netfilter: nf_conntrack: fix memory leak if sysctl registration fails
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / netfilter / nf_conntrack_proto_tcp.c
CommitLineData
9fb9cbb1
YK
1/* (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.
9fb9cbb1
YK
7 */
8
9fb9cbb1 9#include <linux/types.h>
9fb9cbb1 10#include <linux/timer.h>
9fb9cbb1
YK
11#include <linux/module.h>
12#include <linux/in.h>
13#include <linux/tcp.h>
14#include <linux/spinlock.h>
15#include <linux/skbuff.h>
16#include <linux/ipv6.h>
17#include <net/ip6_checksum.h>
534f81a5 18#include <asm/unaligned.h>
9fb9cbb1
YK
19
20#include <net/tcp.h>
21
22#include <linux/netfilter.h>
23#include <linux/netfilter_ipv4.h>
24#include <linux/netfilter_ipv6.h>
25#include <net/netfilter/nf_conntrack.h>
605dcad6 26#include <net/netfilter/nf_conntrack_l4proto.h>
f6180121 27#include <net/netfilter/nf_conntrack_ecache.h>
f01ffbd6 28#include <net/netfilter/nf_log.h>
9d2493f8
CP
29#include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
30#include <net/netfilter/ipv6/nf_conntrack_ipv6.h>
9fb9cbb1 31
601e68e1
YH
32/* "Be conservative in what you do,
33 be liberal in what you accept from others."
9fb9cbb1 34 If it's non-zero, we mark only out of window RST segments as INVALID. */
3aef0fd9 35static int nf_ct_tcp_be_liberal __read_mostly = 0;
9fb9cbb1 36
a09113c2 37/* If it is set to zero, we disable picking up already established
9fb9cbb1 38 connections. */
3aef0fd9 39static int nf_ct_tcp_loose __read_mostly = 1;
9fb9cbb1 40
601e68e1
YH
41/* Max number of the retransmitted packets without receiving an (acceptable)
42 ACK from the destination. If this number is reached, a shorter timer
9fb9cbb1 43 will be started. */
3aef0fd9 44static int nf_ct_tcp_max_retrans __read_mostly = 3;
9fb9cbb1
YK
45
46 /* FIXME: Examine ipfilter's timeouts and conntrack transitions more
47 closely. They're more complex. --RR */
48
82f568fc 49static const char *const tcp_conntrack_names[] = {
9fb9cbb1
YK
50 "NONE",
51 "SYN_SENT",
52 "SYN_RECV",
53 "ESTABLISHED",
54 "FIN_WAIT",
55 "CLOSE_WAIT",
56 "LAST_ACK",
57 "TIME_WAIT",
58 "CLOSE",
874ab923 59 "SYN_SENT2",
9fb9cbb1 60};
601e68e1 61
9fb9cbb1
YK
62#define SECS * HZ
63#define MINS * 60 SECS
64#define HOURS * 60 MINS
65#define DAYS * 24 HOURS
66
33ee4464 67static unsigned int tcp_timeouts[TCP_CONNTRACK_TIMEOUT_MAX] __read_mostly = {
2d646286
PM
68 [TCP_CONNTRACK_SYN_SENT] = 2 MINS,
69 [TCP_CONNTRACK_SYN_RECV] = 60 SECS,
70 [TCP_CONNTRACK_ESTABLISHED] = 5 DAYS,
71 [TCP_CONNTRACK_FIN_WAIT] = 2 MINS,
72 [TCP_CONNTRACK_CLOSE_WAIT] = 60 SECS,
73 [TCP_CONNTRACK_LAST_ACK] = 30 SECS,
74 [TCP_CONNTRACK_TIME_WAIT] = 2 MINS,
75 [TCP_CONNTRACK_CLOSE] = 10 SECS,
874ab923 76 [TCP_CONNTRACK_SYN_SENT2] = 2 MINS,
33ee4464
PNA
77/* RFC1122 says the R2 limit should be at least 100 seconds.
78 Linux uses 15 packets as limit, which corresponds
79 to ~13-30min depending on RTO. */
80 [TCP_CONNTRACK_RETRANS] = 5 MINS,
81 [TCP_CONNTRACK_UNACK] = 5 MINS,
2d646286 82};
601e68e1 83
9fb9cbb1
YK
84#define sNO TCP_CONNTRACK_NONE
85#define sSS TCP_CONNTRACK_SYN_SENT
86#define sSR TCP_CONNTRACK_SYN_RECV
87#define sES TCP_CONNTRACK_ESTABLISHED
88#define sFW TCP_CONNTRACK_FIN_WAIT
89#define sCW TCP_CONNTRACK_CLOSE_WAIT
90#define sLA TCP_CONNTRACK_LAST_ACK
91#define sTW TCP_CONNTRACK_TIME_WAIT
92#define sCL TCP_CONNTRACK_CLOSE
874ab923 93#define sS2 TCP_CONNTRACK_SYN_SENT2
9fb9cbb1
YK
94#define sIV TCP_CONNTRACK_MAX
95#define sIG TCP_CONNTRACK_IGNORE
96
97/* What TCP flags are set from RST/SYN/FIN/ACK. */
98enum tcp_bit_set {
99 TCP_SYN_SET,
100 TCP_SYNACK_SET,
101 TCP_FIN_SET,
102 TCP_ACK_SET,
103 TCP_RST_SET,
104 TCP_NONE_SET,
105};
601e68e1 106
9fb9cbb1
YK
107/*
108 * The TCP state transition table needs a few words...
109 *
110 * We are the man in the middle. All the packets go through us
111 * but might get lost in transit to the destination.
601e68e1 112 * It is assumed that the destinations can't receive segments
9fb9cbb1
YK
113 * we haven't seen.
114 *
115 * The checked segment is in window, but our windows are *not*
116 * equivalent with the ones of the sender/receiver. We always
117 * try to guess the state of the current sender.
118 *
119 * The meaning of the states are:
120 *
121 * NONE: initial state
601e68e1 122 * SYN_SENT: SYN-only packet seen
874ab923 123 * SYN_SENT2: SYN-only packet seen from reply dir, simultaneous open
9fb9cbb1
YK
124 * SYN_RECV: SYN-ACK packet seen
125 * ESTABLISHED: ACK packet seen
126 * FIN_WAIT: FIN packet seen
601e68e1 127 * CLOSE_WAIT: ACK seen (after FIN)
9fb9cbb1
YK
128 * LAST_ACK: FIN seen (after FIN)
129 * TIME_WAIT: last ACK seen
b2155e7f 130 * CLOSE: closed connection (RST)
9fb9cbb1 131 *
9fb9cbb1 132 * Packets marked as IGNORED (sIG):
601e68e1
YH
133 * if they may be either invalid or valid
134 * and the receiver may send back a connection
9fb9cbb1
YK
135 * closing RST or a SYN/ACK.
136 *
137 * Packets marked as INVALID (sIV):
874ab923 138 * if we regard them as truly invalid packets
9fb9cbb1 139 */
a5e73c29 140static const u8 tcp_conntracks[2][6][TCP_CONNTRACK_MAX] = {
9fb9cbb1
YK
141 {
142/* ORIGINAL */
874ab923
JK
143/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
144/*syn*/ { sSS, sSS, sIG, sIG, sIG, sIG, sIG, sSS, sSS, sS2 },
9fb9cbb1
YK
145/*
146 * sNO -> sSS Initialize a new connection
147 * sSS -> sSS Retransmitted SYN
874ab923
JK
148 * sS2 -> sS2 Late retransmitted SYN
149 * sSR -> sIG
9fb9cbb1 150 * sES -> sIG Error: SYNs in window outside the SYN_SENT state
601e68e1 151 * are errors. Receiver will reply with RST
9fb9cbb1
YK
152 * and close the connection.
153 * Or we are not in sync and hold a dead connection.
154 * sFW -> sIG
155 * sCW -> sIG
156 * sLA -> sIG
157 * sTW -> sSS Reopened connection (RFC 1122).
158 * sCL -> sSS
159 */
874ab923
JK
160/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
161/*synack*/ { sIV, sIV, sIG, sIG, sIG, sIG, sIG, sIG, sIG, sSR },
9fb9cbb1 162/*
874ab923
JK
163 * sNO -> sIV Too late and no reason to do anything
164 * sSS -> sIV Client can't send SYN and then SYN/ACK
165 * sS2 -> sSR SYN/ACK sent to SYN2 in simultaneous open
166 * sSR -> sIG
167 * sES -> sIG Error: SYNs in window outside the SYN_SENT state
168 * are errors. Receiver will reply with RST
169 * and close the connection.
170 * Or we are not in sync and hold a dead connection.
171 * sFW -> sIG
172 * sCW -> sIG
173 * sLA -> sIG
174 * sTW -> sIG
175 * sCL -> sIG
9fb9cbb1 176 */
874ab923 177/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
9fb9cbb1
YK
178/*fin*/ { sIV, sIV, sFW, sFW, sLA, sLA, sLA, sTW, sCL, sIV },
179/*
180 * sNO -> sIV Too late and no reason to do anything...
181 * sSS -> sIV Client migth not send FIN in this state:
182 * we enforce waiting for a SYN/ACK reply first.
874ab923 183 * sS2 -> sIV
9fb9cbb1
YK
184 * sSR -> sFW Close started.
185 * sES -> sFW
186 * sFW -> sLA FIN seen in both directions, waiting for
601e68e1 187 * the last ACK.
9fb9cbb1
YK
188 * Migth be a retransmitted FIN as well...
189 * sCW -> sLA
190 * sLA -> sLA Retransmitted FIN. Remain in the same state.
191 * sTW -> sTW
192 * sCL -> sCL
193 */
874ab923 194/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
9fb9cbb1
YK
195/*ack*/ { sES, sIV, sES, sES, sCW, sCW, sTW, sTW, sCL, sIV },
196/*
197 * sNO -> sES Assumed.
198 * sSS -> sIV ACK is invalid: we haven't seen a SYN/ACK yet.
874ab923 199 * sS2 -> sIV
9fb9cbb1
YK
200 * sSR -> sES Established state is reached.
201 * sES -> sES :-)
202 * sFW -> sCW Normal close request answered by ACK.
203 * sCW -> sCW
204 * sLA -> sTW Last ACK detected.
205 * sTW -> sTW Retransmitted last ACK. Remain in the same state.
206 * sCL -> sCL
207 */
874ab923
JK
208/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
209/*rst*/ { sIV, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL },
9fb9cbb1
YK
210/*none*/ { sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV }
211 },
212 {
213/* REPLY */
874ab923
JK
214/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
215/*syn*/ { sIV, sS2, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sS2 },
9fb9cbb1
YK
216/*
217 * sNO -> sIV Never reached.
874ab923
JK
218 * sSS -> sS2 Simultaneous open
219 * sS2 -> sS2 Retransmitted simultaneous SYN
220 * sSR -> sIV Invalid SYN packets sent by the server
221 * sES -> sIV
9fb9cbb1
YK
222 * sFW -> sIV
223 * sCW -> sIV
224 * sLA -> sIV
225 * sTW -> sIV Reopened connection, but server may not do it.
226 * sCL -> sIV
227 */
874ab923 228/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
8a80c79a 229/*synack*/ { sIV, sSR, sIG, sIG, sIG, sIG, sIG, sIG, sIG, sSR },
9fb9cbb1
YK
230/*
231 * sSS -> sSR Standard open.
874ab923 232 * sS2 -> sSR Simultaneous open
8a80c79a 233 * sSR -> sIG Retransmitted SYN/ACK, ignore it.
9fb9cbb1
YK
234 * sES -> sIG Late retransmitted SYN/ACK?
235 * sFW -> sIG Might be SYN/ACK answering ignored SYN
236 * sCW -> sIG
237 * sLA -> sIG
238 * sTW -> sIG
239 * sCL -> sIG
240 */
874ab923 241/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
9fb9cbb1
YK
242/*fin*/ { sIV, sIV, sFW, sFW, sLA, sLA, sLA, sTW, sCL, sIV },
243/*
244 * sSS -> sIV Server might not send FIN in this state.
874ab923 245 * sS2 -> sIV
9fb9cbb1
YK
246 * sSR -> sFW Close started.
247 * sES -> sFW
248 * sFW -> sLA FIN seen in both directions.
249 * sCW -> sLA
250 * sLA -> sLA Retransmitted FIN.
251 * sTW -> sTW
252 * sCL -> sCL
253 */
874ab923
JK
254/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
255/*ack*/ { sIV, sIG, sSR, sES, sCW, sCW, sTW, sTW, sCL, sIG },
9fb9cbb1 256/*
73f30602 257 * sSS -> sIG Might be a half-open connection.
874ab923 258 * sS2 -> sIG
9fb9cbb1
YK
259 * sSR -> sSR Might answer late resent SYN.
260 * sES -> sES :-)
261 * sFW -> sCW Normal close request answered by ACK.
262 * sCW -> sCW
263 * sLA -> sTW Last ACK detected.
264 * sTW -> sTW Retransmitted last ACK.
265 * sCL -> sCL
266 */
874ab923
JK
267/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
268/*rst*/ { sIV, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL },
9fb9cbb1 269/*none*/ { sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV }
601e68e1 270 }
9fb9cbb1
YK
271};
272
d2ba1fde
G
273static inline struct nf_tcp_net *tcp_pernet(struct net *net)
274{
275 return &net->ct.nf_ct_proto.tcp;
276}
277
09f263cd
JE
278static bool tcp_pkt_to_tuple(const struct sk_buff *skb, unsigned int dataoff,
279 struct nf_conntrack_tuple *tuple)
9fb9cbb1 280{
82f568fc
JE
281 const struct tcphdr *hp;
282 struct tcphdr _hdr;
9fb9cbb1
YK
283
284 /* Actually only need first 8 bytes. */
285 hp = skb_header_pointer(skb, dataoff, 8, &_hdr);
286 if (hp == NULL)
09f263cd 287 return false;
9fb9cbb1
YK
288
289 tuple->src.u.tcp.port = hp->source;
290 tuple->dst.u.tcp.port = hp->dest;
291
09f263cd 292 return true;
9fb9cbb1
YK
293}
294
09f263cd
JE
295static bool tcp_invert_tuple(struct nf_conntrack_tuple *tuple,
296 const struct nf_conntrack_tuple *orig)
9fb9cbb1
YK
297{
298 tuple->src.u.tcp.port = orig->dst.u.tcp.port;
299 tuple->dst.u.tcp.port = orig->src.u.tcp.port;
09f263cd 300 return true;
9fb9cbb1
YK
301}
302
303/* Print out the per-protocol part of the tuple. */
304static int tcp_print_tuple(struct seq_file *s,
305 const struct nf_conntrack_tuple *tuple)
306{
307 return seq_printf(s, "sport=%hu dport=%hu ",
308 ntohs(tuple->src.u.tcp.port),
309 ntohs(tuple->dst.u.tcp.port));
310}
311
312/* Print out the private part of the conntrack. */
440f0d58 313static int tcp_print_conntrack(struct seq_file *s, struct nf_conn *ct)
9fb9cbb1
YK
314{
315 enum tcp_conntrack state;
316
440f0d58 317 spin_lock_bh(&ct->lock);
c88130bc 318 state = ct->proto.tcp.state;
440f0d58 319 spin_unlock_bh(&ct->lock);
9fb9cbb1
YK
320
321 return seq_printf(s, "%s ", tcp_conntrack_names[state]);
322}
323
324static unsigned int get_conntrack_index(const struct tcphdr *tcph)
325{
326 if (tcph->rst) return TCP_RST_SET;
327 else if (tcph->syn) return (tcph->ack ? TCP_SYNACK_SET : TCP_SYN_SET);
328 else if (tcph->fin) return TCP_FIN_SET;
329 else if (tcph->ack) return TCP_ACK_SET;
330 else return TCP_NONE_SET;
331}
332
333/* TCP connection tracking based on 'Real Stateful TCP Packet Filtering
334 in IP Filter' by Guido van Rooij.
601e68e1 335
631dd1a8
JM
336 http://www.sane.nl/events/sane2000/papers.html
337 http://www.darkart.com/mirrors/www.obfuscation.org/ipf/
601e68e1 338
9fb9cbb1
YK
339 The boundaries and the conditions are changed according to RFC793:
340 the packet must intersect the window (i.e. segments may be
341 after the right or before the left edge) and thus receivers may ACK
342 segments after the right edge of the window.
343
601e68e1 344 td_maxend = max(sack + max(win,1)) seen in reply packets
9fb9cbb1
YK
345 td_maxwin = max(max(win, 1)) + (sack - ack) seen in sent packets
346 td_maxwin += seq + len - sender.td_maxend
347 if seq + len > sender.td_maxend
348 td_end = max(seq + len) seen in sent packets
601e68e1 349
9fb9cbb1
YK
350 I. Upper bound for valid data: seq <= sender.td_maxend
351 II. Lower bound for valid data: seq + len >= sender.td_end - receiver.td_maxwin
84ebe1cd
JK
352 III. Upper bound for valid (s)ack: sack <= receiver.td_end
353 IV. Lower bound for valid (s)ack: sack >= receiver.td_end - MAXACKWINDOW
9fb9cbb1 354
84ebe1cd
JK
355 where sack is the highest right edge of sack block found in the packet
356 or ack in the case of packet without SACK option.
9fb9cbb1 357
84ebe1cd 358 The upper bound limit for a valid (s)ack is not ignored -
601e68e1 359 we doesn't have to deal with fragments.
9fb9cbb1
YK
360*/
361
362static inline __u32 segment_seq_plus_len(__u32 seq,
363 size_t len,
364 unsigned int dataoff,
82f568fc 365 const struct tcphdr *tcph)
9fb9cbb1
YK
366{
367 /* XXX Should I use payload length field in IP/IPv6 header ?
368 * - YK */
369 return (seq + len - dataoff - tcph->doff*4
370 + (tcph->syn ? 1 : 0) + (tcph->fin ? 1 : 0));
371}
601e68e1 372
9fb9cbb1
YK
373/* Fixme: what about big packets? */
374#define MAXACKWINCONST 66000
375#define MAXACKWINDOW(sender) \
376 ((sender)->td_maxwin > MAXACKWINCONST ? (sender)->td_maxwin \
377 : MAXACKWINCONST)
601e68e1 378
9fb9cbb1
YK
379/*
380 * Simplified tcp_parse_options routine from tcp_input.c
381 */
382static void tcp_options(const struct sk_buff *skb,
383 unsigned int dataoff,
82f568fc 384 const struct tcphdr *tcph,
9fb9cbb1
YK
385 struct ip_ct_tcp_state *state)
386{
387 unsigned char buff[(15 * 4) - sizeof(struct tcphdr)];
82f568fc 388 const unsigned char *ptr;
9fb9cbb1
YK
389 int length = (tcph->doff*4) - sizeof(struct tcphdr);
390
391 if (!length)
392 return;
393
394 ptr = skb_header_pointer(skb, dataoff + sizeof(struct tcphdr),
395 length, buff);
396 BUG_ON(ptr == NULL);
397
601e68e1 398 state->td_scale =
9fb9cbb1
YK
399 state->flags = 0;
400
401 while (length > 0) {
402 int opcode=*ptr++;
403 int opsize;
404
405 switch (opcode) {
406 case TCPOPT_EOL:
407 return;
408 case TCPOPT_NOP: /* Ref: RFC 793 section 3.1 */
409 length--;
410 continue;
411 default:
412 opsize=*ptr++;
413 if (opsize < 2) /* "silly options" */
414 return;
415 if (opsize > length)
4a5cc84a 416 return; /* don't parse partial options */
9fb9cbb1 417
601e68e1 418 if (opcode == TCPOPT_SACK_PERM
9fb9cbb1
YK
419 && opsize == TCPOLEN_SACK_PERM)
420 state->flags |= IP_CT_TCP_FLAG_SACK_PERM;
421 else if (opcode == TCPOPT_WINDOW
422 && opsize == TCPOLEN_WINDOW) {
423 state->td_scale = *(u_int8_t *)ptr;
424
425 if (state->td_scale > 14) {
426 /* See RFC1323 */
427 state->td_scale = 14;
428 }
429 state->flags |=
430 IP_CT_TCP_FLAG_WINDOW_SCALE;
431 }
432 ptr += opsize - 2;
433 length -= opsize;
434 }
435 }
436}
437
438static void tcp_sack(const struct sk_buff *skb, unsigned int dataoff,
82f568fc 439 const struct tcphdr *tcph, __u32 *sack)
9fb9cbb1 440{
601e68e1 441 unsigned char buff[(15 * 4) - sizeof(struct tcphdr)];
82f568fc 442 const unsigned char *ptr;
9fb9cbb1
YK
443 int length = (tcph->doff*4) - sizeof(struct tcphdr);
444 __u32 tmp;
445
446 if (!length)
447 return;
448
449 ptr = skb_header_pointer(skb, dataoff + sizeof(struct tcphdr),
450 length, buff);
451 BUG_ON(ptr == NULL);
452
453 /* Fast path for timestamp-only option */
bb9fc373 454 if (length == TCPOLEN_TSTAMP_ALIGNED
8f05ce91
YH
455 && *(__be32 *)ptr == htonl((TCPOPT_NOP << 24)
456 | (TCPOPT_NOP << 16)
457 | (TCPOPT_TIMESTAMP << 8)
458 | TCPOLEN_TIMESTAMP))
9fb9cbb1
YK
459 return;
460
461 while (length > 0) {
462 int opcode = *ptr++;
463 int opsize, i;
464
465 switch (opcode) {
466 case TCPOPT_EOL:
467 return;
468 case TCPOPT_NOP: /* Ref: RFC 793 section 3.1 */
469 length--;
470 continue;
471 default:
472 opsize = *ptr++;
473 if (opsize < 2) /* "silly options" */
474 return;
475 if (opsize > length)
4a5cc84a 476 return; /* don't parse partial options */
9fb9cbb1 477
601e68e1
YH
478 if (opcode == TCPOPT_SACK
479 && opsize >= (TCPOLEN_SACK_BASE
480 + TCPOLEN_SACK_PERBLOCK)
481 && !((opsize - TCPOLEN_SACK_BASE)
482 % TCPOLEN_SACK_PERBLOCK)) {
483 for (i = 0;
484 i < (opsize - TCPOLEN_SACK_BASE);
485 i += TCPOLEN_SACK_PERBLOCK) {
534f81a5 486 tmp = get_unaligned_be32((__be32 *)(ptr+i)+1);
9fb9cbb1
YK
487
488 if (after(tmp, *sack))
489 *sack = tmp;
490 }
491 return;
492 }
493 ptr += opsize - 2;
494 length -= opsize;
495 }
496 }
497}
498
f9dd09c7
JK
499#ifdef CONFIG_NF_NAT_NEEDED
500static inline s16 nat_offset(const struct nf_conn *ct,
501 enum ip_conntrack_dir dir,
502 u32 seq)
503{
504 typeof(nf_ct_nat_offset) get_offset = rcu_dereference(nf_ct_nat_offset);
505
506 return get_offset != NULL ? get_offset(ct, dir, seq) : 0;
507}
508#define NAT_OFFSET(pf, ct, dir, seq) \
509 (pf == NFPROTO_IPV4 ? nat_offset(ct, dir, seq) : 0)
510#else
511#define NAT_OFFSET(pf, ct, dir, seq) 0
512#endif
513
09f263cd
JE
514static bool tcp_in_window(const struct nf_conn *ct,
515 struct ip_ct_tcp *state,
516 enum ip_conntrack_dir dir,
517 unsigned int index,
518 const struct sk_buff *skb,
519 unsigned int dataoff,
520 const struct tcphdr *tcph,
76108cea 521 u_int8_t pf)
9fb9cbb1 522{
c2a2c7e0 523 struct net *net = nf_ct_net(ct);
d2ba1fde 524 struct nf_tcp_net *tn = tcp_pernet(net);
9fb9cbb1
YK
525 struct ip_ct_tcp_state *sender = &state->seen[dir];
526 struct ip_ct_tcp_state *receiver = &state->seen[!dir];
82f568fc 527 const struct nf_conntrack_tuple *tuple = &ct->tuplehash[dir].tuple;
9fb9cbb1 528 __u32 seq, ack, sack, end, win, swin;
f9dd09c7 529 s16 receiver_offset;
09f263cd 530 bool res;
9fb9cbb1
YK
531
532 /*
533 * Get the required data from the packet.
534 */
535 seq = ntohl(tcph->seq);
536 ack = sack = ntohl(tcph->ack_seq);
537 win = ntohs(tcph->window);
538 end = segment_seq_plus_len(seq, skb->len, dataoff, tcph);
539
540 if (receiver->flags & IP_CT_TCP_FLAG_SACK_PERM)
541 tcp_sack(skb, dataoff, tcph, &sack);
542
f9dd09c7
JK
543 /* Take into account NAT sequence number mangling */
544 receiver_offset = NAT_OFFSET(pf, ct, !dir, ack - 1);
545 ack -= receiver_offset;
546 sack -= receiver_offset;
547
0d53778e
PM
548 pr_debug("tcp_in_window: START\n");
549 pr_debug("tcp_in_window: ");
3c9fba65 550 nf_ct_dump_tuple(tuple);
f9dd09c7
JK
551 pr_debug("seq=%u ack=%u+(%d) sack=%u+(%d) win=%u end=%u\n",
552 seq, ack, receiver_offset, sack, receiver_offset, win, end);
0d53778e
PM
553 pr_debug("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);
9fb9cbb1 559
874ab923 560 if (sender->td_maxwin == 0) {
9fb9cbb1
YK
561 /*
562 * Initialize sender data.
563 */
874ab923 564 if (tcph->syn) {
9fb9cbb1 565 /*
874ab923
JK
566 * SYN-ACK in reply to a SYN
567 * or SYN from reply direction in simultaneous open.
9fb9cbb1 568 */
601e68e1 569 sender->td_end =
9fb9cbb1
YK
570 sender->td_maxend = end;
571 sender->td_maxwin = (win == 0 ? 1 : win);
572
573 tcp_options(skb, dataoff, tcph, sender);
601e68e1 574 /*
9fb9cbb1
YK
575 * RFC 1323:
576 * Both sides must send the Window Scale option
577 * to enable window scaling in either direction.
578 */
579 if (!(sender->flags & IP_CT_TCP_FLAG_WINDOW_SCALE
580 && receiver->flags & IP_CT_TCP_FLAG_WINDOW_SCALE))
601e68e1 581 sender->td_scale =
9fb9cbb1 582 receiver->td_scale = 0;
874ab923
JK
583 if (!tcph->ack)
584 /* Simultaneous open */
585 return true;
9fb9cbb1
YK
586 } else {
587 /*
588 * We are in the middle of a connection,
589 * its history is lost for us.
590 * Let's try to use the data from the packet.
601e68e1 591 */
9fb9cbb1 592 sender->td_end = end;
6ee0b693
CG
593 swin = win << sender->td_scale;
594 sender->td_maxwin = (swin == 0 ? 1 : swin);
9fb9cbb1 595 sender->td_maxend = end + sender->td_maxwin;
fac42a9a
PNA
596 /*
597 * We haven't seen traffic in the other direction yet
598 * but we have to tweak window tracking to pass III
599 * and IV until that happens.
600 */
601 if (receiver->td_maxwin == 0)
602 receiver->td_end = receiver->td_maxend = sack;
9fb9cbb1
YK
603 }
604 } else if (((state->state == TCP_CONNTRACK_SYN_SENT
605 && dir == IP_CT_DIR_ORIGINAL)
606 || (state->state == TCP_CONNTRACK_SYN_RECV
607 && dir == IP_CT_DIR_REPLY))
608 && after(end, sender->td_end)) {
609 /*
610 * RFC 793: "if a TCP is reinitialized ... then it need
601e68e1 611 * not wait at all; it must only be sure to use sequence
9fb9cbb1
YK
612 * numbers larger than those recently used."
613 */
614 sender->td_end =
615 sender->td_maxend = end;
616 sender->td_maxwin = (win == 0 ? 1 : win);
617
618 tcp_options(skb, dataoff, tcph, sender);
619 }
620
621 if (!(tcph->ack)) {
622 /*
623 * If there is no ACK, just pretend it was set and OK.
624 */
625 ack = sack = receiver->td_end;
601e68e1
YH
626 } else if (((tcp_flag_word(tcph) & (TCP_FLAG_ACK|TCP_FLAG_RST)) ==
627 (TCP_FLAG_ACK|TCP_FLAG_RST))
9fb9cbb1
YK
628 && (ack == 0)) {
629 /*
630 * Broken TCP stacks, that set ACK in RST packets as well
631 * with zero ack value.
632 */
633 ack = sack = receiver->td_end;
634 }
635
636 if (seq == end
637 && (!tcph->rst
638 || (seq == 0 && state->state == TCP_CONNTRACK_SYN_SENT)))
639 /*
640 * Packets contains no data: we assume it is valid
641 * and check the ack value only.
642 * However RST segments are always validated by their
643 * SEQ number, except when seq == 0 (reset sent answering
644 * SYN.
645 */
646 seq = end = sender->td_end;
647
0d53778e 648 pr_debug("tcp_in_window: ");
3c9fba65 649 nf_ct_dump_tuple(tuple);
f9dd09c7
JK
650 pr_debug("seq=%u ack=%u+(%d) sack=%u+(%d) win=%u end=%u\n",
651 seq, ack, receiver_offset, sack, receiver_offset, win, end);
0d53778e
PM
652 pr_debug("tcp_in_window: sender end=%u maxend=%u maxwin=%u scale=%i "
653 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
654 sender->td_end, sender->td_maxend, sender->td_maxwin,
655 sender->td_scale,
656 receiver->td_end, receiver->td_maxend, receiver->td_maxwin,
657 receiver->td_scale);
658
659 pr_debug("tcp_in_window: I=%i II=%i III=%i IV=%i\n",
660 before(seq, sender->td_maxend + 1),
661 after(end, sender->td_end - receiver->td_maxwin - 1),
662 before(sack, receiver->td_end + 1),
84ebe1cd 663 after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1));
9fb9cbb1 664
a09113c2
PM
665 if (before(seq, sender->td_maxend + 1) &&
666 after(end, sender->td_end - receiver->td_maxwin - 1) &&
667 before(sack, receiver->td_end + 1) &&
84ebe1cd 668 after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1)) {
601e68e1 669 /*
9fb9cbb1
YK
670 * Take into account window scaling (RFC 1323).
671 */
672 if (!tcph->syn)
673 win <<= sender->td_scale;
674
675 /*
676 * Update sender data.
677 */
678 swin = win + (sack - ack);
679 if (sender->td_maxwin < swin)
680 sender->td_maxwin = swin;
ae375044 681 if (after(end, sender->td_end)) {
9fb9cbb1 682 sender->td_end = end;
ae375044
PM
683 sender->flags |= IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED;
684 }
bfcaa502
JK
685 if (tcph->ack) {
686 if (!(sender->flags & IP_CT_TCP_FLAG_MAXACK_SET)) {
687 sender->td_maxack = ack;
688 sender->flags |= IP_CT_TCP_FLAG_MAXACK_SET;
689 } else if (after(ack, sender->td_maxack))
690 sender->td_maxack = ack;
691 }
692
9fb9cbb1
YK
693 /*
694 * Update receiver data.
695 */
fac42a9a 696 if (receiver->td_maxwin != 0 && after(end, sender->td_maxend))
9fb9cbb1
YK
697 receiver->td_maxwin += end - sender->td_maxend;
698 if (after(sack + win, receiver->td_maxend - 1)) {
699 receiver->td_maxend = sack + win;
700 if (win == 0)
701 receiver->td_maxend++;
702 }
ae375044
PM
703 if (ack == receiver->td_end)
704 receiver->flags &= ~IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED;
9fb9cbb1 705
601e68e1 706 /*
9fb9cbb1
YK
707 * Check retransmissions.
708 */
709 if (index == TCP_ACK_SET) {
710 if (state->last_dir == dir
711 && state->last_seq == seq
712 && state->last_ack == ack
c1fe3ca5
GH
713 && state->last_end == end
714 && state->last_win == win)
9fb9cbb1
YK
715 state->retrans++;
716 else {
717 state->last_dir = dir;
718 state->last_seq = seq;
719 state->last_ack = ack;
720 state->last_end = end;
c1fe3ca5 721 state->last_win = win;
9fb9cbb1
YK
722 state->retrans = 0;
723 }
724 }
09f263cd 725 res = true;
9fb9cbb1 726 } else {
09f263cd 727 res = false;
a09113c2 728 if (sender->flags & IP_CT_TCP_FLAG_BE_LIBERAL ||
d2ba1fde 729 tn->tcp_be_liberal)
09f263cd 730 res = true;
c2a2c7e0 731 if (!res && LOG_INVALID(net, IPPROTO_TCP))
9fb9cbb1
YK
732 nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
733 "nf_ct_tcp: %s ",
734 before(seq, sender->td_maxend + 1) ?
735 after(end, sender->td_end - receiver->td_maxwin - 1) ?
736 before(sack, receiver->td_end + 1) ?
f9dd09c7 737 after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1) ? "BUG"
9fb9cbb1
YK
738 : "ACK is under the lower bound (possible overly delayed ACK)"
739 : "ACK is over the upper bound (ACKed data not seen yet)"
740 : "SEQ is under the lower bound (already ACKed data retransmitted)"
741 : "SEQ is over the upper bound (over the window of the receiver)");
601e68e1
YH
742 }
743
09f263cd 744 pr_debug("tcp_in_window: res=%u sender end=%u maxend=%u maxwin=%u "
0d53778e
PM
745 "receiver end=%u maxend=%u maxwin=%u\n",
746 res, sender->td_end, sender->td_maxend, sender->td_maxwin,
747 receiver->td_end, receiver->td_maxend, receiver->td_maxwin);
9fb9cbb1
YK
748
749 return res;
750}
751
5c8ce7c9 752/* table of valid flag combinations - PUSH, ECE and CWR are always valid */
a3433f35
CG
753static const u8 tcp_valid_flags[(TCPHDR_FIN|TCPHDR_SYN|TCPHDR_RST|TCPHDR_ACK|
754 TCPHDR_URG) + 1] =
9fb9cbb1 755{
a3433f35
CG
756 [TCPHDR_SYN] = 1,
757 [TCPHDR_SYN|TCPHDR_URG] = 1,
758 [TCPHDR_SYN|TCPHDR_ACK] = 1,
759 [TCPHDR_RST] = 1,
760 [TCPHDR_RST|TCPHDR_ACK] = 1,
761 [TCPHDR_FIN|TCPHDR_ACK] = 1,
762 [TCPHDR_FIN|TCPHDR_ACK|TCPHDR_URG] = 1,
763 [TCPHDR_ACK] = 1,
764 [TCPHDR_ACK|TCPHDR_URG] = 1,
9fb9cbb1
YK
765};
766
767/* Protect conntrack agaist broken packets. Code taken from ipt_unclean.c. */
8fea97ec 768static int tcp_error(struct net *net, struct nf_conn *tmpl,
74c51a14 769 struct sk_buff *skb,
9fb9cbb1
YK
770 unsigned int dataoff,
771 enum ip_conntrack_info *ctinfo,
76108cea 772 u_int8_t pf,
96f6bf82 773 unsigned int hooknum)
9fb9cbb1 774{
82f568fc
JE
775 const struct tcphdr *th;
776 struct tcphdr _tcph;
9fb9cbb1
YK
777 unsigned int tcplen = skb->len - dataoff;
778 u_int8_t tcpflags;
779
780 /* Smaller that minimal TCP header? */
781 th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph);
782 if (th == NULL) {
c2a2c7e0 783 if (LOG_INVALID(net, IPPROTO_TCP))
9fb9cbb1
YK
784 nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
785 "nf_ct_tcp: short packet ");
786 return -NF_ACCEPT;
601e68e1
YH
787 }
788
9fb9cbb1
YK
789 /* Not whole TCP header or malformed packet */
790 if (th->doff*4 < sizeof(struct tcphdr) || tcplen < th->doff*4) {
c2a2c7e0 791 if (LOG_INVALID(net, IPPROTO_TCP))
9fb9cbb1
YK
792 nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
793 "nf_ct_tcp: truncated/malformed packet ");
794 return -NF_ACCEPT;
795 }
601e68e1 796
9fb9cbb1
YK
797 /* Checksum invalid? Ignore.
798 * We skip checking packets on the outgoing path
84fa7933 799 * because the checksum is assumed to be correct.
9fb9cbb1
YK
800 */
801 /* FIXME: Source route IP option packets --RR */
c04d0552 802 if (net->ct.sysctl_checksum && hooknum == NF_INET_PRE_ROUTING &&
96f6bf82 803 nf_checksum(skb, hooknum, dataoff, IPPROTO_TCP, pf)) {
c2a2c7e0 804 if (LOG_INVALID(net, IPPROTO_TCP))
9fb9cbb1
YK
805 nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
806 "nf_ct_tcp: bad TCP checksum ");
807 return -NF_ACCEPT;
808 }
809
810 /* Check TCP flags. */
a3433f35 811 tcpflags = (tcp_flag_byte(th) & ~(TCPHDR_ECE|TCPHDR_CWR|TCPHDR_PSH));
9fb9cbb1 812 if (!tcp_valid_flags[tcpflags]) {
c2a2c7e0 813 if (LOG_INVALID(net, IPPROTO_TCP))
9fb9cbb1
YK
814 nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
815 "nf_ct_tcp: invalid TCP flag combination ");
816 return -NF_ACCEPT;
817 }
818
819 return NF_ACCEPT;
820}
821
2c8503f5
PNA
822static unsigned int *tcp_get_timeouts(struct net *net)
823{
824 return tcp_timeouts;
825}
826
9fb9cbb1 827/* Returns verdict for packet, or -1 for invalid. */
c88130bc 828static int tcp_packet(struct nf_conn *ct,
9fb9cbb1
YK
829 const struct sk_buff *skb,
830 unsigned int dataoff,
831 enum ip_conntrack_info ctinfo,
76108cea 832 u_int8_t pf,
2c8503f5
PNA
833 unsigned int hooknum,
834 unsigned int *timeouts)
9fb9cbb1 835{
c2a2c7e0 836 struct net *net = nf_ct_net(ct);
d2ba1fde 837 struct nf_tcp_net *tn = tcp_pernet(net);
0d53778e 838 struct nf_conntrack_tuple *tuple;
9fb9cbb1
YK
839 enum tcp_conntrack new_state, old_state;
840 enum ip_conntrack_dir dir;
82f568fc
JE
841 const struct tcphdr *th;
842 struct tcphdr _tcph;
9fb9cbb1
YK
843 unsigned long timeout;
844 unsigned int index;
845
846 th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph);
847 BUG_ON(th == NULL);
848
440f0d58 849 spin_lock_bh(&ct->lock);
c88130bc 850 old_state = ct->proto.tcp.state;
9fb9cbb1
YK
851 dir = CTINFO2DIR(ctinfo);
852 index = get_conntrack_index(th);
853 new_state = tcp_conntracks[dir][index][old_state];
c88130bc 854 tuple = &ct->tuplehash[dir].tuple;
9fb9cbb1
YK
855
856 switch (new_state) {
17311393
JK
857 case TCP_CONNTRACK_SYN_SENT:
858 if (old_state < TCP_CONNTRACK_TIME_WAIT)
859 break;
b2155e7f
JK
860 /* RFC 1122: "When a connection is closed actively,
861 * it MUST linger in TIME-WAIT state for a time 2xMSL
862 * (Maximum Segment Lifetime). However, it MAY accept
863 * a new SYN from the remote TCP to reopen the connection
864 * directly from TIME-WAIT state, if..."
865 * We ignore the conditions because we are in the
866 * TIME-WAIT state anyway.
867 *
868 * Handle aborted connections: we and the server
869 * think there is an existing connection but the client
870 * aborts it and starts a new one.
871 */
872 if (((ct->proto.tcp.seen[dir].flags
873 | ct->proto.tcp.seen[!dir].flags)
874 & IP_CT_TCP_FLAG_CLOSE_INIT)
c88130bc
PM
875 || (ct->proto.tcp.last_dir == dir
876 && ct->proto.tcp.last_index == TCP_RST_SET)) {
bc34b841
JK
877 /* Attempt to reopen a closed/aborted connection.
878 * Delete this connection and look up again. */
440f0d58 879 spin_unlock_bh(&ct->lock);
2aec609f 880
6b69fe0c
PM
881 /* Only repeat if we can actually remove the timer.
882 * Destruction may already be in progress in process
883 * context and we must give it a chance to terminate.
884 */
2aec609f 885 if (nf_ct_kill(ct))
6b69fe0c 886 return -NF_REPEAT;
ec8d5409 887 return NF_DROP;
17311393
JK
888 }
889 /* Fall through */
9fb9cbb1 890 case TCP_CONNTRACK_IGNORE:
73f30602 891 /* Ignored packets:
b2155e7f
JK
892 *
893 * Our connection entry may be out of sync, so ignore
894 * packets which may signal the real connection between
895 * the client and the server.
73f30602
JK
896 *
897 * a) SYN in ORIGINAL
898 * b) SYN/ACK in REPLY
601e68e1 899 * c) ACK in reply direction after initial SYN in original.
b2155e7f
JK
900 *
901 * If the ignored packet is invalid, the receiver will send
902 * a RST we'll catch below.
73f30602 903 */
9fb9cbb1 904 if (index == TCP_SYNACK_SET
c88130bc
PM
905 && ct->proto.tcp.last_index == TCP_SYN_SET
906 && ct->proto.tcp.last_dir != dir
907 && ntohl(th->ack_seq) == ct->proto.tcp.last_end) {
b2155e7f 908 /* b) This SYN/ACK acknowledges a SYN that we earlier
9fb9cbb1
YK
909 * ignored as invalid. This means that the client and
910 * the server are both in sync, while the firewall is
c4832c7b
PNA
911 * not. We get in sync from the previously annotated
912 * values.
9fb9cbb1 913 */
c4832c7b
PNA
914 old_state = TCP_CONNTRACK_SYN_SENT;
915 new_state = TCP_CONNTRACK_SYN_RECV;
916 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_end =
917 ct->proto.tcp.last_end;
918 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_maxend =
919 ct->proto.tcp.last_end;
920 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_maxwin =
921 ct->proto.tcp.last_win == 0 ?
922 1 : ct->proto.tcp.last_win;
923 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_scale =
924 ct->proto.tcp.last_wscale;
925 ct->proto.tcp.seen[ct->proto.tcp.last_dir].flags =
926 ct->proto.tcp.last_flags;
927 memset(&ct->proto.tcp.seen[dir], 0,
928 sizeof(struct ip_ct_tcp_state));
929 break;
9fb9cbb1 930 }
c88130bc
PM
931 ct->proto.tcp.last_index = index;
932 ct->proto.tcp.last_dir = dir;
933 ct->proto.tcp.last_seq = ntohl(th->seq);
934 ct->proto.tcp.last_end =
9fb9cbb1 935 segment_seq_plus_len(ntohl(th->seq), skb->len, dataoff, th);
c4832c7b
PNA
936 ct->proto.tcp.last_win = ntohs(th->window);
937
938 /* a) This is a SYN in ORIGINAL. The client and the server
939 * may be in sync but we are not. In that case, we annotate
940 * the TCP options and let the packet go through. If it is a
941 * valid SYN packet, the server will reply with a SYN/ACK, and
942 * then we'll get in sync. Otherwise, the server ignores it. */
943 if (index == TCP_SYN_SET && dir == IP_CT_DIR_ORIGINAL) {
944 struct ip_ct_tcp_state seen = {};
945
946 ct->proto.tcp.last_flags =
947 ct->proto.tcp.last_wscale = 0;
948 tcp_options(skb, dataoff, th, &seen);
949 if (seen.flags & IP_CT_TCP_FLAG_WINDOW_SCALE) {
950 ct->proto.tcp.last_flags |=
951 IP_CT_TCP_FLAG_WINDOW_SCALE;
952 ct->proto.tcp.last_wscale = seen.td_scale;
953 }
954 if (seen.flags & IP_CT_TCP_FLAG_SACK_PERM) {
955 ct->proto.tcp.last_flags |=
956 IP_CT_TCP_FLAG_SACK_PERM;
957 }
958 }
440f0d58 959 spin_unlock_bh(&ct->lock);
c2a2c7e0 960 if (LOG_INVALID(net, IPPROTO_TCP))
9fb9cbb1 961 nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
1a4ac987
PNA
962 "nf_ct_tcp: invalid packet ignored in "
963 "state %s ", tcp_conntrack_names[old_state]);
9fb9cbb1
YK
964 return NF_ACCEPT;
965 case TCP_CONNTRACK_MAX:
966 /* Invalid packet */
0d53778e
PM
967 pr_debug("nf_ct_tcp: Invalid dir=%i index=%u ostate=%u\n",
968 dir, get_conntrack_index(th), old_state);
440f0d58 969 spin_unlock_bh(&ct->lock);
c2a2c7e0 970 if (LOG_INVALID(net, IPPROTO_TCP))
9fb9cbb1
YK
971 nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
972 "nf_ct_tcp: invalid state ");
973 return -NF_ACCEPT;
9fb9cbb1 974 case TCP_CONNTRACK_CLOSE:
bfcaa502
JK
975 if (index == TCP_RST_SET
976 && (ct->proto.tcp.seen[!dir].flags & IP_CT_TCP_FLAG_MAXACK_SET)
977 && before(ntohl(th->seq), ct->proto.tcp.seen[!dir].td_maxack)) {
978 /* Invalid RST */
334a47f6 979 spin_unlock_bh(&ct->lock);
bfcaa502
JK
980 if (LOG_INVALID(net, IPPROTO_TCP))
981 nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
982 "nf_ct_tcp: invalid RST ");
983 return -NF_ACCEPT;
984 }
9fb9cbb1 985 if (index == TCP_RST_SET
c88130bc
PM
986 && ((test_bit(IPS_SEEN_REPLY_BIT, &ct->status)
987 && ct->proto.tcp.last_index == TCP_SYN_SET)
988 || (!test_bit(IPS_ASSURED_BIT, &ct->status)
989 && ct->proto.tcp.last_index == TCP_ACK_SET))
990 && ntohl(th->ack_seq) == ct->proto.tcp.last_end) {
93b1fae4 991 /* RST sent to invalid SYN or ACK we had let through
73f30602
JK
992 * at a) and c) above:
993 *
994 * a) SYN was in window then
995 * c) we hold a half-open connection.
996 *
997 * Delete our connection entry.
9fb9cbb1 998 * We skip window checking, because packet might ACK
73f30602 999 * segments we ignored. */
9fb9cbb1
YK
1000 goto in_window;
1001 }
93b1fae4 1002 /* Just fall through */
9fb9cbb1
YK
1003 default:
1004 /* Keep compilers happy. */
1005 break;
1006 }
1007
c88130bc 1008 if (!tcp_in_window(ct, &ct->proto.tcp, dir, index,
9fb9cbb1 1009 skb, dataoff, th, pf)) {
440f0d58 1010 spin_unlock_bh(&ct->lock);
9fb9cbb1
YK
1011 return -NF_ACCEPT;
1012 }
1013 in_window:
1014 /* From now on we have got in-window packets */
c88130bc
PM
1015 ct->proto.tcp.last_index = index;
1016 ct->proto.tcp.last_dir = dir;
9fb9cbb1 1017
0d53778e 1018 pr_debug("tcp_conntracks: ");
3c9fba65 1019 nf_ct_dump_tuple(tuple);
0d53778e
PM
1020 pr_debug("syn=%i ack=%i fin=%i rst=%i old=%i new=%i\n",
1021 (th->syn ? 1 : 0), (th->ack ? 1 : 0),
1022 (th->fin ? 1 : 0), (th->rst ? 1 : 0),
1023 old_state, new_state);
9fb9cbb1 1024
c88130bc 1025 ct->proto.tcp.state = new_state;
9fb9cbb1 1026 if (old_state != new_state
d0c1fd7a 1027 && new_state == TCP_CONNTRACK_FIN_WAIT)
c88130bc 1028 ct->proto.tcp.seen[dir].flags |= IP_CT_TCP_FLAG_CLOSE_INIT;
ae375044 1029
d2ba1fde 1030 if (ct->proto.tcp.retrans >= tn->tcp_max_retrans &&
2c8503f5
PNA
1031 timeouts[new_state] > timeouts[TCP_CONNTRACK_RETRANS])
1032 timeout = timeouts[TCP_CONNTRACK_RETRANS];
ae375044
PM
1033 else if ((ct->proto.tcp.seen[0].flags | ct->proto.tcp.seen[1].flags) &
1034 IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED &&
2c8503f5
PNA
1035 timeouts[new_state] > timeouts[TCP_CONNTRACK_UNACK])
1036 timeout = timeouts[TCP_CONNTRACK_UNACK];
ae375044 1037 else
2c8503f5 1038 timeout = timeouts[new_state];
440f0d58 1039 spin_unlock_bh(&ct->lock);
9fb9cbb1 1040
9fb9cbb1 1041 if (new_state != old_state)
a71996fc 1042 nf_conntrack_event_cache(IPCT_PROTOINFO, ct);
9fb9cbb1 1043
c88130bc 1044 if (!test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) {
9fb9cbb1
YK
1045 /* If only reply is a RST, we can consider ourselves not to
1046 have an established connection: this is a fairly common
1047 problem case, so we can delete the conntrack
1048 immediately. --RR */
1049 if (th->rst) {
718d4ad9 1050 nf_ct_kill_acct(ct, ctinfo, skb);
9fb9cbb1
YK
1051 return NF_ACCEPT;
1052 }
c88130bc 1053 } else if (!test_bit(IPS_ASSURED_BIT, &ct->status)
9fb9cbb1
YK
1054 && (old_state == TCP_CONNTRACK_SYN_RECV
1055 || old_state == TCP_CONNTRACK_ESTABLISHED)
1056 && new_state == TCP_CONNTRACK_ESTABLISHED) {
601e68e1
YH
1057 /* Set ASSURED if we see see valid ack in ESTABLISHED
1058 after SYN_RECV or a valid answer for a picked up
9fb9cbb1 1059 connection. */
c88130bc 1060 set_bit(IPS_ASSURED_BIT, &ct->status);
858b3133 1061 nf_conntrack_event_cache(IPCT_ASSURED, ct);
9fb9cbb1 1062 }
c88130bc 1063 nf_ct_refresh_acct(ct, ctinfo, skb, timeout);
9fb9cbb1
YK
1064
1065 return NF_ACCEPT;
1066}
601e68e1 1067
9fb9cbb1 1068/* Called when a new connection for this protocol found. */
09f263cd 1069static bool tcp_new(struct nf_conn *ct, const struct sk_buff *skb,
2c8503f5 1070 unsigned int dataoff, unsigned int *timeouts)
9fb9cbb1
YK
1071{
1072 enum tcp_conntrack new_state;
82f568fc
JE
1073 const struct tcphdr *th;
1074 struct tcphdr _tcph;
d2ba1fde
G
1075 struct net *net = nf_ct_net(ct);
1076 struct nf_tcp_net *tn = tcp_pernet(net);
82f568fc
JE
1077 const struct ip_ct_tcp_state *sender = &ct->proto.tcp.seen[0];
1078 const struct ip_ct_tcp_state *receiver = &ct->proto.tcp.seen[1];
9fb9cbb1
YK
1079
1080 th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph);
1081 BUG_ON(th == NULL);
1082
1083 /* Don't need lock here: this conntrack not in circulation yet */
e5fc9e7a 1084 new_state = tcp_conntracks[0][get_conntrack_index(th)][TCP_CONNTRACK_NONE];
9fb9cbb1
YK
1085
1086 /* Invalid: delete conntrack */
1087 if (new_state >= TCP_CONNTRACK_MAX) {
0d53778e 1088 pr_debug("nf_ct_tcp: invalid new deleting.\n");
09f263cd 1089 return false;
9fb9cbb1
YK
1090 }
1091
1092 if (new_state == TCP_CONNTRACK_SYN_SENT) {
e5fc9e7a 1093 memset(&ct->proto.tcp, 0, sizeof(ct->proto.tcp));
9fb9cbb1 1094 /* SYN packet */
c88130bc 1095 ct->proto.tcp.seen[0].td_end =
9fb9cbb1
YK
1096 segment_seq_plus_len(ntohl(th->seq), skb->len,
1097 dataoff, th);
c88130bc
PM
1098 ct->proto.tcp.seen[0].td_maxwin = ntohs(th->window);
1099 if (ct->proto.tcp.seen[0].td_maxwin == 0)
1100 ct->proto.tcp.seen[0].td_maxwin = 1;
1101 ct->proto.tcp.seen[0].td_maxend =
1102 ct->proto.tcp.seen[0].td_end;
1103
1104 tcp_options(skb, dataoff, th, &ct->proto.tcp.seen[0]);
d2ba1fde 1105 } else if (tn->tcp_loose == 0) {
9fb9cbb1 1106 /* Don't try to pick up connections. */
09f263cd 1107 return false;
9fb9cbb1 1108 } else {
e5fc9e7a 1109 memset(&ct->proto.tcp, 0, sizeof(ct->proto.tcp));
9fb9cbb1
YK
1110 /*
1111 * We are in the middle of a connection,
1112 * its history is lost for us.
1113 * Let's try to use the data from the packet.
1114 */
c88130bc 1115 ct->proto.tcp.seen[0].td_end =
9fb9cbb1
YK
1116 segment_seq_plus_len(ntohl(th->seq), skb->len,
1117 dataoff, th);
c88130bc
PM
1118 ct->proto.tcp.seen[0].td_maxwin = ntohs(th->window);
1119 if (ct->proto.tcp.seen[0].td_maxwin == 0)
1120 ct->proto.tcp.seen[0].td_maxwin = 1;
1121 ct->proto.tcp.seen[0].td_maxend =
1122 ct->proto.tcp.seen[0].td_end +
1123 ct->proto.tcp.seen[0].td_maxwin;
9fb9cbb1 1124
a09113c2
PM
1125 /* We assume SACK and liberal window checking to handle
1126 * window scaling */
c88130bc
PM
1127 ct->proto.tcp.seen[0].flags =
1128 ct->proto.tcp.seen[1].flags = IP_CT_TCP_FLAG_SACK_PERM |
1129 IP_CT_TCP_FLAG_BE_LIBERAL;
9fb9cbb1 1130 }
601e68e1 1131
9fb9cbb1 1132 /* tcp_packet will set them */
c88130bc 1133 ct->proto.tcp.last_index = TCP_NONE_SET;
601e68e1 1134
0d53778e
PM
1135 pr_debug("tcp_new: sender end=%u maxend=%u maxwin=%u scale=%i "
1136 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
1137 sender->td_end, sender->td_maxend, sender->td_maxwin,
1138 sender->td_scale,
1139 receiver->td_end, receiver->td_maxend, receiver->td_maxwin,
1140 receiver->td_scale);
09f263cd 1141 return true;
9fb9cbb1 1142}
c1d10adb 1143
c0cd1156 1144#if IS_ENABLED(CONFIG_NF_CT_NETLINK)
c1d10adb
PNA
1145
1146#include <linux/netfilter/nfnetlink.h>
1147#include <linux/netfilter/nfnetlink_conntrack.h>
1148
fdf70832 1149static int tcp_to_nlattr(struct sk_buff *skb, struct nlattr *nla,
440f0d58 1150 struct nf_conn *ct)
c1d10adb 1151{
df6fb868 1152 struct nlattr *nest_parms;
c8e2078c 1153 struct nf_ct_tcp_flags tmp = {};
601e68e1 1154
440f0d58 1155 spin_lock_bh(&ct->lock);
df6fb868
PM
1156 nest_parms = nla_nest_start(skb, CTA_PROTOINFO_TCP | NLA_F_NESTED);
1157 if (!nest_parms)
1158 goto nla_put_failure;
1159
4925a459
DM
1160 if (nla_put_u8(skb, CTA_PROTOINFO_TCP_STATE, ct->proto.tcp.state) ||
1161 nla_put_u8(skb, CTA_PROTOINFO_TCP_WSCALE_ORIGINAL,
1162 ct->proto.tcp.seen[0].td_scale) ||
1163 nla_put_u8(skb, CTA_PROTOINFO_TCP_WSCALE_REPLY,
1164 ct->proto.tcp.seen[1].td_scale))
1165 goto nla_put_failure;
c8e2078c
PNA
1166
1167 tmp.flags = ct->proto.tcp.seen[0].flags;
4925a459
DM
1168 if (nla_put(skb, CTA_PROTOINFO_TCP_FLAGS_ORIGINAL,
1169 sizeof(struct nf_ct_tcp_flags), &tmp))
1170 goto nla_put_failure;
c8e2078c
PNA
1171
1172 tmp.flags = ct->proto.tcp.seen[1].flags;
4925a459
DM
1173 if (nla_put(skb, CTA_PROTOINFO_TCP_FLAGS_REPLY,
1174 sizeof(struct nf_ct_tcp_flags), &tmp))
1175 goto nla_put_failure;
440f0d58 1176 spin_unlock_bh(&ct->lock);
c1d10adb 1177
df6fb868 1178 nla_nest_end(skb, nest_parms);
c1d10adb
PNA
1179
1180 return 0;
1181
df6fb868 1182nla_put_failure:
440f0d58 1183 spin_unlock_bh(&ct->lock);
c1d10adb
PNA
1184 return -1;
1185}
1186
f73e924c
PM
1187static const struct nla_policy tcp_nla_policy[CTA_PROTOINFO_TCP_MAX+1] = {
1188 [CTA_PROTOINFO_TCP_STATE] = { .type = NLA_U8 },
1189 [CTA_PROTOINFO_TCP_WSCALE_ORIGINAL] = { .type = NLA_U8 },
1190 [CTA_PROTOINFO_TCP_WSCALE_REPLY] = { .type = NLA_U8 },
1191 [CTA_PROTOINFO_TCP_FLAGS_ORIGINAL] = { .len = sizeof(struct nf_ct_tcp_flags) },
1192 [CTA_PROTOINFO_TCP_FLAGS_REPLY] = { .len = sizeof(struct nf_ct_tcp_flags) },
c1d10adb
PNA
1193};
1194
fdf70832 1195static int nlattr_to_tcp(struct nlattr *cda[], struct nf_conn *ct)
c1d10adb 1196{
2f0d2f10 1197 struct nlattr *pattr = cda[CTA_PROTOINFO_TCP];
df6fb868 1198 struct nlattr *tb[CTA_PROTOINFO_TCP_MAX+1];
f73e924c 1199 int err;
c1d10adb
PNA
1200
1201 /* updates could not contain anything about the private
1202 * protocol info, in that case skip the parsing */
2f0d2f10 1203 if (!pattr)
c1d10adb
PNA
1204 return 0;
1205
2f0d2f10 1206 err = nla_parse_nested(tb, CTA_PROTOINFO_TCP_MAX, pattr, tcp_nla_policy);
f73e924c
PM
1207 if (err < 0)
1208 return err;
c1d10adb 1209
5f7da4d2
PM
1210 if (tb[CTA_PROTOINFO_TCP_STATE] &&
1211 nla_get_u8(tb[CTA_PROTOINFO_TCP_STATE]) >= TCP_CONNTRACK_MAX)
c1d10adb
PNA
1212 return -EINVAL;
1213
440f0d58 1214 spin_lock_bh(&ct->lock);
5f7da4d2
PM
1215 if (tb[CTA_PROTOINFO_TCP_STATE])
1216 ct->proto.tcp.state = nla_get_u8(tb[CTA_PROTOINFO_TCP_STATE]);
c8e2078c 1217
df6fb868 1218 if (tb[CTA_PROTOINFO_TCP_FLAGS_ORIGINAL]) {
c8e2078c 1219 struct nf_ct_tcp_flags *attr =
df6fb868 1220 nla_data(tb[CTA_PROTOINFO_TCP_FLAGS_ORIGINAL]);
c8e2078c
PNA
1221 ct->proto.tcp.seen[0].flags &= ~attr->mask;
1222 ct->proto.tcp.seen[0].flags |= attr->flags & attr->mask;
1223 }
1224
df6fb868 1225 if (tb[CTA_PROTOINFO_TCP_FLAGS_REPLY]) {
c8e2078c 1226 struct nf_ct_tcp_flags *attr =
df6fb868 1227 nla_data(tb[CTA_PROTOINFO_TCP_FLAGS_REPLY]);
c8e2078c
PNA
1228 ct->proto.tcp.seen[1].flags &= ~attr->mask;
1229 ct->proto.tcp.seen[1].flags |= attr->flags & attr->mask;
1230 }
1231
df6fb868
PM
1232 if (tb[CTA_PROTOINFO_TCP_WSCALE_ORIGINAL] &&
1233 tb[CTA_PROTOINFO_TCP_WSCALE_REPLY] &&
c8e2078c
PNA
1234 ct->proto.tcp.seen[0].flags & IP_CT_TCP_FLAG_WINDOW_SCALE &&
1235 ct->proto.tcp.seen[1].flags & IP_CT_TCP_FLAG_WINDOW_SCALE) {
77236b6e
PM
1236 ct->proto.tcp.seen[0].td_scale =
1237 nla_get_u8(tb[CTA_PROTOINFO_TCP_WSCALE_ORIGINAL]);
1238 ct->proto.tcp.seen[1].td_scale =
1239 nla_get_u8(tb[CTA_PROTOINFO_TCP_WSCALE_REPLY]);
c8e2078c 1240 }
440f0d58 1241 spin_unlock_bh(&ct->lock);
c1d10adb
PNA
1242
1243 return 0;
1244}
a400c30e
HE
1245
1246static int tcp_nlattr_size(void)
1247{
1248 return nla_total_size(0) /* CTA_PROTOINFO_TCP */
1249 + nla_policy_len(tcp_nla_policy, CTA_PROTOINFO_TCP_MAX + 1);
1250}
1251
1252static int tcp_nlattr_tuple_size(void)
1253{
1254 return nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1);
1255}
c1d10adb 1256#endif
933a41e7 1257
50978462
PNA
1258#if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
1259
1260#include <linux/netfilter/nfnetlink.h>
1261#include <linux/netfilter/nfnetlink_cttimeout.h>
1262
8264deb8
G
1263static int tcp_timeout_nlattr_to_obj(struct nlattr *tb[],
1264 struct net *net, void *data)
50978462
PNA
1265{
1266 unsigned int *timeouts = data;
8264deb8 1267 struct nf_tcp_net *tn = tcp_pernet(net);
50978462
PNA
1268 int i;
1269
1270 /* set default TCP timeouts. */
1271 for (i=0; i<TCP_CONNTRACK_TIMEOUT_MAX; i++)
8264deb8 1272 timeouts[i] = tn->timeouts[i];
50978462
PNA
1273
1274 if (tb[CTA_TIMEOUT_TCP_SYN_SENT]) {
1275 timeouts[TCP_CONNTRACK_SYN_SENT] =
1276 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_SYN_SENT]))*HZ;
1277 }
1278 if (tb[CTA_TIMEOUT_TCP_SYN_RECV]) {
1279 timeouts[TCP_CONNTRACK_SYN_RECV] =
1280 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_SYN_RECV]))*HZ;
1281 }
1282 if (tb[CTA_TIMEOUT_TCP_ESTABLISHED]) {
1283 timeouts[TCP_CONNTRACK_ESTABLISHED] =
1284 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_ESTABLISHED]))*HZ;
1285 }
1286 if (tb[CTA_TIMEOUT_TCP_FIN_WAIT]) {
1287 timeouts[TCP_CONNTRACK_FIN_WAIT] =
1288 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_FIN_WAIT]))*HZ;
1289 }
1290 if (tb[CTA_TIMEOUT_TCP_CLOSE_WAIT]) {
1291 timeouts[TCP_CONNTRACK_CLOSE_WAIT] =
1292 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_CLOSE_WAIT]))*HZ;
1293 }
1294 if (tb[CTA_TIMEOUT_TCP_LAST_ACK]) {
1295 timeouts[TCP_CONNTRACK_LAST_ACK] =
1296 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_LAST_ACK]))*HZ;
1297 }
1298 if (tb[CTA_TIMEOUT_TCP_TIME_WAIT]) {
1299 timeouts[TCP_CONNTRACK_TIME_WAIT] =
1300 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_TIME_WAIT]))*HZ;
1301 }
1302 if (tb[CTA_TIMEOUT_TCP_CLOSE]) {
1303 timeouts[TCP_CONNTRACK_CLOSE] =
1304 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_CLOSE]))*HZ;
1305 }
1306 if (tb[CTA_TIMEOUT_TCP_SYN_SENT2]) {
1307 timeouts[TCP_CONNTRACK_SYN_SENT2] =
1308 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_SYN_SENT2]))*HZ;
1309 }
1310 if (tb[CTA_TIMEOUT_TCP_RETRANS]) {
1311 timeouts[TCP_CONNTRACK_RETRANS] =
1312 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_RETRANS]))*HZ;
1313 }
1314 if (tb[CTA_TIMEOUT_TCP_UNACK]) {
1315 timeouts[TCP_CONNTRACK_UNACK] =
1316 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_UNACK]))*HZ;
1317 }
1318 return 0;
1319}
1320
1321static int
1322tcp_timeout_obj_to_nlattr(struct sk_buff *skb, const void *data)
1323{
1324 const unsigned int *timeouts = data;
1325
4925a459
DM
1326 if (nla_put_be32(skb, CTA_TIMEOUT_TCP_SYN_SENT,
1327 htonl(timeouts[TCP_CONNTRACK_SYN_SENT] / HZ)) ||
1328 nla_put_be32(skb, CTA_TIMEOUT_TCP_SYN_RECV,
1329 htonl(timeouts[TCP_CONNTRACK_SYN_RECV] / HZ)) ||
1330 nla_put_be32(skb, CTA_TIMEOUT_TCP_ESTABLISHED,
1331 htonl(timeouts[TCP_CONNTRACK_ESTABLISHED] / HZ)) ||
1332 nla_put_be32(skb, CTA_TIMEOUT_TCP_FIN_WAIT,
1333 htonl(timeouts[TCP_CONNTRACK_FIN_WAIT] / HZ)) ||
1334 nla_put_be32(skb, CTA_TIMEOUT_TCP_CLOSE_WAIT,
1335 htonl(timeouts[TCP_CONNTRACK_CLOSE_WAIT] / HZ)) ||
1336 nla_put_be32(skb, CTA_TIMEOUT_TCP_LAST_ACK,
1337 htonl(timeouts[TCP_CONNTRACK_LAST_ACK] / HZ)) ||
1338 nla_put_be32(skb, CTA_TIMEOUT_TCP_TIME_WAIT,
1339 htonl(timeouts[TCP_CONNTRACK_TIME_WAIT] / HZ)) ||
1340 nla_put_be32(skb, CTA_TIMEOUT_TCP_CLOSE,
1341 htonl(timeouts[TCP_CONNTRACK_CLOSE] / HZ)) ||
1342 nla_put_be32(skb, CTA_TIMEOUT_TCP_SYN_SENT2,
1343 htonl(timeouts[TCP_CONNTRACK_SYN_SENT2] / HZ)) ||
1344 nla_put_be32(skb, CTA_TIMEOUT_TCP_RETRANS,
1345 htonl(timeouts[TCP_CONNTRACK_RETRANS] / HZ)) ||
1346 nla_put_be32(skb, CTA_TIMEOUT_TCP_UNACK,
1347 htonl(timeouts[TCP_CONNTRACK_UNACK] / HZ)))
1348 goto nla_put_failure;
50978462
PNA
1349 return 0;
1350
1351nla_put_failure:
1352 return -ENOSPC;
1353}
1354
1355static const struct nla_policy tcp_timeout_nla_policy[CTA_TIMEOUT_TCP_MAX+1] = {
1356 [CTA_TIMEOUT_TCP_SYN_SENT] = { .type = NLA_U32 },
1357 [CTA_TIMEOUT_TCP_SYN_RECV] = { .type = NLA_U32 },
1358 [CTA_TIMEOUT_TCP_ESTABLISHED] = { .type = NLA_U32 },
1359 [CTA_TIMEOUT_TCP_FIN_WAIT] = { .type = NLA_U32 },
1360 [CTA_TIMEOUT_TCP_CLOSE_WAIT] = { .type = NLA_U32 },
1361 [CTA_TIMEOUT_TCP_LAST_ACK] = { .type = NLA_U32 },
1362 [CTA_TIMEOUT_TCP_TIME_WAIT] = { .type = NLA_U32 },
1363 [CTA_TIMEOUT_TCP_CLOSE] = { .type = NLA_U32 },
1364 [CTA_TIMEOUT_TCP_SYN_SENT2] = { .type = NLA_U32 },
1365};
1366#endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
1367
933a41e7 1368#ifdef CONFIG_SYSCTL
933a41e7
PM
1369static struct ctl_table tcp_sysctl_table[] = {
1370 {
933a41e7 1371 .procname = "nf_conntrack_tcp_timeout_syn_sent",
933a41e7
PM
1372 .maxlen = sizeof(unsigned int),
1373 .mode = 0644,
6d9f239a 1374 .proc_handler = proc_dointvec_jiffies,
933a41e7
PM
1375 },
1376 {
933a41e7 1377 .procname = "nf_conntrack_tcp_timeout_syn_recv",
933a41e7
PM
1378 .maxlen = sizeof(unsigned int),
1379 .mode = 0644,
6d9f239a 1380 .proc_handler = proc_dointvec_jiffies,
933a41e7
PM
1381 },
1382 {
933a41e7 1383 .procname = "nf_conntrack_tcp_timeout_established",
933a41e7
PM
1384 .maxlen = sizeof(unsigned int),
1385 .mode = 0644,
6d9f239a 1386 .proc_handler = proc_dointvec_jiffies,
933a41e7
PM
1387 },
1388 {
933a41e7 1389 .procname = "nf_conntrack_tcp_timeout_fin_wait",
933a41e7
PM
1390 .maxlen = sizeof(unsigned int),
1391 .mode = 0644,
6d9f239a 1392 .proc_handler = proc_dointvec_jiffies,
933a41e7
PM
1393 },
1394 {
933a41e7 1395 .procname = "nf_conntrack_tcp_timeout_close_wait",
933a41e7
PM
1396 .maxlen = sizeof(unsigned int),
1397 .mode = 0644,
6d9f239a 1398 .proc_handler = proc_dointvec_jiffies,
933a41e7
PM
1399 },
1400 {
933a41e7 1401 .procname = "nf_conntrack_tcp_timeout_last_ack",
933a41e7
PM
1402 .maxlen = sizeof(unsigned int),
1403 .mode = 0644,
6d9f239a 1404 .proc_handler = proc_dointvec_jiffies,
933a41e7
PM
1405 },
1406 {
933a41e7 1407 .procname = "nf_conntrack_tcp_timeout_time_wait",
933a41e7
PM
1408 .maxlen = sizeof(unsigned int),
1409 .mode = 0644,
6d9f239a 1410 .proc_handler = proc_dointvec_jiffies,
933a41e7
PM
1411 },
1412 {
933a41e7 1413 .procname = "nf_conntrack_tcp_timeout_close",
933a41e7
PM
1414 .maxlen = sizeof(unsigned int),
1415 .mode = 0644,
6d9f239a 1416 .proc_handler = proc_dointvec_jiffies,
933a41e7
PM
1417 },
1418 {
933a41e7 1419 .procname = "nf_conntrack_tcp_timeout_max_retrans",
933a41e7
PM
1420 .maxlen = sizeof(unsigned int),
1421 .mode = 0644,
6d9f239a 1422 .proc_handler = proc_dointvec_jiffies,
933a41e7 1423 },
ae375044
PM
1424 {
1425 .procname = "nf_conntrack_tcp_timeout_unacknowledged",
ae375044
PM
1426 .maxlen = sizeof(unsigned int),
1427 .mode = 0644,
6d9f239a 1428 .proc_handler = proc_dointvec_jiffies,
ae375044 1429 },
933a41e7 1430 {
933a41e7 1431 .procname = "nf_conntrack_tcp_loose",
933a41e7
PM
1432 .maxlen = sizeof(unsigned int),
1433 .mode = 0644,
6d9f239a 1434 .proc_handler = proc_dointvec,
933a41e7
PM
1435 },
1436 {
933a41e7 1437 .procname = "nf_conntrack_tcp_be_liberal",
933a41e7
PM
1438 .maxlen = sizeof(unsigned int),
1439 .mode = 0644,
6d9f239a 1440 .proc_handler = proc_dointvec,
933a41e7
PM
1441 },
1442 {
933a41e7 1443 .procname = "nf_conntrack_tcp_max_retrans",
933a41e7
PM
1444 .maxlen = sizeof(unsigned int),
1445 .mode = 0644,
6d9f239a 1446 .proc_handler = proc_dointvec,
933a41e7 1447 },
f8572d8f 1448 { }
933a41e7 1449};
a999e683
PM
1450
1451#ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT
1452static struct ctl_table tcp_compat_sysctl_table[] = {
1453 {
a999e683 1454 .procname = "ip_conntrack_tcp_timeout_syn_sent",
a999e683
PM
1455 .maxlen = sizeof(unsigned int),
1456 .mode = 0644,
6d9f239a 1457 .proc_handler = proc_dointvec_jiffies,
a999e683 1458 },
874ab923
JK
1459 {
1460 .procname = "ip_conntrack_tcp_timeout_syn_sent2",
874ab923
JK
1461 .maxlen = sizeof(unsigned int),
1462 .mode = 0644,
1463 .proc_handler = proc_dointvec_jiffies,
1464 },
a999e683 1465 {
a999e683 1466 .procname = "ip_conntrack_tcp_timeout_syn_recv",
a999e683
PM
1467 .maxlen = sizeof(unsigned int),
1468 .mode = 0644,
6d9f239a 1469 .proc_handler = proc_dointvec_jiffies,
a999e683
PM
1470 },
1471 {
a999e683 1472 .procname = "ip_conntrack_tcp_timeout_established",
a999e683
PM
1473 .maxlen = sizeof(unsigned int),
1474 .mode = 0644,
6d9f239a 1475 .proc_handler = proc_dointvec_jiffies,
a999e683
PM
1476 },
1477 {
a999e683 1478 .procname = "ip_conntrack_tcp_timeout_fin_wait",
a999e683
PM
1479 .maxlen = sizeof(unsigned int),
1480 .mode = 0644,
6d9f239a 1481 .proc_handler = proc_dointvec_jiffies,
a999e683
PM
1482 },
1483 {
a999e683 1484 .procname = "ip_conntrack_tcp_timeout_close_wait",
a999e683
PM
1485 .maxlen = sizeof(unsigned int),
1486 .mode = 0644,
6d9f239a 1487 .proc_handler = proc_dointvec_jiffies,
a999e683
PM
1488 },
1489 {
a999e683 1490 .procname = "ip_conntrack_tcp_timeout_last_ack",
a999e683
PM
1491 .maxlen = sizeof(unsigned int),
1492 .mode = 0644,
6d9f239a 1493 .proc_handler = proc_dointvec_jiffies,
a999e683
PM
1494 },
1495 {
a999e683 1496 .procname = "ip_conntrack_tcp_timeout_time_wait",
a999e683
PM
1497 .maxlen = sizeof(unsigned int),
1498 .mode = 0644,
6d9f239a 1499 .proc_handler = proc_dointvec_jiffies,
a999e683
PM
1500 },
1501 {
a999e683 1502 .procname = "ip_conntrack_tcp_timeout_close",
a999e683
PM
1503 .maxlen = sizeof(unsigned int),
1504 .mode = 0644,
6d9f239a 1505 .proc_handler = proc_dointvec_jiffies,
a999e683
PM
1506 },
1507 {
a999e683 1508 .procname = "ip_conntrack_tcp_timeout_max_retrans",
a999e683
PM
1509 .maxlen = sizeof(unsigned int),
1510 .mode = 0644,
6d9f239a 1511 .proc_handler = proc_dointvec_jiffies,
a999e683
PM
1512 },
1513 {
a999e683 1514 .procname = "ip_conntrack_tcp_loose",
a999e683
PM
1515 .maxlen = sizeof(unsigned int),
1516 .mode = 0644,
6d9f239a 1517 .proc_handler = proc_dointvec,
a999e683
PM
1518 },
1519 {
a999e683 1520 .procname = "ip_conntrack_tcp_be_liberal",
a999e683
PM
1521 .maxlen = sizeof(unsigned int),
1522 .mode = 0644,
6d9f239a 1523 .proc_handler = proc_dointvec,
a999e683
PM
1524 },
1525 {
a999e683 1526 .procname = "ip_conntrack_tcp_max_retrans",
a999e683
PM
1527 .maxlen = sizeof(unsigned int),
1528 .mode = 0644,
6d9f239a 1529 .proc_handler = proc_dointvec,
a999e683 1530 },
f8572d8f 1531 { }
a999e683
PM
1532};
1533#endif /* CONFIG_NF_CONNTRACK_PROC_COMPAT */
933a41e7
PM
1534#endif /* CONFIG_SYSCTL */
1535
d2ba1fde
G
1536static int tcp_kmemdup_sysctl_table(struct nf_proto_net *pn)
1537{
1538#ifdef CONFIG_SYSCTL
1539 struct nf_tcp_net *tn = (struct nf_tcp_net *)pn;
1540
1541 if (pn->ctl_table)
1542 return 0;
1543
1544 pn->ctl_table = kmemdup(tcp_sysctl_table,
1545 sizeof(tcp_sysctl_table),
1546 GFP_KERNEL);
1547 if (!pn->ctl_table)
1548 return -ENOMEM;
1549
1550 pn->ctl_table[0].data = &tn->timeouts[TCP_CONNTRACK_SYN_SENT];
1551 pn->ctl_table[1].data = &tn->timeouts[TCP_CONNTRACK_SYN_RECV];
1552 pn->ctl_table[2].data = &tn->timeouts[TCP_CONNTRACK_ESTABLISHED];
1553 pn->ctl_table[3].data = &tn->timeouts[TCP_CONNTRACK_FIN_WAIT];
1554 pn->ctl_table[4].data = &tn->timeouts[TCP_CONNTRACK_CLOSE_WAIT];
1555 pn->ctl_table[5].data = &tn->timeouts[TCP_CONNTRACK_LAST_ACK];
1556 pn->ctl_table[6].data = &tn->timeouts[TCP_CONNTRACK_TIME_WAIT];
1557 pn->ctl_table[7].data = &tn->timeouts[TCP_CONNTRACK_CLOSE];
1558 pn->ctl_table[8].data = &tn->timeouts[TCP_CONNTRACK_RETRANS];
1559 pn->ctl_table[9].data = &tn->timeouts[TCP_CONNTRACK_UNACK];
1560 pn->ctl_table[10].data = &tn->tcp_loose;
1561 pn->ctl_table[11].data = &tn->tcp_be_liberal;
1562 pn->ctl_table[12].data = &tn->tcp_max_retrans;
1563#endif
1564 return 0;
1565}
1566
1567static int tcp_kmemdup_compat_sysctl_table(struct nf_proto_net *pn)
1568{
1569#ifdef CONFIG_SYSCTL
1570#ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT
1571 struct nf_tcp_net *tn = (struct nf_tcp_net *)pn;
1572 pn->ctl_compat_table = kmemdup(tcp_compat_sysctl_table,
1573 sizeof(tcp_compat_sysctl_table),
1574 GFP_KERNEL);
1575 if (!pn->ctl_compat_table)
1576 return -ENOMEM;
1577
1578 pn->ctl_compat_table[0].data = &tn->timeouts[TCP_CONNTRACK_SYN_SENT];
1579 pn->ctl_compat_table[1].data = &tn->timeouts[TCP_CONNTRACK_SYN_SENT2];
1580 pn->ctl_compat_table[2].data = &tn->timeouts[TCP_CONNTRACK_SYN_RECV];
1581 pn->ctl_compat_table[3].data = &tn->timeouts[TCP_CONNTRACK_ESTABLISHED];
1582 pn->ctl_compat_table[4].data = &tn->timeouts[TCP_CONNTRACK_FIN_WAIT];
1583 pn->ctl_compat_table[5].data = &tn->timeouts[TCP_CONNTRACK_CLOSE_WAIT];
1584 pn->ctl_compat_table[6].data = &tn->timeouts[TCP_CONNTRACK_LAST_ACK];
1585 pn->ctl_compat_table[7].data = &tn->timeouts[TCP_CONNTRACK_TIME_WAIT];
1586 pn->ctl_compat_table[8].data = &tn->timeouts[TCP_CONNTRACK_CLOSE];
1587 pn->ctl_compat_table[9].data = &tn->timeouts[TCP_CONNTRACK_RETRANS];
1588 pn->ctl_compat_table[10].data = &tn->tcp_loose;
1589 pn->ctl_compat_table[11].data = &tn->tcp_be_liberal;
1590 pn->ctl_compat_table[12].data = &tn->tcp_max_retrans;
1591#endif
1592#endif
1593 return 0;
1594}
1595
f1caad27 1596static int tcpv4_init_net(struct net *net, u_int16_t proto)
d2ba1fde
G
1597{
1598 int i;
1599 int ret = 0;
1600 struct nf_tcp_net *tn = tcp_pernet(net);
1601 struct nf_proto_net *pn = (struct nf_proto_net *)tn;
1602
1603#ifdef CONFIG_SYSCTL
1604 if (!pn->ctl_table) {
1605#else
352e04b9 1606 if (!pn->users++) {
d2ba1fde
G
1607#endif
1608 for (i = 0; i < TCP_CONNTRACK_TIMEOUT_MAX; i++)
1609 tn->timeouts[i] = tcp_timeouts[i];
1610
1611 tn->tcp_loose = nf_ct_tcp_loose;
1612 tn->tcp_be_liberal = nf_ct_tcp_be_liberal;
1613 tn->tcp_max_retrans = nf_ct_tcp_max_retrans;
1614 }
1615
1616 ret = tcp_kmemdup_compat_sysctl_table(pn);
1617
1618 if (ret < 0)
1619 return ret;
1620
1621 ret = tcp_kmemdup_sysctl_table(pn);
1622
1623#ifdef CONFIG_SYSCTL
1624#ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT
1625 if (ret < 0) {
1626 kfree(pn->ctl_compat_table);
1627 pn->ctl_compat_table = NULL;
1628 }
1629#endif
1630#endif
1631 return ret;
1632}
1633
f1caad27 1634static int tcpv6_init_net(struct net *net, u_int16_t proto)
d2ba1fde
G
1635{
1636 int i;
1637 struct nf_tcp_net *tn = tcp_pernet(net);
1638 struct nf_proto_net *pn = (struct nf_proto_net *)tn;
1639
1640#ifdef CONFIG_SYSCTL
1641 if (!pn->ctl_table) {
1642#else
352e04b9 1643 if (!pn->users++) {
d2ba1fde
G
1644#endif
1645 for (i = 0; i < TCP_CONNTRACK_TIMEOUT_MAX; i++)
1646 tn->timeouts[i] = tcp_timeouts[i];
1647 tn->tcp_loose = nf_ct_tcp_loose;
1648 tn->tcp_be_liberal = nf_ct_tcp_be_liberal;
1649 tn->tcp_max_retrans = nf_ct_tcp_max_retrans;
1650 }
1651
1652 return tcp_kmemdup_sysctl_table(pn);
1653}
1654
61075af5 1655struct nf_conntrack_l4proto nf_conntrack_l4proto_tcp4 __read_mostly =
9fb9cbb1
YK
1656{
1657 .l3proto = PF_INET,
605dcad6 1658 .l4proto = IPPROTO_TCP,
9fb9cbb1
YK
1659 .name = "tcp",
1660 .pkt_to_tuple = tcp_pkt_to_tuple,
1661 .invert_tuple = tcp_invert_tuple,
1662 .print_tuple = tcp_print_tuple,
1663 .print_conntrack = tcp_print_conntrack,
1664 .packet = tcp_packet,
2c8503f5 1665 .get_timeouts = tcp_get_timeouts,
9fb9cbb1 1666 .new = tcp_new,
96f6bf82 1667 .error = tcp_error,
c0cd1156 1668#if IS_ENABLED(CONFIG_NF_CT_NETLINK)
fdf70832 1669 .to_nlattr = tcp_to_nlattr,
a400c30e 1670 .nlattr_size = tcp_nlattr_size,
fdf70832
PM
1671 .from_nlattr = nlattr_to_tcp,
1672 .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr,
1673 .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple,
a400c30e 1674 .nlattr_tuple_size = tcp_nlattr_tuple_size,
f73e924c 1675 .nla_policy = nf_ct_port_nla_policy,
c1d10adb 1676#endif
50978462
PNA
1677#if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
1678 .ctnl_timeout = {
1679 .nlattr_to_obj = tcp_timeout_nlattr_to_obj,
1680 .obj_to_nlattr = tcp_timeout_obj_to_nlattr,
1681 .nlattr_max = CTA_TIMEOUT_TCP_MAX,
1682 .obj_size = sizeof(unsigned int) *
1683 TCP_CONNTRACK_TIMEOUT_MAX,
1684 .nla_policy = tcp_timeout_nla_policy,
1685 },
1686#endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
d2ba1fde 1687 .init_net = tcpv4_init_net,
9fb9cbb1 1688};
13b18339 1689EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_tcp4);
9fb9cbb1 1690
61075af5 1691struct nf_conntrack_l4proto nf_conntrack_l4proto_tcp6 __read_mostly =
9fb9cbb1
YK
1692{
1693 .l3proto = PF_INET6,
605dcad6 1694 .l4proto = IPPROTO_TCP,
9fb9cbb1
YK
1695 .name = "tcp",
1696 .pkt_to_tuple = tcp_pkt_to_tuple,
1697 .invert_tuple = tcp_invert_tuple,
1698 .print_tuple = tcp_print_tuple,
1699 .print_conntrack = tcp_print_conntrack,
1700 .packet = tcp_packet,
2c8503f5 1701 .get_timeouts = tcp_get_timeouts,
9fb9cbb1 1702 .new = tcp_new,
96f6bf82 1703 .error = tcp_error,
c0cd1156 1704#if IS_ENABLED(CONFIG_NF_CT_NETLINK)
fdf70832 1705 .to_nlattr = tcp_to_nlattr,
a400c30e 1706 .nlattr_size = tcp_nlattr_size,
fdf70832
PM
1707 .from_nlattr = nlattr_to_tcp,
1708 .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr,
1709 .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple,
a400c30e 1710 .nlattr_tuple_size = tcp_nlattr_tuple_size,
f73e924c 1711 .nla_policy = nf_ct_port_nla_policy,
c1d10adb 1712#endif
50978462
PNA
1713#if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
1714 .ctnl_timeout = {
1715 .nlattr_to_obj = tcp_timeout_nlattr_to_obj,
1716 .obj_to_nlattr = tcp_timeout_obj_to_nlattr,
1717 .nlattr_max = CTA_TIMEOUT_TCP_MAX,
1718 .obj_size = sizeof(unsigned int) *
1719 TCP_CONNTRACK_TIMEOUT_MAX,
1720 .nla_policy = tcp_timeout_nla_policy,
1721 },
1722#endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
d2ba1fde 1723 .init_net = tcpv6_init_net,
9fb9cbb1 1724};
13b18339 1725EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_tcp6);