1 /* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
2 * Patrick Schaaf <bof@bof.de>
3 * Copyright (C) 2003-2013 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License version 2 as
7 * published by the Free Software Foundation.
8 */
9
10 /* Kernel module for IP set management */
11
12 #include <linux/init.h>
13 #include <linux/module.h>
14 #include <linux/moduleparam.h>
15 #include <linux/ip.h>
16 #include <linux/skbuff.h>
17 #include <linux/spinlock.h>
18 #include <linux/rculist.h>
19 #include <net/netlink.h>
20 #include <net/net_namespace.h>
21 #include <net/netns/generic.h>
22
23 #include <linux/netfilter.h>
24 #include <linux/netfilter/x_tables.h>
25 #include <linux/netfilter/nfnetlink.h>
26 #include <linux/netfilter/ipset/ip_set.h>
27
28 static LIST_HEAD(ip_set_type_list); /* all registered set types */
29 static DEFINE_MUTEX(ip_set_type_mutex); /* protects ip_set_type_list */
30 static DEFINE_RWLOCK(ip_set_ref_lock); /* protects the set refs */
31
32 struct ip_set_net {
33 struct ip_set * __rcu *ip_set_list; /* all individual sets */
34 ip_set_id_t ip_set_max; /* max number of sets */
35 bool is_deleted; /* deleted by ip_set_net_exit */
36 bool is_destroyed; /* all sets are destroyed */
37 };
38
39 static int ip_set_net_id __read_mostly;
40
ip_set_pernet(struct net * net)41 static inline struct ip_set_net *ip_set_pernet(struct net *net)
42 {
43 return net_generic(net, ip_set_net_id);
44 }
45
46 #define IP_SET_INC 64
47 #define STRNCMP(a, b) (strncmp(a, b, IPSET_MAXNAMELEN) == 0)
48
49 static unsigned int max_sets;
50
51 module_param(max_sets, int, 0600);
52 MODULE_PARM_DESC(max_sets, "maximal number of sets");
53 MODULE_LICENSE("GPL");
54 MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
55 MODULE_DESCRIPTION("core IP set support");
56 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_IPSET);
57
58 /* When the nfnl mutex is held: */
59 #define ip_set_dereference(p) \
60 rcu_dereference_protected(p, 1)
61 #define ip_set(inst, id) \
62 ip_set_dereference((inst)->ip_set_list)[id]
63
64 /* The set types are implemented in modules and registered set types
65 * can be found in ip_set_type_list. Adding/deleting types is
66 * serialized by ip_set_type_mutex.
67 */
68
69 static inline void
ip_set_type_lock(void)70 ip_set_type_lock(void)
71 {
72 mutex_lock(&ip_set_type_mutex);
73 }
74
75 static inline void
ip_set_type_unlock(void)76 ip_set_type_unlock(void)
77 {
78 mutex_unlock(&ip_set_type_mutex);
79 }
80
81 /* Register and deregister settype */
82
83 static struct ip_set_type *
find_set_type(const char * name,u8 family,u8 revision)84 find_set_type(const char *name, u8 family, u8 revision)
85 {
86 struct ip_set_type *type;
87
88 list_for_each_entry_rcu(type, &ip_set_type_list, list)
89 if (STRNCMP(type->name, name) &&
90 (type->family == family ||
91 type->family == NFPROTO_UNSPEC) &&
92 revision >= type->revision_min &&
93 revision <= type->revision_max)
94 return type;
95 return NULL;
96 }
97
98 /* Unlock, try to load a set type module and lock again */
99 static bool
load_settype(const char * name)100 load_settype(const char *name)
101 {
102 nfnl_unlock(NFNL_SUBSYS_IPSET);
103 pr_debug("try to load ip_set_%s\n", name);
104 if (request_module("ip_set_%s", name) < 0) {
105 pr_warn("Can't find ip_set type %s\n", name);
106 nfnl_lock(NFNL_SUBSYS_IPSET);
107 return false;
108 }
109 nfnl_lock(NFNL_SUBSYS_IPSET);
110 return true;
111 }
112
113 /* Find a set type and reference it */
114 #define find_set_type_get(name, family, revision, found) \
115 __find_set_type_get(name, family, revision, found, false)
116
117 static int
__find_set_type_get(const char * name,u8 family,u8 revision,struct ip_set_type ** found,bool retry)118 __find_set_type_get(const char *name, u8 family, u8 revision,
119 struct ip_set_type **found, bool retry)
120 {
121 struct ip_set_type *type;
122 int err;
123
124 if (retry && !load_settype(name))
125 return -IPSET_ERR_FIND_TYPE;
126
127 rcu_read_lock();
128 *found = find_set_type(name, family, revision);
129 if (*found) {
130 err = !try_module_get((*found)->me) ? -EFAULT : 0;
131 goto unlock;
132 }
133 /* Make sure the type is already loaded
134 * but we don't support the revision
135 */
136 list_for_each_entry_rcu(type, &ip_set_type_list, list)
137 if (STRNCMP(type->name, name)) {
138 err = -IPSET_ERR_FIND_TYPE;
139 goto unlock;
140 }
141 rcu_read_unlock();
142
143 return retry ? -IPSET_ERR_FIND_TYPE :
144 __find_set_type_get(name, family, revision, found, true);
145
146 unlock:
147 rcu_read_unlock();
148 return err;
149 }
150
151 /* Find a given set type by name and family.
152 * If we succeeded, the supported minimal and maximum revisions are
153 * filled out.
154 */
155 #define find_set_type_minmax(name, family, min, max) \
156 __find_set_type_minmax(name, family, min, max, false)
157
158 static int
__find_set_type_minmax(const char * name,u8 family,u8 * min,u8 * max,bool retry)159 __find_set_type_minmax(const char *name, u8 family, u8 *min, u8 *max,
160 bool retry)
161 {
162 struct ip_set_type *type;
163 bool found = false;
164
165 if (retry && !load_settype(name))
166 return -IPSET_ERR_FIND_TYPE;
167
168 *min = 255; *max = 0;
169 rcu_read_lock();
170 list_for_each_entry_rcu(type, &ip_set_type_list, list)
171 if (STRNCMP(type->name, name) &&
172 (type->family == family ||
173 type->family == NFPROTO_UNSPEC)) {
174 found = true;
175 if (type->revision_min < *min)
176 *min = type->revision_min;
177 if (type->revision_max > *max)
178 *max = type->revision_max;
179 }
180 rcu_read_unlock();
181 if (found)
182 return 0;
183
184 return retry ? -IPSET_ERR_FIND_TYPE :
185 __find_set_type_minmax(name, family, min, max, true);
186 }
187
188 #define family_name(f) ((f) == NFPROTO_IPV4 ? "inet" : \
189 (f) == NFPROTO_IPV6 ? "inet6" : "any")
190
191 /* Register a set type structure. The type is identified by
192 * the unique triple of name, family and revision.
193 */
194 int
ip_set_type_register(struct ip_set_type * type)195 ip_set_type_register(struct ip_set_type *type)
196 {
197 int ret = 0;
198
199 if (type->protocol != IPSET_PROTOCOL) {
200 pr_warn("ip_set type %s, family %s, revision %u:%u uses wrong protocol version %u (want %u)\n",
201 type->name, family_name(type->family),
202 type->revision_min, type->revision_max,
203 type->protocol, IPSET_PROTOCOL);
204 return -EINVAL;
205 }
206
207 ip_set_type_lock();
208 if (find_set_type(type->name, type->family, type->revision_min)) {
209 /* Duplicate! */
210 pr_warn("ip_set type %s, family %s with revision min %u already registered!\n",
211 type->name, family_name(type->family),
212 type->revision_min);
213 ip_set_type_unlock();
214 return -EINVAL;
215 }
216 list_add_rcu(&type->list, &ip_set_type_list);
217 pr_debug("type %s, family %s, revision %u:%u registered.\n",
218 type->name, family_name(type->family),
219 type->revision_min, type->revision_max);
220 ip_set_type_unlock();
221
222 return ret;
223 }
224 EXPORT_SYMBOL_GPL(ip_set_type_register);
225
226 /* Unregister a set type. There's a small race with ip_set_create */
227 void
ip_set_type_unregister(struct ip_set_type * type)228 ip_set_type_unregister(struct ip_set_type *type)
229 {
230 ip_set_type_lock();
231 if (!find_set_type(type->name, type->family, type->revision_min)) {
232 pr_warn("ip_set type %s, family %s with revision min %u not registered\n",
233 type->name, family_name(type->family),
234 type->revision_min);
235 ip_set_type_unlock();
236 return;
237 }
238 list_del_rcu(&type->list);
239 pr_debug("type %s, family %s with revision min %u unregistered.\n",
240 type->name, family_name(type->family), type->revision_min);
241 ip_set_type_unlock();
242
243 synchronize_rcu();
244 }
245 EXPORT_SYMBOL_GPL(ip_set_type_unregister);
246
247 /* Utility functions */
248 void *
ip_set_alloc(size_t size)249 ip_set_alloc(size_t size)
250 {
251 void *members = NULL;
252
253 if (size < KMALLOC_MAX_SIZE)
254 members = kzalloc(size, GFP_KERNEL | __GFP_NOWARN);
255
256 if (members) {
257 pr_debug("%p: allocated with kmalloc\n", members);
258 return members;
259 }
260
261 members = vzalloc(size);
262 if (!members)
263 return NULL;
264 pr_debug("%p: allocated with vmalloc\n", members);
265
266 return members;
267 }
268 EXPORT_SYMBOL_GPL(ip_set_alloc);
269
270 void
ip_set_free(void * members)271 ip_set_free(void *members)
272 {
273 pr_debug("%p: free with %s\n", members,
274 is_vmalloc_addr(members) ? "vfree" : "kfree");
275 kvfree(members);
276 }
277 EXPORT_SYMBOL_GPL(ip_set_free);
278
279 static inline bool
flag_nested(const struct nlattr * nla)280 flag_nested(const struct nlattr *nla)
281 {
282 return nla->nla_type & NLA_F_NESTED;
283 }
284
285 static const struct nla_policy ipaddr_policy[IPSET_ATTR_IPADDR_MAX + 1] = {
286 [IPSET_ATTR_IPADDR_IPV4] = { .type = NLA_U32 },
287 [IPSET_ATTR_IPADDR_IPV6] = { .type = NLA_BINARY,
288 .len = sizeof(struct in6_addr) },
289 };
290
291 int
ip_set_get_ipaddr4(struct nlattr * nla,__be32 * ipaddr)292 ip_set_get_ipaddr4(struct nlattr *nla, __be32 *ipaddr)
293 {
294 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1];
295
296 if (unlikely(!flag_nested(nla)))
297 return -IPSET_ERR_PROTOCOL;
298 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla, ipaddr_policy))
299 return -IPSET_ERR_PROTOCOL;
300 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV4)))
301 return -IPSET_ERR_PROTOCOL;
302
303 *ipaddr = nla_get_be32(tb[IPSET_ATTR_IPADDR_IPV4]);
304 return 0;
305 }
306 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr4);
307
308 int
ip_set_get_ipaddr6(struct nlattr * nla,union nf_inet_addr * ipaddr)309 ip_set_get_ipaddr6(struct nlattr *nla, union nf_inet_addr *ipaddr)
310 {
311 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1];
312
313 if (unlikely(!flag_nested(nla)))
314 return -IPSET_ERR_PROTOCOL;
315
316 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla, ipaddr_policy))
317 return -IPSET_ERR_PROTOCOL;
318 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV6)))
319 return -IPSET_ERR_PROTOCOL;
320
321 memcpy(ipaddr, nla_data(tb[IPSET_ATTR_IPADDR_IPV6]),
322 sizeof(struct in6_addr));
323 return 0;
324 }
325 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr6);
326
327 typedef void (*destroyer)(void *);
328 /* ipset data extension types, in size order */
329
330 const struct ip_set_ext_type ip_set_extensions[] = {
331 [IPSET_EXT_ID_COUNTER] = {
332 .type = IPSET_EXT_COUNTER,
333 .flag = IPSET_FLAG_WITH_COUNTERS,
334 .len = sizeof(struct ip_set_counter),
335 .align = __alignof__(struct ip_set_counter),
336 },
337 [IPSET_EXT_ID_TIMEOUT] = {
338 .type = IPSET_EXT_TIMEOUT,
339 .len = sizeof(unsigned long),
340 .align = __alignof__(unsigned long),
341 },
342 [IPSET_EXT_ID_SKBINFO] = {
343 .type = IPSET_EXT_SKBINFO,
344 .flag = IPSET_FLAG_WITH_SKBINFO,
345 .len = sizeof(struct ip_set_skbinfo),
346 .align = __alignof__(struct ip_set_skbinfo),
347 },
348 [IPSET_EXT_ID_COMMENT] = {
349 .type = IPSET_EXT_COMMENT | IPSET_EXT_DESTROY,
350 .flag = IPSET_FLAG_WITH_COMMENT,
351 .len = sizeof(struct ip_set_comment),
352 .align = __alignof__(struct ip_set_comment),
353 .destroy = (destroyer) ip_set_comment_free,
354 },
355 };
356 EXPORT_SYMBOL_GPL(ip_set_extensions);
357
358 static inline bool
add_extension(enum ip_set_ext_id id,u32 flags,struct nlattr * tb[])359 add_extension(enum ip_set_ext_id id, u32 flags, struct nlattr *tb[])
360 {
361 return ip_set_extensions[id].flag ?
362 (flags & ip_set_extensions[id].flag) :
363 !!tb[IPSET_ATTR_TIMEOUT];
364 }
365
366 size_t
ip_set_elem_len(struct ip_set * set,struct nlattr * tb[],size_t len,size_t align)367 ip_set_elem_len(struct ip_set *set, struct nlattr *tb[], size_t len,
368 size_t align)
369 {
370 enum ip_set_ext_id id;
371 u32 cadt_flags = 0;
372
373 if (tb[IPSET_ATTR_CADT_FLAGS])
374 cadt_flags = ip_set_get_h32(tb[IPSET_ATTR_CADT_FLAGS]);
375 if (cadt_flags & IPSET_FLAG_WITH_FORCEADD)
376 set->flags |= IPSET_CREATE_FLAG_FORCEADD;
377 if (!align)
378 align = 1;
379 for (id = 0; id < IPSET_EXT_ID_MAX; id++) {
380 if (!add_extension(id, cadt_flags, tb))
381 continue;
382 len = ALIGN(len, ip_set_extensions[id].align);
383 set->offset[id] = len;
384 set->extensions |= ip_set_extensions[id].type;
385 len += ip_set_extensions[id].len;
386 }
387 return ALIGN(len, align);
388 }
389 EXPORT_SYMBOL_GPL(ip_set_elem_len);
390
391 int
ip_set_get_extensions(struct ip_set * set,struct nlattr * tb[],struct ip_set_ext * ext)392 ip_set_get_extensions(struct ip_set *set, struct nlattr *tb[],
393 struct ip_set_ext *ext)
394 {
395 u64 fullmark;
396
397 if (unlikely(!ip_set_optattr_netorder(tb, IPSET_ATTR_TIMEOUT) ||
398 !ip_set_optattr_netorder(tb, IPSET_ATTR_PACKETS) ||
399 !ip_set_optattr_netorder(tb, IPSET_ATTR_BYTES) ||
400 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBMARK) ||
401 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBPRIO) ||
402 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBQUEUE)))
403 return -IPSET_ERR_PROTOCOL;
404
405 if (tb[IPSET_ATTR_TIMEOUT]) {
406 if (!SET_WITH_TIMEOUT(set))
407 return -IPSET_ERR_TIMEOUT;
408 ext->timeout = ip_set_timeout_uget(tb[IPSET_ATTR_TIMEOUT]);
409 }
410 if (tb[IPSET_ATTR_BYTES] || tb[IPSET_ATTR_PACKETS]) {
411 if (!SET_WITH_COUNTER(set))
412 return -IPSET_ERR_COUNTER;
413 if (tb[IPSET_ATTR_BYTES])
414 ext->bytes = be64_to_cpu(nla_get_be64(
415 tb[IPSET_ATTR_BYTES]));
416 if (tb[IPSET_ATTR_PACKETS])
417 ext->packets = be64_to_cpu(nla_get_be64(
418 tb[IPSET_ATTR_PACKETS]));
419 }
420 if (tb[IPSET_ATTR_COMMENT]) {
421 if (!SET_WITH_COMMENT(set))
422 return -IPSET_ERR_COMMENT;
423 ext->comment = ip_set_comment_uget(tb[IPSET_ATTR_COMMENT]);
424 }
425 if (tb[IPSET_ATTR_SKBMARK]) {
426 if (!SET_WITH_SKBINFO(set))
427 return -IPSET_ERR_SKBINFO;
428 fullmark = be64_to_cpu(nla_get_be64(tb[IPSET_ATTR_SKBMARK]));
429 ext->skbmark = fullmark >> 32;
430 ext->skbmarkmask = fullmark & 0xffffffff;
431 }
432 if (tb[IPSET_ATTR_SKBPRIO]) {
433 if (!SET_WITH_SKBINFO(set))
434 return -IPSET_ERR_SKBINFO;
435 ext->skbprio = be32_to_cpu(nla_get_be32(
436 tb[IPSET_ATTR_SKBPRIO]));
437 }
438 if (tb[IPSET_ATTR_SKBQUEUE]) {
439 if (!SET_WITH_SKBINFO(set))
440 return -IPSET_ERR_SKBINFO;
441 ext->skbqueue = be16_to_cpu(nla_get_be16(
442 tb[IPSET_ATTR_SKBQUEUE]));
443 }
444 return 0;
445 }
446 EXPORT_SYMBOL_GPL(ip_set_get_extensions);
447
448 int
ip_set_put_extensions(struct sk_buff * skb,const struct ip_set * set,const void * e,bool active)449 ip_set_put_extensions(struct sk_buff *skb, const struct ip_set *set,
450 const void *e, bool active)
451 {
452 if (SET_WITH_TIMEOUT(set)) {
453 unsigned long *timeout = ext_timeout(e, set);
454
455 if (nla_put_net32(skb, IPSET_ATTR_TIMEOUT,
456 htonl(active ? ip_set_timeout_get(timeout)
457 : *timeout)))
458 return -EMSGSIZE;
459 }
460 if (SET_WITH_COUNTER(set) &&
461 ip_set_put_counter(skb, ext_counter(e, set)))
462 return -EMSGSIZE;
463 if (SET_WITH_COMMENT(set) &&
464 ip_set_put_comment(skb, ext_comment(e, set)))
465 return -EMSGSIZE;
466 if (SET_WITH_SKBINFO(set) &&
467 ip_set_put_skbinfo(skb, ext_skbinfo(e, set)))
468 return -EMSGSIZE;
469 return 0;
470 }
471 EXPORT_SYMBOL_GPL(ip_set_put_extensions);
472
473 /* Creating/destroying/renaming/swapping affect the existence and
474 * the properties of a set. All of these can be executed from userspace
475 * only and serialized by the nfnl mutex indirectly from nfnetlink.
476 *
477 * Sets are identified by their index in ip_set_list and the index
478 * is used by the external references (set/SET netfilter modules).
479 *
480 * The set behind an index may change by swapping only, from userspace.
481 */
482
483 static inline void
__ip_set_get(struct ip_set * set)484 __ip_set_get(struct ip_set *set)
485 {
486 write_lock_bh(&ip_set_ref_lock);
487 set->ref++;
488 write_unlock_bh(&ip_set_ref_lock);
489 }
490
491 static inline void
__ip_set_put(struct ip_set * set)492 __ip_set_put(struct ip_set *set)
493 {
494 write_lock_bh(&ip_set_ref_lock);
495 BUG_ON(set->ref == 0);
496 set->ref--;
497 write_unlock_bh(&ip_set_ref_lock);
498 }
499
500 /* Add, del and test set entries from kernel.
501 *
502 * The set behind the index must exist and must be referenced
503 * so it can't be destroyed (or changed) under our foot.
504 */
505
506 static inline struct ip_set *
ip_set_rcu_get(struct net * net,ip_set_id_t index)507 ip_set_rcu_get(struct net *net, ip_set_id_t index)
508 {
509 struct ip_set *set;
510 struct ip_set_net *inst = ip_set_pernet(net);
511
512 rcu_read_lock();
513 /* ip_set_list itself needs to be protected */
514 set = rcu_dereference(inst->ip_set_list)[index];
515 rcu_read_unlock();
516
517 return set;
518 }
519
520 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)521 ip_set_test(ip_set_id_t index, const struct sk_buff *skb,
522 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
523 {
524 struct ip_set *set = ip_set_rcu_get(par->net, index);
525 int ret = 0;
526
527 BUG_ON(!set);
528 pr_debug("set %s, index %u\n", set->name, index);
529
530 if (opt->dim < set->type->dimension ||
531 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
532 return 0;
533
534 rcu_read_lock_bh();
535 ret = set->variant->kadt(set, skb, par, IPSET_TEST, opt);
536 rcu_read_unlock_bh();
537
538 if (ret == -EAGAIN) {
539 /* Type requests element to be completed */
540 pr_debug("element must be completed, ADD is triggered\n");
541 spin_lock_bh(&set->lock);
542 set->variant->kadt(set, skb, par, IPSET_ADD, opt);
543 spin_unlock_bh(&set->lock);
544 ret = 1;
545 } else {
546 /* --return-nomatch: invert matched element */
547 if ((opt->cmdflags & IPSET_FLAG_RETURN_NOMATCH) &&
548 (set->type->features & IPSET_TYPE_NOMATCH) &&
549 (ret > 0 || ret == -ENOTEMPTY))
550 ret = -ret;
551 }
552
553 /* Convert error codes to nomatch */
554 return (ret < 0 ? 0 : ret);
555 }
556 EXPORT_SYMBOL_GPL(ip_set_test);
557
558 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)559 ip_set_add(ip_set_id_t index, const struct sk_buff *skb,
560 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
561 {
562 struct ip_set *set = ip_set_rcu_get(par->net, index);
563 int ret;
564
565 BUG_ON(!set);
566 pr_debug("set %s, index %u\n", set->name, index);
567
568 if (opt->dim < set->type->dimension ||
569 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
570 return -IPSET_ERR_TYPE_MISMATCH;
571
572 spin_lock_bh(&set->lock);
573 ret = set->variant->kadt(set, skb, par, IPSET_ADD, opt);
574 spin_unlock_bh(&set->lock);
575
576 return ret;
577 }
578 EXPORT_SYMBOL_GPL(ip_set_add);
579
580 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)581 ip_set_del(ip_set_id_t index, const struct sk_buff *skb,
582 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
583 {
584 struct ip_set *set = ip_set_rcu_get(par->net, index);
585 int ret = 0;
586
587 BUG_ON(!set);
588 pr_debug("set %s, index %u\n", set->name, index);
589
590 if (opt->dim < set->type->dimension ||
591 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
592 return -IPSET_ERR_TYPE_MISMATCH;
593
594 spin_lock_bh(&set->lock);
595 ret = set->variant->kadt(set, skb, par, IPSET_DEL, opt);
596 spin_unlock_bh(&set->lock);
597
598 return ret;
599 }
600 EXPORT_SYMBOL_GPL(ip_set_del);
601
602 /* Find set by name, reference it once. The reference makes sure the
603 * thing pointed to, does not go away under our feet.
604 *
605 */
606 ip_set_id_t
ip_set_get_byname(struct net * net,const char * name,struct ip_set ** set)607 ip_set_get_byname(struct net *net, const char *name, struct ip_set **set)
608 {
609 ip_set_id_t i, index = IPSET_INVALID_ID;
610 struct ip_set *s;
611 struct ip_set_net *inst = ip_set_pernet(net);
612
613 rcu_read_lock();
614 for (i = 0; i < inst->ip_set_max; i++) {
615 s = rcu_dereference(inst->ip_set_list)[i];
616 if (s && STRNCMP(s->name, name)) {
617 __ip_set_get(s);
618 index = i;
619 *set = s;
620 break;
621 }
622 }
623 rcu_read_unlock();
624
625 return index;
626 }
627 EXPORT_SYMBOL_GPL(ip_set_get_byname);
628
629 /* If the given set pointer points to a valid set, decrement
630 * reference count by 1. The caller shall not assume the index
631 * to be valid, after calling this function.
632 *
633 */
634
635 static inline void
__ip_set_put_byindex(struct ip_set_net * inst,ip_set_id_t index)636 __ip_set_put_byindex(struct ip_set_net *inst, ip_set_id_t index)
637 {
638 struct ip_set *set;
639
640 rcu_read_lock();
641 set = rcu_dereference(inst->ip_set_list)[index];
642 if (set)
643 __ip_set_put(set);
644 rcu_read_unlock();
645 }
646
647 void
ip_set_put_byindex(struct net * net,ip_set_id_t index)648 ip_set_put_byindex(struct net *net, ip_set_id_t index)
649 {
650 struct ip_set_net *inst = ip_set_pernet(net);
651
652 __ip_set_put_byindex(inst, index);
653 }
654 EXPORT_SYMBOL_GPL(ip_set_put_byindex);
655
656 /* Get the name of a set behind a set index.
657 * We assume the set is referenced, so it does exist and
658 * can't be destroyed. The set cannot be renamed due to
659 * the referencing either.
660 *
661 */
662 const char *
ip_set_name_byindex(struct net * net,ip_set_id_t index)663 ip_set_name_byindex(struct net *net, ip_set_id_t index)
664 {
665 const struct ip_set *set = ip_set_rcu_get(net, index);
666
667 BUG_ON(!set);
668 BUG_ON(set->ref == 0);
669
670 /* Referenced, so it's safe */
671 return set->name;
672 }
673 EXPORT_SYMBOL_GPL(ip_set_name_byindex);
674
675 /* Routines to call by external subsystems, which do not
676 * call nfnl_lock for us.
677 */
678
679 /* Find set by index, reference it once. The reference makes sure the
680 * thing pointed to, does not go away under our feet.
681 *
682 * The nfnl mutex is used in the function.
683 */
684 ip_set_id_t
ip_set_nfnl_get_byindex(struct net * net,ip_set_id_t index)685 ip_set_nfnl_get_byindex(struct net *net, ip_set_id_t index)
686 {
687 struct ip_set *set;
688 struct ip_set_net *inst = ip_set_pernet(net);
689
690 if (index >= inst->ip_set_max)
691 return IPSET_INVALID_ID;
692
693 nfnl_lock(NFNL_SUBSYS_IPSET);
694 set = ip_set(inst, index);
695 if (set)
696 __ip_set_get(set);
697 else
698 index = IPSET_INVALID_ID;
699 nfnl_unlock(NFNL_SUBSYS_IPSET);
700
701 return index;
702 }
703 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex);
704
705 /* If the given set pointer points to a valid set, decrement
706 * reference count by 1. The caller shall not assume the index
707 * to be valid, after calling this function.
708 *
709 * The nfnl mutex is used in the function.
710 */
711 void
ip_set_nfnl_put(struct net * net,ip_set_id_t index)712 ip_set_nfnl_put(struct net *net, ip_set_id_t index)
713 {
714 struct ip_set *set;
715 struct ip_set_net *inst = ip_set_pernet(net);
716
717 nfnl_lock(NFNL_SUBSYS_IPSET);
718 if (!inst->is_deleted) { /* already deleted from ip_set_net_exit() */
719 set = ip_set(inst, index);
720 if (set)
721 __ip_set_put(set);
722 }
723 nfnl_unlock(NFNL_SUBSYS_IPSET);
724 }
725 EXPORT_SYMBOL_GPL(ip_set_nfnl_put);
726
727 /* Communication protocol with userspace over netlink.
728 *
729 * The commands are serialized by the nfnl mutex.
730 */
731
732 static inline bool
protocol_failed(const struct nlattr * const tb[])733 protocol_failed(const struct nlattr * const tb[])
734 {
735 return !tb[IPSET_ATTR_PROTOCOL] ||
736 nla_get_u8(tb[IPSET_ATTR_PROTOCOL]) != IPSET_PROTOCOL;
737 }
738
739 static inline u32
flag_exist(const struct nlmsghdr * nlh)740 flag_exist(const struct nlmsghdr *nlh)
741 {
742 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST;
743 }
744
745 static struct nlmsghdr *
start_msg(struct sk_buff * skb,u32 portid,u32 seq,unsigned int flags,enum ipset_cmd cmd)746 start_msg(struct sk_buff *skb, u32 portid, u32 seq, unsigned int flags,
747 enum ipset_cmd cmd)
748 {
749 struct nlmsghdr *nlh;
750 struct nfgenmsg *nfmsg;
751
752 nlh = nlmsg_put(skb, portid, seq, cmd | (NFNL_SUBSYS_IPSET << 8),
753 sizeof(*nfmsg), flags);
754 if (!nlh)
755 return NULL;
756
757 nfmsg = nlmsg_data(nlh);
758 nfmsg->nfgen_family = NFPROTO_IPV4;
759 nfmsg->version = NFNETLINK_V0;
760 nfmsg->res_id = 0;
761
762 return nlh;
763 }
764
765 /* Create a set */
766
767 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = {
768 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
769 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
770 .len = IPSET_MAXNAMELEN - 1 },
771 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
772 .len = IPSET_MAXNAMELEN - 1},
773 [IPSET_ATTR_REVISION] = { .type = NLA_U8 },
774 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
775 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
776 };
777
778 static struct ip_set *
find_set_and_id(struct ip_set_net * inst,const char * name,ip_set_id_t * id)779 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id)
780 {
781 struct ip_set *set = NULL;
782 ip_set_id_t i;
783
784 *id = IPSET_INVALID_ID;
785 for (i = 0; i < inst->ip_set_max; i++) {
786 set = ip_set(inst, i);
787 if (set && STRNCMP(set->name, name)) {
788 *id = i;
789 break;
790 }
791 }
792 return (*id == IPSET_INVALID_ID ? NULL : set);
793 }
794
795 static inline struct ip_set *
find_set(struct ip_set_net * inst,const char * name)796 find_set(struct ip_set_net *inst, const char *name)
797 {
798 ip_set_id_t id;
799
800 return find_set_and_id(inst, name, &id);
801 }
802
803 static int
find_free_id(struct ip_set_net * inst,const char * name,ip_set_id_t * index,struct ip_set ** set)804 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index,
805 struct ip_set **set)
806 {
807 struct ip_set *s;
808 ip_set_id_t i;
809
810 *index = IPSET_INVALID_ID;
811 for (i = 0; i < inst->ip_set_max; i++) {
812 s = ip_set(inst, i);
813 if (!s) {
814 if (*index == IPSET_INVALID_ID)
815 *index = i;
816 } else if (STRNCMP(name, s->name)) {
817 /* Name clash */
818 *set = s;
819 return -EEXIST;
820 }
821 }
822 if (*index == IPSET_INVALID_ID)
823 /* No free slot remained */
824 return -IPSET_ERR_MAX_SETS;
825 return 0;
826 }
827
828 static int
ip_set_none(struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[])829 ip_set_none(struct sock *ctnl, struct sk_buff *skb,
830 const struct nlmsghdr *nlh,
831 const struct nlattr * const attr[])
832 {
833 return -EOPNOTSUPP;
834 }
835
836 static int
ip_set_create(struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[])837 ip_set_create(struct sock *ctnl, struct sk_buff *skb,
838 const struct nlmsghdr *nlh,
839 const struct nlattr * const attr[])
840 {
841 struct net *net = sock_net(ctnl);
842 struct ip_set_net *inst = ip_set_pernet(net);
843 struct ip_set *set, *clash = NULL;
844 ip_set_id_t index = IPSET_INVALID_ID;
845 struct nlattr *tb[IPSET_ATTR_CREATE_MAX + 1] = {};
846 const char *name, *typename;
847 u8 family, revision;
848 u32 flags = flag_exist(nlh);
849 int ret = 0;
850
851 if (unlikely(protocol_failed(attr) ||
852 !attr[IPSET_ATTR_SETNAME] ||
853 !attr[IPSET_ATTR_TYPENAME] ||
854 !attr[IPSET_ATTR_REVISION] ||
855 !attr[IPSET_ATTR_FAMILY] ||
856 (attr[IPSET_ATTR_DATA] &&
857 !flag_nested(attr[IPSET_ATTR_DATA]))))
858 return -IPSET_ERR_PROTOCOL;
859
860 name = nla_data(attr[IPSET_ATTR_SETNAME]);
861 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
862 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
863 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]);
864 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
865 name, typename, family_name(family), revision);
866
867 /* First, and without any locks, allocate and initialize
868 * a normal base set structure.
869 */
870 set = kzalloc(sizeof(*set), GFP_KERNEL);
871 if (!set)
872 return -ENOMEM;
873 spin_lock_init(&set->lock);
874 strlcpy(set->name, name, IPSET_MAXNAMELEN);
875 set->family = family;
876 set->revision = revision;
877
878 /* Next, check that we know the type, and take
879 * a reference on the type, to make sure it stays available
880 * while constructing our new set.
881 *
882 * After referencing the type, we try to create the type
883 * specific part of the set without holding any locks.
884 */
885 ret = find_set_type_get(typename, family, revision, &set->type);
886 if (ret)
887 goto out;
888
889 /* Without holding any locks, create private part. */
890 if (attr[IPSET_ATTR_DATA] &&
891 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA],
892 set->type->create_policy)) {
893 ret = -IPSET_ERR_PROTOCOL;
894 goto put_out;
895 }
896
897 ret = set->type->create(net, set, tb, flags);
898 if (ret != 0)
899 goto put_out;
900
901 /* BTW, ret==0 here. */
902
903 /* Here, we have a valid, constructed set and we are protected
904 * by the nfnl mutex. Find the first free index in ip_set_list
905 * and check clashing.
906 */
907 ret = find_free_id(inst, set->name, &index, &clash);
908 if (ret == -EEXIST) {
909 /* If this is the same set and requested, ignore error */
910 if ((flags & IPSET_FLAG_EXIST) &&
911 STRNCMP(set->type->name, clash->type->name) &&
912 set->type->family == clash->type->family &&
913 set->type->revision_min == clash->type->revision_min &&
914 set->type->revision_max == clash->type->revision_max &&
915 set->variant->same_set(set, clash))
916 ret = 0;
917 goto cleanup;
918 } else if (ret == -IPSET_ERR_MAX_SETS) {
919 struct ip_set **list, **tmp;
920 ip_set_id_t i = inst->ip_set_max + IP_SET_INC;
921
922 if (i < inst->ip_set_max || i == IPSET_INVALID_ID)
923 /* Wraparound */
924 goto cleanup;
925
926 list = kcalloc(i, sizeof(struct ip_set *), GFP_KERNEL);
927 if (!list)
928 goto cleanup;
929 /* nfnl mutex is held, both lists are valid */
930 tmp = ip_set_dereference(inst->ip_set_list);
931 memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max);
932 rcu_assign_pointer(inst->ip_set_list, list);
933 /* Make sure all current packets have passed through */
934 synchronize_net();
935 /* Use new list */
936 index = inst->ip_set_max;
937 inst->ip_set_max = i;
938 kfree(tmp);
939 ret = 0;
940 } else if (ret) {
941 goto cleanup;
942 }
943
944 /* Finally! Add our shiny new set to the list, and be done. */
945 pr_debug("create: '%s' created with index %u!\n", set->name, index);
946 ip_set(inst, index) = set;
947
948 return ret;
949
950 cleanup:
951 set->variant->destroy(set);
952 put_out:
953 module_put(set->type->me);
954 out:
955 kfree(set);
956 return ret;
957 }
958
959 /* Destroy sets */
960
961 static const struct nla_policy
962 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = {
963 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
964 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
965 .len = IPSET_MAXNAMELEN - 1 },
966 };
967
968 static void
ip_set_destroy_set(struct ip_set * set)969 ip_set_destroy_set(struct ip_set *set)
970 {
971 pr_debug("set: %s\n", set->name);
972
973 /* Must call it without holding any lock */
974 set->variant->destroy(set);
975 module_put(set->type->me);
976 kfree(set);
977 }
978
979 static int
ip_set_destroy(struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[])980 ip_set_destroy(struct sock *ctnl, struct sk_buff *skb,
981 const struct nlmsghdr *nlh,
982 const struct nlattr * const attr[])
983 {
984 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
985 struct ip_set *s;
986 ip_set_id_t i;
987 int ret = 0;
988
989 if (unlikely(protocol_failed(attr)))
990 return -IPSET_ERR_PROTOCOL;
991
992 /* Commands are serialized and references are
993 * protected by the ip_set_ref_lock.
994 * External systems (i.e. xt_set) must call
995 * ip_set_put|get_nfnl_* functions, that way we
996 * can safely check references here.
997 *
998 * list:set timer can only decrement the reference
999 * counter, so if it's already zero, we can proceed
1000 * without holding the lock.
1001 */
1002 read_lock_bh(&ip_set_ref_lock);
1003 if (!attr[IPSET_ATTR_SETNAME]) {
1004 for (i = 0; i < inst->ip_set_max; i++) {
1005 s = ip_set(inst, i);
1006 if (s && s->ref) {
1007 ret = -IPSET_ERR_BUSY;
1008 goto out;
1009 }
1010 }
1011 inst->is_destroyed = true;
1012 read_unlock_bh(&ip_set_ref_lock);
1013 for (i = 0; i < inst->ip_set_max; i++) {
1014 s = ip_set(inst, i);
1015 if (s) {
1016 ip_set(inst, i) = NULL;
1017 ip_set_destroy_set(s);
1018 }
1019 }
1020 /* Modified by ip_set_destroy() only, which is serialized */
1021 inst->is_destroyed = false;
1022 } else {
1023 s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1024 &i);
1025 if (!s) {
1026 ret = -ENOENT;
1027 goto out;
1028 } else if (s->ref) {
1029 ret = -IPSET_ERR_BUSY;
1030 goto out;
1031 }
1032 ip_set(inst, i) = NULL;
1033 read_unlock_bh(&ip_set_ref_lock);
1034
1035 ip_set_destroy_set(s);
1036 }
1037 return 0;
1038 out:
1039 read_unlock_bh(&ip_set_ref_lock);
1040 return ret;
1041 }
1042
1043 /* Flush sets */
1044
1045 static void
ip_set_flush_set(struct ip_set * set)1046 ip_set_flush_set(struct ip_set *set)
1047 {
1048 pr_debug("set: %s\n", set->name);
1049
1050 spin_lock_bh(&set->lock);
1051 set->variant->flush(set);
1052 spin_unlock_bh(&set->lock);
1053 }
1054
1055 static int
ip_set_flush(struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[])1056 ip_set_flush(struct sock *ctnl, struct sk_buff *skb,
1057 const struct nlmsghdr *nlh,
1058 const struct nlattr * const attr[])
1059 {
1060 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1061 struct ip_set *s;
1062 ip_set_id_t i;
1063
1064 if (unlikely(protocol_failed(attr)))
1065 return -IPSET_ERR_PROTOCOL;
1066
1067 if (!attr[IPSET_ATTR_SETNAME]) {
1068 for (i = 0; i < inst->ip_set_max; i++) {
1069 s = ip_set(inst, i);
1070 if (s)
1071 ip_set_flush_set(s);
1072 }
1073 } else {
1074 s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1075 if (!s)
1076 return -ENOENT;
1077
1078 ip_set_flush_set(s);
1079 }
1080
1081 return 0;
1082 }
1083
1084 /* Rename a set */
1085
1086 static const struct nla_policy
1087 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = {
1088 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1089 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1090 .len = IPSET_MAXNAMELEN - 1 },
1091 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING,
1092 .len = IPSET_MAXNAMELEN - 1 },
1093 };
1094
1095 static int
ip_set_rename(struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[])1096 ip_set_rename(struct sock *ctnl, struct sk_buff *skb,
1097 const struct nlmsghdr *nlh,
1098 const struct nlattr * const attr[])
1099 {
1100 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1101 struct ip_set *set, *s;
1102 const char *name2;
1103 ip_set_id_t i;
1104 int ret = 0;
1105
1106 if (unlikely(protocol_failed(attr) ||
1107 !attr[IPSET_ATTR_SETNAME] ||
1108 !attr[IPSET_ATTR_SETNAME2]))
1109 return -IPSET_ERR_PROTOCOL;
1110
1111 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1112 if (!set)
1113 return -ENOENT;
1114
1115 read_lock_bh(&ip_set_ref_lock);
1116 if (set->ref != 0) {
1117 ret = -IPSET_ERR_REFERENCED;
1118 goto out;
1119 }
1120
1121 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]);
1122 for (i = 0; i < inst->ip_set_max; i++) {
1123 s = ip_set(inst, i);
1124 if (s && STRNCMP(s->name, name2)) {
1125 ret = -IPSET_ERR_EXIST_SETNAME2;
1126 goto out;
1127 }
1128 }
1129 strncpy(set->name, name2, IPSET_MAXNAMELEN);
1130
1131 out:
1132 read_unlock_bh(&ip_set_ref_lock);
1133 return ret;
1134 }
1135
1136 /* Swap two sets so that name/index points to the other.
1137 * References and set names are also swapped.
1138 *
1139 * The commands are serialized by the nfnl mutex and references are
1140 * protected by the ip_set_ref_lock. The kernel interfaces
1141 * do not hold the mutex but the pointer settings are atomic
1142 * so the ip_set_list always contains valid pointers to the sets.
1143 */
1144
1145 static int
ip_set_swap(struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[])1146 ip_set_swap(struct sock *ctnl, struct sk_buff *skb,
1147 const struct nlmsghdr *nlh,
1148 const struct nlattr * const attr[])
1149 {
1150 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1151 struct ip_set *from, *to;
1152 ip_set_id_t from_id, to_id;
1153 char from_name[IPSET_MAXNAMELEN];
1154
1155 if (unlikely(protocol_failed(attr) ||
1156 !attr[IPSET_ATTR_SETNAME] ||
1157 !attr[IPSET_ATTR_SETNAME2]))
1158 return -IPSET_ERR_PROTOCOL;
1159
1160 from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1161 &from_id);
1162 if (!from)
1163 return -ENOENT;
1164
1165 to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]),
1166 &to_id);
1167 if (!to)
1168 return -IPSET_ERR_EXIST_SETNAME2;
1169
1170 /* Features must not change.
1171 * Not an artifical restriction anymore, as we must prevent
1172 * possible loops created by swapping in setlist type of sets.
1173 */
1174 if (!(from->type->features == to->type->features &&
1175 from->family == to->family))
1176 return -IPSET_ERR_TYPE_MISMATCH;
1177
1178 strncpy(from_name, from->name, IPSET_MAXNAMELEN);
1179 strncpy(from->name, to->name, IPSET_MAXNAMELEN);
1180 strncpy(to->name, from_name, IPSET_MAXNAMELEN);
1181
1182 write_lock_bh(&ip_set_ref_lock);
1183 swap(from->ref, to->ref);
1184 ip_set(inst, from_id) = to;
1185 ip_set(inst, to_id) = from;
1186 write_unlock_bh(&ip_set_ref_lock);
1187
1188 return 0;
1189 }
1190
1191 /* List/save set data */
1192
1193 #define DUMP_INIT 0
1194 #define DUMP_ALL 1
1195 #define DUMP_ONE 2
1196 #define DUMP_LAST 3
1197
1198 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1199 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1200
1201 static int
ip_set_dump_done(struct netlink_callback * cb)1202 ip_set_dump_done(struct netlink_callback *cb)
1203 {
1204 if (cb->args[IPSET_CB_ARG0]) {
1205 struct ip_set_net *inst =
1206 (struct ip_set_net *)cb->args[IPSET_CB_NET];
1207 ip_set_id_t index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1208 struct ip_set *set = ip_set(inst, index);
1209
1210 if (set->variant->uref)
1211 set->variant->uref(set, cb, false);
1212 pr_debug("release set %s\n", set->name);
1213 __ip_set_put_byindex(inst, index);
1214 }
1215 return 0;
1216 }
1217
1218 static inline void
dump_attrs(struct nlmsghdr * nlh)1219 dump_attrs(struct nlmsghdr *nlh)
1220 {
1221 const struct nlattr *attr;
1222 int rem;
1223
1224 pr_debug("dump nlmsg\n");
1225 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) {
1226 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len);
1227 }
1228 }
1229
1230 static int
dump_init(struct netlink_callback * cb,struct ip_set_net * inst)1231 dump_init(struct netlink_callback *cb, struct ip_set_net *inst)
1232 {
1233 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb);
1234 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1235 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1236 struct nlattr *attr = (void *)nlh + min_len;
1237 u32 dump_type;
1238 ip_set_id_t index;
1239
1240 /* Second pass, so parser can't fail */
1241 nla_parse(cda, IPSET_ATTR_CMD_MAX,
1242 attr, nlh->nlmsg_len - min_len, ip_set_setname_policy);
1243
1244 if (cda[IPSET_ATTR_SETNAME]) {
1245 struct ip_set *set;
1246
1247 set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]),
1248 &index);
1249 if (!set)
1250 return -ENOENT;
1251
1252 dump_type = DUMP_ONE;
1253 cb->args[IPSET_CB_INDEX] = index;
1254 } else {
1255 dump_type = DUMP_ALL;
1256 }
1257
1258 if (cda[IPSET_ATTR_FLAGS]) {
1259 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]);
1260
1261 dump_type |= (f << 16);
1262 }
1263 cb->args[IPSET_CB_NET] = (unsigned long)inst;
1264 cb->args[IPSET_CB_DUMP] = dump_type;
1265
1266 return 0;
1267 }
1268
1269 static int
ip_set_dump_start(struct sk_buff * skb,struct netlink_callback * cb)1270 ip_set_dump_start(struct sk_buff *skb, struct netlink_callback *cb)
1271 {
1272 ip_set_id_t index = IPSET_INVALID_ID, max;
1273 struct ip_set *set = NULL;
1274 struct nlmsghdr *nlh = NULL;
1275 unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0;
1276 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1277 u32 dump_type, dump_flags;
1278 bool is_destroyed;
1279 int ret = 0;
1280
1281 if (!cb->args[IPSET_CB_DUMP]) {
1282 ret = dump_init(cb, inst);
1283 if (ret < 0) {
1284 nlh = nlmsg_hdr(cb->skb);
1285 /* We have to create and send the error message
1286 * manually :-(
1287 */
1288 if (nlh->nlmsg_flags & NLM_F_ACK)
1289 netlink_ack(cb->skb, nlh, ret);
1290 return ret;
1291 }
1292 }
1293
1294 if (cb->args[IPSET_CB_INDEX] >= inst->ip_set_max)
1295 goto out;
1296
1297 dump_type = DUMP_TYPE(cb->args[IPSET_CB_DUMP]);
1298 dump_flags = DUMP_FLAGS(cb->args[IPSET_CB_DUMP]);
1299 max = dump_type == DUMP_ONE ? cb->args[IPSET_CB_INDEX] + 1
1300 : inst->ip_set_max;
1301 dump_last:
1302 pr_debug("dump type, flag: %u %u index: %ld\n",
1303 dump_type, dump_flags, cb->args[IPSET_CB_INDEX]);
1304 for (; cb->args[IPSET_CB_INDEX] < max; cb->args[IPSET_CB_INDEX]++) {
1305 index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1306 write_lock_bh(&ip_set_ref_lock);
1307 set = ip_set(inst, index);
1308 is_destroyed = inst->is_destroyed;
1309 if (!set || is_destroyed) {
1310 write_unlock_bh(&ip_set_ref_lock);
1311 if (dump_type == DUMP_ONE) {
1312 ret = -ENOENT;
1313 goto out;
1314 }
1315 if (is_destroyed) {
1316 /* All sets are just being destroyed */
1317 ret = 0;
1318 goto out;
1319 }
1320 continue;
1321 }
1322 /* When dumping all sets, we must dump "sorted"
1323 * so that lists (unions of sets) are dumped last.
1324 */
1325 if (dump_type != DUMP_ONE &&
1326 ((dump_type == DUMP_ALL) ==
1327 !!(set->type->features & IPSET_DUMP_LAST))) {
1328 write_unlock_bh(&ip_set_ref_lock);
1329 continue;
1330 }
1331 pr_debug("List set: %s\n", set->name);
1332 if (!cb->args[IPSET_CB_ARG0]) {
1333 /* Start listing: make sure set won't be destroyed */
1334 pr_debug("reference set\n");
1335 set->ref++;
1336 }
1337 write_unlock_bh(&ip_set_ref_lock);
1338 nlh = start_msg(skb, NETLINK_CB(cb->skb).portid,
1339 cb->nlh->nlmsg_seq, flags,
1340 IPSET_CMD_LIST);
1341 if (!nlh) {
1342 ret = -EMSGSIZE;
1343 goto release_refcount;
1344 }
1345 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1346 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name))
1347 goto nla_put_failure;
1348 if (dump_flags & IPSET_FLAG_LIST_SETNAME)
1349 goto next_set;
1350 switch (cb->args[IPSET_CB_ARG0]) {
1351 case 0:
1352 /* Core header data */
1353 if (nla_put_string(skb, IPSET_ATTR_TYPENAME,
1354 set->type->name) ||
1355 nla_put_u8(skb, IPSET_ATTR_FAMILY,
1356 set->family) ||
1357 nla_put_u8(skb, IPSET_ATTR_REVISION,
1358 set->revision))
1359 goto nla_put_failure;
1360 ret = set->variant->head(set, skb);
1361 if (ret < 0)
1362 goto release_refcount;
1363 if (dump_flags & IPSET_FLAG_LIST_HEADER)
1364 goto next_set;
1365 if (set->variant->uref)
1366 set->variant->uref(set, cb, true);
1367 /* Fall through and add elements */
1368 default:
1369 rcu_read_lock_bh();
1370 ret = set->variant->list(set, skb, cb);
1371 rcu_read_unlock_bh();
1372 if (!cb->args[IPSET_CB_ARG0])
1373 /* Set is done, proceed with next one */
1374 goto next_set;
1375 goto release_refcount;
1376 }
1377 }
1378 /* If we dump all sets, continue with dumping last ones */
1379 if (dump_type == DUMP_ALL) {
1380 dump_type = DUMP_LAST;
1381 cb->args[IPSET_CB_DUMP] = dump_type | (dump_flags << 16);
1382 cb->args[IPSET_CB_INDEX] = 0;
1383 if (set && set->variant->uref)
1384 set->variant->uref(set, cb, false);
1385 goto dump_last;
1386 }
1387 goto out;
1388
1389 nla_put_failure:
1390 ret = -EFAULT;
1391 next_set:
1392 if (dump_type == DUMP_ONE)
1393 cb->args[IPSET_CB_INDEX] = IPSET_INVALID_ID;
1394 else
1395 cb->args[IPSET_CB_INDEX]++;
1396 release_refcount:
1397 /* If there was an error or set is done, release set */
1398 if (ret || !cb->args[IPSET_CB_ARG0]) {
1399 set = ip_set(inst, index);
1400 if (set->variant->uref)
1401 set->variant->uref(set, cb, false);
1402 pr_debug("release set %s\n", set->name);
1403 __ip_set_put_byindex(inst, index);
1404 cb->args[IPSET_CB_ARG0] = 0;
1405 }
1406 out:
1407 if (nlh) {
1408 nlmsg_end(skb, nlh);
1409 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len);
1410 dump_attrs(nlh);
1411 }
1412
1413 return ret < 0 ? ret : skb->len;
1414 }
1415
1416 static int
ip_set_dump(struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[])1417 ip_set_dump(struct sock *ctnl, struct sk_buff *skb,
1418 const struct nlmsghdr *nlh,
1419 const struct nlattr * const attr[])
1420 {
1421 if (unlikely(protocol_failed(attr)))
1422 return -IPSET_ERR_PROTOCOL;
1423
1424 {
1425 struct netlink_dump_control c = {
1426 .dump = ip_set_dump_start,
1427 .done = ip_set_dump_done,
1428 };
1429 return netlink_dump_start(ctnl, skb, nlh, &c);
1430 }
1431 }
1432
1433 /* Add, del and test */
1434
1435 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = {
1436 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1437 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1438 .len = IPSET_MAXNAMELEN - 1 },
1439 [IPSET_ATTR_LINENO] = { .type = NLA_U32 },
1440 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
1441 [IPSET_ATTR_ADT] = { .type = NLA_NESTED },
1442 };
1443
1444 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)1445 call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set,
1446 struct nlattr *tb[], enum ipset_adt adt,
1447 u32 flags, bool use_lineno)
1448 {
1449 int ret;
1450 u32 lineno = 0;
1451 bool eexist = flags & IPSET_FLAG_EXIST, retried = false;
1452
1453 do {
1454 spin_lock_bh(&set->lock);
1455 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried);
1456 spin_unlock_bh(&set->lock);
1457 retried = true;
1458 } while (ret == -EAGAIN &&
1459 set->variant->resize &&
1460 (ret = set->variant->resize(set, retried)) == 0);
1461
1462 if (!ret || (ret == -IPSET_ERR_EXIST && eexist))
1463 return 0;
1464 if (lineno && use_lineno) {
1465 /* Error in restore/batch mode: send back lineno */
1466 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb);
1467 struct sk_buff *skb2;
1468 struct nlmsgerr *errmsg;
1469 size_t payload = min(SIZE_MAX,
1470 sizeof(*errmsg) + nlmsg_len(nlh));
1471 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1472 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1473 struct nlattr *cmdattr;
1474 u32 *errline;
1475
1476 skb2 = nlmsg_new(payload, GFP_KERNEL);
1477 if (!skb2)
1478 return -ENOMEM;
1479 rep = __nlmsg_put(skb2, NETLINK_CB(skb).portid,
1480 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0);
1481 errmsg = nlmsg_data(rep);
1482 errmsg->error = ret;
1483 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len);
1484 cmdattr = (void *)&errmsg->msg + min_len;
1485
1486 nla_parse(cda, IPSET_ATTR_CMD_MAX,
1487 cmdattr, nlh->nlmsg_len - min_len,
1488 ip_set_adt_policy);
1489
1490 errline = nla_data(cda[IPSET_ATTR_LINENO]);
1491
1492 *errline = lineno;
1493
1494 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid,
1495 MSG_DONTWAIT);
1496 /* Signal netlink not to send its ACK/errmsg. */
1497 return -EINTR;
1498 }
1499
1500 return ret;
1501 }
1502
1503 static int
ip_set_uadd(struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[])1504 ip_set_uadd(struct sock *ctnl, struct sk_buff *skb,
1505 const struct nlmsghdr *nlh,
1506 const struct nlattr * const attr[])
1507 {
1508 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1509 struct ip_set *set;
1510 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1511 const struct nlattr *nla;
1512 u32 flags = flag_exist(nlh);
1513 bool use_lineno;
1514 int ret = 0;
1515
1516 if (unlikely(protocol_failed(attr) ||
1517 !attr[IPSET_ATTR_SETNAME] ||
1518 !((attr[IPSET_ATTR_DATA] != NULL) ^
1519 (attr[IPSET_ATTR_ADT] != NULL)) ||
1520 (attr[IPSET_ATTR_DATA] &&
1521 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1522 (attr[IPSET_ATTR_ADT] &&
1523 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1524 !attr[IPSET_ATTR_LINENO]))))
1525 return -IPSET_ERR_PROTOCOL;
1526
1527 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1528 if (!set)
1529 return -ENOENT;
1530
1531 use_lineno = !!attr[IPSET_ATTR_LINENO];
1532 if (attr[IPSET_ATTR_DATA]) {
1533 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1534 attr[IPSET_ATTR_DATA],
1535 set->type->adt_policy))
1536 return -IPSET_ERR_PROTOCOL;
1537 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD, flags,
1538 use_lineno);
1539 } else {
1540 int nla_rem;
1541
1542 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1543 memset(tb, 0, sizeof(tb));
1544 if (nla_type(nla) != IPSET_ATTR_DATA ||
1545 !flag_nested(nla) ||
1546 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1547 set->type->adt_policy))
1548 return -IPSET_ERR_PROTOCOL;
1549 ret = call_ad(ctnl, skb, set, tb, IPSET_ADD,
1550 flags, use_lineno);
1551 if (ret < 0)
1552 return ret;
1553 }
1554 }
1555 return ret;
1556 }
1557
1558 static int
ip_set_udel(struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[])1559 ip_set_udel(struct sock *ctnl, struct sk_buff *skb,
1560 const struct nlmsghdr *nlh,
1561 const struct nlattr * const attr[])
1562 {
1563 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1564 struct ip_set *set;
1565 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1566 const struct nlattr *nla;
1567 u32 flags = flag_exist(nlh);
1568 bool use_lineno;
1569 int ret = 0;
1570
1571 if (unlikely(protocol_failed(attr) ||
1572 !attr[IPSET_ATTR_SETNAME] ||
1573 !((attr[IPSET_ATTR_DATA] != NULL) ^
1574 (attr[IPSET_ATTR_ADT] != NULL)) ||
1575 (attr[IPSET_ATTR_DATA] &&
1576 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1577 (attr[IPSET_ATTR_ADT] &&
1578 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1579 !attr[IPSET_ATTR_LINENO]))))
1580 return -IPSET_ERR_PROTOCOL;
1581
1582 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1583 if (!set)
1584 return -ENOENT;
1585
1586 use_lineno = !!attr[IPSET_ATTR_LINENO];
1587 if (attr[IPSET_ATTR_DATA]) {
1588 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1589 attr[IPSET_ATTR_DATA],
1590 set->type->adt_policy))
1591 return -IPSET_ERR_PROTOCOL;
1592 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL, flags,
1593 use_lineno);
1594 } else {
1595 int nla_rem;
1596
1597 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1598 memset(tb, 0, sizeof(*tb));
1599 if (nla_type(nla) != IPSET_ATTR_DATA ||
1600 !flag_nested(nla) ||
1601 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1602 set->type->adt_policy))
1603 return -IPSET_ERR_PROTOCOL;
1604 ret = call_ad(ctnl, skb, set, tb, IPSET_DEL,
1605 flags, use_lineno);
1606 if (ret < 0)
1607 return ret;
1608 }
1609 }
1610 return ret;
1611 }
1612
1613 static int
ip_set_utest(struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[])1614 ip_set_utest(struct sock *ctnl, struct sk_buff *skb,
1615 const struct nlmsghdr *nlh,
1616 const struct nlattr * const attr[])
1617 {
1618 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1619 struct ip_set *set;
1620 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1621 int ret = 0;
1622
1623 if (unlikely(protocol_failed(attr) ||
1624 !attr[IPSET_ATTR_SETNAME] ||
1625 !attr[IPSET_ATTR_DATA] ||
1626 !flag_nested(attr[IPSET_ATTR_DATA])))
1627 return -IPSET_ERR_PROTOCOL;
1628
1629 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1630 if (!set)
1631 return -ENOENT;
1632
1633 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA],
1634 set->type->adt_policy))
1635 return -IPSET_ERR_PROTOCOL;
1636
1637 rcu_read_lock_bh();
1638 ret = set->variant->uadt(set, tb, IPSET_TEST, NULL, 0, 0);
1639 rcu_read_unlock_bh();
1640 /* Userspace can't trigger element to be re-added */
1641 if (ret == -EAGAIN)
1642 ret = 1;
1643
1644 return ret > 0 ? 0 : -IPSET_ERR_EXIST;
1645 }
1646
1647 /* Get headed data of a set */
1648
1649 static int
ip_set_header(struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[])1650 ip_set_header(struct sock *ctnl, struct sk_buff *skb,
1651 const struct nlmsghdr *nlh,
1652 const struct nlattr * const attr[])
1653 {
1654 struct ip_set_net *inst = ip_set_pernet(sock_net(ctnl));
1655 const struct ip_set *set;
1656 struct sk_buff *skb2;
1657 struct nlmsghdr *nlh2;
1658 int ret = 0;
1659
1660 if (unlikely(protocol_failed(attr) ||
1661 !attr[IPSET_ATTR_SETNAME]))
1662 return -IPSET_ERR_PROTOCOL;
1663
1664 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1665 if (!set)
1666 return -ENOENT;
1667
1668 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1669 if (!skb2)
1670 return -ENOMEM;
1671
1672 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1673 IPSET_CMD_HEADER);
1674 if (!nlh2)
1675 goto nlmsg_failure;
1676 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1677 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) ||
1678 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) ||
1679 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
1680 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision))
1681 goto nla_put_failure;
1682 nlmsg_end(skb2, nlh2);
1683
1684 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1685 if (ret < 0)
1686 return ret;
1687
1688 return 0;
1689
1690 nla_put_failure:
1691 nlmsg_cancel(skb2, nlh2);
1692 nlmsg_failure:
1693 kfree_skb(skb2);
1694 return -EMSGSIZE;
1695 }
1696
1697 /* Get type data */
1698
1699 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = {
1700 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1701 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
1702 .len = IPSET_MAXNAMELEN - 1 },
1703 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
1704 };
1705
1706 static int
ip_set_type(struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[])1707 ip_set_type(struct sock *ctnl, struct sk_buff *skb,
1708 const struct nlmsghdr *nlh,
1709 const struct nlattr * const attr[])
1710 {
1711 struct sk_buff *skb2;
1712 struct nlmsghdr *nlh2;
1713 u8 family, min, max;
1714 const char *typename;
1715 int ret = 0;
1716
1717 if (unlikely(protocol_failed(attr) ||
1718 !attr[IPSET_ATTR_TYPENAME] ||
1719 !attr[IPSET_ATTR_FAMILY]))
1720 return -IPSET_ERR_PROTOCOL;
1721
1722 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1723 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1724 ret = find_set_type_minmax(typename, family, &min, &max);
1725 if (ret)
1726 return ret;
1727
1728 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1729 if (!skb2)
1730 return -ENOMEM;
1731
1732 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1733 IPSET_CMD_TYPE);
1734 if (!nlh2)
1735 goto nlmsg_failure;
1736 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL) ||
1737 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) ||
1738 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) ||
1739 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) ||
1740 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min))
1741 goto nla_put_failure;
1742 nlmsg_end(skb2, nlh2);
1743
1744 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len);
1745 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1746 if (ret < 0)
1747 return ret;
1748
1749 return 0;
1750
1751 nla_put_failure:
1752 nlmsg_cancel(skb2, nlh2);
1753 nlmsg_failure:
1754 kfree_skb(skb2);
1755 return -EMSGSIZE;
1756 }
1757
1758 /* Get protocol version */
1759
1760 static const struct nla_policy
1761 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = {
1762 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1763 };
1764
1765 static int
ip_set_protocol(struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const attr[])1766 ip_set_protocol(struct sock *ctnl, struct sk_buff *skb,
1767 const struct nlmsghdr *nlh,
1768 const struct nlattr * const attr[])
1769 {
1770 struct sk_buff *skb2;
1771 struct nlmsghdr *nlh2;
1772 int ret = 0;
1773
1774 if (unlikely(!attr[IPSET_ATTR_PROTOCOL]))
1775 return -IPSET_ERR_PROTOCOL;
1776
1777 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1778 if (!skb2)
1779 return -ENOMEM;
1780
1781 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1782 IPSET_CMD_PROTOCOL);
1783 if (!nlh2)
1784 goto nlmsg_failure;
1785 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL))
1786 goto nla_put_failure;
1787 nlmsg_end(skb2, nlh2);
1788
1789 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1790 if (ret < 0)
1791 return ret;
1792
1793 return 0;
1794
1795 nla_put_failure:
1796 nlmsg_cancel(skb2, nlh2);
1797 nlmsg_failure:
1798 kfree_skb(skb2);
1799 return -EMSGSIZE;
1800 }
1801
1802 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = {
1803 [IPSET_CMD_NONE] = {
1804 .call = ip_set_none,
1805 .attr_count = IPSET_ATTR_CMD_MAX,
1806 },
1807 [IPSET_CMD_CREATE] = {
1808 .call = ip_set_create,
1809 .attr_count = IPSET_ATTR_CMD_MAX,
1810 .policy = ip_set_create_policy,
1811 },
1812 [IPSET_CMD_DESTROY] = {
1813 .call = ip_set_destroy,
1814 .attr_count = IPSET_ATTR_CMD_MAX,
1815 .policy = ip_set_setname_policy,
1816 },
1817 [IPSET_CMD_FLUSH] = {
1818 .call = ip_set_flush,
1819 .attr_count = IPSET_ATTR_CMD_MAX,
1820 .policy = ip_set_setname_policy,
1821 },
1822 [IPSET_CMD_RENAME] = {
1823 .call = ip_set_rename,
1824 .attr_count = IPSET_ATTR_CMD_MAX,
1825 .policy = ip_set_setname2_policy,
1826 },
1827 [IPSET_CMD_SWAP] = {
1828 .call = ip_set_swap,
1829 .attr_count = IPSET_ATTR_CMD_MAX,
1830 .policy = ip_set_setname2_policy,
1831 },
1832 [IPSET_CMD_LIST] = {
1833 .call = ip_set_dump,
1834 .attr_count = IPSET_ATTR_CMD_MAX,
1835 .policy = ip_set_setname_policy,
1836 },
1837 [IPSET_CMD_SAVE] = {
1838 .call = ip_set_dump,
1839 .attr_count = IPSET_ATTR_CMD_MAX,
1840 .policy = ip_set_setname_policy,
1841 },
1842 [IPSET_CMD_ADD] = {
1843 .call = ip_set_uadd,
1844 .attr_count = IPSET_ATTR_CMD_MAX,
1845 .policy = ip_set_adt_policy,
1846 },
1847 [IPSET_CMD_DEL] = {
1848 .call = ip_set_udel,
1849 .attr_count = IPSET_ATTR_CMD_MAX,
1850 .policy = ip_set_adt_policy,
1851 },
1852 [IPSET_CMD_TEST] = {
1853 .call = ip_set_utest,
1854 .attr_count = IPSET_ATTR_CMD_MAX,
1855 .policy = ip_set_adt_policy,
1856 },
1857 [IPSET_CMD_HEADER] = {
1858 .call = ip_set_header,
1859 .attr_count = IPSET_ATTR_CMD_MAX,
1860 .policy = ip_set_setname_policy,
1861 },
1862 [IPSET_CMD_TYPE] = {
1863 .call = ip_set_type,
1864 .attr_count = IPSET_ATTR_CMD_MAX,
1865 .policy = ip_set_type_policy,
1866 },
1867 [IPSET_CMD_PROTOCOL] = {
1868 .call = ip_set_protocol,
1869 .attr_count = IPSET_ATTR_CMD_MAX,
1870 .policy = ip_set_protocol_policy,
1871 },
1872 };
1873
1874 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = {
1875 .name = "ip_set",
1876 .subsys_id = NFNL_SUBSYS_IPSET,
1877 .cb_count = IPSET_MSG_MAX,
1878 .cb = ip_set_netlink_subsys_cb,
1879 };
1880
1881 /* Interface to iptables/ip6tables */
1882
1883 static int
ip_set_sockfn_get(struct sock * sk,int optval,void __user * user,int * len)1884 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len)
1885 {
1886 unsigned int *op;
1887 void *data;
1888 int copylen = *len, ret = 0;
1889 struct net *net = sock_net(sk);
1890 struct ip_set_net *inst = ip_set_pernet(net);
1891
1892 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1893 return -EPERM;
1894 if (optval != SO_IP_SET)
1895 return -EBADF;
1896 if (*len < sizeof(unsigned int))
1897 return -EINVAL;
1898
1899 data = vmalloc(*len);
1900 if (!data)
1901 return -ENOMEM;
1902 if (copy_from_user(data, user, *len) != 0) {
1903 ret = -EFAULT;
1904 goto done;
1905 }
1906 op = (unsigned int *)data;
1907
1908 if (*op < IP_SET_OP_VERSION) {
1909 /* Check the version at the beginning of operations */
1910 struct ip_set_req_version *req_version = data;
1911
1912 if (*len < sizeof(struct ip_set_req_version)) {
1913 ret = -EINVAL;
1914 goto done;
1915 }
1916
1917 if (req_version->version != IPSET_PROTOCOL) {
1918 ret = -EPROTO;
1919 goto done;
1920 }
1921 }
1922
1923 switch (*op) {
1924 case IP_SET_OP_VERSION: {
1925 struct ip_set_req_version *req_version = data;
1926
1927 if (*len != sizeof(struct ip_set_req_version)) {
1928 ret = -EINVAL;
1929 goto done;
1930 }
1931
1932 req_version->version = IPSET_PROTOCOL;
1933 ret = copy_to_user(user, req_version,
1934 sizeof(struct ip_set_req_version));
1935 goto done;
1936 }
1937 case IP_SET_OP_GET_BYNAME: {
1938 struct ip_set_req_get_set *req_get = data;
1939 ip_set_id_t id;
1940
1941 if (*len != sizeof(struct ip_set_req_get_set)) {
1942 ret = -EINVAL;
1943 goto done;
1944 }
1945 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
1946 nfnl_lock(NFNL_SUBSYS_IPSET);
1947 find_set_and_id(inst, req_get->set.name, &id);
1948 req_get->set.index = id;
1949 nfnl_unlock(NFNL_SUBSYS_IPSET);
1950 goto copy;
1951 }
1952 case IP_SET_OP_GET_FNAME: {
1953 struct ip_set_req_get_set_family *req_get = data;
1954 ip_set_id_t id;
1955
1956 if (*len != sizeof(struct ip_set_req_get_set_family)) {
1957 ret = -EINVAL;
1958 goto done;
1959 }
1960 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
1961 nfnl_lock(NFNL_SUBSYS_IPSET);
1962 find_set_and_id(inst, req_get->set.name, &id);
1963 req_get->set.index = id;
1964 if (id != IPSET_INVALID_ID)
1965 req_get->family = ip_set(inst, id)->family;
1966 nfnl_unlock(NFNL_SUBSYS_IPSET);
1967 goto copy;
1968 }
1969 case IP_SET_OP_GET_BYINDEX: {
1970 struct ip_set_req_get_set *req_get = data;
1971 struct ip_set *set;
1972
1973 if (*len != sizeof(struct ip_set_req_get_set) ||
1974 req_get->set.index >= inst->ip_set_max) {
1975 ret = -EINVAL;
1976 goto done;
1977 }
1978 nfnl_lock(NFNL_SUBSYS_IPSET);
1979 set = ip_set(inst, req_get->set.index);
1980 strncpy(req_get->set.name, set ? set->name : "",
1981 IPSET_MAXNAMELEN);
1982 nfnl_unlock(NFNL_SUBSYS_IPSET);
1983 goto copy;
1984 }
1985 default:
1986 ret = -EBADMSG;
1987 goto done;
1988 } /* end of switch(op) */
1989
1990 copy:
1991 ret = copy_to_user(user, data, copylen);
1992
1993 done:
1994 vfree(data);
1995 if (ret > 0)
1996 ret = 0;
1997 return ret;
1998 }
1999
2000 static struct nf_sockopt_ops so_set __read_mostly = {
2001 .pf = PF_INET,
2002 .get_optmin = SO_IP_SET,
2003 .get_optmax = SO_IP_SET + 1,
2004 .get = &ip_set_sockfn_get,
2005 .owner = THIS_MODULE,
2006 };
2007
2008 static int __net_init
ip_set_net_init(struct net * net)2009 ip_set_net_init(struct net *net)
2010 {
2011 struct ip_set_net *inst = ip_set_pernet(net);
2012 struct ip_set **list;
2013
2014 inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX;
2015 if (inst->ip_set_max >= IPSET_INVALID_ID)
2016 inst->ip_set_max = IPSET_INVALID_ID - 1;
2017
2018 list = kcalloc(inst->ip_set_max, sizeof(struct ip_set *), GFP_KERNEL);
2019 if (!list)
2020 return -ENOMEM;
2021 inst->is_deleted = false;
2022 inst->is_destroyed = false;
2023 rcu_assign_pointer(inst->ip_set_list, list);
2024 return 0;
2025 }
2026
2027 static void __net_exit
ip_set_net_exit(struct net * net)2028 ip_set_net_exit(struct net *net)
2029 {
2030 struct ip_set_net *inst = ip_set_pernet(net);
2031
2032 struct ip_set *set = NULL;
2033 ip_set_id_t i;
2034
2035 inst->is_deleted = true; /* flag for ip_set_nfnl_put */
2036
2037 for (i = 0; i < inst->ip_set_max; i++) {
2038 set = ip_set(inst, i);
2039 if (set) {
2040 ip_set(inst, i) = NULL;
2041 ip_set_destroy_set(set);
2042 }
2043 }
2044 kfree(rcu_dereference_protected(inst->ip_set_list, 1));
2045 }
2046
2047 static struct pernet_operations ip_set_net_ops = {
2048 .init = ip_set_net_init,
2049 .exit = ip_set_net_exit,
2050 .id = &ip_set_net_id,
2051 .size = sizeof(struct ip_set_net)
2052 };
2053
2054 static int __init
ip_set_init(void)2055 ip_set_init(void)
2056 {
2057 int ret = nfnetlink_subsys_register(&ip_set_netlink_subsys);
2058
2059 if (ret != 0) {
2060 pr_err("ip_set: cannot register with nfnetlink.\n");
2061 return ret;
2062 }
2063 ret = nf_register_sockopt(&so_set);
2064 if (ret != 0) {
2065 pr_err("SO_SET registry failed: %d\n", ret);
2066 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2067 return ret;
2068 }
2069 ret = register_pernet_subsys(&ip_set_net_ops);
2070 if (ret) {
2071 pr_err("ip_set: cannot register pernet_subsys.\n");
2072 nf_unregister_sockopt(&so_set);
2073 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2074 return ret;
2075 }
2076 pr_info("ip_set: protocol %u\n", IPSET_PROTOCOL);
2077 return 0;
2078 }
2079
2080 static void __exit
ip_set_fini(void)2081 ip_set_fini(void)
2082 {
2083 unregister_pernet_subsys(&ip_set_net_ops);
2084 nf_unregister_sockopt(&so_set);
2085 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2086 pr_debug("these are the famous last words\n");
2087 }
2088
2089 module_init(ip_set_init);
2090 module_exit(ip_set_fini);
2091