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 <linux/rcupdate.h>
26 #include <net/net_namespace.h>
29 #include "nf_internals.h"
31 static DEFINE_MUTEX(afinfo_mutex);
33 const struct nf_afinfo __rcu *nf_afinfo[NFPROTO_NUMPROTO] __read_mostly;
34 EXPORT_SYMBOL(nf_afinfo);
35 const struct nf_ipv6_ops __rcu *nf_ipv6_ops __read_mostly;
36 EXPORT_SYMBOL_GPL(nf_ipv6_ops);
38 DEFINE_PER_CPU(bool, nf_skb_duplicated);
39 EXPORT_SYMBOL_GPL(nf_skb_duplicated);
41 int nf_register_afinfo(const struct nf_afinfo *afinfo)
43 mutex_lock(&afinfo_mutex);
44 RCU_INIT_POINTER(nf_afinfo[afinfo->family], afinfo);
45 mutex_unlock(&afinfo_mutex);
48 EXPORT_SYMBOL_GPL(nf_register_afinfo);
50 void nf_unregister_afinfo(const struct nf_afinfo *afinfo)
52 mutex_lock(&afinfo_mutex);
53 RCU_INIT_POINTER(nf_afinfo[afinfo->family], NULL);
54 mutex_unlock(&afinfo_mutex);
57 EXPORT_SYMBOL_GPL(nf_unregister_afinfo);
59 #ifdef HAVE_JUMP_LABEL
60 struct static_key nf_hooks_needed[NFPROTO_NUMPROTO][NF_MAX_HOOKS];
61 EXPORT_SYMBOL(nf_hooks_needed);
64 static DEFINE_MUTEX(nf_hook_mutex);
65 #define nf_entry_dereference(e) \
66 rcu_dereference_protected(e, lockdep_is_held(&nf_hook_mutex))
68 static struct nf_hook_entry __rcu **nf_hook_entry_head(struct net *net, const struct nf_hook_ops *reg)
70 if (reg->pf != NFPROTO_NETDEV)
71 return net->nf.hooks[reg->pf]+reg->hooknum;
73 #ifdef CONFIG_NETFILTER_INGRESS
74 if (reg->hooknum == NF_NETDEV_INGRESS) {
75 if (reg->dev && dev_net(reg->dev) == net)
76 return ®->dev->nf_hooks_ingress;
82 int nf_register_net_hook(struct net *net, const struct nf_hook_ops *reg)
84 struct nf_hook_entry __rcu **pp;
85 struct nf_hook_entry *entry, *p;
87 if (reg->pf == NFPROTO_NETDEV) {
88 #ifndef CONFIG_NETFILTER_INGRESS
89 if (reg->hooknum == NF_NETDEV_INGRESS)
92 if (reg->hooknum != NF_NETDEV_INGRESS ||
93 !reg->dev || dev_net(reg->dev) != net)
97 pp = nf_hook_entry_head(net, reg);
101 entry = kmalloc(sizeof(*entry), GFP_KERNEL);
105 entry->orig_ops = reg;
109 mutex_lock(&nf_hook_mutex);
111 /* Find the spot in the list */
112 while ((p = nf_entry_dereference(*pp)) != NULL) {
113 if (reg->priority < p->orig_ops->priority)
117 rcu_assign_pointer(entry->next, p);
118 rcu_assign_pointer(*pp, entry);
120 mutex_unlock(&nf_hook_mutex);
121 #ifdef CONFIG_NETFILTER_INGRESS
122 if (reg->pf == NFPROTO_NETDEV && reg->hooknum == NF_NETDEV_INGRESS)
123 net_inc_ingress_queue();
125 #ifdef HAVE_JUMP_LABEL
126 static_key_slow_inc(&nf_hooks_needed[reg->pf][reg->hooknum]);
130 EXPORT_SYMBOL(nf_register_net_hook);
132 void nf_unregister_net_hook(struct net *net, const struct nf_hook_ops *reg)
134 struct nf_hook_entry __rcu **pp;
135 struct nf_hook_entry *p;
137 pp = nf_hook_entry_head(net, reg);
138 if (WARN_ON_ONCE(!pp))
141 mutex_lock(&nf_hook_mutex);
142 while ((p = nf_entry_dereference(*pp)) != NULL) {
143 if (p->orig_ops == reg) {
144 rcu_assign_pointer(*pp, p->next);
149 mutex_unlock(&nf_hook_mutex);
151 WARN(1, "nf_unregister_net_hook: hook not found!\n");
154 #ifdef CONFIG_NETFILTER_INGRESS
155 if (reg->pf == NFPROTO_NETDEV && reg->hooknum == NF_NETDEV_INGRESS)
156 net_dec_ingress_queue();
158 #ifdef HAVE_JUMP_LABEL
159 static_key_slow_dec(&nf_hooks_needed[reg->pf][reg->hooknum]);
162 nf_queue_nf_hook_drop(net, p);
163 /* other cpu might still process nfqueue verdict that used reg */
167 EXPORT_SYMBOL(nf_unregister_net_hook);
169 int nf_register_net_hooks(struct net *net, const struct nf_hook_ops *reg,
175 for (i = 0; i < n; i++) {
176 err = nf_register_net_hook(net, ®[i]);
184 nf_unregister_net_hooks(net, reg, i);
187 EXPORT_SYMBOL(nf_register_net_hooks);
189 void nf_unregister_net_hooks(struct net *net, const struct nf_hook_ops *reg,
193 nf_unregister_net_hook(net, ®[n]);
195 EXPORT_SYMBOL(nf_unregister_net_hooks);
197 static LIST_HEAD(nf_hook_list);
199 static int _nf_register_hook(struct nf_hook_ops *reg)
201 struct net *net, *last;
205 ret = nf_register_net_hook(net, reg);
206 if (ret && ret != -ENOENT)
209 list_add_tail(®->list, &nf_hook_list);
217 nf_unregister_net_hook(net, reg);
222 int nf_register_hook(struct nf_hook_ops *reg)
227 ret = _nf_register_hook(reg);
232 EXPORT_SYMBOL(nf_register_hook);
234 static void _nf_unregister_hook(struct nf_hook_ops *reg)
238 list_del(®->list);
240 nf_unregister_net_hook(net, reg);
243 void nf_unregister_hook(struct nf_hook_ops *reg)
246 _nf_unregister_hook(reg);
249 EXPORT_SYMBOL(nf_unregister_hook);
251 int nf_register_hooks(struct nf_hook_ops *reg, unsigned int n)
256 for (i = 0; i < n; i++) {
257 err = nf_register_hook(®[i]);
265 nf_unregister_hooks(reg, i);
268 EXPORT_SYMBOL(nf_register_hooks);
270 /* Caller MUST take rtnl_lock() */
271 int _nf_register_hooks(struct nf_hook_ops *reg, unsigned int n)
276 for (i = 0; i < n; i++) {
277 err = _nf_register_hook(®[i]);
285 _nf_unregister_hooks(reg, i);
288 EXPORT_SYMBOL(_nf_register_hooks);
290 void nf_unregister_hooks(struct nf_hook_ops *reg, unsigned int n)
293 nf_unregister_hook(®[n]);
295 EXPORT_SYMBOL(nf_unregister_hooks);
297 /* Caller MUST take rtnl_lock */
298 void _nf_unregister_hooks(struct nf_hook_ops *reg, unsigned int n)
301 _nf_unregister_hook(®[n]);
303 EXPORT_SYMBOL(_nf_unregister_hooks);
305 unsigned int nf_iterate(struct sk_buff *skb,
306 struct nf_hook_state *state,
307 struct nf_hook_entry **entryp)
309 unsigned int verdict;
312 * The caller must not block between calls to this
313 * function because of risk of continuing from deleted element.
316 if (state->thresh > (*entryp)->ops.priority) {
317 *entryp = rcu_dereference((*entryp)->next);
321 /* Optimization: we don't need to hold module
322 reference here, since function can't sleep. --RR */
324 verdict = (*entryp)->ops.hook((*entryp)->ops.priv, skb, state);
325 if (verdict != NF_ACCEPT) {
326 #ifdef CONFIG_NETFILTER_DEBUG
327 if (unlikely((verdict & NF_VERDICT_MASK)
329 NFDEBUG("Evil return from %p(%u).\n",
330 (*entryp)->ops.hook, state->hook);
331 *entryp = rcu_dereference((*entryp)->next);
335 if (verdict != NF_REPEAT)
339 *entryp = rcu_dereference((*entryp)->next);
345 /* Returns 1 if okfn() needs to be executed by the caller,
346 * -EPERM for NF_DROP, 0 otherwise. Caller must hold rcu_read_lock. */
347 int nf_hook_slow(struct sk_buff *skb, struct nf_hook_state *state)
349 struct nf_hook_entry *entry;
350 unsigned int verdict;
353 entry = rcu_dereference(state->hook_entries);
355 verdict = nf_iterate(skb, state, &entry);
356 if (verdict == NF_ACCEPT || verdict == NF_STOP) {
358 } else if ((verdict & NF_VERDICT_MASK) == NF_DROP) {
360 ret = NF_DROP_GETERR(verdict);
363 } else if ((verdict & NF_VERDICT_MASK) == NF_QUEUE) {
364 ret = nf_queue(skb, state, &entry, verdict);
365 if (ret == 1 && entry)
368 /* Implicit handling for NF_STOLEN, as well as any other
369 * non conventional verdicts.
375 EXPORT_SYMBOL(nf_hook_slow);
378 int skb_make_writable(struct sk_buff *skb, unsigned int writable_len)
380 if (writable_len > skb->len)
383 /* Not exclusive use of packet? Must copy. */
384 if (!skb_cloned(skb)) {
385 if (writable_len <= skb_headlen(skb))
387 } else if (skb_clone_writable(skb, writable_len))
390 if (writable_len <= skb_headlen(skb))
393 writable_len -= skb_headlen(skb);
395 return !!__pskb_pull_tail(skb, writable_len);
397 EXPORT_SYMBOL(skb_make_writable);
399 /* This needs to be compiled in any case to avoid dependencies between the
400 * nfnetlink_queue code and nf_conntrack.
402 struct nfnl_ct_hook __rcu *nfnl_ct_hook __read_mostly;
403 EXPORT_SYMBOL_GPL(nfnl_ct_hook);
405 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
406 /* This does not belong here, but locally generated errors need it if connection
407 tracking in use: without this, connection may not be in hash table, and hence
408 manufactured ICMP or RST packets will not be associated with it. */
409 void (*ip_ct_attach)(struct sk_buff *, const struct sk_buff *)
411 EXPORT_SYMBOL(ip_ct_attach);
413 void nf_ct_attach(struct sk_buff *new, const struct sk_buff *skb)
415 void (*attach)(struct sk_buff *, const struct sk_buff *);
419 attach = rcu_dereference(ip_ct_attach);
425 EXPORT_SYMBOL(nf_ct_attach);
427 void (*nf_ct_destroy)(struct nf_conntrack *) __rcu __read_mostly;
428 EXPORT_SYMBOL(nf_ct_destroy);
430 void nf_conntrack_destroy(struct nf_conntrack *nfct)
432 void (*destroy)(struct nf_conntrack *);
435 destroy = rcu_dereference(nf_ct_destroy);
436 BUG_ON(destroy == NULL);
440 EXPORT_SYMBOL(nf_conntrack_destroy);
442 /* Built-in default zone used e.g. by modules. */
443 const struct nf_conntrack_zone nf_ct_zone_dflt = {
444 .id = NF_CT_DEFAULT_ZONE_ID,
445 .dir = NF_CT_DEFAULT_ZONE_DIR,
447 EXPORT_SYMBOL_GPL(nf_ct_zone_dflt);
448 #endif /* CONFIG_NF_CONNTRACK */
450 #ifdef CONFIG_NF_NAT_NEEDED
451 void (*nf_nat_decode_session_hook)(struct sk_buff *, struct flowi *);
452 EXPORT_SYMBOL(nf_nat_decode_session_hook);
455 static int nf_register_hook_list(struct net *net)
457 struct nf_hook_ops *elem;
461 list_for_each_entry(elem, &nf_hook_list, list) {
462 ret = nf_register_net_hook(net, elem);
463 if (ret && ret != -ENOENT)
470 list_for_each_entry_continue_reverse(elem, &nf_hook_list, list)
471 nf_unregister_net_hook(net, elem);
476 static void nf_unregister_hook_list(struct net *net)
478 struct nf_hook_ops *elem;
481 list_for_each_entry(elem, &nf_hook_list, list)
482 nf_unregister_net_hook(net, elem);
486 static int __net_init netfilter_net_init(struct net *net)
490 for (i = 0; i < ARRAY_SIZE(net->nf.hooks); i++) {
491 for (h = 0; h < NF_MAX_HOOKS; h++)
492 RCU_INIT_POINTER(net->nf.hooks[i][h], NULL);
495 #ifdef CONFIG_PROC_FS
496 net->nf.proc_netfilter = proc_net_mkdir(net, "netfilter",
498 if (!net->nf.proc_netfilter) {
499 if (!net_eq(net, &init_net))
500 pr_err("cannot create netfilter proc entry");
505 ret = nf_register_hook_list(net);
507 remove_proc_entry("netfilter", net->proc_net);
512 static void __net_exit netfilter_net_exit(struct net *net)
514 nf_unregister_hook_list(net);
515 remove_proc_entry("netfilter", net->proc_net);
518 static struct pernet_operations netfilter_net_ops = {
519 .init = netfilter_net_init,
520 .exit = netfilter_net_exit,
523 int __init netfilter_init(void)
527 ret = register_pernet_subsys(&netfilter_net_ops);
531 ret = netfilter_log_init();
537 unregister_pernet_subsys(&netfilter_net_ops);