]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - net/netfilter/ipset/ip_set_core.c
projects / mirror_ubuntu-jammy-kernel.git / blob
? search:


89009c82a6b2408f637a2716e73b54eea3bafd0f
[mirror_ubuntu-jammy-kernel.git] / net / netfilter / ipset / ip_set_core.c
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
38 static 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 void
71 ip_set_type_lock(void)
72 {
73 mutex_lock(&ip_set_type_mutex);
74 }
75
76 static 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 *
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 lockdep_is_held(&ip_set_type_mutex))
91 if (STRNCMP(type->name, name) &&
92 (type->family == family ||
93 type->family == NFPROTO_UNSPEC) &&
94 revision >= type->revision_min &&
95 revision <= type->revision_max)
96 return type;
97 return NULL;
98 }
99
100 /* Unlock, try to load a set type module and lock again */
101 static bool
102 load_settype(const char *name)
103 {
104 nfnl_unlock(NFNL_SUBSYS_IPSET);
105 pr_debug("try to load ip_set_%s\n", name);
106 if (request_module("ip_set_%s", name) < 0) {
107 pr_warn("Can't find ip_set type %s\n", name);
108 nfnl_lock(NFNL_SUBSYS_IPSET);
109 return false;
110 }
111 nfnl_lock(NFNL_SUBSYS_IPSET);
112 return true;
113 }
114
115 /* Find a set type and reference it */
116 #define find_set_type_get(name, family, revision, found) \
117 __find_set_type_get(name, family, revision, found, false)
118
119 static int
120 __find_set_type_get(const char *name, u8 family, u8 revision,
121 struct ip_set_type **found, bool retry)
122 {
123 struct ip_set_type *type;
124 int err;
125
126 if (retry && !load_settype(name))
127 return -IPSET_ERR_FIND_TYPE;
128
129 rcu_read_lock();
130 *found = find_set_type(name, family, revision);
131 if (*found) {
132 err = !try_module_get((*found)->me) ? -EFAULT : 0;
133 goto unlock;
134 }
135 /* Make sure the type is already loaded
136 * but we don't support the revision
137 */
138 list_for_each_entry_rcu(type, &ip_set_type_list, list)
139 if (STRNCMP(type->name, name)) {
140 err = -IPSET_ERR_FIND_TYPE;
141 goto unlock;
142 }
143 rcu_read_unlock();
144
145 return retry ? -IPSET_ERR_FIND_TYPE :
146 __find_set_type_get(name, family, revision, found, true);
147
148 unlock:
149 rcu_read_unlock();
150 return err;
151 }
152
153 /* Find a given set type by name and family.
154 * If we succeeded, the supported minimal and maximum revisions are
155 * filled out.
156 */
157 #define find_set_type_minmax(name, family, min, max) \
158 __find_set_type_minmax(name, family, min, max, false)
159
160 static int
161 __find_set_type_minmax(const char *name, u8 family, u8 *min, u8 *max,
162 bool retry)
163 {
164 struct ip_set_type *type;
165 bool found = false;
166
167 if (retry && !load_settype(name))
168 return -IPSET_ERR_FIND_TYPE;
169
170 *min = 255; *max = 0;
171 rcu_read_lock();
172 list_for_each_entry_rcu(type, &ip_set_type_list, list)
173 if (STRNCMP(type->name, name) &&
174 (type->family == family ||
175 type->family == NFPROTO_UNSPEC)) {
176 found = true;
177 if (type->revision_min < *min)
178 *min = type->revision_min;
179 if (type->revision_max > *max)
180 *max = type->revision_max;
181 }
182 rcu_read_unlock();
183 if (found)
184 return 0;
185
186 return retry ? -IPSET_ERR_FIND_TYPE :
187 __find_set_type_minmax(name, family, min, max, true);
188 }
189
190 #define family_name(f) ((f) == NFPROTO_IPV4 ? "inet" : \
191 (f) == NFPROTO_IPV6 ? "inet6" : "any")
192
193 /* Register a set type structure. The type is identified by
194 * the unique triple of name, family and revision.
195 */
196 int
197 ip_set_type_register(struct ip_set_type *type)
198 {
199 int ret = 0;
200
201 if (type->protocol != IPSET_PROTOCOL) {
202 pr_warn("ip_set type %s, family %s, revision %u:%u uses wrong protocol version %u (want %u)\n",
203 type->name, family_name(type->family),
204 type->revision_min, type->revision_max,
205 type->protocol, IPSET_PROTOCOL);
206 return -EINVAL;
207 }
208
209 ip_set_type_lock();
210 if (find_set_type(type->name, type->family, type->revision_min)) {
211 /* Duplicate! */
212 pr_warn("ip_set type %s, family %s with revision min %u already registered!\n",
213 type->name, family_name(type->family),
214 type->revision_min);
215 ip_set_type_unlock();
216 return -EINVAL;
217 }
218 list_add_rcu(&type->list, &ip_set_type_list);
219 pr_debug("type %s, family %s, revision %u:%u registered.\n",
220 type->name, family_name(type->family),
221 type->revision_min, type->revision_max);
222 ip_set_type_unlock();
223
224 return ret;
225 }
226 EXPORT_SYMBOL_GPL(ip_set_type_register);
227
228 /* Unregister a set type. There's a small race with ip_set_create */
229 void
230 ip_set_type_unregister(struct ip_set_type *type)
231 {
232 ip_set_type_lock();
233 if (!find_set_type(type->name, type->family, type->revision_min)) {
234 pr_warn("ip_set type %s, family %s with revision min %u not registered\n",
235 type->name, family_name(type->family),
236 type->revision_min);
237 ip_set_type_unlock();
238 return;
239 }
240 list_del_rcu(&type->list);
241 pr_debug("type %s, family %s with revision min %u unregistered.\n",
242 type->name, family_name(type->family), type->revision_min);
243 ip_set_type_unlock();
244
245 synchronize_rcu();
246 }
247 EXPORT_SYMBOL_GPL(ip_set_type_unregister);
248
249 /* Utility functions */
250 void *
251 ip_set_alloc(size_t size)
252 {
253 return kvzalloc(size, GFP_KERNEL_ACCOUNT);
254 }
255 EXPORT_SYMBOL_GPL(ip_set_alloc);
256
257 void
258 ip_set_free(void *members)
259 {
260 pr_debug("%p: free with %s\n", members,
261 is_vmalloc_addr(members) ? "vfree" : "kfree");
262 kvfree(members);
263 }
264 EXPORT_SYMBOL_GPL(ip_set_free);
265
266 static bool
267 flag_nested(const struct nlattr *nla)
268 {
269 return nla->nla_type & NLA_F_NESTED;
270 }
271
272 static const struct nla_policy ipaddr_policy[IPSET_ATTR_IPADDR_MAX + 1] = {
273 [IPSET_ATTR_IPADDR_IPV4] = { .type = NLA_U32 },
274 [IPSET_ATTR_IPADDR_IPV6] = NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
275 };
276
277 int
278 ip_set_get_ipaddr4(struct nlattr *nla, __be32 *ipaddr)
279 {
280 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1];
281
282 if (unlikely(!flag_nested(nla)))
283 return -IPSET_ERR_PROTOCOL;
284 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla,
285 ipaddr_policy, NULL))
286 return -IPSET_ERR_PROTOCOL;
287 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV4)))
288 return -IPSET_ERR_PROTOCOL;
289
290 *ipaddr = nla_get_be32(tb[IPSET_ATTR_IPADDR_IPV4]);
291 return 0;
292 }
293 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr4);
294
295 int
296 ip_set_get_ipaddr6(struct nlattr *nla, union nf_inet_addr *ipaddr)
297 {
298 struct nlattr *tb[IPSET_ATTR_IPADDR_MAX + 1];
299
300 if (unlikely(!flag_nested(nla)))
301 return -IPSET_ERR_PROTOCOL;
302
303 if (nla_parse_nested(tb, IPSET_ATTR_IPADDR_MAX, nla,
304 ipaddr_policy, NULL))
305 return -IPSET_ERR_PROTOCOL;
306 if (unlikely(!ip_set_attr_netorder(tb, IPSET_ATTR_IPADDR_IPV6)))
307 return -IPSET_ERR_PROTOCOL;
308
309 memcpy(ipaddr, nla_data(tb[IPSET_ATTR_IPADDR_IPV6]),
310 sizeof(struct in6_addr));
311 return 0;
312 }
313 EXPORT_SYMBOL_GPL(ip_set_get_ipaddr6);
314
315 static u32
316 ip_set_timeout_get(const unsigned long *timeout)
317 {
318 u32 t;
319
320 if (*timeout == IPSET_ELEM_PERMANENT)
321 return 0;
322
323 t = jiffies_to_msecs(*timeout - jiffies) / MSEC_PER_SEC;
324 /* Zero value in userspace means no timeout */
325 return t == 0 ? 1 : t;
326 }
327
328 static char *
329 ip_set_comment_uget(struct nlattr *tb)
330 {
331 return nla_data(tb);
332 }
333
334 /* Called from uadd only, protected by the set spinlock.
335 * The kadt functions don't use the comment extensions in any way.
336 */
337 void
338 ip_set_init_comment(struct ip_set *set, struct ip_set_comment *comment,
339 const struct ip_set_ext *ext)
340 {
341 struct ip_set_comment_rcu *c = rcu_dereference_protected(comment->c, 1);
342 size_t len = ext->comment ? strlen(ext->comment) : 0;
343
344 if (unlikely(c)) {
345 set->ext_size -= sizeof(*c) + strlen(c->str) + 1;
346 kfree_rcu(c, rcu);
347 rcu_assign_pointer(comment->c, NULL);
348 }
349 if (!len)
350 return;
351 if (unlikely(len > IPSET_MAX_COMMENT_SIZE))
352 len = IPSET_MAX_COMMENT_SIZE;
353 c = kmalloc(sizeof(*c) + len + 1, GFP_ATOMIC);
354 if (unlikely(!c))
355 return;
356 strlcpy(c->str, ext->comment, len + 1);
357 set->ext_size += sizeof(*c) + strlen(c->str) + 1;
358 rcu_assign_pointer(comment->c, c);
359 }
360 EXPORT_SYMBOL_GPL(ip_set_init_comment);
361
362 /* Used only when dumping a set, protected by rcu_read_lock() */
363 static int
364 ip_set_put_comment(struct sk_buff *skb, const struct ip_set_comment *comment)
365 {
366 struct ip_set_comment_rcu *c = rcu_dereference(comment->c);
367
368 if (!c)
369 return 0;
370 return nla_put_string(skb, IPSET_ATTR_COMMENT, c->str);
371 }
372
373 /* Called from uadd/udel, flush or the garbage collectors protected
374 * by the set spinlock.
375 * Called when the set is destroyed and when there can't be any user
376 * of the set data anymore.
377 */
378 static void
379 ip_set_comment_free(struct ip_set *set, void *ptr)
380 {
381 struct ip_set_comment *comment = ptr;
382 struct ip_set_comment_rcu *c;
383
384 c = rcu_dereference_protected(comment->c, 1);
385 if (unlikely(!c))
386 return;
387 set->ext_size -= sizeof(*c) + strlen(c->str) + 1;
388 kfree_rcu(c, rcu);
389 rcu_assign_pointer(comment->c, NULL);
390 }
391
392 typedef void (*destroyer)(struct ip_set *, void *);
393 /* ipset data extension types, in size order */
394
395 const struct ip_set_ext_type ip_set_extensions[] = {
396 [IPSET_EXT_ID_COUNTER] = {
397 .type = IPSET_EXT_COUNTER,
398 .flag = IPSET_FLAG_WITH_COUNTERS,
399 .len = sizeof(struct ip_set_counter),
400 .align = __alignof__(struct ip_set_counter),
401 },
402 [IPSET_EXT_ID_TIMEOUT] = {
403 .type = IPSET_EXT_TIMEOUT,
404 .len = sizeof(unsigned long),
405 .align = __alignof__(unsigned long),
406 },
407 [IPSET_EXT_ID_SKBINFO] = {
408 .type = IPSET_EXT_SKBINFO,
409 .flag = IPSET_FLAG_WITH_SKBINFO,
410 .len = sizeof(struct ip_set_skbinfo),
411 .align = __alignof__(struct ip_set_skbinfo),
412 },
413 [IPSET_EXT_ID_COMMENT] = {
414 .type = IPSET_EXT_COMMENT | IPSET_EXT_DESTROY,
415 .flag = IPSET_FLAG_WITH_COMMENT,
416 .len = sizeof(struct ip_set_comment),
417 .align = __alignof__(struct ip_set_comment),
418 .destroy = ip_set_comment_free,
419 },
420 };
421 EXPORT_SYMBOL_GPL(ip_set_extensions);
422
423 static bool
424 add_extension(enum ip_set_ext_id id, u32 flags, struct nlattr *tb[])
425 {
426 return ip_set_extensions[id].flag ?
427 (flags & ip_set_extensions[id].flag) :
428 !!tb[IPSET_ATTR_TIMEOUT];
429 }
430
431 size_t
432 ip_set_elem_len(struct ip_set *set, struct nlattr *tb[], size_t len,
433 size_t align)
434 {
435 enum ip_set_ext_id id;
436 u32 cadt_flags = 0;
437
438 if (tb[IPSET_ATTR_CADT_FLAGS])
439 cadt_flags = ip_set_get_h32(tb[IPSET_ATTR_CADT_FLAGS]);
440 if (cadt_flags & IPSET_FLAG_WITH_FORCEADD)
441 set->flags |= IPSET_CREATE_FLAG_FORCEADD;
442 if (!align)
443 align = 1;
444 for (id = 0; id < IPSET_EXT_ID_MAX; id++) {
445 if (!add_extension(id, cadt_flags, tb))
446 continue;
447 if (align < ip_set_extensions[id].align)
448 align = ip_set_extensions[id].align;
449 len = ALIGN(len, ip_set_extensions[id].align);
450 set->offset[id] = len;
451 set->extensions |= ip_set_extensions[id].type;
452 len += ip_set_extensions[id].len;
453 }
454 return ALIGN(len, align);
455 }
456 EXPORT_SYMBOL_GPL(ip_set_elem_len);
457
458 int
459 ip_set_get_extensions(struct ip_set *set, struct nlattr *tb[],
460 struct ip_set_ext *ext)
461 {
462 u64 fullmark;
463
464 if (unlikely(!ip_set_optattr_netorder(tb, IPSET_ATTR_TIMEOUT) ||
465 !ip_set_optattr_netorder(tb, IPSET_ATTR_PACKETS) ||
466 !ip_set_optattr_netorder(tb, IPSET_ATTR_BYTES) ||
467 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBMARK) ||
468 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBPRIO) ||
469 !ip_set_optattr_netorder(tb, IPSET_ATTR_SKBQUEUE)))
470 return -IPSET_ERR_PROTOCOL;
471
472 if (tb[IPSET_ATTR_TIMEOUT]) {
473 if (!SET_WITH_TIMEOUT(set))
474 return -IPSET_ERR_TIMEOUT;
475 ext->timeout = ip_set_timeout_uget(tb[IPSET_ATTR_TIMEOUT]);
476 }
477 if (tb[IPSET_ATTR_BYTES] || tb[IPSET_ATTR_PACKETS]) {
478 if (!SET_WITH_COUNTER(set))
479 return -IPSET_ERR_COUNTER;
480 if (tb[IPSET_ATTR_BYTES])
481 ext->bytes = be64_to_cpu(nla_get_be64(
482 tb[IPSET_ATTR_BYTES]));
483 if (tb[IPSET_ATTR_PACKETS])
484 ext->packets = be64_to_cpu(nla_get_be64(
485 tb[IPSET_ATTR_PACKETS]));
486 }
487 if (tb[IPSET_ATTR_COMMENT]) {
488 if (!SET_WITH_COMMENT(set))
489 return -IPSET_ERR_COMMENT;
490 ext->comment = ip_set_comment_uget(tb[IPSET_ATTR_COMMENT]);
491 }
492 if (tb[IPSET_ATTR_SKBMARK]) {
493 if (!SET_WITH_SKBINFO(set))
494 return -IPSET_ERR_SKBINFO;
495 fullmark = be64_to_cpu(nla_get_be64(tb[IPSET_ATTR_SKBMARK]));
496 ext->skbinfo.skbmark = fullmark >> 32;
497 ext->skbinfo.skbmarkmask = fullmark & 0xffffffff;
498 }
499 if (tb[IPSET_ATTR_SKBPRIO]) {
500 if (!SET_WITH_SKBINFO(set))
501 return -IPSET_ERR_SKBINFO;
502 ext->skbinfo.skbprio =
503 be32_to_cpu(nla_get_be32(tb[IPSET_ATTR_SKBPRIO]));
504 }
505 if (tb[IPSET_ATTR_SKBQUEUE]) {
506 if (!SET_WITH_SKBINFO(set))
507 return -IPSET_ERR_SKBINFO;
508 ext->skbinfo.skbqueue =
509 be16_to_cpu(nla_get_be16(tb[IPSET_ATTR_SKBQUEUE]));
510 }
511 return 0;
512 }
513 EXPORT_SYMBOL_GPL(ip_set_get_extensions);
514
515 static u64
516 ip_set_get_bytes(const struct ip_set_counter *counter)
517 {
518 return (u64)atomic64_read(&(counter)->bytes);
519 }
520
521 static u64
522 ip_set_get_packets(const struct ip_set_counter *counter)
523 {
524 return (u64)atomic64_read(&(counter)->packets);
525 }
526
527 static bool
528 ip_set_put_counter(struct sk_buff *skb, const struct ip_set_counter *counter)
529 {
530 return nla_put_net64(skb, IPSET_ATTR_BYTES,
531 cpu_to_be64(ip_set_get_bytes(counter)),
532 IPSET_ATTR_PAD) ||
533 nla_put_net64(skb, IPSET_ATTR_PACKETS,
534 cpu_to_be64(ip_set_get_packets(counter)),
535 IPSET_ATTR_PAD);
536 }
537
538 static bool
539 ip_set_put_skbinfo(struct sk_buff *skb, const struct ip_set_skbinfo *skbinfo)
540 {
541 /* Send nonzero parameters only */
542 return ((skbinfo->skbmark || skbinfo->skbmarkmask) &&
543 nla_put_net64(skb, IPSET_ATTR_SKBMARK,
544 cpu_to_be64((u64)skbinfo->skbmark << 32 |
545 skbinfo->skbmarkmask),
546 IPSET_ATTR_PAD)) ||
547 (skbinfo->skbprio &&
548 nla_put_net32(skb, IPSET_ATTR_SKBPRIO,
549 cpu_to_be32(skbinfo->skbprio))) ||
550 (skbinfo->skbqueue &&
551 nla_put_net16(skb, IPSET_ATTR_SKBQUEUE,
552 cpu_to_be16(skbinfo->skbqueue)));
553 }
554
555 int
556 ip_set_put_extensions(struct sk_buff *skb, const struct ip_set *set,
557 const void *e, bool active)
558 {
559 if (SET_WITH_TIMEOUT(set)) {
560 unsigned long *timeout = ext_timeout(e, set);
561
562 if (nla_put_net32(skb, IPSET_ATTR_TIMEOUT,
563 htonl(active ? ip_set_timeout_get(timeout)
564 : *timeout)))
565 return -EMSGSIZE;
566 }
567 if (SET_WITH_COUNTER(set) &&
568 ip_set_put_counter(skb, ext_counter(e, set)))
569 return -EMSGSIZE;
570 if (SET_WITH_COMMENT(set) &&
571 ip_set_put_comment(skb, ext_comment(e, set)))
572 return -EMSGSIZE;
573 if (SET_WITH_SKBINFO(set) &&
574 ip_set_put_skbinfo(skb, ext_skbinfo(e, set)))
575 return -EMSGSIZE;
576 return 0;
577 }
578 EXPORT_SYMBOL_GPL(ip_set_put_extensions);
579
580 static bool
581 ip_set_match_counter(u64 counter, u64 match, u8 op)
582 {
583 switch (op) {
584 case IPSET_COUNTER_NONE:
585 return true;
586 case IPSET_COUNTER_EQ:
587 return counter == match;
588 case IPSET_COUNTER_NE:
589 return counter != match;
590 case IPSET_COUNTER_LT:
591 return counter < match;
592 case IPSET_COUNTER_GT:
593 return counter > match;
594 }
595 return false;
596 }
597
598 static void
599 ip_set_add_bytes(u64 bytes, struct ip_set_counter *counter)
600 {
601 atomic64_add((long long)bytes, &(counter)->bytes);
602 }
603
604 static void
605 ip_set_add_packets(u64 packets, struct ip_set_counter *counter)
606 {
607 atomic64_add((long long)packets, &(counter)->packets);
608 }
609
610 static void
611 ip_set_update_counter(struct ip_set_counter *counter,
612 const struct ip_set_ext *ext, u32 flags)
613 {
614 if (ext->packets != ULLONG_MAX &&
615 !(flags & IPSET_FLAG_SKIP_COUNTER_UPDATE)) {
616 ip_set_add_bytes(ext->bytes, counter);
617 ip_set_add_packets(ext->packets, counter);
618 }
619 }
620
621 static void
622 ip_set_get_skbinfo(struct ip_set_skbinfo *skbinfo,
623 const struct ip_set_ext *ext,
624 struct ip_set_ext *mext, u32 flags)
625 {
626 mext->skbinfo = *skbinfo;
627 }
628
629 bool
630 ip_set_match_extensions(struct ip_set *set, const struct ip_set_ext *ext,
631 struct ip_set_ext *mext, u32 flags, void *data)
632 {
633 if (SET_WITH_TIMEOUT(set) &&
634 ip_set_timeout_expired(ext_timeout(data, set)))
635 return false;
636 if (SET_WITH_COUNTER(set)) {
637 struct ip_set_counter *counter = ext_counter(data, set);
638
639 ip_set_update_counter(counter, ext, flags);
640
641 if (flags & IPSET_FLAG_MATCH_COUNTERS &&
642 !(ip_set_match_counter(ip_set_get_packets(counter),
643 mext->packets, mext->packets_op) &&
644 ip_set_match_counter(ip_set_get_bytes(counter),
645 mext->bytes, mext->bytes_op)))
646 return false;
647 }
648 if (SET_WITH_SKBINFO(set))
649 ip_set_get_skbinfo(ext_skbinfo(data, set),
650 ext, mext, flags);
651 return true;
652 }
653 EXPORT_SYMBOL_GPL(ip_set_match_extensions);
654
655 /* Creating/destroying/renaming/swapping affect the existence and
656 * the properties of a set. All of these can be executed from userspace
657 * only and serialized by the nfnl mutex indirectly from nfnetlink.
658 *
659 * Sets are identified by their index in ip_set_list and the index
660 * is used by the external references (set/SET netfilter modules).
661 *
662 * The set behind an index may change by swapping only, from userspace.
663 */
664
665 static void
666 __ip_set_get(struct ip_set *set)
667 {
668 write_lock_bh(&ip_set_ref_lock);
669 set->ref++;
670 write_unlock_bh(&ip_set_ref_lock);
671 }
672
673 static void
674 __ip_set_put(struct ip_set *set)
675 {
676 write_lock_bh(&ip_set_ref_lock);
677 BUG_ON(set->ref == 0);
678 set->ref--;
679 write_unlock_bh(&ip_set_ref_lock);
680 }
681
682 /* set->ref can be swapped out by ip_set_swap, netlink events (like dump) need
683 * a separate reference counter
684 */
685 static void
686 __ip_set_put_netlink(struct ip_set *set)
687 {
688 write_lock_bh(&ip_set_ref_lock);
689 BUG_ON(set->ref_netlink == 0);
690 set->ref_netlink--;
691 write_unlock_bh(&ip_set_ref_lock);
692 }
693
694 /* Add, del and test set entries from kernel.
695 *
696 * The set behind the index must exist and must be referenced
697 * so it can't be destroyed (or changed) under our foot.
698 */
699
700 static struct ip_set *
701 ip_set_rcu_get(struct net *net, ip_set_id_t index)
702 {
703 struct ip_set *set;
704 struct ip_set_net *inst = ip_set_pernet(net);
705
706 rcu_read_lock();
707 /* ip_set_list itself needs to be protected */
708 set = rcu_dereference(inst->ip_set_list)[index];
709 rcu_read_unlock();
710
711 return set;
712 }
713
714 static inline void
715 ip_set_lock(struct ip_set *set)
716 {
717 if (!set->variant->region_lock)
718 spin_lock_bh(&set->lock);
719 }
720
721 static inline void
722 ip_set_unlock(struct ip_set *set)
723 {
724 if (!set->variant->region_lock)
725 spin_unlock_bh(&set->lock);
726 }
727
728 int
729 ip_set_test(ip_set_id_t index, const struct sk_buff *skb,
730 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
731 {
732 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
733 int ret = 0;
734
735 BUG_ON(!set);
736 pr_debug("set %s, index %u\n", set->name, index);
737
738 if (opt->dim < set->type->dimension ||
739 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
740 return 0;
741
742 rcu_read_lock_bh();
743 ret = set->variant->kadt(set, skb, par, IPSET_TEST, opt);
744 rcu_read_unlock_bh();
745
746 if (ret == -EAGAIN) {
747 /* Type requests element to be completed */
748 pr_debug("element must be completed, ADD is triggered\n");
749 ip_set_lock(set);
750 set->variant->kadt(set, skb, par, IPSET_ADD, opt);
751 ip_set_unlock(set);
752 ret = 1;
753 } else {
754 /* --return-nomatch: invert matched element */
755 if ((opt->cmdflags & IPSET_FLAG_RETURN_NOMATCH) &&
756 (set->type->features & IPSET_TYPE_NOMATCH) &&
757 (ret > 0 || ret == -ENOTEMPTY))
758 ret = -ret;
759 }
760
761 /* Convert error codes to nomatch */
762 return (ret < 0 ? 0 : ret);
763 }
764 EXPORT_SYMBOL_GPL(ip_set_test);
765
766 int
767 ip_set_add(ip_set_id_t index, const struct sk_buff *skb,
768 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
769 {
770 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
771 int ret;
772
773 BUG_ON(!set);
774 pr_debug("set %s, index %u\n", set->name, index);
775
776 if (opt->dim < set->type->dimension ||
777 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
778 return -IPSET_ERR_TYPE_MISMATCH;
779
780 ip_set_lock(set);
781 ret = set->variant->kadt(set, skb, par, IPSET_ADD, opt);
782 ip_set_unlock(set);
783
784 return ret;
785 }
786 EXPORT_SYMBOL_GPL(ip_set_add);
787
788 int
789 ip_set_del(ip_set_id_t index, const struct sk_buff *skb,
790 const struct xt_action_param *par, struct ip_set_adt_opt *opt)
791 {
792 struct ip_set *set = ip_set_rcu_get(xt_net(par), index);
793 int ret = 0;
794
795 BUG_ON(!set);
796 pr_debug("set %s, index %u\n", set->name, index);
797
798 if (opt->dim < set->type->dimension ||
799 !(opt->family == set->family || set->family == NFPROTO_UNSPEC))
800 return -IPSET_ERR_TYPE_MISMATCH;
801
802 ip_set_lock(set);
803 ret = set->variant->kadt(set, skb, par, IPSET_DEL, opt);
804 ip_set_unlock(set);
805
806 return ret;
807 }
808 EXPORT_SYMBOL_GPL(ip_set_del);
809
810 /* Find set by name, reference it once. The reference makes sure the
811 * thing pointed to, does not go away under our feet.
812 *
813 */
814 ip_set_id_t
815 ip_set_get_byname(struct net *net, const char *name, struct ip_set **set)
816 {
817 ip_set_id_t i, index = IPSET_INVALID_ID;
818 struct ip_set *s;
819 struct ip_set_net *inst = ip_set_pernet(net);
820
821 rcu_read_lock();
822 for (i = 0; i < inst->ip_set_max; i++) {
823 s = rcu_dereference(inst->ip_set_list)[i];
824 if (s && STRNCMP(s->name, name)) {
825 __ip_set_get(s);
826 index = i;
827 *set = s;
828 break;
829 }
830 }
831 rcu_read_unlock();
832
833 return index;
834 }
835 EXPORT_SYMBOL_GPL(ip_set_get_byname);
836
837 /* If the given set pointer points to a valid set, decrement
838 * reference count by 1. The caller shall not assume the index
839 * to be valid, after calling this function.
840 *
841 */
842
843 static void
844 __ip_set_put_byindex(struct ip_set_net *inst, ip_set_id_t index)
845 {
846 struct ip_set *set;
847
848 rcu_read_lock();
849 set = rcu_dereference(inst->ip_set_list)[index];
850 if (set)
851 __ip_set_put(set);
852 rcu_read_unlock();
853 }
854
855 void
856 ip_set_put_byindex(struct net *net, ip_set_id_t index)
857 {
858 struct ip_set_net *inst = ip_set_pernet(net);
859
860 __ip_set_put_byindex(inst, index);
861 }
862 EXPORT_SYMBOL_GPL(ip_set_put_byindex);
863
864 /* Get the name of a set behind a set index.
865 * Set itself is protected by RCU, but its name isn't: to protect against
866 * renaming, grab ip_set_ref_lock as reader (see ip_set_rename()) and copy the
867 * name.
868 */
869 void
870 ip_set_name_byindex(struct net *net, ip_set_id_t index, char *name)
871 {
872 struct ip_set *set = ip_set_rcu_get(net, index);
873
874 BUG_ON(!set);
875
876 read_lock_bh(&ip_set_ref_lock);
877 strncpy(name, set->name, IPSET_MAXNAMELEN);
878 read_unlock_bh(&ip_set_ref_lock);
879 }
880 EXPORT_SYMBOL_GPL(ip_set_name_byindex);
881
882 /* Routines to call by external subsystems, which do not
883 * call nfnl_lock for us.
884 */
885
886 /* Find set by index, reference it once. The reference makes sure the
887 * thing pointed to, does not go away under our feet.
888 *
889 * The nfnl mutex is used in the function.
890 */
891 ip_set_id_t
892 ip_set_nfnl_get_byindex(struct net *net, ip_set_id_t index)
893 {
894 struct ip_set *set;
895 struct ip_set_net *inst = ip_set_pernet(net);
896
897 if (index >= inst->ip_set_max)
898 return IPSET_INVALID_ID;
899
900 nfnl_lock(NFNL_SUBSYS_IPSET);
901 set = ip_set(inst, index);
902 if (set)
903 __ip_set_get(set);
904 else
905 index = IPSET_INVALID_ID;
906 nfnl_unlock(NFNL_SUBSYS_IPSET);
907
908 return index;
909 }
910 EXPORT_SYMBOL_GPL(ip_set_nfnl_get_byindex);
911
912 /* If the given set pointer points to a valid set, decrement
913 * reference count by 1. The caller shall not assume the index
914 * to be valid, after calling this function.
915 *
916 * The nfnl mutex is used in the function.
917 */
918 void
919 ip_set_nfnl_put(struct net *net, ip_set_id_t index)
920 {
921 struct ip_set *set;
922 struct ip_set_net *inst = ip_set_pernet(net);
923
924 nfnl_lock(NFNL_SUBSYS_IPSET);
925 if (!inst->is_deleted) { /* already deleted from ip_set_net_exit() */
926 set = ip_set(inst, index);
927 if (set)
928 __ip_set_put(set);
929 }
930 nfnl_unlock(NFNL_SUBSYS_IPSET);
931 }
932 EXPORT_SYMBOL_GPL(ip_set_nfnl_put);
933
934 /* Communication protocol with userspace over netlink.
935 *
936 * The commands are serialized by the nfnl mutex.
937 */
938
939 static inline u8 protocol(const struct nlattr * const tb[])
940 {
941 return nla_get_u8(tb[IPSET_ATTR_PROTOCOL]);
942 }
943
944 static inline bool
945 protocol_failed(const struct nlattr * const tb[])
946 {
947 return !tb[IPSET_ATTR_PROTOCOL] || protocol(tb) != IPSET_PROTOCOL;
948 }
949
950 static inline bool
951 protocol_min_failed(const struct nlattr * const tb[])
952 {
953 return !tb[IPSET_ATTR_PROTOCOL] || protocol(tb) < IPSET_PROTOCOL_MIN;
954 }
955
956 static inline u32
957 flag_exist(const struct nlmsghdr *nlh)
958 {
959 return nlh->nlmsg_flags & NLM_F_EXCL ? 0 : IPSET_FLAG_EXIST;
960 }
961
962 static struct nlmsghdr *
963 start_msg(struct sk_buff *skb, u32 portid, u32 seq, unsigned int flags,
964 enum ipset_cmd cmd)
965 {
966 struct nlmsghdr *nlh;
967 struct nfgenmsg *nfmsg;
968
969 nlh = nlmsg_put(skb, portid, seq, nfnl_msg_type(NFNL_SUBSYS_IPSET, cmd),
970 sizeof(*nfmsg), flags);
971 if (!nlh)
972 return NULL;
973
974 nfmsg = nlmsg_data(nlh);
975 nfmsg->nfgen_family = NFPROTO_IPV4;
976 nfmsg->version = NFNETLINK_V0;
977 nfmsg->res_id = 0;
978
979 return nlh;
980 }
981
982 /* Create a set */
983
984 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = {
985 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
986 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
987 .len = IPSET_MAXNAMELEN - 1 },
988 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
989 .len = IPSET_MAXNAMELEN - 1},
990 [IPSET_ATTR_REVISION] = { .type = NLA_U8 },
991 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
992 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
993 };
994
995 static struct ip_set *
996 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id)
997 {
998 struct ip_set *set = NULL;
999 ip_set_id_t i;
1000
1001 *id = IPSET_INVALID_ID;
1002 for (i = 0; i < inst->ip_set_max; i++) {
1003 set = ip_set(inst, i);
1004 if (set && STRNCMP(set->name, name)) {
1005 *id = i;
1006 break;
1007 }
1008 }
1009 return (*id == IPSET_INVALID_ID ? NULL : set);
1010 }
1011
1012 static inline struct ip_set *
1013 find_set(struct ip_set_net *inst, const char *name)
1014 {
1015 ip_set_id_t id;
1016
1017 return find_set_and_id(inst, name, &id);
1018 }
1019
1020 static int
1021 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index,
1022 struct ip_set **set)
1023 {
1024 struct ip_set *s;
1025 ip_set_id_t i;
1026
1027 *index = IPSET_INVALID_ID;
1028 for (i = 0; i < inst->ip_set_max; i++) {
1029 s = ip_set(inst, i);
1030 if (!s) {
1031 if (*index == IPSET_INVALID_ID)
1032 *index = i;
1033 } else if (STRNCMP(name, s->name)) {
1034 /* Name clash */
1035 *set = s;
1036 return -EEXIST;
1037 }
1038 }
1039 if (*index == IPSET_INVALID_ID)
1040 /* No free slot remained */
1041 return -IPSET_ERR_MAX_SETS;
1042 return 0;
1043 }
1044
1045 static int ip_set_none(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1046 const struct nlmsghdr *nlh,
1047 const struct nlattr * const attr[],
1048 struct netlink_ext_ack *extack)
1049 {
1050 return -EOPNOTSUPP;
1051 }
1052
1053 static int ip_set_create(struct net *net, struct sock *ctnl,
1054 struct sk_buff *skb, const struct nlmsghdr *nlh,
1055 const struct nlattr * const attr[],
1056 struct netlink_ext_ack *extack)
1057 {
1058 struct ip_set_net *inst = ip_set_pernet(net);
1059 struct ip_set *set, *clash = NULL;
1060 ip_set_id_t index = IPSET_INVALID_ID;
1061 struct nlattr *tb[IPSET_ATTR_CREATE_MAX + 1] = {};
1062 const char *name, *typename;
1063 u8 family, revision;
1064 u32 flags = flag_exist(nlh);
1065 int ret = 0;
1066
1067 if (unlikely(protocol_min_failed(attr) ||
1068 !attr[IPSET_ATTR_SETNAME] ||
1069 !attr[IPSET_ATTR_TYPENAME] ||
1070 !attr[IPSET_ATTR_REVISION] ||
1071 !attr[IPSET_ATTR_FAMILY] ||
1072 (attr[IPSET_ATTR_DATA] &&
1073 !flag_nested(attr[IPSET_ATTR_DATA]))))
1074 return -IPSET_ERR_PROTOCOL;
1075
1076 name = nla_data(attr[IPSET_ATTR_SETNAME]);
1077 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1078 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1079 revision = nla_get_u8(attr[IPSET_ATTR_REVISION]);
1080 pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
1081 name, typename, family_name(family), revision);
1082
1083 /* First, and without any locks, allocate and initialize
1084 * a normal base set structure.
1085 */
1086 set = kzalloc(sizeof(*set), GFP_KERNEL);
1087 if (!set)
1088 return -ENOMEM;
1089 spin_lock_init(&set->lock);
1090 strlcpy(set->name, name, IPSET_MAXNAMELEN);
1091 set->family = family;
1092 set->revision = revision;
1093
1094 /* Next, check that we know the type, and take
1095 * a reference on the type, to make sure it stays available
1096 * while constructing our new set.
1097 *
1098 * After referencing the type, we try to create the type
1099 * specific part of the set without holding any locks.
1100 */
1101 ret = find_set_type_get(typename, family, revision, &set->type);
1102 if (ret)
1103 goto out;
1104
1105 /* Without holding any locks, create private part. */
1106 if (attr[IPSET_ATTR_DATA] &&
1107 nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA],
1108 set->type->create_policy, NULL)) {
1109 ret = -IPSET_ERR_PROTOCOL;
1110 goto put_out;
1111 }
1112 /* Set create flags depending on the type revision */
1113 set->flags |= set->type->create_flags[revision];
1114
1115 ret = set->type->create(net, set, tb, flags);
1116 if (ret != 0)
1117 goto put_out;
1118
1119 /* BTW, ret==0 here. */
1120
1121 /* Here, we have a valid, constructed set and we are protected
1122 * by the nfnl mutex. Find the first free index in ip_set_list
1123 * and check clashing.
1124 */
1125 ret = find_free_id(inst, set->name, &index, &clash);
1126 if (ret == -EEXIST) {
1127 /* If this is the same set and requested, ignore error */
1128 if ((flags & IPSET_FLAG_EXIST) &&
1129 STRNCMP(set->type->name, clash->type->name) &&
1130 set->type->family == clash->type->family &&
1131 set->type->revision_min == clash->type->revision_min &&
1132 set->type->revision_max == clash->type->revision_max &&
1133 set->variant->same_set(set, clash))
1134 ret = 0;
1135 goto cleanup;
1136 } else if (ret == -IPSET_ERR_MAX_SETS) {
1137 struct ip_set **list, **tmp;
1138 ip_set_id_t i = inst->ip_set_max + IP_SET_INC;
1139
1140 if (i < inst->ip_set_max || i == IPSET_INVALID_ID)
1141 /* Wraparound */
1142 goto cleanup;
1143
1144 list = kvcalloc(i, sizeof(struct ip_set *), GFP_KERNEL);
1145 if (!list)
1146 goto cleanup;
1147 /* nfnl mutex is held, both lists are valid */
1148 tmp = ip_set_dereference(inst->ip_set_list);
1149 memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max);
1150 rcu_assign_pointer(inst->ip_set_list, list);
1151 /* Make sure all current packets have passed through */
1152 synchronize_net();
1153 /* Use new list */
1154 index = inst->ip_set_max;
1155 inst->ip_set_max = i;
1156 kvfree(tmp);
1157 ret = 0;
1158 } else if (ret) {
1159 goto cleanup;
1160 }
1161
1162 /* Finally! Add our shiny new set to the list, and be done. */
1163 pr_debug("create: '%s' created with index %u!\n", set->name, index);
1164 ip_set(inst, index) = set;
1165
1166 return ret;
1167
1168 cleanup:
1169 set->variant->destroy(set);
1170 put_out:
1171 module_put(set->type->me);
1172 out:
1173 kfree(set);
1174 return ret;
1175 }
1176
1177 /* Destroy sets */
1178
1179 static const struct nla_policy
1180 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = {
1181 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1182 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1183 .len = IPSET_MAXNAMELEN - 1 },
1184 };
1185
1186 static void
1187 ip_set_destroy_set(struct ip_set *set)
1188 {
1189 pr_debug("set: %s\n", set->name);
1190
1191 /* Must call it without holding any lock */
1192 set->variant->destroy(set);
1193 module_put(set->type->me);
1194 kfree(set);
1195 }
1196
1197 static int ip_set_destroy(struct net *net, struct sock *ctnl,
1198 struct sk_buff *skb, const struct nlmsghdr *nlh,
1199 const struct nlattr * const attr[],
1200 struct netlink_ext_ack *extack)
1201 {
1202 struct ip_set_net *inst = ip_set_pernet(net);
1203 struct ip_set *s;
1204 ip_set_id_t i;
1205 int ret = 0;
1206
1207 if (unlikely(protocol_min_failed(attr)))
1208 return -IPSET_ERR_PROTOCOL;
1209
1210 /* Must wait for flush to be really finished in list:set */
1211 rcu_barrier();
1212
1213 /* Commands are serialized and references are
1214 * protected by the ip_set_ref_lock.
1215 * External systems (i.e. xt_set) must call
1216 * ip_set_put|get_nfnl_* functions, that way we
1217 * can safely check references here.
1218 *
1219 * list:set timer can only decrement the reference
1220 * counter, so if it's already zero, we can proceed
1221 * without holding the lock.
1222 */
1223 read_lock_bh(&ip_set_ref_lock);
1224 if (!attr[IPSET_ATTR_SETNAME]) {
1225 for (i = 0; i < inst->ip_set_max; i++) {
1226 s = ip_set(inst, i);
1227 if (s && (s->ref || s->ref_netlink)) {
1228 ret = -IPSET_ERR_BUSY;
1229 goto out;
1230 }
1231 }
1232 inst->is_destroyed = true;
1233 read_unlock_bh(&ip_set_ref_lock);
1234 for (i = 0; i < inst->ip_set_max; i++) {
1235 s = ip_set(inst, i);
1236 if (s) {
1237 ip_set(inst, i) = NULL;
1238 ip_set_destroy_set(s);
1239 }
1240 }
1241 /* Modified by ip_set_destroy() only, which is serialized */
1242 inst->is_destroyed = false;
1243 } else {
1244 u32 flags = flag_exist(nlh);
1245 s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1246 &i);
1247 if (!s) {
1248 if (!(flags & IPSET_FLAG_EXIST))
1249 ret = -ENOENT;
1250 goto out;
1251 } else if (s->ref || s->ref_netlink) {
1252 ret = -IPSET_ERR_BUSY;
1253 goto out;
1254 }
1255 ip_set(inst, i) = NULL;
1256 read_unlock_bh(&ip_set_ref_lock);
1257
1258 ip_set_destroy_set(s);
1259 }
1260 return 0;
1261 out:
1262 read_unlock_bh(&ip_set_ref_lock);
1263 return ret;
1264 }
1265
1266 /* Flush sets */
1267
1268 static void
1269 ip_set_flush_set(struct ip_set *set)
1270 {
1271 pr_debug("set: %s\n", set->name);
1272
1273 ip_set_lock(set);
1274 set->variant->flush(set);
1275 ip_set_unlock(set);
1276 }
1277
1278 static int ip_set_flush(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1279 const struct nlmsghdr *nlh,
1280 const struct nlattr * const attr[],
1281 struct netlink_ext_ack *extack)
1282 {
1283 struct ip_set_net *inst = ip_set_pernet(net);
1284 struct ip_set *s;
1285 ip_set_id_t i;
1286
1287 if (unlikely(protocol_min_failed(attr)))
1288 return -IPSET_ERR_PROTOCOL;
1289
1290 if (!attr[IPSET_ATTR_SETNAME]) {
1291 for (i = 0; i < inst->ip_set_max; i++) {
1292 s = ip_set(inst, i);
1293 if (s)
1294 ip_set_flush_set(s);
1295 }
1296 } else {
1297 s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1298 if (!s)
1299 return -ENOENT;
1300
1301 ip_set_flush_set(s);
1302 }
1303
1304 return 0;
1305 }
1306
1307 /* Rename a set */
1308
1309 static const struct nla_policy
1310 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = {
1311 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1312 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1313 .len = IPSET_MAXNAMELEN - 1 },
1314 [IPSET_ATTR_SETNAME2] = { .type = NLA_NUL_STRING,
1315 .len = IPSET_MAXNAMELEN - 1 },
1316 };
1317
1318 static int ip_set_rename(struct net *net, struct sock *ctnl,
1319 struct sk_buff *skb, const struct nlmsghdr *nlh,
1320 const struct nlattr * const attr[],
1321 struct netlink_ext_ack *extack)
1322 {
1323 struct ip_set_net *inst = ip_set_pernet(net);
1324 struct ip_set *set, *s;
1325 const char *name2;
1326 ip_set_id_t i;
1327 int ret = 0;
1328
1329 if (unlikely(protocol_min_failed(attr) ||
1330 !attr[IPSET_ATTR_SETNAME] ||
1331 !attr[IPSET_ATTR_SETNAME2]))
1332 return -IPSET_ERR_PROTOCOL;
1333
1334 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1335 if (!set)
1336 return -ENOENT;
1337
1338 write_lock_bh(&ip_set_ref_lock);
1339 if (set->ref != 0 || set->ref_netlink != 0) {
1340 ret = -IPSET_ERR_REFERENCED;
1341 goto out;
1342 }
1343
1344 name2 = nla_data(attr[IPSET_ATTR_SETNAME2]);
1345 for (i = 0; i < inst->ip_set_max; i++) {
1346 s = ip_set(inst, i);
1347 if (s && STRNCMP(s->name, name2)) {
1348 ret = -IPSET_ERR_EXIST_SETNAME2;
1349 goto out;
1350 }
1351 }
1352 strncpy(set->name, name2, IPSET_MAXNAMELEN);
1353
1354 out:
1355 write_unlock_bh(&ip_set_ref_lock);
1356 return ret;
1357 }
1358
1359 /* Swap two sets so that name/index points to the other.
1360 * References and set names are also swapped.
1361 *
1362 * The commands are serialized by the nfnl mutex and references are
1363 * protected by the ip_set_ref_lock. The kernel interfaces
1364 * do not hold the mutex but the pointer settings are atomic
1365 * so the ip_set_list always contains valid pointers to the sets.
1366 */
1367
1368 static int ip_set_swap(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1369 const struct nlmsghdr *nlh,
1370 const struct nlattr * const attr[],
1371 struct netlink_ext_ack *extack)
1372 {
1373 struct ip_set_net *inst = ip_set_pernet(net);
1374 struct ip_set *from, *to;
1375 ip_set_id_t from_id, to_id;
1376 char from_name[IPSET_MAXNAMELEN];
1377
1378 if (unlikely(protocol_min_failed(attr) ||
1379 !attr[IPSET_ATTR_SETNAME] ||
1380 !attr[IPSET_ATTR_SETNAME2]))
1381 return -IPSET_ERR_PROTOCOL;
1382
1383 from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1384 &from_id);
1385 if (!from)
1386 return -ENOENT;
1387
1388 to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]),
1389 &to_id);
1390 if (!to)
1391 return -IPSET_ERR_EXIST_SETNAME2;
1392
1393 /* Features must not change.
1394 * Not an artifical restriction anymore, as we must prevent
1395 * possible loops created by swapping in setlist type of sets.
1396 */
1397 if (!(from->type->features == to->type->features &&
1398 from->family == to->family))
1399 return -IPSET_ERR_TYPE_MISMATCH;
1400
1401 write_lock_bh(&ip_set_ref_lock);
1402
1403 if (from->ref_netlink || to->ref_netlink) {
1404 write_unlock_bh(&ip_set_ref_lock);
1405 return -EBUSY;
1406 }
1407
1408 strncpy(from_name, from->name, IPSET_MAXNAMELEN);
1409 strncpy(from->name, to->name, IPSET_MAXNAMELEN);
1410 strncpy(to->name, from_name, IPSET_MAXNAMELEN);
1411
1412 swap(from->ref, to->ref);
1413 ip_set(inst, from_id) = to;
1414 ip_set(inst, to_id) = from;
1415 write_unlock_bh(&ip_set_ref_lock);
1416
1417 return 0;
1418 }
1419
1420 /* List/save set data */
1421
1422 #define DUMP_INIT 0
1423 #define DUMP_ALL 1
1424 #define DUMP_ONE 2
1425 #define DUMP_LAST 3
1426
1427 #define DUMP_TYPE(arg) (((u32)(arg)) & 0x0000FFFF)
1428 #define DUMP_FLAGS(arg) (((u32)(arg)) >> 16)
1429
1430 int
1431 ip_set_put_flags(struct sk_buff *skb, struct ip_set *set)
1432 {
1433 u32 cadt_flags = 0;
1434
1435 if (SET_WITH_TIMEOUT(set))
1436 if (unlikely(nla_put_net32(skb, IPSET_ATTR_TIMEOUT,
1437 htonl(set->timeout))))
1438 return -EMSGSIZE;
1439 if (SET_WITH_COUNTER(set))
1440 cadt_flags |= IPSET_FLAG_WITH_COUNTERS;
1441 if (SET_WITH_COMMENT(set))
1442 cadt_flags |= IPSET_FLAG_WITH_COMMENT;
1443 if (SET_WITH_SKBINFO(set))
1444 cadt_flags |= IPSET_FLAG_WITH_SKBINFO;
1445 if (SET_WITH_FORCEADD(set))
1446 cadt_flags |= IPSET_FLAG_WITH_FORCEADD;
1447
1448 if (!cadt_flags)
1449 return 0;
1450 return nla_put_net32(skb, IPSET_ATTR_CADT_FLAGS, htonl(cadt_flags));
1451 }
1452 EXPORT_SYMBOL_GPL(ip_set_put_flags);
1453
1454 static int
1455 ip_set_dump_done(struct netlink_callback *cb)
1456 {
1457 if (cb->args[IPSET_CB_ARG0]) {
1458 struct ip_set_net *inst =
1459 (struct ip_set_net *)cb->args[IPSET_CB_NET];
1460 ip_set_id_t index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1461 struct ip_set *set = ip_set_ref_netlink(inst, index);
1462
1463 if (set->variant->uref)
1464 set->variant->uref(set, cb, false);
1465 pr_debug("release set %s\n", set->name);
1466 __ip_set_put_netlink(set);
1467 }
1468 return 0;
1469 }
1470
1471 static inline void
1472 dump_attrs(struct nlmsghdr *nlh)
1473 {
1474 const struct nlattr *attr;
1475 int rem;
1476
1477 pr_debug("dump nlmsg\n");
1478 nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) {
1479 pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len);
1480 }
1481 }
1482
1483 static const struct nla_policy
1484 ip_set_dump_policy[IPSET_ATTR_CMD_MAX + 1] = {
1485 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1486 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1487 .len = IPSET_MAXNAMELEN - 1 },
1488 [IPSET_ATTR_FLAGS] = { .type = NLA_U32 },
1489 };
1490
1491 static int
1492 ip_set_dump_start(struct netlink_callback *cb)
1493 {
1494 struct nlmsghdr *nlh = nlmsg_hdr(cb->skb);
1495 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1496 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1497 struct nlattr *attr = (void *)nlh + min_len;
1498 struct sk_buff *skb = cb->skb;
1499 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1500 u32 dump_type;
1501 int ret;
1502
1503 ret = nla_parse(cda, IPSET_ATTR_CMD_MAX, attr,
1504 nlh->nlmsg_len - min_len,
1505 ip_set_dump_policy, NULL);
1506 if (ret)
1507 goto error;
1508
1509 cb->args[IPSET_CB_PROTO] = nla_get_u8(cda[IPSET_ATTR_PROTOCOL]);
1510 if (cda[IPSET_ATTR_SETNAME]) {
1511 ip_set_id_t index;
1512 struct ip_set *set;
1513
1514 set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]),
1515 &index);
1516 if (!set) {
1517 ret = -ENOENT;
1518 goto error;
1519 }
1520 dump_type = DUMP_ONE;
1521 cb->args[IPSET_CB_INDEX] = index;
1522 } else {
1523 dump_type = DUMP_ALL;
1524 }
1525
1526 if (cda[IPSET_ATTR_FLAGS]) {
1527 u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]);
1528
1529 dump_type |= (f << 16);
1530 }
1531 cb->args[IPSET_CB_NET] = (unsigned long)inst;
1532 cb->args[IPSET_CB_DUMP] = dump_type;
1533
1534 return 0;
1535
1536 error:
1537 /* We have to create and send the error message manually :-( */
1538 if (nlh->nlmsg_flags & NLM_F_ACK) {
1539 netlink_ack(cb->skb, nlh, ret, NULL);
1540 }
1541 return ret;
1542 }
1543
1544 static int
1545 ip_set_dump_do(struct sk_buff *skb, struct netlink_callback *cb)
1546 {
1547 ip_set_id_t index = IPSET_INVALID_ID, max;
1548 struct ip_set *set = NULL;
1549 struct nlmsghdr *nlh = NULL;
1550 unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0;
1551 struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1552 u32 dump_type, dump_flags;
1553 bool is_destroyed;
1554 int ret = 0;
1555
1556 if (!cb->args[IPSET_CB_DUMP])
1557 return -EINVAL;
1558
1559 if (cb->args[IPSET_CB_INDEX] >= inst->ip_set_max)
1560 goto out;
1561
1562 dump_type = DUMP_TYPE(cb->args[IPSET_CB_DUMP]);
1563 dump_flags = DUMP_FLAGS(cb->args[IPSET_CB_DUMP]);
1564 max = dump_type == DUMP_ONE ? cb->args[IPSET_CB_INDEX] + 1
1565 : inst->ip_set_max;
1566 dump_last:
1567 pr_debug("dump type, flag: %u %u index: %ld\n",
1568 dump_type, dump_flags, cb->args[IPSET_CB_INDEX]);
1569 for (; cb->args[IPSET_CB_INDEX] < max; cb->args[IPSET_CB_INDEX]++) {
1570 index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1571 write_lock_bh(&ip_set_ref_lock);
1572 set = ip_set(inst, index);
1573 is_destroyed = inst->is_destroyed;
1574 if (!set || is_destroyed) {
1575 write_unlock_bh(&ip_set_ref_lock);
1576 if (dump_type == DUMP_ONE) {
1577 ret = -ENOENT;
1578 goto out;
1579 }
1580 if (is_destroyed) {
1581 /* All sets are just being destroyed */
1582 ret = 0;
1583 goto out;
1584 }
1585 continue;
1586 }
1587 /* When dumping all sets, we must dump "sorted"
1588 * so that lists (unions of sets) are dumped last.
1589 */
1590 if (dump_type != DUMP_ONE &&
1591 ((dump_type == DUMP_ALL) ==
1592 !!(set->type->features & IPSET_DUMP_LAST))) {
1593 write_unlock_bh(&ip_set_ref_lock);
1594 continue;
1595 }
1596 pr_debug("List set: %s\n", set->name);
1597 if (!cb->args[IPSET_CB_ARG0]) {
1598 /* Start listing: make sure set won't be destroyed */
1599 pr_debug("reference set\n");
1600 set->ref_netlink++;
1601 }
1602 write_unlock_bh(&ip_set_ref_lock);
1603 nlh = start_msg(skb, NETLINK_CB(cb->skb).portid,
1604 cb->nlh->nlmsg_seq, flags,
1605 IPSET_CMD_LIST);
1606 if (!nlh) {
1607 ret = -EMSGSIZE;
1608 goto release_refcount;
1609 }
1610 if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL,
1611 cb->args[IPSET_CB_PROTO]) ||
1612 nla_put_string(skb, IPSET_ATTR_SETNAME, set->name))
1613 goto nla_put_failure;
1614 if (dump_flags & IPSET_FLAG_LIST_SETNAME)
1615 goto next_set;
1616 switch (cb->args[IPSET_CB_ARG0]) {
1617 case 0:
1618 /* Core header data */
1619 if (nla_put_string(skb, IPSET_ATTR_TYPENAME,
1620 set->type->name) ||
1621 nla_put_u8(skb, IPSET_ATTR_FAMILY,
1622 set->family) ||
1623 nla_put_u8(skb, IPSET_ATTR_REVISION,
1624 set->revision))
1625 goto nla_put_failure;
1626 if (cb->args[IPSET_CB_PROTO] > IPSET_PROTOCOL_MIN &&
1627 nla_put_net16(skb, IPSET_ATTR_INDEX, htons(index)))
1628 goto nla_put_failure;
1629 ret = set->variant->head(set, skb);
1630 if (ret < 0)
1631 goto release_refcount;
1632 if (dump_flags & IPSET_FLAG_LIST_HEADER)
1633 goto next_set;
1634 if (set->variant->uref)
1635 set->variant->uref(set, cb, true);
1636 fallthrough;
1637 default:
1638 ret = set->variant->list(set, skb, cb);
1639 if (!cb->args[IPSET_CB_ARG0])
1640 /* Set is done, proceed with next one */
1641 goto next_set;
1642 goto release_refcount;
1643 }
1644 }
1645 /* If we dump all sets, continue with dumping last ones */
1646 if (dump_type == DUMP_ALL) {
1647 dump_type = DUMP_LAST;
1648 cb->args[IPSET_CB_DUMP] = dump_type | (dump_flags << 16);
1649 cb->args[IPSET_CB_INDEX] = 0;
1650 if (set && set->variant->uref)
1651 set->variant->uref(set, cb, false);
1652 goto dump_last;
1653 }
1654 goto out;
1655
1656 nla_put_failure:
1657 ret = -EFAULT;
1658 next_set:
1659 if (dump_type == DUMP_ONE)
1660 cb->args[IPSET_CB_INDEX] = IPSET_INVALID_ID;
1661 else
1662 cb->args[IPSET_CB_INDEX]++;
1663 release_refcount:
1664 /* If there was an error or set is done, release set */
1665 if (ret || !cb->args[IPSET_CB_ARG0]) {
1666 set = ip_set_ref_netlink(inst, index);
1667 if (set->variant->uref)
1668 set->variant->uref(set, cb, false);
1669 pr_debug("release set %s\n", set->name);
1670 __ip_set_put_netlink(set);
1671 cb->args[IPSET_CB_ARG0] = 0;
1672 }
1673 out:
1674 if (nlh) {
1675 nlmsg_end(skb, nlh);
1676 pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len);
1677 dump_attrs(nlh);
1678 }
1679
1680 return ret < 0 ? ret : skb->len;
1681 }
1682
1683 static int ip_set_dump(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1684 const struct nlmsghdr *nlh,
1685 const struct nlattr * const attr[],
1686 struct netlink_ext_ack *extack)
1687 {
1688 if (unlikely(protocol_min_failed(attr)))
1689 return -IPSET_ERR_PROTOCOL;
1690
1691 {
1692 struct netlink_dump_control c = {
1693 .start = ip_set_dump_start,
1694 .dump = ip_set_dump_do,
1695 .done = ip_set_dump_done,
1696 };
1697 return netlink_dump_start(ctnl, skb, nlh, &c);
1698 }
1699 }
1700
1701 /* Add, del and test */
1702
1703 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = {
1704 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1705 [IPSET_ATTR_SETNAME] = { .type = NLA_NUL_STRING,
1706 .len = IPSET_MAXNAMELEN - 1 },
1707 [IPSET_ATTR_LINENO] = { .type = NLA_U32 },
1708 [IPSET_ATTR_DATA] = { .type = NLA_NESTED },
1709 [IPSET_ATTR_ADT] = { .type = NLA_NESTED },
1710 };
1711
1712 static int
1713 call_ad(struct sock *ctnl, struct sk_buff *skb, struct ip_set *set,
1714 struct nlattr *tb[], enum ipset_adt adt,
1715 u32 flags, bool use_lineno)
1716 {
1717 int ret;
1718 u32 lineno = 0;
1719 bool eexist = flags & IPSET_FLAG_EXIST, retried = false;
1720
1721 do {
1722 ip_set_lock(set);
1723 ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried);
1724 ip_set_unlock(set);
1725 retried = true;
1726 } while (ret == -EAGAIN &&
1727 set->variant->resize &&
1728 (ret = set->variant->resize(set, retried)) == 0);
1729
1730 if (!ret || (ret == -IPSET_ERR_EXIST && eexist))
1731 return 0;
1732 if (lineno && use_lineno) {
1733 /* Error in restore/batch mode: send back lineno */
1734 struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb);
1735 struct sk_buff *skb2;
1736 struct nlmsgerr *errmsg;
1737 size_t payload = min(SIZE_MAX,
1738 sizeof(*errmsg) + nlmsg_len(nlh));
1739 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1740 struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1741 struct nlattr *cmdattr;
1742 u32 *errline;
1743
1744 skb2 = nlmsg_new(payload, GFP_KERNEL);
1745 if (!skb2)
1746 return -ENOMEM;
1747 rep = __nlmsg_put(skb2, NETLINK_CB(skb).portid,
1748 nlh->nlmsg_seq, NLMSG_ERROR, payload, 0);
1749 errmsg = nlmsg_data(rep);
1750 errmsg->error = ret;
1751 memcpy(&errmsg->msg, nlh, nlh->nlmsg_len);
1752 cmdattr = (void *)&errmsg->msg + min_len;
1753
1754 ret = nla_parse(cda, IPSET_ATTR_CMD_MAX, cmdattr,
1755 nlh->nlmsg_len - min_len, ip_set_adt_policy,
1756 NULL);
1757
1758 if (ret) {
1759 nlmsg_free(skb2);
1760 return ret;
1761 }
1762 errline = nla_data(cda[IPSET_ATTR_LINENO]);
1763
1764 *errline = lineno;
1765
1766 netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid,
1767 MSG_DONTWAIT);
1768 /* Signal netlink not to send its ACK/errmsg. */
1769 return -EINTR;
1770 }
1771
1772 return ret;
1773 }
1774
1775 static int ip_set_ad(struct net *net, struct sock *ctnl,
1776 struct sk_buff *skb,
1777 enum ipset_adt adt,
1778 const struct nlmsghdr *nlh,
1779 const struct nlattr * const attr[],
1780 struct netlink_ext_ack *extack)
1781 {
1782 struct ip_set_net *inst = ip_set_pernet(net);
1783 struct ip_set *set;
1784 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1785 const struct nlattr *nla;
1786 u32 flags = flag_exist(nlh);
1787 bool use_lineno;
1788 int ret = 0;
1789
1790 if (unlikely(protocol_min_failed(attr) ||
1791 !attr[IPSET_ATTR_SETNAME] ||
1792 !((attr[IPSET_ATTR_DATA] != NULL) ^
1793 (attr[IPSET_ATTR_ADT] != NULL)) ||
1794 (attr[IPSET_ATTR_DATA] &&
1795 !flag_nested(attr[IPSET_ATTR_DATA])) ||
1796 (attr[IPSET_ATTR_ADT] &&
1797 (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1798 !attr[IPSET_ATTR_LINENO]))))
1799 return -IPSET_ERR_PROTOCOL;
1800
1801 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1802 if (!set)
1803 return -ENOENT;
1804
1805 use_lineno = !!attr[IPSET_ATTR_LINENO];
1806 if (attr[IPSET_ATTR_DATA]) {
1807 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1808 attr[IPSET_ATTR_DATA],
1809 set->type->adt_policy, NULL))
1810 return -IPSET_ERR_PROTOCOL;
1811 ret = call_ad(ctnl, skb, set, tb, adt, flags,
1812 use_lineno);
1813 } else {
1814 int nla_rem;
1815
1816 nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1817 if (nla_type(nla) != IPSET_ATTR_DATA ||
1818 !flag_nested(nla) ||
1819 nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1820 set->type->adt_policy, NULL))
1821 return -IPSET_ERR_PROTOCOL;
1822 ret = call_ad(ctnl, skb, set, tb, adt,
1823 flags, use_lineno);
1824 if (ret < 0)
1825 return ret;
1826 }
1827 }
1828 return ret;
1829 }
1830
1831 static int ip_set_uadd(struct net *net, struct sock *ctnl,
1832 struct sk_buff *skb, const struct nlmsghdr *nlh,
1833 const struct nlattr * const attr[],
1834 struct netlink_ext_ack *extack)
1835 {
1836 return ip_set_ad(net, ctnl, skb,
1837 IPSET_ADD, nlh, attr, extack);
1838 }
1839
1840 static int ip_set_udel(struct net *net, struct sock *ctnl,
1841 struct sk_buff *skb, const struct nlmsghdr *nlh,
1842 const struct nlattr * const attr[],
1843 struct netlink_ext_ack *extack)
1844 {
1845 return ip_set_ad(net, ctnl, skb,
1846 IPSET_DEL, nlh, attr, extack);
1847 }
1848
1849 static int ip_set_utest(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1850 const struct nlmsghdr *nlh,
1851 const struct nlattr * const attr[],
1852 struct netlink_ext_ack *extack)
1853 {
1854 struct ip_set_net *inst = ip_set_pernet(net);
1855 struct ip_set *set;
1856 struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1857 int ret = 0;
1858 u32 lineno;
1859
1860 if (unlikely(protocol_min_failed(attr) ||
1861 !attr[IPSET_ATTR_SETNAME] ||
1862 !attr[IPSET_ATTR_DATA] ||
1863 !flag_nested(attr[IPSET_ATTR_DATA])))
1864 return -IPSET_ERR_PROTOCOL;
1865
1866 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1867 if (!set)
1868 return -ENOENT;
1869
1870 if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA],
1871 set->type->adt_policy, NULL))
1872 return -IPSET_ERR_PROTOCOL;
1873
1874 rcu_read_lock_bh();
1875 ret = set->variant->uadt(set, tb, IPSET_TEST, &lineno, 0, 0);
1876 rcu_read_unlock_bh();
1877 /* Userspace can't trigger element to be re-added */
1878 if (ret == -EAGAIN)
1879 ret = 1;
1880
1881 return ret > 0 ? 0 : -IPSET_ERR_EXIST;
1882 }
1883
1884 /* Get headed data of a set */
1885
1886 static int ip_set_header(struct net *net, struct sock *ctnl,
1887 struct sk_buff *skb, const struct nlmsghdr *nlh,
1888 const struct nlattr * const attr[],
1889 struct netlink_ext_ack *extack)
1890 {
1891 struct ip_set_net *inst = ip_set_pernet(net);
1892 const struct ip_set *set;
1893 struct sk_buff *skb2;
1894 struct nlmsghdr *nlh2;
1895 int ret = 0;
1896
1897 if (unlikely(protocol_min_failed(attr) ||
1898 !attr[IPSET_ATTR_SETNAME]))
1899 return -IPSET_ERR_PROTOCOL;
1900
1901 set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1902 if (!set)
1903 return -ENOENT;
1904
1905 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1906 if (!skb2)
1907 return -ENOMEM;
1908
1909 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1910 IPSET_CMD_HEADER);
1911 if (!nlh2)
1912 goto nlmsg_failure;
1913 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
1914 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) ||
1915 nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) ||
1916 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
1917 nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision))
1918 goto nla_put_failure;
1919 nlmsg_end(skb2, nlh2);
1920
1921 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1922 if (ret < 0)
1923 return ret;
1924
1925 return 0;
1926
1927 nla_put_failure:
1928 nlmsg_cancel(skb2, nlh2);
1929 nlmsg_failure:
1930 kfree_skb(skb2);
1931 return -EMSGSIZE;
1932 }
1933
1934 /* Get type data */
1935
1936 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = {
1937 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
1938 [IPSET_ATTR_TYPENAME] = { .type = NLA_NUL_STRING,
1939 .len = IPSET_MAXNAMELEN - 1 },
1940 [IPSET_ATTR_FAMILY] = { .type = NLA_U8 },
1941 };
1942
1943 static int ip_set_type(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1944 const struct nlmsghdr *nlh,
1945 const struct nlattr * const attr[],
1946 struct netlink_ext_ack *extack)
1947 {
1948 struct sk_buff *skb2;
1949 struct nlmsghdr *nlh2;
1950 u8 family, min, max;
1951 const char *typename;
1952 int ret = 0;
1953
1954 if (unlikely(protocol_min_failed(attr) ||
1955 !attr[IPSET_ATTR_TYPENAME] ||
1956 !attr[IPSET_ATTR_FAMILY]))
1957 return -IPSET_ERR_PROTOCOL;
1958
1959 family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1960 typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1961 ret = find_set_type_minmax(typename, family, &min, &max);
1962 if (ret)
1963 return ret;
1964
1965 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1966 if (!skb2)
1967 return -ENOMEM;
1968
1969 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
1970 IPSET_CMD_TYPE);
1971 if (!nlh2)
1972 goto nlmsg_failure;
1973 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
1974 nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) ||
1975 nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) ||
1976 nla_put_u8(skb2, IPSET_ATTR_REVISION, max) ||
1977 nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min))
1978 goto nla_put_failure;
1979 nlmsg_end(skb2, nlh2);
1980
1981 pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len);
1982 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1983 if (ret < 0)
1984 return ret;
1985
1986 return 0;
1987
1988 nla_put_failure:
1989 nlmsg_cancel(skb2, nlh2);
1990 nlmsg_failure:
1991 kfree_skb(skb2);
1992 return -EMSGSIZE;
1993 }
1994
1995 /* Get protocol version */
1996
1997 static const struct nla_policy
1998 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = {
1999 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
2000 };
2001
2002 static int ip_set_protocol(struct net *net, struct sock *ctnl,
2003 struct sk_buff *skb, const struct nlmsghdr *nlh,
2004 const struct nlattr * const attr[],
2005 struct netlink_ext_ack *extack)
2006 {
2007 struct sk_buff *skb2;
2008 struct nlmsghdr *nlh2;
2009 int ret = 0;
2010
2011 if (unlikely(!attr[IPSET_ATTR_PROTOCOL]))
2012 return -IPSET_ERR_PROTOCOL;
2013
2014 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
2015 if (!skb2)
2016 return -ENOMEM;
2017
2018 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
2019 IPSET_CMD_PROTOCOL);
2020 if (!nlh2)
2021 goto nlmsg_failure;
2022 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL))
2023 goto nla_put_failure;
2024 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL_MIN, IPSET_PROTOCOL_MIN))
2025 goto nla_put_failure;
2026 nlmsg_end(skb2, nlh2);
2027
2028 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
2029 if (ret < 0)
2030 return ret;
2031
2032 return 0;
2033
2034 nla_put_failure:
2035 nlmsg_cancel(skb2, nlh2);
2036 nlmsg_failure:
2037 kfree_skb(skb2);
2038 return -EMSGSIZE;
2039 }
2040
2041 /* Get set by name or index, from userspace */
2042
2043 static int ip_set_byname(struct net *net, struct sock *ctnl,
2044 struct sk_buff *skb, const struct nlmsghdr *nlh,
2045 const struct nlattr * const attr[],
2046 struct netlink_ext_ack *extack)
2047 {
2048 struct ip_set_net *inst = ip_set_pernet(net);
2049 struct sk_buff *skb2;
2050 struct nlmsghdr *nlh2;
2051 ip_set_id_t id = IPSET_INVALID_ID;
2052 const struct ip_set *set;
2053 int ret = 0;
2054
2055 if (unlikely(protocol_failed(attr) ||
2056 !attr[IPSET_ATTR_SETNAME]))
2057 return -IPSET_ERR_PROTOCOL;
2058
2059 set = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]), &id);
2060 if (id == IPSET_INVALID_ID)
2061 return -ENOENT;
2062
2063 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
2064 if (!skb2)
2065 return -ENOMEM;
2066
2067 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
2068 IPSET_CMD_GET_BYNAME);
2069 if (!nlh2)
2070 goto nlmsg_failure;
2071 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
2072 nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
2073 nla_put_net16(skb2, IPSET_ATTR_INDEX, htons(id)))
2074 goto nla_put_failure;
2075 nlmsg_end(skb2, nlh2);
2076
2077 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
2078 if (ret < 0)
2079 return ret;
2080
2081 return 0;
2082
2083 nla_put_failure:
2084 nlmsg_cancel(skb2, nlh2);
2085 nlmsg_failure:
2086 kfree_skb(skb2);
2087 return -EMSGSIZE;
2088 }
2089
2090 static const struct nla_policy ip_set_index_policy[IPSET_ATTR_CMD_MAX + 1] = {
2091 [IPSET_ATTR_PROTOCOL] = { .type = NLA_U8 },
2092 [IPSET_ATTR_INDEX] = { .type = NLA_U16 },
2093 };
2094
2095 static int ip_set_byindex(struct net *net, struct sock *ctnl,
2096 struct sk_buff *skb, const struct nlmsghdr *nlh,
2097 const struct nlattr * const attr[],
2098 struct netlink_ext_ack *extack)
2099 {
2100 struct ip_set_net *inst = ip_set_pernet(net);
2101 struct sk_buff *skb2;
2102 struct nlmsghdr *nlh2;
2103 ip_set_id_t id = IPSET_INVALID_ID;
2104 const struct ip_set *set;
2105 int ret = 0;
2106
2107 if (unlikely(protocol_failed(attr) ||
2108 !attr[IPSET_ATTR_INDEX]))
2109 return -IPSET_ERR_PROTOCOL;
2110
2111 id = ip_set_get_h16(attr[IPSET_ATTR_INDEX]);
2112 if (id >= inst->ip_set_max)
2113 return -ENOENT;
2114 set = ip_set(inst, id);
2115 if (set == NULL)
2116 return -ENOENT;
2117
2118 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
2119 if (!skb2)
2120 return -ENOMEM;
2121
2122 nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
2123 IPSET_CMD_GET_BYINDEX);
2124 if (!nlh2)
2125 goto nlmsg_failure;
2126 if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
2127 nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name))
2128 goto nla_put_failure;
2129 nlmsg_end(skb2, nlh2);
2130
2131 ret = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
2132 if (ret < 0)
2133 return ret;
2134
2135 return 0;
2136
2137 nla_put_failure:
2138 nlmsg_cancel(skb2, nlh2);
2139 nlmsg_failure:
2140 kfree_skb(skb2);
2141 return -EMSGSIZE;
2142 }
2143
2144 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = {
2145 [IPSET_CMD_NONE] = {
2146 .call = ip_set_none,
2147 .attr_count = IPSET_ATTR_CMD_MAX,
2148 },
2149 [IPSET_CMD_CREATE] = {
2150 .call = ip_set_create,
2151 .attr_count = IPSET_ATTR_CMD_MAX,
2152 .policy = ip_set_create_policy,
2153 },
2154 [IPSET_CMD_DESTROY] = {
2155 .call = ip_set_destroy,
2156 .attr_count = IPSET_ATTR_CMD_MAX,
2157 .policy = ip_set_setname_policy,
2158 },
2159 [IPSET_CMD_FLUSH] = {
2160 .call = ip_set_flush,
2161 .attr_count = IPSET_ATTR_CMD_MAX,
2162 .policy = ip_set_setname_policy,
2163 },
2164 [IPSET_CMD_RENAME] = {
2165 .call = ip_set_rename,
2166 .attr_count = IPSET_ATTR_CMD_MAX,
2167 .policy = ip_set_setname2_policy,
2168 },
2169 [IPSET_CMD_SWAP] = {
2170 .call = ip_set_swap,
2171 .attr_count = IPSET_ATTR_CMD_MAX,
2172 .policy = ip_set_setname2_policy,
2173 },
2174 [IPSET_CMD_LIST] = {
2175 .call = ip_set_dump,
2176 .attr_count = IPSET_ATTR_CMD_MAX,
2177 .policy = ip_set_dump_policy,
2178 },
2179 [IPSET_CMD_SAVE] = {
2180 .call = ip_set_dump,
2181 .attr_count = IPSET_ATTR_CMD_MAX,
2182 .policy = ip_set_setname_policy,
2183 },
2184 [IPSET_CMD_ADD] = {
2185 .call = ip_set_uadd,
2186 .attr_count = IPSET_ATTR_CMD_MAX,
2187 .policy = ip_set_adt_policy,
2188 },
2189 [IPSET_CMD_DEL] = {
2190 .call = ip_set_udel,
2191 .attr_count = IPSET_ATTR_CMD_MAX,
2192 .policy = ip_set_adt_policy,
2193 },
2194 [IPSET_CMD_TEST] = {
2195 .call = ip_set_utest,
2196 .attr_count = IPSET_ATTR_CMD_MAX,
2197 .policy = ip_set_adt_policy,
2198 },
2199 [IPSET_CMD_HEADER] = {
2200 .call = ip_set_header,
2201 .attr_count = IPSET_ATTR_CMD_MAX,
2202 .policy = ip_set_setname_policy,
2203 },
2204 [IPSET_CMD_TYPE] = {
2205 .call = ip_set_type,
2206 .attr_count = IPSET_ATTR_CMD_MAX,
2207 .policy = ip_set_type_policy,
2208 },
2209 [IPSET_CMD_PROTOCOL] = {
2210 .call = ip_set_protocol,
2211 .attr_count = IPSET_ATTR_CMD_MAX,
2212 .policy = ip_set_protocol_policy,
2213 },
2214 [IPSET_CMD_GET_BYNAME] = {
2215 .call = ip_set_byname,
2216 .attr_count = IPSET_ATTR_CMD_MAX,
2217 .policy = ip_set_setname_policy,
2218 },
2219 [IPSET_CMD_GET_BYINDEX] = {
2220 .call = ip_set_byindex,
2221 .attr_count = IPSET_ATTR_CMD_MAX,
2222 .policy = ip_set_index_policy,
2223 },
2224 };
2225
2226 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = {
2227 .name = "ip_set",
2228 .subsys_id = NFNL_SUBSYS_IPSET,
2229 .cb_count = IPSET_MSG_MAX,
2230 .cb = ip_set_netlink_subsys_cb,
2231 };
2232
2233 /* Interface to iptables/ip6tables */
2234
2235 static int
2236 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len)
2237 {
2238 unsigned int *op;
2239 void *data;
2240 int copylen = *len, ret = 0;
2241 struct net *net = sock_net(sk);
2242 struct ip_set_net *inst = ip_set_pernet(net);
2243
2244 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2245 return -EPERM;
2246 if (optval != SO_IP_SET)
2247 return -EBADF;
2248 if (*len < sizeof(unsigned int))
2249 return -EINVAL;
2250
2251 data = vmalloc(*len);
2252 if (!data)
2253 return -ENOMEM;
2254 if (copy_from_user(data, user, *len) != 0) {
2255 ret = -EFAULT;
2256 goto done;
2257 }
2258 op = data;
2259
2260 if (*op < IP_SET_OP_VERSION) {
2261 /* Check the version at the beginning of operations */
2262 struct ip_set_req_version *req_version = data;
2263
2264 if (*len < sizeof(struct ip_set_req_version)) {
2265 ret = -EINVAL;
2266 goto done;
2267 }
2268
2269 if (req_version->version < IPSET_PROTOCOL_MIN) {
2270 ret = -EPROTO;
2271 goto done;
2272 }
2273 }
2274
2275 switch (*op) {
2276 case IP_SET_OP_VERSION: {
2277 struct ip_set_req_version *req_version = data;
2278
2279 if (*len != sizeof(struct ip_set_req_version)) {
2280 ret = -EINVAL;
2281 goto done;
2282 }
2283
2284 req_version->version = IPSET_PROTOCOL;
2285 if (copy_to_user(user, req_version,
2286 sizeof(struct ip_set_req_version)))
2287 ret = -EFAULT;
2288 goto done;
2289 }
2290 case IP_SET_OP_GET_BYNAME: {
2291 struct ip_set_req_get_set *req_get = data;
2292 ip_set_id_t id;
2293
2294 if (*len != sizeof(struct ip_set_req_get_set)) {
2295 ret = -EINVAL;
2296 goto done;
2297 }
2298 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
2299 nfnl_lock(NFNL_SUBSYS_IPSET);
2300 find_set_and_id(inst, req_get->set.name, &id);
2301 req_get->set.index = id;
2302 nfnl_unlock(NFNL_SUBSYS_IPSET);
2303 goto copy;
2304 }
2305 case IP_SET_OP_GET_FNAME: {
2306 struct ip_set_req_get_set_family *req_get = data;
2307 ip_set_id_t id;
2308
2309 if (*len != sizeof(struct ip_set_req_get_set_family)) {
2310 ret = -EINVAL;
2311 goto done;
2312 }
2313 req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
2314 nfnl_lock(NFNL_SUBSYS_IPSET);
2315 find_set_and_id(inst, req_get->set.name, &id);
2316 req_get->set.index = id;
2317 if (id != IPSET_INVALID_ID)
2318 req_get->family = ip_set(inst, id)->family;
2319 nfnl_unlock(NFNL_SUBSYS_IPSET);
2320 goto copy;
2321 }
2322 case IP_SET_OP_GET_BYINDEX: {
2323 struct ip_set_req_get_set *req_get = data;
2324 struct ip_set *set;
2325
2326 if (*len != sizeof(struct ip_set_req_get_set) ||
2327 req_get->set.index >= inst->ip_set_max) {
2328 ret = -EINVAL;
2329 goto done;
2330 }
2331 nfnl_lock(NFNL_SUBSYS_IPSET);
2332 set = ip_set(inst, req_get->set.index);
2333 ret = strscpy(req_get->set.name, set ? set->name : "",
2334 IPSET_MAXNAMELEN);
2335 nfnl_unlock(NFNL_SUBSYS_IPSET);
2336 if (ret < 0)
2337 goto done;
2338 goto copy;
2339 }
2340 default:
2341 ret = -EBADMSG;
2342 goto done;
2343 } /* end of switch(op) */
2344
2345 copy:
2346 if (copy_to_user(user, data, copylen))
2347 ret = -EFAULT;
2348
2349 done:
2350 vfree(data);
2351 if (ret > 0)
2352 ret = 0;
2353 return ret;
2354 }
2355
2356 static struct nf_sockopt_ops so_set __read_mostly = {
2357 .pf = PF_INET,
2358 .get_optmin = SO_IP_SET,
2359 .get_optmax = SO_IP_SET + 1,
2360 .get = ip_set_sockfn_get,
2361 .owner = THIS_MODULE,
2362 };
2363
2364 static int __net_init
2365 ip_set_net_init(struct net *net)
2366 {
2367 struct ip_set_net *inst = ip_set_pernet(net);
2368 struct ip_set **list;
2369
2370 inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX;
2371 if (inst->ip_set_max >= IPSET_INVALID_ID)
2372 inst->ip_set_max = IPSET_INVALID_ID - 1;
2373
2374 list = kvcalloc(inst->ip_set_max, sizeof(struct ip_set *), GFP_KERNEL);
2375 if (!list)
2376 return -ENOMEM;
2377 inst->is_deleted = false;
2378 inst->is_destroyed = false;
2379 rcu_assign_pointer(inst->ip_set_list, list);
2380 return 0;
2381 }
2382
2383 static void __net_exit
2384 ip_set_net_exit(struct net *net)
2385 {
2386 struct ip_set_net *inst = ip_set_pernet(net);
2387
2388 struct ip_set *set = NULL;
2389 ip_set_id_t i;
2390
2391 inst->is_deleted = true; /* flag for ip_set_nfnl_put */
2392
2393 nfnl_lock(NFNL_SUBSYS_IPSET);
2394 for (i = 0; i < inst->ip_set_max; i++) {
2395 set = ip_set(inst, i);
2396 if (set) {
2397 ip_set(inst, i) = NULL;
2398 ip_set_destroy_set(set);
2399 }
2400 }
2401 nfnl_unlock(NFNL_SUBSYS_IPSET);
2402 kvfree(rcu_dereference_protected(inst->ip_set_list, 1));
2403 }
2404
2405 static struct pernet_operations ip_set_net_ops = {
2406 .init = ip_set_net_init,
2407 .exit = ip_set_net_exit,
2408 .id = &ip_set_net_id,
2409 .size = sizeof(struct ip_set_net),
2410 };
2411
2412 static int __init
2413 ip_set_init(void)
2414 {
2415 int ret = register_pernet_subsys(&ip_set_net_ops);
2416
2417 if (ret) {
2418 pr_err("ip_set: cannot register pernet_subsys.\n");
2419 return ret;
2420 }
2421
2422 ret = nfnetlink_subsys_register(&ip_set_netlink_subsys);
2423 if (ret != 0) {
2424 pr_err("ip_set: cannot register with nfnetlink.\n");
2425 unregister_pernet_subsys(&ip_set_net_ops);
2426 return ret;
2427 }
2428
2429 ret = nf_register_sockopt(&so_set);
2430 if (ret != 0) {
2431 pr_err("SO_SET registry failed: %d\n", ret);
2432 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2433 unregister_pernet_subsys(&ip_set_net_ops);
2434 return ret;
2435 }
2436
2437 return 0;
2438 }
2439
2440 static void __exit
2441 ip_set_fini(void)
2442 {
2443 nf_unregister_sockopt(&so_set);
2444 nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2445
2446 unregister_pernet_subsys(&ip_set_net_ops);
2447 pr_debug("these are the famous last words\n");
2448 }
2449
2450 module_init(ip_set_init);
2451 module_exit(ip_set_fini);
2452
2453 MODULE_DESCRIPTION("ip_set: protocol " __stringify(IPSET_PROTOCOL));
Ubuntu Jammy Linux kernel mirror