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 nf_hook_entry_init(entry
, reg
);
107 mutex_lock(&nf_hook_mutex
);
109 /* Find the spot in the list */
110 for (; (p
= nf_entry_dereference(*pp
)) != NULL
; pp
= &p
->next
) {
111 if (reg
->priority
< nf_hook_entry_priority(p
))
114 rcu_assign_pointer(entry
->next
, p
);
115 rcu_assign_pointer(*pp
, entry
);
117 mutex_unlock(&nf_hook_mutex
);
118 #ifdef CONFIG_NETFILTER_INGRESS
119 if (reg
->pf
== NFPROTO_NETDEV
&& reg
->hooknum
== NF_NETDEV_INGRESS
)
120 net_inc_ingress_queue();
122 #ifdef HAVE_JUMP_LABEL
123 static_key_slow_inc(&nf_hooks_needed
[reg
->pf
][reg
->hooknum
]);
127 EXPORT_SYMBOL(nf_register_net_hook
);
129 void nf_unregister_net_hook(struct net
*net
, const struct nf_hook_ops
*reg
)
131 struct nf_hook_entry __rcu
**pp
;
132 struct nf_hook_entry
*p
;
134 pp
= nf_hook_entry_head(net
, reg
);
135 if (WARN_ON_ONCE(!pp
))
138 mutex_lock(&nf_hook_mutex
);
139 for (; (p
= nf_entry_dereference(*pp
)) != NULL
; pp
= &p
->next
) {
140 if (nf_hook_entry_ops(p
) == reg
) {
141 rcu_assign_pointer(*pp
, p
->next
);
145 mutex_unlock(&nf_hook_mutex
);
147 WARN(1, "nf_unregister_net_hook: hook not found!\n");
150 #ifdef CONFIG_NETFILTER_INGRESS
151 if (reg
->pf
== NFPROTO_NETDEV
&& reg
->hooknum
== NF_NETDEV_INGRESS
)
152 net_dec_ingress_queue();
154 #ifdef HAVE_JUMP_LABEL
155 static_key_slow_dec(&nf_hooks_needed
[reg
->pf
][reg
->hooknum
]);
158 nf_queue_nf_hook_drop(net
, p
);
159 /* other cpu might still process nfqueue verdict that used reg */
163 EXPORT_SYMBOL(nf_unregister_net_hook
);
165 int nf_register_net_hooks(struct net
*net
, const struct nf_hook_ops
*reg
,
171 for (i
= 0; i
< n
; i
++) {
172 err
= nf_register_net_hook(net
, ®
[i
]);
180 nf_unregister_net_hooks(net
, reg
, i
);
183 EXPORT_SYMBOL(nf_register_net_hooks
);
185 void nf_unregister_net_hooks(struct net
*net
, const struct nf_hook_ops
*reg
,
189 nf_unregister_net_hook(net
, ®
[n
]);
191 EXPORT_SYMBOL(nf_unregister_net_hooks
);
193 static LIST_HEAD(nf_hook_list
);
195 static int _nf_register_hook(struct nf_hook_ops
*reg
)
197 struct net
*net
, *last
;
201 ret
= nf_register_net_hook(net
, reg
);
202 if (ret
&& ret
!= -ENOENT
)
205 list_add_tail(®
->list
, &nf_hook_list
);
213 nf_unregister_net_hook(net
, reg
);
218 int nf_register_hook(struct nf_hook_ops
*reg
)
223 ret
= _nf_register_hook(reg
);
228 EXPORT_SYMBOL(nf_register_hook
);
230 static void _nf_unregister_hook(struct nf_hook_ops
*reg
)
234 list_del(®
->list
);
236 nf_unregister_net_hook(net
, reg
);
239 void nf_unregister_hook(struct nf_hook_ops
*reg
)
242 _nf_unregister_hook(reg
);
245 EXPORT_SYMBOL(nf_unregister_hook
);
247 int nf_register_hooks(struct nf_hook_ops
*reg
, unsigned int n
)
252 for (i
= 0; i
< n
; i
++) {
253 err
= nf_register_hook(®
[i
]);
261 nf_unregister_hooks(reg
, i
);
264 EXPORT_SYMBOL(nf_register_hooks
);
266 /* Caller MUST take rtnl_lock() */
267 int _nf_register_hooks(struct nf_hook_ops
*reg
, unsigned int n
)
272 for (i
= 0; i
< n
; i
++) {
273 err
= _nf_register_hook(®
[i
]);
281 _nf_unregister_hooks(reg
, i
);
284 EXPORT_SYMBOL(_nf_register_hooks
);
286 void nf_unregister_hooks(struct nf_hook_ops
*reg
, unsigned int n
)
289 nf_unregister_hook(®
[n
]);
291 EXPORT_SYMBOL(nf_unregister_hooks
);
293 /* Caller MUST take rtnl_lock */
294 void _nf_unregister_hooks(struct nf_hook_ops
*reg
, unsigned int n
)
297 _nf_unregister_hook(®
[n
]);
299 EXPORT_SYMBOL(_nf_unregister_hooks
);
301 /* Returns 1 if okfn() needs to be executed by the caller,
302 * -EPERM for NF_DROP, 0 otherwise. Caller must hold rcu_read_lock. */
303 int nf_hook_slow(struct sk_buff
*skb
, struct nf_hook_state
*state
,
304 struct nf_hook_entry
*entry
)
306 unsigned int verdict
;
310 verdict
= nf_hook_entry_hookfn(entry
, skb
, state
);
311 switch (verdict
& NF_VERDICT_MASK
) {
313 entry
= rcu_dereference(entry
->next
);
317 ret
= NF_DROP_GETERR(verdict
);
322 ret
= nf_queue(skb
, state
, &entry
, verdict
);
323 if (ret
== 1 && entry
)
327 /* Implicit handling for NF_STOLEN, as well as any other
328 * non conventional verdicts.
336 EXPORT_SYMBOL(nf_hook_slow
);
339 int skb_make_writable(struct sk_buff
*skb
, unsigned int writable_len
)
341 if (writable_len
> skb
->len
)
344 /* Not exclusive use of packet? Must copy. */
345 if (!skb_cloned(skb
)) {
346 if (writable_len
<= skb_headlen(skb
))
348 } else if (skb_clone_writable(skb
, writable_len
))
351 if (writable_len
<= skb_headlen(skb
))
354 writable_len
-= skb_headlen(skb
);
356 return !!__pskb_pull_tail(skb
, writable_len
);
358 EXPORT_SYMBOL(skb_make_writable
);
360 /* This needs to be compiled in any case to avoid dependencies between the
361 * nfnetlink_queue code and nf_conntrack.
363 struct nfnl_ct_hook __rcu
*nfnl_ct_hook __read_mostly
;
364 EXPORT_SYMBOL_GPL(nfnl_ct_hook
);
366 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
367 /* This does not belong here, but locally generated errors need it if connection
368 tracking in use: without this, connection may not be in hash table, and hence
369 manufactured ICMP or RST packets will not be associated with it. */
370 void (*ip_ct_attach
)(struct sk_buff
*, const struct sk_buff
*)
372 EXPORT_SYMBOL(ip_ct_attach
);
374 void nf_ct_attach(struct sk_buff
*new, const struct sk_buff
*skb
)
376 void (*attach
)(struct sk_buff
*, const struct sk_buff
*);
380 attach
= rcu_dereference(ip_ct_attach
);
386 EXPORT_SYMBOL(nf_ct_attach
);
388 void (*nf_ct_destroy
)(struct nf_conntrack
*) __rcu __read_mostly
;
389 EXPORT_SYMBOL(nf_ct_destroy
);
391 void nf_conntrack_destroy(struct nf_conntrack
*nfct
)
393 void (*destroy
)(struct nf_conntrack
*);
396 destroy
= rcu_dereference(nf_ct_destroy
);
397 BUG_ON(destroy
== NULL
);
401 EXPORT_SYMBOL(nf_conntrack_destroy
);
403 /* Built-in default zone used e.g. by modules. */
404 const struct nf_conntrack_zone nf_ct_zone_dflt
= {
405 .id
= NF_CT_DEFAULT_ZONE_ID
,
406 .dir
= NF_CT_DEFAULT_ZONE_DIR
,
408 EXPORT_SYMBOL_GPL(nf_ct_zone_dflt
);
409 #endif /* CONFIG_NF_CONNTRACK */
411 #ifdef CONFIG_NF_NAT_NEEDED
412 void (*nf_nat_decode_session_hook
)(struct sk_buff
*, struct flowi
*);
413 EXPORT_SYMBOL(nf_nat_decode_session_hook
);
416 static int nf_register_hook_list(struct net
*net
)
418 struct nf_hook_ops
*elem
;
422 list_for_each_entry(elem
, &nf_hook_list
, list
) {
423 ret
= nf_register_net_hook(net
, elem
);
424 if (ret
&& ret
!= -ENOENT
)
431 list_for_each_entry_continue_reverse(elem
, &nf_hook_list
, list
)
432 nf_unregister_net_hook(net
, elem
);
437 static void nf_unregister_hook_list(struct net
*net
)
439 struct nf_hook_ops
*elem
;
442 list_for_each_entry(elem
, &nf_hook_list
, list
)
443 nf_unregister_net_hook(net
, elem
);
447 static int __net_init
netfilter_net_init(struct net
*net
)
451 for (i
= 0; i
< ARRAY_SIZE(net
->nf
.hooks
); i
++) {
452 for (h
= 0; h
< NF_MAX_HOOKS
; h
++)
453 RCU_INIT_POINTER(net
->nf
.hooks
[i
][h
], NULL
);
456 #ifdef CONFIG_PROC_FS
457 net
->nf
.proc_netfilter
= proc_net_mkdir(net
, "netfilter",
459 if (!net
->nf
.proc_netfilter
) {
460 if (!net_eq(net
, &init_net
))
461 pr_err("cannot create netfilter proc entry");
466 ret
= nf_register_hook_list(net
);
468 remove_proc_entry("netfilter", net
->proc_net
);
473 static void __net_exit
netfilter_net_exit(struct net
*net
)
475 nf_unregister_hook_list(net
);
476 remove_proc_entry("netfilter", net
->proc_net
);
479 static struct pernet_operations netfilter_net_ops
= {
480 .init
= netfilter_net_init
,
481 .exit
= netfilter_net_exit
,
484 int __init
netfilter_init(void)
488 ret
= register_pernet_subsys(&netfilter_net_ops
);
492 ret
= netfilter_log_init();
498 unregister_pernet_subsys(&netfilter_net_ops
);