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_seqadj.h>
29 #include <net/netfilter/nf_conntrack_l3proto.h>
30 #include <net/netfilter/nf_conntrack_zones.h>
31 #include <linux/netfilter/nf_nat.h>
33 static DEFINE_MUTEX(nf_nat_proto_mutex
);
34 static const struct nf_nat_l3proto __rcu
*nf_nat_l3protos
[NFPROTO_NUMPROTO
]
36 static const struct nf_nat_l4proto __rcu
**nf_nat_l4protos
[NFPROTO_NUMPROTO
]
39 struct nf_nat_conn_key
{
40 const struct net
*net
;
41 const struct nf_conntrack_tuple
*tuple
;
42 const struct nf_conntrack_zone
*zone
;
45 static struct rhashtable nf_nat_bysource_table
;
47 inline const struct nf_nat_l3proto
*
48 __nf_nat_l3proto_find(u8 family
)
50 return rcu_dereference(nf_nat_l3protos
[family
]);
53 inline const struct nf_nat_l4proto
*
54 __nf_nat_l4proto_find(u8 family
, u8 protonum
)
56 return rcu_dereference(nf_nat_l4protos
[family
][protonum
]);
58 EXPORT_SYMBOL_GPL(__nf_nat_l4proto_find
);
61 static void __nf_nat_decode_session(struct sk_buff
*skb
, struct flowi
*fl
)
63 const struct nf_nat_l3proto
*l3proto
;
64 const struct nf_conn
*ct
;
65 enum ip_conntrack_info ctinfo
;
66 enum ip_conntrack_dir dir
;
67 unsigned long statusbit
;
70 ct
= nf_ct_get(skb
, &ctinfo
);
74 family
= ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
.src
.l3num
;
76 l3proto
= __nf_nat_l3proto_find(family
);
80 dir
= CTINFO2DIR(ctinfo
);
81 if (dir
== IP_CT_DIR_ORIGINAL
)
82 statusbit
= IPS_DST_NAT
;
84 statusbit
= IPS_SRC_NAT
;
86 l3proto
->decode_session(skb
, ct
, dir
, statusbit
, fl
);
91 int nf_xfrm_me_harder(struct net
*net
, struct sk_buff
*skb
, unsigned int family
)
95 struct dst_entry
*dst
;
98 err
= xfrm_decode_session(skb
, &fl
, family
);
104 dst
= ((struct xfrm_dst
*)dst
)->route
;
107 dst
= xfrm_lookup(net
, dst
, &fl
, skb
->sk
, 0);
112 skb_dst_set(skb
, dst
);
114 /* Change in oif may mean change in hh_len. */
115 hh_len
= skb_dst(skb
)->dev
->hard_header_len
;
116 if (skb_headroom(skb
) < hh_len
&&
117 pskb_expand_head(skb
, hh_len
- skb_headroom(skb
), 0, GFP_ATOMIC
))
121 EXPORT_SYMBOL(nf_xfrm_me_harder
);
122 #endif /* CONFIG_XFRM */
124 static u32
nf_nat_bysource_hash(const void *data
, u32 len
, u32 seed
)
126 const struct nf_conntrack_tuple
*t
;
127 const struct nf_conn
*ct
= data
;
129 t
= &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
;
130 /* Original src, to ensure we map it consistently if poss. */
132 seed
^= net_hash_mix(nf_ct_net(ct
));
133 return jhash2((const u32
*)&t
->src
, sizeof(t
->src
) / sizeof(u32
),
134 t
->dst
.protonum
^ seed
);
137 /* Is this tuple already taken? (not by us) */
139 nf_nat_used_tuple(const struct nf_conntrack_tuple
*tuple
,
140 const struct nf_conn
*ignored_conntrack
)
142 /* Conntrack tracking doesn't keep track of outgoing tuples; only
143 * incoming ones. NAT means they don't have a fixed mapping,
144 * so we invert the tuple and look for the incoming reply.
146 * We could keep a separate hash if this proves too slow.
148 struct nf_conntrack_tuple reply
;
150 nf_ct_invert_tuplepr(&reply
, tuple
);
151 return nf_conntrack_tuple_taken(&reply
, ignored_conntrack
);
153 EXPORT_SYMBOL(nf_nat_used_tuple
);
155 /* If we source map this tuple so reply looks like reply_tuple, will
156 * that meet the constraints of range.
158 static int in_range(const struct nf_nat_l3proto
*l3proto
,
159 const struct nf_nat_l4proto
*l4proto
,
160 const struct nf_conntrack_tuple
*tuple
,
161 const struct nf_nat_range
*range
)
163 /* If we are supposed to map IPs, then we must be in the
164 * range specified, otherwise let this drag us onto a new src IP.
166 if (range
->flags
& NF_NAT_RANGE_MAP_IPS
&&
167 !l3proto
->in_range(tuple
, range
))
170 if (!(range
->flags
& NF_NAT_RANGE_PROTO_SPECIFIED
) ||
171 l4proto
->in_range(tuple
, NF_NAT_MANIP_SRC
,
172 &range
->min_proto
, &range
->max_proto
))
179 same_src(const struct nf_conn
*ct
,
180 const struct nf_conntrack_tuple
*tuple
)
182 const struct nf_conntrack_tuple
*t
;
184 t
= &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
;
185 return (t
->dst
.protonum
== tuple
->dst
.protonum
&&
186 nf_inet_addr_cmp(&t
->src
.u3
, &tuple
->src
.u3
) &&
187 t
->src
.u
.all
== tuple
->src
.u
.all
);
190 static int nf_nat_bysource_cmp(struct rhashtable_compare_arg
*arg
,
193 const struct nf_nat_conn_key
*key
= arg
->key
;
194 const struct nf_conn
*ct
= obj
;
196 return same_src(ct
, key
->tuple
) &&
197 net_eq(nf_ct_net(ct
), key
->net
) &&
198 nf_ct_zone_equal(ct
, key
->zone
, IP_CT_DIR_ORIGINAL
);
201 static struct rhashtable_params nf_nat_bysource_params
= {
202 .head_offset
= offsetof(struct nf_conn
, nat_bysource
),
203 .obj_hashfn
= nf_nat_bysource_hash
,
204 .obj_cmpfn
= nf_nat_bysource_cmp
,
207 .nulls_base
= (1U << RHT_BASE_SHIFT
),
210 /* Only called for SRC manip */
212 find_appropriate_src(struct net
*net
,
213 const struct nf_conntrack_zone
*zone
,
214 const struct nf_nat_l3proto
*l3proto
,
215 const struct nf_nat_l4proto
*l4proto
,
216 const struct nf_conntrack_tuple
*tuple
,
217 struct nf_conntrack_tuple
*result
,
218 const struct nf_nat_range
*range
)
220 const struct nf_conn
*ct
;
221 struct nf_nat_conn_key key
= {
227 ct
= rhashtable_lookup_fast(&nf_nat_bysource_table
, &key
,
228 nf_nat_bysource_params
);
232 nf_ct_invert_tuplepr(result
,
233 &ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
);
234 result
->dst
= tuple
->dst
;
236 return in_range(l3proto
, l4proto
, result
, range
);
239 /* For [FUTURE] fragmentation handling, we want the least-used
240 * src-ip/dst-ip/proto triple. Fairness doesn't come into it. Thus
241 * if the range specifies 1.2.3.4 ports 10000-10005 and 1.2.3.5 ports
242 * 1-65535, we don't do pro-rata allocation based on ports; we choose
243 * the ip with the lowest src-ip/dst-ip/proto usage.
246 find_best_ips_proto(const struct nf_conntrack_zone
*zone
,
247 struct nf_conntrack_tuple
*tuple
,
248 const struct nf_nat_range
*range
,
249 const struct nf_conn
*ct
,
250 enum nf_nat_manip_type maniptype
)
252 union nf_inet_addr
*var_ipp
;
255 u32 minip
, maxip
, j
, dist
;
258 /* No IP mapping? Do nothing. */
259 if (!(range
->flags
& NF_NAT_RANGE_MAP_IPS
))
262 if (maniptype
== NF_NAT_MANIP_SRC
)
263 var_ipp
= &tuple
->src
.u3
;
265 var_ipp
= &tuple
->dst
.u3
;
267 /* Fast path: only one choice. */
268 if (nf_inet_addr_cmp(&range
->min_addr
, &range
->max_addr
)) {
269 *var_ipp
= range
->min_addr
;
273 if (nf_ct_l3num(ct
) == NFPROTO_IPV4
)
274 max
= sizeof(var_ipp
->ip
) / sizeof(u32
) - 1;
276 max
= sizeof(var_ipp
->ip6
) / sizeof(u32
) - 1;
278 /* Hashing source and destination IPs gives a fairly even
279 * spread in practice (if there are a small number of IPs
280 * involved, there usually aren't that many connections
281 * anyway). The consistency means that servers see the same
282 * client coming from the same IP (some Internet Banking sites
283 * like this), even across reboots.
285 j
= jhash2((u32
*)&tuple
->src
.u3
, sizeof(tuple
->src
.u3
) / sizeof(u32
),
286 range
->flags
& NF_NAT_RANGE_PERSISTENT
?
287 0 : (__force u32
)tuple
->dst
.u3
.all
[max
] ^ zone
->id
);
290 for (i
= 0; i
<= max
; i
++) {
291 /* If first bytes of the address are at the maximum, use the
292 * distance. Otherwise use the full range.
295 minip
= ntohl((__force __be32
)range
->min_addr
.all
[i
]);
296 maxip
= ntohl((__force __be32
)range
->max_addr
.all
[i
]);
297 dist
= maxip
- minip
+ 1;
303 var_ipp
->all
[i
] = (__force __u32
)
304 htonl(minip
+ reciprocal_scale(j
, dist
));
305 if (var_ipp
->all
[i
] != range
->max_addr
.all
[i
])
308 if (!(range
->flags
& NF_NAT_RANGE_PERSISTENT
))
309 j
^= (__force u32
)tuple
->dst
.u3
.all
[i
];
313 /* Manipulate the tuple into the range given. For NF_INET_POST_ROUTING,
314 * we change the source to map into the range. For NF_INET_PRE_ROUTING
315 * and NF_INET_LOCAL_OUT, we change the destination to map into the
316 * range. It might not be possible to get a unique tuple, but we try.
317 * At worst (or if we race), we will end up with a final duplicate in
318 * __ip_conntrack_confirm and drop the packet. */
320 get_unique_tuple(struct nf_conntrack_tuple
*tuple
,
321 const struct nf_conntrack_tuple
*orig_tuple
,
322 const struct nf_nat_range
*range
,
324 enum nf_nat_manip_type maniptype
)
326 const struct nf_conntrack_zone
*zone
;
327 const struct nf_nat_l3proto
*l3proto
;
328 const struct nf_nat_l4proto
*l4proto
;
329 struct net
*net
= nf_ct_net(ct
);
331 zone
= nf_ct_zone(ct
);
334 l3proto
= __nf_nat_l3proto_find(orig_tuple
->src
.l3num
);
335 l4proto
= __nf_nat_l4proto_find(orig_tuple
->src
.l3num
,
336 orig_tuple
->dst
.protonum
);
338 /* 1) If this srcip/proto/src-proto-part is currently mapped,
339 * and that same mapping gives a unique tuple within the given
342 * This is only required for source (ie. NAT/masq) mappings.
343 * So far, we don't do local source mappings, so multiple
344 * manips not an issue.
346 if (maniptype
== NF_NAT_MANIP_SRC
&&
347 !(range
->flags
& NF_NAT_RANGE_PROTO_RANDOM_ALL
)) {
348 /* try the original tuple first */
349 if (in_range(l3proto
, l4proto
, orig_tuple
, range
)) {
350 if (!nf_nat_used_tuple(orig_tuple
, ct
)) {
351 *tuple
= *orig_tuple
;
354 } else if (find_appropriate_src(net
, zone
, l3proto
, l4proto
,
355 orig_tuple
, tuple
, range
)) {
356 pr_debug("get_unique_tuple: Found current src map\n");
357 if (!nf_nat_used_tuple(tuple
, ct
))
362 /* 2) Select the least-used IP/proto combination in the given range */
363 *tuple
= *orig_tuple
;
364 find_best_ips_proto(zone
, tuple
, range
, ct
, maniptype
);
366 /* 3) The per-protocol part of the manip is made to map into
367 * the range to make a unique tuple.
370 /* Only bother mapping if it's not already in range and unique */
371 if (!(range
->flags
& NF_NAT_RANGE_PROTO_RANDOM_ALL
)) {
372 if (range
->flags
& NF_NAT_RANGE_PROTO_SPECIFIED
) {
373 if (l4proto
->in_range(tuple
, maniptype
,
375 &range
->max_proto
) &&
376 (range
->min_proto
.all
== range
->max_proto
.all
||
377 !nf_nat_used_tuple(tuple
, ct
)))
379 } else if (!nf_nat_used_tuple(tuple
, ct
)) {
384 /* Last change: get protocol to try to obtain unique tuple. */
385 l4proto
->unique_tuple(l3proto
, tuple
, range
, maniptype
, ct
);
390 struct nf_conn_nat
*nf_ct_nat_ext_add(struct nf_conn
*ct
)
392 struct nf_conn_nat
*nat
= nfct_nat(ct
);
396 if (!nf_ct_is_confirmed(ct
))
397 nat
= nf_ct_ext_add(ct
, NF_CT_EXT_NAT
, GFP_ATOMIC
);
401 EXPORT_SYMBOL_GPL(nf_ct_nat_ext_add
);
404 nf_nat_setup_info(struct nf_conn
*ct
,
405 const struct nf_nat_range
*range
,
406 enum nf_nat_manip_type maniptype
)
408 struct nf_conntrack_tuple curr_tuple
, new_tuple
;
409 struct nf_conn_nat
*nat
;
411 /* nat helper or nfctnetlink also setup binding */
412 nat
= nf_ct_nat_ext_add(ct
);
416 NF_CT_ASSERT(maniptype
== NF_NAT_MANIP_SRC
||
417 maniptype
== NF_NAT_MANIP_DST
);
418 BUG_ON(nf_nat_initialized(ct
, maniptype
));
420 /* What we've got will look like inverse of reply. Normally
421 * this is what is in the conntrack, except for prior
422 * manipulations (future optimization: if num_manips == 0,
423 * orig_tp = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple)
425 nf_ct_invert_tuplepr(&curr_tuple
,
426 &ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
);
428 get_unique_tuple(&new_tuple
, &curr_tuple
, range
, ct
, maniptype
);
430 if (!nf_ct_tuple_equal(&new_tuple
, &curr_tuple
)) {
431 struct nf_conntrack_tuple reply
;
433 /* Alter conntrack table so will recognize replies. */
434 nf_ct_invert_tuplepr(&reply
, &new_tuple
);
435 nf_conntrack_alter_reply(ct
, &reply
);
437 /* Non-atomic: we own this at the moment. */
438 if (maniptype
== NF_NAT_MANIP_SRC
)
439 ct
->status
|= IPS_SRC_NAT
;
441 ct
->status
|= IPS_DST_NAT
;
444 nfct_seqadj_ext_add(ct
);
447 if (maniptype
== NF_NAT_MANIP_SRC
) {
450 err
= rhashtable_insert_fast(&nf_nat_bysource_table
,
452 nf_nat_bysource_params
);
458 if (maniptype
== NF_NAT_MANIP_DST
)
459 ct
->status
|= IPS_DST_NAT_DONE
;
461 ct
->status
|= IPS_SRC_NAT_DONE
;
465 EXPORT_SYMBOL(nf_nat_setup_info
);
468 __nf_nat_alloc_null_binding(struct nf_conn
*ct
, enum nf_nat_manip_type manip
)
470 /* Force range to this IP; let proto decide mapping for
471 * per-proto parts (hence not IP_NAT_RANGE_PROTO_SPECIFIED).
472 * Use reply in case it's already been mangled (eg local packet).
474 union nf_inet_addr ip
=
475 (manip
== NF_NAT_MANIP_SRC
?
476 ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
.dst
.u3
:
477 ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
.src
.u3
);
478 struct nf_nat_range range
= {
479 .flags
= NF_NAT_RANGE_MAP_IPS
,
483 return nf_nat_setup_info(ct
, &range
, manip
);
487 nf_nat_alloc_null_binding(struct nf_conn
*ct
, unsigned int hooknum
)
489 return __nf_nat_alloc_null_binding(ct
, HOOK2MANIP(hooknum
));
491 EXPORT_SYMBOL_GPL(nf_nat_alloc_null_binding
);
493 /* Do packet manipulations according to nf_nat_setup_info. */
494 unsigned int nf_nat_packet(struct nf_conn
*ct
,
495 enum ip_conntrack_info ctinfo
,
496 unsigned int hooknum
,
499 const struct nf_nat_l3proto
*l3proto
;
500 const struct nf_nat_l4proto
*l4proto
;
501 enum ip_conntrack_dir dir
= CTINFO2DIR(ctinfo
);
502 unsigned long statusbit
;
503 enum nf_nat_manip_type mtype
= HOOK2MANIP(hooknum
);
505 if (mtype
== NF_NAT_MANIP_SRC
)
506 statusbit
= IPS_SRC_NAT
;
508 statusbit
= IPS_DST_NAT
;
510 /* Invert if this is reply dir. */
511 if (dir
== IP_CT_DIR_REPLY
)
512 statusbit
^= IPS_NAT_MASK
;
514 /* Non-atomic: these bits don't change. */
515 if (ct
->status
& statusbit
) {
516 struct nf_conntrack_tuple target
;
518 /* We are aiming to look like inverse of other direction. */
519 nf_ct_invert_tuplepr(&target
, &ct
->tuplehash
[!dir
].tuple
);
521 l3proto
= __nf_nat_l3proto_find(target
.src
.l3num
);
522 l4proto
= __nf_nat_l4proto_find(target
.src
.l3num
,
523 target
.dst
.protonum
);
524 if (!l3proto
->manip_pkt(skb
, 0, l4proto
, &target
, mtype
))
529 EXPORT_SYMBOL_GPL(nf_nat_packet
);
531 struct nf_nat_proto_clean
{
536 /* kill conntracks with affected NAT section */
537 static int nf_nat_proto_remove(struct nf_conn
*i
, void *data
)
539 const struct nf_nat_proto_clean
*clean
= data
;
540 struct nf_conn_nat
*nat
= nfct_nat(i
);
545 if ((clean
->l3proto
&& nf_ct_l3num(i
) != clean
->l3proto
) ||
546 (clean
->l4proto
&& nf_ct_protonum(i
) != clean
->l4proto
))
549 return i
->status
& IPS_NAT_MASK
? 1 : 0;
552 static int nf_nat_proto_clean(struct nf_conn
*ct
, void *data
)
554 struct nf_conn_nat
*nat
= nfct_nat(ct
);
556 if (nf_nat_proto_remove(ct
, data
))
562 /* This netns is being destroyed, and conntrack has nat null binding.
563 * Remove it from bysource hash, as the table will be freed soon.
565 * Else, when the conntrack is destoyed, nf_nat_cleanup_conntrack()
566 * will delete entry from already-freed table.
568 if (!del_timer(&ct
->timeout
))
571 ct
->status
&= ~IPS_NAT_DONE_MASK
;
573 rhashtable_remove_fast(&nf_nat_bysource_table
, &ct
->nat_bysource
,
574 nf_nat_bysource_params
);
576 add_timer(&ct
->timeout
);
578 /* don't delete conntrack. Although that would make things a lot
579 * simpler, we'd end up flushing all conntracks on nat rmmod.
584 static void nf_nat_l4proto_clean(u8 l3proto
, u8 l4proto
)
586 struct nf_nat_proto_clean clean
= {
594 nf_ct_iterate_cleanup(net
, nf_nat_proto_remove
, &clean
, 0, 0);
598 static void nf_nat_l3proto_clean(u8 l3proto
)
600 struct nf_nat_proto_clean clean
= {
608 nf_ct_iterate_cleanup(net
, nf_nat_proto_remove
, &clean
, 0, 0);
612 /* Protocol registration. */
613 int nf_nat_l4proto_register(u8 l3proto
, const struct nf_nat_l4proto
*l4proto
)
615 const struct nf_nat_l4proto
**l4protos
;
619 mutex_lock(&nf_nat_proto_mutex
);
620 if (nf_nat_l4protos
[l3proto
] == NULL
) {
621 l4protos
= kmalloc(IPPROTO_MAX
* sizeof(struct nf_nat_l4proto
*),
623 if (l4protos
== NULL
) {
628 for (i
= 0; i
< IPPROTO_MAX
; i
++)
629 RCU_INIT_POINTER(l4protos
[i
], &nf_nat_l4proto_unknown
);
631 /* Before making proto_array visible to lockless readers,
632 * we must make sure its content is committed to memory.
636 nf_nat_l4protos
[l3proto
] = l4protos
;
639 if (rcu_dereference_protected(
640 nf_nat_l4protos
[l3proto
][l4proto
->l4proto
],
641 lockdep_is_held(&nf_nat_proto_mutex
)
642 ) != &nf_nat_l4proto_unknown
) {
646 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
][l4proto
->l4proto
], l4proto
);
648 mutex_unlock(&nf_nat_proto_mutex
);
651 EXPORT_SYMBOL_GPL(nf_nat_l4proto_register
);
653 /* No one stores the protocol anywhere; simply delete it. */
654 void nf_nat_l4proto_unregister(u8 l3proto
, const struct nf_nat_l4proto
*l4proto
)
656 mutex_lock(&nf_nat_proto_mutex
);
657 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
][l4proto
->l4proto
],
658 &nf_nat_l4proto_unknown
);
659 mutex_unlock(&nf_nat_proto_mutex
);
662 nf_nat_l4proto_clean(l3proto
, l4proto
->l4proto
);
664 EXPORT_SYMBOL_GPL(nf_nat_l4proto_unregister
);
666 int nf_nat_l3proto_register(const struct nf_nat_l3proto
*l3proto
)
670 err
= nf_ct_l3proto_try_module_get(l3proto
->l3proto
);
674 mutex_lock(&nf_nat_proto_mutex
);
675 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
->l3proto
][IPPROTO_TCP
],
676 &nf_nat_l4proto_tcp
);
677 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
->l3proto
][IPPROTO_UDP
],
678 &nf_nat_l4proto_udp
);
679 mutex_unlock(&nf_nat_proto_mutex
);
681 RCU_INIT_POINTER(nf_nat_l3protos
[l3proto
->l3proto
], l3proto
);
684 EXPORT_SYMBOL_GPL(nf_nat_l3proto_register
);
686 void nf_nat_l3proto_unregister(const struct nf_nat_l3proto
*l3proto
)
688 mutex_lock(&nf_nat_proto_mutex
);
689 RCU_INIT_POINTER(nf_nat_l3protos
[l3proto
->l3proto
], NULL
);
690 mutex_unlock(&nf_nat_proto_mutex
);
693 nf_nat_l3proto_clean(l3proto
->l3proto
);
694 nf_ct_l3proto_module_put(l3proto
->l3proto
);
696 EXPORT_SYMBOL_GPL(nf_nat_l3proto_unregister
);
698 /* No one using conntrack by the time this called. */
699 static void nf_nat_cleanup_conntrack(struct nf_conn
*ct
)
701 struct nf_conn_nat
*nat
= nf_ct_ext_find(ct
, NF_CT_EXT_NAT
);
706 rhashtable_remove_fast(&nf_nat_bysource_table
, &ct
->nat_bysource
,
707 nf_nat_bysource_params
);
710 static struct nf_ct_ext_type nat_extend __read_mostly
= {
711 .len
= sizeof(struct nf_conn_nat
),
712 .align
= __alignof__(struct nf_conn_nat
),
713 .destroy
= nf_nat_cleanup_conntrack
,
715 .flags
= NF_CT_EXT_F_PREALLOC
,
718 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
720 #include <linux/netfilter/nfnetlink.h>
721 #include <linux/netfilter/nfnetlink_conntrack.h>
723 static const struct nla_policy protonat_nla_policy
[CTA_PROTONAT_MAX
+1] = {
724 [CTA_PROTONAT_PORT_MIN
] = { .type
= NLA_U16
},
725 [CTA_PROTONAT_PORT_MAX
] = { .type
= NLA_U16
},
728 static int nfnetlink_parse_nat_proto(struct nlattr
*attr
,
729 const struct nf_conn
*ct
,
730 struct nf_nat_range
*range
)
732 struct nlattr
*tb
[CTA_PROTONAT_MAX
+1];
733 const struct nf_nat_l4proto
*l4proto
;
736 err
= nla_parse_nested(tb
, CTA_PROTONAT_MAX
, attr
, protonat_nla_policy
);
740 l4proto
= __nf_nat_l4proto_find(nf_ct_l3num(ct
), nf_ct_protonum(ct
));
741 if (l4proto
->nlattr_to_range
)
742 err
= l4proto
->nlattr_to_range(tb
, range
);
747 static const struct nla_policy nat_nla_policy
[CTA_NAT_MAX
+1] = {
748 [CTA_NAT_V4_MINIP
] = { .type
= NLA_U32
},
749 [CTA_NAT_V4_MAXIP
] = { .type
= NLA_U32
},
750 [CTA_NAT_V6_MINIP
] = { .len
= sizeof(struct in6_addr
) },
751 [CTA_NAT_V6_MAXIP
] = { .len
= sizeof(struct in6_addr
) },
752 [CTA_NAT_PROTO
] = { .type
= NLA_NESTED
},
756 nfnetlink_parse_nat(const struct nlattr
*nat
,
757 const struct nf_conn
*ct
, struct nf_nat_range
*range
,
758 const struct nf_nat_l3proto
*l3proto
)
760 struct nlattr
*tb
[CTA_NAT_MAX
+1];
763 memset(range
, 0, sizeof(*range
));
765 err
= nla_parse_nested(tb
, CTA_NAT_MAX
, nat
, nat_nla_policy
);
769 err
= l3proto
->nlattr_to_range(tb
, range
);
773 if (!tb
[CTA_NAT_PROTO
])
776 return nfnetlink_parse_nat_proto(tb
[CTA_NAT_PROTO
], ct
, range
);
779 /* This function is called under rcu_read_lock() */
781 nfnetlink_parse_nat_setup(struct nf_conn
*ct
,
782 enum nf_nat_manip_type manip
,
783 const struct nlattr
*attr
)
785 struct nf_nat_range range
;
786 const struct nf_nat_l3proto
*l3proto
;
789 /* Should not happen, restricted to creating new conntracks
792 if (WARN_ON_ONCE(nf_nat_initialized(ct
, manip
)))
795 /* Make sure that L3 NAT is there by when we call nf_nat_setup_info to
796 * attach the null binding, otherwise this may oops.
798 l3proto
= __nf_nat_l3proto_find(nf_ct_l3num(ct
));
802 /* No NAT information has been passed, allocate the null-binding */
804 return __nf_nat_alloc_null_binding(ct
, manip
);
806 err
= nfnetlink_parse_nat(attr
, ct
, &range
, l3proto
);
810 return nf_nat_setup_info(ct
, &range
, manip
);
814 nfnetlink_parse_nat_setup(struct nf_conn
*ct
,
815 enum nf_nat_manip_type manip
,
816 const struct nlattr
*attr
)
822 static void __net_exit
nf_nat_net_exit(struct net
*net
)
824 struct nf_nat_proto_clean clean
= {};
826 nf_ct_iterate_cleanup(net
, nf_nat_proto_clean
, &clean
, 0, 0);
829 static struct pernet_operations nf_nat_net_ops
= {
830 .exit
= nf_nat_net_exit
,
833 static struct nf_ct_helper_expectfn follow_master_nat
= {
834 .name
= "nat-follow-master",
835 .expectfn
= nf_nat_follow_master
,
838 static int __init
nf_nat_init(void)
842 ret
= rhashtable_init(&nf_nat_bysource_table
, &nf_nat_bysource_params
);
846 ret
= nf_ct_extend_register(&nat_extend
);
848 rhashtable_destroy(&nf_nat_bysource_table
);
849 printk(KERN_ERR
"nf_nat_core: Unable to register extension\n");
853 ret
= register_pernet_subsys(&nf_nat_net_ops
);
857 nf_ct_helper_expectfn_register(&follow_master_nat
);
859 /* Initialize fake conntrack so that NAT will skip it */
860 nf_ct_untracked_status_or(IPS_NAT_DONE_MASK
);
862 BUG_ON(nfnetlink_parse_nat_setup_hook
!= NULL
);
863 RCU_INIT_POINTER(nfnetlink_parse_nat_setup_hook
,
864 nfnetlink_parse_nat_setup
);
866 BUG_ON(nf_nat_decode_session_hook
!= NULL
);
867 RCU_INIT_POINTER(nf_nat_decode_session_hook
, __nf_nat_decode_session
);
872 rhashtable_destroy(&nf_nat_bysource_table
);
873 nf_ct_extend_unregister(&nat_extend
);
877 static void __exit
nf_nat_cleanup(void)
881 unregister_pernet_subsys(&nf_nat_net_ops
);
882 nf_ct_extend_unregister(&nat_extend
);
883 nf_ct_helper_expectfn_unregister(&follow_master_nat
);
884 RCU_INIT_POINTER(nfnetlink_parse_nat_setup_hook
, NULL
);
886 RCU_INIT_POINTER(nf_nat_decode_session_hook
, NULL
);
888 for (i
= 0; i
< NFPROTO_NUMPROTO
; i
++)
889 kfree(nf_nat_l4protos
[i
]);
891 rhashtable_destroy(&nf_nat_bysource_table
);
894 MODULE_LICENSE("GPL");
896 module_init(nf_nat_init
);
897 module_exit(nf_nat_cleanup
);