blob: 308d2abd7ee5815a74a0a2e03bcdb65979d1027b [file] [log] [blame]
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08001/* (C) 1999-2001 Paul `Rusty' Russell
2 * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
3 *
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License version 2 as
6 * published by the Free Software Foundation.
7 *
8 * Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>:
9 * - Real stateful connection tracking
10 * - Modified state transitions table
11 * - Window scaling support added
12 * - SACK support added
13 *
14 * Willy Tarreau:
15 * - State table bugfixes
16 * - More robust state changes
17 * - Tuning timer parameters
18 *
19 * 27 Oct 2004: Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp>
20 * - genelized Layer 3 protocol part.
21 *
22 * Derived from net/ipv4/netfilter/ip_conntrack_proto_tcp.c
23 *
24 * version 2.2
25 */
26
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -080027#include <linux/types.h>
28#include <linux/sched.h>
29#include <linux/timer.h>
30#include <linux/netfilter.h>
31#include <linux/module.h>
32#include <linux/in.h>
33#include <linux/tcp.h>
34#include <linux/spinlock.h>
35#include <linux/skbuff.h>
36#include <linux/ipv6.h>
37#include <net/ip6_checksum.h>
38
39#include <net/tcp.h>
40
41#include <linux/netfilter.h>
42#include <linux/netfilter_ipv4.h>
43#include <linux/netfilter_ipv6.h>
44#include <net/netfilter/nf_conntrack.h>
45#include <net/netfilter/nf_conntrack_protocol.h>
46
47#if 0
48#define DEBUGP printk
49#define DEBUGP_VARS
50#else
51#define DEBUGP(format, args...)
52#endif
53
54/* Protects conntrack->proto.tcp */
55static DEFINE_RWLOCK(tcp_lock);
56
57/* "Be conservative in what you do,
58 be liberal in what you accept from others."
59 If it's non-zero, we mark only out of window RST segments as INVALID. */
60int nf_ct_tcp_be_liberal = 0;
61
62/* When connection is picked up from the middle, how many packets are required
63 to pass in each direction when we assume we are in sync - if any side uses
64 window scaling, we lost the game.
65 If it is set to zero, we disable picking up already established
66 connections. */
67int nf_ct_tcp_loose = 3;
68
69/* Max number of the retransmitted packets without receiving an (acceptable)
70 ACK from the destination. If this number is reached, a shorter timer
71 will be started. */
72int nf_ct_tcp_max_retrans = 3;
73
74 /* FIXME: Examine ipfilter's timeouts and conntrack transitions more
75 closely. They're more complex. --RR */
76
77static const char *tcp_conntrack_names[] = {
78 "NONE",
79 "SYN_SENT",
80 "SYN_RECV",
81 "ESTABLISHED",
82 "FIN_WAIT",
83 "CLOSE_WAIT",
84 "LAST_ACK",
85 "TIME_WAIT",
86 "CLOSE",
87 "LISTEN"
88};
89
90#define SECS * HZ
91#define MINS * 60 SECS
92#define HOURS * 60 MINS
93#define DAYS * 24 HOURS
94
Patrick McHardybabbdb12006-01-09 17:48:09 -080095unsigned int nf_ct_tcp_timeout_syn_sent = 2 MINS;
96unsigned int nf_ct_tcp_timeout_syn_recv = 60 SECS;
97unsigned int nf_ct_tcp_timeout_established = 5 DAYS;
98unsigned int nf_ct_tcp_timeout_fin_wait = 2 MINS;
99unsigned int nf_ct_tcp_timeout_close_wait = 60 SECS;
100unsigned int nf_ct_tcp_timeout_last_ack = 30 SECS;
101unsigned int nf_ct_tcp_timeout_time_wait = 2 MINS;
102unsigned int nf_ct_tcp_timeout_close = 10 SECS;
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800103
104/* RFC1122 says the R2 limit should be at least 100 seconds.
105 Linux uses 15 packets as limit, which corresponds
106 to ~13-30min depending on RTO. */
Patrick McHardybabbdb12006-01-09 17:48:09 -0800107unsigned int nf_ct_tcp_timeout_max_retrans = 5 MINS;
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800108
Patrick McHardybabbdb12006-01-09 17:48:09 -0800109static unsigned int * tcp_timeouts[]
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800110= { NULL, /* TCP_CONNTRACK_NONE */
111 &nf_ct_tcp_timeout_syn_sent, /* TCP_CONNTRACK_SYN_SENT, */
112 &nf_ct_tcp_timeout_syn_recv, /* TCP_CONNTRACK_SYN_RECV, */
113 &nf_ct_tcp_timeout_established, /* TCP_CONNTRACK_ESTABLISHED, */
114 &nf_ct_tcp_timeout_fin_wait, /* TCP_CONNTRACK_FIN_WAIT, */
115 &nf_ct_tcp_timeout_close_wait, /* TCP_CONNTRACK_CLOSE_WAIT, */
116 &nf_ct_tcp_timeout_last_ack, /* TCP_CONNTRACK_LAST_ACK, */
117 &nf_ct_tcp_timeout_time_wait, /* TCP_CONNTRACK_TIME_WAIT, */
118 &nf_ct_tcp_timeout_close, /* TCP_CONNTRACK_CLOSE, */
119 NULL, /* TCP_CONNTRACK_LISTEN */
120 };
121
122#define sNO TCP_CONNTRACK_NONE
123#define sSS TCP_CONNTRACK_SYN_SENT
124#define sSR TCP_CONNTRACK_SYN_RECV
125#define sES TCP_CONNTRACK_ESTABLISHED
126#define sFW TCP_CONNTRACK_FIN_WAIT
127#define sCW TCP_CONNTRACK_CLOSE_WAIT
128#define sLA TCP_CONNTRACK_LAST_ACK
129#define sTW TCP_CONNTRACK_TIME_WAIT
130#define sCL TCP_CONNTRACK_CLOSE
131#define sLI TCP_CONNTRACK_LISTEN
132#define sIV TCP_CONNTRACK_MAX
133#define sIG TCP_CONNTRACK_IGNORE
134
135/* What TCP flags are set from RST/SYN/FIN/ACK. */
136enum tcp_bit_set {
137 TCP_SYN_SET,
138 TCP_SYNACK_SET,
139 TCP_FIN_SET,
140 TCP_ACK_SET,
141 TCP_RST_SET,
142 TCP_NONE_SET,
143};
144
145/*
146 * The TCP state transition table needs a few words...
147 *
148 * We are the man in the middle. All the packets go through us
149 * but might get lost in transit to the destination.
150 * It is assumed that the destinations can't receive segments
151 * we haven't seen.
152 *
153 * The checked segment is in window, but our windows are *not*
154 * equivalent with the ones of the sender/receiver. We always
155 * try to guess the state of the current sender.
156 *
157 * The meaning of the states are:
158 *
159 * NONE: initial state
160 * SYN_SENT: SYN-only packet seen
161 * SYN_RECV: SYN-ACK packet seen
162 * ESTABLISHED: ACK packet seen
163 * FIN_WAIT: FIN packet seen
164 * CLOSE_WAIT: ACK seen (after FIN)
165 * LAST_ACK: FIN seen (after FIN)
166 * TIME_WAIT: last ACK seen
167 * CLOSE: closed connection
168 *
169 * LISTEN state is not used.
170 *
171 * Packets marked as IGNORED (sIG):
172 * if they may be either invalid or valid
173 * and the receiver may send back a connection
174 * closing RST or a SYN/ACK.
175 *
176 * Packets marked as INVALID (sIV):
177 * if they are invalid
178 * or we do not support the request (simultaneous open)
179 */
180static enum tcp_conntrack tcp_conntracks[2][6][TCP_CONNTRACK_MAX] = {
181 {
182/* ORIGINAL */
183/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sLI */
184/*syn*/ { sSS, sSS, sIG, sIG, sIG, sIG, sIG, sSS, sSS, sIV },
185/*
186 * sNO -> sSS Initialize a new connection
187 * sSS -> sSS Retransmitted SYN
188 * sSR -> sIG Late retransmitted SYN?
189 * sES -> sIG Error: SYNs in window outside the SYN_SENT state
190 * are errors. Receiver will reply with RST
191 * and close the connection.
192 * Or we are not in sync and hold a dead connection.
193 * sFW -> sIG
194 * sCW -> sIG
195 * sLA -> sIG
196 * sTW -> sSS Reopened connection (RFC 1122).
197 * sCL -> sSS
198 */
199/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sLI */
200/*synack*/ { sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV },
201/*
202 * A SYN/ACK from the client is always invalid:
203 * - either it tries to set up a simultaneous open, which is
204 * not supported;
205 * - or the firewall has just been inserted between the two hosts
206 * during the session set-up. The SYN will be retransmitted
207 * by the true client (or it'll time out).
208 */
209/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sLI */
210/*fin*/ { sIV, sIV, sFW, sFW, sLA, sLA, sLA, sTW, sCL, sIV },
211/*
212 * sNO -> sIV Too late and no reason to do anything...
213 * sSS -> sIV Client migth not send FIN in this state:
214 * we enforce waiting for a SYN/ACK reply first.
215 * sSR -> sFW Close started.
216 * sES -> sFW
217 * sFW -> sLA FIN seen in both directions, waiting for
218 * the last ACK.
219 * Migth be a retransmitted FIN as well...
220 * sCW -> sLA
221 * sLA -> sLA Retransmitted FIN. Remain in the same state.
222 * sTW -> sTW
223 * sCL -> sCL
224 */
225/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sLI */
226/*ack*/ { sES, sIV, sES, sES, sCW, sCW, sTW, sTW, sCL, sIV },
227/*
228 * sNO -> sES Assumed.
229 * sSS -> sIV ACK is invalid: we haven't seen a SYN/ACK yet.
230 * sSR -> sES Established state is reached.
231 * sES -> sES :-)
232 * sFW -> sCW Normal close request answered by ACK.
233 * sCW -> sCW
234 * sLA -> sTW Last ACK detected.
235 * sTW -> sTW Retransmitted last ACK. Remain in the same state.
236 * sCL -> sCL
237 */
238/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sLI */
239/*rst*/ { sIV, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sIV },
240/*none*/ { sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV }
241 },
242 {
243/* REPLY */
244/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sLI */
245/*syn*/ { sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV },
246/*
247 * sNO -> sIV Never reached.
248 * sSS -> sIV Simultaneous open, not supported
249 * sSR -> sIV Simultaneous open, not supported.
250 * sES -> sIV Server may not initiate a connection.
251 * sFW -> sIV
252 * sCW -> sIV
253 * sLA -> sIV
254 * sTW -> sIV Reopened connection, but server may not do it.
255 * sCL -> sIV
256 */
257/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sLI */
258/*synack*/ { sIV, sSR, sSR, sIG, sIG, sIG, sIG, sIG, sIG, sIV },
259/*
260 * sSS -> sSR Standard open.
261 * sSR -> sSR Retransmitted SYN/ACK.
262 * sES -> sIG Late retransmitted SYN/ACK?
263 * sFW -> sIG Might be SYN/ACK answering ignored SYN
264 * sCW -> sIG
265 * sLA -> sIG
266 * sTW -> sIG
267 * sCL -> sIG
268 */
269/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sLI */
270/*fin*/ { sIV, sIV, sFW, sFW, sLA, sLA, sLA, sTW, sCL, sIV },
271/*
272 * sSS -> sIV Server might not send FIN in this state.
273 * sSR -> sFW Close started.
274 * sES -> sFW
275 * sFW -> sLA FIN seen in both directions.
276 * sCW -> sLA
277 * sLA -> sLA Retransmitted FIN.
278 * sTW -> sTW
279 * sCL -> sCL
280 */
281/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sLI */
Jozsef Kadlecsik73f30602005-12-01 14:28:58 -0800282/*ack*/ { sIV, sIG, sSR, sES, sCW, sCW, sTW, sTW, sCL, sIV },
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800283/*
Jozsef Kadlecsik73f30602005-12-01 14:28:58 -0800284 * sSS -> sIG Might be a half-open connection.
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800285 * sSR -> sSR Might answer late resent SYN.
286 * sES -> sES :-)
287 * sFW -> sCW Normal close request answered by ACK.
288 * sCW -> sCW
289 * sLA -> sTW Last ACK detected.
290 * sTW -> sTW Retransmitted last ACK.
291 * sCL -> sCL
292 */
293/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sLI */
294/*rst*/ { sIV, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sIV },
295/*none*/ { sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV }
296 }
297};
298
299static int tcp_pkt_to_tuple(const struct sk_buff *skb,
300 unsigned int dataoff,
301 struct nf_conntrack_tuple *tuple)
302{
303 struct tcphdr _hdr, *hp;
304
305 /* Actually only need first 8 bytes. */
306 hp = skb_header_pointer(skb, dataoff, 8, &_hdr);
307 if (hp == NULL)
308 return 0;
309
310 tuple->src.u.tcp.port = hp->source;
311 tuple->dst.u.tcp.port = hp->dest;
312
313 return 1;
314}
315
316static int tcp_invert_tuple(struct nf_conntrack_tuple *tuple,
317 const struct nf_conntrack_tuple *orig)
318{
319 tuple->src.u.tcp.port = orig->dst.u.tcp.port;
320 tuple->dst.u.tcp.port = orig->src.u.tcp.port;
321 return 1;
322}
323
324/* Print out the per-protocol part of the tuple. */
325static int tcp_print_tuple(struct seq_file *s,
326 const struct nf_conntrack_tuple *tuple)
327{
328 return seq_printf(s, "sport=%hu dport=%hu ",
329 ntohs(tuple->src.u.tcp.port),
330 ntohs(tuple->dst.u.tcp.port));
331}
332
333/* Print out the private part of the conntrack. */
334static int tcp_print_conntrack(struct seq_file *s,
335 const struct nf_conn *conntrack)
336{
337 enum tcp_conntrack state;
338
339 read_lock_bh(&tcp_lock);
340 state = conntrack->proto.tcp.state;
341 read_unlock_bh(&tcp_lock);
342
343 return seq_printf(s, "%s ", tcp_conntrack_names[state]);
344}
345
346static unsigned int get_conntrack_index(const struct tcphdr *tcph)
347{
348 if (tcph->rst) return TCP_RST_SET;
349 else if (tcph->syn) return (tcph->ack ? TCP_SYNACK_SET : TCP_SYN_SET);
350 else if (tcph->fin) return TCP_FIN_SET;
351 else if (tcph->ack) return TCP_ACK_SET;
352 else return TCP_NONE_SET;
353}
354
355/* TCP connection tracking based on 'Real Stateful TCP Packet Filtering
356 in IP Filter' by Guido van Rooij.
357
358 http://www.nluug.nl/events/sane2000/papers.html
359 http://www.iae.nl/users/guido/papers/tcp_filtering.ps.gz
360
361 The boundaries and the conditions are changed according to RFC793:
362 the packet must intersect the window (i.e. segments may be
363 after the right or before the left edge) and thus receivers may ACK
364 segments after the right edge of the window.
365
366 td_maxend = max(sack + max(win,1)) seen in reply packets
367 td_maxwin = max(max(win, 1)) + (sack - ack) seen in sent packets
368 td_maxwin += seq + len - sender.td_maxend
369 if seq + len > sender.td_maxend
370 td_end = max(seq + len) seen in sent packets
371
372 I. Upper bound for valid data: seq <= sender.td_maxend
373 II. Lower bound for valid data: seq + len >= sender.td_end - receiver.td_maxwin
374 III. Upper bound for valid ack: sack <= receiver.td_end
375 IV. Lower bound for valid ack: ack >= receiver.td_end - MAXACKWINDOW
376
377 where sack is the highest right edge of sack block found in the packet.
378
379 The upper bound limit for a valid ack is not ignored -
380 we doesn't have to deal with fragments.
381*/
382
383static inline __u32 segment_seq_plus_len(__u32 seq,
384 size_t len,
385 unsigned int dataoff,
386 struct tcphdr *tcph)
387{
388 /* XXX Should I use payload length field in IP/IPv6 header ?
389 * - YK */
390 return (seq + len - dataoff - tcph->doff*4
391 + (tcph->syn ? 1 : 0) + (tcph->fin ? 1 : 0));
392}
393
394/* Fixme: what about big packets? */
395#define MAXACKWINCONST 66000
396#define MAXACKWINDOW(sender) \
397 ((sender)->td_maxwin > MAXACKWINCONST ? (sender)->td_maxwin \
398 : MAXACKWINCONST)
399
400/*
401 * Simplified tcp_parse_options routine from tcp_input.c
402 */
403static void tcp_options(const struct sk_buff *skb,
404 unsigned int dataoff,
405 struct tcphdr *tcph,
406 struct ip_ct_tcp_state *state)
407{
408 unsigned char buff[(15 * 4) - sizeof(struct tcphdr)];
409 unsigned char *ptr;
410 int length = (tcph->doff*4) - sizeof(struct tcphdr);
411
412 if (!length)
413 return;
414
415 ptr = skb_header_pointer(skb, dataoff + sizeof(struct tcphdr),
416 length, buff);
417 BUG_ON(ptr == NULL);
418
419 state->td_scale =
420 state->flags = 0;
421
422 while (length > 0) {
423 int opcode=*ptr++;
424 int opsize;
425
426 switch (opcode) {
427 case TCPOPT_EOL:
428 return;
429 case TCPOPT_NOP: /* Ref: RFC 793 section 3.1 */
430 length--;
431 continue;
432 default:
433 opsize=*ptr++;
434 if (opsize < 2) /* "silly options" */
435 return;
436 if (opsize > length)
437 break; /* don't parse partial options */
438
439 if (opcode == TCPOPT_SACK_PERM
440 && opsize == TCPOLEN_SACK_PERM)
441 state->flags |= IP_CT_TCP_FLAG_SACK_PERM;
442 else if (opcode == TCPOPT_WINDOW
443 && opsize == TCPOLEN_WINDOW) {
444 state->td_scale = *(u_int8_t *)ptr;
445
446 if (state->td_scale > 14) {
447 /* See RFC1323 */
448 state->td_scale = 14;
449 }
450 state->flags |=
451 IP_CT_TCP_FLAG_WINDOW_SCALE;
452 }
453 ptr += opsize - 2;
454 length -= opsize;
455 }
456 }
457}
458
459static void tcp_sack(const struct sk_buff *skb, unsigned int dataoff,
460 struct tcphdr *tcph, __u32 *sack)
461{
462 unsigned char buff[(15 * 4) - sizeof(struct tcphdr)];
463 unsigned char *ptr;
464 int length = (tcph->doff*4) - sizeof(struct tcphdr);
465 __u32 tmp;
466
467 if (!length)
468 return;
469
470 ptr = skb_header_pointer(skb, dataoff + sizeof(struct tcphdr),
471 length, buff);
472 BUG_ON(ptr == NULL);
473
474 /* Fast path for timestamp-only option */
475 if (length == TCPOLEN_TSTAMP_ALIGNED*4
476 && *(__u32 *)ptr ==
477 __constant_ntohl((TCPOPT_NOP << 24)
478 | (TCPOPT_NOP << 16)
479 | (TCPOPT_TIMESTAMP << 8)
480 | TCPOLEN_TIMESTAMP))
481 return;
482
483 while (length > 0) {
484 int opcode = *ptr++;
485 int opsize, i;
486
487 switch (opcode) {
488 case TCPOPT_EOL:
489 return;
490 case TCPOPT_NOP: /* Ref: RFC 793 section 3.1 */
491 length--;
492 continue;
493 default:
494 opsize = *ptr++;
495 if (opsize < 2) /* "silly options" */
496 return;
497 if (opsize > length)
498 break; /* don't parse partial options */
499
500 if (opcode == TCPOPT_SACK
501 && opsize >= (TCPOLEN_SACK_BASE
502 + TCPOLEN_SACK_PERBLOCK)
503 && !((opsize - TCPOLEN_SACK_BASE)
504 % TCPOLEN_SACK_PERBLOCK)) {
505 for (i = 0;
506 i < (opsize - TCPOLEN_SACK_BASE);
507 i += TCPOLEN_SACK_PERBLOCK) {
508 memcpy(&tmp, (__u32 *)(ptr + i) + 1,
509 sizeof(__u32));
510 tmp = ntohl(tmp);
511
512 if (after(tmp, *sack))
513 *sack = tmp;
514 }
515 return;
516 }
517 ptr += opsize - 2;
518 length -= opsize;
519 }
520 }
521}
522
523static int tcp_in_window(struct ip_ct_tcp *state,
524 enum ip_conntrack_dir dir,
525 unsigned int index,
526 const struct sk_buff *skb,
527 unsigned int dataoff,
528 struct tcphdr *tcph,
529 int pf)
530{
531 struct ip_ct_tcp_state *sender = &state->seen[dir];
532 struct ip_ct_tcp_state *receiver = &state->seen[!dir];
533 __u32 seq, ack, sack, end, win, swin;
534 int res;
535
536 /*
537 * Get the required data from the packet.
538 */
539 seq = ntohl(tcph->seq);
540 ack = sack = ntohl(tcph->ack_seq);
541 win = ntohs(tcph->window);
542 end = segment_seq_plus_len(seq, skb->len, dataoff, tcph);
543
544 if (receiver->flags & IP_CT_TCP_FLAG_SACK_PERM)
545 tcp_sack(skb, dataoff, tcph, &sack);
546
547 DEBUGP("tcp_in_window: START\n");
548 DEBUGP("tcp_in_window: src=%u.%u.%u.%u:%hu dst=%u.%u.%u.%u:%hu "
549 "seq=%u ack=%u sack=%u win=%u end=%u\n",
550 NIPQUAD(iph->saddr), ntohs(tcph->source),
551 NIPQUAD(iph->daddr), ntohs(tcph->dest),
552 seq, ack, sack, win, end);
553 DEBUGP("tcp_in_window: sender end=%u maxend=%u maxwin=%u scale=%i "
554 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
555 sender->td_end, sender->td_maxend, sender->td_maxwin,
556 sender->td_scale,
557 receiver->td_end, receiver->td_maxend, receiver->td_maxwin,
558 receiver->td_scale);
559
560 if (sender->td_end == 0) {
561 /*
562 * Initialize sender data.
563 */
564 if (tcph->syn && tcph->ack) {
565 /*
566 * Outgoing SYN-ACK in reply to a SYN.
567 */
568 sender->td_end =
569 sender->td_maxend = end;
570 sender->td_maxwin = (win == 0 ? 1 : win);
571
572 tcp_options(skb, dataoff, tcph, sender);
573 /*
574 * RFC 1323:
575 * Both sides must send the Window Scale option
576 * to enable window scaling in either direction.
577 */
578 if (!(sender->flags & IP_CT_TCP_FLAG_WINDOW_SCALE
579 && receiver->flags & IP_CT_TCP_FLAG_WINDOW_SCALE))
580 sender->td_scale =
581 receiver->td_scale = 0;
582 } else {
583 /*
584 * We are in the middle of a connection,
585 * its history is lost for us.
586 * Let's try to use the data from the packet.
587 */
588 sender->td_end = end;
589 sender->td_maxwin = (win == 0 ? 1 : win);
590 sender->td_maxend = end + sender->td_maxwin;
591 }
592 } else if (((state->state == TCP_CONNTRACK_SYN_SENT
593 && dir == IP_CT_DIR_ORIGINAL)
594 || (state->state == TCP_CONNTRACK_SYN_RECV
595 && dir == IP_CT_DIR_REPLY))
596 && after(end, sender->td_end)) {
597 /*
598 * RFC 793: "if a TCP is reinitialized ... then it need
599 * not wait at all; it must only be sure to use sequence
600 * numbers larger than those recently used."
601 */
602 sender->td_end =
603 sender->td_maxend = end;
604 sender->td_maxwin = (win == 0 ? 1 : win);
605
606 tcp_options(skb, dataoff, tcph, sender);
607 }
608
609 if (!(tcph->ack)) {
610 /*
611 * If there is no ACK, just pretend it was set and OK.
612 */
613 ack = sack = receiver->td_end;
614 } else if (((tcp_flag_word(tcph) & (TCP_FLAG_ACK|TCP_FLAG_RST)) ==
615 (TCP_FLAG_ACK|TCP_FLAG_RST))
616 && (ack == 0)) {
617 /*
618 * Broken TCP stacks, that set ACK in RST packets as well
619 * with zero ack value.
620 */
621 ack = sack = receiver->td_end;
622 }
623
624 if (seq == end
625 && (!tcph->rst
626 || (seq == 0 && state->state == TCP_CONNTRACK_SYN_SENT)))
627 /*
628 * Packets contains no data: we assume it is valid
629 * and check the ack value only.
630 * However RST segments are always validated by their
631 * SEQ number, except when seq == 0 (reset sent answering
632 * SYN.
633 */
634 seq = end = sender->td_end;
635
636 DEBUGP("tcp_in_window: src=%u.%u.%u.%u:%hu dst=%u.%u.%u.%u:%hu "
637 "seq=%u ack=%u sack =%u win=%u end=%u\n",
638 NIPQUAD(iph->saddr), ntohs(tcph->source),
639 NIPQUAD(iph->daddr), ntohs(tcph->dest),
640 seq, ack, sack, win, end);
641 DEBUGP("tcp_in_window: sender end=%u maxend=%u maxwin=%u scale=%i "
642 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
643 sender->td_end, sender->td_maxend, sender->td_maxwin,
644 sender->td_scale,
645 receiver->td_end, receiver->td_maxend, receiver->td_maxwin,
646 receiver->td_scale);
647
648 DEBUGP("tcp_in_window: I=%i II=%i III=%i IV=%i\n",
649 before(seq, sender->td_maxend + 1),
650 after(end, sender->td_end - receiver->td_maxwin - 1),
651 before(sack, receiver->td_end + 1),
652 after(ack, receiver->td_end - MAXACKWINDOW(sender)));
653
654 if (sender->loose || receiver->loose ||
655 (before(seq, sender->td_maxend + 1) &&
656 after(end, sender->td_end - receiver->td_maxwin - 1) &&
657 before(sack, receiver->td_end + 1) &&
658 after(ack, receiver->td_end - MAXACKWINDOW(sender)))) {
659 /*
660 * Take into account window scaling (RFC 1323).
661 */
662 if (!tcph->syn)
663 win <<= sender->td_scale;
664
665 /*
666 * Update sender data.
667 */
668 swin = win + (sack - ack);
669 if (sender->td_maxwin < swin)
670 sender->td_maxwin = swin;
671 if (after(end, sender->td_end))
672 sender->td_end = end;
673 /*
674 * Update receiver data.
675 */
676 if (after(end, sender->td_maxend))
677 receiver->td_maxwin += end - sender->td_maxend;
678 if (after(sack + win, receiver->td_maxend - 1)) {
679 receiver->td_maxend = sack + win;
680 if (win == 0)
681 receiver->td_maxend++;
682 }
683
684 /*
685 * Check retransmissions.
686 */
687 if (index == TCP_ACK_SET) {
688 if (state->last_dir == dir
689 && state->last_seq == seq
690 && state->last_ack == ack
691 && state->last_end == end)
692 state->retrans++;
693 else {
694 state->last_dir = dir;
695 state->last_seq = seq;
696 state->last_ack = ack;
697 state->last_end = end;
698 state->retrans = 0;
699 }
700 }
701 /*
702 * Close the window of disabled window tracking :-)
703 */
704 if (sender->loose)
705 sender->loose--;
706
707 res = 1;
708 } else {
709 if (LOG_INVALID(IPPROTO_TCP))
710 nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
711 "nf_ct_tcp: %s ",
712 before(seq, sender->td_maxend + 1) ?
713 after(end, sender->td_end - receiver->td_maxwin - 1) ?
714 before(sack, receiver->td_end + 1) ?
715 after(ack, receiver->td_end - MAXACKWINDOW(sender)) ? "BUG"
716 : "ACK is under the lower bound (possible overly delayed ACK)"
717 : "ACK is over the upper bound (ACKed data not seen yet)"
718 : "SEQ is under the lower bound (already ACKed data retransmitted)"
719 : "SEQ is over the upper bound (over the window of the receiver)");
720
721 res = nf_ct_tcp_be_liberal;
722 }
723
724 DEBUGP("tcp_in_window: res=%i sender end=%u maxend=%u maxwin=%u "
725 "receiver end=%u maxend=%u maxwin=%u\n",
726 res, sender->td_end, sender->td_maxend, sender->td_maxwin,
727 receiver->td_end, receiver->td_maxend, receiver->td_maxwin);
728
729 return res;
730}
731
732#ifdef CONFIG_IP_NF_NAT_NEEDED
733/* Update sender->td_end after NAT successfully mangled the packet */
734/* Caller must linearize skb at tcp header. */
735void nf_conntrack_tcp_update(struct sk_buff *skb,
736 unsigned int dataoff,
737 struct nf_conn *conntrack,
738 int dir)
739{
740 struct tcphdr *tcph = (void *)skb->data + dataoff;
741 __u32 end;
742#ifdef DEBUGP_VARS
743 struct ip_ct_tcp_state *sender = &conntrack->proto.tcp.seen[dir];
744 struct ip_ct_tcp_state *receiver = &conntrack->proto.tcp.seen[!dir];
745#endif
746
747 end = segment_seq_plus_len(ntohl(tcph->seq), skb->len, dataoff, tcph);
748
749 write_lock_bh(&tcp_lock);
750 /*
751 * We have to worry for the ack in the reply packet only...
752 */
753 if (after(end, conntrack->proto.tcp.seen[dir].td_end))
754 conntrack->proto.tcp.seen[dir].td_end = end;
755 conntrack->proto.tcp.last_end = end;
756 write_unlock_bh(&tcp_lock);
757 DEBUGP("tcp_update: sender end=%u maxend=%u maxwin=%u scale=%i "
758 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
759 sender->td_end, sender->td_maxend, sender->td_maxwin,
760 sender->td_scale,
761 receiver->td_end, receiver->td_maxend, receiver->td_maxwin,
762 receiver->td_scale);
763}
764
765#endif
766
767#define TH_FIN 0x01
768#define TH_SYN 0x02
769#define TH_RST 0x04
770#define TH_PUSH 0x08
771#define TH_ACK 0x10
772#define TH_URG 0x20
773#define TH_ECE 0x40
774#define TH_CWR 0x80
775
776/* table of valid flag combinations - ECE and CWR are always valid */
777static u8 tcp_valid_flags[(TH_FIN|TH_SYN|TH_RST|TH_PUSH|TH_ACK|TH_URG) + 1] =
778{
779 [TH_SYN] = 1,
780 [TH_SYN|TH_ACK] = 1,
Vlad Drukkera2d72222005-11-12 12:13:14 -0800781 [TH_SYN|TH_PUSH] = 1,
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800782 [TH_SYN|TH_ACK|TH_PUSH] = 1,
783 [TH_RST] = 1,
784 [TH_RST|TH_ACK] = 1,
785 [TH_RST|TH_ACK|TH_PUSH] = 1,
786 [TH_FIN|TH_ACK] = 1,
787 [TH_ACK] = 1,
788 [TH_ACK|TH_PUSH] = 1,
789 [TH_ACK|TH_URG] = 1,
790 [TH_ACK|TH_URG|TH_PUSH] = 1,
791 [TH_FIN|TH_ACK|TH_PUSH] = 1,
792 [TH_FIN|TH_ACK|TH_URG] = 1,
793 [TH_FIN|TH_ACK|TH_URG|TH_PUSH] = 1,
794};
795
796/* Protect conntrack agaist broken packets. Code taken from ipt_unclean.c. */
797static int tcp_error(struct sk_buff *skb,
798 unsigned int dataoff,
799 enum ip_conntrack_info *ctinfo,
800 int pf,
Patrick McHardy96f6bf82006-04-06 14:19:24 -0700801 unsigned int hooknum)
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800802{
803 struct tcphdr _tcph, *th;
804 unsigned int tcplen = skb->len - dataoff;
805 u_int8_t tcpflags;
806
807 /* Smaller that minimal TCP header? */
808 th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph);
809 if (th == NULL) {
810 if (LOG_INVALID(IPPROTO_TCP))
811 nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
812 "nf_ct_tcp: short packet ");
813 return -NF_ACCEPT;
814 }
815
816 /* Not whole TCP header or malformed packet */
817 if (th->doff*4 < sizeof(struct tcphdr) || tcplen < th->doff*4) {
818 if (LOG_INVALID(IPPROTO_TCP))
819 nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
820 "nf_ct_tcp: truncated/malformed packet ");
821 return -NF_ACCEPT;
822 }
823
824 /* Checksum invalid? Ignore.
825 * We skip checking packets on the outgoing path
Patrick McHardy84fa7932006-08-29 16:44:56 -0700826 * because the checksum is assumed to be correct.
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800827 */
828 /* FIXME: Source route IP option packets --RR */
Patrick McHardy39a27a32006-05-29 18:23:54 -0700829 if (nf_conntrack_checksum &&
830 ((pf == PF_INET && hooknum == NF_IP_PRE_ROUTING) ||
831 (pf == PF_INET6 && hooknum == NF_IP6_PRE_ROUTING)) &&
Patrick McHardy96f6bf82006-04-06 14:19:24 -0700832 nf_checksum(skb, hooknum, dataoff, IPPROTO_TCP, pf)) {
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800833 if (LOG_INVALID(IPPROTO_TCP))
834 nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
835 "nf_ct_tcp: bad TCP checksum ");
836 return -NF_ACCEPT;
837 }
838
839 /* Check TCP flags. */
840 tcpflags = (((u_int8_t *)th)[13] & ~(TH_ECE|TH_CWR));
841 if (!tcp_valid_flags[tcpflags]) {
842 if (LOG_INVALID(IPPROTO_TCP))
843 nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
844 "nf_ct_tcp: invalid TCP flag combination ");
845 return -NF_ACCEPT;
846 }
847
848 return NF_ACCEPT;
849}
850
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800851/* Returns verdict for packet, or -1 for invalid. */
852static int tcp_packet(struct nf_conn *conntrack,
853 const struct sk_buff *skb,
854 unsigned int dataoff,
855 enum ip_conntrack_info ctinfo,
856 int pf,
857 unsigned int hooknum)
858{
859 enum tcp_conntrack new_state, old_state;
860 enum ip_conntrack_dir dir;
861 struct tcphdr *th, _tcph;
862 unsigned long timeout;
863 unsigned int index;
864
865 th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph);
866 BUG_ON(th == NULL);
867
868 write_lock_bh(&tcp_lock);
869 old_state = conntrack->proto.tcp.state;
870 dir = CTINFO2DIR(ctinfo);
871 index = get_conntrack_index(th);
872 new_state = tcp_conntracks[dir][index][old_state];
873
874 switch (new_state) {
875 case TCP_CONNTRACK_IGNORE:
Jozsef Kadlecsik73f30602005-12-01 14:28:58 -0800876 /* Ignored packets:
877 *
878 * a) SYN in ORIGINAL
879 * b) SYN/ACK in REPLY
880 * c) ACK in reply direction after initial SYN in original.
881 */
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800882 if (index == TCP_SYNACK_SET
883 && conntrack->proto.tcp.last_index == TCP_SYN_SET
884 && conntrack->proto.tcp.last_dir != dir
885 && ntohl(th->ack_seq) ==
886 conntrack->proto.tcp.last_end) {
887 /* This SYN/ACK acknowledges a SYN that we earlier
888 * ignored as invalid. This means that the client and
889 * the server are both in sync, while the firewall is
890 * not. We kill this session and block the SYN/ACK so
891 * that the client cannot but retransmit its SYN and
892 * thus initiate a clean new session.
893 */
894 write_unlock_bh(&tcp_lock);
895 if (LOG_INVALID(IPPROTO_TCP))
896 nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
897 "nf_ct_tcp: killing out of sync session ");
898 if (del_timer(&conntrack->timeout))
899 conntrack->timeout.function((unsigned long)
900 conntrack);
901 return -NF_DROP;
902 }
903 conntrack->proto.tcp.last_index = index;
904 conntrack->proto.tcp.last_dir = dir;
905 conntrack->proto.tcp.last_seq = ntohl(th->seq);
906 conntrack->proto.tcp.last_end =
907 segment_seq_plus_len(ntohl(th->seq), skb->len, dataoff, th);
908
909 write_unlock_bh(&tcp_lock);
910 if (LOG_INVALID(IPPROTO_TCP))
911 nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
912 "nf_ct_tcp: invalid packed ignored ");
913 return NF_ACCEPT;
914 case TCP_CONNTRACK_MAX:
915 /* Invalid packet */
916 DEBUGP("nf_ct_tcp: Invalid dir=%i index=%u ostate=%u\n",
917 dir, get_conntrack_index(th),
918 old_state);
919 write_unlock_bh(&tcp_lock);
920 if (LOG_INVALID(IPPROTO_TCP))
921 nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
922 "nf_ct_tcp: invalid state ");
923 return -NF_ACCEPT;
924 case TCP_CONNTRACK_SYN_SENT:
925 if (old_state < TCP_CONNTRACK_TIME_WAIT)
926 break;
927 if ((conntrack->proto.tcp.seen[dir].flags &
928 IP_CT_TCP_FLAG_CLOSE_INIT)
929 || after(ntohl(th->seq),
930 conntrack->proto.tcp.seen[dir].td_end)) {
931 /* Attempt to reopen a closed connection.
932 * Delete this connection and look up again. */
933 write_unlock_bh(&tcp_lock);
934 if (del_timer(&conntrack->timeout))
935 conntrack->timeout.function((unsigned long)
936 conntrack);
937 return -NF_REPEAT;
KOVACS Krisztian3746a2b2005-11-14 15:23:01 -0800938 } else {
939 write_unlock_bh(&tcp_lock);
940 if (LOG_INVALID(IPPROTO_TCP))
941 nf_log_packet(pf, 0, skb, NULL, NULL,
942 NULL, "nf_ct_tcp: invalid SYN");
943 return -NF_ACCEPT;
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800944 }
945 case TCP_CONNTRACK_CLOSE:
946 if (index == TCP_RST_SET
Jozsef Kadlecsik73f30602005-12-01 14:28:58 -0800947 && ((test_bit(IPS_SEEN_REPLY_BIT, &conntrack->status)
948 && conntrack->proto.tcp.last_index == TCP_SYN_SET)
949 || (!test_bit(IPS_ASSURED_BIT, &conntrack->status)
950 && conntrack->proto.tcp.last_index == TCP_ACK_SET))
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800951 && ntohl(th->ack_seq) == conntrack->proto.tcp.last_end) {
Adrian Bunk93b1fae2006-01-10 00:13:33 +0100952 /* RST sent to invalid SYN or ACK we had let through
Jozsef Kadlecsik73f30602005-12-01 14:28:58 -0800953 * at a) and c) above:
954 *
955 * a) SYN was in window then
956 * c) we hold a half-open connection.
957 *
958 * Delete our connection entry.
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800959 * We skip window checking, because packet might ACK
Jozsef Kadlecsik73f30602005-12-01 14:28:58 -0800960 * segments we ignored. */
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800961 goto in_window;
962 }
Adrian Bunk93b1fae2006-01-10 00:13:33 +0100963 /* Just fall through */
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800964 default:
965 /* Keep compilers happy. */
966 break;
967 }
968
969 if (!tcp_in_window(&conntrack->proto.tcp, dir, index,
970 skb, dataoff, th, pf)) {
971 write_unlock_bh(&tcp_lock);
972 return -NF_ACCEPT;
973 }
974 in_window:
975 /* From now on we have got in-window packets */
976 conntrack->proto.tcp.last_index = index;
977
978 DEBUGP("tcp_conntracks: src=%u.%u.%u.%u:%hu dst=%u.%u.%u.%u:%hu "
979 "syn=%i ack=%i fin=%i rst=%i old=%i new=%i\n",
980 NIPQUAD(iph->saddr), ntohs(th->source),
981 NIPQUAD(iph->daddr), ntohs(th->dest),
982 (th->syn ? 1 : 0), (th->ack ? 1 : 0),
983 (th->fin ? 1 : 0), (th->rst ? 1 : 0),
984 old_state, new_state);
985
986 conntrack->proto.tcp.state = new_state;
987 if (old_state != new_state
988 && (new_state == TCP_CONNTRACK_FIN_WAIT
989 || new_state == TCP_CONNTRACK_CLOSE))
990 conntrack->proto.tcp.seen[dir].flags |= IP_CT_TCP_FLAG_CLOSE_INIT;
991 timeout = conntrack->proto.tcp.retrans >= nf_ct_tcp_max_retrans
992 && *tcp_timeouts[new_state] > nf_ct_tcp_timeout_max_retrans
993 ? nf_ct_tcp_timeout_max_retrans : *tcp_timeouts[new_state];
994 write_unlock_bh(&tcp_lock);
995
996 nf_conntrack_event_cache(IPCT_PROTOINFO_VOLATILE, skb);
997 if (new_state != old_state)
998 nf_conntrack_event_cache(IPCT_PROTOINFO, skb);
999
1000 if (!test_bit(IPS_SEEN_REPLY_BIT, &conntrack->status)) {
1001 /* If only reply is a RST, we can consider ourselves not to
1002 have an established connection: this is a fairly common
1003 problem case, so we can delete the conntrack
1004 immediately. --RR */
1005 if (th->rst) {
1006 if (del_timer(&conntrack->timeout))
1007 conntrack->timeout.function((unsigned long)
1008 conntrack);
1009 return NF_ACCEPT;
1010 }
1011 } else if (!test_bit(IPS_ASSURED_BIT, &conntrack->status)
1012 && (old_state == TCP_CONNTRACK_SYN_RECV
1013 || old_state == TCP_CONNTRACK_ESTABLISHED)
1014 && new_state == TCP_CONNTRACK_ESTABLISHED) {
1015 /* Set ASSURED if we see see valid ack in ESTABLISHED
1016 after SYN_RECV or a valid answer for a picked up
1017 connection. */
1018 set_bit(IPS_ASSURED_BIT, &conntrack->status);
1019 nf_conntrack_event_cache(IPCT_STATUS, skb);
1020 }
1021 nf_ct_refresh_acct(conntrack, ctinfo, skb, timeout);
1022
1023 return NF_ACCEPT;
1024}
1025
1026/* Called when a new connection for this protocol found. */
1027static int tcp_new(struct nf_conn *conntrack,
1028 const struct sk_buff *skb,
1029 unsigned int dataoff)
1030{
1031 enum tcp_conntrack new_state;
1032 struct tcphdr *th, _tcph;
1033#ifdef DEBUGP_VARS
1034 struct ip_ct_tcp_state *sender = &conntrack->proto.tcp.seen[0];
1035 struct ip_ct_tcp_state *receiver = &conntrack->proto.tcp.seen[1];
1036#endif
1037
1038 th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph);
1039 BUG_ON(th == NULL);
1040
1041 /* Don't need lock here: this conntrack not in circulation yet */
1042 new_state
1043 = tcp_conntracks[0][get_conntrack_index(th)]
1044 [TCP_CONNTRACK_NONE];
1045
1046 /* Invalid: delete conntrack */
1047 if (new_state >= TCP_CONNTRACK_MAX) {
1048 DEBUGP("nf_ct_tcp: invalid new deleting.\n");
1049 return 0;
1050 }
1051
1052 if (new_state == TCP_CONNTRACK_SYN_SENT) {
1053 /* SYN packet */
1054 conntrack->proto.tcp.seen[0].td_end =
1055 segment_seq_plus_len(ntohl(th->seq), skb->len,
1056 dataoff, th);
1057 conntrack->proto.tcp.seen[0].td_maxwin = ntohs(th->window);
1058 if (conntrack->proto.tcp.seen[0].td_maxwin == 0)
1059 conntrack->proto.tcp.seen[0].td_maxwin = 1;
1060 conntrack->proto.tcp.seen[0].td_maxend =
1061 conntrack->proto.tcp.seen[0].td_end;
1062
1063 tcp_options(skb, dataoff, th, &conntrack->proto.tcp.seen[0]);
1064 conntrack->proto.tcp.seen[1].flags = 0;
1065 conntrack->proto.tcp.seen[0].loose =
1066 conntrack->proto.tcp.seen[1].loose = 0;
1067 } else if (nf_ct_tcp_loose == 0) {
1068 /* Don't try to pick up connections. */
1069 return 0;
1070 } else {
1071 /*
1072 * We are in the middle of a connection,
1073 * its history is lost for us.
1074 * Let's try to use the data from the packet.
1075 */
1076 conntrack->proto.tcp.seen[0].td_end =
1077 segment_seq_plus_len(ntohl(th->seq), skb->len,
1078 dataoff, th);
1079 conntrack->proto.tcp.seen[0].td_maxwin = ntohs(th->window);
1080 if (conntrack->proto.tcp.seen[0].td_maxwin == 0)
1081 conntrack->proto.tcp.seen[0].td_maxwin = 1;
1082 conntrack->proto.tcp.seen[0].td_maxend =
1083 conntrack->proto.tcp.seen[0].td_end +
1084 conntrack->proto.tcp.seen[0].td_maxwin;
1085 conntrack->proto.tcp.seen[0].td_scale = 0;
1086
1087 /* We assume SACK. Should we assume window scaling too? */
1088 conntrack->proto.tcp.seen[0].flags =
1089 conntrack->proto.tcp.seen[1].flags = IP_CT_TCP_FLAG_SACK_PERM;
1090 conntrack->proto.tcp.seen[0].loose =
1091 conntrack->proto.tcp.seen[1].loose = nf_ct_tcp_loose;
1092 }
1093
1094 conntrack->proto.tcp.seen[1].td_end = 0;
1095 conntrack->proto.tcp.seen[1].td_maxend = 0;
1096 conntrack->proto.tcp.seen[1].td_maxwin = 1;
1097 conntrack->proto.tcp.seen[1].td_scale = 0;
1098
1099 /* tcp_packet will set them */
1100 conntrack->proto.tcp.state = TCP_CONNTRACK_NONE;
1101 conntrack->proto.tcp.last_index = TCP_NONE_SET;
1102
1103 DEBUGP("tcp_new: sender end=%u maxend=%u maxwin=%u scale=%i "
1104 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
1105 sender->td_end, sender->td_maxend, sender->td_maxwin,
1106 sender->td_scale,
1107 receiver->td_end, receiver->td_maxend, receiver->td_maxwin,
1108 receiver->td_scale);
1109 return 1;
1110}
Pablo Neira Ayusoc1d10ad2006-01-05 12:19:05 -08001111
1112#if defined(CONFIG_NF_CT_NETLINK) || \
1113 defined(CONFIG_NF_CT_NETLINK_MODULE)
1114
1115#include <linux/netfilter/nfnetlink.h>
1116#include <linux/netfilter/nfnetlink_conntrack.h>
1117
1118static int tcp_to_nfattr(struct sk_buff *skb, struct nfattr *nfa,
1119 const struct nf_conn *ct)
1120{
1121 struct nfattr *nest_parms;
1122
1123 read_lock_bh(&tcp_lock);
1124 nest_parms = NFA_NEST(skb, CTA_PROTOINFO_TCP);
1125 NFA_PUT(skb, CTA_PROTOINFO_TCP_STATE, sizeof(u_int8_t),
1126 &ct->proto.tcp.state);
1127 read_unlock_bh(&tcp_lock);
1128
1129 NFA_NEST_END(skb, nest_parms);
1130
1131 return 0;
1132
1133nfattr_failure:
1134 read_unlock_bh(&tcp_lock);
1135 return -1;
1136}
1137
1138static const size_t cta_min_tcp[CTA_PROTOINFO_TCP_MAX] = {
1139 [CTA_PROTOINFO_TCP_STATE-1] = sizeof(u_int8_t),
1140};
1141
1142static int nfattr_to_tcp(struct nfattr *cda[], struct nf_conn *ct)
1143{
1144 struct nfattr *attr = cda[CTA_PROTOINFO_TCP-1];
1145 struct nfattr *tb[CTA_PROTOINFO_TCP_MAX];
1146
1147 /* updates could not contain anything about the private
1148 * protocol info, in that case skip the parsing */
1149 if (!attr)
1150 return 0;
1151
1152 nfattr_parse_nested(tb, CTA_PROTOINFO_TCP_MAX, attr);
1153
1154 if (nfattr_bad_size(tb, CTA_PROTOINFO_TCP_MAX, cta_min_tcp))
1155 return -EINVAL;
1156
1157 if (!tb[CTA_PROTOINFO_TCP_STATE-1])
1158 return -EINVAL;
1159
1160 write_lock_bh(&tcp_lock);
1161 ct->proto.tcp.state =
1162 *(u_int8_t *)NFA_DATA(tb[CTA_PROTOINFO_TCP_STATE-1]);
1163 write_unlock_bh(&tcp_lock);
1164
1165 return 0;
1166}
1167#endif
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08001168
1169struct nf_conntrack_protocol nf_conntrack_protocol_tcp4 =
1170{
1171 .l3proto = PF_INET,
1172 .proto = IPPROTO_TCP,
1173 .name = "tcp",
1174 .pkt_to_tuple = tcp_pkt_to_tuple,
1175 .invert_tuple = tcp_invert_tuple,
1176 .print_tuple = tcp_print_tuple,
1177 .print_conntrack = tcp_print_conntrack,
1178 .packet = tcp_packet,
1179 .new = tcp_new,
Patrick McHardy96f6bf82006-04-06 14:19:24 -07001180 .error = tcp_error,
Pablo Neira Ayusoc1d10ad2006-01-05 12:19:05 -08001181#if defined(CONFIG_NF_CT_NETLINK) || \
1182 defined(CONFIG_NF_CT_NETLINK_MODULE)
1183 .to_nfattr = tcp_to_nfattr,
1184 .from_nfattr = nfattr_to_tcp,
1185 .tuple_to_nfattr = nf_ct_port_tuple_to_nfattr,
1186 .nfattr_to_tuple = nf_ct_port_nfattr_to_tuple,
1187#endif
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08001188};
1189
1190struct nf_conntrack_protocol nf_conntrack_protocol_tcp6 =
1191{
1192 .l3proto = PF_INET6,
1193 .proto = IPPROTO_TCP,
1194 .name = "tcp",
1195 .pkt_to_tuple = tcp_pkt_to_tuple,
1196 .invert_tuple = tcp_invert_tuple,
1197 .print_tuple = tcp_print_tuple,
1198 .print_conntrack = tcp_print_conntrack,
1199 .packet = tcp_packet,
1200 .new = tcp_new,
Patrick McHardy96f6bf82006-04-06 14:19:24 -07001201 .error = tcp_error,
Pablo Neira Ayusoc1d10ad2006-01-05 12:19:05 -08001202#if defined(CONFIG_NF_CT_NETLINK) || \
1203 defined(CONFIG_NF_CT_NETLINK_MODULE)
1204 .to_nfattr = tcp_to_nfattr,
1205 .from_nfattr = nfattr_to_tcp,
1206 .tuple_to_nfattr = nf_ct_port_tuple_to_nfattr,
1207 .nfattr_to_tuple = nf_ct_port_nfattr_to_tuple,
1208#endif
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08001209};
1210
1211EXPORT_SYMBOL(nf_conntrack_protocol_tcp4);
1212EXPORT_SYMBOL(nf_conntrack_protocol_tcp6);