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 <linux/gfp.h>
16 #include <net/checksum.h>
19 #include <net/tcp.h> /* For tcp_prot in getorigdst */
20 #include <linux/icmp.h>
21 #include <linux/udp.h>
22 #include <linux/jhash.h>
24 #include <linux/netfilter_ipv4.h>
25 #include <net/netfilter/nf_conntrack.h>
26 #include <net/netfilter/nf_conntrack_core.h>
27 #include <net/netfilter/nf_nat.h>
28 #include <net/netfilter/nf_nat_protocol.h>
29 #include <net/netfilter/nf_nat_core.h>
30 #include <net/netfilter/nf_nat_helper.h>
31 #include <net/netfilter/nf_conntrack_helper.h>
32 #include <net/netfilter/nf_conntrack_l3proto.h>
33 #include <net/netfilter/nf_conntrack_l4proto.h>
34 #include <net/netfilter/nf_conntrack_zones.h>
36 static DEFINE_SPINLOCK(nf_nat_lock
);
38 static struct nf_conntrack_l3proto
*l3proto __read_mostly
;
40 #define MAX_IP_NAT_PROTO 256
41 static const struct nf_nat_protocol
*nf_nat_protos
[MAX_IP_NAT_PROTO
]
44 static inline const struct nf_nat_protocol
*
45 __nf_nat_proto_find(u_int8_t protonum
)
47 return rcu_dereference(nf_nat_protos
[protonum
]);
50 const struct nf_nat_protocol
*
51 nf_nat_proto_find_get(u_int8_t protonum
)
53 const struct nf_nat_protocol
*p
;
56 p
= __nf_nat_proto_find(protonum
);
57 if (!try_module_get(p
->me
))
58 p
= &nf_nat_unknown_protocol
;
63 EXPORT_SYMBOL_GPL(nf_nat_proto_find_get
);
66 nf_nat_proto_put(const struct nf_nat_protocol
*p
)
70 EXPORT_SYMBOL_GPL(nf_nat_proto_put
);
72 /* We keep an extra hash for each conntrack, for fast searching. */
73 static inline unsigned int
74 hash_by_src(const struct net
*net
, u16 zone
,
75 const struct nf_conntrack_tuple
*tuple
)
79 /* Original src, to ensure we map it consistently if poss. */
80 hash
= jhash_3words((__force u32
)tuple
->src
.u3
.ip
,
81 (__force u32
)tuple
->src
.u
.all
^ zone
,
82 tuple
->dst
.protonum
, 0);
83 return ((u64
)hash
* net
->ipv4
.nat_htable_size
) >> 32;
86 /* Is this tuple already taken? (not by us) */
88 nf_nat_used_tuple(const struct nf_conntrack_tuple
*tuple
,
89 const struct nf_conn
*ignored_conntrack
)
91 /* Conntrack tracking doesn't keep track of outgoing tuples; only
92 incoming ones. NAT means they don't have a fixed mapping,
93 so we invert the tuple and look for the incoming reply.
95 We could keep a separate hash if this proves too slow. */
96 struct nf_conntrack_tuple reply
;
98 nf_ct_invert_tuplepr(&reply
, tuple
);
99 return nf_conntrack_tuple_taken(&reply
, ignored_conntrack
);
101 EXPORT_SYMBOL(nf_nat_used_tuple
);
103 /* If we source map this tuple so reply looks like reply_tuple, will
104 * that meet the constraints of range. */
106 in_range(const struct nf_conntrack_tuple
*tuple
,
107 const struct nf_nat_range
*range
)
109 const struct nf_nat_protocol
*proto
;
112 /* If we are supposed to map IPs, then we must be in the
113 range specified, otherwise let this drag us onto a new src IP. */
114 if (range
->flags
& IP_NAT_RANGE_MAP_IPS
) {
115 if (ntohl(tuple
->src
.u3
.ip
) < ntohl(range
->min_ip
) ||
116 ntohl(tuple
->src
.u3
.ip
) > ntohl(range
->max_ip
))
121 proto
= __nf_nat_proto_find(tuple
->dst
.protonum
);
122 if (!(range
->flags
& IP_NAT_RANGE_PROTO_SPECIFIED
) ||
123 proto
->in_range(tuple
, IP_NAT_MANIP_SRC
,
124 &range
->min
, &range
->max
))
132 same_src(const struct nf_conn
*ct
,
133 const struct nf_conntrack_tuple
*tuple
)
135 const struct nf_conntrack_tuple
*t
;
137 t
= &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
;
138 return (t
->dst
.protonum
== tuple
->dst
.protonum
&&
139 t
->src
.u3
.ip
== tuple
->src
.u3
.ip
&&
140 t
->src
.u
.all
== tuple
->src
.u
.all
);
143 /* Only called for SRC manip */
145 find_appropriate_src(struct net
*net
, u16 zone
,
146 const struct nf_conntrack_tuple
*tuple
,
147 struct nf_conntrack_tuple
*result
,
148 const struct nf_nat_range
*range
)
150 unsigned int h
= hash_by_src(net
, zone
, tuple
);
151 const struct nf_conn_nat
*nat
;
152 const struct nf_conn
*ct
;
153 const struct hlist_node
*n
;
156 hlist_for_each_entry_rcu(nat
, n
, &net
->ipv4
.nat_bysource
[h
], bysource
) {
158 if (same_src(ct
, tuple
) && nf_ct_zone(ct
) == zone
) {
159 /* Copy source part from reply tuple. */
160 nf_ct_invert_tuplepr(result
,
161 &ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
);
162 result
->dst
= tuple
->dst
;
164 if (in_range(result
, range
)) {
174 /* For [FUTURE] fragmentation handling, we want the least-used
175 src-ip/dst-ip/proto triple. Fairness doesn't come into it. Thus
176 if the range specifies 1.2.3.4 ports 10000-10005 and 1.2.3.5 ports
177 1-65535, we don't do pro-rata allocation based on ports; we choose
178 the ip with the lowest src-ip/dst-ip/proto usage.
181 find_best_ips_proto(u16 zone
, struct nf_conntrack_tuple
*tuple
,
182 const struct nf_nat_range
*range
,
183 const struct nf_conn
*ct
,
184 enum nf_nat_manip_type maniptype
)
188 u_int32_t minip
, maxip
, j
;
190 /* No IP mapping? Do nothing. */
191 if (!(range
->flags
& IP_NAT_RANGE_MAP_IPS
))
194 if (maniptype
== IP_NAT_MANIP_SRC
)
195 var_ipp
= &tuple
->src
.u3
.ip
;
197 var_ipp
= &tuple
->dst
.u3
.ip
;
199 /* Fast path: only one choice. */
200 if (range
->min_ip
== range
->max_ip
) {
201 *var_ipp
= range
->min_ip
;
205 /* Hashing source and destination IPs gives a fairly even
206 * spread in practice (if there are a small number of IPs
207 * involved, there usually aren't that many connections
208 * anyway). The consistency means that servers see the same
209 * client coming from the same IP (some Internet Banking sites
210 * like this), even across reboots. */
211 minip
= ntohl(range
->min_ip
);
212 maxip
= ntohl(range
->max_ip
);
213 j
= jhash_2words((__force u32
)tuple
->src
.u3
.ip
,
214 range
->flags
& IP_NAT_RANGE_PERSISTENT
?
215 0 : (__force u32
)tuple
->dst
.u3
.ip
^ zone
, 0);
216 j
= ((u64
)j
* (maxip
- minip
+ 1)) >> 32;
217 *var_ipp
= htonl(minip
+ j
);
220 /* Manipulate the tuple into the range given. For NF_INET_POST_ROUTING,
221 * we change the source to map into the range. For NF_INET_PRE_ROUTING
222 * and NF_INET_LOCAL_OUT, we change the destination to map into the
223 * range. It might not be possible to get a unique tuple, but we try.
224 * At worst (or if we race), we will end up with a final duplicate in
225 * __ip_conntrack_confirm and drop the packet. */
227 get_unique_tuple(struct nf_conntrack_tuple
*tuple
,
228 const struct nf_conntrack_tuple
*orig_tuple
,
229 const struct nf_nat_range
*range
,
231 enum nf_nat_manip_type maniptype
)
233 struct net
*net
= nf_ct_net(ct
);
234 const struct nf_nat_protocol
*proto
;
235 u16 zone
= nf_ct_zone(ct
);
237 /* 1) If this srcip/proto/src-proto-part is currently mapped,
238 and that same mapping gives a unique tuple within the given
241 This is only required for source (ie. NAT/masq) mappings.
242 So far, we don't do local source mappings, so multiple
243 manips not an issue. */
244 if (maniptype
== IP_NAT_MANIP_SRC
&&
245 !(range
->flags
& IP_NAT_RANGE_PROTO_RANDOM
)) {
246 if (find_appropriate_src(net
, zone
, orig_tuple
, tuple
, range
)) {
247 pr_debug("get_unique_tuple: Found current src map\n");
248 if (!nf_nat_used_tuple(tuple
, ct
))
253 /* 2) Select the least-used IP/proto combination in the given
255 *tuple
= *orig_tuple
;
256 find_best_ips_proto(zone
, tuple
, range
, ct
, maniptype
);
258 /* 3) The per-protocol part of the manip is made to map into
259 the range to make a unique tuple. */
262 proto
= __nf_nat_proto_find(orig_tuple
->dst
.protonum
);
264 /* Only bother mapping if it's not already in range and unique */
265 if (!(range
->flags
& IP_NAT_RANGE_PROTO_RANDOM
) &&
266 (!(range
->flags
& IP_NAT_RANGE_PROTO_SPECIFIED
) ||
267 proto
->in_range(tuple
, maniptype
, &range
->min
, &range
->max
)) &&
268 !nf_nat_used_tuple(tuple
, ct
))
271 /* Last change: get protocol to try to obtain unique tuple. */
272 proto
->unique_tuple(tuple
, range
, maniptype
, ct
);
278 nf_nat_setup_info(struct nf_conn
*ct
,
279 const struct nf_nat_range
*range
,
280 enum nf_nat_manip_type maniptype
)
282 struct net
*net
= nf_ct_net(ct
);
283 struct nf_conntrack_tuple curr_tuple
, new_tuple
;
284 struct nf_conn_nat
*nat
;
285 int have_to_hash
= !(ct
->status
& IPS_NAT_DONE_MASK
);
287 /* nat helper or nfctnetlink also setup binding */
290 nat
= nf_ct_ext_add(ct
, NF_CT_EXT_NAT
, GFP_ATOMIC
);
292 pr_debug("failed to add NAT extension\n");
297 NF_CT_ASSERT(maniptype
== IP_NAT_MANIP_SRC
||
298 maniptype
== IP_NAT_MANIP_DST
);
299 BUG_ON(nf_nat_initialized(ct
, maniptype
));
301 /* What we've got will look like inverse of reply. Normally
302 this is what is in the conntrack, except for prior
303 manipulations (future optimization: if num_manips == 0,
305 conntrack->tuplehash[IP_CT_DIR_ORIGINAL].tuple) */
306 nf_ct_invert_tuplepr(&curr_tuple
,
307 &ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
);
309 get_unique_tuple(&new_tuple
, &curr_tuple
, range
, ct
, maniptype
);
311 if (!nf_ct_tuple_equal(&new_tuple
, &curr_tuple
)) {
312 struct nf_conntrack_tuple reply
;
314 /* Alter conntrack table so will recognize replies. */
315 nf_ct_invert_tuplepr(&reply
, &new_tuple
);
316 nf_conntrack_alter_reply(ct
, &reply
);
318 /* Non-atomic: we own this at the moment. */
319 if (maniptype
== IP_NAT_MANIP_SRC
)
320 ct
->status
|= IPS_SRC_NAT
;
322 ct
->status
|= IPS_DST_NAT
;
325 /* Place in source hash if this is the first time. */
327 unsigned int srchash
;
329 srchash
= hash_by_src(net
, nf_ct_zone(ct
),
330 &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
+ hdrlen
;
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
, hdrlen
+ sizeof(struct icmphdr
),
470 sizeof(struct icmphdr
) + inside
->ip
.ihl
* 4),
471 (u_int16_t
)AF_INET
, inside
->ip
.protocol
,
472 &inner
, l3proto
, l4proto
))
475 /* Change inner back to look like incoming packet. We do the
476 opposite manip on this hook to normal, because it might not
477 pass all hooks (locally-generated ICMP). Consider incoming
478 packet: PREROUTING (DST manip), routing produces ICMP, goes
479 through POSTROUTING (which must correct the DST manip). */
480 if (!manip_pkt(inside
->ip
.protocol
, skb
, hdrlen
+ sizeof(inside
->icmp
),
481 &ct
->tuplehash
[!dir
].tuple
, !manip
))
484 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
) {
485 /* Reloading "inside" here since manip_pkt inner. */
486 inside
= (void *)skb
->data
+ hdrlen
;
487 inside
->icmp
.checksum
= 0;
488 inside
->icmp
.checksum
=
489 csum_fold(skb_checksum(skb
, hdrlen
,
490 skb
->len
- hdrlen
, 0));
493 /* Change outer to look the reply to an incoming packet
494 * (proto 0 means don't invert per-proto part). */
495 if (manip
== IP_NAT_MANIP_SRC
)
496 statusbit
= IPS_SRC_NAT
;
498 statusbit
= IPS_DST_NAT
;
500 /* Invert if this is reply dir. */
501 if (dir
== IP_CT_DIR_REPLY
)
502 statusbit
^= IPS_NAT_MASK
;
504 if (ct
->status
& statusbit
) {
505 nf_ct_invert_tuplepr(&target
, &ct
->tuplehash
[!dir
].tuple
);
506 if (!manip_pkt(0, skb
, 0, &target
, manip
))
512 EXPORT_SYMBOL_GPL(nf_nat_icmp_reply_translation
);
514 /* Protocol registration. */
515 int nf_nat_protocol_register(const struct nf_nat_protocol
*proto
)
519 spin_lock_bh(&nf_nat_lock
);
520 if (nf_nat_protos
[proto
->protonum
] != &nf_nat_unknown_protocol
) {
524 rcu_assign_pointer(nf_nat_protos
[proto
->protonum
], proto
);
526 spin_unlock_bh(&nf_nat_lock
);
529 EXPORT_SYMBOL(nf_nat_protocol_register
);
531 /* Noone stores the protocol anywhere; simply delete it. */
532 void nf_nat_protocol_unregister(const struct nf_nat_protocol
*proto
)
534 spin_lock_bh(&nf_nat_lock
);
535 rcu_assign_pointer(nf_nat_protos
[proto
->protonum
],
536 &nf_nat_unknown_protocol
);
537 spin_unlock_bh(&nf_nat_lock
);
540 EXPORT_SYMBOL(nf_nat_protocol_unregister
);
542 /* Noone using conntrack by the time this called. */
543 static void nf_nat_cleanup_conntrack(struct nf_conn
*ct
)
545 struct nf_conn_nat
*nat
= nf_ct_ext_find(ct
, NF_CT_EXT_NAT
);
547 if (nat
== NULL
|| nat
->ct
== NULL
)
550 NF_CT_ASSERT(nat
->ct
->status
& IPS_NAT_DONE_MASK
);
552 spin_lock_bh(&nf_nat_lock
);
553 hlist_del_rcu(&nat
->bysource
);
554 spin_unlock_bh(&nf_nat_lock
);
557 static void nf_nat_move_storage(void *new, void *old
)
559 struct nf_conn_nat
*new_nat
= new;
560 struct nf_conn_nat
*old_nat
= old
;
561 struct nf_conn
*ct
= old_nat
->ct
;
563 if (!ct
|| !(ct
->status
& IPS_NAT_DONE_MASK
))
566 spin_lock_bh(&nf_nat_lock
);
568 hlist_replace_rcu(&old_nat
->bysource
, &new_nat
->bysource
);
569 spin_unlock_bh(&nf_nat_lock
);
572 static struct nf_ct_ext_type nat_extend __read_mostly
= {
573 .len
= sizeof(struct nf_conn_nat
),
574 .align
= __alignof__(struct nf_conn_nat
),
575 .destroy
= nf_nat_cleanup_conntrack
,
576 .move
= nf_nat_move_storage
,
578 .flags
= NF_CT_EXT_F_PREALLOC
,
581 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
583 #include <linux/netfilter/nfnetlink.h>
584 #include <linux/netfilter/nfnetlink_conntrack.h>
586 static const struct nla_policy protonat_nla_policy
[CTA_PROTONAT_MAX
+1] = {
587 [CTA_PROTONAT_PORT_MIN
] = { .type
= NLA_U16
},
588 [CTA_PROTONAT_PORT_MAX
] = { .type
= NLA_U16
},
591 static int nfnetlink_parse_nat_proto(struct nlattr
*attr
,
592 const struct nf_conn
*ct
,
593 struct nf_nat_range
*range
)
595 struct nlattr
*tb
[CTA_PROTONAT_MAX
+1];
596 const struct nf_nat_protocol
*npt
;
599 err
= nla_parse_nested(tb
, CTA_PROTONAT_MAX
, attr
, protonat_nla_policy
);
603 npt
= nf_nat_proto_find_get(nf_ct_protonum(ct
));
604 if (npt
->nlattr_to_range
)
605 err
= npt
->nlattr_to_range(tb
, range
);
606 nf_nat_proto_put(npt
);
610 static const struct nla_policy nat_nla_policy
[CTA_NAT_MAX
+1] = {
611 [CTA_NAT_MINIP
] = { .type
= NLA_U32
},
612 [CTA_NAT_MAXIP
] = { .type
= NLA_U32
},
616 nfnetlink_parse_nat(const struct nlattr
*nat
,
617 const struct nf_conn
*ct
, struct nf_nat_range
*range
)
619 struct nlattr
*tb
[CTA_NAT_MAX
+1];
622 memset(range
, 0, sizeof(*range
));
624 err
= nla_parse_nested(tb
, CTA_NAT_MAX
, nat
, nat_nla_policy
);
628 if (tb
[CTA_NAT_MINIP
])
629 range
->min_ip
= nla_get_be32(tb
[CTA_NAT_MINIP
]);
631 if (!tb
[CTA_NAT_MAXIP
])
632 range
->max_ip
= range
->min_ip
;
634 range
->max_ip
= nla_get_be32(tb
[CTA_NAT_MAXIP
]);
637 range
->flags
|= IP_NAT_RANGE_MAP_IPS
;
639 if (!tb
[CTA_NAT_PROTO
])
642 err
= nfnetlink_parse_nat_proto(tb
[CTA_NAT_PROTO
], ct
, range
);
650 nfnetlink_parse_nat_setup(struct nf_conn
*ct
,
651 enum nf_nat_manip_type manip
,
652 const struct nlattr
*attr
)
654 struct nf_nat_range range
;
656 if (nfnetlink_parse_nat(attr
, ct
, &range
) < 0)
658 if (nf_nat_initialized(ct
, manip
))
661 return nf_nat_setup_info(ct
, &range
, manip
);
665 nfnetlink_parse_nat_setup(struct nf_conn
*ct
,
666 enum nf_nat_manip_type manip
,
667 const struct nlattr
*attr
)
673 static int __net_init
nf_nat_net_init(struct net
*net
)
675 /* Leave them the same for the moment. */
676 net
->ipv4
.nat_htable_size
= net
->ct
.htable_size
;
677 net
->ipv4
.nat_bysource
= nf_ct_alloc_hashtable(&net
->ipv4
.nat_htable_size
,
678 &net
->ipv4
.nat_vmalloced
, 0);
679 if (!net
->ipv4
.nat_bysource
)
684 /* Clear NAT section of all conntracks, in case we're loaded again. */
685 static int clean_nat(struct nf_conn
*i
, void *data
)
687 struct nf_conn_nat
*nat
= nfct_nat(i
);
691 memset(nat
, 0, sizeof(*nat
));
692 i
->status
&= ~(IPS_NAT_MASK
| IPS_NAT_DONE_MASK
| IPS_SEQ_ADJUST
);
696 static void __net_exit
nf_nat_net_exit(struct net
*net
)
698 nf_ct_iterate_cleanup(net
, &clean_nat
, NULL
);
700 nf_ct_free_hashtable(net
->ipv4
.nat_bysource
, net
->ipv4
.nat_vmalloced
,
701 net
->ipv4
.nat_htable_size
);
704 static struct pernet_operations nf_nat_net_ops
= {
705 .init
= nf_nat_net_init
,
706 .exit
= nf_nat_net_exit
,
709 static int __init
nf_nat_init(void)
714 need_ipv4_conntrack();
716 ret
= nf_ct_extend_register(&nat_extend
);
718 printk(KERN_ERR
"nf_nat_core: Unable to register extension\n");
722 ret
= register_pernet_subsys(&nf_nat_net_ops
);
726 /* Sew in builtin protocols. */
727 spin_lock_bh(&nf_nat_lock
);
728 for (i
= 0; i
< MAX_IP_NAT_PROTO
; i
++)
729 rcu_assign_pointer(nf_nat_protos
[i
], &nf_nat_unknown_protocol
);
730 rcu_assign_pointer(nf_nat_protos
[IPPROTO_TCP
], &nf_nat_protocol_tcp
);
731 rcu_assign_pointer(nf_nat_protos
[IPPROTO_UDP
], &nf_nat_protocol_udp
);
732 rcu_assign_pointer(nf_nat_protos
[IPPROTO_ICMP
], &nf_nat_protocol_icmp
);
733 spin_unlock_bh(&nf_nat_lock
);
735 /* Initialize fake conntrack so that NAT will skip it */
736 nf_ct_untracked_status_or(IPS_NAT_DONE_MASK
);
738 l3proto
= nf_ct_l3proto_find_get((u_int16_t
)AF_INET
);
740 BUG_ON(nf_nat_seq_adjust_hook
!= NULL
);
741 rcu_assign_pointer(nf_nat_seq_adjust_hook
, nf_nat_seq_adjust
);
742 BUG_ON(nfnetlink_parse_nat_setup_hook
!= NULL
);
743 rcu_assign_pointer(nfnetlink_parse_nat_setup_hook
,
744 nfnetlink_parse_nat_setup
);
745 BUG_ON(nf_ct_nat_offset
!= NULL
);
746 rcu_assign_pointer(nf_ct_nat_offset
, nf_nat_get_offset
);
750 nf_ct_extend_unregister(&nat_extend
);
754 static void __exit
nf_nat_cleanup(void)
756 unregister_pernet_subsys(&nf_nat_net_ops
);
757 nf_ct_l3proto_put(l3proto
);
758 nf_ct_extend_unregister(&nat_extend
);
759 rcu_assign_pointer(nf_nat_seq_adjust_hook
, NULL
);
760 rcu_assign_pointer(nfnetlink_parse_nat_setup_hook
, NULL
);
761 rcu_assign_pointer(nf_ct_nat_offset
, NULL
);
765 MODULE_LICENSE("GPL");
766 MODULE_ALIAS("nf-nat-ipv4");
768 module_init(nf_nat_init
);
769 module_exit(nf_nat_cleanup
);