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 /* Change protocol info to have some randomization */
265 if (range
->flags
& IP_NAT_RANGE_PROTO_RANDOM
) {
266 proto
->unique_tuple(tuple
, range
, maniptype
, ct
);
270 /* Only bother mapping if it's not already in range and unique */
271 if ((!(range
->flags
& IP_NAT_RANGE_PROTO_SPECIFIED
) ||
272 proto
->in_range(tuple
, maniptype
, &range
->min
, &range
->max
)) &&
273 !nf_nat_used_tuple(tuple
, ct
))
276 /* Last change: get protocol to try to obtain unique tuple. */
277 proto
->unique_tuple(tuple
, range
, maniptype
, ct
);
283 nf_nat_setup_info(struct nf_conn
*ct
,
284 const struct nf_nat_range
*range
,
285 enum nf_nat_manip_type maniptype
)
287 struct net
*net
= nf_ct_net(ct
);
288 struct nf_conntrack_tuple curr_tuple
, new_tuple
;
289 struct nf_conn_nat
*nat
;
290 int have_to_hash
= !(ct
->status
& IPS_NAT_DONE_MASK
);
292 /* nat helper or nfctnetlink also setup binding */
295 nat
= nf_ct_ext_add(ct
, NF_CT_EXT_NAT
, GFP_ATOMIC
);
297 pr_debug("failed to add NAT extension\n");
302 NF_CT_ASSERT(maniptype
== IP_NAT_MANIP_SRC
||
303 maniptype
== IP_NAT_MANIP_DST
);
304 BUG_ON(nf_nat_initialized(ct
, maniptype
));
306 /* What we've got will look like inverse of reply. Normally
307 this is what is in the conntrack, except for prior
308 manipulations (future optimization: if num_manips == 0,
310 conntrack->tuplehash[IP_CT_DIR_ORIGINAL].tuple) */
311 nf_ct_invert_tuplepr(&curr_tuple
,
312 &ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
);
314 get_unique_tuple(&new_tuple
, &curr_tuple
, range
, ct
, maniptype
);
316 if (!nf_ct_tuple_equal(&new_tuple
, &curr_tuple
)) {
317 struct nf_conntrack_tuple reply
;
319 /* Alter conntrack table so will recognize replies. */
320 nf_ct_invert_tuplepr(&reply
, &new_tuple
);
321 nf_conntrack_alter_reply(ct
, &reply
);
323 /* Non-atomic: we own this at the moment. */
324 if (maniptype
== IP_NAT_MANIP_SRC
)
325 ct
->status
|= IPS_SRC_NAT
;
327 ct
->status
|= IPS_DST_NAT
;
330 /* Place in source hash if this is the first time. */
332 unsigned int srchash
;
334 srchash
= hash_by_src(net
, nf_ct_zone(ct
),
335 &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
);
336 spin_lock_bh(&nf_nat_lock
);
337 /* nf_conntrack_alter_reply might re-allocate exntension aera */
340 hlist_add_head_rcu(&nat
->bysource
,
341 &net
->ipv4
.nat_bysource
[srchash
]);
342 spin_unlock_bh(&nf_nat_lock
);
346 if (maniptype
== IP_NAT_MANIP_DST
)
347 set_bit(IPS_DST_NAT_DONE_BIT
, &ct
->status
);
349 set_bit(IPS_SRC_NAT_DONE_BIT
, &ct
->status
);
353 EXPORT_SYMBOL(nf_nat_setup_info
);
355 /* Returns true if succeeded. */
357 manip_pkt(u_int16_t proto
,
359 unsigned int iphdroff
,
360 const struct nf_conntrack_tuple
*target
,
361 enum nf_nat_manip_type maniptype
)
364 const struct nf_nat_protocol
*p
;
366 if (!skb_make_writable(skb
, iphdroff
+ sizeof(*iph
)))
369 iph
= (void *)skb
->data
+ iphdroff
;
371 /* Manipulate protcol part. */
373 /* rcu_read_lock()ed by nf_hook_slow */
374 p
= __nf_nat_proto_find(proto
);
375 if (!p
->manip_pkt(skb
, iphdroff
, target
, maniptype
))
378 iph
= (void *)skb
->data
+ iphdroff
;
380 if (maniptype
== IP_NAT_MANIP_SRC
) {
381 csum_replace4(&iph
->check
, iph
->saddr
, target
->src
.u3
.ip
);
382 iph
->saddr
= target
->src
.u3
.ip
;
384 csum_replace4(&iph
->check
, iph
->daddr
, target
->dst
.u3
.ip
);
385 iph
->daddr
= target
->dst
.u3
.ip
;
390 /* Do packet manipulations according to nf_nat_setup_info. */
391 unsigned int nf_nat_packet(struct nf_conn
*ct
,
392 enum ip_conntrack_info ctinfo
,
393 unsigned int hooknum
,
396 enum ip_conntrack_dir dir
= CTINFO2DIR(ctinfo
);
397 unsigned long statusbit
;
398 enum nf_nat_manip_type mtype
= HOOK2MANIP(hooknum
);
400 if (mtype
== IP_NAT_MANIP_SRC
)
401 statusbit
= IPS_SRC_NAT
;
403 statusbit
= IPS_DST_NAT
;
405 /* Invert if this is reply dir. */
406 if (dir
== IP_CT_DIR_REPLY
)
407 statusbit
^= IPS_NAT_MASK
;
409 /* Non-atomic: these bits don't change. */
410 if (ct
->status
& statusbit
) {
411 struct nf_conntrack_tuple target
;
413 /* We are aiming to look like inverse of other direction. */
414 nf_ct_invert_tuplepr(&target
, &ct
->tuplehash
[!dir
].tuple
);
416 if (!manip_pkt(target
.dst
.protonum
, skb
, 0, &target
, mtype
))
421 EXPORT_SYMBOL_GPL(nf_nat_packet
);
423 /* Dir is direction ICMP is coming from (opposite to packet it contains) */
424 int nf_nat_icmp_reply_translation(struct nf_conn
*ct
,
425 enum ip_conntrack_info ctinfo
,
426 unsigned int hooknum
,
433 const struct nf_conntrack_l4proto
*l4proto
;
434 struct nf_conntrack_tuple inner
, target
;
435 int hdrlen
= ip_hdrlen(skb
);
436 enum ip_conntrack_dir dir
= CTINFO2DIR(ctinfo
);
437 unsigned long statusbit
;
438 enum nf_nat_manip_type manip
= HOOK2MANIP(hooknum
);
440 if (!skb_make_writable(skb
, hdrlen
+ sizeof(*inside
)))
443 inside
= (void *)skb
->data
+ ip_hdrlen(skb
);
445 /* We're actually going to mangle it beyond trivial checksum
446 adjustment, so make sure the current checksum is correct. */
447 if (nf_ip_checksum(skb
, hooknum
, hdrlen
, 0))
450 /* Must be RELATED */
451 NF_CT_ASSERT(skb
->nfctinfo
== IP_CT_RELATED
||
452 skb
->nfctinfo
== IP_CT_RELATED
+IP_CT_IS_REPLY
);
454 /* Redirects on non-null nats must be dropped, else they'll
455 start talking to each other without our translation, and be
457 if (inside
->icmp
.type
== ICMP_REDIRECT
) {
458 /* If NAT isn't finished, assume it and drop. */
459 if ((ct
->status
& IPS_NAT_DONE_MASK
) != IPS_NAT_DONE_MASK
)
462 if (ct
->status
& IPS_NAT_MASK
)
466 pr_debug("icmp_reply_translation: translating error %p manip %u "
467 "dir %s\n", skb
, manip
,
468 dir
== IP_CT_DIR_ORIGINAL
? "ORIG" : "REPLY");
470 /* rcu_read_lock()ed by nf_hook_slow */
471 l4proto
= __nf_ct_l4proto_find(PF_INET
, inside
->ip
.protocol
);
473 if (!nf_ct_get_tuple(skb
,
474 ip_hdrlen(skb
) + sizeof(struct icmphdr
),
476 sizeof(struct icmphdr
) + inside
->ip
.ihl
* 4),
479 &inner
, l3proto
, l4proto
))
482 /* Change inner back to look like incoming packet. We do the
483 opposite manip on this hook to normal, because it might not
484 pass all hooks (locally-generated ICMP). Consider incoming
485 packet: PREROUTING (DST manip), routing produces ICMP, goes
486 through POSTROUTING (which must correct the DST manip). */
487 if (!manip_pkt(inside
->ip
.protocol
, skb
,
488 ip_hdrlen(skb
) + sizeof(inside
->icmp
),
489 &ct
->tuplehash
[!dir
].tuple
,
493 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
) {
494 /* Reloading "inside" here since manip_pkt inner. */
495 inside
= (void *)skb
->data
+ ip_hdrlen(skb
);
496 inside
->icmp
.checksum
= 0;
497 inside
->icmp
.checksum
=
498 csum_fold(skb_checksum(skb
, hdrlen
,
499 skb
->len
- hdrlen
, 0));
502 /* Change outer to look the reply to an incoming packet
503 * (proto 0 means don't invert per-proto part). */
504 if (manip
== IP_NAT_MANIP_SRC
)
505 statusbit
= IPS_SRC_NAT
;
507 statusbit
= IPS_DST_NAT
;
509 /* Invert if this is reply dir. */
510 if (dir
== IP_CT_DIR_REPLY
)
511 statusbit
^= IPS_NAT_MASK
;
513 if (ct
->status
& statusbit
) {
514 nf_ct_invert_tuplepr(&target
, &ct
->tuplehash
[!dir
].tuple
);
515 if (!manip_pkt(0, skb
, 0, &target
, manip
))
521 EXPORT_SYMBOL_GPL(nf_nat_icmp_reply_translation
);
523 /* Protocol registration. */
524 int nf_nat_protocol_register(const struct nf_nat_protocol
*proto
)
528 spin_lock_bh(&nf_nat_lock
);
529 if (nf_nat_protos
[proto
->protonum
] != &nf_nat_unknown_protocol
) {
533 rcu_assign_pointer(nf_nat_protos
[proto
->protonum
], proto
);
535 spin_unlock_bh(&nf_nat_lock
);
538 EXPORT_SYMBOL(nf_nat_protocol_register
);
540 /* Noone stores the protocol anywhere; simply delete it. */
541 void nf_nat_protocol_unregister(const struct nf_nat_protocol
*proto
)
543 spin_lock_bh(&nf_nat_lock
);
544 rcu_assign_pointer(nf_nat_protos
[proto
->protonum
],
545 &nf_nat_unknown_protocol
);
546 spin_unlock_bh(&nf_nat_lock
);
549 EXPORT_SYMBOL(nf_nat_protocol_unregister
);
551 /* Noone using conntrack by the time this called. */
552 static void nf_nat_cleanup_conntrack(struct nf_conn
*ct
)
554 struct nf_conn_nat
*nat
= nf_ct_ext_find(ct
, NF_CT_EXT_NAT
);
556 if (nat
== NULL
|| nat
->ct
== NULL
)
559 NF_CT_ASSERT(nat
->ct
->status
& IPS_NAT_DONE_MASK
);
561 spin_lock_bh(&nf_nat_lock
);
562 hlist_del_rcu(&nat
->bysource
);
563 spin_unlock_bh(&nf_nat_lock
);
566 static void nf_nat_move_storage(void *new, void *old
)
568 struct nf_conn_nat
*new_nat
= new;
569 struct nf_conn_nat
*old_nat
= old
;
570 struct nf_conn
*ct
= old_nat
->ct
;
572 if (!ct
|| !(ct
->status
& IPS_NAT_DONE_MASK
))
575 spin_lock_bh(&nf_nat_lock
);
577 hlist_replace_rcu(&old_nat
->bysource
, &new_nat
->bysource
);
578 spin_unlock_bh(&nf_nat_lock
);
581 static struct nf_ct_ext_type nat_extend __read_mostly
= {
582 .len
= sizeof(struct nf_conn_nat
),
583 .align
= __alignof__(struct nf_conn_nat
),
584 .destroy
= nf_nat_cleanup_conntrack
,
585 .move
= nf_nat_move_storage
,
587 .flags
= NF_CT_EXT_F_PREALLOC
,
590 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
592 #include <linux/netfilter/nfnetlink.h>
593 #include <linux/netfilter/nfnetlink_conntrack.h>
595 static const struct nla_policy protonat_nla_policy
[CTA_PROTONAT_MAX
+1] = {
596 [CTA_PROTONAT_PORT_MIN
] = { .type
= NLA_U16
},
597 [CTA_PROTONAT_PORT_MAX
] = { .type
= NLA_U16
},
600 static int nfnetlink_parse_nat_proto(struct nlattr
*attr
,
601 const struct nf_conn
*ct
,
602 struct nf_nat_range
*range
)
604 struct nlattr
*tb
[CTA_PROTONAT_MAX
+1];
605 const struct nf_nat_protocol
*npt
;
608 err
= nla_parse_nested(tb
, CTA_PROTONAT_MAX
, attr
, protonat_nla_policy
);
612 npt
= nf_nat_proto_find_get(nf_ct_protonum(ct
));
613 if (npt
->nlattr_to_range
)
614 err
= npt
->nlattr_to_range(tb
, range
);
615 nf_nat_proto_put(npt
);
619 static const struct nla_policy nat_nla_policy
[CTA_NAT_MAX
+1] = {
620 [CTA_NAT_MINIP
] = { .type
= NLA_U32
},
621 [CTA_NAT_MAXIP
] = { .type
= NLA_U32
},
625 nfnetlink_parse_nat(const struct nlattr
*nat
,
626 const struct nf_conn
*ct
, struct nf_nat_range
*range
)
628 struct nlattr
*tb
[CTA_NAT_MAX
+1];
631 memset(range
, 0, sizeof(*range
));
633 err
= nla_parse_nested(tb
, CTA_NAT_MAX
, nat
, nat_nla_policy
);
637 if (tb
[CTA_NAT_MINIP
])
638 range
->min_ip
= nla_get_be32(tb
[CTA_NAT_MINIP
]);
640 if (!tb
[CTA_NAT_MAXIP
])
641 range
->max_ip
= range
->min_ip
;
643 range
->max_ip
= nla_get_be32(tb
[CTA_NAT_MAXIP
]);
646 range
->flags
|= IP_NAT_RANGE_MAP_IPS
;
648 if (!tb
[CTA_NAT_PROTO
])
651 err
= nfnetlink_parse_nat_proto(tb
[CTA_NAT_PROTO
], ct
, range
);
659 nfnetlink_parse_nat_setup(struct nf_conn
*ct
,
660 enum nf_nat_manip_type manip
,
661 const struct nlattr
*attr
)
663 struct nf_nat_range range
;
665 if (nfnetlink_parse_nat(attr
, ct
, &range
) < 0)
667 if (nf_nat_initialized(ct
, manip
))
670 return nf_nat_setup_info(ct
, &range
, manip
);
674 nfnetlink_parse_nat_setup(struct nf_conn
*ct
,
675 enum nf_nat_manip_type manip
,
676 const struct nlattr
*attr
)
682 static int __net_init
nf_nat_net_init(struct net
*net
)
684 /* Leave them the same for the moment. */
685 net
->ipv4
.nat_htable_size
= net
->ct
.htable_size
;
686 net
->ipv4
.nat_bysource
= nf_ct_alloc_hashtable(&net
->ipv4
.nat_htable_size
,
687 &net
->ipv4
.nat_vmalloced
, 0);
688 if (!net
->ipv4
.nat_bysource
)
693 /* Clear NAT section of all conntracks, in case we're loaded again. */
694 static int clean_nat(struct nf_conn
*i
, void *data
)
696 struct nf_conn_nat
*nat
= nfct_nat(i
);
700 memset(nat
, 0, sizeof(*nat
));
701 i
->status
&= ~(IPS_NAT_MASK
| IPS_NAT_DONE_MASK
| IPS_SEQ_ADJUST
);
705 static void __net_exit
nf_nat_net_exit(struct net
*net
)
707 nf_ct_iterate_cleanup(net
, &clean_nat
, NULL
);
709 nf_ct_free_hashtable(net
->ipv4
.nat_bysource
, net
->ipv4
.nat_vmalloced
,
710 net
->ipv4
.nat_htable_size
);
713 static struct pernet_operations nf_nat_net_ops
= {
714 .init
= nf_nat_net_init
,
715 .exit
= nf_nat_net_exit
,
718 static int __init
nf_nat_init(void)
723 need_ipv4_conntrack();
725 ret
= nf_ct_extend_register(&nat_extend
);
727 printk(KERN_ERR
"nf_nat_core: Unable to register extension\n");
731 ret
= register_pernet_subsys(&nf_nat_net_ops
);
735 /* Sew in builtin protocols. */
736 spin_lock_bh(&nf_nat_lock
);
737 for (i
= 0; i
< MAX_IP_NAT_PROTO
; i
++)
738 rcu_assign_pointer(nf_nat_protos
[i
], &nf_nat_unknown_protocol
);
739 rcu_assign_pointer(nf_nat_protos
[IPPROTO_TCP
], &nf_nat_protocol_tcp
);
740 rcu_assign_pointer(nf_nat_protos
[IPPROTO_UDP
], &nf_nat_protocol_udp
);
741 rcu_assign_pointer(nf_nat_protos
[IPPROTO_ICMP
], &nf_nat_protocol_icmp
);
742 spin_unlock_bh(&nf_nat_lock
);
744 /* Initialize fake conntrack so that NAT will skip it */
745 nf_conntrack_untracked
.status
|= IPS_NAT_DONE_MASK
;
747 l3proto
= nf_ct_l3proto_find_get((u_int16_t
)AF_INET
);
749 BUG_ON(nf_nat_seq_adjust_hook
!= NULL
);
750 rcu_assign_pointer(nf_nat_seq_adjust_hook
, nf_nat_seq_adjust
);
751 BUG_ON(nfnetlink_parse_nat_setup_hook
!= NULL
);
752 rcu_assign_pointer(nfnetlink_parse_nat_setup_hook
,
753 nfnetlink_parse_nat_setup
);
754 BUG_ON(nf_ct_nat_offset
!= NULL
);
755 rcu_assign_pointer(nf_ct_nat_offset
, nf_nat_get_offset
);
759 nf_ct_extend_unregister(&nat_extend
);
763 static void __exit
nf_nat_cleanup(void)
765 unregister_pernet_subsys(&nf_nat_net_ops
);
766 nf_ct_l3proto_put(l3proto
);
767 nf_ct_extend_unregister(&nat_extend
);
768 rcu_assign_pointer(nf_nat_seq_adjust_hook
, NULL
);
769 rcu_assign_pointer(nfnetlink_parse_nat_setup_hook
, NULL
);
770 rcu_assign_pointer(nf_ct_nat_offset
, NULL
);
774 MODULE_LICENSE("GPL");
775 MODULE_ALIAS("nf-nat-ipv4");
777 module_init(nf_nat_init
);
778 module_exit(nf_nat_cleanup
);