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>
5 */
6
7 /* Kernel module for IP set management */
8
9 #include <linux/init.h>
10 #include <linux/module.h>
11 #include <linux/moduleparam.h>
12 #include <linux/ip.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>
19
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
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 */
28
29 struct ip_set_net {
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 */
34 };
35
36 static unsigned int ip_set_net_id __read_mostly;
37
ip_set_pernet(struct net * net)38 static inline struct ip_set_net *ip_set_pernet(struct net *net)
39 {
40 return net_generic(net, ip_set_net_id);
41 }
42
43 #define IP_SET_INC 64
44 #define STRNCMP(a, b) (strncmp(a, b, IPSET_MAXNAMELEN) == 0)
45
46 static unsigned int max_sets;
47
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);
54
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]
64
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.
68 */
69
70 static inline void
ip_set_type_lock(void)71 ip_set_type_lock(void)
72 {
73 mutex_lock(&ip_set_type_mutex);
74 }
75
76 static inline void
ip_set_type_unlock(void)77 ip_set_type_unlock(void)
78 {
79 mutex_unlock(&ip_set_type_mutex);
80 }
81
82 /* Register and deregister settype */
83
84 static struct ip_set_type *
find_set_type(const char * name,u8 family,u8 revision)85 find_set_type(const char *name, u8 family, u8 revision)
86 {
87 struct ip_set_type *type;
88
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)
95 return type;
96 return NULL;
97 }
98
99 /* Unlock, try to load a set type module and lock again */
100 static bool
load_settype(const char * name)101 load_settype(const char *name)
102 {
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);
108 return false;
109 }
110 nfnl_lock(NFNL_SUBSYS_IPSET);
111 return true;
112 }
113
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)
117
118 static int
__find_set_type_get(const char * name,u8 family,u8 revision,struct ip_set_type ** found,bool retry)119 __find_set_type_get(const char *name, u8 family, u8 revision,
120 struct ip_set_type **found, bool retry)
121 {
122 struct ip_set_type *type;
123 int err;
124
125 if (retry && !load_settype(name))
126 return -IPSET_ERR_FIND_TYPE;
127
128 rcu_read_lock();
129 *found = find_set_type(name, family, revision);
130 if (*found) {
131 err = !try_module_get((*found)->me) ? -EFAULT : 0;
132 goto unlock;
133 }
134 /* Make sure the type is already loaded
135 * but we don't support the revision
136 */
137 list_for_each_entry_rcu(type, &ip_set_type_list, list)
138 if (STRNCMP(type->name, name)) {
139 err = -IPSET_ERR_FIND_TYPE;
140 goto unlock;
141 }
142 rcu_read_unlock();
143
144 return retry ? -IPSET_ERR_FIND_TYPE :
145 __find_set_type_get(name, family, revision, found, true);
146
147 unlock:
148 rcu_read_unlock();
149 return err;
150 }
151
152 /* Find a given set type by name and family.
153 * If we succeeded, the supported minimal and maximum revisions are
154 * filled out.
155 */
156 #define find_set_type_minmax(name, family, min, max) \
157 __find_set_type_minmax(name, family, min, max, false)
158
159 static int
__find_set_type_minmax(const char * name,u8 family,u8 * min,u8 * max,bool retry)160 __find_set_type_minmax(const char *name, u8 family, u8 *min, u8 *max,
161 bool retry)
162 {
163 struct ip_set_type *type;
164 bool found = false;
165
166 if (retry && !load_settype(name))
167 return -IPSET_ERR_FIND_TYPE;
168
169 *min = 255; *max = 0;
170 rcu_read_lock();
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)) {
175 found = true;
176 if (type->revision_min < *min)
177 *min = type->revision_min;
178 if (type->revision_max > *max)
179 *max = type->revision_max;
180 }
181 rcu_read_unlock();
182 if (found)
183 return 0;
184
185 return retry ? -IPSET_ERR_FIND_TYPE :
186 __find_set_type_minmax(name, family, min, max, true);
187 }
188
189 #define family_name(f) ((f) == NFPROTO_IPV4 ? "inet" : \
190 (f) == NFPROTO_IPV6 ? "inet6" : "any")
191
192 /* Register a set type structure. The type is identified by
193 * the unique triple of name, family and revision.
194 */
195 int
ip_set_type_register(struct ip_set_type * type)196 ip_set_type_register(struct ip_set_type *type)
197 {
198 int ret = 0;
199
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);
205 return -EINVAL;
206 }
207
208 ip_set_type_lock();
209 if (find_set_type(type->name, type->family, type->revision_min)) {
210 /* Duplicate! */
211 pr_warn("ip_set type %s, family %s with revision min %u already registered!\n",
212 type->name, family_name(type->family),
213 type->revision_min);
214 ip_set_type_unlock();
215 return -EINVAL;
216 }
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();
222
223 return ret;
224 }
225 EXPORT_SYMBOL_GPL(ip_set_type_register);
226
227 /* Unregister a set type. There's a small race with ip_set_create */
228 void
ip_set_type_unregister(struct ip_set_type * type)229 ip_set_type_unregister(struct ip_set_type *type)
230 {
231 ip_set_type_lock();
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),
235 type->revision_min);
236 ip_set_type_unlock();
237 return;
238 }
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();
243
244 synchronize_rcu();
245 }
246 EXPORT_SYMBOL_GPL(ip_set_type_unregister);
247
248 /* Utility functions */
249 void *
ip_set_alloc(size_t size)250 ip_set_alloc(size_t size)
251 {
252 void *members = NULL;
253
254 if (size < KMALLOC_MAX_SIZE)
255 members = kzalloc(size, GFP_KERNEL | __GFP_NOWARN);
256
257 if (members) {
258 pr_debug("%p: allocated with kmalloc\n", members);
259 return members;
260 }
261
262 members = vzalloc(size);
263 if (!members)
264 return NULL;
265 pr_debug("%p: allocated with vmalloc\n", members);
266
267 return members;
268 }
269 EXPORT_SYMBOL_GPL(ip_set_alloc);
270
271 void
ip_set_free(void * members)272 ip_set_free(void *members)
273 {
274 pr_debug("%p: free with %s\n", members,
275 is_vmalloc_addr(members) ? "vfree" : "kfree");
276 kvfree(members);
277 }
278 EXPORT_SYMBOL_GPL(ip_set_free);
279
280 static inline bool
flag_nested(const struct nlattr * nla)281 flag_nested(const struct nlattr *nla)
282 {
283 return nla->nla_type & NLA_F_NESTED;
284 }
285
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) },
290 };
291
292 int
ip_set_get_ipaddr4(struct nlattr * nla,__be32 * ipaddr)293 ip_set_get_ipaddr4(struct nlattr *nla, __be32 *ipaddr)
294 {
295 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1];
296
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;
304
305 *ipaddr = nla_get_be32(tb[IPSET_ATTR_IPADDR_IPV4]);
306 return 0;
307 }
308 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr4);
309
310 int
ip_set_get_ipaddr6(struct nlattr * nla,union nf_inet_addr * ipaddr)311 ip_set_get_ipaddr6(struct nlattr *nla, union nf_inet_addr *ipaddr)
312 {
313 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1];
314
315 if (unlikely(!flag_nested(nla)))
316 return -IPSET_ERR_PROTOCOL;
317
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;
323
324 memcpy(ipaddr, nla_data(tb[IPSET_ATTR_IPADDR_IPV6]),
325 sizeof(struct in6_addr));
326 return 0;
327 }
328 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr6);
329
330 typedef void (*destroyer)(struct ip_set *, void *);
331 /* ipset data extension types, in size order */
332
333 const struct ip_set_ext_type ip_set_extensions[] = {
334 [IPSET_EXT_ID_COUNTER] = {
335 .type = IPSET_EXT_COUNTER,
336 .flag = IPSET_FLAG_WITH_COUNTERS,
337 .len = sizeof(struct ip_set_counter),
338 .align = __alignof__(struct ip_set_counter),
339 },
340 [IPSET_EXT_ID_TIMEOUT] = {
341 .type = IPSET_EXT_TIMEOUT,
342 .len = sizeof(unsigned long),
343 .align = __alignof__(unsigned long),
344 },
345 [IPSET_EXT_ID_SKBINFO] = {
346 .type = IPSET_EXT_SKBINFO,
347 .flag = IPSET_FLAG_WITH_SKBINFO,
348 .len = sizeof(struct ip_set_skbinfo),
349 .align = __alignof__(struct ip_set_skbinfo),
350 },
351 [IPSET_EXT_ID_COMMENT] = {
352 .type = IPSET_EXT_COMMENT | IPSET_EXT_DESTROY,
353 .flag = IPSET_FLAG_WITH_COMMENT,
354 .len = sizeof(struct ip_set_comment),
355 .align = __alignof__(struct ip_set_comment),
356 .destroy = (destroyer) ip_set_comment_free,
357 },
358 };
359 EXPORT_SYMBOL_GPL(ip_set_extensions);
360
361 static inline bool
add_extension(enum ip_set_ext_id id,u32 flags,struct nlattr * tb[])362 add_extension(enum ip_set_ext_id id, u32 flags, struct nlattr *tb[])
363 {
364 return ip_set_extensions[id].flag ?
365 (flags & ip_set_extensions[id].flag) :
366 !!tb[IPSET_ATTR_TIMEOUT];
367 }
368
369 size_t
ip_set_elem_len(struct ip_set * set,struct nlattr * tb[],size_t len,size_t align)370 ip_set_elem_len(struct ip_set *set, struct nlattr *tb[], size_t len,
371 size_t align)
372 {
373 enum ip_set_ext_id id;
374 u32 cadt_flags = 0;
375
376 if (tb[IPSET_ATTR_CADT_FLAGS])
377 cadt_flags = ip_set_get_h32(tb[IPSET_ATTR_CADT_FLAGS]);
378 if (cadt_flags & IPSET_FLAG_WITH_FORCEADD)
379 set->flags |= IPSET_CREATE_FLAG_FORCEADD;
380 if (!align)
381 align = 1;
382 for (id = 0; id < IPSET_EXT_ID_MAX; id++) {
383 if (!add_extension(id, cadt_flags, tb))
384 continue;
385 len = ALIGN(len, ip_set_extensions[id].align);
386 set->offset[id] = len;
387 set->extensions |= ip_set_extensions[id].type;
388 len += ip_set_extensions[id].len;
389 }
390 return ALIGN(len, align);
391 }
392 EXPORT_SYMBOL_GPL(ip_set_elem_len);
393
394 int
ip_set_get_extensions(struct ip_set * set,struct nlattr * tb[],struct ip_set_ext * ext)395 ip_set_get_extensions(struct ip_set *set, struct nlattr *tb[],
396 struct ip_set_ext *ext)
397 {
398 u64 fullmark;
399
400 if (unlikely(!ip_set_optattr_netorder(tb, IPSET_ATTR_TIMEOUT) ||
401 !ip_set_optattr_netorder(tb, IPSET_ATTR_PACKETS) ||
402 !ip_set_optattr_netorder(tb, IPSET_ATTR_BYTES) ||
403 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBMARK) ||
404 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBPRIO) ||
405 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBQUEUE)))
406 return -IPSET_ERR_PROTOCOL;
407
408 if (tb[IPSET_ATTR_TIMEOUT]) {
409 if (!SET_WITH_TIMEOUT(set))
410 return -IPSET_ERR_TIMEOUT;
411 ext->timeout = ip_set_timeout_uget(tb[IPSET_ATTR_TIMEOUT]);
412 }
413 if (tb[IPSET_ATTR_BYTES] || tb[IPSET_ATTR_PACKETS]) {
414 if (!SET_WITH_COUNTER(set))
415 return -IPSET_ERR_COUNTER;
416 if (tb[IPSET_ATTR_BYTES])
417 ext->bytes = be64_to_cpu(nla_get_be64(
418 tb[IPSET_ATTR_BYTES]));
419 if (tb[IPSET_ATTR_PACKETS])
420 ext->packets = be64_to_cpu(nla_get_be64(
421 tb[IPSET_ATTR_PACKETS]));
422 }
423 if (tb[IPSET_ATTR_COMMENT]) {
424 if (!SET_WITH_COMMENT(set))
425 return -IPSET_ERR_COMMENT;
426 ext->comment = ip_set_comment_uget(tb[IPSET_ATTR_COMMENT]);
427 }
428 if (tb[IPSET_ATTR_SKBMARK]) {
429 if (!SET_WITH_SKBINFO(set))
430 return -IPSET_ERR_SKBINFO;
431 fullmark = be64_to_cpu(nla_get_be64(tb[IPSET_ATTR_SKBMARK]));
432 ext->skbinfo.skbmark = fullmark >> 32;
433 ext->skbinfo.skbmarkmask = fullmark & 0xffffffff;
434 }
435 if (tb[IPSET_ATTR_SKBPRIO]) {
436 if (!SET_WITH_SKBINFO(set))
437 return -IPSET_ERR_SKBINFO;
438 ext->skbinfo.skbprio =
439 be32_to_cpu(nla_get_be32(tb[IPSET_ATTR_SKBPRIO]));
440 }
441 if (tb[IPSET_ATTR_SKBQUEUE]) {
442 if (!SET_WITH_SKBINFO(set))
443 return -IPSET_ERR_SKBINFO;
444 ext->skbinfo.skbqueue =
445 be16_to_cpu(nla_get_be16(tb[IPSET_ATTR_SKBQUEUE]));
446 }
447 return 0;
448 }
449 EXPORT_SYMBOL_GPL(ip_set_get_extensions);
450
451 int
ip_set_put_extensions(struct sk_buff * skb,const struct ip_set * set,const void * e,bool active)452 ip_set_put_extensions(struct sk_buff *skb, const struct ip_set *set,
453 const void *e, bool active)
454 {
455 if (SET_WITH_TIMEOUT(set)) {
456 unsigned long *timeout = ext_timeout(e, set);
457
458 if (nla_put_net32(skb, IPSET_ATTR_TIMEOUT,
459 htonl(active ? ip_set_timeout_get(timeout)
460 : *timeout)))
461 return -EMSGSIZE;
462 }
463 if (SET_WITH_COUNTER(set) &&
464 ip_set_put_counter(skb, ext_counter(e, set)))
465 return -EMSGSIZE;
466 if (SET_WITH_COMMENT(set) &&
467 ip_set_put_comment(skb, ext_comment(e, set)))
468 return -EMSGSIZE;
469 if (SET_WITH_SKBINFO(set) &&
470 ip_set_put_skbinfo(skb, ext_skbinfo(e, set)))
471 return -EMSGSIZE;
472 return 0;
473 }
474 EXPORT_SYMBOL_GPL(ip_set_put_extensions);
475
476 bool
ip_set_match_extensions(struct ip_set * set,const struct ip_set_ext * ext,struct ip_set_ext * mext,u32 flags,void * data)477 ip_set_match_extensions(struct ip_set *set, const struct ip_set_ext *ext,
478 struct ip_set_ext *mext, u32 flags, void *data)
479 {
480 if (SET_WITH_TIMEOUT(set) &&
481 ip_set_timeout_expired(ext_timeout(data, set)))
482 return false;
483 if (SET_WITH_COUNTER(set)) {
484 struct ip_set_counter *counter = ext_counter(data, set);
485
486 if (flags & IPSET_FLAG_MATCH_COUNTERS &&
487 !(ip_set_match_counter(ip_set_get_packets(counter),
488 mext->packets, mext->packets_op) &&
489 ip_set_match_counter(ip_set_get_bytes(counter),
490 mext->bytes, mext->bytes_op)))
491 return false;
492 ip_set_update_counter(counter, ext, flags);
493 }
494 if (SET_WITH_SKBINFO(set))
495 ip_set_get_skbinfo(ext_skbinfo(data, set),
496 ext, mext, flags);
497 return true;
498 }
499 EXPORT_SYMBOL_GPL(ip_set_match_extensions);
500
501 /* Creating/destroying/renaming/swapping affect the existence and
502 * the properties of a set. All of these can be executed from userspace
503 * only and serialized by the nfnl mutex indirectly from nfnetlink.
504 *
505 * Sets are identified by their index in ip_set_list and the index
506 * is used by the external references (set/SET netfilter modules).
507 *
508 * The set behind an index may change by swapping only, from userspace.
509 */
510
511 static inline void
__ip_set_get(struct ip_set * set)512 __ip_set_get(struct ip_set *set)
513 {
514 write_lock_bh(&ip_set_ref_lock);
515 set->ref++;
516 write_unlock_bh(&ip_set_ref_lock);
517 }
518
519 static inline void
__ip_set_put(struct ip_set * set)520 __ip_set_put(struct ip_set *set)
521 {
522 write_lock_bh(&ip_set_ref_lock);
523 BUG_ON(set->ref == 0);
524 set->ref--;
525 write_unlock_bh(&ip_set_ref_lock);
526 }
527
528 /* set->ref can be swapped out by ip_set_swap, netlink events (like dump) need
529 * a separate reference counter
530 */
531 static inline void
__ip_set_put_netlink(struct ip_set * set)532 __ip_set_put_netlink(struct ip_set *set)
533 {
534 write_lock_bh(&ip_set_ref_lock);
535 BUG_ON(set->ref_netlink == 0);
536 set->ref_netlink--;
537 write_unlock_bh(&ip_set_ref_lock);
538 }
539
540 /* Add, del and test set entries from kernel.
541 *
542 * The set behind the index must exist and must be referenced
543 * so it can't be destroyed (or changed) under our foot.
544 */
545
546 static inline struct ip_set *
ip_set_rcu_get(struct net * net,ip_set_id_t index)547 ip_set_rcu_get(struct net *net, ip_set_id_t index)
548 {
549 struct ip_set *set;
550 struct ip_set_net *inst = ip_set_pernet(net);
551
552 rcu_read_lock();
553 /* ip_set_list itself needs to be protected */
554 set = rcu_dereference(inst->ip_set_list)[index];
555 rcu_read_unlock();
556
557 return set;
558 }
559
560 int
ip_set_test(ip_set_id_t index,const struct sk_buff * skb,const struct xt_action_param * par,struct ip_set_adt_opt * opt)561 ip_set_test(ip_set_id_t index, const struct sk_buff *skb,
562 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
563 {
564 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
565 int ret = 0;
566
567 BUG_ON(!set);
568 pr_debug("set %s, index %u\n", set->name, index);
569
570 if (opt->dim < set->type->dimension ||
571 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
572 return 0;
573
574 rcu_read_lock_bh();
575 ret = set->variant->kadt(set, skb, par, IPSET_TEST, opt);
576 rcu_read_unlock_bh();
577
578 if (ret == -EAGAIN) {
579 /* Type requests element to be completed */
580 pr_debug("element must be completed, ADD is triggered\n");
581 spin_lock_bh(&set->lock);
582 set->variant->kadt(set, skb, par, IPSET_ADD, opt);
583 spin_unlock_bh(&set->lock);
584 ret = 1;
585 } else {
586 /* --return-nomatch: invert matched element */
587 if ((opt->cmdflags & IPSET_FLAG_RETURN_NOMATCH) &&
588 (set->type->features & IPSET_TYPE_NOMATCH) &&
589 (ret > 0 || ret == -ENOTEMPTY))
590 ret = -ret;
591 }
592
593 /* Convert error codes to nomatch */
594 return (ret < 0 ? 0 : ret);
595 }
596 EXPORT_SYMBOL_GPL(ip_set_test);
597
598 int
ip_set_add(ip_set_id_t index,const struct sk_buff * skb,const struct xt_action_param * par,struct ip_set_adt_opt * opt)599 ip_set_add(ip_set_id_t index, const struct sk_buff *skb,
600 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
601 {
602 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
603 int ret;
604
605 BUG_ON(!set);
606 pr_debug("set %s, index %u\n", set->name, index);
607
608 if (opt->dim < set->type->dimension ||
609 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
610 return -IPSET_ERR_TYPE_MISMATCH;
611
612 spin_lock_bh(&set->lock);
613 ret = set->variant->kadt(set, skb, par, IPSET_ADD, opt);
614 spin_unlock_bh(&set->lock);
615
616 return ret;
617 }
618 EXPORT_SYMBOL_GPL(ip_set_add);
619
620 int
ip_set_del(ip_set_id_t index,const struct sk_buff * skb,const struct xt_action_param * par,struct ip_set_adt_opt * opt)621 ip_set_del(ip_set_id_t index, const struct sk_buff *skb,
622 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
623 {
624 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
625 int ret = 0;
626
627 BUG_ON(!set);
628 pr_debug("set %s, index %u\n", set->name, index);
629
630 if (opt->dim < set->type->dimension ||
631 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
632 return -IPSET_ERR_TYPE_MISMATCH;
633
634 spin_lock_bh(&set->lock);
635 ret = set->variant->kadt(set, skb, par, IPSET_DEL, opt);
636 spin_unlock_bh(&set->lock);
637
638 return ret;
639 }
640 EXPORT_SYMBOL_GPL(ip_set_del);
641
642 /* Find set by name, reference it once. The reference makes sure the
643 * thing pointed to, does not go away under our feet.
644 *
645 */
646 ip_set_id_t
ip_set_get_byname(struct net * net,const char * name,struct ip_set ** set)647 ip_set_get_byname(struct net *net, const char *name, struct ip_set **set)
648 {
649 ip_set_id_t i, index = IPSET_INVALID_ID;
650 struct ip_set *s;
651 struct ip_set_net *inst = ip_set_pernet(net);
652
653 rcu_read_lock();
654 for (i = 0; i < inst->ip_set_max; i++) {
655 s = rcu_dereference(inst->ip_set_list)[i];
656 if (s && STRNCMP(s->name, name)) {
657 __ip_set_get(s);
658 index = i;
659 *set = s;
660 break;
661 }
662 }
663 rcu_read_unlock();
664
665 return index;
666 }
667 EXPORT_SYMBOL_GPL(ip_set_get_byname);
668
669 /* If the given set pointer points to a valid set, decrement
670 * reference count by 1. The caller shall not assume the index
671 * to be valid, after calling this function.
672 *
673 */
674
675 static inline void
__ip_set_put_byindex(struct ip_set_net * inst,ip_set_id_t index)676 __ip_set_put_byindex(struct ip_set_net *inst, ip_set_id_t index)
677 {
678 struct ip_set *set;
679
680 rcu_read_lock();
681 set = rcu_dereference(inst->ip_set_list)[index];
682 if (set)
683 __ip_set_put(set);
684 rcu_read_unlock();
685 }
686
687 void
ip_set_put_byindex(struct net * net,ip_set_id_t index)688 ip_set_put_byindex(struct net *net, ip_set_id_t index)
689 {
690 struct ip_set_net *inst = ip_set_pernet(net);
691
692 __ip_set_put_byindex(inst, index);
693 }
694 EXPORT_SYMBOL_GPL(ip_set_put_byindex);
695
696 /* Get the name of a set behind a set index.
697 * Set itself is protected by RCU, but its name isn't: to protect against
698 * renaming, grab ip_set_ref_lock as reader (see ip_set_rename()) and copy the
699 * name.
700 */
701 void
ip_set_name_byindex(struct net * net,ip_set_id_t index,char * name)702 ip_set_name_byindex(struct net *net, ip_set_id_t index, char *name)
703 {
704 struct ip_set *set = ip_set_rcu_get(net, index);
705
706 BUG_ON(!set);
707
708 read_lock_bh(&ip_set_ref_lock);
709 strncpy(name, set->name, IPSET_MAXNAMELEN);
710 read_unlock_bh(&ip_set_ref_lock);
711 }
712 EXPORT_SYMBOL_GPL(ip_set_name_byindex);
713
714 /* Routines to call by external subsystems, which do not
715 * call nfnl_lock for us.
716 */
717
718 /* Find set by index, reference it once. The reference makes sure the
719 * thing pointed to, does not go away under our feet.
720 *
721 * The nfnl mutex is used in the function.
722 */
723 ip_set_id_t
ip_set_nfnl_get_byindex(struct net * net,ip_set_id_t index)724 ip_set_nfnl_get_byindex(struct net *net, ip_set_id_t index)
725 {
726 struct ip_set *set;
727 struct ip_set_net *inst = ip_set_pernet(net);
728
729 if (index >= inst->ip_set_max)
730 return IPSET_INVALID_ID;
731
732 nfnl_lock(NFNL_SUBSYS_IPSET);
733 set = ip_set(inst, index);
734 if (set)
735 __ip_set_get(set);
736 else
737 index = IPSET_INVALID_ID;
738 nfnl_unlock(NFNL_SUBSYS_IPSET);
739
740 return index;
741 }
742 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex);
743
744 /* If the given set pointer points to a valid set, decrement
745 * reference count by 1. The caller shall not assume the index
746 * to be valid, after calling this function.
747 *
748 * The nfnl mutex is used in the function.
749 */
750 void
ip_set_nfnl_put(struct net * net,ip_set_id_t index)751 ip_set_nfnl_put(struct net *net, ip_set_id_t index)
752 {
753 struct ip_set *set;
754 struct ip_set_net *inst = ip_set_pernet(net);
755
756 nfnl_lock(NFNL_SUBSYS_IPSET);
757 if (!inst->is_deleted) { /* already deleted from ip_set_net_exit() */
758 set = ip_set(inst, index);
759 if (set)
760 __ip_set_put(set);
761 }
762 nfnl_unlock(NFNL_SUBSYS_IPSET);
763 }
764 EXPORT_SYMBOL_GPL(ip_set_nfnl_put);
765
766 /* Communication protocol with userspace over netlink.
767 *
768 * The commands are serialized by the nfnl mutex.
769 */
770
protocol(const struct nlattr * const tb[])771 static inline u8 protocol(const struct nlattr * const tb[])
772 {
773 return nla_get_u8(tb[IPSET_ATTR_PROTOCOL]);
774 }
775
776 static inline bool
protocol_failed(const struct nlattr * const tb[])777 protocol_failed(const struct nlattr * const tb[])
778 {
779 return !tb[IPSET_ATTR_PROTOCOL] || protocol(tb) != IPSET_PROTOCOL;
780 }
781
782 static inline bool
protocol_min_failed(const struct nlattr * const tb[])783 protocol_min_failed(const struct nlattr * const tb[])
784 {
785 return !tb[IPSET_ATTR_PROTOCOL] || protocol(tb) < IPSET_PROTOCOL_MIN;
786 }
787
788 static inline u32
flag_exist(const struct nlmsghdr * nlh)789 flag_exist(const struct nlmsghdr *nlh)
790 {
791 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST;
792 }
793
794 static struct nlmsghdr *
start_msg(struct sk_buff * skb,u32 portid,u32 seq,unsigned int flags,enum ipset_cmd cmd)795 start_msg(struct sk_buff *skb, u32 portid, u32 seq, unsigned int flags,
796 enum ipset_cmd cmd)
797 {
798 struct nlmsghdr *nlh;
799 struct nfgenmsg *nfmsg;
800
801 nlh = nlmsg_put(skb, portid, seq, nfnl_msg_type(NFNL_SUBSYS_IPSET, cmd),
802 sizeof(*nfmsg), flags);
803 if (!nlh)
804 return NULL;
805
806 nfmsg = nlmsg_data(nlh);
807 nfmsg->nfgen_family = NFPROTO_IPV4;
808 nfmsg->version = NFNETLINK_V0;
809 nfmsg->res_id = 0;
810
811 return nlh;
812 }
813
814 /* Create a set */
815
816 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = {
817 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
818 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
819 .len = IPSET_MAXNAMELEN - 1 },
820 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
821 .len = IPSET_MAXNAMELEN - 1},
822 [IPSET_ATTR_REVISION] = { .type = NLA_U8 },
823 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
824 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
825 };
826
827 static struct ip_set *
find_set_and_id(struct ip_set_net * inst,const char * name,ip_set_id_t * id)828 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id)
829 {
830 struct ip_set *set = NULL;
831 ip_set_id_t i;
832
833 *id = IPSET_INVALID_ID;
834 for (i = 0; i < inst->ip_set_max; i++) {
835 set = ip_set(inst, i);
836 if (set && STRNCMP(set->name, name)) {
837 *id = i;
838 break;
839 }
840 }
841 return (*id == IPSET_INVALID_ID ? NULL : set);
842 }
843
844 static inline struct ip_set *
find_set(struct ip_set_net * inst,const char * name)845 find_set(struct ip_set_net *inst, const char *name)
846 {
847 ip_set_id_t id;
848
849 return find_set_and_id(inst, name, &id);
850 }
851
852 static int
find_free_id(struct ip_set_net * inst,const char * name,ip_set_id_t * index,struct ip_set ** set)853 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index,
854 struct ip_set **set)
855 {
856 struct ip_set *s;
857 ip_set_id_t i;
858
859 *index = IPSET_INVALID_ID;
860 for (i = 0; i < inst->ip_set_max; i++) {
861 s = ip_set(inst, i);
862 if (!s) {
863 if (*index == IPSET_INVALID_ID)
864 *index = i;
865 } else if (STRNCMP(name, s->name)) {
866 /* Name clash */
867 *set = s;
868 return -EEXIST;
869 }
870 }
871 if (*index == IPSET_INVALID_ID)
872 /* No free slot remained */
873 return -IPSET_ERR_MAX_SETS;
874 return 0;
875 }
876
ip_set_none(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[],struct netlink_ext_ack * extack)877 static int ip_set_none(struct net *net, struct sock *ctnl, struct sk_buff *skb,
878 const struct nlmsghdr *nlh,
879 const struct nlattr * const attr[],
880 struct netlink_ext_ack *extack)
881 {
882 return -EOPNOTSUPP;
883 }
884
ip_set_create(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[],struct netlink_ext_ack * extack)885 static int ip_set_create(struct net *net, struct sock *ctnl,
886 struct sk_buff *skb, const struct nlmsghdr *nlh,
887 const struct nlattr * const attr[],
888 struct netlink_ext_ack *extack)
889 {
890 struct ip_set_net *inst = ip_set_pernet(net);
891 struct ip_set *set, *clash = NULL;
892 ip_set_id_t index = IPSET_INVALID_ID;
893 struct nlattr *tb[IPSET_ATTR_CREATE_MAX + 1] = {};
894 const char *name, *typename;
895 u8 family, revision;
896 u32 flags = flag_exist(nlh);
897 int ret = 0;
898
899 if (unlikely(protocol_min_failed(attr) ||
900 !attr[IPSET_ATTR_SETNAME] ||
901 !attr[IPSET_ATTR_TYPENAME] ||
902 !attr[IPSET_ATTR_REVISION] ||
903 !attr[IPSET_ATTR_FAMILY] ||
904 (attr[IPSET_ATTR_DATA] &&
905 !flag_nested(attr[IPSET_ATTR_DATA]))))
906 return -IPSET_ERR_PROTOCOL;
907
908 name = nla_data(attr[IPSET_ATTR_SETNAME]);
909 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
910 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
911 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]);
912 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
913 name, typename, family_name(family), revision);
914
915 /* First, and without any locks, allocate and initialize
916 * a normal base set structure.
917 */
918 set = kzalloc(sizeof(*set), GFP_KERNEL);
919 if (!set)
920 return -ENOMEM;
921 spin_lock_init(&set->lock);
922 strlcpy(set->name, name, IPSET_MAXNAMELEN);
923 set->family = family;
924 set->revision = revision;
925
926 /* Next, check that we know the type, and take
927 * a reference on the type, to make sure it stays available
928 * while constructing our new set.
929 *
930 * After referencing the type, we try to create the type
931 * specific part of the set without holding any locks.
932 */
933 ret = find_set_type_get(typename, family, revision, &set->type);
934 if (ret)
935 goto out;
936
937 /* Without holding any locks, create private part. */
938 if (attr[IPSET_ATTR_DATA] &&
939 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA],
940 set->type->create_policy, NULL)) {
941 ret = -IPSET_ERR_PROTOCOL;
942 goto put_out;
943 }
944
945 ret = set->type->create(net, set, tb, flags);
946 if (ret != 0)
947 goto put_out;
948
949 /* BTW, ret==0 here. */
950
951 /* Here, we have a valid, constructed set and we are protected
952 * by the nfnl mutex. Find the first free index in ip_set_list
953 * and check clashing.
954 */
955 ret = find_free_id(inst, set->name, &index, &clash);
956 if (ret == -EEXIST) {
957 /* If this is the same set and requested, ignore error */
958 if ((flags & IPSET_FLAG_EXIST) &&
959 STRNCMP(set->type->name, clash->type->name) &&
960 set->type->family == clash->type->family &&
961 set->type->revision_min == clash->type->revision_min &&
962 set->type->revision_max == clash->type->revision_max &&
963 set->variant->same_set(set, clash))
964 ret = 0;
965 goto cleanup;
966 } else if (ret == -IPSET_ERR_MAX_SETS) {
967 struct ip_set **list, **tmp;
968 ip_set_id_t i = inst->ip_set_max + IP_SET_INC;
969
970 if (i < inst->ip_set_max || i == IPSET_INVALID_ID)
971 /* Wraparound */
972 goto cleanup;
973
974 list = kvcalloc(i, sizeof(struct ip_set *), GFP_KERNEL);
975 if (!list)
976 goto cleanup;
977 /* nfnl mutex is held, both lists are valid */
978 tmp = ip_set_dereference(inst->ip_set_list);
979 memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max);
980 rcu_assign_pointer(inst->ip_set_list, list);
981 /* Make sure all current packets have passed through */
982 synchronize_net();
983 /* Use new list */
984 index = inst->ip_set_max;
985 inst->ip_set_max = i;
986 kvfree(tmp);
987 ret = 0;
988 } else if (ret) {
989 goto cleanup;
990 }
991
992 /* Finally! Add our shiny new set to the list, and be done. */
993 pr_debug("create: '%s' created with index %u!\n", set->name, index);
994 ip_set(inst, index) = set;
995
996 return ret;
997
998 cleanup:
999 set->variant->destroy(set);
1000 put_out:
1001 module_put(set->type->me);
1002 out:
1003 kfree(set);
1004 return ret;
1005 }
1006
1007 /* Destroy sets */
1008
1009 static const struct nla_policy
1010 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = {
1011 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1012 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1013 .len = IPSET_MAXNAMELEN - 1 },
1014 };
1015
1016 static void
ip_set_destroy_set(struct ip_set * set)1017 ip_set_destroy_set(struct ip_set *set)
1018 {
1019 pr_debug("set: %s\n", set->name);
1020
1021 /* Must call it without holding any lock */
1022 set->variant->destroy(set);
1023 module_put(set->type->me);
1024 kfree(set);
1025 }
1026
ip_set_destroy(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[],struct netlink_ext_ack * extack)1027 static int ip_set_destroy(struct net *net, struct sock *ctnl,
1028 struct sk_buff *skb, const struct nlmsghdr *nlh,
1029 const struct nlattr * const attr[],
1030 struct netlink_ext_ack *extack)
1031 {
1032 struct ip_set_net *inst = ip_set_pernet(net);
1033 struct ip_set *s;
1034 ip_set_id_t i;
1035 int ret = 0;
1036
1037 if (unlikely(protocol_min_failed(attr)))
1038 return -IPSET_ERR_PROTOCOL;
1039
1040 /* Must wait for flush to be really finished in list:set */
1041 rcu_barrier();
1042
1043 /* Commands are serialized and references are
1044 * protected by the ip_set_ref_lock.
1045 * External systems (i.e. xt_set) must call
1046 * ip_set_put|get_nfnl_* functions, that way we
1047 * can safely check references here.
1048 *
1049 * list:set timer can only decrement the reference
1050 * counter, so if it's already zero, we can proceed
1051 * without holding the lock.
1052 */
1053 read_lock_bh(&ip_set_ref_lock);
1054 if (!attr[IPSET_ATTR_SETNAME]) {
1055 for (i = 0; i < inst->ip_set_max; i++) {
1056 s = ip_set(inst, i);
1057 if (s && (s->ref || s->ref_netlink)) {
1058 ret = -IPSET_ERR_BUSY;
1059 goto out;
1060 }
1061 }
1062 inst->is_destroyed = true;
1063 read_unlock_bh(&ip_set_ref_lock);
1064 for (i = 0; i < inst->ip_set_max; i++) {
1065 s = ip_set(inst, i);
1066 if (s) {
1067 ip_set(inst, i) = NULL;
1068 ip_set_destroy_set(s);
1069 }
1070 }
1071 /* Modified by ip_set_destroy() only, which is serialized */
1072 inst->is_destroyed = false;
1073 } else {
1074 s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1075 &i);
1076 if (!s) {
1077 ret = -ENOENT;
1078 goto out;
1079 } else if (s->ref || s->ref_netlink) {
1080 ret = -IPSET_ERR_BUSY;
1081 goto out;
1082 }
1083 ip_set(inst, i) = NULL;
1084 read_unlock_bh(&ip_set_ref_lock);
1085
1086 ip_set_destroy_set(s);
1087 }
1088 return 0;
1089 out:
1090 read_unlock_bh(&ip_set_ref_lock);
1091 return ret;
1092 }
1093
1094 /* Flush sets */
1095
1096 static void
ip_set_flush_set(struct ip_set * set)1097 ip_set_flush_set(struct ip_set *set)
1098 {
1099 pr_debug("set: %s\n", set->name);
1100
1101 spin_lock_bh(&set->lock);
1102 set->variant->flush(set);
1103 spin_unlock_bh(&set->lock);
1104 }
1105
ip_set_flush(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[],struct netlink_ext_ack * extack)1106 static int ip_set_flush(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1107 const struct nlmsghdr *nlh,
1108 const struct nlattr * const attr[],
1109 struct netlink_ext_ack *extack)
1110 {
1111 struct ip_set_net *inst = ip_set_pernet(net);
1112 struct ip_set *s;
1113 ip_set_id_t i;
1114
1115 if (unlikely(protocol_min_failed(attr)))
1116 return -IPSET_ERR_PROTOCOL;
1117
1118 if (!attr[IPSET_ATTR_SETNAME]) {
1119 for (i = 0; i < inst->ip_set_max; i++) {
1120 s = ip_set(inst, i);
1121 if (s)
1122 ip_set_flush_set(s);
1123 }
1124 } else {
1125 s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1126 if (!s)
1127 return -ENOENT;
1128
1129 ip_set_flush_set(s);
1130 }
1131
1132 return 0;
1133 }
1134
1135 /* Rename a set */
1136
1137 static const struct nla_policy
1138 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = {
1139 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1140 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1141 .len = IPSET_MAXNAMELEN - 1 },
1142 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING,
1143 .len = IPSET_MAXNAMELEN - 1 },
1144 };
1145
ip_set_rename(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[],struct netlink_ext_ack * extack)1146 static int ip_set_rename(struct net *net, struct sock *ctnl,
1147 struct sk_buff *skb, const struct nlmsghdr *nlh,
1148 const struct nlattr * const attr[],
1149 struct netlink_ext_ack *extack)
1150 {
1151 struct ip_set_net *inst = ip_set_pernet(net);
1152 struct ip_set *set, *s;
1153 const char *name2;
1154 ip_set_id_t i;
1155 int ret = 0;
1156
1157 if (unlikely(protocol_min_failed(attr) ||
1158 !attr[IPSET_ATTR_SETNAME] ||
1159 !attr[IPSET_ATTR_SETNAME2]))
1160 return -IPSET_ERR_PROTOCOL;
1161
1162 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1163 if (!set)
1164 return -ENOENT;
1165
1166 write_lock_bh(&ip_set_ref_lock);
1167 if (set->ref != 0 || set->ref_netlink != 0) {
1168 ret = -IPSET_ERR_REFERENCED;
1169 goto out;
1170 }
1171
1172 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]);
1173 for (i = 0; i < inst->ip_set_max; i++) {
1174 s = ip_set(inst, i);
1175 if (s && STRNCMP(s->name, name2)) {
1176 ret = -IPSET_ERR_EXIST_SETNAME2;
1177 goto out;
1178 }
1179 }
1180 strncpy(set->name, name2, IPSET_MAXNAMELEN);
1181
1182 out:
1183 write_unlock_bh(&ip_set_ref_lock);
1184 return ret;
1185 }
1186
1187 /* Swap two sets so that name/index points to the other.
1188 * References and set names are also swapped.
1189 *
1190 * The commands are serialized by the nfnl mutex and references are
1191 * protected by the ip_set_ref_lock. The kernel interfaces
1192 * do not hold the mutex but the pointer settings are atomic
1193 * so the ip_set_list always contains valid pointers to the sets.
1194 */
1195
ip_set_swap(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[],struct netlink_ext_ack * extack)1196 static int ip_set_swap(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1197 const struct nlmsghdr *nlh,
1198 const struct nlattr * const attr[],
1199 struct netlink_ext_ack *extack)
1200 {
1201 struct ip_set_net *inst = ip_set_pernet(net);
1202 struct ip_set *from, *to;
1203 ip_set_id_t from_id, to_id;
1204 char from_name[IPSET_MAXNAMELEN];
1205
1206 if (unlikely(protocol_min_failed(attr) ||
1207 !attr[IPSET_ATTR_SETNAME] ||
1208 !attr[IPSET_ATTR_SETNAME2]))
1209 return -IPSET_ERR_PROTOCOL;
1210
1211 from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1212 &from_id);
1213 if (!from)
1214 return -ENOENT;
1215
1216 to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]),
1217 &to_id);
1218 if (!to)
1219 return -IPSET_ERR_EXIST_SETNAME2;
1220
1221 /* Features must not change.
1222 * Not an artifical restriction anymore, as we must prevent
1223 * possible loops created by swapping in setlist type of sets.
1224 */
1225 if (!(from->type->features == to->type->features &&
1226 from->family == to->family))
1227 return -IPSET_ERR_TYPE_MISMATCH;
1228
1229 write_lock_bh(&ip_set_ref_lock);
1230
1231 if (from->ref_netlink || to->ref_netlink) {
1232 write_unlock_bh(&ip_set_ref_lock);
1233 return -EBUSY;
1234 }
1235
1236 strncpy(from_name, from->name, IPSET_MAXNAMELEN);
1237 strncpy(from->name, to->name, IPSET_MAXNAMELEN);
1238 strncpy(to->name, from_name, IPSET_MAXNAMELEN);
1239
1240 swap(from->ref, to->ref);
1241 ip_set(inst, from_id) = to;
1242 ip_set(inst, to_id) = from;
1243 write_unlock_bh(&ip_set_ref_lock);
1244
1245 return 0;
1246 }
1247
1248 /* List/save set data */
1249
1250 #define DUMP_INIT 0
1251 #define DUMP_ALL 1
1252 #define DUMP_ONE 2
1253 #define DUMP_LAST 3
1254
1255 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1256 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1257
1258 static int
ip_set_dump_done(struct netlink_callback * cb)1259 ip_set_dump_done(struct netlink_callback *cb)
1260 {
1261 if (cb->args[IPSET_CB_ARG0]) {
1262 struct ip_set_net *inst =
1263 (struct ip_set_net *)cb->args[IPSET_CB_NET];
1264 ip_set_id_t index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1265 struct ip_set *set = ip_set_ref_netlink(inst, index);
1266
1267 if (set->variant->uref)
1268 set->variant->uref(set, cb, false);
1269 pr_debug("release set %s\n", set->name);
1270 __ip_set_put_netlink(set);
1271 }
1272 return 0;
1273 }
1274
1275 static inline void
dump_attrs(struct nlmsghdr * nlh)1276 dump_attrs(struct nlmsghdr *nlh)
1277 {
1278 const struct nlattr *attr;
1279 int rem;
1280
1281 pr_debug("dump nlmsg\n");
1282 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) {
1283 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len);
1284 }
1285 }
1286
1287 static const struct nla_policy
1288 ip_set_dump_policy[IPSET_ATTR_CMD_MAX + 1] = {
1289 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1290 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1291 .len = IPSET_MAXNAMELEN - 1 },
1292 [IPSET_ATTR_FLAGS] = { .type = NLA_U32 },
1293 };
1294
1295 static int
dump_init(struct netlink_callback * cb,struct ip_set_net * inst)1296 dump_init(struct netlink_callback *cb, struct ip_set_net *inst)
1297 {
1298 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb);
1299 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1300 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1301 struct nlattr *attr = (void *)nlh + min_len;
1302 u32 dump_type;
1303 ip_set_id_t index;
1304 int ret;
1305
1306 ret = nla_parse(cda, IPSET_ATTR_CMD_MAX, attr,
1307 nlh->nlmsg_len - min_len,
1308 ip_set_dump_policy, NULL);
1309 if (ret)
1310 return ret;
1311
1312 cb->args[IPSET_CB_PROTO] = nla_get_u8(cda[IPSET_ATTR_PROTOCOL]);
1313 if (cda[IPSET_ATTR_SETNAME]) {
1314 struct ip_set *set;
1315
1316 set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]),
1317 &index);
1318 if (!set)
1319 return -ENOENT;
1320
1321 dump_type = DUMP_ONE;
1322 cb->args[IPSET_CB_INDEX] = index;
1323 } else {
1324 dump_type = DUMP_ALL;
1325 }
1326
1327 if (cda[IPSET_ATTR_FLAGS]) {
1328 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]);
1329
1330 dump_type |= (f << 16);
1331 }
1332 cb->args[IPSET_CB_NET] = (unsigned long)inst;
1333 cb->args[IPSET_CB_DUMP] = dump_type;
1334
1335 return 0;
1336 }
1337
1338 static int
ip_set_dump_start(struct sk_buff * skb,struct netlink_callback * cb)1339 ip_set_dump_start(struct sk_buff *skb, struct netlink_callback *cb)
1340 {
1341 ip_set_id_t index = IPSET_INVALID_ID, max;
1342 struct ip_set *set = NULL;
1343 struct nlmsghdr *nlh = NULL;
1344 unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0;
1345 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1346 u32 dump_type, dump_flags;
1347 bool is_destroyed;
1348 int ret = 0;
1349
1350 if (!cb->args[IPSET_CB_DUMP]) {
1351 ret = dump_init(cb, inst);
1352 if (ret < 0) {
1353 nlh = nlmsg_hdr(cb->skb);
1354 /* We have to create and send the error message
1355 * manually :-(
1356 */
1357 if (nlh->nlmsg_flags & NLM_F_ACK)
1358 netlink_ack(cb->skb, nlh, ret, NULL);
1359 return ret;
1360 }
1361 }
1362
1363 if (cb->args[IPSET_CB_INDEX] >= inst->ip_set_max)
1364 goto out;
1365
1366 dump_type = DUMP_TYPE(cb->args[IPSET_CB_DUMP]);
1367 dump_flags = DUMP_FLAGS(cb->args[IPSET_CB_DUMP]);
1368 max = dump_type == DUMP_ONE ? cb->args[IPSET_CB_INDEX] + 1
1369 : inst->ip_set_max;
1370 dump_last:
1371 pr_debug("dump type, flag: %u %u index: %ld\n",
1372 dump_type, dump_flags, cb->args[IPSET_CB_INDEX]);
1373 for (; cb->args[IPSET_CB_INDEX] < max; cb->args[IPSET_CB_INDEX]++) {
1374 index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1375 write_lock_bh(&ip_set_ref_lock);
1376 set = ip_set(inst, index);
1377 is_destroyed = inst->is_destroyed;
1378 if (!set || is_destroyed) {
1379 write_unlock_bh(&ip_set_ref_lock);
1380 if (dump_type == DUMP_ONE) {
1381 ret = -ENOENT;
1382 goto out;
1383 }
1384 if (is_destroyed) {
1385 /* All sets are just being destroyed */
1386 ret = 0;
1387 goto out;
1388 }
1389 continue;
1390 }
1391 /* When dumping all sets, we must dump "sorted"
1392 * so that lists (unions of sets) are dumped last.
1393 */
1394 if (dump_type != DUMP_ONE &&
1395 ((dump_type == DUMP_ALL) ==
1396 !!(set->type->features & IPSET_DUMP_LAST))) {
1397 write_unlock_bh(&ip_set_ref_lock);
1398 continue;
1399 }
1400 pr_debug("List set: %s\n", set->name);
1401 if (!cb->args[IPSET_CB_ARG0]) {
1402 /* Start listing: make sure set won't be destroyed */
1403 pr_debug("reference set\n");
1404 set->ref_netlink++;
1405 }
1406 write_unlock_bh(&ip_set_ref_lock);
1407 nlh = start_msg(skb, NETLINK_CB(cb->skb).portid,
1408 cb->nlh->nlmsg_seq, flags,
1409 IPSET_CMD_LIST);
1410 if (!nlh) {
1411 ret = -EMSGSIZE;
1412 goto release_refcount;
1413 }
1414 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL,
1415 cb->args[IPSET_CB_PROTO]) ||
1416 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name))
1417 goto nla_put_failure;
1418 if (dump_flags & IPSET_FLAG_LIST_SETNAME)
1419 goto next_set;
1420 switch (cb->args[IPSET_CB_ARG0]) {
1421 case 0:
1422 /* Core header data */
1423 if (nla_put_string(skb, IPSET_ATTR_TYPENAME,
1424 set->type->name) ||
1425 nla_put_u8(skb, IPSET_ATTR_FAMILY,
1426 set->family) ||
1427 nla_put_u8(skb, IPSET_ATTR_REVISION,
1428 set->revision))
1429 goto nla_put_failure;
1430 if (cb->args[IPSET_CB_PROTO] > IPSET_PROTOCOL_MIN &&
1431 nla_put_net16(skb, IPSET_ATTR_INDEX, htons(index)))
1432 goto nla_put_failure;
1433 ret = set->variant->head(set, skb);
1434 if (ret < 0)
1435 goto release_refcount;
1436 if (dump_flags & IPSET_FLAG_LIST_HEADER)
1437 goto next_set;
1438 if (set->variant->uref)
1439 set->variant->uref(set, cb, true);
1440 /* fall through */
1441 default:
1442 ret = set->variant->list(set, skb, cb);
1443 if (!cb->args[IPSET_CB_ARG0])
1444 /* Set is done, proceed with next one */
1445 goto next_set;
1446 goto release_refcount;
1447 }
1448 }
1449 /* If we dump all sets, continue with dumping last ones */
1450 if (dump_type == DUMP_ALL) {
1451 dump_type = DUMP_LAST;
1452 cb->args[IPSET_CB_DUMP] = dump_type | (dump_flags << 16);
1453 cb->args[IPSET_CB_INDEX] = 0;
1454 if (set && set->variant->uref)
1455 set->variant->uref(set, cb, false);
1456 goto dump_last;
1457 }
1458 goto out;
1459
1460 nla_put_failure:
1461 ret = -EFAULT;
1462 next_set:
1463 if (dump_type == DUMP_ONE)
1464 cb->args[IPSET_CB_INDEX] = IPSET_INVALID_ID;
1465 else
1466 cb->args[IPSET_CB_INDEX]++;
1467 release_refcount:
1468 /* If there was an error or set is done, release set */
1469 if (ret || !cb->args[IPSET_CB_ARG0]) {
1470 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);
1475 cb->args[IPSET_CB_ARG0] = 0;
1476 }
1477 out:
1478 if (nlh) {
1479 nlmsg_end(skb, nlh);
1480 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len);
1481 dump_attrs(nlh);
1482 }
1483
1484 return ret < 0 ? ret : skb->len;
1485 }
1486
ip_set_dump(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[],struct netlink_ext_ack * extack)1487 static int ip_set_dump(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1488 const struct nlmsghdr *nlh,
1489 const struct nlattr * const attr[],
1490 struct netlink_ext_ack *extack)
1491 {
1492 if (unlikely(protocol_min_failed(attr)))
1493 return -IPSET_ERR_PROTOCOL;
1494
1495 {
1496 struct netlink_dump_control c = {
1497 .dump = ip_set_dump_start,
1498 .done = ip_set_dump_done,
1499 };
1500 return netlink_dump_start(ctnl, skb, nlh, &c);
1501 }
1502 }
1503
1504 /* Add, del and test */
1505
1506 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = {
1507 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1508 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1509 .len = IPSET_MAXNAMELEN - 1 },
1510 [IPSET_ATTR_LINENO] = { .type = NLA_U32 },
1511 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
1512 [IPSET_ATTR_ADT] = { .type = NLA_NESTED },
1513 };
1514
1515 static int
call_ad(struct sock * ctnl,struct sk_buff * skb,struct ip_set * set,struct nlattr * tb[],enum ipset_adt adt,u32 flags,bool use_lineno)1516 call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set,
1517 struct nlattr *tb[], enum ipset_adt adt,
1518 u32 flags, bool use_lineno)
1519 {
1520 int ret;
1521 u32 lineno = 0;
1522 bool eexist = flags & IPSET_FLAG_EXIST, retried = false;
1523
1524 do {
1525 spin_lock_bh(&set->lock);
1526 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried);
1527 spin_unlock_bh(&set->lock);
1528 retried = true;
1529 } while (ret == -EAGAIN &&
1530 set->variant->resize &&
1531 (ret = set->variant->resize(set, retried)) == 0);
1532
1533 if (!ret || (ret == -IPSET_ERR_EXIST && eexist))
1534 return 0;
1535 if (lineno && use_lineno) {
1536 /* Error in restore/batch mode: send back lineno */
1537 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb);
1538 struct sk_buff *skb2;
1539 struct nlmsgerr *errmsg;
1540 size_t payload = min(SIZE_MAX,
1541 sizeof(*errmsg) + nlmsg_len(nlh));
1542 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1543 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1544 struct nlattr *cmdattr;
1545 u32 *errline;
1546
1547 skb2 = nlmsg_new(payload, GFP_KERNEL);
1548 if (!skb2)
1549 return -ENOMEM;
1550 rep = __nlmsg_put(skb2, NETLINK_CB(skb).portid,
1551 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0);
1552 errmsg = nlmsg_data(rep);
1553 errmsg->error = ret;
1554 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len);
1555 cmdattr = (void *)&errmsg->msg + min_len;
1556
1557 ret = nla_parse(cda, IPSET_ATTR_CMD_MAX, cmdattr,
1558 nlh->nlmsg_len - min_len, ip_set_adt_policy,
1559 NULL);
1560
1561 if (ret) {
1562 nlmsg_free(skb2);
1563 return ret;
1564 }
1565 errline = nla_data(cda[IPSET_ATTR_LINENO]);
1566
1567 *errline = lineno;
1568
1569 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid,
1570 MSG_DONTWAIT);
1571 /* Signal netlink not to send its ACK/errmsg. */
1572 return -EINTR;
1573 }
1574
1575 return ret;
1576 }
1577
ip_set_ad(struct net * net,struct sock * ctnl,struct sk_buff * skb,enum ipset_adt adt,const struct nlmsghdr * nlh,const struct nlattr * const attr[],struct netlink_ext_ack * extack)1578 static int ip_set_ad(struct net *net, struct sock *ctnl,
1579 struct sk_buff *skb,
1580 enum ipset_adt adt,
1581 const struct nlmsghdr *nlh,
1582 const struct nlattr * const attr[],
1583 struct netlink_ext_ack *extack)
1584 {
1585 struct ip_set_net *inst = ip_set_pernet(net);
1586 struct ip_set *set;
1587 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1588 const struct nlattr *nla;
1589 u32 flags = flag_exist(nlh);
1590 bool use_lineno;
1591 int ret = 0;
1592
1593 if (unlikely(protocol_min_failed(attr) ||
1594 !attr[IPSET_ATTR_SETNAME] ||
1595 !((attr[IPSET_ATTR_DATA] != NULL) ^
1596 (attr[IPSET_ATTR_ADT] != NULL)) ||
1597 (attr[IPSET_ATTR_DATA] &&
1598 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1599 (attr[IPSET_ATTR_ADT] &&
1600 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1601 !attr[IPSET_ATTR_LINENO]))))
1602 return -IPSET_ERR_PROTOCOL;
1603
1604 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1605 if (!set)
1606 return -ENOENT;
1607
1608 use_lineno = !!attr[IPSET_ATTR_LINENO];
1609 if (attr[IPSET_ATTR_DATA]) {
1610 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1611 attr[IPSET_ATTR_DATA],
1612 set->type->adt_policy, NULL))
1613 return -IPSET_ERR_PROTOCOL;
1614 ret = call_ad(ctnl, skb, set, tb, adt, flags,
1615 use_lineno);
1616 } else {
1617 int nla_rem;
1618
1619 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1620 if (nla_type(nla) != IPSET_ATTR_DATA ||
1621 !flag_nested(nla) ||
1622 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1623 set->type->adt_policy, NULL))
1624 return -IPSET_ERR_PROTOCOL;
1625 ret = call_ad(ctnl, skb, set, tb, adt,
1626 flags, use_lineno);
1627 if (ret < 0)
1628 return ret;
1629 }
1630 }
1631 return ret;
1632 }
1633
ip_set_uadd(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[],struct netlink_ext_ack * extack)1634 static int ip_set_uadd(struct net *net, struct sock *ctnl,
1635 struct sk_buff *skb, const struct nlmsghdr *nlh,
1636 const struct nlattr * const attr[],
1637 struct netlink_ext_ack *extack)
1638 {
1639 return ip_set_ad(net, ctnl, skb,
1640 IPSET_ADD, nlh, attr, extack);
1641 }
1642
ip_set_udel(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[],struct netlink_ext_ack * extack)1643 static int ip_set_udel(struct net *net, struct sock *ctnl,
1644 struct sk_buff *skb, const struct nlmsghdr *nlh,
1645 const struct nlattr * const attr[],
1646 struct netlink_ext_ack *extack)
1647 {
1648 return ip_set_ad(net, ctnl, skb,
1649 IPSET_DEL, nlh, attr, extack);
1650 }
1651
ip_set_utest(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[],struct netlink_ext_ack * extack)1652 static int ip_set_utest(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1653 const struct nlmsghdr *nlh,
1654 const struct nlattr * const attr[],
1655 struct netlink_ext_ack *extack)
1656 {
1657 struct ip_set_net *inst = ip_set_pernet(net);
1658 struct ip_set *set;
1659 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1660 int ret = 0;
1661 u32 lineno;
1662
1663 if (unlikely(protocol_min_failed(attr) ||
1664 !attr[IPSET_ATTR_SETNAME] ||
1665 !attr[IPSET_ATTR_DATA] ||
1666 !flag_nested(attr[IPSET_ATTR_DATA])))
1667 return -IPSET_ERR_PROTOCOL;
1668
1669 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1670 if (!set)
1671 return -ENOENT;
1672
1673 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA],
1674 set->type->adt_policy, NULL))
1675 return -IPSET_ERR_PROTOCOL;
1676
1677 rcu_read_lock_bh();
1678 ret = set->variant->uadt(set, tb, IPSET_TEST, &lineno, 0, 0);
1679 rcu_read_unlock_bh();
1680 /* Userspace can't trigger element to be re-added */
1681 if (ret == -EAGAIN)
1682 ret = 1;
1683
1684 return ret > 0 ? 0 : -IPSET_ERR_EXIST;
1685 }
1686
1687 /* Get headed data of a set */
1688
ip_set_header(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[],struct netlink_ext_ack * extack)1689 static int ip_set_header(struct net *net, struct sock *ctnl,
1690 struct sk_buff *skb, const struct nlmsghdr *nlh,
1691 const struct nlattr * const attr[],
1692 struct netlink_ext_ack *extack)
1693 {
1694 struct ip_set_net *inst = ip_set_pernet(net);
1695 const struct ip_set *set;
1696 struct sk_buff *skb2;
1697 struct nlmsghdr *nlh2;
1698 int ret = 0;
1699
1700 if (unlikely(protocol_min_failed(attr) ||
1701 !attr[IPSET_ATTR_SETNAME]))
1702 return -IPSET_ERR_PROTOCOL;
1703
1704 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1705 if (!set)
1706 return -ENOENT;
1707
1708 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1709 if (!skb2)
1710 return -ENOMEM;
1711
1712 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1713 IPSET_CMD_HEADER);
1714 if (!nlh2)
1715 goto nlmsg_failure;
1716 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
1717 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) ||
1718 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) ||
1719 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
1720 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision))
1721 goto nla_put_failure;
1722 nlmsg_end(skb2, nlh2);
1723
1724 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1725 if (ret < 0)
1726 return ret;
1727
1728 return 0;
1729
1730 nla_put_failure:
1731 nlmsg_cancel(skb2, nlh2);
1732 nlmsg_failure:
1733 kfree_skb(skb2);
1734 return -EMSGSIZE;
1735 }
1736
1737 /* Get type data */
1738
1739 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = {
1740 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1741 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
1742 .len = IPSET_MAXNAMELEN - 1 },
1743 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
1744 };
1745
ip_set_type(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[],struct netlink_ext_ack * extack)1746 static int ip_set_type(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1747 const struct nlmsghdr *nlh,
1748 const struct nlattr * const attr[],
1749 struct netlink_ext_ack *extack)
1750 {
1751 struct sk_buff *skb2;
1752 struct nlmsghdr *nlh2;
1753 u8 family, min, max;
1754 const char *typename;
1755 int ret = 0;
1756
1757 if (unlikely(protocol_min_failed(attr) ||
1758 !attr[IPSET_ATTR_TYPENAME] ||
1759 !attr[IPSET_ATTR_FAMILY]))
1760 return -IPSET_ERR_PROTOCOL;
1761
1762 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1763 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1764 ret = find_set_type_minmax(typename, family, &min, &max);
1765 if (ret)
1766 return ret;
1767
1768 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1769 if (!skb2)
1770 return -ENOMEM;
1771
1772 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1773 IPSET_CMD_TYPE);
1774 if (!nlh2)
1775 goto nlmsg_failure;
1776 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
1777 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) ||
1778 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) ||
1779 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) ||
1780 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min))
1781 goto nla_put_failure;
1782 nlmsg_end(skb2, nlh2);
1783
1784 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len);
1785 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1786 if (ret < 0)
1787 return ret;
1788
1789 return 0;
1790
1791 nla_put_failure:
1792 nlmsg_cancel(skb2, nlh2);
1793 nlmsg_failure:
1794 kfree_skb(skb2);
1795 return -EMSGSIZE;
1796 }
1797
1798 /* Get protocol version */
1799
1800 static const struct nla_policy
1801 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = {
1802 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1803 };
1804
ip_set_protocol(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[],struct netlink_ext_ack * extack)1805 static int ip_set_protocol(struct net *net, struct sock *ctnl,
1806 struct sk_buff *skb, const struct nlmsghdr *nlh,
1807 const struct nlattr * const attr[],
1808 struct netlink_ext_ack *extack)
1809 {
1810 struct sk_buff *skb2;
1811 struct nlmsghdr *nlh2;
1812 int ret = 0;
1813
1814 if (unlikely(!attr[IPSET_ATTR_PROTOCOL]))
1815 return -IPSET_ERR_PROTOCOL;
1816
1817 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1818 if (!skb2)
1819 return -ENOMEM;
1820
1821 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1822 IPSET_CMD_PROTOCOL);
1823 if (!nlh2)
1824 goto nlmsg_failure;
1825 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL))
1826 goto nla_put_failure;
1827 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL_MIN, IPSET_PROTOCOL_MIN))
1828 goto nla_put_failure;
1829 nlmsg_end(skb2, nlh2);
1830
1831 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1832 if (ret < 0)
1833 return ret;
1834
1835 return 0;
1836
1837 nla_put_failure:
1838 nlmsg_cancel(skb2, nlh2);
1839 nlmsg_failure:
1840 kfree_skb(skb2);
1841 return -EMSGSIZE;
1842 }
1843
1844 /* Get set by name or index, from userspace */
1845
ip_set_byname(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[],struct netlink_ext_ack * extack)1846 static int ip_set_byname(struct net *net, struct sock *ctnl,
1847 struct sk_buff *skb, const struct nlmsghdr *nlh,
1848 const struct nlattr * const attr[],
1849 struct netlink_ext_ack *extack)
1850 {
1851 struct ip_set_net *inst = ip_set_pernet(net);
1852 struct sk_buff *skb2;
1853 struct nlmsghdr *nlh2;
1854 ip_set_id_t id = IPSET_INVALID_ID;
1855 const struct ip_set *set;
1856 int ret = 0;
1857
1858 if (unlikely(protocol_failed(attr) ||
1859 !attr[IPSET_ATTR_SETNAME]))
1860 return -IPSET_ERR_PROTOCOL;
1861
1862 set = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]), &id);
1863 if (id == IPSET_INVALID_ID)
1864 return -ENOENT;
1865
1866 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1867 if (!skb2)
1868 return -ENOMEM;
1869
1870 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1871 IPSET_CMD_GET_BYNAME);
1872 if (!nlh2)
1873 goto nlmsg_failure;
1874 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
1875 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
1876 nla_put_net16(skb2, IPSET_ATTR_INDEX, htons(id)))
1877 goto nla_put_failure;
1878 nlmsg_end(skb2, nlh2);
1879
1880 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1881 if (ret < 0)
1882 return ret;
1883
1884 return 0;
1885
1886 nla_put_failure:
1887 nlmsg_cancel(skb2, nlh2);
1888 nlmsg_failure:
1889 kfree_skb(skb2);
1890 return -EMSGSIZE;
1891 }
1892
1893 static const struct nla_policy ip_set_index_policy[IPSET_ATTR_CMD_MAX + 1] = {
1894 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1895 [IPSET_ATTR_INDEX] = { .type = NLA_U16 },
1896 };
1897
ip_set_byindex(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[],struct netlink_ext_ack * extack)1898 static int ip_set_byindex(struct net *net, struct sock *ctnl,
1899 struct sk_buff *skb, const struct nlmsghdr *nlh,
1900 const struct nlattr * const attr[],
1901 struct netlink_ext_ack *extack)
1902 {
1903 struct ip_set_net *inst = ip_set_pernet(net);
1904 struct sk_buff *skb2;
1905 struct nlmsghdr *nlh2;
1906 ip_set_id_t id = IPSET_INVALID_ID;
1907 const struct ip_set *set;
1908 int ret = 0;
1909
1910 if (unlikely(protocol_failed(attr) ||
1911 !attr[IPSET_ATTR_INDEX]))
1912 return -IPSET_ERR_PROTOCOL;
1913
1914 id = ip_set_get_h16(attr[IPSET_ATTR_INDEX]);
1915 if (id >= inst->ip_set_max)
1916 return -ENOENT;
1917 set = ip_set(inst, id);
1918 if (set == NULL)
1919 return -ENOENT;
1920
1921 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1922 if (!skb2)
1923 return -ENOMEM;
1924
1925 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1926 IPSET_CMD_GET_BYINDEX);
1927 if (!nlh2)
1928 goto nlmsg_failure;
1929 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
1930 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name))
1931 goto nla_put_failure;
1932 nlmsg_end(skb2, nlh2);
1933
1934 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1935 if (ret < 0)
1936 return ret;
1937
1938 return 0;
1939
1940 nla_put_failure:
1941 nlmsg_cancel(skb2, nlh2);
1942 nlmsg_failure:
1943 kfree_skb(skb2);
1944 return -EMSGSIZE;
1945 }
1946
1947 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = {
1948 [IPSET_CMD_NONE] = {
1949 .call = ip_set_none,
1950 .attr_count = IPSET_ATTR_CMD_MAX,
1951 },
1952 [IPSET_CMD_CREATE] = {
1953 .call = ip_set_create,
1954 .attr_count = IPSET_ATTR_CMD_MAX,
1955 .policy = ip_set_create_policy,
1956 },
1957 [IPSET_CMD_DESTROY] = {
1958 .call = ip_set_destroy,
1959 .attr_count = IPSET_ATTR_CMD_MAX,
1960 .policy = ip_set_setname_policy,
1961 },
1962 [IPSET_CMD_FLUSH] = {
1963 .call = ip_set_flush,
1964 .attr_count = IPSET_ATTR_CMD_MAX,
1965 .policy = ip_set_setname_policy,
1966 },
1967 [IPSET_CMD_RENAME] = {
1968 .call = ip_set_rename,
1969 .attr_count = IPSET_ATTR_CMD_MAX,
1970 .policy = ip_set_setname2_policy,
1971 },
1972 [IPSET_CMD_SWAP] = {
1973 .call = ip_set_swap,
1974 .attr_count = IPSET_ATTR_CMD_MAX,
1975 .policy = ip_set_setname2_policy,
1976 },
1977 [IPSET_CMD_LIST] = {
1978 .call = ip_set_dump,
1979 .attr_count = IPSET_ATTR_CMD_MAX,
1980 .policy = ip_set_dump_policy,
1981 },
1982 [IPSET_CMD_SAVE] = {
1983 .call = ip_set_dump,
1984 .attr_count = IPSET_ATTR_CMD_MAX,
1985 .policy = ip_set_setname_policy,
1986 },
1987 [IPSET_CMD_ADD] = {
1988 .call = ip_set_uadd,
1989 .attr_count = IPSET_ATTR_CMD_MAX,
1990 .policy = ip_set_adt_policy,
1991 },
1992 [IPSET_CMD_DEL] = {
1993 .call = ip_set_udel,
1994 .attr_count = IPSET_ATTR_CMD_MAX,
1995 .policy = ip_set_adt_policy,
1996 },
1997 [IPSET_CMD_TEST] = {
1998 .call = ip_set_utest,
1999 .attr_count = IPSET_ATTR_CMD_MAX,
2000 .policy = ip_set_adt_policy,
2001 },
2002 [IPSET_CMD_HEADER] = {
2003 .call = ip_set_header,
2004 .attr_count = IPSET_ATTR_CMD_MAX,
2005 .policy = ip_set_setname_policy,
2006 },
2007 [IPSET_CMD_TYPE] = {
2008 .call = ip_set_type,
2009 .attr_count = IPSET_ATTR_CMD_MAX,
2010 .policy = ip_set_type_policy,
2011 },
2012 [IPSET_CMD_PROTOCOL] = {
2013 .call = ip_set_protocol,
2014 .attr_count = IPSET_ATTR_CMD_MAX,
2015 .policy = ip_set_protocol_policy,
2016 },
2017 [IPSET_CMD_GET_BYNAME] = {
2018 .call = ip_set_byname,
2019 .attr_count = IPSET_ATTR_CMD_MAX,
2020 .policy = ip_set_setname_policy,
2021 },
2022 [IPSET_CMD_GET_BYINDEX] = {
2023 .call = ip_set_byindex,
2024 .attr_count = IPSET_ATTR_CMD_MAX,
2025 .policy = ip_set_index_policy,
2026 },
2027 };
2028
2029 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = {
2030 .name = "ip_set",
2031 .subsys_id = NFNL_SUBSYS_IPSET,
2032 .cb_count = IPSET_MSG_MAX,
2033 .cb = ip_set_netlink_subsys_cb,
2034 };
2035
2036 /* Interface to iptables/ip6tables */
2037
2038 static int
ip_set_sockfn_get(struct sock * sk,int optval,void __user * user,int * len)2039 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len)
2040 {
2041 unsigned int *op;
2042 void *data;
2043 int copylen = *len, ret = 0;
2044 struct net *net = sock_net(sk);
2045 struct ip_set_net *inst = ip_set_pernet(net);
2046
2047 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2048 return -EPERM;
2049 if (optval != SO_IP_SET)
2050 return -EBADF;
2051 if (*len < sizeof(unsigned int))
2052 return -EINVAL;
2053
2054 data = vmalloc(*len);
2055 if (!data)
2056 return -ENOMEM;
2057 if (copy_from_user(data, user, *len) != 0) {
2058 ret = -EFAULT;
2059 goto done;
2060 }
2061 op = data;
2062
2063 if (*op < IP_SET_OP_VERSION) {
2064 /* Check the version at the beginning of operations */
2065 struct ip_set_req_version *req_version = data;
2066
2067 if (*len < sizeof(struct ip_set_req_version)) {
2068 ret = -EINVAL;
2069 goto done;
2070 }
2071
2072 if (req_version->version < IPSET_PROTOCOL_MIN) {
2073 ret = -EPROTO;
2074 goto done;
2075 }
2076 }
2077
2078 switch (*op) {
2079 case IP_SET_OP_VERSION: {
2080 struct ip_set_req_version *req_version = data;
2081
2082 if (*len != sizeof(struct ip_set_req_version)) {
2083 ret = -EINVAL;
2084 goto done;
2085 }
2086
2087 req_version->version = IPSET_PROTOCOL;
2088 if (copy_to_user(user, req_version,
2089 sizeof(struct ip_set_req_version)))
2090 ret = -EFAULT;
2091 goto done;
2092 }
2093 case IP_SET_OP_GET_BYNAME: {
2094 struct ip_set_req_get_set *req_get = data;
2095 ip_set_id_t id;
2096
2097 if (*len != sizeof(struct ip_set_req_get_set)) {
2098 ret = -EINVAL;
2099 goto done;
2100 }
2101 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
2102 nfnl_lock(NFNL_SUBSYS_IPSET);
2103 find_set_and_id(inst, req_get->set.name, &id);
2104 req_get->set.index = id;
2105 nfnl_unlock(NFNL_SUBSYS_IPSET);
2106 goto copy;
2107 }
2108 case IP_SET_OP_GET_FNAME: {
2109 struct ip_set_req_get_set_family *req_get = data;
2110 ip_set_id_t id;
2111
2112 if (*len != sizeof(struct ip_set_req_get_set_family)) {
2113 ret = -EINVAL;
2114 goto done;
2115 }
2116 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
2117 nfnl_lock(NFNL_SUBSYS_IPSET);
2118 find_set_and_id(inst, req_get->set.name, &id);
2119 req_get->set.index = id;
2120 if (id != IPSET_INVALID_ID)
2121 req_get->family = ip_set(inst, id)->family;
2122 nfnl_unlock(NFNL_SUBSYS_IPSET);
2123 goto copy;
2124 }
2125 case IP_SET_OP_GET_BYINDEX: {
2126 struct ip_set_req_get_set *req_get = data;
2127 struct ip_set *set;
2128
2129 if (*len != sizeof(struct ip_set_req_get_set) ||
2130 req_get->set.index >= inst->ip_set_max) {
2131 ret = -EINVAL;
2132 goto done;
2133 }
2134 nfnl_lock(NFNL_SUBSYS_IPSET);
2135 set = ip_set(inst, req_get->set.index);
2136 ret = strscpy(req_get->set.name, set ? set->name : "",
2137 IPSET_MAXNAMELEN);
2138 nfnl_unlock(NFNL_SUBSYS_IPSET);
2139 if (ret < 0)
2140 goto done;
2141 goto copy;
2142 }
2143 default:
2144 ret = -EBADMSG;
2145 goto done;
2146 } /* end of switch(op) */
2147
2148 copy:
2149 if (copy_to_user(user, data, copylen))
2150 ret = -EFAULT;
2151
2152 done:
2153 vfree(data);
2154 if (ret > 0)
2155 ret = 0;
2156 return ret;
2157 }
2158
2159 static struct nf_sockopt_ops so_set __read_mostly = {
2160 .pf = PF_INET,
2161 .get_optmin = SO_IP_SET,
2162 .get_optmax = SO_IP_SET + 1,
2163 .get = ip_set_sockfn_get,
2164 .owner = THIS_MODULE,
2165 };
2166
2167 static int __net_init
ip_set_net_init(struct net * net)2168 ip_set_net_init(struct net *net)
2169 {
2170 struct ip_set_net *inst = ip_set_pernet(net);
2171 struct ip_set **list;
2172
2173 inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX;
2174 if (inst->ip_set_max >= IPSET_INVALID_ID)
2175 inst->ip_set_max = IPSET_INVALID_ID - 1;
2176
2177 list = kvcalloc(inst->ip_set_max, sizeof(struct ip_set *), GFP_KERNEL);
2178 if (!list)
2179 return -ENOMEM;
2180 inst->is_deleted = false;
2181 inst->is_destroyed = false;
2182 rcu_assign_pointer(inst->ip_set_list, list);
2183 return 0;
2184 }
2185
2186 static void __net_exit
ip_set_net_exit(struct net * net)2187 ip_set_net_exit(struct net *net)
2188 {
2189 struct ip_set_net *inst = ip_set_pernet(net);
2190
2191 struct ip_set *set = NULL;
2192 ip_set_id_t i;
2193
2194 inst->is_deleted = true; /* flag for ip_set_nfnl_put */
2195
2196 nfnl_lock(NFNL_SUBSYS_IPSET);
2197 for (i = 0; i < inst->ip_set_max; i++) {
2198 set = ip_set(inst, i);
2199 if (set) {
2200 ip_set(inst, i) = NULL;
2201 ip_set_destroy_set(set);
2202 }
2203 }
2204 nfnl_unlock(NFNL_SUBSYS_IPSET);
2205 kvfree(rcu_dereference_protected(inst->ip_set_list, 1));
2206 }
2207
2208 static struct pernet_operations ip_set_net_ops = {
2209 .init = ip_set_net_init,
2210 .exit = ip_set_net_exit,
2211 .id = &ip_set_net_id,
2212 .size = sizeof(struct ip_set_net),
2213 };
2214
2215 static int __init
ip_set_init(void)2216 ip_set_init(void)
2217 {
2218 int ret = register_pernet_subsys(&ip_set_net_ops);
2219
2220 if (ret) {
2221 pr_err("ip_set: cannot register pernet_subsys.\n");
2222 return ret;
2223 }
2224
2225 ret = nfnetlink_subsys_register(&ip_set_netlink_subsys);
2226 if (ret != 0) {
2227 pr_err("ip_set: cannot register with nfnetlink.\n");
2228 unregister_pernet_subsys(&ip_set_net_ops);
2229 return ret;
2230 }
2231
2232 ret = nf_register_sockopt(&so_set);
2233 if (ret != 0) {
2234 pr_err("SO_SET registry failed: %d\n", ret);
2235 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2236 unregister_pernet_subsys(&ip_set_net_ops);
2237 return ret;
2238 }
2239
2240 return 0;
2241 }
2242
2243 static void __exit
ip_set_fini(void)2244 ip_set_fini(void)
2245 {
2246 nf_unregister_sockopt(&so_set);
2247 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2248
2249 unregister_pernet_subsys(&ip_set_net_ops);
2250 pr_debug("these are the famous last words\n");
2251 }
2252
2253 module_init(ip_set_init);
2254 module_exit(ip_set_fini);
2255
2256 MODULE_DESCRIPTION("ip_set: protocol " __stringify(IPSET_PROTOCOL));
2257