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_SPINLOCK(nf_nat_lock
);
35 static DEFINE_MUTEX(nf_nat_proto_mutex
);
36 static const struct nf_nat_l3proto __rcu
*nf_nat_l3protos
[NFPROTO_NUMPROTO
]
38 static const struct nf_nat_l4proto __rcu
**nf_nat_l4protos
[NFPROTO_NUMPROTO
]
42 inline const struct nf_nat_l3proto
*
43 __nf_nat_l3proto_find(u8 family
)
45 return rcu_dereference(nf_nat_l3protos
[family
]);
48 inline const struct nf_nat_l4proto
*
49 __nf_nat_l4proto_find(u8 family
, u8 protonum
)
51 return rcu_dereference(nf_nat_l4protos
[family
][protonum
]);
53 EXPORT_SYMBOL_GPL(__nf_nat_l4proto_find
);
56 static void __nf_nat_decode_session(struct sk_buff
*skb
, struct flowi
*fl
)
58 const struct nf_nat_l3proto
*l3proto
;
59 const struct nf_conn
*ct
;
60 enum ip_conntrack_info ctinfo
;
61 enum ip_conntrack_dir dir
;
62 unsigned long statusbit
;
65 ct
= nf_ct_get(skb
, &ctinfo
);
69 family
= ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
.src
.l3num
;
71 l3proto
= __nf_nat_l3proto_find(family
);
75 dir
= CTINFO2DIR(ctinfo
);
76 if (dir
== IP_CT_DIR_ORIGINAL
)
77 statusbit
= IPS_DST_NAT
;
79 statusbit
= IPS_SRC_NAT
;
81 l3proto
->decode_session(skb
, ct
, dir
, statusbit
, fl
);
86 int nf_xfrm_me_harder(struct sk_buff
*skb
, unsigned int family
)
90 struct dst_entry
*dst
;
93 err
= xfrm_decode_session(skb
, &fl
, family
);
99 dst
= ((struct xfrm_dst
*)dst
)->route
;
102 dst
= xfrm_lookup(dev_net(dst
->dev
), dst
, &fl
, skb
->sk
, 0);
107 skb_dst_set(skb
, dst
);
109 /* Change in oif may mean change in hh_len. */
110 hh_len
= skb_dst(skb
)->dev
->hard_header_len
;
111 if (skb_headroom(skb
) < hh_len
&&
112 pskb_expand_head(skb
, hh_len
- skb_headroom(skb
), 0, GFP_ATOMIC
))
116 EXPORT_SYMBOL(nf_xfrm_me_harder
);
117 #endif /* CONFIG_XFRM */
119 /* We keep an extra hash for each conntrack, for fast searching. */
120 static inline unsigned int
121 hash_by_src(const struct net
*net
, u16 zone
,
122 const struct nf_conntrack_tuple
*tuple
)
126 /* Original src, to ensure we map it consistently if poss. */
127 hash
= jhash2((u32
*)&tuple
->src
, sizeof(tuple
->src
) / sizeof(u32
),
128 tuple
->dst
.protonum
^ zone
^ nf_conntrack_hash_rnd
);
130 return reciprocal_scale(hash
, net
->ct
.nat_htable_size
);
133 /* Is this tuple already taken? (not by us) */
135 nf_nat_used_tuple(const struct nf_conntrack_tuple
*tuple
,
136 const struct nf_conn
*ignored_conntrack
)
138 /* Conntrack tracking doesn't keep track of outgoing tuples; only
139 * incoming ones. NAT means they don't have a fixed mapping,
140 * so we invert the tuple and look for the incoming reply.
142 * We could keep a separate hash if this proves too slow.
144 struct nf_conntrack_tuple reply
;
146 nf_ct_invert_tuplepr(&reply
, tuple
);
147 return nf_conntrack_tuple_taken(&reply
, ignored_conntrack
);
149 EXPORT_SYMBOL(nf_nat_used_tuple
);
151 /* If we source map this tuple so reply looks like reply_tuple, will
152 * that meet the constraints of range.
154 static int in_range(const struct nf_nat_l3proto
*l3proto
,
155 const struct nf_nat_l4proto
*l4proto
,
156 const struct nf_conntrack_tuple
*tuple
,
157 const struct nf_nat_range
*range
)
159 /* If we are supposed to map IPs, then we must be in the
160 * range specified, otherwise let this drag us onto a new src IP.
162 if (range
->flags
& NF_NAT_RANGE_MAP_IPS
&&
163 !l3proto
->in_range(tuple
, range
))
166 if (!(range
->flags
& NF_NAT_RANGE_PROTO_SPECIFIED
) ||
167 l4proto
->in_range(tuple
, NF_NAT_MANIP_SRC
,
168 &range
->min_proto
, &range
->max_proto
))
175 same_src(const struct nf_conn
*ct
,
176 const struct nf_conntrack_tuple
*tuple
)
178 const struct nf_conntrack_tuple
*t
;
180 t
= &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
;
181 return (t
->dst
.protonum
== tuple
->dst
.protonum
&&
182 nf_inet_addr_cmp(&t
->src
.u3
, &tuple
->src
.u3
) &&
183 t
->src
.u
.all
== tuple
->src
.u
.all
);
186 /* Only called for SRC manip */
188 find_appropriate_src(struct net
*net
, u16 zone
,
189 const struct nf_nat_l3proto
*l3proto
,
190 const struct nf_nat_l4proto
*l4proto
,
191 const struct nf_conntrack_tuple
*tuple
,
192 struct nf_conntrack_tuple
*result
,
193 const struct nf_nat_range
*range
)
195 unsigned int h
= hash_by_src(net
, zone
, tuple
);
196 const struct nf_conn_nat
*nat
;
197 const struct nf_conn
*ct
;
199 hlist_for_each_entry_rcu(nat
, &net
->ct
.nat_bysource
[h
], bysource
) {
201 if (same_src(ct
, tuple
) && nf_ct_zone(ct
) == zone
) {
202 /* Copy source part from reply tuple. */
203 nf_ct_invert_tuplepr(result
,
204 &ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
);
205 result
->dst
= tuple
->dst
;
207 if (in_range(l3proto
, l4proto
, result
, range
))
214 /* For [FUTURE] fragmentation handling, we want the least-used
215 * src-ip/dst-ip/proto triple. Fairness doesn't come into it. Thus
216 * if the range specifies 1.2.3.4 ports 10000-10005 and 1.2.3.5 ports
217 * 1-65535, we don't do pro-rata allocation based on ports; we choose
218 * the ip with the lowest src-ip/dst-ip/proto usage.
221 find_best_ips_proto(u16 zone
, struct nf_conntrack_tuple
*tuple
,
222 const struct nf_nat_range
*range
,
223 const struct nf_conn
*ct
,
224 enum nf_nat_manip_type maniptype
)
226 union nf_inet_addr
*var_ipp
;
229 u32 minip
, maxip
, j
, dist
;
232 /* No IP mapping? Do nothing. */
233 if (!(range
->flags
& NF_NAT_RANGE_MAP_IPS
))
236 if (maniptype
== NF_NAT_MANIP_SRC
)
237 var_ipp
= &tuple
->src
.u3
;
239 var_ipp
= &tuple
->dst
.u3
;
241 /* Fast path: only one choice. */
242 if (nf_inet_addr_cmp(&range
->min_addr
, &range
->max_addr
)) {
243 *var_ipp
= range
->min_addr
;
247 if (nf_ct_l3num(ct
) == NFPROTO_IPV4
)
248 max
= sizeof(var_ipp
->ip
) / sizeof(u32
) - 1;
250 max
= sizeof(var_ipp
->ip6
) / sizeof(u32
) - 1;
252 /* Hashing source and destination IPs gives a fairly even
253 * spread in practice (if there are a small number of IPs
254 * involved, there usually aren't that many connections
255 * anyway). The consistency means that servers see the same
256 * client coming from the same IP (some Internet Banking sites
257 * like this), even across reboots.
259 j
= jhash2((u32
*)&tuple
->src
.u3
, sizeof(tuple
->src
.u3
) / sizeof(u32
),
260 range
->flags
& NF_NAT_RANGE_PERSISTENT
?
261 0 : (__force u32
)tuple
->dst
.u3
.all
[max
] ^ zone
);
264 for (i
= 0; i
<= max
; i
++) {
265 /* If first bytes of the address are at the maximum, use the
266 * distance. Otherwise use the full range.
269 minip
= ntohl((__force __be32
)range
->min_addr
.all
[i
]);
270 maxip
= ntohl((__force __be32
)range
->max_addr
.all
[i
]);
271 dist
= maxip
- minip
+ 1;
277 var_ipp
->all
[i
] = (__force __u32
)
278 htonl(minip
+ reciprocal_scale(j
, dist
));
279 if (var_ipp
->all
[i
] != range
->max_addr
.all
[i
])
282 if (!(range
->flags
& NF_NAT_RANGE_PERSISTENT
))
283 j
^= (__force u32
)tuple
->dst
.u3
.all
[i
];
287 /* Manipulate the tuple into the range given. For NF_INET_POST_ROUTING,
288 * we change the source to map into the range. For NF_INET_PRE_ROUTING
289 * and NF_INET_LOCAL_OUT, we change the destination to map into the
290 * range. It might not be possible to get a unique tuple, but we try.
291 * At worst (or if we race), we will end up with a final duplicate in
292 * __ip_conntrack_confirm and drop the packet. */
294 get_unique_tuple(struct nf_conntrack_tuple
*tuple
,
295 const struct nf_conntrack_tuple
*orig_tuple
,
296 const struct nf_nat_range
*range
,
298 enum nf_nat_manip_type maniptype
)
300 const struct nf_nat_l3proto
*l3proto
;
301 const struct nf_nat_l4proto
*l4proto
;
302 struct net
*net
= nf_ct_net(ct
);
303 u16 zone
= nf_ct_zone(ct
);
306 l3proto
= __nf_nat_l3proto_find(orig_tuple
->src
.l3num
);
307 l4proto
= __nf_nat_l4proto_find(orig_tuple
->src
.l3num
,
308 orig_tuple
->dst
.protonum
);
310 /* 1) If this srcip/proto/src-proto-part is currently mapped,
311 * and that same mapping gives a unique tuple within the given
314 * This is only required for source (ie. NAT/masq) mappings.
315 * So far, we don't do local source mappings, so multiple
316 * manips not an issue.
318 if (maniptype
== NF_NAT_MANIP_SRC
&&
319 !(range
->flags
& NF_NAT_RANGE_PROTO_RANDOM_ALL
)) {
320 /* try the original tuple first */
321 if (in_range(l3proto
, l4proto
, orig_tuple
, range
)) {
322 if (!nf_nat_used_tuple(orig_tuple
, ct
)) {
323 *tuple
= *orig_tuple
;
326 } else if (find_appropriate_src(net
, zone
, l3proto
, l4proto
,
327 orig_tuple
, tuple
, range
)) {
328 pr_debug("get_unique_tuple: Found current src map\n");
329 if (!nf_nat_used_tuple(tuple
, ct
))
334 /* 2) Select the least-used IP/proto combination in the given range */
335 *tuple
= *orig_tuple
;
336 find_best_ips_proto(zone
, tuple
, range
, ct
, maniptype
);
338 /* 3) The per-protocol part of the manip is made to map into
339 * the range to make a unique tuple.
342 /* Only bother mapping if it's not already in range and unique */
343 if (!(range
->flags
& NF_NAT_RANGE_PROTO_RANDOM_ALL
)) {
344 if (range
->flags
& NF_NAT_RANGE_PROTO_SPECIFIED
) {
345 if (l4proto
->in_range(tuple
, maniptype
,
347 &range
->max_proto
) &&
348 (range
->min_proto
.all
== range
->max_proto
.all
||
349 !nf_nat_used_tuple(tuple
, ct
)))
351 } else if (!nf_nat_used_tuple(tuple
, ct
)) {
356 /* Last change: get protocol to try to obtain unique tuple. */
357 l4proto
->unique_tuple(l3proto
, tuple
, range
, maniptype
, ct
);
362 struct nf_conn_nat
*nf_ct_nat_ext_add(struct nf_conn
*ct
)
364 struct nf_conn_nat
*nat
= nfct_nat(ct
);
368 if (!nf_ct_is_confirmed(ct
))
369 nat
= nf_ct_ext_add(ct
, NF_CT_EXT_NAT
, GFP_ATOMIC
);
373 EXPORT_SYMBOL_GPL(nf_ct_nat_ext_add
);
376 nf_nat_setup_info(struct nf_conn
*ct
,
377 const struct nf_nat_range
*range
,
378 enum nf_nat_manip_type maniptype
)
380 struct net
*net
= nf_ct_net(ct
);
381 struct nf_conntrack_tuple curr_tuple
, new_tuple
;
382 struct nf_conn_nat
*nat
;
384 /* nat helper or nfctnetlink also setup binding */
385 nat
= nf_ct_nat_ext_add(ct
);
389 NF_CT_ASSERT(maniptype
== NF_NAT_MANIP_SRC
||
390 maniptype
== NF_NAT_MANIP_DST
);
391 BUG_ON(nf_nat_initialized(ct
, maniptype
));
393 /* What we've got will look like inverse of reply. Normally
394 * this is what is in the conntrack, except for prior
395 * manipulations (future optimization: if num_manips == 0,
396 * orig_tp = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple)
398 nf_ct_invert_tuplepr(&curr_tuple
,
399 &ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
);
401 get_unique_tuple(&new_tuple
, &curr_tuple
, range
, ct
, maniptype
);
403 if (!nf_ct_tuple_equal(&new_tuple
, &curr_tuple
)) {
404 struct nf_conntrack_tuple reply
;
406 /* Alter conntrack table so will recognize replies. */
407 nf_ct_invert_tuplepr(&reply
, &new_tuple
);
408 nf_conntrack_alter_reply(ct
, &reply
);
410 /* Non-atomic: we own this at the moment. */
411 if (maniptype
== NF_NAT_MANIP_SRC
)
412 ct
->status
|= IPS_SRC_NAT
;
414 ct
->status
|= IPS_DST_NAT
;
417 nfct_seqadj_ext_add(ct
);
420 if (maniptype
== NF_NAT_MANIP_SRC
) {
421 unsigned int srchash
;
423 srchash
= hash_by_src(net
, nf_ct_zone(ct
),
424 &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
);
425 spin_lock_bh(&nf_nat_lock
);
426 /* nf_conntrack_alter_reply might re-allocate extension aera */
429 hlist_add_head_rcu(&nat
->bysource
,
430 &net
->ct
.nat_bysource
[srchash
]);
431 spin_unlock_bh(&nf_nat_lock
);
435 if (maniptype
== NF_NAT_MANIP_DST
)
436 ct
->status
|= IPS_DST_NAT_DONE
;
438 ct
->status
|= IPS_SRC_NAT_DONE
;
442 EXPORT_SYMBOL(nf_nat_setup_info
);
445 __nf_nat_alloc_null_binding(struct nf_conn
*ct
, enum nf_nat_manip_type manip
)
447 /* Force range to this IP; let proto decide mapping for
448 * per-proto parts (hence not IP_NAT_RANGE_PROTO_SPECIFIED).
449 * Use reply in case it's already been mangled (eg local packet).
451 union nf_inet_addr ip
=
452 (manip
== NF_NAT_MANIP_SRC
?
453 ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
.dst
.u3
:
454 ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
.src
.u3
);
455 struct nf_nat_range range
= {
456 .flags
= NF_NAT_RANGE_MAP_IPS
,
460 return nf_nat_setup_info(ct
, &range
, manip
);
464 nf_nat_alloc_null_binding(struct nf_conn
*ct
, unsigned int hooknum
)
466 return __nf_nat_alloc_null_binding(ct
, HOOK2MANIP(hooknum
));
468 EXPORT_SYMBOL_GPL(nf_nat_alloc_null_binding
);
470 /* Do packet manipulations according to nf_nat_setup_info. */
471 unsigned int nf_nat_packet(struct nf_conn
*ct
,
472 enum ip_conntrack_info ctinfo
,
473 unsigned int hooknum
,
476 const struct nf_nat_l3proto
*l3proto
;
477 const struct nf_nat_l4proto
*l4proto
;
478 enum ip_conntrack_dir dir
= CTINFO2DIR(ctinfo
);
479 unsigned long statusbit
;
480 enum nf_nat_manip_type mtype
= HOOK2MANIP(hooknum
);
482 if (mtype
== NF_NAT_MANIP_SRC
)
483 statusbit
= IPS_SRC_NAT
;
485 statusbit
= IPS_DST_NAT
;
487 /* Invert if this is reply dir. */
488 if (dir
== IP_CT_DIR_REPLY
)
489 statusbit
^= IPS_NAT_MASK
;
491 /* Non-atomic: these bits don't change. */
492 if (ct
->status
& statusbit
) {
493 struct nf_conntrack_tuple target
;
495 /* We are aiming to look like inverse of other direction. */
496 nf_ct_invert_tuplepr(&target
, &ct
->tuplehash
[!dir
].tuple
);
498 l3proto
= __nf_nat_l3proto_find(target
.src
.l3num
);
499 l4proto
= __nf_nat_l4proto_find(target
.src
.l3num
,
500 target
.dst
.protonum
);
501 if (!l3proto
->manip_pkt(skb
, 0, l4proto
, &target
, mtype
))
506 EXPORT_SYMBOL_GPL(nf_nat_packet
);
508 struct nf_nat_proto_clean
{
513 /* kill conntracks with affected NAT section */
514 static int nf_nat_proto_remove(struct nf_conn
*i
, void *data
)
516 const struct nf_nat_proto_clean
*clean
= data
;
517 struct nf_conn_nat
*nat
= nfct_nat(i
);
522 if ((clean
->l3proto
&& nf_ct_l3num(i
) != clean
->l3proto
) ||
523 (clean
->l4proto
&& nf_ct_protonum(i
) != clean
->l4proto
))
526 return i
->status
& IPS_NAT_MASK
? 1 : 0;
529 static int nf_nat_proto_clean(struct nf_conn
*ct
, void *data
)
531 struct nf_conn_nat
*nat
= nfct_nat(ct
);
533 if (nf_nat_proto_remove(ct
, data
))
536 if (!nat
|| !nat
->ct
)
539 /* This netns is being destroyed, and conntrack has nat null binding.
540 * Remove it from bysource hash, as the table will be freed soon.
542 * Else, when the conntrack is destoyed, nf_nat_cleanup_conntrack()
543 * will delete entry from already-freed table.
545 if (!del_timer(&ct
->timeout
))
548 spin_lock_bh(&nf_nat_lock
);
549 hlist_del_rcu(&nat
->bysource
);
550 ct
->status
&= ~IPS_NAT_DONE_MASK
;
552 spin_unlock_bh(&nf_nat_lock
);
554 add_timer(&ct
->timeout
);
556 /* don't delete conntrack. Although that would make things a lot
557 * simpler, we'd end up flushing all conntracks on nat rmmod.
562 static void nf_nat_l4proto_clean(u8 l3proto
, u8 l4proto
)
564 struct nf_nat_proto_clean clean
= {
572 nf_ct_iterate_cleanup(net
, nf_nat_proto_remove
, &clean
, 0, 0);
576 static void nf_nat_l3proto_clean(u8 l3proto
)
578 struct nf_nat_proto_clean clean
= {
586 nf_ct_iterate_cleanup(net
, nf_nat_proto_remove
, &clean
, 0, 0);
590 /* Protocol registration. */
591 int nf_nat_l4proto_register(u8 l3proto
, const struct nf_nat_l4proto
*l4proto
)
593 const struct nf_nat_l4proto
**l4protos
;
597 mutex_lock(&nf_nat_proto_mutex
);
598 if (nf_nat_l4protos
[l3proto
] == NULL
) {
599 l4protos
= kmalloc(IPPROTO_MAX
* sizeof(struct nf_nat_l4proto
*),
601 if (l4protos
== NULL
) {
606 for (i
= 0; i
< IPPROTO_MAX
; i
++)
607 RCU_INIT_POINTER(l4protos
[i
], &nf_nat_l4proto_unknown
);
609 /* Before making proto_array visible to lockless readers,
610 * we must make sure its content is committed to memory.
614 nf_nat_l4protos
[l3proto
] = l4protos
;
617 if (rcu_dereference_protected(
618 nf_nat_l4protos
[l3proto
][l4proto
->l4proto
],
619 lockdep_is_held(&nf_nat_proto_mutex
)
620 ) != &nf_nat_l4proto_unknown
) {
624 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
][l4proto
->l4proto
], l4proto
);
626 mutex_unlock(&nf_nat_proto_mutex
);
629 EXPORT_SYMBOL_GPL(nf_nat_l4proto_register
);
631 /* No one stores the protocol anywhere; simply delete it. */
632 void nf_nat_l4proto_unregister(u8 l3proto
, const struct nf_nat_l4proto
*l4proto
)
634 mutex_lock(&nf_nat_proto_mutex
);
635 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
][l4proto
->l4proto
],
636 &nf_nat_l4proto_unknown
);
637 mutex_unlock(&nf_nat_proto_mutex
);
640 nf_nat_l4proto_clean(l3proto
, l4proto
->l4proto
);
642 EXPORT_SYMBOL_GPL(nf_nat_l4proto_unregister
);
644 int nf_nat_l3proto_register(const struct nf_nat_l3proto
*l3proto
)
648 err
= nf_ct_l3proto_try_module_get(l3proto
->l3proto
);
652 mutex_lock(&nf_nat_proto_mutex
);
653 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
->l3proto
][IPPROTO_TCP
],
654 &nf_nat_l4proto_tcp
);
655 RCU_INIT_POINTER(nf_nat_l4protos
[l3proto
->l3proto
][IPPROTO_UDP
],
656 &nf_nat_l4proto_udp
);
657 mutex_unlock(&nf_nat_proto_mutex
);
659 RCU_INIT_POINTER(nf_nat_l3protos
[l3proto
->l3proto
], l3proto
);
662 EXPORT_SYMBOL_GPL(nf_nat_l3proto_register
);
664 void nf_nat_l3proto_unregister(const struct nf_nat_l3proto
*l3proto
)
666 mutex_lock(&nf_nat_proto_mutex
);
667 RCU_INIT_POINTER(nf_nat_l3protos
[l3proto
->l3proto
], NULL
);
668 mutex_unlock(&nf_nat_proto_mutex
);
671 nf_nat_l3proto_clean(l3proto
->l3proto
);
672 nf_ct_l3proto_module_put(l3proto
->l3proto
);
674 EXPORT_SYMBOL_GPL(nf_nat_l3proto_unregister
);
676 /* No one using conntrack by the time this called. */
677 static void nf_nat_cleanup_conntrack(struct nf_conn
*ct
)
679 struct nf_conn_nat
*nat
= nf_ct_ext_find(ct
, NF_CT_EXT_NAT
);
681 if (nat
== NULL
|| nat
->ct
== NULL
)
684 NF_CT_ASSERT(nat
->ct
->status
& IPS_SRC_NAT_DONE
);
686 spin_lock_bh(&nf_nat_lock
);
687 hlist_del_rcu(&nat
->bysource
);
688 spin_unlock_bh(&nf_nat_lock
);
691 static void nf_nat_move_storage(void *new, void *old
)
693 struct nf_conn_nat
*new_nat
= new;
694 struct nf_conn_nat
*old_nat
= old
;
695 struct nf_conn
*ct
= old_nat
->ct
;
697 if (!ct
|| !(ct
->status
& IPS_SRC_NAT_DONE
))
700 spin_lock_bh(&nf_nat_lock
);
701 hlist_replace_rcu(&old_nat
->bysource
, &new_nat
->bysource
);
702 spin_unlock_bh(&nf_nat_lock
);
705 static struct nf_ct_ext_type nat_extend __read_mostly
= {
706 .len
= sizeof(struct nf_conn_nat
),
707 .align
= __alignof__(struct nf_conn_nat
),
708 .destroy
= nf_nat_cleanup_conntrack
,
709 .move
= nf_nat_move_storage
,
711 .flags
= NF_CT_EXT_F_PREALLOC
,
714 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
716 #include <linux/netfilter/nfnetlink.h>
717 #include <linux/netfilter/nfnetlink_conntrack.h>
719 static const struct nla_policy protonat_nla_policy
[CTA_PROTONAT_MAX
+1] = {
720 [CTA_PROTONAT_PORT_MIN
] = { .type
= NLA_U16
},
721 [CTA_PROTONAT_PORT_MAX
] = { .type
= NLA_U16
},
724 static int nfnetlink_parse_nat_proto(struct nlattr
*attr
,
725 const struct nf_conn
*ct
,
726 struct nf_nat_range
*range
)
728 struct nlattr
*tb
[CTA_PROTONAT_MAX
+1];
729 const struct nf_nat_l4proto
*l4proto
;
732 err
= nla_parse_nested(tb
, CTA_PROTONAT_MAX
, attr
, protonat_nla_policy
);
736 l4proto
= __nf_nat_l4proto_find(nf_ct_l3num(ct
), nf_ct_protonum(ct
));
737 if (l4proto
->nlattr_to_range
)
738 err
= l4proto
->nlattr_to_range(tb
, range
);
743 static const struct nla_policy nat_nla_policy
[CTA_NAT_MAX
+1] = {
744 [CTA_NAT_V4_MINIP
] = { .type
= NLA_U32
},
745 [CTA_NAT_V4_MAXIP
] = { .type
= NLA_U32
},
746 [CTA_NAT_V6_MINIP
] = { .len
= sizeof(struct in6_addr
) },
747 [CTA_NAT_V6_MAXIP
] = { .len
= sizeof(struct in6_addr
) },
748 [CTA_NAT_PROTO
] = { .type
= NLA_NESTED
},
752 nfnetlink_parse_nat(const struct nlattr
*nat
,
753 const struct nf_conn
*ct
, struct nf_nat_range
*range
,
754 const struct nf_nat_l3proto
*l3proto
)
756 struct nlattr
*tb
[CTA_NAT_MAX
+1];
759 memset(range
, 0, sizeof(*range
));
761 err
= nla_parse_nested(tb
, CTA_NAT_MAX
, nat
, nat_nla_policy
);
765 err
= l3proto
->nlattr_to_range(tb
, range
);
769 if (!tb
[CTA_NAT_PROTO
])
772 return nfnetlink_parse_nat_proto(tb
[CTA_NAT_PROTO
], ct
, range
);
775 /* This function is called under rcu_read_lock() */
777 nfnetlink_parse_nat_setup(struct nf_conn
*ct
,
778 enum nf_nat_manip_type manip
,
779 const struct nlattr
*attr
)
781 struct nf_nat_range range
;
782 const struct nf_nat_l3proto
*l3proto
;
785 /* Should not happen, restricted to creating new conntracks
788 if (WARN_ON_ONCE(nf_nat_initialized(ct
, manip
)))
791 /* Make sure that L3 NAT is there by when we call nf_nat_setup_info to
792 * attach the null binding, otherwise this may oops.
794 l3proto
= __nf_nat_l3proto_find(nf_ct_l3num(ct
));
798 /* No NAT information has been passed, allocate the null-binding */
800 return __nf_nat_alloc_null_binding(ct
, manip
);
802 err
= nfnetlink_parse_nat(attr
, ct
, &range
, l3proto
);
806 return nf_nat_setup_info(ct
, &range
, manip
);
810 nfnetlink_parse_nat_setup(struct nf_conn
*ct
,
811 enum nf_nat_manip_type manip
,
812 const struct nlattr
*attr
)
818 static int __net_init
nf_nat_net_init(struct net
*net
)
820 /* Leave them the same for the moment. */
821 net
->ct
.nat_htable_size
= net
->ct
.htable_size
;
822 net
->ct
.nat_bysource
= nf_ct_alloc_hashtable(&net
->ct
.nat_htable_size
, 0);
823 if (!net
->ct
.nat_bysource
)
828 static void __net_exit
nf_nat_net_exit(struct net
*net
)
830 struct nf_nat_proto_clean clean
= {};
832 nf_ct_iterate_cleanup(net
, nf_nat_proto_clean
, &clean
, 0, 0);
834 nf_ct_free_hashtable(net
->ct
.nat_bysource
, net
->ct
.nat_htable_size
);
837 static struct pernet_operations nf_nat_net_ops
= {
838 .init
= nf_nat_net_init
,
839 .exit
= nf_nat_net_exit
,
842 static struct nf_ct_helper_expectfn follow_master_nat
= {
843 .name
= "nat-follow-master",
844 .expectfn
= nf_nat_follow_master
,
847 static int __init
nf_nat_init(void)
851 ret
= nf_ct_extend_register(&nat_extend
);
853 printk(KERN_ERR
"nf_nat_core: Unable to register extension\n");
857 ret
= register_pernet_subsys(&nf_nat_net_ops
);
861 nf_ct_helper_expectfn_register(&follow_master_nat
);
863 /* Initialize fake conntrack so that NAT will skip it */
864 nf_ct_untracked_status_or(IPS_NAT_DONE_MASK
);
866 BUG_ON(nfnetlink_parse_nat_setup_hook
!= NULL
);
867 RCU_INIT_POINTER(nfnetlink_parse_nat_setup_hook
,
868 nfnetlink_parse_nat_setup
);
870 BUG_ON(nf_nat_decode_session_hook
!= NULL
);
871 RCU_INIT_POINTER(nf_nat_decode_session_hook
, __nf_nat_decode_session
);
876 nf_ct_extend_unregister(&nat_extend
);
880 static void __exit
nf_nat_cleanup(void)
884 unregister_pernet_subsys(&nf_nat_net_ops
);
885 nf_ct_extend_unregister(&nat_extend
);
886 nf_ct_helper_expectfn_unregister(&follow_master_nat
);
887 RCU_INIT_POINTER(nfnetlink_parse_nat_setup_hook
, NULL
);
889 RCU_INIT_POINTER(nf_nat_decode_session_hook
, NULL
);
891 for (i
= 0; i
< NFPROTO_NUMPROTO
; i
++)
892 kfree(nf_nat_l4protos
[i
]);
896 MODULE_LICENSE("GPL");
898 module_init(nf_nat_init
);
899 module_exit(nf_nat_cleanup
);