~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/netfilter/ipset/ip_set_core.c

Version: ~ [ linux-5.5 ] ~ [ linux-5.4.15 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.98 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.167 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.211 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.211 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.81 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

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

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp