2 * (C) 1999-2001 Paul `Rusty' Russell
3 * (C) 2002-2006 Netfilter Core Team <coreteam@netfilter.org>
4 * (C) 2011 Patrick McHardy <kaber@trash.net>
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>
17 #include <linux/jhash.h>
18 #include <linux/rtnetlink.h>
20 #include <net/netfilter/nf_conntrack.h>
21 #include <net/netfilter/nf_conntrack_core.h>
22 #include <net/netfilter/nf_nat.h>
23 #include <net/netfilter/nf_nat_l3proto.h>
24 #include <net/netfilter/nf_nat_l4proto.h>
25 #include <net/netfilter/nf_nat_core.h>
26 #include <net/netfilter/nf_nat_helper.h>
27 #include <net/netfilter/nf_conntrack_helper.h>
28 #include <net/netfilter/nf_conntrack_l3proto.h>
29 #include <net/netfilter/nf_conntrack_zones.h>
30 #include <linux/netfilter/nf_nat.h>
32 static DEFINE_SPINLOCK(nf_nat_lock
);
34 static DEFINE_MUTEX(nf_nat_proto_mutex
);
35 static const struct nf_nat_l3proto __rcu
*nf_nat_l3protos
[NFPROTO_NUMPROTO
]
37 static const struct nf_nat_l4proto __rcu
**nf_nat_l4protos
[NFPROTO_NUMPROTO
]
41 inline const struct nf_nat_l3proto
*
42 __nf_nat_l3proto_find(u8 family
)
44 return rcu_dereference(nf_nat_l3protos
[family
]);
47 inline const struct nf_nat_l4proto
*
48 __nf_nat_l4proto_find(u8 family
, u8 protonum
)
50 return rcu_dereference(nf_nat_l4protos
[family
][protonum
]);
52 EXPORT_SYMBOL_GPL(__nf_nat_l4proto_find
);
55 static void __nf_nat_decode_session(struct sk_buff
*skb
, struct flowi
*fl
)
57 const struct nf_nat_l3proto
*l3proto
;
58 const struct nf_conn
*ct
;
59 enum ip_conntrack_info ctinfo
;
60 enum ip_conntrack_dir dir
;
61 unsigned long statusbit
;
64 ct
= nf_ct_get(skb
, &ctinfo
);
68 family
= ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
.src
.l3num
;
70 l3proto
= __nf_nat_l3proto_find(family
);
74 dir
= CTINFO2DIR(ctinfo
);
75 if (dir
== IP_CT_DIR_ORIGINAL
)
76 statusbit
= IPS_DST_NAT
;
78 statusbit
= IPS_SRC_NAT
;
80 l3proto
->decode_session(skb
, ct
, dir
, statusbit
, fl
);
85 int nf_xfrm_me_harder(struct sk_buff
*skb
, unsigned int family
)
89 struct dst_entry
*dst
;
92 err
= xfrm_decode_session(skb
, &fl
, family
);
97 dst
= ((struct xfrm_dst
*)dst
)->route
;
100 dst
= xfrm_lookup(dev_net(dst
->dev
), dst
, &fl
, skb
->sk
, 0);
105 skb_dst_set(skb
, dst
);
107 /* Change in oif may mean change in hh_len. */
108 hh_len
= skb_dst(skb
)->dev
->hard_header_len
;
109 if (skb_headroom(skb
) < hh_len
&&
110 pskb_expand_head(skb
, hh_len
- skb_headroom(skb
), 0, GFP_ATOMIC
))
114 EXPORT_SYMBOL(nf_xfrm_me_harder
);
115 #endif /* CONFIG_XFRM */
117 /* We keep an extra hash for each conntrack, for fast searching. */
118 static inline unsigned int
119 hash_by_src(const struct net
*net
, u16 zone
,
120 const struct nf_conntrack_tuple
*tuple
)
124 /* Original src, to ensure we map it consistently if poss. */
125 hash
= jhash2((u32
*)&tuple
->src
, sizeof(tuple
->src
) / sizeof(u32
),
126 tuple
->dst
.protonum
^ zone
^ nf_conntrack_hash_rnd
);
127 return ((u64
)hash
* net
->ct
.nat_htable_size
) >> 32;
130 /* Is this tuple already taken? (not by us) */
132 nf_nat_used_tuple(const struct nf_conntrack_tuple
*tuple
,
133 const struct nf_conn
*ignored_conntrack
)
135 /* Conntrack tracking doesn't keep track of outgoing tuples; only
136 * incoming ones. NAT means they don't have a fixed mapping,
137 * so we invert the tuple and look for the incoming reply.
139 * We could keep a separate hash if this proves too slow.
141 struct nf_conntrack_tuple reply
;
143 nf_ct_invert_tuplepr(&reply
, tuple
);
144 return nf_conntrack_tuple_taken(&reply
, ignored_conntrack
);
146 EXPORT_SYMBOL(nf_nat_used_tuple
);
148 /* If we source map this tuple so reply looks like reply_tuple, will
149 * that meet the constraints of range.
151 static int in_range(const struct nf_nat_l3proto
*l3proto
,
152 const struct nf_nat_l4proto
*l4proto
,
153 const struct nf_conntrack_tuple
*tuple
,
154 const struct nf_nat_range
*range
)
156 /* If we are supposed to map IPs, then we must be in the
157 * range specified, otherwise let this drag us onto a new src IP.
159 if (range
->flags
& NF_NAT_RANGE_MAP_IPS
&&
160 !l3proto
->in_range(tuple
, range
))
163 if (!(range
->flags
& NF_NAT_RANGE_PROTO_SPECIFIED
) ||
164 l4proto
->in_range(tuple
, NF_NAT_MANIP_SRC
,
165 &range
->min_proto
, &range
->max_proto
))
172 same_src(const struct nf_conn
*ct
,
173 const struct nf_conntrack_tuple
*tuple
)
175 const struct nf_conntrack_tuple
*t
;
177 t
= &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
;
178 return (t
->dst
.protonum
== tuple
->dst
.protonum
&&
179 nf_inet_addr_cmp(&t
->src
.u3
, &tuple
->src
.u3
) &&
180 t
->src
.u
.all
== tuple
->src
.u
.all
);
183 /* Only called for SRC manip */
185 find_appropriate_src(struct net
*net
, u16 zone
,
186 const struct nf_nat_l3proto
*l3proto
,
187 const struct nf_nat_l4proto
*l4proto
,
188 const struct nf_conntrack_tuple
*tuple
,
189 struct nf_conntrack_tuple
*result
,
190 const struct nf_nat_range
*range
)
192 unsigned int h
= hash_by_src(net
, zone
, tuple
);
193 const struct nf_conn_nat
*nat
;
194 const struct nf_conn
*ct
;
196 hlist_for_each_entry_rcu(nat
, &net
->ct
.nat_bysource
[h
], bysource
) {
198 if (same_src(ct
, tuple
) && nf_ct_zone(ct
) == zone
) {
199 /* Copy source part from reply tuple. */
200 nf_ct_invert_tuplepr(result
,
201 &ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
);
202 result
->dst
= tuple
->dst
;
204 if (in_range(l3proto
, l4proto
, result
, range
))
211 /* For [FUTURE] fragmentation handling, we want the least-used
212 * src-ip/dst-ip/proto triple. Fairness doesn't come into it. Thus
213 * if the range specifies 1.2.3.4 ports 10000-10005 and 1.2.3.5 ports
214 * 1-65535, we don't do pro-rata allocation based on ports; we choose
215 * the ip with the lowest src-ip/dst-ip/proto usage.
218 find_best_ips_proto(u16 zone
, struct nf_conntrack_tuple
*tuple
,
219 const struct nf_nat_range
*range
,
220 const struct nf_conn
*ct
,
221 enum nf_nat_manip_type maniptype
)
223 union nf_inet_addr
*var_ipp
;
226 u32 minip
, maxip
, j
, dist
;
229 /* No IP mapping? Do nothing. */
230 if (!(range
->flags
& NF_NAT_RANGE_MAP_IPS
))
233 if (maniptype
== NF_NAT_MANIP_SRC
)
234 var_ipp
= &tuple
->src
.u3
;
236 var_ipp
= &tuple
->dst
.u3
;
238 /* Fast path: only one choice. */
239 if (nf_inet_addr_cmp(&range
->min_addr
, &range
->max_addr
)) {
240 *var_ipp
= range
->min_addr
;
244 if (nf_ct_l3num(ct
) == NFPROTO_IPV4
)
245 max
= sizeof(var_ipp
->ip
) / sizeof(u32
) - 1;
247 max
= sizeof(var_ipp
->ip6
) / sizeof(u32
) - 1;
249 /* Hashing source and destination IPs gives a fairly even
250 * spread in practice (if there are a small number of IPs
251 * involved, there usually aren't that many connections
252 * anyway). The consistency means that servers see the same
253 * client coming from the same IP (some Internet Banking sites
254 * like this), even across reboots.
256 j
= jhash2((u32
*)&tuple
->src
.u3
, sizeof(tuple
->src
.u3
) / sizeof(u32
),
257 range
->flags
& NF_NAT_RANGE_PERSISTENT
?
258 0 : (__force u32
)tuple
->dst
.u3
.all
[max
] ^ zone
);
261 for (i
= 0; i
<= max
; i
++) {
262 /* If first bytes of the address are at the maximum, use the
263 * distance. Otherwise use the full range.
266 minip
= ntohl((__force __be32
)range
->min_addr
.all
[i
]);
267 maxip
= ntohl((__force __be32
)range
->max_addr
.all
[i
]);
268 dist
= maxip
- minip
+ 1;
274 var_ipp
->all
[i
] = (__force __u32
)
275 htonl(minip
+ (((u64
)j
* dist
) >> 32));
276 if (var_ipp
->all
[i
] != range
->max_addr
.all
[i
])
279 if (!(range
->flags
& NF_NAT_RANGE_PERSISTENT
))
280 j
^= (__force u32
)tuple
->dst
.u3
.all
[i
];
284 /* Manipulate the tuple into the range given. For NF_INET_POST_ROUTING,
285 * we change the source to map into the range. For NF_INET_PRE_ROUTING
286 * and NF_INET_LOCAL_OUT, we change the destination to map into the
287 * range. It might not be possible to get a unique tuple, but we try.
288 * At worst (or if we race), we will end up with a final duplicate in
289 * __ip_conntrack_confirm and drop the packet. */
291 get_unique_tuple(struct nf_conntrack_tuple
*tuple
,
292 const struct nf_conntrack_tuple
*orig_tuple
,
293 const struct nf_nat_range
*range
,
295 enum nf_nat_manip_type maniptype
)
297 const struct nf_nat_l3proto
*l3proto
;
298 const struct nf_nat_l4proto
*l4proto
;
299 struct net
*net
= nf_ct_net(ct
);
300 u16 zone
= nf_ct_zone(ct
);
303 l3proto
= __nf_nat_l3proto_find(orig_tuple
->src
.l3num
);
304 l4proto
= __nf_nat_l4proto_find(orig_tuple
->src
.l3num
,
305 orig_tuple
->dst
.protonum
);
307 /* 1) If this srcip/proto/src-proto-part is currently mapped,
308 * and that same mapping gives a unique tuple within the given
311 * This is only required for source (ie. NAT/masq) mappings.
312 * So far, we don't do local source mappings, so multiple
313 * manips not an issue.
315 if (maniptype
== NF_NAT_MANIP_SRC
&&
316 !(range
->flags
& NF_NAT_RANGE_PROTO_RANDOM
)) {
317 /* try the original tuple first */
318 if (in_range(l3proto
, l4proto
, orig_tuple
, range
)) {
319 if (!nf_nat_used_tuple(orig_tuple
, ct
)) {
320 *tuple
= *orig_tuple
;
323 } else if (find_appropriate_src(net
, zone
, l3proto
, l4proto
,
324 orig_tuple
, tuple
, range
)) {
325 pr_debug("get_unique_tuple: Found current src map\n");
326 if (!nf_nat_used_tuple(tuple
, ct
))
331 /* 2) Select the least-used IP/proto combination in the given range */
332 *tuple
= *orig_tuple
;
333 find_best_ips_proto(zone
, tuple
, range
, ct
, maniptype
);
335 /* 3) The per-protocol part of the manip is made to map into
336 * the range to make a unique tuple.
339 /* Only bother mapping if it's not already in range and unique */
340 if (!(range
->flags
& NF_NAT_RANGE_PROTO_RANDOM
)) {
341 if (range
->flags
& NF_NAT_RANGE_PROTO_SPECIFIED
) {
342 if (l4proto
->in_range(tuple
, maniptype
,
344 &range
->max_proto
) &&
345 (range
->min_proto
.all
== range
->max_proto
.all
||
346 !nf_nat_used_tuple(tuple
, ct
)))
348 } else if (!nf_nat_used_tuple(tuple
, ct
)) {
353 /* Last change: get protocol to try to obtain unique tuple. */
354 l4proto
->unique_tuple(l3proto
, tuple
, range
, maniptype
, ct
);
360 nf_nat_setup_info(struct nf_conn
*ct
,
361 const struct nf_nat_range
*range
,
362 enum nf_nat_manip_type maniptype
)
364 struct net
*net
= nf_ct_net(ct
);
365 struct nf_conntrack_tuple curr_tuple
, new_tuple
;
366 struct nf_conn_nat
*nat
;
368 /* nat helper or nfctnetlink also setup binding */
371 nat
= nf_ct_ext_add(ct
, NF_CT_EXT_NAT
, GFP_ATOMIC
);
373 pr_debug("failed to add NAT extension\n");
378 NF_CT_ASSERT(maniptype
== NF_NAT_MANIP_SRC
||
379 maniptype
== NF_NAT_MANIP_DST
);
380 BUG_ON(nf_nat_initialized(ct
, maniptype
));
382 /* What we've got will look like inverse of reply. Normally
383 * this is what is in the conntrack, except for prior
384 * manipulations (future optimization: if num_manips == 0,
385 * orig_tp = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple)
387 nf_ct_invert_tuplepr(&curr_tuple
,
388 &ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
);
390 get_unique_tuple(&new_tuple
, &curr_tuple
, range
, ct
, maniptype
);
392 if (!nf_ct_tuple_equal(&new_tuple
, &curr_tuple
)) {
393 struct nf_conntrack_tuple reply
;
395 /* Alter conntrack table so will recognize replies. */
396 nf_ct_invert_tuplepr(&reply
, &new_tuple
);
397 nf_conntrack_alter_reply(ct
, &reply
);
399 /* Non-atomic: we own this at the moment. */
400 if (maniptype
== NF_NAT_MANIP_SRC
)
401 ct
->status
|= IPS_SRC_NAT
;
403 ct
->status
|= IPS_DST_NAT
;
406 if (maniptype
== NF_NAT_MANIP_SRC
) {
407 unsigned int srchash
;
409 srchash
= hash_by_src(net
, nf_ct_zone(ct
),
410 &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
);
411 spin_lock_bh(&nf_nat_lock
);
412 /* nf_conntrack_alter_reply might re-allocate extension aera */
415 hlist_add_head_rcu(&nat
->bysource
,
416 &net
->ct
.nat_bysource
[srchash
]);
417 spin_unlock_bh(&nf_nat_lock
);
421 if (maniptype
== NF_NAT_MANIP_DST
)
422 ct
->status
|= IPS_DST_NAT_DONE
;
424 ct
->status
|= IPS_SRC_NAT_DONE
;
428 EXPORT_SYMBOL(nf_nat_setup_info
);
430 /* Do packet manipulations according to nf_nat_setup_info. */
431 unsigned int nf_nat_packet(struct nf_conn
*ct
,
432 enum ip_conntrack_info ctinfo
,
433 unsigned int hooknum
,
436 const struct nf_nat_l3proto
*l3proto
;
437 const struct nf_nat_l4proto
*l4proto
;
438 enum ip_conntrack_dir dir
= CTINFO2DIR(ctinfo
);
439 unsigned long statusbit
;
440 enum nf_nat_manip_type mtype
= HOOK2MANIP(hooknum
);
442 if (mtype
== NF_NAT_MANIP_SRC
)
443 statusbit
= IPS_SRC_NAT
;
445 statusbit
= IPS_DST_NAT
;
447 /* Invert if this is reply dir. */
448 if (dir
== IP_CT_DIR_REPLY
)
449 statusbit
^= IPS_NAT_MASK
;
451 /* Non-atomic: these bits don't change. */
452 if (ct
->status
& statusbit
) {
453 struct nf_conntrack_tuple target
;
455 /* We are aiming to look like inverse of other direction. */
456 nf_ct_invert_tuplepr(&target
, &ct
->tuplehash
[!dir
].tuple
);
458 l3proto
= __nf_nat_l3proto_find(target
.src
.l3num
);
459 l4proto
= __nf_nat_l4proto_find(target
.src
.l3num
,
460 target
.dst
.protonum
);
461 if (!l3proto
->manip_pkt(skb
, 0, l4proto
, &target
, mtype
))
466 EXPORT_SYMBOL_GPL(nf_nat_packet
);
468 struct nf_nat_proto_clean
{
473 /* kill conntracks with affected NAT section */
474 static int nf_nat_proto_remove(struct nf_conn
*i
, void *data
)
476 const struct nf_nat_proto_clean
*clean
= data
;
477 struct nf_conn_nat
*nat
= nfct_nat(i
);
482 if ((clean
->l3proto
&& nf_ct_l3num(i
) != clean
->l3proto
) ||
483 (clean
->l4proto
&& nf_ct_protonum(i
) != clean
->l4proto
))
486 return i
->status
& IPS_NAT_MASK
? 1 : 0;
489 static void nf_nat_l4proto_clean(u8 l3proto
, u8 l4proto
)
491 struct nf_nat_proto_clean clean
= {
499 nf_ct_iterate_cleanup(net
, nf_nat_proto_remove
, &clean
);
503 static void nf_nat_l3proto_clean(u8 l3proto
)
505 struct nf_nat_proto_clean clean
= {
513 nf_ct_iterate_cleanup(net
, nf_nat_proto_remove
, &clean
);
517 /* Protocol registration. */
518 int nf_nat_l4proto_register(u8 l3proto
, const struct nf_nat_l4proto
*l4proto
)
520 const struct nf_nat_l4proto
**l4protos
;
524 mutex_lock(&nf_nat_proto_mutex
);
525 if (nf_nat_l4protos
[l3proto
] == NULL
) {
526 l4protos
= kmalloc(IPPROTO_MAX
* sizeof(struct nf_nat_l4proto
*),
528 if (l4protos
== NULL
) {
533 for (i
= 0; i
< IPPROTO_MAX
; i
++)
534 RCU_INIT_POINTER(l4protos
[i
], &nf_nat_l4proto_unknown
);
536 /* Before making proto_array visible to lockless readers,
537 * we must make sure its content is committed to memory.
541 nf_nat_l4protos
[l3proto
] = l4protos
;
544 if (rcu_dereference_protected(
545 nf_nat_l4protos
[l3proto
][l4proto
->l4proto
],
546 lockdep_is_held(&nf_nat_proto_mutex
)
547 ) != &nf_nat_l4proto_unknown
) {
551 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
][l4proto
->l4proto
], l4proto
);
553 mutex_unlock(&nf_nat_proto_mutex
);
556 EXPORT_SYMBOL_GPL(nf_nat_l4proto_register
);
558 /* No one stores the protocol anywhere; simply delete it. */
559 void nf_nat_l4proto_unregister(u8 l3proto
, const struct nf_nat_l4proto
*l4proto
)
561 mutex_lock(&nf_nat_proto_mutex
);
562 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
][l4proto
->l4proto
],
563 &nf_nat_l4proto_unknown
);
564 mutex_unlock(&nf_nat_proto_mutex
);
567 nf_nat_l4proto_clean(l3proto
, l4proto
->l4proto
);
569 EXPORT_SYMBOL_GPL(nf_nat_l4proto_unregister
);
571 int nf_nat_l3proto_register(const struct nf_nat_l3proto
*l3proto
)
575 err
= nf_ct_l3proto_try_module_get(l3proto
->l3proto
);
579 mutex_lock(&nf_nat_proto_mutex
);
580 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
->l3proto
][IPPROTO_TCP
],
581 &nf_nat_l4proto_tcp
);
582 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
->l3proto
][IPPROTO_UDP
],
583 &nf_nat_l4proto_udp
);
584 mutex_unlock(&nf_nat_proto_mutex
);
586 RCU_INIT_POINTER(nf_nat_l3protos
[l3proto
->l3proto
], l3proto
);
589 EXPORT_SYMBOL_GPL(nf_nat_l3proto_register
);
591 void nf_nat_l3proto_unregister(const struct nf_nat_l3proto
*l3proto
)
593 mutex_lock(&nf_nat_proto_mutex
);
594 RCU_INIT_POINTER(nf_nat_l3protos
[l3proto
->l3proto
], NULL
);
595 mutex_unlock(&nf_nat_proto_mutex
);
598 nf_nat_l3proto_clean(l3proto
->l3proto
);
599 nf_ct_l3proto_module_put(l3proto
->l3proto
);
601 EXPORT_SYMBOL_GPL(nf_nat_l3proto_unregister
);
603 /* No one using conntrack by the time this called. */
604 static void nf_nat_cleanup_conntrack(struct nf_conn
*ct
)
606 struct nf_conn_nat
*nat
= nf_ct_ext_find(ct
, NF_CT_EXT_NAT
);
608 if (nat
== NULL
|| nat
->ct
== NULL
)
611 NF_CT_ASSERT(nat
->ct
->status
& IPS_SRC_NAT_DONE
);
613 spin_lock_bh(&nf_nat_lock
);
614 hlist_del_rcu(&nat
->bysource
);
615 spin_unlock_bh(&nf_nat_lock
);
618 static void nf_nat_move_storage(void *new, void *old
)
620 struct nf_conn_nat
*new_nat
= new;
621 struct nf_conn_nat
*old_nat
= old
;
622 struct nf_conn
*ct
= old_nat
->ct
;
624 if (!ct
|| !(ct
->status
& IPS_SRC_NAT_DONE
))
627 spin_lock_bh(&nf_nat_lock
);
628 hlist_replace_rcu(&old_nat
->bysource
, &new_nat
->bysource
);
629 spin_unlock_bh(&nf_nat_lock
);
632 static struct nf_ct_ext_type nat_extend __read_mostly
= {
633 .len
= sizeof(struct nf_conn_nat
),
634 .align
= __alignof__(struct nf_conn_nat
),
635 .destroy
= nf_nat_cleanup_conntrack
,
636 .move
= nf_nat_move_storage
,
638 .flags
= NF_CT_EXT_F_PREALLOC
,
641 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
643 #include <linux/netfilter/nfnetlink.h>
644 #include <linux/netfilter/nfnetlink_conntrack.h>
646 static const struct nla_policy protonat_nla_policy
[CTA_PROTONAT_MAX
+1] = {
647 [CTA_PROTONAT_PORT_MIN
] = { .type
= NLA_U16
},
648 [CTA_PROTONAT_PORT_MAX
] = { .type
= NLA_U16
},
651 static int nfnetlink_parse_nat_proto(struct nlattr
*attr
,
652 const struct nf_conn
*ct
,
653 struct nf_nat_range
*range
)
655 struct nlattr
*tb
[CTA_PROTONAT_MAX
+1];
656 const struct nf_nat_l4proto
*l4proto
;
659 err
= nla_parse_nested(tb
, CTA_PROTONAT_MAX
, attr
, protonat_nla_policy
);
663 l4proto
= __nf_nat_l4proto_find(nf_ct_l3num(ct
), nf_ct_protonum(ct
));
664 if (l4proto
->nlattr_to_range
)
665 err
= l4proto
->nlattr_to_range(tb
, range
);
670 static const struct nla_policy nat_nla_policy
[CTA_NAT_MAX
+1] = {
671 [CTA_NAT_V4_MINIP
] = { .type
= NLA_U32
},
672 [CTA_NAT_V4_MAXIP
] = { .type
= NLA_U32
},
673 [CTA_NAT_V6_MINIP
] = { .len
= sizeof(struct in6_addr
) },
674 [CTA_NAT_V6_MAXIP
] = { .len
= sizeof(struct in6_addr
) },
675 [CTA_NAT_PROTO
] = { .type
= NLA_NESTED
},
679 nfnetlink_parse_nat(const struct nlattr
*nat
,
680 const struct nf_conn
*ct
, struct nf_nat_range
*range
)
682 const struct nf_nat_l3proto
*l3proto
;
683 struct nlattr
*tb
[CTA_NAT_MAX
+1];
686 memset(range
, 0, sizeof(*range
));
688 err
= nla_parse_nested(tb
, CTA_NAT_MAX
, nat
, nat_nla_policy
);
693 l3proto
= __nf_nat_l3proto_find(nf_ct_l3num(ct
));
694 if (l3proto
== NULL
) {
698 err
= l3proto
->nlattr_to_range(tb
, range
);
702 if (!tb
[CTA_NAT_PROTO
])
705 err
= nfnetlink_parse_nat_proto(tb
[CTA_NAT_PROTO
], ct
, range
);
712 nfnetlink_parse_nat_setup(struct nf_conn
*ct
,
713 enum nf_nat_manip_type manip
,
714 const struct nlattr
*attr
)
716 struct nf_nat_range range
;
719 err
= nfnetlink_parse_nat(attr
, ct
, &range
);
722 if (nf_nat_initialized(ct
, manip
))
725 return nf_nat_setup_info(ct
, &range
, manip
);
729 nfnetlink_parse_nat_setup(struct nf_conn
*ct
,
730 enum nf_nat_manip_type manip
,
731 const struct nlattr
*attr
)
737 static int __net_init
nf_nat_net_init(struct net
*net
)
739 /* Leave them the same for the moment. */
740 net
->ct
.nat_htable_size
= net
->ct
.htable_size
;
741 net
->ct
.nat_bysource
= nf_ct_alloc_hashtable(&net
->ct
.nat_htable_size
, 0);
742 if (!net
->ct
.nat_bysource
)
747 static void __net_exit
nf_nat_net_exit(struct net
*net
)
749 struct nf_nat_proto_clean clean
= {};
751 nf_ct_iterate_cleanup(net
, &nf_nat_proto_remove
, &clean
);
753 nf_ct_free_hashtable(net
->ct
.nat_bysource
, net
->ct
.nat_htable_size
);
756 static struct pernet_operations nf_nat_net_ops
= {
757 .init
= nf_nat_net_init
,
758 .exit
= nf_nat_net_exit
,
761 static struct nf_ct_helper_expectfn follow_master_nat
= {
762 .name
= "nat-follow-master",
763 .expectfn
= nf_nat_follow_master
,
766 static struct nfq_ct_nat_hook nfq_ct_nat
= {
767 .seq_adjust
= nf_nat_tcp_seq_adjust
,
770 static int __init
nf_nat_init(void)
774 ret
= nf_ct_extend_register(&nat_extend
);
776 printk(KERN_ERR
"nf_nat_core: Unable to register extension\n");
780 ret
= register_pernet_subsys(&nf_nat_net_ops
);
784 nf_ct_helper_expectfn_register(&follow_master_nat
);
786 /* Initialize fake conntrack so that NAT will skip it */
787 nf_ct_untracked_status_or(IPS_NAT_DONE_MASK
);
789 BUG_ON(nf_nat_seq_adjust_hook
!= NULL
);
790 RCU_INIT_POINTER(nf_nat_seq_adjust_hook
, nf_nat_seq_adjust
);
791 BUG_ON(nfnetlink_parse_nat_setup_hook
!= NULL
);
792 RCU_INIT_POINTER(nfnetlink_parse_nat_setup_hook
,
793 nfnetlink_parse_nat_setup
);
794 BUG_ON(nf_ct_nat_offset
!= NULL
);
795 RCU_INIT_POINTER(nf_ct_nat_offset
, nf_nat_get_offset
);
796 RCU_INIT_POINTER(nfq_ct_nat_hook
, &nfq_ct_nat
);
798 BUG_ON(nf_nat_decode_session_hook
!= NULL
);
799 RCU_INIT_POINTER(nf_nat_decode_session_hook
, __nf_nat_decode_session
);
804 nf_ct_extend_unregister(&nat_extend
);
808 static void __exit
nf_nat_cleanup(void)
812 unregister_pernet_subsys(&nf_nat_net_ops
);
813 nf_ct_extend_unregister(&nat_extend
);
814 nf_ct_helper_expectfn_unregister(&follow_master_nat
);
815 RCU_INIT_POINTER(nf_nat_seq_adjust_hook
, NULL
);
816 RCU_INIT_POINTER(nfnetlink_parse_nat_setup_hook
, NULL
);
817 RCU_INIT_POINTER(nf_ct_nat_offset
, NULL
);
818 RCU_INIT_POINTER(nfq_ct_nat_hook
, NULL
);
820 RCU_INIT_POINTER(nf_nat_decode_session_hook
, NULL
);
822 for (i
= 0; i
< NFPROTO_NUMPROTO
; i
++)
823 kfree(nf_nat_l4protos
[i
]);
827 MODULE_LICENSE("GPL");
829 module_init(nf_nat_init
);
830 module_exit(nf_nat_cleanup
);