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 rhltable 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 if (!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
))
204 static struct rhashtable_params nf_nat_bysource_params
= {
205 .head_offset
= offsetof(struct nf_conn
, nat_bysource
),
206 .obj_hashfn
= nf_nat_bysource_hash
,
207 .obj_cmpfn
= nf_nat_bysource_cmp
,
212 /* Only called for SRC manip */
214 find_appropriate_src(struct net
*net
,
215 const struct nf_conntrack_zone
*zone
,
216 const struct nf_nat_l3proto
*l3proto
,
217 const struct nf_nat_l4proto
*l4proto
,
218 const struct nf_conntrack_tuple
*tuple
,
219 struct nf_conntrack_tuple
*result
,
220 const struct nf_nat_range
*range
)
222 const struct nf_conn
*ct
;
223 struct nf_nat_conn_key key
= {
228 struct rhlist_head
*hl
;
230 hl
= rhltable_lookup(&nf_nat_bysource_table
, &key
,
231 nf_nat_bysource_params
);
235 ct
= container_of(hl
, typeof(*ct
), nat_bysource
);
237 nf_ct_invert_tuplepr(result
,
238 &ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
);
239 result
->dst
= tuple
->dst
;
241 return in_range(l3proto
, l4proto
, result
, range
);
244 /* For [FUTURE] fragmentation handling, we want the least-used
245 * src-ip/dst-ip/proto triple. Fairness doesn't come into it. Thus
246 * if the range specifies 1.2.3.4 ports 10000-10005 and 1.2.3.5 ports
247 * 1-65535, we don't do pro-rata allocation based on ports; we choose
248 * the ip with the lowest src-ip/dst-ip/proto usage.
251 find_best_ips_proto(const struct nf_conntrack_zone
*zone
,
252 struct nf_conntrack_tuple
*tuple
,
253 const struct nf_nat_range
*range
,
254 const struct nf_conn
*ct
,
255 enum nf_nat_manip_type maniptype
)
257 union nf_inet_addr
*var_ipp
;
260 u32 minip
, maxip
, j
, dist
;
263 /* No IP mapping? Do nothing. */
264 if (!(range
->flags
& NF_NAT_RANGE_MAP_IPS
))
267 if (maniptype
== NF_NAT_MANIP_SRC
)
268 var_ipp
= &tuple
->src
.u3
;
270 var_ipp
= &tuple
->dst
.u3
;
272 /* Fast path: only one choice. */
273 if (nf_inet_addr_cmp(&range
->min_addr
, &range
->max_addr
)) {
274 *var_ipp
= range
->min_addr
;
278 if (nf_ct_l3num(ct
) == NFPROTO_IPV4
)
279 max
= sizeof(var_ipp
->ip
) / sizeof(u32
) - 1;
281 max
= sizeof(var_ipp
->ip6
) / sizeof(u32
) - 1;
283 /* Hashing source and destination IPs gives a fairly even
284 * spread in practice (if there are a small number of IPs
285 * involved, there usually aren't that many connections
286 * anyway). The consistency means that servers see the same
287 * client coming from the same IP (some Internet Banking sites
288 * like this), even across reboots.
290 j
= jhash2((u32
*)&tuple
->src
.u3
, sizeof(tuple
->src
.u3
) / sizeof(u32
),
291 range
->flags
& NF_NAT_RANGE_PERSISTENT
?
292 0 : (__force u32
)tuple
->dst
.u3
.all
[max
] ^ zone
->id
);
295 for (i
= 0; i
<= max
; i
++) {
296 /* If first bytes of the address are at the maximum, use the
297 * distance. Otherwise use the full range.
300 minip
= ntohl((__force __be32
)range
->min_addr
.all
[i
]);
301 maxip
= ntohl((__force __be32
)range
->max_addr
.all
[i
]);
302 dist
= maxip
- minip
+ 1;
308 var_ipp
->all
[i
] = (__force __u32
)
309 htonl(minip
+ reciprocal_scale(j
, dist
));
310 if (var_ipp
->all
[i
] != range
->max_addr
.all
[i
])
313 if (!(range
->flags
& NF_NAT_RANGE_PERSISTENT
))
314 j
^= (__force u32
)tuple
->dst
.u3
.all
[i
];
318 /* Manipulate the tuple into the range given. For NF_INET_POST_ROUTING,
319 * we change the source to map into the range. For NF_INET_PRE_ROUTING
320 * and NF_INET_LOCAL_OUT, we change the destination to map into the
321 * range. It might not be possible to get a unique tuple, but we try.
322 * At worst (or if we race), we will end up with a final duplicate in
323 * __ip_conntrack_confirm and drop the packet. */
325 get_unique_tuple(struct nf_conntrack_tuple
*tuple
,
326 const struct nf_conntrack_tuple
*orig_tuple
,
327 const struct nf_nat_range
*range
,
329 enum nf_nat_manip_type maniptype
)
331 const struct nf_conntrack_zone
*zone
;
332 const struct nf_nat_l3proto
*l3proto
;
333 const struct nf_nat_l4proto
*l4proto
;
334 struct net
*net
= nf_ct_net(ct
);
336 zone
= nf_ct_zone(ct
);
339 l3proto
= __nf_nat_l3proto_find(orig_tuple
->src
.l3num
);
340 l4proto
= __nf_nat_l4proto_find(orig_tuple
->src
.l3num
,
341 orig_tuple
->dst
.protonum
);
343 /* 1) If this srcip/proto/src-proto-part is currently mapped,
344 * and that same mapping gives a unique tuple within the given
347 * This is only required for source (ie. NAT/masq) mappings.
348 * So far, we don't do local source mappings, so multiple
349 * manips not an issue.
351 if (maniptype
== NF_NAT_MANIP_SRC
&&
352 !(range
->flags
& NF_NAT_RANGE_PROTO_RANDOM_ALL
)) {
353 /* try the original tuple first */
354 if (in_range(l3proto
, l4proto
, orig_tuple
, range
)) {
355 if (!nf_nat_used_tuple(orig_tuple
, ct
)) {
356 *tuple
= *orig_tuple
;
359 } else if (find_appropriate_src(net
, zone
, l3proto
, l4proto
,
360 orig_tuple
, tuple
, range
)) {
361 pr_debug("get_unique_tuple: Found current src map\n");
362 if (!nf_nat_used_tuple(tuple
, ct
))
367 /* 2) Select the least-used IP/proto combination in the given range */
368 *tuple
= *orig_tuple
;
369 find_best_ips_proto(zone
, tuple
, range
, ct
, maniptype
);
371 /* 3) The per-protocol part of the manip is made to map into
372 * the range to make a unique tuple.
375 /* Only bother mapping if it's not already in range and unique */
376 if (!(range
->flags
& NF_NAT_RANGE_PROTO_RANDOM_ALL
)) {
377 if (range
->flags
& NF_NAT_RANGE_PROTO_SPECIFIED
) {
378 if (l4proto
->in_range(tuple
, maniptype
,
380 &range
->max_proto
) &&
381 (range
->min_proto
.all
== range
->max_proto
.all
||
382 !nf_nat_used_tuple(tuple
, ct
)))
384 } else if (!nf_nat_used_tuple(tuple
, ct
)) {
389 /* Last change: get protocol to try to obtain unique tuple. */
390 l4proto
->unique_tuple(l3proto
, tuple
, range
, maniptype
, ct
);
395 struct nf_conn_nat
*nf_ct_nat_ext_add(struct nf_conn
*ct
)
397 struct nf_conn_nat
*nat
= nfct_nat(ct
);
401 if (!nf_ct_is_confirmed(ct
))
402 nat
= nf_ct_ext_add(ct
, NF_CT_EXT_NAT
, GFP_ATOMIC
);
406 EXPORT_SYMBOL_GPL(nf_ct_nat_ext_add
);
409 nf_nat_setup_info(struct nf_conn
*ct
,
410 const struct nf_nat_range
*range
,
411 enum nf_nat_manip_type maniptype
)
413 struct nf_conntrack_tuple curr_tuple
, new_tuple
;
414 struct nf_conn_nat
*nat
;
416 /* nat helper or nfctnetlink also setup binding */
417 nat
= nf_ct_nat_ext_add(ct
);
421 NF_CT_ASSERT(maniptype
== NF_NAT_MANIP_SRC
||
422 maniptype
== NF_NAT_MANIP_DST
);
423 BUG_ON(nf_nat_initialized(ct
, maniptype
));
425 /* What we've got will look like inverse of reply. Normally
426 * this is what is in the conntrack, except for prior
427 * manipulations (future optimization: if num_manips == 0,
428 * orig_tp = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple)
430 nf_ct_invert_tuplepr(&curr_tuple
,
431 &ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
);
433 get_unique_tuple(&new_tuple
, &curr_tuple
, range
, ct
, maniptype
);
435 if (!nf_ct_tuple_equal(&new_tuple
, &curr_tuple
)) {
436 struct nf_conntrack_tuple reply
;
438 /* Alter conntrack table so will recognize replies. */
439 nf_ct_invert_tuplepr(&reply
, &new_tuple
);
440 nf_conntrack_alter_reply(ct
, &reply
);
442 /* Non-atomic: we own this at the moment. */
443 if (maniptype
== NF_NAT_MANIP_SRC
)
444 ct
->status
|= IPS_SRC_NAT
;
446 ct
->status
|= IPS_DST_NAT
;
449 if (!nfct_seqadj_ext_add(ct
))
453 if (maniptype
== NF_NAT_MANIP_SRC
) {
454 struct nf_nat_conn_key key
= {
455 .net
= nf_ct_net(ct
),
456 .tuple
= &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
,
457 .zone
= nf_ct_zone(ct
),
461 err
= rhltable_insert_key(&nf_nat_bysource_table
,
464 nf_nat_bysource_params
);
470 if (maniptype
== NF_NAT_MANIP_DST
)
471 ct
->status
|= IPS_DST_NAT_DONE
;
473 ct
->status
|= IPS_SRC_NAT_DONE
;
477 EXPORT_SYMBOL(nf_nat_setup_info
);
480 __nf_nat_alloc_null_binding(struct nf_conn
*ct
, enum nf_nat_manip_type manip
)
482 /* Force range to this IP; let proto decide mapping for
483 * per-proto parts (hence not IP_NAT_RANGE_PROTO_SPECIFIED).
484 * Use reply in case it's already been mangled (eg local packet).
486 union nf_inet_addr ip
=
487 (manip
== NF_NAT_MANIP_SRC
?
488 ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
.dst
.u3
:
489 ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
.src
.u3
);
490 struct nf_nat_range range
= {
491 .flags
= NF_NAT_RANGE_MAP_IPS
,
495 return nf_nat_setup_info(ct
, &range
, manip
);
499 nf_nat_alloc_null_binding(struct nf_conn
*ct
, unsigned int hooknum
)
501 return __nf_nat_alloc_null_binding(ct
, HOOK2MANIP(hooknum
));
503 EXPORT_SYMBOL_GPL(nf_nat_alloc_null_binding
);
505 /* Do packet manipulations according to nf_nat_setup_info. */
506 unsigned int nf_nat_packet(struct nf_conn
*ct
,
507 enum ip_conntrack_info ctinfo
,
508 unsigned int hooknum
,
511 const struct nf_nat_l3proto
*l3proto
;
512 const struct nf_nat_l4proto
*l4proto
;
513 enum ip_conntrack_dir dir
= CTINFO2DIR(ctinfo
);
514 unsigned long statusbit
;
515 enum nf_nat_manip_type mtype
= HOOK2MANIP(hooknum
);
517 if (mtype
== NF_NAT_MANIP_SRC
)
518 statusbit
= IPS_SRC_NAT
;
520 statusbit
= IPS_DST_NAT
;
522 /* Invert if this is reply dir. */
523 if (dir
== IP_CT_DIR_REPLY
)
524 statusbit
^= IPS_NAT_MASK
;
526 /* Non-atomic: these bits don't change. */
527 if (ct
->status
& statusbit
) {
528 struct nf_conntrack_tuple target
;
530 /* We are aiming to look like inverse of other direction. */
531 nf_ct_invert_tuplepr(&target
, &ct
->tuplehash
[!dir
].tuple
);
533 l3proto
= __nf_nat_l3proto_find(target
.src
.l3num
);
534 l4proto
= __nf_nat_l4proto_find(target
.src
.l3num
,
535 target
.dst
.protonum
);
536 if (!l3proto
->manip_pkt(skb
, 0, l4proto
, &target
, mtype
))
541 EXPORT_SYMBOL_GPL(nf_nat_packet
);
543 struct nf_nat_proto_clean
{
548 /* kill conntracks with affected NAT section */
549 static int nf_nat_proto_remove(struct nf_conn
*i
, void *data
)
551 const struct nf_nat_proto_clean
*clean
= data
;
552 struct nf_conn_nat
*nat
= nfct_nat(i
);
557 if ((clean
->l3proto
&& nf_ct_l3num(i
) != clean
->l3proto
) ||
558 (clean
->l4proto
&& nf_ct_protonum(i
) != clean
->l4proto
))
561 return i
->status
& IPS_NAT_MASK
? 1 : 0;
564 static int nf_nat_proto_clean(struct nf_conn
*ct
, void *data
)
566 struct nf_conn_nat
*nat
= nfct_nat(ct
);
568 if (nf_nat_proto_remove(ct
, data
))
574 /* This netns is being destroyed, and conntrack has nat null binding.
575 * Remove it from bysource hash, as the table will be freed soon.
577 * Else, when the conntrack is destoyed, nf_nat_cleanup_conntrack()
578 * will delete entry from already-freed table.
580 ct
->status
&= ~IPS_NAT_DONE_MASK
;
581 rhltable_remove(&nf_nat_bysource_table
, &ct
->nat_bysource
,
582 nf_nat_bysource_params
);
584 /* don't delete conntrack. Although that would make things a lot
585 * simpler, we'd end up flushing all conntracks on nat rmmod.
590 static void nf_nat_l4proto_clean(u8 l3proto
, u8 l4proto
)
592 struct nf_nat_proto_clean clean
= {
600 nf_ct_iterate_cleanup(net
, nf_nat_proto_remove
, &clean
, 0, 0);
604 static void nf_nat_l3proto_clean(u8 l3proto
)
606 struct nf_nat_proto_clean clean
= {
614 nf_ct_iterate_cleanup(net
, nf_nat_proto_remove
, &clean
, 0, 0);
618 /* Protocol registration. */
619 int nf_nat_l4proto_register(u8 l3proto
, const struct nf_nat_l4proto
*l4proto
)
621 const struct nf_nat_l4proto
**l4protos
;
625 mutex_lock(&nf_nat_proto_mutex
);
626 if (nf_nat_l4protos
[l3proto
] == NULL
) {
627 l4protos
= kmalloc(IPPROTO_MAX
* sizeof(struct nf_nat_l4proto
*),
629 if (l4protos
== NULL
) {
634 for (i
= 0; i
< IPPROTO_MAX
; i
++)
635 RCU_INIT_POINTER(l4protos
[i
], &nf_nat_l4proto_unknown
);
637 /* Before making proto_array visible to lockless readers,
638 * we must make sure its content is committed to memory.
642 nf_nat_l4protos
[l3proto
] = l4protos
;
645 if (rcu_dereference_protected(
646 nf_nat_l4protos
[l3proto
][l4proto
->l4proto
],
647 lockdep_is_held(&nf_nat_proto_mutex
)
648 ) != &nf_nat_l4proto_unknown
) {
652 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
][l4proto
->l4proto
], l4proto
);
654 mutex_unlock(&nf_nat_proto_mutex
);
657 EXPORT_SYMBOL_GPL(nf_nat_l4proto_register
);
659 /* No one stores the protocol anywhere; simply delete it. */
660 void nf_nat_l4proto_unregister(u8 l3proto
, const struct nf_nat_l4proto
*l4proto
)
662 mutex_lock(&nf_nat_proto_mutex
);
663 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
][l4proto
->l4proto
],
664 &nf_nat_l4proto_unknown
);
665 mutex_unlock(&nf_nat_proto_mutex
);
668 nf_nat_l4proto_clean(l3proto
, l4proto
->l4proto
);
670 EXPORT_SYMBOL_GPL(nf_nat_l4proto_unregister
);
672 int nf_nat_l3proto_register(const struct nf_nat_l3proto
*l3proto
)
676 err
= nf_ct_l3proto_try_module_get(l3proto
->l3proto
);
680 mutex_lock(&nf_nat_proto_mutex
);
681 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
->l3proto
][IPPROTO_TCP
],
682 &nf_nat_l4proto_tcp
);
683 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
->l3proto
][IPPROTO_UDP
],
684 &nf_nat_l4proto_udp
);
685 #ifdef CONFIG_NF_NAT_PROTO_DCCP
686 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
->l3proto
][IPPROTO_DCCP
],
687 &nf_nat_l4proto_dccp
);
689 #ifdef CONFIG_NF_NAT_PROTO_SCTP
690 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
->l3proto
][IPPROTO_SCTP
],
691 &nf_nat_l4proto_sctp
);
693 #ifdef CONFIG_NF_NAT_PROTO_UDPLITE
694 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
->l3proto
][IPPROTO_UDPLITE
],
695 &nf_nat_l4proto_udplite
);
697 mutex_unlock(&nf_nat_proto_mutex
);
699 RCU_INIT_POINTER(nf_nat_l3protos
[l3proto
->l3proto
], l3proto
);
702 EXPORT_SYMBOL_GPL(nf_nat_l3proto_register
);
704 void nf_nat_l3proto_unregister(const struct nf_nat_l3proto
*l3proto
)
706 mutex_lock(&nf_nat_proto_mutex
);
707 RCU_INIT_POINTER(nf_nat_l3protos
[l3proto
->l3proto
], NULL
);
708 mutex_unlock(&nf_nat_proto_mutex
);
711 nf_nat_l3proto_clean(l3proto
->l3proto
);
712 nf_ct_l3proto_module_put(l3proto
->l3proto
);
714 EXPORT_SYMBOL_GPL(nf_nat_l3proto_unregister
);
716 /* No one using conntrack by the time this called. */
717 static void nf_nat_cleanup_conntrack(struct nf_conn
*ct
)
719 struct nf_conn_nat
*nat
= nf_ct_ext_find(ct
, NF_CT_EXT_NAT
);
724 rhltable_remove(&nf_nat_bysource_table
, &ct
->nat_bysource
,
725 nf_nat_bysource_params
);
728 static struct nf_ct_ext_type nat_extend __read_mostly
= {
729 .len
= sizeof(struct nf_conn_nat
),
730 .align
= __alignof__(struct nf_conn_nat
),
731 .destroy
= nf_nat_cleanup_conntrack
,
733 .flags
= NF_CT_EXT_F_PREALLOC
,
736 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
738 #include <linux/netfilter/nfnetlink.h>
739 #include <linux/netfilter/nfnetlink_conntrack.h>
741 static const struct nla_policy protonat_nla_policy
[CTA_PROTONAT_MAX
+1] = {
742 [CTA_PROTONAT_PORT_MIN
] = { .type
= NLA_U16
},
743 [CTA_PROTONAT_PORT_MAX
] = { .type
= NLA_U16
},
746 static int nfnetlink_parse_nat_proto(struct nlattr
*attr
,
747 const struct nf_conn
*ct
,
748 struct nf_nat_range
*range
)
750 struct nlattr
*tb
[CTA_PROTONAT_MAX
+1];
751 const struct nf_nat_l4proto
*l4proto
;
754 err
= nla_parse_nested(tb
, CTA_PROTONAT_MAX
, attr
, protonat_nla_policy
);
758 l4proto
= __nf_nat_l4proto_find(nf_ct_l3num(ct
), nf_ct_protonum(ct
));
759 if (l4proto
->nlattr_to_range
)
760 err
= l4proto
->nlattr_to_range(tb
, range
);
765 static const struct nla_policy nat_nla_policy
[CTA_NAT_MAX
+1] = {
766 [CTA_NAT_V4_MINIP
] = { .type
= NLA_U32
},
767 [CTA_NAT_V4_MAXIP
] = { .type
= NLA_U32
},
768 [CTA_NAT_V6_MINIP
] = { .len
= sizeof(struct in6_addr
) },
769 [CTA_NAT_V6_MAXIP
] = { .len
= sizeof(struct in6_addr
) },
770 [CTA_NAT_PROTO
] = { .type
= NLA_NESTED
},
774 nfnetlink_parse_nat(const struct nlattr
*nat
,
775 const struct nf_conn
*ct
, struct nf_nat_range
*range
,
776 const struct nf_nat_l3proto
*l3proto
)
778 struct nlattr
*tb
[CTA_NAT_MAX
+1];
781 memset(range
, 0, sizeof(*range
));
783 err
= nla_parse_nested(tb
, CTA_NAT_MAX
, nat
, nat_nla_policy
);
787 err
= l3proto
->nlattr_to_range(tb
, range
);
791 if (!tb
[CTA_NAT_PROTO
])
794 return nfnetlink_parse_nat_proto(tb
[CTA_NAT_PROTO
], ct
, range
);
797 /* This function is called under rcu_read_lock() */
799 nfnetlink_parse_nat_setup(struct nf_conn
*ct
,
800 enum nf_nat_manip_type manip
,
801 const struct nlattr
*attr
)
803 struct nf_nat_range range
;
804 const struct nf_nat_l3proto
*l3proto
;
807 /* Should not happen, restricted to creating new conntracks
810 if (WARN_ON_ONCE(nf_nat_initialized(ct
, manip
)))
813 /* Make sure that L3 NAT is there by when we call nf_nat_setup_info to
814 * attach the null binding, otherwise this may oops.
816 l3proto
= __nf_nat_l3proto_find(nf_ct_l3num(ct
));
820 /* No NAT information has been passed, allocate the null-binding */
822 return __nf_nat_alloc_null_binding(ct
, manip
);
824 err
= nfnetlink_parse_nat(attr
, ct
, &range
, l3proto
);
828 return nf_nat_setup_info(ct
, &range
, manip
) == NF_DROP
? -ENOMEM
: 0;
832 nfnetlink_parse_nat_setup(struct nf_conn
*ct
,
833 enum nf_nat_manip_type manip
,
834 const struct nlattr
*attr
)
840 static void __net_exit
nf_nat_net_exit(struct net
*net
)
842 struct nf_nat_proto_clean clean
= {};
844 nf_ct_iterate_cleanup(net
, nf_nat_proto_clean
, &clean
, 0, 0);
847 static struct pernet_operations nf_nat_net_ops
= {
848 .exit
= nf_nat_net_exit
,
851 static struct nf_ct_helper_expectfn follow_master_nat
= {
852 .name
= "nat-follow-master",
853 .expectfn
= nf_nat_follow_master
,
856 static int __init
nf_nat_init(void)
860 ret
= rhltable_init(&nf_nat_bysource_table
, &nf_nat_bysource_params
);
864 ret
= nf_ct_extend_register(&nat_extend
);
866 rhltable_destroy(&nf_nat_bysource_table
);
867 printk(KERN_ERR
"nf_nat_core: Unable to register extension\n");
871 ret
= register_pernet_subsys(&nf_nat_net_ops
);
875 nf_ct_helper_expectfn_register(&follow_master_nat
);
877 /* Initialize fake conntrack so that NAT will skip it */
878 nf_ct_untracked_status_or(IPS_NAT_DONE_MASK
);
880 BUG_ON(nfnetlink_parse_nat_setup_hook
!= NULL
);
881 RCU_INIT_POINTER(nfnetlink_parse_nat_setup_hook
,
882 nfnetlink_parse_nat_setup
);
884 BUG_ON(nf_nat_decode_session_hook
!= NULL
);
885 RCU_INIT_POINTER(nf_nat_decode_session_hook
, __nf_nat_decode_session
);
890 rhltable_destroy(&nf_nat_bysource_table
);
891 nf_ct_extend_unregister(&nat_extend
);
895 static void __exit
nf_nat_cleanup(void)
899 unregister_pernet_subsys(&nf_nat_net_ops
);
900 nf_ct_extend_unregister(&nat_extend
);
901 nf_ct_helper_expectfn_unregister(&follow_master_nat
);
902 RCU_INIT_POINTER(nfnetlink_parse_nat_setup_hook
, NULL
);
904 RCU_INIT_POINTER(nf_nat_decode_session_hook
, NULL
);
906 for (i
= 0; i
< NFPROTO_NUMPROTO
; i
++)
907 kfree(nf_nat_l4protos
[i
]);
909 rhltable_destroy(&nf_nat_bysource_table
);
912 MODULE_LICENSE("GPL");
914 module_init(nf_nat_init
);
915 module_exit(nf_nat_cleanup
);