u8(skb2
, IPSET_ATTR_REVISION_MIN
, min
))
1986 goto nla_put_failure
;
1987 nlmsg_end(skb2
, nlh2
);
1989 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2
->nlmsg_len
);
1990 ret
= netlink_unicast(ctnl
, skb2
, NETLINK_CB(skb
).portid
, MSG_DONTWAIT
);
1997 nlmsg_cancel(skb2
, nlh2
);
2003 /* Get protocol version */
2005 static const struct nla_policy
2006 ip_set_protocol_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
2007 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
2010 static int ip_set_protocol(struct net
*net
, struct sock
*ctnl
,
2011 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
2012 const struct nlattr
* const attr
[],
2013 struct netlink_ext_ack
*extack
)
2015 struct sk_buff
*skb2
;
2016 struct nlmsghdr
*nlh2
;
2019 if (unlikely(!attr
[IPSET_ATTR_PROTOCOL
]))
2020 return -IPSET_ERR_PROTOCOL
;
2022 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
2026 nlh2
= start_msg(skb2
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
2027 IPSET_CMD_PROTOCOL
);
2030 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, IPSET_PROTOCOL
))
2031 goto nla_put_failure
;
2032 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL_MIN
, IPSET_PROTOCOL_MIN
))
2033 goto nla_put_failure
;
2034 nlmsg_end(skb2
, nlh2
);
2036 ret
= netlink_unicast(ctnl
, skb2
, NETLINK_CB(skb
).portid
, MSG_DONTWAIT
);
2043 nlmsg_cancel(skb2
, nlh2
);
2049 /* Get set by name or index, from userspace */
2051 static int ip_set_byname(struct net
*net
, struct sock
*ctnl
,
2052 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
2053 const struct nlattr
* const attr
[],
2054 struct netlink_ext_ack
*extack
)
2056 struct ip_set_net
*inst
= ip_set_pernet(net
);
2057 struct sk_buff
*skb2
;
2058 struct nlmsghdr
*nlh2
;
2059 ip_set_id_t id
= IPSET_INVALID_ID
;
2060 const struct ip_set
*set
;
2063 if (unlikely(protocol_failed(attr
) ||
2064 !attr
[IPSET_ATTR_SETNAME
]))
2065 return -IPSET_ERR_PROTOCOL
;
2067 set
= find_set_and_id(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]), &id
);
2068 if (id
== IPSET_INVALID_ID
)
2071 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
2075 nlh2
= start_msg(skb2
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
2076 IPSET_CMD_GET_BYNAME
);
2079 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, protocol(attr
)) ||
2080 nla_put_u8(skb2
, IPSET_ATTR_FAMILY
, set
->family
) ||
2081 nla_put_net16(skb2
, IPSET_ATTR_INDEX
, htons(id
)))
2082 goto nla_put_failure
;
2083 nlmsg_end(skb2
, nlh2
);
2085 ret
= netlink_unicast(ctnl
, skb2
, NETLINK_CB(skb
).portid
, MSG_DONTWAIT
);
2092 nlmsg_cancel(skb2
, nlh2
);
2098 static const struct nla_policy ip_set_index_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
2099 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
2100 [IPSET_ATTR_INDEX
] = { .type
= NLA_U16
},
2103 static int ip_set_byindex(struct net
*net
, struct sock
*ctnl
,
2104 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
2105 const struct nlattr
* const attr
[],
2106 struct netlink_ext_ack
*extack
)
2108 struct ip_set_net
*inst
= ip_set_pernet(net
);
2109 struct sk_buff
*skb2
;
2110 struct nlmsghdr
*nlh2
;
2111 ip_set_id_t id
= IPSET_INVALID_ID
;
2112 const struct ip_set
*set
;
2115 if (unlikely(protocol_failed(attr
) ||
2116 !attr
[IPSET_ATTR_INDEX
]))
2117 return -IPSET_ERR_PROTOCOL
;
2119 id
= ip_set_get_h16(attr
[IPSET_ATTR_INDEX
]);
2120 if (id
>= inst
->ip_set_max
)
2122 set
= ip_set(inst
, id
);
2126 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
2130 nlh2
= start_msg(skb2
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
2131 IPSET_CMD_GET_BYINDEX
);
2134 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, protocol(attr
)) ||
2135 nla_put_string(skb2
, IPSET_ATTR_SETNAME
, set
->name
))
2136 goto nla_put_failure
;
2137 nlmsg_end(skb2
, nlh2
);
2139 ret
= netlink_unicast(ctnl
, skb2
, NETLINK_CB(skb
).portid
, MSG_DONTWAIT
);
2146 nlmsg_cancel(skb2
, nlh2
);
2152 static const struct nfnl_callback ip_set_netlink_subsys_cb
[IPSET_MSG_MAX
] = {
2153 [IPSET_CMD_NONE
] = {
2154 .call
= ip_set_none
,
2155 .attr_count
= IPSET_ATTR_CMD_MAX
,
2157 [IPSET_CMD_CREATE
] = {
2158 .call
= ip_set_create
,
2159 .attr_count
= IPSET_ATTR_CMD_MAX
,
2160 .policy
= ip_set_create_policy
,
2162 [IPSET_CMD_DESTROY
] = {
2163 .call
= ip_set_destroy
,
2164 .attr_count
= IPSET_ATTR_CMD_MAX
,
2165 .policy
= ip_set_setname_policy
,
2167 [IPSET_CMD_FLUSH
] = {
2168 .call
= ip_set_flush
,
2169 .attr_count
= IPSET_ATTR_CMD_MAX
,
2170 .policy
= ip_set_setname_policy
,
2172 [IPSET_CMD_RENAME
] = {
2173 .call
= ip_set_rename
,
2174 .attr_count
= IPSET_ATTR_CMD_MAX
,
2175 .policy
= ip_set_setname2_policy
,
2177 [IPSET_CMD_SWAP
] = {
2178 .call
= ip_set_swap
,
2179 .attr_count
= IPSET_ATTR_CMD_MAX
,
2180 .policy
= ip_set_setname2_policy
,
2182 [IPSET_CMD_LIST
] = {
2183 .call
= ip_set_dump
,
2184 .attr_count
= IPSET_ATTR_CMD_MAX
,
2185 .policy
= ip_set_dump_policy
,
2187 [IPSET_CMD_SAVE
] = {
2188 .call
= ip_set_dump
,
2189 .attr_count
= IPSET_ATTR_CMD_MAX
,
2190 .policy
= ip_set_setname_policy
,
2193 .call
= ip_set_uadd
,
2194 .attr_count
= IPSET_ATTR_CMD_MAX
,
2195 .policy
= ip_set_adt_policy
,
2198 .call
= ip_set_udel
,
2199 .attr_count
= IPSET_ATTR_CMD_MAX
,
2200 .policy
= ip_set_adt_policy
,
2202 [IPSET_CMD_TEST
] = {
2203 .call
= ip_set_utest
,
2204 .attr_count
= IPSET_ATTR_CMD_MAX
,
2205 .policy
= ip_set_adt_policy
,
2207 [IPSET_CMD_HEADER
] = {
2208 .call
= ip_set_header
,
2209 .attr_count
= IPSET_ATTR_CMD_MAX
,
2210 .policy
= ip_set_setname_policy
,
2212 [IPSET_CMD_TYPE
] = {
2213 .call
= ip_set_type
,
2214 .attr_count
= IPSET_ATTR_CMD_MAX
,
2215 .policy
= ip_set_type_policy
,
2217 [IPSET_CMD_PROTOCOL
] = {
2218 .call
= ip_set_protocol
,
2219 .attr_count
= IPSET_ATTR_CMD_MAX
,
2220 .policy
= ip_set_protocol_policy
,
2222 [IPSET_CMD_GET_BYNAME
] = {
2223 .call
= ip_set_byname
,
2224 .attr_count
= IPSET_ATTR_CMD_MAX
,
2225 .policy
= ip_set_setname_policy
,
2227 [IPSET_CMD_GET_BYINDEX
] = {
2228 .call
= ip_set_byindex
,
2229 .attr_count
= IPSET_ATTR_CMD_MAX
,
2230 .policy
= ip_set_index_policy
,
2234 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly
= {
2236 .subsys_id
= NFNL_SUBSYS_IPSET
,
2237 .cb_count
= IPSET_MSG_MAX
,
2238 .cb
= ip_set_netlink_subsys_cb
,
2241 /* Interface to iptables/ip6tables */
2244 ip_set_sockfn_get(struct sock
*sk
, int optval
, void __user
*user
, int *len
)
2248 int copylen
= *len
, ret
= 0;
2249 struct net
*net
= sock_net(sk
);
2250 struct ip_set_net
*inst
= ip_set_pernet(net
);
2252 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
2254 if (optval
!= SO_IP_SET
)
2256 if (*len
< sizeof(unsigned int))
2259 data
= vmalloc(*len
);
2262 if (copy_from_user(data
, user
, *len
) != 0) {
2268 if (*op
< IP_SET_OP_VERSION
) {
2269 /* Check the version at the beginning of operations */
2270 struct ip_set_req_version
*req_version
= data
;
2272 if (*len
< sizeof(struct ip_set_req_version
)) {
2277 if (req_version
->version
< IPSET_PROTOCOL_MIN
) {
2284 case IP_SET_OP_VERSION
: {
2285 struct ip_set_req_version
*req_version
= data
;
2287 if (*len
!= sizeof(struct ip_set_req_version
)) {
2292 req_version
->version
= IPSET_PROTOCOL
;
2293 if (copy_to_user(user
, req_version
,
2294 sizeof(struct ip_set_req_version
)))
2298 case IP_SET_OP_GET_BYNAME
: {
2299 struct ip_set_req_get_set
*req_get
= data
;
2302 if (*len
!= sizeof(struct ip_set_req_get_set
)) {
2306 req_get
->set
.name
[IPSET_MAXNAMELEN
- 1] = '\0';
2307 nfnl_lock(NFNL_SUBSYS_IPSET
);
2308 find_set_and_id(inst
, req_get
->set
.name
, &id
);
2309 req_get
->set
.index
= id
;
2310 nfnl_unlock(NFNL_SUBSYS_IPSET
);
2313 case IP_SET_OP_GET_FNAME
: {
2314 struct ip_set_req_get_set_family
*req_get
= data
;
2317 if (*len
!= sizeof(struct ip_set_req_get_set_family
)) {
2321 req_get
->set
.name
[IPSET_MAXNAMELEN
- 1] = '\0';
2322 nfnl_lock(NFNL_SUBSYS_IPSET
);
2323 find_set_and_id(inst
, req_get
->set
.name
, &id
);
2324 req_get
->set
.index
= id
;
2325 if (id
!= IPSET_INVALID_ID
)
2326 req_get
->family
= ip_set(inst
, id
)->family
;
2327 nfnl_unlock(NFNL_SUBSYS_IPSET
);
2330 case IP_SET_OP_GET_BYINDEX
: {
2331 struct ip_set_req_get_set
*req_get
= data
;
2334 if (*len
!= sizeof(struct ip_set_req_get_set
) ||
2335 req_get
->set
.index
>= inst
->ip_set_max
) {
2339 nfnl_lock(NFNL_SUBSYS_IPSET
);
2340 set
= ip_set(inst
, req_get
->set
.index
);
2341 ret
= strscpy(req_get
->set
.name
, set
? set
->name
: "",
2343 nfnl_unlock(NFNL_SUBSYS_IPSET
);
2351 } /* end of switch(op) */
2354 if (copy_to_user(user
, data
, copylen
))
2364 static struct nf_sockopt_ops so_set __read_mostly
= {
2366 .get_optmin
= SO_IP_SET
,
2367 .get_optmax
= SO_IP_SET
+ 1,
2368 .get
= ip_set_sockfn_get
,
2369 .owner
= THIS_MODULE
,
2372 static int __net_init
2373 ip_set_net_init(struct net
*net
)
2375 struct ip_set_net
*inst
= ip_set_pernet(net
);
2376 struct ip_set
**list
;
2378 inst
->ip_set_max
= max_sets
? max_sets
: CONFIG_IP_SET_MAX
;
2379 if (inst
->ip_set_max
>= IPSET_INVALID_ID
)
2380 inst
->ip_set_max
= IPSET_INVALID_ID
- 1;
2382 list
= kvcalloc(inst
->ip_set_max
, sizeof(struct ip_set
*), GFP_KERNEL
);
2385 inst
->is_deleted
= false;
2386 inst
->is_destroyed
= false;
2387 rcu_assign_pointer(inst
->ip_set_list
, list
);
2391 static void __net_exit
2392 ip_set_net_exit(struct net
*net
)
2394 struct ip_set_net
*inst
= ip_set_pernet(net
);
2396 struct ip_set
*set
= NULL
;
2399 inst
->is_deleted
= true; /* flag for ip_set_nfnl_put */
2401 nfnl_lock(NFNL_SUBSYS_IPSET
);
2402 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
2403 set
= ip_set(inst
, i
);
2405 ip_set(inst
, i
) = NULL
;
2406 ip_set_destroy_set(set
);
2409 nfnl_unlock(NFNL_SUBSYS_IPSET
);
2410 kvfree(rcu_dereference_protected(inst
->ip_set_list
, 1));
2413 static struct pernet_operations ip_set_net_ops
= {
2414 .init
= ip_set_net_init
,
2415 .exit
= ip_set_net_exit
,
2416 .id
= &ip_set_net_id
,
2417 .size
= sizeof(struct ip_set_net
),
2423 int ret
= register_pernet_subsys(&ip_set_net_ops
);
2426 pr_err("ip_set: cannot register pernet_subsys.\n");
2430 ret
= nfnetlink_subsys_register(&ip_set_netlink_subsys
);
2432 pr_err("ip_set: cannot register with nfnetlink.\n");
2433 unregister_pernet_subsys(&ip_set_net_ops
);
2437 ret
= nf_register_sockopt(&so_set
);
2439 pr_err("SO_SET registry failed: %d\n", ret
);
2440 nfnetlink_subsys_unregister(&ip_set_netlink_subsys
);
2441 unregister_pernet_subsys(&ip_set_net_ops
);
2451 nf_unregister_sockopt(&so_set
);
2452 nfnetlink_subsys_unregister(&ip_set_netlink_subsys
);
2454 unregister_pernet_subsys(&ip_set_net_ops
);
2455 pr_debug("these are the famous last words\n");
2458 module_init(ip_set_init
);
2459 module_exit(ip_set_fini
);
2461 MODULE_DESCRIPTION("ip_set: protocol " __stringify(IPSET_PROTOCOL
));