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
;
91 if (xfrm_decode_session(skb
, &fl
, family
) < 0)
96 dst
= ((struct xfrm_dst
*)dst
)->route
;
99 dst
= xfrm_lookup(dev_net(dst
->dev
), dst
, &fl
, skb
->sk
, 0);
104 skb_dst_set(skb
, dst
);
106 /* Change in oif may mean change in hh_len. */
107 hh_len
= skb_dst(skb
)->dev
->hard_header_len
;
108 if (skb_headroom(skb
) < hh_len
&&
109 pskb_expand_head(skb
, hh_len
- skb_headroom(skb
), 0, GFP_ATOMIC
))
113 EXPORT_SYMBOL(nf_xfrm_me_harder
);
114 #endif /* CONFIG_XFRM */
116 /* We keep an extra hash for each conntrack, for fast searching. */
117 static inline unsigned int
118 hash_by_src(const struct net
*net
, u16 zone
,
119 const struct nf_conntrack_tuple
*tuple
)
123 /* Original src, to ensure we map it consistently if poss. */
124 hash
= jhash2((u32
*)&tuple
->src
, sizeof(tuple
->src
) / sizeof(u32
),
125 tuple
->dst
.protonum
^ zone
^ nf_conntrack_hash_rnd
);
126 return ((u64
)hash
* net
->ct
.nat_htable_size
) >> 32;
129 /* Is this tuple already taken? (not by us) */
131 nf_nat_used_tuple(const struct nf_conntrack_tuple
*tuple
,
132 const struct nf_conn
*ignored_conntrack
)
134 /* Conntrack tracking doesn't keep track of outgoing tuples; only
135 * incoming ones. NAT means they don't have a fixed mapping,
136 * so we invert the tuple and look for the incoming reply.
138 * We could keep a separate hash if this proves too slow.
140 struct nf_conntrack_tuple reply
;
142 nf_ct_invert_tuplepr(&reply
, tuple
);
143 return nf_conntrack_tuple_taken(&reply
, ignored_conntrack
);
145 EXPORT_SYMBOL(nf_nat_used_tuple
);
147 /* If we source map this tuple so reply looks like reply_tuple, will
148 * that meet the constraints of range.
150 static int in_range(const struct nf_nat_l3proto
*l3proto
,
151 const struct nf_nat_l4proto
*l4proto
,
152 const struct nf_conntrack_tuple
*tuple
,
153 const struct nf_nat_range
*range
)
155 /* If we are supposed to map IPs, then we must be in the
156 * range specified, otherwise let this drag us onto a new src IP.
158 if (range
->flags
& NF_NAT_RANGE_MAP_IPS
&&
159 !l3proto
->in_range(tuple
, range
))
162 if (!(range
->flags
& NF_NAT_RANGE_PROTO_SPECIFIED
) ||
163 l4proto
->in_range(tuple
, NF_NAT_MANIP_SRC
,
164 &range
->min_proto
, &range
->max_proto
))
171 same_src(const struct nf_conn
*ct
,
172 const struct nf_conntrack_tuple
*tuple
)
174 const struct nf_conntrack_tuple
*t
;
176 t
= &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
;
177 return (t
->dst
.protonum
== tuple
->dst
.protonum
&&
178 nf_inet_addr_cmp(&t
->src
.u3
, &tuple
->src
.u3
) &&
179 t
->src
.u
.all
== tuple
->src
.u
.all
);
182 /* Only called for SRC manip */
184 find_appropriate_src(struct net
*net
, u16 zone
,
185 const struct nf_nat_l3proto
*l3proto
,
186 const struct nf_nat_l4proto
*l4proto
,
187 const struct nf_conntrack_tuple
*tuple
,
188 struct nf_conntrack_tuple
*result
,
189 const struct nf_nat_range
*range
)
191 unsigned int h
= hash_by_src(net
, zone
, tuple
);
192 const struct nf_conn_nat
*nat
;
193 const struct nf_conn
*ct
;
195 hlist_for_each_entry_rcu(nat
, &net
->ct
.nat_bysource
[h
], bysource
) {
197 if (same_src(ct
, tuple
) && nf_ct_zone(ct
) == zone
) {
198 /* Copy source part from reply tuple. */
199 nf_ct_invert_tuplepr(result
,
200 &ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
);
201 result
->dst
= tuple
->dst
;
203 if (in_range(l3proto
, l4proto
, result
, range
))
210 /* For [FUTURE] fragmentation handling, we want the least-used
211 * src-ip/dst-ip/proto triple. Fairness doesn't come into it. Thus
212 * if the range specifies 1.2.3.4 ports 10000-10005 and 1.2.3.5 ports
213 * 1-65535, we don't do pro-rata allocation based on ports; we choose
214 * the ip with the lowest src-ip/dst-ip/proto usage.
217 find_best_ips_proto(u16 zone
, struct nf_conntrack_tuple
*tuple
,
218 const struct nf_nat_range
*range
,
219 const struct nf_conn
*ct
,
220 enum nf_nat_manip_type maniptype
)
222 union nf_inet_addr
*var_ipp
;
225 u32 minip
, maxip
, j
, dist
;
228 /* No IP mapping? Do nothing. */
229 if (!(range
->flags
& NF_NAT_RANGE_MAP_IPS
))
232 if (maniptype
== NF_NAT_MANIP_SRC
)
233 var_ipp
= &tuple
->src
.u3
;
235 var_ipp
= &tuple
->dst
.u3
;
237 /* Fast path: only one choice. */
238 if (nf_inet_addr_cmp(&range
->min_addr
, &range
->max_addr
)) {
239 *var_ipp
= range
->min_addr
;
243 if (nf_ct_l3num(ct
) == NFPROTO_IPV4
)
244 max
= sizeof(var_ipp
->ip
) / sizeof(u32
) - 1;
246 max
= sizeof(var_ipp
->ip6
) / sizeof(u32
) - 1;
248 /* Hashing source and destination IPs gives a fairly even
249 * spread in practice (if there are a small number of IPs
250 * involved, there usually aren't that many connections
251 * anyway). The consistency means that servers see the same
252 * client coming from the same IP (some Internet Banking sites
253 * like this), even across reboots.
255 j
= jhash2((u32
*)&tuple
->src
.u3
, sizeof(tuple
->src
.u3
) / sizeof(u32
),
256 range
->flags
& NF_NAT_RANGE_PERSISTENT
?
257 0 : (__force u32
)tuple
->dst
.u3
.all
[max
] ^ zone
);
260 for (i
= 0; i
<= max
; i
++) {
261 /* If first bytes of the address are at the maximum, use the
262 * distance. Otherwise use the full range.
265 minip
= ntohl((__force __be32
)range
->min_addr
.all
[i
]);
266 maxip
= ntohl((__force __be32
)range
->max_addr
.all
[i
]);
267 dist
= maxip
- minip
+ 1;
273 var_ipp
->all
[i
] = (__force __u32
)
274 htonl(minip
+ (((u64
)j
* dist
) >> 32));
275 if (var_ipp
->all
[i
] != range
->max_addr
.all
[i
])
278 if (!(range
->flags
& NF_NAT_RANGE_PERSISTENT
))
279 j
^= (__force u32
)tuple
->dst
.u3
.all
[i
];
283 /* Manipulate the tuple into the range given. For NF_INET_POST_ROUTING,
284 * we change the source to map into the range. For NF_INET_PRE_ROUTING
285 * and NF_INET_LOCAL_OUT, we change the destination to map into the
286 * range. It might not be possible to get a unique tuple, but we try.
287 * At worst (or if we race), we will end up with a final duplicate in
288 * __ip_conntrack_confirm and drop the packet. */
290 get_unique_tuple(struct nf_conntrack_tuple
*tuple
,
291 const struct nf_conntrack_tuple
*orig_tuple
,
292 const struct nf_nat_range
*range
,
294 enum nf_nat_manip_type maniptype
)
296 const struct nf_nat_l3proto
*l3proto
;
297 const struct nf_nat_l4proto
*l4proto
;
298 struct net
*net
= nf_ct_net(ct
);
299 u16 zone
= nf_ct_zone(ct
);
302 l3proto
= __nf_nat_l3proto_find(orig_tuple
->src
.l3num
);
303 l4proto
= __nf_nat_l4proto_find(orig_tuple
->src
.l3num
,
304 orig_tuple
->dst
.protonum
);
306 /* 1) If this srcip/proto/src-proto-part is currently mapped,
307 * and that same mapping gives a unique tuple within the given
310 * This is only required for source (ie. NAT/masq) mappings.
311 * So far, we don't do local source mappings, so multiple
312 * manips not an issue.
314 if (maniptype
== NF_NAT_MANIP_SRC
&&
315 !(range
->flags
& NF_NAT_RANGE_PROTO_RANDOM
)) {
316 /* try the original tuple first */
317 if (in_range(l3proto
, l4proto
, orig_tuple
, range
)) {
318 if (!nf_nat_used_tuple(orig_tuple
, ct
)) {
319 *tuple
= *orig_tuple
;
322 } else if (find_appropriate_src(net
, zone
, l3proto
, l4proto
,
323 orig_tuple
, tuple
, range
)) {
324 pr_debug("get_unique_tuple: Found current src map\n");
325 if (!nf_nat_used_tuple(tuple
, ct
))
330 /* 2) Select the least-used IP/proto combination in the given range */
331 *tuple
= *orig_tuple
;
332 find_best_ips_proto(zone
, tuple
, range
, ct
, maniptype
);
334 /* 3) The per-protocol part of the manip is made to map into
335 * the range to make a unique tuple.
338 /* Only bother mapping if it's not already in range and unique */
339 if (!(range
->flags
& NF_NAT_RANGE_PROTO_RANDOM
)) {
340 if (range
->flags
& NF_NAT_RANGE_PROTO_SPECIFIED
) {
341 if (l4proto
->in_range(tuple
, maniptype
,
343 &range
->max_proto
) &&
344 (range
->min_proto
.all
== range
->max_proto
.all
||
345 !nf_nat_used_tuple(tuple
, ct
)))
347 } else if (!nf_nat_used_tuple(tuple
, ct
)) {
352 /* Last change: get protocol to try to obtain unique tuple. */
353 l4proto
->unique_tuple(l3proto
, tuple
, range
, maniptype
, ct
);
359 nf_nat_setup_info(struct nf_conn
*ct
,
360 const struct nf_nat_range
*range
,
361 enum nf_nat_manip_type maniptype
)
363 struct net
*net
= nf_ct_net(ct
);
364 struct nf_conntrack_tuple curr_tuple
, new_tuple
;
365 struct nf_conn_nat
*nat
;
367 /* nat helper or nfctnetlink also setup binding */
370 nat
= nf_ct_ext_add(ct
, NF_CT_EXT_NAT
, GFP_ATOMIC
);
372 pr_debug("failed to add NAT extension\n");
377 NF_CT_ASSERT(maniptype
== NF_NAT_MANIP_SRC
||
378 maniptype
== NF_NAT_MANIP_DST
);
379 BUG_ON(nf_nat_initialized(ct
, maniptype
));
381 /* What we've got will look like inverse of reply. Normally
382 * this is what is in the conntrack, except for prior
383 * manipulations (future optimization: if num_manips == 0,
384 * orig_tp = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple)
386 nf_ct_invert_tuplepr(&curr_tuple
,
387 &ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
);
389 get_unique_tuple(&new_tuple
, &curr_tuple
, range
, ct
, maniptype
);
391 if (!nf_ct_tuple_equal(&new_tuple
, &curr_tuple
)) {
392 struct nf_conntrack_tuple reply
;
394 /* Alter conntrack table so will recognize replies. */
395 nf_ct_invert_tuplepr(&reply
, &new_tuple
);
396 nf_conntrack_alter_reply(ct
, &reply
);
398 /* Non-atomic: we own this at the moment. */
399 if (maniptype
== NF_NAT_MANIP_SRC
)
400 ct
->status
|= IPS_SRC_NAT
;
402 ct
->status
|= IPS_DST_NAT
;
405 if (maniptype
== NF_NAT_MANIP_SRC
) {
406 unsigned int srchash
;
408 srchash
= hash_by_src(net
, nf_ct_zone(ct
),
409 &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
);
410 spin_lock_bh(&nf_nat_lock
);
411 /* nf_conntrack_alter_reply might re-allocate extension aera */
414 hlist_add_head_rcu(&nat
->bysource
,
415 &net
->ct
.nat_bysource
[srchash
]);
416 spin_unlock_bh(&nf_nat_lock
);
420 if (maniptype
== NF_NAT_MANIP_DST
)
421 ct
->status
|= IPS_DST_NAT_DONE
;
423 ct
->status
|= IPS_SRC_NAT_DONE
;
427 EXPORT_SYMBOL(nf_nat_setup_info
);
429 /* Do packet manipulations according to nf_nat_setup_info. */
430 unsigned int nf_nat_packet(struct nf_conn
*ct
,
431 enum ip_conntrack_info ctinfo
,
432 unsigned int hooknum
,
435 const struct nf_nat_l3proto
*l3proto
;
436 const struct nf_nat_l4proto
*l4proto
;
437 enum ip_conntrack_dir dir
= CTINFO2DIR(ctinfo
);
438 unsigned long statusbit
;
439 enum nf_nat_manip_type mtype
= HOOK2MANIP(hooknum
);
441 if (mtype
== NF_NAT_MANIP_SRC
)
442 statusbit
= IPS_SRC_NAT
;
444 statusbit
= IPS_DST_NAT
;
446 /* Invert if this is reply dir. */
447 if (dir
== IP_CT_DIR_REPLY
)
448 statusbit
^= IPS_NAT_MASK
;
450 /* Non-atomic: these bits don't change. */
451 if (ct
->status
& statusbit
) {
452 struct nf_conntrack_tuple target
;
454 /* We are aiming to look like inverse of other direction. */
455 nf_ct_invert_tuplepr(&target
, &ct
->tuplehash
[!dir
].tuple
);
457 l3proto
= __nf_nat_l3proto_find(target
.src
.l3num
);
458 l4proto
= __nf_nat_l4proto_find(target
.src
.l3num
,
459 target
.dst
.protonum
);
460 if (!l3proto
->manip_pkt(skb
, 0, l4proto
, &target
, mtype
))
465 EXPORT_SYMBOL_GPL(nf_nat_packet
);
467 struct nf_nat_proto_clean
{
473 /* Clear NAT section of all conntracks, in case we're loaded again. */
474 static int nf_nat_proto_clean(struct nf_conn
*i
, void *data
)
476 const struct nf_nat_proto_clean
*clean
= data
;
477 struct nf_conn_nat
*nat
= nfct_nat(i
);
481 if (!(i
->status
& IPS_SRC_NAT_DONE
))
483 if ((clean
->l3proto
&& nf_ct_l3num(i
) != clean
->l3proto
) ||
484 (clean
->l4proto
&& nf_ct_protonum(i
) != clean
->l4proto
))
488 spin_lock_bh(&nf_nat_lock
);
489 hlist_del_rcu(&nat
->bysource
);
490 spin_unlock_bh(&nf_nat_lock
);
492 memset(nat
, 0, sizeof(*nat
));
493 i
->status
&= ~(IPS_NAT_MASK
| IPS_NAT_DONE_MASK
|
499 static void nf_nat_l4proto_clean(u8 l3proto
, u8 l4proto
)
501 struct nf_nat_proto_clean clean
= {
508 /* Step 1 - remove from bysource hash */
511 nf_ct_iterate_cleanup(net
, nf_nat_proto_clean
, &clean
);
514 /* Step 2 - clean NAT section */
517 nf_ct_iterate_cleanup(net
, nf_nat_proto_clean
, &clean
);
521 static void nf_nat_l3proto_clean(u8 l3proto
)
523 struct nf_nat_proto_clean clean
= {
529 /* Step 1 - remove from bysource hash */
532 nf_ct_iterate_cleanup(net
, nf_nat_proto_clean
, &clean
);
535 /* Step 2 - clean NAT section */
538 nf_ct_iterate_cleanup(net
, nf_nat_proto_clean
, &clean
);
542 /* Protocol registration. */
543 int nf_nat_l4proto_register(u8 l3proto
, const struct nf_nat_l4proto
*l4proto
)
545 const struct nf_nat_l4proto
**l4protos
;
549 mutex_lock(&nf_nat_proto_mutex
);
550 if (nf_nat_l4protos
[l3proto
] == NULL
) {
551 l4protos
= kmalloc(IPPROTO_MAX
* sizeof(struct nf_nat_l4proto
*),
553 if (l4protos
== NULL
) {
558 for (i
= 0; i
< IPPROTO_MAX
; i
++)
559 RCU_INIT_POINTER(l4protos
[i
], &nf_nat_l4proto_unknown
);
561 /* Before making proto_array visible to lockless readers,
562 * we must make sure its content is committed to memory.
566 nf_nat_l4protos
[l3proto
] = l4protos
;
569 if (rcu_dereference_protected(
570 nf_nat_l4protos
[l3proto
][l4proto
->l4proto
],
571 lockdep_is_held(&nf_nat_proto_mutex
)
572 ) != &nf_nat_l4proto_unknown
) {
576 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
][l4proto
->l4proto
], l4proto
);
578 mutex_unlock(&nf_nat_proto_mutex
);
581 EXPORT_SYMBOL_GPL(nf_nat_l4proto_register
);
583 /* No one stores the protocol anywhere; simply delete it. */
584 void nf_nat_l4proto_unregister(u8 l3proto
, const struct nf_nat_l4proto
*l4proto
)
586 mutex_lock(&nf_nat_proto_mutex
);
587 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
][l4proto
->l4proto
],
588 &nf_nat_l4proto_unknown
);
589 mutex_unlock(&nf_nat_proto_mutex
);
592 nf_nat_l4proto_clean(l3proto
, l4proto
->l4proto
);
594 EXPORT_SYMBOL_GPL(nf_nat_l4proto_unregister
);
596 int nf_nat_l3proto_register(const struct nf_nat_l3proto
*l3proto
)
600 err
= nf_ct_l3proto_try_module_get(l3proto
->l3proto
);
604 mutex_lock(&nf_nat_proto_mutex
);
605 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
->l3proto
][IPPROTO_TCP
],
606 &nf_nat_l4proto_tcp
);
607 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
->l3proto
][IPPROTO_UDP
],
608 &nf_nat_l4proto_udp
);
609 mutex_unlock(&nf_nat_proto_mutex
);
611 RCU_INIT_POINTER(nf_nat_l3protos
[l3proto
->l3proto
], l3proto
);
614 EXPORT_SYMBOL_GPL(nf_nat_l3proto_register
);
616 void nf_nat_l3proto_unregister(const struct nf_nat_l3proto
*l3proto
)
618 mutex_lock(&nf_nat_proto_mutex
);
619 RCU_INIT_POINTER(nf_nat_l3protos
[l3proto
->l3proto
], NULL
);
620 mutex_unlock(&nf_nat_proto_mutex
);
623 nf_nat_l3proto_clean(l3proto
->l3proto
);
624 nf_ct_l3proto_module_put(l3proto
->l3proto
);
626 EXPORT_SYMBOL_GPL(nf_nat_l3proto_unregister
);
628 /* No one using conntrack by the time this called. */
629 static void nf_nat_cleanup_conntrack(struct nf_conn
*ct
)
631 struct nf_conn_nat
*nat
= nf_ct_ext_find(ct
, NF_CT_EXT_NAT
);
633 if (nat
== NULL
|| nat
->ct
== NULL
)
636 NF_CT_ASSERT(nat
->ct
->status
& IPS_SRC_NAT_DONE
);
638 spin_lock_bh(&nf_nat_lock
);
639 hlist_del_rcu(&nat
->bysource
);
640 spin_unlock_bh(&nf_nat_lock
);
643 static void nf_nat_move_storage(void *new, void *old
)
645 struct nf_conn_nat
*new_nat
= new;
646 struct nf_conn_nat
*old_nat
= old
;
647 struct nf_conn
*ct
= old_nat
->ct
;
649 if (!ct
|| !(ct
->status
& IPS_SRC_NAT_DONE
))
652 spin_lock_bh(&nf_nat_lock
);
653 hlist_replace_rcu(&old_nat
->bysource
, &new_nat
->bysource
);
654 spin_unlock_bh(&nf_nat_lock
);
657 static struct nf_ct_ext_type nat_extend __read_mostly
= {
658 .len
= sizeof(struct nf_conn_nat
),
659 .align
= __alignof__(struct nf_conn_nat
),
660 .destroy
= nf_nat_cleanup_conntrack
,
661 .move
= nf_nat_move_storage
,
663 .flags
= NF_CT_EXT_F_PREALLOC
,
666 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
668 #include <linux/netfilter/nfnetlink.h>
669 #include <linux/netfilter/nfnetlink_conntrack.h>
671 static const struct nla_policy protonat_nla_policy
[CTA_PROTONAT_MAX
+1] = {
672 [CTA_PROTONAT_PORT_MIN
] = { .type
= NLA_U16
},
673 [CTA_PROTONAT_PORT_MAX
] = { .type
= NLA_U16
},
676 static int nfnetlink_parse_nat_proto(struct nlattr
*attr
,
677 const struct nf_conn
*ct
,
678 struct nf_nat_range
*range
)
680 struct nlattr
*tb
[CTA_PROTONAT_MAX
+1];
681 const struct nf_nat_l4proto
*l4proto
;
684 err
= nla_parse_nested(tb
, CTA_PROTONAT_MAX
, attr
, protonat_nla_policy
);
688 l4proto
= __nf_nat_l4proto_find(nf_ct_l3num(ct
), nf_ct_protonum(ct
));
689 if (l4proto
->nlattr_to_range
)
690 err
= l4proto
->nlattr_to_range(tb
, range
);
695 static const struct nla_policy nat_nla_policy
[CTA_NAT_MAX
+1] = {
696 [CTA_NAT_V4_MINIP
] = { .type
= NLA_U32
},
697 [CTA_NAT_V4_MAXIP
] = { .type
= NLA_U32
},
698 [CTA_NAT_V6_MINIP
] = { .len
= sizeof(struct in6_addr
) },
699 [CTA_NAT_V6_MAXIP
] = { .len
= sizeof(struct in6_addr
) },
700 [CTA_NAT_PROTO
] = { .type
= NLA_NESTED
},
704 nfnetlink_parse_nat(const struct nlattr
*nat
,
705 const struct nf_conn
*ct
, struct nf_nat_range
*range
)
707 const struct nf_nat_l3proto
*l3proto
;
708 struct nlattr
*tb
[CTA_NAT_MAX
+1];
711 memset(range
, 0, sizeof(*range
));
713 err
= nla_parse_nested(tb
, CTA_NAT_MAX
, nat
, nat_nla_policy
);
718 l3proto
= __nf_nat_l3proto_find(nf_ct_l3num(ct
));
719 if (l3proto
== NULL
) {
723 err
= l3proto
->nlattr_to_range(tb
, range
);
727 if (!tb
[CTA_NAT_PROTO
])
730 err
= nfnetlink_parse_nat_proto(tb
[CTA_NAT_PROTO
], ct
, range
);
737 nfnetlink_parse_nat_setup(struct nf_conn
*ct
,
738 enum nf_nat_manip_type manip
,
739 const struct nlattr
*attr
)
741 struct nf_nat_range range
;
744 err
= nfnetlink_parse_nat(attr
, ct
, &range
);
747 if (nf_nat_initialized(ct
, manip
))
750 return nf_nat_setup_info(ct
, &range
, manip
);
754 nfnetlink_parse_nat_setup(struct nf_conn
*ct
,
755 enum nf_nat_manip_type manip
,
756 const struct nlattr
*attr
)
762 static int __net_init
nf_nat_net_init(struct net
*net
)
764 /* Leave them the same for the moment. */
765 net
->ct
.nat_htable_size
= net
->ct
.htable_size
;
766 net
->ct
.nat_bysource
= nf_ct_alloc_hashtable(&net
->ct
.nat_htable_size
, 0);
767 if (!net
->ct
.nat_bysource
)
772 static void __net_exit
nf_nat_net_exit(struct net
*net
)
774 struct nf_nat_proto_clean clean
= {};
776 nf_ct_iterate_cleanup(net
, &nf_nat_proto_clean
, &clean
);
778 nf_ct_free_hashtable(net
->ct
.nat_bysource
, net
->ct
.nat_htable_size
);
781 static struct pernet_operations nf_nat_net_ops
= {
782 .init
= nf_nat_net_init
,
783 .exit
= nf_nat_net_exit
,
786 static struct nf_ct_helper_expectfn follow_master_nat
= {
787 .name
= "nat-follow-master",
788 .expectfn
= nf_nat_follow_master
,
791 static struct nfq_ct_nat_hook nfq_ct_nat
= {
792 .seq_adjust
= nf_nat_tcp_seq_adjust
,
795 static int __init
nf_nat_init(void)
799 ret
= nf_ct_extend_register(&nat_extend
);
801 printk(KERN_ERR
"nf_nat_core: Unable to register extension\n");
805 ret
= register_pernet_subsys(&nf_nat_net_ops
);
809 nf_ct_helper_expectfn_register(&follow_master_nat
);
811 /* Initialize fake conntrack so that NAT will skip it */
812 nf_ct_untracked_status_or(IPS_NAT_DONE_MASK
);
814 BUG_ON(nf_nat_seq_adjust_hook
!= NULL
);
815 RCU_INIT_POINTER(nf_nat_seq_adjust_hook
, nf_nat_seq_adjust
);
816 BUG_ON(nfnetlink_parse_nat_setup_hook
!= NULL
);
817 RCU_INIT_POINTER(nfnetlink_parse_nat_setup_hook
,
818 nfnetlink_parse_nat_setup
);
819 BUG_ON(nf_ct_nat_offset
!= NULL
);
820 RCU_INIT_POINTER(nf_ct_nat_offset
, nf_nat_get_offset
);
821 RCU_INIT_POINTER(nfq_ct_nat_hook
, &nfq_ct_nat
);
823 BUG_ON(nf_nat_decode_session_hook
!= NULL
);
824 RCU_INIT_POINTER(nf_nat_decode_session_hook
, __nf_nat_decode_session
);
829 nf_ct_extend_unregister(&nat_extend
);
833 static void __exit
nf_nat_cleanup(void)
837 unregister_pernet_subsys(&nf_nat_net_ops
);
838 nf_ct_extend_unregister(&nat_extend
);
839 nf_ct_helper_expectfn_unregister(&follow_master_nat
);
840 RCU_INIT_POINTER(nf_nat_seq_adjust_hook
, NULL
);
841 RCU_INIT_POINTER(nfnetlink_parse_nat_setup_hook
, NULL
);
842 RCU_INIT_POINTER(nf_ct_nat_offset
, NULL
);
843 RCU_INIT_POINTER(nfq_ct_nat_hook
, NULL
);
845 RCU_INIT_POINTER(nf_nat_decode_session_hook
, NULL
);
847 for (i
= 0; i
< NFPROTO_NUMPROTO
; i
++)
848 kfree(nf_nat_l4protos
[i
]);
852 MODULE_LICENSE("GPL");
854 module_init(nf_nat_init
);
855 module_exit(nf_nat_cleanup
);