1 // SPDX-License-Identifier: GPL-2.0-only
2 /* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
3  *                         Patrick Schaaf <bof@bof.de>
4  * Copyright (C) 2003-2013 Jozsef Kadlecsik <kadlec@netfilter.org>
5  */
6 
7 /* Kernel module for IP set management */
8 
9 #include <linux/init.h>
10 #include <linux/module.h>
11 #include <linux/moduleparam.h>
12 #include <linux/ip.h>
13 #include <linux/skbuff.h>
14 #include <linux/spinlock.h>
15 #include <linux/rculist.h>
16 #include <net/netlink.h>
17 #include <net/net_namespace.h>
18 #include <net/netns/generic.h>
19 
20 #include <linux/netfilter.h>
21 #include <linux/netfilter/x_tables.h>
22 #include <linux/netfilter/nfnetlink.h>
23 #include <linux/netfilter/ipset/ip_set.h>
24 
25 static LIST_HEAD(ip_set_type_list);		/* all registered set types */
26 static DEFINE_MUTEX(ip_set_type_mutex);		/* protects ip_set_type_list */
27 static DEFINE_RWLOCK(ip_set_ref_lock);		/* protects the set refs */
28 
29 struct ip_set_net {
30 	struct ip_set * __rcu *ip_set_list;	/* all individual sets */
31 	ip_set_id_t	ip_set_max;	/* max number of sets */
32 	bool		is_deleted;	/* deleted by ip_set_net_exit */
33 	bool		is_destroyed;	/* all sets are destroyed */
34 };
35 
36 static unsigned int ip_set_net_id __read_mostly;
37 
ip_set_pernet(struct net * net)38 static 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
ip_set_type_lock(void)71 ip_set_type_lock(void)
72 {
73 	mutex_lock(&ip_set_type_mutex);
74 }
75 
76 static void
ip_set_type_unlock(void)77 ip_set_type_unlock(void)
78 {
79 	mutex_unlock(&ip_set_type_mutex);
80 }
81 
82 /* Register and deregister settype */
83 
84 static struct ip_set_type *
find_set_type(const char * name,u8 family,u8 revision)85 find_set_type(const char *name, u8 family, u8 revision)
86 {
87 	struct ip_set_type *type;
88 
89 	list_for_each_entry_rcu(type, &ip_set_type_list, list,
90 				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
load_settype(const char * name)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
__find_set_type_get(const char * name,u8 family,u8 revision,struct ip_set_type ** found,bool retry)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
__find_set_type_minmax(const char * name,u8 family,u8 * min,u8 * max,bool retry)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
ip_set_type_register(struct ip_set_type * type)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
ip_set_type_unregister(struct ip_set_type * type)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 *
ip_set_alloc(size_t size)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
ip_set_free(void * members)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
flag_nested(const struct nlattr * nla)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
ip_set_get_ipaddr4(struct nlattr * nla,__be32 * ipaddr)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
ip_set_get_ipaddr6(struct nlattr * nla,union nf_inet_addr * ipaddr)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
ip_set_timeout_get(const unsigned long * timeout)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 *
ip_set_comment_uget(struct nlattr * tb)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
ip_set_init_comment(struct ip_set * set,struct ip_set_comment * comment,const struct ip_set_ext * ext)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 	strscpy(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
ip_set_put_comment(struct sk_buff * skb,const struct ip_set_comment * comment)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
ip_set_comment_free(struct ip_set * set,void * ptr)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
add_extension(enum ip_set_ext_id id,u32 flags,struct nlattr * tb[])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
ip_set_elem_len(struct ip_set * set,struct nlattr * tb[],size_t len,size_t align)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
ip_set_get_extensions(struct ip_set * set,struct nlattr * tb[],struct ip_set_ext * ext)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
ip_set_get_bytes(const struct ip_set_counter * counter)516 ip_set_get_bytes(const struct ip_set_counter *counter)
517 {
518 	return (u64)atomic64_read(&(counter)->bytes);
519 }
520 
521 static u64
ip_set_get_packets(const struct ip_set_counter * counter)522 ip_set_get_packets(const struct ip_set_counter *counter)
523 {
524 	return (u64)atomic64_read(&(counter)->packets);
525 }
526 
527 static bool
ip_set_put_counter(struct sk_buff * skb,const struct ip_set_counter * counter)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
ip_set_put_skbinfo(struct sk_buff * skb,const struct ip_set_skbinfo * skbinfo)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
ip_set_put_extensions(struct sk_buff * skb,const struct ip_set * set,const void * e,bool active)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
ip_set_match_counter(u64 counter,u64 match,u8 op)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
ip_set_add_bytes(u64 bytes,struct ip_set_counter * counter)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
ip_set_add_packets(u64 packets,struct ip_set_counter * counter)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
ip_set_update_counter(struct ip_set_counter * counter,const struct ip_set_ext * ext,u32 flags)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
ip_set_get_skbinfo(struct ip_set_skbinfo * skbinfo,const struct ip_set_ext * ext,struct ip_set_ext * mext,u32 flags)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
ip_set_match_extensions(struct ip_set * set,const struct ip_set_ext * ext,struct ip_set_ext * mext,u32 flags,void * data)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
__ip_set_get(struct ip_set * set)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
__ip_set_put(struct ip_set * set)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
__ip_set_put_netlink(struct ip_set * set)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 *
ip_set_rcu_get(struct net * net,ip_set_id_t index)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
ip_set_lock(struct ip_set * set)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
ip_set_unlock(struct ip_set * set)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
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)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
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)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
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)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
ip_set_get_byname(struct net * net,const char * name,struct ip_set ** set)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
__ip_set_put_byindex(struct ip_set_net * inst,ip_set_id_t index)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
ip_set_put_byindex(struct net * net,ip_set_id_t index)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
ip_set_name_byindex(struct net * net,ip_set_id_t index,char * name)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
ip_set_nfnl_get_byindex(struct net * net,ip_set_id_t index)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
ip_set_nfnl_put(struct net * net,ip_set_id_t index)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 
protocol(const struct nlattr * const tb[])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
protocol_failed(const struct nlattr * const tb[])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
protocol_min_failed(const struct nlattr * const tb[])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
flag_exist(const struct nlmsghdr * nlh)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 *
start_msg(struct sk_buff * skb,u32 portid,u32 seq,unsigned int flags,enum ipset_cmd cmd)963 start_msg(struct sk_buff *skb, u32 portid, u32 seq, unsigned int flags,
964 	  enum ipset_cmd cmd)
965 {
966 	return nfnl_msg_put(skb, portid, seq,
967 			    nfnl_msg_type(NFNL_SUBSYS_IPSET, cmd), flags,
968 			    NFPROTO_IPV4, NFNETLINK_V0, 0);
969 }
970 
971 /* Create a set */
972 
973 static const struct nla_policy ip_set_create_policy[IPSET_ATTR_CMD_MAX + 1] = {
974 	[IPSET_ATTR_PROTOCOL]	= { .type = NLA_U8 },
975 	[IPSET_ATTR_SETNAME]	= { .type = NLA_NUL_STRING,
976 				    .len = IPSET_MAXNAMELEN - 1 },
977 	[IPSET_ATTR_TYPENAME]	= { .type = NLA_NUL_STRING,
978 				    .len = IPSET_MAXNAMELEN - 1},
979 	[IPSET_ATTR_REVISION]	= { .type = NLA_U8 },
980 	[IPSET_ATTR_FAMILY]	= { .type = NLA_U8 },
981 	[IPSET_ATTR_DATA]	= { .type = NLA_NESTED },
982 };
983 
984 static struct ip_set *
find_set_and_id(struct ip_set_net * inst,const char * name,ip_set_id_t * id)985 find_set_and_id(struct ip_set_net *inst, const char *name, ip_set_id_t *id)
986 {
987 	struct ip_set *set = NULL;
988 	ip_set_id_t i;
989 
990 	*id = IPSET_INVALID_ID;
991 	for (i = 0; i < inst->ip_set_max; i++) {
992 		set = ip_set(inst, i);
993 		if (set && STRNCMP(set->name, name)) {
994 			*id = i;
995 			break;
996 		}
997 	}
998 	return (*id == IPSET_INVALID_ID ? NULL : set);
999 }
1000 
1001 static inline struct ip_set *
find_set(struct ip_set_net * inst,const char * name)1002 find_set(struct ip_set_net *inst, const char *name)
1003 {
1004 	ip_set_id_t id;
1005 
1006 	return find_set_and_id(inst, name, &id);
1007 }
1008 
1009 static int
find_free_id(struct ip_set_net * inst,const char * name,ip_set_id_t * index,struct ip_set ** set)1010 find_free_id(struct ip_set_net *inst, const char *name, ip_set_id_t *index,
1011 	     struct ip_set **set)
1012 {
1013 	struct ip_set *s;
1014 	ip_set_id_t i;
1015 
1016 	*index = IPSET_INVALID_ID;
1017 	for (i = 0;  i < inst->ip_set_max; i++) {
1018 		s = ip_set(inst, i);
1019 		if (!s) {
1020 			if (*index == IPSET_INVALID_ID)
1021 				*index = i;
1022 		} else if (STRNCMP(name, s->name)) {
1023 			/* Name clash */
1024 			*set = s;
1025 			return -EEXIST;
1026 		}
1027 	}
1028 	if (*index == IPSET_INVALID_ID)
1029 		/* No free slot remained */
1030 		return -IPSET_ERR_MAX_SETS;
1031 	return 0;
1032 }
1033 
ip_set_none(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const attr[])1034 static int ip_set_none(struct sk_buff *skb, const struct nfnl_info *info,
1035 		       const struct nlattr * const attr[])
1036 {
1037 	return -EOPNOTSUPP;
1038 }
1039 
ip_set_create(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const attr[])1040 static int ip_set_create(struct sk_buff *skb, const struct nfnl_info *info,
1041 			 const struct nlattr * const attr[])
1042 {
1043 	struct ip_set_net *inst = ip_set_pernet(info->net);
1044 	struct ip_set *set, *clash = NULL;
1045 	ip_set_id_t index = IPSET_INVALID_ID;
1046 	struct nlattr *tb[IPSET_ATTR_CREATE_MAX + 1] = {};
1047 	const char *name, *typename;
1048 	u8 family, revision;
1049 	u32 flags = flag_exist(info->nlh);
1050 	int ret = 0;
1051 
1052 	if (unlikely(protocol_min_failed(attr) ||
1053 		     !attr[IPSET_ATTR_SETNAME] ||
1054 		     !attr[IPSET_ATTR_TYPENAME] ||
1055 		     !attr[IPSET_ATTR_REVISION] ||
1056 		     !attr[IPSET_ATTR_FAMILY] ||
1057 		     (attr[IPSET_ATTR_DATA] &&
1058 		      !flag_nested(attr[IPSET_ATTR_DATA]))))
1059 		return -IPSET_ERR_PROTOCOL;
1060 
1061 	name = nla_data(attr[IPSET_ATTR_SETNAME]);
1062 	typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1063 	family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1064 	revision = nla_get_u8(attr[IPSET_ATTR_REVISION]);
1065 	pr_debug("setname: %s, typename: %s, family: %s, revision: %u\n",
1066 		 name, typename, family_name(family), revision);
1067 
1068 	/* First, and without any locks, allocate and initialize
1069 	 * a normal base set structure.
1070 	 */
1071 	set = kzalloc(sizeof(*set), GFP_KERNEL);
1072 	if (!set)
1073 		return -ENOMEM;
1074 	spin_lock_init(&set->lock);
1075 	strscpy(set->name, name, IPSET_MAXNAMELEN);
1076 	set->family = family;
1077 	set->revision = revision;
1078 
1079 	/* Next, check that we know the type, and take
1080 	 * a reference on the type, to make sure it stays available
1081 	 * while constructing our new set.
1082 	 *
1083 	 * After referencing the type, we try to create the type
1084 	 * specific part of the set without holding any locks.
1085 	 */
1086 	ret = find_set_type_get(typename, family, revision, &set->type);
1087 	if (ret)
1088 		goto out;
1089 
1090 	/* Without holding any locks, create private part. */
1091 	if (attr[IPSET_ATTR_DATA] &&
1092 	    nla_parse_nested(tb, IPSET_ATTR_CREATE_MAX, attr[IPSET_ATTR_DATA],
1093 			     set->type->create_policy, NULL)) {
1094 		ret = -IPSET_ERR_PROTOCOL;
1095 		goto put_out;
1096 	}
1097 	/* Set create flags depending on the type revision */
1098 	set->flags |= set->type->create_flags[revision];
1099 
1100 	ret = set->type->create(info->net, set, tb, flags);
1101 	if (ret != 0)
1102 		goto put_out;
1103 
1104 	/* BTW, ret==0 here. */
1105 
1106 	/* Here, we have a valid, constructed set and we are protected
1107 	 * by the nfnl mutex. Find the first free index in ip_set_list
1108 	 * and check clashing.
1109 	 */
1110 	ret = find_free_id(inst, set->name, &index, &clash);
1111 	if (ret == -EEXIST) {
1112 		/* If this is the same set and requested, ignore error */
1113 		if ((flags & IPSET_FLAG_EXIST) &&
1114 		    STRNCMP(set->type->name, clash->type->name) &&
1115 		    set->type->family == clash->type->family &&
1116 		    set->type->revision_min == clash->type->revision_min &&
1117 		    set->type->revision_max == clash->type->revision_max &&
1118 		    set->variant->same_set(set, clash))
1119 			ret = 0;
1120 		goto cleanup;
1121 	} else if (ret == -IPSET_ERR_MAX_SETS) {
1122 		struct ip_set **list, **tmp;
1123 		ip_set_id_t i = inst->ip_set_max + IP_SET_INC;
1124 
1125 		if (i < inst->ip_set_max || i == IPSET_INVALID_ID)
1126 			/* Wraparound */
1127 			goto cleanup;
1128 
1129 		list = kvcalloc(i, sizeof(struct ip_set *), GFP_KERNEL);
1130 		if (!list)
1131 			goto cleanup;
1132 		/* nfnl mutex is held, both lists are valid */
1133 		tmp = ip_set_dereference(inst->ip_set_list);
1134 		memcpy(list, tmp, sizeof(struct ip_set *) * inst->ip_set_max);
1135 		rcu_assign_pointer(inst->ip_set_list, list);
1136 		/* Make sure all current packets have passed through */
1137 		synchronize_net();
1138 		/* Use new list */
1139 		index = inst->ip_set_max;
1140 		inst->ip_set_max = i;
1141 		kvfree(tmp);
1142 		ret = 0;
1143 	} else if (ret) {
1144 		goto cleanup;
1145 	}
1146 
1147 	/* Finally! Add our shiny new set to the list, and be done. */
1148 	pr_debug("create: '%s' created with index %u!\n", set->name, index);
1149 	ip_set(inst, index) = set;
1150 
1151 	return ret;
1152 
1153 cleanup:
1154 	set->variant->destroy(set);
1155 put_out:
1156 	module_put(set->type->me);
1157 out:
1158 	kfree(set);
1159 	return ret;
1160 }
1161 
1162 /* Destroy sets */
1163 
1164 static const struct nla_policy
1165 ip_set_setname_policy[IPSET_ATTR_CMD_MAX + 1] = {
1166 	[IPSET_ATTR_PROTOCOL]	= { .type = NLA_U8 },
1167 	[IPSET_ATTR_SETNAME]	= { .type = NLA_NUL_STRING,
1168 				    .len = IPSET_MAXNAMELEN - 1 },
1169 };
1170 
1171 static void
ip_set_destroy_set(struct ip_set * set)1172 ip_set_destroy_set(struct ip_set *set)
1173 {
1174 	pr_debug("set: %s\n",  set->name);
1175 
1176 	/* Must call it without holding any lock */
1177 	set->variant->destroy(set);
1178 	module_put(set->type->me);
1179 	kfree(set);
1180 }
1181 
ip_set_destroy(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const attr[])1182 static int ip_set_destroy(struct sk_buff *skb, const struct nfnl_info *info,
1183 			  const struct nlattr * const attr[])
1184 {
1185 	struct ip_set_net *inst = ip_set_pernet(info->net);
1186 	struct ip_set *s;
1187 	ip_set_id_t i;
1188 	int ret = 0;
1189 
1190 	if (unlikely(protocol_min_failed(attr)))
1191 		return -IPSET_ERR_PROTOCOL;
1192 
1193 	/* Must wait for flush to be really finished in list:set */
1194 	rcu_barrier();
1195 
1196 	/* Commands are serialized and references are
1197 	 * protected by the ip_set_ref_lock.
1198 	 * External systems (i.e. xt_set) must call
1199 	 * ip_set_put|get_nfnl_* functions, that way we
1200 	 * can safely check references here.
1201 	 *
1202 	 * list:set timer can only decrement the reference
1203 	 * counter, so if it's already zero, we can proceed
1204 	 * without holding the lock.
1205 	 */
1206 	read_lock_bh(&ip_set_ref_lock);
1207 	if (!attr[IPSET_ATTR_SETNAME]) {
1208 		for (i = 0; i < inst->ip_set_max; i++) {
1209 			s = ip_set(inst, i);
1210 			if (s && (s->ref || s->ref_netlink)) {
1211 				ret = -IPSET_ERR_BUSY;
1212 				goto out;
1213 			}
1214 		}
1215 		inst->is_destroyed = true;
1216 		read_unlock_bh(&ip_set_ref_lock);
1217 		for (i = 0; i < inst->ip_set_max; i++) {
1218 			s = ip_set(inst, i);
1219 			if (s) {
1220 				ip_set(inst, i) = NULL;
1221 				ip_set_destroy_set(s);
1222 			}
1223 		}
1224 		/* Modified by ip_set_destroy() only, which is serialized */
1225 		inst->is_destroyed = false;
1226 	} else {
1227 		u32 flags = flag_exist(info->nlh);
1228 		s = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1229 				    &i);
1230 		if (!s) {
1231 			if (!(flags & IPSET_FLAG_EXIST))
1232 				ret = -ENOENT;
1233 			goto out;
1234 		} else if (s->ref || s->ref_netlink) {
1235 			ret = -IPSET_ERR_BUSY;
1236 			goto out;
1237 		}
1238 		ip_set(inst, i) = NULL;
1239 		read_unlock_bh(&ip_set_ref_lock);
1240 
1241 		ip_set_destroy_set(s);
1242 	}
1243 	return 0;
1244 out:
1245 	read_unlock_bh(&ip_set_ref_lock);
1246 	return ret;
1247 }
1248 
1249 /* Flush sets */
1250 
1251 static void
ip_set_flush_set(struct ip_set * set)1252 ip_set_flush_set(struct ip_set *set)
1253 {
1254 	pr_debug("set: %s\n",  set->name);
1255 
1256 	ip_set_lock(set);
1257 	set->variant->flush(set);
1258 	ip_set_unlock(set);
1259 }
1260 
ip_set_flush(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const attr[])1261 static int ip_set_flush(struct sk_buff *skb, const struct nfnl_info *info,
1262 			const struct nlattr * const attr[])
1263 {
1264 	struct ip_set_net *inst = ip_set_pernet(info->net);
1265 	struct ip_set *s;
1266 	ip_set_id_t i;
1267 
1268 	if (unlikely(protocol_min_failed(attr)))
1269 		return -IPSET_ERR_PROTOCOL;
1270 
1271 	if (!attr[IPSET_ATTR_SETNAME]) {
1272 		for (i = 0; i < inst->ip_set_max; i++) {
1273 			s = ip_set(inst, i);
1274 			if (s)
1275 				ip_set_flush_set(s);
1276 		}
1277 	} else {
1278 		s = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1279 		if (!s)
1280 			return -ENOENT;
1281 
1282 		ip_set_flush_set(s);
1283 	}
1284 
1285 	return 0;
1286 }
1287 
1288 /* Rename a set */
1289 
1290 static const struct nla_policy
1291 ip_set_setname2_policy[IPSET_ATTR_CMD_MAX + 1] = {
1292 	[IPSET_ATTR_PROTOCOL]	= { .type = NLA_U8 },
1293 	[IPSET_ATTR_SETNAME]	= { .type = NLA_NUL_STRING,
1294 				    .len = IPSET_MAXNAMELEN - 1 },
1295 	[IPSET_ATTR_SETNAME2]	= { .type = NLA_NUL_STRING,
1296 				    .len = IPSET_MAXNAMELEN - 1 },
1297 };
1298 
ip_set_rename(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const attr[])1299 static int ip_set_rename(struct sk_buff *skb, const struct nfnl_info *info,
1300 			 const struct nlattr * const attr[])
1301 {
1302 	struct ip_set_net *inst = ip_set_pernet(info->net);
1303 	struct ip_set *set, *s;
1304 	const char *name2;
1305 	ip_set_id_t i;
1306 	int ret = 0;
1307 
1308 	if (unlikely(protocol_min_failed(attr) ||
1309 		     !attr[IPSET_ATTR_SETNAME] ||
1310 		     !attr[IPSET_ATTR_SETNAME2]))
1311 		return -IPSET_ERR_PROTOCOL;
1312 
1313 	set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1314 	if (!set)
1315 		return -ENOENT;
1316 
1317 	write_lock_bh(&ip_set_ref_lock);
1318 	if (set->ref != 0 || set->ref_netlink != 0) {
1319 		ret = -IPSET_ERR_REFERENCED;
1320 		goto out;
1321 	}
1322 
1323 	name2 = nla_data(attr[IPSET_ATTR_SETNAME2]);
1324 	for (i = 0; i < inst->ip_set_max; i++) {
1325 		s = ip_set(inst, i);
1326 		if (s && STRNCMP(s->name, name2)) {
1327 			ret = -IPSET_ERR_EXIST_SETNAME2;
1328 			goto out;
1329 		}
1330 	}
1331 	strncpy(set->name, name2, IPSET_MAXNAMELEN);
1332 
1333 out:
1334 	write_unlock_bh(&ip_set_ref_lock);
1335 	return ret;
1336 }
1337 
1338 /* Swap two sets so that name/index points to the other.
1339  * References and set names are also swapped.
1340  *
1341  * The commands are serialized by the nfnl mutex and references are
1342  * protected by the ip_set_ref_lock. The kernel interfaces
1343  * do not hold the mutex but the pointer settings are atomic
1344  * so the ip_set_list always contains valid pointers to the sets.
1345  */
1346 
ip_set_swap(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const attr[])1347 static int ip_set_swap(struct sk_buff *skb, const struct nfnl_info *info,
1348 		       const struct nlattr * const attr[])
1349 {
1350 	struct ip_set_net *inst = ip_set_pernet(info->net);
1351 	struct ip_set *from, *to;
1352 	ip_set_id_t from_id, to_id;
1353 	char from_name[IPSET_MAXNAMELEN];
1354 
1355 	if (unlikely(protocol_min_failed(attr) ||
1356 		     !attr[IPSET_ATTR_SETNAME] ||
1357 		     !attr[IPSET_ATTR_SETNAME2]))
1358 		return -IPSET_ERR_PROTOCOL;
1359 
1360 	from = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]),
1361 			       &from_id);
1362 	if (!from)
1363 		return -ENOENT;
1364 
1365 	to = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME2]),
1366 			     &to_id);
1367 	if (!to)
1368 		return -IPSET_ERR_EXIST_SETNAME2;
1369 
1370 	/* Features must not change.
1371 	 * Not an artifical restriction anymore, as we must prevent
1372 	 * possible loops created by swapping in setlist type of sets.
1373 	 */
1374 	if (!(from->type->features == to->type->features &&
1375 	      from->family == to->family))
1376 		return -IPSET_ERR_TYPE_MISMATCH;
1377 
1378 	write_lock_bh(&ip_set_ref_lock);
1379 
1380 	if (from->ref_netlink || to->ref_netlink) {
1381 		write_unlock_bh(&ip_set_ref_lock);
1382 		return -EBUSY;
1383 	}
1384 
1385 	strncpy(from_name, from->name, IPSET_MAXNAMELEN);
1386 	strncpy(from->name, to->name, IPSET_MAXNAMELEN);
1387 	strncpy(to->name, from_name, IPSET_MAXNAMELEN);
1388 
1389 	swap(from->ref, to->ref);
1390 	ip_set(inst, from_id) = to;
1391 	ip_set(inst, to_id) = from;
1392 	write_unlock_bh(&ip_set_ref_lock);
1393 
1394 	return 0;
1395 }
1396 
1397 /* List/save set data */
1398 
1399 #define DUMP_INIT	0
1400 #define DUMP_ALL	1
1401 #define DUMP_ONE	2
1402 #define DUMP_LAST	3
1403 
1404 #define DUMP_TYPE(arg)		(((u32)(arg)) & 0x0000FFFF)
1405 #define DUMP_FLAGS(arg)		(((u32)(arg)) >> 16)
1406 
1407 int
ip_set_put_flags(struct sk_buff * skb,struct ip_set * set)1408 ip_set_put_flags(struct sk_buff *skb, struct ip_set *set)
1409 {
1410 	u32 cadt_flags = 0;
1411 
1412 	if (SET_WITH_TIMEOUT(set))
1413 		if (unlikely(nla_put_net32(skb, IPSET_ATTR_TIMEOUT,
1414 					   htonl(set->timeout))))
1415 			return -EMSGSIZE;
1416 	if (SET_WITH_COUNTER(set))
1417 		cadt_flags |= IPSET_FLAG_WITH_COUNTERS;
1418 	if (SET_WITH_COMMENT(set))
1419 		cadt_flags |= IPSET_FLAG_WITH_COMMENT;
1420 	if (SET_WITH_SKBINFO(set))
1421 		cadt_flags |= IPSET_FLAG_WITH_SKBINFO;
1422 	if (SET_WITH_FORCEADD(set))
1423 		cadt_flags |= IPSET_FLAG_WITH_FORCEADD;
1424 
1425 	if (!cadt_flags)
1426 		return 0;
1427 	return nla_put_net32(skb, IPSET_ATTR_CADT_FLAGS, htonl(cadt_flags));
1428 }
1429 EXPORT_SYMBOL_GPL(ip_set_put_flags);
1430 
1431 static int
ip_set_dump_done(struct netlink_callback * cb)1432 ip_set_dump_done(struct netlink_callback *cb)
1433 {
1434 	if (cb->args[IPSET_CB_ARG0]) {
1435 		struct ip_set_net *inst =
1436 			(struct ip_set_net *)cb->args[IPSET_CB_NET];
1437 		ip_set_id_t index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1438 		struct ip_set *set = ip_set_ref_netlink(inst, index);
1439 
1440 		if (set->variant->uref)
1441 			set->variant->uref(set, cb, false);
1442 		pr_debug("release set %s\n", set->name);
1443 		__ip_set_put_netlink(set);
1444 	}
1445 	return 0;
1446 }
1447 
1448 static inline void
dump_attrs(struct nlmsghdr * nlh)1449 dump_attrs(struct nlmsghdr *nlh)
1450 {
1451 	const struct nlattr *attr;
1452 	int rem;
1453 
1454 	pr_debug("dump nlmsg\n");
1455 	nlmsg_for_each_attr(attr, nlh, sizeof(struct nfgenmsg), rem) {
1456 		pr_debug("type: %u, len %u\n", nla_type(attr), attr->nla_len);
1457 	}
1458 }
1459 
1460 static const struct nla_policy
1461 ip_set_dump_policy[IPSET_ATTR_CMD_MAX + 1] = {
1462 	[IPSET_ATTR_PROTOCOL]	= { .type = NLA_U8 },
1463 	[IPSET_ATTR_SETNAME]	= { .type = NLA_NUL_STRING,
1464 				    .len = IPSET_MAXNAMELEN - 1 },
1465 	[IPSET_ATTR_FLAGS]	= { .type = NLA_U32 },
1466 };
1467 
1468 static int
ip_set_dump_start(struct netlink_callback * cb)1469 ip_set_dump_start(struct netlink_callback *cb)
1470 {
1471 	struct nlmsghdr *nlh = nlmsg_hdr(cb->skb);
1472 	int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1473 	struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1474 	struct nlattr *attr = (void *)nlh + min_len;
1475 	struct sk_buff *skb = cb->skb;
1476 	struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1477 	u32 dump_type;
1478 	int ret;
1479 
1480 	ret = nla_parse(cda, IPSET_ATTR_CMD_MAX, attr,
1481 			nlh->nlmsg_len - min_len,
1482 			ip_set_dump_policy, NULL);
1483 	if (ret)
1484 		goto error;
1485 
1486 	cb->args[IPSET_CB_PROTO] = nla_get_u8(cda[IPSET_ATTR_PROTOCOL]);
1487 	if (cda[IPSET_ATTR_SETNAME]) {
1488 		ip_set_id_t index;
1489 		struct ip_set *set;
1490 
1491 		set = find_set_and_id(inst, nla_data(cda[IPSET_ATTR_SETNAME]),
1492 				      &index);
1493 		if (!set) {
1494 			ret = -ENOENT;
1495 			goto error;
1496 		}
1497 		dump_type = DUMP_ONE;
1498 		cb->args[IPSET_CB_INDEX] = index;
1499 	} else {
1500 		dump_type = DUMP_ALL;
1501 	}
1502 
1503 	if (cda[IPSET_ATTR_FLAGS]) {
1504 		u32 f = ip_set_get_h32(cda[IPSET_ATTR_FLAGS]);
1505 
1506 		dump_type |= (f << 16);
1507 	}
1508 	cb->args[IPSET_CB_NET] = (unsigned long)inst;
1509 	cb->args[IPSET_CB_DUMP] = dump_type;
1510 
1511 	return 0;
1512 
1513 error:
1514 	/* We have to create and send the error message manually :-( */
1515 	if (nlh->nlmsg_flags & NLM_F_ACK) {
1516 		netlink_ack(cb->skb, nlh, ret, NULL);
1517 	}
1518 	return ret;
1519 }
1520 
1521 static int
ip_set_dump_do(struct sk_buff * skb,struct netlink_callback * cb)1522 ip_set_dump_do(struct sk_buff *skb, struct netlink_callback *cb)
1523 {
1524 	ip_set_id_t index = IPSET_INVALID_ID, max;
1525 	struct ip_set *set = NULL;
1526 	struct nlmsghdr *nlh = NULL;
1527 	unsigned int flags = NETLINK_CB(cb->skb).portid ? NLM_F_MULTI : 0;
1528 	struct ip_set_net *inst = ip_set_pernet(sock_net(skb->sk));
1529 	u32 dump_type, dump_flags;
1530 	bool is_destroyed;
1531 	int ret = 0;
1532 
1533 	if (!cb->args[IPSET_CB_DUMP])
1534 		return -EINVAL;
1535 
1536 	if (cb->args[IPSET_CB_INDEX] >= inst->ip_set_max)
1537 		goto out;
1538 
1539 	dump_type = DUMP_TYPE(cb->args[IPSET_CB_DUMP]);
1540 	dump_flags = DUMP_FLAGS(cb->args[IPSET_CB_DUMP]);
1541 	max = dump_type == DUMP_ONE ? cb->args[IPSET_CB_INDEX] + 1
1542 				    : inst->ip_set_max;
1543 dump_last:
1544 	pr_debug("dump type, flag: %u %u index: %ld\n",
1545 		 dump_type, dump_flags, cb->args[IPSET_CB_INDEX]);
1546 	for (; cb->args[IPSET_CB_INDEX] < max; cb->args[IPSET_CB_INDEX]++) {
1547 		index = (ip_set_id_t)cb->args[IPSET_CB_INDEX];
1548 		write_lock_bh(&ip_set_ref_lock);
1549 		set = ip_set(inst, index);
1550 		is_destroyed = inst->is_destroyed;
1551 		if (!set || is_destroyed) {
1552 			write_unlock_bh(&ip_set_ref_lock);
1553 			if (dump_type == DUMP_ONE) {
1554 				ret = -ENOENT;
1555 				goto out;
1556 			}
1557 			if (is_destroyed) {
1558 				/* All sets are just being destroyed */
1559 				ret = 0;
1560 				goto out;
1561 			}
1562 			continue;
1563 		}
1564 		/* When dumping all sets, we must dump "sorted"
1565 		 * so that lists (unions of sets) are dumped last.
1566 		 */
1567 		if (dump_type != DUMP_ONE &&
1568 		    ((dump_type == DUMP_ALL) ==
1569 		     !!(set->type->features & IPSET_DUMP_LAST))) {
1570 			write_unlock_bh(&ip_set_ref_lock);
1571 			continue;
1572 		}
1573 		pr_debug("List set: %s\n", set->name);
1574 		if (!cb->args[IPSET_CB_ARG0]) {
1575 			/* Start listing: make sure set won't be destroyed */
1576 			pr_debug("reference set\n");
1577 			set->ref_netlink++;
1578 		}
1579 		write_unlock_bh(&ip_set_ref_lock);
1580 		nlh = start_msg(skb, NETLINK_CB(cb->skb).portid,
1581 				cb->nlh->nlmsg_seq, flags,
1582 				IPSET_CMD_LIST);
1583 		if (!nlh) {
1584 			ret = -EMSGSIZE;
1585 			goto release_refcount;
1586 		}
1587 		if (nla_put_u8(skb, IPSET_ATTR_PROTOCOL,
1588 			       cb->args[IPSET_CB_PROTO]) ||
1589 		    nla_put_string(skb, IPSET_ATTR_SETNAME, set->name))
1590 			goto nla_put_failure;
1591 		if (dump_flags & IPSET_FLAG_LIST_SETNAME)
1592 			goto next_set;
1593 		switch (cb->args[IPSET_CB_ARG0]) {
1594 		case 0:
1595 			/* Core header data */
1596 			if (nla_put_string(skb, IPSET_ATTR_TYPENAME,
1597 					   set->type->name) ||
1598 			    nla_put_u8(skb, IPSET_ATTR_FAMILY,
1599 				       set->family) ||
1600 			    nla_put_u8(skb, IPSET_ATTR_REVISION,
1601 				       set->revision))
1602 				goto nla_put_failure;
1603 			if (cb->args[IPSET_CB_PROTO] > IPSET_PROTOCOL_MIN &&
1604 			    nla_put_net16(skb, IPSET_ATTR_INDEX, htons(index)))
1605 				goto nla_put_failure;
1606 			ret = set->variant->head(set, skb);
1607 			if (ret < 0)
1608 				goto release_refcount;
1609 			if (dump_flags & IPSET_FLAG_LIST_HEADER)
1610 				goto next_set;
1611 			if (set->variant->uref)
1612 				set->variant->uref(set, cb, true);
1613 			fallthrough;
1614 		default:
1615 			ret = set->variant->list(set, skb, cb);
1616 			if (!cb->args[IPSET_CB_ARG0])
1617 				/* Set is done, proceed with next one */
1618 				goto next_set;
1619 			goto release_refcount;
1620 		}
1621 	}
1622 	/* If we dump all sets, continue with dumping last ones */
1623 	if (dump_type == DUMP_ALL) {
1624 		dump_type = DUMP_LAST;
1625 		cb->args[IPSET_CB_DUMP] = dump_type | (dump_flags << 16);
1626 		cb->args[IPSET_CB_INDEX] = 0;
1627 		if (set && set->variant->uref)
1628 			set->variant->uref(set, cb, false);
1629 		goto dump_last;
1630 	}
1631 	goto out;
1632 
1633 nla_put_failure:
1634 	ret = -EFAULT;
1635 next_set:
1636 	if (dump_type == DUMP_ONE)
1637 		cb->args[IPSET_CB_INDEX] = IPSET_INVALID_ID;
1638 	else
1639 		cb->args[IPSET_CB_INDEX]++;
1640 release_refcount:
1641 	/* If there was an error or set is done, release set */
1642 	if (ret || !cb->args[IPSET_CB_ARG0]) {
1643 		set = ip_set_ref_netlink(inst, index);
1644 		if (set->variant->uref)
1645 			set->variant->uref(set, cb, false);
1646 		pr_debug("release set %s\n", set->name);
1647 		__ip_set_put_netlink(set);
1648 		cb->args[IPSET_CB_ARG0] = 0;
1649 	}
1650 out:
1651 	if (nlh) {
1652 		nlmsg_end(skb, nlh);
1653 		pr_debug("nlmsg_len: %u\n", nlh->nlmsg_len);
1654 		dump_attrs(nlh);
1655 	}
1656 
1657 	return ret < 0 ? ret : skb->len;
1658 }
1659 
ip_set_dump(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const attr[])1660 static int ip_set_dump(struct sk_buff *skb, const struct nfnl_info *info,
1661 		       const struct nlattr * const attr[])
1662 {
1663 	if (unlikely(protocol_min_failed(attr)))
1664 		return -IPSET_ERR_PROTOCOL;
1665 
1666 	{
1667 		struct netlink_dump_control c = {
1668 			.start = ip_set_dump_start,
1669 			.dump = ip_set_dump_do,
1670 			.done = ip_set_dump_done,
1671 		};
1672 		return netlink_dump_start(info->sk, skb, info->nlh, &c);
1673 	}
1674 }
1675 
1676 /* Add, del and test */
1677 
1678 static const struct nla_policy ip_set_adt_policy[IPSET_ATTR_CMD_MAX + 1] = {
1679 	[IPSET_ATTR_PROTOCOL]	= { .type = NLA_U8 },
1680 	[IPSET_ATTR_SETNAME]	= { .type = NLA_NUL_STRING,
1681 				    .len = IPSET_MAXNAMELEN - 1 },
1682 	[IPSET_ATTR_LINENO]	= { .type = NLA_U32 },
1683 	[IPSET_ATTR_DATA]	= { .type = NLA_NESTED },
1684 	[IPSET_ATTR_ADT]	= { .type = NLA_NESTED },
1685 };
1686 
1687 static int
call_ad(struct net * net,struct sock * ctnl,struct sk_buff * skb,struct ip_set * set,struct nlattr * tb[],enum ipset_adt adt,u32 flags,bool use_lineno)1688 call_ad(struct net *net, struct sock *ctnl, struct sk_buff *skb,
1689 	struct ip_set *set, struct nlattr *tb[], enum ipset_adt adt,
1690 	u32 flags, bool use_lineno)
1691 {
1692 	int ret;
1693 	u32 lineno = 0;
1694 	bool eexist = flags & IPSET_FLAG_EXIST, retried = false;
1695 
1696 	do {
1697 		ip_set_lock(set);
1698 		ret = set->variant->uadt(set, tb, adt, &lineno, flags, retried);
1699 		ip_set_unlock(set);
1700 		retried = true;
1701 	} while (ret == -ERANGE ||
1702 		 (ret == -EAGAIN &&
1703 		  set->variant->resize &&
1704 		  (ret = set->variant->resize(set, retried)) == 0));
1705 
1706 	if (!ret || (ret == -IPSET_ERR_EXIST && eexist))
1707 		return 0;
1708 	if (lineno && use_lineno) {
1709 		/* Error in restore/batch mode: send back lineno */
1710 		struct nlmsghdr *rep, *nlh = nlmsg_hdr(skb);
1711 		struct sk_buff *skb2;
1712 		struct nlmsgerr *errmsg;
1713 		size_t payload = min(SIZE_MAX,
1714 				     sizeof(*errmsg) + nlmsg_len(nlh));
1715 		int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
1716 		struct nlattr *cda[IPSET_ATTR_CMD_MAX + 1];
1717 		struct nlattr *cmdattr;
1718 		u32 *errline;
1719 
1720 		skb2 = nlmsg_new(payload, GFP_KERNEL);
1721 		if (!skb2)
1722 			return -ENOMEM;
1723 		rep = nlmsg_put(skb2, NETLINK_CB(skb).portid,
1724 				nlh->nlmsg_seq, NLMSG_ERROR, payload, 0);
1725 		errmsg = nlmsg_data(rep);
1726 		errmsg->error = ret;
1727 		unsafe_memcpy(&errmsg->msg, nlh, nlh->nlmsg_len,
1728 			      /* Bounds checked by the skb layer. */);
1729 
1730 		cmdattr = (void *)&errmsg->msg + min_len;
1731 
1732 		ret = nla_parse(cda, IPSET_ATTR_CMD_MAX, cmdattr,
1733 				nlh->nlmsg_len - min_len, ip_set_adt_policy,
1734 				NULL);
1735 
1736 		if (ret) {
1737 			nlmsg_free(skb2);
1738 			return ret;
1739 		}
1740 		errline = nla_data(cda[IPSET_ATTR_LINENO]);
1741 
1742 		*errline = lineno;
1743 
1744 		nfnetlink_unicast(skb2, net, NETLINK_CB(skb).portid);
1745 		/* Signal netlink not to send its ACK/errmsg.  */
1746 		return -EINTR;
1747 	}
1748 
1749 	return ret;
1750 }
1751 
ip_set_ad(struct net * net,struct sock * ctnl,struct sk_buff * skb,enum ipset_adt adt,const struct nlmsghdr * nlh,const struct nlattr * const attr[],struct netlink_ext_ack * extack)1752 static int ip_set_ad(struct net *net, struct sock *ctnl,
1753 		     struct sk_buff *skb,
1754 		     enum ipset_adt adt,
1755 		     const struct nlmsghdr *nlh,
1756 		     const struct nlattr * const attr[],
1757 		     struct netlink_ext_ack *extack)
1758 {
1759 	struct ip_set_net *inst = ip_set_pernet(net);
1760 	struct ip_set *set;
1761 	struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1762 	const struct nlattr *nla;
1763 	u32 flags = flag_exist(nlh);
1764 	bool use_lineno;
1765 	int ret = 0;
1766 
1767 	if (unlikely(protocol_min_failed(attr) ||
1768 		     !attr[IPSET_ATTR_SETNAME] ||
1769 		     !((attr[IPSET_ATTR_DATA] != NULL) ^
1770 		       (attr[IPSET_ATTR_ADT] != NULL)) ||
1771 		     (attr[IPSET_ATTR_DATA] &&
1772 		      !flag_nested(attr[IPSET_ATTR_DATA])) ||
1773 		     (attr[IPSET_ATTR_ADT] &&
1774 		      (!flag_nested(attr[IPSET_ATTR_ADT]) ||
1775 		       !attr[IPSET_ATTR_LINENO]))))
1776 		return -IPSET_ERR_PROTOCOL;
1777 
1778 	set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1779 	if (!set)
1780 		return -ENOENT;
1781 
1782 	use_lineno = !!attr[IPSET_ATTR_LINENO];
1783 	if (attr[IPSET_ATTR_DATA]) {
1784 		if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX,
1785 				     attr[IPSET_ATTR_DATA],
1786 				     set->type->adt_policy, NULL))
1787 			return -IPSET_ERR_PROTOCOL;
1788 		ret = call_ad(net, ctnl, skb, set, tb, adt, flags,
1789 			      use_lineno);
1790 	} else {
1791 		int nla_rem;
1792 
1793 		nla_for_each_nested(nla, attr[IPSET_ATTR_ADT], nla_rem) {
1794 			if (nla_type(nla) != IPSET_ATTR_DATA ||
1795 			    !flag_nested(nla) ||
1796 			    nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, nla,
1797 					     set->type->adt_policy, NULL))
1798 				return -IPSET_ERR_PROTOCOL;
1799 			ret = call_ad(net, ctnl, skb, set, tb, adt,
1800 				      flags, use_lineno);
1801 			if (ret < 0)
1802 				return ret;
1803 		}
1804 	}
1805 	return ret;
1806 }
1807 
ip_set_uadd(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const attr[])1808 static int ip_set_uadd(struct sk_buff *skb, const struct nfnl_info *info,
1809 		       const struct nlattr * const attr[])
1810 {
1811 	return ip_set_ad(info->net, info->sk, skb,
1812 			 IPSET_ADD, info->nlh, attr, info->extack);
1813 }
1814 
ip_set_udel(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const attr[])1815 static int ip_set_udel(struct sk_buff *skb, const struct nfnl_info *info,
1816 		       const struct nlattr * const attr[])
1817 {
1818 	return ip_set_ad(info->net, info->sk, skb,
1819 			 IPSET_DEL, info->nlh, attr, info->extack);
1820 }
1821 
ip_set_utest(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const attr[])1822 static int ip_set_utest(struct sk_buff *skb, const struct nfnl_info *info,
1823 			const struct nlattr * const attr[])
1824 {
1825 	struct ip_set_net *inst = ip_set_pernet(info->net);
1826 	struct ip_set *set;
1827 	struct nlattr *tb[IPSET_ATTR_ADT_MAX + 1] = {};
1828 	int ret = 0;
1829 	u32 lineno;
1830 
1831 	if (unlikely(protocol_min_failed(attr) ||
1832 		     !attr[IPSET_ATTR_SETNAME] ||
1833 		     !attr[IPSET_ATTR_DATA] ||
1834 		     !flag_nested(attr[IPSET_ATTR_DATA])))
1835 		return -IPSET_ERR_PROTOCOL;
1836 
1837 	set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1838 	if (!set)
1839 		return -ENOENT;
1840 
1841 	if (nla_parse_nested(tb, IPSET_ATTR_ADT_MAX, attr[IPSET_ATTR_DATA],
1842 			     set->type->adt_policy, NULL))
1843 		return -IPSET_ERR_PROTOCOL;
1844 
1845 	rcu_read_lock_bh();
1846 	ret = set->variant->uadt(set, tb, IPSET_TEST, &lineno, 0, 0);
1847 	rcu_read_unlock_bh();
1848 	/* Userspace can't trigger element to be re-added */
1849 	if (ret == -EAGAIN)
1850 		ret = 1;
1851 
1852 	return ret > 0 ? 0 : -IPSET_ERR_EXIST;
1853 }
1854 
1855 /* Get headed data of a set */
1856 
ip_set_header(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const attr[])1857 static int ip_set_header(struct sk_buff *skb, const struct nfnl_info *info,
1858 			 const struct nlattr * const attr[])
1859 {
1860 	struct ip_set_net *inst = ip_set_pernet(info->net);
1861 	const struct ip_set *set;
1862 	struct sk_buff *skb2;
1863 	struct nlmsghdr *nlh2;
1864 
1865 	if (unlikely(protocol_min_failed(attr) ||
1866 		     !attr[IPSET_ATTR_SETNAME]))
1867 		return -IPSET_ERR_PROTOCOL;
1868 
1869 	set = find_set(inst, nla_data(attr[IPSET_ATTR_SETNAME]));
1870 	if (!set)
1871 		return -ENOENT;
1872 
1873 	skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1874 	if (!skb2)
1875 		return -ENOMEM;
1876 
1877 	nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, info->nlh->nlmsg_seq, 0,
1878 			 IPSET_CMD_HEADER);
1879 	if (!nlh2)
1880 		goto nlmsg_failure;
1881 	if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
1882 	    nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name) ||
1883 	    nla_put_string(skb2, IPSET_ATTR_TYPENAME, set->type->name) ||
1884 	    nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
1885 	    nla_put_u8(skb2, IPSET_ATTR_REVISION, set->revision))
1886 		goto nla_put_failure;
1887 	nlmsg_end(skb2, nlh2);
1888 
1889 	return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid);
1890 
1891 nla_put_failure:
1892 	nlmsg_cancel(skb2, nlh2);
1893 nlmsg_failure:
1894 	kfree_skb(skb2);
1895 	return -EMSGSIZE;
1896 }
1897 
1898 /* Get type data */
1899 
1900 static const struct nla_policy ip_set_type_policy[IPSET_ATTR_CMD_MAX + 1] = {
1901 	[IPSET_ATTR_PROTOCOL]	= { .type = NLA_U8 },
1902 	[IPSET_ATTR_TYPENAME]	= { .type = NLA_NUL_STRING,
1903 				    .len = IPSET_MAXNAMELEN - 1 },
1904 	[IPSET_ATTR_FAMILY]	= { .type = NLA_U8 },
1905 };
1906 
ip_set_type(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const attr[])1907 static int ip_set_type(struct sk_buff *skb, const struct nfnl_info *info,
1908 		       const struct nlattr * const attr[])
1909 {
1910 	struct sk_buff *skb2;
1911 	struct nlmsghdr *nlh2;
1912 	u8 family, min, max;
1913 	const char *typename;
1914 	int ret = 0;
1915 
1916 	if (unlikely(protocol_min_failed(attr) ||
1917 		     !attr[IPSET_ATTR_TYPENAME] ||
1918 		     !attr[IPSET_ATTR_FAMILY]))
1919 		return -IPSET_ERR_PROTOCOL;
1920 
1921 	family = nla_get_u8(attr[IPSET_ATTR_FAMILY]);
1922 	typename = nla_data(attr[IPSET_ATTR_TYPENAME]);
1923 	ret = find_set_type_minmax(typename, family, &min, &max);
1924 	if (ret)
1925 		return ret;
1926 
1927 	skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1928 	if (!skb2)
1929 		return -ENOMEM;
1930 
1931 	nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, info->nlh->nlmsg_seq, 0,
1932 			 IPSET_CMD_TYPE);
1933 	if (!nlh2)
1934 		goto nlmsg_failure;
1935 	if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
1936 	    nla_put_string(skb2, IPSET_ATTR_TYPENAME, typename) ||
1937 	    nla_put_u8(skb2, IPSET_ATTR_FAMILY, family) ||
1938 	    nla_put_u8(skb2, IPSET_ATTR_REVISION, max) ||
1939 	    nla_put_u8(skb2, IPSET_ATTR_REVISION_MIN, min))
1940 		goto nla_put_failure;
1941 	nlmsg_end(skb2, nlh2);
1942 
1943 	pr_debug("Send TYPE, nlmsg_len: %u\n", nlh2->nlmsg_len);
1944 	return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid);
1945 
1946 nla_put_failure:
1947 	nlmsg_cancel(skb2, nlh2);
1948 nlmsg_failure:
1949 	kfree_skb(skb2);
1950 	return -EMSGSIZE;
1951 }
1952 
1953 /* Get protocol version */
1954 
1955 static const struct nla_policy
1956 ip_set_protocol_policy[IPSET_ATTR_CMD_MAX + 1] = {
1957 	[IPSET_ATTR_PROTOCOL]	= { .type = NLA_U8 },
1958 };
1959 
ip_set_protocol(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const attr[])1960 static int ip_set_protocol(struct sk_buff *skb, const struct nfnl_info *info,
1961 			   const struct nlattr * const attr[])
1962 {
1963 	struct sk_buff *skb2;
1964 	struct nlmsghdr *nlh2;
1965 
1966 	if (unlikely(!attr[IPSET_ATTR_PROTOCOL]))
1967 		return -IPSET_ERR_PROTOCOL;
1968 
1969 	skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1970 	if (!skb2)
1971 		return -ENOMEM;
1972 
1973 	nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, info->nlh->nlmsg_seq, 0,
1974 			 IPSET_CMD_PROTOCOL);
1975 	if (!nlh2)
1976 		goto nlmsg_failure;
1977 	if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, IPSET_PROTOCOL))
1978 		goto nla_put_failure;
1979 	if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL_MIN, IPSET_PROTOCOL_MIN))
1980 		goto nla_put_failure;
1981 	nlmsg_end(skb2, nlh2);
1982 
1983 	return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid);
1984 
1985 nla_put_failure:
1986 	nlmsg_cancel(skb2, nlh2);
1987 nlmsg_failure:
1988 	kfree_skb(skb2);
1989 	return -EMSGSIZE;
1990 }
1991 
1992 /* Get set by name or index, from userspace */
1993 
ip_set_byname(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const attr[])1994 static int ip_set_byname(struct sk_buff *skb, const struct nfnl_info *info,
1995 			 const struct nlattr * const attr[])
1996 {
1997 	struct ip_set_net *inst = ip_set_pernet(info->net);
1998 	struct sk_buff *skb2;
1999 	struct nlmsghdr *nlh2;
2000 	ip_set_id_t id = IPSET_INVALID_ID;
2001 	const struct ip_set *set;
2002 
2003 	if (unlikely(protocol_failed(attr) ||
2004 		     !attr[IPSET_ATTR_SETNAME]))
2005 		return -IPSET_ERR_PROTOCOL;
2006 
2007 	set = find_set_and_id(inst, nla_data(attr[IPSET_ATTR_SETNAME]), &id);
2008 	if (id == IPSET_INVALID_ID)
2009 		return -ENOENT;
2010 
2011 	skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
2012 	if (!skb2)
2013 		return -ENOMEM;
2014 
2015 	nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, info->nlh->nlmsg_seq, 0,
2016 			 IPSET_CMD_GET_BYNAME);
2017 	if (!nlh2)
2018 		goto nlmsg_failure;
2019 	if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
2020 	    nla_put_u8(skb2, IPSET_ATTR_FAMILY, set->family) ||
2021 	    nla_put_net16(skb2, IPSET_ATTR_INDEX, htons(id)))
2022 		goto nla_put_failure;
2023 	nlmsg_end(skb2, nlh2);
2024 
2025 	return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid);
2026 
2027 nla_put_failure:
2028 	nlmsg_cancel(skb2, nlh2);
2029 nlmsg_failure:
2030 	kfree_skb(skb2);
2031 	return -EMSGSIZE;
2032 }
2033 
2034 static const struct nla_policy ip_set_index_policy[IPSET_ATTR_CMD_MAX + 1] = {
2035 	[IPSET_ATTR_PROTOCOL]	= { .type = NLA_U8 },
2036 	[IPSET_ATTR_INDEX]	= { .type = NLA_U16 },
2037 };
2038 
ip_set_byindex(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const attr[])2039 static int ip_set_byindex(struct sk_buff *skb, const struct nfnl_info *info,
2040 			  const struct nlattr * const attr[])
2041 {
2042 	struct ip_set_net *inst = ip_set_pernet(info->net);
2043 	struct sk_buff *skb2;
2044 	struct nlmsghdr *nlh2;
2045 	ip_set_id_t id = IPSET_INVALID_ID;
2046 	const struct ip_set *set;
2047 
2048 	if (unlikely(protocol_failed(attr) ||
2049 		     !attr[IPSET_ATTR_INDEX]))
2050 		return -IPSET_ERR_PROTOCOL;
2051 
2052 	id = ip_set_get_h16(attr[IPSET_ATTR_INDEX]);
2053 	if (id >= inst->ip_set_max)
2054 		return -ENOENT;
2055 	set = ip_set(inst, id);
2056 	if (set == NULL)
2057 		return -ENOENT;
2058 
2059 	skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
2060 	if (!skb2)
2061 		return -ENOMEM;
2062 
2063 	nlh2 = start_msg(skb2, NETLINK_CB(skb).portid, info->nlh->nlmsg_seq, 0,
2064 			 IPSET_CMD_GET_BYINDEX);
2065 	if (!nlh2)
2066 		goto nlmsg_failure;
2067 	if (nla_put_u8(skb2, IPSET_ATTR_PROTOCOL, protocol(attr)) ||
2068 	    nla_put_string(skb2, IPSET_ATTR_SETNAME, set->name))
2069 		goto nla_put_failure;
2070 	nlmsg_end(skb2, nlh2);
2071 
2072 	return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid);
2073 
2074 nla_put_failure:
2075 	nlmsg_cancel(skb2, nlh2);
2076 nlmsg_failure:
2077 	kfree_skb(skb2);
2078 	return -EMSGSIZE;
2079 }
2080 
2081 static const struct nfnl_callback ip_set_netlink_subsys_cb[IPSET_MSG_MAX] = {
2082 	[IPSET_CMD_NONE]	= {
2083 		.call		= ip_set_none,
2084 		.type		= NFNL_CB_MUTEX,
2085 		.attr_count	= IPSET_ATTR_CMD_MAX,
2086 	},
2087 	[IPSET_CMD_CREATE]	= {
2088 		.call		= ip_set_create,
2089 		.type		= NFNL_CB_MUTEX,
2090 		.attr_count	= IPSET_ATTR_CMD_MAX,
2091 		.policy		= ip_set_create_policy,
2092 	},
2093 	[IPSET_CMD_DESTROY]	= {
2094 		.call		= ip_set_destroy,
2095 		.type		= NFNL_CB_MUTEX,
2096 		.attr_count	= IPSET_ATTR_CMD_MAX,
2097 		.policy		= ip_set_setname_policy,
2098 	},
2099 	[IPSET_CMD_FLUSH]	= {
2100 		.call		= ip_set_flush,
2101 		.type		= NFNL_CB_MUTEX,
2102 		.attr_count	= IPSET_ATTR_CMD_MAX,
2103 		.policy		= ip_set_setname_policy,
2104 	},
2105 	[IPSET_CMD_RENAME]	= {
2106 		.call		= ip_set_rename,
2107 		.type		= NFNL_CB_MUTEX,
2108 		.attr_count	= IPSET_ATTR_CMD_MAX,
2109 		.policy		= ip_set_setname2_policy,
2110 	},
2111 	[IPSET_CMD_SWAP]	= {
2112 		.call		= ip_set_swap,
2113 		.type		= NFNL_CB_MUTEX,
2114 		.attr_count	= IPSET_ATTR_CMD_MAX,
2115 		.policy		= ip_set_setname2_policy,
2116 	},
2117 	[IPSET_CMD_LIST]	= {
2118 		.call		= ip_set_dump,
2119 		.type		= NFNL_CB_MUTEX,
2120 		.attr_count	= IPSET_ATTR_CMD_MAX,
2121 		.policy		= ip_set_dump_policy,
2122 	},
2123 	[IPSET_CMD_SAVE]	= {
2124 		.call		= ip_set_dump,
2125 		.type		= NFNL_CB_MUTEX,
2126 		.attr_count	= IPSET_ATTR_CMD_MAX,
2127 		.policy		= ip_set_setname_policy,
2128 	},
2129 	[IPSET_CMD_ADD]	= {
2130 		.call		= ip_set_uadd,
2131 		.type		= NFNL_CB_MUTEX,
2132 		.attr_count	= IPSET_ATTR_CMD_MAX,
2133 		.policy		= ip_set_adt_policy,
2134 	},
2135 	[IPSET_CMD_DEL]	= {
2136 		.call		= ip_set_udel,
2137 		.type		= NFNL_CB_MUTEX,
2138 		.attr_count	= IPSET_ATTR_CMD_MAX,
2139 		.policy		= ip_set_adt_policy,
2140 	},
2141 	[IPSET_CMD_TEST]	= {
2142 		.call		= ip_set_utest,
2143 		.type		= NFNL_CB_MUTEX,
2144 		.attr_count	= IPSET_ATTR_CMD_MAX,
2145 		.policy		= ip_set_adt_policy,
2146 	},
2147 	[IPSET_CMD_HEADER]	= {
2148 		.call		= ip_set_header,
2149 		.type		= NFNL_CB_MUTEX,
2150 		.attr_count	= IPSET_ATTR_CMD_MAX,
2151 		.policy		= ip_set_setname_policy,
2152 	},
2153 	[IPSET_CMD_TYPE]	= {
2154 		.call		= ip_set_type,
2155 		.type		= NFNL_CB_MUTEX,
2156 		.attr_count	= IPSET_ATTR_CMD_MAX,
2157 		.policy		= ip_set_type_policy,
2158 	},
2159 	[IPSET_CMD_PROTOCOL]	= {
2160 		.call		= ip_set_protocol,
2161 		.type		= NFNL_CB_MUTEX,
2162 		.attr_count	= IPSET_ATTR_CMD_MAX,
2163 		.policy		= ip_set_protocol_policy,
2164 	},
2165 	[IPSET_CMD_GET_BYNAME]	= {
2166 		.call		= ip_set_byname,
2167 		.type		= NFNL_CB_MUTEX,
2168 		.attr_count	= IPSET_ATTR_CMD_MAX,
2169 		.policy		= ip_set_setname_policy,
2170 	},
2171 	[IPSET_CMD_GET_BYINDEX]	= {
2172 		.call		= ip_set_byindex,
2173 		.type		= NFNL_CB_MUTEX,
2174 		.attr_count	= IPSET_ATTR_CMD_MAX,
2175 		.policy		= ip_set_index_policy,
2176 	},
2177 };
2178 
2179 static struct nfnetlink_subsystem ip_set_netlink_subsys __read_mostly = {
2180 	.name		= "ip_set",
2181 	.subsys_id	= NFNL_SUBSYS_IPSET,
2182 	.cb_count	= IPSET_MSG_MAX,
2183 	.cb		= ip_set_netlink_subsys_cb,
2184 };
2185 
2186 /* Interface to iptables/ip6tables */
2187 
2188 static int
ip_set_sockfn_get(struct sock * sk,int optval,void __user * user,int * len)2189 ip_set_sockfn_get(struct sock *sk, int optval, void __user *user, int *len)
2190 {
2191 	unsigned int *op;
2192 	void *data;
2193 	int copylen = *len, ret = 0;
2194 	struct net *net = sock_net(sk);
2195 	struct ip_set_net *inst = ip_set_pernet(net);
2196 
2197 	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2198 		return -EPERM;
2199 	if (optval != SO_IP_SET)
2200 		return -EBADF;
2201 	if (*len < sizeof(unsigned int))
2202 		return -EINVAL;
2203 
2204 	data = vmalloc(*len);
2205 	if (!data)
2206 		return -ENOMEM;
2207 	if (copy_from_user(data, user, *len) != 0) {
2208 		ret = -EFAULT;
2209 		goto done;
2210 	}
2211 	op = data;
2212 
2213 	if (*op < IP_SET_OP_VERSION) {
2214 		/* Check the version at the beginning of operations */
2215 		struct ip_set_req_version *req_version = data;
2216 
2217 		if (*len < sizeof(struct ip_set_req_version)) {
2218 			ret = -EINVAL;
2219 			goto done;
2220 		}
2221 
2222 		if (req_version->version < IPSET_PROTOCOL_MIN) {
2223 			ret = -EPROTO;
2224 			goto done;
2225 		}
2226 	}
2227 
2228 	switch (*op) {
2229 	case IP_SET_OP_VERSION: {
2230 		struct ip_set_req_version *req_version = data;
2231 
2232 		if (*len != sizeof(struct ip_set_req_version)) {
2233 			ret = -EINVAL;
2234 			goto done;
2235 		}
2236 
2237 		req_version->version = IPSET_PROTOCOL;
2238 		if (copy_to_user(user, req_version,
2239 				 sizeof(struct ip_set_req_version)))
2240 			ret = -EFAULT;
2241 		goto done;
2242 	}
2243 	case IP_SET_OP_GET_BYNAME: {
2244 		struct ip_set_req_get_set *req_get = data;
2245 		ip_set_id_t id;
2246 
2247 		if (*len != sizeof(struct ip_set_req_get_set)) {
2248 			ret = -EINVAL;
2249 			goto done;
2250 		}
2251 		req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
2252 		nfnl_lock(NFNL_SUBSYS_IPSET);
2253 		find_set_and_id(inst, req_get->set.name, &id);
2254 		req_get->set.index = id;
2255 		nfnl_unlock(NFNL_SUBSYS_IPSET);
2256 		goto copy;
2257 	}
2258 	case IP_SET_OP_GET_FNAME: {
2259 		struct ip_set_req_get_set_family *req_get = data;
2260 		ip_set_id_t id;
2261 
2262 		if (*len != sizeof(struct ip_set_req_get_set_family)) {
2263 			ret = -EINVAL;
2264 			goto done;
2265 		}
2266 		req_get->set.name[IPSET_MAXNAMELEN - 1] = '\0';
2267 		nfnl_lock(NFNL_SUBSYS_IPSET);
2268 		find_set_and_id(inst, req_get->set.name, &id);
2269 		req_get->set.index = id;
2270 		if (id != IPSET_INVALID_ID)
2271 			req_get->family = ip_set(inst, id)->family;
2272 		nfnl_unlock(NFNL_SUBSYS_IPSET);
2273 		goto copy;
2274 	}
2275 	case IP_SET_OP_GET_BYINDEX: {
2276 		struct ip_set_req_get_set *req_get = data;
2277 		struct ip_set *set;
2278 
2279 		if (*len != sizeof(struct ip_set_req_get_set) ||
2280 		    req_get->set.index >= inst->ip_set_max) {
2281 			ret = -EINVAL;
2282 			goto done;
2283 		}
2284 		nfnl_lock(NFNL_SUBSYS_IPSET);
2285 		set = ip_set(inst, req_get->set.index);
2286 		ret = strscpy(req_get->set.name, set ? set->name : "",
2287 			      IPSET_MAXNAMELEN);
2288 		nfnl_unlock(NFNL_SUBSYS_IPSET);
2289 		if (ret < 0)
2290 			goto done;
2291 		goto copy;
2292 	}
2293 	default:
2294 		ret = -EBADMSG;
2295 		goto done;
2296 	}	/* end of switch(op) */
2297 
2298 copy:
2299 	if (copy_to_user(user, data, copylen))
2300 		ret = -EFAULT;
2301 
2302 done:
2303 	vfree(data);
2304 	if (ret > 0)
2305 		ret = 0;
2306 	return ret;
2307 }
2308 
2309 static struct nf_sockopt_ops so_set __read_mostly = {
2310 	.pf		= PF_INET,
2311 	.get_optmin	= SO_IP_SET,
2312 	.get_optmax	= SO_IP_SET + 1,
2313 	.get		= ip_set_sockfn_get,
2314 	.owner		= THIS_MODULE,
2315 };
2316 
2317 static int __net_init
ip_set_net_init(struct net * net)2318 ip_set_net_init(struct net *net)
2319 {
2320 	struct ip_set_net *inst = ip_set_pernet(net);
2321 	struct ip_set **list;
2322 
2323 	inst->ip_set_max = max_sets ? max_sets : CONFIG_IP_SET_MAX;
2324 	if (inst->ip_set_max >= IPSET_INVALID_ID)
2325 		inst->ip_set_max = IPSET_INVALID_ID - 1;
2326 
2327 	list = kvcalloc(inst->ip_set_max, sizeof(struct ip_set *), GFP_KERNEL);
2328 	if (!list)
2329 		return -ENOMEM;
2330 	inst->is_deleted = false;
2331 	inst->is_destroyed = false;
2332 	rcu_assign_pointer(inst->ip_set_list, list);
2333 	return 0;
2334 }
2335 
2336 static void __net_exit
ip_set_net_exit(struct net * net)2337 ip_set_net_exit(struct net *net)
2338 {
2339 	struct ip_set_net *inst = ip_set_pernet(net);
2340 
2341 	struct ip_set *set = NULL;
2342 	ip_set_id_t i;
2343 
2344 	inst->is_deleted = true; /* flag for ip_set_nfnl_put */
2345 
2346 	nfnl_lock(NFNL_SUBSYS_IPSET);
2347 	for (i = 0; i < inst->ip_set_max; i++) {
2348 		set = ip_set(inst, i);
2349 		if (set) {
2350 			ip_set(inst, i) = NULL;
2351 			ip_set_destroy_set(set);
2352 		}
2353 	}
2354 	nfnl_unlock(NFNL_SUBSYS_IPSET);
2355 	kvfree(rcu_dereference_protected(inst->ip_set_list, 1));
2356 }
2357 
2358 static struct pernet_operations ip_set_net_ops = {
2359 	.init	= ip_set_net_init,
2360 	.exit   = ip_set_net_exit,
2361 	.id	= &ip_set_net_id,
2362 	.size	= sizeof(struct ip_set_net),
2363 };
2364 
2365 static int __init
ip_set_init(void)2366 ip_set_init(void)
2367 {
2368 	int ret = register_pernet_subsys(&ip_set_net_ops);
2369 
2370 	if (ret) {
2371 		pr_err("ip_set: cannot register pernet_subsys.\n");
2372 		return ret;
2373 	}
2374 
2375 	ret = nfnetlink_subsys_register(&ip_set_netlink_subsys);
2376 	if (ret != 0) {
2377 		pr_err("ip_set: cannot register with nfnetlink.\n");
2378 		unregister_pernet_subsys(&ip_set_net_ops);
2379 		return ret;
2380 	}
2381 
2382 	ret = nf_register_sockopt(&so_set);
2383 	if (ret != 0) {
2384 		pr_err("SO_SET registry failed: %d\n", ret);
2385 		nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2386 		unregister_pernet_subsys(&ip_set_net_ops);
2387 		return ret;
2388 	}
2389 
2390 	return 0;
2391 }
2392 
2393 static void __exit
ip_set_fini(void)2394 ip_set_fini(void)
2395 {
2396 	nf_unregister_sockopt(&so_set);
2397 	nfnetlink_subsys_unregister(&ip_set_netlink_subsys);
2398 
2399 	unregister_pernet_subsys(&ip_set_net_ops);
2400 	pr_debug("these are the famous last words\n");
2401 }
2402 
2403 module_init(ip_set_init);
2404 module_exit(ip_set_fini);
2405 
2406 MODULE_DESCRIPTION("ip_set: protocol " __stringify(IPSET_PROTOCOL));
2407