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
= nf_ct_l3num(ct
);
75 l3proto
= __nf_nat_l3proto_find(family
);
79 dir
= CTINFO2DIR(ctinfo
);
80 if (dir
== IP_CT_DIR_ORIGINAL
)
81 statusbit
= IPS_DST_NAT
;
83 statusbit
= IPS_SRC_NAT
;
85 l3proto
->decode_session(skb
, ct
, dir
, statusbit
, fl
);
88 int nf_xfrm_me_harder(struct net
*net
, struct sk_buff
*skb
, unsigned int family
)
92 struct dst_entry
*dst
;
95 err
= xfrm_decode_session(skb
, &fl
, family
);
101 dst
= ((struct xfrm_dst
*)dst
)->route
;
104 dst
= xfrm_lookup(net
, dst
, &fl
, skb
->sk
, 0);
109 skb_dst_set(skb
, dst
);
111 /* Change in oif may mean change in hh_len. */
112 hh_len
= skb_dst(skb
)->dev
->hard_header_len
;
113 if (skb_headroom(skb
) < hh_len
&&
114 pskb_expand_head(skb
, hh_len
- skb_headroom(skb
), 0, GFP_ATOMIC
))
118 EXPORT_SYMBOL(nf_xfrm_me_harder
);
119 #endif /* CONFIG_XFRM */
121 static u32
nf_nat_bysource_hash(const void *data
, u32 len
, u32 seed
)
123 const struct nf_conntrack_tuple
*t
;
124 const struct nf_conn
*ct
= data
;
126 t
= &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
;
127 /* Original src, to ensure we map it consistently if poss. */
129 seed
^= net_hash_mix(nf_ct_net(ct
));
130 return jhash2((const u32
*)&t
->src
, sizeof(t
->src
) / sizeof(u32
),
131 t
->dst
.protonum
^ seed
);
134 /* Is this tuple already taken? (not by us) */
136 nf_nat_used_tuple(const struct nf_conntrack_tuple
*tuple
,
137 const struct nf_conn
*ignored_conntrack
)
139 /* Conntrack tracking doesn't keep track of outgoing tuples; only
140 * incoming ones. NAT means they don't have a fixed mapping,
141 * so we invert the tuple and look for the incoming reply.
143 * We could keep a separate hash if this proves too slow.
145 struct nf_conntrack_tuple reply
;
147 nf_ct_invert_tuplepr(&reply
, tuple
);
148 return nf_conntrack_tuple_taken(&reply
, ignored_conntrack
);
150 EXPORT_SYMBOL(nf_nat_used_tuple
);
152 /* If we source map this tuple so reply looks like reply_tuple, will
153 * that meet the constraints of range.
155 static int in_range(const struct nf_nat_l3proto
*l3proto
,
156 const struct nf_nat_l4proto
*l4proto
,
157 const struct nf_conntrack_tuple
*tuple
,
158 const struct nf_nat_range
*range
)
160 /* If we are supposed to map IPs, then we must be in the
161 * range specified, otherwise let this drag us onto a new src IP.
163 if (range
->flags
& NF_NAT_RANGE_MAP_IPS
&&
164 !l3proto
->in_range(tuple
, range
))
167 if (!(range
->flags
& NF_NAT_RANGE_PROTO_SPECIFIED
) ||
168 l4proto
->in_range(tuple
, NF_NAT_MANIP_SRC
,
169 &range
->min_proto
, &range
->max_proto
))
176 same_src(const struct nf_conn
*ct
,
177 const struct nf_conntrack_tuple
*tuple
)
179 const struct nf_conntrack_tuple
*t
;
181 t
= &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
;
182 return (t
->dst
.protonum
== tuple
->dst
.protonum
&&
183 nf_inet_addr_cmp(&t
->src
.u3
, &tuple
->src
.u3
) &&
184 t
->src
.u
.all
== tuple
->src
.u
.all
);
187 static int nf_nat_bysource_cmp(struct rhashtable_compare_arg
*arg
,
190 const struct nf_nat_conn_key
*key
= arg
->key
;
191 const struct nf_conn
*ct
= obj
;
193 if (!same_src(ct
, key
->tuple
) ||
194 !net_eq(nf_ct_net(ct
), key
->net
) ||
195 !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
,
209 /* Only called for SRC manip */
211 find_appropriate_src(struct net
*net
,
212 const struct nf_conntrack_zone
*zone
,
213 const struct nf_nat_l3proto
*l3proto
,
214 const struct nf_nat_l4proto
*l4proto
,
215 const struct nf_conntrack_tuple
*tuple
,
216 struct nf_conntrack_tuple
*result
,
217 const struct nf_nat_range
*range
)
219 const struct nf_conn
*ct
;
220 struct nf_nat_conn_key key
= {
225 struct rhlist_head
*hl
;
227 hl
= rhltable_lookup(&nf_nat_bysource_table
, &key
,
228 nf_nat_bysource_params
);
232 ct
= container_of(hl
, typeof(*ct
), nat_bysource
);
234 nf_ct_invert_tuplepr(result
,
235 &ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
);
236 result
->dst
= tuple
->dst
;
238 return in_range(l3proto
, l4proto
, result
, range
);
241 /* For [FUTURE] fragmentation handling, we want the least-used
242 * src-ip/dst-ip/proto triple. Fairness doesn't come into it. Thus
243 * if the range specifies 1.2.3.4 ports 10000-10005 and 1.2.3.5 ports
244 * 1-65535, we don't do pro-rata allocation based on ports; we choose
245 * the ip with the lowest src-ip/dst-ip/proto usage.
248 find_best_ips_proto(const struct nf_conntrack_zone
*zone
,
249 struct nf_conntrack_tuple
*tuple
,
250 const struct nf_nat_range
*range
,
251 const struct nf_conn
*ct
,
252 enum nf_nat_manip_type maniptype
)
254 union nf_inet_addr
*var_ipp
;
257 u32 minip
, maxip
, j
, dist
;
260 /* No IP mapping? Do nothing. */
261 if (!(range
->flags
& NF_NAT_RANGE_MAP_IPS
))
264 if (maniptype
== NF_NAT_MANIP_SRC
)
265 var_ipp
= &tuple
->src
.u3
;
267 var_ipp
= &tuple
->dst
.u3
;
269 /* Fast path: only one choice. */
270 if (nf_inet_addr_cmp(&range
->min_addr
, &range
->max_addr
)) {
271 *var_ipp
= range
->min_addr
;
275 if (nf_ct_l3num(ct
) == NFPROTO_IPV4
)
276 max
= sizeof(var_ipp
->ip
) / sizeof(u32
) - 1;
278 max
= sizeof(var_ipp
->ip6
) / sizeof(u32
) - 1;
280 /* Hashing source and destination IPs gives a fairly even
281 * spread in practice (if there are a small number of IPs
282 * involved, there usually aren't that many connections
283 * anyway). The consistency means that servers see the same
284 * client coming from the same IP (some Internet Banking sites
285 * like this), even across reboots.
287 j
= jhash2((u32
*)&tuple
->src
.u3
, sizeof(tuple
->src
.u3
) / sizeof(u32
),
288 range
->flags
& NF_NAT_RANGE_PERSISTENT
?
289 0 : (__force u32
)tuple
->dst
.u3
.all
[max
] ^ zone
->id
);
292 for (i
= 0; i
<= max
; i
++) {
293 /* If first bytes of the address are at the maximum, use the
294 * distance. Otherwise use the full range.
297 minip
= ntohl((__force __be32
)range
->min_addr
.all
[i
]);
298 maxip
= ntohl((__force __be32
)range
->max_addr
.all
[i
]);
299 dist
= maxip
- minip
+ 1;
305 var_ipp
->all
[i
] = (__force __u32
)
306 htonl(minip
+ reciprocal_scale(j
, dist
));
307 if (var_ipp
->all
[i
] != range
->max_addr
.all
[i
])
310 if (!(range
->flags
& NF_NAT_RANGE_PERSISTENT
))
311 j
^= (__force u32
)tuple
->dst
.u3
.all
[i
];
315 /* Manipulate the tuple into the range given. For NF_INET_POST_ROUTING,
316 * we change the source to map into the range. For NF_INET_PRE_ROUTING
317 * and NF_INET_LOCAL_OUT, we change the destination to map into the
318 * range. It might not be possible to get a unique tuple, but we try.
319 * At worst (or if we race), we will end up with a final duplicate in
320 * __ip_conntrack_confirm and drop the packet. */
322 get_unique_tuple(struct nf_conntrack_tuple
*tuple
,
323 const struct nf_conntrack_tuple
*orig_tuple
,
324 const struct nf_nat_range
*range
,
326 enum nf_nat_manip_type maniptype
)
328 const struct nf_conntrack_zone
*zone
;
329 const struct nf_nat_l3proto
*l3proto
;
330 const struct nf_nat_l4proto
*l4proto
;
331 struct net
*net
= nf_ct_net(ct
);
333 zone
= nf_ct_zone(ct
);
336 l3proto
= __nf_nat_l3proto_find(orig_tuple
->src
.l3num
);
337 l4proto
= __nf_nat_l4proto_find(orig_tuple
->src
.l3num
,
338 orig_tuple
->dst
.protonum
);
340 /* 1) If this srcip/proto/src-proto-part is currently mapped,
341 * and that same mapping gives a unique tuple within the given
344 * This is only required for source (ie. NAT/masq) mappings.
345 * So far, we don't do local source mappings, so multiple
346 * manips not an issue.
348 if (maniptype
== NF_NAT_MANIP_SRC
&&
349 !(range
->flags
& NF_NAT_RANGE_PROTO_RANDOM_ALL
)) {
350 /* try the original tuple first */
351 if (in_range(l3proto
, l4proto
, orig_tuple
, range
)) {
352 if (!nf_nat_used_tuple(orig_tuple
, ct
)) {
353 *tuple
= *orig_tuple
;
356 } else if (find_appropriate_src(net
, zone
, l3proto
, l4proto
,
357 orig_tuple
, tuple
, range
)) {
358 pr_debug("get_unique_tuple: Found current src map\n");
359 if (!nf_nat_used_tuple(tuple
, ct
))
364 /* 2) Select the least-used IP/proto combination in the given range */
365 *tuple
= *orig_tuple
;
366 find_best_ips_proto(zone
, tuple
, range
, ct
, maniptype
);
368 /* 3) The per-protocol part of the manip is made to map into
369 * the range to make a unique tuple.
372 /* Only bother mapping if it's not already in range and unique */
373 if (!(range
->flags
& NF_NAT_RANGE_PROTO_RANDOM_ALL
)) {
374 if (range
->flags
& NF_NAT_RANGE_PROTO_SPECIFIED
) {
375 if (l4proto
->in_range(tuple
, maniptype
,
377 &range
->max_proto
) &&
378 (range
->min_proto
.all
== range
->max_proto
.all
||
379 !nf_nat_used_tuple(tuple
, ct
)))
381 } else if (!nf_nat_used_tuple(tuple
, ct
)) {
386 /* Last change: get protocol to try to obtain unique tuple. */
387 l4proto
->unique_tuple(l3proto
, tuple
, range
, maniptype
, ct
);
392 struct nf_conn_nat
*nf_ct_nat_ext_add(struct nf_conn
*ct
)
394 struct nf_conn_nat
*nat
= nfct_nat(ct
);
398 if (!nf_ct_is_confirmed(ct
))
399 nat
= nf_ct_ext_add(ct
, NF_CT_EXT_NAT
, GFP_ATOMIC
);
403 EXPORT_SYMBOL_GPL(nf_ct_nat_ext_add
);
406 nf_nat_setup_info(struct nf_conn
*ct
,
407 const struct nf_nat_range
*range
,
408 enum nf_nat_manip_type maniptype
)
410 struct nf_conntrack_tuple curr_tuple
, new_tuple
;
412 /* Can't setup nat info for confirmed ct. */
413 if (nf_ct_is_confirmed(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 if (!nfct_seqadj_ext_add(ct
))
448 if (maniptype
== NF_NAT_MANIP_SRC
) {
449 struct nf_nat_conn_key key
= {
450 .net
= nf_ct_net(ct
),
451 .tuple
= &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
,
452 .zone
= nf_ct_zone(ct
),
456 err
= rhltable_insert_key(&nf_nat_bysource_table
,
459 nf_nat_bysource_params
);
465 if (maniptype
== NF_NAT_MANIP_DST
)
466 ct
->status
|= IPS_DST_NAT_DONE
;
468 ct
->status
|= IPS_SRC_NAT_DONE
;
472 EXPORT_SYMBOL(nf_nat_setup_info
);
475 __nf_nat_alloc_null_binding(struct nf_conn
*ct
, enum nf_nat_manip_type manip
)
477 /* Force range to this IP; let proto decide mapping for
478 * per-proto parts (hence not IP_NAT_RANGE_PROTO_SPECIFIED).
479 * Use reply in case it's already been mangled (eg local packet).
481 union nf_inet_addr ip
=
482 (manip
== NF_NAT_MANIP_SRC
?
483 ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
.dst
.u3
:
484 ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
.src
.u3
);
485 struct nf_nat_range range
= {
486 .flags
= NF_NAT_RANGE_MAP_IPS
,
490 return nf_nat_setup_info(ct
, &range
, manip
);
494 nf_nat_alloc_null_binding(struct nf_conn
*ct
, unsigned int hooknum
)
496 return __nf_nat_alloc_null_binding(ct
, HOOK2MANIP(hooknum
));
498 EXPORT_SYMBOL_GPL(nf_nat_alloc_null_binding
);
500 /* Do packet manipulations according to nf_nat_setup_info. */
501 unsigned int nf_nat_packet(struct nf_conn
*ct
,
502 enum ip_conntrack_info ctinfo
,
503 unsigned int hooknum
,
506 const struct nf_nat_l3proto
*l3proto
;
507 const struct nf_nat_l4proto
*l4proto
;
508 enum ip_conntrack_dir dir
= CTINFO2DIR(ctinfo
);
509 unsigned long statusbit
;
510 enum nf_nat_manip_type mtype
= HOOK2MANIP(hooknum
);
512 if (mtype
== NF_NAT_MANIP_SRC
)
513 statusbit
= IPS_SRC_NAT
;
515 statusbit
= IPS_DST_NAT
;
517 /* Invert if this is reply dir. */
518 if (dir
== IP_CT_DIR_REPLY
)
519 statusbit
^= IPS_NAT_MASK
;
521 /* Non-atomic: these bits don't change. */
522 if (ct
->status
& statusbit
) {
523 struct nf_conntrack_tuple target
;
525 /* We are aiming to look like inverse of other direction. */
526 nf_ct_invert_tuplepr(&target
, &ct
->tuplehash
[!dir
].tuple
);
528 l3proto
= __nf_nat_l3proto_find(target
.src
.l3num
);
529 l4proto
= __nf_nat_l4proto_find(target
.src
.l3num
,
530 target
.dst
.protonum
);
531 if (!l3proto
->manip_pkt(skb
, 0, l4proto
, &target
, mtype
))
536 EXPORT_SYMBOL_GPL(nf_nat_packet
);
538 struct nf_nat_proto_clean
{
543 /* kill conntracks with affected NAT section */
544 static int nf_nat_proto_remove(struct nf_conn
*i
, void *data
)
546 const struct nf_nat_proto_clean
*clean
= data
;
548 if ((clean
->l3proto
&& nf_ct_l3num(i
) != clean
->l3proto
) ||
549 (clean
->l4proto
&& nf_ct_protonum(i
) != clean
->l4proto
))
552 return i
->status
& IPS_NAT_MASK
? 1 : 0;
555 static int nf_nat_proto_clean(struct nf_conn
*ct
, void *data
)
557 if (nf_nat_proto_remove(ct
, data
))
560 if ((ct
->status
& IPS_SRC_NAT_DONE
) == 0)
563 /* This netns is being destroyed, and conntrack has nat null binding.
564 * Remove it from bysource hash, as the table will be freed soon.
566 * Else, when the conntrack is destoyed, nf_nat_cleanup_conntrack()
567 * will delete entry from already-freed table.
569 clear_bit(IPS_SRC_NAT_DONE_BIT
, &ct
->status
);
570 rhltable_remove(&nf_nat_bysource_table
, &ct
->nat_bysource
,
571 nf_nat_bysource_params
);
573 /* don't delete conntrack. Although that would make things a lot
574 * simpler, we'd end up flushing all conntracks on nat rmmod.
579 static void nf_nat_l4proto_clean(u8 l3proto
, u8 l4proto
)
581 struct nf_nat_proto_clean clean
= {
586 nf_ct_iterate_destroy(nf_nat_proto_remove
, &clean
);
589 static void nf_nat_l3proto_clean(u8 l3proto
)
591 struct nf_nat_proto_clean clean
= {
595 nf_ct_iterate_destroy(nf_nat_proto_remove
, &clean
);
598 /* Protocol registration. */
599 int nf_nat_l4proto_register(u8 l3proto
, const struct nf_nat_l4proto
*l4proto
)
601 const struct nf_nat_l4proto
**l4protos
;
605 mutex_lock(&nf_nat_proto_mutex
);
606 if (nf_nat_l4protos
[l3proto
] == NULL
) {
607 l4protos
= kmalloc(IPPROTO_MAX
* sizeof(struct nf_nat_l4proto
*),
609 if (l4protos
== NULL
) {
614 for (i
= 0; i
< IPPROTO_MAX
; i
++)
615 RCU_INIT_POINTER(l4protos
[i
], &nf_nat_l4proto_unknown
);
617 /* Before making proto_array visible to lockless readers,
618 * we must make sure its content is committed to memory.
622 nf_nat_l4protos
[l3proto
] = l4protos
;
625 if (rcu_dereference_protected(
626 nf_nat_l4protos
[l3proto
][l4proto
->l4proto
],
627 lockdep_is_held(&nf_nat_proto_mutex
)
628 ) != &nf_nat_l4proto_unknown
) {
632 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
][l4proto
->l4proto
], l4proto
);
634 mutex_unlock(&nf_nat_proto_mutex
);
637 EXPORT_SYMBOL_GPL(nf_nat_l4proto_register
);
639 /* No one stores the protocol anywhere; simply delete it. */
640 void nf_nat_l4proto_unregister(u8 l3proto
, const struct nf_nat_l4proto
*l4proto
)
642 mutex_lock(&nf_nat_proto_mutex
);
643 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
][l4proto
->l4proto
],
644 &nf_nat_l4proto_unknown
);
645 mutex_unlock(&nf_nat_proto_mutex
);
648 nf_nat_l4proto_clean(l3proto
, l4proto
->l4proto
);
650 EXPORT_SYMBOL_GPL(nf_nat_l4proto_unregister
);
652 int nf_nat_l3proto_register(const struct nf_nat_l3proto
*l3proto
)
656 err
= nf_ct_l3proto_try_module_get(l3proto
->l3proto
);
660 mutex_lock(&nf_nat_proto_mutex
);
661 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
->l3proto
][IPPROTO_TCP
],
662 &nf_nat_l4proto_tcp
);
663 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
->l3proto
][IPPROTO_UDP
],
664 &nf_nat_l4proto_udp
);
665 #ifdef CONFIG_NF_NAT_PROTO_DCCP
666 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
->l3proto
][IPPROTO_DCCP
],
667 &nf_nat_l4proto_dccp
);
669 #ifdef CONFIG_NF_NAT_PROTO_SCTP
670 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
->l3proto
][IPPROTO_SCTP
],
671 &nf_nat_l4proto_sctp
);
673 #ifdef CONFIG_NF_NAT_PROTO_UDPLITE
674 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
->l3proto
][IPPROTO_UDPLITE
],
675 &nf_nat_l4proto_udplite
);
677 mutex_unlock(&nf_nat_proto_mutex
);
679 RCU_INIT_POINTER(nf_nat_l3protos
[l3proto
->l3proto
], l3proto
);
682 EXPORT_SYMBOL_GPL(nf_nat_l3proto_register
);
684 void nf_nat_l3proto_unregister(const struct nf_nat_l3proto
*l3proto
)
686 mutex_lock(&nf_nat_proto_mutex
);
687 RCU_INIT_POINTER(nf_nat_l3protos
[l3proto
->l3proto
], NULL
);
688 mutex_unlock(&nf_nat_proto_mutex
);
691 nf_nat_l3proto_clean(l3proto
->l3proto
);
692 nf_ct_l3proto_module_put(l3proto
->l3proto
);
694 EXPORT_SYMBOL_GPL(nf_nat_l3proto_unregister
);
696 /* No one using conntrack by the time this called. */
697 static void nf_nat_cleanup_conntrack(struct nf_conn
*ct
)
699 if (ct
->status
& IPS_SRC_NAT_DONE
)
700 rhltable_remove(&nf_nat_bysource_table
, &ct
->nat_bysource
,
701 nf_nat_bysource_params
);
704 static struct nf_ct_ext_type nat_extend __read_mostly
= {
705 .len
= sizeof(struct nf_conn_nat
),
706 .align
= __alignof__(struct nf_conn_nat
),
707 .destroy
= nf_nat_cleanup_conntrack
,
711 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
713 #include <linux/netfilter/nfnetlink.h>
714 #include <linux/netfilter/nfnetlink_conntrack.h>
716 static const struct nla_policy protonat_nla_policy
[CTA_PROTONAT_MAX
+1] = {
717 [CTA_PROTONAT_PORT_MIN
] = { .type
= NLA_U16
},
718 [CTA_PROTONAT_PORT_MAX
] = { .type
= NLA_U16
},
721 static int nfnetlink_parse_nat_proto(struct nlattr
*attr
,
722 const struct nf_conn
*ct
,
723 struct nf_nat_range
*range
)
725 struct nlattr
*tb
[CTA_PROTONAT_MAX
+1];
726 const struct nf_nat_l4proto
*l4proto
;
729 err
= nla_parse_nested(tb
, CTA_PROTONAT_MAX
, attr
,
730 protonat_nla_policy
, NULL
);
734 l4proto
= __nf_nat_l4proto_find(nf_ct_l3num(ct
), nf_ct_protonum(ct
));
735 if (l4proto
->nlattr_to_range
)
736 err
= l4proto
->nlattr_to_range(tb
, range
);
741 static const struct nla_policy nat_nla_policy
[CTA_NAT_MAX
+1] = {
742 [CTA_NAT_V4_MINIP
] = { .type
= NLA_U32
},
743 [CTA_NAT_V4_MAXIP
] = { .type
= NLA_U32
},
744 [CTA_NAT_V6_MINIP
] = { .len
= sizeof(struct in6_addr
) },
745 [CTA_NAT_V6_MAXIP
] = { .len
= sizeof(struct in6_addr
) },
746 [CTA_NAT_PROTO
] = { .type
= NLA_NESTED
},
750 nfnetlink_parse_nat(const struct nlattr
*nat
,
751 const struct nf_conn
*ct
, struct nf_nat_range
*range
,
752 const struct nf_nat_l3proto
*l3proto
)
754 struct nlattr
*tb
[CTA_NAT_MAX
+1];
757 memset(range
, 0, sizeof(*range
));
759 err
= nla_parse_nested(tb
, CTA_NAT_MAX
, nat
, nat_nla_policy
, NULL
);
763 err
= l3proto
->nlattr_to_range(tb
, range
);
767 if (!tb
[CTA_NAT_PROTO
])
770 return nfnetlink_parse_nat_proto(tb
[CTA_NAT_PROTO
], ct
, range
);
773 /* This function is called under rcu_read_lock() */
775 nfnetlink_parse_nat_setup(struct nf_conn
*ct
,
776 enum nf_nat_manip_type manip
,
777 const struct nlattr
*attr
)
779 struct nf_nat_range range
;
780 const struct nf_nat_l3proto
*l3proto
;
783 /* Should not happen, restricted to creating new conntracks
786 if (WARN_ON_ONCE(nf_nat_initialized(ct
, manip
)))
789 /* Make sure that L3 NAT is there by when we call nf_nat_setup_info to
790 * attach the null binding, otherwise this may oops.
792 l3proto
= __nf_nat_l3proto_find(nf_ct_l3num(ct
));
796 /* No NAT information has been passed, allocate the null-binding */
798 return __nf_nat_alloc_null_binding(ct
, manip
) == NF_DROP
? -ENOMEM
: 0;
800 err
= nfnetlink_parse_nat(attr
, ct
, &range
, l3proto
);
804 return nf_nat_setup_info(ct
, &range
, manip
) == NF_DROP
? -ENOMEM
: 0;
808 nfnetlink_parse_nat_setup(struct nf_conn
*ct
,
809 enum nf_nat_manip_type manip
,
810 const struct nlattr
*attr
)
816 static struct nf_ct_helper_expectfn follow_master_nat
= {
817 .name
= "nat-follow-master",
818 .expectfn
= nf_nat_follow_master
,
821 static int __init
nf_nat_init(void)
825 ret
= rhltable_init(&nf_nat_bysource_table
, &nf_nat_bysource_params
);
829 ret
= nf_ct_extend_register(&nat_extend
);
831 rhltable_destroy(&nf_nat_bysource_table
);
832 printk(KERN_ERR
"nf_nat_core: Unable to register extension\n");
836 nf_ct_helper_expectfn_register(&follow_master_nat
);
838 BUG_ON(nfnetlink_parse_nat_setup_hook
!= NULL
);
839 RCU_INIT_POINTER(nfnetlink_parse_nat_setup_hook
,
840 nfnetlink_parse_nat_setup
);
842 BUG_ON(nf_nat_decode_session_hook
!= NULL
);
843 RCU_INIT_POINTER(nf_nat_decode_session_hook
, __nf_nat_decode_session
);
848 static void __exit
nf_nat_cleanup(void)
850 struct nf_nat_proto_clean clean
= {};
853 nf_ct_iterate_destroy(nf_nat_proto_clean
, &clean
);
855 nf_ct_extend_unregister(&nat_extend
);
856 nf_ct_helper_expectfn_unregister(&follow_master_nat
);
857 RCU_INIT_POINTER(nfnetlink_parse_nat_setup_hook
, NULL
);
859 RCU_INIT_POINTER(nf_nat_decode_session_hook
, NULL
);
863 for (i
= 0; i
< NFPROTO_NUMPROTO
; i
++)
864 kfree(nf_nat_l4protos
[i
]);
866 rhltable_destroy(&nf_nat_bysource_table
);
869 MODULE_LICENSE("GPL");
871 module_init(nf_nat_init
);
872 module_exit(nf_nat_cleanup
);