1 /* netfilter.c: look after the filters for various protocols.
2 * Heavily influenced by the old firewall.c by David Bonn and Alan Cox.
4 * Thanks to Rob `CmdrTaco' Malda for not influencing this code in any
7 * Rusty Russell (C)2000 -- This code is GPL.
8 * Patrick McHardy (c) 2006-2012
10 #include <linux/kernel.h>
11 #include <linux/netfilter.h>
12 #include <net/protocol.h>
13 #include <linux/init.h>
14 #include <linux/skbuff.h>
15 #include <linux/wait.h>
16 #include <linux/module.h>
17 #include <linux/interrupt.h>
19 #include <linux/netdevice.h>
20 #include <linux/netfilter_ipv6.h>
21 #include <linux/inetdevice.h>
22 #include <linux/proc_fs.h>
23 #include <linux/mutex.h>
24 #include <linux/slab.h>
25 #include <net/net_namespace.h>
28 #include "nf_internals.h"
30 static DEFINE_MUTEX(afinfo_mutex);
32 const struct nf_afinfo __rcu *nf_afinfo[NFPROTO_NUMPROTO] __read_mostly;
33 EXPORT_SYMBOL(nf_afinfo);
34 const struct nf_ipv6_ops __rcu *nf_ipv6_ops __read_mostly;
35 EXPORT_SYMBOL_GPL(nf_ipv6_ops);
37 DEFINE_PER_CPU(bool, nf_skb_duplicated);
38 EXPORT_SYMBOL_GPL(nf_skb_duplicated);
40 int nf_register_afinfo(const struct nf_afinfo *afinfo)
42 mutex_lock(&afinfo_mutex);
43 RCU_INIT_POINTER(nf_afinfo[afinfo->family], afinfo);
44 mutex_unlock(&afinfo_mutex);
47 EXPORT_SYMBOL_GPL(nf_register_afinfo);
49 void nf_unregister_afinfo(const struct nf_afinfo *afinfo)
51 mutex_lock(&afinfo_mutex);
52 RCU_INIT_POINTER(nf_afinfo[afinfo->family], NULL);
53 mutex_unlock(&afinfo_mutex);
56 EXPORT_SYMBOL_GPL(nf_unregister_afinfo);
58 #ifdef HAVE_JUMP_LABEL
59 struct static_key nf_hooks_needed[NFPROTO_NUMPROTO][NF_MAX_HOOKS];
60 EXPORT_SYMBOL(nf_hooks_needed);
63 static DEFINE_MUTEX(nf_hook_mutex);
65 static struct list_head *nf_find_hook_list(struct net *net,
66 const struct nf_hook_ops *reg)
68 struct list_head *hook_list = NULL;
70 if (reg->pf != NFPROTO_NETDEV)
71 hook_list = &net->nf.hooks[reg->pf][reg->hooknum];
72 else if (reg->hooknum == NF_NETDEV_INGRESS) {
73 #ifdef CONFIG_NETFILTER_INGRESS
74 if (reg->dev && dev_net(reg->dev) == net)
75 hook_list = ®->dev->nf_hooks_ingress;
81 struct nf_hook_entry {
82 const struct nf_hook_ops *orig_ops;
83 struct nf_hook_ops ops;
86 int nf_register_net_hook(struct net *net, const struct nf_hook_ops *reg)
88 struct list_head *hook_list;
89 struct nf_hook_entry *entry;
90 struct nf_hook_ops *elem;
92 entry = kmalloc(sizeof(*entry), GFP_KERNEL);
96 entry->orig_ops = reg;
99 hook_list = nf_find_hook_list(net, reg);
105 mutex_lock(&nf_hook_mutex);
106 list_for_each_entry(elem, hook_list, list) {
107 if (reg->priority < elem->priority)
110 list_add_rcu(&entry->ops.list, elem->list.prev);
111 mutex_unlock(&nf_hook_mutex);
112 #ifdef CONFIG_NETFILTER_INGRESS
113 if (reg->pf == NFPROTO_NETDEV && reg->hooknum == NF_NETDEV_INGRESS)
114 net_inc_ingress_queue();
116 #ifdef HAVE_JUMP_LABEL
117 static_key_slow_inc(&nf_hooks_needed[reg->pf][reg->hooknum]);
121 EXPORT_SYMBOL(nf_register_net_hook);
123 void nf_unregister_net_hook(struct net *net, const struct nf_hook_ops *reg)
125 struct list_head *hook_list;
126 struct nf_hook_entry *entry;
127 struct nf_hook_ops *elem;
129 hook_list = nf_find_hook_list(net, reg);
133 mutex_lock(&nf_hook_mutex);
134 list_for_each_entry(elem, hook_list, list) {
135 entry = container_of(elem, struct nf_hook_entry, ops);
136 if (entry->orig_ops == reg) {
137 list_del_rcu(&entry->ops.list);
141 mutex_unlock(&nf_hook_mutex);
142 if (&elem->list == hook_list) {
143 WARN(1, "nf_unregister_net_hook: hook not found!\n");
146 #ifdef CONFIG_NETFILTER_INGRESS
147 if (reg->pf == NFPROTO_NETDEV && reg->hooknum == NF_NETDEV_INGRESS)
148 net_dec_ingress_queue();
150 #ifdef HAVE_JUMP_LABEL
151 static_key_slow_dec(&nf_hooks_needed[reg->pf][reg->hooknum]);
154 nf_queue_nf_hook_drop(net, &entry->ops);
157 EXPORT_SYMBOL(nf_unregister_net_hook);
159 int nf_register_net_hooks(struct net *net, const struct nf_hook_ops *reg,
165 for (i = 0; i < n; i++) {
166 err = nf_register_net_hook(net, ®[i]);
174 nf_unregister_net_hooks(net, reg, i);
177 EXPORT_SYMBOL(nf_register_net_hooks);
179 void nf_unregister_net_hooks(struct net *net, const struct nf_hook_ops *reg,
183 nf_unregister_net_hook(net, ®[n]);
185 EXPORT_SYMBOL(nf_unregister_net_hooks);
187 static LIST_HEAD(nf_hook_list);
189 int nf_register_hook(struct nf_hook_ops *reg)
191 struct net *net, *last;
196 ret = nf_register_net_hook(net, reg);
197 if (ret && ret != -ENOENT)
200 list_add_tail(®->list, &nf_hook_list);
209 nf_unregister_net_hook(net, reg);
214 EXPORT_SYMBOL(nf_register_hook);
216 void nf_unregister_hook(struct nf_hook_ops *reg)
221 list_del(®->list);
223 nf_unregister_net_hook(net, reg);
226 EXPORT_SYMBOL(nf_unregister_hook);
228 int nf_register_hooks(struct nf_hook_ops *reg, unsigned int n)
233 for (i = 0; i < n; i++) {
234 err = nf_register_hook(®[i]);
242 nf_unregister_hooks(reg, i);
245 EXPORT_SYMBOL(nf_register_hooks);
247 void nf_unregister_hooks(struct nf_hook_ops *reg, unsigned int n)
250 nf_unregister_hook(®[n]);
252 EXPORT_SYMBOL(nf_unregister_hooks);
254 unsigned int nf_iterate(struct list_head *head,
256 struct nf_hook_state *state,
257 struct nf_hook_ops **elemp)
259 unsigned int verdict;
262 * The caller must not block between calls to this
263 * function because of risk of continuing from deleted element.
265 list_for_each_entry_continue_rcu((*elemp), head, list) {
266 if (state->thresh > (*elemp)->priority)
269 /* Optimization: we don't need to hold module
270 reference here, since function can't sleep. --RR */
272 verdict = (*elemp)->hook(*elemp, skb, state);
273 if (verdict != NF_ACCEPT) {
274 #ifdef CONFIG_NETFILTER_DEBUG
275 if (unlikely((verdict & NF_VERDICT_MASK)
277 NFDEBUG("Evil return from %p(%u).\n",
278 (*elemp)->hook, state->hook);
282 if (verdict != NF_REPEAT)
291 /* Returns 1 if okfn() needs to be executed by the caller,
292 * -EPERM for NF_DROP, 0 otherwise. */
293 int nf_hook_slow(struct sk_buff *skb, struct nf_hook_state *state)
295 struct nf_hook_ops *elem;
296 unsigned int verdict;
299 /* We may already have this, but read-locks nest anyway */
302 elem = list_entry_rcu(state->hook_list, struct nf_hook_ops, list);
304 verdict = nf_iterate(state->hook_list, skb, state, &elem);
305 if (verdict == NF_ACCEPT || verdict == NF_STOP) {
307 } else if ((verdict & NF_VERDICT_MASK) == NF_DROP) {
309 ret = NF_DROP_GETERR(verdict);
312 } else if ((verdict & NF_VERDICT_MASK) == NF_QUEUE) {
313 int err = nf_queue(skb, elem, state,
314 verdict >> NF_VERDICT_QBITS);
316 if (err == -ECANCELED)
319 (verdict & NF_VERDICT_FLAG_QUEUE_BYPASS))
327 EXPORT_SYMBOL(nf_hook_slow);
330 int skb_make_writable(struct sk_buff *skb, unsigned int writable_len)
332 if (writable_len > skb->len)
335 /* Not exclusive use of packet? Must copy. */
336 if (!skb_cloned(skb)) {
337 if (writable_len <= skb_headlen(skb))
339 } else if (skb_clone_writable(skb, writable_len))
342 if (writable_len <= skb_headlen(skb))
345 writable_len -= skb_headlen(skb);
347 return !!__pskb_pull_tail(skb, writable_len);
349 EXPORT_SYMBOL(skb_make_writable);
351 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
352 /* This does not belong here, but locally generated errors need it if connection
353 tracking in use: without this, connection may not be in hash table, and hence
354 manufactured ICMP or RST packets will not be associated with it. */
355 void (*ip_ct_attach)(struct sk_buff *, const struct sk_buff *)
357 EXPORT_SYMBOL(ip_ct_attach);
359 void nf_ct_attach(struct sk_buff *new, const struct sk_buff *skb)
361 void (*attach)(struct sk_buff *, const struct sk_buff *);
365 attach = rcu_dereference(ip_ct_attach);
371 EXPORT_SYMBOL(nf_ct_attach);
373 void (*nf_ct_destroy)(struct nf_conntrack *) __rcu __read_mostly;
374 EXPORT_SYMBOL(nf_ct_destroy);
376 void nf_conntrack_destroy(struct nf_conntrack *nfct)
378 void (*destroy)(struct nf_conntrack *);
381 destroy = rcu_dereference(nf_ct_destroy);
382 BUG_ON(destroy == NULL);
386 EXPORT_SYMBOL(nf_conntrack_destroy);
388 struct nfq_ct_hook __rcu *nfq_ct_hook __read_mostly;
389 EXPORT_SYMBOL_GPL(nfq_ct_hook);
391 /* Built-in default zone used e.g. by modules. */
392 const struct nf_conntrack_zone nf_ct_zone_dflt = {
393 .id = NF_CT_DEFAULT_ZONE_ID,
394 .dir = NF_CT_DEFAULT_ZONE_DIR,
396 EXPORT_SYMBOL_GPL(nf_ct_zone_dflt);
397 #endif /* CONFIG_NF_CONNTRACK */
399 #ifdef CONFIG_NF_NAT_NEEDED
400 void (*nf_nat_decode_session_hook)(struct sk_buff *, struct flowi *);
401 EXPORT_SYMBOL(nf_nat_decode_session_hook);
404 static int nf_register_hook_list(struct net *net)
406 struct nf_hook_ops *elem;
410 list_for_each_entry(elem, &nf_hook_list, list) {
411 ret = nf_register_net_hook(net, elem);
412 if (ret && ret != -ENOENT)
419 list_for_each_entry_continue_reverse(elem, &nf_hook_list, list)
420 nf_unregister_net_hook(net, elem);
425 static void nf_unregister_hook_list(struct net *net)
427 struct nf_hook_ops *elem;
430 list_for_each_entry(elem, &nf_hook_list, list)
431 nf_unregister_net_hook(net, elem);
435 static int __net_init netfilter_net_init(struct net *net)
439 for (i = 0; i < ARRAY_SIZE(net->nf.hooks); i++) {
440 for (h = 0; h < NF_MAX_HOOKS; h++)
441 INIT_LIST_HEAD(&net->nf.hooks[i][h]);
444 #ifdef CONFIG_PROC_FS
445 net->nf.proc_netfilter = proc_net_mkdir(net, "netfilter",
447 if (!net->nf.proc_netfilter) {
448 if (!net_eq(net, &init_net))
449 pr_err("cannot create netfilter proc entry");
454 ret = nf_register_hook_list(net);
456 remove_proc_entry("netfilter", net->proc_net);
461 static void __net_exit netfilter_net_exit(struct net *net)
463 nf_unregister_hook_list(net);
464 remove_proc_entry("netfilter", net->proc_net);
467 static struct pernet_operations netfilter_net_ops = {
468 .init = netfilter_net_init,
469 .exit = netfilter_net_exit,
472 int __init netfilter_init(void)
476 ret = register_pernet_subsys(&netfilter_net_ops);
480 ret = netfilter_log_init();
486 unregister_pernet_subsys(&netfilter_net_ops);