1 // SPDX-License-Identifier: GPL-2.0-only
2 /* (C) 1999-2001 Paul `Rusty' Russell
3 * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
4 * (C) 2002-2013 Jozsef Kadlecsik <kadlec@netfilter.org>
5 * (C) 2006-2012 Patrick McHardy <kaber@trash.net>
6 */
7
8 #include <linux/types.h>
9 #include <linux/timer.h>
10 #include <linux/module.h>
11 #include <linux/in.h>
12 #include <linux/tcp.h>
13 #include <linux/spinlock.h>
14 #include <linux/skbuff.h>
15 #include <linux/ipv6.h>
16 #include <net/ip6_checksum.h>
17 #include <asm/unaligned.h>
18
19 #include <net/tcp.h>
20
21 #include <linux/netfilter.h>
22 #include <linux/netfilter_ipv4.h>
23 #include <linux/netfilter_ipv6.h>
24 #include <net/netfilter/nf_conntrack.h>
25 #include <net/netfilter/nf_conntrack_l4proto.h>
26 #include <net/netfilter/nf_conntrack_ecache.h>
27 #include <net/netfilter/nf_conntrack_seqadj.h>
28 #include <net/netfilter/nf_conntrack_synproxy.h>
29 #include <net/netfilter/nf_conntrack_timeout.h>
30 #include <net/netfilter/nf_log.h>
31 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
32 #include <net/netfilter/ipv6/nf_conntrack_ipv6.h>
33
34 /* FIXME: Examine ipfilter's timeouts and conntrack transitions more
35 closely. They're more complex. --RR */
36
37 static const char *const tcp_conntrack_names[] = {
38 "NONE",
39 "SYN_SENT",
40 "SYN_RECV",
41 "ESTABLISHED",
42 "FIN_WAIT",
43 "CLOSE_WAIT",
44 "LAST_ACK",
45 "TIME_WAIT",
46 "CLOSE",
47 "SYN_SENT2",
48 };
49
50 #define SECS * HZ
51 #define MINS * 60 SECS
52 #define HOURS * 60 MINS
53 #define DAYS * 24 HOURS
54
55 static const unsigned int tcp_timeouts[TCP_CONNTRACK_TIMEOUT_MAX] = {
56 [TCP_CONNTRACK_SYN_SENT] = 2 MINS,
57 [TCP_CONNTRACK_SYN_RECV] = 60 SECS,
58 [TCP_CONNTRACK_ESTABLISHED] = 5 DAYS,
59 [TCP_CONNTRACK_FIN_WAIT] = 2 MINS,
60 [TCP_CONNTRACK_CLOSE_WAIT] = 60 SECS,
61 [TCP_CONNTRACK_LAST_ACK] = 30 SECS,
62 [TCP_CONNTRACK_TIME_WAIT] = 2 MINS,
63 [TCP_CONNTRACK_CLOSE] = 10 SECS,
64 [TCP_CONNTRACK_SYN_SENT2] = 2 MINS,
65 /* RFC1122 says the R2 limit should be at least 100 seconds.
66 Linux uses 15 packets as limit, which corresponds
67 to ~13-30min depending on RTO. */
68 [TCP_CONNTRACK_RETRANS] = 5 MINS,
69 [TCP_CONNTRACK_UNACK] = 5 MINS,
70 };
71
72 #define sNO TCP_CONNTRACK_NONE
73 #define sSS TCP_CONNTRACK_SYN_SENT
74 #define sSR TCP_CONNTRACK_SYN_RECV
75 #define sES TCP_CONNTRACK_ESTABLISHED
76 #define sFW TCP_CONNTRACK_FIN_WAIT
77 #define sCW TCP_CONNTRACK_CLOSE_WAIT
78 #define sLA TCP_CONNTRACK_LAST_ACK
79 #define sTW TCP_CONNTRACK_TIME_WAIT
80 #define sCL TCP_CONNTRACK_CLOSE
81 #define sS2 TCP_CONNTRACK_SYN_SENT2
82 #define sIV TCP_CONNTRACK_MAX
83 #define sIG TCP_CONNTRACK_IGNORE
84
85 /* What TCP flags are set from RST/SYN/FIN/ACK. */
86 enum tcp_bit_set {
87 TCP_SYN_SET,
88 TCP_SYNACK_SET,
89 TCP_FIN_SET,
90 TCP_ACK_SET,
91 TCP_RST_SET,
92 TCP_NONE_SET,
93 };
94
95 /*
96 * The TCP state transition table needs a few words...
97 *
98 * We are the man in the middle. All the packets go through us
99 * but might get lost in transit to the destination.
100 * It is assumed that the destinations can't receive segments
101 * we haven't seen.
102 *
103 * The checked segment is in window, but our windows are *not*
104 * equivalent with the ones of the sender/receiver. We always
105 * try to guess the state of the current sender.
106 *
107 * The meaning of the states are:
108 *
109 * NONE: initial state
110 * SYN_SENT: SYN-only packet seen
111 * SYN_SENT2: SYN-only packet seen from reply dir, simultaneous open
112 * SYN_RECV: SYN-ACK packet seen
113 * ESTABLISHED: ACK packet seen
114 * FIN_WAIT: FIN packet seen
115 * CLOSE_WAIT: ACK seen (after FIN)
116 * LAST_ACK: FIN seen (after FIN)
117 * TIME_WAIT: last ACK seen
118 * CLOSE: closed connection (RST)
119 *
120 * Packets marked as IGNORED (sIG):
121 * if they may be either invalid or valid
122 * and the receiver may send back a connection
123 * closing RST or a SYN/ACK.
124 *
125 * Packets marked as INVALID (sIV):
126 * if we regard them as truly invalid packets
127 */
128 static const u8 tcp_conntracks[2][6][TCP_CONNTRACK_MAX] = {
129 {
130 /* ORIGINAL */
131 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
132 /*syn*/ { sSS, sSS, sIG, sIG, sIG, sIG, sIG, sSS, sSS, sS2 },
133 /*
134 * sNO -> sSS Initialize a new connection
135 * sSS -> sSS Retransmitted SYN
136 * sS2 -> sS2 Late retransmitted SYN
137 * sSR -> sIG
138 * sES -> sIG Error: SYNs in window outside the SYN_SENT state
139 * are errors. Receiver will reply with RST
140 * and close the connection.
141 * Or we are not in sync and hold a dead connection.
142 * sFW -> sIG
143 * sCW -> sIG
144 * sLA -> sIG
145 * sTW -> sSS Reopened connection (RFC 1122).
146 * sCL -> sSS
147 */
148 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
149 /*synack*/ { sIV, sIV, sSR, sIV, sIV, sIV, sIV, sIV, sIV, sSR },
150 /*
151 * sNO -> sIV Too late and no reason to do anything
152 * sSS -> sIV Client can't send SYN and then SYN/ACK
153 * sS2 -> sSR SYN/ACK sent to SYN2 in simultaneous open
154 * sSR -> sSR Late retransmitted SYN/ACK in simultaneous open
155 * sES -> sIV Invalid SYN/ACK packets sent by the client
156 * sFW -> sIV
157 * sCW -> sIV
158 * sLA -> sIV
159 * sTW -> sIV
160 * sCL -> sIV
161 */
162 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
163 /*fin*/ { sIV, sIV, sFW, sFW, sLA, sLA, sLA, sTW, sCL, sIV },
164 /*
165 * sNO -> sIV Too late and no reason to do anything...
166 * sSS -> sIV Client migth not send FIN in this state:
167 * we enforce waiting for a SYN/ACK reply first.
168 * sS2 -> sIV
169 * sSR -> sFW Close started.
170 * sES -> sFW
171 * sFW -> sLA FIN seen in both directions, waiting for
172 * the last ACK.
173 * Migth be a retransmitted FIN as well...
174 * sCW -> sLA
175 * sLA -> sLA Retransmitted FIN. Remain in the same state.
176 * sTW -> sTW
177 * sCL -> sCL
178 */
179 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
180 /*ack*/ { sES, sIV, sES, sES, sCW, sCW, sTW, sTW, sCL, sIV },
181 /*
182 * sNO -> sES Assumed.
183 * sSS -> sIV ACK is invalid: we haven't seen a SYN/ACK yet.
184 * sS2 -> sIV
185 * sSR -> sES Established state is reached.
186 * sES -> sES :-)
187 * sFW -> sCW Normal close request answered by ACK.
188 * sCW -> sCW
189 * sLA -> sTW Last ACK detected (RFC5961 challenged)
190 * sTW -> sTW Retransmitted last ACK. Remain in the same state.
191 * sCL -> sCL
192 */
193 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
194 /*rst*/ { sIV, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL },
195 /*none*/ { sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV }
196 },
197 {
198 /* REPLY */
199 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
200 /*syn*/ { sIV, sS2, sIV, sIV, sIV, sIV, sIV, sSS, sIV, sS2 },
201 /*
202 * sNO -> sIV Never reached.
203 * sSS -> sS2 Simultaneous open
204 * sS2 -> sS2 Retransmitted simultaneous SYN
205 * sSR -> sIV Invalid SYN packets sent by the server
206 * sES -> sIV
207 * sFW -> sIV
208 * sCW -> sIV
209 * sLA -> sIV
210 * sTW -> sSS Reopened connection, but server may have switched role
211 * sCL -> sIV
212 */
213 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
214 /*synack*/ { sIV, sSR, sIG, sIG, sIG, sIG, sIG, sIG, sIG, sSR },
215 /*
216 * sSS -> sSR Standard open.
217 * sS2 -> sSR Simultaneous open
218 * sSR -> sIG Retransmitted SYN/ACK, ignore it.
219 * sES -> sIG Late retransmitted SYN/ACK?
220 * sFW -> sIG Might be SYN/ACK answering ignored SYN
221 * sCW -> sIG
222 * sLA -> sIG
223 * sTW -> sIG
224 * sCL -> sIG
225 */
226 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
227 /*fin*/ { sIV, sIV, sFW, sFW, sLA, sLA, sLA, sTW, sCL, sIV },
228 /*
229 * sSS -> sIV Server might not send FIN in this state.
230 * sS2 -> sIV
231 * sSR -> sFW Close started.
232 * sES -> sFW
233 * sFW -> sLA FIN seen in both directions.
234 * sCW -> sLA
235 * sLA -> sLA Retransmitted FIN.
236 * sTW -> sTW
237 * sCL -> sCL
238 */
239 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
240 /*ack*/ { sIV, sIG, sSR, sES, sCW, sCW, sTW, sTW, sCL, sIG },
241 /*
242 * sSS -> sIG Might be a half-open connection.
243 * sS2 -> sIG
244 * sSR -> sSR Might answer late resent SYN.
245 * sES -> sES :-)
246 * sFW -> sCW Normal close request answered by ACK.
247 * sCW -> sCW
248 * sLA -> sTW Last ACK detected (RFC5961 challenged)
249 * sTW -> sTW Retransmitted last ACK.
250 * sCL -> sCL
251 */
252 /* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
253 /*rst*/ { sIV, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL },
254 /*none*/ { sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV }
255 }
256 };
257
258 #ifdef CONFIG_NF_CONNTRACK_PROCFS
259 /* Print out the private part of the conntrack. */
tcp_print_conntrack(struct seq_file * s,struct nf_conn * ct)260 static void tcp_print_conntrack(struct seq_file *s, struct nf_conn *ct)
261 {
262 if (test_bit(IPS_OFFLOAD_BIT, &ct->status))
263 return;
264
265 seq_printf(s, "%s ", tcp_conntrack_names[ct->proto.tcp.state]);
266 }
267 #endif
268
get_conntrack_index(const struct tcphdr * tcph)269 static unsigned int get_conntrack_index(const struct tcphdr *tcph)
270 {
271 if (tcph->rst) return TCP_RST_SET;
272 else if (tcph->syn) return (tcph->ack ? TCP_SYNACK_SET : TCP_SYN_SET);
273 else if (tcph->fin) return TCP_FIN_SET;
274 else if (tcph->ack) return TCP_ACK_SET;
275 else return TCP_NONE_SET;
276 }
277
278 /* TCP connection tracking based on 'Real Stateful TCP Packet Filtering
279 in IP Filter' by Guido van Rooij.
280
281 http://www.sane.nl/events/sane2000/papers.html
282 http://www.darkart.com/mirrors/www.obfuscation.org/ipf/
283
284 The boundaries and the conditions are changed according to RFC793:
285 the packet must intersect the window (i.e. segments may be
286 after the right or before the left edge) and thus receivers may ACK
287 segments after the right edge of the window.
288
289 td_maxend = max(sack + max(win,1)) seen in reply packets
290 td_maxwin = max(max(win, 1)) + (sack - ack) seen in sent packets
291 td_maxwin += seq + len - sender.td_maxend
292 if seq + len > sender.td_maxend
293 td_end = max(seq + len) seen in sent packets
294
295 I. Upper bound for valid data: seq <= sender.td_maxend
296 II. Lower bound for valid data: seq + len >= sender.td_end - receiver.td_maxwin
297 III. Upper bound for valid (s)ack: sack <= receiver.td_end
298 IV. Lower bound for valid (s)ack: sack >= receiver.td_end - MAXACKWINDOW
299
300 where sack is the highest right edge of sack block found in the packet
301 or ack in the case of packet without SACK option.
302
303 The upper bound limit for a valid (s)ack is not ignored -
304 we doesn't have to deal with fragments.
305 */
306
segment_seq_plus_len(__u32 seq,size_t len,unsigned int dataoff,const struct tcphdr * tcph)307 static inline __u32 segment_seq_plus_len(__u32 seq,
308 size_t len,
309 unsigned int dataoff,
310 const struct tcphdr *tcph)
311 {
312 /* XXX Should I use payload length field in IP/IPv6 header ?
313 * - YK */
314 return (seq + len - dataoff - tcph->doff*4
315 + (tcph->syn ? 1 : 0) + (tcph->fin ? 1 : 0));
316 }
317
318 /* Fixme: what about big packets? */
319 #define MAXACKWINCONST 66000
320 #define MAXACKWINDOW(sender) \
321 ((sender)->td_maxwin > MAXACKWINCONST ? (sender)->td_maxwin \
322 : MAXACKWINCONST)
323
324 /*
325 * Simplified tcp_parse_options routine from tcp_input.c
326 */
tcp_options(const struct sk_buff * skb,unsigned int dataoff,const struct tcphdr * tcph,struct ip_ct_tcp_state * state)327 static void tcp_options(const struct sk_buff *skb,
328 unsigned int dataoff,
329 const struct tcphdr *tcph,
330 struct ip_ct_tcp_state *state)
331 {
332 unsigned char buff[(15 * 4) - sizeof(struct tcphdr)];
333 const unsigned char *ptr;
334 int length = (tcph->doff*4) - sizeof(struct tcphdr);
335
336 if (!length)
337 return;
338
339 ptr = skb_header_pointer(skb, dataoff + sizeof(struct tcphdr),
340 length, buff);
341 if (!ptr)
342 return;
343
344 state->td_scale = 0;
345 state->flags &= IP_CT_TCP_FLAG_BE_LIBERAL;
346
347 while (length > 0) {
348 int opcode=*ptr++;
349 int opsize;
350
351 switch (opcode) {
352 case TCPOPT_EOL:
353 return;
354 case TCPOPT_NOP: /* Ref: RFC 793 section 3.1 */
355 length--;
356 continue;
357 default:
358 if (length < 2)
359 return;
360 opsize=*ptr++;
361 if (opsize < 2) /* "silly options" */
362 return;
363 if (opsize > length)
364 return; /* don't parse partial options */
365
366 if (opcode == TCPOPT_SACK_PERM
367 && opsize == TCPOLEN_SACK_PERM)
368 state->flags |= IP_CT_TCP_FLAG_SACK_PERM;
369 else if (opcode == TCPOPT_WINDOW
370 && opsize == TCPOLEN_WINDOW) {
371 state->td_scale = *(u_int8_t *)ptr;
372
373 if (state->td_scale > TCP_MAX_WSCALE)
374 state->td_scale = TCP_MAX_WSCALE;
375
376 state->flags |=
377 IP_CT_TCP_FLAG_WINDOW_SCALE;
378 }
379 ptr += opsize - 2;
380 length -= opsize;
381 }
382 }
383 }
384
tcp_sack(const struct sk_buff * skb,unsigned int dataoff,const struct tcphdr * tcph,__u32 * sack)385 static void tcp_sack(const struct sk_buff *skb, unsigned int dataoff,
386 const struct tcphdr *tcph, __u32 *sack)
387 {
388 unsigned char buff[(15 * 4) - sizeof(struct tcphdr)];
389 const unsigned char *ptr;
390 int length = (tcph->doff*4) - sizeof(struct tcphdr);
391 __u32 tmp;
392
393 if (!length)
394 return;
395
396 ptr = skb_header_pointer(skb, dataoff + sizeof(struct tcphdr),
397 length, buff);
398 if (!ptr)
399 return;
400
401 /* Fast path for timestamp-only option */
402 if (length == TCPOLEN_TSTAMP_ALIGNED
403 && *(__be32 *)ptr == htonl((TCPOPT_NOP << 24)
404 | (TCPOPT_NOP << 16)
405 | (TCPOPT_TIMESTAMP << 8)
406 | TCPOLEN_TIMESTAMP))
407 return;
408
409 while (length > 0) {
410 int opcode = *ptr++;
411 int opsize, i;
412
413 switch (opcode) {
414 case TCPOPT_EOL:
415 return;
416 case TCPOPT_NOP: /* Ref: RFC 793 section 3.1 */
417 length--;
418 continue;
419 default:
420 if (length < 2)
421 return;
422 opsize = *ptr++;
423 if (opsize < 2) /* "silly options" */
424 return;
425 if (opsize > length)
426 return; /* don't parse partial options */
427
428 if (opcode == TCPOPT_SACK
429 && opsize >= (TCPOLEN_SACK_BASE
430 + TCPOLEN_SACK_PERBLOCK)
431 && !((opsize - TCPOLEN_SACK_BASE)
432 % TCPOLEN_SACK_PERBLOCK)) {
433 for (i = 0;
434 i < (opsize - TCPOLEN_SACK_BASE);
435 i += TCPOLEN_SACK_PERBLOCK) {
436 tmp = get_unaligned_be32((__be32 *)(ptr+i)+1);
437
438 if (after(tmp, *sack))
439 *sack = tmp;
440 }
441 return;
442 }
443 ptr += opsize - 2;
444 length -= opsize;
445 }
446 }
447 }
448
tcp_in_window(struct nf_conn * ct,enum ip_conntrack_dir dir,unsigned int index,const struct sk_buff * skb,unsigned int dataoff,const struct tcphdr * tcph,const struct nf_hook_state * hook_state)449 static bool tcp_in_window(struct nf_conn *ct,
450 enum ip_conntrack_dir dir,
451 unsigned int index,
452 const struct sk_buff *skb,
453 unsigned int dataoff,
454 const struct tcphdr *tcph,
455 const struct nf_hook_state *hook_state)
456 {
457 struct ip_ct_tcp *state = &ct->proto.tcp;
458 struct net *net = nf_ct_net(ct);
459 struct nf_tcp_net *tn = nf_tcp_pernet(net);
460 struct ip_ct_tcp_state *sender = &state->seen[dir];
461 struct ip_ct_tcp_state *receiver = &state->seen[!dir];
462 const struct nf_conntrack_tuple *tuple = &ct->tuplehash[dir].tuple;
463 __u32 seq, ack, sack, end, win, swin;
464 u16 win_raw;
465 s32 receiver_offset;
466 bool res, in_recv_win;
467
468 /*
469 * Get the required data from the packet.
470 */
471 seq = ntohl(tcph->seq);
472 ack = sack = ntohl(tcph->ack_seq);
473 win_raw = ntohs(tcph->window);
474 win = win_raw;
475 end = segment_seq_plus_len(seq, skb->len, dataoff, tcph);
476
477 if (receiver->flags & IP_CT_TCP_FLAG_SACK_PERM)
478 tcp_sack(skb, dataoff, tcph, &sack);
479
480 /* Take into account NAT sequence number mangling */
481 receiver_offset = nf_ct_seq_offset(ct, !dir, ack - 1);
482 ack -= receiver_offset;
483 sack -= receiver_offset;
484
485 pr_debug("tcp_in_window: START\n");
486 pr_debug("tcp_in_window: ");
487 nf_ct_dump_tuple(tuple);
488 pr_debug("seq=%u ack=%u+(%d) sack=%u+(%d) win=%u end=%u\n",
489 seq, ack, receiver_offset, sack, receiver_offset, win, end);
490 pr_debug("tcp_in_window: sender end=%u maxend=%u maxwin=%u scale=%i "
491 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
492 sender->td_end, sender->td_maxend, sender->td_maxwin,
493 sender->td_scale,
494 receiver->td_end, receiver->td_maxend, receiver->td_maxwin,
495 receiver->td_scale);
496
497 if (sender->td_maxwin == 0) {
498 /*
499 * Initialize sender data.
500 */
501 if (tcph->syn) {
502 /*
503 * SYN-ACK in reply to a SYN
504 * or SYN from reply direction in simultaneous open.
505 */
506 sender->td_end =
507 sender->td_maxend = end;
508 sender->td_maxwin = (win == 0 ? 1 : win);
509
510 tcp_options(skb, dataoff, tcph, sender);
511 /*
512 * RFC 1323:
513 * Both sides must send the Window Scale option
514 * to enable window scaling in either direction.
515 */
516 if (!(sender->flags & IP_CT_TCP_FLAG_WINDOW_SCALE
517 && receiver->flags & IP_CT_TCP_FLAG_WINDOW_SCALE))
518 sender->td_scale =
519 receiver->td_scale = 0;
520 if (!tcph->ack)
521 /* Simultaneous open */
522 return true;
523 } else {
524 /*
525 * We are in the middle of a connection,
526 * its history is lost for us.
527 * Let's try to use the data from the packet.
528 */
529 sender->td_end = end;
530 swin = win << sender->td_scale;
531 sender->td_maxwin = (swin == 0 ? 1 : swin);
532 sender->td_maxend = end + sender->td_maxwin;
533 if (receiver->td_maxwin == 0) {
534 /* We haven't seen traffic in the other
535 * direction yet but we have to tweak window
536 * tracking to pass III and IV until that
537 * happens.
538 */
539 receiver->td_end = receiver->td_maxend = sack;
540 } else if (sack == receiver->td_end + 1) {
541 /* Likely a reply to a keepalive.
542 * Needed for III.
543 */
544 receiver->td_end++;
545 }
546
547 }
548 } else if (((state->state == TCP_CONNTRACK_SYN_SENT
549 && dir == IP_CT_DIR_ORIGINAL)
550 || (state->state == TCP_CONNTRACK_SYN_RECV
551 && dir == IP_CT_DIR_REPLY))
552 && after(end, sender->td_end)) {
553 /*
554 * RFC 793: "if a TCP is reinitialized ... then it need
555 * not wait at all; it must only be sure to use sequence
556 * numbers larger than those recently used."
557 */
558 sender->td_end =
559 sender->td_maxend = end;
560 sender->td_maxwin = (win == 0 ? 1 : win);
561
562 tcp_options(skb, dataoff, tcph, sender);
563 }
564
565 if (!(tcph->ack)) {
566 /*
567 * If there is no ACK, just pretend it was set and OK.
568 */
569 ack = sack = receiver->td_end;
570 } else if (((tcp_flag_word(tcph) & (TCP_FLAG_ACK|TCP_FLAG_RST)) ==
571 (TCP_FLAG_ACK|TCP_FLAG_RST))
572 && (ack == 0)) {
573 /*
574 * Broken TCP stacks, that set ACK in RST packets as well
575 * with zero ack value.
576 */
577 ack = sack = receiver->td_end;
578 }
579
580 if (tcph->rst && seq == 0 && state->state == TCP_CONNTRACK_SYN_SENT)
581 /*
582 * RST sent answering SYN.
583 */
584 seq = end = sender->td_end;
585
586 pr_debug("tcp_in_window: ");
587 nf_ct_dump_tuple(tuple);
588 pr_debug("seq=%u ack=%u+(%d) sack=%u+(%d) win=%u end=%u\n",
589 seq, ack, receiver_offset, sack, receiver_offset, win, end);
590 pr_debug("tcp_in_window: sender end=%u maxend=%u maxwin=%u scale=%i "
591 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
592 sender->td_end, sender->td_maxend, sender->td_maxwin,
593 sender->td_scale,
594 receiver->td_end, receiver->td_maxend, receiver->td_maxwin,
595 receiver->td_scale);
596
597 /* Is the ending sequence in the receive window (if available)? */
598 in_recv_win = !receiver->td_maxwin ||
599 after(end, sender->td_end - receiver->td_maxwin - 1);
600
601 pr_debug("tcp_in_window: I=%i II=%i III=%i IV=%i\n",
602 before(seq, sender->td_maxend + 1),
603 (in_recv_win ? 1 : 0),
604 before(sack, receiver->td_end + 1),
605 after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1));
606
607 if (before(seq, sender->td_maxend + 1) &&
608 in_recv_win &&
609 before(sack, receiver->td_end + 1) &&
610 after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1)) {
611 /*
612 * Take into account window scaling (RFC 1323).
613 */
614 if (!tcph->syn)
615 win <<= sender->td_scale;
616
617 /*
618 * Update sender data.
619 */
620 swin = win + (sack - ack);
621 if (sender->td_maxwin < swin)
622 sender->td_maxwin = swin;
623 if (after(end, sender->td_end)) {
624 sender->td_end = end;
625 sender->flags |= IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED;
626 }
627 if (tcph->ack) {
628 if (!(sender->flags & IP_CT_TCP_FLAG_MAXACK_SET)) {
629 sender->td_maxack = ack;
630 sender->flags |= IP_CT_TCP_FLAG_MAXACK_SET;
631 } else if (after(ack, sender->td_maxack))
632 sender->td_maxack = ack;
633 }
634
635 /*
636 * Update receiver data.
637 */
638 if (receiver->td_maxwin != 0 && after(end, sender->td_maxend))
639 receiver->td_maxwin += end - sender->td_maxend;
640 if (after(sack + win, receiver->td_maxend - 1)) {
641 receiver->td_maxend = sack + win;
642 if (win == 0)
643 receiver->td_maxend++;
644 }
645 if (ack == receiver->td_end)
646 receiver->flags &= ~IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED;
647
648 /*
649 * Check retransmissions.
650 */
651 if (index == TCP_ACK_SET) {
652 if (state->last_dir == dir
653 && state->last_seq == seq
654 && state->last_ack == ack
655 && state->last_end == end
656 && state->last_win == win_raw)
657 state->retrans++;
658 else {
659 state->last_dir = dir;
660 state->last_seq = seq;
661 state->last_ack = ack;
662 state->last_end = end;
663 state->last_win = win_raw;
664 state->retrans = 0;
665 }
666 }
667 res = true;
668 } else {
669 res = false;
670 if (sender->flags & IP_CT_TCP_FLAG_BE_LIBERAL ||
671 tn->tcp_be_liberal)
672 res = true;
673 if (!res) {
674 bool seq_ok = before(seq, sender->td_maxend + 1);
675
676 if (!seq_ok) {
677 u32 overshot = end - sender->td_maxend + 1;
678 bool ack_ok;
679
680 ack_ok = after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1);
681
682 if (in_recv_win &&
683 ack_ok &&
684 overshot <= receiver->td_maxwin &&
685 before(sack, receiver->td_end + 1)) {
686 /* Work around TCPs that send more bytes than allowed by
687 * the receive window.
688 *
689 * If the (marked as invalid) packet is allowed to pass by
690 * the ruleset and the peer acks this data, then its possible
691 * all future packets will trigger 'ACK is over upper bound' check.
692 *
693 * Thus if only the sequence check fails then do update td_end so
694 * possible ACK for this data can update internal state.
695 */
696 sender->td_end = end;
697 sender->flags |= IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED;
698
699 nf_ct_l4proto_log_invalid(skb, ct, hook_state,
700 "%u bytes more than expected", overshot);
701 return res;
702 }
703 }
704
705 nf_ct_l4proto_log_invalid(skb, ct, hook_state,
706 "%s",
707 before(seq, sender->td_maxend + 1) ?
708 in_recv_win ?
709 before(sack, receiver->td_end + 1) ?
710 after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1) ? "BUG"
711 : "ACK is under the lower bound (possible overly delayed ACK)"
712 : "ACK is over the upper bound (ACKed data not seen yet)"
713 : "SEQ is under the lower bound (already ACKed data retransmitted)"
714 : "SEQ is over the upper bound (over the window of the receiver)");
715 }
716 }
717
718 pr_debug("tcp_in_window: res=%u sender end=%u maxend=%u maxwin=%u "
719 "receiver end=%u maxend=%u maxwin=%u\n",
720 res, sender->td_end, sender->td_maxend, sender->td_maxwin,
721 receiver->td_end, receiver->td_maxend, receiver->td_maxwin);
722
723 return res;
724 }
725
726 /* table of valid flag combinations - PUSH, ECE and CWR are always valid */
727 static const u8 tcp_valid_flags[(TCPHDR_FIN|TCPHDR_SYN|TCPHDR_RST|TCPHDR_ACK|
728 TCPHDR_URG) + 1] =
729 {
730 [TCPHDR_SYN] = 1,
731 [TCPHDR_SYN|TCPHDR_URG] = 1,
732 [TCPHDR_SYN|TCPHDR_ACK] = 1,
733 [TCPHDR_RST] = 1,
734 [TCPHDR_RST|TCPHDR_ACK] = 1,
735 [TCPHDR_FIN|TCPHDR_ACK] = 1,
736 [TCPHDR_FIN|TCPHDR_ACK|TCPHDR_URG] = 1,
737 [TCPHDR_ACK] = 1,
738 [TCPHDR_ACK|TCPHDR_URG] = 1,
739 };
740
tcp_error_log(const struct sk_buff * skb,const struct nf_hook_state * state,const char * msg)741 static void tcp_error_log(const struct sk_buff *skb,
742 const struct nf_hook_state *state,
743 const char *msg)
744 {
745 nf_l4proto_log_invalid(skb, state, IPPROTO_TCP, "%s", msg);
746 }
747
748 /* Protect conntrack agaist broken packets. Code taken from ipt_unclean.c. */
tcp_error(const struct tcphdr * th,struct sk_buff * skb,unsigned int dataoff,const struct nf_hook_state * state)749 static bool tcp_error(const struct tcphdr *th,
750 struct sk_buff *skb,
751 unsigned int dataoff,
752 const struct nf_hook_state *state)
753 {
754 unsigned int tcplen = skb->len - dataoff;
755 u8 tcpflags;
756
757 /* Not whole TCP header or malformed packet */
758 if (th->doff*4 < sizeof(struct tcphdr) || tcplen < th->doff*4) {
759 tcp_error_log(skb, state, "truncated packet");
760 return true;
761 }
762
763 /* Checksum invalid? Ignore.
764 * We skip checking packets on the outgoing path
765 * because the checksum is assumed to be correct.
766 */
767 /* FIXME: Source route IP option packets --RR */
768 if (state->net->ct.sysctl_checksum &&
769 state->hook == NF_INET_PRE_ROUTING &&
770 nf_checksum(skb, state->hook, dataoff, IPPROTO_TCP, state->pf)) {
771 tcp_error_log(skb, state, "bad checksum");
772 return true;
773 }
774
775 /* Check TCP flags. */
776 tcpflags = (tcp_flag_byte(th) & ~(TCPHDR_ECE|TCPHDR_CWR|TCPHDR_PSH));
777 if (!tcp_valid_flags[tcpflags]) {
778 tcp_error_log(skb, state, "invalid tcp flag combination");
779 return true;
780 }
781
782 return false;
783 }
784
tcp_new(struct nf_conn * ct,const struct sk_buff * skb,unsigned int dataoff,const struct tcphdr * th)785 static noinline bool tcp_new(struct nf_conn *ct, const struct sk_buff *skb,
786 unsigned int dataoff,
787 const struct tcphdr *th)
788 {
789 enum tcp_conntrack new_state;
790 struct net *net = nf_ct_net(ct);
791 const struct nf_tcp_net *tn = nf_tcp_pernet(net);
792 const struct ip_ct_tcp_state *sender = &ct->proto.tcp.seen[0];
793 const struct ip_ct_tcp_state *receiver = &ct->proto.tcp.seen[1];
794
795 /* Don't need lock here: this conntrack not in circulation yet */
796 new_state = tcp_conntracks[0][get_conntrack_index(th)][TCP_CONNTRACK_NONE];
797
798 /* Invalid: delete conntrack */
799 if (new_state >= TCP_CONNTRACK_MAX) {
800 pr_debug("nf_ct_tcp: invalid new deleting.\n");
801 return false;
802 }
803
804 if (new_state == TCP_CONNTRACK_SYN_SENT) {
805 memset(&ct->proto.tcp, 0, sizeof(ct->proto.tcp));
806 /* SYN packet */
807 ct->proto.tcp.seen[0].td_end =
808 segment_seq_plus_len(ntohl(th->seq), skb->len,
809 dataoff, th);
810 ct->proto.tcp.seen[0].td_maxwin = ntohs(th->window);
811 if (ct->proto.tcp.seen[0].td_maxwin == 0)
812 ct->proto.tcp.seen[0].td_maxwin = 1;
813 ct->proto.tcp.seen[0].td_maxend =
814 ct->proto.tcp.seen[0].td_end;
815
816 tcp_options(skb, dataoff, th, &ct->proto.tcp.seen[0]);
817 } else if (tn->tcp_loose == 0) {
818 /* Don't try to pick up connections. */
819 return false;
820 } else {
821 memset(&ct->proto.tcp, 0, sizeof(ct->proto.tcp));
822 /*
823 * We are in the middle of a connection,
824 * its history is lost for us.
825 * Let's try to use the data from the packet.
826 */
827 ct->proto.tcp.seen[0].td_end =
828 segment_seq_plus_len(ntohl(th->seq), skb->len,
829 dataoff, th);
830 ct->proto.tcp.seen[0].td_maxwin = ntohs(th->window);
831 if (ct->proto.tcp.seen[0].td_maxwin == 0)
832 ct->proto.tcp.seen[0].td_maxwin = 1;
833 ct->proto.tcp.seen[0].td_maxend =
834 ct->proto.tcp.seen[0].td_end +
835 ct->proto.tcp.seen[0].td_maxwin;
836
837 /* We assume SACK and liberal window checking to handle
838 * window scaling */
839 ct->proto.tcp.seen[0].flags =
840 ct->proto.tcp.seen[1].flags = IP_CT_TCP_FLAG_SACK_PERM |
841 IP_CT_TCP_FLAG_BE_LIBERAL;
842 }
843
844 /* tcp_packet will set them */
845 ct->proto.tcp.last_index = TCP_NONE_SET;
846
847 pr_debug("%s: sender end=%u maxend=%u maxwin=%u scale=%i "
848 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
849 __func__,
850 sender->td_end, sender->td_maxend, sender->td_maxwin,
851 sender->td_scale,
852 receiver->td_end, receiver->td_maxend, receiver->td_maxwin,
853 receiver->td_scale);
854 return true;
855 }
856
tcp_can_early_drop(const struct nf_conn * ct)857 static bool tcp_can_early_drop(const struct nf_conn *ct)
858 {
859 switch (ct->proto.tcp.state) {
860 case TCP_CONNTRACK_FIN_WAIT:
861 case TCP_CONNTRACK_LAST_ACK:
862 case TCP_CONNTRACK_TIME_WAIT:
863 case TCP_CONNTRACK_CLOSE:
864 case TCP_CONNTRACK_CLOSE_WAIT:
865 return true;
866 default:
867 break;
868 }
869
870 return false;
871 }
872
nf_conntrack_tcp_set_closing(struct nf_conn * ct)873 void nf_conntrack_tcp_set_closing(struct nf_conn *ct)
874 {
875 enum tcp_conntrack old_state;
876 const unsigned int *timeouts;
877 u32 timeout;
878
879 if (!nf_ct_is_confirmed(ct))
880 return;
881
882 spin_lock_bh(&ct->lock);
883 old_state = ct->proto.tcp.state;
884 ct->proto.tcp.state = TCP_CONNTRACK_CLOSE;
885
886 if (old_state == TCP_CONNTRACK_CLOSE ||
887 test_bit(IPS_FIXED_TIMEOUT_BIT, &ct->status)) {
888 spin_unlock_bh(&ct->lock);
889 return;
890 }
891
892 timeouts = nf_ct_timeout_lookup(ct);
893 if (!timeouts) {
894 const struct nf_tcp_net *tn;
895
896 tn = nf_tcp_pernet(nf_ct_net(ct));
897 timeouts = tn->timeouts;
898 }
899
900 timeout = timeouts[TCP_CONNTRACK_CLOSE];
901 WRITE_ONCE(ct->timeout, timeout + nfct_time_stamp);
902
903 spin_unlock_bh(&ct->lock);
904
905 nf_conntrack_event_cache(IPCT_PROTOINFO, ct);
906 }
907
nf_ct_tcp_state_reset(struct ip_ct_tcp_state * state)908 static void nf_ct_tcp_state_reset(struct ip_ct_tcp_state *state)
909 {
910 state->td_end = 0;
911 state->td_maxend = 0;
912 state->td_maxwin = 0;
913 state->td_maxack = 0;
914 state->td_scale = 0;
915 state->flags &= IP_CT_TCP_FLAG_BE_LIBERAL;
916 }
917
918 /* Returns verdict for packet, or -1 for invalid. */
nf_conntrack_tcp_packet(struct nf_conn * ct,struct sk_buff * skb,unsigned int dataoff,enum ip_conntrack_info ctinfo,const struct nf_hook_state * state)919 int nf_conntrack_tcp_packet(struct nf_conn *ct,
920 struct sk_buff *skb,
921 unsigned int dataoff,
922 enum ip_conntrack_info ctinfo,
923 const struct nf_hook_state *state)
924 {
925 struct net *net = nf_ct_net(ct);
926 struct nf_tcp_net *tn = nf_tcp_pernet(net);
927 struct nf_conntrack_tuple *tuple;
928 enum tcp_conntrack new_state, old_state;
929 unsigned int index, *timeouts;
930 enum ip_conntrack_dir dir;
931 const struct tcphdr *th;
932 struct tcphdr _tcph;
933 unsigned long timeout;
934
935 th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph);
936 if (th == NULL)
937 return -NF_ACCEPT;
938
939 if (tcp_error(th, skb, dataoff, state))
940 return -NF_ACCEPT;
941
942 if (!nf_ct_is_confirmed(ct) && !tcp_new(ct, skb, dataoff, th))
943 return -NF_ACCEPT;
944
945 spin_lock_bh(&ct->lock);
946 old_state = ct->proto.tcp.state;
947 dir = CTINFO2DIR(ctinfo);
948 index = get_conntrack_index(th);
949 new_state = tcp_conntracks[dir][index][old_state];
950 tuple = &ct->tuplehash[dir].tuple;
951
952 switch (new_state) {
953 case TCP_CONNTRACK_SYN_SENT:
954 if (old_state < TCP_CONNTRACK_TIME_WAIT)
955 break;
956 /* RFC 1122: "When a connection is closed actively,
957 * it MUST linger in TIME-WAIT state for a time 2xMSL
958 * (Maximum Segment Lifetime). However, it MAY accept
959 * a new SYN from the remote TCP to reopen the connection
960 * directly from TIME-WAIT state, if..."
961 * We ignore the conditions because we are in the
962 * TIME-WAIT state anyway.
963 *
964 * Handle aborted connections: we and the server
965 * think there is an existing connection but the client
966 * aborts it and starts a new one.
967 */
968 if (((ct->proto.tcp.seen[dir].flags
969 | ct->proto.tcp.seen[!dir].flags)
970 & IP_CT_TCP_FLAG_CLOSE_INIT)
971 || (ct->proto.tcp.last_dir == dir
972 && ct->proto.tcp.last_index == TCP_RST_SET)) {
973 /* Attempt to reopen a closed/aborted connection.
974 * Delete this connection and look up again. */
975 spin_unlock_bh(&ct->lock);
976
977 /* Only repeat if we can actually remove the timer.
978 * Destruction may already be in progress in process
979 * context and we must give it a chance to terminate.
980 */
981 if (nf_ct_kill(ct))
982 return -NF_REPEAT;
983 return NF_DROP;
984 }
985 fallthrough;
986 case TCP_CONNTRACK_IGNORE:
987 /* Ignored packets:
988 *
989 * Our connection entry may be out of sync, so ignore
990 * packets which may signal the real connection between
991 * the client and the server.
992 *
993 * a) SYN in ORIGINAL
994 * b) SYN/ACK in REPLY
995 * c) ACK in reply direction after initial SYN in original.
996 *
997 * If the ignored packet is invalid, the receiver will send
998 * a RST we'll catch below.
999 */
1000 if (index == TCP_SYNACK_SET
1001 && ct->proto.tcp.last_index == TCP_SYN_SET
1002 && ct->proto.tcp.last_dir != dir
1003 && ntohl(th->ack_seq) == ct->proto.tcp.last_end) {
1004 /* b) This SYN/ACK acknowledges a SYN that we earlier
1005 * ignored as invalid. This means that the client and
1006 * the server are both in sync, while the firewall is
1007 * not. We get in sync from the previously annotated
1008 * values.
1009 */
1010 old_state = TCP_CONNTRACK_SYN_SENT;
1011 new_state = TCP_CONNTRACK_SYN_RECV;
1012 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_end =
1013 ct->proto.tcp.last_end;
1014 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_maxend =
1015 ct->proto.tcp.last_end;
1016 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_maxwin =
1017 ct->proto.tcp.last_win == 0 ?
1018 1 : ct->proto.tcp.last_win;
1019 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_scale =
1020 ct->proto.tcp.last_wscale;
1021 ct->proto.tcp.last_flags &= ~IP_CT_EXP_CHALLENGE_ACK;
1022 ct->proto.tcp.seen[ct->proto.tcp.last_dir].flags =
1023 ct->proto.tcp.last_flags;
1024 nf_ct_tcp_state_reset(&ct->proto.tcp.seen[dir]);
1025 break;
1026 }
1027 ct->proto.tcp.last_index = index;
1028 ct->proto.tcp.last_dir = dir;
1029 ct->proto.tcp.last_seq = ntohl(th->seq);
1030 ct->proto.tcp.last_end =
1031 segment_seq_plus_len(ntohl(th->seq), skb->len, dataoff, th);
1032 ct->proto.tcp.last_win = ntohs(th->window);
1033
1034 /* a) This is a SYN in ORIGINAL. The client and the server
1035 * may be in sync but we are not. In that case, we annotate
1036 * the TCP options and let the packet go through. If it is a
1037 * valid SYN packet, the server will reply with a SYN/ACK, and
1038 * then we'll get in sync. Otherwise, the server potentially
1039 * responds with a challenge ACK if implementing RFC5961.
1040 */
1041 if (index == TCP_SYN_SET && dir == IP_CT_DIR_ORIGINAL) {
1042 struct ip_ct_tcp_state seen = {};
1043
1044 ct->proto.tcp.last_flags =
1045 ct->proto.tcp.last_wscale = 0;
1046 tcp_options(skb, dataoff, th, &seen);
1047 if (seen.flags & IP_CT_TCP_FLAG_WINDOW_SCALE) {
1048 ct->proto.tcp.last_flags |=
1049 IP_CT_TCP_FLAG_WINDOW_SCALE;
1050 ct->proto.tcp.last_wscale = seen.td_scale;
1051 }
1052 if (seen.flags & IP_CT_TCP_FLAG_SACK_PERM) {
1053 ct->proto.tcp.last_flags |=
1054 IP_CT_TCP_FLAG_SACK_PERM;
1055 }
1056 /* Mark the potential for RFC5961 challenge ACK,
1057 * this pose a special problem for LAST_ACK state
1058 * as ACK is intrepretated as ACKing last FIN.
1059 */
1060 if (old_state == TCP_CONNTRACK_LAST_ACK)
1061 ct->proto.tcp.last_flags |=
1062 IP_CT_EXP_CHALLENGE_ACK;
1063 }
1064 spin_unlock_bh(&ct->lock);
1065 nf_ct_l4proto_log_invalid(skb, ct, state,
1066 "packet (index %d) in dir %d ignored, state %s",
1067 index, dir,
1068 tcp_conntrack_names[old_state]);
1069 return NF_ACCEPT;
1070 case TCP_CONNTRACK_MAX:
1071 /* Special case for SYN proxy: when the SYN to the server or
1072 * the SYN/ACK from the server is lost, the client may transmit
1073 * a keep-alive packet while in SYN_SENT state. This needs to
1074 * be associated with the original conntrack entry in order to
1075 * generate a new SYN with the correct sequence number.
1076 */
1077 if (nfct_synproxy(ct) && old_state == TCP_CONNTRACK_SYN_SENT &&
1078 index == TCP_ACK_SET && dir == IP_CT_DIR_ORIGINAL &&
1079 ct->proto.tcp.last_dir == IP_CT_DIR_ORIGINAL &&
1080 ct->proto.tcp.seen[dir].td_end - 1 == ntohl(th->seq)) {
1081 pr_debug("nf_ct_tcp: SYN proxy client keep alive\n");
1082 spin_unlock_bh(&ct->lock);
1083 return NF_ACCEPT;
1084 }
1085
1086 /* Invalid packet */
1087 pr_debug("nf_ct_tcp: Invalid dir=%i index=%u ostate=%u\n",
1088 dir, get_conntrack_index(th), old_state);
1089 spin_unlock_bh(&ct->lock);
1090 nf_ct_l4proto_log_invalid(skb, ct, state, "invalid state");
1091 return -NF_ACCEPT;
1092 case TCP_CONNTRACK_TIME_WAIT:
1093 /* RFC5961 compliance cause stack to send "challenge-ACK"
1094 * e.g. in response to spurious SYNs. Conntrack MUST
1095 * not believe this ACK is acking last FIN.
1096 */
1097 if (old_state == TCP_CONNTRACK_LAST_ACK &&
1098 index == TCP_ACK_SET &&
1099 ct->proto.tcp.last_dir != dir &&
1100 ct->proto.tcp.last_index == TCP_SYN_SET &&
1101 (ct->proto.tcp.last_flags & IP_CT_EXP_CHALLENGE_ACK)) {
1102 /* Detected RFC5961 challenge ACK */
1103 ct->proto.tcp.last_flags &= ~IP_CT_EXP_CHALLENGE_ACK;
1104 spin_unlock_bh(&ct->lock);
1105 nf_ct_l4proto_log_invalid(skb, ct, state, "challenge-ack ignored");
1106 return NF_ACCEPT; /* Don't change state */
1107 }
1108 break;
1109 case TCP_CONNTRACK_SYN_SENT2:
1110 /* tcp_conntracks table is not smart enough to handle
1111 * simultaneous open.
1112 */
1113 ct->proto.tcp.last_flags |= IP_CT_TCP_SIMULTANEOUS_OPEN;
1114 break;
1115 case TCP_CONNTRACK_SYN_RECV:
1116 if (dir == IP_CT_DIR_REPLY && index == TCP_ACK_SET &&
1117 ct->proto.tcp.last_flags & IP_CT_TCP_SIMULTANEOUS_OPEN)
1118 new_state = TCP_CONNTRACK_ESTABLISHED;
1119 break;
1120 case TCP_CONNTRACK_CLOSE:
1121 if (index != TCP_RST_SET)
1122 break;
1123
1124 /* If we are closing, tuple might have been re-used already.
1125 * last_index, last_ack, and all other ct fields used for
1126 * sequence/window validation are outdated in that case.
1127 *
1128 * As the conntrack can already be expired by GC under pressure,
1129 * just skip validation checks.
1130 */
1131 if (tcp_can_early_drop(ct))
1132 goto in_window;
1133
1134 /* td_maxack might be outdated if we let a SYN through earlier */
1135 if ((ct->proto.tcp.seen[!dir].flags & IP_CT_TCP_FLAG_MAXACK_SET) &&
1136 ct->proto.tcp.last_index != TCP_SYN_SET) {
1137 u32 seq = ntohl(th->seq);
1138
1139 /* If we are not in established state and SEQ=0 this is most
1140 * likely an answer to a SYN we let go through above (last_index
1141 * can be updated due to out-of-order ACKs).
1142 */
1143 if (seq == 0 && !nf_conntrack_tcp_established(ct))
1144 break;
1145
1146 if (before(seq, ct->proto.tcp.seen[!dir].td_maxack) &&
1147 !tn->tcp_ignore_invalid_rst) {
1148 /* Invalid RST */
1149 spin_unlock_bh(&ct->lock);
1150 nf_ct_l4proto_log_invalid(skb, ct, state, "invalid rst");
1151 return -NF_ACCEPT;
1152 }
1153
1154 if (!nf_conntrack_tcp_established(ct) ||
1155 seq == ct->proto.tcp.seen[!dir].td_maxack)
1156 break;
1157
1158 /* Check if rst is part of train, such as
1159 * foo:80 > bar:4379: P, 235946583:235946602(19) ack 42
1160 * foo:80 > bar:4379: R, 235946602:235946602(0) ack 42
1161 */
1162 if (ct->proto.tcp.last_index == TCP_ACK_SET &&
1163 ct->proto.tcp.last_dir == dir &&
1164 seq == ct->proto.tcp.last_end)
1165 break;
1166
1167 /* ... RST sequence number doesn't match exactly, keep
1168 * established state to allow a possible challenge ACK.
1169 */
1170 new_state = old_state;
1171 }
1172 if (((test_bit(IPS_SEEN_REPLY_BIT, &ct->status)
1173 && ct->proto.tcp.last_index == TCP_SYN_SET)
1174 || (!test_bit(IPS_ASSURED_BIT, &ct->status)
1175 && ct->proto.tcp.last_index == TCP_ACK_SET))
1176 && ntohl(th->ack_seq) == ct->proto.tcp.last_end) {
1177 /* RST sent to invalid SYN or ACK we had let through
1178 * at a) and c) above:
1179 *
1180 * a) SYN was in window then
1181 * c) we hold a half-open connection.
1182 *
1183 * Delete our connection entry.
1184 * We skip window checking, because packet might ACK
1185 * segments we ignored. */
1186 goto in_window;
1187 }
1188 break;
1189 default:
1190 /* Keep compilers happy. */
1191 break;
1192 }
1193
1194 if (!tcp_in_window(ct, dir, index,
1195 skb, dataoff, th, state)) {
1196 spin_unlock_bh(&ct->lock);
1197 return -NF_ACCEPT;
1198 }
1199 in_window:
1200 /* From now on we have got in-window packets */
1201 ct->proto.tcp.last_index = index;
1202 ct->proto.tcp.last_dir = dir;
1203
1204 pr_debug("tcp_conntracks: ");
1205 nf_ct_dump_tuple(tuple);
1206 pr_debug("syn=%i ack=%i fin=%i rst=%i old=%i new=%i\n",
1207 (th->syn ? 1 : 0), (th->ack ? 1 : 0),
1208 (th->fin ? 1 : 0), (th->rst ? 1 : 0),
1209 old_state, new_state);
1210
1211 ct->proto.tcp.state = new_state;
1212 if (old_state != new_state
1213 && new_state == TCP_CONNTRACK_FIN_WAIT)
1214 ct->proto.tcp.seen[dir].flags |= IP_CT_TCP_FLAG_CLOSE_INIT;
1215
1216 timeouts = nf_ct_timeout_lookup(ct);
1217 if (!timeouts)
1218 timeouts = tn->timeouts;
1219
1220 if (ct->proto.tcp.retrans >= tn->tcp_max_retrans &&
1221 timeouts[new_state] > timeouts[TCP_CONNTRACK_RETRANS])
1222 timeout = timeouts[TCP_CONNTRACK_RETRANS];
1223 else if (unlikely(index == TCP_RST_SET))
1224 timeout = timeouts[TCP_CONNTRACK_CLOSE];
1225 else if ((ct->proto.tcp.seen[0].flags | ct->proto.tcp.seen[1].flags) &
1226 IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED &&
1227 timeouts[new_state] > timeouts[TCP_CONNTRACK_UNACK])
1228 timeout = timeouts[TCP_CONNTRACK_UNACK];
1229 else if (ct->proto.tcp.last_win == 0 &&
1230 timeouts[new_state] > timeouts[TCP_CONNTRACK_RETRANS])
1231 timeout = timeouts[TCP_CONNTRACK_RETRANS];
1232 else
1233 timeout = timeouts[new_state];
1234 spin_unlock_bh(&ct->lock);
1235
1236 if (new_state != old_state)
1237 nf_conntrack_event_cache(IPCT_PROTOINFO, ct);
1238
1239 if (!test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) {
1240 /* If only reply is a RST, we can consider ourselves not to
1241 have an established connection: this is a fairly common
1242 problem case, so we can delete the conntrack
1243 immediately. --RR */
1244 if (th->rst) {
1245 nf_ct_kill_acct(ct, ctinfo, skb);
1246 return NF_ACCEPT;
1247 }
1248
1249 if (index == TCP_SYN_SET && old_state == TCP_CONNTRACK_SYN_SENT) {
1250 /* do not renew timeout on SYN retransmit.
1251 *
1252 * Else port reuse by client or NAT middlebox can keep
1253 * entry alive indefinitely (including nat info).
1254 */
1255 return NF_ACCEPT;
1256 }
1257
1258 /* ESTABLISHED without SEEN_REPLY, i.e. mid-connection
1259 * pickup with loose=1. Avoid large ESTABLISHED timeout.
1260 */
1261 if (new_state == TCP_CONNTRACK_ESTABLISHED &&
1262 timeout > timeouts[TCP_CONNTRACK_UNACK])
1263 timeout = timeouts[TCP_CONNTRACK_UNACK];
1264 } else if (!test_bit(IPS_ASSURED_BIT, &ct->status)
1265 && (old_state == TCP_CONNTRACK_SYN_RECV
1266 || old_state == TCP_CONNTRACK_ESTABLISHED)
1267 && new_state == TCP_CONNTRACK_ESTABLISHED) {
1268 /* Set ASSURED if we see valid ack in ESTABLISHED
1269 after SYN_RECV or a valid answer for a picked up
1270 connection. */
1271 set_bit(IPS_ASSURED_BIT, &ct->status);
1272 nf_conntrack_event_cache(IPCT_ASSURED, ct);
1273 }
1274 nf_ct_refresh_acct(ct, ctinfo, skb, timeout);
1275
1276 return NF_ACCEPT;
1277 }
1278
1279 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
1280
1281 #include <linux/netfilter/nfnetlink.h>
1282 #include <linux/netfilter/nfnetlink_conntrack.h>
1283
tcp_to_nlattr(struct sk_buff * skb,struct nlattr * nla,struct nf_conn * ct,bool destroy)1284 static int tcp_to_nlattr(struct sk_buff *skb, struct nlattr *nla,
1285 struct nf_conn *ct, bool destroy)
1286 {
1287 struct nlattr *nest_parms;
1288 struct nf_ct_tcp_flags tmp = {};
1289
1290 spin_lock_bh(&ct->lock);
1291 nest_parms = nla_nest_start(skb, CTA_PROTOINFO_TCP);
1292 if (!nest_parms)
1293 goto nla_put_failure;
1294
1295 if (nla_put_u8(skb, CTA_PROTOINFO_TCP_STATE, ct->proto.tcp.state))
1296 goto nla_put_failure;
1297
1298 if (destroy)
1299 goto skip_state;
1300
1301 if (nla_put_u8(skb, CTA_PROTOINFO_TCP_WSCALE_ORIGINAL,
1302 ct->proto.tcp.seen[0].td_scale) ||
1303 nla_put_u8(skb, CTA_PROTOINFO_TCP_WSCALE_REPLY,
1304 ct->proto.tcp.seen[1].td_scale))
1305 goto nla_put_failure;
1306
1307 tmp.flags = ct->proto.tcp.seen[0].flags;
1308 if (nla_put(skb, CTA_PROTOINFO_TCP_FLAGS_ORIGINAL,
1309 sizeof(struct nf_ct_tcp_flags), &tmp))
1310 goto nla_put_failure;
1311
1312 tmp.flags = ct->proto.tcp.seen[1].flags;
1313 if (nla_put(skb, CTA_PROTOINFO_TCP_FLAGS_REPLY,
1314 sizeof(struct nf_ct_tcp_flags), &tmp))
1315 goto nla_put_failure;
1316 skip_state:
1317 spin_unlock_bh(&ct->lock);
1318 nla_nest_end(skb, nest_parms);
1319
1320 return 0;
1321
1322 nla_put_failure:
1323 spin_unlock_bh(&ct->lock);
1324 return -1;
1325 }
1326
1327 static const struct nla_policy tcp_nla_policy[CTA_PROTOINFO_TCP_MAX+1] = {
1328 [CTA_PROTOINFO_TCP_STATE] = { .type = NLA_U8 },
1329 [CTA_PROTOINFO_TCP_WSCALE_ORIGINAL] = { .type = NLA_U8 },
1330 [CTA_PROTOINFO_TCP_WSCALE_REPLY] = { .type = NLA_U8 },
1331 [CTA_PROTOINFO_TCP_FLAGS_ORIGINAL] = { .len = sizeof(struct nf_ct_tcp_flags) },
1332 [CTA_PROTOINFO_TCP_FLAGS_REPLY] = { .len = sizeof(struct nf_ct_tcp_flags) },
1333 };
1334
1335 #define TCP_NLATTR_SIZE ( \
1336 NLA_ALIGN(NLA_HDRLEN + 1) + \
1337 NLA_ALIGN(NLA_HDRLEN + 1) + \
1338 NLA_ALIGN(NLA_HDRLEN + sizeof(struct nf_ct_tcp_flags)) + \
1339 NLA_ALIGN(NLA_HDRLEN + sizeof(struct nf_ct_tcp_flags)))
1340
nlattr_to_tcp(struct nlattr * cda[],struct nf_conn * ct)1341 static int nlattr_to_tcp(struct nlattr *cda[], struct nf_conn *ct)
1342 {
1343 struct nlattr *pattr = cda[CTA_PROTOINFO_TCP];
1344 struct nlattr *tb[CTA_PROTOINFO_TCP_MAX+1];
1345 int err;
1346
1347 /* updates could not contain anything about the private
1348 * protocol info, in that case skip the parsing */
1349 if (!pattr)
1350 return 0;
1351
1352 err = nla_parse_nested_deprecated(tb, CTA_PROTOINFO_TCP_MAX, pattr,
1353 tcp_nla_policy, NULL);
1354 if (err < 0)
1355 return err;
1356
1357 if (tb[CTA_PROTOINFO_TCP_STATE] &&
1358 nla_get_u8(tb[CTA_PROTOINFO_TCP_STATE]) >= TCP_CONNTRACK_MAX)
1359 return -EINVAL;
1360
1361 spin_lock_bh(&ct->lock);
1362 if (tb[CTA_PROTOINFO_TCP_STATE])
1363 ct->proto.tcp.state = nla_get_u8(tb[CTA_PROTOINFO_TCP_STATE]);
1364
1365 if (tb[CTA_PROTOINFO_TCP_FLAGS_ORIGINAL]) {
1366 struct nf_ct_tcp_flags *attr =
1367 nla_data(tb[CTA_PROTOINFO_TCP_FLAGS_ORIGINAL]);
1368 ct->proto.tcp.seen[0].flags &= ~attr->mask;
1369 ct->proto.tcp.seen[0].flags |= attr->flags & attr->mask;
1370 }
1371
1372 if (tb[CTA_PROTOINFO_TCP_FLAGS_REPLY]) {
1373 struct nf_ct_tcp_flags *attr =
1374 nla_data(tb[CTA_PROTOINFO_TCP_FLAGS_REPLY]);
1375 ct->proto.tcp.seen[1].flags &= ~attr->mask;
1376 ct->proto.tcp.seen[1].flags |= attr->flags & attr->mask;
1377 }
1378
1379 if (tb[CTA_PROTOINFO_TCP_WSCALE_ORIGINAL] &&
1380 tb[CTA_PROTOINFO_TCP_WSCALE_REPLY] &&
1381 ct->proto.tcp.seen[0].flags & IP_CT_TCP_FLAG_WINDOW_SCALE &&
1382 ct->proto.tcp.seen[1].flags & IP_CT_TCP_FLAG_WINDOW_SCALE) {
1383 ct->proto.tcp.seen[0].td_scale =
1384 nla_get_u8(tb[CTA_PROTOINFO_TCP_WSCALE_ORIGINAL]);
1385 ct->proto.tcp.seen[1].td_scale =
1386 nla_get_u8(tb[CTA_PROTOINFO_TCP_WSCALE_REPLY]);
1387 }
1388 spin_unlock_bh(&ct->lock);
1389
1390 return 0;
1391 }
1392
tcp_nlattr_tuple_size(void)1393 static unsigned int tcp_nlattr_tuple_size(void)
1394 {
1395 static unsigned int size __read_mostly;
1396
1397 if (!size)
1398 size = nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1);
1399
1400 return size;
1401 }
1402 #endif
1403
1404 #ifdef CONFIG_NF_CONNTRACK_TIMEOUT
1405
1406 #include <linux/netfilter/nfnetlink.h>
1407 #include <linux/netfilter/nfnetlink_cttimeout.h>
1408
tcp_timeout_nlattr_to_obj(struct nlattr * tb[],struct net * net,void * data)1409 static int tcp_timeout_nlattr_to_obj(struct nlattr *tb[],
1410 struct net *net, void *data)
1411 {
1412 struct nf_tcp_net *tn = nf_tcp_pernet(net);
1413 unsigned int *timeouts = data;
1414 int i;
1415
1416 if (!timeouts)
1417 timeouts = tn->timeouts;
1418 /* set default TCP timeouts. */
1419 for (i=0; i<TCP_CONNTRACK_TIMEOUT_MAX; i++)
1420 timeouts[i] = tn->timeouts[i];
1421
1422 if (tb[CTA_TIMEOUT_TCP_SYN_SENT]) {
1423 timeouts[TCP_CONNTRACK_SYN_SENT] =
1424 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_SYN_SENT]))*HZ;
1425 }
1426
1427 if (tb[CTA_TIMEOUT_TCP_SYN_RECV]) {
1428 timeouts[TCP_CONNTRACK_SYN_RECV] =
1429 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_SYN_RECV]))*HZ;
1430 }
1431 if (tb[CTA_TIMEOUT_TCP_ESTABLISHED]) {
1432 timeouts[TCP_CONNTRACK_ESTABLISHED] =
1433 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_ESTABLISHED]))*HZ;
1434 }
1435 if (tb[CTA_TIMEOUT_TCP_FIN_WAIT]) {
1436 timeouts[TCP_CONNTRACK_FIN_WAIT] =
1437 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_FIN_WAIT]))*HZ;
1438 }
1439 if (tb[CTA_TIMEOUT_TCP_CLOSE_WAIT]) {
1440 timeouts[TCP_CONNTRACK_CLOSE_WAIT] =
1441 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_CLOSE_WAIT]))*HZ;
1442 }
1443 if (tb[CTA_TIMEOUT_TCP_LAST_ACK]) {
1444 timeouts[TCP_CONNTRACK_LAST_ACK] =
1445 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_LAST_ACK]))*HZ;
1446 }
1447 if (tb[CTA_TIMEOUT_TCP_TIME_WAIT]) {
1448 timeouts[TCP_CONNTRACK_TIME_WAIT] =
1449 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_TIME_WAIT]))*HZ;
1450 }
1451 if (tb[CTA_TIMEOUT_TCP_CLOSE]) {
1452 timeouts[TCP_CONNTRACK_CLOSE] =
1453 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_CLOSE]))*HZ;
1454 }
1455 if (tb[CTA_TIMEOUT_TCP_SYN_SENT2]) {
1456 timeouts[TCP_CONNTRACK_SYN_SENT2] =
1457 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_SYN_SENT2]))*HZ;
1458 }
1459 if (tb[CTA_TIMEOUT_TCP_RETRANS]) {
1460 timeouts[TCP_CONNTRACK_RETRANS] =
1461 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_RETRANS]))*HZ;
1462 }
1463 if (tb[CTA_TIMEOUT_TCP_UNACK]) {
1464 timeouts[TCP_CONNTRACK_UNACK] =
1465 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_UNACK]))*HZ;
1466 }
1467
1468 timeouts[CTA_TIMEOUT_TCP_UNSPEC] = timeouts[CTA_TIMEOUT_TCP_SYN_SENT];
1469 return 0;
1470 }
1471
1472 static int
tcp_timeout_obj_to_nlattr(struct sk_buff * skb,const void * data)1473 tcp_timeout_obj_to_nlattr(struct sk_buff *skb, const void *data)
1474 {
1475 const unsigned int *timeouts = data;
1476
1477 if (nla_put_be32(skb, CTA_TIMEOUT_TCP_SYN_SENT,
1478 htonl(timeouts[TCP_CONNTRACK_SYN_SENT] / HZ)) ||
1479 nla_put_be32(skb, CTA_TIMEOUT_TCP_SYN_RECV,
1480 htonl(timeouts[TCP_CONNTRACK_SYN_RECV] / HZ)) ||
1481 nla_put_be32(skb, CTA_TIMEOUT_TCP_ESTABLISHED,
1482 htonl(timeouts[TCP_CONNTRACK_ESTABLISHED] / HZ)) ||
1483 nla_put_be32(skb, CTA_TIMEOUT_TCP_FIN_WAIT,
1484 htonl(timeouts[TCP_CONNTRACK_FIN_WAIT] / HZ)) ||
1485 nla_put_be32(skb, CTA_TIMEOUT_TCP_CLOSE_WAIT,
1486 htonl(timeouts[TCP_CONNTRACK_CLOSE_WAIT] / HZ)) ||
1487 nla_put_be32(skb, CTA_TIMEOUT_TCP_LAST_ACK,
1488 htonl(timeouts[TCP_CONNTRACK_LAST_ACK] / HZ)) ||
1489 nla_put_be32(skb, CTA_TIMEOUT_TCP_TIME_WAIT,
1490 htonl(timeouts[TCP_CONNTRACK_TIME_WAIT] / HZ)) ||
1491 nla_put_be32(skb, CTA_TIMEOUT_TCP_CLOSE,
1492 htonl(timeouts[TCP_CONNTRACK_CLOSE] / HZ)) ||
1493 nla_put_be32(skb, CTA_TIMEOUT_TCP_SYN_SENT2,
1494 htonl(timeouts[TCP_CONNTRACK_SYN_SENT2] / HZ)) ||
1495 nla_put_be32(skb, CTA_TIMEOUT_TCP_RETRANS,
1496 htonl(timeouts[TCP_CONNTRACK_RETRANS] / HZ)) ||
1497 nla_put_be32(skb, CTA_TIMEOUT_TCP_UNACK,
1498 htonl(timeouts[TCP_CONNTRACK_UNACK] / HZ)))
1499 goto nla_put_failure;
1500 return 0;
1501
1502 nla_put_failure:
1503 return -ENOSPC;
1504 }
1505
1506 static const struct nla_policy tcp_timeout_nla_policy[CTA_TIMEOUT_TCP_MAX+1] = {
1507 [CTA_TIMEOUT_TCP_SYN_SENT] = { .type = NLA_U32 },
1508 [CTA_TIMEOUT_TCP_SYN_RECV] = { .type = NLA_U32 },
1509 [CTA_TIMEOUT_TCP_ESTABLISHED] = { .type = NLA_U32 },
1510 [CTA_TIMEOUT_TCP_FIN_WAIT] = { .type = NLA_U32 },
1511 [CTA_TIMEOUT_TCP_CLOSE_WAIT] = { .type = NLA_U32 },
1512 [CTA_TIMEOUT_TCP_LAST_ACK] = { .type = NLA_U32 },
1513 [CTA_TIMEOUT_TCP_TIME_WAIT] = { .type = NLA_U32 },
1514 [CTA_TIMEOUT_TCP_CLOSE] = { .type = NLA_U32 },
1515 [CTA_TIMEOUT_TCP_SYN_SENT2] = { .type = NLA_U32 },
1516 [CTA_TIMEOUT_TCP_RETRANS] = { .type = NLA_U32 },
1517 [CTA_TIMEOUT_TCP_UNACK] = { .type = NLA_U32 },
1518 };
1519 #endif /* CONFIG_NF_CONNTRACK_TIMEOUT */
1520
nf_conntrack_tcp_init_net(struct net * net)1521 void nf_conntrack_tcp_init_net(struct net *net)
1522 {
1523 struct nf_tcp_net *tn = nf_tcp_pernet(net);
1524 int i;
1525
1526 for (i = 0; i < TCP_CONNTRACK_TIMEOUT_MAX; i++)
1527 tn->timeouts[i] = tcp_timeouts[i];
1528
1529 /* timeouts[0] is unused, make it same as SYN_SENT so
1530 * ->timeouts[0] contains 'new' timeout, like udp or icmp.
1531 */
1532 tn->timeouts[0] = tcp_timeouts[TCP_CONNTRACK_SYN_SENT];
1533
1534 /* If it is set to zero, we disable picking up already established
1535 * connections.
1536 */
1537 tn->tcp_loose = 1;
1538
1539 /* "Be conservative in what you do,
1540 * be liberal in what you accept from others."
1541 * If it's non-zero, we mark only out of window RST segments as INVALID.
1542 */
1543 tn->tcp_be_liberal = 0;
1544
1545 /* If it's non-zero, we turn off RST sequence number check */
1546 tn->tcp_ignore_invalid_rst = 0;
1547
1548 /* Max number of the retransmitted packets without receiving an (acceptable)
1549 * ACK from the destination. If this number is reached, a shorter timer
1550 * will be started.
1551 */
1552 tn->tcp_max_retrans = 3;
1553
1554 #if IS_ENABLED(CONFIG_NF_FLOW_TABLE)
1555 tn->offload_timeout = 30 * HZ;
1556 #endif
1557 }
1558
1559 const struct nf_conntrack_l4proto nf_conntrack_l4proto_tcp =
1560 {
1561 .l4proto = IPPROTO_TCP,
1562 #ifdef CONFIG_NF_CONNTRACK_PROCFS
1563 .print_conntrack = tcp_print_conntrack,
1564 #endif
1565 .can_early_drop = tcp_can_early_drop,
1566 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
1567 .to_nlattr = tcp_to_nlattr,
1568 .from_nlattr = nlattr_to_tcp,
1569 .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr,
1570 .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple,
1571 .nlattr_tuple_size = tcp_nlattr_tuple_size,
1572 .nlattr_size = TCP_NLATTR_SIZE,
1573 .nla_policy = nf_ct_port_nla_policy,
1574 #endif
1575 #ifdef CONFIG_NF_CONNTRACK_TIMEOUT
1576 .ctnl_timeout = {
1577 .nlattr_to_obj = tcp_timeout_nlattr_to_obj,
1578 .obj_to_nlattr = tcp_timeout_obj_to_nlattr,
1579 .nlattr_max = CTA_TIMEOUT_TCP_MAX,
1580 .obj_size = sizeof(unsigned int) *
1581 TCP_CONNTRACK_TIMEOUT_MAX,
1582 .nla_policy = tcp_timeout_nla_policy,
1583 },
1584 #endif /* CONFIG_NF_CONNTRACK_TIMEOUT */
1585 };
1586