1 /* ip_nat_helper.c - generic support functions for NAT helpers
3 * (C) 2000-2002 Harald Welte <laforge@netfilter.org>
4 * (C) 2003-2006 Netfilter Core Team <coreteam@netfilter.org>
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.
10 #include <linux/module.h>
11 #include <linux/kmod.h>
12 #include <linux/types.h>
13 #include <linux/timer.h>
14 #include <linux/skbuff.h>
15 #include <linux/tcp.h>
16 #include <linux/udp.h>
17 #include <net/checksum.h>
19 #include <net/route.h>
21 #include <linux/netfilter_ipv4.h>
22 #include <net/netfilter/nf_conntrack.h>
23 #include <net/netfilter/nf_conntrack_helper.h>
24 #include <net/netfilter/nf_conntrack_ecache.h>
25 #include <net/netfilter/nf_conntrack_expect.h>
26 #include <net/netfilter/nf_nat.h>
27 #include <net/netfilter/nf_nat_protocol.h>
28 #include <net/netfilter/nf_nat_core.h>
29 #include <net/netfilter/nf_nat_helper.h>
31 #define DUMP_OFFSET(x) \
32 pr_debug("offset_before=%d, offset_after=%d, correction_pos=%u\n", \
33 x->offset_before, x->offset_after, x->correction_pos);
35 static DEFINE_SPINLOCK(nf_nat_seqofs_lock
);
37 /* Setup TCP sequence correction given this change at this sequence */
39 adjust_tcp_sequence(u32 seq
,
42 enum ip_conntrack_info ctinfo
)
45 struct nf_nat_seq
*this_way
, *other_way
;
46 struct nf_conn_nat
*nat
= nfct_nat(ct
);
48 pr_debug("adjust_tcp_sequence: seq = %u, sizediff = %d\n", seq
, seq
);
50 dir
= CTINFO2DIR(ctinfo
);
52 this_way
= &nat
->seq
[dir
];
53 other_way
= &nat
->seq
[!dir
];
55 pr_debug("nf_nat_resize_packet: Seq_offset before: ");
56 DUMP_OFFSET(this_way
);
58 spin_lock_bh(&nf_nat_seqofs_lock
);
60 /* SYN adjust. If it's uninitialized, or this is after last
61 * correction, record it: we don't handle more than one
62 * adjustment in the window, but do deal with common case of a
64 if (this_way
->offset_before
== this_way
->offset_after
||
65 before(this_way
->correction_pos
, seq
)) {
66 this_way
->correction_pos
= seq
;
67 this_way
->offset_before
= this_way
->offset_after
;
68 this_way
->offset_after
+= sizediff
;
70 spin_unlock_bh(&nf_nat_seqofs_lock
);
72 pr_debug("nf_nat_resize_packet: Seq_offset after: ");
73 DUMP_OFFSET(this_way
);
76 /* Frobs data inside this packet, which is linear. */
77 static void mangle_contents(struct sk_buff
*skb
,
79 unsigned int match_offset
,
80 unsigned int match_len
,
81 const char *rep_buffer
,
86 BUG_ON(skb_is_nonlinear(skb
));
87 data
= skb_network_header(skb
) + dataoff
;
89 /* move post-replacement */
90 memmove(data
+ match_offset
+ rep_len
,
91 data
+ match_offset
+ match_len
,
92 skb
->tail
- (skb
->network_header
+ dataoff
+
93 match_offset
+ match_len
));
95 /* insert data from buffer */
96 memcpy(data
+ match_offset
, rep_buffer
, rep_len
);
99 if (rep_len
> match_len
) {
100 pr_debug("nf_nat_mangle_packet: Extending packet by "
101 "%u from %u bytes\n", rep_len
- match_len
, skb
->len
);
102 skb_put(skb
, rep_len
- match_len
);
104 pr_debug("nf_nat_mangle_packet: Shrinking packet from "
105 "%u from %u bytes\n", match_len
- rep_len
, skb
->len
);
106 __skb_trim(skb
, skb
->len
+ rep_len
- match_len
);
109 /* fix IP hdr checksum information */
110 ip_hdr(skb
)->tot_len
= htons(skb
->len
);
111 ip_send_check(ip_hdr(skb
));
114 /* Unusual, but possible case. */
115 static int enlarge_skb(struct sk_buff
*skb
, unsigned int extra
)
117 if (skb
->len
+ extra
> 65535)
120 if (pskb_expand_head(skb
, 0, extra
- skb_tailroom(skb
), GFP_ATOMIC
))
126 /* Generic function for mangling variable-length address changes inside
127 * NATed TCP connections (like the PORT XXX,XXX,XXX,XXX,XXX,XXX
130 * Takes care about all the nasty sequence number changes, checksumming,
131 * skb enlargement, ...
135 nf_nat_mangle_tcp_packet(struct sk_buff
*skb
,
137 enum ip_conntrack_info ctinfo
,
138 unsigned int match_offset
,
139 unsigned int match_len
,
140 const char *rep_buffer
,
141 unsigned int rep_len
)
143 struct rtable
*rt
= skb
->rtable
;
148 if (!skb_make_writable(skb
, skb
->len
))
151 if (rep_len
> match_len
&&
152 rep_len
- match_len
> skb_tailroom(skb
) &&
153 !enlarge_skb(skb
, rep_len
- match_len
))
156 SKB_LINEAR_ASSERT(skb
);
159 tcph
= (void *)iph
+ iph
->ihl
*4;
161 oldlen
= skb
->len
- iph
->ihl
*4;
162 mangle_contents(skb
, iph
->ihl
*4 + tcph
->doff
*4,
163 match_offset
, match_len
, rep_buffer
, rep_len
);
165 datalen
= skb
->len
- iph
->ihl
*4;
166 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
) {
167 if (!(rt
->rt_flags
& RTCF_LOCAL
) &&
168 skb
->dev
->features
& NETIF_F_V4_CSUM
) {
169 skb
->ip_summed
= CHECKSUM_PARTIAL
;
170 skb
->csum_start
= skb_headroom(skb
) +
171 skb_network_offset(skb
) +
173 skb
->csum_offset
= offsetof(struct tcphdr
, check
);
174 tcph
->check
= ~tcp_v4_check(datalen
,
175 iph
->saddr
, iph
->daddr
, 0);
178 tcph
->check
= tcp_v4_check(datalen
,
179 iph
->saddr
, iph
->daddr
,
184 inet_proto_csum_replace2(&tcph
->check
, skb
,
185 htons(oldlen
), htons(datalen
), 1);
187 if (rep_len
!= match_len
) {
188 set_bit(IPS_SEQ_ADJUST_BIT
, &ct
->status
);
189 adjust_tcp_sequence(ntohl(tcph
->seq
),
190 (int)rep_len
- (int)match_len
,
192 /* Tell TCP window tracking about seq change */
193 nf_conntrack_tcp_update(skb
, ip_hdrlen(skb
),
194 ct
, CTINFO2DIR(ctinfo
));
196 nf_conntrack_event_cache(IPCT_NATSEQADJ
, ct
);
200 EXPORT_SYMBOL(nf_nat_mangle_tcp_packet
);
202 /* Generic function for mangling variable-length address changes inside
203 * NATed UDP connections (like the CONNECT DATA XXXXX MESG XXXXX INDEX XXXXX
204 * command in the Amanda protocol)
206 * Takes care about all the nasty sequence number changes, checksumming,
207 * skb enlargement, ...
209 * XXX - This function could be merged with nf_nat_mangle_tcp_packet which
210 * should be fairly easy to do.
213 nf_nat_mangle_udp_packet(struct sk_buff
*skb
,
215 enum ip_conntrack_info ctinfo
,
216 unsigned int match_offset
,
217 unsigned int match_len
,
218 const char *rep_buffer
,
219 unsigned int rep_len
)
221 struct rtable
*rt
= skb
->rtable
;
226 /* UDP helpers might accidentally mangle the wrong packet */
228 if (skb
->len
< iph
->ihl
*4 + sizeof(*udph
) +
229 match_offset
+ match_len
)
232 if (!skb_make_writable(skb
, skb
->len
))
235 if (rep_len
> match_len
&&
236 rep_len
- match_len
> skb_tailroom(skb
) &&
237 !enlarge_skb(skb
, rep_len
- match_len
))
241 udph
= (void *)iph
+ iph
->ihl
*4;
243 oldlen
= skb
->len
- iph
->ihl
*4;
244 mangle_contents(skb
, iph
->ihl
*4 + sizeof(*udph
),
245 match_offset
, match_len
, rep_buffer
, rep_len
);
247 /* update the length of the UDP packet */
248 datalen
= skb
->len
- iph
->ihl
*4;
249 udph
->len
= htons(datalen
);
251 /* fix udp checksum if udp checksum was previously calculated */
252 if (!udph
->check
&& skb
->ip_summed
!= CHECKSUM_PARTIAL
)
255 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
) {
256 if (!(rt
->rt_flags
& RTCF_LOCAL
) &&
257 skb
->dev
->features
& NETIF_F_V4_CSUM
) {
258 skb
->ip_summed
= CHECKSUM_PARTIAL
;
259 skb
->csum_start
= skb_headroom(skb
) +
260 skb_network_offset(skb
) +
262 skb
->csum_offset
= offsetof(struct udphdr
, check
);
263 udph
->check
= ~csum_tcpudp_magic(iph
->saddr
, iph
->daddr
,
264 datalen
, IPPROTO_UDP
,
268 udph
->check
= csum_tcpudp_magic(iph
->saddr
, iph
->daddr
,
269 datalen
, IPPROTO_UDP
,
273 udph
->check
= CSUM_MANGLED_0
;
276 inet_proto_csum_replace2(&udph
->check
, skb
,
277 htons(oldlen
), htons(datalen
), 1);
281 EXPORT_SYMBOL(nf_nat_mangle_udp_packet
);
283 /* Adjust one found SACK option including checksum correction */
285 sack_adjust(struct sk_buff
*skb
,
287 unsigned int sackoff
,
288 unsigned int sackend
,
289 struct nf_nat_seq
*natseq
)
291 while (sackoff
< sackend
) {
292 struct tcp_sack_block_wire
*sack
;
293 __be32 new_start_seq
, new_end_seq
;
295 sack
= (void *)skb
->data
+ sackoff
;
296 if (after(ntohl(sack
->start_seq
) - natseq
->offset_before
,
297 natseq
->correction_pos
))
298 new_start_seq
= htonl(ntohl(sack
->start_seq
)
299 - natseq
->offset_after
);
301 new_start_seq
= htonl(ntohl(sack
->start_seq
)
302 - natseq
->offset_before
);
304 if (after(ntohl(sack
->end_seq
) - natseq
->offset_before
,
305 natseq
->correction_pos
))
306 new_end_seq
= htonl(ntohl(sack
->end_seq
)
307 - natseq
->offset_after
);
309 new_end_seq
= htonl(ntohl(sack
->end_seq
)
310 - natseq
->offset_before
);
312 pr_debug("sack_adjust: start_seq: %d->%d, end_seq: %d->%d\n",
313 ntohl(sack
->start_seq
), new_start_seq
,
314 ntohl(sack
->end_seq
), new_end_seq
);
316 inet_proto_csum_replace4(&tcph
->check
, skb
,
317 sack
->start_seq
, new_start_seq
, 0);
318 inet_proto_csum_replace4(&tcph
->check
, skb
,
319 sack
->end_seq
, new_end_seq
, 0);
320 sack
->start_seq
= new_start_seq
;
321 sack
->end_seq
= new_end_seq
;
322 sackoff
+= sizeof(*sack
);
326 /* TCP SACK sequence number adjustment */
327 static inline unsigned int
328 nf_nat_sack_adjust(struct sk_buff
*skb
,
331 enum ip_conntrack_info ctinfo
)
333 unsigned int dir
, optoff
, optend
;
334 struct nf_conn_nat
*nat
= nfct_nat(ct
);
336 optoff
= ip_hdrlen(skb
) + sizeof(struct tcphdr
);
337 optend
= ip_hdrlen(skb
) + tcph
->doff
* 4;
339 if (!skb_make_writable(skb
, optend
))
342 dir
= CTINFO2DIR(ctinfo
);
344 while (optoff
< optend
) {
345 /* Usually: option, length. */
346 unsigned char *op
= skb
->data
+ optoff
;
355 /* no partial options */
356 if (optoff
+ 1 == optend
||
357 optoff
+ op
[1] > optend
||
360 if (op
[0] == TCPOPT_SACK
&&
361 op
[1] >= 2+TCPOLEN_SACK_PERBLOCK
&&
362 ((op
[1] - 2) % TCPOLEN_SACK_PERBLOCK
) == 0)
363 sack_adjust(skb
, tcph
, optoff
+2,
364 optoff
+op
[1], &nat
->seq
[!dir
]);
371 /* TCP sequence number adjustment. Returns 1 on success, 0 on failure */
373 nf_nat_seq_adjust(struct sk_buff
*skb
,
375 enum ip_conntrack_info ctinfo
)
379 __be32 newseq
, newack
;
380 struct nf_conn_nat
*nat
= nfct_nat(ct
);
381 struct nf_nat_seq
*this_way
, *other_way
;
383 dir
= CTINFO2DIR(ctinfo
);
385 this_way
= &nat
->seq
[dir
];
386 other_way
= &nat
->seq
[!dir
];
388 if (!skb_make_writable(skb
, ip_hdrlen(skb
) + sizeof(*tcph
)))
391 tcph
= (void *)skb
->data
+ ip_hdrlen(skb
);
392 if (after(ntohl(tcph
->seq
), this_way
->correction_pos
))
393 newseq
= htonl(ntohl(tcph
->seq
) + this_way
->offset_after
);
395 newseq
= htonl(ntohl(tcph
->seq
) + this_way
->offset_before
);
397 if (after(ntohl(tcph
->ack_seq
) - other_way
->offset_before
,
398 other_way
->correction_pos
))
399 newack
= htonl(ntohl(tcph
->ack_seq
) - other_way
->offset_after
);
401 newack
= htonl(ntohl(tcph
->ack_seq
) - other_way
->offset_before
);
403 inet_proto_csum_replace4(&tcph
->check
, skb
, tcph
->seq
, newseq
, 0);
404 inet_proto_csum_replace4(&tcph
->check
, skb
, tcph
->ack_seq
, newack
, 0);
406 pr_debug("Adjusting sequence number from %u->%u, ack from %u->%u\n",
407 ntohl(tcph
->seq
), ntohl(newseq
), ntohl(tcph
->ack_seq
),
411 tcph
->ack_seq
= newack
;
413 if (!nf_nat_sack_adjust(skb
, tcph
, ct
, ctinfo
))
416 nf_conntrack_tcp_update(skb
, ip_hdrlen(skb
), ct
, dir
);
421 /* Setup NAT on this expected conntrack so it follows master. */
422 /* If we fail to get a free NAT slot, we'll get dropped on confirm */
423 void nf_nat_follow_master(struct nf_conn
*ct
,
424 struct nf_conntrack_expect
*exp
)
426 struct nf_nat_range range
;
428 /* This must be a fresh one. */
429 BUG_ON(ct
->status
& IPS_NAT_DONE_MASK
);
431 /* Change src to where master sends to */
432 range
.flags
= IP_NAT_RANGE_MAP_IPS
;
433 range
.min_ip
= range
.max_ip
434 = ct
->master
->tuplehash
[!exp
->dir
].tuple
.dst
.u3
.ip
;
435 nf_nat_setup_info(ct
, &range
, IP_NAT_MANIP_SRC
);
437 /* For DST manip, map port here to where it's expected. */
438 range
.flags
= (IP_NAT_RANGE_MAP_IPS
| IP_NAT_RANGE_PROTO_SPECIFIED
);
439 range
.min
= range
.max
= exp
->saved_proto
;
440 range
.min_ip
= range
.max_ip
441 = ct
->master
->tuplehash
[!exp
->dir
].tuple
.src
.u3
.ip
;
442 nf_nat_setup_info(ct
, &range
, IP_NAT_MANIP_DST
);
444 EXPORT_SYMBOL(nf_nat_follow_master
);