1 /* NAT for netfilter; shared with compatibility layer. */
3 /* (C) 1999-2001 Paul `Rusty' Russell
4 * (C) 2002-2006 Netfilter Core Team <coreteam@netfilter.org>
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
11 #include <linux/module.h>
12 #include <linux/types.h>
13 #include <linux/timer.h>
14 #include <linux/skbuff.h>
15 #include <net/checksum.h>
18 #include <net/tcp.h> /* For tcp_prot in getorigdst */
19 #include <linux/icmp.h>
20 #include <linux/udp.h>
21 #include <linux/jhash.h>
23 #include <linux/netfilter_ipv4.h>
24 #include <net/netfilter/nf_conntrack.h>
25 #include <net/netfilter/nf_conntrack_core.h>
26 #include <net/netfilter/nf_nat.h>
27 #include <net/netfilter/nf_nat_protocol.h>
28 #include <net/netfilter/nf_nat_core.h>
29 #include <net/netfilter/nf_nat_helper.h>
30 #include <net/netfilter/nf_conntrack_helper.h>
31 #include <net/netfilter/nf_conntrack_l3proto.h>
32 #include <net/netfilter/nf_conntrack_l4proto.h>
34 static DEFINE_SPINLOCK(nf_nat_lock
);
36 static struct nf_conntrack_l3proto
*l3proto __read_mostly
;
38 #define MAX_IP_NAT_PROTO 256
39 static const struct nf_nat_protocol
*nf_nat_protos
[MAX_IP_NAT_PROTO
]
42 static inline const struct nf_nat_protocol
*
43 __nf_nat_proto_find(u_int8_t protonum
)
45 return rcu_dereference(nf_nat_protos
[protonum
]);
48 const struct nf_nat_protocol
*
49 nf_nat_proto_find_get(u_int8_t protonum
)
51 const struct nf_nat_protocol
*p
;
54 p
= __nf_nat_proto_find(protonum
);
55 if (!try_module_get(p
->me
))
56 p
= &nf_nat_unknown_protocol
;
61 EXPORT_SYMBOL_GPL(nf_nat_proto_find_get
);
64 nf_nat_proto_put(const struct nf_nat_protocol
*p
)
68 EXPORT_SYMBOL_GPL(nf_nat_proto_put
);
70 /* We keep an extra hash for each conntrack, for fast searching. */
71 static inline unsigned int
72 hash_by_src(const struct net
*net
, const struct nf_conntrack_tuple
*tuple
)
76 /* Original src, to ensure we map it consistently if poss. */
77 hash
= jhash_3words((__force u32
)tuple
->src
.u3
.ip
,
78 (__force u32
)tuple
->src
.u
.all
,
79 tuple
->dst
.protonum
, 0);
80 return ((u64
)hash
* net
->ipv4
.nat_htable_size
) >> 32;
83 /* Is this tuple already taken? (not by us) */
85 nf_nat_used_tuple(const struct nf_conntrack_tuple
*tuple
,
86 const struct nf_conn
*ignored_conntrack
)
88 /* Conntrack tracking doesn't keep track of outgoing tuples; only
89 incoming ones. NAT means they don't have a fixed mapping,
90 so we invert the tuple and look for the incoming reply.
92 We could keep a separate hash if this proves too slow. */
93 struct nf_conntrack_tuple reply
;
95 nf_ct_invert_tuplepr(&reply
, tuple
);
96 return nf_conntrack_tuple_taken(&reply
, ignored_conntrack
);
98 EXPORT_SYMBOL(nf_nat_used_tuple
);
100 /* If we source map this tuple so reply looks like reply_tuple, will
101 * that meet the constraints of range. */
103 in_range(const struct nf_conntrack_tuple
*tuple
,
104 const struct nf_nat_range
*range
)
106 const struct nf_nat_protocol
*proto
;
109 /* If we are supposed to map IPs, then we must be in the
110 range specified, otherwise let this drag us onto a new src IP. */
111 if (range
->flags
& IP_NAT_RANGE_MAP_IPS
) {
112 if (ntohl(tuple
->src
.u3
.ip
) < ntohl(range
->min_ip
) ||
113 ntohl(tuple
->src
.u3
.ip
) > ntohl(range
->max_ip
))
118 proto
= __nf_nat_proto_find(tuple
->dst
.protonum
);
119 if (!(range
->flags
& IP_NAT_RANGE_PROTO_SPECIFIED
) ||
120 proto
->in_range(tuple
, IP_NAT_MANIP_SRC
,
121 &range
->min
, &range
->max
))
129 same_src(const struct nf_conn
*ct
,
130 const struct nf_conntrack_tuple
*tuple
)
132 const struct nf_conntrack_tuple
*t
;
134 t
= &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
;
135 return (t
->dst
.protonum
== tuple
->dst
.protonum
&&
136 t
->src
.u3
.ip
== tuple
->src
.u3
.ip
&&
137 t
->src
.u
.all
== tuple
->src
.u
.all
);
140 /* Only called for SRC manip */
142 find_appropriate_src(struct net
*net
,
143 const struct nf_conntrack_tuple
*tuple
,
144 struct nf_conntrack_tuple
*result
,
145 const struct nf_nat_range
*range
)
147 unsigned int h
= hash_by_src(net
, tuple
);
148 const struct nf_conn_nat
*nat
;
149 const struct nf_conn
*ct
;
150 const struct hlist_node
*n
;
153 hlist_for_each_entry_rcu(nat
, n
, &net
->ipv4
.nat_bysource
[h
], bysource
) {
155 if (same_src(ct
, tuple
)) {
156 /* Copy source part from reply tuple. */
157 nf_ct_invert_tuplepr(result
,
158 &ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
);
159 result
->dst
= tuple
->dst
;
161 if (in_range(result
, range
)) {
171 /* For [FUTURE] fragmentation handling, we want the least-used
172 src-ip/dst-ip/proto triple. Fairness doesn't come into it. Thus
173 if the range specifies 1.2.3.4 ports 10000-10005 and 1.2.3.5 ports
174 1-65535, we don't do pro-rata allocation based on ports; we choose
175 the ip with the lowest src-ip/dst-ip/proto usage.
178 find_best_ips_proto(struct nf_conntrack_tuple
*tuple
,
179 const struct nf_nat_range
*range
,
180 const struct nf_conn
*ct
,
181 enum nf_nat_manip_type maniptype
)
185 u_int32_t minip
, maxip
, j
;
187 /* No IP mapping? Do nothing. */
188 if (!(range
->flags
& IP_NAT_RANGE_MAP_IPS
))
191 if (maniptype
== IP_NAT_MANIP_SRC
)
192 var_ipp
= &tuple
->src
.u3
.ip
;
194 var_ipp
= &tuple
->dst
.u3
.ip
;
196 /* Fast path: only one choice. */
197 if (range
->min_ip
== range
->max_ip
) {
198 *var_ipp
= range
->min_ip
;
202 /* Hashing source and destination IPs gives a fairly even
203 * spread in practice (if there are a small number of IPs
204 * involved, there usually aren't that many connections
205 * anyway). The consistency means that servers see the same
206 * client coming from the same IP (some Internet Banking sites
207 * like this), even across reboots. */
208 minip
= ntohl(range
->min_ip
);
209 maxip
= ntohl(range
->max_ip
);
210 j
= jhash_2words((__force u32
)tuple
->src
.u3
.ip
,
211 range
->flags
& IP_NAT_RANGE_PERSISTENT
?
212 0 : (__force u32
)tuple
->dst
.u3
.ip
, 0);
213 j
= ((u64
)j
* (maxip
- minip
+ 1)) >> 32;
214 *var_ipp
= htonl(minip
+ j
);
217 /* Manipulate the tuple into the range given. For NF_INET_POST_ROUTING,
218 * we change the source to map into the range. For NF_INET_PRE_ROUTING
219 * and NF_INET_LOCAL_OUT, we change the destination to map into the
220 * range. It might not be possible to get a unique tuple, but we try.
221 * At worst (or if we race), we will end up with a final duplicate in
222 * __ip_conntrack_confirm and drop the packet. */
224 get_unique_tuple(struct nf_conntrack_tuple
*tuple
,
225 const struct nf_conntrack_tuple
*orig_tuple
,
226 const struct nf_nat_range
*range
,
228 enum nf_nat_manip_type maniptype
)
230 struct net
*net
= nf_ct_net(ct
);
231 const struct nf_nat_protocol
*proto
;
233 /* 1) If this srcip/proto/src-proto-part is currently mapped,
234 and that same mapping gives a unique tuple within the given
237 This is only required for source (ie. NAT/masq) mappings.
238 So far, we don't do local source mappings, so multiple
239 manips not an issue. */
240 if (maniptype
== IP_NAT_MANIP_SRC
&&
241 !(range
->flags
& IP_NAT_RANGE_PROTO_RANDOM
)) {
242 if (find_appropriate_src(net
, orig_tuple
, tuple
, range
)) {
243 pr_debug("get_unique_tuple: Found current src map\n");
244 if (!nf_nat_used_tuple(tuple
, ct
))
249 /* 2) Select the least-used IP/proto combination in the given
251 *tuple
= *orig_tuple
;
252 find_best_ips_proto(tuple
, range
, ct
, maniptype
);
254 /* 3) The per-protocol part of the manip is made to map into
255 the range to make a unique tuple. */
258 proto
= __nf_nat_proto_find(orig_tuple
->dst
.protonum
);
260 /* Change protocol info to have some randomization */
261 if (range
->flags
& IP_NAT_RANGE_PROTO_RANDOM
) {
262 proto
->unique_tuple(tuple
, range
, maniptype
, ct
);
266 /* Only bother mapping if it's not already in range and unique */
267 if ((!(range
->flags
& IP_NAT_RANGE_PROTO_SPECIFIED
) ||
268 proto
->in_range(tuple
, maniptype
, &range
->min
, &range
->max
)) &&
269 !nf_nat_used_tuple(tuple
, ct
))
272 /* Last change: get protocol to try to obtain unique tuple. */
273 proto
->unique_tuple(tuple
, range
, maniptype
, ct
);
279 nf_nat_setup_info(struct nf_conn
*ct
,
280 const struct nf_nat_range
*range
,
281 enum nf_nat_manip_type maniptype
)
283 struct net
*net
= nf_ct_net(ct
);
284 struct nf_conntrack_tuple curr_tuple
, new_tuple
;
285 struct nf_conn_nat
*nat
;
286 int have_to_hash
= !(ct
->status
& IPS_NAT_DONE_MASK
);
288 /* nat helper or nfctnetlink also setup binding */
291 nat
= nf_ct_ext_add(ct
, NF_CT_EXT_NAT
, GFP_ATOMIC
);
293 pr_debug("failed to add NAT extension\n");
298 NF_CT_ASSERT(maniptype
== IP_NAT_MANIP_SRC
||
299 maniptype
== IP_NAT_MANIP_DST
);
300 BUG_ON(nf_nat_initialized(ct
, maniptype
));
302 /* What we've got will look like inverse of reply. Normally
303 this is what is in the conntrack, except for prior
304 manipulations (future optimization: if num_manips == 0,
306 conntrack->tuplehash[IP_CT_DIR_ORIGINAL].tuple) */
307 nf_ct_invert_tuplepr(&curr_tuple
,
308 &ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
);
310 get_unique_tuple(&new_tuple
, &curr_tuple
, range
, ct
, maniptype
);
312 if (!nf_ct_tuple_equal(&new_tuple
, &curr_tuple
)) {
313 struct nf_conntrack_tuple reply
;
315 /* Alter conntrack table so will recognize replies. */
316 nf_ct_invert_tuplepr(&reply
, &new_tuple
);
317 nf_conntrack_alter_reply(ct
, &reply
);
319 /* Non-atomic: we own this at the moment. */
320 if (maniptype
== IP_NAT_MANIP_SRC
)
321 ct
->status
|= IPS_SRC_NAT
;
323 ct
->status
|= IPS_DST_NAT
;
326 /* Place in source hash if this is the first time. */
328 unsigned int srchash
;
330 srchash
= hash_by_src(net
, &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
);
331 spin_lock_bh(&nf_nat_lock
);
332 /* nf_conntrack_alter_reply might re-allocate exntension aera */
335 hlist_add_head_rcu(&nat
->bysource
,
336 &net
->ipv4
.nat_bysource
[srchash
]);
337 spin_unlock_bh(&nf_nat_lock
);
341 if (maniptype
== IP_NAT_MANIP_DST
)
342 set_bit(IPS_DST_NAT_DONE_BIT
, &ct
->status
);
344 set_bit(IPS_SRC_NAT_DONE_BIT
, &ct
->status
);
348 EXPORT_SYMBOL(nf_nat_setup_info
);
350 /* Returns true if succeeded. */
352 manip_pkt(u_int16_t proto
,
354 unsigned int iphdroff
,
355 const struct nf_conntrack_tuple
*target
,
356 enum nf_nat_manip_type maniptype
)
359 const struct nf_nat_protocol
*p
;
361 if (!skb_make_writable(skb
, iphdroff
+ sizeof(*iph
)))
364 iph
= (void *)skb
->data
+ iphdroff
;
366 /* Manipulate protcol part. */
368 /* rcu_read_lock()ed by nf_hook_slow */
369 p
= __nf_nat_proto_find(proto
);
370 if (!p
->manip_pkt(skb
, iphdroff
, target
, maniptype
))
373 iph
= (void *)skb
->data
+ iphdroff
;
375 if (maniptype
== IP_NAT_MANIP_SRC
) {
376 csum_replace4(&iph
->check
, iph
->saddr
, target
->src
.u3
.ip
);
377 iph
->saddr
= target
->src
.u3
.ip
;
379 csum_replace4(&iph
->check
, iph
->daddr
, target
->dst
.u3
.ip
);
380 iph
->daddr
= target
->dst
.u3
.ip
;
385 /* Do packet manipulations according to nf_nat_setup_info. */
386 unsigned int nf_nat_packet(struct nf_conn
*ct
,
387 enum ip_conntrack_info ctinfo
,
388 unsigned int hooknum
,
391 enum ip_conntrack_dir dir
= CTINFO2DIR(ctinfo
);
392 unsigned long statusbit
;
393 enum nf_nat_manip_type mtype
= HOOK2MANIP(hooknum
);
395 if (mtype
== IP_NAT_MANIP_SRC
)
396 statusbit
= IPS_SRC_NAT
;
398 statusbit
= IPS_DST_NAT
;
400 /* Invert if this is reply dir. */
401 if (dir
== IP_CT_DIR_REPLY
)
402 statusbit
^= IPS_NAT_MASK
;
404 /* Non-atomic: these bits don't change. */
405 if (ct
->status
& statusbit
) {
406 struct nf_conntrack_tuple target
;
408 /* We are aiming to look like inverse of other direction. */
409 nf_ct_invert_tuplepr(&target
, &ct
->tuplehash
[!dir
].tuple
);
411 if (!manip_pkt(target
.dst
.protonum
, skb
, 0, &target
, mtype
))
416 EXPORT_SYMBOL_GPL(nf_nat_packet
);
418 /* Dir is direction ICMP is coming from (opposite to packet it contains) */
419 int nf_nat_icmp_reply_translation(struct nf_conn
*ct
,
420 enum ip_conntrack_info ctinfo
,
421 unsigned int hooknum
,
428 const struct nf_conntrack_l4proto
*l4proto
;
429 struct nf_conntrack_tuple inner
, target
;
430 int hdrlen
= ip_hdrlen(skb
);
431 enum ip_conntrack_dir dir
= CTINFO2DIR(ctinfo
);
432 unsigned long statusbit
;
433 enum nf_nat_manip_type manip
= HOOK2MANIP(hooknum
);
435 if (!skb_make_writable(skb
, hdrlen
+ sizeof(*inside
)))
438 inside
= (void *)skb
->data
+ ip_hdrlen(skb
);
440 /* We're actually going to mangle it beyond trivial checksum
441 adjustment, so make sure the current checksum is correct. */
442 if (nf_ip_checksum(skb
, hooknum
, hdrlen
, 0))
445 /* Must be RELATED */
446 NF_CT_ASSERT(skb
->nfctinfo
== IP_CT_RELATED
||
447 skb
->nfctinfo
== IP_CT_RELATED
+IP_CT_IS_REPLY
);
449 /* Redirects on non-null nats must be dropped, else they'll
450 start talking to each other without our translation, and be
452 if (inside
->icmp
.type
== ICMP_REDIRECT
) {
453 /* If NAT isn't finished, assume it and drop. */
454 if ((ct
->status
& IPS_NAT_DONE_MASK
) != IPS_NAT_DONE_MASK
)
457 if (ct
->status
& IPS_NAT_MASK
)
461 pr_debug("icmp_reply_translation: translating error %p manip %u "
462 "dir %s\n", skb
, manip
,
463 dir
== IP_CT_DIR_ORIGINAL
? "ORIG" : "REPLY");
465 /* rcu_read_lock()ed by nf_hook_slow */
466 l4proto
= __nf_ct_l4proto_find(PF_INET
, inside
->ip
.protocol
);
468 if (!nf_ct_get_tuple(skb
,
469 ip_hdrlen(skb
) + sizeof(struct icmphdr
),
471 sizeof(struct icmphdr
) + inside
->ip
.ihl
* 4),
474 &inner
, l3proto
, l4proto
))
477 /* Change inner back to look like incoming packet. We do the
478 opposite manip on this hook to normal, because it might not
479 pass all hooks (locally-generated ICMP). Consider incoming
480 packet: PREROUTING (DST manip), routing produces ICMP, goes
481 through POSTROUTING (which must correct the DST manip). */
482 if (!manip_pkt(inside
->ip
.protocol
, skb
,
483 ip_hdrlen(skb
) + sizeof(inside
->icmp
),
484 &ct
->tuplehash
[!dir
].tuple
,
488 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
) {
489 /* Reloading "inside" here since manip_pkt inner. */
490 inside
= (void *)skb
->data
+ ip_hdrlen(skb
);
491 inside
->icmp
.checksum
= 0;
492 inside
->icmp
.checksum
=
493 csum_fold(skb_checksum(skb
, hdrlen
,
494 skb
->len
- hdrlen
, 0));
497 /* Change outer to look the reply to an incoming packet
498 * (proto 0 means don't invert per-proto part). */
499 if (manip
== IP_NAT_MANIP_SRC
)
500 statusbit
= IPS_SRC_NAT
;
502 statusbit
= IPS_DST_NAT
;
504 /* Invert if this is reply dir. */
505 if (dir
== IP_CT_DIR_REPLY
)
506 statusbit
^= IPS_NAT_MASK
;
508 if (ct
->status
& statusbit
) {
509 nf_ct_invert_tuplepr(&target
, &ct
->tuplehash
[!dir
].tuple
);
510 if (!manip_pkt(0, skb
, 0, &target
, manip
))
516 EXPORT_SYMBOL_GPL(nf_nat_icmp_reply_translation
);
518 /* Protocol registration. */
519 int nf_nat_protocol_register(const struct nf_nat_protocol
*proto
)
523 spin_lock_bh(&nf_nat_lock
);
524 if (nf_nat_protos
[proto
->protonum
] != &nf_nat_unknown_protocol
) {
528 rcu_assign_pointer(nf_nat_protos
[proto
->protonum
], proto
);
530 spin_unlock_bh(&nf_nat_lock
);
533 EXPORT_SYMBOL(nf_nat_protocol_register
);
535 /* Noone stores the protocol anywhere; simply delete it. */
536 void nf_nat_protocol_unregister(const struct nf_nat_protocol
*proto
)
538 spin_lock_bh(&nf_nat_lock
);
539 rcu_assign_pointer(nf_nat_protos
[proto
->protonum
],
540 &nf_nat_unknown_protocol
);
541 spin_unlock_bh(&nf_nat_lock
);
544 EXPORT_SYMBOL(nf_nat_protocol_unregister
);
546 /* Noone using conntrack by the time this called. */
547 static void nf_nat_cleanup_conntrack(struct nf_conn
*ct
)
549 struct nf_conn_nat
*nat
= nf_ct_ext_find(ct
, NF_CT_EXT_NAT
);
551 if (nat
== NULL
|| nat
->ct
== NULL
)
554 NF_CT_ASSERT(nat
->ct
->status
& IPS_NAT_DONE_MASK
);
556 spin_lock_bh(&nf_nat_lock
);
557 hlist_del_rcu(&nat
->bysource
);
558 spin_unlock_bh(&nf_nat_lock
);
561 static void nf_nat_move_storage(void *new, void *old
)
563 struct nf_conn_nat
*new_nat
= new;
564 struct nf_conn_nat
*old_nat
= old
;
565 struct nf_conn
*ct
= old_nat
->ct
;
567 if (!ct
|| !(ct
->status
& IPS_NAT_DONE_MASK
))
570 spin_lock_bh(&nf_nat_lock
);
572 hlist_replace_rcu(&old_nat
->bysource
, &new_nat
->bysource
);
573 spin_unlock_bh(&nf_nat_lock
);
576 static struct nf_ct_ext_type nat_extend __read_mostly
= {
577 .len
= sizeof(struct nf_conn_nat
),
578 .align
= __alignof__(struct nf_conn_nat
),
579 .destroy
= nf_nat_cleanup_conntrack
,
580 .move
= nf_nat_move_storage
,
582 .flags
= NF_CT_EXT_F_PREALLOC
,
585 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
587 #include <linux/netfilter/nfnetlink.h>
588 #include <linux/netfilter/nfnetlink_conntrack.h>
590 static const struct nla_policy protonat_nla_policy
[CTA_PROTONAT_MAX
+1] = {
591 [CTA_PROTONAT_PORT_MIN
] = { .type
= NLA_U16
},
592 [CTA_PROTONAT_PORT_MAX
] = { .type
= NLA_U16
},
595 static int nfnetlink_parse_nat_proto(struct nlattr
*attr
,
596 const struct nf_conn
*ct
,
597 struct nf_nat_range
*range
)
599 struct nlattr
*tb
[CTA_PROTONAT_MAX
+1];
600 const struct nf_nat_protocol
*npt
;
603 err
= nla_parse_nested(tb
, CTA_PROTONAT_MAX
, attr
, protonat_nla_policy
);
607 npt
= nf_nat_proto_find_get(nf_ct_protonum(ct
));
608 if (npt
->nlattr_to_range
)
609 err
= npt
->nlattr_to_range(tb
, range
);
610 nf_nat_proto_put(npt
);
614 static const struct nla_policy nat_nla_policy
[CTA_NAT_MAX
+1] = {
615 [CTA_NAT_MINIP
] = { .type
= NLA_U32
},
616 [CTA_NAT_MAXIP
] = { .type
= NLA_U32
},
620 nfnetlink_parse_nat(const struct nlattr
*nat
,
621 const struct nf_conn
*ct
, struct nf_nat_range
*range
)
623 struct nlattr
*tb
[CTA_NAT_MAX
+1];
626 memset(range
, 0, sizeof(*range
));
628 err
= nla_parse_nested(tb
, CTA_NAT_MAX
, nat
, nat_nla_policy
);
632 if (tb
[CTA_NAT_MINIP
])
633 range
->min_ip
= nla_get_be32(tb
[CTA_NAT_MINIP
]);
635 if (!tb
[CTA_NAT_MAXIP
])
636 range
->max_ip
= range
->min_ip
;
638 range
->max_ip
= nla_get_be32(tb
[CTA_NAT_MAXIP
]);
641 range
->flags
|= IP_NAT_RANGE_MAP_IPS
;
643 if (!tb
[CTA_NAT_PROTO
])
646 err
= nfnetlink_parse_nat_proto(tb
[CTA_NAT_PROTO
], ct
, range
);
654 nfnetlink_parse_nat_setup(struct nf_conn
*ct
,
655 enum nf_nat_manip_type manip
,
656 const struct nlattr
*attr
)
658 struct nf_nat_range range
;
660 if (nfnetlink_parse_nat(attr
, ct
, &range
) < 0)
662 if (nf_nat_initialized(ct
, manip
))
665 return nf_nat_setup_info(ct
, &range
, manip
);
669 nfnetlink_parse_nat_setup(struct nf_conn
*ct
,
670 enum nf_nat_manip_type manip
,
671 const struct nlattr
*attr
)
677 static int __net_init
nf_nat_net_init(struct net
*net
)
679 /* Leave them the same for the moment. */
680 net
->ipv4
.nat_htable_size
= net
->ct
.htable_size
;
681 net
->ipv4
.nat_bysource
= nf_ct_alloc_hashtable(&net
->ipv4
.nat_htable_size
,
682 &net
->ipv4
.nat_vmalloced
, 0);
683 if (!net
->ipv4
.nat_bysource
)
688 /* Clear NAT section of all conntracks, in case we're loaded again. */
689 static int clean_nat(struct nf_conn
*i
, void *data
)
691 struct nf_conn_nat
*nat
= nfct_nat(i
);
695 memset(nat
, 0, sizeof(*nat
));
696 i
->status
&= ~(IPS_NAT_MASK
| IPS_NAT_DONE_MASK
| IPS_SEQ_ADJUST
);
700 static void __net_exit
nf_nat_net_exit(struct net
*net
)
702 nf_ct_iterate_cleanup(net
, &clean_nat
, NULL
);
704 nf_ct_free_hashtable(net
->ipv4
.nat_bysource
, net
->ipv4
.nat_vmalloced
,
705 net
->ipv4
.nat_htable_size
);
708 static struct pernet_operations nf_nat_net_ops
= {
709 .init
= nf_nat_net_init
,
710 .exit
= nf_nat_net_exit
,
713 static int __init
nf_nat_init(void)
718 need_ipv4_conntrack();
720 ret
= nf_ct_extend_register(&nat_extend
);
722 printk(KERN_ERR
"nf_nat_core: Unable to register extension\n");
726 ret
= register_pernet_subsys(&nf_nat_net_ops
);
730 /* Sew in builtin protocols. */
731 spin_lock_bh(&nf_nat_lock
);
732 for (i
= 0; i
< MAX_IP_NAT_PROTO
; i
++)
733 rcu_assign_pointer(nf_nat_protos
[i
], &nf_nat_unknown_protocol
);
734 rcu_assign_pointer(nf_nat_protos
[IPPROTO_TCP
], &nf_nat_protocol_tcp
);
735 rcu_assign_pointer(nf_nat_protos
[IPPROTO_UDP
], &nf_nat_protocol_udp
);
736 rcu_assign_pointer(nf_nat_protos
[IPPROTO_ICMP
], &nf_nat_protocol_icmp
);
737 spin_unlock_bh(&nf_nat_lock
);
739 /* Initialize fake conntrack so that NAT will skip it */
740 nf_conntrack_untracked
.status
|= IPS_NAT_DONE_MASK
;
742 l3proto
= nf_ct_l3proto_find_get((u_int16_t
)AF_INET
);
744 BUG_ON(nf_nat_seq_adjust_hook
!= NULL
);
745 rcu_assign_pointer(nf_nat_seq_adjust_hook
, nf_nat_seq_adjust
);
746 BUG_ON(nfnetlink_parse_nat_setup_hook
!= NULL
);
747 rcu_assign_pointer(nfnetlink_parse_nat_setup_hook
,
748 nfnetlink_parse_nat_setup
);
749 BUG_ON(nf_ct_nat_offset
!= NULL
);
750 rcu_assign_pointer(nf_ct_nat_offset
, nf_nat_get_offset
);
754 nf_ct_extend_unregister(&nat_extend
);
758 static void __exit
nf_nat_cleanup(void)
760 unregister_pernet_subsys(&nf_nat_net_ops
);
761 nf_ct_l3proto_put(l3proto
);
762 nf_ct_extend_unregister(&nat_extend
);
763 rcu_assign_pointer(nf_nat_seq_adjust_hook
, NULL
);
764 rcu_assign_pointer(nfnetlink_parse_nat_setup_hook
, NULL
);
765 rcu_assign_pointer(nf_ct_nat_offset
, NULL
);
769 MODULE_LICENSE("GPL");
770 MODULE_ALIAS("nf-nat-ipv4");
772 module_init(nf_nat_init
);
773 module_exit(nf_nat_cleanup
);