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>
24 #include <linux/netfilter/ipset/ip_set_compiler.h>
26 static LIST_HEAD(ip_set_type_list
); /* all registered set types */
27 static DEFINE_MUTEX(ip_set_type_mutex
); /* protects ip_set_type_list */
28 static DEFINE_RWLOCK(ip_set_ref_lock
); /* protects the set refs */
31 struct ip_set
* __rcu
*ip_set_list
; /* all individual sets */
32 ip_set_id_t ip_set_max
; /* max number of sets */
33 bool is_deleted
; /* deleted by ip_set_net_exit */
34 bool is_destroyed
; /* all sets are destroyed */
37 static unsigned int ip_set_net_id __read_mostly
;
39 static struct ip_set_net
*ip_set_pernet(struct net
*net
)
41 return net_generic(net
, ip_set_net_id
);
45 #define STRNCMP(a, b) (strncmp(a, b, IPSET_MAXNAMELEN) == 0)
47 static unsigned int max_sets
;
49 module_param(max_sets
, int, 0600);
50 MODULE_PARM_DESC(max_sets
, "maximal number of sets");
51 MODULE_LICENSE("GPL");
52 MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@netfilter.org>");
53 MODULE_DESCRIPTION("ip_set: protocol " __stringify(IPSET_PROTOCOL
));
54 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_IPSET
);
56 /* When the nfnl mutex or ip_set_ref_lock is held: */
57 #define ip_set_dereference(p) \
58 rcu_dereference_protected(p, \
59 lockdep_nfnl_is_held(NFNL_SUBSYS_IPSET) || \
60 lockdep_is_held(&ip_set_ref_lock))
61 #define ip_set(inst, id) \
62 ip_set_dereference((inst)->ip_set_list)[id]
63 #define ip_set_ref_netlink(inst,id) \
64 rcu_dereference_raw((inst)->ip_set_list)[id]
66 /* The set types are implemented in modules and registered set types
67 * can be found in ip_set_type_list. Adding/deleting types is
68 * serialized by ip_set_type_mutex.
72 ip_set_type_lock(void)
74 mutex_lock(&ip_set_type_mutex
);
78 ip_set_type_unlock(void)
80 mutex_unlock(&ip_set_type_mutex
);
83 /* Register and deregister settype */
85 static struct ip_set_type
*
86 find_set_type(const char *name
, u8 family
, u8 revision
)
88 struct ip_set_type
*type
;
90 list_for_each_entry_rcu_compat(type
, &ip_set_type_list
, list
,
91 lockdep_is_held(&ip_set_type_mutex
))
92 if (STRNCMP(type
->name
, name
) &&
93 (type
->family
== family
||
94 type
->family
== NFPROTO_UNSPEC
) &&
95 revision
>= type
->revision_min
&&
96 revision
<= type
->revision_max
)
101 /* Unlock, try to load a set type module and lock again */
103 load_settype(const char *name
)
105 nfnl_unlock(NFNL_SUBSYS_IPSET
);
106 pr_debug("try to load ip_set_%s\n", name
);
107 if (request_module("ip_set_%s", name
) < 0) {
108 pr_warn("Can't find ip_set type %s\n", name
);
109 nfnl_lock(NFNL_SUBSYS_IPSET
);
112 nfnl_lock(NFNL_SUBSYS_IPSET
);
116 /* Find a set type and reference it */
117 #define find_set_type_get(name, family, revision, found) \
118 __find_set_type_get(name, family, revision, found, false)
121 __find_set_type_get(const char *name
, u8 family
, u8 revision
,
122 struct ip_set_type
**found
, bool retry
)
124 struct ip_set_type
*type
;
127 if (retry
&& !load_settype(name
))
128 return -IPSET_ERR_FIND_TYPE
;
131 *found
= find_set_type(name
, family
, revision
);
133 err
= !try_module_get((*found
)->me
) ? -EFAULT
: 0;
136 /* Make sure the type is already loaded
137 * but we don't support the revision
139 list_for_each_entry_rcu(type
, &ip_set_type_list
, list
)
140 if (STRNCMP(type
->name
, name
)) {
141 err
= -IPSET_ERR_FIND_TYPE
;
146 return retry
? -IPSET_ERR_FIND_TYPE
:
147 __find_set_type_get(name
, family
, revision
, found
, true);
154 /* Find a given set type by name and family.
155 * If we succeeded, the supported minimal and maximum revisions are
158 #define find_set_type_minmax(name, family, min, max) \
159 __find_set_type_minmax(name, family, min, max, false)
162 __find_set_type_minmax(const char *name
, u8 family
, u8
*min
, u8
*max
,
165 struct ip_set_type
*type
;
168 if (retry
&& !load_settype(name
))
169 return -IPSET_ERR_FIND_TYPE
;
171 *min
= 255; *max
= 0;
173 list_for_each_entry_rcu(type
, &ip_set_type_list
, list
)
174 if (STRNCMP(type
->name
, name
) &&
175 (type
->family
== family
||
176 type
->family
== NFPROTO_UNSPEC
)) {
178 if (type
->revision_min
< *min
)
179 *min
= type
->revision_min
;
180 if (type
->revision_max
> *max
)
181 *max
= type
->revision_max
;
187 return retry
? -IPSET_ERR_FIND_TYPE
:
188 __find_set_type_minmax(name
, family
, min
, max
, true);
191 #define family_name(f) ((f) == NFPROTO_IPV4 ? "inet" : \
192 (f) == NFPROTO_IPV6 ? "inet6" : "any")
194 /* Register a set type structure. The type is identified by
195 * the unique triple of name, family and revision.
198 ip_set_type_register(struct ip_set_type
*type
)
202 if (type
->protocol
!= IPSET_PROTOCOL
) {
203 pr_warn("ip_set type %s, family %s, revision %u:%u uses wrong protocol version %u (want %u)\n",
204 type
->name
, family_name(type
->family
),
205 type
->revision_min
, type
->revision_max
,
206 type
->protocol
, IPSET_PROTOCOL
);
211 if (find_set_type(type
->name
, type
->family
, type
->revision_min
)) {
213 pr_warn("ip_set type %s, family %s with revision min %u already registered!\n",
214 type
->name
, family_name(type
->family
),
216 ip_set_type_unlock();
219 list_add_rcu(&type
->list
, &ip_set_type_list
);
220 pr_debug("type %s, family %s, revision %u:%u registered.\n",
221 type
->name
, family_name(type
->family
),
222 type
->revision_min
, type
->revision_max
);
223 ip_set_type_unlock();
227 EXPORT_SYMBOL_GPL(ip_set_type_register
);
229 /* Unregister a set type. There's a small race with ip_set_create */
231 ip_set_type_unregister(struct ip_set_type
*type
)
234 if (!find_set_type(type
->name
, type
->family
, type
->revision_min
)) {
235 pr_warn("ip_set type %s, family %s with revision min %u not registered\n",
236 type
->name
, family_name(type
->family
),
238 ip_set_type_unlock();
241 list_del_rcu(&type
->list
);
242 pr_debug("type %s, family %s with revision min %u unregistered.\n",
243 type
->name
, family_name(type
->family
), type
->revision_min
);
244 ip_set_type_unlock();
248 EXPORT_SYMBOL_GPL(ip_set_type_unregister
);
250 /* Utility functions */
252 ip_set_alloc(size_t size
)
254 return kvzalloc(size
, GFP_KERNEL_ACCOUNT
);
256 EXPORT_SYMBOL_GPL(ip_set_alloc
);
259 ip_set_free(void *members
)
261 pr_debug("%p: free with %s\n", members
,
262 is_vmalloc_addr(members
) ? "vfree" : "kfree");
265 EXPORT_SYMBOL_GPL(ip_set_free
);
268 flag_nested(const struct nlattr
*nla
)
270 return nla
->nla_type
& NLA_F_NESTED
;
273 static const struct nla_policy ipaddr_policy
[IPSET_ATTR_IPADDR_MAX
+ 1] = {
274 [IPSET_ATTR_IPADDR_IPV4
] = { .type
= NLA_U32
},
275 [IPSET_ATTR_IPADDR_IPV6
] = NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr
)),
279 ip_set_get_ipaddr4(struct nlattr
*nla
, __be32
*ipaddr
)
281 struct nlattr
*tb
[IPSET_ATTR_IPADDR_MAX
+ 1];
283 if (unlikely(!flag_nested(nla
)))
284 return -IPSET_ERR_PROTOCOL
;
285 if (NLA_PARSE_NESTED(tb
, IPSET_ATTR_IPADDR_MAX
, nla
,
286 ipaddr_policy
, NULL
))
287 return -IPSET_ERR_PROTOCOL
;
288 if (unlikely(!ip_set_attr_netorder(tb
, IPSET_ATTR_IPADDR_IPV4
)))
289 return -IPSET_ERR_PROTOCOL
;
291 *ipaddr
= nla_get_be32(tb
[IPSET_ATTR_IPADDR_IPV4
]);
294 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr4
);
297 ip_set_get_ipaddr6(struct nlattr
*nla
, union nf_inet_addr
*ipaddr
)
299 struct nlattr
*tb
[IPSET_ATTR_IPADDR_MAX
+ 1];
301 if (unlikely(!flag_nested(nla
)))
302 return -IPSET_ERR_PROTOCOL
;
304 if (NLA_PARSE_NESTED(tb
, IPSET_ATTR_IPADDR_MAX
, nla
,
305 ipaddr_policy
, NULL
))
306 return -IPSET_ERR_PROTOCOL
;
307 if (unlikely(!ip_set_attr_netorder(tb
, IPSET_ATTR_IPADDR_IPV6
)))
308 return -IPSET_ERR_PROTOCOL
;
310 memcpy(ipaddr
, nla_data(tb
[IPSET_ATTR_IPADDR_IPV6
]),
311 sizeof(struct in6_addr
));
314 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr6
);
317 ip_set_timeout_get(const unsigned long *timeout
)
321 if (*timeout
== IPSET_ELEM_PERMANENT
)
324 t
= jiffies_to_msecs(*timeout
- jiffies
) / MSEC_PER_SEC
;
325 /* Zero value in userspace means no timeout */
326 return t
== 0 ? 1 : t
;
330 ip_set_comment_uget(struct nlattr
*tb
)
335 /* Called from uadd only, protected by the set spinlock.
336 * The kadt functions don't use the comment extensions in any way.
339 ip_set_init_comment(struct ip_set
*set
, struct ip_set_comment
*comment
,
340 const struct ip_set_ext
*ext
)
342 struct ip_set_comment_rcu
*c
= rcu_dereference_protected(comment
->c
, 1);
343 size_t len
= ext
->comment
? strlen(ext
->comment
) : 0;
346 set
->ext_size
-= sizeof(*c
) + strlen(c
->str
) + 1;
348 rcu_assign_pointer(comment
->c
, NULL
);
352 if (unlikely(len
> IPSET_MAX_COMMENT_SIZE
))
353 len
= IPSET_MAX_COMMENT_SIZE
;
354 c
= kmalloc(sizeof(*c
) + len
+ 1, GFP_ATOMIC
);
357 strscpy(c
->str
, ext
->comment
, len
+ 1);
358 set
->ext_size
+= sizeof(*c
) + strlen(c
->str
) + 1;
359 rcu_assign_pointer(comment
->c
, c
);
361 EXPORT_SYMBOL_GPL(ip_set_init_comment
);
363 /* Used only when dumping a set, protected by rcu_read_lock() */
365 ip_set_put_comment(struct sk_buff
*skb
, const struct ip_set_comment
*comment
)
367 struct ip_set_comment_rcu
*c
= rcu_dereference(comment
->c
);
371 return nla_put_string(skb
, IPSET_ATTR_COMMENT
, c
->str
);
374 /* Called from uadd/udel, flush or the garbage collectors protected
375 * by the set spinlock.
376 * Called when the set is destroyed and when there can't be any user
377 * of the set data anymore.
380 ip_set_comment_free(struct ip_set
*set
, void *ptr
)
382 struct ip_set_comment
*comment
= ptr
;
383 struct ip_set_comment_rcu
*c
;
385 c
= rcu_dereference_protected(comment
->c
, 1);
388 set
->ext_size
-= sizeof(*c
) + strlen(c
->str
) + 1;
390 rcu_assign_pointer(comment
->c
, NULL
);
393 typedef void (*destroyer
)(struct ip_set
*, void *);
394 /* ipset data extension types, in size order */
396 const struct ip_set_ext_type ip_set_extensions
[] = {
397 [IPSET_EXT_ID_COUNTER
] = {
398 .type
= IPSET_EXT_COUNTER
,
399 .flag
= IPSET_FLAG_WITH_COUNTERS
,
400 .len
= sizeof(struct ip_set_counter
),
401 .align
= __alignof__(struct ip_set_counter
),
403 [IPSET_EXT_ID_TIMEOUT
] = {
404 .type
= IPSET_EXT_TIMEOUT
,
405 .len
= sizeof(unsigned long),
406 .align
= __alignof__(unsigned long),
408 [IPSET_EXT_ID_SKBINFO
] = {
409 .type
= IPSET_EXT_SKBINFO
,
410 .flag
= IPSET_FLAG_WITH_SKBINFO
,
411 .len
= sizeof(struct ip_set_skbinfo
),
412 .align
= __alignof__(struct ip_set_skbinfo
),
414 [IPSET_EXT_ID_COMMENT
] = {
415 .type
= IPSET_EXT_COMMENT
| IPSET_EXT_DESTROY
,
416 .flag
= IPSET_FLAG_WITH_COMMENT
,
417 .len
= sizeof(struct ip_set_comment
),
418 .align
= __alignof__(struct ip_set_comment
),
419 .destroy
= ip_set_comment_free
,
422 EXPORT_SYMBOL_GPL(ip_set_extensions
);
425 add_extension(enum ip_set_ext_id id
, u32 flags
, struct nlattr
*tb
[])
427 return ip_set_extensions
[id
].flag
?
428 (flags
& ip_set_extensions
[id
].flag
) :
429 !!tb
[IPSET_ATTR_TIMEOUT
];
433 ip_set_elem_len(struct ip_set
*set
, struct nlattr
*tb
[], size_t len
,
436 enum ip_set_ext_id id
;
439 if (tb
[IPSET_ATTR_CADT_FLAGS
])
440 cadt_flags
= ip_set_get_h32(tb
[IPSET_ATTR_CADT_FLAGS
]);
441 if (cadt_flags
& IPSET_FLAG_WITH_FORCEADD
)
442 set
->flags
|= IPSET_CREATE_FLAG_FORCEADD
;
445 for (id
= 0; id
< IPSET_EXT_ID_MAX
; id
++) {
446 if (!add_extension(id
, cadt_flags
, tb
))
448 if (align
< ip_set_extensions
[id
].align
)
449 align
= ip_set_extensions
[id
].align
;
450 len
= ALIGN(len
, ip_set_extensions
[id
].align
);
451 set
->offset
[id
] = len
;
452 set
->extensions
|= ip_set_extensions
[id
].type
;
453 len
+= ip_set_extensions
[id
].len
;
455 return ALIGN(len
, align
);
457 EXPORT_SYMBOL_GPL(ip_set_elem_len
);
460 ip_set_get_extensions(struct ip_set
*set
, struct nlattr
*tb
[],
461 struct ip_set_ext
*ext
)
465 if (unlikely(!ip_set_optattr_netorder(tb
, IPSET_ATTR_TIMEOUT
) ||
466 !ip_set_optattr_netorder(tb
, IPSET_ATTR_PACKETS
) ||
467 !ip_set_optattr_netorder(tb
, IPSET_ATTR_BYTES
) ||
468 !ip_set_optattr_netorder(tb
, IPSET_ATTR_SKBMARK
) ||
469 !ip_set_optattr_netorder(tb
, IPSET_ATTR_SKBPRIO
) ||
470 !ip_set_optattr_netorder(tb
, IPSET_ATTR_SKBQUEUE
)))
471 return -IPSET_ERR_PROTOCOL
;
473 if (tb
[IPSET_ATTR_TIMEOUT
]) {
474 if (!SET_WITH_TIMEOUT(set
))
475 return -IPSET_ERR_TIMEOUT
;
476 ext
->timeout
= ip_set_timeout_uget(tb
[IPSET_ATTR_TIMEOUT
]);
478 if (tb
[IPSET_ATTR_BYTES
] || tb
[IPSET_ATTR_PACKETS
]) {
479 if (!SET_WITH_COUNTER(set
))
480 return -IPSET_ERR_COUNTER
;
481 if (tb
[IPSET_ATTR_BYTES
])
482 ext
->bytes
= be64_to_cpu(nla_get_be64(
483 tb
[IPSET_ATTR_BYTES
]));
484 if (tb
[IPSET_ATTR_PACKETS
])
485 ext
->packets
= be64_to_cpu(nla_get_be64(
486 tb
[IPSET_ATTR_PACKETS
]));
488 if (tb
[IPSET_ATTR_COMMENT
]) {
489 if (!SET_WITH_COMMENT(set
))
490 return -IPSET_ERR_COMMENT
;
491 ext
->comment
= ip_set_comment_uget(tb
[IPSET_ATTR_COMMENT
]);
493 if (tb
[IPSET_ATTR_SKBMARK
]) {
494 if (!SET_WITH_SKBINFO(set
))
495 return -IPSET_ERR_SKBINFO
;
496 fullmark
= be64_to_cpu(nla_get_be64(tb
[IPSET_ATTR_SKBMARK
]));
497 ext
->skbinfo
.skbmark
= fullmark
>> 32;
498 ext
->skbinfo
.skbmarkmask
= fullmark
& 0xffffffff;
500 if (tb
[IPSET_ATTR_SKBPRIO
]) {
501 if (!SET_WITH_SKBINFO(set
))
502 return -IPSET_ERR_SKBINFO
;
503 ext
->skbinfo
.skbprio
=
504 be32_to_cpu(nla_get_be32(tb
[IPSET_ATTR_SKBPRIO
]));
506 if (tb
[IPSET_ATTR_SKBQUEUE
]) {
507 if (!SET_WITH_SKBINFO(set
))
508 return -IPSET_ERR_SKBINFO
;
509 ext
->skbinfo
.skbqueue
=
510 be16_to_cpu(nla_get_be16(tb
[IPSET_ATTR_SKBQUEUE
]));
514 EXPORT_SYMBOL_GPL(ip_set_get_extensions
);
517 ip_set_get_bytes(const struct ip_set_counter
*counter
)
519 return (u64
)atomic64_read(&(counter
)->bytes
);
523 ip_set_get_packets(const struct ip_set_counter
*counter
)
525 return (u64
)atomic64_read(&(counter
)->packets
);
529 ip_set_put_counter(struct sk_buff
*skb
, const struct ip_set_counter
*counter
)
531 return IPSET_NLA_PUT_NET64(skb
, IPSET_ATTR_BYTES
,
532 cpu_to_be64(ip_set_get_bytes(counter
)),
534 IPSET_NLA_PUT_NET64(skb
, IPSET_ATTR_PACKETS
,
535 cpu_to_be64(ip_set_get_packets(counter
)),
540 ip_set_put_skbinfo(struct sk_buff
*skb
, const struct ip_set_skbinfo
*skbinfo
)
542 /* Send nonzero parameters only */
543 return ((skbinfo
->skbmark
|| skbinfo
->skbmarkmask
) &&
544 IPSET_NLA_PUT_NET64(skb
, IPSET_ATTR_SKBMARK
,
545 cpu_to_be64((u64
)skbinfo
->skbmark
<< 32 |
546 skbinfo
->skbmarkmask
),
549 nla_put_net32(skb
, IPSET_ATTR_SKBPRIO
,
550 cpu_to_be32(skbinfo
->skbprio
))) ||
551 (skbinfo
->skbqueue
&&
552 nla_put_net16(skb
, IPSET_ATTR_SKBQUEUE
,
553 cpu_to_be16(skbinfo
->skbqueue
)));
557 ip_set_put_extensions(struct sk_buff
*skb
, const struct ip_set
*set
,
558 const void *e
, bool active
)
560 if (SET_WITH_TIMEOUT(set
)) {
561 unsigned long *timeout
= ext_timeout(e
, set
);
563 if (nla_put_net32(skb
, IPSET_ATTR_TIMEOUT
,
564 htonl(active
? ip_set_timeout_get(timeout
)
568 if (SET_WITH_COUNTER(set
) &&
569 ip_set_put_counter(skb
, ext_counter(e
, set
)))
571 if (SET_WITH_COMMENT(set
) &&
572 ip_set_put_comment(skb
, ext_comment(e
, set
)))
574 if (SET_WITH_SKBINFO(set
) &&
575 ip_set_put_skbinfo(skb
, ext_skbinfo(e
, set
)))
579 EXPORT_SYMBOL_GPL(ip_set_put_extensions
);
582 ip_set_match_counter(u64 counter
, u64 match
, u8 op
)
585 case IPSET_COUNTER_NONE
:
587 case IPSET_COUNTER_EQ
:
588 return counter
== match
;
589 case IPSET_COUNTER_NE
:
590 return counter
!= match
;
591 case IPSET_COUNTER_LT
:
592 return counter
< match
;
593 case IPSET_COUNTER_GT
:
594 return counter
> match
;
600 ip_set_add_bytes(u64 bytes
, struct ip_set_counter
*counter
)
602 atomic64_add((long long)bytes
, &(counter
)->bytes
);
606 ip_set_add_packets(u64 packets
, struct ip_set_counter
*counter
)
608 atomic64_add((long long)packets
, &(counter
)->packets
);
612 ip_set_update_counter(struct ip_set_counter
*counter
,
613 const struct ip_set_ext
*ext
, u32 flags
)
615 if (ext
->packets
!= ULLONG_MAX
&&
616 !(flags
& IPSET_FLAG_SKIP_COUNTER_UPDATE
)) {
617 ip_set_add_bytes(ext
->bytes
, counter
);
618 ip_set_add_packets(ext
->packets
, counter
);
623 ip_set_get_skbinfo(struct ip_set_skbinfo
*skbinfo
,
624 const struct ip_set_ext
*ext
,
625 struct ip_set_ext
*mext
, u32 flags
)
627 mext
->skbinfo
= *skbinfo
;
631 ip_set_match_extensions(struct ip_set
*set
, const struct ip_set_ext
*ext
,
632 struct ip_set_ext
*mext
, u32 flags
, void *data
)
634 if (SET_WITH_TIMEOUT(set
) &&
635 ip_set_timeout_expired(ext_timeout(data
, set
)))
637 if (SET_WITH_COUNTER(set
)) {
638 struct ip_set_counter
*counter
= ext_counter(data
, set
);
640 ip_set_update_counter(counter
, ext
, flags
);
642 if (flags
& IPSET_FLAG_MATCH_COUNTERS
&&
643 !(ip_set_match_counter(ip_set_get_packets(counter
),
644 mext
->packets
, mext
->packets_op
) &&
645 ip_set_match_counter(ip_set_get_bytes(counter
),
646 mext
->bytes
, mext
->bytes_op
)))
649 if (SET_WITH_SKBINFO(set
))
650 ip_set_get_skbinfo(ext_skbinfo(data
, set
),
654 EXPORT_SYMBOL_GPL(ip_set_match_extensions
);
656 /* Creating/destroying/renaming/swapping affect the existence and
657 * the properties of a set. All of these can be executed from userspace
658 * only and serialized by the nfnl mutex indirectly from nfnetlink.
660 * Sets are identified by their index in ip_set_list and the index
661 * is used by the external references (set/SET netfilter modules).
663 * The set behind an index may change by swapping only, from userspace.
667 __ip_set_get(struct ip_set
*set
)
669 write_lock_bh(&ip_set_ref_lock
);
671 write_unlock_bh(&ip_set_ref_lock
);
675 __ip_set_put(struct ip_set
*set
)
677 write_lock_bh(&ip_set_ref_lock
);
678 BUG_ON(set
->ref
== 0);
680 write_unlock_bh(&ip_set_ref_lock
);
683 /* set->ref can be swapped out by ip_set_swap, netlink events (like dump) need
684 * a separate reference counter
687 __ip_set_get_netlink(struct ip_set
*set
)
689 write_lock_bh(&ip_set_ref_lock
);
691 write_unlock_bh(&ip_set_ref_lock
);
695 __ip_set_put_netlink(struct ip_set
*set
)
697 write_lock_bh(&ip_set_ref_lock
);
698 BUG_ON(set
->ref_netlink
== 0);
700 write_unlock_bh(&ip_set_ref_lock
);
703 /* Add, del and test set entries from kernel.
705 * The set behind the index must exist and must be referenced
706 * so it can't be destroyed (or changed) under our foot.
709 static struct ip_set
*
710 ip_set_rcu_get(struct net
*net
, ip_set_id_t index
)
713 struct ip_set_net
*inst
= ip_set_pernet(net
);
716 /* ip_set_list and the set pointer need to be protected */
717 set
= rcu_dereference(inst
->ip_set_list
)[index
];
723 ip_set_rcu_put(struct ip_set
*set __always_unused
)
729 ip_set_lock(struct ip_set
*set
)
731 if (!set
->variant
->region_lock
)
732 spin_lock_bh(&set
->lock
);
736 ip_set_unlock(struct ip_set
*set
)
738 if (!set
->variant
->region_lock
)
739 spin_unlock_bh(&set
->lock
);
743 ip_set_test(ip_set_id_t index
, const struct sk_buff
*skb
,
744 const struct xt_action_param
*par
, struct ip_set_adt_opt
*opt
)
746 struct ip_set
*set
= ip_set_rcu_get(IPSET_DEV_NET(par
), index
);
750 pr_debug("set %s, index %u\n", set
->name
, index
);
752 if (opt
->dim
< set
->type
->dimension
||
753 !(opt
->family
== set
->family
|| set
->family
== NFPROTO_UNSPEC
)) {
758 ret
= set
->variant
->kadt(set
, skb
, par
, IPSET_TEST
, opt
);
760 if (ret
== -EAGAIN
) {
761 /* Type requests element to be completed */
762 pr_debug("element must be completed, ADD is triggered\n");
764 set
->variant
->kadt(set
, skb
, par
, IPSET_ADD
, opt
);
768 /* --return-nomatch: invert matched element */
769 if ((opt
->cmdflags
& IPSET_FLAG_RETURN_NOMATCH
) &&
770 (set
->type
->features
& IPSET_TYPE_NOMATCH
) &&
771 (ret
> 0 || ret
== -ENOTEMPTY
))
776 /* Convert error codes to nomatch */
777 return (ret
< 0 ? 0 : ret
);
779 EXPORT_SYMBOL_GPL(ip_set_test
);
782 ip_set_add(ip_set_id_t index
, const struct sk_buff
*skb
,
783 const struct xt_action_param
*par
, struct ip_set_adt_opt
*opt
)
785 struct ip_set
*set
= ip_set_rcu_get(IPSET_DEV_NET(par
), index
);
789 pr_debug("set %s, index %u\n", set
->name
, index
);
791 if (opt
->dim
< set
->type
->dimension
||
792 !(opt
->family
== set
->family
|| set
->family
== NFPROTO_UNSPEC
)) {
794 return -IPSET_ERR_TYPE_MISMATCH
;
798 ret
= set
->variant
->kadt(set
, skb
, par
, IPSET_ADD
, opt
);
804 EXPORT_SYMBOL_GPL(ip_set_add
);
807 ip_set_del(ip_set_id_t index
, const struct sk_buff
*skb
,
808 const struct xt_action_param
*par
, struct ip_set_adt_opt
*opt
)
810 struct ip_set
*set
= ip_set_rcu_get(IPSET_DEV_NET(par
), index
);
814 pr_debug("set %s, index %u\n", set
->name
, index
);
816 if (opt
->dim
< set
->type
->dimension
||
817 !(opt
->family
== set
->family
|| set
->family
== NFPROTO_UNSPEC
)) {
819 return -IPSET_ERR_TYPE_MISMATCH
;
823 ret
= set
->variant
->kadt(set
, skb
, par
, IPSET_DEL
, opt
);
829 EXPORT_SYMBOL_GPL(ip_set_del
);
831 /* Find set by name, reference it once. The reference makes sure the
832 * thing pointed to, does not go away under our feet.
836 ip_set_get_byname(struct net
*net
, const char *name
, struct ip_set
**set
)
838 ip_set_id_t i
, index
= IPSET_INVALID_ID
;
840 struct ip_set_net
*inst
= ip_set_pernet(net
);
843 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
844 s
= rcu_dereference(inst
->ip_set_list
)[i
];
845 if (s
&& STRNCMP(s
->name
, name
)) {
856 EXPORT_SYMBOL_GPL(ip_set_get_byname
);
858 /* If the given set pointer points to a valid set, decrement
859 * reference count by 1. The caller shall not assume the index
860 * to be valid, after calling this function.
865 __ip_set_put_byindex(struct ip_set_net
*inst
, ip_set_id_t index
)
870 set
= rcu_dereference(inst
->ip_set_list
)[index
];
877 ip_set_put_byindex(struct net
*net
, ip_set_id_t index
)
879 struct ip_set_net
*inst
= ip_set_pernet(net
);
881 __ip_set_put_byindex(inst
, index
);
883 EXPORT_SYMBOL_GPL(ip_set_put_byindex
);
885 /* Get the name of a set behind a set index.
886 * Set itself is protected by RCU, but its name isn't: to protect against
887 * renaming, grab ip_set_ref_lock as reader (see ip_set_rename()) and copy the
891 ip_set_name_byindex(struct net
*net
, ip_set_id_t index
, char *name
)
893 struct ip_set
*set
= ip_set_rcu_get(net
, index
);
897 read_lock_bh(&ip_set_ref_lock
);
898 strscpy_pad(name
, set
->name
, IPSET_MAXNAMELEN
);
899 read_unlock_bh(&ip_set_ref_lock
);
902 EXPORT_SYMBOL_GPL(ip_set_name_byindex
);
904 /* Routines to call by external subsystems, which do not
905 * call nfnl_lock for us.
908 /* Find set by index, reference it once. The reference makes sure the
909 * thing pointed to, does not go away under our feet.
911 * The nfnl mutex is used in the function.
914 ip_set_nfnl_get_byindex(struct net
*net
, ip_set_id_t index
)
917 struct ip_set_net
*inst
= ip_set_pernet(net
);
919 if (index
>= inst
->ip_set_max
)
920 return IPSET_INVALID_ID
;
922 nfnl_lock(NFNL_SUBSYS_IPSET
);
923 set
= ip_set(inst
, index
);
927 index
= IPSET_INVALID_ID
;
928 nfnl_unlock(NFNL_SUBSYS_IPSET
);
932 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex
);
934 /* If the given set pointer points to a valid set, decrement
935 * reference count by 1. The caller shall not assume the index
936 * to be valid, after calling this function.
938 * The nfnl mutex is used in the function.
941 ip_set_nfnl_put(struct net
*net
, ip_set_id_t index
)
944 struct ip_set_net
*inst
= ip_set_pernet(net
);
946 nfnl_lock(NFNL_SUBSYS_IPSET
);
947 if (!inst
->is_deleted
) { /* already deleted from ip_set_net_exit() */
948 set
= ip_set(inst
, index
);
952 nfnl_unlock(NFNL_SUBSYS_IPSET
);
954 EXPORT_SYMBOL_GPL(ip_set_nfnl_put
);
956 /* Communication protocol with userspace over netlink.
958 * The commands are serialized by the nfnl mutex.
961 static inline u8
protocol(const struct nlattr
* const tb
[])
963 return nla_get_u8(tb
[IPSET_ATTR_PROTOCOL
]);
967 protocol_failed(const struct nlattr
* const tb
[])
969 return !tb
[IPSET_ATTR_PROTOCOL
] || protocol(tb
) != IPSET_PROTOCOL
;
973 protocol_min_failed(const struct nlattr
* const tb
[])
975 return !tb
[IPSET_ATTR_PROTOCOL
] || protocol(tb
) < IPSET_PROTOCOL_MIN
;
979 flag_exist(const struct nlmsghdr
*nlh
)
981 return nlh
->nlmsg_flags
& NLM_F_EXCL
? 0 : IPSET_FLAG_EXIST
;
984 static struct nlmsghdr
*
985 start_msg(struct sk_buff
*skb
, u32 portid
, u32 seq
, unsigned int flags
,
988 return nfnl_msg_put(skb
, portid
, seq
,
989 nfnl_msg_type(NFNL_SUBSYS_IPSET
, cmd
), flags
,
990 NFPROTO_IPV4
, NFNETLINK_V0
, 0);
995 static const struct nla_policy ip_set_create_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
996 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
997 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
998 .len
= IPSET_MAXNAMELEN
- 1 },
999 [IPSET_ATTR_TYPENAME
] = { .type
= NLA_NUL_STRING
,
1000 .len
= IPSET_MAXNAMELEN
- 1},
1001 [IPSET_ATTR_REVISION
] = { .type
= NLA_U8
},
1002 [IPSET_ATTR_FAMILY
] = { .type
= NLA_U8
},
1003 [IPSET_ATTR_DATA
] = { .type
= NLA_NESTED
},
1006 static struct ip_set
*
1007 find_set_and_id(struct ip_set_net
*inst
, const char *name
, ip_set_id_t
*id
)
1009 struct ip_set
*set
= NULL
;
1012 *id
= IPSET_INVALID_ID
;
1013 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1014 set
= ip_set(inst
, i
);
1015 if (set
&& STRNCMP(set
->name
, name
)) {
1020 return (*id
== IPSET_INVALID_ID
? NULL
: set
);
1023 static inline struct ip_set
*
1024 find_set(struct ip_set_net
*inst
, const char *name
)
1028 return find_set_and_id(inst
, name
, &id
);
1032 find_free_id(struct ip_set_net
*inst
, const char *name
, ip_set_id_t
*index
,
1033 struct ip_set
**set
)
1038 *index
= IPSET_INVALID_ID
;
1039 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1040 s
= ip_set(inst
, i
);
1042 if (*index
== IPSET_INVALID_ID
)
1044 } else if (STRNCMP(name
, s
->name
)) {
1050 if (*index
== IPSET_INVALID_ID
)
1051 /* No free slot remained */
1052 return -IPSET_ERR_MAX_SETS
;
1057 IPSET_CBFN(ip_set_none
, struct net
*net
, struct sock
*ctnl
,
1058 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1059 const struct nlattr
* const attr
[],
1060 struct netlink_ext_ack
*extack
,
1061 const struct nfnl_info
*info
)
1067 IPSET_CBFN(ip_set_create
, struct net
*n
, struct sock
*ctnl
,
1068 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1069 const struct nlattr
* const attr
[],
1070 struct netlink_ext_ack
*extack
,
1071 const struct nfnl_info
*info
)
1073 struct net
*net
= IPSET_SOCK_NET(n
, ctnl
, info
);
1074 struct ip_set_net
*inst
= ip_set_pernet(net
);
1075 struct ip_set
*set
, *clash
= NULL
;
1076 ip_set_id_t index
= IPSET_INVALID_ID
;
1077 struct nlattr
*tb
[IPSET_ATTR_CREATE_MAX
+ 1] = {};
1078 const char *name
, *typename
;
1079 u8 family
, revision
;
1080 u32 flags
= flag_exist(INFO_NLH(info
, nlh
));
1083 if (unlikely(protocol_min_failed(attr
) ||
1084 !attr
[IPSET_ATTR_SETNAME
] ||
1085 !attr
[IPSET_ATTR_TYPENAME
] ||
1086 !attr
[IPSET_ATTR_REVISION
] ||
1087 !attr
[IPSET_ATTR_FAMILY
] ||
1088 (attr
[IPSET_ATTR_DATA
] &&
1089 !flag_nested(attr
[IPSET_ATTR_DATA
]))))
1090 return -IPSET_ERR_PROTOCOL
;
1092 name
= nla_data(attr
[IPSET_ATTR_SETNAME
]);
1093 typename
= nla_data(attr
[IPSET_ATTR_TYPENAME
]);
1094 family
= nla_get_u8(attr
[IPSET_ATTR_FAMILY
]);
1095 revision
= nla_get_u8(attr
[IPSET_ATTR_REVISION
]);
1096 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
1097 name
, typename
, family_name(family
), revision
);
1099 /* First, and without any locks, allocate and initialize
1100 * a normal base set structure.
1102 set
= kzalloc(sizeof(*set
), GFP_KERNEL
);
1105 spin_lock_init(&set
->lock
);
1106 strscpy(set
->name
, name
, IPSET_MAXNAMELEN
);
1107 set
->family
= family
;
1108 set
->revision
= revision
;
1110 /* Next, check that we know the type, and take
1111 * a reference on the type, to make sure it stays available
1112 * while constructing our new set.
1114 * After referencing the type, we try to create the type
1115 * specific part of the set without holding any locks.
1117 ret
= find_set_type_get(typename
, family
, revision
, &set
->type
);
1121 /* Without holding any locks, create private part. */
1122 if (attr
[IPSET_ATTR_DATA
] &&
1123 NLA_PARSE_NESTED(tb
, IPSET_ATTR_CREATE_MAX
, attr
[IPSET_ATTR_DATA
],
1124 set
->type
->create_policy
, NULL
)) {
1125 ret
= -IPSET_ERR_PROTOCOL
;
1128 /* Set create flags depending on the type revision */
1129 set
->flags
|= set
->type
->create_flags
[revision
];
1131 ret
= set
->type
->create(INFO_NET(info
, net
), set
, tb
, flags
);
1135 /* BTW, ret==0 here. */
1137 /* Here, we have a valid, constructed set and we are protected
1138 * by the nfnl mutex. Find the first free index in ip_set_list
1139 * and check clashing.
1141 ret
= find_free_id(inst
, set
->name
, &index
, &clash
);
1142 if (ret
== -EEXIST
) {
1143 /* If this is the same set and requested, ignore error */
1144 if ((flags
& IPSET_FLAG_EXIST
) &&
1145 STRNCMP(set
->type
->name
, clash
->type
->name
) &&
1146 set
->type
->family
== clash
->type
->family
&&
1147 set
->type
->revision_min
== clash
->type
->revision_min
&&
1148 set
->type
->revision_max
== clash
->type
->revision_max
&&
1149 set
->variant
->same_set(set
, clash
))
1152 } else if (ret
== -IPSET_ERR_MAX_SETS
) {
1153 struct ip_set
**list
, **tmp
;
1154 ip_set_id_t i
= inst
->ip_set_max
+ IP_SET_INC
;
1156 if (i
< inst
->ip_set_max
|| i
== IPSET_INVALID_ID
)
1160 list
= kvcalloc(i
, sizeof(struct ip_set
*), GFP_KERNEL
);
1163 /* nfnl mutex is held, both lists are valid */
1164 tmp
= ip_set_dereference(inst
->ip_set_list
);
1165 memcpy(list
, tmp
, sizeof(struct ip_set
*) * inst
->ip_set_max
);
1166 rcu_assign_pointer(inst
->ip_set_list
, list
);
1167 /* Make sure all current packets have passed through */
1170 index
= inst
->ip_set_max
;
1171 inst
->ip_set_max
= i
;
1178 /* Finally! Add our shiny new set to the list, and be done. */
1179 pr_debug("create: '%s' created with index %u!\n", set
->name
, index
);
1180 ip_set(inst
, index
) = set
;
1185 set
->variant
->destroy(set
);
1187 module_put(set
->type
->me
);
1195 static const struct nla_policy
1196 ip_set_setname_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1197 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1198 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
1199 .len
= IPSET_MAXNAMELEN
- 1 },
1203 ip_set_destroy_set(struct ip_set
*set
)
1205 pr_debug("set: %s\n", set
->name
);
1207 /* Must call it without holding any lock */
1208 set
->variant
->destroy(set
);
1209 module_put(set
->type
->me
);
1214 IPSET_CBFN(ip_set_destroy
, struct net
*net
, struct sock
*ctnl
,
1215 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1216 const struct nlattr
* const attr
[],
1217 struct netlink_ext_ack
*extack
,
1218 const struct nfnl_info
*info
)
1220 struct ip_set_net
*inst
= ip_set_pernet(IPSET_SOCK_NET(net
, ctnl
, info
));
1225 if (unlikely(protocol_min_failed(attr
)))
1226 return -IPSET_ERR_PROTOCOL
;
1228 /* Make sure all readers of the old set pointers are completed. */
1231 /* Must wait for flush to be really finished in list:set */
1234 /* Commands are serialized and references are
1235 * protected by the ip_set_ref_lock.
1236 * External systems (i.e. xt_set) must call
1237 * ip_set_put|get_nfnl_* functions, that way we
1238 * can safely check references here.
1240 * list:set timer can only decrement the reference
1241 * counter, so if it's already zero, we can proceed
1242 * without holding the lock.
1244 read_lock_bh(&ip_set_ref_lock
);
1245 if (!attr
[IPSET_ATTR_SETNAME
]) {
1246 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1247 s
= ip_set(inst
, i
);
1248 if (s
&& (s
->ref
|| s
->ref_netlink
)) {
1249 ret
= -IPSET_ERR_BUSY
;
1253 inst
->is_destroyed
= true;
1254 read_unlock_bh(&ip_set_ref_lock
);
1255 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1256 s
= ip_set(inst
, i
);
1258 ip_set(inst
, i
) = NULL
;
1259 ip_set_destroy_set(s
);
1262 /* Modified by ip_set_destroy() only, which is serialized */
1263 inst
->is_destroyed
= false;
1265 u32 flags
= flag_exist(INFO_NLH(info
, nlh
));
1266 s
= find_set_and_id(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]),
1269 if (!(flags
& IPSET_FLAG_EXIST
))
1272 } else if (s
->ref
|| s
->ref_netlink
) {
1273 ret
= -IPSET_ERR_BUSY
;
1276 ip_set(inst
, i
) = NULL
;
1277 read_unlock_bh(&ip_set_ref_lock
);
1279 ip_set_destroy_set(s
);
1283 read_unlock_bh(&ip_set_ref_lock
);
1290 ip_set_flush_set(struct ip_set
*set
)
1292 pr_debug("set: %s\n", set
->name
);
1295 set
->variant
->flush(set
);
1300 IPSET_CBFN(ip_set_flush
, struct net
*net
, struct sock
*ctnl
,
1301 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1302 const struct nlattr
* const attr
[],
1303 struct netlink_ext_ack
*extack
,
1304 const struct nfnl_info
*info
)
1306 struct ip_set_net
*inst
= ip_set_pernet(IPSET_SOCK_NET(net
, ctnl
, info
));
1310 if (unlikely(protocol_min_failed(attr
)))
1311 return -IPSET_ERR_PROTOCOL
;
1313 if (!attr
[IPSET_ATTR_SETNAME
]) {
1314 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1315 s
= ip_set(inst
, i
);
1317 ip_set_flush_set(s
);
1320 s
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1324 ip_set_flush_set(s
);
1332 static const struct nla_policy
1333 ip_set_setname2_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1334 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1335 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
1336 .len
= IPSET_MAXNAMELEN
- 1 },
1337 [IPSET_ATTR_SETNAME2
] = { .type
= NLA_NUL_STRING
,
1338 .len
= IPSET_MAXNAMELEN
- 1 },
1342 IPSET_CBFN(ip_set_rename
, struct net
*net
, struct sock
*ctnl
,
1343 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1344 const struct nlattr
* const attr
[],
1345 struct netlink_ext_ack
*extack
,
1346 const struct nfnl_info
*info
)
1348 struct ip_set_net
*inst
= ip_set_pernet(IPSET_SOCK_NET(net
, ctnl
, info
));
1349 struct ip_set
*set
, *s
;
1354 if (unlikely(protocol_min_failed(attr
) ||
1355 !attr
[IPSET_ATTR_SETNAME
] ||
1356 !attr
[IPSET_ATTR_SETNAME2
]))
1357 return -IPSET_ERR_PROTOCOL
;
1359 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1363 write_lock_bh(&ip_set_ref_lock
);
1364 if (set
->ref
!= 0 || set
->ref_netlink
!= 0) {
1365 ret
= -IPSET_ERR_REFERENCED
;
1369 name2
= nla_data(attr
[IPSET_ATTR_SETNAME2
]);
1370 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
1371 s
= ip_set(inst
, i
);
1372 if (s
&& STRNCMP(s
->name
, name2
)) {
1373 ret
= -IPSET_ERR_EXIST_SETNAME2
;
1377 strscpy_pad(set
->name
, name2
, IPSET_MAXNAMELEN
);
1380 write_unlock_bh(&ip_set_ref_lock
);
1384 /* Swap two sets so that name/index points to the other.
1385 * References and set names are also swapped.
1387 * The commands are serialized by the nfnl mutex and references are
1388 * protected by the ip_set_ref_lock. The kernel interfaces
1389 * do not hold the mutex but the pointer settings are atomic
1390 * so the ip_set_list always contains valid pointers to the sets.
1394 IPSET_CBFN(ip_set_swap
, struct net
*net
, struct sock
*ctnl
,
1395 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1396 const struct nlattr
* const attr
[],
1397 struct netlink_ext_ack
*extack
,
1398 const struct nfnl_info
*info
)
1400 struct ip_set_net
*inst
= ip_set_pernet(IPSET_SOCK_NET(net
, ctnl
, info
));
1401 struct ip_set
*from
, *to
;
1402 ip_set_id_t from_id
, to_id
;
1403 char from_name
[IPSET_MAXNAMELEN
];
1405 if (unlikely(protocol_min_failed(attr
) ||
1406 !attr
[IPSET_ATTR_SETNAME
] ||
1407 !attr
[IPSET_ATTR_SETNAME2
]))
1408 return -IPSET_ERR_PROTOCOL
;
1410 from
= find_set_and_id(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]),
1415 to
= find_set_and_id(inst
, nla_data(attr
[IPSET_ATTR_SETNAME2
]),
1418 return -IPSET_ERR_EXIST_SETNAME2
;
1420 /* Features must not change.
1421 * Not an artifical restriction anymore, as we must prevent
1422 * possible loops created by swapping in setlist type of sets.
1424 if (!(from
->type
->features
== to
->type
->features
&&
1425 from
->family
== to
->family
))
1426 return -IPSET_ERR_TYPE_MISMATCH
;
1428 write_lock_bh(&ip_set_ref_lock
);
1430 if (from
->ref_netlink
|| to
->ref_netlink
) {
1431 write_unlock_bh(&ip_set_ref_lock
);
1435 strscpy_pad(from_name
, from
->name
, IPSET_MAXNAMELEN
);
1436 strscpy_pad(from
->name
, to
->name
, IPSET_MAXNAMELEN
);
1437 strscpy_pad(to
->name
, from_name
, IPSET_MAXNAMELEN
);
1439 swap(from
->ref
, to
->ref
);
1440 ip_set(inst
, from_id
) = to
;
1441 ip_set(inst
, to_id
) = from
;
1442 write_unlock_bh(&ip_set_ref_lock
);
1447 /* List/save set data */
1454 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1455 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1458 ip_set_put_flags(struct sk_buff
*skb
, struct ip_set
*set
)
1462 if (SET_WITH_TIMEOUT(set
))
1463 if (unlikely(nla_put_net32(skb
, IPSET_ATTR_TIMEOUT
,
1464 htonl(set
->timeout
))))
1466 if (SET_WITH_COUNTER(set
))
1467 cadt_flags
|= IPSET_FLAG_WITH_COUNTERS
;
1468 if (SET_WITH_COMMENT(set
))
1469 cadt_flags
|= IPSET_FLAG_WITH_COMMENT
;
1470 if (SET_WITH_SKBINFO(set
))
1471 cadt_flags
|= IPSET_FLAG_WITH_SKBINFO
;
1472 if (SET_WITH_FORCEADD(set
))
1473 cadt_flags
|= IPSET_FLAG_WITH_FORCEADD
;
1477 return nla_put_net32(skb
, IPSET_ATTR_CADT_FLAGS
, htonl(cadt_flags
));
1479 EXPORT_SYMBOL_GPL(ip_set_put_flags
);
1482 ip_set_dump_done(struct netlink_callback
*cb
)
1484 if (cb
->args
[IPSET_CB_ARG0
]) {
1485 struct ip_set_net
*inst
=
1486 (struct ip_set_net
*)cb
->args
[IPSET_CB_NET
];
1487 ip_set_id_t index
= (ip_set_id_t
)cb
->args
[IPSET_CB_INDEX
];
1488 struct ip_set
*set
= ip_set_ref_netlink(inst
, index
);
1490 if (set
->variant
->uref
)
1491 set
->variant
->uref(set
, cb
, false);
1492 pr_debug("release set %s\n", set
->name
);
1493 __ip_set_put_netlink(set
);
1499 dump_attrs(struct nlmsghdr
*nlh
)
1501 const struct nlattr
*attr
;
1504 pr_debug("dump nlmsg\n");
1505 nlmsg_for_each_attr(attr
, nlh
, sizeof(struct nfgenmsg
), rem
) {
1506 pr_debug("type: %u, len %u\n", nla_type(attr
), attr
->nla_len
);
1510 static const struct nla_policy
1511 ip_set_dump_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1512 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1513 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
1514 .len
= IPSET_MAXNAMELEN
- 1 },
1515 [IPSET_ATTR_FLAGS
] = { .type
= NLA_U32
},
1519 ip_set_dump_start(struct netlink_callback
*cb
)
1521 struct nlmsghdr
*nlh
= nlmsg_hdr(cb
->skb
);
1522 int min_len
= nlmsg_total_size(sizeof(struct nfgenmsg
));
1523 struct nlattr
*cda
[IPSET_ATTR_CMD_MAX
+ 1];
1524 struct nlattr
*attr
= (void *)nlh
+ min_len
;
1525 struct sk_buff
*skb
= cb
->skb
;
1526 struct ip_set_net
*inst
= ip_set_pernet(sock_net(skb
->sk
));
1530 ret
= NLA_PARSE(cda
, IPSET_ATTR_CMD_MAX
, attr
,
1531 nlh
->nlmsg_len
- min_len
,
1532 ip_set_dump_policy
, NULL
);
1536 cb
->args
[IPSET_CB_PROTO
] = nla_get_u8(cda
[IPSET_ATTR_PROTOCOL
]);
1537 if (cda
[IPSET_ATTR_SETNAME
]) {
1541 #if HAVE_NETLINK_DUMP_START_ARGS != 4
1542 read_lock_bh(&ip_set_ref_lock
);
1544 set
= find_set_and_id(inst
, nla_data(cda
[IPSET_ATTR_SETNAME
]),
1546 #if HAVE_NETLINK_DUMP_START_ARGS != 4
1547 read_unlock_bh(&ip_set_ref_lock
);
1553 dump_type
= DUMP_ONE
;
1554 cb
->args
[IPSET_CB_INDEX
] = index
;
1556 dump_type
= DUMP_ALL
;
1559 if (cda
[IPSET_ATTR_FLAGS
]) {
1560 u32 f
= ip_set_get_h32(cda
[IPSET_ATTR_FLAGS
]);
1562 dump_type
|= (f
<< 16);
1564 cb
->args
[IPSET_CB_NET
] = (unsigned long)inst
;
1565 cb
->args
[IPSET_CB_DUMP
] = dump_type
;
1570 /* We have to create and send the error message manually :-( */
1571 if (nlh
->nlmsg_flags
& NLM_F_ACK
) {
1572 NETLINK_ACK(cb
->skb
, nlh
, ret
, NULL
);
1578 ip_set_dump_do(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1580 ip_set_id_t index
= IPSET_INVALID_ID
, max
;
1581 struct ip_set
*set
= NULL
;
1582 struct nlmsghdr
*nlh
= NULL
;
1583 unsigned int flags
= NETLINK_PORTID(cb
->skb
) ? NLM_F_MULTI
: 0;
1584 struct ip_set_net
*inst
= ip_set_pernet(sock_net(skb
->sk
));
1585 u32 dump_type
, dump_flags
;
1589 if (!cb
->args
[IPSET_CB_DUMP
]) {
1590 #if HAVE_NETLINK_DUMP_START_ARGS == 4
1593 ret
= ip_set_dump_start(cb
);
1600 if (cb
->args
[IPSET_CB_INDEX
] >= inst
->ip_set_max
)
1603 dump_type
= DUMP_TYPE(cb
->args
[IPSET_CB_DUMP
]);
1604 dump_flags
= DUMP_FLAGS(cb
->args
[IPSET_CB_DUMP
]);
1605 max
= dump_type
== DUMP_ONE
? cb
->args
[IPSET_CB_INDEX
] + 1
1608 pr_debug("dump type, flag: %u %u index: %ld\n",
1609 dump_type
, dump_flags
, cb
->args
[IPSET_CB_INDEX
]);
1610 for (; cb
->args
[IPSET_CB_INDEX
] < max
; cb
->args
[IPSET_CB_INDEX
]++) {
1611 index
= (ip_set_id_t
)cb
->args
[IPSET_CB_INDEX
];
1612 write_lock_bh(&ip_set_ref_lock
);
1613 set
= ip_set(inst
, index
);
1614 is_destroyed
= inst
->is_destroyed
;
1615 if (!set
|| is_destroyed
) {
1616 write_unlock_bh(&ip_set_ref_lock
);
1617 if (dump_type
== DUMP_ONE
) {
1622 /* All sets are just being destroyed */
1628 /* When dumping all sets, we must dump "sorted"
1629 * so that lists (unions of sets) are dumped last.
1631 if (dump_type
!= DUMP_ONE
&&
1632 ((dump_type
== DUMP_ALL
) ==
1633 !!(set
->type
->features
& IPSET_DUMP_LAST
))) {
1634 write_unlock_bh(&ip_set_ref_lock
);
1637 pr_debug("List set: %s\n", set
->name
);
1638 if (!cb
->args
[IPSET_CB_ARG0
]) {
1639 /* Start listing: make sure set won't be destroyed */
1640 pr_debug("reference set\n");
1643 write_unlock_bh(&ip_set_ref_lock
);
1644 nlh
= start_msg(skb
, NETLINK_PORTID(cb
->skb
),
1645 cb
->nlh
->nlmsg_seq
, flags
,
1649 goto release_refcount
;
1651 if (nla_put_u8(skb
, IPSET_ATTR_PROTOCOL
,
1652 cb
->args
[IPSET_CB_PROTO
]) ||
1653 nla_put_string(skb
, IPSET_ATTR_SETNAME
, set
->name
))
1654 goto nla_put_failure
;
1655 if (dump_flags
& IPSET_FLAG_LIST_SETNAME
)
1657 switch (cb
->args
[IPSET_CB_ARG0
]) {
1659 /* Core header data */
1660 if (nla_put_string(skb
, IPSET_ATTR_TYPENAME
,
1662 nla_put_u8(skb
, IPSET_ATTR_FAMILY
,
1664 nla_put_u8(skb
, IPSET_ATTR_REVISION
,
1666 goto nla_put_failure
;
1667 if (cb
->args
[IPSET_CB_PROTO
] > IPSET_PROTOCOL_MIN
&&
1668 nla_put_net16(skb
, IPSET_ATTR_INDEX
, htons(index
)))
1669 goto nla_put_failure
;
1670 ret
= set
->variant
->head(set
, skb
);
1672 goto release_refcount
;
1673 if (dump_flags
& IPSET_FLAG_LIST_HEADER
)
1675 if (set
->variant
->uref
)
1676 set
->variant
->uref(set
, cb
, true);
1679 ret
= set
->variant
->list(set
, skb
, cb
);
1680 if (!cb
->args
[IPSET_CB_ARG0
])
1681 /* Set is done, proceed with next one */
1683 goto release_refcount
;
1686 /* If we dump all sets, continue with dumping last ones */
1687 if (dump_type
== DUMP_ALL
) {
1688 dump_type
= DUMP_LAST
;
1689 cb
->args
[IPSET_CB_DUMP
] = dump_type
| (dump_flags
<< 16);
1690 cb
->args
[IPSET_CB_INDEX
] = 0;
1691 if (set
&& set
->variant
->uref
)
1692 set
->variant
->uref(set
, cb
, false);
1700 if (dump_type
== DUMP_ONE
)
1701 cb
->args
[IPSET_CB_INDEX
] = IPSET_INVALID_ID
;
1703 cb
->args
[IPSET_CB_INDEX
]++;
1705 /* If there was an error or set is done, release set */
1706 if (ret
|| !cb
->args
[IPSET_CB_ARG0
]) {
1707 set
= ip_set_ref_netlink(inst
, index
);
1708 if (set
->variant
->uref
)
1709 set
->variant
->uref(set
, cb
, false);
1710 pr_debug("release set %s\n", set
->name
);
1711 __ip_set_put_netlink(set
);
1712 cb
->args
[IPSET_CB_ARG0
] = 0;
1716 nlmsg_end(skb
, nlh
);
1717 pr_debug("nlmsg_len: %u\n", nlh
->nlmsg_len
);
1721 return ret
< 0 ? ret
: skb
->len
;
1725 IPSET_CBFN(ip_set_dump
, struct net
*net
, struct sock
*ctnl
,
1726 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1727 const struct nlattr
* const attr
[],
1728 struct netlink_ext_ack
*extack
,
1729 const struct nfnl_info
*info
)
1731 if (unlikely(protocol_min_failed(attr
)))
1732 return -IPSET_ERR_PROTOCOL
;
1734 #if HAVE_NETLINK_DUMP_START_ARGS == 5
1735 return netlink_dump_start(ctnl
, skb
, nlh
,
1738 #elif HAVE_NETLINK_DUMP_START_ARGS == 6
1739 return netlink_dump_start(ctnl
, skb
, nlh
,
1741 ip_set_dump_done
, 0);
1744 struct netlink_dump_control c
= {
1745 #if HAVE_NETLINK_DUMP_START_ARGS == 4
1746 .start
= ip_set_dump_start
,
1748 .dump
= ip_set_dump_do
,
1749 .done
= ip_set_dump_done
,
1751 return netlink_dump_start(INFO_SK(info
, ctnl
), skb
, INFO_NLH(info
, nlh
), &c
);
1756 /* Add, del and test */
1758 static const struct nla_policy ip_set_adt_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
1759 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
1760 [IPSET_ATTR_SETNAME
] = { .type
= NLA_NUL_STRING
,
1761 .len
= IPSET_MAXNAMELEN
- 1 },
1762 [IPSET_ATTR_LINENO
] = { .type
= NLA_U32
},
1763 [IPSET_ATTR_DATA
] = { .type
= NLA_NESTED
},
1764 [IPSET_ATTR_ADT
] = { .type
= NLA_NESTED
},
1768 CALL_AD(struct net
*net
, struct sock
*ctnl
, struct sk_buff
*skb
,
1769 struct ip_set
*set
, struct nlattr
*tb
[], enum ipset_adt adt
,
1770 u32 flags
, bool use_lineno
)
1774 bool eexist
= flags
& IPSET_FLAG_EXIST
, retried
= false;
1778 __ip_set_get_netlink(set
);
1779 nfnl_unlock(NFNL_SUBSYS_IPSET
);
1781 nfnl_lock(NFNL_SUBSYS_IPSET
);
1782 __ip_set_put_netlink(set
);
1786 ret
= set
->variant
->uadt(set
, tb
, adt
, &lineno
, flags
, retried
);
1789 } while (ret
== -ERANGE
||
1791 set
->variant
->resize
&&
1792 (ret
= set
->variant
->resize(set
, retried
)) == 0));
1794 if (!ret
|| (ret
== -IPSET_ERR_EXIST
&& eexist
))
1796 if (lineno
&& use_lineno
) {
1797 /* Error in restore/batch mode: send back lineno */
1798 struct nlmsghdr
*rep
, *nlh
= nlmsg_hdr(skb
);
1799 struct sk_buff
*skb2
;
1800 struct nlmsgerr
*errmsg
;
1801 size_t payload
= min(SIZE_MAX
,
1802 sizeof(*errmsg
) + nlmsg_len(nlh
));
1803 int min_len
= nlmsg_total_size(sizeof(struct nfgenmsg
));
1804 struct nlattr
*cda
[IPSET_ATTR_CMD_MAX
+ 1];
1805 struct nlattr
*cmdattr
;
1808 skb2
= nlmsg_new(payload
, GFP_KERNEL
);
1811 rep
= nlmsg_put(skb2
, NETLINK_PORTID(skb
),
1812 nlh
->nlmsg_seq
, NLMSG_ERROR
, payload
, 0);
1813 errmsg
= nlmsg_data(rep
);
1814 errmsg
->error
= ret
;
1815 unsafe_memcpy(&errmsg
->msg
, nlh
, nlh
->nlmsg_len
,
1816 /* Bounds checked by the skb layer. */);
1817 cmdattr
= (void *)&errmsg
->msg
+ min_len
;
1819 ret
= NLA_PARSE(cda
, IPSET_ATTR_CMD_MAX
, cmdattr
,
1820 nlh
->nlmsg_len
- min_len
, ip_set_adt_policy
,
1827 errline
= nla_data(cda
[IPSET_ATTR_LINENO
]);
1831 NFNETLINK_UNICAST(ctnl
, skb2
, net
, NETLINK_PORTID(skb
));
1832 /* Signal netlink not to send its ACK/errmsg. */
1839 static int IPSET_CBFN_AD(ip_set_ad
, struct net
*net
, struct sock
*ctnl
,
1840 struct sk_buff
*skb
,
1842 const struct nlmsghdr
*nlh
,
1843 const struct nlattr
* const attr
[],
1844 struct netlink_ext_ack
*extack
, const struct nfnl_info
*info
)
1846 struct ip_set_net
*inst
= ip_set_pernet(IPSET_SOCK_NET(net
, ctnl
, info
));
1848 struct nlattr
*tb
[IPSET_ATTR_ADT_MAX
+ 1] = {};
1849 const struct nlattr
*nla
;
1850 u32 flags
= flag_exist(INFO_NLH(info
, nlh
));
1854 if (unlikely(protocol_min_failed(attr
) ||
1855 !attr
[IPSET_ATTR_SETNAME
] ||
1856 !((attr
[IPSET_ATTR_DATA
] != NULL
) ^
1857 (attr
[IPSET_ATTR_ADT
] != NULL
)) ||
1858 (attr
[IPSET_ATTR_DATA
] &&
1859 !flag_nested(attr
[IPSET_ATTR_DATA
])) ||
1860 (attr
[IPSET_ATTR_ADT
] &&
1861 (!flag_nested(attr
[IPSET_ATTR_ADT
]) ||
1862 !attr
[IPSET_ATTR_LINENO
]))))
1863 return -IPSET_ERR_PROTOCOL
;
1865 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1869 use_lineno
= !!attr
[IPSET_ATTR_LINENO
];
1870 if (attr
[IPSET_ATTR_DATA
]) {
1871 if (NLA_PARSE_NESTED(tb
, IPSET_ATTR_ADT_MAX
,
1872 attr
[IPSET_ATTR_DATA
],
1873 set
->type
->adt_policy
, NULL
))
1874 return -IPSET_ERR_PROTOCOL
;
1875 ret
= CALL_AD(net
, ctnl
, skb
, set
, tb
, adt
, flags
,
1880 nla_for_each_nested(nla
, attr
[IPSET_ATTR_ADT
], nla_rem
) {
1881 if (nla_type(nla
) != IPSET_ATTR_DATA
||
1882 !flag_nested(nla
) ||
1883 NLA_PARSE_NESTED(tb
, IPSET_ATTR_ADT_MAX
, nla
,
1884 set
->type
->adt_policy
, NULL
))
1885 return -IPSET_ERR_PROTOCOL
;
1886 ret
= CALL_AD(net
, ctnl
, skb
, set
, tb
, adt
,
1896 IPSET_CBFN(ip_set_uadd
, struct net
*net
, struct sock
*ctnl
,
1897 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1898 const struct nlattr
* const attr
[],
1899 struct netlink_ext_ack
*extack
,
1900 const struct nfnl_info
*info
)
1902 return IPSET_CBFN_AD(ip_set_ad
, INFO_NET(info
, net
), INFO_SK(info
, ctnl
), skb
,
1903 IPSET_ADD
, INFO_NLH(info
, nlh
), attr
, extack
, info
);
1907 IPSET_CBFN(ip_set_udel
, struct net
*net
, struct sock
*ctnl
,
1908 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1909 const struct nlattr
* const attr
[],
1910 struct netlink_ext_ack
*extack
,
1911 const struct nfnl_info
*info
)
1913 return IPSET_CBFN_AD(ip_set_ad
, INFO_NET(info
, net
), INFO_SK(info
, ctnl
), skb
,
1914 IPSET_DEL
, INFO_NLH(info
, nlh
), attr
, extack
, info
);
1918 IPSET_CBFN(ip_set_utest
, struct net
*net
, struct sock
*ctnl
,
1919 struct sk_buff
*skb
,
1920 const struct nlmsghdr
*nlh
,
1921 const struct nlattr
* const attr
[],
1922 struct netlink_ext_ack
*extack
,
1923 const struct nfnl_info
*info
)
1925 struct ip_set_net
*inst
= ip_set_pernet(IPSET_SOCK_NET(net
, ctnl
, info
));
1927 struct nlattr
*tb
[IPSET_ATTR_ADT_MAX
+ 1] = {};
1931 if (unlikely(protocol_min_failed(attr
) ||
1932 !attr
[IPSET_ATTR_SETNAME
] ||
1933 !attr
[IPSET_ATTR_DATA
] ||
1934 !flag_nested(attr
[IPSET_ATTR_DATA
])))
1935 return -IPSET_ERR_PROTOCOL
;
1937 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1941 if (NLA_PARSE_NESTED(tb
, IPSET_ATTR_ADT_MAX
, attr
[IPSET_ATTR_DATA
],
1942 set
->type
->adt_policy
, NULL
))
1943 return -IPSET_ERR_PROTOCOL
;
1946 ret
= set
->variant
->uadt(set
, tb
, IPSET_TEST
, &lineno
, 0, 0);
1947 rcu_read_unlock_bh();
1948 /* Userspace can't trigger element to be re-added */
1952 return ret
> 0 ? 0 : -IPSET_ERR_EXIST
;
1955 /* Get headed data of a set */
1958 IPSET_CBFN(ip_set_header
, struct net
*net
, struct sock
*ctnl
,
1959 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
1960 const struct nlattr
* const attr
[],
1961 struct netlink_ext_ack
*extack
,
1962 const struct nfnl_info
*info
)
1964 struct ip_set_net
*inst
= ip_set_pernet(IPSET_SOCK_NET(net
, ctnl
, info
));
1965 const struct ip_set
*set
;
1966 struct sk_buff
*skb2
;
1967 struct nlmsghdr
*nlh2
;
1969 if (unlikely(protocol_min_failed(attr
) ||
1970 !attr
[IPSET_ATTR_SETNAME
]))
1971 return -IPSET_ERR_PROTOCOL
;
1973 set
= find_set(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]));
1977 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
1981 nlh2
= start_msg(skb2
, NETLINK_PORTID(skb
), INFO_NLH(info
, nlh
)->nlmsg_seq
, 0,
1985 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, protocol(attr
)) ||
1986 nla_put_string(skb2
, IPSET_ATTR_SETNAME
, set
->name
) ||
1987 nla_put_string(skb2
, IPSET_ATTR_TYPENAME
, set
->type
->name
) ||
1988 nla_put_u8(skb2
, IPSET_ATTR_FAMILY
, set
->family
) ||
1989 nla_put_u8(skb2
, IPSET_ATTR_REVISION
, set
->revision
))
1990 goto nla_put_failure
;
1991 nlmsg_end(skb2
, nlh2
);
1993 return NFNETLINK_UNICAST(INFO_SK(info
, ctnl
), skb2
, INFO_NET(info
, net
), NETLINK_PORTID(skb
));
1996 nlmsg_cancel(skb2
, nlh2
);
2004 static const struct nla_policy ip_set_type_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
2005 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
2006 [IPSET_ATTR_TYPENAME
] = { .type
= NLA_NUL_STRING
,
2007 .len
= IPSET_MAXNAMELEN
- 1 },
2008 [IPSET_ATTR_FAMILY
] = { .type
= NLA_U8
},
2012 IPSET_CBFN(ip_set_type
, struct net
*net
, struct sock
*ctnl
,
2013 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
2014 const struct nlattr
* const attr
[],
2015 struct netlink_ext_ack
*extack
,
2016 const struct nfnl_info
*info
)
2018 struct sk_buff
*skb2
;
2019 struct nlmsghdr
*nlh2
;
2020 u8 family
, min
, max
;
2021 const char *typename
;
2024 if (unlikely(protocol_min_failed(attr
) ||
2025 !attr
[IPSET_ATTR_TYPENAME
] ||
2026 !attr
[IPSET_ATTR_FAMILY
]))
2027 return -IPSET_ERR_PROTOCOL
;
2029 family
= nla_get_u8(attr
[IPSET_ATTR_FAMILY
]);
2030 typename
= nla_data(attr
[IPSET_ATTR_TYPENAME
]);
2031 ret
= find_set_type_minmax(typename
, family
, &min
, &max
);
2035 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
2039 nlh2
= start_msg(skb2
, NETLINK_PORTID(skb
), INFO_NLH(info
, nlh
)->nlmsg_seq
, 0,
2043 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, protocol(attr
)) ||
2044 nla_put_string(skb2
, IPSET_ATTR_TYPENAME
, typename
) ||
2045 nla_put_u8(skb2
, IPSET_ATTR_FAMILY
, family
) ||
2046 nla_put_u8(skb2
, IPSET_ATTR_REVISION
, max
) ||
2047 nla_put_u8(skb2
, IPSET_ATTR_REVISION_MIN
, min
))
2048 goto nla_put_failure
;
2049 nlmsg_end(skb2
, nlh2
);
2051 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2
->nlmsg_len
);
2052 return NFNETLINK_UNICAST(INFO_SK(info
, ctnl
), skb2
, INFO_NET(info
, net
), NETLINK_PORTID(skb
));
2055 nlmsg_cancel(skb2
, nlh2
);
2061 /* Get protocol version */
2063 static const struct nla_policy
2064 ip_set_protocol_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
2065 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
2069 IPSET_CBFN(ip_set_protocol
, struct net
*net
, struct sock
*ctnl
,
2070 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
2071 const struct nlattr
* const attr
[],
2072 struct netlink_ext_ack
*extack
,
2073 const struct nfnl_info
*info
)
2075 struct sk_buff
*skb2
;
2076 struct nlmsghdr
*nlh2
;
2078 if (unlikely(!attr
[IPSET_ATTR_PROTOCOL
]))
2079 return -IPSET_ERR_PROTOCOL
;
2081 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
2085 nlh2
= start_msg(skb2
, NETLINK_PORTID(skb
), INFO_NLH(info
, nlh
)->nlmsg_seq
, 0,
2086 IPSET_CMD_PROTOCOL
);
2089 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, IPSET_PROTOCOL
))
2090 goto nla_put_failure
;
2091 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL_MIN
, IPSET_PROTOCOL_MIN
))
2092 goto nla_put_failure
;
2093 nlmsg_end(skb2
, nlh2
);
2095 return NFNETLINK_UNICAST(INFO_SK(info
, ctnl
), skb2
, INFO_NET(info
, net
), NETLINK_PORTID(skb
));
2098 nlmsg_cancel(skb2
, nlh2
);
2104 /* Get set by name or index, from userspace */
2107 IPSET_CBFN(ip_set_byname
, struct net
*net
, struct sock
*ctnl
,
2108 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
2109 const struct nlattr
* const attr
[],
2110 struct netlink_ext_ack
*extack
,
2111 const struct nfnl_info
*info
)
2113 struct ip_set_net
*inst
= ip_set_pernet(IPSET_SOCK_NET(net
, ctnl
, info
));
2114 struct sk_buff
*skb2
;
2115 struct nlmsghdr
*nlh2
;
2116 ip_set_id_t id
= IPSET_INVALID_ID
;
2117 const struct ip_set
*set
;
2119 if (unlikely(protocol_failed(attr
) ||
2120 !attr
[IPSET_ATTR_SETNAME
]))
2121 return -IPSET_ERR_PROTOCOL
;
2123 set
= find_set_and_id(inst
, nla_data(attr
[IPSET_ATTR_SETNAME
]), &id
);
2124 if (id
== IPSET_INVALID_ID
)
2127 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
2131 nlh2
= start_msg(skb2
, NETLINK_PORTID(skb
), INFO_NLH(info
, nlh
)->nlmsg_seq
, 0,
2132 IPSET_CMD_GET_BYNAME
);
2135 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, protocol(attr
)) ||
2136 nla_put_u8(skb2
, IPSET_ATTR_FAMILY
, set
->family
) ||
2137 nla_put_net16(skb2
, IPSET_ATTR_INDEX
, htons(id
)))
2138 goto nla_put_failure
;
2139 nlmsg_end(skb2
, nlh2
);
2141 return NFNETLINK_UNICAST(INFO_SK(info
, ctnl
), skb2
, INFO_NET(info
, net
), NETLINK_PORTID(skb
));
2144 nlmsg_cancel(skb2
, nlh2
);
2150 static const struct nla_policy ip_set_index_policy
[IPSET_ATTR_CMD_MAX
+ 1] = {
2151 [IPSET_ATTR_PROTOCOL
] = { .type
= NLA_U8
},
2152 [IPSET_ATTR_INDEX
] = { .type
= NLA_U16
},
2156 IPSET_CBFN(ip_set_byindex
, struct net
*net
, struct sock
*ctnl
,
2157 struct sk_buff
*skb
, const struct nlmsghdr
*nlh
,
2158 const struct nlattr
* const attr
[],
2159 struct netlink_ext_ack
*extack
,
2160 const struct nfnl_info
*info
)
2162 struct ip_set_net
*inst
= ip_set_pernet(IPSET_SOCK_NET(net
, ctnl
, info
));
2163 struct sk_buff
*skb2
;
2164 struct nlmsghdr
*nlh2
;
2165 ip_set_id_t id
= IPSET_INVALID_ID
;
2166 const struct ip_set
*set
;
2168 if (unlikely(protocol_failed(attr
) ||
2169 !attr
[IPSET_ATTR_INDEX
]))
2170 return -IPSET_ERR_PROTOCOL
;
2172 id
= ip_set_get_h16(attr
[IPSET_ATTR_INDEX
]);
2173 if (id
>= inst
->ip_set_max
)
2175 set
= ip_set(inst
, id
);
2179 skb2
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
2183 nlh2
= start_msg(skb2
, NETLINK_PORTID(skb
), INFO_NLH(info
, nlh
)->nlmsg_seq
, 0,
2184 IPSET_CMD_GET_BYINDEX
);
2187 if (nla_put_u8(skb2
, IPSET_ATTR_PROTOCOL
, protocol(attr
)) ||
2188 nla_put_string(skb2
, IPSET_ATTR_SETNAME
, set
->name
))
2189 goto nla_put_failure
;
2190 nlmsg_end(skb2
, nlh2
);
2192 return NFNETLINK_UNICAST(INFO_SK(info
, ctnl
), skb2
, INFO_NET(info
, net
), NETLINK_PORTID(skb
));
2195 nlmsg_cancel(skb2
, nlh2
);
2201 static const struct nfnl_callback ip_set_netlink_subsys_cb
[IPSET_MSG_MAX
] = {
2202 [IPSET_CMD_NONE
] = {
2203 .call
= ip_set_none
,
2204 SET_NFNL_CALLBACK_TYPE(NFNL_CB_MUTEX
)
2205 .attr_count
= IPSET_ATTR_CMD_MAX
,
2207 [IPSET_CMD_CREATE
] = {
2208 .call
= ip_set_create
,
2209 SET_NFNL_CALLBACK_TYPE(NFNL_CB_MUTEX
)
2210 .attr_count
= IPSET_ATTR_CMD_MAX
,
2211 .policy
= ip_set_create_policy
,
2213 [IPSET_CMD_DESTROY
] = {
2214 .call
= ip_set_destroy
,
2215 SET_NFNL_CALLBACK_TYPE(NFNL_CB_MUTEX
)
2216 .attr_count
= IPSET_ATTR_CMD_MAX
,
2217 .policy
= ip_set_setname_policy
,
2219 [IPSET_CMD_FLUSH
] = {
2220 .call
= ip_set_flush
,
2221 SET_NFNL_CALLBACK_TYPE(NFNL_CB_MUTEX
)
2222 .attr_count
= IPSET_ATTR_CMD_MAX
,
2223 .policy
= ip_set_setname_policy
,
2225 [IPSET_CMD_RENAME
] = {
2226 .call
= ip_set_rename
,
2227 SET_NFNL_CALLBACK_TYPE(NFNL_CB_MUTEX
)
2228 .attr_count
= IPSET_ATTR_CMD_MAX
,
2229 .policy
= ip_set_setname2_policy
,
2231 [IPSET_CMD_SWAP
] = {
2232 .call
= ip_set_swap
,
2233 SET_NFNL_CALLBACK_TYPE(NFNL_CB_MUTEX
)
2234 .attr_count
= IPSET_ATTR_CMD_MAX
,
2235 .policy
= ip_set_setname2_policy
,
2237 [IPSET_CMD_LIST
] = {
2238 .call
= ip_set_dump
,
2239 SET_NFNL_CALLBACK_TYPE(NFNL_CB_MUTEX
)
2240 .attr_count
= IPSET_ATTR_CMD_MAX
,
2241 .policy
= ip_set_dump_policy
,
2243 [IPSET_CMD_SAVE
] = {
2244 .call
= ip_set_dump
,
2245 SET_NFNL_CALLBACK_TYPE(NFNL_CB_MUTEX
)
2246 .attr_count
= IPSET_ATTR_CMD_MAX
,
2247 .policy
= ip_set_setname_policy
,
2250 .call
= ip_set_uadd
,
2251 SET_NFNL_CALLBACK_TYPE(NFNL_CB_MUTEX
)
2252 .attr_count
= IPSET_ATTR_CMD_MAX
,
2253 .policy
= ip_set_adt_policy
,
2256 .call
= ip_set_udel
,
2257 SET_NFNL_CALLBACK_TYPE(NFNL_CB_MUTEX
)
2258 .attr_count
= IPSET_ATTR_CMD_MAX
,
2259 .policy
= ip_set_adt_policy
,
2261 [IPSET_CMD_TEST
] = {
2262 .call
= ip_set_utest
,
2263 SET_NFNL_CALLBACK_TYPE(NFNL_CB_MUTEX
)
2264 .attr_count
= IPSET_ATTR_CMD_MAX
,
2265 .policy
= ip_set_adt_policy
,
2267 [IPSET_CMD_HEADER
] = {
2268 .call
= ip_set_header
,
2269 SET_NFNL_CALLBACK_TYPE(NFNL_CB_MUTEX
)
2270 .attr_count
= IPSET_ATTR_CMD_MAX
,
2271 .policy
= ip_set_setname_policy
,
2273 [IPSET_CMD_TYPE
] = {
2274 .call
= ip_set_type
,
2275 SET_NFNL_CALLBACK_TYPE(NFNL_CB_MUTEX
)
2276 .attr_count
= IPSET_ATTR_CMD_MAX
,
2277 .policy
= ip_set_type_policy
,
2279 [IPSET_CMD_PROTOCOL
] = {
2280 .call
= ip_set_protocol
,
2281 SET_NFNL_CALLBACK_TYPE(NFNL_CB_MUTEX
)
2282 .attr_count
= IPSET_ATTR_CMD_MAX
,
2283 .policy
= ip_set_protocol_policy
,
2285 [IPSET_CMD_GET_BYNAME
] = {
2286 .call
= ip_set_byname
,
2287 SET_NFNL_CALLBACK_TYPE(NFNL_CB_MUTEX
)
2288 .attr_count
= IPSET_ATTR_CMD_MAX
,
2289 .policy
= ip_set_setname_policy
,
2291 [IPSET_CMD_GET_BYINDEX
] = {
2292 .call
= ip_set_byindex
,
2293 SET_NFNL_CALLBACK_TYPE(NFNL_CB_MUTEX
)
2294 .attr_count
= IPSET_ATTR_CMD_MAX
,
2295 .policy
= ip_set_index_policy
,
2299 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly
= {
2301 .subsys_id
= NFNL_SUBSYS_IPSET
,
2302 .cb_count
= IPSET_MSG_MAX
,
2303 .cb
= ip_set_netlink_subsys_cb
,
2306 /* Interface to iptables/ip6tables */
2309 ip_set_sockfn_get(struct sock
*sk
, int optval
, void __user
*user
, int *len
)
2313 int copylen
= *len
, ret
= 0;
2314 struct net
*net
= sock_net(sk
);
2315 struct ip_set_net
*inst
= ip_set_pernet(net
);
2317 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
2319 if (optval
!= SO_IP_SET
)
2321 if (*len
< sizeof(unsigned int))
2324 data
= vmalloc(*len
);
2327 if (copy_from_user(data
, user
, *len
) != 0) {
2333 if (*op
< IP_SET_OP_VERSION
) {
2334 /* Check the version at the beginning of operations */
2335 struct ip_set_req_version
*req_version
= data
;
2337 if (*len
< sizeof(struct ip_set_req_version
)) {
2342 if (req_version
->version
< IPSET_PROTOCOL_MIN
) {
2349 case IP_SET_OP_VERSION
: {
2350 struct ip_set_req_version
*req_version
= data
;
2352 if (*len
!= sizeof(struct ip_set_req_version
)) {
2357 req_version
->version
= IPSET_PROTOCOL
;
2358 if (copy_to_user(user
, req_version
,
2359 sizeof(struct ip_set_req_version
)))
2363 case IP_SET_OP_GET_BYNAME
: {
2364 struct ip_set_req_get_set
*req_get
= data
;
2367 if (*len
!= sizeof(struct ip_set_req_get_set
)) {
2371 req_get
->set
.name
[IPSET_MAXNAMELEN
- 1] = '\0';
2372 nfnl_lock(NFNL_SUBSYS_IPSET
);
2373 find_set_and_id(inst
, req_get
->set
.name
, &id
);
2374 req_get
->set
.index
= id
;
2375 nfnl_unlock(NFNL_SUBSYS_IPSET
);
2378 case IP_SET_OP_GET_FNAME
: {
2379 struct ip_set_req_get_set_family
*req_get
= data
;
2382 if (*len
!= sizeof(struct ip_set_req_get_set_family
)) {
2386 req_get
->set
.name
[IPSET_MAXNAMELEN
- 1] = '\0';
2387 nfnl_lock(NFNL_SUBSYS_IPSET
);
2388 find_set_and_id(inst
, req_get
->set
.name
, &id
);
2389 req_get
->set
.index
= id
;
2390 if (id
!= IPSET_INVALID_ID
)
2391 req_get
->family
= ip_set(inst
, id
)->family
;
2392 nfnl_unlock(NFNL_SUBSYS_IPSET
);
2395 case IP_SET_OP_GET_BYINDEX
: {
2396 struct ip_set_req_get_set
*req_get
= data
;
2399 if (*len
!= sizeof(struct ip_set_req_get_set
) ||
2400 req_get
->set
.index
>= inst
->ip_set_max
) {
2404 nfnl_lock(NFNL_SUBSYS_IPSET
);
2405 set
= ip_set(inst
, req_get
->set
.index
);
2406 ret
= strscpy(req_get
->set
.name
, set
? set
->name
: "",
2408 nfnl_unlock(NFNL_SUBSYS_IPSET
);
2416 } /* end of switch(op) */
2419 if (copy_to_user(user
, data
, copylen
))
2429 static struct nf_sockopt_ops so_set __read_mostly
= {
2431 .get_optmin
= SO_IP_SET
,
2432 .get_optmax
= SO_IP_SET
+ 1,
2433 .get
= ip_set_sockfn_get
,
2434 .owner
= THIS_MODULE
,
2437 static int __net_init
2438 ip_set_net_init(struct net
*net
)
2440 struct ip_set_net
*inst
;
2441 struct ip_set
**list
;
2443 #ifdef HAVE_NET_OPS_ID
2444 inst
= ip_set_pernet(net
);
2448 inst
= kzalloc(sizeof(struct ip_set_net
), GFP_KERNEL
);
2451 err
= net_assign_generic(net
, ip_set_net_id
, inst
);
2455 inst
->ip_set_max
= max_sets
? max_sets
: CONFIG_IP_SET_MAX
;
2456 if (inst
->ip_set_max
>= IPSET_INVALID_ID
)
2457 inst
->ip_set_max
= IPSET_INVALID_ID
- 1;
2459 list
= kvcalloc(inst
->ip_set_max
, sizeof(struct ip_set
*), GFP_KERNEL
);
2461 #ifdef HAVE_NET_OPS_ID
2466 inst
->is_deleted
= false;
2467 inst
->is_destroyed
= false;
2468 rcu_assign_pointer(inst
->ip_set_list
, list
);
2471 #ifndef HAVE_NET_OPS_ID
2478 static void __net_exit
2479 ip_set_net_exit(struct net
*net
)
2481 struct ip_set_net
*inst
= ip_set_pernet(net
);
2483 struct ip_set
*set
= NULL
;
2486 inst
->is_deleted
= true; /* flag for ip_set_nfnl_put */
2488 nfnl_lock(NFNL_SUBSYS_IPSET
);
2489 for (i
= 0; i
< inst
->ip_set_max
; i
++) {
2490 set
= ip_set(inst
, i
);
2492 ip_set(inst
, i
) = NULL
;
2493 ip_set_destroy_set(set
);
2496 nfnl_unlock(NFNL_SUBSYS_IPSET
);
2497 kvfree(rcu_dereference_protected(inst
->ip_set_list
, 1));
2498 #ifndef HAVE_NET_OPS_ID
2503 static struct pernet_operations ip_set_net_ops
= {
2504 .init
= ip_set_net_init
,
2505 .exit
= ip_set_net_exit
,
2506 #ifdef HAVE_NET_OPS_ID
2507 .id
= &ip_set_net_id
,
2508 .size
= sizeof(struct ip_set_net
),
2509 #ifdef HAVE_NET_OPS_ASYNC
2515 #ifdef HAVE_NET_OPS_ID
2516 #define REGISTER_PERNET_SUBSYS(s) \
2517 register_pernet_subsys(s)
2518 #define UNREGISTER_PERNET_SUBSYS(s) \
2519 unregister_pernet_subsys(s);
2521 #define REGISTER_PERNET_SUBSYS(s) \
2522 register_pernet_gen_device(&ip_set_net_id, s)
2523 #define UNREGISTER_PERNET_SUBSYS(s) \
2524 unregister_pernet_gen_device(ip_set_net_id, s);
2531 int ret
= REGISTER_PERNET_SUBSYS(&ip_set_net_ops
);
2534 pr_err("ip_set: cannot register pernet_subsys.\n");
2538 ret
= nfnetlink_subsys_register(&ip_set_netlink_subsys
);
2540 pr_err("ip_set: cannot register with nfnetlink.\n");
2541 UNREGISTER_PERNET_SUBSYS(&ip_set_net_ops
);
2545 ret
= nf_register_sockopt(&so_set
);
2547 pr_err("SO_SET registry failed: %d\n", ret
);
2548 nfnetlink_subsys_unregister(&ip_set_netlink_subsys
);
2549 UNREGISTER_PERNET_SUBSYS(&ip_set_net_ops
);
2559 nf_unregister_sockopt(&so_set
);
2560 nfnetlink_subsys_unregister(&ip_set_netlink_subsys
);
2562 UNREGISTER_PERNET_SUBSYS(&ip_set_net_ops
);
2563 pr_debug("these are the famous last words\n");
2566 module_init(ip_set_init
);
2567 module_exit(ip_set_fini
);