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
)
44 enum ip_conntrack_dir dir
= CTINFO2DIR(ctinfo
);
45 struct nf_conn_nat
*nat
= nfct_nat(ct
);
46 struct nf_nat_seq
*this_way
= &nat
->seq
[dir
];
48 pr_debug("adjust_tcp_sequence: seq = %u, sizediff = %d\n",
51 pr_debug("adjust_tcp_sequence: Seq_offset before: ");
52 DUMP_OFFSET(this_way
);
54 spin_lock_bh(&nf_nat_seqofs_lock
);
56 /* SYN adjust. If it's uninitialized, or this is after last
57 * correction, record it: we don't handle more than one
58 * adjustment in the window, but do deal with common case of a
60 if (this_way
->offset_before
== this_way
->offset_after
||
61 before(this_way
->correction_pos
, seq
)) {
62 this_way
->correction_pos
= seq
;
63 this_way
->offset_before
= this_way
->offset_after
;
64 this_way
->offset_after
+= sizediff
;
66 spin_unlock_bh(&nf_nat_seqofs_lock
);
68 pr_debug("adjust_tcp_sequence: Seq_offset after: ");
69 DUMP_OFFSET(this_way
);
72 /* Get the offset value, for conntrack */
73 s16
nf_nat_get_offset(const struct nf_conn
*ct
,
74 enum ip_conntrack_dir dir
,
77 struct nf_conn_nat
*nat
= nfct_nat(ct
);
78 struct nf_nat_seq
*this_way
;
84 this_way
= &nat
->seq
[dir
];
85 spin_lock_bh(&nf_nat_seqofs_lock
);
86 offset
= after(seq
, this_way
->correction_pos
)
87 ? this_way
->offset_after
: this_way
->offset_before
;
88 spin_unlock_bh(&nf_nat_seqofs_lock
);
92 EXPORT_SYMBOL_GPL(nf_nat_get_offset
);
94 /* Frobs data inside this packet, which is linear. */
95 static void mangle_contents(struct sk_buff
*skb
,
97 unsigned int match_offset
,
98 unsigned int match_len
,
99 const char *rep_buffer
,
100 unsigned int rep_len
)
104 BUG_ON(skb_is_nonlinear(skb
));
105 data
= skb_network_header(skb
) + dataoff
;
107 /* move post-replacement */
108 memmove(data
+ match_offset
+ rep_len
,
109 data
+ match_offset
+ match_len
,
110 skb
->tail
- (skb
->network_header
+ dataoff
+
111 match_offset
+ match_len
));
113 /* insert data from buffer */
114 memcpy(data
+ match_offset
, rep_buffer
, rep_len
);
116 /* update skb info */
117 if (rep_len
> match_len
) {
118 pr_debug("nf_nat_mangle_packet: Extending packet by "
119 "%u from %u bytes\n", rep_len
- match_len
, skb
->len
);
120 skb_put(skb
, rep_len
- match_len
);
122 pr_debug("nf_nat_mangle_packet: Shrinking packet from "
123 "%u from %u bytes\n", match_len
- rep_len
, skb
->len
);
124 __skb_trim(skb
, skb
->len
+ rep_len
- match_len
);
127 /* fix IP hdr checksum information */
128 ip_hdr(skb
)->tot_len
= htons(skb
->len
);
129 ip_send_check(ip_hdr(skb
));
132 /* Unusual, but possible case. */
133 static int enlarge_skb(struct sk_buff
*skb
, unsigned int extra
)
135 if (skb
->len
+ extra
> 65535)
138 if (pskb_expand_head(skb
, 0, extra
- skb_tailroom(skb
), GFP_ATOMIC
))
144 /* Generic function for mangling variable-length address changes inside
145 * NATed TCP connections (like the PORT XXX,XXX,XXX,XXX,XXX,XXX
148 * Takes care about all the nasty sequence number changes, checksumming,
149 * skb enlargement, ...
153 nf_nat_mangle_tcp_packet(struct sk_buff
*skb
,
155 enum ip_conntrack_info ctinfo
,
156 unsigned int match_offset
,
157 unsigned int match_len
,
158 const char *rep_buffer
,
159 unsigned int rep_len
)
161 struct rtable
*rt
= skb_rtable(skb
);
166 if (!skb_make_writable(skb
, skb
->len
))
169 if (rep_len
> match_len
&&
170 rep_len
- match_len
> skb_tailroom(skb
) &&
171 !enlarge_skb(skb
, rep_len
- match_len
))
174 SKB_LINEAR_ASSERT(skb
);
177 tcph
= (void *)iph
+ iph
->ihl
*4;
179 oldlen
= skb
->len
- iph
->ihl
*4;
180 mangle_contents(skb
, iph
->ihl
*4 + tcph
->doff
*4,
181 match_offset
, match_len
, rep_buffer
, rep_len
);
183 datalen
= skb
->len
- iph
->ihl
*4;
184 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
) {
185 if (!(rt
->rt_flags
& RTCF_LOCAL
) &&
186 skb
->dev
->features
& NETIF_F_V4_CSUM
) {
187 skb
->ip_summed
= CHECKSUM_PARTIAL
;
188 skb
->csum_start
= skb_headroom(skb
) +
189 skb_network_offset(skb
) +
191 skb
->csum_offset
= offsetof(struct tcphdr
, check
);
192 tcph
->check
= ~tcp_v4_check(datalen
,
193 iph
->saddr
, iph
->daddr
, 0);
196 tcph
->check
= tcp_v4_check(datalen
,
197 iph
->saddr
, iph
->daddr
,
202 inet_proto_csum_replace2(&tcph
->check
, skb
,
203 htons(oldlen
), htons(datalen
), 1);
205 if (rep_len
!= match_len
) {
206 set_bit(IPS_SEQ_ADJUST_BIT
, &ct
->status
);
207 adjust_tcp_sequence(ntohl(tcph
->seq
),
208 (int)rep_len
- (int)match_len
,
210 nf_conntrack_event_cache(IPCT_NATSEQADJ
, ct
);
214 EXPORT_SYMBOL(nf_nat_mangle_tcp_packet
);
216 /* Generic function for mangling variable-length address changes inside
217 * NATed UDP connections (like the CONNECT DATA XXXXX MESG XXXXX INDEX XXXXX
218 * command in the Amanda protocol)
220 * Takes care about all the nasty sequence number changes, checksumming,
221 * skb enlargement, ...
223 * XXX - This function could be merged with nf_nat_mangle_tcp_packet which
224 * should be fairly easy to do.
227 nf_nat_mangle_udp_packet(struct sk_buff
*skb
,
229 enum ip_conntrack_info ctinfo
,
230 unsigned int match_offset
,
231 unsigned int match_len
,
232 const char *rep_buffer
,
233 unsigned int rep_len
)
235 struct rtable
*rt
= skb_rtable(skb
);
240 /* UDP helpers might accidentally mangle the wrong packet */
242 if (skb
->len
< iph
->ihl
*4 + sizeof(*udph
) +
243 match_offset
+ match_len
)
246 if (!skb_make_writable(skb
, skb
->len
))
249 if (rep_len
> match_len
&&
250 rep_len
- match_len
> skb_tailroom(skb
) &&
251 !enlarge_skb(skb
, rep_len
- match_len
))
255 udph
= (void *)iph
+ iph
->ihl
*4;
257 oldlen
= skb
->len
- iph
->ihl
*4;
258 mangle_contents(skb
, iph
->ihl
*4 + sizeof(*udph
),
259 match_offset
, match_len
, rep_buffer
, rep_len
);
261 /* update the length of the UDP packet */
262 datalen
= skb
->len
- iph
->ihl
*4;
263 udph
->len
= htons(datalen
);
265 /* fix udp checksum if udp checksum was previously calculated */
266 if (!udph
->check
&& skb
->ip_summed
!= CHECKSUM_PARTIAL
)
269 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
) {
270 if (!(rt
->rt_flags
& RTCF_LOCAL
) &&
271 skb
->dev
->features
& NETIF_F_V4_CSUM
) {
272 skb
->ip_summed
= CHECKSUM_PARTIAL
;
273 skb
->csum_start
= skb_headroom(skb
) +
274 skb_network_offset(skb
) +
276 skb
->csum_offset
= offsetof(struct udphdr
, check
);
277 udph
->check
= ~csum_tcpudp_magic(iph
->saddr
, iph
->daddr
,
278 datalen
, IPPROTO_UDP
,
282 udph
->check
= csum_tcpudp_magic(iph
->saddr
, iph
->daddr
,
283 datalen
, IPPROTO_UDP
,
287 udph
->check
= CSUM_MANGLED_0
;
290 inet_proto_csum_replace2(&udph
->check
, skb
,
291 htons(oldlen
), htons(datalen
), 1);
295 EXPORT_SYMBOL(nf_nat_mangle_udp_packet
);
297 /* Adjust one found SACK option including checksum correction */
299 sack_adjust(struct sk_buff
*skb
,
301 unsigned int sackoff
,
302 unsigned int sackend
,
303 struct nf_nat_seq
*natseq
)
305 while (sackoff
< sackend
) {
306 struct tcp_sack_block_wire
*sack
;
307 __be32 new_start_seq
, new_end_seq
;
309 sack
= (void *)skb
->data
+ sackoff
;
310 if (after(ntohl(sack
->start_seq
) - natseq
->offset_before
,
311 natseq
->correction_pos
))
312 new_start_seq
= htonl(ntohl(sack
->start_seq
)
313 - natseq
->offset_after
);
315 new_start_seq
= htonl(ntohl(sack
->start_seq
)
316 - natseq
->offset_before
);
318 if (after(ntohl(sack
->end_seq
) - natseq
->offset_before
,
319 natseq
->correction_pos
))
320 new_end_seq
= htonl(ntohl(sack
->end_seq
)
321 - natseq
->offset_after
);
323 new_end_seq
= htonl(ntohl(sack
->end_seq
)
324 - natseq
->offset_before
);
326 pr_debug("sack_adjust: start_seq: %d->%d, end_seq: %d->%d\n",
327 ntohl(sack
->start_seq
), new_start_seq
,
328 ntohl(sack
->end_seq
), new_end_seq
);
330 inet_proto_csum_replace4(&tcph
->check
, skb
,
331 sack
->start_seq
, new_start_seq
, 0);
332 inet_proto_csum_replace4(&tcph
->check
, skb
,
333 sack
->end_seq
, new_end_seq
, 0);
334 sack
->start_seq
= new_start_seq
;
335 sack
->end_seq
= new_end_seq
;
336 sackoff
+= sizeof(*sack
);
340 /* TCP SACK sequence number adjustment */
341 static inline unsigned int
342 nf_nat_sack_adjust(struct sk_buff
*skb
,
345 enum ip_conntrack_info ctinfo
)
347 unsigned int dir
, optoff
, optend
;
348 struct nf_conn_nat
*nat
= nfct_nat(ct
);
350 optoff
= ip_hdrlen(skb
) + sizeof(struct tcphdr
);
351 optend
= ip_hdrlen(skb
) + tcph
->doff
* 4;
353 if (!skb_make_writable(skb
, optend
))
356 dir
= CTINFO2DIR(ctinfo
);
358 while (optoff
< optend
) {
359 /* Usually: option, length. */
360 unsigned char *op
= skb
->data
+ optoff
;
369 /* no partial options */
370 if (optoff
+ 1 == optend
||
371 optoff
+ op
[1] > optend
||
374 if (op
[0] == TCPOPT_SACK
&&
375 op
[1] >= 2+TCPOLEN_SACK_PERBLOCK
&&
376 ((op
[1] - 2) % TCPOLEN_SACK_PERBLOCK
) == 0)
377 sack_adjust(skb
, tcph
, optoff
+2,
378 optoff
+op
[1], &nat
->seq
[!dir
]);
385 /* TCP sequence number adjustment. Returns 1 on success, 0 on failure */
387 nf_nat_seq_adjust(struct sk_buff
*skb
,
389 enum ip_conntrack_info ctinfo
)
393 __be32 newseq
, newack
;
395 struct nf_conn_nat
*nat
= nfct_nat(ct
);
396 struct nf_nat_seq
*this_way
, *other_way
;
398 dir
= CTINFO2DIR(ctinfo
);
400 this_way
= &nat
->seq
[dir
];
401 other_way
= &nat
->seq
[!dir
];
403 if (!skb_make_writable(skb
, ip_hdrlen(skb
) + sizeof(*tcph
)))
406 tcph
= (void *)skb
->data
+ ip_hdrlen(skb
);
407 if (after(ntohl(tcph
->seq
), this_way
->correction_pos
))
408 seqoff
= this_way
->offset_after
;
410 seqoff
= this_way
->offset_before
;
412 if (after(ntohl(tcph
->ack_seq
) - other_way
->offset_before
,
413 other_way
->correction_pos
))
414 ackoff
= other_way
->offset_after
;
416 ackoff
= other_way
->offset_before
;
418 newseq
= htonl(ntohl(tcph
->seq
) + seqoff
);
419 newack
= htonl(ntohl(tcph
->ack_seq
) - ackoff
);
421 inet_proto_csum_replace4(&tcph
->check
, skb
, tcph
->seq
, newseq
, 0);
422 inet_proto_csum_replace4(&tcph
->check
, skb
, tcph
->ack_seq
, newack
, 0);
424 pr_debug("Adjusting sequence number from %u->%u, ack from %u->%u\n",
425 ntohl(tcph
->seq
), ntohl(newseq
), ntohl(tcph
->ack_seq
),
429 tcph
->ack_seq
= newack
;
431 return nf_nat_sack_adjust(skb
, tcph
, ct
, ctinfo
);
434 /* Setup NAT on this expected conntrack so it follows master. */
435 /* If we fail to get a free NAT slot, we'll get dropped on confirm */
436 void nf_nat_follow_master(struct nf_conn
*ct
,
437 struct nf_conntrack_expect
*exp
)
439 struct nf_nat_range range
;
441 /* This must be a fresh one. */
442 BUG_ON(ct
->status
& IPS_NAT_DONE_MASK
);
444 /* Change src to where master sends to */
445 range
.flags
= IP_NAT_RANGE_MAP_IPS
;
446 range
.min_ip
= range
.max_ip
447 = ct
->master
->tuplehash
[!exp
->dir
].tuple
.dst
.u3
.ip
;
448 nf_nat_setup_info(ct
, &range
, IP_NAT_MANIP_SRC
);
450 /* For DST manip, map port here to where it's expected. */
451 range
.flags
= (IP_NAT_RANGE_MAP_IPS
| IP_NAT_RANGE_PROTO_SPECIFIED
);
452 range
.min
= range
.max
= exp
->saved_proto
;
453 range
.min_ip
= range
.max_ip
454 = ct
->master
->tuplehash
[!exp
->dir
].tuple
.src
.u3
.ip
;
455 nf_nat_setup_info(ct
, &range
, IP_NAT_MANIP_DST
);
457 EXPORT_SYMBOL(nf_nat_follow_master
);