Merge tag 'ceph-for-5.13-rc1' of git://github.com/ceph/ceph-client
[linux-2.6-microblaze.git] / net / netfilter / nfnetlink.c
1 /* Netfilter messages via netlink socket. Allows for user space
2  * protocol helpers and general trouble making from userspace.
3  *
4  * (C) 2001 by Jay Schulist <jschlst@samba.org>,
5  * (C) 2002-2005 by Harald Welte <laforge@gnumonks.org>
6  * (C) 2005-2017 by Pablo Neira Ayuso <pablo@netfilter.org>
7  *
8  * Initial netfilter messages via netlink development funded and
9  * generally made possible by Network Robots, Inc. (www.networkrobots.com)
10  *
11  * Further development of this code funded by Astaro AG (http://www.astaro.com)
12  *
13  * This software may be used and distributed according to the terms
14  * of the GNU General Public License, incorporated herein by reference.
15  */
16
17 #include <linux/module.h>
18 #include <linux/types.h>
19 #include <linux/socket.h>
20 #include <linux/kernel.h>
21 #include <linux/string.h>
22 #include <linux/sockios.h>
23 #include <linux/net.h>
24 #include <linux/skbuff.h>
25 #include <linux/uaccess.h>
26 #include <net/sock.h>
27 #include <linux/init.h>
28 #include <linux/sched/signal.h>
29
30 #include <net/netlink.h>
31 #include <net/netns/generic.h>
32 #include <linux/netfilter/nfnetlink.h>
33
34 MODULE_LICENSE("GPL");
35 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
36 MODULE_ALIAS_NET_PF_PROTO(PF_NETLINK, NETLINK_NETFILTER);
37 MODULE_DESCRIPTION("Netfilter messages via netlink socket");
38
39 #define nfnl_dereference_protected(id) \
40         rcu_dereference_protected(table[(id)].subsys, \
41                                   lockdep_nfnl_is_held((id)))
42
43 #define NFNL_MAX_ATTR_COUNT     32
44
45 static unsigned int nfnetlink_pernet_id __read_mostly;
46
47 struct nfnl_net {
48         struct sock *nfnl;
49 };
50
51 static struct {
52         struct mutex                            mutex;
53         const struct nfnetlink_subsystem __rcu  *subsys;
54 } table[NFNL_SUBSYS_COUNT];
55
56 static struct lock_class_key nfnl_lockdep_keys[NFNL_SUBSYS_COUNT];
57
58 static const char *const nfnl_lockdep_names[NFNL_SUBSYS_COUNT] = {
59         [NFNL_SUBSYS_NONE] = "nfnl_subsys_none",
60         [NFNL_SUBSYS_CTNETLINK] = "nfnl_subsys_ctnetlink",
61         [NFNL_SUBSYS_CTNETLINK_EXP] = "nfnl_subsys_ctnetlink_exp",
62         [NFNL_SUBSYS_QUEUE] = "nfnl_subsys_queue",
63         [NFNL_SUBSYS_ULOG] = "nfnl_subsys_ulog",
64         [NFNL_SUBSYS_OSF] = "nfnl_subsys_osf",
65         [NFNL_SUBSYS_IPSET] = "nfnl_subsys_ipset",
66         [NFNL_SUBSYS_ACCT] = "nfnl_subsys_acct",
67         [NFNL_SUBSYS_CTNETLINK_TIMEOUT] = "nfnl_subsys_cttimeout",
68         [NFNL_SUBSYS_CTHELPER] = "nfnl_subsys_cthelper",
69         [NFNL_SUBSYS_NFTABLES] = "nfnl_subsys_nftables",
70         [NFNL_SUBSYS_NFT_COMPAT] = "nfnl_subsys_nftcompat",
71 };
72
73 static const int nfnl_group2type[NFNLGRP_MAX+1] = {
74         [NFNLGRP_CONNTRACK_NEW]         = NFNL_SUBSYS_CTNETLINK,
75         [NFNLGRP_CONNTRACK_UPDATE]      = NFNL_SUBSYS_CTNETLINK,
76         [NFNLGRP_CONNTRACK_DESTROY]     = NFNL_SUBSYS_CTNETLINK,
77         [NFNLGRP_CONNTRACK_EXP_NEW]     = NFNL_SUBSYS_CTNETLINK_EXP,
78         [NFNLGRP_CONNTRACK_EXP_UPDATE]  = NFNL_SUBSYS_CTNETLINK_EXP,
79         [NFNLGRP_CONNTRACK_EXP_DESTROY] = NFNL_SUBSYS_CTNETLINK_EXP,
80         [NFNLGRP_NFTABLES]              = NFNL_SUBSYS_NFTABLES,
81         [NFNLGRP_ACCT_QUOTA]            = NFNL_SUBSYS_ACCT,
82         [NFNLGRP_NFTRACE]               = NFNL_SUBSYS_NFTABLES,
83 };
84
85 static struct nfnl_net *nfnl_pernet(struct net *net)
86 {
87         return net_generic(net, nfnetlink_pernet_id);
88 }
89
90 void nfnl_lock(__u8 subsys_id)
91 {
92         mutex_lock(&table[subsys_id].mutex);
93 }
94 EXPORT_SYMBOL_GPL(nfnl_lock);
95
96 void nfnl_unlock(__u8 subsys_id)
97 {
98         mutex_unlock(&table[subsys_id].mutex);
99 }
100 EXPORT_SYMBOL_GPL(nfnl_unlock);
101
102 #ifdef CONFIG_PROVE_LOCKING
103 bool lockdep_nfnl_is_held(u8 subsys_id)
104 {
105         return lockdep_is_held(&table[subsys_id].mutex);
106 }
107 EXPORT_SYMBOL_GPL(lockdep_nfnl_is_held);
108 #endif
109
110 int nfnetlink_subsys_register(const struct nfnetlink_subsystem *n)
111 {
112         u8 cb_id;
113
114         /* Sanity-check attr_count size to avoid stack buffer overflow. */
115         for (cb_id = 0; cb_id < n->cb_count; cb_id++)
116                 if (WARN_ON(n->cb[cb_id].attr_count > NFNL_MAX_ATTR_COUNT))
117                         return -EINVAL;
118
119         nfnl_lock(n->subsys_id);
120         if (table[n->subsys_id].subsys) {
121                 nfnl_unlock(n->subsys_id);
122                 return -EBUSY;
123         }
124         rcu_assign_pointer(table[n->subsys_id].subsys, n);
125         nfnl_unlock(n->subsys_id);
126
127         return 0;
128 }
129 EXPORT_SYMBOL_GPL(nfnetlink_subsys_register);
130
131 int nfnetlink_subsys_unregister(const struct nfnetlink_subsystem *n)
132 {
133         nfnl_lock(n->subsys_id);
134         table[n->subsys_id].subsys = NULL;
135         nfnl_unlock(n->subsys_id);
136         synchronize_rcu();
137         return 0;
138 }
139 EXPORT_SYMBOL_GPL(nfnetlink_subsys_unregister);
140
141 static inline const struct nfnetlink_subsystem *nfnetlink_get_subsys(u16 type)
142 {
143         u8 subsys_id = NFNL_SUBSYS_ID(type);
144
145         if (subsys_id >= NFNL_SUBSYS_COUNT)
146                 return NULL;
147
148         return rcu_dereference(table[subsys_id].subsys);
149 }
150
151 static inline const struct nfnl_callback *
152 nfnetlink_find_client(u16 type, const struct nfnetlink_subsystem *ss)
153 {
154         u8 cb_id = NFNL_MSG_TYPE(type);
155
156         if (cb_id >= ss->cb_count)
157                 return NULL;
158
159         return &ss->cb[cb_id];
160 }
161
162 int nfnetlink_has_listeners(struct net *net, unsigned int group)
163 {
164         struct nfnl_net *nfnlnet = nfnl_pernet(net);
165
166         return netlink_has_listeners(nfnlnet->nfnl, group);
167 }
168 EXPORT_SYMBOL_GPL(nfnetlink_has_listeners);
169
170 int nfnetlink_send(struct sk_buff *skb, struct net *net, u32 portid,
171                    unsigned int group, int echo, gfp_t flags)
172 {
173         struct nfnl_net *nfnlnet = nfnl_pernet(net);
174
175         return nlmsg_notify(nfnlnet->nfnl, skb, portid, group, echo, flags);
176 }
177 EXPORT_SYMBOL_GPL(nfnetlink_send);
178
179 int nfnetlink_set_err(struct net *net, u32 portid, u32 group, int error)
180 {
181         struct nfnl_net *nfnlnet = nfnl_pernet(net);
182
183         return netlink_set_err(nfnlnet->nfnl, portid, group, error);
184 }
185 EXPORT_SYMBOL_GPL(nfnetlink_set_err);
186
187 int nfnetlink_unicast(struct sk_buff *skb, struct net *net, u32 portid)
188 {
189         struct nfnl_net *nfnlnet = nfnl_pernet(net);
190         int err;
191
192         err = nlmsg_unicast(nfnlnet->nfnl, skb, portid);
193         if (err == -EAGAIN)
194                 err = -ENOBUFS;
195
196         return err;
197 }
198 EXPORT_SYMBOL_GPL(nfnetlink_unicast);
199
200 void nfnetlink_broadcast(struct net *net, struct sk_buff *skb, __u32 portid,
201                          __u32 group, gfp_t allocation)
202 {
203         struct nfnl_net *nfnlnet = nfnl_pernet(net);
204
205         netlink_broadcast(nfnlnet->nfnl, skb, portid, group, allocation);
206 }
207 EXPORT_SYMBOL_GPL(nfnetlink_broadcast);
208
209 /* Process one complete nfnetlink message. */
210 static int nfnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
211                              struct netlink_ext_ack *extack)
212 {
213         struct net *net = sock_net(skb->sk);
214         const struct nfnl_callback *nc;
215         const struct nfnetlink_subsystem *ss;
216         int type, err;
217
218         /* All the messages must at least contain nfgenmsg */
219         if (nlmsg_len(nlh) < sizeof(struct nfgenmsg))
220                 return 0;
221
222         type = nlh->nlmsg_type;
223 replay:
224         rcu_read_lock();
225
226         ss = nfnetlink_get_subsys(type);
227         if (!ss) {
228 #ifdef CONFIG_MODULES
229                 rcu_read_unlock();
230                 request_module("nfnetlink-subsys-%d", NFNL_SUBSYS_ID(type));
231                 rcu_read_lock();
232                 ss = nfnetlink_get_subsys(type);
233                 if (!ss)
234 #endif
235                 {
236                         rcu_read_unlock();
237                         return -EINVAL;
238                 }
239         }
240
241         nc = nfnetlink_find_client(type, ss);
242         if (!nc) {
243                 rcu_read_unlock();
244                 return -EINVAL;
245         }
246
247         {
248                 int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
249                 struct nfnl_net *nfnlnet = nfnl_pernet(net);
250                 u8 cb_id = NFNL_MSG_TYPE(nlh->nlmsg_type);
251                 struct nlattr *cda[NFNL_MAX_ATTR_COUNT + 1];
252                 struct nlattr *attr = (void *)nlh + min_len;
253                 int attrlen = nlh->nlmsg_len - min_len;
254                 __u8 subsys_id = NFNL_SUBSYS_ID(type);
255                 struct nfnl_info info = {
256                         .net    = net,
257                         .sk     = nfnlnet->nfnl,
258                         .nlh    = nlh,
259                         .extack = extack,
260                 };
261
262                 /* Sanity-check NFNL_MAX_ATTR_COUNT */
263                 if (ss->cb[cb_id].attr_count > NFNL_MAX_ATTR_COUNT) {
264                         rcu_read_unlock();
265                         return -ENOMEM;
266                 }
267
268                 err = nla_parse_deprecated(cda, ss->cb[cb_id].attr_count,
269                                            attr, attrlen,
270                                            ss->cb[cb_id].policy, extack);
271                 if (err < 0) {
272                         rcu_read_unlock();
273                         return err;
274                 }
275
276                 if (!nc->call) {
277                         rcu_read_unlock();
278                         return -EINVAL;
279                 }
280
281                 switch (nc->type) {
282                 case NFNL_CB_RCU:
283                         err = nc->call(skb, &info, (const struct nlattr **)cda);
284                         rcu_read_unlock();
285                         break;
286                 case NFNL_CB_MUTEX:
287                         rcu_read_unlock();
288                         nfnl_lock(subsys_id);
289                         if (nfnl_dereference_protected(subsys_id) != ss ||
290                             nfnetlink_find_client(type, ss) != nc) {
291                                 err = -EAGAIN;
292                                 break;
293                         }
294                         err = nc->call(skb, &info, (const struct nlattr **)cda);
295                         nfnl_unlock(subsys_id);
296                         break;
297                 default:
298                         err = -EINVAL;
299                         break;
300                 }
301                 if (err == -EAGAIN)
302                         goto replay;
303                 return err;
304         }
305 }
306
307 struct nfnl_err {
308         struct list_head        head;
309         struct nlmsghdr         *nlh;
310         int                     err;
311         struct netlink_ext_ack  extack;
312 };
313
314 static int nfnl_err_add(struct list_head *list, struct nlmsghdr *nlh, int err,
315                         const struct netlink_ext_ack *extack)
316 {
317         struct nfnl_err *nfnl_err;
318
319         nfnl_err = kmalloc(sizeof(struct nfnl_err), GFP_KERNEL);
320         if (nfnl_err == NULL)
321                 return -ENOMEM;
322
323         nfnl_err->nlh = nlh;
324         nfnl_err->err = err;
325         nfnl_err->extack = *extack;
326         list_add_tail(&nfnl_err->head, list);
327
328         return 0;
329 }
330
331 static void nfnl_err_del(struct nfnl_err *nfnl_err)
332 {
333         list_del(&nfnl_err->head);
334         kfree(nfnl_err);
335 }
336
337 static void nfnl_err_reset(struct list_head *err_list)
338 {
339         struct nfnl_err *nfnl_err, *next;
340
341         list_for_each_entry_safe(nfnl_err, next, err_list, head)
342                 nfnl_err_del(nfnl_err);
343 }
344
345 static void nfnl_err_deliver(struct list_head *err_list, struct sk_buff *skb)
346 {
347         struct nfnl_err *nfnl_err, *next;
348
349         list_for_each_entry_safe(nfnl_err, next, err_list, head) {
350                 netlink_ack(skb, nfnl_err->nlh, nfnl_err->err,
351                             &nfnl_err->extack);
352                 nfnl_err_del(nfnl_err);
353         }
354 }
355
356 enum {
357         NFNL_BATCH_FAILURE      = (1 << 0),
358         NFNL_BATCH_DONE         = (1 << 1),
359         NFNL_BATCH_REPLAY       = (1 << 2),
360 };
361
362 static void nfnetlink_rcv_batch(struct sk_buff *skb, struct nlmsghdr *nlh,
363                                 u16 subsys_id, u32 genid)
364 {
365         struct sk_buff *oskb = skb;
366         struct net *net = sock_net(skb->sk);
367         const struct nfnetlink_subsystem *ss;
368         const struct nfnl_callback *nc;
369         struct netlink_ext_ack extack;
370         LIST_HEAD(err_list);
371         u32 status;
372         int err;
373
374         if (subsys_id >= NFNL_SUBSYS_COUNT)
375                 return netlink_ack(skb, nlh, -EINVAL, NULL);
376 replay:
377         status = 0;
378 replay_abort:
379         skb = netlink_skb_clone(oskb, GFP_KERNEL);
380         if (!skb)
381                 return netlink_ack(oskb, nlh, -ENOMEM, NULL);
382
383         nfnl_lock(subsys_id);
384         ss = nfnl_dereference_protected(subsys_id);
385         if (!ss) {
386 #ifdef CONFIG_MODULES
387                 nfnl_unlock(subsys_id);
388                 request_module("nfnetlink-subsys-%d", subsys_id);
389                 nfnl_lock(subsys_id);
390                 ss = nfnl_dereference_protected(subsys_id);
391                 if (!ss)
392 #endif
393                 {
394                         nfnl_unlock(subsys_id);
395                         netlink_ack(oskb, nlh, -EOPNOTSUPP, NULL);
396                         return kfree_skb(skb);
397                 }
398         }
399
400         if (!ss->valid_genid || !ss->commit || !ss->abort) {
401                 nfnl_unlock(subsys_id);
402                 netlink_ack(oskb, nlh, -EOPNOTSUPP, NULL);
403                 return kfree_skb(skb);
404         }
405
406         if (!try_module_get(ss->owner)) {
407                 nfnl_unlock(subsys_id);
408                 netlink_ack(oskb, nlh, -EOPNOTSUPP, NULL);
409                 return kfree_skb(skb);
410         }
411
412         if (!ss->valid_genid(net, genid)) {
413                 module_put(ss->owner);
414                 nfnl_unlock(subsys_id);
415                 netlink_ack(oskb, nlh, -ERESTART, NULL);
416                 return kfree_skb(skb);
417         }
418
419         nfnl_unlock(subsys_id);
420
421         while (skb->len >= nlmsg_total_size(0)) {
422                 int msglen, type;
423
424                 if (fatal_signal_pending(current)) {
425                         nfnl_err_reset(&err_list);
426                         err = -EINTR;
427                         status = NFNL_BATCH_FAILURE;
428                         goto done;
429                 }
430
431                 memset(&extack, 0, sizeof(extack));
432                 nlh = nlmsg_hdr(skb);
433                 err = 0;
434
435                 if (nlh->nlmsg_len < NLMSG_HDRLEN ||
436                     skb->len < nlh->nlmsg_len ||
437                     nlmsg_len(nlh) < sizeof(struct nfgenmsg)) {
438                         nfnl_err_reset(&err_list);
439                         status |= NFNL_BATCH_FAILURE;
440                         goto done;
441                 }
442
443                 /* Only requests are handled by the kernel */
444                 if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) {
445                         err = -EINVAL;
446                         goto ack;
447                 }
448
449                 type = nlh->nlmsg_type;
450                 if (type == NFNL_MSG_BATCH_BEGIN) {
451                         /* Malformed: Batch begin twice */
452                         nfnl_err_reset(&err_list);
453                         status |= NFNL_BATCH_FAILURE;
454                         goto done;
455                 } else if (type == NFNL_MSG_BATCH_END) {
456                         status |= NFNL_BATCH_DONE;
457                         goto done;
458                 } else if (type < NLMSG_MIN_TYPE) {
459                         err = -EINVAL;
460                         goto ack;
461                 }
462
463                 /* We only accept a batch with messages for the same
464                  * subsystem.
465                  */
466                 if (NFNL_SUBSYS_ID(type) != subsys_id) {
467                         err = -EINVAL;
468                         goto ack;
469                 }
470
471                 nc = nfnetlink_find_client(type, ss);
472                 if (!nc) {
473                         err = -EINVAL;
474                         goto ack;
475                 }
476
477                 if (nc->type != NFNL_CB_BATCH) {
478                         err = -EINVAL;
479                         goto ack;
480                 }
481
482                 {
483                         int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
484                         struct nfnl_net *nfnlnet = nfnl_pernet(net);
485                         struct nlattr *cda[NFNL_MAX_ATTR_COUNT + 1];
486                         struct nlattr *attr = (void *)nlh + min_len;
487                         u8 cb_id = NFNL_MSG_TYPE(nlh->nlmsg_type);
488                         int attrlen = nlh->nlmsg_len - min_len;
489                         struct nfnl_info info = {
490                                 .net    = net,
491                                 .sk     = nfnlnet->nfnl,
492                                 .nlh    = nlh,
493                                 .extack = &extack,
494                         };
495
496                         /* Sanity-check NFTA_MAX_ATTR */
497                         if (ss->cb[cb_id].attr_count > NFNL_MAX_ATTR_COUNT) {
498                                 err = -ENOMEM;
499                                 goto ack;
500                         }
501
502                         err = nla_parse_deprecated(cda,
503                                                    ss->cb[cb_id].attr_count,
504                                                    attr, attrlen,
505                                                    ss->cb[cb_id].policy, NULL);
506                         if (err < 0)
507                                 goto ack;
508
509                         err = nc->call(skb, &info, (const struct nlattr **)cda);
510
511                         /* The lock was released to autoload some module, we
512                          * have to abort and start from scratch using the
513                          * original skb.
514                          */
515                         if (err == -EAGAIN) {
516                                 status |= NFNL_BATCH_REPLAY;
517                                 goto done;
518                         }
519                 }
520 ack:
521                 if (nlh->nlmsg_flags & NLM_F_ACK || err) {
522                         /* Errors are delivered once the full batch has been
523                          * processed, this avoids that the same error is
524                          * reported several times when replaying the batch.
525                          */
526                         if (nfnl_err_add(&err_list, nlh, err, &extack) < 0) {
527                                 /* We failed to enqueue an error, reset the
528                                  * list of errors and send OOM to userspace
529                                  * pointing to the batch header.
530                                  */
531                                 nfnl_err_reset(&err_list);
532                                 netlink_ack(oskb, nlmsg_hdr(oskb), -ENOMEM,
533                                             NULL);
534                                 status |= NFNL_BATCH_FAILURE;
535                                 goto done;
536                         }
537                         /* We don't stop processing the batch on errors, thus,
538                          * userspace gets all the errors that the batch
539                          * triggers.
540                          */
541                         if (err)
542                                 status |= NFNL_BATCH_FAILURE;
543                 }
544
545                 msglen = NLMSG_ALIGN(nlh->nlmsg_len);
546                 if (msglen > skb->len)
547                         msglen = skb->len;
548                 skb_pull(skb, msglen);
549         }
550 done:
551         if (status & NFNL_BATCH_REPLAY) {
552                 ss->abort(net, oskb, NFNL_ABORT_AUTOLOAD);
553                 nfnl_err_reset(&err_list);
554                 kfree_skb(skb);
555                 module_put(ss->owner);
556                 goto replay;
557         } else if (status == NFNL_BATCH_DONE) {
558                 err = ss->commit(net, oskb);
559                 if (err == -EAGAIN) {
560                         status |= NFNL_BATCH_REPLAY;
561                         goto done;
562                 } else if (err) {
563                         ss->abort(net, oskb, NFNL_ABORT_NONE);
564                         netlink_ack(oskb, nlmsg_hdr(oskb), err, NULL);
565                 }
566         } else {
567                 enum nfnl_abort_action abort_action;
568
569                 if (status & NFNL_BATCH_FAILURE)
570                         abort_action = NFNL_ABORT_NONE;
571                 else
572                         abort_action = NFNL_ABORT_VALIDATE;
573
574                 err = ss->abort(net, oskb, abort_action);
575                 if (err == -EAGAIN) {
576                         nfnl_err_reset(&err_list);
577                         kfree_skb(skb);
578                         module_put(ss->owner);
579                         status |= NFNL_BATCH_FAILURE;
580                         goto replay_abort;
581                 }
582         }
583         if (ss->cleanup)
584                 ss->cleanup(net);
585
586         nfnl_err_deliver(&err_list, oskb);
587         kfree_skb(skb);
588         module_put(ss->owner);
589 }
590
591 static const struct nla_policy nfnl_batch_policy[NFNL_BATCH_MAX + 1] = {
592         [NFNL_BATCH_GENID]      = { .type = NLA_U32 },
593 };
594
595 static void nfnetlink_rcv_skb_batch(struct sk_buff *skb, struct nlmsghdr *nlh)
596 {
597         int min_len = nlmsg_total_size(sizeof(struct nfgenmsg));
598         struct nlattr *attr = (void *)nlh + min_len;
599         struct nlattr *cda[NFNL_BATCH_MAX + 1];
600         int attrlen = nlh->nlmsg_len - min_len;
601         struct nfgenmsg *nfgenmsg;
602         int msglen, err;
603         u32 gen_id = 0;
604         u16 res_id;
605
606         msglen = NLMSG_ALIGN(nlh->nlmsg_len);
607         if (msglen > skb->len)
608                 msglen = skb->len;
609
610         if (skb->len < NLMSG_HDRLEN + sizeof(struct nfgenmsg))
611                 return;
612
613         err = nla_parse_deprecated(cda, NFNL_BATCH_MAX, attr, attrlen,
614                                    nfnl_batch_policy, NULL);
615         if (err < 0) {
616                 netlink_ack(skb, nlh, err, NULL);
617                 return;
618         }
619         if (cda[NFNL_BATCH_GENID])
620                 gen_id = ntohl(nla_get_be32(cda[NFNL_BATCH_GENID]));
621
622         nfgenmsg = nlmsg_data(nlh);
623         skb_pull(skb, msglen);
624         /* Work around old nft using host byte order */
625         if (nfgenmsg->res_id == NFNL_SUBSYS_NFTABLES)
626                 res_id = NFNL_SUBSYS_NFTABLES;
627         else
628                 res_id = ntohs(nfgenmsg->res_id);
629
630         nfnetlink_rcv_batch(skb, nlh, res_id, gen_id);
631 }
632
633 static void nfnetlink_rcv(struct sk_buff *skb)
634 {
635         struct nlmsghdr *nlh = nlmsg_hdr(skb);
636
637         if (skb->len < NLMSG_HDRLEN ||
638             nlh->nlmsg_len < NLMSG_HDRLEN ||
639             skb->len < nlh->nlmsg_len)
640                 return;
641
642         if (!netlink_net_capable(skb, CAP_NET_ADMIN)) {
643                 netlink_ack(skb, nlh, -EPERM, NULL);
644                 return;
645         }
646
647         if (nlh->nlmsg_type == NFNL_MSG_BATCH_BEGIN)
648                 nfnetlink_rcv_skb_batch(skb, nlh);
649         else
650                 netlink_rcv_skb(skb, nfnetlink_rcv_msg);
651 }
652
653 #ifdef CONFIG_MODULES
654 static int nfnetlink_bind(struct net *net, int group)
655 {
656         const struct nfnetlink_subsystem *ss;
657         int type;
658
659         if (group <= NFNLGRP_NONE || group > NFNLGRP_MAX)
660                 return 0;
661
662         type = nfnl_group2type[group];
663
664         rcu_read_lock();
665         ss = nfnetlink_get_subsys(type << 8);
666         rcu_read_unlock();
667         if (!ss)
668                 request_module_nowait("nfnetlink-subsys-%d", type);
669         return 0;
670 }
671 #endif
672
673 static int __net_init nfnetlink_net_init(struct net *net)
674 {
675         struct nfnl_net *nfnlnet = nfnl_pernet(net);
676         struct netlink_kernel_cfg cfg = {
677                 .groups = NFNLGRP_MAX,
678                 .input  = nfnetlink_rcv,
679 #ifdef CONFIG_MODULES
680                 .bind   = nfnetlink_bind,
681 #endif
682         };
683
684         nfnlnet->nfnl = netlink_kernel_create(net, NETLINK_NETFILTER, &cfg);
685         if (!nfnlnet->nfnl)
686                 return -ENOMEM;
687         return 0;
688 }
689
690 static void __net_exit nfnetlink_net_exit_batch(struct list_head *net_exit_list)
691 {
692         struct nfnl_net *nfnlnet;
693         struct net *net;
694
695         list_for_each_entry(net, net_exit_list, exit_list) {
696                 nfnlnet = nfnl_pernet(net);
697
698                 netlink_kernel_release(nfnlnet->nfnl);
699         }
700 }
701
702 static struct pernet_operations nfnetlink_net_ops = {
703         .init           = nfnetlink_net_init,
704         .exit_batch     = nfnetlink_net_exit_batch,
705         .id             = &nfnetlink_pernet_id,
706         .size           = sizeof(struct nfnl_net),
707 };
708
709 static int __init nfnetlink_init(void)
710 {
711         int i;
712
713         for (i = NFNLGRP_NONE + 1; i <= NFNLGRP_MAX; i++)
714                 BUG_ON(nfnl_group2type[i] == NFNL_SUBSYS_NONE);
715
716         for (i=0; i<NFNL_SUBSYS_COUNT; i++)
717                 __mutex_init(&table[i].mutex, nfnl_lockdep_names[i], &nfnl_lockdep_keys[i]);
718
719         return register_pernet_subsys(&nfnetlink_net_ops);
720 }
721
722 static void __exit nfnetlink_exit(void)
723 {
724         unregister_pernet_subsys(&nfnetlink_net_ops);
725 }
726 module_init(nfnetlink_init);
727 module_exit(nfnetlink_exit);