xen-blkfront: don't add indirect pages to list when !feature_persistent
[linux-2.6-block.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>
f229f6ce
PM
3 * (C) 2002-2013 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
4 * (C) 2006-2012 Patrick McHardy <kaber@trash.net>
9fb9cbb1
YK
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
9fb9cbb1
YK
9 */
10
9fb9cbb1 11#include <linux/types.h>
9fb9cbb1 12#include <linux/timer.h>
9fb9cbb1
YK
13#include <linux/module.h>
14#include <linux/in.h>
15#include <linux/tcp.h>
16#include <linux/spinlock.h>
17#include <linux/skbuff.h>
18#include <linux/ipv6.h>
19#include <net/ip6_checksum.h>
534f81a5 20#include <asm/unaligned.h>
9fb9cbb1
YK
21
22#include <net/tcp.h>
23
24#include <linux/netfilter.h>
25#include <linux/netfilter_ipv4.h>
26#include <linux/netfilter_ipv6.h>
27#include <net/netfilter/nf_conntrack.h>
605dcad6 28#include <net/netfilter/nf_conntrack_l4proto.h>
f6180121 29#include <net/netfilter/nf_conntrack_ecache.h>
41d73ec0 30#include <net/netfilter/nf_conntrack_seqadj.h>
48b1de4c 31#include <net/netfilter/nf_conntrack_synproxy.h>
f01ffbd6 32#include <net/netfilter/nf_log.h>
9d2493f8
CP
33#include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
34#include <net/netfilter/ipv6/nf_conntrack_ipv6.h>
9fb9cbb1 35
601e68e1
YH
36/* "Be conservative in what you do,
37 be liberal in what you accept from others."
9fb9cbb1 38 If it's non-zero, we mark only out of window RST segments as INVALID. */
3aef0fd9 39static int nf_ct_tcp_be_liberal __read_mostly = 0;
9fb9cbb1 40
a09113c2 41/* If it is set to zero, we disable picking up already established
9fb9cbb1 42 connections. */
3aef0fd9 43static int nf_ct_tcp_loose __read_mostly = 1;
9fb9cbb1 44
601e68e1
YH
45/* Max number of the retransmitted packets without receiving an (acceptable)
46 ACK from the destination. If this number is reached, a shorter timer
9fb9cbb1 47 will be started. */
3aef0fd9 48static int nf_ct_tcp_max_retrans __read_mostly = 3;
9fb9cbb1
YK
49
50 /* FIXME: Examine ipfilter's timeouts and conntrack transitions more
51 closely. They're more complex. --RR */
52
82f568fc 53static const char *const tcp_conntrack_names[] = {
9fb9cbb1
YK
54 "NONE",
55 "SYN_SENT",
56 "SYN_RECV",
57 "ESTABLISHED",
58 "FIN_WAIT",
59 "CLOSE_WAIT",
60 "LAST_ACK",
61 "TIME_WAIT",
62 "CLOSE",
874ab923 63 "SYN_SENT2",
9fb9cbb1 64};
601e68e1 65
9fb9cbb1
YK
66#define SECS * HZ
67#define MINS * 60 SECS
68#define HOURS * 60 MINS
69#define DAYS * 24 HOURS
70
33ee4464 71static unsigned int tcp_timeouts[TCP_CONNTRACK_TIMEOUT_MAX] __read_mostly = {
2d646286
PM
72 [TCP_CONNTRACK_SYN_SENT] = 2 MINS,
73 [TCP_CONNTRACK_SYN_RECV] = 60 SECS,
74 [TCP_CONNTRACK_ESTABLISHED] = 5 DAYS,
75 [TCP_CONNTRACK_FIN_WAIT] = 2 MINS,
76 [TCP_CONNTRACK_CLOSE_WAIT] = 60 SECS,
77 [TCP_CONNTRACK_LAST_ACK] = 30 SECS,
78 [TCP_CONNTRACK_TIME_WAIT] = 2 MINS,
79 [TCP_CONNTRACK_CLOSE] = 10 SECS,
874ab923 80 [TCP_CONNTRACK_SYN_SENT2] = 2 MINS,
33ee4464
PNA
81/* RFC1122 says the R2 limit should be at least 100 seconds.
82 Linux uses 15 packets as limit, which corresponds
83 to ~13-30min depending on RTO. */
84 [TCP_CONNTRACK_RETRANS] = 5 MINS,
85 [TCP_CONNTRACK_UNACK] = 5 MINS,
2d646286 86};
601e68e1 87
9fb9cbb1
YK
88#define sNO TCP_CONNTRACK_NONE
89#define sSS TCP_CONNTRACK_SYN_SENT
90#define sSR TCP_CONNTRACK_SYN_RECV
91#define sES TCP_CONNTRACK_ESTABLISHED
92#define sFW TCP_CONNTRACK_FIN_WAIT
93#define sCW TCP_CONNTRACK_CLOSE_WAIT
94#define sLA TCP_CONNTRACK_LAST_ACK
95#define sTW TCP_CONNTRACK_TIME_WAIT
96#define sCL TCP_CONNTRACK_CLOSE
874ab923 97#define sS2 TCP_CONNTRACK_SYN_SENT2
9fb9cbb1
YK
98#define sIV TCP_CONNTRACK_MAX
99#define sIG TCP_CONNTRACK_IGNORE
100
101/* What TCP flags are set from RST/SYN/FIN/ACK. */
102enum tcp_bit_set {
103 TCP_SYN_SET,
104 TCP_SYNACK_SET,
105 TCP_FIN_SET,
106 TCP_ACK_SET,
107 TCP_RST_SET,
108 TCP_NONE_SET,
109};
601e68e1 110
9fb9cbb1
YK
111/*
112 * The TCP state transition table needs a few words...
113 *
114 * We are the man in the middle. All the packets go through us
115 * but might get lost in transit to the destination.
601e68e1 116 * It is assumed that the destinations can't receive segments
9fb9cbb1
YK
117 * we haven't seen.
118 *
119 * The checked segment is in window, but our windows are *not*
120 * equivalent with the ones of the sender/receiver. We always
121 * try to guess the state of the current sender.
122 *
123 * The meaning of the states are:
124 *
125 * NONE: initial state
601e68e1 126 * SYN_SENT: SYN-only packet seen
874ab923 127 * SYN_SENT2: SYN-only packet seen from reply dir, simultaneous open
9fb9cbb1
YK
128 * SYN_RECV: SYN-ACK packet seen
129 * ESTABLISHED: ACK packet seen
130 * FIN_WAIT: FIN packet seen
601e68e1 131 * CLOSE_WAIT: ACK seen (after FIN)
9fb9cbb1
YK
132 * LAST_ACK: FIN seen (after FIN)
133 * TIME_WAIT: last ACK seen
b2155e7f 134 * CLOSE: closed connection (RST)
9fb9cbb1 135 *
9fb9cbb1 136 * Packets marked as IGNORED (sIG):
601e68e1
YH
137 * if they may be either invalid or valid
138 * and the receiver may send back a connection
9fb9cbb1
YK
139 * closing RST or a SYN/ACK.
140 *
141 * Packets marked as INVALID (sIV):
874ab923 142 * if we regard them as truly invalid packets
9fb9cbb1 143 */
a5e73c29 144static const u8 tcp_conntracks[2][6][TCP_CONNTRACK_MAX] = {
9fb9cbb1
YK
145 {
146/* ORIGINAL */
874ab923
JK
147/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
148/*syn*/ { sSS, sSS, sIG, sIG, sIG, sIG, sIG, sSS, sSS, sS2 },
9fb9cbb1
YK
149/*
150 * sNO -> sSS Initialize a new connection
151 * sSS -> sSS Retransmitted SYN
874ab923
JK
152 * sS2 -> sS2 Late retransmitted SYN
153 * sSR -> sIG
9fb9cbb1 154 * sES -> sIG Error: SYNs in window outside the SYN_SENT state
601e68e1 155 * are errors. Receiver will reply with RST
9fb9cbb1
YK
156 * and close the connection.
157 * Or we are not in sync and hold a dead connection.
158 * sFW -> sIG
159 * sCW -> sIG
160 * sLA -> sIG
161 * sTW -> sSS Reopened connection (RFC 1122).
162 * sCL -> sSS
163 */
874ab923 164/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
64f509ce 165/*synack*/ { sIV, sIV, sSR, sIV, sIV, sIV, sIV, sIV, sIV, sSR },
9fb9cbb1 166/*
874ab923
JK
167 * sNO -> sIV Too late and no reason to do anything
168 * sSS -> sIV Client can't send SYN and then SYN/ACK
169 * sS2 -> sSR SYN/ACK sent to SYN2 in simultaneous open
64f509ce
JK
170 * sSR -> sSR Late retransmitted SYN/ACK in simultaneous open
171 * sES -> sIV Invalid SYN/ACK packets sent by the client
172 * sFW -> sIV
173 * sCW -> sIV
174 * sLA -> sIV
175 * sTW -> sIV
176 * sCL -> sIV
9fb9cbb1 177 */
874ab923 178/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
9fb9cbb1
YK
179/*fin*/ { sIV, sIV, sFW, sFW, sLA, sLA, sLA, sTW, sCL, sIV },
180/*
181 * sNO -> sIV Too late and no reason to do anything...
182 * sSS -> sIV Client migth not send FIN in this state:
183 * we enforce waiting for a SYN/ACK reply first.
874ab923 184 * sS2 -> sIV
9fb9cbb1
YK
185 * sSR -> sFW Close started.
186 * sES -> sFW
187 * sFW -> sLA FIN seen in both directions, waiting for
601e68e1 188 * the last ACK.
9fb9cbb1
YK
189 * Migth be a retransmitted FIN as well...
190 * sCW -> sLA
191 * sLA -> sLA Retransmitted FIN. Remain in the same state.
192 * sTW -> sTW
193 * sCL -> sCL
194 */
874ab923 195/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
9fb9cbb1
YK
196/*ack*/ { sES, sIV, sES, sES, sCW, sCW, sTW, sTW, sCL, sIV },
197/*
198 * sNO -> sES Assumed.
199 * sSS -> sIV ACK is invalid: we haven't seen a SYN/ACK yet.
874ab923 200 * sS2 -> sIV
9fb9cbb1
YK
201 * sSR -> sES Established state is reached.
202 * sES -> sES :-)
203 * sFW -> sCW Normal close request answered by ACK.
204 * sCW -> sCW
205 * sLA -> sTW Last ACK detected.
206 * sTW -> sTW Retransmitted last ACK. Remain in the same state.
207 * sCL -> sCL
208 */
874ab923
JK
209/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
210/*rst*/ { sIV, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL },
9fb9cbb1
YK
211/*none*/ { sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV }
212 },
213 {
214/* REPLY */
874ab923 215/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
e37ad9fd 216/*syn*/ { sIV, sS2, sIV, sIV, sIV, sIV, sIV, sSS, sIV, sS2 },
9fb9cbb1
YK
217/*
218 * sNO -> sIV Never reached.
874ab923
JK
219 * sSS -> sS2 Simultaneous open
220 * sS2 -> sS2 Retransmitted simultaneous SYN
221 * sSR -> sIV Invalid SYN packets sent by the server
222 * sES -> sIV
9fb9cbb1
YK
223 * sFW -> sIV
224 * sCW -> sIV
225 * sLA -> sIV
e37ad9fd 226 * sTW -> sSS Reopened connection, but server may have switched role
9fb9cbb1
YK
227 * sCL -> sIV
228 */
874ab923 229/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
8a80c79a 230/*synack*/ { sIV, sSR, sIG, sIG, sIG, sIG, sIG, sIG, sIG, sSR },
9fb9cbb1
YK
231/*
232 * sSS -> sSR Standard open.
874ab923 233 * sS2 -> sSR Simultaneous open
8a80c79a 234 * sSR -> sIG Retransmitted SYN/ACK, ignore it.
9fb9cbb1
YK
235 * sES -> sIG Late retransmitted SYN/ACK?
236 * sFW -> sIG Might be SYN/ACK answering ignored SYN
237 * sCW -> sIG
238 * sLA -> sIG
239 * sTW -> sIG
240 * sCL -> sIG
241 */
874ab923 242/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
9fb9cbb1
YK
243/*fin*/ { sIV, sIV, sFW, sFW, sLA, sLA, sLA, sTW, sCL, sIV },
244/*
245 * sSS -> sIV Server might not send FIN in this state.
874ab923 246 * sS2 -> sIV
9fb9cbb1
YK
247 * sSR -> sFW Close started.
248 * sES -> sFW
249 * sFW -> sLA FIN seen in both directions.
250 * sCW -> sLA
251 * sLA -> sLA Retransmitted FIN.
252 * sTW -> sTW
253 * sCL -> sCL
254 */
874ab923
JK
255/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
256/*ack*/ { sIV, sIG, sSR, sES, sCW, sCW, sTW, sTW, sCL, sIG },
9fb9cbb1 257/*
73f30602 258 * sSS -> sIG Might be a half-open connection.
874ab923 259 * sS2 -> sIG
9fb9cbb1
YK
260 * sSR -> sSR Might answer late resent SYN.
261 * sES -> sES :-)
262 * sFW -> sCW Normal close request answered by ACK.
263 * sCW -> sCW
264 * sLA -> sTW Last ACK detected.
265 * sTW -> sTW Retransmitted last ACK.
266 * sCL -> sCL
267 */
874ab923
JK
268/* sNO, sSS, sSR, sES, sFW, sCW, sLA, sTW, sCL, sS2 */
269/*rst*/ { sIV, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL, sCL },
9fb9cbb1 270/*none*/ { sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV }
601e68e1 271 }
9fb9cbb1
YK
272};
273
d2ba1fde
G
274static inline struct nf_tcp_net *tcp_pernet(struct net *net)
275{
276 return &net->ct.nf_ct_proto.tcp;
277}
278
09f263cd
JE
279static bool tcp_pkt_to_tuple(const struct sk_buff *skb, unsigned int dataoff,
280 struct nf_conntrack_tuple *tuple)
9fb9cbb1 281{
82f568fc
JE
282 const struct tcphdr *hp;
283 struct tcphdr _hdr;
9fb9cbb1
YK
284
285 /* Actually only need first 8 bytes. */
286 hp = skb_header_pointer(skb, dataoff, 8, &_hdr);
287 if (hp == NULL)
09f263cd 288 return false;
9fb9cbb1
YK
289
290 tuple->src.u.tcp.port = hp->source;
291 tuple->dst.u.tcp.port = hp->dest;
292
09f263cd 293 return true;
9fb9cbb1
YK
294}
295
09f263cd
JE
296static bool tcp_invert_tuple(struct nf_conntrack_tuple *tuple,
297 const struct nf_conntrack_tuple *orig)
9fb9cbb1
YK
298{
299 tuple->src.u.tcp.port = orig->dst.u.tcp.port;
300 tuple->dst.u.tcp.port = orig->src.u.tcp.port;
09f263cd 301 return true;
9fb9cbb1
YK
302}
303
304/* Print out the per-protocol part of the tuple. */
824f1fbe
JP
305static void tcp_print_tuple(struct seq_file *s,
306 const struct nf_conntrack_tuple *tuple)
9fb9cbb1 307{
824f1fbe
JP
308 seq_printf(s, "sport=%hu dport=%hu ",
309 ntohs(tuple->src.u.tcp.port),
310 ntohs(tuple->dst.u.tcp.port));
9fb9cbb1
YK
311}
312
313/* Print out the private part of the conntrack. */
37246a58 314static void tcp_print_conntrack(struct seq_file *s, struct nf_conn *ct)
9fb9cbb1
YK
315{
316 enum tcp_conntrack state;
317
440f0d58 318 spin_lock_bh(&ct->lock);
c88130bc 319 state = ct->proto.tcp.state;
440f0d58 320 spin_unlock_bh(&ct->lock);
9fb9cbb1 321
37246a58 322 seq_printf(s, "%s ", tcp_conntrack_names[state]);
9fb9cbb1
YK
323}
324
325static unsigned int get_conntrack_index(const struct tcphdr *tcph)
326{
327 if (tcph->rst) return TCP_RST_SET;
328 else if (tcph->syn) return (tcph->ack ? TCP_SYNACK_SET : TCP_SYN_SET);
329 else if (tcph->fin) return TCP_FIN_SET;
330 else if (tcph->ack) return TCP_ACK_SET;
331 else return TCP_NONE_SET;
332}
333
334/* TCP connection tracking based on 'Real Stateful TCP Packet Filtering
335 in IP Filter' by Guido van Rooij.
601e68e1 336
631dd1a8
JM
337 http://www.sane.nl/events/sane2000/papers.html
338 http://www.darkart.com/mirrors/www.obfuscation.org/ipf/
601e68e1 339
9fb9cbb1
YK
340 The boundaries and the conditions are changed according to RFC793:
341 the packet must intersect the window (i.e. segments may be
342 after the right or before the left edge) and thus receivers may ACK
343 segments after the right edge of the window.
344
601e68e1 345 td_maxend = max(sack + max(win,1)) seen in reply packets
9fb9cbb1
YK
346 td_maxwin = max(max(win, 1)) + (sack - ack) seen in sent packets
347 td_maxwin += seq + len - sender.td_maxend
348 if seq + len > sender.td_maxend
349 td_end = max(seq + len) seen in sent packets
601e68e1 350
9fb9cbb1
YK
351 I. Upper bound for valid data: seq <= sender.td_maxend
352 II. Lower bound for valid data: seq + len >= sender.td_end - receiver.td_maxwin
84ebe1cd
JK
353 III. Upper bound for valid (s)ack: sack <= receiver.td_end
354 IV. Lower bound for valid (s)ack: sack >= receiver.td_end - MAXACKWINDOW
9fb9cbb1 355
84ebe1cd
JK
356 where sack is the highest right edge of sack block found in the packet
357 or ack in the case of packet without SACK option.
9fb9cbb1 358
84ebe1cd 359 The upper bound limit for a valid (s)ack is not ignored -
601e68e1 360 we doesn't have to deal with fragments.
9fb9cbb1
YK
361*/
362
363static inline __u32 segment_seq_plus_len(__u32 seq,
364 size_t len,
365 unsigned int dataoff,
82f568fc 366 const struct tcphdr *tcph)
9fb9cbb1
YK
367{
368 /* XXX Should I use payload length field in IP/IPv6 header ?
369 * - YK */
370 return (seq + len - dataoff - tcph->doff*4
371 + (tcph->syn ? 1 : 0) + (tcph->fin ? 1 : 0));
372}
601e68e1 373
9fb9cbb1
YK
374/* Fixme: what about big packets? */
375#define MAXACKWINCONST 66000
376#define MAXACKWINDOW(sender) \
377 ((sender)->td_maxwin > MAXACKWINCONST ? (sender)->td_maxwin \
378 : MAXACKWINCONST)
601e68e1 379
9fb9cbb1
YK
380/*
381 * Simplified tcp_parse_options routine from tcp_input.c
382 */
383static void tcp_options(const struct sk_buff *skb,
384 unsigned int dataoff,
82f568fc 385 const struct tcphdr *tcph,
9fb9cbb1
YK
386 struct ip_ct_tcp_state *state)
387{
388 unsigned char buff[(15 * 4) - sizeof(struct tcphdr)];
82f568fc 389 const unsigned char *ptr;
9fb9cbb1
YK
390 int length = (tcph->doff*4) - sizeof(struct tcphdr);
391
392 if (!length)
393 return;
394
395 ptr = skb_header_pointer(skb, dataoff + sizeof(struct tcphdr),
396 length, buff);
397 BUG_ON(ptr == NULL);
398
601e68e1 399 state->td_scale =
9fb9cbb1
YK
400 state->flags = 0;
401
402 while (length > 0) {
403 int opcode=*ptr++;
404 int opsize;
405
406 switch (opcode) {
407 case TCPOPT_EOL:
408 return;
409 case TCPOPT_NOP: /* Ref: RFC 793 section 3.1 */
410 length--;
411 continue;
412 default:
413 opsize=*ptr++;
414 if (opsize < 2) /* "silly options" */
415 return;
416 if (opsize > length)
4a5cc84a 417 return; /* don't parse partial options */
9fb9cbb1 418
601e68e1 419 if (opcode == TCPOPT_SACK_PERM
9fb9cbb1
YK
420 && opsize == TCPOLEN_SACK_PERM)
421 state->flags |= IP_CT_TCP_FLAG_SACK_PERM;
422 else if (opcode == TCPOPT_WINDOW
423 && opsize == TCPOLEN_WINDOW) {
424 state->td_scale = *(u_int8_t *)ptr;
425
426 if (state->td_scale > 14) {
427 /* See RFC1323 */
428 state->td_scale = 14;
429 }
430 state->flags |=
431 IP_CT_TCP_FLAG_WINDOW_SCALE;
432 }
433 ptr += opsize - 2;
434 length -= opsize;
435 }
436 }
437}
438
439static void tcp_sack(const struct sk_buff *skb, unsigned int dataoff,
82f568fc 440 const struct tcphdr *tcph, __u32 *sack)
9fb9cbb1 441{
601e68e1 442 unsigned char buff[(15 * 4) - sizeof(struct tcphdr)];
82f568fc 443 const unsigned char *ptr;
9fb9cbb1
YK
444 int length = (tcph->doff*4) - sizeof(struct tcphdr);
445 __u32 tmp;
446
447 if (!length)
448 return;
449
450 ptr = skb_header_pointer(skb, dataoff + sizeof(struct tcphdr),
451 length, buff);
452 BUG_ON(ptr == NULL);
453
454 /* Fast path for timestamp-only option */
bb9fc373 455 if (length == TCPOLEN_TSTAMP_ALIGNED
8f05ce91
YH
456 && *(__be32 *)ptr == htonl((TCPOPT_NOP << 24)
457 | (TCPOPT_NOP << 16)
458 | (TCPOPT_TIMESTAMP << 8)
459 | TCPOLEN_TIMESTAMP))
9fb9cbb1
YK
460 return;
461
462 while (length > 0) {
463 int opcode = *ptr++;
464 int opsize, i;
465
466 switch (opcode) {
467 case TCPOPT_EOL:
468 return;
469 case TCPOPT_NOP: /* Ref: RFC 793 section 3.1 */
470 length--;
471 continue;
472 default:
473 opsize = *ptr++;
474 if (opsize < 2) /* "silly options" */
475 return;
476 if (opsize > length)
4a5cc84a 477 return; /* don't parse partial options */
9fb9cbb1 478
601e68e1
YH
479 if (opcode == TCPOPT_SACK
480 && opsize >= (TCPOLEN_SACK_BASE
481 + TCPOLEN_SACK_PERBLOCK)
482 && !((opsize - TCPOLEN_SACK_BASE)
483 % TCPOLEN_SACK_PERBLOCK)) {
484 for (i = 0;
485 i < (opsize - TCPOLEN_SACK_BASE);
486 i += TCPOLEN_SACK_PERBLOCK) {
534f81a5 487 tmp = get_unaligned_be32((__be32 *)(ptr+i)+1);
9fb9cbb1
YK
488
489 if (after(tmp, *sack))
490 *sack = tmp;
491 }
492 return;
493 }
494 ptr += opsize - 2;
495 length -= opsize;
496 }
497 }
498}
499
09f263cd
JE
500static bool tcp_in_window(const struct nf_conn *ct,
501 struct ip_ct_tcp *state,
502 enum ip_conntrack_dir dir,
503 unsigned int index,
504 const struct sk_buff *skb,
505 unsigned int dataoff,
506 const struct tcphdr *tcph,
76108cea 507 u_int8_t pf)
9fb9cbb1 508{
c2a2c7e0 509 struct net *net = nf_ct_net(ct);
d2ba1fde 510 struct nf_tcp_net *tn = tcp_pernet(net);
9fb9cbb1
YK
511 struct ip_ct_tcp_state *sender = &state->seen[dir];
512 struct ip_ct_tcp_state *receiver = &state->seen[!dir];
82f568fc 513 const struct nf_conntrack_tuple *tuple = &ct->tuplehash[dir].tuple;
9fb9cbb1 514 __u32 seq, ack, sack, end, win, swin;
2d89c68a 515 s32 receiver_offset;
356d7d88 516 bool res, in_recv_win;
9fb9cbb1
YK
517
518 /*
519 * Get the required data from the packet.
520 */
521 seq = ntohl(tcph->seq);
522 ack = sack = ntohl(tcph->ack_seq);
523 win = ntohs(tcph->window);
524 end = segment_seq_plus_len(seq, skb->len, dataoff, tcph);
525
526 if (receiver->flags & IP_CT_TCP_FLAG_SACK_PERM)
527 tcp_sack(skb, dataoff, tcph, &sack);
528
f9dd09c7 529 /* Take into account NAT sequence number mangling */
41d73ec0 530 receiver_offset = nf_ct_seq_offset(ct, !dir, ack - 1);
f9dd09c7
JK
531 ack -= receiver_offset;
532 sack -= receiver_offset;
533
0d53778e
PM
534 pr_debug("tcp_in_window: START\n");
535 pr_debug("tcp_in_window: ");
3c9fba65 536 nf_ct_dump_tuple(tuple);
f9dd09c7
JK
537 pr_debug("seq=%u ack=%u+(%d) sack=%u+(%d) win=%u end=%u\n",
538 seq, ack, receiver_offset, sack, receiver_offset, win, end);
0d53778e
PM
539 pr_debug("tcp_in_window: sender end=%u maxend=%u maxwin=%u scale=%i "
540 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
541 sender->td_end, sender->td_maxend, sender->td_maxwin,
542 sender->td_scale,
543 receiver->td_end, receiver->td_maxend, receiver->td_maxwin,
544 receiver->td_scale);
9fb9cbb1 545
874ab923 546 if (sender->td_maxwin == 0) {
9fb9cbb1
YK
547 /*
548 * Initialize sender data.
549 */
874ab923 550 if (tcph->syn) {
9fb9cbb1 551 /*
874ab923
JK
552 * SYN-ACK in reply to a SYN
553 * or SYN from reply direction in simultaneous open.
9fb9cbb1 554 */
601e68e1 555 sender->td_end =
9fb9cbb1
YK
556 sender->td_maxend = end;
557 sender->td_maxwin = (win == 0 ? 1 : win);
558
559 tcp_options(skb, dataoff, tcph, sender);
601e68e1 560 /*
9fb9cbb1
YK
561 * RFC 1323:
562 * Both sides must send the Window Scale option
563 * to enable window scaling in either direction.
564 */
565 if (!(sender->flags & IP_CT_TCP_FLAG_WINDOW_SCALE
566 && receiver->flags & IP_CT_TCP_FLAG_WINDOW_SCALE))
601e68e1 567 sender->td_scale =
9fb9cbb1 568 receiver->td_scale = 0;
874ab923
JK
569 if (!tcph->ack)
570 /* Simultaneous open */
571 return true;
9fb9cbb1
YK
572 } else {
573 /*
574 * We are in the middle of a connection,
575 * its history is lost for us.
576 * Let's try to use the data from the packet.
601e68e1 577 */
9fb9cbb1 578 sender->td_end = end;
6ee0b693
CG
579 swin = win << sender->td_scale;
580 sender->td_maxwin = (swin == 0 ? 1 : swin);
9fb9cbb1 581 sender->td_maxend = end + sender->td_maxwin;
fac42a9a
PNA
582 /*
583 * We haven't seen traffic in the other direction yet
584 * but we have to tweak window tracking to pass III
585 * and IV until that happens.
586 */
587 if (receiver->td_maxwin == 0)
588 receiver->td_end = receiver->td_maxend = sack;
9fb9cbb1
YK
589 }
590 } else if (((state->state == TCP_CONNTRACK_SYN_SENT
591 && dir == IP_CT_DIR_ORIGINAL)
592 || (state->state == TCP_CONNTRACK_SYN_RECV
593 && dir == IP_CT_DIR_REPLY))
594 && after(end, sender->td_end)) {
595 /*
596 * RFC 793: "if a TCP is reinitialized ... then it need
601e68e1 597 * not wait at all; it must only be sure to use sequence
9fb9cbb1
YK
598 * numbers larger than those recently used."
599 */
600 sender->td_end =
601 sender->td_maxend = end;
602 sender->td_maxwin = (win == 0 ? 1 : win);
603
604 tcp_options(skb, dataoff, tcph, sender);
605 }
606
607 if (!(tcph->ack)) {
608 /*
609 * If there is no ACK, just pretend it was set and OK.
610 */
611 ack = sack = receiver->td_end;
601e68e1
YH
612 } else if (((tcp_flag_word(tcph) & (TCP_FLAG_ACK|TCP_FLAG_RST)) ==
613 (TCP_FLAG_ACK|TCP_FLAG_RST))
9fb9cbb1
YK
614 && (ack == 0)) {
615 /*
616 * Broken TCP stacks, that set ACK in RST packets as well
617 * with zero ack value.
618 */
619 ack = sack = receiver->td_end;
620 }
621
4a70bbfa 622 if (tcph->rst && seq == 0 && state->state == TCP_CONNTRACK_SYN_SENT)
9fb9cbb1 623 /*
4a70bbfa 624 * RST sent answering SYN.
9fb9cbb1
YK
625 */
626 seq = end = sender->td_end;
627
0d53778e 628 pr_debug("tcp_in_window: ");
3c9fba65 629 nf_ct_dump_tuple(tuple);
f9dd09c7
JK
630 pr_debug("seq=%u ack=%u+(%d) sack=%u+(%d) win=%u end=%u\n",
631 seq, ack, receiver_offset, sack, receiver_offset, win, end);
0d53778e
PM
632 pr_debug("tcp_in_window: sender end=%u maxend=%u maxwin=%u scale=%i "
633 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
634 sender->td_end, sender->td_maxend, sender->td_maxwin,
635 sender->td_scale,
636 receiver->td_end, receiver->td_maxend, receiver->td_maxwin,
637 receiver->td_scale);
638
356d7d88
YC
639 /* Is the ending sequence in the receive window (if available)? */
640 in_recv_win = !receiver->td_maxwin ||
641 after(end, sender->td_end - receiver->td_maxwin - 1);
642
0d53778e
PM
643 pr_debug("tcp_in_window: I=%i II=%i III=%i IV=%i\n",
644 before(seq, sender->td_maxend + 1),
356d7d88 645 (in_recv_win ? 1 : 0),
0d53778e 646 before(sack, receiver->td_end + 1),
84ebe1cd 647 after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1));
9fb9cbb1 648
a09113c2 649 if (before(seq, sender->td_maxend + 1) &&
356d7d88 650 in_recv_win &&
a09113c2 651 before(sack, receiver->td_end + 1) &&
84ebe1cd 652 after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1)) {
601e68e1 653 /*
9fb9cbb1
YK
654 * Take into account window scaling (RFC 1323).
655 */
656 if (!tcph->syn)
657 win <<= sender->td_scale;
658
659 /*
660 * Update sender data.
661 */
662 swin = win + (sack - ack);
663 if (sender->td_maxwin < swin)
664 sender->td_maxwin = swin;
ae375044 665 if (after(end, sender->td_end)) {
9fb9cbb1 666 sender->td_end = end;
ae375044
PM
667 sender->flags |= IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED;
668 }
bfcaa502
JK
669 if (tcph->ack) {
670 if (!(sender->flags & IP_CT_TCP_FLAG_MAXACK_SET)) {
671 sender->td_maxack = ack;
672 sender->flags |= IP_CT_TCP_FLAG_MAXACK_SET;
673 } else if (after(ack, sender->td_maxack))
674 sender->td_maxack = ack;
675 }
676
9fb9cbb1
YK
677 /*
678 * Update receiver data.
679 */
fac42a9a 680 if (receiver->td_maxwin != 0 && after(end, sender->td_maxend))
9fb9cbb1
YK
681 receiver->td_maxwin += end - sender->td_maxend;
682 if (after(sack + win, receiver->td_maxend - 1)) {
683 receiver->td_maxend = sack + win;
684 if (win == 0)
685 receiver->td_maxend++;
686 }
ae375044
PM
687 if (ack == receiver->td_end)
688 receiver->flags &= ~IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED;
9fb9cbb1 689
601e68e1 690 /*
9fb9cbb1
YK
691 * Check retransmissions.
692 */
693 if (index == TCP_ACK_SET) {
694 if (state->last_dir == dir
695 && state->last_seq == seq
696 && state->last_ack == ack
c1fe3ca5
GH
697 && state->last_end == end
698 && state->last_win == win)
9fb9cbb1
YK
699 state->retrans++;
700 else {
701 state->last_dir = dir;
702 state->last_seq = seq;
703 state->last_ack = ack;
704 state->last_end = end;
c1fe3ca5 705 state->last_win = win;
9fb9cbb1
YK
706 state->retrans = 0;
707 }
708 }
09f263cd 709 res = true;
9fb9cbb1 710 } else {
09f263cd 711 res = false;
a09113c2 712 if (sender->flags & IP_CT_TCP_FLAG_BE_LIBERAL ||
d2ba1fde 713 tn->tcp_be_liberal)
09f263cd 714 res = true;
c2a2c7e0 715 if (!res && LOG_INVALID(net, IPPROTO_TCP))
30e0c6a6 716 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
9fb9cbb1
YK
717 "nf_ct_tcp: %s ",
718 before(seq, sender->td_maxend + 1) ?
356d7d88 719 in_recv_win ?
9fb9cbb1 720 before(sack, receiver->td_end + 1) ?
f9dd09c7 721 after(sack, receiver->td_end - MAXACKWINDOW(sender) - 1) ? "BUG"
9fb9cbb1
YK
722 : "ACK is under the lower bound (possible overly delayed ACK)"
723 : "ACK is over the upper bound (ACKed data not seen yet)"
724 : "SEQ is under the lower bound (already ACKed data retransmitted)"
725 : "SEQ is over the upper bound (over the window of the receiver)");
601e68e1
YH
726 }
727
09f263cd 728 pr_debug("tcp_in_window: res=%u sender end=%u maxend=%u maxwin=%u "
0d53778e
PM
729 "receiver end=%u maxend=%u maxwin=%u\n",
730 res, sender->td_end, sender->td_maxend, sender->td_maxwin,
731 receiver->td_end, receiver->td_maxend, receiver->td_maxwin);
9fb9cbb1
YK
732
733 return res;
734}
735
5c8ce7c9 736/* table of valid flag combinations - PUSH, ECE and CWR are always valid */
a3433f35
CG
737static const u8 tcp_valid_flags[(TCPHDR_FIN|TCPHDR_SYN|TCPHDR_RST|TCPHDR_ACK|
738 TCPHDR_URG) + 1] =
9fb9cbb1 739{
a3433f35
CG
740 [TCPHDR_SYN] = 1,
741 [TCPHDR_SYN|TCPHDR_URG] = 1,
742 [TCPHDR_SYN|TCPHDR_ACK] = 1,
743 [TCPHDR_RST] = 1,
744 [TCPHDR_RST|TCPHDR_ACK] = 1,
745 [TCPHDR_FIN|TCPHDR_ACK] = 1,
746 [TCPHDR_FIN|TCPHDR_ACK|TCPHDR_URG] = 1,
747 [TCPHDR_ACK] = 1,
748 [TCPHDR_ACK|TCPHDR_URG] = 1,
9fb9cbb1
YK
749};
750
751/* Protect conntrack agaist broken packets. Code taken from ipt_unclean.c. */
8fea97ec 752static int tcp_error(struct net *net, struct nf_conn *tmpl,
74c51a14 753 struct sk_buff *skb,
9fb9cbb1
YK
754 unsigned int dataoff,
755 enum ip_conntrack_info *ctinfo,
76108cea 756 u_int8_t pf,
96f6bf82 757 unsigned int hooknum)
9fb9cbb1 758{
82f568fc
JE
759 const struct tcphdr *th;
760 struct tcphdr _tcph;
9fb9cbb1
YK
761 unsigned int tcplen = skb->len - dataoff;
762 u_int8_t tcpflags;
763
764 /* Smaller that minimal TCP header? */
765 th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph);
766 if (th == NULL) {
c2a2c7e0 767 if (LOG_INVALID(net, IPPROTO_TCP))
30e0c6a6 768 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
9fb9cbb1
YK
769 "nf_ct_tcp: short packet ");
770 return -NF_ACCEPT;
601e68e1
YH
771 }
772
9fb9cbb1
YK
773 /* Not whole TCP header or malformed packet */
774 if (th->doff*4 < sizeof(struct tcphdr) || tcplen < th->doff*4) {
c2a2c7e0 775 if (LOG_INVALID(net, IPPROTO_TCP))
30e0c6a6 776 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
9fb9cbb1
YK
777 "nf_ct_tcp: truncated/malformed packet ");
778 return -NF_ACCEPT;
779 }
601e68e1 780
9fb9cbb1
YK
781 /* Checksum invalid? Ignore.
782 * We skip checking packets on the outgoing path
84fa7933 783 * because the checksum is assumed to be correct.
9fb9cbb1
YK
784 */
785 /* FIXME: Source route IP option packets --RR */
c04d0552 786 if (net->ct.sysctl_checksum && hooknum == NF_INET_PRE_ROUTING &&
96f6bf82 787 nf_checksum(skb, hooknum, dataoff, IPPROTO_TCP, pf)) {
c2a2c7e0 788 if (LOG_INVALID(net, IPPROTO_TCP))
30e0c6a6 789 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
9fb9cbb1
YK
790 "nf_ct_tcp: bad TCP checksum ");
791 return -NF_ACCEPT;
792 }
793
794 /* Check TCP flags. */
a3433f35 795 tcpflags = (tcp_flag_byte(th) & ~(TCPHDR_ECE|TCPHDR_CWR|TCPHDR_PSH));
9fb9cbb1 796 if (!tcp_valid_flags[tcpflags]) {
c2a2c7e0 797 if (LOG_INVALID(net, IPPROTO_TCP))
30e0c6a6 798 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
9fb9cbb1
YK
799 "nf_ct_tcp: invalid TCP flag combination ");
800 return -NF_ACCEPT;
801 }
802
803 return NF_ACCEPT;
804}
805
2c8503f5
PNA
806static unsigned int *tcp_get_timeouts(struct net *net)
807{
be0593c6 808 return tcp_pernet(net)->timeouts;
2c8503f5
PNA
809}
810
9fb9cbb1 811/* Returns verdict for packet, or -1 for invalid. */
c88130bc 812static int tcp_packet(struct nf_conn *ct,
9fb9cbb1
YK
813 const struct sk_buff *skb,
814 unsigned int dataoff,
815 enum ip_conntrack_info ctinfo,
76108cea 816 u_int8_t pf,
2c8503f5
PNA
817 unsigned int hooknum,
818 unsigned int *timeouts)
9fb9cbb1 819{
c2a2c7e0 820 struct net *net = nf_ct_net(ct);
d2ba1fde 821 struct nf_tcp_net *tn = tcp_pernet(net);
0d53778e 822 struct nf_conntrack_tuple *tuple;
9fb9cbb1
YK
823 enum tcp_conntrack new_state, old_state;
824 enum ip_conntrack_dir dir;
82f568fc
JE
825 const struct tcphdr *th;
826 struct tcphdr _tcph;
9fb9cbb1
YK
827 unsigned long timeout;
828 unsigned int index;
829
830 th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph);
831 BUG_ON(th == NULL);
832
440f0d58 833 spin_lock_bh(&ct->lock);
c88130bc 834 old_state = ct->proto.tcp.state;
9fb9cbb1
YK
835 dir = CTINFO2DIR(ctinfo);
836 index = get_conntrack_index(th);
837 new_state = tcp_conntracks[dir][index][old_state];
c88130bc 838 tuple = &ct->tuplehash[dir].tuple;
9fb9cbb1
YK
839
840 switch (new_state) {
17311393
JK
841 case TCP_CONNTRACK_SYN_SENT:
842 if (old_state < TCP_CONNTRACK_TIME_WAIT)
843 break;
b2155e7f
JK
844 /* RFC 1122: "When a connection is closed actively,
845 * it MUST linger in TIME-WAIT state for a time 2xMSL
846 * (Maximum Segment Lifetime). However, it MAY accept
847 * a new SYN from the remote TCP to reopen the connection
848 * directly from TIME-WAIT state, if..."
849 * We ignore the conditions because we are in the
850 * TIME-WAIT state anyway.
851 *
852 * Handle aborted connections: we and the server
853 * think there is an existing connection but the client
854 * aborts it and starts a new one.
855 */
856 if (((ct->proto.tcp.seen[dir].flags
857 | ct->proto.tcp.seen[!dir].flags)
858 & IP_CT_TCP_FLAG_CLOSE_INIT)
c88130bc
PM
859 || (ct->proto.tcp.last_dir == dir
860 && ct->proto.tcp.last_index == TCP_RST_SET)) {
bc34b841
JK
861 /* Attempt to reopen a closed/aborted connection.
862 * Delete this connection and look up again. */
440f0d58 863 spin_unlock_bh(&ct->lock);
2aec609f 864
6b69fe0c
PM
865 /* Only repeat if we can actually remove the timer.
866 * Destruction may already be in progress in process
867 * context and we must give it a chance to terminate.
868 */
2aec609f 869 if (nf_ct_kill(ct))
6b69fe0c 870 return -NF_REPEAT;
ec8d5409 871 return NF_DROP;
17311393
JK
872 }
873 /* Fall through */
9fb9cbb1 874 case TCP_CONNTRACK_IGNORE:
73f30602 875 /* Ignored packets:
b2155e7f
JK
876 *
877 * Our connection entry may be out of sync, so ignore
878 * packets which may signal the real connection between
879 * the client and the server.
73f30602
JK
880 *
881 * a) SYN in ORIGINAL
882 * b) SYN/ACK in REPLY
601e68e1 883 * c) ACK in reply direction after initial SYN in original.
b2155e7f
JK
884 *
885 * If the ignored packet is invalid, the receiver will send
886 * a RST we'll catch below.
73f30602 887 */
9fb9cbb1 888 if (index == TCP_SYNACK_SET
c88130bc
PM
889 && ct->proto.tcp.last_index == TCP_SYN_SET
890 && ct->proto.tcp.last_dir != dir
891 && ntohl(th->ack_seq) == ct->proto.tcp.last_end) {
b2155e7f 892 /* b) This SYN/ACK acknowledges a SYN that we earlier
9fb9cbb1
YK
893 * ignored as invalid. This means that the client and
894 * the server are both in sync, while the firewall is
c4832c7b
PNA
895 * not. We get in sync from the previously annotated
896 * values.
9fb9cbb1 897 */
c4832c7b
PNA
898 old_state = TCP_CONNTRACK_SYN_SENT;
899 new_state = TCP_CONNTRACK_SYN_RECV;
900 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_end =
901 ct->proto.tcp.last_end;
902 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_maxend =
903 ct->proto.tcp.last_end;
904 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_maxwin =
905 ct->proto.tcp.last_win == 0 ?
906 1 : ct->proto.tcp.last_win;
907 ct->proto.tcp.seen[ct->proto.tcp.last_dir].td_scale =
908 ct->proto.tcp.last_wscale;
909 ct->proto.tcp.seen[ct->proto.tcp.last_dir].flags =
910 ct->proto.tcp.last_flags;
911 memset(&ct->proto.tcp.seen[dir], 0,
912 sizeof(struct ip_ct_tcp_state));
913 break;
9fb9cbb1 914 }
c88130bc
PM
915 ct->proto.tcp.last_index = index;
916 ct->proto.tcp.last_dir = dir;
917 ct->proto.tcp.last_seq = ntohl(th->seq);
918 ct->proto.tcp.last_end =
9fb9cbb1 919 segment_seq_plus_len(ntohl(th->seq), skb->len, dataoff, th);
c4832c7b
PNA
920 ct->proto.tcp.last_win = ntohs(th->window);
921
922 /* a) This is a SYN in ORIGINAL. The client and the server
923 * may be in sync but we are not. In that case, we annotate
924 * the TCP options and let the packet go through. If it is a
925 * valid SYN packet, the server will reply with a SYN/ACK, and
926 * then we'll get in sync. Otherwise, the server ignores it. */
927 if (index == TCP_SYN_SET && dir == IP_CT_DIR_ORIGINAL) {
928 struct ip_ct_tcp_state seen = {};
929
930 ct->proto.tcp.last_flags =
931 ct->proto.tcp.last_wscale = 0;
932 tcp_options(skb, dataoff, th, &seen);
933 if (seen.flags & IP_CT_TCP_FLAG_WINDOW_SCALE) {
934 ct->proto.tcp.last_flags |=
935 IP_CT_TCP_FLAG_WINDOW_SCALE;
936 ct->proto.tcp.last_wscale = seen.td_scale;
937 }
938 if (seen.flags & IP_CT_TCP_FLAG_SACK_PERM) {
939 ct->proto.tcp.last_flags |=
940 IP_CT_TCP_FLAG_SACK_PERM;
941 }
942 }
440f0d58 943 spin_unlock_bh(&ct->lock);
c2a2c7e0 944 if (LOG_INVALID(net, IPPROTO_TCP))
30e0c6a6 945 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
1a4ac987
PNA
946 "nf_ct_tcp: invalid packet ignored in "
947 "state %s ", tcp_conntrack_names[old_state]);
9fb9cbb1
YK
948 return NF_ACCEPT;
949 case TCP_CONNTRACK_MAX:
48b1de4c
PM
950 /* Special case for SYN proxy: when the SYN to the server or
951 * the SYN/ACK from the server is lost, the client may transmit
952 * a keep-alive packet while in SYN_SENT state. This needs to
953 * be associated with the original conntrack entry in order to
954 * generate a new SYN with the correct sequence number.
955 */
956 if (nfct_synproxy(ct) && old_state == TCP_CONNTRACK_SYN_SENT &&
957 index == TCP_ACK_SET && dir == IP_CT_DIR_ORIGINAL &&
958 ct->proto.tcp.last_dir == IP_CT_DIR_ORIGINAL &&
959 ct->proto.tcp.seen[dir].td_end - 1 == ntohl(th->seq)) {
960 pr_debug("nf_ct_tcp: SYN proxy client keep alive\n");
961 spin_unlock_bh(&ct->lock);
962 return NF_ACCEPT;
963 }
964
9fb9cbb1 965 /* Invalid packet */
0d53778e
PM
966 pr_debug("nf_ct_tcp: Invalid dir=%i index=%u ostate=%u\n",
967 dir, get_conntrack_index(th), old_state);
440f0d58 968 spin_unlock_bh(&ct->lock);
c2a2c7e0 969 if (LOG_INVALID(net, IPPROTO_TCP))
30e0c6a6 970 nf_log_packet(net, pf, 0, skb, NULL, NULL, NULL,
9fb9cbb1
YK
971 "nf_ct_tcp: invalid state ");
972 return -NF_ACCEPT;
9fb9cbb1 973 case TCP_CONNTRACK_CLOSE:
bfcaa502
JK
974 if (index == TCP_RST_SET
975 && (ct->proto.tcp.seen[!dir].flags & IP_CT_TCP_FLAG_MAXACK_SET)
976 && before(ntohl(th->seq), ct->proto.tcp.seen[!dir].td_maxack)) {
977 /* Invalid RST */
334a47f6 978 spin_unlock_bh(&ct->lock);
bfcaa502 979 if (LOG_INVALID(net, IPPROTO_TCP))
30e0c6a6
G
980 nf_log_packet(net, pf, 0, skb, NULL, NULL,
981 NULL, "nf_ct_tcp: invalid RST ");
bfcaa502
JK
982 return -NF_ACCEPT;
983 }
9fb9cbb1 984 if (index == TCP_RST_SET
c88130bc
PM
985 && ((test_bit(IPS_SEEN_REPLY_BIT, &ct->status)
986 && ct->proto.tcp.last_index == TCP_SYN_SET)
987 || (!test_bit(IPS_ASSURED_BIT, &ct->status)
988 && ct->proto.tcp.last_index == TCP_ACK_SET))
989 && ntohl(th->ack_seq) == ct->proto.tcp.last_end) {
93b1fae4 990 /* RST sent to invalid SYN or ACK we had let through
73f30602
JK
991 * at a) and c) above:
992 *
993 * a) SYN was in window then
994 * c) we hold a half-open connection.
995 *
996 * Delete our connection entry.
9fb9cbb1 997 * We skip window checking, because packet might ACK
73f30602 998 * segments we ignored. */
9fb9cbb1
YK
999 goto in_window;
1000 }
93b1fae4 1001 /* Just fall through */
9fb9cbb1
YK
1002 default:
1003 /* Keep compilers happy. */
1004 break;
1005 }
1006
c88130bc 1007 if (!tcp_in_window(ct, &ct->proto.tcp, dir, index,
9fb9cbb1 1008 skb, dataoff, th, pf)) {
440f0d58 1009 spin_unlock_bh(&ct->lock);
9fb9cbb1
YK
1010 return -NF_ACCEPT;
1011 }
1012 in_window:
1013 /* From now on we have got in-window packets */
c88130bc
PM
1014 ct->proto.tcp.last_index = index;
1015 ct->proto.tcp.last_dir = dir;
9fb9cbb1 1016
0d53778e 1017 pr_debug("tcp_conntracks: ");
3c9fba65 1018 nf_ct_dump_tuple(tuple);
0d53778e
PM
1019 pr_debug("syn=%i ack=%i fin=%i rst=%i old=%i new=%i\n",
1020 (th->syn ? 1 : 0), (th->ack ? 1 : 0),
1021 (th->fin ? 1 : 0), (th->rst ? 1 : 0),
1022 old_state, new_state);
9fb9cbb1 1023
c88130bc 1024 ct->proto.tcp.state = new_state;
9fb9cbb1 1025 if (old_state != new_state
d0c1fd7a 1026 && new_state == TCP_CONNTRACK_FIN_WAIT)
c88130bc 1027 ct->proto.tcp.seen[dir].flags |= IP_CT_TCP_FLAG_CLOSE_INIT;
ae375044 1028
d2ba1fde 1029 if (ct->proto.tcp.retrans >= tn->tcp_max_retrans &&
2c8503f5
PNA
1030 timeouts[new_state] > timeouts[TCP_CONNTRACK_RETRANS])
1031 timeout = timeouts[TCP_CONNTRACK_RETRANS];
ae375044
PM
1032 else if ((ct->proto.tcp.seen[0].flags | ct->proto.tcp.seen[1].flags) &
1033 IP_CT_TCP_FLAG_DATA_UNACKNOWLEDGED &&
2c8503f5
PNA
1034 timeouts[new_state] > timeouts[TCP_CONNTRACK_UNACK])
1035 timeout = timeouts[TCP_CONNTRACK_UNACK];
ae375044 1036 else
2c8503f5 1037 timeout = timeouts[new_state];
440f0d58 1038 spin_unlock_bh(&ct->lock);
9fb9cbb1 1039
9fb9cbb1 1040 if (new_state != old_state)
a71996fc 1041 nf_conntrack_event_cache(IPCT_PROTOINFO, ct);
9fb9cbb1 1042
c88130bc 1043 if (!test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) {
9fb9cbb1
YK
1044 /* If only reply is a RST, we can consider ourselves not to
1045 have an established connection: this is a fairly common
1046 problem case, so we can delete the conntrack
1047 immediately. --RR */
1048 if (th->rst) {
718d4ad9 1049 nf_ct_kill_acct(ct, ctinfo, skb);
9fb9cbb1
YK
1050 return NF_ACCEPT;
1051 }
6547a221
FW
1052 /* ESTABLISHED without SEEN_REPLY, i.e. mid-connection
1053 * pickup with loose=1. Avoid large ESTABLISHED timeout.
1054 */
1055 if (new_state == TCP_CONNTRACK_ESTABLISHED &&
1056 timeout > timeouts[TCP_CONNTRACK_UNACK])
1057 timeout = timeouts[TCP_CONNTRACK_UNACK];
c88130bc 1058 } else if (!test_bit(IPS_ASSURED_BIT, &ct->status)
9fb9cbb1
YK
1059 && (old_state == TCP_CONNTRACK_SYN_RECV
1060 || old_state == TCP_CONNTRACK_ESTABLISHED)
1061 && new_state == TCP_CONNTRACK_ESTABLISHED) {
601e68e1
YH
1062 /* Set ASSURED if we see see valid ack in ESTABLISHED
1063 after SYN_RECV or a valid answer for a picked up
9fb9cbb1 1064 connection. */
c88130bc 1065 set_bit(IPS_ASSURED_BIT, &ct->status);
858b3133 1066 nf_conntrack_event_cache(IPCT_ASSURED, ct);
9fb9cbb1 1067 }
c88130bc 1068 nf_ct_refresh_acct(ct, ctinfo, skb, timeout);
9fb9cbb1
YK
1069
1070 return NF_ACCEPT;
1071}
601e68e1 1072
9fb9cbb1 1073/* Called when a new connection for this protocol found. */
09f263cd 1074static bool tcp_new(struct nf_conn *ct, const struct sk_buff *skb,
2c8503f5 1075 unsigned int dataoff, unsigned int *timeouts)
9fb9cbb1
YK
1076{
1077 enum tcp_conntrack new_state;
82f568fc
JE
1078 const struct tcphdr *th;
1079 struct tcphdr _tcph;
d2ba1fde
G
1080 struct net *net = nf_ct_net(ct);
1081 struct nf_tcp_net *tn = tcp_pernet(net);
82f568fc
JE
1082 const struct ip_ct_tcp_state *sender = &ct->proto.tcp.seen[0];
1083 const struct ip_ct_tcp_state *receiver = &ct->proto.tcp.seen[1];
9fb9cbb1
YK
1084
1085 th = skb_header_pointer(skb, dataoff, sizeof(_tcph), &_tcph);
1086 BUG_ON(th == NULL);
1087
1088 /* Don't need lock here: this conntrack not in circulation yet */
e5fc9e7a 1089 new_state = tcp_conntracks[0][get_conntrack_index(th)][TCP_CONNTRACK_NONE];
9fb9cbb1
YK
1090
1091 /* Invalid: delete conntrack */
1092 if (new_state >= TCP_CONNTRACK_MAX) {
0d53778e 1093 pr_debug("nf_ct_tcp: invalid new deleting.\n");
09f263cd 1094 return false;
9fb9cbb1
YK
1095 }
1096
1097 if (new_state == TCP_CONNTRACK_SYN_SENT) {
e5fc9e7a 1098 memset(&ct->proto.tcp, 0, sizeof(ct->proto.tcp));
9fb9cbb1 1099 /* SYN packet */
c88130bc 1100 ct->proto.tcp.seen[0].td_end =
9fb9cbb1
YK
1101 segment_seq_plus_len(ntohl(th->seq), skb->len,
1102 dataoff, th);
c88130bc
PM
1103 ct->proto.tcp.seen[0].td_maxwin = ntohs(th->window);
1104 if (ct->proto.tcp.seen[0].td_maxwin == 0)
1105 ct->proto.tcp.seen[0].td_maxwin = 1;
1106 ct->proto.tcp.seen[0].td_maxend =
1107 ct->proto.tcp.seen[0].td_end;
1108
1109 tcp_options(skb, dataoff, th, &ct->proto.tcp.seen[0]);
d2ba1fde 1110 } else if (tn->tcp_loose == 0) {
9fb9cbb1 1111 /* Don't try to pick up connections. */
09f263cd 1112 return false;
9fb9cbb1 1113 } else {
e5fc9e7a 1114 memset(&ct->proto.tcp, 0, sizeof(ct->proto.tcp));
9fb9cbb1
YK
1115 /*
1116 * We are in the middle of a connection,
1117 * its history is lost for us.
1118 * Let's try to use the data from the packet.
1119 */
c88130bc 1120 ct->proto.tcp.seen[0].td_end =
9fb9cbb1
YK
1121 segment_seq_plus_len(ntohl(th->seq), skb->len,
1122 dataoff, th);
c88130bc
PM
1123 ct->proto.tcp.seen[0].td_maxwin = ntohs(th->window);
1124 if (ct->proto.tcp.seen[0].td_maxwin == 0)
1125 ct->proto.tcp.seen[0].td_maxwin = 1;
1126 ct->proto.tcp.seen[0].td_maxend =
1127 ct->proto.tcp.seen[0].td_end +
1128 ct->proto.tcp.seen[0].td_maxwin;
9fb9cbb1 1129
a09113c2
PM
1130 /* We assume SACK and liberal window checking to handle
1131 * window scaling */
c88130bc
PM
1132 ct->proto.tcp.seen[0].flags =
1133 ct->proto.tcp.seen[1].flags = IP_CT_TCP_FLAG_SACK_PERM |
1134 IP_CT_TCP_FLAG_BE_LIBERAL;
9fb9cbb1 1135 }
601e68e1 1136
9fb9cbb1 1137 /* tcp_packet will set them */
c88130bc 1138 ct->proto.tcp.last_index = TCP_NONE_SET;
601e68e1 1139
0d53778e
PM
1140 pr_debug("tcp_new: sender end=%u maxend=%u maxwin=%u scale=%i "
1141 "receiver end=%u maxend=%u maxwin=%u scale=%i\n",
1142 sender->td_end, sender->td_maxend, sender->td_maxwin,
1143 sender->td_scale,
1144 receiver->td_end, receiver->td_maxend, receiver->td_maxwin,
1145 receiver->td_scale);
09f263cd 1146 return true;
9fb9cbb1 1147}
c1d10adb 1148
c0cd1156 1149#if IS_ENABLED(CONFIG_NF_CT_NETLINK)
c1d10adb
PNA
1150
1151#include <linux/netfilter/nfnetlink.h>
1152#include <linux/netfilter/nfnetlink_conntrack.h>
1153
fdf70832 1154static int tcp_to_nlattr(struct sk_buff *skb, struct nlattr *nla,
440f0d58 1155 struct nf_conn *ct)
c1d10adb 1156{
df6fb868 1157 struct nlattr *nest_parms;
c8e2078c 1158 struct nf_ct_tcp_flags tmp = {};
601e68e1 1159
440f0d58 1160 spin_lock_bh(&ct->lock);
df6fb868
PM
1161 nest_parms = nla_nest_start(skb, CTA_PROTOINFO_TCP | NLA_F_NESTED);
1162 if (!nest_parms)
1163 goto nla_put_failure;
1164
4925a459
DM
1165 if (nla_put_u8(skb, CTA_PROTOINFO_TCP_STATE, ct->proto.tcp.state) ||
1166 nla_put_u8(skb, CTA_PROTOINFO_TCP_WSCALE_ORIGINAL,
1167 ct->proto.tcp.seen[0].td_scale) ||
1168 nla_put_u8(skb, CTA_PROTOINFO_TCP_WSCALE_REPLY,
1169 ct->proto.tcp.seen[1].td_scale))
1170 goto nla_put_failure;
c8e2078c
PNA
1171
1172 tmp.flags = ct->proto.tcp.seen[0].flags;
4925a459
DM
1173 if (nla_put(skb, CTA_PROTOINFO_TCP_FLAGS_ORIGINAL,
1174 sizeof(struct nf_ct_tcp_flags), &tmp))
1175 goto nla_put_failure;
c8e2078c
PNA
1176
1177 tmp.flags = ct->proto.tcp.seen[1].flags;
4925a459
DM
1178 if (nla_put(skb, CTA_PROTOINFO_TCP_FLAGS_REPLY,
1179 sizeof(struct nf_ct_tcp_flags), &tmp))
1180 goto nla_put_failure;
440f0d58 1181 spin_unlock_bh(&ct->lock);
c1d10adb 1182
df6fb868 1183 nla_nest_end(skb, nest_parms);
c1d10adb
PNA
1184
1185 return 0;
1186
df6fb868 1187nla_put_failure:
440f0d58 1188 spin_unlock_bh(&ct->lock);
c1d10adb
PNA
1189 return -1;
1190}
1191
f73e924c
PM
1192static const struct nla_policy tcp_nla_policy[CTA_PROTOINFO_TCP_MAX+1] = {
1193 [CTA_PROTOINFO_TCP_STATE] = { .type = NLA_U8 },
1194 [CTA_PROTOINFO_TCP_WSCALE_ORIGINAL] = { .type = NLA_U8 },
1195 [CTA_PROTOINFO_TCP_WSCALE_REPLY] = { .type = NLA_U8 },
1196 [CTA_PROTOINFO_TCP_FLAGS_ORIGINAL] = { .len = sizeof(struct nf_ct_tcp_flags) },
1197 [CTA_PROTOINFO_TCP_FLAGS_REPLY] = { .len = sizeof(struct nf_ct_tcp_flags) },
c1d10adb
PNA
1198};
1199
fdf70832 1200static int nlattr_to_tcp(struct nlattr *cda[], struct nf_conn *ct)
c1d10adb 1201{
2f0d2f10 1202 struct nlattr *pattr = cda[CTA_PROTOINFO_TCP];
df6fb868 1203 struct nlattr *tb[CTA_PROTOINFO_TCP_MAX+1];
f73e924c 1204 int err;
c1d10adb
PNA
1205
1206 /* updates could not contain anything about the private
1207 * protocol info, in that case skip the parsing */
2f0d2f10 1208 if (!pattr)
c1d10adb
PNA
1209 return 0;
1210
2f0d2f10 1211 err = nla_parse_nested(tb, CTA_PROTOINFO_TCP_MAX, pattr, tcp_nla_policy);
f73e924c
PM
1212 if (err < 0)
1213 return err;
c1d10adb 1214
5f7da4d2
PM
1215 if (tb[CTA_PROTOINFO_TCP_STATE] &&
1216 nla_get_u8(tb[CTA_PROTOINFO_TCP_STATE]) >= TCP_CONNTRACK_MAX)
c1d10adb
PNA
1217 return -EINVAL;
1218
440f0d58 1219 spin_lock_bh(&ct->lock);
5f7da4d2
PM
1220 if (tb[CTA_PROTOINFO_TCP_STATE])
1221 ct->proto.tcp.state = nla_get_u8(tb[CTA_PROTOINFO_TCP_STATE]);
c8e2078c 1222
df6fb868 1223 if (tb[CTA_PROTOINFO_TCP_FLAGS_ORIGINAL]) {
c8e2078c 1224 struct nf_ct_tcp_flags *attr =
df6fb868 1225 nla_data(tb[CTA_PROTOINFO_TCP_FLAGS_ORIGINAL]);
c8e2078c
PNA
1226 ct->proto.tcp.seen[0].flags &= ~attr->mask;
1227 ct->proto.tcp.seen[0].flags |= attr->flags & attr->mask;
1228 }
1229
df6fb868 1230 if (tb[CTA_PROTOINFO_TCP_FLAGS_REPLY]) {
c8e2078c 1231 struct nf_ct_tcp_flags *attr =
df6fb868 1232 nla_data(tb[CTA_PROTOINFO_TCP_FLAGS_REPLY]);
c8e2078c
PNA
1233 ct->proto.tcp.seen[1].flags &= ~attr->mask;
1234 ct->proto.tcp.seen[1].flags |= attr->flags & attr->mask;
1235 }
1236
df6fb868
PM
1237 if (tb[CTA_PROTOINFO_TCP_WSCALE_ORIGINAL] &&
1238 tb[CTA_PROTOINFO_TCP_WSCALE_REPLY] &&
c8e2078c
PNA
1239 ct->proto.tcp.seen[0].flags & IP_CT_TCP_FLAG_WINDOW_SCALE &&
1240 ct->proto.tcp.seen[1].flags & IP_CT_TCP_FLAG_WINDOW_SCALE) {
77236b6e
PM
1241 ct->proto.tcp.seen[0].td_scale =
1242 nla_get_u8(tb[CTA_PROTOINFO_TCP_WSCALE_ORIGINAL]);
1243 ct->proto.tcp.seen[1].td_scale =
1244 nla_get_u8(tb[CTA_PROTOINFO_TCP_WSCALE_REPLY]);
c8e2078c 1245 }
440f0d58 1246 spin_unlock_bh(&ct->lock);
c1d10adb
PNA
1247
1248 return 0;
1249}
a400c30e
HE
1250
1251static int tcp_nlattr_size(void)
1252{
1253 return nla_total_size(0) /* CTA_PROTOINFO_TCP */
1254 + nla_policy_len(tcp_nla_policy, CTA_PROTOINFO_TCP_MAX + 1);
1255}
1256
1257static int tcp_nlattr_tuple_size(void)
1258{
1259 return nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1);
1260}
c1d10adb 1261#endif
933a41e7 1262
50978462
PNA
1263#if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
1264
1265#include <linux/netfilter/nfnetlink.h>
1266#include <linux/netfilter/nfnetlink_cttimeout.h>
1267
8264deb8
G
1268static int tcp_timeout_nlattr_to_obj(struct nlattr *tb[],
1269 struct net *net, void *data)
50978462
PNA
1270{
1271 unsigned int *timeouts = data;
8264deb8 1272 struct nf_tcp_net *tn = tcp_pernet(net);
50978462
PNA
1273 int i;
1274
1275 /* set default TCP timeouts. */
1276 for (i=0; i<TCP_CONNTRACK_TIMEOUT_MAX; i++)
8264deb8 1277 timeouts[i] = tn->timeouts[i];
50978462
PNA
1278
1279 if (tb[CTA_TIMEOUT_TCP_SYN_SENT]) {
1280 timeouts[TCP_CONNTRACK_SYN_SENT] =
1281 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_SYN_SENT]))*HZ;
1282 }
1283 if (tb[CTA_TIMEOUT_TCP_SYN_RECV]) {
1284 timeouts[TCP_CONNTRACK_SYN_RECV] =
1285 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_SYN_RECV]))*HZ;
1286 }
1287 if (tb[CTA_TIMEOUT_TCP_ESTABLISHED]) {
1288 timeouts[TCP_CONNTRACK_ESTABLISHED] =
1289 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_ESTABLISHED]))*HZ;
1290 }
1291 if (tb[CTA_TIMEOUT_TCP_FIN_WAIT]) {
1292 timeouts[TCP_CONNTRACK_FIN_WAIT] =
1293 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_FIN_WAIT]))*HZ;
1294 }
1295 if (tb[CTA_TIMEOUT_TCP_CLOSE_WAIT]) {
1296 timeouts[TCP_CONNTRACK_CLOSE_WAIT] =
1297 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_CLOSE_WAIT]))*HZ;
1298 }
1299 if (tb[CTA_TIMEOUT_TCP_LAST_ACK]) {
1300 timeouts[TCP_CONNTRACK_LAST_ACK] =
1301 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_LAST_ACK]))*HZ;
1302 }
1303 if (tb[CTA_TIMEOUT_TCP_TIME_WAIT]) {
1304 timeouts[TCP_CONNTRACK_TIME_WAIT] =
1305 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_TIME_WAIT]))*HZ;
1306 }
1307 if (tb[CTA_TIMEOUT_TCP_CLOSE]) {
1308 timeouts[TCP_CONNTRACK_CLOSE] =
1309 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_CLOSE]))*HZ;
1310 }
1311 if (tb[CTA_TIMEOUT_TCP_SYN_SENT2]) {
1312 timeouts[TCP_CONNTRACK_SYN_SENT2] =
1313 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_SYN_SENT2]))*HZ;
1314 }
1315 if (tb[CTA_TIMEOUT_TCP_RETRANS]) {
1316 timeouts[TCP_CONNTRACK_RETRANS] =
1317 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_RETRANS]))*HZ;
1318 }
1319 if (tb[CTA_TIMEOUT_TCP_UNACK]) {
1320 timeouts[TCP_CONNTRACK_UNACK] =
1321 ntohl(nla_get_be32(tb[CTA_TIMEOUT_TCP_UNACK]))*HZ;
1322 }
1323 return 0;
1324}
1325
1326static int
1327tcp_timeout_obj_to_nlattr(struct sk_buff *skb, const void *data)
1328{
1329 const unsigned int *timeouts = data;
1330
4925a459
DM
1331 if (nla_put_be32(skb, CTA_TIMEOUT_TCP_SYN_SENT,
1332 htonl(timeouts[TCP_CONNTRACK_SYN_SENT] / HZ)) ||
1333 nla_put_be32(skb, CTA_TIMEOUT_TCP_SYN_RECV,
1334 htonl(timeouts[TCP_CONNTRACK_SYN_RECV] / HZ)) ||
1335 nla_put_be32(skb, CTA_TIMEOUT_TCP_ESTABLISHED,
1336 htonl(timeouts[TCP_CONNTRACK_ESTABLISHED] / HZ)) ||
1337 nla_put_be32(skb, CTA_TIMEOUT_TCP_FIN_WAIT,
1338 htonl(timeouts[TCP_CONNTRACK_FIN_WAIT] / HZ)) ||
1339 nla_put_be32(skb, CTA_TIMEOUT_TCP_CLOSE_WAIT,
1340 htonl(timeouts[TCP_CONNTRACK_CLOSE_WAIT] / HZ)) ||
1341 nla_put_be32(skb, CTA_TIMEOUT_TCP_LAST_ACK,
1342 htonl(timeouts[TCP_CONNTRACK_LAST_ACK] / HZ)) ||
1343 nla_put_be32(skb, CTA_TIMEOUT_TCP_TIME_WAIT,
1344 htonl(timeouts[TCP_CONNTRACK_TIME_WAIT] / HZ)) ||
1345 nla_put_be32(skb, CTA_TIMEOUT_TCP_CLOSE,
1346 htonl(timeouts[TCP_CONNTRACK_CLOSE] / HZ)) ||
1347 nla_put_be32(skb, CTA_TIMEOUT_TCP_SYN_SENT2,
1348 htonl(timeouts[TCP_CONNTRACK_SYN_SENT2] / HZ)) ||
1349 nla_put_be32(skb, CTA_TIMEOUT_TCP_RETRANS,
1350 htonl(timeouts[TCP_CONNTRACK_RETRANS] / HZ)) ||
1351 nla_put_be32(skb, CTA_TIMEOUT_TCP_UNACK,
1352 htonl(timeouts[TCP_CONNTRACK_UNACK] / HZ)))
1353 goto nla_put_failure;
50978462
PNA
1354 return 0;
1355
1356nla_put_failure:
1357 return -ENOSPC;
1358}
1359
1360static const struct nla_policy tcp_timeout_nla_policy[CTA_TIMEOUT_TCP_MAX+1] = {
1361 [CTA_TIMEOUT_TCP_SYN_SENT] = { .type = NLA_U32 },
1362 [CTA_TIMEOUT_TCP_SYN_RECV] = { .type = NLA_U32 },
1363 [CTA_TIMEOUT_TCP_ESTABLISHED] = { .type = NLA_U32 },
1364 [CTA_TIMEOUT_TCP_FIN_WAIT] = { .type = NLA_U32 },
1365 [CTA_TIMEOUT_TCP_CLOSE_WAIT] = { .type = NLA_U32 },
1366 [CTA_TIMEOUT_TCP_LAST_ACK] = { .type = NLA_U32 },
1367 [CTA_TIMEOUT_TCP_TIME_WAIT] = { .type = NLA_U32 },
1368 [CTA_TIMEOUT_TCP_CLOSE] = { .type = NLA_U32 },
1369 [CTA_TIMEOUT_TCP_SYN_SENT2] = { .type = NLA_U32 },
6d1fafca
FW
1370 [CTA_TIMEOUT_TCP_RETRANS] = { .type = NLA_U32 },
1371 [CTA_TIMEOUT_TCP_UNACK] = { .type = NLA_U32 },
50978462
PNA
1372};
1373#endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
1374
933a41e7 1375#ifdef CONFIG_SYSCTL
933a41e7
PM
1376static struct ctl_table tcp_sysctl_table[] = {
1377 {
933a41e7 1378 .procname = "nf_conntrack_tcp_timeout_syn_sent",
933a41e7
PM
1379 .maxlen = sizeof(unsigned int),
1380 .mode = 0644,
6d9f239a 1381 .proc_handler = proc_dointvec_jiffies,
933a41e7
PM
1382 },
1383 {
933a41e7 1384 .procname = "nf_conntrack_tcp_timeout_syn_recv",
933a41e7
PM
1385 .maxlen = sizeof(unsigned int),
1386 .mode = 0644,
6d9f239a 1387 .proc_handler = proc_dointvec_jiffies,
933a41e7
PM
1388 },
1389 {
933a41e7 1390 .procname = "nf_conntrack_tcp_timeout_established",
933a41e7
PM
1391 .maxlen = sizeof(unsigned int),
1392 .mode = 0644,
6d9f239a 1393 .proc_handler = proc_dointvec_jiffies,
933a41e7
PM
1394 },
1395 {
933a41e7 1396 .procname = "nf_conntrack_tcp_timeout_fin_wait",
933a41e7
PM
1397 .maxlen = sizeof(unsigned int),
1398 .mode = 0644,
6d9f239a 1399 .proc_handler = proc_dointvec_jiffies,
933a41e7
PM
1400 },
1401 {
933a41e7 1402 .procname = "nf_conntrack_tcp_timeout_close_wait",
933a41e7
PM
1403 .maxlen = sizeof(unsigned int),
1404 .mode = 0644,
6d9f239a 1405 .proc_handler = proc_dointvec_jiffies,
933a41e7
PM
1406 },
1407 {
933a41e7 1408 .procname = "nf_conntrack_tcp_timeout_last_ack",
933a41e7
PM
1409 .maxlen = sizeof(unsigned int),
1410 .mode = 0644,
6d9f239a 1411 .proc_handler = proc_dointvec_jiffies,
933a41e7
PM
1412 },
1413 {
933a41e7 1414 .procname = "nf_conntrack_tcp_timeout_time_wait",
933a41e7
PM
1415 .maxlen = sizeof(unsigned int),
1416 .mode = 0644,
6d9f239a 1417 .proc_handler = proc_dointvec_jiffies,
933a41e7
PM
1418 },
1419 {
933a41e7 1420 .procname = "nf_conntrack_tcp_timeout_close",
933a41e7
PM
1421 .maxlen = sizeof(unsigned int),
1422 .mode = 0644,
6d9f239a 1423 .proc_handler = proc_dointvec_jiffies,
933a41e7
PM
1424 },
1425 {
933a41e7 1426 .procname = "nf_conntrack_tcp_timeout_max_retrans",
933a41e7
PM
1427 .maxlen = sizeof(unsigned int),
1428 .mode = 0644,
6d9f239a 1429 .proc_handler = proc_dointvec_jiffies,
933a41e7 1430 },
ae375044
PM
1431 {
1432 .procname = "nf_conntrack_tcp_timeout_unacknowledged",
ae375044
PM
1433 .maxlen = sizeof(unsigned int),
1434 .mode = 0644,
6d9f239a 1435 .proc_handler = proc_dointvec_jiffies,
ae375044 1436 },
933a41e7 1437 {
933a41e7 1438 .procname = "nf_conntrack_tcp_loose",
933a41e7
PM
1439 .maxlen = sizeof(unsigned int),
1440 .mode = 0644,
6d9f239a 1441 .proc_handler = proc_dointvec,
933a41e7
PM
1442 },
1443 {
933a41e7 1444 .procname = "nf_conntrack_tcp_be_liberal",
933a41e7
PM
1445 .maxlen = sizeof(unsigned int),
1446 .mode = 0644,
6d9f239a 1447 .proc_handler = proc_dointvec,
933a41e7
PM
1448 },
1449 {
933a41e7 1450 .procname = "nf_conntrack_tcp_max_retrans",
933a41e7
PM
1451 .maxlen = sizeof(unsigned int),
1452 .mode = 0644,
6d9f239a 1453 .proc_handler = proc_dointvec,
933a41e7 1454 },
f8572d8f 1455 { }
933a41e7 1456};
a999e683
PM
1457
1458#ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT
1459static struct ctl_table tcp_compat_sysctl_table[] = {
1460 {
a999e683 1461 .procname = "ip_conntrack_tcp_timeout_syn_sent",
a999e683
PM
1462 .maxlen = sizeof(unsigned int),
1463 .mode = 0644,
6d9f239a 1464 .proc_handler = proc_dointvec_jiffies,
a999e683 1465 },
874ab923
JK
1466 {
1467 .procname = "ip_conntrack_tcp_timeout_syn_sent2",
874ab923
JK
1468 .maxlen = sizeof(unsigned int),
1469 .mode = 0644,
1470 .proc_handler = proc_dointvec_jiffies,
1471 },
a999e683 1472 {
a999e683 1473 .procname = "ip_conntrack_tcp_timeout_syn_recv",
a999e683
PM
1474 .maxlen = sizeof(unsigned int),
1475 .mode = 0644,
6d9f239a 1476 .proc_handler = proc_dointvec_jiffies,
a999e683
PM
1477 },
1478 {
a999e683 1479 .procname = "ip_conntrack_tcp_timeout_established",
a999e683
PM
1480 .maxlen = sizeof(unsigned int),
1481 .mode = 0644,
6d9f239a 1482 .proc_handler = proc_dointvec_jiffies,
a999e683
PM
1483 },
1484 {
a999e683 1485 .procname = "ip_conntrack_tcp_timeout_fin_wait",
a999e683
PM
1486 .maxlen = sizeof(unsigned int),
1487 .mode = 0644,
6d9f239a 1488 .proc_handler = proc_dointvec_jiffies,
a999e683
PM
1489 },
1490 {
a999e683 1491 .procname = "ip_conntrack_tcp_timeout_close_wait",
a999e683
PM
1492 .maxlen = sizeof(unsigned int),
1493 .mode = 0644,
6d9f239a 1494 .proc_handler = proc_dointvec_jiffies,
a999e683
PM
1495 },
1496 {
a999e683 1497 .procname = "ip_conntrack_tcp_timeout_last_ack",
a999e683
PM
1498 .maxlen = sizeof(unsigned int),
1499 .mode = 0644,
6d9f239a 1500 .proc_handler = proc_dointvec_jiffies,
a999e683
PM
1501 },
1502 {
a999e683 1503 .procname = "ip_conntrack_tcp_timeout_time_wait",
a999e683
PM
1504 .maxlen = sizeof(unsigned int),
1505 .mode = 0644,
6d9f239a 1506 .proc_handler = proc_dointvec_jiffies,
a999e683
PM
1507 },
1508 {
a999e683 1509 .procname = "ip_conntrack_tcp_timeout_close",
a999e683
PM
1510 .maxlen = sizeof(unsigned int),
1511 .mode = 0644,
6d9f239a 1512 .proc_handler = proc_dointvec_jiffies,
a999e683
PM
1513 },
1514 {
a999e683 1515 .procname = "ip_conntrack_tcp_timeout_max_retrans",
a999e683
PM
1516 .maxlen = sizeof(unsigned int),
1517 .mode = 0644,
6d9f239a 1518 .proc_handler = proc_dointvec_jiffies,
a999e683
PM
1519 },
1520 {
a999e683 1521 .procname = "ip_conntrack_tcp_loose",
a999e683
PM
1522 .maxlen = sizeof(unsigned int),
1523 .mode = 0644,
6d9f239a 1524 .proc_handler = proc_dointvec,
a999e683
PM
1525 },
1526 {
a999e683 1527 .procname = "ip_conntrack_tcp_be_liberal",
a999e683
PM
1528 .maxlen = sizeof(unsigned int),
1529 .mode = 0644,
6d9f239a 1530 .proc_handler = proc_dointvec,
a999e683
PM
1531 },
1532 {
a999e683 1533 .procname = "ip_conntrack_tcp_max_retrans",
a999e683
PM
1534 .maxlen = sizeof(unsigned int),
1535 .mode = 0644,
6d9f239a 1536 .proc_handler = proc_dointvec,
a999e683 1537 },
f8572d8f 1538 { }
a999e683
PM
1539};
1540#endif /* CONFIG_NF_CONNTRACK_PROC_COMPAT */
933a41e7
PM
1541#endif /* CONFIG_SYSCTL */
1542
efa758fe
G
1543static int tcp_kmemdup_sysctl_table(struct nf_proto_net *pn,
1544 struct nf_tcp_net *tn)
d2ba1fde
G
1545{
1546#ifdef CONFIG_SYSCTL
d2ba1fde
G
1547 if (pn->ctl_table)
1548 return 0;
1549
1550 pn->ctl_table = kmemdup(tcp_sysctl_table,
1551 sizeof(tcp_sysctl_table),
1552 GFP_KERNEL);
1553 if (!pn->ctl_table)
1554 return -ENOMEM;
1555
1556 pn->ctl_table[0].data = &tn->timeouts[TCP_CONNTRACK_SYN_SENT];
1557 pn->ctl_table[1].data = &tn->timeouts[TCP_CONNTRACK_SYN_RECV];
1558 pn->ctl_table[2].data = &tn->timeouts[TCP_CONNTRACK_ESTABLISHED];
1559 pn->ctl_table[3].data = &tn->timeouts[TCP_CONNTRACK_FIN_WAIT];
1560 pn->ctl_table[4].data = &tn->timeouts[TCP_CONNTRACK_CLOSE_WAIT];
1561 pn->ctl_table[5].data = &tn->timeouts[TCP_CONNTRACK_LAST_ACK];
1562 pn->ctl_table[6].data = &tn->timeouts[TCP_CONNTRACK_TIME_WAIT];
1563 pn->ctl_table[7].data = &tn->timeouts[TCP_CONNTRACK_CLOSE];
1564 pn->ctl_table[8].data = &tn->timeouts[TCP_CONNTRACK_RETRANS];
1565 pn->ctl_table[9].data = &tn->timeouts[TCP_CONNTRACK_UNACK];
1566 pn->ctl_table[10].data = &tn->tcp_loose;
1567 pn->ctl_table[11].data = &tn->tcp_be_liberal;
1568 pn->ctl_table[12].data = &tn->tcp_max_retrans;
1569#endif
1570 return 0;
1571}
1572
efa758fe
G
1573static int tcp_kmemdup_compat_sysctl_table(struct nf_proto_net *pn,
1574 struct nf_tcp_net *tn)
d2ba1fde
G
1575{
1576#ifdef CONFIG_SYSCTL
1577#ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT
d2ba1fde
G
1578 pn->ctl_compat_table = kmemdup(tcp_compat_sysctl_table,
1579 sizeof(tcp_compat_sysctl_table),
1580 GFP_KERNEL);
1581 if (!pn->ctl_compat_table)
1582 return -ENOMEM;
1583
1584 pn->ctl_compat_table[0].data = &tn->timeouts[TCP_CONNTRACK_SYN_SENT];
1585 pn->ctl_compat_table[1].data = &tn->timeouts[TCP_CONNTRACK_SYN_SENT2];
1586 pn->ctl_compat_table[2].data = &tn->timeouts[TCP_CONNTRACK_SYN_RECV];
1587 pn->ctl_compat_table[3].data = &tn->timeouts[TCP_CONNTRACK_ESTABLISHED];
1588 pn->ctl_compat_table[4].data = &tn->timeouts[TCP_CONNTRACK_FIN_WAIT];
1589 pn->ctl_compat_table[5].data = &tn->timeouts[TCP_CONNTRACK_CLOSE_WAIT];
1590 pn->ctl_compat_table[6].data = &tn->timeouts[TCP_CONNTRACK_LAST_ACK];
1591 pn->ctl_compat_table[7].data = &tn->timeouts[TCP_CONNTRACK_TIME_WAIT];
1592 pn->ctl_compat_table[8].data = &tn->timeouts[TCP_CONNTRACK_CLOSE];
1593 pn->ctl_compat_table[9].data = &tn->timeouts[TCP_CONNTRACK_RETRANS];
1594 pn->ctl_compat_table[10].data = &tn->tcp_loose;
1595 pn->ctl_compat_table[11].data = &tn->tcp_be_liberal;
1596 pn->ctl_compat_table[12].data = &tn->tcp_max_retrans;
1597#endif
1598#endif
1599 return 0;
1600}
1601
efa758fe 1602static int tcp_init_net(struct net *net, u_int16_t proto)
d2ba1fde 1603{
efa758fe 1604 int ret;
d2ba1fde 1605 struct nf_tcp_net *tn = tcp_pernet(net);
efa758fe
G
1606 struct nf_proto_net *pn = &tn->pn;
1607
1608 if (!pn->users) {
1609 int i;
d2ba1fde 1610
d2ba1fde
G
1611 for (i = 0; i < TCP_CONNTRACK_TIMEOUT_MAX; i++)
1612 tn->timeouts[i] = tcp_timeouts[i];
1613
1614 tn->tcp_loose = nf_ct_tcp_loose;
1615 tn->tcp_be_liberal = nf_ct_tcp_be_liberal;
1616 tn->tcp_max_retrans = nf_ct_tcp_max_retrans;
1617 }
1618
efa758fe
G
1619 if (proto == AF_INET) {
1620 ret = tcp_kmemdup_compat_sysctl_table(pn, tn);
1621 if (ret < 0)
1622 return ret;
d2ba1fde 1623
efa758fe
G
1624 ret = tcp_kmemdup_sysctl_table(pn, tn);
1625 if (ret < 0)
1626 nf_ct_kfree_compat_sysctl_table(pn);
1627 } else
1628 ret = tcp_kmemdup_sysctl_table(pn, tn);
d2ba1fde 1629
d2ba1fde
G
1630 return ret;
1631}
1632
08911475
PNA
1633static struct nf_proto_net *tcp_get_net_proto(struct net *net)
1634{
1635 return &net->ct.nf_ct_proto.tcp.pn;
1636}
1637
61075af5 1638struct nf_conntrack_l4proto nf_conntrack_l4proto_tcp4 __read_mostly =
9fb9cbb1
YK
1639{
1640 .l3proto = PF_INET,
605dcad6 1641 .l4proto = IPPROTO_TCP,
9fb9cbb1
YK
1642 .name = "tcp",
1643 .pkt_to_tuple = tcp_pkt_to_tuple,
1644 .invert_tuple = tcp_invert_tuple,
1645 .print_tuple = tcp_print_tuple,
1646 .print_conntrack = tcp_print_conntrack,
1647 .packet = tcp_packet,
2c8503f5 1648 .get_timeouts = tcp_get_timeouts,
9fb9cbb1 1649 .new = tcp_new,
96f6bf82 1650 .error = tcp_error,
c0cd1156 1651#if IS_ENABLED(CONFIG_NF_CT_NETLINK)
fdf70832 1652 .to_nlattr = tcp_to_nlattr,
a400c30e 1653 .nlattr_size = tcp_nlattr_size,
fdf70832
PM
1654 .from_nlattr = nlattr_to_tcp,
1655 .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr,
1656 .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple,
a400c30e 1657 .nlattr_tuple_size = tcp_nlattr_tuple_size,
f73e924c 1658 .nla_policy = nf_ct_port_nla_policy,
c1d10adb 1659#endif
50978462
PNA
1660#if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
1661 .ctnl_timeout = {
1662 .nlattr_to_obj = tcp_timeout_nlattr_to_obj,
1663 .obj_to_nlattr = tcp_timeout_obj_to_nlattr,
1664 .nlattr_max = CTA_TIMEOUT_TCP_MAX,
1665 .obj_size = sizeof(unsigned int) *
1666 TCP_CONNTRACK_TIMEOUT_MAX,
1667 .nla_policy = tcp_timeout_nla_policy,
1668 },
1669#endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
efa758fe 1670 .init_net = tcp_init_net,
08911475 1671 .get_net_proto = tcp_get_net_proto,
9fb9cbb1 1672};
13b18339 1673EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_tcp4);
9fb9cbb1 1674
61075af5 1675struct nf_conntrack_l4proto nf_conntrack_l4proto_tcp6 __read_mostly =
9fb9cbb1
YK
1676{
1677 .l3proto = PF_INET6,
605dcad6 1678 .l4proto = IPPROTO_TCP,
9fb9cbb1
YK
1679 .name = "tcp",
1680 .pkt_to_tuple = tcp_pkt_to_tuple,
1681 .invert_tuple = tcp_invert_tuple,
1682 .print_tuple = tcp_print_tuple,
1683 .print_conntrack = tcp_print_conntrack,
1684 .packet = tcp_packet,
2c8503f5 1685 .get_timeouts = tcp_get_timeouts,
9fb9cbb1 1686 .new = tcp_new,
96f6bf82 1687 .error = tcp_error,
c0cd1156 1688#if IS_ENABLED(CONFIG_NF_CT_NETLINK)
fdf70832 1689 .to_nlattr = tcp_to_nlattr,
a400c30e 1690 .nlattr_size = tcp_nlattr_size,
fdf70832
PM
1691 .from_nlattr = nlattr_to_tcp,
1692 .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr,
1693 .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple,
a400c30e 1694 .nlattr_tuple_size = tcp_nlattr_tuple_size,
f73e924c 1695 .nla_policy = nf_ct_port_nla_policy,
c1d10adb 1696#endif
50978462
PNA
1697#if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
1698 .ctnl_timeout = {
1699 .nlattr_to_obj = tcp_timeout_nlattr_to_obj,
1700 .obj_to_nlattr = tcp_timeout_obj_to_nlattr,
1701 .nlattr_max = CTA_TIMEOUT_TCP_MAX,
1702 .obj_size = sizeof(unsigned int) *
1703 TCP_CONNTRACK_TIMEOUT_MAX,
1704 .nla_policy = tcp_timeout_nla_policy,
1705 },
1706#endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
efa758fe 1707 .init_net = tcp_init_net,
08911475 1708 .get_net_proto = tcp_get_net_proto,
9fb9cbb1 1709};
13b18339 1710EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_tcp6);